LLM4Binary/decompile-bench · Datasets at Hugging Face

name stringlengths 1 473k	code stringlengths 7 647k	asm stringlengths 4 3.39M	file stringlengths 8 196
ncnn::im2col_sgemm_pack8to1_int8_sse_xop(ncnn::Mat const&, ncnn::Mat&, ncnn::Mat const&, ncnn::Option const&)	void im2col_sgemm_pack8to1_int8_sse_xop(const Mat& bottom_im2col, Mat& top_blob, const Mat& kernel, const Option& opt) { im2col_sgemm_pack8to1_int8_sse(bottom_im2col, top_blob, kernel, opt); }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x88, %rsp movq %rcx, %rax movq %rdx, 0x10(%rsp) movq %rdi, %r12 movslq 0x2c(%rdi), %r13 movl 0x30(%rdi), %ebx movl 0x38(%rdi), %r15d movq %rsi, 0x28(%rsp) movslq 0x38(%rsi), %rcx movq %rcx, 0x18(%rsp) leaq 0x30(%rsp), %rdi andq $0x0, 0x40(%rdi) movl %r13d, %ecx shrl %ecx movl %r13d, %r10d andl $0x1, %r10d addl %ecx, %r10d cmpq $0x2, %r13 setge %cl vpxor %xmm0, %xmm0, %xmm0 vmovdqu %xmm0, 0xc(%rdi) vmovdqa %xmm0, (%rdi) vmovdqu %xmm0, 0x2c(%rdi) vmovdqa %xmm0, 0x20(%rdi) cmovll %r13d, %r10d movl %ebx, %esi shll %cl, %esi movq 0x10(%rax), %rax movq %rax, (%rsp) pushq $0x8 popq %r8 pushq $0x8 popq %r9 movl %r15d, %edx movl %r10d, %ecx callq 0x628f2 movl %r13d, %eax sarl %eax xorl %ecx, %ecx testl %ebx, %ebx movl %ebx, %edx movl $0x0, %ebx movl %edx, 0x24(%rsp) cmovgl %edx, %ebx testl %r15d, %r15d movl $0x0, %edi cmovgl %r15d, %edi testl %eax, %eax cmovlel %ecx, %eax leaq (,%r13,8), %r8 xorl %edx, %edx cmpq %rax, %rdx je 0x138d13 movq 0x70(%rsp), %r9 imulq %rdx, %r9 imulq 0x40(%rsp), %r9 addq 0x30(%rsp), %r9 xorl %r10d, %r10d cmpq %rdi, %r10 je 0x138d0a movq 0x40(%r12), %r11 movq (%r12), %rsi imulq 0x10(%r12), %r11 addq %rcx, %rsi imulq %r10, %r11 addq %rsi, %r11 movl %ebx, %ebp subl $0x1, %ebp jb 0x138d05 vmovdqu (%r11), %xmm0 vmovdqu %xmm0, (%r9) addq $0x10, %r9 addq %r8, %r11 jmp 0x138ced incq %r10 jmp 0x138ccd incq %rdx addq $0x10, %rcx jmp 0x138cb1 movq %r13, %rcx andq $-0x2, %rcx movq 0x30(%rsp), %r9 leaq (,%rcx,8), %r10 cmpq %r13, %rcx jge 0x138d8c movl %ecx, %eax cltd pushq $0x2 popq %rsi idivl %esi addl %eax, %edx movslq %edx, %rax imulq 0x70(%rsp), %rax imulq 0x40(%rsp), %rax addq %r9, %rax movq (%r12), %rdx addq %r10, %rdx xorl %ebp, %ebp cmpq %rdi, %rbp je 0x138d83 movq 0x40(%r12), %r14 imulq 0x10(%r12), %r14 imulq %rbp, %r14 addq %rdx, %r14 movl %ebx, %esi subl $0x1, %esi jb 0x138d7e movq (%r14), %r11 movq %r11, (%rax) addq $0x8, %rax addq %r8, %r14 jmp 0x138d6a incq %rbp jmp 0x138d51 incq %rcx addq $0x8, %r10 jmp 0x138d27 movq 0x18(%rsp), %rax movl %eax, %ebx sarl $0x2, %ebx movq 0x28(%rsp), %rcx movq (%rcx), %rdx movq %rdx, 0x80(%rsp) movq 0x10(%rcx), %rdx imulq 0x40(%rcx), %rdx movq %rdx, 0x78(%rsp) imull 0x24(%rsp), %r15d xorl %esi, %esi testl %r15d, %r15d cmovlel %esi, %r15d testl %ebx, %ebx cmovlel %esi, %ebx vpxor %xmm0, %xmm0, %xmm0 cmpq %rbx, %rsi je 0x1390aa leaq (,%rsi,4), %rdi movq 0x78(%rsp), %rcx imulq %rcx, %rdi movq 0x80(%rsp), %rax addq %rax, %rdi leaq 0x1(,%rsi,4), %r8 imulq %rcx, %r8 addq %rax, %r8 leaq 0x2(,%rsi,4), %r9 imulq %rcx, %r9 addq %rax, %r9 leaq 0x3(,%rsi,4), %r10 imulq %rcx, %r10 addq %rax, %r10 movq 0x10(%rsp), %rax movq 0x40(%rax), %r14 imulq 0x10(%rax), %r14 movq 0x30(%rsp), %r12 imulq %rsi, %r14 addq (%rax), %r14 movq 0x40(%rsp), %rbp imulq 0x70(%rsp), %rbp xorl %r11d, %r11d movq %r11, %rax orq $0x1, %rax cmpq %r13, %rax jge 0x138f9f vpxor %xmm1, %xmm1, %xmm1 xorl %eax, %eax movl %r15d, %ecx vpxor %xmm2, %xmm2, %xmm2 vpxor %xmm3, %xmm3, %xmm3 vpxor %xmm4, %xmm4, %xmm4 vpxor %xmm5, %xmm5, %xmm5 vpxor %xmm6, %xmm6, %xmm6 vpxor %xmm7, %xmm7, %xmm7 vpxor %xmm8, %xmm8, %xmm8 subl $0x1, %ecx jb 0x138ef7 vmovdqu (%r12,%rax), %xmm9 vpcmpgtb %xmm9, %xmm0, %xmm10 vpunpcklbw %xmm10, %xmm9, %xmm11 # xmm11 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3],xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7] vpunpckhbw %xmm10, %xmm9, %xmm9 # xmm9 = xmm9[8],xmm10[8],xmm9[9],xmm10[9],xmm9[10],xmm10[10],xmm9[11],xmm10[11],xmm9[12],xmm10[12],xmm9[13],xmm10[13],xmm9[14],xmm10[14],xmm9[15],xmm10[15] vmovdqu (%r14,%rax,2), %xmm10 vmovdqu 0x10(%r14,%rax,2), %xmm12 vpcmpgtb %xmm10, %xmm0, %xmm13 vpcmpgtb %xmm12, %xmm0, %xmm14 vpunpcklbw %xmm13, %xmm10, %xmm15 # xmm15 = xmm10[0],xmm13[0],xmm10[1],xmm13[1],xmm10[2],xmm13[2],xmm10[3],xmm13[3],xmm10[4],xmm13[4],xmm10[5],xmm13[5],xmm10[6],xmm13[6],xmm10[7],xmm13[7] vpunpckhbw %xmm13, %xmm10, %xmm10 # xmm10 = xmm10[8],xmm13[8],xmm10[9],xmm13[9],xmm10[10],xmm13[10],xmm10[11],xmm13[11],xmm10[12],xmm13[12],xmm10[13],xmm13[13],xmm10[14],xmm13[14],xmm10[15],xmm13[15] vpunpcklbw %xmm14, %xmm12, %xmm13 # xmm13 = xmm12[0],xmm14[0],xmm12[1],xmm14[1],xmm12[2],xmm14[2],xmm12[3],xmm14[3],xmm12[4],xmm14[4],xmm12[5],xmm14[5],xmm12[6],xmm14[6],xmm12[7],xmm14[7] vpunpckhbw %xmm14, %xmm12, %xmm12 # xmm12 = xmm12[8],xmm14[8],xmm12[9],xmm14[9],xmm12[10],xmm14[10],xmm12[11],xmm14[11],xmm12[12],xmm14[12],xmm12[13],xmm14[13],xmm12[14],xmm14[14],xmm12[15],xmm14[15] vpmadcswd %xmm8, %xmm15, %xmm11, %xmm8 vpmadcswd %xmm7, %xmm10, %xmm11, %xmm7 vpmadcswd %xmm6, %xmm13, %xmm11, %xmm6 vpmadcswd %xmm5, %xmm12, %xmm11, %xmm5 vpmadcswd %xmm4, %xmm15, %xmm9, %xmm4 vpmadcswd %xmm3, %xmm10, %xmm9, %xmm3 vpmadcswd %xmm2, %xmm13, %xmm9, %xmm2 vpmadcswd %xmm1, %xmm12, %xmm9, %xmm1 addq $0x10, %rax jmp 0x138e7c vpunpckldq %xmm7, %xmm8, %xmm9 # xmm9 = xmm8[0],xmm7[0],xmm8[1],xmm7[1] vpunpckldq %xmm5, %xmm6, %xmm10 # xmm10 = xmm6[0],xmm5[0],xmm6[1],xmm5[1] vpunpckhdq %xmm7, %xmm8, %xmm7 # xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3] vpunpckhdq %xmm5, %xmm6, %xmm5 # xmm5 = xmm6[2],xmm5[2],xmm6[3],xmm5[3] vpunpckldq %xmm3, %xmm4, %xmm6 # xmm6 = xmm4[0],xmm3[0],xmm4[1],xmm3[1] vpunpckldq %xmm1, %xmm2, %xmm8 # xmm8 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] vpunpckhdq %xmm3, %xmm4, %xmm3 # xmm3 = xmm4[2],xmm3[2],xmm4[3],xmm3[3] vpunpckhdq %xmm1, %xmm2, %xmm1 # xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3] vpunpcklqdq %xmm10, %xmm9, %xmm2 # xmm2 = xmm9[0],xmm10[0] vpunpckhqdq %xmm10, %xmm9, %xmm4 # xmm4 = xmm9[1],xmm10[1] vpaddd %xmm4, %xmm2, %xmm2 vpunpcklqdq %xmm5, %xmm7, %xmm4 # xmm4 = xmm7[0],xmm5[0] vpunpckhqdq %xmm5, %xmm7, %xmm5 # xmm5 = xmm7[1],xmm5[1] vpaddd %xmm4, %xmm5, %xmm4 vpaddd %xmm4, %xmm2, %xmm2 vpunpcklqdq %xmm8, %xmm6, %xmm4 # xmm4 = xmm6[0],xmm8[0] vpunpckhqdq %xmm8, %xmm6, %xmm5 # xmm5 = xmm6[1],xmm8[1] vpaddd %xmm5, %xmm4, %xmm4 vpunpcklqdq %xmm1, %xmm3, %xmm5 # xmm5 = xmm3[0],xmm1[0] vmovd %xmm2, (%rdi) vpextrd $0x1, %xmm2, (%r8) vpextrd $0x2, %xmm2, (%r9) vpextrd $0x3, %xmm2, (%r10) vpunpckhqdq %xmm1, %xmm3, %xmm1 # xmm1 = xmm3[1],xmm1[1] vpaddd %xmm5, %xmm1, %xmm1 vpaddd %xmm1, %xmm4, %xmm1 vmovd %xmm1, 0x4(%rdi) vpextrd $0x1, %xmm1, 0x4(%r8) vpextrd $0x2, %xmm1, 0x4(%r9) vpextrd $0x3, %xmm1, 0x4(%r10) addq $0x8, %rdi addq $0x8, %r8 addq $0x8, %r9 addq $0x8, %r10 addq $0x2, %r11 addq %rbp, %r12 jmp 0x138e46 movq 0x30(%rsp), %r14 movq 0x40(%rsp), %r12 movq 0x10(%rsp), %rax movq 0x40(%rax), %rdx imulq %rsi, %rdx imulq 0x10(%rax), %rdx addq (%rax), %rdx imulq 0x70(%rsp), %r12 cmpl %r13d, %r11d jge 0x1390a2 movl %r11d, %ecx shrl %ecx movl %r11d, %eax andl $0x1, %eax addl %ecx, %eax imulq %r12, %rax addq %r14, %rax vpxor %xmm1, %xmm1, %xmm1 xorl %ecx, %ecx movq %rdx, %rbp vpxor %xmm2, %xmm2, %xmm2 vpxor %xmm3, %xmm3, %xmm3 vpxor %xmm4, %xmm4, %xmm4 cmpl %ecx, %r15d je 0x139048 vpmovsxbw (%rax,%rcx,8), %xmm5 vmovdqu (%rbp), %xmm6 vmovdqu 0x10(%rbp), %xmm7 vpcmpgtb %xmm6, %xmm0, %xmm8 vpcmpgtb %xmm7, %xmm0, %xmm9 vpunpcklbw %xmm8, %xmm6, %xmm10 # xmm10 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3],xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7] vpunpckhbw %xmm8, %xmm6, %xmm6 # xmm6 = xmm6[8],xmm8[8],xmm6[9],xmm8[9],xmm6[10],xmm8[10],xmm6[11],xmm8[11],xmm6[12],xmm8[12],xmm6[13],xmm8[13],xmm6[14],xmm8[14],xmm6[15],xmm8[15] vpunpcklbw %xmm9, %xmm7, %xmm8 # xmm8 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3],xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7] vpunpckhbw %xmm9, %xmm7, %xmm7 # xmm7 = xmm7[8],xmm9[8],xmm7[9],xmm9[9],xmm7[10],xmm9[10],xmm7[11],xmm9[11],xmm7[12],xmm9[12],xmm7[13],xmm9[13],xmm7[14],xmm9[14],xmm7[15],xmm9[15] vpmadcswd %xmm1, %xmm10, %xmm5, %xmm1 vpmadcswd %xmm2, %xmm6, %xmm5, %xmm2 vpmadcswd %xmm3, %xmm8, %xmm5, %xmm3 vpmadcswd %xmm4, %xmm7, %xmm5, %xmm4 addq $0x20, %rbp incq %rcx jmp 0x138ff6 vpunpckldq %xmm2, %xmm1, %xmm5 # xmm5 = xmm1[0],xmm2[0],xmm1[1],xmm2[1] vpunpckldq %xmm4, %xmm3, %xmm6 # xmm6 = xmm3[0],xmm4[0],xmm3[1],xmm4[1] vpunpckhdq %xmm2, %xmm1, %xmm1 # xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3] vpunpckhdq %xmm4, %xmm3, %xmm2 # xmm2 = xmm3[2],xmm4[2],xmm3[3],xmm4[3] vpunpcklqdq %xmm6, %xmm5, %xmm3 # xmm3 = xmm5[0],xmm6[0] vpunpckhqdq %xmm6, %xmm5, %xmm4 # xmm4 = xmm5[1],xmm6[1] vpaddd %xmm4, %xmm3, %xmm3 vpunpcklqdq %xmm2, %xmm1, %xmm4 # xmm4 = xmm1[0],xmm2[0] vpunpckhqdq %xmm2, %xmm1, %xmm1 # xmm1 = xmm1[1],xmm2[1] vpaddd %xmm4, %xmm1, %xmm1 vpaddd %xmm1, %xmm3, %xmm1 vmovd %xmm1, (%rdi) vpextrd $0x1, %xmm1, (%r8) vpextrd $0x2, %xmm1, (%r9) vpextrd $0x3, %xmm1, (%r10) addq $0x4, %rdi addq $0x4, %r8 addq $0x4, %r9 addq $0x4, %r10 incl %r11d jmp 0x138fc4 incq %rsi jmp 0x138dcc movq 0x18(%rsp), %rcx andq $-0x4, %rcx movq 0x28(%rsp), %rax movq 0x10(%rax), %rsi imulq 0x40(%rax), %rsi movq (%rax), %rdi movl %r15d, %r8d pushq $0x4 popq %r9 vpxor %xmm0, %xmm0, %xmm0 cmpq 0x18(%rsp), %rcx jge 0x139213 movq 0x30(%rsp), %r11 movq 0x40(%rsp), %r14 imulq 0x70(%rsp), %r14 movl %ecx, %eax cltd idivl %r9d movq %rsi, %r10 imulq %rcx, %r10 addl %eax, %edx movslq %edx, %rax movq 0x10(%rsp), %rdx movq 0x40(%rdx), %r15 imulq %rax, %r15 imulq 0x10(%rdx), %r15 addq %rdi, %r10 addq (%rdx), %r15 xorl %edx, %edx movq %rdx, %rbx orq $0x1, %rbx cmpq %r13, %rbx jge 0x139196 movq %rdx, %r12 shrq %r12 imulq %r14, %r12 addq %r11, %r12 vpxor %xmm1, %xmm1, %xmm1 xorl %ebp, %ebp vpxor %xmm2, %xmm2, %xmm2 cmpl %ebp, %r8d je 0x13916b vmovdqu (%r12), %xmm3 vpcmpgtb %xmm3, %xmm0, %xmm4 vpunpcklbw %xmm4, %xmm3, %xmm5 # xmm5 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7] vpunpckhbw %xmm4, %xmm3, %xmm3 # xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15] vpmovsxbw (%r15,%rbp,8), %xmm4 vpmadcswd %xmm1, %xmm4, %xmm5, %xmm1 vpmadcswd %xmm2, %xmm4, %xmm3, %xmm2 addq $0x10, %r12 incq %rbp jmp 0x139139 vpshufd $0xee, %xmm1, %xmm3 # xmm3 = xmm1[2,3,2,3] vpaddd %xmm1, %xmm3, %xmm1 vpshufd $0xee, %xmm2, %xmm3 # xmm3 = xmm2[2,3,2,3] vpaddd %xmm2, %xmm3, %xmm2 vphaddd %xmm2, %xmm1, %xmm1 vpshufd $0xe8, %xmm1, %xmm1 # xmm1 = xmm1[0,2,2,3] vmovq %xmm1, (%r10) addq $0x8, %r10 addq $0x2, %rdx jmp 0x139116 movq 0x30(%rsp), %r11 movq 0x40(%rsp), %r14 movq 0x10(%rsp), %rbx imulq 0x40(%rbx), %rax imulq 0x10(%rbx), %rax addq (%rbx), %rax imulq 0x70(%rsp), %r14 cmpl %r13d, %edx jge 0x13920b movl %edx, %ebx shrl %ebx movl %edx, %r15d andl $0x1, %r15d addl %ebx, %r15d imulq %r14, %r15 addq %r11, %r15 vpxor %xmm1, %xmm1, %xmm1 xorl %ebx, %ebx cmpl %ebx, %r8d je 0x1391f4 vpmovsxbw (%r15,%rbx,8), %xmm2 vpmovsxbw (%rax,%rbx,8), %xmm3 vpmadcswd %xmm1, %xmm3, %xmm2, %xmm1 incq %rbx jmp 0x1391d8 vphaddd %xmm1, %xmm1, %xmm1 vphaddd %xmm1, %xmm1, %xmm1 vmovd %xmm1, (%r10) addq $0x4, %r10 incl %edx jmp 0x1391b8 incq %rcx jmp 0x1390cf movq 0x38(%rsp), %rax testq %rax, %rax je 0x139241 lock decl (%rax) jne 0x139241 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x139239 movq (%rdi), %rax callq 0x18(%rax) jmp 0x139241 movq %rsi, %rdi callq 0x5f3e0 addq $0x88, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x13928e movq %rax, %rbx movq 0x38(%rsp), %rax testq %rax, %rax je 0x139286 lock decl (%rax) jne 0x139286 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi jne 0x139280 movq %rsi, %rdi callq 0x5f3e0 jmp 0x139286 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/x86/convolution_x86_xop.cpp
ncnn::crop_pack4_sse(ncnn::Mat const&, ncnn::Mat&, int, int)	static void crop_pack4_sse(const Mat& src, Mat& dst, int top, int left) { int w = dst.w; int h = dst.h; int right = src.w - dst.w - left; const float* ptr = src.row(top) + left * 4; float* outptr = dst; for (int y = 0; y < h; y++) { for (int x = 0; x < w; x++) { __m128 _p = _mm_loadu_ps(ptr); _mm_storeu_ps(outptr, _p); ptr += 4; outptr += 4; } ptr += (left + right) * 4; } }	movl 0x2c(%rsi), %eax movl 0x30(%rsi), %r8d movslq 0x2c(%rdi), %r9 movslq %edx, %rdx imulq %r9, %rdx subl %eax, %r9d imulq 0x10(%rdi), %rdx addq (%rdi), %rdx shll $0x2, %ecx movslq %ecx, %rcx leaq (%rdx,%rcx,4), %rcx movq (%rsi), %rdx shll $0x2, %r9d movslq %r9d, %rsi xorl %edi, %edi testl %eax, %eax cmovlel %edi, %eax testl %r8d, %r8d cmovlel %edi, %r8d shlq $0x2, %rsi movl %eax, %r9d cmpl %r8d, %edi je 0x1409f4 subl $0x1, %r9d jb 0x1409ed movups (%rcx), %xmm0 movups %xmm0, (%rdx) addq $0x10, %rcx addq $0x10, %rdx jmp 0x1409d7 addq %rsi, %rcx incl %edi jmp 0x1409cf retq nop	/csukuangfj[P]ncnn/src/layer/x86/crop_x86.cpp
ncnn::Deconvolution_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Deconvolution_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { // deconvolv with NxN kernel // value = value + bias int w = bottom_blob.w; int h = bottom_blob.h; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; // NCNN_LOGE("Deconvolution input %d x %d pad = %d %d ksize=%d %d stride=%d %d", w, h, pad_w, pad_h, kernel_w, kernel_h, stride_w, stride_h); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; int outw = (w - 1) * stride_w + kernel_extent_w + output_pad_right; int outh = (h - 1) * stride_h + kernel_extent_h + output_pad_bottom; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; int out_channels = num_output / out_elempack; Mat top_blob_bordered; if (pad_left > 0 \|\| pad_right > 0 \|\| pad_top > 0 \|\| pad_bottom > 0 \|\| (output_w > 0 && output_h > 0)) { top_blob_bordered.create(outw, outh, out_channels, out_elemsize, out_elempack, opt.workspace_allocator); } else { top_blob_bordered = top_blob; top_blob_bordered.create(outw, outh, out_channels, out_elemsize, out_elempack, opt.blob_allocator); } if (top_blob_bordered.empty()) return -100; const int maxk = kernel_w * kernel_h; if (opt.use_sgemm_convolution) { // sgemm Mat bottom_blob_2 = bottom_blob; { bottom_blob_2.w = bottom_blob.w * bottom_blob.h; bottom_blob_2.h = 1; } Mat top_col2im; Option opt_b = opt; opt_b.blob_allocator = top_blob_bordered.allocator; gemm->forward(bottom_blob_2, top_col2im, opt_b); { // col2im const int gap = (outw * stride_h - w * stride_w) * out_elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (out_elempack == 16) { #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < out_channels; p++) { const float* sptr = top_col2im.row(p * maxk); Mat outm = top_blob_bordered.channel(p); if (bias_data.empty()) { outm.fill(_mm512_setzero_ps()); } else { outm.fill(_mm512_load_ps((const float)bias_data + p 16)); } for (int u = 0; u < kernel_h; u++) { for (int v = 0; v < kernel_w; v++) { float* ptr = outm.row(dilation_h * u) + dilation_w * v * 16; for (int i = 0; i < h; i++) { for (int j = 0; j < w; j++) { __m512 _val = _mm512_load_ps(ptr); __m512 _s = _mm512_load_ps(sptr); _val = _mm512_add_ps(_val, _s); _mm512_store_ps(ptr, _val); ptr += stride_w * 16; sptr += 16; } ptr += gap; } } } } } #endif // __AVX512F__ if (out_elempack == 8) { #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < out_channels; p++) { const float* sptr = top_col2im.row(p * maxk); Mat outm = top_blob_bordered.channel(p); if (bias_data.empty()) { outm.fill(_mm256_setzero_ps()); } else { outm.fill(_mm256_load_ps((const float)bias_data + p 8)); } for (int u = 0; u < kernel_h; u++) { for (int v = 0; v < kernel_w; v++) { float* ptr = outm.row(dilation_h * u) + dilation_w * v * 8; for (int i = 0; i < h; i++) { for (int j = 0; j < w; j++) { __m256 _val = _mm256_load_ps(ptr); __m256 _s = _mm256_load_ps(sptr); _val = _mm256_add_ps(_val, _s); _mm256_store_ps(ptr, _val); ptr += stride_w * 8; sptr += 8; } ptr += gap; } } } } } #endif // __AVX__ if (out_elempack == 4) { #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < out_channels; p++) { const float* sptr = top_col2im.row(p * maxk); Mat outm = top_blob_bordered.channel(p); if (bias_data.empty()) { outm.fill(_mm_setzero_ps()); } else { outm.fill(_mm_load_ps((const float)bias_data + p 4)); } for (int u = 0; u < kernel_h; u++) { for (int v = 0; v < kernel_w; v++) { float* ptr = outm.row(dilation_h * u) + dilation_w * v * 4; for (int i = 0; i < h; i++) { for (int j = 0; j < w; j++) { __m128 _val = _mm_load_ps(ptr); __m128 _s = _mm_load_ps(sptr); _val = _mm_add_ps(_val, _s); _mm_store_ps(ptr, _val); ptr += stride_w * 4; sptr += 4; } ptr += gap; } } } } } #endif // __SSE2__ if (out_elempack == 1) { #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < out_channels; p++) { const float* sptr = top_col2im.row(p * maxk); Mat outm = top_blob_bordered.channel(p); const float bias = bias_data.empty() ? 0.f : bias_data[p]; outm.fill(bias); for (int u = 0; u < kernel_h; u++) { for (int v = 0; v < kernel_w; v++) { float* ptr = outm.row(dilation_h * u) + dilation_w * v; for (int i = 0; i < h; i++) { for (int j = 0; j < w; j++) { ptr[0] += sptr[0]; ptr += stride_w; sptr += 1; } ptr += gap; } } } } } } if (activation) { activation->forward_inplace(top_blob_bordered, opt); } } else { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16 && out_elempack == 16) { deconvolution_pack16_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 8 && out_elempack == 16) { deconvolution_pack8to16_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 16 && out_elempack == 8) { deconvolution_pack16to8_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 4 && out_elempack == 16) { deconvolution_pack4to16_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 16 && out_elempack == 4) { deconvolution_pack16to4_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 1 && out_elempack == 16) { deconvolution_pack1to16_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 16 && out_elempack == 1) { deconvolution_pack16to1_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } #endif // __AVX512F__ if (elempack == 8 && out_elempack == 8) { deconvolution_pack8_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 4 && out_elempack == 8) { deconvolution_pack4to8_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 8 && out_elempack == 4) { deconvolution_pack8to4_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 1 && out_elempack == 8) { deconvolution_pack1to8_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 8 && out_elempack == 1) { deconvolution_pack8to1_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } #endif // __AVX__ if (elempack == 4 && out_elempack == 4) { deconvolution_pack4_sse(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 1 && out_elempack == 4) { deconvolution_pack1to4_sse(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 4 && out_elempack == 1) { deconvolution_pack4to1_sse(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } #endif // __SSE2__ if (elempack == 1 && out_elempack == 1) { // num_output #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < num_output; p++) { float* outptr = top_blob_bordered.channel(p); // shadowed variable for less openmp task args const int w = bottom_blob.w; const int h = bottom_blob.h; const int channels = bottom_blob.c; const int outw = top_blob_bordered.w; const int outh = top_blob_bordered.h; for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { float sum = 0.f; if (bias_term) { sum = bias_data[p]; } const float* kptr = (const float)weight_data_tm.channel(p); // channels for (int q = 0; q < channels; q++) { const Mat m = bottom_blob.channel(q); for (int y = 0; y < kernel_h; y++) { int sys = (i + y dilation_h - (kernel_extent_h - 1)); if (sys < 0 \|\| sys % stride_h != 0) continue; int sy = sys / stride_h; if (sy >= h) continue; const float* sptr = m.row(sy); for (int x = 0; x < kernel_w; x++) { int sxs = (j + x * dilation_w - (kernel_extent_w - 1)); if (sxs < 0 \|\| sxs % stride_w != 0) continue; int sx = sxs / stride_w; if (sx >= w) continue; float val = sptr[sx]; int k = y * kernel_w + x; float w = kptr[k]; sum += val * w; } } kptr += maxk; } sum = activation_ss(sum, activation_type, activation_params); outptr[j] = sum; } outptr += outw; } } } } cut_padding(top_blob_bordered, top_blob, opt); if (top_blob.empty()) return -100; return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x258, %rsp # imm = 0x258 movq %rdx, 0x160(%rsp) movq %rdi, %r15 movq (%rdi), %rdi movq -0x18(%rdi), %rax movl 0xd4(%r15,%rax), %edx decl %edx imull 0xdc(%r15,%rax), %edx movl %edx, 0x144(%rsp) movl 0x2c(%rsi), %r13d leal -0x1(%r13), %r12d imull 0xe4(%r15,%rax), %r12d movl 0xd8(%r15,%rax), %edx decl %edx imull 0xe0(%r15,%rax), %edx movl %edx, 0x140(%rsp) movl 0x30(%rsi), %r14d leal -0x1(%r14), %ebp imull 0xe8(%r15,%rax), %ebp movl 0xd0(%r15,%rax), %r8d movl 0xfc(%r15,%rax), %r11d movl 0x100(%r15,%rax), %r10d testb $0x3, %r8b sete %r9b movq %rcx, 0x130(%rsp) andb 0x27(%rcx), %r9b movq 0x10(%rsi), %rax movq %rsi, 0x70(%rsp) movslq 0x18(%rsi), %rsi xorl %edx, %edx movq %rsi, 0x148(%rsp) divq %rsi movq %rax, %rbx movb %r9b, 0xf(%rsp) movzbl %r9b, %ecx leal (%rcx,%rcx,2), %r9d incl %r9d addb %cl, %cl shlq %cl, %rbx movl %r8d, %eax cltd idivl %r9d movq %rax, 0xa0(%rsp) andq $0x0, 0x120(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0xe0(%rsp) movups %xmm0, 0xec(%rsp) movaps %xmm0, 0x100(%rsp) movups %xmm0, 0x10c(%rsp) movq -0x18(%rdi), %rax cmpl $0x0, 0xec(%r15,%rax) pushq $0x10 popq %rsi movq %rcx, 0x10(%rsp) jg 0x154c2e cmpl $0x0, 0xf0(%r15,%rax) jg 0x154c2e cmpl $0x0, 0xf4(%r15,%rax) jg 0x154c2e cmpl $0x0, 0xf8(%r15,%rax) jle 0x156a97 addl 0x144(%rsp), %r12d addl 0x140(%rsp), %ebp addl %r11d, %r12d incl %r12d leal 0x1(%r10,%rbp), %edx movq 0x130(%rsp), %rax movq (%rax,%rsi), %rax movq %rax, (%rsp) leaq 0xe0(%rsp), %rdi movl %r12d, %esi movq 0xa0(%rsp), %rcx movq %rbx, %r8 callq 0x628f2 pushq $-0x64 popq %rbx cmpq $0x0, 0xe0(%rsp) je 0x156a4c movslq 0x118(%rsp), %rcx movq 0x120(%rsp), %rax imulq %rcx, %rax testq %rax, %rax je 0x156a4c movq %rcx, %rsi movq (%r15), %rdx movq -0x18(%rdx), %rax movslq 0xd4(%r15,%rax), %rcx movq %rax, 0x40(%rsp) movl 0xd8(%r15,%rax), %eax movl %eax, 0x48(%rsp) movq %rcx, %rdi imull %ecx, %eax movq 0x130(%rsp), %rcx cmpb $0x1, 0x1d(%rcx) cltq movq %rax, 0x158(%rsp) jne 0x154fda movq 0x70(%rsp), %rdi movq (%rdi), %rax movq %rax, 0x178(%rsp) movq 0x8(%rdi), %rcx movq %rcx, 0x180(%rsp) movq 0x10(%rdi), %rax movq %rax, 0x188(%rsp) movl 0x18(%rdi), %eax movl %eax, 0x190(%rsp) movq 0x20(%rdi), %rax movq %rax, 0x198(%rsp) movups 0x28(%rdi), %xmm0 movl 0x2c(%rdi), %edx movl 0x30(%rdi), %eax movups %xmm0, 0x1a0(%rsp) movl 0x38(%rdi), %esi movl %esi, 0x1b0(%rsp) movq 0x40(%rdi), %rsi movq %rsi, 0x1b8(%rsp) testq %rcx, %rcx je 0x154d5b lock incl (%rcx) movl 0x2c(%rdi), %edx movl 0x30(%rdi), %eax imull %edx, %eax leaq 0x178(%rsp), %rsi leaq 0x1c0(%rsp), %rdx andq $0x0, 0x40(%rdx) movl %eax, 0x2c(%rsi) movl $0x1, 0x30(%rsi) xorps %xmm0, %xmm0 movaps %xmm0, (%rdx) movups %xmm0, 0xc(%rdx) movaps %xmm0, 0x20(%rdx) movups %xmm0, 0x2c(%rdx) movq 0x130(%rsp), %rax movups (%rax), %xmm0 movups 0x10(%rax), %xmm1 movups 0x20(%rax), %xmm2 movups 0x30(%rax), %xmm3 leaq 0x210(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x100(%rsp), %rax movq %rax, 0x8(%rcx) movq 0x10(%r15), %rdi movq (%rdi), %rax callq 0x38(%rax) movq (%r15), %rsi movq -0x18(%rsi), %r8 imull 0xe8(%r15,%r8), %r12d movl 0xe4(%r15,%r8), %eax imull %r13d, %eax subl %eax, %r12d movq 0x10(%rsp), %rcx shll %cl, %r12d movslq %r12d, %rdi cmpb $0x0, 0xf(%rsp) je 0x156751 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d testl %r14d, %r14d cmovlel %ecx, %r14d movq 0xa0(%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, 0xa0(%rsp) xorps %xmm0, %xmm0 xorl %edx, %edx cmpq 0xa0(%rsp), %rdx je 0x15697a movq %rdx, %rax imulq 0x158(%rsp), %rax movslq 0x1ec(%rsp), %r10 imulq %rax, %r10 imulq 0x1d0(%rsp), %r10 addq 0x1c0(%rsp), %r10 movslq 0x10c(%rsp), %r12 movslq 0x110(%rsp), %rcx movq 0x120(%rsp), %r11 movq %rdx, 0x10(%rsp) imulq %rdx, %r11 movq 0xf0(%rsp), %rbx imulq %rbx, %r11 addq 0xe0(%rsp), %r11 movl 0x114(%rsp), %ebp imulq %r12, %rcx movq %rbx, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rbx cmpl $0x4, 0x108(%rsp) cmoveq %rcx, %rax movq 0x1a8(%r15,%r8), %rcx testq %rcx, %rcx je 0x154f17 movslq 0x1e0(%r15,%r8), %rdx imulq 0x1e8(%r15,%r8), %rdx testq %rdx, %rdx je 0x154f17 movq 0x10(%rsp), %rdx shlq $0x4, %rdx movaps (%rcx,%rdx), %xmm1 imull %eax, %ebp testl %ebp, %ebp movl $0x0, %eax cmovlel %eax, %ebp movq %r11, %rax subl $0x1, %ebp jb 0x154f35 movups %xmm1, (%rax) addq $0x10, %rax jmp 0x154f09 imull %eax, %ebp testl %ebp, %ebp movl $0x0, %eax cmovlel %eax, %ebp movq %r11, %rax subl $0x1, %ebp jb 0x154f35 movups %xmm0, (%rax) addq $0x10, %rax jmp 0x154f27 imulq %r12, %rbx movq -0x18(%rsi), %r8 xorl %eax, %eax cmpl 0xd8(%r15,%r8), %eax jge 0x154fcd xorl %edx, %edx cmpl 0xd4(%r15,%r8), %edx jge 0x154fc6 movslq 0xe0(%r15,%r8), %rcx movslq %eax, %r12 imulq %rcx, %r12 imulq %rbx, %r12 addq %r11, %r12 movl 0xdc(%r15,%r8), %ecx imull %edx, %ecx shll $0x2, %ecx movslq %ecx, %rcx leaq (%r12,%rcx,4), %r12 xorl %ebp, %ebp cmpl %r14d, %ebp je 0x154fc2 movl %r13d, %ecx subl $0x1, %ecx jb 0x154fba movaps (%r10), %xmm1 addps (%r12), %xmm1 movaps %xmm1, (%r12) movq -0x18(%rsi), %r8 movslq 0xe4(%r15,%r8), %r9 shlq $0x4, %r9 addq %r9, %r12 addq $0x10, %r10 jmp 0x154f8e leaq (%r12,%rdi,4), %r12 incl %ebp jmp 0x154f86 incl %edx jmp 0x154f4f incl %eax jmp 0x154f3f movq 0x10(%rsp), %rdx incq %rdx jmp 0x154e32 movq %rdx, 0x150(%rsp) cmpl $0x4, 0x148(%rsp) setne %al movb 0xf(%rsp), %cl xorb $0x1, %cl movb %cl, 0xf(%rsp) orb %cl, %al movq 0x70(%rsp), %rdx movl 0x48(%rsp), %r13d jne 0x15572e movq %rdi, %r14 movl %r14d, %eax movq 0x40(%rsp), %rcx movl 0xdc(%r15,%rcx), %r11d movl 0xe0(%r15,%rcx), %edi movl %edi, 0x50(%rsp) movl 0xe4(%r15,%rcx), %edi movl 0xe8(%r15,%rcx), %r8d movl %r8d, 0xa0(%rsp) movl 0x114(%r15,%rcx), %r9d movq 0x1a8(%r15,%rcx), %r10 movl %r14d, %ecx imull %r13d, %ecx shll $0x4, %ecx xorl %ebx, %ebx testl %r14d, %r14d movl $0x0, %r8d cmovgl %r14d, %r8d testl %r13d, %r13d movl $0x0, %r12d cmovgl %r13d, %r12d movq %r12, 0x30(%rsp) movslq %ecx, %rcx testl %esi, %esi cmovlel %ebx, %esi movq %rsi, 0x138(%rsp) xorl %esi, %esi shlq $0x2, %rcx movq %rcx, 0xc8(%rsp) shlq $0x6, %r14 pushq $0x1 popq %rcx subl %eax, %ecx imull %r11d, %ecx movl %ecx, 0x60(%rsp) shlq $0x6, %r8 decl %r9d movq %r9, 0xc0(%rsp) xorps %xmm0, %xmm0 movaps 0x29bf4b(%rip), %xmm15 # 0x3f1010 movaps 0x29bf53(%rip), %xmm9 # 0x3f1020 movaps 0x29bf5b(%rip), %xmm8 # 0x3f1030 movaps 0x298fc4(%rip), %xmm5 # 0x3ee0a0 movaps 0x29bf5d(%rip), %xmm6 # 0x3f1040 movq %r10, 0xb8(%rsp) movaps 0x29bf7d(%rip), %xmm11 # 0x3f1070 movaps 0x29bf85(%rip), %xmm12 # 0x3f1080 movaps 0x29bf8d(%rip), %xmm13 # 0x3f1090 cmpq 0x138(%rsp), %rsi je 0x15572e movq 0x120(%rsp), %rax imulq %rsi, %rax imulq 0xf0(%rsp), %rax addq 0xe0(%rsp), %rax movl 0x2c(%rdx), %r12d movl 0x30(%rdx), %ecx movl %ecx, 0x38(%rsp) movl 0x38(%rdx), %ecx movq %rsi, 0x80(%rsp) shlq $0x4, %rsi movq %rsi, 0x78(%rsp) xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx movq %rcx, 0x58(%rsp) movl 0x10c(%rsp), %ecx testl %ecx, %ecx cmovlel %edx, %ecx movl %ecx, 0x68(%rsp) movl 0x110(%rsp), %ecx testl %ecx, %ecx cmovlel %edx, %ecx movl %ecx, 0x64(%rsp) movl 0x50(%rsp), %esi cmpl 0x64(%rsp), %edx je 0x155719 xorl %ecx, %ecx movl 0x60(%rsp), %r9d movl %r9d, 0x20(%rsp) movl %edx, 0x10(%rsp) cmpl 0x68(%rsp), %ecx je 0x155712 testq %r10, %r10 movq %rax, 0x18(%rsp) movl %ecx, 0x28(%rsp) je 0x1551bb movq 0x78(%rsp), %rax movups (%r10,%rax), %xmm1 jmp 0x1551be xorps %xmm1, %xmm1 movq 0x18(%r15), %rax movq 0x70(%rsp), %rdx movslq 0x2c(%rdx), %rbx movq (%rdx), %rcx movq %rcx, 0x90(%rsp) movq 0x10(%rdx), %rcx movq 0x40(%rdx), %rdx imulq %rcx, %rdx movq %rdx, 0x88(%rsp) imulq %rcx, %rbx movq 0x58(%r15), %rcx imulq 0x80(%rsp), %rcx imulq 0x28(%r15), %rcx leaq (%rax,%rcx), %r9 addq $0x30, %r9 xorl %eax, %eax movq 0x30(%rsp), %rcx cmpq 0x58(%rsp), %rax je 0x155338 movq 0x88(%rsp), %rbp movq %rax, 0x98(%rsp) imulq %rax, %rbp addq 0x90(%rsp), %rbp movq %r9, 0xd0(%rsp) xorl %r10d, %r10d cmpq %rcx, %r10 je 0x155318 movl %r10d, %eax subl %r13d, %eax incl %eax imull %esi, %eax addl 0x10(%rsp), %eax js 0x15530d cltd idivl 0xa0(%rsp) testl %edx, %edx jne 0x15530d cmpl 0x38(%rsp), %eax jge 0x15530d movq %r14, %r13 movslq %eax, %rsi imulq %rbx, %rsi addq %rbp, %rsi movl 0x20(%rsp), %ecx xorl %r14d, %r14d cmpq %r14, %r8 je 0x1552fc testl %ecx, %ecx js 0x1552f3 movl %ecx, %eax cltd idivl %edi testl %edx, %edx jne 0x1552f3 cmpl %r12d, %eax jge 0x1552f3 shll $0x2, %eax cltq movss (%rsi,%rax,4), %xmm2 movss 0x4(%rsi,%rax,4), %xmm3 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps -0x30(%r9,%r14), %xmm2 movss 0x8(%rsi,%rax,4), %xmm7 addps %xmm1, %xmm2 movss 0xc(%rsi,%rax,4), %xmm1 shufps $0x0, %xmm3, %xmm3 # xmm3 = xmm3[0,0,0,0] shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] shufps $0x0, %xmm1, %xmm1 # xmm1 = xmm1[0,0,0,0] mulps -0x20(%r9,%r14), %xmm3 mulps -0x10(%r9,%r14), %xmm7 addps %xmm3, %xmm7 addps %xmm2, %xmm7 mulps (%r9,%r14), %xmm1 addps %xmm7, %xmm1 addq $0x40, %r14 addl %r11d, %ecx jmp 0x15528d movq %r13, %r14 movl 0x48(%rsp), %r13d movl 0x50(%rsp), %esi movq 0x30(%rsp), %rcx incq %r10 addq %r14, %r9 jmp 0x155241 movq 0x98(%rsp), %rax incq %rax movq 0xd0(%rsp), %r9 addq 0xc8(%rsp), %r9 jmp 0x15520f movq 0xc0(%rsp), %rax cmpl $0x5, %eax ja 0x1556fc leaq 0x29dd18(%rip), %rcx # 0x3f3068 movslq (%rcx,%rax,4), %rax addq %rcx, %rax movaps %xmm1, %xmm7 maxps %xmm0, %xmm7 movq 0xb8(%rsp), %r10 movl 0x10(%rsp), %edx jmpq %rax movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax minps %xmm0, %xmm1 movss (%rax), %xmm2 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm2 addps %xmm2, %xmm7 jmp 0x1556e1 movaps %xmm1, %xmm2 minps %xmm15, %xmm2 maxps %xmm9, %xmm2 movaps %xmm2, %xmm14 mulps %xmm8, %xmm14 addps %xmm5, %xmm14 cvttps2dq %xmm14, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm14 andps %xmm6, %xmm14 subps %xmm14, %xmm7 movaps %xmm15, %xmm8 cvttps2dq %xmm7, %xmm15 movaps 0x29bc86(%rip), %xmm0 # 0x3f1050 mulps %xmm0, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm7 mulps %xmm2, %xmm7 movaps %xmm2, %xmm14 movaps 0x29bc7e(%rip), %xmm10 # 0x3f1060 mulps %xmm10, %xmm14 addps %xmm11, %xmm14 mulps %xmm2, %xmm14 addps %xmm12, %xmm14 mulps %xmm2, %xmm14 addps %xmm13, %xmm14 mulps %xmm2, %xmm14 xorps %xmm4, %xmm4 movaps %xmm5, %xmm0 movaps 0x29bc95(%rip), %xmm5 # 0x3f10a0 addps %xmm5, %xmm14 mulps %xmm2, %xmm14 addps %xmm0, %xmm14 mulps %xmm7, %xmm14 addps %xmm6, %xmm2 addps %xmm14, %xmm2 pslld $0x17, %xmm15 paddd %xmm6, %xmm15 mulps %xmm2, %xmm15 addps %xmm6, %xmm15 movaps %xmm15, %xmm2 maxps 0x29bc6f(%rip), %xmm15 # 0x3f10b0 movaps %xmm15, %xmm7 andps 0x29bc73(%rip), %xmm15 # 0x3f10c0 orps %xmm0, %xmm15 movaps %xmm15, %xmm3 cmpltps 0x29bc83(%rip), %xmm3 # 0x3f10e0 movaps %xmm3, %xmm14 andps %xmm15, %xmm14 movaps 0x29bc83(%rip), %xmm10 # 0x3f10f0 addps %xmm10, %xmm15 addps %xmm14, %xmm15 cmpleps %xmm4, %xmm2 psrld $0x17, %xmm7 paddd 0x29bc4a(%rip), %xmm7 # 0x3f10d0 cvtdq2ps %xmm7, %xmm14 andps %xmm6, %xmm3 subps %xmm3, %xmm14 movaps %xmm15, %xmm7 mulps 0x29bc64(%rip), %xmm7 # 0x3f1100 addps 0x29bc6d(%rip), %xmm7 # 0x3f1110 mulps %xmm15, %xmm7 addps 0x29bc72(%rip), %xmm7 # 0x3f1120 mulps %xmm15, %xmm7 addps 0x29bc77(%rip), %xmm7 # 0x3f1130 mulps %xmm15, %xmm7 addps 0x29bc7c(%rip), %xmm7 # 0x3f1140 mulps %xmm15, %xmm7 addps 0x29bc81(%rip), %xmm7 # 0x3f1150 mulps %xmm15, %xmm7 addps 0x29bc86(%rip), %xmm7 # 0x3f1160 mulps %xmm15, %xmm7 addps 0x29bc8b(%rip), %xmm7 # 0x3f1170 mulps %xmm15, %xmm7 addps 0x29bc90(%rip), %xmm7 # 0x3f1180 mulps %xmm15, %xmm7 movaps 0x29bb55(%rip), %xmm4 # 0x3f1050 mulps %xmm4, %xmm14 addps %xmm15, %xmm14 mulps %xmm15, %xmm15 addps 0x29bc82(%rip), %xmm7 # 0x3f1190 mulps %xmm15, %xmm7 movaps %xmm8, %xmm15 movaps 0x29bb12(%rip), %xmm8 # 0x3f1030 addps %xmm7, %xmm14 mulps 0x29c796(%rip), %xmm14 # 0x3f1cc0 movaps %xmm2, %xmm3 andnps %xmm14, %xmm3 andps 0x29c798(%rip), %xmm2 # 0x3f1cd0 orps %xmm3, %xmm2 minps %xmm15, %xmm2 maxps %xmm9, %xmm2 movaps %xmm2, %xmm3 mulps %xmm8, %xmm3 addps %xmm0, %xmm3 cvttps2dq %xmm3, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm3 andps %xmm6, %xmm3 subps %xmm3, %xmm7 cvttps2dq %xmm7, %xmm14 mulps %xmm4, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm3 mulps %xmm2, %xmm3 movaps %xmm2, %xmm7 mulps 0x29bae7(%rip), %xmm7 # 0x3f1060 addps %xmm11, %xmm7 mulps %xmm2, %xmm7 addps %xmm12, %xmm7 mulps %xmm2, %xmm7 addps %xmm13, %xmm7 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 movaps %xmm0, %xmm5 xorps %xmm0, %xmm0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm3, %xmm7 addps %xmm6, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm6, %xmm14 mulps %xmm2, %xmm14 addps %xmm6, %xmm14 rcpps %xmm14, %xmm2 movaps %xmm2, %xmm7 addps %xmm2, %xmm7 mulps %xmm7, %xmm14 movaps 0x29c712(%rip), %xmm3 # 0x3f1ce0 subps %xmm14, %xmm3 mulps %xmm2, %xmm3 addps %xmm10, %xmm7 addps %xmm3, %xmm7 jmp 0x1556de movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax movss (%rax), %xmm2 movss 0x4(%rax), %xmm7 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] maxps %xmm2, %xmm1 minps %xmm7, %xmm1 movaps %xmm1, %xmm7 jmp 0x1556e1 xorps 0x298a7c(%rip), %xmm1 # 0x3ee090 minps %xmm15, %xmm1 maxps %xmm9, %xmm1 movaps %xmm1, %xmm7 mulps %xmm8, %xmm7 addps %xmm5, %xmm7 cvttps2dq %xmm7, %xmm2 cvtdq2ps %xmm2, %xmm2 cmpltps %xmm2, %xmm7 andps %xmm6, %xmm7 subps %xmm7, %xmm2 cvttps2dq %xmm2, %xmm14 mulps 0x29c6ad(%rip), %xmm2 # 0x3f1cf0 addps %xmm1, %xmm2 movaps %xmm2, %xmm1 mulps %xmm2, %xmm1 movaps %xmm2, %xmm7 mulps 0x29ba0a(%rip), %xmm7 # 0x3f1060 addps 0x29ba13(%rip), %xmm7 # 0x3f1070 mulps %xmm2, %xmm7 addps 0x29ba19(%rip), %xmm7 # 0x3f1080 mulps %xmm2, %xmm7 addps 0x29ba1f(%rip), %xmm7 # 0x3f1090 mulps %xmm2, %xmm7 addps 0x29ba25(%rip), %xmm7 # 0x3f10a0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm1, %xmm7 addps %xmm6, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm6, %xmm14 mulps %xmm2, %xmm14 addps %xmm6, %xmm14 rcpps %xmm14, %xmm1 mulps %xmm1, %xmm14 movaps %xmm6, %xmm7 subps %xmm14, %xmm7 mulps %xmm1, %xmm7 addps %xmm1, %xmm7 jmp 0x1556e1 movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax movss (%rax), %xmm7 movss 0x4(%rax), %xmm2 shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm7 addps %xmm2, %xmm7 maxps %xmm0, %xmm7 minps %xmm6, %xmm7 mulps %xmm1, %xmm7 movq 0x18(%rsp), %rax movl 0x28(%rsp), %ecx movups %xmm7, (%rax) addq $0x10, %rax incl %ecx incl 0x20(%rsp) jmp 0x155197 movaps %xmm1, %xmm7 movq 0xb8(%rsp), %r10 movq 0x18(%rsp), %rax movl 0x10(%rsp), %edx jmp 0x1556e6 incl %edx jmp 0x15517d movq 0x80(%rsp), %rsi incq %rsi movq 0x70(%rsp), %rdx jmp 0x155103 cmpl $0x1, 0x148(%rsp) setne %al orb 0xf(%rsp), %al jne 0x155e33 movq 0x150(%rsp), %rax movq -0x18(%rax), %rcx movslq 0xd4(%r15,%rcx), %rax movl 0xd8(%r15,%rcx), %ebp movl 0xdc(%r15,%rcx), %r8d movl 0xe0(%r15,%rcx), %r14d movl 0xe4(%r15,%rcx), %r10d movl 0xe8(%r15,%rcx), %ebx movl 0x114(%r15,%rcx), %edi movl 0x118(%rsp), %r9d movq %rcx, 0x40(%rsp) movq 0x1a8(%r15,%rcx), %rcx movq %rcx, 0xc0(%rsp) movl %eax, %ecx imull %ebp, %ecx shll $0x2, %ecx movslq %ecx, %rcx xorl %r11d, %r11d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi testl %ebp, %ebp movl $0x0, %r13d cmovgl %ebp, %r13d testl %r9d, %r9d cmovlel %r11d, %r9d movq %r9, 0x138(%rsp) shlq $0x2, %rcx movq %rcx, 0x18(%rsp) pushq $0x1 popq %rcx subl %eax, %ecx movq %rax, %r11 xorl %eax, %eax shlq $0x4, %r11 imull %r8d, %ecx movl %ecx, 0x60(%rsp) shlq $0x4, %rsi decl %edi xorps %xmm0, %xmm0 movaps 0x29b809(%rip), %xmm15 # 0x3f1010 movaps 0x29b811(%rip), %xmm9 # 0x3f1020 movaps 0x29b819(%rip), %xmm8 # 0x3f1030 movaps 0x298882(%rip), %xmm5 # 0x3ee0a0 movaps 0x29b81b(%rip), %xmm6 # 0x3f1040 movq %rdi, 0xb8(%rsp) movq %r13, 0x50(%rsp) movq %r11, 0x48(%rsp) movaps 0x29b831(%rip), %xmm11 # 0x3f1070 movaps 0x29b839(%rip), %xmm12 # 0x3f1080 movaps 0x29b841(%rip), %xmm13 # 0x3f1090 cmpq 0x138(%rsp), %rax je 0x155e33 movq 0x120(%rsp), %rcx imulq %rax, %rcx imulq 0xf0(%rsp), %rcx addq 0xe0(%rsp), %rcx movq %rcx, 0x28(%rsp) movl 0x2c(%rdx), %ecx movl %ecx, 0x10(%rsp) movl 0x30(%rdx), %ecx movl %ecx, 0xa0(%rsp) movl 0x38(%rdx), %ecx movq %rax, 0x80(%rsp) shlq $0x4, %rax movq %rax, 0x78(%rsp) xorl %r12d, %r12d testl %ecx, %ecx cmovlel %r12d, %ecx movq %rcx, 0x58(%rsp) movl 0x10c(%rsp), %eax testl %eax, %eax cmovlel %r12d, %eax movl %eax, 0x68(%rsp) movl 0x110(%rsp), %eax testl %eax, %eax cmovlel %r12d, %eax movl %eax, 0x64(%rsp) movq 0xc0(%rsp), %rax cmpl 0x64(%rsp), %r12d je 0x155e1e xorl %edx, %edx movl 0x60(%rsp), %ecx movl %ecx, 0x20(%rsp) cmpl 0x68(%rsp), %edx je 0x155e16 testq %rax, %rax je 0x15590b movq 0x78(%rsp), %rcx movups (%rax,%rcx), %xmm1 jmp 0x15590e xorps %xmm1, %xmm1 movq 0x70(%rsp), %rcx movslq 0x2c(%rcx), %rdi movq (%rcx), %rax movq %rax, 0x88(%rsp) movq 0x10(%rcx), %rax movq 0x40(%rcx), %rcx imulq %rax, %rcx movq %rcx, 0xc8(%rsp) movq 0x58(%r15), %rcx imulq 0x80(%rsp), %rcx imulq 0x28(%r15), %rcx imulq %rax, %rdi movq %rdi, 0x30(%rsp) addq 0x18(%r15), %rcx xorl %eax, %eax movl %edx, 0x90(%rsp) cmpq 0x58(%rsp), %rax je 0x155a55 movq 0xc8(%rsp), %rdx movq %rax, 0x98(%rsp) imulq %rax, %rdx addq 0x88(%rsp), %rdx movq %rdx, 0x38(%rsp) movq %rcx, 0xd0(%rsp) movq %rcx, %rdi xorl %r9d, %r9d cmpq %r13, %r9 je 0x155a31 movl %r9d, %eax subl %ebp, %eax incl %eax imull %r14d, %eax addl %r12d, %eax js 0x155a26 cltd idivl %ebx testl %edx, %edx jne 0x155a26 cmpl 0xa0(%rsp), %eax jge 0x155a26 movl %ebx, %r11d movl %r14d, %ebx movl %ebp, %r13d movslq %eax, %r14 imulq 0x30(%rsp), %r14 addq 0x38(%rsp), %r14 movl 0x20(%rsp), %ecx xorl %ebp, %ebp cmpq %rbp, %rsi je 0x155a13 testl %ecx, %ecx js 0x155a0a movl %ecx, %eax cltd idivl %r10d testl %edx, %edx jne 0x155a0a cmpl 0x10(%rsp), %eax jge 0x155a0a cltq movss (%r14,%rax,4), %xmm2 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps (%rdi,%rbp), %xmm2 addps %xmm2, %xmm1 addq $0x10, %rbp addl %r8d, %ecx jmp 0x1559de movl %r13d, %ebp movl %ebx, %r14d movl %r11d, %ebx movq 0x50(%rsp), %r13 movq 0x48(%rsp), %r11 incq %r9 addq %r11, %rdi jmp 0x155998 movq 0x98(%rsp), %rax incq %rax movq 0xd0(%rsp), %rcx addq 0x18(%rsp), %rcx movl 0x90(%rsp), %edx jmp 0x15595e movq 0xb8(%rsp), %rax cmpl $0x5, %eax ja 0x155d16 leaq 0x29d613(%rip), %rcx # 0x3f3080 movslq (%rcx,%rax,4), %rax addq %rcx, %rax movaps %xmm1, %xmm7 maxps %xmm0, %xmm7 jmpq %rax movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax minps %xmm0, %xmm1 movss (%rax), %xmm2 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm2 addps %xmm2, %xmm7 jmp 0x155df2 movaps %xmm1, %xmm2 minps %xmm15, %xmm2 maxps %xmm9, %xmm2 movaps %xmm2, %xmm14 mulps %xmm8, %xmm14 addps %xmm5, %xmm14 cvttps2dq %xmm14, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm14 andps %xmm6, %xmm14 subps %xmm14, %xmm7 movaps %xmm15, %xmm8 cvttps2dq %xmm7, %xmm15 movaps 0x29b575(%rip), %xmm0 # 0x3f1050 mulps %xmm0, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm7 mulps %xmm2, %xmm7 movaps %xmm2, %xmm14 movaps 0x29b56d(%rip), %xmm10 # 0x3f1060 mulps %xmm10, %xmm14 addps %xmm11, %xmm14 mulps %xmm2, %xmm14 addps %xmm12, %xmm14 mulps %xmm2, %xmm14 addps %xmm13, %xmm14 mulps %xmm2, %xmm14 xorps %xmm4, %xmm4 movaps %xmm5, %xmm0 movaps 0x29b584(%rip), %xmm5 # 0x3f10a0 addps %xmm5, %xmm14 mulps %xmm2, %xmm14 addps %xmm0, %xmm14 mulps %xmm7, %xmm14 addps %xmm6, %xmm2 addps %xmm14, %xmm2 pslld $0x17, %xmm15 paddd %xmm6, %xmm15 mulps %xmm2, %xmm15 addps %xmm6, %xmm15 movaps %xmm15, %xmm2 maxps 0x29b55e(%rip), %xmm15 # 0x3f10b0 movaps %xmm15, %xmm7 andps 0x29b562(%rip), %xmm15 # 0x3f10c0 orps %xmm0, %xmm15 movaps %xmm15, %xmm3 cmpltps 0x29b572(%rip), %xmm3 # 0x3f10e0 movaps %xmm3, %xmm14 andps %xmm15, %xmm14 movaps 0x29b572(%rip), %xmm10 # 0x3f10f0 addps %xmm10, %xmm15 addps %xmm14, %xmm15 cmpleps %xmm4, %xmm2 psrld $0x17, %xmm7 paddd 0x29b539(%rip), %xmm7 # 0x3f10d0 cvtdq2ps %xmm7, %xmm14 andps %xmm6, %xmm3 subps %xmm3, %xmm14 movaps %xmm15, %xmm7 mulps 0x29b553(%rip), %xmm7 # 0x3f1100 addps 0x29b55c(%rip), %xmm7 # 0x3f1110 mulps %xmm15, %xmm7 addps 0x29b561(%rip), %xmm7 # 0x3f1120 mulps %xmm15, %xmm7 addps 0x29b566(%rip), %xmm7 # 0x3f1130 mulps %xmm15, %xmm7 addps 0x29b56b(%rip), %xmm7 # 0x3f1140 mulps %xmm15, %xmm7 addps 0x29b570(%rip), %xmm7 # 0x3f1150 mulps %xmm15, %xmm7 addps 0x29b575(%rip), %xmm7 # 0x3f1160 mulps %xmm15, %xmm7 addps 0x29b57a(%rip), %xmm7 # 0x3f1170 mulps %xmm15, %xmm7 addps 0x29b57f(%rip), %xmm7 # 0x3f1180 mulps %xmm15, %xmm7 movaps 0x29b444(%rip), %xmm4 # 0x3f1050 mulps %xmm4, %xmm14 addps %xmm15, %xmm14 mulps %xmm15, %xmm15 addps 0x29b571(%rip), %xmm7 # 0x3f1190 mulps %xmm15, %xmm7 movaps %xmm8, %xmm15 movaps 0x29b401(%rip), %xmm8 # 0x3f1030 addps %xmm7, %xmm14 mulps 0x29c085(%rip), %xmm14 # 0x3f1cc0 movaps %xmm2, %xmm3 andnps %xmm14, %xmm3 andps 0x29c087(%rip), %xmm2 # 0x3f1cd0 orps %xmm3, %xmm2 minps %xmm15, %xmm2 maxps %xmm9, %xmm2 movaps %xmm2, %xmm3 mulps %xmm8, %xmm3 addps %xmm0, %xmm3 cvttps2dq %xmm3, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm3 andps %xmm6, %xmm3 subps %xmm3, %xmm7 cvttps2dq %xmm7, %xmm14 mulps %xmm4, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm3 mulps %xmm2, %xmm3 movaps %xmm2, %xmm7 mulps 0x29b3d6(%rip), %xmm7 # 0x3f1060 addps %xmm11, %xmm7 mulps %xmm2, %xmm7 addps %xmm12, %xmm7 mulps %xmm2, %xmm7 addps %xmm13, %xmm7 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 movaps %xmm0, %xmm5 xorps %xmm0, %xmm0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm3, %xmm7 addps %xmm6, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm6, %xmm14 mulps %xmm2, %xmm14 addps %xmm6, %xmm14 rcpps %xmm14, %xmm2 movaps %xmm2, %xmm7 addps %xmm2, %xmm7 mulps %xmm7, %xmm14 movaps 0x29c001(%rip), %xmm3 # 0x3f1ce0 subps %xmm14, %xmm3 mulps %xmm2, %xmm3 addps %xmm10, %xmm7 addps %xmm3, %xmm7 jmp 0x155def movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax movss (%rax), %xmm2 movss 0x4(%rax), %xmm7 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] maxps %xmm2, %xmm1 minps %xmm7, %xmm1 movaps %xmm1, %xmm7 jmp 0x155df2 xorps 0x29836b(%rip), %xmm1 # 0x3ee090 minps %xmm15, %xmm1 maxps %xmm9, %xmm1 movaps %xmm1, %xmm7 mulps %xmm8, %xmm7 addps %xmm5, %xmm7 cvttps2dq %xmm7, %xmm2 cvtdq2ps %xmm2, %xmm2 cmpltps %xmm2, %xmm7 andps %xmm6, %xmm7 subps %xmm7, %xmm2 cvttps2dq %xmm2, %xmm14 mulps 0x29bf9c(%rip), %xmm2 # 0x3f1cf0 addps %xmm1, %xmm2 movaps %xmm2, %xmm1 mulps %xmm2, %xmm1 movaps %xmm2, %xmm7 mulps 0x29b2f9(%rip), %xmm7 # 0x3f1060 addps 0x29b302(%rip), %xmm7 # 0x3f1070 mulps %xmm2, %xmm7 addps 0x29b308(%rip), %xmm7 # 0x3f1080 mulps %xmm2, %xmm7 addps 0x29b30e(%rip), %xmm7 # 0x3f1090 mulps %xmm2, %xmm7 addps 0x29b314(%rip), %xmm7 # 0x3f10a0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm1, %xmm7 addps %xmm6, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm6, %xmm14 mulps %xmm2, %xmm14 addps %xmm6, %xmm14 rcpps %xmm14, %xmm1 mulps %xmm1, %xmm14 movaps %xmm6, %xmm7 subps %xmm14, %xmm7 mulps %xmm1, %xmm7 addps %xmm1, %xmm7 jmp 0x155df2 movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax movss (%rax), %xmm7 movss 0x4(%rax), %xmm2 shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm7 addps %xmm2, %xmm7 maxps %xmm0, %xmm7 minps %xmm6, %xmm7 mulps %xmm1, %xmm7 movq 0x28(%rsp), %rax movups %xmm7, (%rax) addq $0x10, %rax movq %rax, 0x28(%rsp) incl %edx incl 0x20(%rsp) movq 0xc0(%rsp), %rax jmp 0x1558f1 incl %r12d jmp 0x1558dc movq 0x80(%rsp), %rax incq %rax movq 0x70(%rsp), %rdx jmp 0x15584f cmpl $0x4, 0x148(%rsp) jne 0x1562f5 cmpb $0x0, 0xf(%rsp) je 0x1562f5 movq 0x150(%rsp), %rax movq -0x18(%rax), %rsi movslq 0xd4(%r15,%rsi), %rax movl 0xd8(%r15,%rsi), %edi movl 0xdc(%r15,%rsi), %r14d movl 0xe0(%r15,%rsi), %r10d movl 0x114(%r15,%rsi), %r8d movl 0x118(%rsp), %r9d movq 0xf0(%rsp), %rcx imulq 0x120(%rsp), %rcx movq %rcx, 0x170(%rsp) movl %eax, %ecx imull %edi, %ecx shll $0x2, %ecx xorl %r11d, %r11d testl %eax, %eax movl $0x0, %ebx cmovgl %eax, %ebx testl %edi, %edi movl $0x0, %edx cmovgl %edi, %edx movq %rdx, 0x30(%rsp) movq 0x70(%rsp), %rdx movl 0x38(%rdx), %r12d testl %r12d, %r12d cmovlel %r11d, %r12d movq %r12, 0x58(%rsp) movl 0x10c(%rsp), %ebp testl %ebp, %ebp cmovlel %r11d, %ebp movl %ebp, 0x80(%rsp) movl 0x110(%rsp), %ebp testl %ebp, %ebp cmovlel %r11d, %ebp movl %ebp, 0x64(%rsp) movslq %ecx, %rcx testl %r9d, %r9d cmovlel %r11d, %r9d movq %r9, 0x150(%rsp) shlq $0x2, %rcx movq %rcx, 0xc8(%rsp) pushq $0x1 popq %rcx subl %eax, %ecx shlq $0x4, %rax movq %rax, 0x10(%rsp) imull %r14d, %ecx movl %ecx, 0x60(%rsp) shlq $0x4, %rbx decl %r8d movq %r8, 0xb8(%rsp) movl 0xe4(%r15,%rsi), %r12d movl 0xe8(%r15,%rsi), %eax movl %eax, 0xa0(%rsp) movq %rsi, 0x138(%rsp) movq 0x1a8(%r15,%rsi), %rcx movq 0xe0(%rsp), %rax movq %rax, 0x168(%rsp) movl 0x2c(%rdx), %r13d movl 0x30(%rdx), %eax movl %eax, 0x38(%rsp) movl %edi, 0x50(%rsp) movl %r10d, 0x48(%rsp) movq %rcx, 0x68(%rsp) cmpq 0x150(%rsp), %r11 je 0x1562f5 movq 0x170(%rsp), %rdx movq %r11, 0x78(%rsp) imulq %r11, %rdx addq 0x168(%rsp), %rdx movq %rdx, 0x18(%rsp) xorl %ebp, %ebp cmpl 0x64(%rsp), %ebp je 0x1562e8 movq 0x138(%rsp), %rax movq 0x118(%r15,%rax), %rax movq %rax, 0x40(%rsp) movq 0x58(%r15), %rax imulq 0x78(%rsp), %rax imulq 0x28(%r15), %rax addq 0x18(%r15), %rax movq %rax, 0xc0(%rsp) xorl %eax, %eax movl 0x60(%rsp), %edx movl %edx, 0x20(%rsp) cmpl 0x80(%rsp), %eax je 0x1562e1 movl %eax, 0x28(%rsp) testq %rcx, %rcx je 0x156026 movq 0x78(%rsp), %rax movss (%rcx,%rax,4), %xmm0 jmp 0x156029 xorps %xmm0, %xmm0 movq 0x70(%rsp), %rcx movslq 0x2c(%rcx), %rsi movq (%rcx), %rax movq %rax, 0x90(%rsp) movq 0x10(%rcx), %rax movq 0x40(%rcx), %rcx imulq %rax, %rcx movq %rcx, 0x88(%rsp) imulq %rax, %rsi xorps %xmm1, %xmm1 movq 0xc0(%rsp), %r9 xorl %eax, %eax movq 0x30(%rsp), %rcx cmpq 0x58(%rsp), %rax je 0x15613d movq 0x88(%rsp), %r11 movq %rax, 0x98(%rsp) imulq %rax, %r11 addq 0x90(%rsp), %r11 movq %r9, 0xd0(%rsp) xorl %r8d, %r8d cmpq %rcx, %r8 je 0x15611d movl %r8d, %eax subl %edi, %eax incl %eax imull %r10d, %eax addl %ebp, %eax js 0x156110 cltd idivl 0xa0(%rsp) testl %edx, %edx jne 0x156110 cmpl 0x38(%rsp), %eax jge 0x156110 movslq %eax, %rdi imulq %rsi, %rdi addq %r11, %rdi movl 0x20(%rsp), %ecx xorl %r10d, %r10d cmpq %r10, %rbx je 0x156102 testl %ecx, %ecx js 0x1560f9 movl %ecx, %eax cltd idivl %r12d testl %edx, %edx jne 0x1560f9 cmpl %r13d, %eax jge 0x1560f9 shll $0x2, %eax cltq movaps (%r9,%r10), %xmm2 mulps (%rdi,%rax,4), %xmm2 addps %xmm2, %xmm1 addq $0x10, %r10 addl %r14d, %ecx jmp 0x1560d0 movl 0x50(%rsp), %edi movl 0x48(%rsp), %r10d movq 0x30(%rsp), %rcx incq %r8 addq 0x10(%rsp), %r9 jmp 0x156099 movq 0x98(%rsp), %rax incq %rax movq 0xd0(%rsp), %r9 addq 0xc8(%rsp), %r9 jmp 0x156067 movaps %xmm1, %xmm4 unpckhpd %xmm1, %xmm4 # xmm4 = xmm4[1],xmm1[1] addps %xmm1, %xmm4 addss %xmm4, %xmm0 shufps $0x55, %xmm4, %xmm4 # xmm4 = xmm4[1,1,1,1] addss %xmm0, %xmm4 movq 0xb8(%rsp), %rax cmpl $0x5, %eax ja 0x1562b8 leaq 0x29cf2d(%rip), %rcx # 0x3f3098 movslq (%rcx,%rax,4), %rax addq %rcx, %rax movq 0x68(%rsp), %rcx jmpq %rax maxss 0x297e8f(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x1562c0 movaps %xmm4, %xmm0 movaps %xmm4, 0xd0(%rsp) callq 0x5f410 addss 0x298ae7(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x68(%rsp), %rcx movl 0x48(%rsp), %r10d movl 0x50(%rsp), %edi mulss 0xd0(%rsp), %xmm0 jmp 0x1562c0 movq 0x40(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x1562b3 jmp 0x1562c0 movss 0x29afca(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x297e94(%rip), %xmm0 # 0x3ee090 cmpltss 0x29afb7(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x68(%rsp), %rcx movl 0x48(%rsp), %r10d movl 0x50(%rsp), %edi movaps %xmm0, %xmm1 movss 0x298a56(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x1562c0 xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x298a36(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x40(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x1562c0 movq 0x40(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x297e0e(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x1562c0 movss 0x2989f2(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x156181 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 jmp 0x1562c0 movaps %xmm4, %xmm0 movq 0x68(%rsp), %rcx movq 0x18(%rsp), %rax movss %xmm0, (%rax) addq $0x4, %rax movq %rax, 0x18(%rsp) movl 0x28(%rsp), %eax incl %eax incl 0x20(%rsp) jmp 0x156004 incl %ebp jmp 0x155fc0 movq 0x78(%rsp), %r11 incq %r11 jmp 0x155f92 cmpl $0x1, 0x148(%rsp) setne %al movb 0xf(%rsp), %cl xorb $0x1, %cl orb %al, %cl movq 0x70(%rsp), %rax jne 0x156a07 movq (%r15), %rcx movq %rcx, 0xc0(%rsp) movq 0xe0(%rsp), %rcx movq %rcx, 0xb8(%rsp) movq 0x120(%rsp), %rcx imulq 0xf0(%rsp), %rcx movq %rcx, 0x78(%rsp) movl 0x2c(%rax), %ebp movl 0x30(%rax), %ecx movl %ecx, 0x10(%rsp) movl 0x38(%rax), %eax movslq 0x10c(%rsp), %rcx xorl %edx, %edx testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x20(%rsp) testl %ecx, %ecx movl $0x0, %esi movq %rcx, 0x80(%rsp) cmovgl %ecx, %esi movq %rsi, 0x90(%rsp) movl 0x110(%rsp), %eax testl %eax, %eax cmovlel %edx, %eax movl %eax, 0x68(%rsp) shlq $0x2, 0x158(%rsp) movq 0xc0(%rsp), %rax movq -0x18(%rax), %rax movslq 0xd0(%r15,%rax), %rax cmpq %rax, %rdx jge 0x156a07 movq 0x78(%rsp), %rax imulq %rdx, %rax addq 0xb8(%rsp), %rax movq %rax, 0x58(%rsp) xorl %r12d, %r12d movq %rdx, 0x18(%rsp) cmpl 0x68(%rsp), %r12d je 0x156749 movq (%r15), %rax movq %rax, 0x88(%rsp) movl %r12d, 0x28(%rsp) movq 0x58(%r15), %rcx imulq %rdx, %rcx imulq 0x28(%r15), %rcx subl 0x140(%rsp), %r12d addq 0x18(%r15), %rcx movq %rcx, 0xc8(%rsp) xorl %edx, %edx cmpq 0x90(%rsp), %rdx je 0x156721 movq 0x88(%rsp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r15,%rax) movq %rax, 0x98(%rsp) je 0x156453 movq 0x1a8(%r15,%rax), %rax movq 0x18(%rsp), %rcx movss (%rax,%rcx,4), %xmm4 jmp 0x156456 xorps %xmm4, %xmm4 movq 0x70(%rsp), %rcx movslq 0x2c(%rcx), %rsi movq (%rcx), %rax movq %rax, 0x30(%rsp) movq 0x10(%rcx), %rax movq 0x40(%rcx), %rcx imulq %rax, %rcx movq %rcx, 0x50(%rsp) movq (%r15), %rcx movq %rcx, 0x48(%rsp) imulq %rax, %rsi movq %rsi, 0xa0(%rsp) movq %rdx, 0xd0(%rsp) movl %edx, %r10d subl 0x144(%rsp), %r10d movq 0xc8(%rsp), %r11 xorl %eax, %eax cmpq 0x20(%rsp), %rax je 0x156595 movq 0x50(%rsp), %r8 movq %rax, 0x38(%rsp) imulq %rax, %r8 addq 0x30(%rsp), %r8 movq 0x48(%rsp), %rax movq -0x18(%rax), %r14 movl 0xd8(%r15,%r14), %edi xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x156580 movl 0xe0(%r15,%r14), %eax imull %r9d, %eax addl %r12d, %eax js 0x156578 cltd idivl 0xe8(%r15,%r14) testl %edx, %edx jne 0x156578 cmpl 0x10(%rsp), %eax jge 0x156578 movslq %eax, %rcx imulq 0xa0(%rsp), %rcx addq %r8, %rcx movl 0xd4(%r15,%r14), %eax testl %eax, %eax movl $0x0, %ebx cmovgl %eax, %ebx imull %r9d, %eax cltq leaq (%r11,%rax,4), %r13 xorl %esi, %esi cmpq %rsi, %rbx je 0x156578 movl 0xdc(%r15,%r14), %eax imull %esi, %eax addl %r10d, %eax js 0x156573 cltd idivl 0xe4(%r15,%r14) testl %edx, %edx jne 0x156573 cmpl %ebp, %eax jge 0x156573 cltq movss (%r13,%rsi,4), %xmm0 mulss (%rcx,%rax,4), %xmm0 addss %xmm0, %xmm4 incq %rsi jmp 0x15653b incl %r9d jmp 0x1564e1 movq 0x38(%rsp), %rax incq %rax addq 0x158(%rsp), %r11 jmp 0x1564a9 movq 0x98(%rsp), %rsi movl 0x114(%r15,%rsi), %eax decl %eax cmpl $0x5, %eax ja 0x156704 leaq 0x29caf9(%rip), %rcx # 0x3f30b0 movslq (%rcx,%rax,4), %rax addq %rcx, %rax movq 0xd0(%rsp), %rdx jmpq %rax maxss 0x297a40(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x15670f movaps %xmm4, %xmm0 movaps %xmm4, 0xa0(%rsp) callq 0x5f410 addss 0x298698(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0xd0(%rsp), %rdx mulss 0xa0(%rsp), %xmm0 jmp 0x15670f movq 0x118(%r15,%rsi), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x1566ff jmp 0x15670f movss 0x29ab7e(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x297a48(%rip), %xmm0 # 0x3ee090 cmpltss 0x29ab6b(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0xd0(%rsp), %rdx movaps %xmm0, %xmm1 movss 0x298610(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x15670f movq 0x118(%r15,%rsi), %rax movss (%rax), %xmm1 xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm2 andnps %xmm1, %xmm2 movss 0x2985e1(%rip), %xmm1 # 0x3eec88 andps %xmm1, %xmm0 orps %xmm2, %xmm0 mulss %xmm4, %xmm0 jmp 0x15670f movq 0x118(%r15,%rsi), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x2979c2(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x15670f movss 0x2985a6(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x1565d0 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 jmp 0x15670f movaps %xmm4, %xmm0 movq 0xd0(%rsp), %rdx movq 0x58(%rsp), %rax movss %xmm0, (%rax,%rdx,4) incq %rdx jmp 0x156412 movq 0x80(%rsp), %rax movq 0x58(%rsp), %rcx leaq (%rcx,%rax,4), %rcx movq %rcx, 0x58(%rsp) movl 0x28(%rsp), %r12d incl %r12d movq 0x18(%rsp), %rdx jmp 0x1563d4 incq %rdx jmp 0x156399 movslq 0x1ec(%rsp), %rcx imulq 0x1d0(%rsp), %rcx movq 0x1c0(%rsp), %rax movq %rax, 0x30(%rsp) movslq 0x10c(%rsp), %r11 movslq 0x110(%rsp), %rax movq 0xf0(%rsp), %rdx movq 0x120(%rsp), %rbp imulq %rdx, %rbp imulq %r11, %rax movq %rdx, %r9 movq %rax, 0x50(%rsp) imulq %rax, %r9 addq $0xf, %r9 andq $-0x10, %r9 movq %r9, 0x98(%rsp) movq %rdx, 0x48(%rsp) imulq %rdx, %r11 xorl %edx, %edx testl %r13d, %r13d cmovlel %edx, %r13d movl 0x114(%rsp), %eax movl %eax, 0x58(%rsp) testl %r14d, %r14d cmovlel %edx, %r14d movq 0xe0(%rsp), %r9 movq 0xa0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xa0(%rsp) movl 0x108(%rsp), %eax movl %eax, 0x88(%rsp) imulq 0x158(%rsp), %rcx movq %rcx, 0x20(%rsp) movq %r9, 0x90(%rsp) movq %r9, 0x38(%rsp) xorl %eax, %eax movq %rbp, 0xd0(%rsp) cmpq 0xa0(%rsp), %rax je 0x15697a movq %rax, 0x10(%rsp) movq 0x98(%rsp), %rax xorl %edx, %edx divq 0x48(%rsp) cmpl $0x4, 0x88(%rsp) cmoveq 0x50(%rsp), %rax movq 0x1a8(%r15,%r8), %rcx xorps %xmm0, %xmm0 testq %rcx, %rcx je 0x15688e movslq 0x1e0(%r15,%r8), %rdx imulq 0x1e8(%r15,%r8), %rdx testq %rdx, %rdx movq 0x38(%rsp), %r8 je 0x156893 movq 0x10(%rsp), %rdx movss (%rcx,%rdx,4), %xmm0 jmp 0x156893 movq 0x38(%rsp), %r8 movq 0x10(%rsp), %rcx movq %rcx, %rdx imulq 0x20(%rsp), %rdx addq 0x30(%rsp), %rdx imulq %rcx, %rbp addq 0x90(%rsp), %rbp imull 0x58(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x1568d2 movss %xmm0, (%r8,%rcx,4) incq %rcx jmp 0x1568c3 movq -0x18(%rsi), %r8 xorl %eax, %eax cmpl 0xd8(%r15,%r8), %eax jge 0x156960 xorl %r9d, %r9d cmpl 0xd4(%r15,%r8), %r9d jge 0x156959 movslq 0xe0(%r15,%r8), %rcx movslq %eax, %r10 imulq %rcx, %r10 imulq %r11, %r10 addq %rbp, %r10 movslq 0xdc(%r15,%r8), %rcx movslq %r9d, %r12 imulq %rcx, %r12 leaq (%r10,%r12,4), %rcx xorl %r10d, %r10d cmpl %r14d, %r10d je 0x156954 movl %r13d, %r12d subl $0x1, %r12d jb 0x15694b movss (%rcx), %xmm0 addss (%rdx), %xmm0 movss %xmm0, (%rcx) movq -0x18(%rsi), %r8 movslq 0xe4(%r15,%r8), %rbx leaq (%rcx,%rbx,4), %rcx addq $0x4, %rdx jmp 0x156923 leaq (%rcx,%rdi,4), %rcx incl %r10d jmp 0x15691b incl %r9d jmp 0x1568e5 incl %eax jmp 0x1568d8 movq 0x10(%rsp), %rax incq %rax movq 0xd0(%rsp), %rbp addq %rbp, 0x38(%rsp) jmp 0x156827 movq 0x8(%r15), %rdi testq %rdi, %rdi je 0x156999 movq (%rdi), %rax leaq 0xe0(%rsp), %rsi movq 0x130(%rsp), %rdx callq 0x48(%rax) movq 0x1c8(%rsp), %rax testq %rax, %rax je 0x1569d0 lock decl (%rax) jne 0x1569d0 movq 0x1c0(%rsp), %rsi movq 0x1e0(%rsp), %rdi testq %rdi, %rdi je 0x1569c8 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1569d0 movq %rsi, %rdi callq 0x5f3e0 movq 0x180(%rsp), %rax testq %rax, %rax je 0x156a07 lock decl (%rax) jne 0x156a07 movq 0x178(%rsp), %rsi movq 0x198(%rsp), %rdi testq %rdi, %rdi je 0x1569ff movq (%rdi), %rax callq 0x18(%rax) jmp 0x156a07 movq %rsi, %rdi callq 0x5f3e0 movq (%r15), %rax addq -0x18(%rax), %r15 leaq 0xe0(%rsp), %rsi movq %r15, %rdi movq 0x160(%rsp), %rbx movq %rbx, %rdx movq 0x130(%rsp), %rcx callq 0x153cf6 cmpq $0x0, (%rbx) pushq $-0x64 popq %rcx je 0x156a4a movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax xorl %ebx, %ebx testq %rax, %rax jne 0x156a4c movl %ecx, %ebx movq 0xe8(%rsp), %rax testq %rax, %rax je 0x156a83 lock decl (%rax) jne 0x156a83 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x156a7b movq (%rdi), %rax callq 0x18(%rax) jmp 0x156a83 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0x258, %rsp # imm = 0x258 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq cmpl $0x0, 0x104(%r15,%rax) jle 0x156ab1 cmpl $0x0, 0x108(%r15,%rax) jg 0x154c2e pushq $0x8 popq %rsi leaq 0xe0(%rsp), %rax movq 0x160(%rsp), %rcx cmpq %rcx, %rax je 0x154c2e movq 0x8(%rcx), %rax testq %rax, %rax je 0x156b43 lock incl (%rax) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x156b43 lock decl (%rax) jne 0x156b43 movl %r9d, 0x30(%rsp) movq %r11, 0x20(%rsp) movq %r10, 0x38(%rsp) movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x156b29 movq (%rdi), %rax callq 0x18(%rax) movq 0x38(%rsp), %r10 movq 0x20(%rsp), %r11 movl 0x30(%rsp), %r9d pushq $0x8 popq %rsi jmp 0x156b43 movq %rsi, %rdi callq 0x5f3e0 movq 0x38(%rsp), %r10 movq 0x20(%rsp), %r11 movl 0x30(%rsp), %r9d pushq $0x8 popq %rsi movq 0x160(%rsp), %rcx movups (%rcx), %xmm0 movaps %xmm0, 0xe0(%rsp) movq 0x10(%rcx), %rax movq %rax, 0xf0(%rsp) movl 0x18(%rcx), %eax movl %eax, 0xf8(%rsp) movq 0x20(%rcx), %rax movq %rax, 0x100(%rsp) movups 0x28(%rcx), %xmm0 movups %xmm0, 0x108(%rsp) movl 0x38(%rcx), %eax movl %eax, 0x118(%rsp) movq 0x40(%rcx), %rax movq %rax, 0x120(%rsp) jmp 0x154c2e jmp 0x156c69 jmp 0x156c69 jmp 0x156c69 movq %rax, %rbx movq 0x1c8(%rsp), %rax testq %rax, %rax je 0x156be8 lock decl (%rax) jne 0x156be8 movq 0x1c0(%rsp), %rsi movq 0x1e0(%rsp), %rdi testq %rdi, %rdi jne 0x156be2 movq %rsi, %rdi callq 0x5f3e0 jmp 0x156be8 movq (%rdi), %rax callq 0x18(%rax) movq 0x180(%rsp), %rax testq %rax, %rax je 0x156c2a lock decl (%rax) jne 0x156c2a movq 0x178(%rsp), %rsi movq 0x198(%rsp), %rdi testq %rdi, %rdi jne 0x156c19 movq %rsi, %rdi callq 0x5f3e0 jmp 0x156c2a movq (%rdi), %rax callq 0x18(%rax) jmp 0x156c2a jmp 0x156c69 jmp 0x156c69 jmp 0x156c27 movq %rax, %rbx movq 0xe8(%rsp), %rax testq %rax, %rax je 0x156c61 lock decl (%rax) jne 0x156c61 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi jne 0x156c5b movq %rsi, %rdi callq 0x5f3e0 jmp 0x156c61 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/x86/deconvolution_x86.cpp
virtual thunk to ncnn::Deconvolution_x86_avx::create_pipeline(ncnn::Option const&)	int Deconvolution_x86_avx::create_pipeline(const Option& opt) { activation = create_activation_layer(activation_type, activation_params, opt); const int maxk = kernel_w * kernel_h; int num_input = weight_data_size / maxk / num_output; int elempack = 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = num_input % 16 == 0 ? 16 : num_input % 8 == 0 ? 8 : num_input % 4 == 0 ? 4 : 1; out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = num_input % 8 == 0 ? 8 : num_input % 4 == 0 ? 4 : 1; out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else elempack = num_input % 4 == 0 ? 4 : 1; out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ if (opt.use_sgemm_convolution) { const int maxk = kernel_w * kernel_h; gemm = ncnn::create_layer(ncnn::LayerType::Gemm); ncnn::ParamDict pd; pd.set(2, 1); // transA pd.set(3, 0); // transB pd.set(4, 1); // constantA pd.set(5, 0); // constantB pd.set(6, 1); // constantC pd.set(7, maxk * num_output); // M = maxknum_output pd.set(8, 0); // N = size pd.set(9, num_input); // K = inch pd.set(10, -1); // constant_broadcast_type_C = null pd.set(11, 0); // output_N1M pd.set(12, out_elempack); gemm->load_param(pd); // maxk-inch-outch to pa-maxk-outch/pa-inch Mat tmp; { Mat weight_data_r2 = weight_data.reshape(maxk, num_input, num_output); tmp.create(maxk num_output, num_input); for (int p = 0; p < num_input; p += 1) { float* g00 = tmp.row(p); for (int q = 0; q + (out_elempack - 1) < num_output; q += out_elempack) { for (int k = 0; k < maxk; k++) { for (int i = 0; i < out_elempack; i++) { const float* k00 = weight_data_r2.channel(q + i).row(p); g00[0] = k00[k]; g00++; } } } } } ncnn::Mat weights[1]; weights[0] = tmp; gemm->load_model(ModelBinFromMatArray(weights)); gemm->create_pipeline(opt); } else { Mat weight_data_transposed(weight_data.w); { float* pt = weight_data_transposed; const float* p = weight_data; for (int i = 0; i < num_input * num_output; i++) { for (int k = 0; k < maxk; k++) { pt[maxk - 1 - k] = p[k]; } p += maxk; pt += maxk; } } // src = kw-kh-inch-outch // dst = pb-pa-kw-kh-inch/pa-outch/pb Mat weight_data_r2 = weight_data_transposed.reshape(maxk, num_input, num_output); weight_data_tm.create(maxk, num_input / elempack, num_output / out_elempack, (size_t)4u * elempack * out_elempack, elempack * out_elempack); for (int q = 0; q + (out_elempack - 1) < num_output; q += out_elempack) { float* g00 = weight_data_tm.channel(q / out_elempack); for (int p = 0; p + (elempack - 1) < num_input; p += elempack) { for (int k = 0; k < maxk; k++) { for (int i = 0; i < elempack; i++) { for (int j = 0; j < out_elempack; j++) { const float* k00 = weight_data_r2.channel(q + j).row(p + i); g00[0] = k00[k]; g00++; } } } } } } if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %rax movq (%rdi), %rax addq -0x30(%rax), %rdi callq 0x164cde xorl %eax, %eax popq %rcx retq nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/deconvolution_x86_avx.cpp
ncnn::Deconvolution_x86_avx::destroy_pipeline(ncnn::Option const&)	int Deconvolution_x86_avx::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } if (gemm) { gemm->destroy_pipeline(opt); delete gemm; gemm = 0; } return 0; }	pushq %r14 pushq %rbx pushq %rax movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x165840 movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x16583b movq (%rdi), %rax callq 0x8(%rax) andq $0x0, 0x8(%rbx) movq 0x10(%rbx), %rdi testq %rdi, %rdi je 0x165866 movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x10(%rbx), %rdi testq %rdi, %rdi je 0x165861 movq (%rdi), %rax callq 0x8(%rax) andq $0x0, 0x10(%rbx) xorl %eax, %eax addq $0x8, %rsp popq %rbx popq %r14 retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/deconvolution_x86_avx.cpp
ncnn::InnerProduct_x86::create_pipeline_int8_x86(ncnn::Option const&)	int InnerProduct_x86::create_pipeline_int8_x86(const Option& opt) { const int num_input = weight_data_size / num_output; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { out_elempack = num_output % 8 == 0 ? 8 : 1; } #endif // __SSE2__ // src = inch-outch // dst = pb-inch-outch/pb { Mat weight_data_r2 = weight_data.reshape(num_input, num_output); weight_data_tm.create(num_input, num_output / out_elempack, (size_t)out_elempack, out_elempack); for (int q = 0; q + (out_elempack - 1) < num_output; q += out_elempack) { signed char* g0 = weight_data_tm.row<signed char>(q / out_elempack); for (int p = 0; p < num_input; p++) { for (int j = 0; j < out_elempack; j++) { g0++ = weight_data_r2.row<signed char>(q + j)[p]; } } } } scale_in_data.create(num_output); for (int p = 0; p < num_output; p++) { // dequantize float scale_in; if (weight_data_int8_scales[p] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[0] weight_data_int8_scales[p]); scale_in_data[p] = scale_in; } if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x48, %rsp movq %rsi, %rbx movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rax leaq 0x130(%rdi,%rax), %rsi movl -0x60(%rsi), %ecx movl -0x58(%rsi), %eax cltd idivl %ecx movl %eax, %r15d testb $0x7, %cl sete %bpl andb 0x27(%rbx), %bpl pushq $0x8 popq %rax pushq $0x1 popq %r12 cmovnel %eax, %r12d xorl %r13d, %r13d movq %rsp, %rdi movl %r15d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x10(%r14), %rdi movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd0(%r14,%rax), %eax cltd idivl %r12d movl %r15d, %esi movl %eax, %edx movq %r12, %rcx movl %r12d, %r8d xorl %r9d, %r9d callq 0x627de leal -0x1(%r12), %eax movq (%r14), %rdx movzbl %bpl, %ecx leal (%rcx,%rcx,2), %ecx testl %r15d, %r15d cmovlel %r13d, %r15d leaq (%rax,%r13), %rsi movq -0x18(%rdx), %rdi movslq 0xd0(%r14,%rdi), %rdi cmpq %rdi, %rsi jge 0x175b27 movl %r13d, %esi shrl %cl, %esi movslq 0x3c(%r14), %rdi imulq %rdi, %rsi imulq 0x20(%r14), %rsi addq 0x10(%r14), %rsi xorl %edi, %edi cmpq %r15, %rdi je 0x175b22 xorl %r8d, %r8d cmpq %r8, %r12 je 0x175b1a leaq (%r8,%r13), %r9 movslq 0x2c(%rsp), %r10 imulq %r9, %r10 imulq 0x10(%rsp), %r10 addq (%rsp), %r10 movb (%rdi,%r10), %r9b movb %r9b, (%rsi,%r8) incq %r8 jmp 0x175af1 incq %rdi addq %r8, %rsi jmp 0x175ae9 addq %r12, %r13 jmp 0x175abc movq 0x8(%rsp), %rax testq %rax, %rax je 0x175b54 lock decl (%rax) jne 0x175b54 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x175b4c movq (%rdi), %rax callq 0x18(%rax) jmp 0x175b54 movq %rsi, %rdi callq 0x5f3e0 leaq 0x58(%r14), %rdi movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd0(%r14,%rax), %esi xorl %r12d, %r12d pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq 0x58(%r14), %rcx movss 0x279105(%rip), %xmm0 # 0x3eec88 movq -0x18(%rax), %rdx leaq (%r14,%rdx), %r15 movslq 0xd0(%r14,%rdx), %rdx cmpq %rdx, %r12 jge 0x175bca movq 0x1c0(%r15), %rdx movss (%rdx,%r12,4), %xmm1 xorps %xmm2, %xmm2 ucomiss %xmm1, %xmm2 je 0x175bbf movq 0x208(%r15), %rdx mulss (%rdx), %xmm1 movaps %xmm0, %xmm2 divss %xmm1, %xmm2 movss %xmm2, (%rcx,%r12,4) incq %r12 jmp 0x175b83 cmpb $0x0, (%rbx) je 0x175c2c leaq 0x130(%r15), %rbx movq 0x138(%r15), %rax testq %rax, %rax je 0x175c0a lock decl (%rax) jne 0x175c0a movq 0x130(%r15), %rsi movq 0x150(%r15), %rdi testq %rdi, %rdi je 0x175c02 movq (%rdi), %rax callq 0x18(%rax) jmp 0x175c0a movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x170(%r15) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rbx) movups %xmm0, (%rbx) movups %xmm0, 0x158(%r15) andl $0x0, 0x168(%r15) xorl %eax, %eax addq $0x48, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x175c77 movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x175c6f lock decl (%rax) jne 0x175c6f movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x175c69 movq %rsi, %rdi callq 0x5f3e0 jmp 0x175c6f movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/x86/innerproduct_x86.cpp
ncnn::Input::Input()	Input::Input() { one_blob_only = true; support_inplace = true; support_vulkan = true; support_packing = true; support_bf16_storage = true; support_image_storage = true; }	pushq %rbx movq %rdi, %rbx callq 0x77e84 leaq 0x2a7e68(%rip), %rax # 0x4836d0 movq %rax, (%rbx) movb $0x1, %al movb %al, 0xf(%rbx) movb %al, 0xc(%rbx) movl $0x1010101, 0x8(%rbx) # imm = 0x1010101 popq %rbx retq	/csukuangfj[P]ncnn/src/layer/input.cpp
ncnn::Log::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int Log::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; int size = w * h; if (base == -1.f) { #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); for (int i = 0; i < size; i++) { ptr[i] = logf(shift + ptr[i] * scale); } } } else { float log_base_inv = 1.f / logf(base); #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); for (int i = 0; i < size; i++) { ptr[i] = logf(shift + ptr[i] * scale) * log_base_inv; } } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x18, %rsp movq %rsi, %r13 movq %rdi, %rbx movl 0x30(%rsi), %r15d movl 0x38(%rsi), %r12d imull 0x2c(%rsi), %r15d movss 0xd0(%rdi), %xmm0 movss 0x21584b(%rip), %xmm1 # 0x3f11f0 ucomiss %xmm0, %xmm1 jne 0x1dba13 movq (%r13), %r14 movq 0x40(%r13), %rax xorl %ebp, %ebp testl %r15d, %r15d cmovlel %ebp, %r15d testl %r12d, %r12d cmovlel %ebp, %r12d imulq 0x10(%r13), %rax movq %rax, 0x8(%rsp) cmpq %r12, %rbp je 0x1dba94 movq %r12, %r13 xorl %r12d, %r12d cmpq %r12, %r15 je 0x1dba06 movss 0xd4(%rbx), %xmm0 mulss (%r14,%r12,4), %xmm0 addss 0xd8(%rbx), %xmm0 callq 0x5f200 movss %xmm0, (%r14,%r12,4) incq %r12 jmp 0x1db9db incq %rbp addq 0x8(%rsp), %r14 movq %r13, %r12 jmp 0x1db9cc callq 0x5f200 movq %r13, %rcx movq (%r13), %r13 xorl %ebp, %ebp testl %r15d, %r15d cmovlel %ebp, %r15d movq 0x40(%rcx), %rax testl %r12d, %r12d cmovlel %ebp, %r12d movss 0x21324d(%rip), %xmm1 # 0x3eec88 divss %xmm0, %xmm1 movss %xmm1, 0x8(%rsp) imulq 0x10(%rcx), %rax movq %rax, 0x10(%rsp) cmpq %r12, %rbp je 0x1dba94 xorl %r14d, %r14d cmpq %r14, %r15 je 0x1dba8a movss 0xd4(%rbx), %xmm0 mulss (%r13,%r14,4), %xmm0 addss 0xd8(%rbx), %xmm0 callq 0x5f200 mulss 0x8(%rsp), %xmm0 movss %xmm0, (%r13,%r14,4) incq %r14 jmp 0x1dba57 incq %rbp addq 0x10(%rsp), %r13 jmp 0x1dba4f xorl %eax, %eax addq $0x18, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq nop	/csukuangfj[P]ncnn/src/layer/log.cpp
ncnn::LRN::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int LRN::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); for (int i = 0; i < size; i++) { outptr[i] = ptr[i] * ptr[i]; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum float* ssptr = square_sum.channel(q); for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 \|\| p >= channels) continue; const float* sptr = square_blob.channel(p); for (int i = 0; i < size; i++) { ssptr[i] += sptr[i]; } } float* ptr = bottom_top_blob.channel(q); for (int i = 0; i < size; i++) { ptr[i] = ptr[i] * powf(bias + alpha_div_size * ssptr[i], -beta); } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; opt_b.use_packing_layout = false; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size * local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x158, %rsp # imm = 0x158 movq %rsi, %rbp movl 0x2c(%rsi), %esi movl 0x30(%rbp), %ecx movl 0x38(%rbp), %r13d movq 0x10(%rbp), %r15 leaq 0x90(%rsp), %rax andq $0x0, 0x40(%rax) movq %rdx, %r14 movq %rdi, %rbx xorps %xmm0, %xmm0 movaps %xmm0, (%rax) movups %xmm0, 0xc(%rax) movaps %xmm0, 0x20(%rax) movups %xmm0, 0x2c(%rax) movq 0x10(%rdx), %r9 movq %rax, %rdi movq %rsi, %r12 movq %rcx, (%rsp) movl %ecx, %edx movl %r13d, %ecx movq %r15, %r8 callq 0x63810 movq 0x90(%rsp), %rcx pushq $-0x64 popq %r10 testq %rcx, %rcx je 0x1dbc70 movq 0xd0(%rsp), %rdx movslq 0xc8(%rsp), %rax movq %rdx, %rsi imulq %rax, %rsi testq %rsi, %rsi je 0x1dbc70 movq %r15, %r11 movq %r14, 0x18(%rsp) movq %r12, %rsi movslq %esi, %rdi movq %rdi, 0x78(%rsp) movq (%rsp), %r14 imull %esi, %r14d movq (%rbp), %rsi movq 0x40(%rbp), %rdi imulq 0xa0(%rsp), %rdx xorl %r8d, %r8d testl %r14d, %r14d cmovlel %r8d, %r14d testl %r13d, %r13d movl $0x0, %r15d cmovgl %r13d, %r15d imulq 0x10(%rbp), %rdi cmpq %r15, %r8 je 0x1dbcb1 xorl %r9d, %r9d cmpq %r9, %r14 je 0x1dbc65 movss (%rsi,%r9,4), %xmm0 mulss %xmm0, %xmm0 movss %xmm0, (%rcx,%r9,4) incq %r9 jmp 0x1dbc4b incq %r8 addq %rdx, %rcx addq %rdi, %rsi jmp 0x1dbc43 movl %r10d, %r14d movq 0x98(%rsp), %rax testq %rax, %rax je 0x1dbe5c lock decl (%rax) jne 0x1dbe5c movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi je 0x1dbe54 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dbe5c movl 0xd0(%rbx), %ecx testl %ecx, %ecx je 0x1dbddf xorl %r14d, %r14d cmpl $0x1, %ecx jne 0x1dbc73 movq 0x90(%rsp), %rdx movq 0x98(%rsp), %rcx movq %rdx, 0x20(%rsp) movq %rcx, 0x28(%rsp) movq 0xa0(%rsp), %rdx movq %rdx, 0x30(%rsp) movl 0xa8(%rsp), %edx movl %edx, 0x38(%rsp) movq 0xb0(%rsp), %rdx movq %rdx, 0x40(%rsp) movups 0xb8(%rsp), %xmm0 movups %xmm0, 0x48(%rsp) movl %eax, 0x58(%rsp) movq 0xd0(%rsp), %rax movq %rax, 0x60(%rsp) testq %rcx, %rcx je 0x1dbd2c lock incl (%rcx) movl 0xd4(%rbx), %r13d cmpl $0x2, %r13d jl 0x1dbea9 movl %r10d, %r14d movl %r13d, %edx shrl %edx movq 0x18(%rsp), %r11 movups (%r11), %xmm0 movups 0x10(%r11), %xmm1 movups 0x20(%r11), %xmm2 movups 0x30(%r11), %xmm3 leaq 0x110(%rsp), %rax movaps %xmm3, 0x30(%rax) movaps %xmm2, 0x20(%rax) movaps %xmm1, 0x10(%rax) movaps %xmm0, (%rax) movq 0x10(%r11), %rcx movq %rcx, 0x8(%rax) movb $0x0, 0x27(%rax) movl %edx, %ecx notl %ecx addl %r13d, %ecx leaq 0x90(%rsp), %rdi leaq 0x20(%rsp), %rsi xorps %xmm0, %xmm0 movl %edx, %r8d movl %ecx, %r9d pushq %rax pushq $0x0 callq 0x6466c popq %rax popq %rcx cmpq $0x0, 0x20(%rsp) movl %r14d, %ecx je 0x1dc271 movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x1dc271 movq %rbp, 0x10(%rsp) movl 0x4c(%rsp), %ebp movl 0xd4(%rbx), %r13d jmp 0x1dbeb1 leaq 0x20(%rsp), %rdi andq $0x0, 0x40(%rdi) xorps %xmm0, %xmm0 movaps %xmm0, (%rdi) movups %xmm0, 0xc(%rdi) movaps %xmm0, 0x20(%rdi) movups %xmm0, 0x2c(%rdi) movq 0x18(%rsp), %rax movq 0x10(%rax), %r9 movl %r12d, %esi movq (%rsp), %rdx movl %r13d, %ecx movq %r11, %r8 callq 0x63810 movq 0x20(%rsp), %rax testq %rax, %rax pushq $-0x64 popq %rdi je 0x1dbe71 movq 0x60(%rsp), %rcx movslq 0x58(%rsp), %rdx movq %rcx, %rsi imulq %rdx, %rsi testq %rsi, %rsi je 0x1dbe71 imull %edx, %ecx xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpl %edx, %ecx je 0x1dc0f2 andl $0x0, (%rax,%rdx,4) incq %rdx jmp 0x1dbe43 movq %rsi, %rdi callq 0x5f3e0 movl %r14d, %eax addq $0x158, %rsp # imm = 0x158 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc2a7 lock decl (%rax) jne 0x1dc2a7 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dc29c movq (%rdi), %rax callq 0x18(%rax) pushq $-0x64 popq %rdi jmp 0x1dc2a7 movq %rbp, 0x10(%rsp) movl %r12d, %ebp imull %r13d, %r13d movss 0xd8(%rbx), %xmm0 movss %xmm0, 0x8(%rsp) leaq 0x110(%rsp), %rdi leaq 0xf(%rsp), %rdx movq %r13, %rsi callq 0x73bbe cvtsi2ss %r13d, %xmm0 movq 0x110(%rsp), %r14 movl 0xd4(%rbx), %ecx subl %ecx, %ebp xorl %eax, %eax xorl %edx, %edx xorl %esi, %esi cmpl %ecx, %esi jge 0x1dbf24 cltq leaq (%r14,%rax,4), %r8 movl %edx, %r9d xorl %edi, %edi cmpl %ecx, %edi jge 0x1dbf19 leaq (%r9,%rdi), %rcx movl %ecx, (%r8,%rdi,4) movl 0xd4(%rbx), %ecx incq %rdi jmp 0x1dbf02 addl %ebp, %edx addl %edi, %edx incl %esi addq %rdi, %rax jmp 0x1dbef3 movq %rbx, 0x18(%rsp) movq 0x10(%rsp), %rcx movq 0x40(%rcx), %rax imulq 0x10(%rcx), %rax movq %rax, 0xf8(%rsp) movslq 0x4c(%rsp), %rdx movq 0x30(%rsp), %rax movq 0x60(%rsp), %rsi imulq %rax, %rsi movq %rsi, 0xf0(%rsp) imulq %rax, %rdx movq %rdx, 0x108(%rsp) xorl %eax, %eax movq %r12, %rbx testl %ebx, %ebx cmovlel %eax, %ebx movq (%rcx), %rcx movq %rcx, 0xe8(%rsp) movq (%rsp), %r12 testl %r12d, %r12d cmovlel %eax, %r12d movq %r12, (%rsp) xorl %eax, %eax movq 0x20(%rsp), %rcx movq %rcx, 0xe0(%rsp) movss 0x212ce7(%rip), %xmm1 # 0x3eec88 divss %xmm0, %xmm1 movss %xmm1, 0x10(%rsp) movq %r15, 0x70(%rsp) cmpq %r15, %rax je 0x1dc0b0 movq 0xf8(%rsp), %r12 imulq %rax, %r12 addq 0xe8(%rsp), %r12 movq 0xf0(%rsp), %rcx movq %rax, 0x100(%rsp) imulq %rax, %rcx addq 0xe0(%rsp), %rcx movq %rcx, 0x80(%rsp) xorl %eax, %eax cmpq (%rsp), %rax je 0x1dc09b movq 0x108(%rsp), %rbp movq %rax, 0x88(%rsp) imulq %rax, %rbp addq 0x80(%rsp), %rbp xorl %r15d, %r15d cmpq %rbx, %r15 je 0x1dc082 leaq (,%r15,4), %rax addq %rbp, %rax xorps %xmm0, %xmm0 xorl %ecx, %ecx cmpq %rcx, %r13 je 0x1dc044 movslq (%r14,%rcx,4), %rdx addss (%rax,%rdx,4), %xmm0 incq %rcx jmp 0x1dc031 mulss 0x8(%rsp), %xmm0 mulss 0x10(%rsp), %xmm0 movq 0x18(%rsp), %rax addss 0xe0(%rax), %xmm0 movss 0xdc(%rax), %xmm1 xorps 0x212024(%rip), %xmm1 # 0x3ee090 callq 0x5f0e0 mulss (%r12,%r15,4), %xmm0 movss %xmm0, (%r12,%r15,4) incq %r15 jmp 0x1dc01c movq 0x78(%rsp), %rax leaq (%r12,%rax,4), %r12 movq 0x88(%rsp), %rax incq %rax jmp 0x1dbff3 movq 0x100(%rsp), %rax incq %rax movq 0x70(%rsp), %r15 jmp 0x1dbfb0 leaq 0x110(%rsp), %rdi callq 0x624be movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc2b7 lock decl (%rax) jne 0x1dc2b7 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dc29a movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dc2b7 movq %r15, 0x70(%rsp) movl 0xd4(%rbx), %ecx cvtsi2ss %ecx, %xmm0 movss 0xd8(%rbx), %xmm1 movss %xmm1, 0x18(%rsp) movq 0x20(%rsp), %r12 movq 0x60(%rsp), %rdi pushq $-0x2 popq %rsi movl %ecx, %eax cltd idivl %esi movq %rax, (%rsp) pushq $0x2 popq %rsi movl %ecx, %eax cltd idivl %esi movq (%rbp), %r15 movq 0x40(%rbp), %rcx cltq movq %rax, 0x78(%rsp) movss 0x212b44(%rip), %xmm1 # 0x3eec88 divss %xmm0, %xmm1 movss %xmm1, 0x8(%rsp) imulq 0x30(%rsp), %rdi movq %rdi, 0x88(%rsp) imulq 0x10(%rbp), %rcx movq %rcx, 0x80(%rsp) xorl %edx, %edx cmpq 0x70(%rsp), %rdx je 0x1dc24b movq (%rsp), %rax addl %edx, %eax movq 0x78(%rsp), %rcx movq %rdx, 0x10(%rsp) addq %rdx, %rcx movq 0x90(%rsp), %rdx movq 0xd0(%rsp), %rsi imulq 0xa0(%rsp), %rsi movslq %eax, %rdi cmpq %rdi, %rcx jl 0x1dc1e8 testl %eax, %eax sets %dil cmpl %r13d, %eax setge %r8b orb %dil, %r8b jne 0x1dc1e4 movl %eax, %edi imulq %rsi, %rdi addq %rdx, %rdi xorl %r8d, %r8d cmpq %r8, %r14 je 0x1dc1e4 movss (%r12,%r8,4), %xmm0 addss (%rdi,%r8,4), %xmm0 movss %xmm0, (%r12,%r8,4) incq %r8 jmp 0x1dc1c8 incl %eax jmp 0x1dc1a2 xorl %ebp, %ebp cmpq %rbp, %r14 je 0x1dc22e movss (%r12,%rbp,4), %xmm0 mulss 0x18(%rsp), %xmm0 mulss 0x8(%rsp), %xmm0 addss 0xe0(%rbx), %xmm0 movss 0xdc(%rbx), %xmm1 xorps 0x211e78(%rip), %xmm1 # 0x3ee090 callq 0x5f0e0 mulss (%r15,%rbp,4), %xmm0 movss %xmm0, (%r15,%rbp,4) incq %rbp jmp 0x1dc1ea movq 0x10(%rsp), %rdx incq %rdx addq 0x88(%rsp), %r12 addq 0x80(%rsp), %r15 jmp 0x1dc16b movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc2b7 lock decl (%rax) jne 0x1dc2b7 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dc2af movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dc2b7 movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc2ca lock decl (%rax) jne 0x1dc2ca movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dc2bf movq (%rdi), %rax callq 0x18(%rax) movl %r14d, %ecx jmp 0x1dc2ca jmp 0x1dc2af movq %rsi, %rdi callq 0x5f3e0 pushq $-0x64 popq %rdi movl %edi, %r14d jmp 0x1dbc73 movq %rsi, %rdi callq 0x5f3e0 xorl %r14d, %r14d jmp 0x1dbc73 movq %rsi, %rdi callq 0x5f3e0 movl %r14d, %ecx movl %ecx, %r14d jmp 0x1dbc73 jmp 0x1dc38e jmp 0x1dc38e jmp 0x1dc38e jmp 0x1dc38e jmp 0x1dc2e8 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc34f lock decl (%rax) jne 0x1dc34f movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1dc30b jmp 0x1dc336 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dc34f jmp 0x1dc38e movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc34f lock decl (%rax) jne 0x1dc34f movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1dc340 movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dc34f movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dc34f jmp 0x1dc38e jmp 0x1dc38e movq %rax, %rbx movq 0x98(%rsp), %rax testq %rax, %rax je 0x1dc386 lock decl (%rax) jne 0x1dc386 movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi jne 0x1dc380 movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dc386 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/lrn.cpp
virtual thunk to ncnn::LRN_x86::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int LRN_x86::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { outptr = ptr * ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 \|\| p >= channels) continue; const float sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { ssptr += sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { ptr = ptr * powf(bias + alpha_div_size * ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }	movq (%rdi), %rax addq -0x58(%rax), %rdi jmp 0x1dc3b0	/csukuangfj[P]ncnn/src/layer/x86/lrn_x86.cpp
ncnn::LRN_x86_avx512::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int LRN_x86_avx512::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { outptr = ptr * ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 \|\| p >= channels) continue; const float sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { ssptr += sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { ptr = ptr * powf(bias + alpha_div_size * ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x178, %rsp # imm = 0x178 movq %rsi, %r15 movl 0x2c(%rsi), %esi movl 0x30(%r15), %ecx movl 0x38(%r15), %r12d movq 0x10(%r15), %r13 leaq 0x90(%rsp), %rax andq $0x0, 0x40(%rax) movq %rdx, %r14 movq %rdi, %rbx vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rax) vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, 0x20(%rax) vmovups %xmm0, 0x2c(%rax) movq 0x10(%rdx), %r9 movq %rax, %rdi movq %rsi, (%rsp) movq %rcx, 0x10(%rsp) movl %ecx, %edx movl %r12d, %ecx movq %r13, %r8 callq 0x63810 pushq $-0x64 popq %rbp cmpq $0x0, 0x90(%rsp) je 0x1dd6c2 movslq 0xc8(%rsp), %rax imulq 0xd0(%rsp), %rax testq %rax, %rax je 0x1dd6c2 movq %r13, 0x110(%rsp) movq %r14, 0x80(%rsp) movq (%rsp), %rcx movslq %ecx, %rax movq %rax, 0x108(%rsp) movq 0x10(%rsp), %rax movl %eax, %r13d imull %ecx, %r13d xorl %eax, %eax testl %r12d, %r12d movl $0x0, %ecx cmovgl %r12d, %ecx movq %rcx, 0x68(%rsp) movq %rbx, 0x18(%rsp) cmpq 0x68(%rsp), %rax je 0x1dcdff movq (%r15), %rcx movq 0x10(%r15), %r8 movq 0x40(%r15), %rdi movq %rdi, %r9 imulq %rax, %r9 imulq %r8, %r9 addq %rcx, %r9 movq 0x90(%rsp), %rdx movq 0xa0(%rsp), %r11 movq 0xd0(%rsp), %r10 movq %r10, %rbx imulq %rax, %rbx imulq %r11, %rbx addq %rdx, %rbx xorl %esi, %esi xorl %r14d, %r14d leal 0x7(%r14), %ebp cmpl %r13d, %ebp jge 0x1dcdc4 vmovups (%r9), %ymm0 vmulps %ymm0, %ymm0, %ymm0 vmovups %ymm0, (%rbx) addq $0x20, %r9 addq $0x20, %rbx addl $0x8, %r14d addq $0x8, %rsi jmp 0x1dcd9c imulq %r11, %r10 imulq %rax, %r10 addq %r10, %rdx imulq %r8, %rdi imulq %rax, %rdi addq %rdi, %rcx movq 0x18(%rsp), %rbx cmpl %r13d, %esi jge 0x1dcdf7 vmovss (%rcx,%rsi,4), %xmm0 vmulss %xmm0, %xmm0, %xmm0 vmovss %xmm0, (%rdx,%rsi,4) incq %rsi jmp 0x1dcddf incq %rax jmp 0x1dcd4d movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd0(%rbx,%rcx), %ecx testl %ecx, %ecx je 0x1dcf42 xorl %ebp, %ebp cmpl $0x1, %ecx jne 0x1dd6c2 movq 0x98(%rsp), %rcx vmovaps 0x90(%rsp), %xmm0 vmovaps %xmm0, 0x20(%rsp) movq 0xa0(%rsp), %rdx movq %rdx, 0x30(%rsp) movl 0xa8(%rsp), %edx movl %edx, 0x38(%rsp) movq 0xb0(%rsp), %rdx movq %rdx, 0x40(%rsp) vmovups 0xb8(%rsp), %xmm0 vmovups %xmm0, 0x48(%rsp) movl 0xc8(%rsp), %edx movl %edx, 0x58(%rsp) movq 0xd0(%rsp), %rdx movq %rdx, 0x60(%rsp) testq %rcx, %rcx je 0x1dce8e lock incl (%rcx) movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %r13d cmpl $0x2, %r13d jl 0x1dcffe shrl %r13d movq 0x80(%rsp), %rcx vmovups (%rcx), %zmm0 leaq 0x130(%rsp), %r10 vmovups %zmm0, (%r10) movq 0x10(%rcx), %rcx movq %rcx, 0x8(%r10) movq -0x18(%rax), %rax movl %r13d, %ecx notl %ecx addl 0xd4(%rbx,%rax), %ecx leaq 0x90(%rsp), %rdi leaq 0x20(%rsp), %rsi vxorps %xmm0, %xmm0, %xmm0 movl %r13d, %edx movl %r13d, %r8d movl %ecx, %r9d pushq %r10 pushq $0x0 vzeroupper callq 0x6466c popq %rax popq %rcx cmpq $0x0, 0x20(%rsp) pushq $-0x64 popq %rbp je 0x1dd68b movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x1dd68b movq %r15, %r14 movl 0x4c(%rsp), %r15d movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %r13d jmp 0x1dd008 leaq 0x20(%rsp), %rdi andq $0x0, 0x40(%rdi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rdi) vmovups %xmm0, 0xc(%rdi) vmovaps %xmm0, 0x20(%rdi) vmovups %xmm0, 0x2c(%rdi) movq 0x80(%rsp), %rax movq 0x10(%rax), %r9 movq (%rsp), %rsi movq 0x10(%rsp), %rdx movl %r12d, %ecx movq 0x110(%rsp), %r8 vzeroupper callq 0x63810 movq 0x20(%rsp), %rax testq %rax, %rax je 0x1dcfc6 movq 0x60(%rsp), %rcx movslq 0x58(%rsp), %rdx movq %rcx, %rsi imulq %rdx, %rsi testq %rsi, %rsi je 0x1dcfc6 imull %edx, %ecx xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpl %edx, %ecx je 0x1dd26a andl $0x0, (%rax,%rdx,4) incq %rdx jmp 0x1dcfb5 movq 0x28(%rsp), %rax testq %rax, %rax pushq $-0x64 popq %rbp je 0x1dd6c2 lock decl (%rax) jne 0x1dd6c2 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dd6b3 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dd6c2 movq %r15, %r14 movq (%rsp), %rax movl %eax, %r15d imull %r13d, %r13d vmovss 0xd8(%rbx,%rcx), %xmm0 vmovss %xmm0, 0x80(%rsp) leaq 0x130(%rsp), %rdi leaq 0xf(%rsp), %rdx movq %r13, %rsi vzeroupper callq 0x73bbe vcvtsi2ss %r13d, %xmm1, %xmm0 movq 0x130(%rsp), %rbp movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %edx subl %edx, %r15d xorl %ecx, %ecx xorl %esi, %esi xorl %edi, %edi cmpl %edx, %edi jge 0x1dd09c movslq %ecx, %rcx leaq (,%rcx,4), %r9 addq %rbp, %r9 movl %esi, %r10d xorl %r8d, %r8d cmpl %edx, %r8d jge 0x1dd08f leaq (%r10,%r8), %rdx movl %edx, (%r9,%r8,4) movq -0x18(%rax), %rdx movl 0xd4(%rbx,%rdx), %edx incq %r8 jmp 0x1dd072 addl %r15d, %esi addl %r8d, %esi incl %edi addq %r8, %rcx jmp 0x1dd05a movq (%r14), %rax movq %rax, 0xf0(%rsp) movq 0x40(%r14), %rax imulq 0x10(%r14), %rax movq %rax, 0xe8(%rsp) movslq 0x4c(%rsp), %rcx movq 0x30(%rsp), %rax movq 0x60(%rsp), %rdx imulq %rax, %rdx movq %rdx, 0xe0(%rsp) imulq %rax, %rcx movq %rcx, 0x100(%rsp) xorl %ecx, %ecx movq (%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, (%rsp) movq 0x20(%rsp), %rax movq %rax, 0xd8(%rsp) movq 0x10(%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, 0x10(%rsp) vmovss 0x211b76(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0x110(%rsp) cmpq 0x68(%rsp), %rcx je 0x1dd228 movq 0xe8(%rsp), %r14 imulq %rcx, %r14 addq 0xf0(%rsp), %r14 movq 0xe0(%rsp), %rax movq %rcx, 0xf8(%rsp) imulq %rcx, %rax addq 0xd8(%rsp), %rax movq %rax, 0x70(%rsp) xorl %eax, %eax cmpq 0x10(%rsp), %rax je 0x1dd218 movq 0x100(%rsp), %r15 movq %rax, 0x78(%rsp) imulq %rax, %r15 addq 0x70(%rsp), %r15 movq (%rbx), %r12 xorl %ebx, %ebx cmpq (%rsp), %rbx je 0x1dd1fa leaq (%r15,%rbx,4), %rax vxorps %xmm0, %xmm0, %xmm0 xorl %ecx, %ecx cmpq %rcx, %r13 je 0x1dd1ab movslq (%rbp,%rcx,4), %rdx vaddss (%rax,%rdx,4), %xmm0, %xmm0 incq %rcx jmp 0x1dd197 movq -0x18(%r12), %rax vmulss 0x80(%rsp), %xmm0, %xmm0 movq 0x18(%rsp), %rcx vmovss 0x110(%rsp), %xmm1 vfmadd213ss 0xe0(%rcx,%rax), %xmm1, %xmm0 # xmm0 = (xmm1 xmm0) + mem vmovss 0xdc(%rcx,%rax), %xmm1 vxorps 0x213fd0(%rip){1to4}, %xmm1, %xmm1 # 0x3f11b4 callq 0x5f0e0 vmulss (%r14,%rbx,4), %xmm0, %xmm0 vmovss %xmm0, (%r14,%rbx,4) incq %rbx jmp 0x1dd187 movq 0x108(%rsp), %rax leaq (%r14,%rax,4), %r14 movq 0x78(%rsp), %rax incq %rax movq 0x18(%rsp), %rbx jmp 0x1dd161 movq 0xf8(%rsp), %rcx incq %rcx jmp 0x1dd11f leaq 0x130(%rsp), %rdi callq 0x624be movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd6c0 lock decl (%rax) jne 0x1dd6c0 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dd6b1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dd6c0 movq %r15, 0x78(%rsp) movq (%rbx), %rax movq %rax, (%rsp) movq -0x18(%rax), %rax vmovss 0xd8(%rbx,%rax), %xmm0 vcvtsi2ssl 0xd4(%rbx,%rax), %xmm1, %xmm1 vdivss %xmm1, %xmm0, %xmm0 vmovaps %xmm0, 0x80(%rsp) vbroadcastss %xmm0, %ymm0 vmovups %ymm0, 0x110(%rsp) xorl %r15d, %r15d pushq $0x2 popq %r14 movl %r12d, 0x70(%rsp) cmpq 0x68(%rsp), %r15 je 0x1dd662 movq (%rsp), %rax movq -0x18(%rax), %r9 movl 0xd4(%rbx,%r9), %esi movl %esi, %eax cltd pushq $-0x2 popq %rcx idivl %ecx movl %eax, %ecx addl %r15d, %ecx movl %esi, %eax cltd idivl %r14d addl %r15d, %eax cmpl %eax, %ecx jg 0x1dd3ae movl %ecx, %ecx testl %ecx, %ecx sets %al cmpl %r12d, %ecx setge %dl orb %al, %dl jne 0x1dd3a7 movq 0x90(%rsp), %rax movq 0xa0(%rsp), %r8 movq 0xd0(%rsp), %rdi movq %rdi, %r9 imulq %rcx, %r9 imulq %r8, %r9 addq %rax, %r9 movq 0x60(%rsp), %r10 imulq %r15, %r10 imulq 0x30(%rsp), %r10 movq 0x20(%rsp), %rdx leaq (%rdx,%r10), %r11 xorl %esi, %esi xorl %ebx, %ebx leal 0x7(%rbx), %ebp cmpl %r13d, %ebp jge 0x1dd36b vmovups (%r11), %ymm0 vaddps (%r9), %ymm0, %ymm0 vmovups %ymm0, (%r11) addq $0x20, %r9 addq $0x20, %r11 addl $0x8, %ebx addq $0x8, %rsi jmp 0x1dd343 addq %r10, %rdx imulq %r8, %rdi imulq %rcx, %rdi addq %rdi, %rax movq 0x18(%rsp), %rbx cmpl %r13d, %esi jge 0x1dd397 vmovss (%rdx,%rsi,4), %xmm0 vaddss (%rax,%rsi,4), %xmm0, %xmm0 vmovss %xmm0, (%rdx,%rsi,4) incq %rsi jmp 0x1dd37e movq (%rsp), %rax movq -0x18(%rax), %r9 movl 0xd4(%rbx,%r9), %esi incl %ecx jmp 0x1dd2db movq 0x78(%rsp), %rax movq (%rax), %r12 movq 0x10(%rax), %rcx movq 0x40(%rax), %rax movq %rax, %rdi imulq %r15, %rdi imulq %rcx, %rdi addq %r12, %rdi movq 0x20(%rsp), %rbp movq 0x30(%rsp), %rsi movq 0x60(%rsp), %rdx movq %rdx, %r8 imulq %r15, %r8 imulq %rsi, %r8 addq %rbp, %r8 vbroadcastss 0xe0(%rbx,%r9), %ymm0 vmovss 0xdc(%rbx,%r9), %xmm1 vxorps 0x213dad(%rip){1to4}, %xmm1, %xmm1 # 0x3f11b4 vbroadcastss %xmm1, %ymm1 xorl %r14d, %r14d xorl %r9d, %r9d leal 0x7(%r9), %r10d cmpl %r13d, %r10d jge 0x1dd5e8 vmovups (%r8), %ymm2 vfmadd132ps 0x110(%rsp), %ymm0, %ymm2 # ymm2 = (ymm2 mem) + ymm0 vcmpleps 0x215587(%rip), %ymm2, %k1 # 0x3f29c0 vmaxps 0x213d9d(%rip){1to8}, %ymm2, %ymm2 # 0x3f11e0 vpsrld $0x17, %ymm2, %ymm3 vpbroadcastd 0x213d93(%rip), %ymm4 # 0x3f11e4 vpternlogd $0xea, 0x210bb8(%rip){1to8}, %ymm4, %ymm2 # 0x3ee014 vcmpltps 0x213d85(%rip){1to8}, %ymm2, %k2 # 0x3f11ec vaddps 0x213d7f(%rip){1to8}, %ymm2, %ymm4 # 0x3f11f0 vaddps %ymm2, %ymm4, %ymm4 {%k2} vmulps %ymm4, %ymm4, %ymm2 vbroadcastss 0x213d70(%rip), %ymm5 # 0x3f11f4 vfmadd213ps 0x213d6a(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d64(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d5e(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d58(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d52(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d4c(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d46(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d40(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vmulps %ymm4, %ymm2, %ymm6 vmulps %ymm5, %ymm6, %ymm5 vpaddd 0x213d02(%rip){1to8}, %ymm3, %ymm3 # 0x3f11e8 vcvtdq2ps %ymm3, %ymm3 vbroadcastss 0x211795(%rip), %ymm6 # 0x3eec88 vsubps %ymm6, %ymm3, %ymm3 {%k2} vbroadcastss 0x213cc6(%rip), %ymm7 # 0x3f11c8 vfmadd231ps %ymm7, %ymm3, %ymm5 # ymm5 = (ymm3 * ymm7) + ymm5 vbroadcastss 0x210b04(%rip), %ymm8 # 0x3ee014 vfmsub231ps %ymm2, %ymm8, %ymm5 # ymm5 = (ymm8 * ymm2) - ymm5 vsubps %ymm4, %ymm5, %ymm2 vbroadcastss 0x213ca2(%rip), %ymm5 # 0x3f11c4 vfmsub231ps %ymm3, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm3) - ymm2 vpcmpeqd %ymm3, %ymm3, %ymm3 vmovaps %ymm3, %ymm2 {%k1} vmulps %ymm1, %ymm2, %ymm2 vminps 0x213c79(%rip){1to8}, %ymm2, %ymm2 # 0x3f11b8 vmaxps 0x213c73(%rip){1to8}, %ymm2, %ymm2 # 0x3f11bc vmovaps %ymm8, %ymm3 vfmadd231ps 0x213c69(%rip){1to8}, %ymm2, %ymm3 # ymm3 = (ymm2 * mem) + ymm3 vroundps $0x1, %ymm3, %ymm4 vcmpltps %ymm4, %ymm3, %k1 vsubps %ymm6, %ymm4, %ymm4 {%k1} vfmsub231ps %ymm5, %ymm4, %ymm2 # ymm2 = (ymm4 * ymm5) - ymm2 vfnmsub231ps %ymm7, %ymm4, %ymm2 # ymm2 = -(ymm4 * ymm7) - ymm2 vmulps %ymm2, %ymm2, %ymm3 vbroadcastss 0x213c4b(%rip), %ymm5 # 0x3f11cc vfmadd213ps 0x213c45(%rip){1to8}, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + mem vfmadd213ps 0x213c3f(%rip){1to8}, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + mem vfmadd213ps 0x213c39(%rip){1to8}, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + mem vfmadd213ps 0x213c33(%rip){1to8}, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + mem vfmadd213ps %ymm8, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm8 vfmadd213ps %ymm2, %ymm3, %ymm5 # ymm5 = (ymm3 * ymm5) + ymm2 vcvttps2dq %ymm4, %ymm2 vpslld $0x17, %ymm2, %ymm2 vpaddd 0x2116c2(%rip){1to8}, %ymm2, %ymm2 # 0x3eec88 vmulps (%rdi), %ymm2, %ymm2 vfmadd213ps %ymm2, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm2) + ymm2 vmovups %ymm2, (%rdi) addq $0x20, %r8 addq $0x20, %rdi addl $0x8, %r9d addq $0x8, %r14 jmp 0x1dd412 imulq %rcx, %rax imulq %r15, %rax addq %rax, %r12 imulq %rsi, %rdx imulq %r15, %rdx addq %rdx, %rbp cmpl %r13d, %r14d jge 0x1dd651 movq (%rsp), %rax movq -0x18(%rax), %rax vmovss (%rbp,%r14,4), %xmm0 vmovaps 0x80(%rsp), %xmm1 vfmadd213ss 0xe0(%rbx,%rax), %xmm1, %xmm0 # xmm0 = (xmm1 * xmm0) + mem vmovss 0xdc(%rbx,%rax), %xmm1 vxorps 0x213b7c(%rip){1to4}, %xmm1, %xmm1 # 0x3f11b4 vzeroupper callq 0x5f0e0 vmulss (%r12,%r14,4), %xmm0, %xmm0 vmovss %xmm0, (%r12,%r14,4) incq %r14 jmp 0x1dd5fe incq %r15 movl 0x70(%rsp), %r12d pushq $0x2 popq %r14 jmp 0x1dd2b3 movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd6c0 lock decl (%rax) jne 0x1dd6c0 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dd6b5 movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x1dd6c0 movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd6c2 lock decl (%rax) jne 0x1dd6c2 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dd716 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dd6c2 jmp 0x1dd6b5 jmp 0x1dd716 movq %rsi, %rdi vzeroupper callq 0x5f3e0 xorl %ebp, %ebp movq 0x98(%rsp), %rax testq %rax, %rax je 0x1dd6ff lock decl (%rax) jne 0x1dd6ff movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi je 0x1dd6f4 movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x1dd6ff movq %rsi, %rdi vzeroupper callq 0x5f3e0 movl %ebp, %eax addq $0x178, %rsp # imm = 0x178 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dd6c2 jmp 0x1dd7dc jmp 0x1dd7dc jmp 0x1dd7dc jmp 0x1dd7dc jmp 0x1dd736 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd79d lock decl (%rax) jne 0x1dd79d movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1dd759 jmp 0x1dd784 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dd79d jmp 0x1dd7dc movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd79d lock decl (%rax) jne 0x1dd79d movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1dd78e movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dd79d movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dd79d jmp 0x1dd7dc jmp 0x1dd7dc movq %rax, %rbx movq 0x98(%rsp), %rax testq %rax, %rax je 0x1dd7d4 lock decl (%rax) jne 0x1dd7d4 movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi jne 0x1dd7ce movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dd7d4 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/build_O2/src/layer/x86/lrn_x86_avx512.cpp
virtual thunk to ncnn::LRN_x86_avx512::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int LRN_x86_avx512::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { outptr = ptr * ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 \|\| p >= channels) continue; const float sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { ssptr += sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { ptr = ptr * powf(bias + alpha_div_size * ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }	movq (%rdi), %rax addq -0x58(%rax), %rdi jmp 0x1dcc74	/csukuangfj[P]ncnn/build_O2/src/layer/x86/lrn_x86_avx512.cpp
ncnn::LRN_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int LRN_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { outptr = ptr * ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 \|\| p >= channels) continue; const float sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { ssptr += sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { ptr = ptr * powf(bias + alpha_div_size * ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x178, %rsp # imm = 0x178 movq %rsi, %r15 movl 0x2c(%rsi), %esi movl 0x30(%r15), %ecx movl 0x38(%r15), %r12d movq 0x10(%r15), %r13 leaq 0x90(%rsp), %rax andq $0x0, 0x40(%rax) movq %rdx, %r14 movq %rdi, %rbx vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rax) vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, 0x20(%rax) vmovups %xmm0, 0x2c(%rax) movq 0x10(%rdx), %r9 movq %rax, %rdi movq %rsi, (%rsp) movq %rcx, 0x10(%rsp) movl %ecx, %edx movl %r12d, %ecx movq %r13, %r8 callq 0x63810 pushq $-0x64 popq %rbp cmpq $0x0, 0x90(%rsp) je 0x1de30e movslq 0xc8(%rsp), %rax imulq 0xd0(%rsp), %rax testq %rax, %rax je 0x1de30e movq %r13, 0x110(%rsp) movq %r14, 0x80(%rsp) movq (%rsp), %rcx movslq %ecx, %rax movq %rax, 0x108(%rsp) movq 0x10(%rsp), %rax movl %eax, %r13d imull %ecx, %r13d xorl %eax, %eax testl %r12d, %r12d movl $0x0, %ecx cmovgl %r12d, %ecx movq %rcx, 0x68(%rsp) movq %rbx, 0x18(%rsp) cmpq 0x68(%rsp), %rax je 0x1dd9c3 movq (%r15), %rcx movq 0x10(%r15), %r8 movq 0x40(%r15), %rdi movq %rdi, %r9 imulq %rax, %r9 imulq %r8, %r9 addq %rcx, %r9 movq 0x90(%rsp), %rdx movq 0xa0(%rsp), %r11 movq 0xd0(%rsp), %r10 movq %r10, %rbx imulq %rax, %rbx imulq %r11, %rbx addq %rdx, %rbx xorl %esi, %esi xorl %r14d, %r14d leal 0x7(%r14), %ebp cmpl %r13d, %ebp jge 0x1dd988 vmovups (%r9), %ymm0 vmulps %ymm0, %ymm0, %ymm0 vmovups %ymm0, (%rbx) addq $0x20, %r9 addq $0x20, %rbx addl $0x8, %r14d addq $0x8, %rsi jmp 0x1dd960 imulq %r11, %r10 imulq %rax, %r10 addq %r10, %rdx imulq %r8, %rdi imulq %rax, %rdi addq %rdi, %rcx movq 0x18(%rsp), %rbx cmpl %r13d, %esi jge 0x1dd9bb vmovss (%rcx,%rsi,4), %xmm0 vmulss %xmm0, %xmm0, %xmm0 vmovss %xmm0, (%rdx,%rsi,4) incq %rsi jmp 0x1dd9a3 incq %rax jmp 0x1dd911 movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd0(%rbx,%rcx), %ecx testl %ecx, %ecx je 0x1ddb0e xorl %ebp, %ebp cmpl $0x1, %ecx jne 0x1de30e movq 0x98(%rsp), %rcx vmovaps 0x90(%rsp), %xmm0 vmovaps %xmm0, 0x20(%rsp) movq 0xa0(%rsp), %rdx movq %rdx, 0x30(%rsp) movl 0xa8(%rsp), %edx movl %edx, 0x38(%rsp) movq 0xb0(%rsp), %rdx movq %rdx, 0x40(%rsp) vmovups 0xb8(%rsp), %xmm0 vmovups %xmm0, 0x48(%rsp) movl 0xc8(%rsp), %edx movl %edx, 0x58(%rsp) movq 0xd0(%rsp), %rdx movq %rdx, 0x60(%rsp) testq %rcx, %rcx je 0x1dda52 lock incl (%rcx) movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %r13d cmpl $0x2, %r13d jl 0x1ddbca shrl %r13d movq 0x80(%rsp), %rcx vmovups (%rcx), %ymm0 vmovups 0x20(%rcx), %ymm1 leaq 0x130(%rsp), %r10 vmovups %ymm1, 0x20(%r10) vmovups %ymm0, (%r10) movq 0x10(%rcx), %rcx movq %rcx, 0x8(%r10) movq -0x18(%rax), %rax movl %r13d, %ecx notl %ecx addl 0xd4(%rbx,%rax), %ecx leaq 0x90(%rsp), %rdi leaq 0x20(%rsp), %rsi vxorps %xmm0, %xmm0, %xmm0 movl %r13d, %edx movl %r13d, %r8d movl %ecx, %r9d pushq %r10 pushq $0x0 vzeroupper callq 0x6466c popq %rax popq %rcx cmpq $0x0, 0x20(%rsp) pushq $-0x64 popq %rbp je 0x1de2d7 movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x1de2d7 movq %r15, %r14 movl 0x4c(%rsp), %r15d movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %r13d jmp 0x1ddbd4 leaq 0x20(%rsp), %rdi andq $0x0, 0x40(%rdi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rdi) vmovups %xmm0, 0xc(%rdi) vmovaps %xmm0, 0x20(%rdi) vmovups %xmm0, 0x2c(%rdi) movq 0x80(%rsp), %rax movq 0x10(%rax), %r9 movq (%rsp), %rsi movq 0x10(%rsp), %rdx movl %r12d, %ecx movq 0x110(%rsp), %r8 vzeroupper callq 0x63810 movq 0x20(%rsp), %rax testq %rax, %rax je 0x1ddb92 movq 0x60(%rsp), %rcx movslq 0x58(%rsp), %rdx movq %rcx, %rsi imulq %rdx, %rsi testq %rsi, %rsi je 0x1ddb92 imull %edx, %ecx xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpl %edx, %ecx je 0x1dde3b andl $0x0, (%rax,%rdx,4) incq %rdx jmp 0x1ddb81 movq 0x28(%rsp), %rax testq %rax, %rax pushq $-0x64 popq %rbp je 0x1de30e lock decl (%rax) jne 0x1de30e movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1de2ff movq (%rdi), %rax callq 0x18(%rax) jmp 0x1de30e movq %r15, %r14 movq (%rsp), %rax movl %eax, %r15d imull %r13d, %r13d vmovss 0xd8(%rbx,%rcx), %xmm0 vmovss %xmm0, 0x80(%rsp) leaq 0x130(%rsp), %rdi leaq 0xf(%rsp), %rdx movq %r13, %rsi vzeroupper callq 0x73bbe vcvtsi2ss %r13d, %xmm2, %xmm0 movq 0x130(%rsp), %rbp movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %edx subl %edx, %r15d xorl %ecx, %ecx xorl %esi, %esi xorl %edi, %edi cmpl %edx, %edi jge 0x1ddc68 movslq %ecx, %rcx leaq (,%rcx,4), %r9 addq %rbp, %r9 movl %esi, %r10d xorl %r8d, %r8d cmpl %edx, %r8d jge 0x1ddc5b leaq (%r10,%r8), %rdx movl %edx, (%r9,%r8,4) movq -0x18(%rax), %rdx movl 0xd4(%rbx,%rdx), %edx incq %r8 jmp 0x1ddc3e addl %r15d, %esi addl %r8d, %esi incl %edi addq %r8, %rcx jmp 0x1ddc26 movq 0x40(%r14), %rax imulq 0x10(%r14), %rax movq %rax, 0xf0(%rsp) movslq 0x4c(%rsp), %rcx movq 0x30(%rsp), %rax movq 0x60(%rsp), %rdx imulq %rax, %rdx movq %rdx, 0xe8(%rsp) imulq %rax, %rcx movq %rcx, 0x100(%rsp) xorl %ecx, %ecx movq (%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, (%rsp) movq (%r14), %rax movq %rax, 0xe0(%rsp) xorl %ecx, %ecx movq 0x10(%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, 0x10(%rsp) movq 0x20(%rsp), %rax movq %rax, 0xd8(%rsp) vmovss 0x210fa8(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0x110(%rsp) cmpq 0x68(%rsp), %rcx je 0x1dddf9 movq 0xf0(%rsp), %r14 imulq %rcx, %r14 addq 0xe0(%rsp), %r14 movq 0xe8(%rsp), %rax movq %rcx, 0xf8(%rsp) imulq %rcx, %rax addq 0xd8(%rsp), %rax movq %rax, 0x70(%rsp) xorl %eax, %eax cmpq 0x10(%rsp), %rax je 0x1ddde9 movq 0x100(%rsp), %r15 movq %rax, 0x78(%rsp) imulq %rax, %r15 addq 0x70(%rsp), %r15 movq (%rbx), %r12 xorl %ebx, %ebx cmpq (%rsp), %rbx je 0x1dddcb leaq (%r15,%rbx,4), %rax vxorps %xmm0, %xmm0, %xmm0 xorl %ecx, %ecx cmpq %rcx, %r13 je 0x1ddd79 movslq (%rbp,%rcx,4), %rdx vaddss (%rax,%rdx,4), %xmm0, %xmm0 incq %rcx jmp 0x1ddd65 movq -0x18(%r12), %rax vmulss 0x80(%rsp), %xmm0, %xmm0 movq 0x18(%rsp), %rcx vmovss 0x110(%rsp), %xmm1 vfmadd213ss 0xe0(%rcx,%rax), %xmm1, %xmm0 # xmm0 = (xmm1 xmm0) + mem vmovss 0xdc(%rcx,%rax), %xmm1 vbroadcastss 0x213403(%rip), %xmm2 # 0x3f11b4 vxorps %xmm2, %xmm1, %xmm1 callq 0x5f0e0 vmulss (%r14,%rbx,4), %xmm0, %xmm0 vmovss %xmm0, (%r14,%rbx,4) incq %rbx jmp 0x1ddd55 movq 0x108(%rsp), %rax leaq (%r14,%rax,4), %r14 movq 0x78(%rsp), %rax incq %rax movq 0x18(%rsp), %rbx jmp 0x1ddd2f movq 0xf8(%rsp), %rcx incq %rcx jmp 0x1ddced leaq 0x130(%rsp), %rdi callq 0x624be movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de30c lock decl (%rax) jne 0x1de30c movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1de2fd movq (%rdi), %rax callq 0x18(%rax) jmp 0x1de30c movq %r15, 0x78(%rsp) movq (%rbx), %rax movq %rax, (%rsp) movq -0x18(%rax), %rax vmovss 0xd8(%rbx,%rax), %xmm0 vcvtsi2ssl 0xd4(%rbx,%rax), %xmm1, %xmm1 vdivss %xmm1, %xmm0, %xmm0 vmovaps %xmm0, 0x80(%rsp) vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovups %ymm0, 0x110(%rsp) xorl %r15d, %r15d pushq $0x2 popq %r14 movl %r12d, 0x70(%rsp) cmpq 0x68(%rsp), %r15 je 0x1de2ae movq (%rsp), %rax movq -0x18(%rax), %r9 movl 0xd4(%rbx,%r9), %esi movl %esi, %eax cltd pushq $-0x2 popq %rcx idivl %ecx movl %eax, %ecx addl %r15d, %ecx movl %esi, %eax cltd idivl %r14d addl %r15d, %eax cmpl %eax, %ecx jg 0x1ddf85 movl %ecx, %ecx testl %ecx, %ecx sets %al cmpl %r12d, %ecx setge %dl orb %al, %dl jne 0x1ddf7e movq 0x90(%rsp), %rax movq 0xa0(%rsp), %r8 movq 0xd0(%rsp), %rdi movq %rdi, %r9 imulq %rcx, %r9 imulq %r8, %r9 addq %rax, %r9 movq 0x60(%rsp), %r10 imulq %r15, %r10 imulq 0x30(%rsp), %r10 movq 0x20(%rsp), %rdx leaq (%rdx,%r10), %r11 xorl %esi, %esi xorl %ebx, %ebx leal 0x7(%rbx), %ebp cmpl %r13d, %ebp jge 0x1ddf42 vmovups (%r11), %ymm0 vaddps (%r9), %ymm0, %ymm0 vmovups %ymm0, (%r11) addq $0x20, %r9 addq $0x20, %r11 addl $0x8, %ebx addq $0x8, %rsi jmp 0x1ddf1a addq %r10, %rdx imulq %r8, %rdi imulq %rcx, %rdi addq %rdi, %rax movq 0x18(%rsp), %rbx cmpl %r13d, %esi jge 0x1ddf6e vmovss (%rdx,%rsi,4), %xmm0 vaddss (%rax,%rsi,4), %xmm0, %xmm0 vmovss %xmm0, (%rdx,%rsi,4) incq %rsi jmp 0x1ddf55 movq (%rsp), %rax movq -0x18(%rax), %r9 movl 0xd4(%rbx,%r9), %esi incl %ecx jmp 0x1ddeb2 movq 0x78(%rsp), %rax movq (%rax), %r12 movq 0x10(%rax), %rcx movq 0x40(%rax), %rax movq %rax, %rdi imulq %r15, %rdi imulq %rcx, %rdi addq %r12, %rdi movq 0x20(%rsp), %rbp movq 0x30(%rsp), %rsi movq 0x60(%rsp), %rdx movq %rdx, %r8 imulq %r15, %r8 imulq %rsi, %r8 addq %rbp, %r8 vbroadcastss 0xe0(%rbx,%r9), %ymm0 vbroadcastss 0xdc(%rbx,%r9), %xmm1 vbroadcastss 0x2131d7(%rip), %xmm2 # 0x3f11b4 vxorps %xmm2, %xmm1, %xmm1 vinsertf128 $0x1, %xmm1, %ymm1, %ymm1 xorl %r14d, %r14d xorl %r9d, %r9d leal 0x7(%r9), %r10d cmpl %r13d, %r10d jge 0x1de231 vmovups (%r8), %ymm2 vfmadd132ps 0x110(%rsp), %ymm0, %ymm2 # ymm2 = (ymm2 mem) + ymm0 vbroadcastss 0x2131ce(%rip), %ymm3 # 0x3f11e0 vmaxps %ymm3, %ymm2, %ymm3 vpsrld $0x17, %xmm3, %xmm4 vextractf128 $0x1, %ymm3, %xmm5 vpsrld $0x17, %xmm5, %xmm5 vbroadcastss 0x2131b5(%rip), %ymm6 # 0x3f11e4 vandps %ymm6, %ymm3, %ymm3 vbroadcastss 0x20ffd8(%rip), %ymm8 # 0x3ee014 vorps %ymm3, %ymm8, %ymm3 vbroadcastss 0x2131a3(%rip), %ymm6 # 0x3f11ec vcmpleps %ymm3, %ymm6, %ymm6 vandnps %ymm3, %ymm6, %ymm7 vbroadcastss 0x213195(%rip), %ymm9 # 0x3f11f0 vaddps %ymm3, %ymm9, %ymm3 vaddps %ymm7, %ymm3, %ymm3 vextractf128 $0x1, %ymm6, %xmm7 vpsubd %xmm7, %xmm5, %xmm5 vbroadcastss 0x21388e(%rip), %xmm7 # 0x3f1904 vpaddd %xmm7, %xmm5, %xmm5 vpsubd %xmm6, %xmm4, %xmm4 vpaddd %xmm7, %xmm4, %xmm4 vinsertf128 $0x1, %xmm5, %ymm4, %ymm4 vmulps %ymm3, %ymm3, %ymm5 vbroadcastss 0x21315f(%rip), %ymm6 # 0x3f11f4 vbroadcastss 0x21315a(%rip), %ymm7 # 0x3f11f8 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213150(%rip), %ymm7 # 0x3f11fc vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213146(%rip), %ymm7 # 0x3f1200 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x21313c(%rip), %ymm7 # 0x3f1204 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213132(%rip), %ymm7 # 0x3f1208 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213128(%rip), %ymm7 # 0x3f120c vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x21311e(%rip), %ymm7 # 0x3f1210 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213114(%rip), %ymm7 # 0x3f1214 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vmulps %ymm3, %ymm5, %ymm7 vmulps %ymm6, %ymm7, %ymm6 vcvtdq2ps %ymm4, %ymm4 vbroadcastss 0x2130ae(%rip), %ymm7 # 0x3f11c8 vfmadd231ps %ymm7, %ymm4, %ymm6 # ymm6 = (ymm4 * ymm7) + ymm6 vfmsub231ps %ymm5, %ymm8, %ymm6 # ymm6 = (ymm8 * ymm5) - ymm6 vcmpleps 0x214893(%rip), %ymm2, %ymm2 # 0x3f29c0 vsubps %ymm3, %ymm6, %ymm3 vbroadcastss 0x21308a(%rip), %ymm5 # 0x3f11c4 vfmsub231ps %ymm4, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm4) - ymm3 vorps %ymm3, %ymm2, %ymm2 vmulps %ymm1, %ymm2, %ymm2 vbroadcastss 0x213068(%rip), %ymm3 # 0x3f11b8 vminps %ymm3, %ymm2, %ymm2 vbroadcastss 0x21305f(%rip), %ymm3 # 0x3f11bc vmaxps %ymm3, %ymm2, %ymm2 vbroadcastss 0x213056(%rip), %ymm3 # 0x3f11c0 vfmadd213ps %ymm8, %ymm2, %ymm3 # ymm3 = (ymm2 * ymm3) + ymm8 vroundps $0x1, %ymm3, %ymm4 vcmpltps %ymm4, %ymm3, %ymm3 vbroadcastss 0x210b05(%rip), %ymm6 # 0x3eec88 vandps %ymm6, %ymm3, %ymm3 vsubps %ymm3, %ymm4, %ymm3 vfmsub231ps %ymm5, %ymm3, %ymm2 # ymm2 = (ymm3 * ymm5) - ymm2 vfnmsub231ps %ymm7, %ymm3, %ymm2 # ymm2 = -(ymm3 * ymm7) - ymm2 vmulps %ymm2, %ymm2, %ymm4 vbroadcastss 0x21302a(%rip), %ymm5 # 0x3f11cc vbroadcastss 0x213025(%rip), %ymm6 # 0x3f11d0 vfmadd213ps %ymm6, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm6 vbroadcastss 0x21301b(%rip), %ymm6 # 0x3f11d4 vfmadd213ps %ymm6, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm6 vbroadcastss 0x213011(%rip), %ymm6 # 0x3f11d8 vfmadd213ps %ymm6, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm6 vbroadcastss 0x213007(%rip), %ymm6 # 0x3f11dc vfmadd213ps %ymm6, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm6 vfmadd213ps %ymm8, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm8 vfmadd213ps %ymm2, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + ymm2 vcvttps2dq %ymm3, %ymm2 vpslld $0x17, %xmm2, %xmm3 vextractf128 $0x1, %ymm2, %xmm2 vpslld $0x17, %xmm2, %xmm2 vbroadcastss 0x210a87(%rip), %xmm4 # 0x3eec88 vpaddd %xmm4, %xmm2, %xmm2 vpaddd %xmm4, %xmm3, %xmm3 vinsertf128 $0x1, %xmm2, %ymm3, %ymm2 vmulps (%rdi), %ymm2, %ymm2 vfmadd213ps %ymm2, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm2) + ymm2 vmovups %ymm2, (%rdi) addq $0x20, %r8 addq $0x20, %rdi addl $0x8, %r9d addq $0x8, %r14 jmp 0x1ddfed imulq %rcx, %rax imulq %r15, %rax addq %rax, %r12 imulq %rsi, %rdx imulq %r15, %rdx addq %rdx, %rbp cmpl %r13d, %r14d jge 0x1de29d movq (%rsp), %rax movq -0x18(%rax), %rax vmovss (%rbp,%r14,4), %xmm0 vmovaps 0x80(%rsp), %xmm1 vfmadd213ss 0xe0(%rbx,%rax), %xmm1, %xmm0 # xmm0 = (xmm1 * xmm0) + mem vmovss 0xdc(%rbx,%rax), %xmm1 vbroadcastss 0x212f34(%rip), %xmm2 # 0x3f11b4 vxorps %xmm2, %xmm1, %xmm1 vzeroupper callq 0x5f0e0 vmulss (%r12,%r14,4), %xmm0, %xmm0 vmovss %xmm0, (%r12,%r14,4) incq %r14 jmp 0x1de247 incq %r15 movl 0x70(%rsp), %r12d pushq $0x2 popq %r14 jmp 0x1dde8a movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de30c lock decl (%rax) jne 0x1de30c movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1de301 movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x1de30c movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de30e lock decl (%rax) jne 0x1de30e movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1de362 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1de30e jmp 0x1de301 jmp 0x1de362 movq %rsi, %rdi vzeroupper callq 0x5f3e0 xorl %ebp, %ebp movq 0x98(%rsp), %rax testq %rax, %rax je 0x1de34b lock decl (%rax) jne 0x1de34b movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi je 0x1de340 movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x1de34b movq %rsi, %rdi vzeroupper callq 0x5f3e0 movl %ebp, %eax addq $0x178, %rsp # imm = 0x178 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movq %rsi, %rdi callq 0x5f3e0 jmp 0x1de30e jmp 0x1de428 jmp 0x1de428 jmp 0x1de428 jmp 0x1de428 jmp 0x1de382 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de3e9 lock decl (%rax) jne 0x1de3e9 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1de3a5 jmp 0x1de3d0 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1de3e9 jmp 0x1de428 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de3e9 lock decl (%rax) jne 0x1de3e9 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1de3da movq %rsi, %rdi callq 0x5f3e0 jmp 0x1de3e9 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1de3e9 jmp 0x1de428 jmp 0x1de428 movq %rax, %rbx movq 0x98(%rsp), %rax testq %rax, %rax je 0x1de420 lock decl (%rax) jne 0x1de420 movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi jne 0x1de41a movq %rsi, %rdi callq 0x5f3e0 jmp 0x1de420 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/build_O2/src/layer/x86/lrn_x86_fma.cpp
virtual thunk to ncnn::LRN_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int LRN_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { outptr = ptr * ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 \|\| p >= channels) continue; const float sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { ssptr += sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { ptr = ptr * powf(bias + alpha_div_size * ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }	movq (%rdi), %rax addq -0x58(%rax), %rdi jmp 0x1dd838	/csukuangfj[P]ncnn/build_O2/src/layer/x86/lrn_x86_fma.cpp
ncnn::MemoryData::load_model(ncnn::ModelBin const&)	int MemoryData::load_model(const ModelBin& mb) { if (d != 0) { data = mb.load(w, h, d, c, 1); } else if (c != 0) { data = mb.load(w, h, c, 1); } else if (h != 0) { data = mb.load(w, h, 1); } else if (w != 0) { data = mb.load(w, 1); } else // 0 0 0 { data.create(1); } if (data.empty()) return -100; return 0; }	pushq %r14 pushq %rbx subq $0x58, %rsp movq %rdi, %rbx movl 0xd8(%rdi), %r8d testl %r8d, %r8d je 0x1df233 movl 0xd0(%rbx), %edx movl 0xd4(%rbx), %ecx movl 0xdc(%rbx), %r9d movq (%rsi), %rax movl $0x1, (%rsp) leaq 0x10(%rsp), %r14 movq %r14, %rdi callq 0x28(%rax) leaq 0xe0(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x1df407 testq %rax, %rax je 0x1df1f8 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x1df3ab lock decl (%rax) jne 0x1df3ab movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x1df3a3 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df3ab movl 0xdc(%rbx), %r8d testl %r8d, %r8d je 0x1df2b4 movl 0xd0(%rbx), %edx movl 0xd4(%rbx), %ecx movq (%rsi), %rax leaq 0x10(%rsp), %r14 pushq $0x1 popq %r9 movq %r14, %rdi callq 0x20(%rax) leaq 0xe0(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x1df4b4 testq %rax, %rax je 0x1df279 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x1df458 lock decl (%rax) jne 0x1df458 movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x1df450 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df458 movl 0xd4(%rbx), %ecx testl %ecx, %ecx je 0x1df32d movl 0xd0(%rbx), %edx movq (%rsi), %rax leaq 0x10(%rsp), %r14 pushq $0x1 popq %r8 movq %r14, %rdi callq 0x18(%rax) leaq 0xe0(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x1df548 testq %rax, %rax je 0x1df2f2 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x1df4ec lock decl (%rax) jne 0x1df4ec movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x1df4e4 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df4ec movl 0xd0(%rbx), %edx testl %edx, %edx je 0x1df437 movq (%rsi), %rax leaq 0x10(%rsp), %r14 pushq $0x1 popq %rcx movq %r14, %rdi callq 0x10(%rax) leaq 0xe0(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x1df5dc testq %rax, %rax je 0x1df368 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x1df580 lock decl (%rax) jne 0x1df580 movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x1df578 movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df580 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x1df605 lock decl (%rax) jne 0x1df605 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df5fd movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df605 leaq 0xe0(%rbx), %rdi pushq $0x1 popq %rsi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa jmp 0x1df605 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x1df605 lock decl (%rax) jne 0x1df605 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df5fd movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df605 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x1df605 lock decl (%rax) jne 0x1df605 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df5fd movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df605 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x1df605 lock decl (%rax) jne 0x1df605 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df5fd movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df605 movq %rsi, %rdi callq 0x5f3e0 cmpq $0x0, 0xe0(%rbx) je 0x1df625 movslq 0x118(%rbx), %rcx imulq 0x120(%rbx), %rcx xorl %eax, %eax testq %rcx, %rcx jne 0x1df628 pushq $-0x64 popq %rax addq $0x58, %rsp popq %rbx popq %r14 retq movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x1df6fd lock decl (%rax) jne 0x1df6fd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df6ed movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df6fd jmp 0x1df707 jmp 0x1df707 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x1df6fd lock decl (%rax) jne 0x1df6fd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df6ed movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df6fd jmp 0x1df707 jmp 0x1df707 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x1df6fd lock decl (%rax) jne 0x1df6fd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df6ed movq (%rdi), %rax callq 0x18(%rax) jmp 0x1df6fd jmp 0x1df707 jmp 0x1df707 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x1df6fd lock decl (%rax) jne 0x1df6fd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi jne 0x1df6f7 movq %rsi, %rdi callq 0x5f3e0 jmp 0x1df6fd movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x1df707 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/memorydata.cpp
ncnn::MVN::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int MVN::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int size = w * h; top_blob.create(w, h, channels, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; // prepare sum per channel Mat sum(channels, elemsize, opt.workspace_allocator); if (sum.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_blob.channel(q); float s = 0.f; for (int i = 0; i < size; i++) { s += ptr[i]; } sum[q] = s; } if (across_channels) { // compute mean across channels float mean = 0.f; for (int q = 0; q < channels; q++) { mean += sum[q]; } mean = mean / (channels * size); // subtract mean #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_blob.channel(q); float* outptr = top_blob.channel(q); for (int i = 0; i < size; i++) { outptr[i] = ptr[i] - mean; } } } else { // subtract mean #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_blob.channel(q); float* outptr = top_blob.channel(q); float mean = sum[q] / size; for (int i = 0; i < size; i++) { outptr[i] = ptr[i] - mean; } } } if (normalize_variance) { // prepare squared sum per channel Mat sqsum(channels, elemsize, opt.workspace_allocator); if (sqsum.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = top_blob.channel(q); float s = 0.f; for (int i = 0; i < size; i++) { s += ptr[i] * ptr[i]; } sqsum[q] = s; } if (across_channels) { // compute squared mean across channels float sqmean = 0.f; for (int q = 0; q < channels; q++) { sqmean += sqsum[q]; } sqmean = sqmean / (channels * size); // normalize variance float norm_var = sqrtf(sqmean) + eps; float norm_var_inv = 1.f / norm_var; // apply normalize_variance #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* outptr = top_blob.channel(q); for (int i = 0; i < size; i++) { outptr[i] = outptr[i] * norm_var_inv; } } } else { // apply normalize_variance #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* outptr = top_blob.channel(q); float sqmean = sqsum[q] / size; float norm_var = sqrt(sqmean) + eps; float norm_var_inv = 1.f / norm_var; for (int i = 0; i < size; i++) { outptr[i] = outptr[i] * norm_var_inv; } } } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0xb8, %rsp movq %rdx, %r14 movq %rsi, %r12 movq %rdi, 0x8(%rsp) movl 0x2c(%rsi), %ebp movl 0x30(%rsi), %r15d movl 0x38(%rsi), %r13d movq 0x10(%rsi), %rbx movq %rcx, 0x10(%rsp) movq 0x8(%rcx), %r9 movq %rdx, %rdi movl %ebp, %esi movl %r15d, %edx movl %r13d, %ecx movq %rbx, %r8 callq 0x63810 pushq $-0x64 popq %rax cmpq $0x0, (%r14) je 0x1dfd1f movslq 0x38(%r14), %rcx imulq 0x40(%r14), %rcx testq %rcx, %rcx je 0x1dfd1f movl %r13d, %esi movl %r15d, %r13d movq 0x10(%rsp), %rax movq 0x10(%rax), %rcx leaq 0x70(%rsp), %r15 andq $0x0, 0x40(%r15) xorps %xmm0, %xmm0 movaps %xmm0, (%r15) movups %xmm0, 0xc(%r15) movaps %xmm0, 0x20(%r15) movups %xmm0, 0x2c(%r15) movq %r15, %rdi movq %rbx, %rdx movl %esi, %ebx movq %rdx, 0x18(%rsp) callq 0x635fa movq (%r15), %r8 pushq $-0x64 popq %rax testq %r8, %r8 je 0x1dfce8 movslq 0xa8(%rsp), %rcx imulq 0xb0(%rsp), %rcx testq %rcx, %rcx je 0x1dfce8 imull %ebp, %r13d movq (%r12), %rcx movq 0x40(%r12), %rdx imulq 0x10(%r12), %rdx xorl %esi, %esi testl %r13d, %r13d movl $0x0, %r15d movl %r13d, 0x4(%rsp) cmovgl %r13d, %r15d movl %ebx, %r13d testl %ebx, %ebx movl $0x0, %ebp cmovgl %ebx, %ebp movq 0x8(%rsp), %r10 cmpq %rbp, %rsi je 0x1df9f8 xorps %xmm0, %xmm0 xorl %edi, %edi cmpq %rdi, %r15 je 0x1df9ea addss (%rcx,%rdi,4), %xmm0 incq %rdi jmp 0x1df9db movss %xmm0, (%r8,%rsi,4) incq %rsi addq %rdx, %rcx jmp 0x1df9d1 cmpl $0x0, 0xd4(%r10) je 0x1dfa79 xorps %xmm0, %xmm0 movq 0x70(%rsp), %rdx xorl %ecx, %ecx cmpq %rcx, %rbp je 0x1dfa1b addss (%rdx,%rcx,4), %xmm0 incq %rcx jmp 0x1dfa0c movl %r13d, %ecx imull 0x4(%rsp), %ecx cvtsi2ss %ecx, %xmm1 divss %xmm1, %xmm0 movq (%r12), %r9 movq 0x40(%r12), %rcx movq 0x40(%r14), %rdx imulq 0x10(%r14), %rdx movq (%r14), %rsi imulq 0x10(%r12), %rcx xorl %edi, %edi cmpq %rbp, %rdi je 0x1dfae5 xorl %r8d, %r8d cmpq %r8, %r15 je 0x1dfa6e movss (%r9,%r8,4), %xmm1 subss %xmm0, %xmm1 movss %xmm1, (%rsi,%r8,4) incq %r8 jmp 0x1dfa54 incq %rdi addq %rdx, %rsi addq %rcx, %r9 jmp 0x1dfa48 movq (%r12), %r11 movq 0x40(%r12), %rcx movq (%r14), %rdx movq 0x40(%r14), %rsi imulq 0x10(%r14), %rsi movq 0x70(%rsp), %rdi cvtsi2ssl 0x4(%rsp), %xmm1 movss 0x20f1e7(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 imulq 0x10(%r12), %rcx xorl %r8d, %r8d cmpq %rbp, %r8 je 0x1dfae5 movss (%rdi,%r8,4), %xmm1 mulss %xmm0, %xmm1 xorl %r9d, %r9d cmpq %r9, %r15 je 0x1dfada movss (%r11,%r9,4), %xmm2 subss %xmm1, %xmm2 movss %xmm2, (%rdx,%r9,4) incq %r9 jmp 0x1dfac0 incq %r8 addq %rsi, %rdx addq %rcx, %r11 jmp 0x1dfaae cmpl $0x0, 0xd0(%r10) je 0x1dfce6 movl %eax, %ebx movq 0x10(%rsp), %rax movq 0x10(%rax), %rcx leaq 0x20(%rsp), %rdi andq $0x0, 0x40(%rdi) xorps %xmm0, %xmm0 movaps %xmm0, (%rdi) movups %xmm0, 0xc(%rdi) movaps %xmm0, 0x20(%rdi) movups %xmm0, 0x2c(%rdi) movl %r13d, %esi movq 0x18(%rsp), %rdx callq 0x635fa movq 0x20(%rsp), %r9 testq %r9, %r9 movl %ebx, %eax je 0x1dfb85 movslq 0x58(%rsp), %rcx imulq 0x60(%rsp), %rcx testq %rcx, %rcx je 0x1dfb85 movq (%r14), %rcx movq 0x40(%r14), %rdx imulq 0x10(%r14), %rdx xorl %esi, %esi movq 0x8(%rsp), %r8 cmpq %rbp, %rsi je 0x1dfbbc xorps %xmm0, %xmm0 xorl %edi, %edi cmpq %rdi, %r15 je 0x1dfb77 movss (%rcx,%rdi,4), %xmm1 mulss %xmm1, %xmm1 addss %xmm1, %xmm0 incq %rdi jmp 0x1dfb60 movss %xmm0, (%r9,%rsi,4) incq %rsi addq %rdx, %rcx jmp 0x1dfb56 movq 0x28(%rsp), %rcx testq %rcx, %rcx je 0x1dfce8 lock decl (%rcx) jne 0x1dfce8 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dfcd2 movq (%rdi), %rax callq 0x18(%rax) movl %ebx, %eax jmp 0x1dfce8 cmpl $0x0, 0xd4(%r8) je 0x1dfc40 xorps %xmm0, %xmm0 movq 0x20(%rsp), %rax xorl %ecx, %ecx cmpq %rcx, %rbp je 0x1dfbdf addss (%rax,%rcx,4), %xmm0 incq %rcx jmp 0x1dfbd0 imull 0x4(%rsp), %r13d cvtsi2ss %r13d, %xmm1 divss %xmm1, %xmm0 sqrtss %xmm0, %xmm1 addss 0xd8(%r8), %xmm1 movq 0x40(%r14), %rax imulq 0x10(%r14), %rax movq (%r14), %rcx movss 0x20f079(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 xorl %edx, %edx cmpq %rbp, %rdx je 0x1dfcac xorl %esi, %esi cmpq %rsi, %r15 je 0x1dfc38 movss (%rcx,%rsi,4), %xmm1 mulss %xmm0, %xmm1 movss %xmm1, (%rcx,%rsi,4) incq %rsi jmp 0x1dfc20 incq %rdx addq %rax, %rcx jmp 0x1dfc15 movq (%r14), %rax movq 0x40(%r14), %rcx imulq 0x10(%r14), %rcx movq 0x20(%rsp), %rdx cvtsi2ssl 0x4(%rsp), %xmm2 movss 0x20f029(%rip), %xmm0 # 0x3eec88 movaps %xmm0, %xmm1 divss %xmm2, %xmm1 xorl %esi, %esi cmpq %rbp, %rsi je 0x1dfcac movss (%rdx,%rsi,4), %xmm2 mulss %xmm1, %xmm2 sqrtss %xmm2, %xmm3 addss 0xd8(%r8), %xmm3 movaps %xmm0, %xmm2 divss %xmm3, %xmm2 xorl %edi, %edi cmpq %rdi, %r15 je 0x1dfca4 movss (%rax,%rdi,4), %xmm3 mulss %xmm2, %xmm3 movss %xmm3, (%rax,%rdi,4) incq %rdi jmp 0x1dfc8c incq %rsi addq %rcx, %rax jmp 0x1dfc68 movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dfce6 lock decl (%rax) jne 0x1dfce6 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dfcde movq (%rdi), %rax callq 0x18(%rax) jmp 0x1dfce6 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax jmp 0x1dfce8 movq %rsi, %rdi callq 0x5f3e0 xorl %eax, %eax movq 0x78(%rsp), %rcx testq %rcx, %rcx je 0x1dfd1f lock decl (%rcx) jne 0x1dfd1f movl %eax, %ebx movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x1dfd15 movq (%rdi), %rax callq 0x18(%rax) movl %ebx, %eax jmp 0x1dfd1f movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0xb8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x1dfd35 jmp 0x1dfd35 movq %rax, %rdi callq 0x61d68 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x1dfd71 lock decl (%rax) jne 0x1dfd71 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x1dfd6b movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dfd71 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x1dfd35 nop	/csukuangfj[P]ncnn/src/layer/mvn.cpp
ncnn::Pooling::load_param(ncnn::ParamDict const&)	int Pooling::load_param(const ParamDict& pd) { pooling_type = pd.get(0, 0); kernel_w = pd.get(1, 0); kernel_h = pd.get(11, kernel_w); stride_w = pd.get(2, 1); stride_h = pd.get(12, stride_w); pad_left = pd.get(3, 0); pad_right = pd.get(14, pad_left); pad_top = pd.get(13, pad_left); pad_bottom = pd.get(15, pad_top); global_pooling = pd.get(4, 0); pad_mode = pd.get(5, 0); avgpool_count_include_pad = pd.get(6, 0); adaptive_pooling = pd.get(7, 0); out_w = pd.get(8, 0); out_h = pd.get(18, out_w); return 0; }	pushq %rbp pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rbp movq %r14, %rdi movl %ebp, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd4(%rbx) pushq $0xb popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi movl %ebp, %edx callq 0x718a6 movl %eax, 0xdc(%rbx) pushq $0xc popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xe0(%rbx) pushq $0x3 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe4(%rbx) pushq $0xe popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xe8(%rbx) movl 0xe4(%rbx), %edx pushq $0xd popq %rsi movq %r14, %rdi callq 0x718a6 movl %eax, 0xec(%rbx) pushq $0xf popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xf0(%rbx) pushq $0x4 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xf4(%rbx) pushq $0x5 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xf8(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xfc(%rbx) pushq $0x7 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x100(%rbx) pushq $0x8 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x104(%rbx) pushq $0x12 popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0x108(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %rbp retq nop	/csukuangfj[P]ncnn/src/layer/pooling.cpp
ncnn::Pooling_x86_fma::create_pipeline(ncnn::Option const&)	int Pooling_x86_fma::create_pipeline(const Option& /opt/) { if (adaptive_pooling) { support_packing = false; support_bf16_storage = false; support_fp16_storage = false; support_int8_storage = false; support_tensor_storage = false; } return 0; }	movq (%rdi), %rax movq -0x18(%rax), %rcx cmpl $0x0, 0x100(%rdi,%rcx) je 0x1e431f xorl %edx, %edx movb %dl, 0xb(%rdi,%rcx) movq -0x18(%rax), %rcx movb %dl, 0xc(%rdi,%rcx) movq -0x18(%rax), %rcx movb %dl, 0xd(%rdi,%rcx) movq -0x18(%rax), %rcx movb %dl, 0xe(%rdi,%rcx) movq -0x18(%rax), %rax movb %dl, 0x10(%rdi,%rax) xorl %eax, %eax retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/pooling_x86_fma.cpp
ncnn::binary_op_broadcast_inner(ncnn::Mat const&, ncnn::Mat const&, ncnn::Mat&, int, ncnn::Option const&)	static int binary_op_broadcast_inner(const Mat& a, const Mat& b, Mat& c, int op_type, const Option& opt) { // squeeze inner axes Mat b2 = b; if (b.dims == 2 && b.w == 1) b2 = b.reshape(b.h); else if (b.dims == 3 && b.h == 1) b2 = b.reshape(b.c); else if (b.dims == 3 && b.w == 1) b2 = b.reshape(b.h, b.c); else if (b.dims == 4 && b.d == 1) b2 = b.reshape(b.c); else if (b.dims == 4 && b.h == 1) b2 = b.reshape(b.d, b.c); else if (b.dims == 4 && b.w == 1) b2 = b.reshape(b.h, b.d, b.c); using namespace BinaryOp_x86_fma_functor; if (op_type == BinaryOp::Operation_ADD) return binary_op_broadcast_inner<binary_op_add>(a, b2, c, opt); if (op_type == BinaryOp::Operation_SUB) return binary_op_broadcast_inner<binary_op_sub>(a, b2, c, opt); if (op_type == BinaryOp::Operation_MUL) return binary_op_broadcast_inner<binary_op_mul>(a, b2, c, opt); if (op_type == BinaryOp::Operation_DIV) return binary_op_broadcast_inner<binary_op_div>(a, b2, c, opt); if (op_type == BinaryOp::Operation_MAX) return binary_op_broadcast_inner<binary_op_max>(a, b2, c, opt); if (op_type == BinaryOp::Operation_MIN) return binary_op_broadcast_inner<binary_op_min>(a, b2, c, opt); if (op_type == BinaryOp::Operation_POW) return binary_op_broadcast_inner<binary_op_pow>(a, b2, c, opt); if (op_type == BinaryOp::Operation_RSUB) return binary_op_broadcast_inner<binary_op_rsub>(a, b2, c, opt); if (op_type == BinaryOp::Operation_RDIV) return binary_op_broadcast_inner<binary_op_rdiv>(a, b2, c, opt); if (op_type == BinaryOp::Operation_RPOW) return binary_op_broadcast_inner<binary_op_rpow>(a, b2, c, opt); if (op_type == BinaryOp::Operation_ATAN2) return binary_op_broadcast_inner<binary_op_atan2>(a, b2, c, opt); if (op_type == BinaryOp::Operation_RATAN2) return binary_op_broadcast_inner<binary_op_ratan2>(a, b2, c, opt); // should never reach here return 0; }	pushq %rbp movq %rsp, %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx andq $-0x20, %rsp subq $0x1a0, %rsp # imm = 0x1A0 movl %ecx, %r13d movq %rdx, 0x48(%rsp) movq %rdi, 0x20(%rsp) movq (%rsi), %rbx movq 0x8(%rsi), %rcx movq 0x10(%rsi), %rax movq %rax, 0x78(%rsp) movq 0x20(%rsi), %rax movq %rax, 0x50(%rsp) vmovq 0x28(%rsi), %xmm1 movq 0x40(%rsi), %rax movq %rax, 0x98(%rsp) vmovdqa %xmm1, %xmm0 testq %rcx, %rcx je 0x264d73 lock incl (%rcx) vmovq 0x28(%rsi), %xmm0 movq %rcx, 0x58(%rsp) vmovd %xmm0, %ecx movl %ecx, %eax xorl $0x2, %eax vpextrd $0x1, %xmm0, %edi xorl $0x1, %edi orl %edi, %eax jne 0x264de3 movl 0x30(%rsi), %edx leaq 0x140(%rsp), %rdi xorl %ecx, %ecx callq 0x62c8a movq 0x148(%rsp), %rax testq %rax, %rax je 0x264db0 lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x264f79 lock decl (%rax) jne 0x264f79 movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x264f71 movq (%rdi), %rax movq %rbx, %rsi callq 0x18(%rax) jmp 0x264f79 movl %ecx, %eax xorl $0x3, %eax movl 0x30(%rsi), %edx movl %edx, %r8d xorl $0x1, %r8d movl %eax, %r9d orl %r8d, %r9d jne 0x264e4f movl 0x38(%rsi), %edx leaq 0x140(%rsp), %rdi xorl %ecx, %ecx callq 0x62c8a movq 0x148(%rsp), %rax testq %rax, %rax je 0x264e1c lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x26506e lock decl (%rax) jne 0x26506e movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x265066 movq (%rdi), %rax movq %rbx, %rsi callq 0x18(%rax) jmp 0x26506e orl %edi, %eax jne 0x264ea9 movl 0x38(%rsi), %ecx leaq 0x140(%rsp), %rdi xorl %r8d, %r8d callq 0x62e4e movq 0x148(%rsp), %rax testq %rax, %rax je 0x264e76 lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x265102 lock decl (%rax) jne 0x265102 movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2650fa movq (%rdi), %rax movq %rbx, %rsi callq 0x18(%rax) jmp 0x265102 xorl $0x4, %ecx movl 0x34(%rsi), %eax movl %eax, %r9d xorl $0x1, %r9d orl %ecx, %r9d jne 0x264f10 movl 0x38(%rsi), %edx leaq 0x140(%rsp), %rdi xorl %ecx, %ecx callq 0x62c8a movq 0x148(%rsp), %rax testq %rax, %rax je 0x264edd lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x265196 lock decl (%rax) jne 0x265196 movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x26518e movq (%rdi), %rax movq %rbx, %rsi callq 0x18(%rax) jmp 0x265196 orl %ecx, %r8d jne 0x265005 movl 0x38(%rsi), %ecx leaq 0x140(%rsp), %rdi movl %eax, %edx xorl %r8d, %r8d callq 0x62e4e movq 0x148(%rsp), %rax testq %rax, %rax je 0x264f3e lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x26522a lock decl (%rax) jne 0x26522a movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x265222 movq (%rdi), %rax movq %rbx, %rsi callq 0x18(%rax) jmp 0x26522a movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq 0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c orl %edi, %ecx jne 0x265351 movl 0x38(%rsi), %r8d leaq 0x140(%rsp), %rdi movl %eax, %ecx xorl %r9d, %r9d callq 0x63020 movq 0x148(%rsp), %rax testq %rax, %rax je 0x265033 lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2652be lock decl (%rax) jne 0x2652be movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2652b6 movq (%rdi), %rax movq %rbx, %rsi callq 0x18(%rax) jmp 0x2652be movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq 0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq 0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq 0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq 0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq 0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rsi, %rdi callq 0x5f3e0 vmovdqa 0x80(%rsp), %xmm1 movq %r14, 0x58(%rsp) movq %rbx, 0x18(%rsp) cmpl $0xb, %r13d ja 0x26bab3 movl %r13d, %eax leaq 0x192102(%rip), %rcx # 0x3f746c movslq (%rcx,%rax,4), %rax addq %rcx, %rax vmovd %xmm1, 0x38(%rsp) vpextrd $0x1, %xmm1, 0x34(%rsp) jmpq %rax movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x268001 cmpl $0x1, 0x38(%rsp) jne 0x268001 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x267ff9 vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x2653fd movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x265416 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x265416 movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x26541c vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x2654a8 vaddps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x26546a vaddps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x26548e imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2654eb vaddss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2654d5 incq %r14 movq 0x18(%rsp), %r10 jmp 0x2653d8 movq 0x20(%rsp), %rax movl 0x2c(%rax), %ecx movl %ecx, 0x80(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x40(%rsp) movl 0x34(%rax), %ecx movq %rcx, 0xf8(%rsp) movl 0x38(%rax), %ecx movq %rcx, 0xf0(%rsp) movl 0x18(%rax), %r12d movl 0x28(%rax), %eax cmpl $0x2, %eax movq 0x18(%rsp), %rbx movl %r12d, 0xdc(%rsp) jne 0x26966d cmpl $0x1, 0x38(%rsp) jne 0x26966d movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %r13d, %r13d movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x269665 vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0xb0(%rsp) jne 0x2655ad movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x2655cf vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x2655cf movq %r13, %rax shlq $0x5, %rax vmovups (%rbx,%rax), %ymm0 jmp 0x2655d5 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, %r12 movq %rcx, 0xa0(%rsp) imulq %rcx, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0x28(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, 0xa8(%rsp) movq %rcx, 0xc8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x60(%rsp) addq %rcx, %r13 vmovaps %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x2656db vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x26564e vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x26569a movq 0xa8(%rsp), %r13 movq 0x138(%rsp), %rax imulq %r13, %rax imulq 0xc8(%rsp), %rax addq %rax, 0x60(%rsp) movq 0xd0(%rsp), %rax imulq %r13, %rax imulq 0xa0(%rsp), %rax addq %rax, 0x28(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d vmovdqa 0xb0(%rsp), %xmm0 jge 0x26575d movq 0x28(%rsp), %rax vmovd (%rax,%r14,4), %xmm1 vzeroupper callq 0x5f170 movq 0x60(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x26572c incq %r13 jmp 0x265573 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x266967 cmpl $0x1, 0x38(%rsp) jne 0x266967 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x26695f vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x2657e1 movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x2657fa vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2657fa movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x265800 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x26588c vmaxps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x26584e vmaxps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x265872 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2658cf vmaxss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2658b9 incq %r14 movq 0x18(%rsp), %r10 jmp 0x2657bc movq 0x20(%rsp), %rax movl 0x2c(%rax), %ecx movl %ecx, 0x80(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x40(%rsp) movl 0x34(%rax), %ecx movq %rcx, 0xf8(%rsp) movl 0x38(%rax), %ecx movq %rcx, 0xf0(%rsp) movl 0x18(%rax), %r12d movl 0x28(%rax), %eax cmpl $0x2, %eax movq 0x18(%rsp), %rbx movl %r12d, 0xdc(%rsp) jne 0x2698b9 cmpl $0x1, 0x38(%rsp) jne 0x2698b9 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %r13d, %r13d movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x2698b1 vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0xb0(%rsp) jne 0x265991 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x2659b3 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x2659b3 movq %r13, %rax shlq $0x5, %rax vmovups (%rbx,%rax), %ymm0 jmp 0x2659b9 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, %r12 movq %rcx, 0xa0(%rsp) imulq %rcx, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0x28(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, 0xa8(%rsp) movq %rcx, 0xc8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x60(%rsp) addq %rcx, %r13 vmovaps %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x265abf vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x265a32 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x265a7e movq 0xa8(%rsp), %r13 movq 0x138(%rsp), %rax imulq %r13, %rax imulq 0xc8(%rsp), %rax addq %rax, 0x60(%rsp) movq 0xd0(%rsp), %rax imulq %r13, %rax imulq 0xa0(%rsp), %rax addq %rax, 0x28(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d vmovdqa 0xb0(%rsp), %xmm0 jge 0x265b41 movq 0x28(%rsp), %rax vmovd (%rax,%r14,4), %xmm1 vzeroupper callq 0x5f0e0 movq 0x60(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x265b10 incq %r13 jmp 0x265957 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x266ed2 cmpl $0x1, 0x38(%rsp) jne 0x266ed2 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x266eca vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x265bc5 movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x265bde vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x265bde movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x265be4 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x265c70 vmulps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x265c32 vmulps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x265c56 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x265cb3 vmulss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x265c9d incq %r14 movq 0x18(%rsp), %r10 jmp 0x265ba0 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x26743d cmpl $0x1, 0x38(%rsp) jne 0x26743d movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) vbroadcastss 0x188f68(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x188f5f(%rip), %xmm1 # 0x3eec88 vmovss 0x188f57(%rip), %xmm2 # 0x3eec88 cmpq 0x40(%rsp), %r14 je 0x267435 vmovss (%r10,%r14,4), %xmm3 cmpl $0x4, %edi jne 0x265d56 movq %r14, %rsi shlq $0x4, %rsi vmovups (%r10,%rsi), %xmm4 jmp 0x265d6f vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x265d6f movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm5 jmp 0x265d75 vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r15 imulq %r15, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %r12d, %r12d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x265df5 vmulps (%r13), %ymm5, %ymm6 vmovups %ymm6, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r12 jmp 0x265dd1 vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%rsi), %edx cmpl %r8d, %edx jge 0x265e27 vmulps (%r13), %xmm4, %xmm5 vmovups %xmm5, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r12 jmp 0x265e03 vdivss %xmm3, %xmm2, %xmm3 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r15 imulq 0xb0(%rsp), %r15 addq %r15, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r12d jge 0x265e66 vmulss (%r11,%r12,4), %xmm3, %xmm4 vmovss %xmm4, (%rbx,%r12,4) incq %r12 jmp 0x265e50 incq %r14 movq 0x18(%rsp), %r10 jmp 0x265d31 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x267a96 cmpl $0x1, 0x38(%rsp) jne 0x267a96 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x267a8e vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x265eef movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x265f08 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x265f08 movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x265f0e vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x265f9a vsubps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x265f5c vsubps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x265f80 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x265fdd vsubss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x265fc7 incq %r14 movq 0x18(%rsp), %r10 jmp 0x265eca movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x26816b cmpl $0x1, 0x38(%rsp) jne 0x26816b movl %edi, %r8d imull %eax, %r8d xorl %r15d, %r15d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r15 je 0x268163 vmovd (%r10,%r15,4), %xmm0 cmpl $0x4, %edi jne 0x266066 movq %r15, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x26607f vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26607f movq %r15, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x266085 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r15, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r15, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r14d, %r14d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x266119 vmovups (%r13), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r14 jmp 0x2660d3 vmovups (%r13), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r14 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x2660fb imulq %r15, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r15, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r14d jge 0x266160 vmovss (%r11,%r14,4), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%rbx,%r14,4) incq %r14 jmp 0x266146 incq %r15 movq 0x18(%rsp), %r10 jmp 0x266041 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x268b07 cmpl $0x1, 0x38(%rsp) jne 0x268b07 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x268aff vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x2661e9 movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x266202 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266202 movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x266208 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x266294 vminps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x266256 vminps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x26627a imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2662d7 vminss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2662c1 incq %r14 movq 0x18(%rsp), %r10 jmp 0x2661c4 movq 0x20(%rsp), %rax movl 0x2c(%rax), %ecx movl %ecx, 0x80(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x40(%rsp) movl 0x34(%rax), %ecx movq %rcx, 0xf8(%rsp) movl 0x38(%rax), %ecx movq %rcx, 0xf0(%rsp) movl 0x18(%rax), %r12d movl 0x28(%rax), %eax cmpl $0x2, %eax movq 0x18(%rsp), %rbx movl %r12d, 0xdc(%rsp) jne 0x269b05 cmpl $0x1, 0x38(%rsp) jne 0x269b05 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %r13d, %r13d movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x269afd vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0x60(%rsp) jne 0x266396 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x2663b8 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x2663b8 movq %r13, %rax shlq $0x5, %rax vmovdqu (%rbx,%rax), %ymm0 jmp 0x2663be vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, %r12 movq %rcx, 0xa0(%rsp) imulq %rcx, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0xb0(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, 0xa8(%rsp) movq %rcx, 0xc8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x28(%rsp) addq %rcx, %r13 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x2664c7 vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x26643a vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x266486 movq 0xa8(%rsp), %r13 movq 0x138(%rsp), %rax imulq %r13, %rax imulq 0xc8(%rsp), %rax addq %rax, 0x28(%rsp) movq 0xd0(%rsp), %rax imulq %r13, %rax imulq 0xa0(%rsp), %rax addq %rax, 0xb0(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d jge 0x26654c movq 0xb0(%rsp), %rax vmovd (%rax,%r14,4), %xmm0 vmovdqa 0x60(%rsp), %xmm1 vzeroupper callq 0x5f0e0 movq 0x28(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x26651b incq %r13 jmp 0x26635f movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x269072 cmpl $0x1, 0x38(%rsp) jne 0x269072 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x26906a vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x2665d0 movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x2665e9 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2665e9 movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x2665ef vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x26669f vmovups (%r13), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x26663d vmovups (%r13), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x266673 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2666e2 vdivss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2666cc incq %r14 movq 0x18(%rsp), %r10 jmp 0x2665ab movq 0x20(%rsp), %rax movl 0x2c(%rax), %ecx movl %ecx, 0x80(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x40(%rsp) movl 0x34(%rax), %ecx movq %rcx, 0xf8(%rsp) movl 0x38(%rax), %ecx movq %rcx, 0xf0(%rsp) movl 0x18(%rax), %r12d movl 0x28(%rax), %eax cmpl $0x2, %eax movq 0x18(%rsp), %rbx movl %r12d, 0xdc(%rsp) jne 0x269d54 cmpl $0x1, 0x38(%rsp) jne 0x269d54 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %r13d, %r13d movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x269d4c vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0x60(%rsp) jne 0x2667a1 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x2667c3 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x2667c3 movq %r13, %rax shlq $0x5, %rax vmovdqu (%rbx,%rax), %ymm0 jmp 0x2667c9 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, %r12 movq %rcx, 0xa0(%rsp) imulq %rcx, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0xb0(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, 0xa8(%rsp) movq %rcx, 0xc8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x28(%rsp) addq %rcx, %r13 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x2668d2 vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x266845 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x266891 movq 0xa8(%rsp), %r13 movq 0x138(%rsp), %rax imulq %r13, %rax imulq 0xc8(%rsp), %rax addq %rax, 0x28(%rsp) movq 0xd0(%rsp), %rax imulq %r13, %rax imulq 0xa0(%rsp), %rax addq %rax, 0xb0(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d jge 0x266957 movq 0xb0(%rsp), %rax vmovd (%rax,%r14,4), %xmm0 vmovdqa 0x60(%rsp), %xmm1 vzeroupper callq 0x5f170 movq 0x28(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x266926 incq %r13 jmp 0x26676a movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x266ad1 cmpl $0x1, 0x38(%rsp) jne 0x266ad1 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x266ac9 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x2669c9 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x2669e2 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2669e2 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x2669e8 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x266a77 vmaxps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x266a37 vmaxps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x266a5d imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x266abc vmaxss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x266aa6 incq %r9 movq 0x18(%rsp), %r10 jmp 0x2669a4 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x266c17 cmpl $0x2, 0x38(%rsp) jne 0x266c17 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x266c0f movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x266bff vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x266b7c movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x266b95 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266b95 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x266b9b vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x266bd3 vmaxps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x266b9d vmaxps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x266bbd jmp 0x266bf2 vmaxss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x266bdd incq %r13 jmp 0x266b59 incq %r9 movq 0x80(%rsp), %r13 jmp 0x266b17 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x266d62 cmpl $0x2, 0x38(%rsp) jne 0x266d62 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x266d5a movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x266d4a vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x266cc7 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x266ce0 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266ce0 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x266ce6 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x266d1e vmaxps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x266ce8 vmaxps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x266d08 jmp 0x266d3d vmaxss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x266d28 incq %r13 jmp 0x266ca4 incq %r9 movq 0x80(%rsp), %r13 jmp 0x266c62 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x266ec2 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x266eba vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x266e3d movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x266e56 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266e56 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x266e5c vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x266e91 vmaxps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x266e5e vmaxps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x266e7c jmp 0x266eae vmaxss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x266e9a incq %r12 jmp 0x266e1a incq %r14 jmp 0x266e04 incq %r8 jmp 0x266dbb movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x26703c cmpl $0x1, 0x38(%rsp) jne 0x26703c movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x267034 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x266f34 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x266f4d vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266f4d movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x266f53 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x266fe2 vmulps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x266fa2 vmulps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x266fc8 imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x267027 vmulss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x267011 incq %r9 movq 0x18(%rsp), %r10 jmp 0x266f0f movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x267182 cmpl $0x2, 0x38(%rsp) jne 0x267182 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x26717a movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x26716a vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x2670e7 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x267100 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267100 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267106 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x26713e vmulps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x267108 vmulps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x267128 jmp 0x26715d vmulss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x267148 incq %r13 jmp 0x2670c4 incq %r9 movq 0x80(%rsp), %r13 jmp 0x267082 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x2672cd cmpl $0x2, 0x38(%rsp) jne 0x2672cd movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x2672c5 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x2672b5 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x267232 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x26724b vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26724b movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267251 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x267289 vmulps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x267253 vmulps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x267273 jmp 0x2672a8 vmulss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x267293 incq %r13 jmp 0x26720f incq %r9 movq 0x80(%rsp), %r13 jmp 0x2671cd movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x26742d movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x267425 vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x2673a8 movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2673c1 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2673c1 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2673c7 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x2673fc vmulps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x2673c9 vmulps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x2673e7 jmp 0x267419 vmulss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x267405 incq %r12 jmp 0x267385 incq %r14 jmp 0x26736f incq %r8 jmp 0x267326 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x2675e1 cmpl $0x1, 0x38(%rsp) jne 0x2675e1 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) vbroadcastss 0x187805(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x1877fc(%rip), %xmm1 # 0x3eec88 vmovss 0x1877f4(%rip), %xmm2 # 0x3eec88 cmpq 0x40(%rsp), %r9 je 0x2675d9 vmovss (%r10,%r9,4), %xmm3 cmpl $0x4, %edi jne 0x2674b9 movq %r9, %rdx shlq $0x4, %rdx vmovups (%r10,%rdx), %xmm4 jmp 0x2674d2 vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x2674d2 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm5 jmp 0x2674d8 vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 movq %r12, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %r15d, %r15d xorl %r10d, %r10d leal 0x7(%r10), %edx cmpl %r8d, %edx jge 0x26755a vmulps (%r13), %ymm5, %ymm6 vmovups %ymm6, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r10d addq $0x8, %r15 jmp 0x267535 vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%r10), %edx cmpl %r8d, %edx jge 0x26758d vmulps (%r13), %xmm4, %xmm5 vmovups %xmm5, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r10d addq $0x4, %r15 jmp 0x267568 vdivss %xmm3, %xmm2, %xmm3 imulq 0x28(%rsp), %r12 imulq %r9, %r12 addq %r12, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2675cc vmulss (%r11,%r15,4), %xmm3, %xmm4 vmovss %xmm4, (%rbx,%r15,4) incq %r15 jmp 0x2675b6 incq %r9 movq 0x18(%rsp), %r10 jmp 0x267494 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x267763 cmpl $0x2, 0x38(%rsp) jne 0x267763 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx vbroadcastss 0x187658(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x18764f(%rip), %xmm1 # 0x3eec88 vmovss 0x187647(%rip), %xmm2 # 0x3eec88 cmpq %r11, %r9 je 0x26775b movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x26774b vmovss (%r15,%r13,4), %xmm3 cmpl $0x4, %edi jne 0x2676a6 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm4 jmp 0x2676bf vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x2676bf movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm5 jmp 0x2676c5 vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x2676f5 vmulps (%r14), %ymm5, %ymm6 vmovups %ymm6, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x2676d5 vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%rsi), %edx cmpl %r8d, %edx jge 0x267723 vmulps (%r14), %xmm4, %xmm5 vmovups %xmm5, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi jmp 0x267703 vdivss %xmm3, %xmm2, %xmm3 cmpl %r8d, %esi jge 0x267743 vmulss (%r14), %xmm3, %xmm4 vmovss %xmm4, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi jmp 0x267727 incq %r13 jmp 0x267683 incq %r9 movq 0x80(%rsp), %r13 jmp 0x267641 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x2678ea cmpl $0x2, 0x38(%rsp) jne 0x2678ea movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx vbroadcastss 0x1874d1(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x1874c8(%rip), %xmm1 # 0x3eec88 vmovss 0x1874c0(%rip), %xmm2 # 0x3eec88 cmpq %r11, %r9 je 0x2678e2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x2678d2 vmovss (%r15,%r13,4), %xmm3 cmpl $0x4, %edi jne 0x26782d movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm4 jmp 0x267846 vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x267846 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm5 jmp 0x26784c vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x26787c vmulps (%r14), %ymm5, %ymm6 vmovups %ymm6, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x26785c vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%rsi), %edx cmpl %r8d, %edx jge 0x2678aa vmulps (%r14), %xmm4, %xmm5 vmovups %xmm5, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi jmp 0x26788a vdivss %xmm3, %xmm2, %xmm3 cmpl %r8d, %esi jge 0x2678ca vmulss (%r14), %xmm3, %xmm4 vmovss %xmm4, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi jmp 0x2678ae incq %r13 jmp 0x26780a incq %r9 movq 0x80(%rsp), %r13 jmp 0x2677c8 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 vbroadcastss 0x18733c(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x187333(%rip), %xmm1 # 0x3eec88 vmovss 0x18732b(%rip), %xmm2 # 0x3eec88 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x267a86 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x267a7e vmovss (%r15,%r12,4), %xmm3 cmpl $0x4, %edi jne 0x2679df movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm4 jmp 0x2679f8 vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x2679f8 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm5 jmp 0x2679fe vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x267a2c vmulps (%r10), %ymm5, %ymm6 vmovups %ymm6, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x267a0e vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%rsi), %edx cmpl %eax, %edx jge 0x267a58 vmulps (%r10), %xmm4, %xmm5 vmovups %xmm5, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi jmp 0x267a3a vdivss %xmm3, %xmm2, %xmm3 cmpl %eax, %esi jge 0x267a76 vmulss (%r10), %xmm3, %xmm4 vmovss %xmm4, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi jmp 0x267a5c incq %r12 jmp 0x2679bc incq %r14 jmp 0x2679a6 incq %r8 jmp 0x26795d movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x267c00 cmpl $0x1, 0x38(%rsp) jne 0x267c00 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x267bf8 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x267af8 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x267b11 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267b11 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x267b17 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x267ba6 vsubps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x267b66 vsubps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x267b8c imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x267beb vsubss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x267bd5 incq %r9 movq 0x18(%rsp), %r10 jmp 0x267ad3 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x267d46 cmpl $0x2, 0x38(%rsp) jne 0x267d46 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x267d3e movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x267d2e vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x267cab movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x267cc4 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267cc4 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267cca vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x267d02 vsubps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x267ccc vsubps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x267cec jmp 0x267d21 vsubss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x267d0c incq %r13 jmp 0x267c88 incq %r9 movq 0x80(%rsp), %r13 jmp 0x267c46 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x267e91 cmpl $0x2, 0x38(%rsp) jne 0x267e91 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x267e89 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x267e79 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x267df6 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x267e0f vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267e0f movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267e15 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x267e4d vsubps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x267e17 vsubps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x267e37 jmp 0x267e6c vsubss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x267e57 incq %r13 jmp 0x267dd3 incq %r9 movq 0x80(%rsp), %r13 jmp 0x267d91 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x267ff1 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x267fe9 vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x267f6c movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x267f85 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267f85 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267f8b vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x267fc0 vsubps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x267f8d vsubps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x267fab jmp 0x267fdd vsubss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x267fc9 incq %r12 jmp 0x267f49 incq %r14 jmp 0x267f33 incq %r8 jmp 0x267eea movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x2682e1 cmpl $0x1, 0x38(%rsp) jne 0x2682e1 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x2682d9 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x268063 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x26807c vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26807c movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x268082 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x268111 vaddps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x2680d1 vaddps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x2680f7 imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x268156 vaddss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x268140 incq %r9 movq 0x18(%rsp), %r10 jmp 0x26803e movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x268427 cmpl $0x1, 0x38(%rsp) jne 0x268427 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x26841f vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x2681cd movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x2681e6 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2681e6 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x2681ec vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x268283 vmovups (%r13), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x26823b vmovups (%r13), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x268265 imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2682cc vmovss (%r11,%r15,4), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2682b2 incq %r9 movq 0x18(%rsp), %r10 jmp 0x2681a8 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x268579 cmpl $0x2, 0x38(%rsp) jne 0x268579 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268571 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x26840f vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x26838c movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2683a5 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2683a5 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2683ab vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x2683e3 vaddps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x2683ad vaddps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x2683cd jmp 0x268402 vaddss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x2683ed incq %r13 jmp 0x268369 incq %r9 movq 0x80(%rsp), %r13 jmp 0x268327 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x2686c4 cmpl $0x2, 0x38(%rsp) jne 0x2686c4 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x2686bc movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x268561 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x2684d2 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2684eb vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2684eb movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2684f1 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x268531 vmovups (%r14), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x2684f3 vmovups (%r14), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x268517 jmp 0x268554 vmovss (%r14), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x26853b incq %r13 jmp 0x2684af incq %r9 movq 0x80(%rsp), %r13 jmp 0x26846d movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26881b cmpl $0x2, 0x38(%rsp) jne 0x26881b movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268813 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x2686ac vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x268629 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268642 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268642 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268648 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x268680 vaddps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x26864a vaddps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x26866a jmp 0x26869f vaddss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x26868a incq %r13 jmp 0x268606 incq %r9 movq 0x80(%rsp), %r13 jmp 0x2685c4 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26898b cmpl $0x2, 0x38(%rsp) jne 0x26898b movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268983 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x268803 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x268774 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x26878d vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26878d movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268793 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x2687d3 vmovups (%r14), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x268795 vmovups (%r14), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x2687b9 jmp 0x2687f6 vmovss (%r14), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x2687dd incq %r13 jmp 0x268751 incq %r9 movq 0x80(%rsp), %r13 jmp 0x26870f movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x26897b movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x268973 vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x2688f6 movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x26890f vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26890f movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268915 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x26894a vaddps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x268917 vaddps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x268935 jmp 0x268967 vaddss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x268953 incq %r12 jmp 0x2688d3 incq %r14 jmp 0x2688bd incq %r8 jmp 0x268874 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x268af7 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x268aef vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x268a66 movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268a7f vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268a7f movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268a85 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x268ac2 vmovups (%r10), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x268a87 vmovups (%r10), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x268aa9 jmp 0x268ae3 vmovss (%r10), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x268acb incq %r12 jmp 0x268a43 incq %r14 jmp 0x268a2d incq %r8 jmp 0x2689e4 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x268c71 cmpl $0x1, 0x38(%rsp) jne 0x268c71 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x268c69 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x268b69 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x268b82 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268b82 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x268b88 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x268c17 vminps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x268bd7 vminps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x268bfd imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x268c5c vminss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x268c46 incq %r9 movq 0x18(%rsp), %r10 jmp 0x268b44 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x268db7 cmpl $0x2, 0x38(%rsp) jne 0x268db7 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268daf movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x268d9f vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x268d1c movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268d35 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268d35 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268d3b vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x268d73 vminps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x268d3d vminps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x268d5d jmp 0x268d92 vminss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x268d7d incq %r13 jmp 0x268cf9 incq %r9 movq 0x80(%rsp), %r13 jmp 0x268cb7 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x268f02 cmpl $0x2, 0x38(%rsp) jne 0x268f02 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268efa movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x268eea vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x268e67 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268e80 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268e80 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268e86 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x268ebe vminps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x268e88 vminps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x268ea8 jmp 0x268edd vminss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x268ec8 incq %r13 jmp 0x268e44 incq %r9 movq 0x80(%rsp), %r13 jmp 0x268e02 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x269062 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x26905a vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x268fdd movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268ff6 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268ff6 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268ffc vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x269031 vminps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x268ffe vminps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x26901c jmp 0x26904e vminss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x26903a incq %r12 jmp 0x268fba incq %r14 jmp 0x268fa4 incq %r8 jmp 0x268f5b movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x269200 cmpl $0x1, 0x38(%rsp) jne 0x269200 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x2691f8 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x2690d4 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x2690ed vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2690ed movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x2690f3 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x2691a6 vmovups (%r13), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x269142 vmovups (%r13), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x26917a imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2691eb vdivss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2691d5 incq %r9 movq 0x18(%rsp), %r10 jmp 0x2690af movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x26936a cmpl $0x2, 0x38(%rsp) jne 0x26936a movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x269362 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x269352 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x2692ab movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2692c4 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2692c4 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2692ca vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x269326 vmovups (%r14), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x2692cc vmovups (%r14), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x2692fe jmp 0x269345 vdivss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x269330 incq %r13 jmp 0x269288 incq %r9 movq 0x80(%rsp), %r13 jmp 0x269246 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x2694d9 cmpl $0x2, 0x38(%rsp) jne 0x2694d9 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x2694d1 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x2694c1 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x26941a movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x269433 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x269433 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x269439 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x269495 vmovups (%r14), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x26943b vmovups (%r14), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x26946d jmp 0x2694b4 vdivss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x26949f incq %r13 jmp 0x2693f7 incq %r9 movq 0x80(%rsp), %r13 jmp 0x2693b5 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x26965d movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x269655 vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x2695b4 movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2695cd vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2695cd movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2695d3 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x26962c vmovups (%r10), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x2695d5 vmovups (%r10), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x269605 jmp 0x269649 vdivss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x269635 incq %r12 jmp 0x269591 incq %r14 jmp 0x26957b incq %r8 jmp 0x269532 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax leal -0x3(%rax), %ecx cmpl $0x1, %ecx ja 0x269fa3 cmpl $0x1, 0x38(%rsp) jne 0x269fa3 movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d imull 0xf8(%rsp), %r15d xorl %r13d, %r13d movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x269f9b vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0xb0(%rsp) jne 0x2696f6 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x269718 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x269718 movq %r13, %rax shlq $0x5, %rax vmovups (%rbx,%rax), %ymm0 jmp 0x26971e vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movq 0x40(%rax), %r12 movq %r12, 0xa0(%rsp) imulq %r13, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0x28(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movq 0x40(%rax), %rdx movq %r13, %rcx movq %rdx, 0xc8(%rsp) movq %rdx, %r13 movq %rcx, 0xa8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x60(%rsp) addq %rcx, %r13 vmovaps %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x269827 vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x26979a vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x2697e6 movq 0xc8(%rsp), %rax imulq 0x138(%rsp), %rax movq 0xa8(%rsp), %r13 imulq %r13, %rax addq %rax, 0x60(%rsp) movq 0xa0(%rsp), %rax imulq 0xd0(%rsp), %rax imulq %r13, %rax addq %rax, 0x28(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d vmovdqa 0xb0(%rsp), %xmm0 jge 0x2698a9 movq 0x28(%rsp), %rax vmovd (%rax,%r14,4), %xmm1 vzeroupper callq 0x5f170 movq 0x60(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x269878 incq %r13 jmp 0x2696bc movq 0x20(%rsp), %rax movl 0x28(%rax), %eax leal -0x3(%rax), %ecx cmpl $0x1, %ecx ja 0x26a1bc cmpl $0x1, 0x38(%rsp) jne 0x26a1bc movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d imull 0xf8(%rsp), %r15d xorl %r13d, %r13d movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x26a1b4 vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0xb0(%rsp) jne 0x269942 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x269964 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x269964 movq %r13, %rax shlq $0x5, %rax vmovups (%rbx,%rax), %ymm0 jmp 0x26996a vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movq 0x40(%rax), %r12 movq %r12, 0xa0(%rsp) imulq %r13, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0x28(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movq 0x40(%rax), %rdx movq %r13, %rcx movq %rdx, 0xc8(%rsp) movq %rdx, %r13 movq %rcx, 0xa8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x60(%rsp) addq %rcx, %r13 vmovaps %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x269a73 vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x2699e6 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x269a32 movq 0xc8(%rsp), %rax imulq 0x138(%rsp), %rax movq 0xa8(%rsp), %r13 imulq %r13, %rax addq %rax, 0x60(%rsp) movq 0xa0(%rsp), %rax imulq 0xd0(%rsp), %rax imulq %r13, %rax addq %rax, 0x28(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d vmovdqa 0xb0(%rsp), %xmm0 jge 0x269af5 movq 0x28(%rsp), %rax vmovd (%rax,%r14,4), %xmm1 vzeroupper callq 0x5f0e0 movq 0x60(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x269ac4 incq %r13 jmp 0x269908 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax leal -0x3(%rax), %ecx cmpl $0x1, %ecx ja 0x26a3d5 cmpl $0x1, 0x38(%rsp) jne 0x26a3d5 movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d imull 0xf8(%rsp), %r15d xorl %r13d, %r13d movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x26a3cd vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0x60(%rsp) jne 0x269b8b movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x269bad vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x269bad movq %r13, %rax shlq $0x5, %rax vmovdqu (%rbx,%rax), %ymm0 jmp 0x269bb3 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movq 0x40(%rax), %r12 movq %r12, 0xa0(%rsp) imulq %r13, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0xb0(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movq 0x40(%rax), %rdx movq %r13, %rcx movq %rdx, 0xc8(%rsp) movq %rdx, %r13 movq %rcx, 0xa8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x28(%rsp) addq %rcx, %r13 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x269cbf vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x269c32 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x269c7e movq 0xc8(%rsp), %rax imulq 0x138(%rsp), %rax movq 0xa8(%rsp), %r13 imulq %r13, %rax addq %rax, 0x28(%rsp) movq 0xa0(%rsp), %rax imulq 0xd0(%rsp), %rax imulq %r13, %rax addq %rax, 0xb0(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d jge 0x269d44 movq 0xb0(%rsp), %rax vmovd (%rax,%r14,4), %xmm0 vmovdqa 0x60(%rsp), %xmm1 vzeroupper callq 0x5f0e0 movq 0x28(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x269d13 incq %r13 jmp 0x269b54 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax leal -0x3(%rax), %ecx cmpl $0x1, %ecx ja 0x26a5ee cmpl $0x1, 0x38(%rsp) jne 0x26a5ee movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d imull 0xf8(%rsp), %r15d xorl %r13d, %r13d movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x26a5e6 vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0x60(%rsp) jne 0x269dda movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x269dfc vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x269dfc movq %r13, %rax shlq $0x5, %rax vmovdqu (%rbx,%rax), %ymm0 jmp 0x269e02 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movq 0x40(%rax), %r12 movq %r12, 0xa0(%rsp) imulq %r13, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0xb0(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movq 0x40(%rax), %rdx movq %r13, %rcx movq %rdx, 0xc8(%rsp) movq %rdx, %r13 movq %rcx, 0xa8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x28(%rsp) addq %rcx, %r13 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x269f0e vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x269e81 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x269ecd movq 0xc8(%rsp), %rax imulq 0x138(%rsp), %rax movq 0xa8(%rsp), %r13 imulq %r13, %rax addq %rax, 0x28(%rsp) movq 0xa0(%rsp), %rax imulq 0xd0(%rsp), %rax imulq %r13, %rax addq %rax, 0xb0(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d jge 0x269f93 movq 0xb0(%rsp), %rax vmovd (%rax,%r14,4), %xmm0 vmovdqa 0x60(%rsp), %xmm1 vzeroupper callq 0x5f170 movq 0x28(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x269f62 incq %r13 jmp 0x269da3 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x3, %eax jne 0x26a807 cmpl $0x2, 0x38(%rsp) jne 0x26a807 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %ecx, %ecx movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26a7ff movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26a19f vmovd (%rsi,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26a0a0 movq %rcx, %rax shlq $0x4, %rax vmovdqu (%rsi,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26a0c2 vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a0c2 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a0c8 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a154 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a0d4 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a118 jmp 0x26a185 vmovd (%r13), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f170 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a15f movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a065 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a006 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x3, %eax jne 0x26aa29 cmpl $0x2, 0x38(%rsp) jne 0x26aa29 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %ecx, %ecx movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26aa21 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26a3b8 vmovd (%rsi,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26a2b9 movq %rcx, %rax shlq $0x4, %rax vmovdqu (%rsi,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26a2db vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a2db movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a2e1 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a36d vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a2ed vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a331 jmp 0x26a39e vmovd (%r13), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a378 movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a27e movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a21f movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x3, %eax jne 0x26ac4b cmpl $0x2, 0x38(%rsp) jne 0x26ac4b movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %ecx, %ecx movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26ac43 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26a5d1 vmovss (%rsi,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26a4d2 movq %rcx, %rax shlq $0x4, %rax vmovups (%rsi,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26a4f4 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a4f4 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a4fa vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a586 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a506 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a54a jmp 0x26a5b7 vmovd (%r13), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a591 movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a497 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a438 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x3, %eax jne 0x26ae6d cmpl $0x2, 0x38(%rsp) jne 0x26ae6d movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %ecx, %ecx movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26ae65 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26a7ea vmovss (%rsi,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26a6eb movq %rcx, %rax shlq $0x4, %rax vmovups (%rsi,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26a70d vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a70d movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a713 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a79f vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a71f vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a763 jmp 0x26a7d0 vmovd (%r13), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f170 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a7aa movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a6b0 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a651 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26b08f cmpl $0x2, 0x38(%rsp) jne 0x26b08f movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d xorl %ecx, %ecx movq 0xf8(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26b087 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26aa0c vmovd (%rsi,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26a90d movq %rcx, %rax shlq $0x4, %rax vmovdqu (%rsi,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26a92f vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a92f movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a935 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a9c1 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a941 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a985 jmp 0x26a9f2 vmovd (%r13), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f170 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a9cc movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a8d2 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a873 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26b31a cmpl $0x2, 0x38(%rsp) jne 0x26b31a movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d xorl %ecx, %ecx movq 0xf8(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26b312 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26ac2e vmovd (%rsi,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26ab2f movq %rcx, %rax shlq $0x4, %rax vmovdqu (%rsi,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26ab51 vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26ab51 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26ab57 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26abe3 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26ab63 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26aba7 jmp 0x26ac14 vmovd (%r13), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26abee movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26aaf4 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26aa95 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26b5a5 cmpl $0x2, 0x38(%rsp) jne 0x26b5a5 movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d xorl %ecx, %ecx movq 0xf8(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26b59d movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26ae50 vmovss (%rsi,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26ad51 movq %rcx, %rax shlq $0x4, %rax vmovups (%rsi,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26ad73 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26ad73 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26ad79 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26ae05 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26ad85 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26adc9 jmp 0x26ae36 vmovd (%r13), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26ae10 movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26ad16 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26acb7 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26b830 cmpl $0x2, 0x38(%rsp) jne 0x26b830 movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d xorl %ecx, %ecx movq 0xf8(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26b828 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26b072 vmovss (%rsi,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26af73 movq %rcx, %rax shlq $0x4, %rax vmovups (%rsi,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26af95 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26af95 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26af9b vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26b027 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26afa7 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26afeb jmp 0x26b058 vmovd (%r13), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f170 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26b032 movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26af38 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26aed9 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 movl 0x80(%rsp), %eax imull %r12d, %eax movl %eax, 0x80(%rsp) xorl %edx, %edx movq 0x40(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x40(%rsp) movq 0xf8(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf8(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf0(%rsp) movq 0x98(%rsp), %rax movq 0x78(%rsp), %rcx imulq %rcx, %rax movq %rax, 0x98(%rsp) movslq 0x34(%rsp), %rax imulq %rcx, %rax movq %rax, 0x78(%rsp) leaq 0x17(%rsp), %r13 leaq 0x100(%rsp), %r14 cmpq 0xf0(%rsp), %rdx je 0x26bab3 movq 0x20(%rsp), %rax movq 0x40(%rax), %rbx imulq %rdx, %rbx imulq 0x10(%rax), %rbx addq (%rax), %rbx movq 0x48(%rsp), %rax movq 0x40(%rax), %r15 imulq %rdx, %r15 imulq 0x10(%rax), %r15 addq (%rax), %r15 movq 0x98(%rsp), %rax movq %rdx, 0xa0(%rsp) imulq %rdx, %rax addq 0x18(%rsp), %rax movq %rax, 0x38(%rsp) xorl %eax, %eax cmpq 0xf8(%rsp), %rax je 0x26b302 movq 0x78(%rsp), %rcx imulq %rax, %rcx addq 0x38(%rsp), %rcx movq %rcx, 0x60(%rsp) xorl %ecx, %ecx movq %rax, 0xa8(%rsp) cmpq 0x40(%rsp), %rcx je 0x26b2fa movq 0x60(%rsp), %rax vmovd (%rax,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26b1ee movq 0x28(%rsp), %rax shlq $0x4, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26b217 vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26b217 movq 0x28(%rsp), %rax shlq $0x5, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %ymm0 jmp 0x26b21d vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r12d, %r12d leal 0x7(%r12), %eax cmpl 0x80(%rsp), %eax jge 0x26b29e vmovdqu (%rbx), %ymm0 vmovdqa %ymm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%r15) addq $0x20, %rbx addq $0x20, %r15 addl $0x8, %r12d jmp 0x26b229 vmovdqu (%rbx), %xmm0 vmovdqa %xmm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%r15) addq $0x10, %rbx addq $0x10, %r15 addl $0x4, %r12d leal 0x3(%r12), %eax cmpl 0x80(%rsp), %eax jl 0x26b26a jmp 0x26b2d3 vmovd (%rbx), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f170 vmovd %xmm0, (%r15) addq $0x4, %rbx addq $0x4, %r15 incl %r12d cmpl 0x80(%rsp), %r12d jl 0x26b2ae movq 0x28(%rsp), %rcx incq %rcx movl 0xdc(%rsp), %r12d movq 0xa8(%rsp), %rax jmp 0x26b1a7 incq %rax jmp 0x26b17c movq 0xa0(%rsp), %rdx incq %rdx jmp 0x26b124 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 movl 0x80(%rsp), %eax imull %r12d, %eax movl %eax, 0x80(%rsp) xorl %edx, %edx movq 0x40(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x40(%rsp) movq 0xf8(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf8(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf0(%rsp) movq 0x98(%rsp), %rax movq 0x78(%rsp), %rcx imulq %rcx, %rax movq %rax, 0x98(%rsp) movslq 0x34(%rsp), %rax imulq %rcx, %rax movq %rax, 0x78(%rsp) leaq 0x17(%rsp), %r13 leaq 0x100(%rsp), %r14 cmpq 0xf0(%rsp), %rdx je 0x26bab3 movq 0x20(%rsp), %rax movq 0x40(%rax), %rbx imulq %rdx, %rbx imulq 0x10(%rax), %rbx addq (%rax), %rbx movq 0x48(%rsp), %rax movq 0x40(%rax), %r15 imulq %rdx, %r15 imulq 0x10(%rax), %r15 addq (%rax), %r15 movq 0x98(%rsp), %rax movq %rdx, 0xa0(%rsp) imulq %rdx, %rax addq 0x18(%rsp), %rax movq %rax, 0x38(%rsp) xorl %eax, %eax cmpq 0xf8(%rsp), %rax je 0x26b58d movq 0x78(%rsp), %rcx imulq %rax, %rcx addq 0x38(%rsp), %rcx movq %rcx, 0x60(%rsp) xorl %ecx, %ecx movq %rax, 0xa8(%rsp) cmpq 0x40(%rsp), %rcx je 0x26b585 movq 0x60(%rsp), %rax vmovd (%rax,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26b479 movq 0x28(%rsp), %rax shlq $0x4, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26b4a2 vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26b4a2 movq 0x28(%rsp), %rax shlq $0x5, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %ymm0 jmp 0x26b4a8 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r12d, %r12d leal 0x7(%r12), %eax cmpl 0x80(%rsp), %eax jge 0x26b529 vmovdqu (%rbx), %ymm0 vmovdqa %ymm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%r15) addq $0x20, %rbx addq $0x20, %r15 addl $0x8, %r12d jmp 0x26b4b4 vmovdqu (%rbx), %xmm0 vmovdqa %xmm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%r15) addq $0x10, %rbx addq $0x10, %r15 addl $0x4, %r12d leal 0x3(%r12), %eax cmpl 0x80(%rsp), %eax jl 0x26b4f5 jmp 0x26b55e vmovd (%rbx), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%r15) addq $0x4, %rbx addq $0x4, %r15 incl %r12d cmpl 0x80(%rsp), %r12d jl 0x26b539 movq 0x28(%rsp), %rcx incq %rcx movl 0xdc(%rsp), %r12d movq 0xa8(%rsp), %rax jmp 0x26b432 incq %rax jmp 0x26b407 movq 0xa0(%rsp), %rdx incq %rdx jmp 0x26b3af movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 movl 0x80(%rsp), %eax imull %r12d, %eax movl %eax, 0x80(%rsp) xorl %edx, %edx movq 0x40(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x40(%rsp) movq 0xf8(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf8(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf0(%rsp) movq 0x98(%rsp), %rax movq 0x78(%rsp), %rcx imulq %rcx, %rax movq %rax, 0x98(%rsp) movslq 0x34(%rsp), %rax imulq %rcx, %rax movq %rax, 0x78(%rsp) leaq 0x17(%rsp), %r13 leaq 0x100(%rsp), %r14 cmpq 0xf0(%rsp), %rdx je 0x26bab3 movq 0x20(%rsp), %rax movq 0x40(%rax), %rbx imulq %rdx, %rbx imulq 0x10(%rax), %rbx addq (%rax), %rbx movq 0x48(%rsp), %rax movq 0x40(%rax), %r15 imulq %rdx, %r15 imulq 0x10(%rax), %r15 addq (%rax), %r15 movq 0x98(%rsp), %rax movq %rdx, 0xa0(%rsp) imulq %rdx, %rax addq 0x18(%rsp), %rax movq %rax, 0x38(%rsp) xorl %eax, %eax cmpq 0xf8(%rsp), %rax je 0x26b818 movq 0x78(%rsp), %rcx imulq %rax, %rcx addq 0x38(%rsp), %rcx movq %rcx, 0x60(%rsp) xorl %ecx, %ecx movq %rax, 0xa8(%rsp) cmpq 0x40(%rsp), %rcx je 0x26b810 movq 0x60(%rsp), %rax vmovss (%rax,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26b704 movq 0x28(%rsp), %rax shlq $0x4, %rax movq 0x60(%rsp), %rcx vmovups (%rcx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26b72d vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26b72d movq 0x28(%rsp), %rax shlq $0x5, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %ymm0 jmp 0x26b733 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r12d, %r12d leal 0x7(%r12), %eax cmpl 0x80(%rsp), %eax jge 0x26b7b4 vmovdqu (%rbx), %ymm0 vmovdqa %ymm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%r15) addq $0x20, %rbx addq $0x20, %r15 addl $0x8, %r12d jmp 0x26b73f vmovdqu (%rbx), %xmm0 vmovdqa %xmm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%r15) addq $0x10, %rbx addq $0x10, %r15 addl $0x4, %r12d leal 0x3(%r12), %eax cmpl 0x80(%rsp), %eax jl 0x26b780 jmp 0x26b7e9 vmovd (%rbx), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%r15) addq $0x4, %rbx addq $0x4, %r15 incl %r12d cmpl 0x80(%rsp), %r12d jl 0x26b7c4 movq 0x28(%rsp), %rcx incq %rcx movl 0xdc(%rsp), %r12d movq 0xa8(%rsp), %rax jmp 0x26b6bd incq %rax jmp 0x26b692 movq 0xa0(%rsp), %rdx incq %rdx jmp 0x26b63a movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 movl 0x80(%rsp), %eax imull %r12d, %eax movl %eax, 0x80(%rsp) xorl %edx, %edx movq 0x40(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x40(%rsp) movq 0xf8(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf8(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf0(%rsp) movq 0x98(%rsp), %rax movq 0x78(%rsp), %rcx imulq %rcx, %rax movq %rax, 0x98(%rsp) movslq 0x34(%rsp), %rax imulq %rcx, %rax movq %rax, 0x78(%rsp) leaq 0x17(%rsp), %r13 leaq 0x100(%rsp), %r14 cmpq 0xf0(%rsp), %rdx je 0x26bab3 movq 0x20(%rsp), %rax movq 0x40(%rax), %rbx imulq %rdx, %rbx imulq 0x10(%rax), %rbx addq (%rax), %rbx movq 0x48(%rsp), %rax movq 0x40(%rax), %r15 imulq %rdx, %r15 imulq 0x10(%rax), %r15 addq (%rax), %r15 movq 0x98(%rsp), %rax movq %rdx, 0xa0(%rsp) imulq %rdx, %rax addq 0x18(%rsp), %rax movq %rax, 0x38(%rsp) xorl %eax, %eax cmpq 0xf8(%rsp), %rax je 0x26baa3 movq 0x78(%rsp), %rcx imulq %rax, %rcx addq 0x38(%rsp), %rcx movq %rcx, 0x60(%rsp) xorl %ecx, %ecx movq %rax, 0xa8(%rsp) cmpq 0x40(%rsp), %rcx je 0x26ba9b movq 0x60(%rsp), %rax vmovss (%rax,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26b98f movq 0x28(%rsp), %rax shlq $0x4, %rax movq 0x60(%rsp), %rcx vmovups (%rcx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26b9b8 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26b9b8 movq 0x28(%rsp), %rax shlq $0x5, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %ymm0 jmp 0x26b9be vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r12d, %r12d leal 0x7(%r12), %eax cmpl 0x80(%rsp), %eax jge 0x26ba3f vmovdqu (%rbx), %ymm0 vmovdqa %ymm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%r15) addq $0x20, %rbx addq $0x20, %r15 addl $0x8, %r12d jmp 0x26b9ca vmovdqu (%rbx), %xmm0 vmovdqa %xmm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%r15) addq $0x10, %rbx addq $0x10, %r15 addl $0x4, %r12d leal 0x3(%r12), %eax cmpl 0x80(%rsp), %eax jl 0x26ba0b jmp 0x26ba74 vmovd (%rbx), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f170 vmovd %xmm0, (%r15) addq $0x4, %rbx addq $0x4, %r15 incl %r12d cmpl 0x80(%rsp), %r12d jl 0x26ba4f movq 0x28(%rsp), %rcx incq %rcx movl 0xdc(%rsp), %r12d movq 0xa8(%rsp), %rax jmp 0x26b948 incq %rax jmp 0x26b91d movq 0xa0(%rsp), %rdx incq %rdx jmp 0x26b8c5 movq 0x58(%rsp), %rax testq %rax, %rax je 0x26bae9 lock decl (%rax) jne 0x26bae9 movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x26badc movq (%rdi), %rax movq 0x18(%rsp), %rsi vzeroupper callq 0x18(%rax) jmp 0x26bae9 movq 0x18(%rsp), %rdi vzeroupper callq 0x5f3e0 leaq -0x28(%rbp), %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq 0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq 0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq 0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bcbe jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq 0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bcbe jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq 0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bcbe jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi jne 0x26bca8 movq %rsi, %rdi callq 0x5f3e0 jmp 0x26bcc1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bcbe jmp 0x26bd56 jmp 0x26bcbe movq %rax, %r15 movq %rbx, %rsi jmp 0x26bd1c jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 movq %rax, %r15 movq 0x18(%rsp), %rsi movq 0x50(%rsp), %rdi movq 0x58(%rsp), %rax testq %rax, %rax je 0x26bd4b lock decl (%rax) jne 0x26bd4b testq %rdi, %rdi jne 0x26bd42 movq %rsi, %rdi vzeroupper callq 0x5f3e0 jmp 0x26bd4b movq (%rdi), %rax vzeroupper callq 0x18(%rax) movq %r15, %rdi vzeroupper callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp
virtual thunk to ncnn::BinaryOp_x86_fma::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int BinaryOp_x86_fma::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const bool b_is_scalar = bottom_blobs[1].w * bottom_blobs[1].h * bottom_blobs[1].d * bottom_blobs[1].c * bottom_blobs[1].elempack == 1; const bool a_rank_is_lower = bottom_blobs[0].dims < bottom_blobs[1].dims && !b_is_scalar; const bool a_size_is_lower = bottom_blobs[0].w * bottom_blobs[0].h * bottom_blobs[0].d * bottom_blobs[0].c * bottom_blobs[0].elempack < bottom_blobs[1].w * bottom_blobs[1].h * bottom_blobs[1].d * bottom_blobs[1].c * bottom_blobs[1].elempack; const bool a_is_lower = a_rank_is_lower \|\| (!a_rank_is_lower && a_size_is_lower); const Mat& A = a_is_lower ? bottom_blobs[1] : bottom_blobs[0]; const Mat& B = a_is_lower ? bottom_blobs[0] : bottom_blobs[1]; const int op_type_r = a_is_lower ? get_reverse_op_type(op_type) : op_type; Mat& top_blob = top_blobs[0]; top_blob.create_like(A, opt.blob_allocator); if (top_blob.empty()) return -100; // B is a scalar if (B.w * B.h * B.d * B.c * B.elempack == 1) { return binary_op_scalar(A, B[0], top_blob, op_type_r, opt); } // no broadcast if (A.dims == B.dims && A.w == B.w && A.h == B.h && A.d == B.d && A.c == B.c && A.elempack == B.elempack) { return binary_op_no_broadcast(A, B, top_blob, op_type_r, opt); } // broadcast B for inner axis if ((B.dims < A.dims) \|\| (A.dims == 2 && B.w == 1 && B.h == A.h) \|\| (A.dims == 3 && B.w == 1 && B.h == 1 && B.c == A.c) \|\| (A.dims == 3 && B.w == 1 && B.h == A.h && B.c == A.c) \|\| (A.dims == 4 && B.w == 1 && B.h == 1 && B.d == 1 && B.c == A.c) \|\| (A.dims == 4 && B.w == 1 && B.h == 1 && B.d == A.d && B.c == A.c) \|\| (A.dims == 4 && B.w == 1 && B.h == A.h && B.d == A.d && B.c == A.c)) { return binary_op_broadcast_inner(A, B, top_blob, op_type_r, opt); } // broadcast B for outer axis if (B.elempack == 1 && ((A.dims == 2 && B.w == A.w && B.h == 1) \|\| (A.dims == 3 && B.w == A.w && B.h == 1 && B.c == 1) \|\| (A.dims == 3 && B.w == A.w && B.h == A.h && B.c == 1) \|\| (A.dims == 4 && B.w == A.w && B.h == 1 && B.d == 1 && B.c == 1) \|\| (A.dims == 4 && B.w == A.w && B.h == A.h && B.d == 1 && B.c == 1) \|\| (A.dims == 4 && B.w == A.w && B.h == A.h && B.d == A.d && B.c == 1))) { return binary_op_broadcast_outer(A, B, top_blob, op_type_r, opt); } // some special broadcast rule here if (A.dims == 3 && B.dims == 3 && A.w == B.w && B.h == 1 && A.c == B.c) { return binary_op_broadcast_20(A, B, top_blob, op_type_r, opt); } return 0; }	movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x25fe94	/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp
ncnn::BinaryOp_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int BinaryOp_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { using namespace BinaryOp_x86_fma_functor; if (op_type == Operation_ADD) return binary_op_scalar_inplace<binary_op_add>(bottom_top_blob, b, opt); if (op_type == Operation_SUB) return binary_op_scalar_inplace<binary_op_sub>(bottom_top_blob, b, opt); if (op_type == Operation_MUL) return binary_op_scalar_inplace<binary_op_mul>(bottom_top_blob, b, opt); if (op_type == Operation_DIV) return binary_op_scalar_inplace<binary_op_div>(bottom_top_blob, b, opt); if (op_type == Operation_MAX) return binary_op_scalar_inplace<binary_op_max>(bottom_top_blob, b, opt); if (op_type == Operation_MIN) return binary_op_scalar_inplace<binary_op_min>(bottom_top_blob, b, opt); if (op_type == Operation_POW) return binary_op_scalar_inplace<binary_op_pow>(bottom_top_blob, b, opt); if (op_type == Operation_RSUB) return binary_op_scalar_inplace<binary_op_rsub>(bottom_top_blob, b, opt); if (op_type == Operation_RDIV) return binary_op_scalar_inplace<binary_op_rdiv>(bottom_top_blob, b, opt); if (op_type == Operation_RPOW) return binary_op_scalar_inplace<binary_op_rpow>(bottom_top_blob, b, opt); if (op_type == Operation_ATAN2) return binary_op_scalar_inplace<binary_op_atan2>(bottom_top_blob, b, opt); if (op_type == Operation_RATAN2) return binary_op_scalar_inplace<binary_op_ratan2>(bottom_top_blob, b, opt); return 0; }	pushq %rbp movq %rsp, %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx andq $-0x20, %rsp subq $0xe0, %rsp movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %eax cmpq $0xb, %rax ja 0x26c8c4 movq %rsi, %r14 leaq 0x18b6f8(%rip), %rcx # 0x3f749c movslq (%rcx,%rax,4), %rax addq %rcx, %rax movq %rsi, 0x38(%rsp) jmpq %rax vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26be37 vaddps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26be01 vaddps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26be21 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26be5d vaddss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26be4a incq %rdx jmp 0x26bdda vbroadcastss 0xd8(%rdi,%rdx), %ymm0 vmovaps %ymm0, 0x40(%rsp) movl 0x30(%r14), %ebx movl 0x38(%r14), %eax imull 0x2c(%r14), %ebx imull 0x34(%r14), %ebx imull 0x18(%r14), %ebx xorl %r15d, %r15d testl %eax, %eax cmovlel %r15d, %eax movq %rax, 0x18(%rsp) cmpq 0x18(%rsp), %r15 je 0x26c8c4 movq (%r14), %r12 movq 0x10(%r14), %rax movq 0x40(%r14), %r14 movq %r14, 0x28(%rsp) movq %r15, 0x30(%rsp) imulq %r15, %r14 movq %rax, 0x20(%rsp) imulq %rax, %r14 addq %r12, %r14 vmovaps 0x40(%rsp), %ymm0 vmovaps %ymm0, 0xa0(%rsp) xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %ebx, %eax jge 0x26bf1c vmovups (%r14), %ymm0 vmovaps %ymm0, 0x60(%rsp) leaq 0x17(%rsp), %rdi leaq 0x60(%rsp), %rsi leaq 0xa0(%rsp), %rdx callq 0x26d916 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x26bedf vmovaps 0x40(%rsp), %ymm0 vmovaps %xmm0, 0x60(%rsp) leal 0x3(%r15), %eax cmpl %ebx, %eax jge 0x26bf6b vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x90(%rsp) leaq 0x17(%rsp), %rdi leaq 0x90(%rsp), %rsi leaq 0x60(%rsp), %rdx vzeroupper callq 0x26daa0 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 jmp 0x26bf28 movq 0x28(%rsp), %rax imulq 0x20(%rsp), %rax movq 0x30(%rsp), %r15 imulq %r15, %rax addq %rax, %r12 movq 0x38(%rsp), %r14 cmpl %ebx, %r13d jge 0x26bfab vmovss (%r12,%r13,4), %xmm1 vmovaps 0x40(%rsp), %ymm0 vzeroupper callq 0x5f170 vmovss %xmm0, (%r12,%r13,4) incq %r13 jmp 0x26bf87 incq %r15 jmp 0x26be9a vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c038 vmaxps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c002 vmaxps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c022 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c05e vmaxss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c04b incq %rdx jmp 0x26bfdb vbroadcastss 0xd8(%rdi,%rdx), %ymm0 vmovaps %ymm0, 0x40(%rsp) movl 0x30(%r14), %ebx movl 0x38(%r14), %eax imull 0x2c(%r14), %ebx imull 0x34(%r14), %ebx imull 0x18(%r14), %ebx xorl %r15d, %r15d testl %eax, %eax cmovlel %r15d, %eax movq %rax, 0x18(%rsp) cmpq 0x18(%rsp), %r15 je 0x26c8c4 movq (%r14), %r12 movq 0x10(%r14), %rax movq 0x40(%r14), %r14 movq %r14, 0x28(%rsp) movq %r15, 0x30(%rsp) imulq %r15, %r14 movq %rax, 0x20(%rsp) imulq %rax, %r14 addq %r12, %r14 vmovaps 0x40(%rsp), %ymm0 vmovaps %ymm0, 0xa0(%rsp) xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %ebx, %eax jge 0x26c11d vmovups (%r14), %ymm0 vmovaps %ymm0, 0x60(%rsp) leaq 0x17(%rsp), %rdi leaq 0x60(%rsp), %rsi leaq 0xa0(%rsp), %rdx callq 0x26d220 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x26c0e0 vmovaps 0x40(%rsp), %ymm0 vmovaps %xmm0, 0x60(%rsp) leal 0x3(%r15), %eax cmpl %ebx, %eax jge 0x26c16c vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x90(%rsp) leaq 0x17(%rsp), %rdi leaq 0x90(%rsp), %rsi leaq 0x60(%rsp), %rdx vzeroupper callq 0x26d430 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 jmp 0x26c129 movq 0x28(%rsp), %rax imulq 0x20(%rsp), %rax movq 0x30(%rsp), %r15 imulq %r15, %rax addq %rax, %r12 movq 0x38(%rsp), %r14 cmpl %ebx, %r13d jge 0x26c1ac vmovss (%r12,%r13,4), %xmm1 vmovaps 0x40(%rsp), %ymm0 vzeroupper callq 0x5f0e0 vmovss %xmm0, (%r12,%r13,4) incq %r13 jmp 0x26c188 incq %r15 jmp 0x26c09b vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c239 vmulps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c203 vmulps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c223 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c25f vmulss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c24c incq %rdx jmp 0x26c1dc movl 0x30(%r14), %eax imull 0x2c(%r14), %eax imull 0x34(%r14), %eax movl 0x38(%r14), %ecx imull 0x18(%r14), %eax vmovss 0x182a02(%rip), %xmm0 # 0x3eec88 vdivss 0xd8(%rdi,%rdx), %xmm0, %xmm0 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c2fe vmulps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c2c8 vmulps (%r10), %xmm1, %xmm3 vmovaps %xmm3, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c2e8 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c324 vmulss (%rsi,%rdi,4), %xmm0, %xmm3 vmovss %xmm3, (%rsi,%rdi,4) incq %rdi jmp 0x26c311 incq %rdx jmp 0x26c2a1 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c3b1 vsubps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c37b vsubps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c39b imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c3d7 vsubss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c3c4 incq %rdx jmp 0x26c354 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c46c vmovups (%r10), %ymm1 vsubps %ymm0, %ymm1, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c42e vmovaps (%r10), %xmm1 vsubps %xmm0, %xmm1, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c452 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c496 vmovss (%rsi,%rdi,4), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c47f incq %rdx jmp 0x26c407 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c523 vminps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c4ed vminps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c50d imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c549 vminss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c536 incq %rdx jmp 0x26c4c6 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 vmovaps %ymm0, 0x40(%rsp) movl 0x30(%r14), %ebx movl 0x38(%r14), %eax imull 0x2c(%r14), %ebx imull 0x34(%r14), %ebx imull 0x18(%r14), %ebx xorl %r15d, %r15d testl %eax, %eax cmovlel %r15d, %eax movq %rax, 0x18(%rsp) cmpq 0x18(%rsp), %r15 je 0x26c8c4 movq (%r14), %r12 movq 0x10(%r14), %rax movq 0x40(%r14), %r14 movq %r14, 0x28(%rsp) movq %r15, 0x30(%rsp) imulq %r15, %r14 movq %rax, 0x20(%rsp) imulq %rax, %r14 addq %r12, %r14 vmovaps 0x40(%rsp), %ymm0 vmovaps %ymm0, 0xa0(%rsp) xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %ebx, %eax jge 0x26c608 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x60(%rsp) leaq 0x17(%rsp), %rdi leaq 0x60(%rsp), %rsi leaq 0xa0(%rsp), %rdx callq 0x26ce36 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x26c5cb vmovaps 0x40(%rsp), %ymm0 vmovaps %xmm0, 0x60(%rsp) leal 0x3(%r15), %eax cmpl %ebx, %eax jge 0x26c657 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x90(%rsp) leaq 0x17(%rsp), %rdi leaq 0x90(%rsp), %rsi leaq 0x60(%rsp), %rdx vzeroupper callq 0x26d046 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 jmp 0x26c614 movq 0x28(%rsp), %rax imulq 0x20(%rsp), %rax movq 0x30(%rsp), %r15 imulq %r15, %rax addq %rax, %r12 movq 0x38(%rsp), %r14 cmpl %ebx, %r13d jge 0x26c697 vmovss (%r12,%r13,4), %xmm0 vmovaps 0x40(%rsp), %ymm1 vzeroupper callq 0x5f0e0 vmovss %xmm0, (%r12,%r13,4) incq %r13 jmp 0x26c673 incq %r15 jmp 0x26c586 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 vmovaps %ymm0, 0x40(%rsp) movl 0x30(%r14), %ebx movl 0x38(%r14), %eax imull 0x2c(%r14), %ebx imull 0x34(%r14), %ebx imull 0x18(%r14), %ebx xorl %r15d, %r15d testl %eax, %eax cmovlel %r15d, %eax movq %rax, 0x18(%rsp) cmpq 0x18(%rsp), %r15 je 0x26c8c4 movq (%r14), %r12 movq 0x10(%r14), %rax movq 0x40(%r14), %r14 movq %r14, 0x28(%rsp) movq %r15, 0x30(%rsp) imulq %r15, %r14 movq %rax, 0x20(%rsp) imulq %rax, %r14 addq %r12, %r14 vmovaps 0x40(%rsp), %ymm0 vmovaps %ymm0, 0xa0(%rsp) xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %ebx, %eax jge 0x26c756 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x60(%rsp) leaq 0x17(%rsp), %rdi leaq 0x60(%rsp), %rsi leaq 0xa0(%rsp), %rdx callq 0x26d60a vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x26c719 vmovaps 0x40(%rsp), %ymm0 vmovaps %xmm0, 0x60(%rsp) leal 0x3(%r15), %eax cmpl %ebx, %eax jge 0x26c7a5 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x90(%rsp) leaq 0x17(%rsp), %rdi leaq 0x90(%rsp), %rsi leaq 0x60(%rsp), %rdx vzeroupper callq 0x26d794 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 jmp 0x26c762 movq 0x28(%rsp), %rax imulq 0x20(%rsp), %rax movq 0x30(%rsp), %r15 imulq %r15, %rax addq %rax, %r12 movq 0x38(%rsp), %r14 cmpl %ebx, %r13d jge 0x26c7e5 vmovss (%r12,%r13,4), %xmm0 vmovaps 0x40(%rsp), %ymm1 vzeroupper callq 0x5f170 vmovss %xmm0, (%r12,%r13,4) incq %r13 jmp 0x26c7c1 incq %r15 jmp 0x26c6d4 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c896 vmovups (%r10), %ymm1 vrcpps %ymm1, %ymm2 vmulps %ymm2, %ymm0, %ymm3 vfmsub213ps %ymm0, %ymm3, %ymm1 # ymm1 = (ymm3 ymm1) - ymm0 vfnmadd213ps %ymm3, %ymm2, %ymm1 # ymm1 = -(ymm2 * ymm1) + ymm3 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c83c vmovaps (%r10), %xmm1 vrcpps %xmm1, %xmm2 vmulps %xmm2, %xmm0, %xmm3 vfmsub213ps %xmm0, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm1) - xmm0 vfnmadd213ps %xmm3, %xmm2, %xmm1 # xmm1 = -(xmm2 * xmm1) + xmm3 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c86e imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c8bc vdivss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c8a9 incq %rdx jmp 0x26c815 xorl %eax, %eax leaq -0x28(%rbp), %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp
virtual thunk to ncnn::BinaryOp_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int BinaryOp_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { using namespace BinaryOp_x86_fma_functor; if (op_type == Operation_ADD) return binary_op_scalar_inplace<binary_op_add>(bottom_top_blob, b, opt); if (op_type == Operation_SUB) return binary_op_scalar_inplace<binary_op_sub>(bottom_top_blob, b, opt); if (op_type == Operation_MUL) return binary_op_scalar_inplace<binary_op_mul>(bottom_top_blob, b, opt); if (op_type == Operation_DIV) return binary_op_scalar_inplace<binary_op_div>(bottom_top_blob, b, opt); if (op_type == Operation_MAX) return binary_op_scalar_inplace<binary_op_max>(bottom_top_blob, b, opt); if (op_type == Operation_MIN) return binary_op_scalar_inplace<binary_op_min>(bottom_top_blob, b, opt); if (op_type == Operation_POW) return binary_op_scalar_inplace<binary_op_pow>(bottom_top_blob, b, opt); if (op_type == Operation_RSUB) return binary_op_scalar_inplace<binary_op_rsub>(bottom_top_blob, b, opt); if (op_type == Operation_RDIV) return binary_op_scalar_inplace<binary_op_rdiv>(bottom_top_blob, b, opt); if (op_type == Operation_RPOW) return binary_op_scalar_inplace<binary_op_rpow>(bottom_top_blob, b, opt); if (op_type == Operation_ATAN2) return binary_op_scalar_inplace<binary_op_atan2>(bottom_top_blob, b, opt); if (op_type == Operation_RATAN2) return binary_op_scalar_inplace<binary_op_ratan2>(bottom_top_blob, b, opt); return 0; }	pushq %rax movq (%rdi), %rax addq -0x58(%rax), %rdi callq 0x26bd6a xorl %eax, %eax popq %rcx retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp
int ncnn::binary_op_no_broadcast<ncnn::BinaryOp_x86_fma_functor::binary_op_sub>(ncnn::Mat const&, ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&)	static int binary_op_no_broadcast(const Mat& a, const Mat& b, Mat& c, const Option& opt) { Op op; const int channels = a.c; const int size = a.w * a.h * a.d * a.elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = a.channel(q); const float* ptr1 = b.channel(q); float* outptr = c.channel(q); int i = 0; #if __SSE2__ #if __AVX__ #if __AVX512F__ for (; i + 15 < size; i += 16) { __m512 _p = _mm512_loadu_ps(ptr); __m512 _p1 = _mm512_loadu_ps(ptr1); __m512 _outp = op.func_pack16(_p, _p1); _mm512_storeu_ps(outptr, _outp); ptr += 16; ptr1 += 16; outptr += 16; } #endif // __AVX512F__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _p1 = _mm256_loadu_ps(ptr1); __m256 _outp = op.func_pack8(_p, _p1); _mm256_storeu_ps(outptr, _outp); ptr += 8; ptr1 += 8; outptr += 8; } #endif // __AVX__ for (; i + 3 < size; i += 4) { __m128 _p = _mm_load_ps(ptr); __m128 _p1 = _mm_load_ps(ptr1); __m128 _outp = op.func_pack4(_p, _p1); _mm_store_ps(outptr, _outp); ptr += 4; ptr1 += 4; outptr += 4; } #endif // __SSE2__ for (; i < size; i++) { outptr = op.func(ptr, *ptr1); ptr += 1; ptr1 += 1; outptr += 1; } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %rbx movl 0x30(%rdi), %eax movl 0x38(%rdi), %ecx imull 0x2c(%rdi), %eax imull 0x34(%rdi), %eax imull 0x18(%rdi), %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx cmpq %rcx, %r8 je 0x26c9bc movq 0x10(%rsi), %r9 imulq 0x40(%rsi), %r9 movq 0x10(%rdi), %r10 imulq %r8, %r9 addq (%rsi), %r9 movq 0x10(%rdx), %r11 imulq 0x40(%rdx), %r11 imulq %r8, %r11 addq (%rdx), %r11 imulq 0x40(%rdi), %r10 imulq %r8, %r10 addq (%rdi), %r10 xorl %ebx, %ebx xorl %r14d, %r14d xorl %r15d, %r15d leal 0x7(%r15), %ebp cmpl %eax, %ebp jge 0x26c991 vmovups (%r10,%r14), %ymm0 vsubps (%r9,%r14), %ymm0, %ymm0 vmovups %ymm0, (%r11,%r14) addl $0x8, %r15d addq $0x20, %r14 addq $0x8, %rbx jmp 0x26c94b vmovaps (%r10,%r14), %xmm0 vsubps (%r9,%r14), %xmm0, %xmm0 vmovaps %xmm0, (%r11,%r14) addl $0x4, %r15d addq $0x10, %r14 addq $0x4, %rbx leal 0x3(%r15), %ebp cmpl %eax, %ebp jl 0x26c973 jmp 0x26c9b0 vmovss (%r10,%rbx,4), %xmm0 vsubss (%r9,%rbx,4), %xmm0, %xmm0 vmovss %xmm0, (%r11,%rbx,4) incq %rbx cmpl %eax, %ebx jl 0x26c99b incq %r8 jmp 0x26c90a popq %rbx popq %r14 popq %r15 popq %rbp vzeroupper retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp
int ncnn::binary_op_no_broadcast<ncnn::BinaryOp_x86_fma_functor::binary_op_div>(ncnn::Mat const&, ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&)	static int binary_op_no_broadcast(const Mat& a, const Mat& b, Mat& c, const Option& opt) { Op op; const int channels = a.c; const int size = a.w * a.h * a.d * a.elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = a.channel(q); const float* ptr1 = b.channel(q); float* outptr = c.channel(q); int i = 0; #if __SSE2__ #if __AVX__ #if __AVX512F__ for (; i + 15 < size; i += 16) { __m512 _p = _mm512_loadu_ps(ptr); __m512 _p1 = _mm512_loadu_ps(ptr1); __m512 _outp = op.func_pack16(_p, _p1); _mm512_storeu_ps(outptr, _outp); ptr += 16; ptr1 += 16; outptr += 16; } #endif // __AVX512F__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _p1 = _mm256_loadu_ps(ptr1); __m256 _outp = op.func_pack8(_p, _p1); _mm256_storeu_ps(outptr, _outp); ptr += 8; ptr1 += 8; outptr += 8; } #endif // __AVX__ for (; i + 3 < size; i += 4) { __m128 _p = _mm_load_ps(ptr); __m128 _p1 = _mm_load_ps(ptr1); __m128 _outp = op.func_pack4(_p, _p1); _mm_store_ps(outptr, _outp); ptr += 4; ptr1 += 4; outptr += 4; } #endif // __SSE2__ for (; i < size; i++) { outptr = op.func(ptr, *ptr1); ptr += 1; ptr1 += 1; outptr += 1; } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %rbx movl 0x30(%rdi), %eax movl 0x38(%rdi), %ecx imull 0x2c(%rdi), %eax imull 0x34(%rdi), %eax imull 0x18(%rdi), %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx cmpq %rcx, %r8 je 0x26cabd movq 0x10(%rsi), %r9 imulq 0x40(%rsi), %r9 movq 0x10(%rdi), %r10 imulq %r8, %r9 addq (%rsi), %r9 movq 0x10(%rdx), %r11 imulq 0x40(%rdx), %r11 imulq %r8, %r11 addq (%rdx), %r11 imulq 0x40(%rdi), %r10 imulq %r8, %r10 addq (%rdi), %r10 xorl %ebx, %ebx xorl %r14d, %r14d xorl %r15d, %r15d leal 0x7(%r15), %ebp cmpl %eax, %ebp jge 0x26ca92 vmovups (%r9,%r14), %ymm0 vrcpps %ymm0, %ymm1 vmovups (%r10,%r14), %ymm2 vmulps %ymm1, %ymm2, %ymm3 vfmsub213ps %ymm2, %ymm3, %ymm0 # ymm0 = (ymm3 * ymm0) - ymm2 vfnmadd213ps %ymm3, %ymm1, %ymm0 # ymm0 = -(ymm1 * ymm0) + ymm3 vmovups %ymm0, (%r11,%r14) addl $0x8, %r15d addq $0x20, %r14 addq $0x8, %rbx jmp 0x26ca28 vmovaps (%r10,%r14), %xmm0 vmovaps (%r9,%r14), %xmm1 vrcpps %xmm1, %xmm2 vmulps %xmm2, %xmm0, %xmm3 vfmsub213ps %xmm0, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm1) - xmm0 vfnmadd213ps %xmm3, %xmm2, %xmm1 # xmm1 = -(xmm2 * xmm1) + xmm3 vmovaps %xmm1, (%r11,%r14) addl $0x4, %r15d addq $0x10, %r14 addq $0x4, %rbx leal 0x3(%r15), %ebp cmpl %eax, %ebp jl 0x26ca62 jmp 0x26cab1 vmovss (%r10,%rbx,4), %xmm0 vdivss (%r9,%rbx,4), %xmm0, %xmm0 vmovss %xmm0, (%r11,%rbx,4) incq %rbx cmpl %eax, %ebx jl 0x26ca9c incq %r8 jmp 0x26c9e7 popq %rbx popq %r14 popq %r15 popq %rbp vzeroupper retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp
ncnn::UnaryOp_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int UnaryOp_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { using namespace UnaryOp_x86_fma_functor; if (op_type == Operation_ABS) return unary_op_inplace<unary_op_abs>(bottom_top_blob, opt); if (op_type == Operation_NEG) return unary_op_inplace<unary_op_neg>(bottom_top_blob, opt); if (op_type == Operation_FLOOR) return unary_op_inplace<unary_op_floor>(bottom_top_blob, opt); if (op_type == Operation_CEIL) return unary_op_inplace<unary_op_ceil>(bottom_top_blob, opt); if (op_type == Operation_SQUARE) return unary_op_inplace<unary_op_square>(bottom_top_blob, opt); if (op_type == Operation_SQRT) return unary_op_inplace<unary_op_sqrt>(bottom_top_blob, opt); if (op_type == Operation_RSQRT) return unary_op_inplace<unary_op_rsqrt>(bottom_top_blob, opt); if (op_type == Operation_EXP) return unary_op_inplace<unary_op_exp>(bottom_top_blob, opt); if (op_type == Operation_LOG) return unary_op_inplace<unary_op_log>(bottom_top_blob, opt); if (op_type == Operation_SIN) return unary_op_inplace<unary_op_sin>(bottom_top_blob, opt); if (op_type == Operation_COS) return unary_op_inplace<unary_op_cos>(bottom_top_blob, opt); if (op_type == Operation_TAN) return unary_op_inplace<unary_op_tan>(bottom_top_blob, opt); if (op_type == Operation_ASIN) return unary_op_inplace<unary_op_asin>(bottom_top_blob, opt); if (op_type == Operation_ACOS) return unary_op_inplace<unary_op_acos>(bottom_top_blob, opt); if (op_type == Operation_ATAN) return unary_op_inplace<unary_op_atan>(bottom_top_blob, opt); if (op_type == Operation_RECIPROCAL) return unary_op_inplace<unary_op_reciprocal>(bottom_top_blob, opt); if (op_type == Operation_TANH) return unary_op_inplace<unary_op_tanh>(bottom_top_blob, opt); if (op_type == Operation_LOG10) return unary_op_inplace<unary_op_log10>(bottom_top_blob, opt); return 0; }	pushq %rbp movq %rsp, %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx andq $-0x20, %rsp subq $0x80, %rsp movq %rsi, 0x18(%rsp) movq (%rdi), %rax movq -0x18(%rax), %rax movl 0xd0(%rdi,%rax), %eax cmpq $0x11, %rax ja 0x281716 leaq 0x177d68(%rip), %rcx # 0x3f8348 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq %rax movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x1712f0(%rip), %ymm0 # 0x3f1900 vbroadcastss 0x1712e7(%rip), %xmm1 # 0x3f1900 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x28067b vandps (%r10), %ymm0, %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280645 vandps (%r10), %xmm1, %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280665 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2806a5 vmovss (%rsi,%rdi,4), %xmm2 vandps %xmm1, %xmm2, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x28068e incq %rdx jmp 0x280619 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280775 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2828e6 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x280711 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2829f4 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280747 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2807b3 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f160 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x280795 incq %r15 jmp 0x2806d6 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280883 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282726 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28081f vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x282806 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280855 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2808c1 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f540 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2808a3 incq %r15 jmp 0x2807e4 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280991 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282016 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28092d vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2821ce vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280963 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2809cf vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f4c0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2809b1 incq %r15 jmp 0x2808f2 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x280a5f vmovups (%r10), %ymm0 vmulps %ymm0, %ymm0, %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280a21 vmovaps (%r10), %xmm0 vmulps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280a45 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x280a89 vmovss (%rsi,%rdi,4), %xmm0 vmulss %xmm0, %xmm0, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x280a72 incq %rdx jmp 0x2809f5 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280b59 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282330 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x280af5 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x282422 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280b2b movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x280b97 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f240 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x280b79 incq %r15 jmp 0x280aba movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280c67 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x281b3e vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x280c03 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x281c96 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280c39 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x280ca5 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f3b0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x280c87 incq %r15 jmp 0x280bc8 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x280d2f vroundps $0x1, (%r10), %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280cf7 vroundps $0x1, (%r10), %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280d18 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x280d57 vroundss $0x9, (%rsi,%rdi,4), %xmm1, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x280d42 incq %rdx jmp 0x280ccb movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16df02(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x16def9(%rip), %xmm1 # 0x3eec88 vmovss 0x16def1(%rip), %xmm2 # 0x3eec88 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x280e15 vmovups (%r10), %ymm3 vrcpps %ymm3, %ymm4 vfmsub213ps %ymm0, %ymm4, %ymm3 # ymm3 = (ymm4 ymm3) - ymm0 vfnmadd132ps %ymm4, %ymm4, %ymm3 # ymm3 = -(ymm3 * ymm4) + ymm4 vmovups %ymm3, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280dc3 vmovaps (%r10), %xmm3 vrcpps %xmm3, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) - xmm1 vfnmadd132ps %xmm4, %xmm4, %xmm3 # xmm3 = -(xmm3 * xmm4) + xmm4 vmovaps %xmm3, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280df1 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x280e3b vdivss (%rsi,%rdi,4), %xmm2, %xmm3 vmovss %xmm3, (%rsi,%rdi,4) incq %rdi jmp 0x280e28 incq %rdx jmp 0x280d97 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x280ec5 vroundps $0x2, (%r10), %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280e8d vroundps $0x2, (%r10), %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280eae imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x280eed vroundss $0xa, (%rsi,%rdi,4), %xmm1, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x280ed8 incq %rdx jmp 0x280e61 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280fbd vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x28173c vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x280f59 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x281826 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280f8f movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x280ffb vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f410 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x280fdd incq %r15 jmp 0x280f1e movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x17018a(%rip), %ymm0 # 0x3f11b4 vbroadcastss 0x170181(%rip), %xmm1 # 0x3f11b4 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x281095 vxorps (%r10), %ymm0, %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x28105f vxorps (%r10), %xmm1, %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x28107f imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2810bf vmovss (%rsi,%rdi,4), %xmm2 vxorps %xmm1, %xmm2, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x2810a8 incq %rdx jmp 0x281033 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x28118f vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282afa vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28112b vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x282c36 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x281161 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2811cd vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f560 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2811af incq %r15 jmp 0x2810f0 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x1745c8(%rip), %ymm0 # 0x3f57c4 vbroadcastss 0x17071f(%rip), %ymm1 # 0x3f1924 vbroadcastss 0x1706f2(%rip), %ymm2 # 0x3f1900 vbroadcastss 0x16ffc9(%rip), %ymm3 # 0x3f11e0 vbroadcastss 0x1745a4(%rip), %xmm4 # 0x3f57c4 vbroadcastss 0x1706fb(%rip), %xmm5 # 0x3f1924 vbroadcastss 0x1706ce(%rip), %xmm6 # 0x3f1900 vbroadcastss 0x16ffa5(%rip), %xmm7 # 0x3f11e0 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x2812eb vmovups (%r10), %ymm8 vrsqrtps %ymm8, %ymm9 vmulps %ymm9, %ymm8, %ymm10 vfmadd213ps %ymm0, %ymm10, %ymm9 # ymm9 = (ymm10 * ymm9) + ymm0 vmulps %ymm1, %ymm10, %ymm10 vmulps %ymm9, %ymm10, %ymm9 vandps %ymm2, %ymm8, %ymm8 vcmpleps %ymm8, %ymm3, %ymm8 vandps %ymm9, %ymm8, %ymm8 vmovups %ymm8, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x281267 vmovaps (%r10), %xmm8 vrsqrtps %xmm8, %xmm9 vmulps %xmm9, %xmm8, %xmm10 vfmadd213ps %xmm4, %xmm10, %xmm9 # xmm9 = (xmm10 * xmm9) + xmm4 vmulps %xmm5, %xmm10, %xmm10 vmulps %xmm9, %xmm10, %xmm9 vandps %xmm6, %xmm8, %xmm8 vcmpleps %xmm8, %xmm7, %xmm8 vandps %xmm9, %xmm8, %xmm8 vmovaps %xmm8, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x2812ae imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x281311 vsqrtss (%rsi,%rdi,4), %xmm11, %xmm8 vmovss %xmm8, (%rsi,%rdi,4) incq %rdi jmp 0x2812fe incq %rdx jmp 0x28123b movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x2813e1 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282512 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28137d vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x28261c vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x2813b3 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x28141f vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f0a0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x281401 incq %r15 jmp 0x281342 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vmovss 0x174377(%rip), %xmm0 # 0x3f57c4 vmovss 0x1704cf(%rip), %xmm1 # 0x3f1924 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x2814b7 vrsqrtps (%r10), %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x281481 vrsqrtps (%r10), %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x2814a1 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2814f2 vmovss (%rsi,%rdi,4), %xmm2 vrsqrtss %xmm2, %xmm2, %xmm3 vmulss %xmm3, %xmm2, %xmm2 vfmadd213ss %xmm0, %xmm3, %xmm2 # xmm2 = (xmm3 * xmm2) + xmm0 vmulss %xmm1, %xmm3, %xmm3 vmulss %xmm2, %xmm3, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x2814ca incq %rdx jmp 0x281455 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x2815c2 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x281dac vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28155e vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x281f00 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x281594 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x281600 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f4a0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2815e2 incq %r15 jmp 0x281523 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x2816d0 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2818f4 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28166c vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x281a22 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x2816a2 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x28170e vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f200 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2816f0 incq %r15 jmp 0x281631 xorl %eax, %eax leaq -0x28(%rbp), %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/unaryop_x86_fma.cpp
ncnn::UnaryOp_x86_fma_functor::unary_op_exp::func_pack8(float vector[8] const&) const	static NCNN_FORCEINLINE __m256 exp256_ps(__m256 x) { __m256 tmp = _mm256_setzero_ps(), fx; __m256i imm0; __m256 one = (__m256)_ps256_1; x = _mm256_min_ps(x, (__m256)_ps256_exp_hi); x = _mm256_max_ps(x, (__m256)_ps256_exp_lo); /* express exp(x) as exp(g + nlog(2)) / fx = _mm256_comp_fmadd_ps(x, (__m256)_ps256_cephes_LOG2EF, (__m256)_ps256_0p5); /* how to perform a floorf with SSE: just below / //imm0 = _mm256_cvttps_epi32(fx); //tmp = _mm256_cvtepi32_ps(imm0); tmp = _mm256_floor_ps(fx); / if greater, subtract 1 / //__m256 mask = _mm256_cmpgt_ps(tmp, fx); __m256 mask = _mm256_cmp_ps(tmp, fx, _CMP_GT_OS); mask = _mm256_and_ps(mask, one); fx = _mm256_sub_ps(tmp, mask); // x = x - fx exp_C1 x = _mm256_comp_fnmadd_ps(fx, (__m256)_ps256_cephes_exp_C1, x); // x = x - fx * exp_C2 x = _mm256_comp_fnmadd_ps(fx, (__m256)_ps256_cephes_exp_C2, x); tmp = _mm256_mul_ps(x, x); __m256 y = (__m256)_ps256_cephes_exp_p0; y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p1); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p2); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p3); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p4); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p5); y = _mm256_comp_fmadd_ps(y, tmp, x); y = _mm256_add_ps(y, one); /* build 2^n / imm0 = _mm256_cvttps_epi32(fx); // another two AVX2 instructions imm0 = _mm256_comp_add_epi32(imm0, (__m256i*)_pi32_256_0x7f); imm0 = _mm256_comp_slli_epi32(imm0, 23); __m256 pow2n = _mm256_castsi256_ps(imm0); y = _mm256_mul_ps(y, pow2n); return y; }	vbroadcastss 0x16fa73(%rip), %ymm0 # 0x3f11b8 vminps (%rsi), %ymm0, %ymm0 vbroadcastss 0x16fa6a(%rip), %ymm1 # 0x3f11bc vmaxps %ymm1, %ymm0, %ymm1 vbroadcastss 0x16c8b5(%rip), %ymm0 # 0x3ee014 vbroadcastss 0x16fa58(%rip), %ymm2 # 0x3f11c0 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vroundps $0x1, %ymm2, %ymm3 vbroadcastss 0x16d50c(%rip), %ymm4 # 0x3eec88 vcmpltps %ymm3, %ymm2, %ymm2 vandps %ymm4, %ymm2, %ymm2 vbroadcastss 0x16fa36(%rip), %ymm4 # 0x3f11c4 vsubps %ymm2, %ymm3, %ymm2 vfmsub213ps %ymm1, %ymm2, %ymm4 # ymm4 = (ymm2 * ymm4) - ymm1 vbroadcastss 0x170f7c(%rip), %ymm1 # 0x3f271c vbroadcastss 0x16fa27(%rip), %ymm3 # 0x3f11d0 vbroadcastss 0x16fa1a(%rip), %ymm5 # 0x3f11cc vfmsub213ps %ymm4, %ymm2, %ymm1 # ymm1 = (ymm2 * ymm1) - ymm4 vfmadd213ps %ymm3, %ymm1, %ymm5 # ymm5 = (ymm1 * ymm5) + ymm3 vbroadcastss 0x16fa0f(%rip), %ymm3 # 0x3f11d4 vmulps %ymm1, %ymm1, %ymm4 vfmadd231ps %ymm5, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm5) + ymm3 vbroadcastss 0x16fa01(%rip), %ymm5 # 0x3f11d8 vfmadd231ps %ymm3, %ymm1, %ymm5 # ymm5 = (ymm1 * ymm3) + ymm5 vbroadcastss 0x16f9f7(%rip), %ymm3 # 0x3f11dc vfmadd231ps %ymm5, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm5) + ymm3 vfmadd213ps %ymm0, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm3) + ymm0 vfmadd213ps %ymm1, %ymm4, %ymm3 # ymm3 = (ymm4 * ymm3) + ymm1 vcvttps2dq %ymm2, %ymm0 vpslld $0x17, %xmm0, %xmm1 vextractf128 $0x1, %ymm0, %xmm0 vpslld $0x17, %xmm0, %xmm0 vbroadcastss 0x16d477(%rip), %xmm2 # 0x3eec88 vpaddd %xmm2, %xmm0, %xmm0 vpaddd %xmm2, %xmm1, %xmm1 vinsertf128 $0x1, %xmm0, %ymm1, %ymm0 vfmadd213ps %ymm0, %ymm3, %ymm0 # ymm0 = (ymm3 * ymm0) + ymm0 retq nop	/csukuangfj[P]ncnn/src/layer/x86/avx_mathfun.h
ncnn::UnaryOp_x86_avx::forward_inplace(ncnn::Mat&, ncnn::Option const&) const	int UnaryOp_x86_avx::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { using namespace UnaryOp_x86_avx_functor; if (op_type == Operation_ABS) return unary_op_inplace<unary_op_abs>(bottom_top_blob, opt); if (op_type == Operation_NEG) return unary_op_inplace<unary_op_neg>(bottom_top_blob, opt); if (op_type == Operation_FLOOR) return unary_op_inplace<unary_op_floor>(bottom_top_blob, opt); if (op_type == Operation_CEIL) return unary_op_inplace<unary_op_ceil>(bottom_top_blob, opt); if (op_type == Operation_SQUARE) return unary_op_inplace<unary_op_square>(bottom_top_blob, opt); if (op_type == Operation_SQRT) return unary_op_inplace<unary_op_sqrt>(bottom_top_blob, opt); if (op_type == Operation_RSQRT) return unary_op_inplace<unary_op_rsqrt>(bottom_top_blob, opt); if (op_type == Operation_EXP) return unary_op_inplace<unary_op_exp>(bottom_top_blob, opt); if (op_type == Operation_LOG) return unary_op_inplace<unary_op_log>(bottom_top_blob, opt); if (op_type == Operation_SIN) return unary_op_inplace<unary_op_sin>(bottom_top_blob, opt); if (op_type == Operation_COS) return unary_op_inplace<unary_op_cos>(bottom_top_blob, opt); if (op_type == Operation_TAN) return unary_op_inplace<unary_op_tan>(bottom_top_blob, opt); if (op_type == Operation_ASIN) return unary_op_inplace<unary_op_asin>(bottom_top_blob, opt); if (op_type == Operation_ACOS) return unary_op_inplace<unary_op_acos>(bottom_top_blob, opt); if (op_type == Operation_ATAN) return unary_op_inplace<unary_op_atan>(bottom_top_blob, opt); if (op_type == Operation_RECIPROCAL) return unary_op_inplace<unary_op_reciprocal>(bottom_top_blob, opt); if (op_type == Operation_TANH) return unary_op_inplace<unary_op_tanh>(bottom_top_blob, opt); if (op_type == Operation_LOG10) return unary_op_inplace<unary_op_log10>(bottom_top_blob, opt); return 0; }	pushq %rbp movq %rsp, %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx andq $-0x20, %rsp subq $0x80, %rsp movq %rsi, 0x18(%rsp) movq (%rdi), %rax movq -0x18(%rax), %rax movl 0xd0(%rdi,%rax), %eax cmpq $0x11, %rax ja 0x283f35 leaq 0x1755e4(%rip), %rcx # 0x3f83c8 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq *%rax movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16eaec(%rip), %ymm0 # 0x3f1900 vbroadcastss 0x16eae3(%rip), %xmm1 # 0x3f1900 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x282e7f vandps (%r10), %ymm0, %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x282e49 vandps (%r10), %xmm1, %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x282e69 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x282ea9 vmovss (%rsi,%rdi,4), %xmm2 vandps %xmm1, %xmm2, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x282e92 incq %rdx jmp 0x282e1d movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x282f79 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2851c4 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x282f15 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2852fa vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x282f4b movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x282fb7 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f160 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x282f99 incq %r15 jmp 0x282eda movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283087 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x284fc0 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283023 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2850c2 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283059 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2830c5 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f540 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2830a7 incq %r15 jmp 0x282fe8 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283195 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x28484e vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283131 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284a08 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283167 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2831d3 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f4c0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2831b5 incq %r15 jmp 0x2830f6 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283263 vmovups (%r10), %ymm0 vmulps %ymm0, %ymm0, %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x283225 vmovaps (%r10), %xmm0 vmulps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x283249 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x28328d vmovss (%rsi,%rdi,4), %xmm0 vmulss %xmm0, %xmm0, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x283276 incq %rdx jmp 0x2831f9 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x28335d vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x284b6c vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x2832f9 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284c78 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x28332f movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x28339b vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f240 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x28337d incq %r15 jmp 0x2832be movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x28346b vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2843a2 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283407 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2844f0 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x28343d movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2834a9 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f3b0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x28348b incq %r15 jmp 0x2833cc movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283533 vroundps $0x1, (%r10), %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x2834fb vroundps $0x1, (%r10), %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x28351c imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x28355b vroundss $0x9, (%rsi,%rdi,4), %xmm1, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x283546 incq %rdx jmp 0x2834cf movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16b6fe(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x16b6f5(%rip), %xmm1 # 0x3eec88 vmovss 0x16b6ed(%rip), %xmm2 # 0x3eec88 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283625 vmovups (%r10), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm3, %ymm3 vsubps %ymm3, %ymm0, %ymm3 vmulps %ymm3, %ymm4, %ymm3 vaddps %ymm3, %ymm4, %ymm3 vmovups %ymm3, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x2835c7 vmovaps (%r10), %xmm3 vrcpps %xmm3, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vsubps %xmm3, %xmm1, %xmm3 vmulps %xmm3, %xmm4, %xmm3 vaddps %xmm3, %xmm4, %xmm3 vmovaps %xmm3, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x2835fb imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x28364b vdivss (%rsi,%rdi,4), %xmm2, %xmm3 vmovss %xmm3, (%rsi,%rdi,4) incq %rdi jmp 0x283638 incq %rdx jmp 0x28359b movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x2836d5 vroundps $0x2, (%r10), %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x28369d vroundps $0x2, (%r10), %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x2836be imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2836fd vroundss $0xa, (%rsi,%rdi,4), %xmm1, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x2836e8 incq %rdx jmp 0x283671 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x2837cd vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x283f5c vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283769 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284052 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x28379f movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x28380b vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f410 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2837ed incq %r15 jmp 0x28372e movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16d97a(%rip), %ymm0 # 0x3f11b4 vbroadcastss 0x16d971(%rip), %xmm1 # 0x3f11b4 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x2838a5 vxorps (%r10), %ymm0, %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x28386f vxorps (%r10), %xmm1, %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x28388f imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2838cf vmovss (%rsi,%rdi,4), %xmm2 vxorps %xmm1, %xmm2, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x2838b8 incq %rdx jmp 0x283843 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x28399f vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x285414 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28393b vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x285564 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283971 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2839dd vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f560 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2839bf incq %r15 jmp 0x283900 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16df18(%rip), %ymm0 # 0x3f1924 vbroadcastss 0x171daf(%rip), %ymm1 # 0x3f57c4 vbroadcastss 0x16dee2(%rip), %ymm2 # 0x3f1900 vbroadcastss 0x16d7b9(%rip), %ymm3 # 0x3f11e0 vbroadcastss 0x16def4(%rip), %xmm4 # 0x3f1924 vbroadcastss 0x171d8b(%rip), %xmm5 # 0x3f57c4 vbroadcastss 0x16debe(%rip), %xmm6 # 0x3f1900 vbroadcastss 0x16d795(%rip), %xmm7 # 0x3f11e0 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283b07 vmovups (%r10), %ymm8 vrsqrtps %ymm8, %ymm9 vmulps %ymm9, %ymm8, %ymm10 vmulps %ymm0, %ymm10, %ymm11 vmulps %ymm9, %ymm10, %ymm9 vaddps %ymm1, %ymm9, %ymm9 vmulps %ymm9, %ymm11, %ymm9 vandps %ymm2, %ymm8, %ymm8 vcmpleps %ymm8, %ymm3, %ymm8 vandps %ymm9, %ymm8, %ymm8 vmovups %ymm8, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x283a77 vmovaps (%r10), %xmm8 vrsqrtps %xmm8, %xmm9 vmulps %xmm9, %xmm8, %xmm10 vmulps %xmm4, %xmm10, %xmm11 vmulps %xmm9, %xmm10, %xmm9 vaddps %xmm5, %xmm9, %xmm9 vmulps %xmm9, %xmm11, %xmm9 vandps %xmm6, %xmm8, %xmm8 vcmpleps %xmm8, %xmm7, %xmm8 vandps %xmm9, %xmm8, %xmm8 vmovaps %xmm8, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x283ac6 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x283b2d vsqrtss (%rsi,%rdi,4), %xmm12, %xmm8 vmovss %xmm8, (%rsi,%rdi,4) incq %rdi jmp 0x283b1a incq %rdx jmp 0x283a4b movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283bfd vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x284d84 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283b99 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284ea2 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283bcf movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x283c3b vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f0a0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x283c1d incq %r15 jmp 0x283b5e movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vmovss 0x171b5b(%rip), %xmm0 # 0x3f57c4 vmovss 0x16dcb3(%rip), %xmm1 # 0x3f1924 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283cd3 vrsqrtps (%r10), %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x283c9d vrsqrtps (%r10), %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x283cbd imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x283d11 vmovss (%rsi,%rdi,4), %xmm2 vrsqrtss %xmm2, %xmm2, %xmm3 vmulss %xmm3, %xmm2, %xmm2 vmulss %xmm3, %xmm2, %xmm2 vaddss %xmm0, %xmm2, %xmm2 vmulss %xmm1, %xmm3, %xmm3 vmulss %xmm2, %xmm3, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x283ce6 incq %rdx jmp 0x283c71 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283de1 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2845fa vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283d7d vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284744 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283db3 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x283e1f vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f4a0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x283e01 incq %r15 jmp 0x283d42 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283eef vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x28412c vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283e8b vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284270 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283ec1 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x283f2d vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f200 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x283f0f incq %r15 jmp 0x283e50 xorl %eax, %eax leaq -0x28(%rbp), %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/unaryop_x86_avx.cpp
ncnn::UnaryOp_x86_avx_functor::unary_op_exp::func_pack8(float vector[8] const&) const	static NCNN_FORCEINLINE __m256 exp256_ps(__m256 x) { __m256 tmp = _mm256_setzero_ps(), fx; __m256i imm0; __m256 one = (__m256)_ps256_1; x = _mm256_min_ps(x, (__m256)_ps256_exp_hi); x = _mm256_max_ps(x, (__m256)_ps256_exp_lo); /* express exp(x) as exp(g + nlog(2)) / fx = _mm256_comp_fmadd_ps(x, (__m256)_ps256_cephes_LOG2EF, (__m256)_ps256_0p5); /* how to perform a floorf with SSE: just below / //imm0 = _mm256_cvttps_epi32(fx); //tmp = _mm256_cvtepi32_ps(imm0); tmp = _mm256_floor_ps(fx); / if greater, subtract 1 / //__m256 mask = _mm256_cmpgt_ps(tmp, fx); __m256 mask = _mm256_cmp_ps(tmp, fx, _CMP_GT_OS); mask = _mm256_and_ps(mask, one); fx = _mm256_sub_ps(tmp, mask); // x = x - fx exp_C1 x = _mm256_comp_fnmadd_ps(fx, (__m256)_ps256_cephes_exp_C1, x); // x = x - fx * exp_C2 x = _mm256_comp_fnmadd_ps(fx, (__m256)_ps256_cephes_exp_C2, x); tmp = _mm256_mul_ps(x, x); __m256 y = (__m256)_ps256_cephes_exp_p0; y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p1); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p2); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p3); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p4); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_exp_p5); y = _mm256_comp_fmadd_ps(y, tmp, x); y = _mm256_add_ps(y, one); /* build 2^n / imm0 = _mm256_cvttps_epi32(fx); // another two AVX2 instructions imm0 = _mm256_comp_add_epi32(imm0, (__m256i*)_pi32_256_0x7f); imm0 = _mm256_comp_slli_epi32(imm0, 23); __m256 pow2n = _mm256_castsi256_ps(imm0); y = _mm256_mul_ps(y, pow2n); return y; }	vbroadcastss 0x16d253(%rip), %ymm0 # 0x3f11b8 vminps (%rsi), %ymm0, %ymm0 vbroadcastss 0x16d24a(%rip), %ymm1 # 0x3f11bc vmaxps %ymm1, %ymm0, %ymm0 vbroadcastss 0x16d241(%rip), %ymm1 # 0x3f11c0 vmulps %ymm1, %ymm0, %ymm1 vbroadcastss 0x16a088(%rip), %ymm2 # 0x3ee014 vaddps %ymm2, %ymm1, %ymm1 vroundps $0x1, %ymm1, %ymm3 vcmpltps %ymm3, %ymm1, %ymm1 vbroadcastss 0x16ace4(%rip), %ymm4 # 0x3eec88 vandps %ymm4, %ymm1, %ymm1 vsubps %ymm1, %ymm3, %ymm1 vbroadcastss 0x16ea2b(%rip), %ymm3 # 0x3f29e0 vmulps %ymm3, %ymm1, %ymm3 vaddps %ymm3, %ymm0, %ymm0 vbroadcastss 0x16d206(%rip), %ymm3 # 0x3f11cc vmulps %ymm0, %ymm0, %ymm5 vmulps %ymm3, %ymm0, %ymm3 vbroadcastss 0x16d1f9(%rip), %ymm6 # 0x3f11d0 vaddps %ymm6, %ymm3, %ymm3 vmulps %ymm0, %ymm3, %ymm3 vbroadcastss 0x16d1ec(%rip), %ymm6 # 0x3f11d4 vaddps %ymm6, %ymm3, %ymm3 vmulps %ymm0, %ymm3, %ymm3 vbroadcastss 0x16d1df(%rip), %ymm6 # 0x3f11d8 vaddps %ymm6, %ymm3, %ymm3 vmulps %ymm0, %ymm3, %ymm3 vbroadcastss 0x16d1d2(%rip), %ymm6 # 0x3f11dc vaddps %ymm6, %ymm3, %ymm3 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm2, %ymm3, %ymm2 vmulps %ymm2, %ymm5, %ymm2 vaddps %ymm4, %ymm0, %ymm0 vaddps %ymm2, %ymm0, %ymm0 vcvttps2dq %ymm1, %ymm1 vpslld $0x17, %xmm1, %xmm2 vextractf128 $0x1, %ymm1, %xmm1 vpslld $0x17, %xmm1, %xmm1 vbroadcastss 0x16ac49(%rip), %xmm3 # 0x3eec88 vpaddd %xmm3, %xmm1, %xmm1 vpaddd %xmm3, %xmm2, %xmm2 vinsertf128 $0x1, %xmm1, %ymm2, %ymm1 vmulps %ymm1, %ymm0, %ymm0 retq	/csukuangfj[P]ncnn/src/layer/x86/avx_mathfun.h
ncnn::UnaryOp_x86_avx_functor::unary_op_cos::func_pack8(float vector[8] const&) const	static NCNN_FORCEINLINE __m256 cos256_ps(__m256 x) { // any x __m256 xmm1, xmm2 = _mm256_setzero_ps(), xmm3, y; __m256i imm0, imm2; #ifndef __AVX2__ __m128i imm0_1, imm0_2; __m128i imm2_1, imm2_2; #endif /* take the absolute value / x = _mm256_and_ps(x, (__m256)_ps256_inv_sign_mask); / scale by 4/Pi / y = _mm256_mul_ps(x, (__m256)_ps256_cephes_FOPI); #ifdef __AVX2__ / store the integer part of y in mm0 / imm2 = _mm256_cvttps_epi32(y); / j=(j+1) & (~1) (see the cephes sources) / imm2 = _mm256_comp_add_epi32(imm2, (__m256i)_pi32_256_1); imm2 = _mm256_and_si256(imm2, (__m256i)_pi32_256_inv1); y = _mm256_cvtepi32_ps(imm2); imm2 = _mm256_comp_sub_epi32(imm2, (__m256i)_pi32_256_2); / get the swap sign flag / imm0 = _mm256_andnot_si256(imm2, (__m256i)_pi32_256_4); imm0 = _mm256_comp_slli_epi32(imm0, 29); / get the polynom selection mask / imm2 = _mm256_and_si256(imm2, (__m256i)_pi32_256_2); imm2 = _mm256_cmpeq_epi32(imm2, (__m256i)_pi32_256_0); #else / we use SSE2 routines to perform the integer ops / COPY_IMM_TO_XMM(_mm256_cvttps_epi32(y), imm2_1, imm2_2); imm2_1 = _mm_add_epi32(imm2_1, (__m128i)_pi32avx_1); imm2_2 = _mm_add_epi32(imm2_2, (__m128i)_pi32avx_1); imm2_1 = _mm_and_si128(imm2_1, (__m128i)_pi32avx_inv1); imm2_2 = _mm_and_si128(imm2_2, (__m128i)_pi32avx_inv1); COPY_XMM_TO_IMM(imm2_1, imm2_2, imm2); y = _mm256_cvtepi32_ps(imm2); imm2_1 = _mm_sub_epi32(imm2_1, (__m128i)_pi32avx_2); imm2_2 = _mm_sub_epi32(imm2_2, (__m128i)_pi32avx_2); imm0_1 = _mm_andnot_si128(imm2_1, (__m128i)_pi32avx_4); imm0_2 = _mm_andnot_si128(imm2_2, (__m128i)_pi32avx_4); imm0_1 = _mm_slli_epi32(imm0_1, 29); imm0_2 = _mm_slli_epi32(imm0_2, 29); COPY_XMM_TO_IMM(imm0_1, imm0_2, imm0); imm2_1 = _mm_and_si128(imm2_1, (__m128i)_pi32avx_2); imm2_2 = _mm_and_si128(imm2_2, (__m128i)_pi32avx_2); imm2_1 = _mm_cmpeq_epi32(imm2_1, _mm_setzero_si128()); imm2_2 = _mm_cmpeq_epi32(imm2_2, _mm_setzero_si128()); COPY_XMM_TO_IMM(imm2_1, imm2_2, imm2); #endif __m256 sign_bit = _mm256_castsi256_ps(imm0); __m256 poly_mask = _mm256_castsi256_ps(imm2); / The magic pass: "Extended precision modular arithmetic" x = ((x - y * DP1) - y * DP2) - y * DP3; / xmm1 = (__m256)_ps256_minus_cephes_DP1; xmm2 = (__m256)_ps256_minus_cephes_DP2; xmm3 = (__m256)_ps256_minus_cephes_DP3; x = _mm256_comp_fmadd_ps(y, xmm1, x); x = _mm256_comp_fmadd_ps(y, xmm2, x); x = _mm256_comp_fmadd_ps(y, xmm3, x); / Evaluate the first polynom (0 <= x <= Pi/4) / y = (__m256)_ps256_coscof_p0; __m256 z = _mm256_mul_ps(x, x); y = _mm256_comp_fmadd_ps(y, z, (__m256)_ps256_coscof_p1); y = _mm256_comp_fmadd_ps(y, z, (__m256)_ps256_coscof_p2); y = _mm256_mul_ps(y, z); y = _mm256_mul_ps(y, z); // y = y - z 0.5 y = _mm256_comp_fnmadd_ps(z, (__m256)_ps256_0p5, y); y = _mm256_add_ps(y, (__m256)_ps256_1); /* Evaluate the second polynom (Pi/4 <= x <= 0) / __m256 y2 = (__m256)_ps256_sincof_p0; y2 = _mm256_mul_ps(y2, z); y2 = _mm256_add_ps(y2, (__m256)_ps256_sincof_p1); y2 = _mm256_mul_ps(y2, z); y2 = _mm256_add_ps(y2, (__m256)_ps256_sincof_p2); y2 = _mm256_mul_ps(y2, z); y2 = _mm256_mul_ps(y2, x); y2 = _mm256_add_ps(y2, x); / select the correct result from the two polynoms / xmm3 = poly_mask; y2 = _mm256_and_ps(xmm3, y2); //, xmm3); y = _mm256_andnot_ps(xmm3, y); y = _mm256_add_ps(y, y2); / update the sign */ y = _mm256_xor_ps(y, sign_bit); return y; }	vbroadcastss 0x16d2fd(%rip), %ymm0 # 0x3f1900 vandps (%rsi), %ymm0, %ymm2 vbroadcastss 0x173240(%rip), %ymm0 # 0x3f7850 vmulps %ymm0, %ymm2, %ymm0 vcvttps2dq %ymm0, %ymm0 vpcmpeqd %xmm1, %xmm1, %xmm1 vpsubd %xmm1, %xmm0, %xmm3 vextractf128 $0x1, %ymm0, %xmm0 vpsubd %xmm1, %xmm0, %xmm0 vmovddup 0x173d7e(%rip), %xmm1 # xmm1 = mem[0,0] vpand %xmm1, %xmm3, %xmm3 vpand %xmm1, %xmm0, %xmm0 vinsertf128 $0x1, %xmm0, %ymm3, %ymm1 vcvtdq2ps %ymm1, %ymm4 vbroadcastss 0x173207(%rip), %xmm1 # 0x3f7854 vpaddd %xmm1, %xmm3, %xmm3 vpaddd %xmm1, %xmm0, %xmm5 vpslld $0x1d, %xmm3, %xmm0 vmovddup 0x173d56(%rip), %xmm1 # xmm1 = mem[0,0] vpandn %xmm1, %xmm0, %xmm0 vpslld $0x1d, %xmm5, %xmm6 vpandn %xmm1, %xmm6, %xmm1 vinsertf128 $0x1, %xmm1, %ymm0, %ymm0 vbroadcastss 0x1731da(%rip), %xmm6 # 0x3f7858 vpand %xmm6, %xmm3, %xmm1 vbroadcastss 0x173d35(%rip), %ymm3 # 0x3f83c0 vpand %xmm6, %xmm5, %xmm5 vmulps %ymm3, %ymm4, %ymm3 vaddps %ymm2, %ymm3, %ymm2 vmulps %ymm2, %ymm2, %ymm3 vbroadcastss 0x1731c4(%rip), %ymm4 # 0x3f7868 vbroadcastss 0x1731bf(%rip), %ymm6 # 0x3f786c vmulps %ymm4, %ymm3, %ymm4 vaddps %ymm6, %ymm4, %ymm4 vbroadcastss 0x1731b2(%rip), %ymm6 # 0x3f7870 vmulps %ymm3, %ymm4, %ymm4 vaddps %ymm6, %ymm4, %ymm4 vbroadcastss 0x16d255(%rip), %ymm6 # 0x3f1924 vmulps %ymm3, %ymm4, %ymm4 vaddps %ymm6, %ymm4, %ymm4 vbroadcastss 0x16a5a8(%rip), %ymm6 # 0x3eec88 vmulps %ymm3, %ymm4, %ymm4 vaddps %ymm6, %ymm4, %ymm4 vbroadcastss 0x173183(%rip), %ymm6 # 0x3f7874 vmulps %ymm6, %ymm3, %ymm6 vbroadcastss 0x17317a(%rip), %ymm7 # 0x3f7878 vaddps %ymm7, %ymm6, %ymm6 vmulps %ymm3, %ymm6, %ymm6 vbroadcastss 0x17316d(%rip), %ymm7 # 0x3f787c vaddps %ymm7, %ymm6, %ymm6 vmulps %ymm2, %ymm3, %ymm3 vmulps %ymm6, %ymm3, %ymm3 vaddps %ymm2, %ymm3, %ymm2 vinsertf128 $0x1, %xmm5, %ymm1, %ymm1 vcvtdq2ps %ymm1, %ymm1 vxorps %xmm3, %xmm3, %xmm3 vcmpeqps %ymm3, %ymm1, %ymm1 vandps %ymm2, %ymm1, %ymm2 vandnps %ymm4, %ymm1, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vxorps %ymm1, %ymm0, %ymm0 retq nop	/csukuangfj[P]ncnn/src/layer/x86/avx_mathfun.h
ncnn::UnaryOp_x86_avx_functor::unary_op_log10::func_pack8(float vector[8] const&) const	static NCNN_FORCEINLINE __m256 log256_ps(__m256 x) { __m256i imm0; __m256 one = (__m256)_ps256_1; //__m256 invalid_mask = _mm256_cmple_ps(x, _mm256_setzero_ps()); __m256 invalid_mask = _mm256_cmp_ps(x, _mm256_setzero_ps(), _CMP_LE_OS); x = _mm256_max_ps(x, (__m256)_ps256_min_norm_pos); /* cut off denormalized stuff / // can be done with AVX2 imm0 = _mm256_comp_srli_epi32(_mm256_castps_si256(x), 23); / keep only the fractional part / x = _mm256_and_ps(x, (__m256)_ps256_inv_mant_mask); x = _mm256_or_ps(x, (__m256)_ps256_0p5); // this is again another AVX2 instruction imm0 = _mm256_comp_sub_epi32(imm0, (__m256i)_pi32_256_0x7f); __m256 e = _mm256_cvtepi32_ps(imm0); e = _mm256_add_ps(e, one); / part2: if( x < SQRTHF ) { e -= 1; x = x + x - 1.0; } else { x = x - 1.0; } / //__m256 mask = _mm256_cmplt_ps(x, (__m256)_ps256_cephes_SQRTHF); __m256 mask = _mm256_cmp_ps(x, (__m256)_ps256_cephes_SQRTHF, _CMP_LT_OS); __m256 tmp = _mm256_and_ps(x, mask); x = _mm256_sub_ps(x, one); e = _mm256_sub_ps(e, _mm256_and_ps(one, mask)); x = _mm256_add_ps(x, tmp); __m256 z = _mm256_mul_ps(x, x); __m256 y = (__m256)_ps256_cephes_log_p0; y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_log_p1); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_log_p2); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_log_p3); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_log_p4); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_log_p5); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_log_p6); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_log_p7); y = _mm256_comp_fmadd_ps(y, x, (__m256)_ps256_cephes_log_p8); y = _mm256_mul_ps(y, x); y = _mm256_mul_ps(y, z); y = _mm256_comp_fmadd_ps(e, (__m256)_ps256_cephes_log_q1, y); //y = -z 0.5 + y y = _mm256_comp_fnmadd_ps(z, (__m256)_ps256_0p5, y); x = _mm256_add_ps(x, y); x = _mm256_comp_fmadd_ps(e, (__m256)_ps256_cephes_log_q2, x); y = _mm256_or_ps(x, invalid_mask); // negative arg will be NAN return y; }	vbroadcastss 0x16bdc3(%rip), %ymm0 # 0x3f11e0 vmaxps (%rsi), %ymm0, %ymm0 vpsrld $0x17, %xmm0, %xmm1 vextractf128 $0x1, %ymm0, %xmm2 vpsrld $0x17, %xmm2, %xmm2 vbroadcastss 0x16bdaa(%rip), %ymm3 # 0x3f11e4 vandps %ymm3, %ymm0, %ymm0 vbroadcastss 0x168bcd(%rip), %ymm3 # 0x3ee014 vorps %ymm3, %ymm0, %ymm0 vbroadcastss 0x16c4b0(%rip), %xmm3 # 0x3f1904 vbroadcastss 0x16bd8f(%rip), %ymm4 # 0x3f11ec vcmpleps %ymm0, %ymm4, %ymm4 vandnps %ymm0, %ymm4, %ymm5 vbroadcastss 0x16bd81(%rip), %ymm6 # 0x3f11f0 vaddps %ymm6, %ymm0, %ymm0 vaddps %ymm5, %ymm0, %ymm0 vextractf128 $0x1, %ymm4, %xmm5 vpsubd %xmm5, %xmm2, %xmm2 vpaddd %xmm3, %xmm2, %xmm2 vpsubd %xmm4, %xmm1, %xmm1 vpaddd %xmm3, %xmm1, %xmm1 vinsertf128 $0x1, %xmm2, %ymm1, %ymm1 vcvtdq2ps %ymm1, %ymm2 vmulps %ymm0, %ymm0, %ymm1 vbroadcastss 0x16bd50(%rip), %ymm3 # 0x3f11f4 vbroadcastss 0x16bd4b(%rip), %ymm4 # 0x3f11f8 vmulps %ymm3, %ymm0, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd3e(%rip), %ymm4 # 0x3f11fc vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd31(%rip), %ymm4 # 0x3f1200 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd24(%rip), %ymm4 # 0x3f1204 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd17(%rip), %ymm4 # 0x3f1208 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd0a(%rip), %ymm4 # 0x3f120c vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bcfd(%rip), %ymm4 # 0x3f1210 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bcf0(%rip), %ymm4 # 0x3f1214 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16c3eb(%rip), %ymm4 # 0x3f1920 vmulps %ymm0, %ymm3, %ymm3 vmulps %ymm4, %ymm2, %ymm2 vbroadcastss 0x16c3de(%rip), %ymm4 # 0x3f1924 vaddps %ymm4, %ymm3, %ymm3 vmulps %ymm3, %ymm1, %ymm1 vaddps %ymm0, %ymm2, %ymm0 vaddps %ymm1, %ymm0, %ymm0 vbroadcastss 0x172e65(%rip), %ymm1 # 0x3f83c4 vmulps %ymm1, %ymm0, %ymm0 retq	/csukuangfj[P]ncnn/src/layer/x86/avx_mathfun.h
ncnn::ConvolutionDepthWise::load_model(ncnn::ModelBin const&)	int ConvolutionDepthWise::load_model(const ModelBin& mb) { if (dynamic_weight) return 0; weight_data = mb.load(weight_data_size, 0); if (weight_data.empty()) return -100; if (bias_term) { bias_data = mb.load(num_output, 1); if (bias_data.empty()) return -100; } #if NCNN_INT8 if (int8_scale_term == 1 \|\| int8_scale_term == 101) { weight_data_int8_scales = mb.load(group, 1); bottom_blob_int8_scales = mb.load(1, 1); float bottom_blob_int8_scale = bottom_blob_int8_scales[0]; bottom_blob_int8_scales = Mat(group); bottom_blob_int8_scales.fill(bottom_blob_int8_scale); } else if (int8_scale_term == 2 \|\| int8_scale_term == 102) { weight_data_int8_scales = mb.load(1, 1); bottom_blob_int8_scales = mb.load(1, 1); // extend group if only one provided float weight_data_int8_scale = weight_data_int8_scales[0]; weight_data_int8_scales = Mat(group); weight_data_int8_scales.fill(weight_data_int8_scale); float bottom_blob_int8_scale = bottom_blob_int8_scales[0]; bottom_blob_int8_scales = Mat(group); bottom_blob_int8_scales.fill(bottom_blob_int8_scale); } if (int8_scale_term > 100) { top_blob_int8_scales = mb.load(1, 1); float top_blob_int8_scale = top_blob_int8_scales[0]; top_blob_int8_scales = Mat(group); top_blob_int8_scales.fill(top_blob_int8_scale); } #endif // NCNN_INT8 return 0; }	pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x60, %rsp cmpl $0x0, 0x160(%rdi) je 0x285b4c xorl %r12d, %r12d movl %r12d, %eax addq $0x60, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 retq movq %rsi, %r15 movq %rdi, %rbx movl 0x104(%rdi), %edx movq (%rsi), %rax leaq 0x10(%rsp), %r14 movq %r14, %rdi xorl %ecx, %ecx callq 0x10(%rax) leaq 0x168(%rbx), %r13 movq 0x8(%r14), %rax cmpq %r14, %r13 je 0x285c14 testq %rax, %rax je 0x285b84 lock incl (%rax) movq 0x170(%rbx), %rax testq %rax, %rax je 0x285bb8 lock decl (%rax) jne 0x285bb8 movq 0x168(%rbx), %rsi movq 0x188(%rbx), %rdi testq %rdi, %rdi je 0x285bb0 movq (%rdi), %rax callq 0x18(%rax) jmp 0x285bb8 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x168(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x170(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x178(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x180(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x188(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x190(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x1a0(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x1a8(%rbx) testq %rax, %rax je 0x285c3d lock decl (%rax) jne 0x285c3d movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x285c35 movq (%rdi), %rax callq 0x18(%rax) jmp 0x285c3d movq %rsi, %rdi callq 0x5f3e0 pushq $-0x64 popq %r12 cmpq $0x0, (%r13) je 0x285b3b movslq 0x1a0(%rbx), %rax imulq 0x1a8(%rbx), %rax testq %rax, %rax je 0x285b3b cmpl $0x0, 0x100(%rbx) je 0x285d7f movl 0xd0(%rbx), %edx movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi callq 0x10(%rax) leaq 0x1b0(%rbx), %r13 movq 0x18(%rsp), %rax cmpq %r14, %r13 je 0x285d33 testq %rax, %rax je 0x285ca3 lock incl (%rax) movq 0x1b8(%rbx), %rax testq %rax, %rax je 0x285cd7 lock decl (%rax) jne 0x285cd7 movq 0x1b0(%rbx), %rsi movq 0x1d0(%rbx), %rdi testq %rdi, %rdi je 0x285ccf movq (%rdi), %rax callq 0x18(%rax) jmp 0x285cd7 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x1b0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x1b8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x1c0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x1c8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x1d0(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x1d8(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x1e8(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x1f0(%rbx) testq %rax, %rax je 0x285d5c lock decl (%rax) jne 0x285d5c movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x285d54 movq (%rdi), %rax callq 0x18(%rax) jmp 0x285d5c movq %rsi, %rdi callq 0x5f3e0 cmpq $0x0, (%r13) je 0x285b3b movslq 0x1e8(%rbx), %rax imulq 0x1f0(%rbx), %rax testq %rax, %rax je 0x285b3b movl 0x10c(%rbx), %eax cmpl $0x1, %eax je 0x285d9d cmpl $0x2, %eax je 0x285e0a cmpl $0x66, %eax je 0x285e0a cmpl $0x65, %eax jne 0x286582 movl 0x108(%rbx), %edx movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi callq 0x10(%rax) leaq 0x1f8(%rbx), %rcx movq 0x18(%rsp), %rax cmpq %r14, %rcx je 0x285fbe testq %rax, %rax je 0x285dcf lock incl (%rax) movq 0x200(%rbx), %rax testq %rax, %rax je 0x285f62 lock decl (%rax) jne 0x285f62 movq 0x1f8(%rbx), %rsi movq 0x218(%rbx), %rdi testq %rdi, %rdi je 0x285f5a movq (%rdi), %rax callq 0x18(%rax) jmp 0x285f62 movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi movl %ecx, %edx callq 0x10(%rax) leaq 0x1f8(%rbx), %r13 movq 0x18(%rsp), %rax cmpq %r14, %r13 je 0x285ec8 testq %rax, %rax je 0x285e38 lock incl (%rax) movq 0x200(%rbx), %rax testq %rax, %rax je 0x285e6c lock decl (%rax) jne 0x285e6c movq 0x1f8(%rbx), %rsi movq 0x218(%rbx), %rdi testq %rdi, %rdi je 0x285e64 movq (%rdi), %rax callq 0x18(%rax) jmp 0x285e6c movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x1f8(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x200(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x208(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x210(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x218(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x220(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x230(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x238(%rbx) testq %rax, %rax je 0x285ef1 lock decl (%rax) jne 0x285ef1 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x285ee9 movq (%rdi), %rax callq 0x18(%rax) jmp 0x285ef1 movq %rsi, %rdi callq 0x5f3e0 movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi movl %ecx, %edx callq 0x10(%rax) leaq 0x240(%rbx), %r12 movq 0x18(%rsp), %rax cmpq %r14, %r12 je 0x2860b4 testq %rax, %rax je 0x285f1f lock incl (%rax) movq 0x248(%rbx), %rax testq %rax, %rax je 0x286058 lock decl (%rax) jne 0x286058 movq 0x240(%rbx), %rsi movq 0x260(%rbx), %rdi testq %rdi, %rdi je 0x286050 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286058 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x1f8(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x200(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x208(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x210(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x218(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x220(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x230(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x238(%rbx) testq %rax, %rax je 0x285fe7 lock decl (%rax) jne 0x285fe7 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x285fdf movq (%rdi), %rax callq 0x18(%rax) jmp 0x285fe7 movq %rsi, %rdi callq 0x5f3e0 movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi movl %ecx, %edx callq 0x10(%rax) leaq 0x240(%rbx), %r12 movq 0x18(%rsp), %rax cmpq %r14, %r12 je 0x2861d3 testq %rax, %rax je 0x286015 lock incl (%rax) movq 0x248(%rbx), %rax testq %rax, %rax je 0x286177 lock decl (%rax) jne 0x286177 movq 0x240(%rbx), %rsi movq 0x260(%rbx), %rdi testq %rdi, %rdi je 0x28616f movq (%rdi), %rax callq 0x18(%rax) jmp 0x286177 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x240(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x248(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x250(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x258(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x260(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x268(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x278(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x280(%rbx) testq %rax, %rax je 0x2860dd lock decl (%rax) jne 0x2860dd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2860d5 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2860dd movq %rsi, %rdi callq 0x5f3e0 movq 0x1f8(%rbx), %rax movss (%rax), %xmm0 movss %xmm0, 0xc(%rsp) movl 0x108(%rbx), %esi andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) movaps %xmm0, 0x30(%rsp) movups %xmm0, 0x3c(%rsp) pushq $0x4 popq %rdx movq %r14, %rdi xorl %ecx, %ecx callq 0x635fa movq 0x18(%rsp), %rax cmpq %r14, %r13 je 0x2862f2 testq %rax, %rax je 0x286134 lock incl (%rax) movq 0x200(%rbx), %rax testq %rax, %rax je 0x286296 lock decl (%rax) jne 0x286296 movq 0x1f8(%rbx), %rsi movq 0x218(%rbx), %rdi testq %rdi, %rdi je 0x28628e movq (%rdi), %rax callq 0x18(%rax) jmp 0x286296 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x240(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x248(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x250(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x258(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x260(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x268(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x278(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x280(%rbx) testq %rax, %rax je 0x2861fc lock decl (%rax) jne 0x2861fc movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2861f4 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2861fc movq %rsi, %rdi callq 0x5f3e0 movq 0x240(%rbx), %rax movss (%rax), %xmm0 movss %xmm0, 0xc(%rsp) movl 0x108(%rbx), %esi andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) movaps %xmm0, 0x30(%rsp) movups %xmm0, 0x3c(%rsp) pushq $0x4 popq %rdx movq %r14, %rdi xorl %ecx, %ecx callq 0x635fa movq 0x18(%rsp), %rax cmpq %r14, %r12 je 0x28644c testq %rax, %rax je 0x286253 lock incl (%rax) movq 0x248(%rbx), %rax testq %rax, %rax je 0x2863f0 lock decl (%rax) jne 0x2863f0 movq 0x240(%rbx), %rsi movq 0x260(%rbx), %rdi testq %rdi, %rdi je 0x2863e8 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2863f0 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x1f8(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x200(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x208(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x210(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x218(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x220(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x230(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x238(%rbx) testq %rax, %rax movss 0xc(%rsp), %xmm0 je 0x28632d lock decl (%rax) jne 0x28632d movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x28631f movq (%rdi), %rax callq 0x18(%rax) movss 0xc(%rsp), %xmm0 jmp 0x28632d movq %rsi, %rdi callq 0x5f3e0 movss 0xc(%rsp), %xmm0 movl 0x230(%rbx), %eax imull 0x238(%rbx), %eax movq 0x1f8(%rbx), %rcx xorl %edx, %edx testl %eax, %eax cmovlel %edx, %eax cmpl %edx, %eax je 0x286356 movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x286348 movq 0x240(%rbx), %rax movss (%rax), %xmm0 movss %xmm0, 0xc(%rsp) movl 0x108(%rbx), %esi andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) movaps %xmm0, 0x30(%rsp) movups %xmm0, 0x3c(%rsp) pushq $0x4 popq %rdx movq %r14, %rdi xorl %ecx, %ecx callq 0x635fa movq 0x18(%rsp), %rax cmpq %r14, %r12 je 0x286518 testq %rax, %rax je 0x2863ad lock incl (%rax) movq 0x248(%rbx), %rax testq %rax, %rax je 0x2864bc lock decl (%rax) jne 0x2864bc movq 0x240(%rbx), %rsi movq 0x260(%rbx), %rdi testq %rdi, %rdi je 0x2864b4 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2864bc movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x240(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x248(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x250(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x258(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x260(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x268(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x278(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x280(%rbx) testq %rax, %rax movss 0xc(%rsp), %xmm0 je 0x286487 lock decl (%rax) jne 0x286487 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286479 movq (%rdi), %rax callq 0x18(%rax) movss 0xc(%rsp), %xmm0 jmp 0x286487 movq %rsi, %rdi callq 0x5f3e0 movss 0xc(%rsp), %xmm0 movl 0x278(%rbx), %eax imull 0x280(%rbx), %eax movq 0x240(%rbx), %rcx xorl %edx, %edx testl %eax, %eax cmovlel %edx, %eax cmpl %edx, %eax je 0x28657c movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x2864a2 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x240(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x248(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x250(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x258(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x260(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x268(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x278(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x280(%rbx) testq %rax, %rax movss 0xc(%rsp), %xmm0 je 0x286553 lock decl (%rax) jne 0x286553 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286545 movq (%rdi), %rax callq 0x18(%rax) movss 0xc(%rsp), %xmm0 jmp 0x286553 movq %rsi, %rdi callq 0x5f3e0 movss 0xc(%rsp), %xmm0 movl 0x278(%rbx), %eax imull 0x280(%rbx), %eax movq 0x240(%rbx), %rcx xorl %edx, %edx testl %eax, %eax cmovlel %edx, %eax cmpl %edx, %eax je 0x28657c movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x28656e movl 0x10c(%rbx), %eax cmpl $0x65, %eax jl 0x285b38 movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi movl %ecx, %edx callq 0x10(%rax) leaq 0x288(%rbx), %r15 movq 0x18(%rsp), %rax cmpq %r14, %r15 je 0x286649 testq %rax, %rax je 0x2865b9 lock incl (%rax) movq 0x290(%rbx), %rax testq %rax, %rax je 0x2865ed lock decl (%rax) jne 0x2865ed movq 0x288(%rbx), %rsi movq 0x2a8(%rbx), %rdi testq %rdi, %rdi je 0x2865e5 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2865ed movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x288(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x290(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x298(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x2a0(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x2a8(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x2b0(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x2c0(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x2c8(%rbx) testq %rax, %rax je 0x286672 lock decl (%rax) jne 0x286672 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x28666a movq (%rdi), %rax callq 0x18(%rax) jmp 0x286672 movq %rsi, %rdi callq 0x5f3e0 movq 0x288(%rbx), %rax movss (%rax), %xmm0 movss %xmm0, 0xc(%rsp) movl 0x108(%rbx), %esi andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) movaps %xmm0, 0x30(%rsp) movups %xmm0, 0x3c(%rsp) pushq $0x4 popq %rdx movq %r14, %rdi xorl %ecx, %ecx callq 0x635fa movq 0x18(%rsp), %rax cmpq %r14, %r15 je 0x286759 testq %rax, %rax je 0x2866c9 lock incl (%rax) movq 0x290(%rbx), %rax testq %rax, %rax je 0x2866fd lock decl (%rax) jne 0x2866fd movq 0x288(%rbx), %rsi movq 0x2a8(%rbx), %rdi testq %rdi, %rdi je 0x2866f5 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2866fd movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x288(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x290(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x298(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x2a0(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x2a8(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x2b0(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x2c0(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x2c8(%rbx) testq %rax, %rax movss 0xc(%rsp), %xmm0 je 0x286794 lock decl (%rax) jne 0x286794 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286786 movq (%rdi), %rax callq 0x18(%rax) movss 0xc(%rsp), %xmm0 jmp 0x286794 movq %rsi, %rdi callq 0x5f3e0 movss 0xc(%rsp), %xmm0 movl 0x2c0(%rbx), %eax imull 0x2c8(%rbx), %eax movq 0x288(%rbx), %rcx xorl %r12d, %r12d testl %eax, %eax cmovlel %r12d, %eax xorl %edx, %edx cmpl %edx, %eax je 0x285b3b movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x2867b3 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi jne 0x286a50 movq %rsi, %rdi callq 0x5f3e0 movq %rbx, %rdi callq 0x5f340 movq (%rdi), %rax callq 0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/convolutiondepthwise.cpp
ncnn::ConvolutionDepthWise::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { // convolv with NxN kernel // value = value + bias #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return forward_int8(bottom_blob, top_blob, opt); } #endif // NCNN_LOGE("ConvolutionDepthWise input %d x %d pad = %d %d ksize=%d %d stride=%d %d", w, h, pad_w, pad_h, kernel_w, kernel_h, stride_w, stride_h); Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; const int w = bottom_blob_bordered.w; const int h = bottom_blob_bordered.h; const size_t elemsize = bottom_blob_bordered.elemsize; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; const int outw = (w - kernel_extent_w) / stride_w + 1; const int outh = (h - kernel_extent_h) / stride_h + 1; top_blob.create(outw, outh, num_output, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; int ret = convolutiondepthwise(bottom_blob_bordered, top_blob, weight_data, bias_data, kernel_w, kernel_h, stride_w, stride_h, dilation_w, dilation_h, group, activation_type, activation_params, opt); if (ret != 0) return ret; return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x48, %rsp movq %rcx, %r12 movq %rdx, %r15 movq %rdi, %r14 cmpb $0x1, 0x1e(%rcx) jne 0x286fbf cmpq $0x1, 0x178(%r14) jne 0x286fbf movq %r14, %rdi movq %r15, %rdx movq %r12, %rcx addq $0x48, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x287170 movq %rsp, %rdx andq $0x0, 0x40(%rdx) xorps %xmm0, %xmm0 movaps %xmm0, (%rdx) movups %xmm0, 0xc(%rdx) movaps %xmm0, 0x20(%rdx) movups %xmm0, 0x2c(%rdx) movq %r14, %rdi movq %r12, %rcx callq 0x287daa pushq $-0x64 popq %rbx cmpq $0x0, (%rsp) je 0x2870eb movslq 0x38(%rsp), %rax imulq 0x40(%rsp), %rax testq %rax, %rax je 0x2870eb movq 0x10(%rsp), %r8 movl 0xd4(%r14), %eax decl %eax imull 0xdc(%r14), %eax movl 0xd0(%r14), %ecx notl %eax movl 0xd8(%r14), %esi decl %esi imull 0xe0(%r14), %esi addl 0x2c(%rsp), %eax cltd idivl 0xe4(%r14) movl %eax, %edi notl %esi addl 0x30(%rsp), %esi movl %esi, %eax cltd idivl 0xe8(%r14) leal 0x1(%rdi), %esi leal 0x1(%rax), %edx movq 0x8(%r12), %r9 movq %r15, %rdi callq 0x63810 cmpq $0x0, (%r15) je 0x2870eb movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2870eb leaq 0x168(%r14), %rdx leaq 0x1b0(%r14), %rcx movl 0xd4(%r14), %r8d movl 0xd8(%r14), %r9d movl 0xe4(%r14), %eax movl 0xe8(%r14), %r10d movl 0xdc(%r14), %r11d movl 0xe0(%r14), %r12d movl 0x108(%r14), %r13d movl 0x110(%r14), %ebp addq $0x118, %r14 # imm = 0x118 xorl %ebx, %ebx subq $0x8, %rsp leaq 0x8(%rsp), %rdi movq %r15, %rsi pushq %r14 pushq %rbp pushq %r13 pushq %r12 pushq %r11 pushq %r10 pushq %rax callq 0x287dbf addq $0x40, %rsp movq 0x8(%rsp), %rax testq %rax, %rax je 0x287118 lock decl (%rax) jne 0x287118 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x287110 movq (%rdi), %rax callq 0x18(%rax) jmp 0x287118 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0x48, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x28712f jmp 0x287167 jmp 0x28712f movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x28715f lock decl (%rax) jne 0x28715f movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x287159 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28715f movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/convolutiondepthwise.cpp
ncnn::ConvolutionDepthWise::forward_int8(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { // convolv with NxN kernel // value = value + bias int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; if (channels % group != 0 \|\| num_output % group != 0) { // reject invalid group return -100; } // NCNN_LOGE("ConvolutionDepthWise input %d x %d pad = %d %d ksize=%d %d stride=%d %d", w, h, pad_w, pad_h, kernel_w, kernel_h, stride_w, stride_h); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_int8 = bottom_blob; if (elemsize != 1) { const int channels_g = channels / group; Mat scales(channels); { float* ps = scales; for (int g = 0; g < group; g++) { float scale = bottom_blob_int8_scales[g]; for (int q = 0; q < channels_g; q++) { ps++ = scale; } } } Option opt_q = opt; opt_q.blob_allocator = opt.workspace_allocator; quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q); } Mat bottom_blob_bordered; make_padding(bottom_blob_int8, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; const int maxk = kernel_w kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } // int8 bool use_int8_requantize = int8_scale_term > 100; size_t out_elemsize = use_int8_requantize ? 1u : 4u; top_blob.create(outw, outh, num_output, out_elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels == group && group == num_output) { #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { signed char* outptr = top_blob.channel(g); const signed char* kptr = (const signed char)weight_data + maxk g; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* sptr = m.row<signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr[0] = sums8; outptr += 1; } else { // dequantize ((float)outptr)[0] = sumfp32; outptr += 4; } } } } } else { // group convolution const int channels_g = channels / group; const int num_output_g = num_output / group; #ifdef _WIN32 #pragma omp parallel for num_threads(opt.num_threads) #else // _WIN32 #pragma omp parallel for collapse(2) num_threads(opt.num_threads) #endif // _WIN32 for (int g = 0; g < group; g++) { for (int p = 0; p < num_output_g; p++) { signed char outptr = top_blob.channel(g * num_output_g + p); const signed char* weight_data_ptr = (const signed char)weight_data + maxk channels_g * num_output_g * g; for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* kptr = weight_data_ptr + maxk * channels_g * p; // channels_g for (int q = 0; q < channels_g; q++) { const Mat m = bottom_blob_bordered.channel(channels_g * g + q); const signed char* sptr = m.row<signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } kptr += maxk; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g * num_output_g + p]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr[0] = sums8; outptr += 1; } else { // dequantize ((float*)outptr)[0] = sumfp32; outptr += 4; } } } } } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x1a8, %rsp # imm = 0x1A8 movq %rcx, %r14 movq %rdx, %r8 movl 0x38(%rsi), %ebp movl 0x108(%rdi), %ecx movl %ebp, %eax cltd idivl %ecx pushq $-0x64 popq %rbx testl %edx, %edx jne 0x2871a9 movl 0xd0(%rdi), %eax cltd idivl %ecx testl %edx, %edx je 0x2871bd movl %ebx, %eax addq $0x1a8, %rsp # imm = 0x1A8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq movq %rsi, %r13 movq %r8, 0x58(%rsp) movl 0xd4(%rdi), %r15d decl %r15d imull 0xdc(%rdi), %r15d movl 0xd8(%rdi), %r12d decl %r12d movq %rdi, 0x8(%rsp) imull 0xe0(%rdi), %r12d movq 0x10(%rsi), %rax movq (%rsi), %rcx movq %rcx, 0x118(%rsp) movq 0x8(%rsi), %rcx movq %rcx, 0x120(%rsp) movq %rax, 0x128(%rsp) movl 0x18(%rsi), %edx movl %edx, 0x130(%rsp) movq 0x20(%rsi), %rdx movq %rdx, 0x138(%rsp) movups 0x28(%rsi), %xmm0 movups %xmm0, 0x140(%rsp) movl %ebp, 0x150(%rsp) movq 0x40(%rsi), %rdx movq %rdx, 0x158(%rsp) testq %rcx, %rcx je 0x28724e lock incl (%rcx) cmpq $0x1, %rax movl %ebp, 0x10(%rsp) je 0x28735c movq %r14, 0x18(%rsp) movq 0x8(%rsp), %r14 movl 0x108(%r14), %eax movl %eax, 0x30(%rsp) leaq 0x70(%rsp), %rdi andq $0x0, 0x40(%rdi) xorps %xmm0, %xmm0 movaps %xmm0, (%rdi) movups %xmm0, 0xc(%rdi) movaps %xmm0, 0x20(%rdi) movups %xmm0, 0x2c(%rdi) xorl %ebx, %ebx pushq $0x4 popq %rdx movl %ebp, %esi xorl %ecx, %ecx callq 0x635fa movl %ebp, %eax cltd idivl 0x30(%rsp) movq 0x70(%rsp), %rcx movl 0x108(%r14), %edx testl %eax, %eax cmovlel %ebx, %eax movq 0x240(%r14), %rsi testl %edx, %edx cmovlel %ebx, %edx movq 0x18(%rsp), %r14 cmpq %rdx, %rbx je 0x2872e4 movss (%rsi,%rbx,4), %xmm0 movl %eax, %edi subl $0x1, %edi jb 0x2872df movss %xmm0, (%rcx) addq $0x4, %rcx jmp 0x2872d0 incq %rbx jmp 0x2872c4 movups (%r14), %xmm0 movups 0x10(%r14), %xmm1 movups 0x20(%r14), %xmm2 movups 0x30(%r14), %xmm3 leaq 0x160(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x10(%r14), %rax movq %rax, 0x8(%rcx) leaq 0x118(%rsp), %rsi leaq 0x70(%rsp), %rdx movq %r13, %rdi callq 0x652e3 movq 0x78(%rsp), %rax testq %rax, %rax je 0x28735c lock decl (%rax) jne 0x28735c movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x287354 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28735c movq %rsi, %rdi callq 0x5f3e0 leaq 0x70(%rsp), %rdx andq $0x0, 0x40(%rdx) xorps %xmm0, %xmm0 movaps %xmm0, (%rdx) movups %xmm0, 0xc(%rdx) movaps %xmm0, 0x20(%rdx) movups %xmm0, 0x2c(%rdx) leaq 0x118(%rsp), %rsi movq 0x8(%rsp), %rdi movq %r14, %rcx callq 0x287daa pushq $-0x64 popq %rbx cmpq $0x0, 0x70(%rsp) je 0x287c5f movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x287c5f movq %r14, 0x18(%rsp) notl %r15d movl 0x9c(%rsp), %ebx addl %ebx, %r15d movl %r15d, %eax cltd movq 0x8(%rsp), %r14 idivl 0xe4(%r14) movl %eax, %r15d notl %r12d addl 0xa0(%rsp), %r12d movl %r12d, %eax cltd idivl 0xe8(%r14) movl %eax, %r12d movslq 0xd4(%r14), %rax movslq 0xd8(%r14), %r13 imulq %rax, %r13 leaq 0x160(%rsp), %rdi leaq 0x27(%rsp), %rdx movq %r13, %rsi callq 0x73bbe movq %r15, 0x28(%rsp) leal 0x1(%r15), %esi movq %r12, 0x60(%rsp) leal 0x1(%r12), %edx imull 0xe0(%r14), %ebx movq 0x160(%rsp), %rbp movl 0xd4(%r14), %eax movl 0xdc(%r14), %ecx imull %eax, %ecx subl %ecx, %ebx xorl %ecx, %ecx xorl %edi, %edi xorl %r8d, %r8d movq 0x58(%rsp), %r14 movq 0x8(%rsp), %r9 cmpl 0xd8(%r9), %ecx jge 0x2874a4 movslq %r8d, %r8 leaq (,%r8,4), %r10 addq %rbp, %r10 xorl %r9d, %r9d cmpl %eax, %r9d jge 0x28749b movl %edi, (%r10,%r9,4) movq 0x8(%rsp), %rax addl 0xdc(%rax), %edi movl 0xd4(%rax), %eax incq %r9 jmp 0x28747c addl %ebx, %edi incl %ecx addq %r9, %r8 jmp 0x28745d movq 0x8(%rsp), %rax movl 0xd0(%rax), %ecx movl 0x10c(%rax), %edi xorl %eax, %eax movl %edi, 0x44(%rsp) cmpl $0x65, %edi setl %al leaq 0x1(%rax,%rax,2), %r8 movq 0x18(%rsp), %rax movq 0x8(%rax), %r9 movq %r14, %rdi callq 0x63810 pushq $-0x64 popq %rax movq %rax, 0x50(%rsp) cmpq $0x0, (%r14) movq 0x28(%rsp), %r8 movl 0x10(%rsp), %eax je 0x287c4d movslq 0x38(%r14), %rcx imulq 0x40(%r14), %rcx testq %rcx, %rcx je 0x287c4d movq 0x8(%rsp), %rdx movl 0xd0(%rdx), %ecx movl 0x108(%rdx), %edi movl %eax, %esi xorl %edi, %esi movl %edi, %edx xorl %ecx, %edx orl %esi, %edx jne 0x28784a movq $0x0, 0x50(%rsp) testl %r13d, %r13d movl $0x0, %ebx cmovgl %r13d, %ebx movq $0x0, 0x48(%rsp) xorl %r9d, %r9d movq 0x8(%rsp), %rdx cltq cmpq %rax, %r9 jge 0x287c4d movq 0x40(%r14), %r12 imulq %r9, %r12 imulq 0x10(%r14), %r12 addq (%r14), %r12 movslq 0x9c(%rsp), %rcx movq 0xb0(%rsp), %rsi imulq %r9, %rsi movq 0x80(%rsp), %rax imulq %rax, %rsi addq 0x70(%rsp), %rsi movq %rsi, 0x30(%rsp) imulq %rax, %rcx movq %rcx, 0x10(%rsp) movq 0x168(%rdx), %r15 addq 0x48(%rsp), %r15 xorl %eax, %eax movq %r9, 0x38(%rsp) cmpl 0x60(%rsp), %eax jg 0x287832 xorl %r14d, %r14d movl %eax, 0xc0(%rsp) cltq movq %rax, 0x18(%rsp) cmpl %r8d, %r14d jg 0x287824 movslq 0xe8(%rdx), %rax imulq 0x18(%rsp), %rax imulq 0x10(%rsp), %rax movslq 0xe4(%rdx), %rdx movslq %r14d, %rcx imulq %rdx, %rcx addq 0x30(%rsp), %rcx addq %rax, %rcx xorl %edx, %edx xorl %eax, %eax cmpq %rdx, %rbx je 0x28761a movslq (%rbp,%rdx,4), %rsi movsbl (%rcx,%rsi), %esi movsbl (%r15,%rdx), %edi imull %esi, %edi addl %edi, %eax incq %rdx jmp 0x2875fd movq 0x8(%rsp), %rdx movq 0x1f8(%rdx), %rcx movss (%rcx,%r9,4), %xmm1 xorps %xmm0, %xmm0 ucomiss %xmm1, %xmm0 je 0x28764d movq 0x240(%rdx), %rcx mulss (%rcx,%r9,4), %xmm1 movss 0x16763f(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 cvtsi2ss %eax, %xmm4 mulss %xmm0, %xmm4 cmpl $0x0, 0x100(%rdx) je 0x28766b movq 0x1b0(%rdx), %rax addss (%rax,%r9,4), %xmm4 movl 0x110(%rdx), %eax decl %eax cmpl $0x5, %eax ja 0x287690 leaq 0x170dd9(%rip), %rcx # 0x3f8458 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq %rax maxss 0x166980(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2877ca movaps %xmm4, %xmm0 movaps %xmm4, 0xd0(%rsp) callq 0x5f410 addss 0x1675d8(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x38(%rsp), %r9 movq 0x28(%rsp), %r8 movq 0x8(%rsp), %rdx mulss 0xd0(%rsp), %xmm0 jmp 0x2877ca movq 0x118(%rdx), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2877c7 jmp 0x2877ca movss 0x169ab8(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x166982(%rip), %xmm0 # 0x3ee090 cmpltss 0x169aa5(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x38(%rsp), %r9 movq 0x28(%rsp), %r8 movq 0x8(%rsp), %rdx movaps %xmm0, %xmm1 movss 0x167543(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2877ca movq 0x118(%rdx), %rax xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x16751f(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2877ca movq 0x118(%rdx), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x1668fa(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2877ca movss 0x1674de(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x287690 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 cmpl $0x65, 0x44(%rsp) jl 0x287810 movq 0x288(%rdx), %rax mulss (%rax,%r9,4), %xmm0 callq 0x5f2d0 movq 0x38(%rsp), %r9 movq 0x28(%rsp), %r8 movq 0x8(%rsp), %rdx cvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movb %al, (%r12) pushq $0x1 jmp 0x287818 movss %xmm0, (%r12) pushq $0x4 popq %rax addq %rax, %r12 incl %r14d jmp 0x2875c7 movl 0xc0(%rsp), %eax incl %eax jmp 0x2875ac incq %r9 movl 0x108(%rdx), %eax addq %r13, 0x48(%rsp) movq 0x58(%rsp), %r14 jmp 0x28754b cltd idivl %edi movl %eax, %esi movl %ecx, %eax cltd idivl %edi movl %esi, %ecx imull %r13d, %ecx movl %eax, %edx imull %ecx, %edx movl %edx, 0x6c(%rsp) movq $0x0, 0x50(%rsp) testl %r13d, %r13d movl $0x0, %r12d cmovgl %r13d, %r12d testl %esi, %esi movl $0x0, %r15d cmovgl %esi, %r15d movslq %ecx, %rcx movq %rcx, 0xf8(%rsp) testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x110(%rsp) movslq %esi, %rcx movq %rcx, 0xf0(%rsp) cltq movq %rax, 0xe8(%rsp) xorl %ebx, %ebx movq 0x8(%rsp), %rdx movq 0x60(%rsp), %rcx movslq %edi, %rax cmpq %rax, %rbx jge 0x287c4d movl 0x6c(%rsp), %eax imull %ebx, %eax movq %rbx, %rsi imulq 0xe8(%rsp), %rsi movq %rsi, 0x108(%rsp) cltq movq %rax, 0x100(%rsp) movq %rbx, 0x10(%rsp) imulq 0xf0(%rsp), %rbx xorl %eax, %eax cmpq 0x110(%rsp), %rax je 0x287c3a movq %rax, %rsi movq 0x108(%rsp), %rax leaq (%rsi,%rax), %r9 movq 0x58(%rsp), %rax movq 0x40(%rax), %rdi movq %r9, 0xd0(%rsp) imulq %r9, %rdi imulq 0x10(%rax), %rdi addq (%rax), %rdi movq 0x168(%rdx), %rax addq 0x100(%rsp), %rax movq %rsi, 0x48(%rsp) imulq 0xf8(%rsp), %rsi addq %rax, %rsi movq %rsi, 0x38(%rsp) xorl %eax, %eax cmpl %ecx, %eax jg 0x287c2d movl %eax, 0x30(%rsp) xorl %r14d, %r14d cmpl %r8d, %r14d jg 0x287c1d movq %rdi, 0x18(%rsp) movslq 0x9c(%rsp), %rcx movq %rdx, %r8 movq 0x80(%rsp), %rdx movq 0xb0(%rsp), %rax movslq 0xe8(%r8), %rsi movslq 0x30(%rsp), %rdi imulq %rsi, %rdi imulq %rdx, %rcx imulq %rdi, %rcx addq 0x70(%rsp), %rcx imulq %rdx, %rax movslq 0xe4(%r8), %rsi movslq %r14d, %rdx imulq %rsi, %rdx addq %rcx, %rdx xorl %esi, %esi xorl %ecx, %ecx movq 0x38(%rsp), %rdi cmpq %r15, %rsi je 0x287a13 leaq (%rsi,%rbx), %r8 imulq %rax, %r8 addq %rdx, %r8 xorl %r9d, %r9d cmpq %r9, %r12 je 0x287a0b movslq (%rbp,%r9,4), %r10 movsbl (%r8,%r10), %r10d movsbl (%rdi,%r9), %r11d imull %r10d, %r11d addl %r11d, %ecx incq %r9 jmp 0x2879eb addq %r13, %rdi incq %rsi jmp 0x2879d8 movq 0x8(%rsp), %rdx movq 0x1f8(%rdx), %rax movq 0x10(%rsp), %rsi movss (%rax,%rsi,4), %xmm1 xorps %xmm0, %xmm0 ucomiss %xmm1, %xmm0 je 0x287a49 movq 0x240(%rdx), %rax mulss (%rax,%rsi,4), %xmm1 movss 0x167243(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 cvtsi2ss %ecx, %xmm4 mulss %xmm0, %xmm4 cmpl $0x0, 0x100(%rdx) je 0x287a6e movq 0x1b0(%rdx), %rax movq 0xd0(%rsp), %rcx addss (%rax,%rcx,4), %xmm4 movl 0x110(%rdx), %eax decl %eax cmpl $0x5, %eax ja 0x287a93 leaq 0x1709be(%rip), %rcx # 0x3f8440 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq %rax maxss 0x16657d(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x287bc3 movaps %xmm4, %xmm0 movaps %xmm4, 0xc0(%rsp) callq 0x5f410 addss 0x1671d5(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x10(%rsp), %rsi movq 0x8(%rsp), %rdx mulss 0xc0(%rsp), %xmm0 jmp 0x287bc3 movq 0x118(%rdx), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x287bc0 jmp 0x287bc3 movss 0x1696ba(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x166584(%rip), %xmm0 # 0x3ee090 cmpltss 0x1696a7(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x10(%rsp), %rsi movq 0x8(%rsp), %rdx movaps %xmm0, %xmm1 movss 0x16714a(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x287bc3 movq 0x118(%rdx), %rax xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x167126(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x287bc3 movq 0x118(%rdx), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x166501(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x287bc3 movss 0x1670e5(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x287a93 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 cmpl $0x65, 0x44(%rsp) jl 0x287c01 movq 0x288(%rdx), %rax mulss (%rax,%rsi,4), %xmm0 callq 0x5f2d0 movq 0x8(%rsp), %rdx cvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movq 0x18(%rsp), %rdi movb %al, (%rdi) pushq $0x1 jmp 0x287c0c movq 0x18(%rsp), %rdi movss %xmm0, (%rdi) pushq $0x4 popq %rax movq 0x28(%rsp), %r8 addq %rax, %rdi incl %r14d jmp 0x287974 movl 0x30(%rsp), %eax incl %eax movq 0x60(%rsp), %rcx jmp 0x287965 movq 0x48(%rsp), %rax incq %rax jmp 0x287904 movq 0x10(%rsp), %rbx incq %rbx movl 0x108(%rdx), %edi jmp 0x2878c3 leaq 0x160(%rsp), %rdi callq 0x624be movq 0x50(%rsp), %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x287c90 lock decl (%rax) jne 0x287c90 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x287c88 movq (%rdi), %rax callq 0x18(%rax) jmp 0x287c90 movq %rsi, %rdi callq 0x5f3e0 movq 0x120(%rsp), %rax testq %rax, %rax je 0x2871a9 lock decl (%rax) jne 0x2871a9 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x287cca movq (%rdi), %rax callq 0x18(%rax) jmp 0x2871a9 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2871a9 jmp 0x287da2 jmp 0x287da2 jmp 0x287da2 movq %rax, %rbx leaq 0x160(%rsp), %rdi callq 0x624be jmp 0x287d30 jmp 0x287d2d movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x287d61 lock decl (%rax) jne 0x287d61 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x287d51 movq (%rdi), %rax callq 0x18(%rax) jmp 0x287d61 jmp 0x287da2 movq %rax, %rbx jmp 0x287d61 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x287d61 lock decl (%rax) jne 0x287d61 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x287d5b movq %rsi, %rdi callq 0x5f3e0 jmp 0x287d61 movq (%rdi), %rax callq 0x18(%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x287d98 lock decl (%rax) jne 0x287d98 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi jne 0x287d92 movq %rsi, %rdi callq 0x5f3e0 jmp 0x287d98 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x287da2 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/convolutiondepthwise.cpp
ncnn::convolutiondepthwise(ncnn::Mat const&, ncnn::Mat&, ncnn::Mat const&, ncnn::Mat const&, int, int, int, int, int, int, int, int, ncnn::Mat const&, ncnn::Option const&)	static int convolutiondepthwise(const Mat& bottom_blob, Mat& top_blob, const Mat& weight_data, const Mat& bias_data, int kernel_w, int kernel_h, int stride_w, int stride_h, int dilation_w, int dilation_h, int group, int activation_type, const Mat& activation_params, const Option& opt) { const int w = bottom_blob.w; const int inch = bottom_blob.c; const int outw = top_blob.w; const int outh = top_blob.h; const int outch = top_blob.c; const int bias_term = bias_data.empty() ? 0 : 1; const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } // depth-wise if (inch == group && group == outch) { #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data + maxk g; const Mat m = bottom_blob.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { float sum = 0.f; if (bias_term) sum = bias_data[g]; const float* sptr = m.row(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; float w = kptr[k]; sum += val * w; } outptr[j] = activation_ss(sum, activation_type, activation_params); } outptr += outw; } } } else { // group convolution const int inch_g = inch / group; const int outch_g = outch / group; #ifdef _WIN32 #pragma omp parallel for num_threads(opt.num_threads) #else #pragma omp parallel for collapse(2) num_threads(opt.num_threads) #endif for (int g = 0; g < group; g++) { for (int p = 0; p < outch_g; p++) { float* outptr = top_blob.channel(g * outch_g + p); const float* weight_data_ptr = (const float)weight_data + maxk inch_g * outch_g * g; // shadowed variable for less openmp task args const int outw = top_blob.w; const int outh = top_blob.h; for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { float sum = 0.f; if (bias_term) sum = bias_data[outch_g * g + p]; const float* kptr = weight_data_ptr + maxk * inch_g * p; for (int q = 0; q < inch_g; q++) { const Mat m = bottom_blob.channel(inch_g * g + q); const float* sptr = m.row(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; float w = kptr[k]; sum += val * w; } kptr += maxk; } outptr[j] = activation_ss(sum, activation_type, activation_params); } outptr += outw; } } } } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x128, %rsp # imm = 0x128 movl %r9d, %ebp movl %r8d, %r15d movq %rdx, 0xa0(%rsp) movq %rsi, 0x58(%rsp) movl 0x170(%rsp), %ebx movq %rdi, 0x28(%rsp) movl 0x2c(%rdi), %r12d cmpq $0x0, (%rcx) movq %rcx, 0xa8(%rsp) je 0x287e14 movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax sete 0xf(%rsp) jmp 0x287e19 movb $0x1, 0xf(%rsp) movl 0x180(%rsp), %r13d movq 0x28(%rsp), %rax movl 0x38(%rax), %eax movl %eax, 0x40(%rsp) movq 0x58(%rsp), %rax movslq 0x2c(%rax), %rcx movq %rcx, 0x48(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x18(%rsp) movl 0x38(%rax), %eax movl %eax, 0x10(%rsp) movl %ebp, %eax imull %r15d, %eax movslq %eax, %r14 leaq 0x110(%rsp), %rdi leaq 0x27(%rsp), %rdx movq %r14, %rsi callq 0x73bbe movl %ebp, %r9d imull 0x178(%rsp), %r12d movl %ebx, %eax imull %r15d, %eax subl %eax, %r12d xorl %eax, %eax testl %r15d, %r15d cmovlel %eax, %r15d testl %ebp, %ebp cmovlel %eax, %r9d leaq 0x110(%rsp), %rcx movq (%rcx), %rbp movl %r15d, %ecx xorl %edx, %edx xorl %esi, %esi cmpl %r9d, %eax je 0x287ec8 movslq %ecx, %rdi movslq %esi, %r8 leal (%r15,%r8), %esi cmpq %r8, %rdi je 0x287ebe movl %edx, (%rbp,%r8,4) incq %r8 addl %ebx, %edx jmp 0x287ead addl %r12d, %edx incl %eax addl %r15d, %ecx jmp 0x287e9e movl 0x40(%rsp), %eax movl %eax, %edx xorl %r13d, %edx movl 0x10(%rsp), %esi movl %esi, %ecx xorl %r13d, %ecx orl %edx, %ecx movb 0xf(%rsp), %r10b movl 0x188(%rsp), %r11d jne 0x288230 movq 0x58(%rsp), %rax movq (%rax), %rcx movq %rcx, 0x80(%rsp) movq 0x40(%rax), %rcx imulq 0x10(%rax), %rcx movq %rcx, 0x78(%rsp) movq %r13, %rdx movq 0xa0(%rsp), %rax movq (%rax), %r13 movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rax movq (%rsi), %rcx movq %rcx, 0x70(%rsp) movq 0x10(%rsi), %rcx movq 0x40(%rsi), %rsi imulq %rcx, %rsi movq %rsi, 0x68(%rsp) imulq %rcx, %rax xorl %esi, %esi testl %r14d, %r14d movl $0x0, %r12d cmovgl %r14d, %r12d movslq 0x160(%rsp), %rdi movq 0x48(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r8d cmovgl %ecx, %r8d movslq 0x168(%rsp), %rcx movq 0x18(%rsp), %r15 testl %r15d, %r15d cmovlel %esi, %r15d movq %r15, 0x18(%rsp) testl %edx, %edx cmovlel %esi, %edx movq %rdx, 0x60(%rsp) imulq %rax, %rcx movq %rcx, 0x88(%rsp) shlq $0x2, %r14 leaq 0x1704e5(%rip), %r9 # 0x3f8488 movq %rdi, 0x28(%rsp) movq %r8, 0x50(%rsp) cmpq 0x60(%rsp), %rsi je 0x288651 movq 0x78(%rsp), %rax imulq %rsi, %rax addq 0x80(%rsp), %rax movq %rax, 0x10(%rsp) movq 0x68(%rsp), %rax imulq %rsi, %rax addq 0x70(%rsp), %rax movq %rax, 0xb0(%rsp) xorl %ebx, %ebx movq %rsi, 0x30(%rsp) cmpq 0x18(%rsp), %rbx je 0x288225 movq 0xa8(%rsp), %rax movq (%rax), %rax movq %rax, 0x40(%rsp) movq %rbx, 0x90(%rsp) imulq 0x88(%rsp), %rbx addq 0xb0(%rsp), %rbx movq 0x190(%rsp), %rax movq (%rax), %rax movq %rax, 0x38(%rsp) xorl %r15d, %r15d cmpq %r8, %r15 je 0x288202 xorps %xmm4, %xmm4 testb %r10b, %r10b jne 0x28804d movq 0x40(%rsp), %rax movss (%rax,%rsi,4), %xmm4 movq %r15, %rax imulq %rdi, %rax leaq (%rbx,%rax,4), %rax xorl %ecx, %ecx cmpq %rcx, %r12 je 0x288079 movslq (%rbp,%rcx,4), %rdx movss (%r13,%rcx,4), %xmm0 mulss (%rax,%rdx,4), %xmm0 addss %xmm0, %xmm4 incq %rcx jmp 0x28805a leal -0x1(%r11), %eax cmpl $0x5, %eax ja 0x288093 movslq (%r9,%rax,4), %rax addq %r9, %rax jmpq %rax maxss 0x165f7d(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2881ef movaps %xmm4, %xmm0 movaps %xmm4, 0xc0(%rsp) callq 0x5f410 addss 0x166bd5(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 leaq 0x1703c4(%rip), %r9 # 0x3f8488 movq 0x50(%rsp), %r8 movq 0x28(%rsp), %rdi movq 0x30(%rsp), %rsi movl 0x188(%rsp), %r11d movb 0xf(%rsp), %r10b mulss 0xc0(%rsp), %xmm0 jmp 0x2881ef movq 0x38(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2881ec jmp 0x2881ef movss 0x1690a3(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x165f6d(%rip), %xmm0 # 0x3ee090 cmpltss 0x169090(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 leaq 0x170341(%rip), %r9 # 0x3f8488 movq 0x50(%rsp), %r8 movq 0x28(%rsp), %rdi movq 0x30(%rsp), %rsi movl 0x188(%rsp), %r11d movb 0xf(%rsp), %r10b movaps %xmm0, %xmm1 movss 0x166b1a(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2881ef xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x166afd(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x38(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2881ef movq 0x38(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x165ed5(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2881ef movss 0x166ab9(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x288093 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 movq 0x10(%rsp), %rax movss %xmm0, (%rax,%r15,4) incq %r15 jmp 0x288032 movq 0x48(%rsp), %rax movq 0x10(%rsp), %rcx leaq (%rcx,%rax,4), %rcx movq %rcx, 0x10(%rsp) movq 0x90(%rsp), %rbx incq %rbx jmp 0x287feb incq %rsi addq %r14, %r13 jmp 0x287fad cltd idivl %r13d movl %eax, %ecx movl %esi, %eax cltd idivl %r13d movl %ecx, %edx imull %r14d, %edx xorl %esi, %esi testl %r14d, %r14d movl $0x0, %ebx cmovgl %r14d, %ebx testl %ecx, %ecx movl $0x0, %r12d cmovgl %ecx, %r12d movslq 0x160(%rsp), %rdi movq %rdi, 0x50(%rsp) movslq 0x168(%rsp), %rdi movq %rdi, 0x70(%rsp) testl %eax, %eax movl $0x0, %edi cmovgl %eax, %edi movq %rdi, 0x68(%rsp) movslq %edx, %rdx movslq %ecx, %rcx cltq testl %r13d, %r13d cmovlel %esi, %r13d movq %rax, 0xd8(%rsp) movq %rcx, 0xe0(%rsp) imull %ecx, %eax imull %r14d, %eax movl %eax, 0x98(%rsp) shlq $0x2, %rdx movq %rdx, 0xf0(%rsp) shlq $0x2, %r14 xorl %eax, %eax xorl %ecx, %ecx movq %r13, 0x60(%rsp) cmpq %r13, %rcx je 0x288651 movl %eax, 0x9c(%rsp) movslq %eax, %rdx shlq $0x2, %rdx movq %rcx, %rsi movq 0x58(%rsp), %rax movq 0x40(%rax), %rdi imulq 0x10(%rax), %rdi movq %rdi, 0x100(%rsp) imulq 0xd8(%rsp), %rsi movq %rsi, 0x108(%rsp) movslq 0x2c(%rax), %rsi movq %rcx, 0xe8(%rsp) movq %rcx, %r13 imulq 0xe0(%rsp), %r13 testl %esi, %esi movl $0x0, %edi movq %rsi, 0x78(%rsp) cmovgl %esi, %edi movq %rdi, 0x40(%rsp) movl 0x30(%rax), %esi testl %esi, %esi movl $0x0, %ecx cmovlel %ecx, %esi movq %rsi, 0x88(%rsp) movq (%rax), %rax movq %rax, 0xf8(%rsp) movq 0xa0(%rsp), %rax addq (%rax), %rdx movq %rdx, 0x30(%rsp) xorl %ecx, %ecx cmpq 0x68(%rsp), %rcx je 0x28862e movq 0x108(%rsp), %rax leaq (%rcx,%rax), %rdx movq 0x100(%rsp), %rsi movq %rdx, 0x90(%rsp) imulq %rdx, %rsi addq 0xf8(%rsp), %rsi movq %rsi, 0x10(%rsp) xorl %edx, %edx movq %rcx, 0x80(%rsp) cmpq 0x88(%rsp), %rdx je 0x288614 movq 0xa8(%rsp), %rax movq (%rax), %rax movq %rax, 0xc0(%rsp) movq %rdx, 0x48(%rsp) imulq 0x70(%rsp), %rdx movq %rdx, 0x38(%rsp) movq 0x190(%rsp), %rax movq (%rax), %rax movq %rax, 0x18(%rsp) xorl %r15d, %r15d cmpq 0x40(%rsp), %r15 je 0x2885ec xorps %xmm4, %xmm4 testb %r10b, %r10b jne 0x288418 movq 0x90(%rsp), %rax movq 0xc0(%rsp), %rcx movss (%rcx,%rax,4), %xmm4 movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rcx imulq 0x38(%rsp), %rcx movq 0x10(%rsi), %rdx movq 0x40(%rsi), %rax imulq %rdx, %rax imulq %rdx, %rcx addq (%rsi), %rcx movq %r15, %rdx imulq 0x50(%rsp), %rdx leaq (%rcx,%rdx,4), %rcx movq 0x30(%rsp), %rdx xorl %esi, %esi cmpq %r12, %rsi je 0x288488 leaq (%rsi,%r13), %rdi imulq %rax, %rdi addq %rcx, %rdi xorl %r8d, %r8d cmpq %r8, %rbx je 0x288480 movslq (%rbp,%r8,4), %r9 movss (%rdx,%r8,4), %xmm0 mulss (%rdi,%r9,4), %xmm0 addss %xmm0, %xmm4 incq %r8 jmp 0x288461 incq %rsi addq %r14, %rdx jmp 0x28844e leal -0x1(%r11), %eax cmpl $0x5, %eax ja 0x2884a9 leaq 0x16ffd8(%rip), %rcx # 0x3f8470 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq %rax maxss 0x165b67(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2885d9 movaps %xmm4, %xmm0 movaps %xmm4, 0xb0(%rsp) callq 0x5f410 addss 0x1667bf(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movl 0x188(%rsp), %r11d movb 0xf(%rsp), %r10b mulss 0xb0(%rsp), %xmm0 jmp 0x2885d9 movq 0x18(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2885d6 jmp 0x2885d9 movss 0x168ca3(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x165b6d(%rip), %xmm0 # 0x3ee090 cmpltss 0x168c90(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movl 0x188(%rsp), %r11d movb 0xf(%rsp), %r10b movaps %xmm0, %xmm1 movss 0x166730(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2885d9 xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x166713(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x18(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2885d9 movq 0x18(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x165aeb(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2885d9 movss 0x1666cf(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x2884a9 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 movq 0x10(%rsp), %rax movss %xmm0, (%rax,%r15,4) incq %r15 jmp 0x2883f0 movq 0x78(%rsp), %rax movq 0x10(%rsp), %rcx leaq (%rcx,%rax,4), %rcx movq %rcx, 0x10(%rsp) movq 0x48(%rsp), %rdx incq %rdx movq 0x80(%rsp), %rcx jmp 0x2883ac incq %rcx movq 0x30(%rsp), %rax addq 0xf0(%rsp), %rax movq %rax, 0x30(%rsp) jmp 0x28836a movq 0xe8(%rsp), %rcx incq %rcx movl 0x9c(%rsp), %eax addl 0x98(%rsp), %eax movq 0x60(%rsp), %r13 jmp 0x2882cb leaq 0x110(%rsp), %rdi callq 0x624be addq $0x128, %rsp # imm = 0x128 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq	/csukuangfj[P]ncnn/src/layer/convolutiondepthwise.cpp
ncnn::ConvolutionDepthWise_x86::create_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x50, %rsp movq (%rdi), %rax movq -0x18(%rax), %r13 cmpl $0x0, 0x160(%rdi,%r13) je 0x288dd5 xorl %eax, %eax addq $0x50, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 retq movq %rsi, %r14 movq %rdi, %rbx movl 0x110(%rdi,%r13), %ecx decl %ecx cmpl $0x5, %ecx ja 0x289171 leaq 0x16f6cb(%rip), %rax # 0x3f84c0 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx jmpq %rcx pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x288f68 pushq $0x47 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x288f68 pushq $0x36 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax movss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax movss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) jmp 0x288f68 pushq $0x1e popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x288f68 pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax movss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) jmp 0x288f68 pushq $0x43 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax movss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax movss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) leaq 0x8(%rsp), %rdi callq 0x71614 movq (%r15), %rax movq %r15, %rdi movq %r14, %rsi callq 0x20(%rax) movq (%rbx), %rax movq %r15, 0x8(%rbx) movq -0x18(%rax), %r15 cmpb $0x1, 0x1e(%r14) jne 0x288fab cmpq $0x1, 0x178(%rbx,%r15) jne 0x288fab movq %rbx, %rdi movq %r14, %rsi callq 0x289278 jmp 0x288dc5 movl 0xd0(%rbx,%r15), %esi movl 0xd4(%rbx,%r15), %r10d movl 0xd8(%rbx,%r15), %r11d movl %r11d, %r8d imull %r10d, %r8d movl 0x104(%rbx,%r15), %eax movl 0x108(%rbx,%r15), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x28911e imull %ecx, %eax cmpl %ecx, %eax jne 0x28911e leaq (%rbx,%r15), %rsi testb $0x3, %cl jne 0x289075 cmpb $0x0, 0x27(%r14) je 0x289075 addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x4 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x289129 lock decl (%rax) jne 0x289129 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x2891a8 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289129 xorl $0x3, %r10d xorl $0x3, %r11d orl %r10d, %r11d jne 0x28911e cmpl $0x1, 0xdc(%rsi) jne 0x2890aa cmpl $0x1, 0xe0(%rsi) jne 0x2890aa cmpl $0x1, 0xe4(%rsi) jne 0x2890aa cmpl $0x1, 0xe8(%rsi) je 0x2890ce cmpl $0x1, 0xdc(%rsi) jne 0x28911e cmpl $0x1, 0xe0(%rsi) jne 0x28911e cmpl $0x2, 0xe4(%rsi) jne 0x28911e cmpl $0x2, 0xe8(%rsi) jne 0x28911e addq $0x168, %r15 # imm = 0x168 cmpq $0x28, %r15 je 0x289129 movq 0x8(%rbx,%r15), %rax testq %rax, %rax je 0x2890e8 lock incl (%rax) leaq 0x28(%rbx), %r12 movq 0x30(%rbx), %rax testq %rax, %rax je 0x2891bd lock decl (%rax) jne 0x2891bd movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x2891b5 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2891bd movq %rbx, %rdi movq %r14, %rsi callq 0x2894d8 cmpb $0x1, (%r14) jne 0x288dc5 movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x289181 lock decl (%rax) jne 0x289181 movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x289179 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289181 xorl %r15d, %r15d jmp 0x288f81 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rbx) movups %xmm0, (%rbx) movups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x288dc5 movq %rsi, %rdi callq 0x5f3e0 jmp 0x289129 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%r12) movups %xmm0, (%r12) andl $0x0, 0x60(%rbx) movups %xmm0, 0x50(%rbx) movups (%rbx,%r15), %xmm0 movups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r15), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r15), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r15), %rax movq %rax, 0x48(%rbx) movups 0x28(%rbx,%r15), %xmm0 movups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r15), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r15), %rax movq %rax, 0x68(%rbx) jmp 0x289129 jmp 0x289256 jmp 0x289262 jmp 0x289262 jmp 0x289262 movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x28926f lock decl (%rax) jne 0x28926f movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi jne 0x28924e movq %rsi, %rdi callq 0x5f3e0 jmp 0x28926f movq (%rdi), %rax callq 0x18(%rax) jmp 0x28926f movq %rax, %rdi callq 0x61d68 jmp 0x289262 jmp 0x289262 movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq %rbx, %rdi callq 0x5f340 nop	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
ncnn::ConvolutionDepthWise_x86::create_pipeline_int8_x86(ncnn::Option const&)	int ConvolutionDepthWise_x86::create_pipeline_int8_x86(const Option& opt) { const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { elempack = channels % 8 == 0 ? 8 : 1; } #endif // __SSE2__ if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } if (elempack == 1) { weight_data_tm = weight_data; } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %r15 pushq %r14 pushq %r12 pushq %rbx subq $0x48, %rsp movq %rsi, %r15 movq %rdi, %rbx movq (%rdi), %rax movq -0x18(%rax), %r14 movl 0xd8(%rdi,%r14), %r8d imull 0xd4(%rdi,%r14), %r8d movl 0xd0(%rdi,%r14), %esi movl 0x104(%rdi,%r14), %eax movl 0x108(%rdi,%r14), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x28935a imull %ecx, %eax cmpl %ecx, %eax jne 0x28935a testb $0x7, %cl jne 0x2893ad cmpb $0x0, 0x27(%r15) je 0x2893ad addq %rbx, %r14 addq $0x168, %r14 # imm = 0x168 movq %rsp, %r12 movq %r12, %rdi movq %r14, %rsi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e addq $0x28, %rbx pushq $0x8 popq %rdx movq %r12, %rdi movq %rbx, %rsi movq %r15, %rcx callq 0x64e3b movq 0x8(%rsp), %rax testq %rax, %rax je 0x289487 lock decl (%rax) jne 0x289487 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x28941e movq (%rdi), %rax callq 0x18(%rax) jmp 0x289487 movq %rbx, %rdi movq %r15, %rsi callq 0x2894d8 cmpb $0x1, (%r15) jne 0x289487 movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x2893fa lock decl (%rax) jne 0x2893fa movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x2893f2 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2893fa addq $0x168, %r14 # imm = 0x168 cmpq $0x28, %r14 je 0x289487 movq 0x8(%rbx,%r14), %rax testq %rax, %rax je 0x2893cb lock incl (%rax) leaq 0x28(%rbx), %r15 movq 0x30(%rbx), %rax testq %rax, %rax je 0x289430 lock decl (%rax) jne 0x289430 movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x289428 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289430 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rbx) movups %xmm0, (%rbx) movups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x289487 movq %rsi, %rdi callq 0x5f3e0 jmp 0x289487 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%r15) movups %xmm0, (%r15) andl $0x0, 0x60(%rbx) movups %xmm0, 0x50(%rbx) movups (%rbx,%r14), %xmm0 movups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r14), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r14), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r14), %rax movq %rax, 0x48(%rbx) movups 0x28(%rbx,%r14), %xmm0 movups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r14), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r14), %rax movq %rax, 0x68(%rbx) xorl %eax, %eax addq $0x48, %rsp popq %rbx popq %r12 popq %r14 popq %r15 retq jmp 0x2894cf movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2894c7 lock decl (%rax) jne 0x2894c7 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2894c1 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2894c7 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
ncnn::ConvolutionDepthWise_x86::create_group_ops(ncnn::Option const&)	int ConvolutionDepthWise_x86::create_group_ops(const Option& opt) { // create Convolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Convolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[5]; weights[0] = weight_data_g; weights[1] = bias_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[2] = weight_data_int8_scales_g; weights[3] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[4] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[4]; weights[0] = weight_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[1] = weight_data_int8_scales_g; weights[2] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[3] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x278, %rsp # imm = 0x278 movq %rsi, 0x250(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x104(%rdi,%rdx), %eax movl 0x108(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x10(%rdi), %rax movq %rax, 0x1e0(%rsp) xorl %r12d, %r12d movq 0x10(%r14), %rax movq 0x18(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x289575 movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x289570 movq (%rdi), %rax callq 0x8(%rax) incq %r12 jmp 0x289547 imull %r15d, %ebx cmpq %rax, %rcx je 0x289582 movq %rax, 0x18(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x108(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0x1e0(%rsp), %rdi callq 0x6fbc2 leaq 0x118(%r14), %rax movq %rax, 0x258(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0x6c(%rsp) movslq %ebx, %rax movq %rax, 0x260(%rsp) movslq %r15d, %rax movq %rax, 0x248(%rsp) pushq $0x1 popq %rbx xorl %edx, %edx movl %r15d, 0x14(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rcx cmpq %rcx, %rdx jge 0x28a2ad movq %rdx, %rcx movq %rdx, 0x8(%rsp) movq 0x260(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdx, 0x80(%rsp) movl %esi, 0x88(%rsp) movq %rax, 0x90(%rsp) movl %ebx, 0x98(%rsp) movl %edi, 0x9c(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa0(%rsp) movl %ebx, 0xa8(%rsp) movq %rdi, 0xb0(%rsp) leaq 0x200(%rsp), %rdi leaq 0x70(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2896d3 lock decl (%rax) jne 0x2896d3 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2896cb movq (%rdi), %rax callq 0x18(%rax) jmp 0x2896d3 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x289743 movq 0x8(%rsp), %r15 movq 0x248(%rsp), %rcx imulq %rcx, %r15 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0x1f8(%rsp) imulq %rsi, %r15 addq 0x1b0(%r14,%rax), %r15 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x1c(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0x1f0(%rsp) movl %ebx, %r13d movl 0x14(%rsp), %eax movl %eax, 0x18(%rsp) movq %rcx, 0x1e8(%rsp) jmp 0x28977d xorl %r15d, %r15d movq $0x0, 0x1f8(%rsp) movl $0x0, 0x1c(%rsp) movq $0x0, 0x1f0(%rsp) xorl %r13d, %r13d movl $0x0, 0x18(%rsp) movq $0x0, 0x1e8(%rsp) pushq $0x6 popq %rdi callq 0x782bf movq %rax, %rbp leaq 0x268(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq %r12, %rdi xorl %esi, %esi movl 0x14(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r12, %rdi movl %ebx, %esi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r12, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r12, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r12, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r12, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r12, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r12, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r12, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x6 popq %rsi movl 0x6c(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r12, %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x110(%r14,%rax), %edx movq %r12, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x258(%rsp), %rdx movq %r12, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r12, %rsi callq 0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x28995e pushq $0x40 popq %rax xorps %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) movups %xmm0, 0x30(%rsp,%rax) movups %xmm0, 0x3c(%rsp,%rax) movups %xmm0, 0x50(%rsp,%rax) movups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x1a8, %rax # imm = 0x1A8 jne 0x2898f7 movq 0x208(%rsp), %rax testq %rax, %rax je 0x28992d lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2899da lock decl (%rax) jne 0x2899da movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2899d2 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2899da pushq $0x40 popq %rax xorps %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) movups %xmm0, 0x30(%rsp,%rax) movups %xmm0, 0x3c(%rsp,%rax) movups %xmm0, 0x50(%rsp,%rax) movups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x160, %rax # imm = 0x160 jne 0x289964 movq 0x208(%rsp), %rax testq %rax, %rax je 0x28999a lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x289d81 lock decl (%rax) jne 0x289d81 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x289d79 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289d81 movq %rsi, %rdi callq 0x5f3e0 movaps 0x200(%rsp), %xmm0 movaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) movups 0x228(%rsp), %xmm0 movups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x289a7a lock decl (%rax) jne 0x289a7a movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x289a72 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289a7a movq %rsi, %rdi callq 0x5f3e0 movq %r15, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq 0x1f8(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x1c(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x1f0(%rsp), %rax movq %rax, 0xd8(%rsp) movl %r13d, 0xe0(%rsp) movl 0x18(%rsp), %eax movl %eax, 0xe4(%rsp) movl %r13d, 0xe8(%rsp) movl %r13d, 0xec(%rsp) movl %r13d, 0xf0(%rsp) movq 0x1e8(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x289d03 andq $0x0, 0x60(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x20(%rsp) movups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax movups %xmm0, 0xc(%rax) movaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx movss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x289b7a movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x289b6c movq 0x28(%rsp), %rax testq %rax, %rax je 0x289b87 lock incl (%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x289bbe lock decl (%rax) jne 0x289bbe movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x289bb6 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289bbe movq %rsi, %rdi callq 0x5f3e0 movaps 0x20(%rsp), %xmm0 movaps %xmm0, 0x100(%rsp) movq 0x30(%rsp), %rax movq %rax, 0x110(%rsp) movl 0x38(%rsp), %eax movl %eax, 0x118(%rsp) movq 0x40(%rsp), %rax movq %rax, 0x120(%rsp) movups 0x48(%rsp), %xmm0 movups %xmm0, 0x128(%rsp) movl 0x58(%rsp), %eax movl %eax, 0x138(%rsp) movq 0x60(%rsp), %rax movq %rax, 0x140(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %r12d movq 0x260(%r14,%rax), %rbx movq 0x150(%rsp), %rax testq %rax, %rax je 0x289c7c lock decl (%rax) jne 0x289c7c movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x289c74 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289c7c movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %r12d, 0x160(%rsp) movq %rbx, 0x168(%rsp) movaps 0x16e784(%rip), %xmm0 # 0x3f8430 movaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x289cfc lock decl (%rax) jne 0x289cfc movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x289cf4 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289cfc movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x289f17 movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x198(%rsp), %rax testq %rax, %rax je 0x289ec6 lock decl (%rax) jne 0x289ec6 movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x289ebe movq (%rdi), %rax callq 0x18(%rax) jmp 0x289ec6 movq %rsi, %rdi callq 0x5f3e0 movaps 0x200(%rsp), %xmm0 movaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) movups 0x228(%rsp), %xmm0 movups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x28a0ef andq $0x0, 0x60(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x20(%rsp) movups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax movups %xmm0, 0xc(%rax) movaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx movss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x289e73 movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x289e65 movq 0x28(%rsp), %rax testq %rax, %rax je 0x289e80 lock incl (%rax) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x289fab lock decl (%rax) jne 0x289fab movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x289fa3 movq (%rdi), %rax callq 0x18(%rax) jmp 0x289fab movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x190(%rsp) andq $0x0, 0x198(%rsp) movq %r15, 0x1a0(%rsp) movl %ebx, 0x1a8(%rsp) movq %r12, 0x1b0(%rsp) movaps 0x16e53b(%rip), %xmm0 # 0x3f8430 movups %xmm0, 0x1b8(%rsp) movl $0x1, 0x1c8(%rsp) movq $0x1, 0x1d0(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0x120, %r15d # imm = 0x120 movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x289f75 lock decl (%rax) jne 0x289f75 movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x289f6d movq (%rdi), %rax callq 0x18(%rax) jmp 0x289f75 movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x289f44 jmp 0x28a236 movq %rsi, %rdi callq 0x5f3e0 movaps 0x20(%rsp), %xmm0 movups %xmm0, 0xb8(%rsp) movq 0x30(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x38(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x40(%rsp), %rax movq %rax, 0xd8(%rsp) movups 0x48(%rsp), %xmm0 movaps %xmm0, 0xe0(%rsp) movl 0x58(%rsp), %eax movl %eax, 0xf0(%rsp) movq 0x60(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %ebx movq 0x260(%r14,%rax), %r12 movq 0x108(%rsp), %rax testq %rax, %rax je 0x28a069 lock decl (%rax) jne 0x28a069 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x28a061 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28a069 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x100(%rsp) andq $0x0, 0x108(%rsp) movq %r15, 0x110(%rsp) movl %ebx, 0x118(%rsp) movq %r12, 0x120(%rsp) movaps 0x16e398(%rip), %xmm0 # 0x3f8430 movups %xmm0, 0x128(%rsp) movl $0x1, 0x138(%rsp) movq $0x1, 0x140(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x28a0e8 lock decl (%rax) jne 0x28a0e8 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x28a0e0 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28a0e8 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x28a1af movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x150(%rsp), %rax testq %rax, %rax je 0x28a15e lock decl (%rax) jne 0x28a15e movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x28a156 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28a15e movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %ebx, 0x160(%rsp) movq %r12, 0x168(%rsp) movaps 0x16e2a3(%rip), %xmm0 # 0x3f8430 movaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0xd8, %r15d movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x28a20d lock decl (%rax) jne 0x28a20d movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x28a205 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28a20d movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x28a1dc movq (%rbp), %rax movq %rbp, %rdi movq 0x250(%rsp), %rsi callq 0x20(%rax) leaq 0x268(%rsp), %rdi movq 0x1e0(%rsp), %rax movq (%rax), %rax movq 0x8(%rsp), %rcx movq %rbp, (%rax,%rcx,8) callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x28a2a0 lock decl (%rax) jne 0x28a2a0 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi je 0x28a298 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28a2a0 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rdx incq %rdx jmp 0x2895ed xorl %eax, %eax addq $0x278, %rsp # imm = 0x278 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x28a375 jmp 0x28a500 jmp 0x28a2cd movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x28a387 lock decl (%rax) jne 0x28a387 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x28a303 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28a387 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28a387 jmp 0x28a500 jmp 0x28a3f9 jmp 0x28a375 jmp 0x28a319 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x28a40b lock decl (%rax) jne 0x28a40b movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x28a34f movq %rsi, %rdi callq 0x5f3e0 jmp 0x28a40b movq (%rdi), %rax callq 0x18(%rax) jmp 0x28a40b jmp 0x28a500 jmp 0x28a500 jmp 0x28a500 jmp 0x28a500 jmp 0x28a375 jmp 0x28a3f9 movq %rax, %rbx jmp 0x28a387 movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0xd8, %r14d xorps %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x28a3c7 lock decl (%rax) jne 0x28a3c7 movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x28a3bc movq (%rdi), %rax callq 0x18(%rax) xorps %xmm0, %xmm0 jmp 0x28a3c7 movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x28a390 jmp 0x28a4b4 jmp 0x28a500 jmp 0x28a3f9 movq %rax, %rbx jmp 0x28a40b movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0x120, %r14d # imm = 0x120 xorps %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x28a44b lock decl (%rax) jne 0x28a44b movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x28a440 movq (%rdi), %rax callq 0x18(%rax) xorps %xmm0, %xmm0 jmp 0x28a44b movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x28a414 jmp 0x28a4b4 jmp 0x28a500 jmp 0x28a4a8 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x28a4f8 lock decl (%rax) jne 0x28a4f8 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x28a4e8 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28a4f8 jmp 0x28a500 movq %rax, %rbx jmp 0x28a4c1 jmp 0x28a500 jmp 0x28a500 movq %rax, %rbx leaq 0x268(%rsp), %rdi callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x28a4f8 lock decl (%rax) jne 0x28a4f8 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi jne 0x28a4f2 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28a4f8 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86::create_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %rax movq (%rdi), %rax addq -0x30(%rax), %rdi callq 0x288da6 xorl %eax, %eax popq %rcx retq nop	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
ncnn::ConvolutionDepthWise_x86::destroy_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }	pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x28a54b movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x28a546 movq (%rdi), %rax callq 0x8(%rax) andq $0x0, 0x8(%rbx) xorl %r15d, %r15d movq 0x10(%rbx), %rax movq 0x18(%rbx), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r15 jge 0x28a58d movq (%rax,%r15,8), %rdi movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x10(%rbx), %rax movq (%rax,%r15,8), %rdi testq %rdi, %rdi je 0x28a588 movq (%rdi), %rax callq 0x8(%rax) incq %r15 jmp 0x28a54e cmpq %rax, %rcx je 0x28a596 movq %rax, 0x18(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %r15 retq	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86::destroy_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }	pushq %rax movq (%rdi), %rax addq -0x38(%rax), %rdi callq 0x28a51a xorl %eax, %eax popq %rcx retq nop	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
ncnn::ConvolutionDepthWise_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float)bias_data) + g 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float)bias_data) + g 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float)bias_data) + g 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x2c8, %rsp # imm = 0x2C8 movq %rcx, %r14 movq (%rdi), %rax movq -0x18(%rax), %rcx cmpb $0x1, 0x1e(%r14) jne 0x28a5f5 cmpl $0x0, 0x10c(%rdi,%rcx) je 0x28a5f5 movq %r14, %rcx addq $0x2c8, %rsp # imm = 0x2C8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x28d5aa movq %rdx, 0x48(%rsp) movl 0x38(%rsi), %edx movq %rdx, 0x50(%rsp) movq 0x10(%rsi), %r13 movslq 0x18(%rsi), %r15 movl 0xd4(%rdi,%rcx), %ebx decl %ebx imull 0xdc(%rdi,%rcx), %ebx movq %rdi, %r8 movl 0xd8(%rdi,%rcx), %ebp decl %ebp imull 0xe0(%rdi,%rcx), %ebp leaq 0xb0(%rsp), %rdx andq $0x0, 0x40(%rdx) xorps %xmm0, %xmm0 movaps %xmm0, (%rdx) movups %xmm0, 0xc(%rdx) movaps %xmm0, 0x20(%rdx) movups %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi movq %r8, 0x88(%rsp) addq %r8, %rdi movq %r14, %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0xb0(%rsp) je 0x28d3c9 movslq 0xe8(%rsp), %rax imulq 0xf0(%rsp), %rax testq %rax, %rax je 0x28d3c9 movq %r15, %r10 notl %ebx movl 0xdc(%rsp), %eax movl %eax, 0x10(%rsp) addl %eax, %ebx movq 0x88(%rsp), %rsi movq (%rsi), %rax movq -0x18(%rax), %rcx movl %ebx, %eax cltd idivl 0xe4(%rsi,%rcx) movl %eax, %edi notl %ebp addl 0xe0(%rsp), %ebp movl %ebp, %eax movq %rsi, %rbp cltd idivl 0xe8(%rsi,%rcx) movq %rdi, 0xa0(%rsp) leal 0x1(%rdi), %esi movq %rax, 0x20(%rsp) leal 0x1(%rax), %r8d movl 0xd0(%rbp,%rcx), %edi testb $0x3, %dil sete %al andb 0x27(%r14), %al movzbl %al, %r15d leal (%r15,%r15,2), %r9d incl %r9d movq %r13, %rax xorl %edx, %edx movq %r10, %rbx divq %r10 movq %rax, %r13 addb %r15b, %r15b movl %r15d, %ecx shlq %cl, %r13 movl %edi, %eax cltd idivl %r9d movq %r14, 0x190(%rsp) movq 0x8(%r14), %rcx movq %rcx, (%rsp) movq 0x48(%rsp), %r14 movq %r14, %rdi movl %esi, 0x60(%rsp) movl %r8d, 0x70(%rsp) movl %r8d, %edx movl %eax, %ecx movq %r13, %r8 movl %r9d, 0x30(%rsp) callq 0x628f2 movq (%r14), %r10 testq %r10, %r10 je 0x28d3c9 movq %r13, %r11 movq 0x40(%r14), %r13 movslq 0x38(%r14), %rax imulq %r13, %rax testq %rax, %rax je 0x28d3c9 movq %rbx, %r9 movl %r9d, %eax imull 0x50(%rsp), %eax movq (%rbp), %rcx movq -0x18(%rcx), %r8 movl 0xd0(%rbp,%r8), %ecx movl 0x108(%rbp,%r8), %esi cmpl %esi, %eax jne 0x28beda cmpl %ecx, %eax jne 0x28beda cmpl $0x1, %r9d je 0x28b2ef movl %eax, %ecx cmpl $0x4, %r9d jne 0x28beda movl 0xd4(%rbp,%r8), %eax movl 0xd8(%rbp,%r8), %r15d cmpl $0x5, %eax je 0x28b672 cmpl $0x3, %eax jne 0x28ccf0 cmpl $0x3, %r15d jne 0x28ccf0 cmpl $0x1, 0xdc(%rbp,%r8) jne 0x28c434 cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28c434 cmpl $0x1, 0xe4(%rbp,%r8) jne 0x28c434 cmpl $0x1, 0xe8(%rbp,%r8) jne 0x28c434 movl 0x2c(%r14), %eax movl 0x30(%r14), %ecx movl 0xe8(%rsp), %edx movq 0x1b0(%rbp,%r8), %rsi xorl %edi, %edi testl %ecx, %ecx cmovlel %edi, %ecx testl %edx, %edx cmovlel %edi, %edx movq %rdx, 0x1e8(%rsp) movl $0x80, %r8d movq %rsi, 0x1e0(%rsp) cmpq 0x1e8(%rsp), %rdi je 0x28d415 testq %rsi, %rsi je 0x28a884 movq %rdi, %rdx shlq $0x4, %rdx movups (%rsi,%rdx), %xmm0 jmp 0x28a887 xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movq 0x40(%r14), %r9 imulq %rdi, %r9 imulq 0x10(%r14), %r9 addq (%r14), %r9 movq 0x28(%rbp), %rdx movslq 0x54(%rbp), %rsi imulq %rdi, %rsi imulq 0x38(%rbp), %rsi movslq 0xdc(%rsp), %r10 movq 0xf0(%rsp), %r13 movq %rdi, %r14 imulq %rdi, %r13 movq 0xc0(%rsp), %r11 imulq %r11, %r13 addq 0xb0(%rsp), %r13 imulq %r11, %r10 leaq (%r10,%r13), %rbp leaq (,%r10,2), %r11 addq %r13, %r11 movaps (%rdx,%rsi), %xmm0 movaps %xmm0, 0x20(%rsp) movaps 0x10(%rdx,%rsi), %xmm0 movaps %xmm0, 0x50(%rsp) movaps 0x20(%rdx,%rsi), %xmm0 movaps %xmm0, 0x110(%rsp) movaps 0x30(%rdx,%rsi), %xmm0 movaps %xmm0, 0x90(%rsp) movaps 0x40(%rdx,%rsi), %xmm0 movaps %xmm0, 0x30(%rsp) movaps 0x50(%rdx,%rsi), %xmm0 movaps %xmm0, 0xa0(%rsp) movaps 0x60(%rdx,%rsi), %xmm0 movaps %xmm0, 0x60(%rsp) movaps 0x70(%rdx,%rsi), %xmm0 movaps %xmm0, 0x100(%rsp) movaps 0x80(%rdx,%rsi), %xmm0 movaps %xmm0, 0x70(%rsp) xorl %r10d, %r10d cmpl %ecx, %r10d je 0x28b2cf leaq 0x20(%r13), %r15 leaq 0x20(%rbp), %rbx leaq 0x20(%r11), %rdx xorl %esi, %esi xorl %r12d, %r12d leal 0x7(%r12), %edi cmpl %eax, %edi jge 0x28b0bf movaps (%r13,%rsi), %xmm0 mulps 0x20(%rsp), %xmm0 addps 0x10(%rsp), %xmm0 movaps 0x10(%r13,%rsi), %xmm3 movaps 0x20(%r13,%rsi), %xmm1 movaps 0x30(%r13,%rsi), %xmm9 movaps %xmm9, 0x180(%rsp) movaps %xmm3, %xmm2 mulps 0x50(%rsp), %xmm2 movaps %xmm1, %xmm4 movaps 0x110(%rsp), %xmm8 mulps %xmm8, %xmm4 addps %xmm2, %xmm4 addps %xmm0, %xmm4 movaps (%rbp,%rsi), %xmm12 movaps 0x90(%rsp), %xmm7 mulps %xmm7, %xmm12 movaps 0x10(%rbp,%rsi), %xmm0 movaps 0x20(%rbp,%rsi), %xmm2 movaps 0x30(%rbp,%rsi), %xmm5 movaps %xmm5, 0x120(%rsp) movaps %xmm0, %xmm13 mulps 0x30(%rsp), %xmm13 addps %xmm12, %xmm13 movaps %xmm2, %xmm14 mulps 0xa0(%rsp), %xmm14 addps %xmm13, %xmm14 addps %xmm4, %xmm14 movaps (%r11,%rsi), %xmm15 movaps 0x60(%rsp), %xmm11 mulps %xmm11, %xmm15 movaps 0x10(%r11,%rsi), %xmm5 movaps 0x20(%r11,%rsi), %xmm4 movaps 0x30(%r11,%rsi), %xmm12 movaps %xmm5, %xmm13 movaps 0x100(%rsp), %xmm10 mulps %xmm10, %xmm13 addps %xmm15, %xmm13 movaps %xmm4, %xmm15 mulps 0x70(%rsp), %xmm15 addps %xmm13, %xmm15 addps %xmm14, %xmm15 movaps %xmm15, (%r9,%rsi) mulps 0x20(%rsp), %xmm3 addps 0x10(%rsp), %xmm3 movaps %xmm1, %xmm13 movaps 0x50(%rsp), %xmm15 mulps %xmm15, %xmm13 movaps %xmm9, %xmm14 mulps %xmm8, %xmm14 mulps %xmm7, %xmm0 addps %xmm13, %xmm0 addps %xmm3, %xmm0 movaps %xmm2, %xmm3 mulps 0x30(%rsp), %xmm3 movaps 0x120(%rsp), %xmm9 movaps %xmm9, %xmm13 movaps 0xa0(%rsp), %xmm6 mulps %xmm6, %xmm13 addps %xmm14, %xmm13 mulps %xmm11, %xmm5 addps %xmm3, %xmm5 movaps %xmm4, %xmm3 mulps %xmm10, %xmm3 addps %xmm5, %xmm3 addps %xmm0, %xmm3 movaps %xmm12, %xmm5 movaps 0x70(%rsp), %xmm10 mulps %xmm10, %xmm5 addps %xmm13, %xmm5 addps %xmm3, %xmm5 movaps 0x40(%r13,%rsi), %xmm0 movaps 0x40(%rbp,%rsi), %xmm3 movaps 0x40(%r11,%rsi), %xmm13 movaps %xmm5, 0x10(%r9,%rsi) mulps 0x20(%rsp), %xmm1 addps 0x10(%rsp), %xmm1 movaps 0x180(%rsp), %xmm11 movaps %xmm11, %xmm5 mulps %xmm15, %xmm5 movaps %xmm0, %xmm14 mulps %xmm8, %xmm14 mulps %xmm7, %xmm2 movaps 0x30(%rsp), %xmm8 mulps %xmm8, %xmm9 addps %xmm5, %xmm9 movaps %xmm3, %xmm5 mulps %xmm6, %xmm5 addps %xmm14, %xmm5 mulps 0x60(%rsp), %xmm4 addps %xmm2, %xmm4 addps %xmm1, %xmm4 movaps %xmm12, %xmm1 movaps 0x100(%rsp), %xmm6 mulps %xmm6, %xmm1 addps %xmm9, %xmm1 addps %xmm4, %xmm1 movaps %xmm13, %xmm2 mulps %xmm10, %xmm2 addps %xmm5, %xmm2 addps %xmm1, %xmm2 movaps 0x50(%r13,%rsi), %xmm4 movaps 0x50(%rbp,%rsi), %xmm1 movaps 0x50(%r11,%rsi), %xmm14 movaps %xmm2, 0x20(%r9,%rsi) movaps %xmm11, %xmm2 mulps 0x20(%rsp), %xmm2 addps 0x10(%rsp), %xmm2 movaps %xmm2, %xmm9 movaps %xmm0, %xmm2 mulps 0x50(%rsp), %xmm2 movaps %xmm4, %xmm5 mulps 0x110(%rsp), %xmm5 movaps 0x120(%rsp), %xmm10 mulps %xmm7, %xmm10 movaps %xmm3, %xmm15 mulps %xmm8, %xmm15 addps %xmm2, %xmm15 movaps %xmm1, %xmm2 movaps %xmm1, %xmm8 mulps 0xa0(%rsp), %xmm2 addps %xmm5, %xmm2 movaps 0x60(%rsp), %xmm7 mulps %xmm7, %xmm12 addps %xmm10, %xmm12 addps %xmm9, %xmm12 movaps %xmm13, %xmm5 mulps %xmm6, %xmm5 addps %xmm15, %xmm5 addps %xmm12, %xmm5 movaps %xmm14, %xmm10 movaps 0x70(%rsp), %xmm6 mulps %xmm6, %xmm10 addps %xmm2, %xmm10 addps %xmm5, %xmm10 movaps 0x60(%r13,%rsi), %xmm15 movaps 0x60(%rbp,%rsi), %xmm1 movaps %xmm1, 0x120(%rsp) movaps 0x60(%r11,%rsi), %xmm12 movaps %xmm10, 0x30(%r9,%rsi) mulps 0x20(%rsp), %xmm0 addps 0x10(%rsp), %xmm0 movaps %xmm4, %xmm5 movaps 0x50(%rsp), %xmm9 mulps %xmm9, %xmm5 movaps %xmm15, %xmm10 movaps 0x110(%rsp), %xmm2 mulps %xmm2, %xmm10 mulps 0x90(%rsp), %xmm3 movaps %xmm8, %xmm11 mulps 0x30(%rsp), %xmm11 addps %xmm5, %xmm11 movaps %xmm1, %xmm5 mulps 0xa0(%rsp), %xmm5 addps %xmm10, %xmm5 mulps %xmm7, %xmm13 addps %xmm3, %xmm13 addps %xmm0, %xmm13 movaps %xmm14, %xmm0 mulps 0x100(%rsp), %xmm0 addps %xmm11, %xmm0 addps %xmm13, %xmm0 movaps %xmm12, %xmm3 mulps %xmm6, %xmm3 addps %xmm5, %xmm3 addps %xmm0, %xmm3 movaps 0x70(%r13,%rsi), %xmm11 movaps 0x70(%rbp,%rsi), %xmm13 movaps 0x70(%r11,%rsi), %xmm10 movaps %xmm3, 0x40(%r9,%rsi) mulps 0x20(%rsp), %xmm4 addps 0x10(%rsp), %xmm4 movaps %xmm15, %xmm0 mulps %xmm9, %xmm0 movaps %xmm9, %xmm7 movaps %xmm11, %xmm3 mulps %xmm2, %xmm3 movaps 0x90(%rsp), %xmm6 mulps %xmm6, %xmm8 movaps 0x120(%rsp), %xmm5 movaps 0x30(%rsp), %xmm2 mulps %xmm2, %xmm5 addps %xmm0, %xmm5 movaps %xmm13, %xmm0 movaps %xmm13, 0x180(%rsp) movaps 0xa0(%rsp), %xmm9 mulps %xmm9, %xmm0 addps %xmm3, %xmm0 mulps 0x60(%rsp), %xmm14 addps %xmm8, %xmm14 addps %xmm4, %xmm14 movaps %xmm12, %xmm1 movaps 0x100(%rsp), %xmm8 mulps %xmm8, %xmm1 addps %xmm5, %xmm1 addps %xmm14, %xmm1 movaps %xmm10, %xmm4 mulps 0x70(%rsp), %xmm4 addps %xmm0, %xmm4 addps %xmm1, %xmm4 movaps 0x80(%r13,%rsi), %xmm3 movaps 0x80(%rbp,%rsi), %xmm1 movaps 0x80(%r11,%rsi), %xmm0 movaps %xmm4, 0x50(%r9,%rsi) movaps %xmm11, %xmm4 mulps %xmm7, %xmm4 movaps %xmm3, %xmm5 movaps 0x110(%rsp), %xmm7 mulps %xmm7, %xmm5 movaps %xmm13, %xmm14 mulps %xmm2, %xmm14 addps %xmm4, %xmm14 movaps %xmm1, %xmm4 mulps %xmm9, %xmm4 addps %xmm5, %xmm4 movaps 0x120(%rsp), %xmm2 mulps %xmm6, %xmm2 movaps 0x60(%rsp), %xmm13 mulps %xmm13, %xmm12 addps %xmm2, %xmm12 mulps 0x20(%rsp), %xmm15 addps 0x10(%rsp), %xmm15 addps %xmm15, %xmm12 movaps %xmm10, %xmm2 mulps %xmm8, %xmm2 addps %xmm14, %xmm2 addps %xmm12, %xmm2 movaps %xmm0, %xmm5 movaps 0x70(%rsp), %xmm6 mulps %xmm6, %xmm5 addps %xmm4, %xmm5 addps %xmm2, %xmm5 movaps 0x90(%r13,%rsi), %xmm2 mulps %xmm7, %xmm2 movaps 0x90(%rbp,%rsi), %xmm4 mulps %xmm9, %xmm4 addps %xmm2, %xmm4 movaps 0x90(%r11,%rsi), %xmm2 mulps %xmm6, %xmm2 addps %xmm4, %xmm2 movaps %xmm5, 0x60(%r9,%rsi) mulps 0x50(%rsp), %xmm3 mulps 0x30(%rsp), %xmm1 addps %xmm3, %xmm1 movaps 0x180(%rsp), %xmm3 mulps 0x90(%rsp), %xmm3 mulps %xmm13, %xmm10 addps %xmm3, %xmm10 mulps 0x20(%rsp), %xmm11 addps 0x10(%rsp), %xmm11 addps %xmm11, %xmm10 mulps %xmm8, %xmm0 addps %xmm1, %xmm0 addps %xmm10, %xmm0 addps %xmm2, %xmm0 movaps %xmm0, 0x70(%r9,%rsi) addl $0x8, %r12d addq %r8, %rsi addq %r8, %r15 addq %r8, %rbx addq %r8, %rdx jmp 0x28a970 movaps (%r13,%rsi), %xmm0 mulps 0x20(%rsp), %xmm0 addps 0x10(%rsp), %xmm0 movaps 0x10(%r13,%rsi), %xmm3 movaps 0x20(%r13,%rsi), %xmm13 movaps 0x30(%r13,%rsi), %xmm11 movaps %xmm3, %xmm1 mulps 0x50(%rsp), %xmm1 movaps %xmm13, %xmm4 movaps 0x110(%rsp), %xmm2 mulps %xmm2, %xmm4 movaps %xmm2, %xmm7 addps %xmm1, %xmm4 addps %xmm0, %xmm4 movaps (%rbp,%rsi), %xmm1 mulps 0x90(%rsp), %xmm1 movaps 0x10(%rbp,%rsi), %xmm2 movaps 0x20(%rbp,%rsi), %xmm8 movaps 0x30(%rbp,%rsi), %xmm9 movaps %xmm2, %xmm5 mulps 0x30(%rsp), %xmm5 addps %xmm1, %xmm5 movaps %xmm8, %xmm14 movaps 0xa0(%rsp), %xmm6 mulps %xmm6, %xmm14 addps %xmm5, %xmm14 addps %xmm4, %xmm14 movaps (%r11,%rsi), %xmm5 mulps 0x60(%rsp), %xmm5 movaps 0x10(%r11,%rsi), %xmm4 movaps 0x20(%r11,%rsi), %xmm1 movaps 0x30(%r11,%rsi), %xmm12 movaps %xmm4, %xmm15 movaps 0x100(%rsp), %xmm10 mulps %xmm10, %xmm15 addps %xmm5, %xmm15 movaps %xmm1, %xmm5 movaps 0x70(%rsp), %xmm0 mulps %xmm0, %xmm5 addps %xmm15, %xmm5 addps %xmm14, %xmm5 movaps %xmm5, (%r9,%rsi) mulps 0x20(%rsp), %xmm3 addps 0x10(%rsp), %xmm3 movaps %xmm13, %xmm5 movaps 0x50(%rsp), %xmm15 mulps %xmm15, %xmm5 movaps %xmm11, %xmm14 mulps %xmm7, %xmm14 movaps 0x90(%rsp), %xmm7 mulps %xmm7, %xmm2 addps %xmm5, %xmm2 addps %xmm3, %xmm2 movaps %xmm8, %xmm3 movaps 0x30(%rsp), %xmm0 mulps %xmm0, %xmm3 movaps %xmm9, %xmm5 movaps %xmm9, 0x120(%rsp) mulps %xmm6, %xmm5 addps %xmm14, %xmm5 movaps 0x60(%rsp), %xmm6 mulps %xmm6, %xmm4 addps %xmm3, %xmm4 movaps %xmm1, %xmm3 mulps %xmm10, %xmm3 addps %xmm4, %xmm3 addps %xmm2, %xmm3 movaps %xmm12, %xmm14 mulps 0x70(%rsp), %xmm14 addps %xmm5, %xmm14 addps %xmm3, %xmm14 movaps 0x40(%r13,%rsi), %xmm4 movaps 0x40(%rbp,%rsi), %xmm3 movaps 0x40(%r11,%rsi), %xmm2 movaps %xmm14, 0x10(%r9,%rsi) movaps %xmm11, %xmm5 mulps %xmm15, %xmm5 movaps %xmm4, %xmm14 movaps 0x110(%rsp), %xmm10 mulps %xmm10, %xmm14 movaps %xmm9, %xmm15 mulps %xmm0, %xmm15 addps %xmm5, %xmm15 movaps %xmm3, %xmm5 movaps 0xa0(%rsp), %xmm9 mulps %xmm9, %xmm5 addps %xmm14, %xmm5 movaps 0x10(%rsp), %xmm14 mulps %xmm7, %xmm8 mulps %xmm6, %xmm1 addps %xmm8, %xmm1 mulps 0x20(%rsp), %xmm13 addps %xmm14, %xmm13 addps %xmm13, %xmm1 movaps %xmm12, %xmm0 movaps 0x100(%rsp), %xmm6 mulps %xmm6, %xmm0 addps %xmm15, %xmm0 addps %xmm1, %xmm0 movaps %xmm2, %xmm1 movaps 0x70(%rsp), %xmm7 mulps %xmm7, %xmm1 addps %xmm5, %xmm1 addps %xmm0, %xmm1 movaps 0x50(%r13,%rsi), %xmm0 mulps %xmm10, %xmm0 movaps 0x50(%rbp,%rsi), %xmm5 mulps %xmm9, %xmm5 addps %xmm0, %xmm5 movaps 0x50(%r11,%rsi), %xmm0 mulps %xmm7, %xmm0 addps %xmm5, %xmm0 movaps %xmm1, 0x20(%r9,%rsi) mulps 0x50(%rsp), %xmm4 mulps 0x30(%rsp), %xmm3 addps %xmm4, %xmm3 movaps 0x120(%rsp), %xmm1 mulps 0x90(%rsp), %xmm1 mulps 0x60(%rsp), %xmm12 addps %xmm1, %xmm12 mulps 0x20(%rsp), %xmm11 movaps %xmm14, 0x10(%rsp) addps %xmm14, %xmm11 addps %xmm11, %xmm12 mulps %xmm6, %xmm2 addps %xmm3, %xmm2 addps %xmm12, %xmm2 addps %xmm0, %xmm2 movaps %xmm2, 0x30(%r9,%rsi) addl $0x4, %r12d addq $0x40, %rsi addq $0x40, %r15 addq $0x40, %rbx addq $0x40, %rdx leal 0x3(%r12), %edi cmpl %eax, %edi jl 0x28ae58 jmp 0x28b20f movaps 0x10(%r13,%rsi), %xmm4 movaps 0x20(%r13,%rsi), %xmm10 movaps 0x10(%rbp,%rsi), %xmm0 movaps 0x20(%rbp,%rsi), %xmm6 movaps 0x10(%r11,%rsi), %xmm3 movaps 0x20(%r11,%rsi), %xmm1 movaps (%r13,%rsi), %xmm5 mulps 0x20(%rsp), %xmm5 addps 0x10(%rsp), %xmm5 movaps %xmm4, %xmm11 mulps 0x50(%rsp), %xmm11 movaps %xmm10, %xmm12 movaps 0x110(%rsp), %xmm7 mulps %xmm7, %xmm12 addps %xmm11, %xmm12 addps %xmm5, %xmm12 movaps (%rbp,%rsi), %xmm5 movaps 0x90(%rsp), %xmm14 mulps %xmm14, %xmm5 movaps %xmm0, %xmm11 mulps 0x30(%rsp), %xmm11 addps %xmm5, %xmm11 movaps %xmm6, %xmm5 movaps 0xa0(%rsp), %xmm15 mulps %xmm15, %xmm5 addps %xmm11, %xmm5 addps %xmm12, %xmm5 movaps (%r11,%rsi), %xmm11 movaps 0x60(%rsp), %xmm8 mulps %xmm8, %xmm11 movaps %xmm3, %xmm12 movaps 0x100(%rsp), %xmm9 mulps %xmm9, %xmm12 addps %xmm11, %xmm12 movaps %xmm1, %xmm13 movaps 0x70(%rsp), %xmm2 mulps %xmm2, %xmm13 addps %xmm12, %xmm13 addps %xmm5, %xmm13 movaps 0x30(%r13,%rsi), %xmm5 mulps %xmm7, %xmm5 movaps 0x30(%rbp,%rsi), %xmm12 mulps %xmm15, %xmm12 addps %xmm5, %xmm12 movaps 0x30(%r11,%rsi), %xmm11 mulps %xmm2, %xmm11 addps %xmm12, %xmm11 movaps %xmm13, (%r9,%rsi) mulps 0x50(%rsp), %xmm10 mulps %xmm14, %xmm0 addps %xmm10, %xmm0 mulps 0x20(%rsp), %xmm4 addps 0x10(%rsp), %xmm4 addps %xmm4, %xmm0 mulps 0x30(%rsp), %xmm6 mulps %xmm8, %xmm3 addps %xmm6, %xmm3 mulps %xmm9, %xmm1 addps %xmm3, %xmm1 addps %xmm0, %xmm1 addps %xmm11, %xmm1 movaps %xmm1, 0x10(%r9,%rsi) addl $0x2, %r12d addq $0x20, %rsi addq $0x20, %r15 addq $0x20, %rbx addq $0x20, %rdx leal 0x1(%r12), %edi cmpl %eax, %edi jl 0x28b0d1 addq %rsi, %r9 cmpl %eax, %r12d jge 0x28b2be movaps -0x20(%r15), %xmm0 mulps 0x20(%rsp), %xmm0 addps 0x10(%rsp), %xmm0 movaps -0x10(%r15), %xmm1 mulps 0x50(%rsp), %xmm1 movaps (%r15), %xmm2 mulps 0x110(%rsp), %xmm2 addps %xmm1, %xmm2 addps %xmm0, %xmm2 movaps -0x20(%rbx), %xmm0 mulps 0x90(%rsp), %xmm0 movaps -0x10(%rbx), %xmm1 mulps 0x30(%rsp), %xmm1 addps %xmm0, %xmm1 movaps (%rbx), %xmm0 mulps 0xa0(%rsp), %xmm0 addps %xmm1, %xmm0 addps %xmm2, %xmm0 movaps -0x20(%rdx), %xmm1 mulps 0x60(%rsp), %xmm1 movaps -0x10(%rdx), %xmm2 mulps 0x100(%rsp), %xmm2 addps %xmm1, %xmm2 movaps (%rdx), %xmm1 mulps 0x70(%rsp), %xmm1 addps %xmm2, %xmm1 addps %xmm0, %xmm1 movaps %xmm1, (%r9) addq $0x10, %r9 incl %r12d addq $0x10, %r15 addq $0x10, %rbx addq $0x10, %rdx jmp 0x28b21f incl %r10d movq %r15, %r13 movq %rbx, %rbp movq %rdx, %r11 jmp 0x28a956 movq %r14, %rdi incq %rdi movq 0x48(%rsp), %r14 movq 0x88(%rsp), %rbp movq 0x1e0(%rsp), %rsi jmp 0x28a864 cmpl $0x3, 0xd4(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x3, 0xd8(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x1, 0xdc(%rbp,%r8) movq %r10, 0x20(%rsp) jne 0x28bd03 cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28bd03 cmpl $0x1, 0xe4(%rbp,%r8) jne 0x28bd03 cmpl $0x1, 0xe8(%rbp,%r8) jne 0x28bd03 movslq 0xdc(%rsp), %rax movl 0x2c(%r14), %edx movl 0x30(%r14), %edi movl 0xe8(%rsp), %ecx movq 0x28(%rbp), %r9 imulq 0x10(%r14), %r13 movq 0x1b0(%rbp,%r8), %rsi movslq %edx, %r8 movq 0xb0(%rsp), %r11 movq %r11, 0x110(%rsp) movq 0xc0(%rsp), %r11 imulq 0xf0(%rsp), %r11 movq %r11, 0x100(%rsp) leaq (%rax,%rax), %r11 movq %r11, 0x120(%rsp) leaq (%rax,%rax,2), %r11 movq %r11, 0x180(%rsp) movq %rax, 0x70(%rsp) addl $0x2, %eax movslq %eax, %rbp xorl %eax, %eax testl %ecx, %ecx cmovlel %eax, %ecx movq %rcx, 0x50(%rsp) shlq $0x2, %rbp movq %r8, 0xa0(%rsp) leaq (,%r8,4), %rcx movq %rcx, 0x10(%rsp) movq %rsi, 0x90(%rsp) movq %r13, 0x60(%rsp) cmpq 0x50(%rsp), %rax je 0x28d41b testq %rsi, %rsi je 0x28b412 movss (%rsi,%rax,4), %xmm0 jmp 0x28b415 xorps %xmm0, %xmm0 movq 0xa0(%rsp), %r14 movq %r13, %rsi imulq %rax, %rsi addq %r10, %rsi imulq $0x24, %rax, %r13 movq 0x100(%rsp), %r15 movq %rax, 0x30(%rsp) imulq %rax, %r15 addq 0x110(%rsp), %r15 movq 0x70(%rsp), %rax leaq (%r15,%rax,4), %rax movq 0x120(%rsp), %rcx leaq (%r15,%rcx,4), %r8 movq 0x180(%rsp), %rcx leaq (%r15,%rcx,4), %r11 xorl %r10d, %r10d movq %rsi, %rbx movl %r10d, %ecx orl $0x1, %ecx cmpl %edi, %ecx jge 0x28b645 leaq (%rsi,%r14,4), %rsi xorl %r12d, %r12d movl %edx, %ecx testl %ecx, %ecx jle 0x28b582 movsd 0x4(%r15,%r12), %xmm3 movsd 0x4(%rax,%r12), %xmm6 movsd (%r8,%r12), %xmm5 movups (%r9,%r13), %xmm1 movups 0x10(%r9,%r13), %xmm2 movss (%r15,%r12), %xmm4 unpcklps %xmm3, %xmm4 # xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1] movss (%rax,%r12), %xmm7 movaps %xmm7, %xmm8 shufps $0xd4, %xmm3, %xmm8 # xmm8 = xmm8[0,1],xmm3[1,3] shufps $0x24, %xmm8, %xmm4 # xmm4 = xmm4[0,1],xmm8[2,0] movaps %xmm6, %xmm9 movlhps %xmm5, %xmm9 # xmm9 = xmm9[0],xmm5[0] mulps %xmm1, %xmm4 mulps %xmm2, %xmm9 addps %xmm4, %xmm9 movss 0x8(%r8,%r12), %xmm8 movss 0x20(%r9,%r13), %xmm3 movaps %xmm3, %xmm4 mulss %xmm8, %xmm4 movaps %xmm9, %xmm10 unpckhpd %xmm9, %xmm10 # xmm10 = xmm10[1],xmm9[1] addps %xmm9, %xmm10 movaps %xmm10, %xmm9 shufps $0x55, %xmm10, %xmm9 # xmm9 = xmm9[1,1],xmm10[1,1] addss %xmm10, %xmm9 addss %xmm0, %xmm4 addss %xmm9, %xmm4 movsd (%r11,%r12), %xmm9 movaps %xmm1, %xmm10 unpcklps %xmm6, %xmm10 # xmm10 = xmm10[0],xmm6[0],xmm10[1],xmm6[1] shufps $0xd4, %xmm5, %xmm8 # xmm8 = xmm8[0,1],xmm5[1,3] shufps $0xd4, %xmm6, %xmm5 # xmm5 = xmm5[0,1],xmm6[1,3] shufps $0x24, %xmm5, %xmm10 # xmm10 = xmm10[0,1],xmm5[2,0] shufps $0x42, %xmm9, %xmm8 # xmm8 = xmm8[2,0],xmm9[0,1] movss %xmm7, %xmm1 # xmm1 = xmm7[0],xmm1[1,2,3] mulps %xmm10, %xmm1 mulps %xmm2, %xmm8 addps %xmm1, %xmm8 mulss 0x8(%r11,%r12), %xmm3 movaps %xmm8, %xmm1 unpckhpd %xmm8, %xmm1 # xmm1 = xmm1[1],xmm8[1] addps %xmm8, %xmm1 movaps %xmm1, %xmm2 shufps $0x55, %xmm1, %xmm2 # xmm2 = xmm2[1,1],xmm1[1,1] addss %xmm1, %xmm2 addss %xmm0, %xmm3 addss %xmm2, %xmm3 movss %xmm4, (%rbx,%r12) movss %xmm3, (%rsi,%r12) decl %ecx addq $0x4, %r12 jmp 0x28b482 addq %rbp, %r15 addq %r12, %r15 addq %rbp, %rax addq %r12, %rax addq %rbp, %r8 addq %r12, %r8 addq %rbp, %r11 addq %r12, %r11 addq 0x10(%rsp), %rbx addq %r12, %rbx addl $0x2, %r10d addq %r12, %rsi jmp 0x28b46b xorl %ecx, %ecx movl %edx, %esi testl %esi, %esi jle 0x28b62a movsd 0x4(%r15,%rcx), %xmm1 movups (%r9,%r13), %xmm2 movups 0x10(%r9,%r13), %xmm3 movss (%r15,%rcx), %xmm4 unpcklps %xmm1, %xmm4 # xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1] movss (%rax,%rcx), %xmm5 shufps $0xd4, %xmm1, %xmm5 # xmm5 = xmm5[0,1],xmm1[1,3] shufps $0x24, %xmm5, %xmm4 # xmm4 = xmm4[0,1],xmm5[2,0] mulps %xmm2, %xmm4 movsd 0x4(%rax,%rcx), %xmm1 movhps (%r8,%rcx), %xmm1 # xmm1 = xmm1[0,1],mem[0,1] mulps %xmm3, %xmm1 movss 0x20(%r9,%r13), %xmm2 mulss 0x8(%r8,%rcx), %xmm2 addps %xmm4, %xmm1 movaps %xmm1, %xmm3 unpckhpd %xmm1, %xmm3 # xmm3 = xmm3[1],xmm1[1] addps %xmm1, %xmm3 movaps %xmm3, %xmm1 shufps $0x55, %xmm3, %xmm1 # xmm1 = xmm1[1,1],xmm3[1,1] addss %xmm3, %xmm1 addss %xmm0, %xmm2 addss %xmm1, %xmm2 movss %xmm2, (%rbx,%rcx) decl %esi addq $0x4, %rcx jmp 0x28b5b2 addq %rcx, %r15 addq $0x8, %r15 addq %rcx, %rax addq $0x8, %rax addq %rcx, %r8 addq $0x8, %r8 incl %r10d addq %rcx, %rbx cmpl %edi, %r10d jl 0x28b5ae movq 0x30(%rsp), %rax incq %rax movq 0x48(%rsp), %r14 movq 0x20(%rsp), %r10 movq 0x60(%rsp), %r13 movq 0x90(%rsp), %rsi jmp 0x28b3fb cmpl $0x5, %r15d jne 0x28ccf0 cmpl $0x1, 0xdc(%rbp,%r8) jne 0x28c9de cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28c9de cmpl $0x1, 0xe4(%rbp,%r8) jne 0x28c9de cmpl $0x1, 0xe8(%rbp,%r8) jne 0x28c9de movl 0x2c(%r14), %ecx movl 0x30(%r14), %edx movl 0xdc(%rsp), %eax movl 0xe8(%rsp), %esi movq 0x1b0(%rbp,%r8), %r9 movl %edx, %r8d leal 0x10(,%rax,4), %eax movslq %eax, %rdi leal (,%rcx,4), %eax cltq xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx movq %rcx, 0x10(%rsp) testl %esi, %esi cmovlel %edx, %esi movq %rsi, 0x50(%rsp) shlq $0x2, %rax movq %rax, 0x20(%rsp) shlq $0x2, %rdi movq %r9, 0x60(%rsp) cmpq 0x50(%rsp), %rdx je 0x28d415 testq %r9, %r9 je 0x28b732 movq %rdx, %rax shlq $0x4, %rax movups (%r9,%rax), %xmm0 jmp 0x28b735 xorps %xmm0, %xmm0 movq 0x40(%r14), %r10 imulq %rdx, %r10 movq 0x10(%r14), %rax imulq %rax, %r10 addq (%r14), %r10 movslq 0x2c(%r14), %r9 movq 0x28(%rbp), %r11 movslq 0x54(%rbp), %rbx imulq %rdx, %rbx imulq 0x38(%rbp), %rbx imulq %rax, %r9 addq %r10, %r9 movslq 0xdc(%rsp), %rax movq 0xf0(%rsp), %r15 movq %rdx, 0x30(%rsp) imulq %rdx, %r15 movq 0xc0(%rsp), %rcx imulq %rcx, %r15 addq 0xb0(%rsp), %r15 imulq %rcx, %rax leaq (%r15,%rax), %r12 leaq (%r15,%rax,2), %r13 leaq (%rax,%rax,2), %rbp addq %r15, %rbp leaq (%r15,%rax,4), %rdx leaq (%rax,%rax,4), %rax addq %r15, %rax xorl %esi, %esi movl %esi, %ecx orl $0x1, %ecx cmpl %r8d, %ecx jge 0x28baf4 movl %r8d, %r14d movq 0x10(%rsp), %rcx xorl %r8d, %r8d subl $0x1, %ecx jb 0x28bab4 movaps (%r11,%rbx), %xmm6 movaps 0x10(%r11,%rbx), %xmm8 movaps 0x20(%r11,%rbx), %xmm2 movaps 0x30(%r11,%rbx), %xmm3 movaps 0x40(%r11,%rbx), %xmm5 movaps (%r15,%r8), %xmm1 mulps %xmm6, %xmm1 addps %xmm0, %xmm1 movaps 0x10(%r15,%r8), %xmm4 mulps %xmm8, %xmm4 movaps 0x20(%r15,%r8), %xmm7 mulps %xmm2, %xmm7 addps %xmm4, %xmm7 addps %xmm1, %xmm7 movaps 0x30(%r15,%r8), %xmm1 mulps %xmm3, %xmm1 movaps 0x40(%r15,%r8), %xmm14 mulps %xmm5, %xmm14 addps %xmm1, %xmm14 movaps (%r12,%r8), %xmm1 movaps 0x10(%r12,%r8), %xmm13 movaps 0x20(%r12,%r8), %xmm12 movaps 0x30(%r12,%r8), %xmm10 movaps 0x40(%r12,%r8), %xmm4 mulps %xmm1, %xmm6 addps %xmm0, %xmm6 mulps %xmm13, %xmm8 mulps %xmm12, %xmm2 addps %xmm8, %xmm2 addps %xmm6, %xmm2 mulps %xmm10, %xmm3 mulps %xmm4, %xmm5 addps %xmm3, %xmm5 movaps 0x50(%r11,%rbx), %xmm6 movaps 0x60(%r11,%rbx), %xmm11 movaps 0x70(%r11,%rbx), %xmm9 movaps 0x80(%r11,%rbx), %xmm8 movaps 0x90(%r11,%rbx), %xmm3 mulps %xmm6, %xmm1 addps %xmm14, %xmm1 addps %xmm7, %xmm1 mulps %xmm11, %xmm13 mulps %xmm9, %xmm12 addps %xmm13, %xmm12 mulps %xmm8, %xmm10 addps %xmm12, %xmm10 mulps %xmm3, %xmm4 addps %xmm10, %xmm4 addps %xmm1, %xmm4 movaps (%r13,%r8), %xmm13 movaps 0x10(%r13,%r8), %xmm12 movaps 0x20(%r13,%r8), %xmm10 movaps 0x30(%r13,%r8), %xmm7 movaps 0x40(%r13,%r8), %xmm1 mulps %xmm13, %xmm6 addps %xmm5, %xmm6 addps %xmm2, %xmm6 mulps %xmm12, %xmm11 mulps %xmm10, %xmm9 addps %xmm11, %xmm9 mulps %xmm7, %xmm8 addps %xmm9, %xmm8 mulps %xmm1, %xmm3 addps %xmm8, %xmm3 addps %xmm6, %xmm3 movaps 0xa0(%r11,%rbx), %xmm14 movaps 0xb0(%r11,%rbx), %xmm11 movaps 0xc0(%r11,%rbx), %xmm8 movaps 0xd0(%r11,%rbx), %xmm5 movaps 0xe0(%r11,%rbx), %xmm2 mulps %xmm14, %xmm13 mulps %xmm11, %xmm12 addps %xmm13, %xmm12 mulps %xmm8, %xmm10 addps %xmm12, %xmm10 mulps %xmm5, %xmm7 addps %xmm10, %xmm7 mulps %xmm2, %xmm1 addps %xmm7, %xmm1 addps %xmm4, %xmm1 movaps (%rbp,%r8), %xmm13 movaps 0x10(%rbp,%r8), %xmm10 movaps 0x20(%rbp,%r8), %xmm9 movaps 0x30(%rbp,%r8), %xmm6 movaps 0x40(%rbp,%r8), %xmm4 mulps %xmm13, %xmm14 mulps %xmm10, %xmm11 addps %xmm14, %xmm11 mulps %xmm9, %xmm8 addps %xmm11, %xmm8 mulps %xmm6, %xmm5 addps %xmm8, %xmm5 mulps %xmm4, %xmm2 addps %xmm5, %xmm2 addps %xmm3, %xmm2 movaps 0xf0(%r11,%rbx), %xmm14 movaps 0x100(%r11,%rbx), %xmm12 movaps 0x110(%r11,%rbx), %xmm8 movaps 0x120(%r11,%rbx), %xmm7 movaps 0x130(%r11,%rbx), %xmm3 mulps %xmm14, %xmm13 mulps %xmm12, %xmm10 addps %xmm13, %xmm10 mulps %xmm8, %xmm9 addps %xmm10, %xmm9 mulps %xmm7, %xmm6 addps %xmm9, %xmm6 mulps %xmm3, %xmm4 addps %xmm6, %xmm4 movaps (%rdx,%r8), %xmm6 movaps 0x10(%rdx,%r8), %xmm13 movaps 0x20(%rdx,%r8), %xmm11 movaps 0x30(%rdx,%r8), %xmm9 movaps 0x40(%rdx,%r8), %xmm5 mulps %xmm6, %xmm14 mulps %xmm13, %xmm12 addps %xmm14, %xmm12 mulps %xmm11, %xmm8 addps %xmm12, %xmm8 mulps %xmm9, %xmm7 addps %xmm8, %xmm7 mulps %xmm5, %xmm3 addps %xmm7, %xmm3 movaps 0x140(%r11,%rbx), %xmm7 movaps 0x150(%r11,%rbx), %xmm14 movaps 0x160(%r11,%rbx), %xmm12 movaps 0x170(%r11,%rbx), %xmm10 movaps 0x180(%r11,%rbx), %xmm8 mulps %xmm7, %xmm6 addps %xmm4, %xmm6 addps %xmm1, %xmm6 mulps %xmm14, %xmm13 mulps %xmm12, %xmm11 addps %xmm13, %xmm11 mulps %xmm10, %xmm9 addps %xmm11, %xmm9 mulps %xmm8, %xmm5 addps %xmm9, %xmm5 addps %xmm6, %xmm5 mulps (%rax,%r8), %xmm7 addps %xmm3, %xmm7 mulps 0x10(%rax,%r8), %xmm14 mulps 0x20(%rax,%r8), %xmm12 addps %xmm2, %xmm7 addps %xmm14, %xmm12 mulps 0x30(%rax,%r8), %xmm10 addps %xmm12, %xmm10 mulps 0x40(%rax,%r8), %xmm8 addps %xmm10, %xmm8 addps %xmm7, %xmm8 movaps %xmm5, (%r10,%r8) movaps %xmm8, (%r9,%r8) addq $0x10, %r8 jmp 0x28b7ca addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp addq %rdi, %rdx addq %r8, %rdx addq %rdi, %rax addq %r8, %rax movq 0x20(%rsp), %rcx addq %rcx, %r10 addq %r8, %r10 addq %rcx, %r9 addq %r8, %r9 addl $0x2, %esi movl %r14d, %r8d jmp 0x28b7b1 movq 0x60(%rsp), %r9 cmpl %r8d, %esi jge 0x28bce9 movq 0x10(%rsp), %rax movl %eax, %ecx xorl %eax, %eax subl $0x1, %ecx jb 0x28bcbe movaps (%r11,%rbx), %xmm1 movaps 0x10(%r11,%rbx), %xmm2 movaps 0x20(%r11,%rbx), %xmm3 movaps 0x30(%r11,%rbx), %xmm4 mulps (%r15,%rax), %xmm1 movaps 0x40(%r11,%rbx), %xmm5 addps %xmm0, %xmm1 mulps 0x10(%r15,%rax), %xmm2 mulps 0x20(%r15,%rax), %xmm3 addps %xmm2, %xmm3 addps %xmm1, %xmm3 mulps 0x30(%r15,%rax), %xmm4 mulps 0x40(%r15,%rax), %xmm5 addps %xmm4, %xmm5 movaps 0x50(%r11,%rbx), %xmm1 movaps 0x60(%r11,%rbx), %xmm4 movaps 0x70(%r11,%rbx), %xmm6 movaps 0x80(%r11,%rbx), %xmm7 movaps 0x90(%r11,%rbx), %xmm2 mulps (%r12,%rax), %xmm1 addps %xmm5, %xmm1 addps %xmm3, %xmm1 mulps 0x10(%r12,%rax), %xmm4 mulps 0x20(%r12,%rax), %xmm6 mulps 0x30(%r12,%rax), %xmm7 addps %xmm4, %xmm6 addps %xmm6, %xmm7 mulps 0x40(%r12,%rax), %xmm2 addps %xmm7, %xmm2 addps %xmm1, %xmm2 movaps 0xa0(%r11,%rbx), %xmm3 movaps 0xb0(%r11,%rbx), %xmm4 movaps 0xc0(%r11,%rbx), %xmm5 movaps 0xd0(%r11,%rbx), %xmm6 mulps (%r13,%rax), %xmm3 movaps 0xe0(%r11,%rbx), %xmm1 mulps 0x10(%r13,%rax), %xmm4 addps %xmm3, %xmm4 mulps 0x20(%r13,%rax), %xmm5 mulps 0x30(%r13,%rax), %xmm6 addps %xmm4, %xmm5 addps %xmm5, %xmm6 mulps 0x40(%r13,%rax), %xmm1 addps %xmm6, %xmm1 addps %xmm2, %xmm1 movaps 0xf0(%r11,%rbx), %xmm2 movaps 0x100(%r11,%rbx), %xmm3 movaps 0x110(%r11,%rbx), %xmm4 movaps 0x120(%r11,%rbx), %xmm5 mulps (%rbp,%rax), %xmm2 movaps 0x130(%r11,%rbx), %xmm6 mulps 0x10(%rbp,%rax), %xmm3 addps %xmm2, %xmm3 mulps 0x20(%rbp,%rax), %xmm4 mulps 0x30(%rbp,%rax), %xmm5 addps %xmm3, %xmm4 addps %xmm4, %xmm5 mulps 0x40(%rbp,%rax), %xmm6 addps %xmm5, %xmm6 movaps 0x140(%r11,%rbx), %xmm2 movaps 0x150(%r11,%rbx), %xmm3 movaps 0x160(%r11,%rbx), %xmm4 movaps 0x170(%r11,%rbx), %xmm5 movaps 0x180(%r11,%rbx), %xmm7 mulps (%rdx,%rax), %xmm2 addps %xmm6, %xmm2 addps %xmm1, %xmm2 mulps 0x10(%rdx,%rax), %xmm3 mulps 0x20(%rdx,%rax), %xmm4 mulps 0x30(%rdx,%rax), %xmm5 addps %xmm3, %xmm4 addps %xmm4, %xmm5 mulps 0x40(%rdx,%rax), %xmm7 addps %xmm5, %xmm7 addps %xmm2, %xmm7 movaps %xmm7, (%r10,%rax) addq $0x10, %rax jmp 0x28bb0b addq %rax, %r15 addq $0x40, %r15 leaq 0x40(%r12,%rax), %r12 addq %rax, %r13 addq $0x40, %r13 addq %rax, %rbp addq $0x40, %rbp addq %rax, %rdx addq $0x40, %rdx incl %esi addq %rax, %r10 jmp 0x28baf9 movq 0x30(%rsp), %rdx incq %rdx movq 0x48(%rsp), %r14 movq 0x88(%rsp), %rbp jmp 0x28b714 cmpl $0x1, 0xdc(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x1, 0xe0(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x2, 0xe4(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x2, 0xe8(%rbp,%r8) movl %eax, %ecx jne 0x28beda movslq 0xdc(%rsp), %rdx movl 0x2c(%r14), %eax movl 0x30(%r14), %edi movl 0xe8(%rsp), %esi movl %edx, %ecx subl %eax, %ecx addl %ecx, %ecx movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r8), %r8 imulq 0x10(%r14), %r13 movq 0xb0(%rsp), %r11 movq %r11, 0x60(%rsp) movq 0xc0(%rsp), %r11 imulq 0xf0(%rsp), %r11 movq %r11, 0x70(%rsp) movq %rdx, 0x30(%rsp) addq %rdx, %rdx movq %rdx, 0x90(%rsp) movslq %ecx, %r15 xorl %r12d, %r12d testl %edi, %edi cmovlel %r12d, %edi testl %esi, %esi cmovlel %r12d, %esi movq %rsi, 0x10(%rsp) shlq $0x2, %r15 movq %r8, 0x50(%rsp) cmpq 0x10(%rsp), %r12 je 0x28d415 testq %r8, %r8 je 0x28bde0 movss (%r8,%r12,4), %xmm0 jmp 0x28bde3 xorps %xmm0, %xmm0 movq %r13, %r8 imulq %r12, %r13 addq %r10, %r13 imulq $0x24, %r12, %rbp movq 0x70(%rsp), %r11 imulq %r12, %r11 addq 0x60(%rsp), %r11 movq 0x30(%rsp), %rcx leaq (%r11,%rcx,4), %r10 movq 0x90(%rsp), %rcx leaq (%r11,%rcx,4), %rbx xorl %esi, %esi cmpl %edi, %esi je 0x28beb8 xorl %edx, %edx movl %eax, %ecx testl %ecx, %ecx jle 0x28be9f movsd 0x4(%r11,%rdx), %xmm1 movups (%r9,%rbp), %xmm2 movups 0x10(%r9,%rbp), %xmm3 movss (%r11,%rdx), %xmm4 unpcklps %xmm1, %xmm4 # xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1] movss (%r10,%rdx), %xmm5 shufps $0xd4, %xmm1, %xmm5 # xmm5 = xmm5[0,1],xmm1[1,3] shufps $0x24, %xmm5, %xmm4 # xmm4 = xmm4[0,1],xmm5[2,0] mulps %xmm2, %xmm4 movsd 0x4(%r10,%rdx), %xmm1 movhps (%rbx,%rdx), %xmm1 # xmm1 = xmm1[0,1],mem[0,1] mulps %xmm3, %xmm1 addps %xmm4, %xmm1 movss 0x20(%r9,%rbp), %xmm2 mulss 0x8(%rbx,%rdx), %xmm2 addss %xmm0, %xmm2 movaps %xmm1, %xmm3 unpckhpd %xmm1, %xmm3 # xmm3 = xmm3[1],xmm1[1] addps %xmm1, %xmm3 movaps %xmm3, %xmm1 shufps $0x55, %xmm3, %xmm1 # xmm1 = xmm1[1,1],xmm3[1,1] addss %xmm3, %xmm1 addss %xmm2, %xmm1 movss %xmm1, (%r13) addq $0x4, %r13 decl %ecx addq $0x8, %rdx jmp 0x28be22 addq %r15, %r11 addq %rdx, %r11 addq %r15, %r10 addq %rdx, %r10 addq %r15, %rbx addq %rdx, %rbx incl %esi jmp 0x28be16 incq %r12 movq 0x48(%rsp), %r14 movq 0x88(%rsp), %rbp movq 0x20(%rsp), %r10 movq %r8, %r13 movq 0x50(%rsp), %r8 jmp 0x28bdc8 cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %r10d movq 0x190(%rsp), %r14 cmpb $0x1, 0x27(%r14) jne 0x28bf18 xorl %eax, %eax testb $0x3, %dil sete %al leal (%rax,%rax,2), %eax incl %eax movl %eax, 0x20(%rsp) xorl %eax, %eax testb $0x3, %r10b sete %al leal (%rax,%rax,2), %eax incl %eax jmp 0x28bf1f pushq $0x1 popq %rax movl %eax, 0x20(%rsp) movl %eax, 0x10(%rsp) movl 0x30(%rsp), %ebx movq %r11, %r13 movq 0xb0(%rsp), %rcx movq 0xb8(%rsp), %rax movq %rcx, 0x198(%rsp) movq %rax, 0x1a0(%rsp) movq 0xc0(%rsp), %rcx movq %rcx, 0x1a8(%rsp) movl 0xc8(%rsp), %ecx movl %ecx, 0x1b0(%rsp) movq 0xd0(%rsp), %rcx movq %rcx, 0x1b8(%rsp) movups 0xd8(%rsp), %xmm0 movups %xmm0, 0x1c0(%rsp) movl 0xe8(%rsp), %ecx movl %ecx, 0x1d0(%rsp) movq 0xf0(%rsp), %rcx movq %rcx, 0x1d8(%rsp) testq %rax, %rax je 0x28bfae lock incl (%rax) cmpl 0x20(%rsp), %r9d jle 0x28c00a movl %r10d, %r12d movl %edi, %ebp movups (%r14), %xmm0 movups 0x10(%r14), %xmm1 movups 0x20(%r14), %xmm2 movups 0x30(%r14), %xmm3 leaq 0x130(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x10(%r14), %rax movq %rax, 0x8(%rcx) leaq 0xb0(%rsp), %rdi leaq 0x198(%rsp), %rsi movl 0x20(%rsp), %edx callq 0x64e3b movl %ebp, %edi movl %r12d, %r10d movq 0x48(%rsp), %rdx movq (%rdx), %rax movq %rax, 0x130(%rsp) movq 0x8(%rdx), %rax movq %rax, 0x138(%rsp) movq 0x10(%rdx), %rcx movq %rcx, 0x140(%rsp) movl 0x18(%rdx), %ecx movl %ecx, 0x148(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x150(%rsp) movups 0x28(%rdx), %xmm0 movups %xmm0, 0x158(%rsp) movl 0x38(%rdx), %ecx movl %ecx, 0x168(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x170(%rsp) testq %rax, %rax je 0x28c072 lock incl (%rax) cmpl %ebx, 0x10(%rsp) jae 0x28c0fe movq 0x88(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xd0(%rcx,%rax), %eax cltd movl 0x10(%rsp), %r9d idivl %r9d movl %r15d, %ecx shrq %cl, %r13 movl %edi, %ebx movl %r9d, %r8d imulq %r13, %r8 movq 0x10(%r14), %rcx movq %rcx, (%rsp) leaq 0x130(%rsp), %rdi movl 0x60(%rsp), %esi movl 0x70(%rsp), %edx movl %eax, %ecx movl %r10d, %ebp callq 0x628f2 pushq $-0x64 popq %r12 cmpq $0x0, 0x130(%rsp) je 0x28d35b movl %ebp, %r10d movl %ebx, %edi movslq 0x168(%rsp), %rax imulq 0x170(%rsp), %rax testq %rax, %rax je 0x28d35b xorl %r15d, %r15d xorl %r12d, %r12d xorl %ebx, %ebx movq 0x88(%rsp), %r9 movq (%r9), %rax movq -0x18(%rax), %rax movslq 0x108(%r9,%rax), %rax cmpq %rax, %rbx jge 0x28c3ab movl %r15d, %eax cltd movl 0x20(%rsp), %esi idivl %esi movl %eax, %ecx movl %edi, %ebp movl %edi, %eax cltd idivl %esi movl %eax, %edx movslq 0x1c4(%rsp), %rdi movslq 0x1c8(%rsp), %rsi movslq 0x1cc(%rsp), %rax movslq %ecx, %r8 imulq 0x1d8(%rsp), %r8 movq 0x1a8(%rsp), %rcx imulq %rcx, %r8 addq 0x198(%rsp), %r8 movq %r8, 0x238(%rsp) movl 0x1b0(%rsp), %r8d andq $0x0, 0x240(%rsp) movq %rcx, 0x248(%rsp) movl %r8d, 0x250(%rsp) movq 0x1b8(%rsp), %r8 movq %r8, 0x258(%rsp) movl %edi, 0x264(%rsp) movl %esi, 0x268(%rsp) movl %eax, 0x26c(%rsp) movl %edx, 0x270(%rsp) imulq %rdi, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x278(%rsp) movl 0x1c0(%rsp), %eax movl %eax, 0x260(%rsp) movl %r12d, %eax cltd movl 0x10(%rsp), %esi idivl %esi movl %eax, %ecx movl %r10d, %r13d movl %r10d, %eax cltd idivl %esi movl %eax, %edx movslq 0x15c(%rsp), %r8 movslq 0x160(%rsp), %rdi movslq 0x164(%rsp), %rax movslq %ecx, %rcx imulq 0x170(%rsp), %rcx movq 0x140(%rsp), %rsi imulq %rsi, %rcx addq 0x130(%rsp), %rcx movq %rcx, 0x1f0(%rsp) movl 0x148(%rsp), %ecx andq $0x0, 0x1f8(%rsp) movq %rsi, 0x200(%rsp) movl %ecx, 0x208(%rsp) movq 0x150(%rsp), %rcx movq %rcx, 0x210(%rsp) movl %r8d, 0x21c(%rsp) movl %edi, 0x220(%rsp) movl %eax, 0x224(%rsp) movl %edx, 0x228(%rsp) imulq %r8, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x230(%rsp) movl 0x158(%rsp), %eax movl %eax, 0x218(%rsp) movq 0x10(%r9), %rax movq (%rax,%rbx,8), %rdi movups (%r14), %xmm0 movups 0x10(%r14), %xmm1 movups 0x20(%r14), %xmm2 movups 0x30(%r14), %xmm3 movaps %xmm0, 0x280(%rsp) movaps %xmm3, 0x2b0(%rsp) movaps %xmm2, 0x2a0(%rsp) movaps %xmm1, 0x290(%rsp) movq %rcx, 0x288(%rsp) movq (%rdi), %rax leaq 0x238(%rsp), %rsi leaq 0x1f0(%rsp), %rdx leaq 0x280(%rsp), %rcx callq 0x38(%rax) movq 0x1f8(%rsp), %rax testq %rax, %rax je 0x28c361 lock decl (%rax) jne 0x28c361 movq 0x1f0(%rsp), %rsi movq 0x210(%rsp), %rdi testq %rdi, %rdi je 0x28c359 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28c361 movq %rsi, %rdi callq 0x5f3e0 movq 0x240(%rsp), %rax testq %rax, %rax je 0x28c398 lock decl (%rax) jne 0x28c398 movq 0x238(%rsp), %rsi movq 0x258(%rsp), %rdi testq %rdi, %rdi je 0x28c390 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28c398 movq %rsi, %rdi callq 0x5f3e0 incq %rbx movl %r13d, %r10d addl %r13d, %r12d movl %ebp, %edi addl %ebp, %r15d jmp 0x28c106 movl 0x30(%rsp), %edx cmpl %edx, 0x10(%rsp) jae 0x28c3d2 leaq 0x130(%rsp), %rdi movq 0x48(%rsp), %rsi movq %r14, %rcx callq 0x64e3b xorl %r12d, %r12d jmp 0x28d35b xorl %r12d, %r12d leaq 0x130(%rsp), %rax movq 0x48(%rsp), %rcx cmpq %rcx, %rax je 0x28d35b movq 0x138(%rsp), %rax testq %rax, %rax je 0x28c3fb lock incl (%rax) movq 0x8(%rcx), %rax testq %rax, %rax je 0x28d307 lock decl (%rax) jne 0x28d307 movq 0x48(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x28d2fc movq (%rdi), %rax callq 0x18(%rax) xorl %r12d, %r12d jmp 0x28d307 pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rbp,%r8) jne 0x28ccf0 cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28ccf0 cmpl $0x2, 0xe4(%rbp,%r8) jne 0x28ccf0 cmpl $0x2, 0xe8(%rbp,%r8) jne 0x28ccf0 movl 0x2c(%r14), %eax movl 0x30(%r14), %ecx movl 0xdc(%rsp), %edi movl 0xe8(%rsp), %edx subl %eax, %edi shll $0x3, %edi movq 0x1b0(%rbp,%r8), %rsi movslq %edi, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x28d415 testq %rsi, %rsi je 0x28c4c9 movq %r8, %r9 shlq $0x4, %r9 movups (%rsi,%r9), %xmm15 jmp 0x28c4cd xorps %xmm15, %xmm15 movq 0x40(%r14), %r9 imulq %r8, %r9 imulq 0x10(%r14), %r9 addq (%r14), %r9 movq 0x28(%rbp), %r15 movslq 0x54(%rbp), %r12 imulq %r8, %r12 imulq 0x38(%rbp), %r12 movslq 0xdc(%rsp), %rbx movq 0xf0(%rsp), %r10 imulq %r8, %r10 movq 0xc0(%rsp), %r11 imulq %r11, %r10 addq 0xb0(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx movaps (%r15,%r12), %xmm3 movaps 0x10(%r15,%r12), %xmm4 movaps 0x20(%r15,%r12), %xmm5 movaps 0x30(%r15,%r12), %xmm14 movaps 0x40(%r15,%r12), %xmm7 movaps 0x50(%r15,%r12), %xmm6 movaps 0x60(%r15,%r12), %xmm8 movaps 0x70(%r15,%r12), %xmm0 movaps %xmm0, 0x30(%rsp) movaps 0x80(%r15,%r12), %xmm0 movaps %xmm0, 0x60(%rsp) xorl %ebp, %ebp movaps %xmm5, 0x70(%rsp) movaps %xmm6, 0x90(%rsp) movaps %xmm8, 0xa0(%rsp) movaps %xmm3, 0x20(%rsp) movaps %xmm15, 0x50(%rsp) movaps %xmm4, 0x10(%rsp) cmpl %ecx, %ebp je 0x28c9ce xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r13d cmpl %eax, %r13d jge 0x28c7f1 movaps (%r10,%r15), %xmm11 mulps %xmm3, %xmm11 addps %xmm15, %xmm11 movaps 0x10(%r10,%r15), %xmm12 mulps %xmm4, %xmm12 movaps 0x20(%r10,%r15), %xmm13 movaps 0x40(%r10,%r15), %xmm10 movaps %xmm13, %xmm15 movaps 0x70(%rsp), %xmm6 mulps %xmm6, %xmm15 addps %xmm12, %xmm15 addps %xmm11, %xmm15 movaps (%r11,%r15), %xmm11 mulps %xmm14, %xmm11 movaps 0x10(%r11,%r15), %xmm12 mulps %xmm7, %xmm12 addps %xmm11, %xmm12 movaps 0x20(%r11,%r15), %xmm0 movaps 0x40(%r11,%r15), %xmm11 movaps %xmm7, %xmm8 movaps %xmm14, %xmm7 movaps %xmm0, %xmm14 movaps 0x90(%rsp), %xmm9 mulps %xmm9, %xmm14 addps %xmm12, %xmm14 addps %xmm15, %xmm14 movaps (%rbx,%r15), %xmm12 movaps 0xa0(%rsp), %xmm5 mulps %xmm5, %xmm12 movaps 0x10(%rbx,%r15), %xmm1 movaps 0x30(%rsp), %xmm4 mulps %xmm4, %xmm1 addps %xmm12, %xmm1 movaps 0x20(%rbx,%r15), %xmm2 movaps 0x40(%rbx,%r15), %xmm12 movaps %xmm2, %xmm15 movaps 0x60(%rsp), %xmm3 mulps %xmm3, %xmm15 addps %xmm1, %xmm15 addps %xmm14, %xmm15 movaps 0x30(%r10,%r15), %xmm1 mulps 0x10(%rsp), %xmm1 movaps 0x30(%r11,%r15), %xmm14 mulps %xmm8, %xmm14 addps %xmm1, %xmm14 movaps 0x30(%rbx,%r15), %xmm1 mulps %xmm4, %xmm1 addps %xmm14, %xmm1 movaps %xmm15, (%r9) mulps 0x20(%rsp), %xmm13 addps 0x50(%rsp), %xmm13 movaps %xmm10, %xmm14 mulps %xmm6, %xmm14 mulps %xmm7, %xmm0 movaps %xmm11, %xmm15 mulps %xmm9, %xmm15 addps %xmm14, %xmm15 mulps %xmm5, %xmm2 addps %xmm0, %xmm2 addps %xmm13, %xmm2 addps %xmm1, %xmm2 movaps %xmm12, %xmm0 mulps %xmm3, %xmm0 addps %xmm15, %xmm0 addps %xmm2, %xmm0 movaps 0x60(%r10,%r15), %xmm14 movaps 0x60(%r11,%r15), %xmm15 movaps 0x60(%rbx,%r15), %xmm13 movaps 0x50(%r10,%r15), %xmm1 mulps 0x10(%rsp), %xmm1 movaps 0x50(%r11,%r15), %xmm2 mulps %xmm8, %xmm2 addps %xmm1, %xmm2 movaps 0x50(%rbx,%r15), %xmm1 mulps %xmm4, %xmm1 addps %xmm2, %xmm1 movaps %xmm0, 0x10(%r9) movaps %xmm14, %xmm0 mulps %xmm6, %xmm0 movaps %xmm15, %xmm2 mulps %xmm9, %xmm2 addps %xmm0, %xmm2 mulps %xmm7, %xmm11 mulps %xmm5, %xmm12 addps %xmm11, %xmm12 mulps 0x20(%rsp), %xmm10 addps 0x50(%rsp), %xmm10 addps %xmm10, %xmm12 addps %xmm1, %xmm12 movaps %xmm13, %xmm0 mulps %xmm3, %xmm0 addps %xmm2, %xmm0 addps %xmm12, %xmm0 movaps 0x70(%r10,%r15), %xmm1 mulps 0x10(%rsp), %xmm1 movaps 0x80(%r10,%r15), %xmm2 mulps %xmm6, %xmm2 movaps 0x70(%r11,%r15), %xmm10 mulps %xmm8, %xmm10 addps %xmm1, %xmm10 movaps 0x80(%r11,%r15), %xmm1 mulps %xmm9, %xmm1 addps %xmm2, %xmm1 movaps 0x70(%rbx,%r15), %xmm2 mulps %xmm4, %xmm2 addps %xmm10, %xmm2 movaps 0x80(%rbx,%r15), %xmm10 mulps %xmm3, %xmm10 addps %xmm1, %xmm10 movaps %xmm0, 0x20(%r9) mulps %xmm7, %xmm15 mulps %xmm5, %xmm13 addps %xmm15, %xmm13 movaps 0x50(%rsp), %xmm15 movaps 0x20(%rsp), %xmm3 movaps 0x10(%rsp), %xmm4 mulps %xmm3, %xmm14 addps %xmm15, %xmm14 addps %xmm14, %xmm13 movaps %xmm7, %xmm14 movaps %xmm8, %xmm7 addps %xmm2, %xmm13 addps %xmm10, %xmm13 movaps %xmm13, 0x30(%r9) addq $0x40, %r9 addl $0x4, %r12d subq $-0x80, %r15 jmp 0x28c59a movaps 0x90(%rsp), %xmm6 movaps 0xa0(%rsp), %xmm8 leal 0x1(%r12), %r13d cmpl %eax, %r13d jge 0x28c91e movaps 0x20(%r10,%r15), %xmm11 movaps 0x20(%r11,%r15), %xmm12 movaps 0x20(%rbx,%r15), %xmm10 movaps (%r10,%r15), %xmm0 mulps %xmm3, %xmm0 addps %xmm15, %xmm0 movaps 0x10(%r10,%r15), %xmm1 mulps %xmm4, %xmm1 movaps %xmm11, %xmm2 movaps 0x70(%rsp), %xmm13 mulps %xmm13, %xmm2 addps %xmm1, %xmm2 addps %xmm0, %xmm2 movaps (%r11,%r15), %xmm0 mulps %xmm14, %xmm0 movaps 0x10(%r11,%r15), %xmm1 mulps %xmm7, %xmm1 addps %xmm0, %xmm1 movaps %xmm12, %xmm0 mulps %xmm6, %xmm0 addps %xmm1, %xmm0 addps %xmm2, %xmm0 movaps (%rbx,%r15), %xmm1 mulps %xmm8, %xmm1 movaps 0x10(%rbx,%r15), %xmm2 movaps 0x30(%rsp), %xmm9 mulps %xmm9, %xmm2 addps %xmm1, %xmm2 movaps %xmm10, %xmm1 movaps 0x60(%rsp), %xmm5 mulps %xmm5, %xmm1 addps %xmm2, %xmm1 addps %xmm0, %xmm1 movaps 0x30(%r10,%r15), %xmm0 mulps %xmm4, %xmm0 movaps 0x40(%r10,%r15), %xmm2 mulps %xmm13, %xmm2 movaps 0x30(%r11,%r15), %xmm13 mulps %xmm7, %xmm13 addps %xmm0, %xmm13 movaps 0x40(%r11,%r15), %xmm0 mulps %xmm6, %xmm0 addps %xmm2, %xmm0 movaps 0x30(%rbx,%r15), %xmm2 mulps %xmm9, %xmm2 addps %xmm13, %xmm2 movaps 0x40(%rbx,%r15), %xmm13 mulps %xmm5, %xmm13 addps %xmm0, %xmm13 movaps %xmm1, (%r9) mulps %xmm14, %xmm12 mulps %xmm8, %xmm10 addps %xmm12, %xmm10 mulps %xmm3, %xmm11 addps %xmm15, %xmm11 addps %xmm11, %xmm10 addps %xmm2, %xmm10 addps %xmm13, %xmm10 movaps %xmm10, 0x10(%r9) addq $0x20, %r9 addl $0x2, %r12d addq $0x40, %r15 jmp 0x28c802 movq 0x48(%rsp), %r14 movaps 0x70(%rsp), %xmm5 cmpl %eax, %r12d jge 0x28c9b5 movaps (%r10,%r15), %xmm0 mulps %xmm3, %xmm0 addps %xmm15, %xmm0 movaps 0x10(%r10,%r15), %xmm1 mulps %xmm4, %xmm1 movaps 0x20(%r10,%r15), %xmm2 mulps %xmm5, %xmm2 addps %xmm1, %xmm2 addps %xmm0, %xmm2 movaps (%r11,%r15), %xmm0 mulps %xmm14, %xmm0 movaps 0x10(%r11,%r15), %xmm1 mulps %xmm7, %xmm1 addps %xmm0, %xmm1 movaps 0x20(%r11,%r15), %xmm0 mulps %xmm6, %xmm0 addps %xmm1, %xmm0 addps %xmm2, %xmm0 movaps (%rbx,%r15), %xmm1 mulps %xmm8, %xmm1 movaps 0x10(%rbx,%r15), %xmm2 mulps 0x30(%rsp), %xmm2 addps %xmm1, %xmm2 movaps 0x20(%rbx,%r15), %xmm1 mulps 0x60(%rsp), %xmm1 addps %xmm2, %xmm1 addps %xmm0, %xmm1 movaps %xmm1, (%r9) addq $0x10, %r9 incl %r12d addq $0x20, %r15 jmp 0x28c928 addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x28c58c incq %r8 movq 0x88(%rsp), %rbp jmp 0x28c4ad pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rbp,%r8) jne 0x28ccf0 cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28ccf0 cmpl $0x2, 0xe4(%rbp,%r8) jne 0x28ccf0 cmpl $0x2, 0xe8(%rbp,%r8) jne 0x28ccf0 movl 0x2c(%r14), %edi movl 0x30(%r14), %ecx movl 0xdc(%rsp), %eax movl 0xe8(%rsp), %edx subl %edi, %eax shll $0x3, %eax movq 0x1b0(%rbp,%r8), %rsi xorl %r15d, %r15d testl %edi, %edi cmovlel %r15d, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %edx, %edx cmovlel %r15d, %edx movq %rdx, 0x10(%rsp) shlq $0x2, %r8 movq %rsi, 0x20(%rsp) cmpq 0x10(%rsp), %r15 je 0x28d415 testq %rsi, %rsi je 0x28ca84 movq %r15, %rax shlq $0x4, %rax movups (%rsi,%rax), %xmm0 jmp 0x28ca87 xorps %xmm0, %xmm0 movq 0x40(%r14), %r9 imulq %r15, %r9 imulq 0x10(%r14), %r9 addq (%r14), %r9 movq 0x28(%rbp), %r10 movslq 0x54(%rbp), %r11 imulq %r15, %r11 imulq 0x38(%rbp), %r11 movq 0xf0(%rsp), %rbx movq %r15, %r14 imulq %r15, %rbx movq 0xc0(%rsp), %rax imulq %rax, %rbx addq 0xb0(%rsp), %rbx movslq 0xdc(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x28ccd3 movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x28ccae movaps (%r10,%r11), %xmm2 movaps 0x10(%r10,%r11), %xmm3 movaps 0x20(%r10,%r11), %xmm1 movaps 0x30(%r10,%r11), %xmm4 movaps 0x40(%r10,%r11), %xmm5 mulps (%rbx,%rsi), %xmm2 mulps 0x10(%rbx,%rsi), %xmm3 mulps 0x20(%rbx,%rsi), %xmm1 addps %xmm0, %xmm2 addps %xmm3, %xmm1 mulps 0x30(%rbx,%rsi), %xmm4 addps %xmm2, %xmm1 mulps 0x40(%rbx,%rsi), %xmm5 addps %xmm4, %xmm5 movaps 0x50(%r10,%r11), %xmm3 movaps 0x60(%r10,%r11), %xmm4 movaps 0x70(%r10,%r11), %xmm6 movaps 0x80(%r10,%r11), %xmm7 movaps 0x90(%r10,%r11), %xmm2 mulps (%r15,%rsi), %xmm3 addps %xmm5, %xmm3 mulps 0x10(%r15,%rsi), %xmm4 mulps 0x20(%r15,%rsi), %xmm6 addps %xmm1, %xmm3 addps %xmm4, %xmm6 mulps 0x30(%r15,%rsi), %xmm7 addps %xmm6, %xmm7 mulps 0x40(%r15,%rsi), %xmm2 addps %xmm7, %xmm2 addps %xmm3, %xmm2 movaps 0xa0(%r10,%r11), %xmm3 movaps 0xb0(%r10,%r11), %xmm4 movaps 0xc0(%r10,%r11), %xmm5 movaps 0xd0(%r10,%r11), %xmm6 movaps 0xe0(%r10,%r11), %xmm1 mulps (%r12,%rsi), %xmm3 mulps 0x10(%r12,%rsi), %xmm4 mulps 0x20(%r12,%rsi), %xmm5 addps %xmm3, %xmm4 addps %xmm4, %xmm5 mulps 0x30(%r12,%rsi), %xmm6 addps %xmm5, %xmm6 mulps 0x40(%r12,%rsi), %xmm1 addps %xmm6, %xmm1 addps %xmm2, %xmm1 movaps 0xf0(%r10,%r11), %xmm2 movaps 0x100(%r10,%r11), %xmm3 movaps 0x110(%r10,%r11), %xmm4 movaps 0x120(%r10,%r11), %xmm5 movaps 0x130(%r10,%r11), %xmm6 mulps (%r13,%rsi), %xmm2 mulps 0x10(%r13,%rsi), %xmm3 mulps 0x20(%r13,%rsi), %xmm4 addps %xmm2, %xmm3 addps %xmm3, %xmm4 mulps 0x30(%r13,%rsi), %xmm5 addps %xmm4, %xmm5 mulps 0x40(%r13,%rsi), %xmm6 addps %xmm5, %xmm6 movaps 0x140(%r10,%r11), %xmm2 movaps 0x150(%r10,%r11), %xmm3 movaps 0x160(%r10,%r11), %xmm4 movaps 0x170(%r10,%r11), %xmm5 movaps 0x180(%r10,%r11), %xmm7 mulps (%rbp,%rsi), %xmm2 addps %xmm6, %xmm2 mulps 0x10(%rbp,%rsi), %xmm3 mulps 0x20(%rbp,%rsi), %xmm4 addps %xmm1, %xmm2 addps %xmm3, %xmm4 mulps 0x30(%rbp,%rsi), %xmm5 addps %xmm4, %xmm5 mulps 0x40(%rbp,%rsi), %xmm7 addps %xmm5, %xmm7 addps %xmm2, %xmm7 movaps %xmm7, (%r9) addq $0x10, %r9 addq $0x20, %rsi jmp 0x28caf8 addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x28caec movq %r14, %r15 incq %r15 movq 0x48(%rsp), %r14 movq 0x88(%rsp), %rbp movq 0x20(%rsp), %rsi jmp 0x28ca67 imull %eax, %r15d movslq %r15d, %rsi leaq 0x198(%rsp), %rdi leaq 0x130(%rsp), %rdx callq 0x73bbe movq (%rbp), %rcx movq -0x18(%rcx), %rdx movl 0x10(%rsp), %r11d imull 0xe0(%rbp,%rdx), %r11d movq 0x198(%rsp), %rax movl 0xdc(%rbp,%rdx), %esi imull 0xd4(%rbp,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%rbp,%rdx), %r8d jge 0x28cd80 movslq %edi, %rdi leaq (%rax,%rdi,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%rbp,%rdx), %r9d jge 0x28cd75 movl %esi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%rbp,%rdx), %esi incq %r9 jmp 0x28cd57 addl %r11d, %esi incl %r8d addq %r9, %rdi jmp 0x28cd43 leal (,%r15,4), %ecx movl %ecx, 0x90(%rsp) movl 0x60(%rsp), %edx shll $0x2, %edx xorl %ecx, %ecx testl %r15d, %r15d cmovlel %ecx, %r15d movslq %edx, %rdx movq %rdx, 0x30(%rsp) movq 0x50(%rsp), %r12 testl %r12d, %r12d cmovlel %ecx, %r12d movq %r12, 0x50(%rsp) movslq 0xa0(%rsp), %rdi shlq $0x2, %r15 leaq 0x16b70d(%rip), %r8 # 0x3f84d8 xorps %xmm0, %xmm0 movaps 0x16423a(%rip), %xmm15 # 0x3f1010 movaps 0x164243(%rip), %xmm4 # 0x3f1020 movaps 0x16424c(%rip), %xmm6 # 0x3f1030 movaps 0x1612b5(%rip), %xmm5 # 0x3ee0a0 movaps 0x16424d(%rip), %xmm8 # 0x3f1040 xorl %edx, %edx movaps 0x164253(%rip), %xmm9 # 0x3f1050 movaps 0x16426b(%rip), %xmm11 # 0x3f1070 movaps 0x164273(%rip), %xmm12 # 0x3f1080 movaps 0x16427b(%rip), %xmm13 # 0x3f1090 cmpq 0x50(%rsp), %rdx je 0x28d2e7 movq %rcx, 0x60(%rsp) movslq %ecx, %r10 movq 0x40(%r14), %r11 imulq %rdx, %r11 imulq 0x10(%r14), %r11 shlq $0x2, %r10 addq (%r14), %r11 movslq 0xdc(%rsp), %rbx movq 0xf0(%rsp), %r13 imulq %rdx, %r13 movq 0xc0(%rsp), %rcx imulq %rcx, %r13 addq 0xb0(%rsp), %r13 imulq %rcx, %rbx movq %rdx, 0x70(%rsp) shlq $0x4, %rdx movq %rdx, 0x10(%rsp) addq 0x28(%rbp), %r10 xorl %ecx, %ecx cmpl 0x20(%rsp), %ecx jg 0x28d2c9 movq (%rbp), %rdx xorl %r9d, %r9d cmpq %rdi, %r9 jg 0x28d2b9 movq -0x18(%rdx), %r12 cmpl $0x0, 0x100(%rbp,%r12) je 0x28ceb9 movq 0x1b0(%rbp,%r12), %rsi movq 0x10(%rsp), %r14 movups (%rsi,%r14), %xmm1 jmp 0x28cebc xorps %xmm1, %xmm1 movslq 0xe8(%rbp,%r12), %rsi movq %rbp, %r14 movslq %ecx, %rbp imulq %rsi, %rbp imulq %rbx, %rbp addq %r13, %rbp movl 0xe4(%r14,%r12), %esi imull %r9d, %esi shll $0x2, %esi movslq %esi, %rsi leaq (,%rsi,4), %rsi addq %rbp, %rsi xorl %ebp, %ebp cmpq %rbp, %r15 je 0x28cf17 movslq (%rax,%rbp), %r14 shlq $0x4, %r14 movups (%rsi,%r14), %xmm2 movups (%r10,%rbp,4), %xmm3 mulps %xmm2, %xmm3 addps %xmm3, %xmm1 addq $0x4, %rbp jmp 0x28cef4 movq 0x88(%rsp), %rbp movl 0x110(%rbp,%r12), %esi decl %esi cmpl $0x5, %esi ja 0x28d1cc movslq (%r8,%rsi,4), %rsi addq %r8, %rsi movaps %xmm1, %xmm7 maxps %xmm0, %xmm7 jmpq %rsi movq 0x118(%rbp,%r12), %rsi minps %xmm0, %xmm1 movss (%rsi), %xmm2 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm2 addps %xmm2, %xmm7 jmp 0x28d2a5 movaps %xmm1, %xmm2 minps %xmm15, %xmm2 maxps %xmm4, %xmm2 movaps %xmm2, %xmm14 mulps %xmm6, %xmm14 addps %xmm5, %xmm14 cvttps2dq %xmm14, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm14 andps %xmm8, %xmm14 subps %xmm14, %xmm7 movaps %xmm15, %xmm4 cvttps2dq %xmm7, %xmm15 mulps %xmm9, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm7 mulps %xmm2, %xmm7 movaps %xmm2, %xmm14 movaps 0x1640b4(%rip), %xmm10 # 0x3f1060 mulps %xmm10, %xmm14 addps %xmm11, %xmm14 mulps %xmm2, %xmm14 addps %xmm12, %xmm14 mulps %xmm2, %xmm14 addps %xmm13, %xmm14 mulps %xmm2, %xmm14 xorps %xmm6, %xmm6 movaps %xmm5, %xmm0 movaps 0x1640cb(%rip), %xmm5 # 0x3f10a0 addps %xmm5, %xmm14 mulps %xmm2, %xmm14 addps %xmm0, %xmm14 mulps %xmm7, %xmm14 addps %xmm8, %xmm2 addps %xmm14, %xmm2 pslld $0x17, %xmm15 paddd %xmm8, %xmm15 mulps %xmm2, %xmm15 addps %xmm8, %xmm15 movaps %xmm15, %xmm2 maxps 0x1640a4(%rip), %xmm15 # 0x3f10b0 movaps %xmm15, %xmm7 andps 0x1640a8(%rip), %xmm15 # 0x3f10c0 orps %xmm0, %xmm15 movaps %xmm15, %xmm3 cmpltps 0x1640b8(%rip), %xmm3 # 0x3f10e0 movaps %xmm3, %xmm14 andps %xmm15, %xmm14 movaps 0x1640b8(%rip), %xmm10 # 0x3f10f0 addps %xmm10, %xmm15 addps %xmm14, %xmm15 cmpleps %xmm6, %xmm2 psrld $0x17, %xmm7 paddd 0x16407f(%rip), %xmm7 # 0x3f10d0 cvtdq2ps %xmm7, %xmm14 andps %xmm8, %xmm3 subps %xmm3, %xmm14 movaps %xmm15, %xmm7 mulps 0x164098(%rip), %xmm7 # 0x3f1100 addps 0x1640a1(%rip), %xmm7 # 0x3f1110 mulps %xmm15, %xmm7 addps 0x1640a6(%rip), %xmm7 # 0x3f1120 mulps %xmm15, %xmm7 addps 0x1640ab(%rip), %xmm7 # 0x3f1130 mulps %xmm15, %xmm7 addps 0x1640b0(%rip), %xmm7 # 0x3f1140 mulps %xmm15, %xmm7 addps 0x1640b5(%rip), %xmm7 # 0x3f1150 mulps %xmm15, %xmm7 addps 0x1640ba(%rip), %xmm7 # 0x3f1160 mulps %xmm15, %xmm7 addps 0x1640bf(%rip), %xmm7 # 0x3f1170 mulps %xmm15, %xmm7 addps 0x1640c4(%rip), %xmm7 # 0x3f1180 mulps %xmm15, %xmm7 mulps %xmm9, %xmm14 addps %xmm15, %xmm14 mulps %xmm15, %xmm15 addps 0x1640bd(%rip), %xmm7 # 0x3f1190 mulps %xmm15, %xmm7 movaps %xmm4, %xmm15 movaps 0x163f3e(%rip), %xmm4 # 0x3f1020 movaps 0x163f47(%rip), %xmm6 # 0x3f1030 addps %xmm7, %xmm14 mulps 0x164bcb(%rip), %xmm14 # 0x3f1cc0 movaps %xmm2, %xmm3 andnps %xmm14, %xmm3 andps 0x164bcd(%rip), %xmm2 # 0x3f1cd0 orps %xmm3, %xmm2 minps %xmm15, %xmm2 maxps %xmm4, %xmm2 movaps %xmm2, %xmm3 mulps %xmm6, %xmm3 addps %xmm0, %xmm3 cvttps2dq %xmm3, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm3 andps %xmm8, %xmm3 subps %xmm3, %xmm7 cvttps2dq %xmm7, %xmm14 mulps %xmm9, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm3 mulps %xmm2, %xmm3 movaps %xmm2, %xmm7 mulps 0x163f1c(%rip), %xmm7 # 0x3f1060 addps %xmm11, %xmm7 mulps %xmm2, %xmm7 addps %xmm12, %xmm7 mulps %xmm2, %xmm7 addps %xmm13, %xmm7 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 movaps %xmm0, %xmm5 xorps %xmm0, %xmm0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm3, %xmm7 addps %xmm8, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm8, %xmm14 mulps %xmm2, %xmm14 addps %xmm8, %xmm14 rcpps %xmm14, %xmm2 movaps %xmm2, %xmm7 addps %xmm2, %xmm7 mulps %xmm7, %xmm14 movaps 0x164b46(%rip), %xmm3 # 0x3f1ce0 subps %xmm14, %xmm3 mulps %xmm2, %xmm3 addps %xmm10, %xmm7 addps %xmm3, %xmm7 jmp 0x28d2a2 movq 0x118(%rbp,%r12), %rsi movss (%rsi), %xmm2 movss 0x4(%rsi), %xmm7 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] maxps %xmm2, %xmm1 minps %xmm7, %xmm1 movaps %xmm1, %xmm7 jmp 0x28d2a5 xorps 0x160eb5(%rip), %xmm1 # 0x3ee090 minps %xmm15, %xmm1 maxps %xmm4, %xmm1 movaps %xmm1, %xmm7 mulps %xmm6, %xmm7 addps %xmm5, %xmm7 cvttps2dq %xmm7, %xmm2 cvtdq2ps %xmm2, %xmm2 cmpltps %xmm2, %xmm7 andps %xmm8, %xmm7 subps %xmm7, %xmm2 cvttps2dq %xmm2, %xmm14 mulps 0x164ae7(%rip), %xmm2 # 0x3f1cf0 addps %xmm1, %xmm2 movaps %xmm2, %xmm1 mulps %xmm2, %xmm1 movaps %xmm2, %xmm7 mulps 0x163e44(%rip), %xmm7 # 0x3f1060 addps 0x163e4d(%rip), %xmm7 # 0x3f1070 mulps %xmm2, %xmm7 addps 0x163e53(%rip), %xmm7 # 0x3f1080 mulps %xmm2, %xmm7 addps 0x163e59(%rip), %xmm7 # 0x3f1090 mulps %xmm2, %xmm7 addps 0x163e5f(%rip), %xmm7 # 0x3f10a0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm1, %xmm7 addps %xmm8, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm8, %xmm14 mulps %xmm2, %xmm14 addps %xmm8, %xmm14 rcpps %xmm14, %xmm1 mulps %xmm1, %xmm14 movaps %xmm8, %xmm7 subps %xmm14, %xmm7 mulps %xmm1, %xmm7 addps %xmm1, %xmm7 jmp 0x28d2a5 movq 0x118(%rbp,%r12), %rsi movss (%rsi), %xmm7 movss 0x4(%rsi), %xmm2 shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm7 addps %xmm2, %xmm7 maxps %xmm0, %xmm7 minps %xmm8, %xmm7 mulps %xmm1, %xmm7 movq %r9, %rsi shlq $0x4, %rsi movups %xmm7, (%r11,%rsi) incq %r9 jmp 0x28ce8d movq 0x30(%rsp), %rdx leaq (%r11,%rdx,4), %r11 incl %ecx jmp 0x28ce7c movq 0x70(%rsp), %rdx incq %rdx movq 0x60(%rsp), %rcx addl 0x90(%rsp), %ecx movq 0x48(%rsp), %r14 jmp 0x28ce15 leaq 0x198(%rsp), %rdi callq 0x624be xorl %r12d, %r12d jmp 0x28d3c9 movq %rsi, %rdi callq 0x5f3e0 xorl %r12d, %r12d movaps 0x130(%rsp), %xmm0 movq 0x48(%rsp), %rcx movups %xmm0, (%rcx) movq 0x140(%rsp), %rax movq %rax, 0x10(%rcx) movl 0x148(%rsp), %eax movl %eax, 0x18(%rcx) movq 0x150(%rsp), %rax movq %rax, 0x20(%rcx) movups 0x158(%rsp), %xmm0 movups %xmm0, 0x28(%rcx) movl 0x168(%rsp), %eax movl %eax, 0x38(%rcx) movq 0x170(%rsp), %rax movq %rax, 0x40(%rcx) movq 0x138(%rsp), %rax testq %rax, %rax je 0x28d392 lock decl (%rax) jne 0x28d392 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi je 0x28d38a movq (%rdi), %rax callq 0x18(%rax) jmp 0x28d392 movq %rsi, %rdi callq 0x5f3e0 movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x28d3c9 lock decl (%rax) jne 0x28d3c9 movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi je 0x28d3c1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28d3c9 movq %rsi, %rdi callq 0x5f3e0 movq 0xb8(%rsp), %rax testq %rax, %rax je 0x28d400 lock decl (%rax) jne 0x28d400 movq 0xb0(%rsp), %rsi movq 0xd0(%rsp), %rdi testq %rdi, %rdi je 0x28d3f8 movq (%rdi), %rax callq 0x18(%rax) jmp 0x28d400 movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x2c8, %rsp # imm = 0x2C8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq movq 0x8(%rbp), %rdi jmp 0x28d427 movq 0x88(%rsp), %rax movq 0x8(%rax), %rdi testq %rdi, %rdi je 0x28d2f4 movq (%rdi), %rax movq %r14, %rsi movq 0x190(%rsp), %rdx callq 0x48(%rax) jmp 0x28d2f4 jmp 0x28d473 jmp 0x28d5a1 jmp 0x28d5a1 movq %rax, %rbx jmp 0x28d523 movq %rax, %rbx jmp 0x28d4ec jmp 0x28d5a1 jmp 0x28d473 jmp 0x28d5a1 jmp 0x28d5a1 movq %rax, %rbx jmp 0x28d55a movq %rax, %rbx movq 0x1f8(%rsp), %rax testq %rax, %rax je 0x28d4b5 lock decl (%rax) jne 0x28d4b5 movq 0x1f0(%rsp), %rsi movq 0x210(%rsp), %rdi testq %rdi, %rdi jne 0x28d4af movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d4b5 movq (%rdi), %rax callq 0x18(%rax) movq 0x240(%rsp), %rax testq %rax, %rax je 0x28d4ec lock decl (%rax) jne 0x28d4ec movq 0x238(%rsp), %rsi movq 0x258(%rsp), %rdi testq %rdi, %rdi jne 0x28d4e6 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d4ec movq (%rdi), %rax callq 0x18(%rax) movq 0x138(%rsp), %rax testq %rax, %rax je 0x28d523 lock decl (%rax) jne 0x28d523 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi jne 0x28d51d movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d523 movq (%rdi), %rax callq 0x18(%rax) movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x28d55a lock decl (%rax) jne 0x28d55a movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi jne 0x28d554 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d55a movq (%rdi), %rax callq 0x18(%rax) movq 0xb8(%rsp), %rax testq %rax, %rax je 0x28d591 lock decl (%rax) jne 0x28d591 movq 0xb0(%rsp), %rsi movq 0xd0(%rsp), %rdi testq %rdi, %rdi jne 0x28d58b movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d591 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x28d5a1 jmp 0x28d5a1 jmp 0x28d5a1 jmp 0x28d5a1 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float)bias_data) + g 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float)bias_data) + g 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float)bias_data) + g 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x28a5b0	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
ncnn::ConvolutionDepthWise_x86::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int ConvolutionDepthWise_x86::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w = weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w = bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x168, %rsp # imm = 0x168 movq %rsi, %rbp movq %rdi, %r13 movq (%rsi), %r14 leaq 0x48(%r14), %rdi movq (%rdx), %rax movq %rax, 0xb8(%rsp) movl 0x60(%r14), %ebx movl 0x74(%r14), %eax movl %eax, 0x1c(%rsp) movl 0x78(%r14), %eax movl %eax, 0x18(%rsp) imull 0x80(%r14), %ebx leaq 0x70(%rsp), %rsi andq $0x0, 0x40(%rsi) xorps %xmm0, %xmm0 movaps %xmm0, (%rsi) movups %xmm0, 0xc(%rsi) movaps %xmm0, 0x20(%rsi) movups %xmm0, 0x2c(%rsi) movq %rcx, %r15 movq %rcx, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2904be movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x2904be movslq 0x88(%rsp), %rcx movl 0x9c(%rsp), %eax imull %ecx, %eax movl %eax, 0x9c(%rsp) movq 0x80(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x80(%rsp) movl $0x1, 0x88(%rsp) andq $0x0, 0x60(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x20(%rsp) movups %xmm0, 0x2c(%rsp) movaps %xmm0, 0x40(%rsp) movups %xmm0, 0x4c(%rsp) movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r13,%rax) je 0x29001e movl $0x90, %edi addq (%rbp), %rdi leaq 0x20(%rsp), %rsi movq %r15, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x20(%rsp) je 0x290490 movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x290490 movslq 0x38(%rsp), %rcx movl 0x4c(%rsp), %eax imull %ecx, %eax movl %eax, 0x4c(%rsp) movq 0x30(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x30(%rsp) movl $0x1, 0x38(%rsp) pushq $0x2a popq %rdi callq 0x782bf movq %rax, %r12 leaq 0x8(%rsp), %rdi callq 0x71548 leaq 0x8(%rsp), %rdi xorl %esi, %esi movl %ebx, %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi movl 0x1c(%rsp), %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0xb popq %rsi movl 0x18(%rsp), %edx callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xdc(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xd popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xec(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x4 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xf popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xe popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x10 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movss 0xfc(%r13,%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x12 popq %rsi callq 0x71952 movq (%r13), %rax movq -0x18(%rax), %rax movl 0x100(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x5 popq %rsi callq 0x7193a movl 0x9c(%rsp), %edx leaq 0x8(%rsp), %rdi pushq $0x6 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x108(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x7 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x10c(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x110(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax leaq (%rax,%r13), %rdx addq $0x118, %rdx # imm = 0x118 leaq 0x8(%rsp), %rdi pushq $0xa popq %rsi callq 0x7196c movq (%r12), %rax leaq 0x8(%rsp), %rsi movq %r12, %rdi callq 0x10(%rax) andq $0x0, 0x110(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0xd0(%rsp) movups %xmm0, 0xdc(%rsp) movaps %xmm0, 0xf0(%rsp) movups %xmm0, 0xfc(%rsp) andq $0x0, 0x158(%rsp) movups %xmm0, 0x118(%rsp) movups %xmm0, 0x124(%rsp) movups %xmm0, 0x138(%rsp) movups %xmm0, 0x144(%rsp) movq 0x78(%rsp), %rax testq %rax, %rax je 0x290283 lock incl (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2902ba lock decl (%rax) jne 0x2902ba movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x2902b2 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2902ba movq %rsi, %rdi callq 0x5f3e0 movaps 0x70(%rsp), %xmm0 movaps %xmm0, 0xd0(%rsp) movq 0x80(%rsp), %rax movq %rax, 0xe0(%rsp) movl 0x88(%rsp), %eax movl %eax, 0xe8(%rsp) movq 0x90(%rsp), %rax movq %rax, 0xf0(%rsp) movups 0x98(%rsp), %xmm0 movups %xmm0, 0xf8(%rsp) movl 0xa8(%rsp), %eax movl %eax, 0x108(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x110(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax je 0x290330 lock incl (%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x290367 lock decl (%rax) jne 0x290367 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x29035f movq (%rdi), %rax callq 0x18(%rax) jmp 0x290367 movq %rsi, %rdi callq 0x5f3e0 movaps 0x20(%rsp), %xmm0 leaq 0xd0(%rsp), %rsi movups %xmm0, 0x48(%rsi) movq 0x30(%rsp), %rax movq %rax, 0x58(%rsi) movl 0x38(%rsp), %eax movl %eax, 0x60(%rsi) movq 0x40(%rsp), %rax movq %rax, 0x68(%rsi) movups 0x48(%rsp), %xmm0 movaps %xmm0, 0x70(%rsi) movl 0x58(%rsp), %eax movl %eax, 0x80(%rsi) movq 0x60(%rsp), %rax movq %rax, 0x88(%rsi) leaq 0xc0(%rsp), %rdi callq 0x6b00e movq (%r12), %rax leaq 0xc0(%rsp), %rsi movq %r12, %rdi callq 0x18(%rax) leaq 0xc0(%rsp), %rdi callq 0x6b03a movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq 0x20(%rax) movq (%r12), %rax movq %r12, %rdi movq %r14, %rsi movq 0xb8(%rsp), %rdx movq %r15, %rcx callq 0x38(%rax) movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq 0x28(%rax) movq (%r12), %rax movq %r12, %rdi callq 0x8(%rax) pushq $0x48 popq %rbx xorps %xmm0, %xmm0 movq 0xd8(%rsp,%rbx), %rax testq %rax, %rax je 0x29045b lock decl (%rax) jne 0x29045b movq 0xd0(%rsp,%rbx), %rsi movq 0xf0(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x290450 movq (%rdi), %rax callq 0x18(%rax) xorps %xmm0, %xmm0 jmp 0x29045b movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%rbx), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x29041e leaq 0x8(%rsp), %rdi callq 0x71614 xorl %r12d, %r12d movq 0x28(%rsp), %rax testq %rax, %rax je 0x2904be lock decl (%rax) jne 0x2904be movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2904b6 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2904be movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2904ef lock decl (%rax) jne 0x2904ef movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2904e7 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2904ef movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x168, %rsp # imm = 0x168 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x29061e jmp 0x29061e movq %rax, %rbx leaq 0xc0(%rsp), %rdi callq 0x6b03a jmp 0x290531 jmp 0x29052e jmp 0x290526 jmp 0x290526 movq %rax, %rbx jmp 0x2905b5 movq %rax, %rbx pushq $0x48 popq %r14 xorps %xmm0, %xmm0 movq 0xd8(%rsp,%r14), %rax testq %rax, %rax je 0x290575 lock decl (%rax) jne 0x290575 movq 0xd0(%rsp,%r14), %rsi movq 0xf0(%rsp,%r14), %rdi testq %rdi, %rdi je 0x29056a movq (%rdi), %rax callq 0x18(%rax) xorps %xmm0, %xmm0 jmp 0x290575 movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x290538 jmp 0x2905ab jmp 0x29061e movq %rax, %rbx jmp 0x2905e3 jmp 0x29061e movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq 0x28(%rsp), %rax testq %rax, %rax je 0x2905e3 lock decl (%rax) jne 0x2905e3 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2905dd movq %rsi, %rdi callq 0x5f3e0 jmp 0x2905e3 movq (%rdi), %rax callq 0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x290614 lock decl (%rax) jne 0x290614 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x29060e movq %rsi, %rdi callq 0x5f3e0 jmp 0x290614 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x29061e movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int ConvolutionDepthWise_x86::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w = weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w = bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }	movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x28febe	/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp
ncnn::ConvolutionDepthWise_x86_avx512::create_group_ops(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx512::create_group_ops(const Option& opt) { // create Convolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Convolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[5]; weights[0] = weight_data_g; weights[1] = bias_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[2] = weight_data_int8_scales_g; weights[3] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[4] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[4]; weights[0] = weight_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[1] = weight_data_int8_scales_g; weights[2] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[3] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x278, %rsp # imm = 0x278 movq %rsi, 0x250(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x104(%rdi,%rdx), %eax movl 0x108(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x10(%rdi), %rax movq %rax, 0x1e0(%rsp) xorl %r12d, %r12d movq 0x10(%r14), %rax movq 0x18(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x29104f movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x29104a movq (%rdi), %rax callq 0x8(%rax) incq %r12 jmp 0x291021 imull %r15d, %ebx cmpq %rax, %rcx je 0x29105c movq %rax, 0x18(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x108(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0x1e0(%rsp), %rdi callq 0x6fbc2 leaq 0x118(%r14), %rax movq %rax, 0x258(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0x6c(%rsp) movslq %ebx, %rax movq %rax, 0x260(%rsp) movslq %r15d, %rax movq %rax, 0x248(%rsp) pushq $0x1 popq %rbx xorl %edx, %edx movl %r15d, 0x14(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rcx cmpq %rcx, %rdx jge 0x291dc6 movq %rdx, %rcx movq %rdx, 0x8(%rsp) movq 0x260(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdx, 0x80(%rsp) movl %esi, 0x88(%rsp) movq %rax, 0x90(%rsp) movl %ebx, 0x98(%rsp) movl %edi, 0x9c(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa0(%rsp) movl %ebx, 0xa8(%rsp) movq %rdi, 0xb0(%rsp) leaq 0x200(%rsp), %rdi leaq 0x70(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2911ad lock decl (%rax) jne 0x2911ad movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2911a5 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2911ad movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x29121d movq 0x8(%rsp), %r15 movq 0x248(%rsp), %rcx imulq %rcx, %r15 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0x1f8(%rsp) imulq %rsi, %r15 addq 0x1b0(%r14,%rax), %r15 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x1c(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0x1f0(%rsp) movl %ebx, %r13d movl 0x14(%rsp), %eax movl %eax, 0x18(%rsp) movq %rcx, 0x1e8(%rsp) jmp 0x291257 xorl %r15d, %r15d movq $0x0, 0x1f8(%rsp) movl $0x0, 0x1c(%rsp) movq $0x0, 0x1f0(%rsp) xorl %r13d, %r13d movl $0x0, 0x18(%rsp) movq $0x0, 0x1e8(%rsp) pushq $0x6 popq %rdi callq 0x782bf movq %rax, %rbp leaq 0x268(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq %r12, %rdi xorl %esi, %esi movl 0x14(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r12, %rdi movl %ebx, %esi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r12, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r12, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r12, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r12, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r12, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r12, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r12, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x6 popq %rsi movl 0x6c(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r12, %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x110(%r14,%rax), %edx movq %r12, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x258(%rsp), %rdx movq %r12, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r12, %rsi callq 0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x291444 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x1a8, %rax # imm = 0x1A8 jne 0x2913d2 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29140c lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2914c5 lock decl (%rax) jne 0x2914c5 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2914bd movq (%rdi), %rax callq 0x18(%rax) jmp 0x2914c5 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x160, %rax # imm = 0x160 jne 0x29144b movq 0x208(%rsp), %rax testq %rax, %rax je 0x291485 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29187c lock decl (%rax) jne 0x29187c movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x291874 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29187c movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x291569 lock decl (%rax) jne 0x291569 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x291561 movq (%rdi), %rax callq 0x18(%rax) jmp 0x291569 movq %rsi, %rdi callq 0x5f3e0 movq %r15, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq 0x1f8(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x1c(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x1f0(%rsp), %rax movq %rax, 0xd8(%rsp) movl %r13d, 0xe0(%rsp) movl 0x18(%rsp), %eax movl %eax, 0xe4(%rsp) movl %r13d, 0xe8(%rsp) movl %r13d, 0xec(%rsp) movl %r13d, 0xf0(%rsp) movq 0x1e8(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2917fe andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x29166e vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x291660 movq 0x28(%rsp), %rax testq %rax, %rax je 0x29167b lock incl (%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x2916b2 lock decl (%rax) jne 0x2916b2 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x2916aa movq (%rdi), %rax callq 0x18(%rax) jmp 0x2916b2 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovaps %xmm0, 0x100(%rsp) movq 0x30(%rsp), %rax movq %rax, 0x110(%rsp) movl 0x38(%rsp), %eax movl %eax, 0x118(%rsp) movq 0x40(%rsp), %rax movq %rax, 0x120(%rsp) vmovups 0x48(%rsp), %xmm0 vmovups %xmm0, 0x128(%rsp) movl 0x58(%rsp), %eax movl %eax, 0x138(%rsp) movq 0x60(%rsp), %rax movq %rax, 0x140(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %r12d movq 0x260(%r14,%rax), %rbx movq 0x150(%rsp), %rax testq %rax, %rax je 0x291774 lock decl (%rax) jne 0x291774 movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x29176c movq (%rdi), %rax callq 0x18(%rax) jmp 0x291774 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %r12d, 0x160(%rsp) movq %rbx, 0x168(%rsp) vbroadcastss 0x166e5a(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x2917f7 lock decl (%rax) jne 0x2917f7 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2917ef movq (%rdi), %rax callq 0x18(%rax) jmp 0x2917f7 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x291a1e movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x198(%rsp), %rax testq %rax, %rax je 0x2919ca lock decl (%rax) jne 0x2919ca movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x2919c2 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2919ca movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x291c01 andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x291977 vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x291969 movq 0x28(%rsp), %rax testq %rax, %rax je 0x291984 lock incl (%rax) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x291ab6 lock decl (%rax) jne 0x291ab6 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x291aae movq (%rdi), %rax callq 0x18(%rax) jmp 0x291ab6 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x190(%rsp) andq $0x0, 0x198(%rsp) movq %r15, 0x1a0(%rsp) movl %ebx, 0x1a8(%rsp) movq %r12, 0x1b0(%rsp) vbroadcastss 0x166c05(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x1b8(%rsp) movl $0x1, 0x1c8(%rsp) movq $0x1, 0x1d0(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0x120, %r15d # imm = 0x120 movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x291a7c lock decl (%rax) jne 0x291a7c movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x291a74 movq (%rdi), %rax callq 0x18(%rax) jmp 0x291a7c movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x291a4b jmp 0x291d4f movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovups %xmm0, 0xb8(%rsp) movq 0x30(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x38(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x40(%rsp), %rax movq %rax, 0xd8(%rsp) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movl 0x58(%rsp), %eax movl %eax, 0xf0(%rsp) movq 0x60(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %ebx movq 0x260(%r14,%rax), %r12 movq 0x108(%rsp), %rax testq %rax, %rax je 0x291b78 lock decl (%rax) jne 0x291b78 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x291b70 movq (%rdi), %rax callq 0x18(%rax) jmp 0x291b78 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x100(%rsp) andq $0x0, 0x108(%rsp) movq %r15, 0x110(%rsp) movl %ebx, 0x118(%rsp) movq %r12, 0x120(%rsp) vbroadcastss 0x166a57(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x128(%rsp) movl $0x1, 0x138(%rsp) movq $0x1, 0x140(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x291bfa lock decl (%rax) jne 0x291bfa movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x291bf2 movq (%rdi), %rax callq 0x18(%rax) jmp 0x291bfa movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x291cc4 movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x150(%rsp), %rax testq %rax, %rax je 0x291c70 lock decl (%rax) jne 0x291c70 movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x291c68 movq (%rdi), %rax callq 0x18(%rax) jmp 0x291c70 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %ebx, 0x160(%rsp) movq %r12, 0x168(%rsp) vbroadcastss 0x16695f(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0xd8, %r15d movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x291d22 lock decl (%rax) jne 0x291d22 movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x291d1a movq (%rdi), %rax callq 0x18(%rax) jmp 0x291d22 movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x291cf1 movq (%rbp), %rax movq %rbp, %rdi movq 0x250(%rsp), %rsi callq 0x20(%rax) leaq 0x268(%rsp), %rdi movq 0x1e0(%rsp), %rax movq (%rax), %rax movq 0x8(%rsp), %rcx movq %rbp, (%rax,%rcx,8) callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x291db9 lock decl (%rax) jne 0x291db9 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi je 0x291db1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x291db9 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rdx incq %rdx jmp 0x2910c7 xorl %eax, %eax addq $0x278, %rsp # imm = 0x278 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x291e8e jmp 0x292025 jmp 0x291de6 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x291ea0 lock decl (%rax) jne 0x291ea0 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x291e1c movq %rsi, %rdi callq 0x5f3e0 jmp 0x291ea0 movq (%rdi), %rax callq 0x18(%rax) jmp 0x291ea0 jmp 0x292025 jmp 0x291f18 jmp 0x291e8e jmp 0x291e32 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x291f2a lock decl (%rax) jne 0x291f2a movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x291e68 movq %rsi, %rdi callq 0x5f3e0 jmp 0x291f2a movq (%rdi), %rax callq 0x18(%rax) jmp 0x291f2a jmp 0x292025 jmp 0x292025 jmp 0x292025 jmp 0x292025 jmp 0x291e8e jmp 0x291f18 movq %rax, %rbx jmp 0x291ea0 movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0xd8, %r14d vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x291ee3 lock decl (%rax) jne 0x291ee3 movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x291ed7 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x291ee3 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x291eaa jmp 0x291fd9 jmp 0x292025 jmp 0x291f18 movq %rax, %rbx jmp 0x291f2a movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0x120, %r14d # imm = 0x120 vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x291f6d lock decl (%rax) jne 0x291f6d movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x291f61 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x291f6d movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x291f34 jmp 0x291fd9 jmp 0x292025 jmp 0x291fcd movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x29201d lock decl (%rax) jne 0x29201d movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29200d movq (%rdi), %rax callq 0x18(%rax) jmp 0x29201d jmp 0x292025 movq %rax, %rbx jmp 0x291fe6 jmp 0x292025 jmp 0x292025 movq %rax, %rbx leaq 0x268(%rsp), %rdi callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29201d lock decl (%rax) jne 0x29201d movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi jne 0x292017 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29201d movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp
ncnn::ConvolutionDepthWise_x86_avx512::destroy_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx512::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }	pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x292071 movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x29206c movq (%rdi), %rax callq 0x8(%rax) andq $0x0, 0x8(%rbx) xorl %r15d, %r15d movq 0x10(%rbx), %rax movq 0x18(%rbx), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r15 jge 0x2920b3 movq (%rax,%r15,8), %rdi movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x10(%rbx), %rax movq (%rax,%r15,8), %rdi testq %rdi, %rdi je 0x2920ae movq (%rdi), %rax callq 0x8(%rax) incq %r15 jmp 0x292074 cmpq %rax, %rcx je 0x2920bc movq %rax, 0x18(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %r15 retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_avx512::destroy_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx512::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }	pushq %rax movq (%rdi), %rax addq -0x38(%rax), %rdi callq 0x292040 xorl %eax, %eax popq %rcx retq nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp
ncnn::ConvolutionDepthWise_x86_avx512::forward_int8_x86(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx512::forward_int8_x86(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; int elempack = bottom_blob.elempack; int elembits = bottom_blob.elembits(); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_int8 = bottom_blob; if (elembits != 8) { const int channels_g = channels * elempack / group; Mat scales(channels * elempack); { float* ps = scales; for (int g = 0; g < group; g++) { float scale = bottom_blob_int8_scales[g]; for (int q = 0; q < channels_g; q++) { ps++ = scale; } } } Option opt_q = opt; opt_q.blob_allocator = opt.workspace_allocator; quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q); } Mat bottom_blob_bordered; make_padding(bottom_blob_int8, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; channels = bottom_blob_bordered.c; elempack = bottom_blob_bordered.elempack; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; // depth-wise if (channels elempack == group && group == num_output) { int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { out_elempack = num_output % 8 == 0 ? 8 : 1; } #endif // __SSE2__ bool use_int8_requantize = int8_scale_term > 100; size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #if __SSE2__ if (elempack == 8) { { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128i _sum0 = _mm_setzero_si128(); __m128i _sum1 = _mm_setzero_si128(); const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { // TODO use _mm_cvtepi8_epi16 on sse4.1 __m128i _val = _mm_loadl_epi64((const __m128i)(sptr + space_ofs[k] 8)); _val = _mm_unpacklo_epi8(_val, _mm_cmpgt_epi8(_mm_setzero_si128(), _val)); __m128i _w = _mm_loadl_epi64((const __m128i)(kptr + k 8)); _w = _mm_unpacklo_epi8(_w, _mm_cmpgt_epi8(_mm_setzero_si128(), _w)); __m128i _sl = _mm_mullo_epi16(_val, _w); __m128i _sh = _mm_mulhi_epi16(_val, _w); __m128i _s0 = _mm_unpacklo_epi16(_sl, _sh); __m128i _s1 = _mm_unpackhi_epi16(_sl, _sh); _sum0 = _mm_add_epi32(_sum0, _s0); _sum1 = _mm_add_epi32(_sum1, _s1); } __m128 _scale_in0; __m128 _scale_in1; { __m128 _bottom_blob_int8_scales0 = _mm_loadu_ps((const float)bottom_blob_int8_scales + g 8); __m128 _bottom_blob_int8_scales1 = _mm_loadu_ps((const float)bottom_blob_int8_scales + g 8 + 4); __m128 _weight_data_int8_scales0 = _mm_loadu_ps((const float)weight_data_int8_scales + g 8); __m128 _weight_data_int8_scales1 = _mm_loadu_ps((const float)weight_data_int8_scales + g 8 + 4); _scale_in0 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales0, _weight_data_int8_scales0)); _scale_in1 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales1, _weight_data_int8_scales1)); __m128 _m0 = _mm_cmpneq_ps(_weight_data_int8_scales0, _mm_setzero_ps()); __m128 _m1 = _mm_cmpneq_ps(_weight_data_int8_scales1, _mm_setzero_ps()); _scale_in0 = _mm_and_ps(_scale_in0, _m0); _scale_in1 = _mm_and_ps(_scale_in1, _m1); } __m128 _sumfp32_0 = _mm_mul_ps(_mm_cvtepi32_ps(_sum0), _scale_in0); __m128 _sumfp32_1 = _mm_mul_ps(_mm_cvtepi32_ps(_sum1), _scale_in1); if (bias_term) { __m128 _bias0 = _mm_loadu_ps((const float)bias_data + g 8); __m128 _bias1 = _mm_loadu_ps((const float)bias_data + g 8 + 4); _sumfp32_0 = _mm_add_ps(_sumfp32_0, _bias0); _sumfp32_1 = _mm_add_ps(_sumfp32_1, _bias1); } _sumfp32_0 = activation_sse(_sumfp32_0, activation_type, activation_params); _sumfp32_1 = activation_sse(_sumfp32_1, activation_type, activation_params); if (use_int8_requantize) { // requantize and relu __m128 _scale_out0 = _mm_loadu_ps((const float)top_blob_int8_scales + g 8); __m128 _scale_out1 = _mm_loadu_ps((const float)top_blob_int8_scales + g 8 + 4); _sumfp32_0 = _mm_mul_ps(_sumfp32_0, _scale_out0); _sumfp32_1 = _mm_mul_ps(_sumfp32_1, _scale_out1); int64_t _sum8 = float2int8_sse(_sumfp32_0, _sumfp32_1); (int64_t)outptr_s8 = _sum8; outptr_s8 += 8; } else { // dequantize and relu _mm_storeu_ps(outptr_f32, _sumfp32_0); _mm_storeu_ps(outptr_f32 + 4, _sumfp32_1); outptr_f32 += 8; } } } } } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && stride_w == 1 && stride_h == 1 && dilation_w == 1 && dilation_h == 1 && (activation_type == 0 \|\| activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s1_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s1_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2 && (activation_type == 0 \|\| activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s2_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s2_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char)weight_data_tm + maxk g; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr_s8[0] = sums8; outptr_s8 += 1; } else { // dequantize outptr_f32[0] = sumfp32; outptr_f32 += 1; } } } } } } return 0; } bool use_int8_requantize = int8_scale_term > 100; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { if (use_int8_requantize) out_elempack = num_output % 8 == 0 ? 8 : 1; else out_elempack = num_output % 4 == 0 ? 4 : 1; } #endif // __SSE2__ size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { g_elempack = channels_g % 8 == 0 ? 8 : 1; if (use_int8_requantize) out_g_elempack = num_output_g % 8 == 0 ? 8 : 1; else out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x318, %rsp # imm = 0x318 movq %rcx, 0x30(%rsp) movq %rdx, 0x20(%rsp) movq %rsi, %r15 movl 0x38(%rsi), %ecx movl 0x18(%rsi), %r12d movq 0x10(%rsi), %rsi testl %r12d, %r12d je 0x2975b9 leal (,%rsi,8), %eax cltd idivl %r12d cmpl $0x8, %eax sete %dl jmp 0x2975bb xorl %edx, %edx movq (%rdi), %rax movq -0x18(%rax), %r9 movq %rdi, %r8 movl 0xd4(%rdi,%r9), %ebp movl 0xd8(%rdi,%r9), %ebx decl %ebp imull 0xdc(%rdi,%r9), %ebp decl %ebx imull 0xe0(%rdi,%r9), %ebx movq 0x8(%r15), %rdi vmovups (%r15), %xmm0 vmovaps %xmm0, 0x1e0(%rsp) movq %rsi, 0x1f0(%rsp) movl %r12d, 0x1f8(%rsp) movq 0x20(%r15), %rsi movq %rsi, 0x200(%rsp) vmovdqu 0x28(%r15), %xmm0 vmovdqu %xmm0, 0x208(%rsp) movl %ecx, 0x218(%rsp) movq 0x40(%r15), %rsi movq %rsi, 0x220(%rsp) testq %rdi, %rdi je 0x297646 lock incl (%rdi) movq (%r8), %rax movq %r8, 0x10(%rsp) testb %dl, %dl je 0x29765e movq 0x20(%rsp), %r15 movq 0x10(%rsp), %rcx jmp 0x297770 imull %ecx, %r12d movq -0x18(%rax), %rax movq 0x10(%rsp), %r13 movl 0x108(%r13,%rax), %eax movl %eax, 0x18(%rsp) leaq 0x70(%rsp), %rdi andq $0x0, 0x40(%rdi) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdi) vmovdqu %xmm0, 0xc(%rdi) vmovdqa %xmm0, 0x20(%rdi) vmovdqu %xmm0, 0x2c(%rdi) xorl %r14d, %r14d pushq $0x4 popq %rdx movl %r12d, %esi xorl %ecx, %ecx callq 0x635fa movl %r12d, %eax cltd idivl 0x18(%rsp) movq 0x70(%rsp), %rcx movq (%r13), %rdx testl %eax, %eax cmovlel %r14d, %eax movq 0x10(%rsp), %r8 movq -0x18(%rdx), %rsi movslq 0x108(%r8,%rsi), %rdi cmpq %rdi, %r14 jge 0x2976f9 movq 0x240(%r8,%rsi), %rsi vmovd (%rsi,%r14,4), %xmm0 movl %eax, %esi subl $0x1, %esi jb 0x2976f4 vmovd %xmm0, (%rcx) addq $0x4, %rcx jmp 0x2976e5 incq %r14 jmp 0x2976c4 movq 0x30(%rsp), %rax vmovups (%rax), %zmm0 leaq 0x130(%rsp), %rcx vmovups %zmm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x1e0(%rsp), %rsi leaq 0x70(%rsp), %rdx movq %r15, %rdi vzeroupper callq 0x652e3 movq 0x78(%rsp), %rax testq %rax, %rax movq 0x20(%rsp), %r15 je 0x297768 lock decl (%rax) jne 0x297768 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x297760 movq (%rdi), %rax callq 0x18(%rax) jmp 0x297768 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rcx movq (%rcx), %rax leaq 0x70(%rsp), %rdx andq $0x0, 0x40(%rdx) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdx) vmovdqu %xmm0, 0xc(%rdx) vmovdqa %xmm0, 0x20(%rdx) vmovdqu %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %rcx, %rdi leaq 0x1e0(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %rax cmpq $0x0, 0x70(%rsp) movl %eax, 0x48(%rsp) je 0x298fc6 movslq 0xa8(%rsp), %r14 movq 0xb0(%rsp), %rax imulq %r14, %rax testq %rax, %rax je 0x298fc6 notl %ebp movl 0x88(%rsp), %r13d movl 0x9c(%rsp), %r12d addl %r12d, %ebp movq 0x10(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx movl %ebp, %eax movq %rdx, %rbp cltd idivl 0xe4(%rbp,%rcx) movl %eax, %esi notl %ebx addl 0xa0(%rsp), %ebx movl %ebx, %eax cltd idivl 0xe8(%rbp,%rcx) movl %eax, %edx movq %rsi, 0x18(%rsp) incl %esi movq %rdx, 0x60(%rsp) leal 0x1(%rdx), %r10d movl %r13d, %ebx imull %r14d, %ebx pushq $0x8 popq %r11 cmpl 0x108(%rbp,%rcx), %ebx jne 0x297980 cmpl 0xd0(%rbp,%rcx), %ebx jne 0x297980 movq 0x30(%rsp), %rdi cmpb $0x0, 0x27(%rdi) pushq $0x1 popq %rax cmovel %eax, %r11d testb $0x7, %bl cmovnel %eax, %r11d movl 0x10c(%rbp,%rcx), %eax leal (,%r11,4), %r8d movl %eax, 0x38(%rsp) cmpl $0x65, %eax cmovgel %r11d, %r8d movl %ebx, %eax cltd idivl %r11d movq 0x8(%rdi), %rcx movq %rcx, (%rsp) movq %r15, %rdi movl %r10d, %edx movl %eax, %ecx movl %r11d, %r9d callq 0x628f2 cmpq $0x0, (%r15) je 0x298fc6 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x298fc6 cmpl $0x1, %r13d je 0x297f2e movl $0x0, 0x48(%rsp) cmpl $0x8, %r13d jne 0x298fc6 movq (%rbp), %rax movq -0x18(%rax), %rax movslq 0xd4(%rbp,%rax), %rcx movslq 0xd8(%rbp,%rax), %r15 imulq %rcx, %r15 leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %r15, %rsi callq 0x73bbe movq (%rbp), %rcx movq -0x18(%rcx), %rdx imull 0xe0(%rbp,%rdx), %r12d movq 0x130(%rsp), %rax movl 0xdc(%rbp,%rdx), %esi imull 0xd4(%rbp,%rdx), %esi subl %esi, %r12d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%rbp,%rdx), %esi jge 0x298067 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%rbp,%rdx), %r9d jge 0x297976 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%rbp,%rdx), %edi incq %r9 jmp 0x297958 addl %r12d, %edi incl %esi addq %r9, %r8 jmp 0x297941 movl 0xd0(%rbp,%rcx), %eax movl 0x10c(%rbp,%rcx), %r14d movq 0x30(%rsp), %rcx cmpb $0x1, 0x27(%rcx) jne 0x2979ac cmpl $0x65, %r14d jl 0x2979b2 testb $0x7, %al pushq $0x1 popq %r9 cmovel %r11d, %r9d jmp 0x2979c0 pushq $0x1 popq %r9 jmp 0x2979c0 xorl %ecx, %ecx testb $0x3, %al sete %cl leal (%rcx,%rcx,2), %r9d incl %r9d leal (,%r9,4), %r12d cmpl $0x65, %r14d cmovgel %r9d, %r12d cltd idivl %r9d movq 0x30(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq %r15, %rdi movl %esi, 0x28(%rsp) movl %r10d, 0x40(%rsp) movl %r10d, %edx movl %eax, %ecx movq %r12, %r8 movq %r9, 0x38(%rsp) callq 0x628f2 cmpq $0x0, (%r15) je 0x298fc6 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x298fc6 movq (%rbp), %rax movq -0x18(%rax), %rax movl 0xd0(%rbp,%rax), %ecx movl 0x108(%rbp,%rax), %esi movl %ebx, %eax cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %ebp movq 0x30(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x297a73 testb $0x7, %dil pushq $0x1 popq %rbx movl %ebx, %r15d pushq $0x8 popq %rax cmovel %eax, %r15d cmpl $0x65, %r14d movl 0x28(%rsp), %esi movl 0x40(%rsp), %r9d jl 0x297a85 testb $0x7, %bpl cmovel %eax, %ebx jmp 0x297a93 pushq $0x1 popq %r15 movl %r15d, %ebx movl 0x28(%rsp), %esi movl 0x40(%rsp), %r9d jmp 0x297a93 xorl %eax, %eax testb $0x3, %bpl sete %al leal (%rax,%rax,2), %ebx incl %ebx movq 0x78(%rsp), %rax vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0x130(%rsp) movq 0x80(%rsp), %rcx movq %rcx, 0x140(%rsp) movl 0x88(%rsp), %ecx movl %ecx, 0x148(%rsp) movq 0x90(%rsp), %rcx movq %rcx, 0x150(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0x158(%rsp) movl 0xa8(%rsp), %ecx movl %ecx, 0x168(%rsp) movq 0xb0(%rsp), %rcx movq %rcx, 0x170(%rsp) testq %rax, %rax je 0x297b0d lock incl (%rax) cmpl %r15d, %r13d movl %edi, 0x18(%rsp) jle 0x297b5c movq 0x30(%rsp), %rax vmovups (%rax), %zmm0 leaq 0xe0(%rsp), %rcx vmovups %zmm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x70(%rsp), %rdi leaq 0x130(%rsp), %rsi movl %r15d, %edx vzeroupper callq 0x64e3b movl 0x18(%rsp), %edi movl 0x28(%rsp), %esi movl 0x40(%rsp), %r9d movq 0x20(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movq 0x10(%rdx), %rcx movq %rcx, 0xf0(%rsp) movl 0x18(%rdx), %ecx movl %ecx, 0xf8(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x100(%rsp) vmovups 0x28(%rdx), %xmm0 vmovups %xmm0, 0x108(%rsp) movl 0x38(%rdx), %ecx movl %ecx, 0x118(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x120(%rsp) testq %rax, %rax je 0x297bc0 lock incl (%rax) cmpl 0x38(%rsp), %ebx jae 0x297c3f movq 0x10(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xd0(%rcx,%rax), %eax cltd idivl %ebx movl %eax, %ecx movzbl %r12b, %eax divb 0x38(%rsp) movzbl %al, %r8d imull %ebx, %r8d movq 0x30(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0xe0(%rsp), %rdi movl %r9d, %edx movl %ebx, %r9d callq 0x628f2 movl 0x18(%rsp), %edi pushq $-0x64 popq %r14 cmpq $0x0, 0xe0(%rsp) je 0x298f53 movslq 0x118(%rsp), %rax imulq 0x120(%rsp), %rax testq %rax, %rax je 0x298f53 xorl %r12d, %r12d xorl %r13d, %r13d xorl %r14d, %r14d movq 0x10(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rax movslq 0x108(%r8,%rax), %rax cmpq %rax, %r14 jge 0x297eaa movl %r12d, %eax cltd idivl %r15d movl %eax, %ecx movl %edi, %eax cltd idivl %r15d movslq %ecx, %rdx imulq 0x170(%rsp), %rdx movq 0x140(%rsp), %rcx imulq %rcx, %rdx addq 0x130(%rsp), %rdx movq %rdx, 0x190(%rsp) andq $0x0, 0x198(%rsp) movl 0x148(%rsp), %edx movq 0x150(%rsp), %rsi movq %rcx, 0x1a0(%rsp) movl %edx, 0x1a8(%rsp) movq %rsi, 0x1b0(%rsp) movl %eax, 0x1c8(%rsp) vmovups 0x158(%rsp), %xmm0 movslq 0x164(%rsp), %rax movslq 0x15c(%rsp), %rdx movslq 0x160(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x1d0(%rsp) vmovups %xmm0, 0x1b8(%rsp) movl %r13d, %eax cltd idivl %ebx movl %eax, %ecx movl %ebp, %eax cltd idivl %ebx movslq %ecx, %rdx imulq 0x120(%rsp), %rdx movq 0xf0(%rsp), %rcx imulq %rcx, %rdx addq 0xe0(%rsp), %rdx movq %rdx, 0x228(%rsp) andq $0x0, 0x230(%rsp) movl 0xf8(%rsp), %edx movq 0x100(%rsp), %rsi movq %rcx, 0x238(%rsp) movl %edx, 0x240(%rsp) movq %rsi, 0x248(%rsp) movl %eax, 0x260(%rsp) vmovups 0x108(%rsp), %xmm0 movslq 0x114(%rsp), %rax movslq 0x10c(%rsp), %rdx movslq 0x110(%rsp), %rdi imulq %rdx, %rdi imulq %rcx, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x268(%rsp) vmovups %xmm0, 0x250(%rsp) movq 0x10(%r8), %rax movq (%rax,%r14,8), %rdi movq 0x30(%rsp), %rax vmovups (%rax), %zmm0 vmovups %zmm0, 0x2d0(%rsp) movq %rsi, 0x2d8(%rsp) movq (%rdi), %rax leaq 0x190(%rsp), %rsi leaq 0x228(%rsp), %rdx leaq 0x2d0(%rsp), %rcx vzeroupper callq 0x38(%rax) movq 0x230(%rsp), %rax testq %rax, %rax movl 0x18(%rsp), %edi je 0x297e5d lock decl (%rax) jne 0x297e5d movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi je 0x297e51 movq (%rdi), %rax callq 0x18(%rax) movl 0x18(%rsp), %edi jmp 0x297e5d movq %rsi, %rdi callq 0x5f3e0 movl 0x18(%rsp), %edi movq 0x198(%rsp), %rax testq %rax, %rax je 0x297e9c lock decl (%rax) jne 0x297e9c movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x297e90 movq (%rdi), %rax callq 0x18(%rax) movl 0x18(%rsp), %edi jmp 0x297e9c movq %rsi, %rdi callq 0x5f3e0 movl 0x18(%rsp), %edi incq %r14 addl %ebp, %r13d addl %edi, %r12d jmp 0x297c48 movq 0x38(%rsp), %rdx xorl %r14d, %r14d cmpl %edx, %ebx jae 0x297ed2 leaq 0xe0(%rsp), %rdi movq 0x20(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x64e3b jmp 0x298f53 leaq 0xe0(%rsp), %rax movq 0x20(%rsp), %rcx cmpq %rcx, %rax je 0x298f53 movq 0xe8(%rsp), %rax testq %rax, %rax je 0x297ef8 lock incl (%rax) movq 0x8(%rcx), %rax testq %rax, %rax je 0x298efb lock decl (%rax) jne 0x298efb movq 0x20(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x298ef3 movq (%rdi), %rax callq 0x18(%rax) jmp 0x298efb movq (%rbp), %rax movq -0x18(%rax), %rdx movl 0xd4(%rbp,%rdx), %ecx movl 0xd8(%rbp,%rdx), %r13d movl %ecx, %esi xorl $0x3, %esi movl %r13d, %edi xorl $0x3, %edi orl %esi, %edi jne 0x298ab0 cmpl $0x1, 0xe4(%rbp,%rdx) jne 0x29899d cmpl $0x1, 0xe8(%rbp,%rdx) jne 0x29899d cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x29899d cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x29899d cmpl $0x1, 0x110(%rbp,%rdx) ja 0x29899d movl $0x0, 0x48(%rsp) cmpl $0x65, 0x38(%rsp) jl 0x299046 vpxor %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %r13 vmovdqa %xmm0, (%r13) andq $0x0, 0x10(%r13) xorl %ebx, %ebx movq %rbp, %rcx leaq 0xe0(%rsp), %rbp leaq 0x190(%rsp), %r12 movq -0x18(%rax), %r14 movslq 0x108(%rcx,%r14), %rax cmpq %rax, %rbx jge 0x299347 movq 0x1f8(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm1 vpxor %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x298022 movq 0x240(%rcx,%r14), %rax vmulss (%rax,%rbx,4), %xmm1, %xmm0 vmovss 0x156c6a(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rcx,%r14), %rax vmovd (%rax,%rbx,4), %xmm0 vmovd %xmm0, 0x190(%rsp) movq %r13, %rdi movq %rbp, %rsi callq 0x1ea12c movq %r13, %rdi movq %r12, %rsi callq 0x1ea12c incq %rbx movq 0x10(%rsp), %rcx movq (%rcx), %rax jmp 0x297fdd leal (,%r15,8), %ecx movl %ecx, 0x40(%rsp) xorl %edx, %edx testl %r15d, %r15d cmovlel %edx, %r15d testl %r14d, %r14d cmovlel %edx, %r14d vpxor %xmm0, %xmm0, %xmm0 vbroadcastss 0x159124(%rip), %xmm1 # 0x3f11b4 vpbroadcastd 0x155f7b(%rip), %xmm2 # 0x3ee014 vpbroadcastw 0x1604fd(%rip), %xmm25 # 0x3f85a0 vpbroadcastw 0x1604f5(%rip), %xmm26 # 0x3f85a2 vbroadcastss 0x159101(%rip), %xmm27 # 0x3f11b8 vbroadcastss 0x1590fb(%rip), %xmm28 # 0x3f11bc vbroadcastss 0x1590f6(%rip), %xmm7 # 0x3f11c0 vbroadcastss 0x155f41(%rip), %xmm8 # 0x3ee014 vbroadcastss 0x156bac(%rip), %xmm9 # 0x3eec88 vbroadcastss 0x1590df(%rip), %xmm10 # 0x3f11c4 vbroadcastss 0x1590da(%rip), %xmm11 # 0x3f11c8 vbroadcastss 0x1590d8(%rip), %xmm24 # 0x3f11d0 vbroadcastss 0x1590cb(%rip), %xmm13 # 0x3f11cc vbroadcastss 0x1590ca(%rip), %xmm14 # 0x3f11d4 vbroadcastss 0x1590c5(%rip), %xmm15 # 0x3f11d8 vbroadcastss 0x1590bf(%rip), %xmm16 # 0x3f11dc vpbroadcastd 0x156b61(%rip), %xmm17 # 0x3eec88 xorl %esi, %esi movq %r14, 0x28(%rsp) vbroadcastss 0x1590bc(%rip), %xmm20 # 0x3f11f4 vbroadcastss 0x1590ba(%rip), %xmm23 # 0x3f11fc cmpq %r14, %rsi je 0x298ed9 movq 0x20(%rsp), %rcx movq 0x10(%rcx), %r8 imulq %rsi, %r8 imulq 0x40(%rcx), %r8 movq %rdx, 0x30(%rsp) movslq %edx, %r9 addq (%rcx), %r8 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rsi, %r11 movq 0x80(%rsp), %rcx imulq %rcx, %r11 addq 0x70(%rsp), %r11 imulq %rcx, %r10 movq %rsi, 0x48(%rsp) leaq (,%rsi,8), %rbx addq 0x28(%rbp), %r9 xorl %ecx, %ecx movq %r8, %rdx cmpl 0x60(%rsp), %ecx jg 0x298982 movq (%rbp), %rdi xorl %r13d, %r13d cmpl 0x18(%rsp), %r13d jg 0x29897b movq -0x18(%rdi), %rsi movslq 0xe8(%rbp,%rsi), %r14 movslq %ecx, %r12 imulq %r14, %r12 imulq %r10, %r12 movl 0xe4(%rbp,%rsi), %ebp imull %r13d, %ebp shll $0x3, %ebp movslq %ebp, %r14 addq %r11, %r14 addq %r12, %r14 vpxor %xmm3, %xmm3, %xmm3 xorl %r12d, %r12d vpxor %xmm4, %xmm4, %xmm4 cmpq %r12, %r15 je 0x29823f movslq (%rax,%r12,4), %rbp vmovq (%r14,%rbp,8), %xmm5 vpcmpgtb %xmm5, %xmm0, %xmm6 vpunpcklbw %xmm6, %xmm5, %xmm5 # xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7] vmovq (%r9,%r12,8), %xmm6 vpcmpgtb %xmm6, %xmm0, %xmm12 vpunpcklbw %xmm12, %xmm6, %xmm6 # xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7] vpmullw %xmm5, %xmm6, %xmm12 vpmulhw %xmm6, %xmm5, %xmm5 vpunpcklwd %xmm5, %xmm12, %xmm6 # xmm6 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3] vpunpckhwd %xmm5, %xmm12, %xmm5 # xmm5 = xmm12[4],xmm5[4],xmm12[5],xmm5[5],xmm12[6],xmm5[6],xmm12[7],xmm5[7] vpaddd %xmm6, %xmm4, %xmm4 vpaddd %xmm5, %xmm3, %xmm3 incq %r12 jmp 0x2981fc movq 0x10(%rsp), %rbp movq 0x1f8(%rbp,%rsi), %r14 movq 0x240(%rbp,%rsi), %r12 vmovups (%r14,%rbx,4), %xmm21 vmovups 0x10(%r14,%rbx,4), %xmm22 vmulps (%r12,%rbx,4), %xmm21, %xmm5 vmulps 0x10(%r12,%rbx,4), %xmm22, %xmm6 vrcpps %xmm5, %xmm5 vrcpps %xmm6, %xmm6 vcmpneqps %xmm0, %xmm21, %k1 vcmpneqps %xmm0, %xmm22, %k2 vmovaps %xmm5, %xmm5 {%k1} {z} vmovaps %xmm6, %xmm6 {%k2} {z} vcvtdq2ps %xmm4, %xmm4 vmulps %xmm4, %xmm5, %xmm22 vcvtdq2ps %xmm3, %xmm3 vmulps %xmm3, %xmm6, %xmm21 cmpl $0x0, 0x100(%rbp,%rsi) je 0x2982c9 movq 0x1b0(%rbp,%rsi), %r14 vaddps (%r14,%rbx,4), %xmm22, %xmm22 vaddps 0x10(%r14,%rbx,4), %xmm21, %xmm21 movl 0x110(%rbp,%rsi), %r14d decl %r14d cmpl $0x5, %r14d ja 0x2988fa leaq 0x1602d7(%rip), %r12 # 0x3f85bc movslq (%r12,%r14,4), %r14 addq %r12, %r14 vmaxps %xmm0, %xmm22, %xmm3 vmaxps %xmm0, %xmm21, %xmm4 jmpq %r14 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm5 vminps %xmm0, %xmm22, %xmm6 vfmadd231ps %xmm6, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm6) + xmm3 vminps %xmm0, %xmm21, %xmm6 vfmadd231ps %xmm6, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm6) + xmm4 jmp 0x298906 vminps %xmm27, %xmm22, %xmm3 vmaxps %xmm28, %xmm3, %xmm3 vmovaps %xmm7, %xmm4 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm5 vcvtdq2ps %xmm5, %xmm5 vcmpltps %xmm5, %xmm4, %k1 vsubps %xmm9, %xmm5, %xmm5 {%k1} vfmsub231ps %xmm10, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm10) - xmm3 vfnmsub231ps %xmm11, %xmm5, %xmm3 # xmm3 = -(xmm5 * xmm11) - xmm3 vmulps %xmm3, %xmm3, %xmm4 vmovaps %xmm13, %xmm6 vfmadd213ps %xmm24, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm24 vfmadd213ps %xmm14, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm14 vfmadd213ps %xmm15, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm15 vfmadd213ps %xmm16, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm16 vfmadd213ps %xmm8, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm8 vfmadd213ps %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vaddps %xmm6, %xmm9, %xmm3 vcvttps2dq %xmm5, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm17, %xmm4, %xmm4 vfmadd213ps %xmm9, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm9 vcmpleps %xmm0, %xmm4, %k1 vbroadcastss 0x158e38(%rip), %xmm29 # 0x3f11e0 vmaxps %xmm29, %xmm4, %xmm3 vpsrld $0x17, %xmm3, %xmm4 vpbroadcastd 0x158e27(%rip), %xmm30 # 0x3f11e4 vpternlogd $0xec, %xmm30, %xmm2, %xmm3 vmovaps %xmm1, %xmm30 vmovaps %xmm13, %xmm1 vmovdqa64 %xmm17, %xmm13 vmovaps %xmm16, %xmm17 vmovaps %xmm15, %xmm16 vmovaps %xmm14, %xmm15 vmovaps %xmm7, %xmm14 vmovaps %xmm24, %xmm7 vpbroadcastd 0x158def(%rip), %xmm24 # 0x3f11e8 vpaddd %xmm24, %xmm4, %xmm4 vcvtdq2ps %xmm4, %xmm4 vbroadcastss 0x158ddf(%rip), %xmm31 # 0x3f11ec vcmpltps %xmm31, %xmm3, %k2 vbroadcastss 0x158dd2(%rip), %xmm18 # 0x3f11f0 vaddps %xmm18, %xmm3, %xmm5 vsubps %xmm9, %xmm4, %xmm4 {%k2} vaddps %xmm3, %xmm5, %xmm5 {%k2} vmulps %xmm5, %xmm5, %xmm3 vmovaps %xmm20, %xmm6 vbroadcastss 0x158db4(%rip), %xmm19 # 0x3f11f8 vfmadd213ps %xmm19, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm19 vfmadd213ps %xmm23, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm23 vbroadcastss 0x158da6(%rip), %xmm18 # 0x3f1200 vfmadd213ps %xmm18, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm18 vbroadcastss 0x158d9a(%rip), %xmm19 # 0x3f1204 vfmadd213ps %xmm19, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm19 vbroadcastss 0x158d8e(%rip), %xmm24 # 0x3f1208 vfmadd213ps %xmm24, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm24 vbroadcastss 0x158d82(%rip), %xmm29 # 0x3f120c vfmadd213ps %xmm29, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm29 vbroadcastss 0x158d76(%rip), %xmm31 # 0x3f1210 vfmadd213ps %xmm31, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm31 vbroadcastss 0x158d6b(%rip), %xmm12 # 0x3f1214 vfmadd213ps %xmm12, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm12 vmulps %xmm5, %xmm3, %xmm12 vmulps %xmm6, %xmm12, %xmm6 vfmadd231ps %xmm11, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm11) + xmm6 vfmsub231ps %xmm3, %xmm8, %xmm6 # xmm6 = (xmm8 * xmm3) - xmm6 vsubps %xmm5, %xmm6, %xmm3 vfmsub231ps %xmm4, %xmm10, %xmm3 # xmm3 = (xmm10 * xmm4) - xmm3 vbroadcastss 0x15a246(%rip), %xmm4 # 0x3f2718 vmulps %xmm4, %xmm3, %xmm3 vbroadcastss 0x159421(%rip), %xmm4 # 0x3f1900 vmovaps %xmm4, %xmm3 {%k1} vminps %xmm27, %xmm3, %xmm3 vmaxps %xmm28, %xmm3, %xmm3 vmovaps %xmm14, %xmm4 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm5 vcvtdq2ps %xmm5, %xmm5 vcmpltps %xmm5, %xmm4, %k1 vsubps %xmm9, %xmm5, %xmm5 {%k1} vfmsub231ps %xmm10, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm10) - xmm3 vfnmsub231ps %xmm11, %xmm5, %xmm3 # xmm3 = -(xmm5 * xmm11) - xmm3 vmulps %xmm3, %xmm3, %xmm4 vmovaps %xmm1, %xmm6 vfmadd213ps %xmm7, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm7 vfmadd213ps %xmm15, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm15 vfmadd213ps %xmm16, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm16 vfmadd213ps %xmm17, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm17 vfmadd213ps %xmm8, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm8 vfmadd213ps %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vaddps %xmm6, %xmm9, %xmm4 vcvttps2dq %xmm5, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm3, %xmm13, %xmm3 vfmadd213ps %xmm9, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) + xmm9 vrcpps %xmm3, %xmm4 vaddps %xmm4, %xmm4, %xmm5 vbroadcastss 0x15a1a0(%rip), %xmm6 # 0x3f2708 vfmsub213ps %xmm6, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm3) - xmm6 vfnmadd213ps %xmm5, %xmm4, %xmm3 # xmm3 = -(xmm4 * xmm3) + xmm5 vminps %xmm27, %xmm21, %xmm4 vmaxps %xmm28, %xmm4, %xmm4 vmovaps %xmm14, %xmm5 vfmadd213ps %xmm8, %xmm4, %xmm5 # xmm5 = (xmm4 * xmm5) + xmm8 vcvttps2dq %xmm5, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm5, %k1 vsubps %xmm9, %xmm6, %xmm6 {%k1} vfmsub231ps %xmm10, %xmm6, %xmm4 # xmm4 = (xmm6 * xmm10) - xmm4 vfnmsub231ps %xmm11, %xmm6, %xmm4 # xmm4 = -(xmm6 * xmm11) - xmm4 vmulps %xmm4, %xmm4, %xmm5 vmovaps %xmm1, %xmm12 vfmadd213ps %xmm7, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm7 vfmadd213ps %xmm15, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm15 vfmadd213ps %xmm16, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm16 vfmadd213ps %xmm17, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm17 vfmadd213ps %xmm8, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm8 vfmadd213ps %xmm4, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm4 vaddps %xmm9, %xmm12, %xmm4 vcvttps2dq %xmm6, %xmm5 vpslld $0x17, %xmm5, %xmm5 vpaddd %xmm5, %xmm13, %xmm5 vfmadd213ps %xmm9, %xmm4, %xmm5 # xmm5 = (xmm4 * xmm5) + xmm9 vcmpleps %xmm0, %xmm5, %k1 vmaxps 0x158bea(%rip){1to4}, %xmm5, %xmm4 # 0x3f11e0 vpsrld $0x17, %xmm4, %xmm5 vpternlogd $0xec, 0x158bde(%rip){1to4}, %xmm2, %xmm4 # 0x3f11e4 vcmpltps 0x158bdb(%rip){1to4}, %xmm4, %k2 # 0x3f11ec vaddps 0x158bd5(%rip){1to4}, %xmm4, %xmm6 # 0x3f11f0 vaddps %xmm4, %xmm6, %xmm6 {%k2} vmulps %xmm6, %xmm6, %xmm4 vmovaps %xmm20, %xmm12 vfmadd213ps 0x158bc3(%rip){1to4}, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + mem vfmadd213ps %xmm23, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm23 vfmadd213ps %xmm18, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm18 vfmadd213ps %xmm19, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm19 vfmadd213ps %xmm24, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm24 vfmadd213ps %xmm29, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm29 vfmadd213ps %xmm31, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm31 vfmadd213ps 0x158bb1(%rip){1to4}, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + mem vmulps %xmm6, %xmm4, %xmm29 vmulps %xmm12, %xmm29, %xmm12 vpaddd 0x158b6f(%rip){1to4}, %xmm5, %xmm5 # 0x3f11e8 vmovaps %xmm7, %xmm24 vmovaps %xmm14, %xmm7 vmovaps %xmm15, %xmm14 vmovaps %xmm16, %xmm15 vmovaps %xmm17, %xmm16 vmovdqa64 %xmm13, %xmm17 vmovaps %xmm1, %xmm13 vmovaps %xmm30, %xmm1 vcvtdq2ps %xmm5, %xmm5 vsubps %xmm9, %xmm5, %xmm5 {%k2} vfmadd231ps %xmm11, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm11) + xmm12 vfmsub231ps %xmm4, %xmm8, %xmm12 # xmm12 = (xmm8 * xmm4) - xmm12 vsubps %xmm6, %xmm12, %xmm4 vfmsub231ps %xmm5, %xmm10, %xmm4 # xmm4 = (xmm10 * xmm5) - xmm4 vmulps 0x15a04d(%rip){1to4}, %xmm4, %xmm4 # 0x3f2718 vbroadcastss 0x15922c(%rip), %xmm5 # 0x3f1900 vmovaps %xmm5, %xmm4 {%k1} vminps %xmm27, %xmm4, %xmm4 vmaxps %xmm28, %xmm4, %xmm4 vmovaps %xmm7, %xmm5 vfmadd213ps %xmm8, %xmm4, %xmm5 # xmm5 = (xmm4 * xmm5) + xmm8 vcvttps2dq %xmm5, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm5, %k1 vsubps %xmm9, %xmm6, %xmm6 {%k1} vfmsub231ps %xmm10, %xmm6, %xmm4 # xmm4 = (xmm6 * xmm10) - xmm4 vfnmsub231ps %xmm11, %xmm6, %xmm4 # xmm4 = -(xmm6 * xmm11) - xmm4 vmulps %xmm4, %xmm4, %xmm5 vmovaps %xmm13, %xmm12 vfmadd213ps %xmm24, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm24 vfmadd213ps %xmm14, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm14 vfmadd213ps %xmm15, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm15 vfmadd213ps %xmm16, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm16 vfmadd213ps %xmm8, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm8 vfmadd213ps %xmm4, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm4 vaddps %xmm9, %xmm12, %xmm5 vcvttps2dq %xmm6, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm17, %xmm4, %xmm4 vfmadd213ps %xmm9, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm9 vrcpps %xmm4, %xmm5 vaddps %xmm5, %xmm5, %xmm6 vfmsub213ps 0x159fa6(%rip){1to4}, %xmm6, %xmm4 # xmm4 = (xmm6 * xmm4) - mem vfnmadd213ps %xmm6, %xmm5, %xmm4 # xmm4 = -(xmm5 * xmm4) + xmm6 vfmsub213ps %xmm22, %xmm22, %xmm3 # xmm3 = (xmm22 * xmm3) - xmm22 vfmsub213ps %xmm21, %xmm21, %xmm4 # xmm4 = (xmm21 * xmm4) - xmm21 jmp 0x298906 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm4 vbroadcastss 0x4(%r14), %xmm5 vmaxps %xmm4, %xmm22, %xmm3 vminps %xmm5, %xmm3, %xmm3 vmaxps %xmm4, %xmm21, %xmm4 vminps %xmm5, %xmm4, %xmm4 jmp 0x298906 vxorps %xmm1, %xmm22, %xmm3 vminps %xmm27, %xmm3, %xmm3 vmaxps %xmm28, %xmm3, %xmm3 vmovaps %xmm7, %xmm4 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm5 vcvtdq2ps %xmm5, %xmm5 vcmpltps %xmm5, %xmm4, %k1 vsubps %xmm9, %xmm5, %xmm5 {%k1} vfmsub231ps %xmm10, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm10) - xmm3 vfnmsub231ps %xmm11, %xmm5, %xmm3 # xmm3 = -(xmm5 * xmm11) - xmm3 vmulps %xmm3, %xmm3, %xmm4 vmovaps %xmm13, %xmm6 vfmadd213ps %xmm24, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm24 vfmadd213ps %xmm14, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm14 vfmadd213ps %xmm15, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm15 vfmadd213ps %xmm16, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm16 vfmadd213ps %xmm8, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm8 vfmadd213ps %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vaddps %xmm6, %xmm9, %xmm4 vcvttps2dq %xmm5, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm17, %xmm3, %xmm3 vfmadd213ps %xmm9, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) + xmm9 vrcpps %xmm3, %xmm4 vfmsub213ps %xmm9, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) - xmm9 vfnmadd132ps %xmm4, %xmm4, %xmm3 # xmm3 = -(xmm3 * xmm4) + xmm4 vxorps %xmm1, %xmm21, %xmm4 vminps %xmm27, %xmm4, %xmm4 vmaxps %xmm28, %xmm4, %xmm4 vmovaps %xmm7, %xmm5 vfmadd213ps %xmm8, %xmm4, %xmm5 # xmm5 = (xmm4 * xmm5) + xmm8 vcvttps2dq %xmm5, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm5, %k1 vsubps %xmm9, %xmm6, %xmm6 {%k1} vfmsub231ps %xmm10, %xmm6, %xmm4 # xmm4 = (xmm6 * xmm10) - xmm4 vfnmsub231ps %xmm11, %xmm6, %xmm4 # xmm4 = -(xmm6 * xmm11) - xmm4 vmulps %xmm4, %xmm4, %xmm5 vmovaps %xmm13, %xmm12 vfmadd213ps %xmm24, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm24 vfmadd213ps %xmm14, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm14 vfmadd213ps %xmm15, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm15 vfmadd213ps %xmm16, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm16 vfmadd213ps %xmm8, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm8 vfmadd213ps %xmm4, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm4 vaddps %xmm9, %xmm12, %xmm5 vcvttps2dq %xmm6, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm17, %xmm4, %xmm4 vfmadd213ps %xmm9, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm9 vrcpps %xmm4, %xmm5 vfmsub213ps %xmm9, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm9 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 jmp 0x298906 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm4 vbroadcastss 0x4(%r14), %xmm5 vmovaps %xmm22, %xmm3 vfmadd213ps %xmm5, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) + xmm5 vmaxps %xmm0, %xmm3, %xmm3 vminps %xmm3, %xmm9, %xmm3 vmulps %xmm22, %xmm3, %xmm3 vfmadd213ps %xmm5, %xmm21, %xmm4 # xmm4 = (xmm21 * xmm4) + xmm5 vmaxps %xmm0, %xmm4, %xmm4 vminps %xmm4, %xmm9, %xmm4 vmulps %xmm21, %xmm4, %xmm4 jmp 0x298906 vmovaps %xmm22, %xmm3 vmovaps %xmm21, %xmm4 cmpl $0x65, 0x38(%rsp) jl 0x298964 movq 0x288(%rbp,%rsi), %rsi vmulps (%rsi,%rbx,4), %xmm3, %xmm3 vmulps 0x10(%rsi,%rbx,4), %xmm4, %xmm4 vmovdqa %xmm2, %xmm5 vpternlogd $0xf8, %xmm1, %xmm3, %xmm5 vmovdqa %xmm2, %xmm6 vpternlogd $0xf8, %xmm1, %xmm4, %xmm6 vaddps %xmm5, %xmm3, %xmm3 vaddps %xmm6, %xmm4, %xmm4 vcvttps2dq %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpackssdw %xmm4, %xmm3, %xmm3 vpminsw %xmm25, %xmm3, %xmm3 vpmaxsw %xmm26, %xmm3, %xmm3 vpacksswb %xmm3, %xmm3, %xmm3 vmovq %xmm3, (%rdx) addq $0x8, %rdx jmp 0x298973 vmovups %xmm3, (%r8) vmovups %xmm4, 0x10(%r8) addq $0x20, %r8 incl %r13d jmp 0x2981b8 incl %ecx jmp 0x2981a7 movq 0x48(%rsp), %rsi incq %rsi movq 0x30(%rsp), %rdx addl 0x40(%rsp), %edx movq 0x28(%rsp), %r14 jmp 0x298142 pushq $0x3 popq %r13 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x298ab0 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x298ab0 cmpl $0x2, 0xe4(%rbp,%rdx) jne 0x298ab0 cmpl $0x2, 0xe8(%rbp,%rdx) jne 0x298ab0 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x298ab0 movl $0x0, 0x48(%rsp) cmpl $0x65, 0x38(%rsp) jl 0x299539 vpxor %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %r13 vmovdqa %xmm0, (%r13) andq $0x0, 0x10(%r13) xorl %ebx, %ebx movq %rbp, %rcx leaq 0xe0(%rsp), %rbp leaq 0x190(%rsp), %r12 movq -0x18(%rax), %r14 movslq 0x108(%rcx,%r14), %rax cmpq %rax, %rbx jge 0x299868 movq 0x1f8(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm1 vpxor %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x298a6b movq 0x240(%rcx,%r14), %rax vmulss (%rax,%rbx,4), %xmm1, %xmm0 vmovss 0x156221(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rcx,%r14), %rax vmovd (%rax,%rbx,4), %xmm0 vmovd %xmm0, 0x190(%rsp) movq %r13, %rdi movq %rbp, %rsi callq 0x1ea12c movq %r13, %rdi movq %r12, %rsi callq 0x1ea12c incq %rbx movq 0x10(%rsp), %rcx movq (%rcx), %rax jmp 0x298a26 imull %ecx, %r13d movslq %r13d, %rsi leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %rsi, 0xc0(%rsp) callq 0x73bbe movq (%rbp), %rax movq %rax, 0x58(%rsp) movq -0x18(%rax), %rax imull 0xe0(%rbp,%rax), %r12d movq 0x130(%rsp), %rbx movl 0xdc(%rbp,%rax), %ecx imull 0xd4(%rbp,%rax), %ecx subl %ecx, %r12d xorl %ecx, %ecx xorl %edx, %edx xorl %esi, %esi cmpl 0xd8(%rbp,%rax), %esi jge 0x298b48 movslq %ecx, %rcx leaq (%rbx,%rcx,4), %r8 xorl %edi, %edi cmpl 0xd4(%rbp,%rax), %edi jge 0x298b3e movl %edx, (%r8,%rdi,4) movq 0x58(%rsp), %rax movq -0x18(%rax), %rax addl 0xdc(%rbp,%rax), %edx incq %rdi jmp 0x298b1c addl %r12d, %edx incl %esi addq %rdi, %rcx jmp 0x298b0a xorl %edx, %edx testl %r13d, %r13d movl $0x0, %ecx movq %rcx, 0x50(%rsp) cmovlel %edx, %r13d vbroadcastss 0x158650(%rip), %xmm4 # 0x3f11b4 leaq 0x15fa39(%rip), %r8 # 0x3f85a4 xorl %ebp, %ebp movq 0x10(%rsp), %rcx movslq 0x108(%rcx,%rax), %rax cmpq %rax, %rbp jge 0x298ed9 movq 0x10(%r15), %r9 imulq %rbp, %r9 imulq 0x40(%r15), %r9 addq (%r15), %r9 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rbp, %r11 movq 0x80(%rsp), %rax imulq %rax, %r11 addq 0x70(%rsp), %r11 imulq %rax, %r10 movq 0x28(%rcx), %r15 addq 0x50(%rsp), %r15 xorl %eax, %eax movq %r9, 0x28(%rsp) movq %r10, 0x40(%rsp) movq %r11, 0x68(%rsp) movq %rbp, 0xc8(%rsp) cmpl 0x60(%rsp), %eax jg 0x298eb1 movl %eax, 0x30(%rsp) movq 0x10(%rsp), %rax movq (%rax), %rax movq %rax, 0x48(%rsp) xorl %r14d, %r14d cmpl 0x18(%rsp), %r14d jg 0x298ea6 movq 0x48(%rsp), %rax movq -0x18(%rax), %r12 movq 0x10(%rsp), %rcx movslq 0xe8(%rcx,%r12), %rax movslq 0x30(%rsp), %rdx imulq %rax, %rdx imulq %r10, %rdx movslq 0xe4(%rcx,%r12), %rax movslq %r14d, %rcx imulq %rax, %rcx addq %r11, %rcx addq %rdx, %rcx xorl %edx, %edx xorl %eax, %eax cmpq %rdx, %r13 je 0x298c5f movslq (%rbx,%rdx,4), %rsi movsbl (%rcx,%rsi), %esi movsbl (%r15,%rdx), %edi imull %esi, %edi addl %edi, %eax incq %rdx jmp 0x298c43 movq 0x10(%rsp), %rdx movq 0x1f8(%rdx,%r12), %rcx vmovss (%rcx,%rbp,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x298c94 movq 0x240(%rdx,%r12), %rcx vmulss (%rcx,%rbp,4), %xmm1, %xmm0 vmovss 0x155ff8(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vcvtsi2ss %eax, %xmm6, %xmm1 vmulss %xmm1, %xmm0, %xmm5 cmpl $0x0, 0x100(%rdx,%r12) je 0x298cb4 movq 0x1b0(%rdx,%r12), %rax vaddss (%rax,%rbp,4), %xmm5, %xmm5 movl 0x110(%rdx,%r12), %eax decl %eax cmpl $0x5, %eax ja 0x298e2c movslq (%r8,%rax,4), %rax addq %r8, %rax jmpq %rax vmaxss 0x155338(%rip), %xmm5, %xmm0 # 0x3ee010 jmp 0x298e30 vmovaps %xmm5, %xmm0 movq %r9, 0xd8(%rsp) vmovss %xmm5, 0xd0(%rsp) callq 0x5f410 vaddss 0x155f89(%rip), %xmm0, %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x68(%rsp), %r11 movq 0x40(%rsp), %r10 movq 0xd8(%rsp), %r9 leaq 0x15f882(%rip), %r8 # 0x3f85a4 vbroadcastss 0x158489(%rip), %xmm4 # 0x3f11b4 vmulss 0xd0(%rsp), %xmm0, %xmm0 jmp 0x298e30 movq 0x10(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss 0x4(%rax), %xmm1 vmaxss (%rax), %xmm5, %xmm0 vucomiss %xmm1, %xmm0 jbe 0x298e30 vmovaps %xmm1, %xmm0 jmp 0x298e30 vmovss 0x15844e(%rip), %xmm2 # 0x3f11b8 vminss %xmm2, %xmm5, %xmm1 vxorps %xmm4, %xmm1, %xmm0 vcmpltss 0x15843f(%rip), %xmm1, %k1 # 0x3f11bc vmovss %xmm2, %xmm0, %xmm0 {%k1} movq %r9, %rbp callq 0x5f410 movq 0x68(%rsp), %r11 movq 0x40(%rsp), %r10 movq %rbp, %r9 movq 0xc8(%rsp), %rbp leaq 0x15f7fd(%rip), %r8 # 0x3f85a4 vbroadcastss 0x158404(%rip), %xmm4 # 0x3f11b4 vmovss 0x155ed0(%rip), %xmm1 # 0x3eec88 vaddss %xmm1, %xmm0, %xmm0 vdivss %xmm0, %xmm1, %xmm0 jmp 0x298e30 movq 0x10(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss (%rax), %xmm0 vcmpgtss 0x155232(%rip), %xmm5, %k1 # 0x3ee010 vmovss 0x155ea0(%rip), %xmm0 {%k1} # 0x3eec88 vmulss %xmm5, %xmm0, %xmm0 jmp 0x298e30 movq 0x10(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss (%rax), %xmm1 vmovss 0x4(%rax), %xmm2 vxorps %xmm4, %xmm2, %xmm0 vdivss %xmm1, %xmm0, %xmm3 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm3, %xmm5 jb 0x298e30 vmovss 0x155e6a(%rip), %xmm0 # 0x3eec88 vdivss %xmm1, %xmm0, %xmm0 vaddss %xmm0, %xmm3, %xmm0 vucomiss %xmm0, %xmm5 jbe 0x298e9b vmovaps %xmm5, %xmm0 cmpl $0x65, 0x38(%rsp) jl 0x298e81 movq 0x10(%rsp), %rax movq 0x288(%rax,%r12), %rax vmulss (%rax,%rbp,4), %xmm0, %xmm0 vpbroadcastd 0x15f7b2(%rip), %xmm1 # 0x3f8604 vpternlogd $0xf8, %xmm4, %xmm0, %xmm1 vaddss %xmm1, %xmm0, %xmm0 vroundss $0xb, %xmm0, %xmm0, %xmm0 vcvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movb %al, (%r9) incq %r9 jmp 0x298e93 movq 0x28(%rsp), %rax vmovss %xmm0, (%rax) addq $0x4, %rax movq %rax, 0x28(%rsp) incl %r14d jmp 0x298bfc vfmadd213ss %xmm2, %xmm5, %xmm1 # xmm1 = (xmm5 xmm1) + xmm2 vmulss %xmm5, %xmm1, %xmm0 jmp 0x298e30 movl 0x30(%rsp), %eax incl %eax jmp 0x298bde incq %rbp movq 0x58(%rsp), %rax movq -0x18(%rax), %rax movq 0x50(%rsp), %rcx addq 0xc0(%rsp), %rcx movq %rcx, 0x50(%rsp) movq 0x20(%rsp), %r15 jmp 0x298b6d leaq 0x130(%rsp), %rdi callq 0x624be movl $0x0, 0x48(%rsp) jmp 0x298fc6 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0xe0(%rsp), %xmm0 movq 0x20(%rsp), %rcx vmovups %xmm0, (%rcx) movq 0xf0(%rsp), %rax movq %rax, 0x10(%rcx) movl 0xf8(%rsp), %eax movl %eax, 0x18(%rcx) movq 0x100(%rsp), %rax movq %rax, 0x20(%rcx) vmovups 0x108(%rsp), %xmm0 vmovups %xmm0, 0x28(%rcx) movl 0x118(%rsp), %eax movl %eax, 0x38(%rcx) movq 0x120(%rsp), %rax movq %rax, 0x40(%rcx) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x298f8a lock decl (%rax) jne 0x298f8a movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x298f82 movq (%rdi), %rax callq 0x18(%rax) jmp 0x298f8a movq %rsi, %rdi callq 0x5f3e0 movl %r14d, 0x48(%rsp) movq 0x138(%rsp), %rax testq %rax, %rax je 0x298fc6 lock decl (%rax) jne 0x298fc6 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi je 0x298fbe movq (%rdi), %rax callq 0x18(%rax) jmp 0x298fc6 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x298ff7 lock decl (%rax) jne 0x298ff7 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x298fef movq (%rdi), %rax callq 0x18(%rax) jmp 0x298ff7 movq %rsi, %rdi callq 0x5f3e0 movq 0x1e8(%rsp), %rax testq %rax, %rax movl 0x48(%rsp), %ebx je 0x299032 lock decl (%rax) jne 0x299032 movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi je 0x29902a movq (%rdi), %rax callq 0x18(%rax) jmp 0x299032 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0x318, %rsp # imm = 0x318 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq vpxor %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovdqa %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2990bc movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x155bed(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x299066 leaq 0x2a0(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdx movq 0x20(%rsp), %rcx movslq 0x2c(%rcx), %rdi movslq 0x30(%rcx), %rax movl 0x38(%rcx), %esi movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movl 0x34(%rcx), %r8d movl %r8d, 0x68(%rsp) movq (%rcx), %r8 movq 0x10(%rcx), %r11 movq 0x40(%rcx), %r15 imulq %r11, %r15 movq %rax, %rbx imulq %rdi, %rbx movq %r11, 0xd0(%rsp) movq %rbx, 0x60(%rsp) imulq %rbx, %r11 addq $0xf, %r11 movq 0x80(%rsp), %rbx imulq 0xb0(%rsp), %rbx movq %rbx, 0x50(%rsp) andq $-0x10, %r11 movq %r11, 0x58(%rsp) xorl %r11d, %r11d testl %eax, %eax movl $0x0, %ebp cmovgl %eax, %ebp movl 0x28(%rcx), %eax movl %eax, 0xc0(%rsp) testl %esi, %esi cmovlel %r11d, %esi movq %rsi, 0x38(%rsp) vpxor %xmm0, %xmm0, %xmm0 movq 0x70(%rsp), %rax movq %rax, 0x188(%rsp) movq %rdx, 0x28(%rsp) leaq (%rdx,%rdx), %rax movq %rax, 0x180(%rsp) movq 0x2a0(%rsp), %rax movq %rax, 0x178(%rsp) movq %r8, 0xd8(%rsp) xorl %esi, %esi movq %r10, 0x40(%rsp) movq %r15, 0xc8(%rsp) cmpq 0x38(%rsp), %rsi je 0x29933a movq 0x58(%rsp), %rax xorl %edx, %edx divq 0xd0(%rsp) cmpl $0x4, 0xc0(%rsp) cmoveq 0x60(%rsp), %rax testq %r10, %r10 je 0x2991e8 vmovss (%r10,%rsi,4), %xmm2 jmp 0x2991ec vxorps %xmm2, %xmm2, %xmm2 imulq %rsi, %r15 addq 0xd8(%rsp), %r15 movq 0x178(%rsp), %rcx vmovss (%rcx,%rsi,4), %xmm1 imull 0x68(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x299225 vmovss %xmm2, (%r8,%rcx,4) incq %rcx jmp 0x299216 leaq (%rsi,%rsi,8), %rax movq 0x50(%rsp), %r13 movq %rsi, 0x18(%rsp) imulq %rsi, %r13 addq 0x188(%rsp), %r13 movq 0x28(%rsp), %rcx leaq (%rcx,%r13), %r12 movq 0x180(%rsp), %rcx leaq (%rcx,%r13), %r11 xorl %ebx, %ebx cmpl %ebp, %ebx je 0x29931d xorl %ecx, %ecx movq %r15, %rdx movl %edi, %r10d testl %r10d, %r10d jle 0x2992fe movzwl (%r13,%rcx), %r14d vmovd %r14d, %xmm2 vpinsrb $0x2, 0x2(%r13,%rcx), %xmm2, %xmm2 vpinsrb $0x3, (%r12,%rcx), %xmm2, %xmm2 vpinsrb $0x4, 0x1(%r12,%rcx), %xmm2, %xmm2 vpinsrb $0x5, 0x2(%r12,%rcx), %xmm2, %xmm2 vpinsrb $0x6, (%r11,%rcx), %xmm2, %xmm2 vpinsrb $0x7, 0x1(%r11,%rcx), %xmm2, %xmm2 vpmovsxbw (%r9,%rax), %xmm3 vpmovsxbw %xmm2, %xmm2 vpmaddwd %xmm2, %xmm3, %xmm2 vphaddd %xmm2, %xmm0, %xmm2 movsbl 0x2(%r11,%rcx), %r14d movsbl 0x8(%r9,%rax), %esi imull %r14d, %esi vphaddd %xmm2, %xmm2, %xmm2 vphaddd %xmm2, %xmm2, %xmm2 vmovd %xmm2, %r14d addl %esi, %r14d vcvtsi2ss %r14d, %xmm4, %xmm2 vfmadd213ss (%r15,%rcx,4), %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + mem vmovss %xmm2, (%r15,%rcx,4) addq $0x4, %rdx decl %r10d incq %rcx jmp 0x299266 addq %rcx, %r13 addq $0x2, %r13 addq %rcx, %r12 addq $0x2, %r12 addq %rcx, %r11 addq $0x2, %r11 incl %ebx movq %rdx, %r15 jmp 0x299256 movq 0x18(%rsp), %rsi incq %rsi movq 0xc8(%rsp), %r15 addq %r15, %r8 movq 0x40(%rsp), %r10 jmp 0x2991b3 leaq 0x2a0(%rsp), %rdi jmp 0x29984c leaq 0x2b8(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rsi movq 0x20(%rsp), %r15 movl 0x2c(%r15), %eax movl 0x30(%r15), %edx movl 0x38(%r15), %r9d movq 0x10(%rsp), %rcx movq 0x28(%rcx), %rdi movq 0x1b0(%rcx,%r14), %r8 movq %rsi, 0x38(%rsp) leaq (%rsi,%rsi), %rcx movq %rcx, 0x28(%rsp) xorl %r10d, %r10d testl %edx, %edx cmovlel %r10d, %edx testl %r9d, %r9d cmovlel %r10d, %r9d movq %r9, 0x18(%rsp) vpxor %xmm0, %xmm0, %xmm0 vpbroadcastd 0x15f24e(%rip), %xmm1 # 0x3f8604 vpbroadcastd 0x157df5(%rip), %xmm2 # 0x3f11b4 cmpq 0x18(%rsp), %r10 je 0x29952c testq %r8, %r8 je 0x2993d7 vmovss (%r8,%r10,4), %xmm3 jmp 0x2993db vxorps %xmm3, %xmm3, %xmm3 movq 0x40(%r15), %r11 imulq %r10, %r11 imulq 0x10(%r15), %r11 addq (%r15), %r11 movq 0x2b8(%rsp), %rcx vmovss (%rcx,%r10,8), %xmm4 vmovss 0x4(%rcx,%r10,8), %xmm5 movq 0xb0(%rsp), %rbx imulq %r10, %rbx imulq 0x80(%rsp), %rbx addq 0x70(%rsp), %rbx leaq (%r10,%r10,8), %r14 movq 0x38(%rsp), %rcx leaq (%rbx,%rcx), %r15 movq 0x28(%rsp), %rcx leaq (%rbx,%rcx), %r12 xorl %ebp, %ebp cmpl %edx, %ebp je 0x29951f xorl %r13d, %r13d movl %eax, %ecx testl %ecx, %ecx jle 0x299500 movzwl (%rbx,%r13), %r9d vmovd %r9d, %xmm6 vpinsrb $0x2, 0x2(%rbx,%r13), %xmm6, %xmm6 vpinsrb $0x3, (%r15,%r13), %xmm6, %xmm6 vpinsrb $0x4, 0x1(%r15,%r13), %xmm6, %xmm6 vpinsrb $0x5, 0x2(%r15,%r13), %xmm6, %xmm6 vpinsrb $0x6, (%r12,%r13), %xmm6, %xmm6 vpinsrb $0x7, 0x1(%r12,%r13), %xmm6, %xmm6 vpmovsxbw (%rdi,%r14), %xmm7 vpmovsxbw %xmm6, %xmm6 vpmaddwd %xmm6, %xmm7, %xmm6 vphaddd %xmm6, %xmm0, %xmm6 movsbl 0x2(%r12,%r13), %r9d movsbl 0x8(%rdi,%r14), %esi imull %r9d, %esi vphaddd %xmm6, %xmm6, %xmm6 vphaddd %xmm6, %xmm6, %xmm6 vmovd %xmm6, %r9d addl %esi, %r9d vcvtsi2ss %r9d, %xmm8, %xmm6 vfmadd213ss %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vmulss %xmm5, %xmm6, %xmm6 vmovdqa %xmm1, %xmm7 vpternlogd $0xf8, %xmm2, %xmm6, %xmm7 vaddss %xmm7, %xmm6, %xmm6 vroundss $0xb, %xmm6, %xmm6, %xmm6 vcvttss2si %xmm6, %esi cmpl $-0x7e, %esi pushq $-0x7f popq %r9 cmovll %r9d, %esi cmpl $0x7f, %esi pushq $0x7f popq %r9 cmovgel %r9d, %esi movb %sil, (%r11,%r13) decl %ecx incq %r13 jmp 0x29943f addq %r13, %rbx addq $0x2, %rbx addq %r13, %r15 addq $0x2, %r15 addq %r13, %r12 addq $0x2, %r12 incl %ebp addq %r13, %r11 jmp 0x299432 incq %r10 movq 0x20(%rsp), %r15 jmp 0x2993bf leaq 0x2b8(%rsp), %rdi jmp 0x299a63 vpxor %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovdqa %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2995af movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x1556fa(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x299559 leaq 0x270(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rsi movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rdi movslq 0x30(%rdx), %rcx movl 0x38(%rdx), %r8d movl %esi, %eax subl %edi, %eax addl %eax, %eax movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movl 0x34(%rdx), %r11d movl %r11d, 0xd8(%rsp) movq 0x10(%rdx), %r11 movq 0x40(%rdx), %r14 imulq %r11, %r14 movq %rcx, %rbx movq %rdi, 0x18(%rsp) imulq %rdi, %rbx movq %r11, 0xd0(%rsp) movq %rbx, 0x60(%rsp) imulq %rbx, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0x58(%rsp) movq 0x80(%rsp), %rdi imulq 0xb0(%rsp), %rdi movq %rdi, 0x50(%rsp) xorl %edi, %edi testl %ecx, %ecx movl $0x0, %ebp cmovgl %ecx, %ebp movq (%rdx), %r13 testl %r8d, %r8d cmovlel %edi, %r8d movq %r8, 0x28(%rsp) movslq %eax, %r11 vpxor %xmm0, %xmm0, %xmm0 movl 0x28(%rdx), %eax movl %eax, 0x188(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x180(%rsp) movq %rsi, 0x40(%rsp) leaq (%rsi,%rsi), %rax movq %rax, 0x178(%rsp) movq 0x270(%rsp), %rax movq %rax, 0x1d8(%rsp) movq %r13, 0xc0(%rsp) xorl %edi, %edi movq %r10, 0x68(%rsp) movq %r14, 0xc8(%rsp) cmpq 0x28(%rsp), %rdi je 0x299844 movq 0x58(%rsp), %rax xorl %edx, %edx divq 0xd0(%rsp) cmpl $0x4, 0x188(%rsp) cmoveq 0x60(%rsp), %rax testq %r10, %r10 je 0x2996ed vmovss (%r10,%rdi,4), %xmm2 jmp 0x2996f1 vxorps %xmm2, %xmm2, %xmm2 imulq %rdi, %r14 addq 0xc0(%rsp), %r14 movq 0x1d8(%rsp), %rcx vmovss (%rcx,%rdi,4), %xmm1 imull 0xd8(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x29972e vmovss %xmm2, (%r13,%rcx,4) incq %rcx jmp 0x29971e leaq (%rdi,%rdi,8), %rax movq 0x50(%rsp), %rsi movq %rdi, 0x38(%rsp) imulq %rdi, %rsi addq 0x180(%rsp), %rsi movq 0x40(%rsp), %rcx leaq (%rsi,%rcx), %rbx movq 0x178(%rsp), %rcx leaq (%rsi,%rcx), %r8 xorl %r15d, %r15d cmpl %ebp, %r15d je 0x299827 xorl %r10d, %r10d movq %r14, %rdx movq 0x18(%rsp), %rcx testl %ecx, %ecx jle 0x29980a movzwl (%rsi,%r10), %r12d vmovd %r12d, %xmm2 vpinsrb $0x2, 0x2(%rsi,%r10), %xmm2, %xmm2 vpinsrb $0x3, (%rbx,%r10), %xmm2, %xmm2 vpinsrb $0x4, 0x1(%rbx,%r10), %xmm2, %xmm2 vpinsrb $0x5, 0x2(%rbx,%r10), %xmm2, %xmm2 vpinsrb $0x6, (%r8,%r10), %xmm2, %xmm2 vpinsrb $0x7, 0x1(%r8,%r10), %xmm2, %xmm2 vpmovsxbw (%r9,%rax), %xmm3 vpmovsxbw %xmm2, %xmm2 vpmaddwd %xmm2, %xmm3, %xmm2 vphaddd %xmm2, %xmm0, %xmm2 movsbl 0x2(%r8,%r10), %r12d movsbl 0x8(%r9,%rax), %edi imull %r12d, %edi vphaddd %xmm2, %xmm2, %xmm2 vphaddd %xmm2, %xmm2, %xmm2 vmovd %xmm2, %r12d addl %edi, %r12d vcvtsi2ss %r12d, %xmm4, %xmm2 vfmadd213ss (%r14,%r10,2), %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + mem vmovss %xmm2, (%r14,%r10,2) addq $0x4, %rdx decl %ecx addq $0x2, %r10 jmp 0x299774 addq %r11, %rsi addq %r10, %rsi addq %r11, %rbx addq %r10, %rbx addq %r11, %r8 addq %r10, %r8 incl %r15d movq %rdx, %r14 jmp 0x299760 movq 0x38(%rsp), %rdi incq %rdi movq 0xc8(%rsp), %r14 addq %r14, %r13 movq 0x68(%rsp), %r10 jmp 0x2996b8 leaq 0x270(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x20(%rsp), %r15 jmp 0x299a75 leaq 0x288(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %r8 movq 0x20(%rsp), %r15 movl 0x2c(%r15), %eax movl 0x30(%r15), %edx movl 0x38(%r15), %r9d movl %r8d, %ecx subl %eax, %ecx addl %ecx, %ecx movq 0x10(%rsp), %rsi movq 0x28(%rsi), %rdi movq 0x1b0(%rsi,%r14), %rsi movq %r8, 0x38(%rsp) addq %r8, %r8 movq %r8, 0x40(%rsp) movslq %ecx, %r10 xorl %r11d, %r11d testl %edx, %edx cmovlel %r11d, %edx testl %r9d, %r9d cmovlel %r11d, %r9d movq %r9, 0x18(%rsp) vpxor %xmm0, %xmm0, %xmm0 vpbroadcastd 0x15ed24(%rip), %xmm1 # 0x3f8604 vpbroadcastd 0x1578cb(%rip), %xmm2 # 0x3f11b4 movq %rsi, 0x28(%rsp) cmpq 0x18(%rsp), %r11 je 0x299a5b testq %rsi, %rsi je 0x299906 vmovss (%rsi,%r11,4), %xmm3 jmp 0x29990a vxorps %xmm3, %xmm3, %xmm3 movq 0x40(%r15), %rbx imulq %r11, %rbx imulq 0x10(%r15), %rbx addq (%r15), %rbx movq 0x288(%rsp), %rcx vmovss (%rcx,%r11,8), %xmm4 vmovss 0x4(%rcx,%r11,8), %xmm5 movq 0xb0(%rsp), %r14 imulq %r11, %r14 imulq 0x80(%rsp), %r14 addq 0x70(%rsp), %r14 leaq (%r11,%r11,8), %r15 movq 0x38(%rsp), %rcx leaq (%r14,%rcx), %r12 movq 0x40(%rsp), %rcx leaq (%r14,%rcx), %r13 xorl %ebp, %ebp cmpl %edx, %ebp je 0x299a49 xorl %r9d, %r9d movl %eax, %ecx testl %ecx, %ecx jle 0x299a30 movzwl (%r14,%r9), %esi vmovd %esi, %xmm6 vpinsrb $0x2, 0x2(%r14,%r9), %xmm6, %xmm6 vpinsrb $0x3, (%r12,%r9), %xmm6, %xmm6 vpinsrb $0x4, 0x1(%r12,%r9), %xmm6, %xmm6 vpinsrb $0x5, 0x2(%r12,%r9), %xmm6, %xmm6 vpinsrb $0x6, (%r13,%r9), %xmm6, %xmm6 vpinsrb $0x7, 0x1(%r13,%r9), %xmm6, %xmm6 vpmovsxbw (%rdi,%r15), %xmm7 vpmovsxbw %xmm6, %xmm6 vpmaddwd %xmm6, %xmm7, %xmm6 vphaddd %xmm6, %xmm0, %xmm6 movsbl 0x2(%r13,%r9), %esi movsbl 0x8(%rdi,%r15), %r8d imull %esi, %r8d vphaddd %xmm6, %xmm6, %xmm6 vphaddd %xmm6, %xmm6, %xmm6 vmovd %xmm6, %esi addl %r8d, %esi vcvtsi2ss %esi, %xmm8, %xmm6 vfmadd213ss %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vmulss %xmm5, %xmm6, %xmm6 vmovdqa %xmm1, %xmm7 vpternlogd $0xf8, %xmm2, %xmm6, %xmm7 vaddss %xmm7, %xmm6, %xmm6 vroundss $0xb, %xmm6, %xmm6, %xmm6 vcvttss2si %xmm6, %esi cmpl $-0x7e, %esi pushq $-0x7f popq %r8 cmovll %r8d, %esi cmpl $0x7f, %esi pushq $0x7f popq %r8 cmovgel %r8d, %esi movb %sil, (%rbx) incq %rbx decl %ecx addq $0x2, %r9 jmp 0x29996e addq %r10, %r14 addq %r9, %r14 addq %r10, %r12 addq %r9, %r12 addq %r10, %r13 addq %r9, %r13 incl %ebp jmp 0x299961 incq %r11 movq 0x20(%rsp), %r15 movq 0x28(%rsp), %rsi jmp 0x2998ee leaq 0x288(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x10(%rsp), %rax movq 0x8(%rax), %rdi testq %rdi, %rdi je 0x298fc6 movq (%rdi), %rax movq %r15, %rsi movq 0x30(%rsp), %rdx callq 0x48(%rax) jmp 0x298fc6 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 movq %rax, %rbx leaq 0x130(%rsp), %rdi callq 0x621c2 jmp 0x299c2a jmp 0x299b43 jmp 0x299b43 jmp 0x299ca4 jmp 0x299ca4 movq %rax, %rbx jmp 0x299bf3 jmp 0x299ca4 jmp 0x299b43 movq %rax, %rbx jmp 0x299bbc jmp 0x299ca4 jmp 0x299ca4 jmp 0x299b43 jmp 0x299ca4 jmp 0x299ca4 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x299c5b lock decl (%rax) jne 0x299c5b movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x299c4b movq (%rdi), %rax callq 0x18(%rax) jmp 0x299c5b jmp 0x299ca4 movq %rax, %rbx jmp 0x299c5b movq %rax, %rbx jmp 0x299c2a movq %rax, %rbx movq 0x230(%rsp), %rax testq %rax, %rax je 0x299b85 lock decl (%rax) jne 0x299b85 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi jne 0x299b7f movq %rsi, %rdi callq 0x5f3e0 jmp 0x299b85 movq (%rdi), %rax callq 0x18(%rax) movq 0x198(%rsp), %rax testq %rax, %rax je 0x299bbc lock decl (%rax) jne 0x299bbc movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi jne 0x299bb6 movq %rsi, %rdi callq 0x5f3e0 jmp 0x299bbc movq (%rdi), %rax callq 0x18(%rax) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x299bf3 lock decl (%rax) jne 0x299bf3 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi jne 0x299bed movq %rsi, %rdi callq 0x5f3e0 jmp 0x299bf3 movq (%rdi), %rax callq 0x18(%rax) movq 0x138(%rsp), %rax testq %rax, %rax je 0x299c2a lock decl (%rax) jne 0x299c2a movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi jne 0x299c24 movq %rsi, %rdi callq 0x5f3e0 jmp 0x299c2a movq (%rdi), %rax callq 0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x299c5b lock decl (%rax) jne 0x299c5b movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x299c55 movq %rsi, %rdi callq 0x5f3e0 jmp 0x299c5b movq (%rdi), %rax callq 0x18(%rax) movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x299c92 lock decl (%rax) jne 0x299c92 movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi jne 0x299c8c movq %rsi, %rdi callq 0x5f3e0 jmp 0x299c92 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x299ca4 jmp 0x299ca4 jmp 0x299ca4 jmp 0x299ca4 jmp 0x299ca4 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_avx512::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx512::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float)bias_data) + g 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float)bias_data) + g 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float)bias_data) + g 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2920d6	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp
ncnn::ConvolutionDepthWise_x86_avx512::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx512::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w = weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w = bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x168, %rsp # imm = 0x168 movq %rsi, %rbp movq %rdi, %r13 movq (%rsi), %r14 leaq 0x48(%r14), %rdi movq (%rdx), %rax movq %rax, 0xb8(%rsp) movl 0x60(%r14), %ebx movl 0x74(%r14), %eax movl %eax, 0x1c(%rsp) movl 0x78(%r14), %eax movl %eax, 0x18(%rsp) imull 0x80(%r14), %ebx leaq 0x70(%rsp), %rsi andq $0x0, 0x40(%rsi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rsi) vmovups %xmm0, 0xc(%rsi) vmovaps %xmm0, 0x20(%rsi) vmovups %xmm0, 0x2c(%rsi) movq %rcx, %r15 movq %rcx, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x29a2d9 movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x29a2d9 movslq 0x88(%rsp), %rcx movl 0x9c(%rsp), %eax imull %ecx, %eax movl %eax, 0x9c(%rsp) movq 0x80(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x80(%rsp) movl $0x1, 0x88(%rsp) andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) vmovaps %xmm0, 0x40(%rsp) vmovups %xmm0, 0x4c(%rsp) movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r13,%rax) je 0x299e22 movl $0x90, %edi addq (%rbp), %rdi leaq 0x20(%rsp), %rsi movq %r15, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x20(%rsp) je 0x29a2ab movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x29a2ab movslq 0x38(%rsp), %rcx movl 0x4c(%rsp), %eax imull %ecx, %eax movl %eax, 0x4c(%rsp) movq 0x30(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x30(%rsp) movl $0x1, 0x38(%rsp) pushq $0x2a popq %rdi callq 0x782bf movq %rax, %r12 leaq 0x8(%rsp), %rdi callq 0x71548 leaq 0x8(%rsp), %rdi xorl %esi, %esi movl %ebx, %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi movl 0x1c(%rsp), %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0xb popq %rsi movl 0x18(%rsp), %edx callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xdc(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xd popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xec(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x4 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xf popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xe popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x10 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax vmovss 0xfc(%r13,%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x12 popq %rsi callq 0x71952 movq (%r13), %rax movq -0x18(%rax), %rax movl 0x100(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x5 popq %rsi callq 0x7193a movl 0x9c(%rsp), %edx leaq 0x8(%rsp), %rdi pushq $0x6 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x108(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x7 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x10c(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x110(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax leaq (%rax,%r13), %rdx addq $0x118, %rdx # imm = 0x118 leaq 0x8(%rsp), %rdi pushq $0xa popq %rsi callq 0x7196c movq (%r12), %rax leaq 0x8(%rsp), %rsi movq %r12, %rdi callq 0x10(%rax) andq $0x0, 0x110(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0xd0(%rsp) vmovups %xmm0, 0xdc(%rsp) vmovaps %xmm0, 0xf0(%rsp) vmovups %xmm0, 0xfc(%rsp) andq $0x0, 0x158(%rsp) vmovups %xmm0, 0x118(%rsp) vmovups %xmm0, 0x124(%rsp) vmovups %xmm0, 0x138(%rsp) vmovups %xmm0, 0x144(%rsp) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29a090 lock incl (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x29a0c7 lock decl (%rax) jne 0x29a0c7 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x29a0bf movq (%rdi), %rax callq 0x18(%rax) jmp 0x29a0c7 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0xd0(%rsp) movq 0x80(%rsp), %rax movq %rax, 0xe0(%rsp) movl 0x88(%rsp), %eax movl %eax, 0xe8(%rsp) movq 0x90(%rsp), %rax movq %rax, 0xf0(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0xf8(%rsp) movl 0xa8(%rsp), %eax movl %eax, 0x108(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x110(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax je 0x29a141 lock incl (%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x29a178 lock decl (%rax) jne 0x29a178 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x29a170 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29a178 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 leaq 0xd0(%rsp), %rsi vmovups %xmm0, 0x48(%rsi) movq 0x30(%rsp), %rax movq %rax, 0x58(%rsi) movl 0x38(%rsp), %eax movl %eax, 0x60(%rsi) movq 0x40(%rsp), %rax movq %rax, 0x68(%rsi) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsi) movl 0x58(%rsp), %eax movl %eax, 0x80(%rsi) movq 0x60(%rsp), %rax movq %rax, 0x88(%rsi) leaq 0xc0(%rsp), %rdi callq 0x6b00e movq (%r12), %rax leaq 0xc0(%rsp), %rsi movq %r12, %rdi callq 0x18(%rax) leaq 0xc0(%rsp), %rdi callq 0x6b03a movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq 0x20(%rax) movq (%r12), %rax movq %r12, %rdi movq %r14, %rsi movq 0xb8(%rsp), %rdx movq %r15, %rcx callq 0x38(%rax) movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq 0x28(%rax) movq (%r12), %rax movq %r12, %rdi callq 0x8(%rax) pushq $0x48 popq %rbx vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%rbx), %rax testq %rax, %rax je 0x29a273 lock decl (%rax) jne 0x29a273 movq 0xd0(%rsp,%rbx), %rsi movq 0xf0(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x29a267 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x29a273 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%rbx), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x29a234 leaq 0x8(%rsp), %rdi callq 0x71614 xorl %r12d, %r12d movq 0x28(%rsp), %rax testq %rax, %rax je 0x29a2d9 lock decl (%rax) jne 0x29a2d9 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x29a2d1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29a2d9 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x29a30a lock decl (%rax) jne 0x29a30a movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29a302 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29a30a movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x168, %rsp # imm = 0x168 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x29a43f jmp 0x29a43f movq %rax, %rbx leaq 0xc0(%rsp), %rdi callq 0x6b03a jmp 0x29a34c jmp 0x29a349 jmp 0x29a341 jmp 0x29a341 movq %rax, %rbx jmp 0x29a3d6 movq %rax, %rbx pushq $0x48 popq %r14 vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%r14), %rax testq %rax, %rax je 0x29a393 lock decl (%rax) jne 0x29a393 movq 0xd0(%rsp,%r14), %rsi movq 0xf0(%rsp,%r14), %rdi testq %rdi, %rdi je 0x29a387 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x29a393 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x29a354 jmp 0x29a3cc jmp 0x29a43f movq %rax, %rbx jmp 0x29a404 jmp 0x29a43f movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq 0x28(%rsp), %rax testq %rax, %rax je 0x29a404 lock decl (%rax) jne 0x29a404 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x29a3fe movq %rsi, %rdi callq 0x5f3e0 jmp 0x29a404 movq (%rdi), %rax callq 0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29a435 lock decl (%rax) jne 0x29a435 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x29a42f movq %rsi, %rdi callq 0x5f3e0 jmp 0x29a435 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x29a43f movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_avx512::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx512::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w = weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w = bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }	movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x299cb8	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp
ncnn::ConvolutionDepthWise_x86_fma::create_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_fma::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x50, %rsp movq (%rdi), %rax movq -0x18(%rax), %r13 cmpl $0x0, 0x160(%rdi,%r13) je 0x29a51d xorl %eax, %eax addq $0x50, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 retq movq %rsi, %r14 movq %rdi, %rbx movl 0x110(%rdi,%r13), %ecx decl %ecx cmpl $0x5, %ecx ja 0x29a874 leaq 0x15e0e3(%rip), %rax # 0x3f8620 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx jmpq %rcx pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x29a6b0 pushq $0x47 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x29a6b0 pushq $0x36 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax vmovss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) jmp 0x29a6b0 pushq $0x1e popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x29a6b0 pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) jmp 0x29a6b0 pushq $0x43 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax vmovss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) leaq 0x8(%rsp), %rdi callq 0x71614 movq (%r15), %rax movq %r15, %rdi movq %r14, %rsi callq 0x20(%rax) movq (%rbx), %rax movq %r15, 0x8(%rbx) movq -0x18(%rax), %r15 cmpb $0x1, 0x1e(%r14) jne 0x29a6f3 cmpq $0x1, 0x178(%rbx,%r15) jne 0x29a6f3 movq %rbx, %rdi movq %r14, %rsi callq 0x29aa72 jmp 0x29a50d movl 0xd0(%rbx,%r15), %esi movl 0xd4(%rbx,%r15), %r10d movl 0xd8(%rbx,%r15), %r11d movl %r11d, %r8d imull %r10d, %r8d movl 0x104(%rbx,%r15), %eax movl 0x108(%rbx,%r15), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x29a812 imull %ecx, %eax cmpl %ecx, %eax jne 0x29a812 leaq (%rbx,%r15), %rsi cmpb $0x1, 0x27(%r14) jne 0x29a769 testb $0x7, %cl je 0x29a87c testb $0x3, %cl je 0x29a8dd xorl $0x3, %r10d xorl $0x3, %r11d orl %r10d, %r11d jne 0x29a812 cmpl $0x1, 0xdc(%rsi) jne 0x29a79e cmpl $0x1, 0xe0(%rsi) jne 0x29a79e cmpl $0x1, 0xe4(%rsi) jne 0x29a79e cmpl $0x1, 0xe8(%rsi) je 0x29a7c2 cmpl $0x1, 0xdc(%rsi) jne 0x29a812 cmpl $0x1, 0xe0(%rsi) jne 0x29a812 cmpl $0x2, 0xe4(%rsi) jne 0x29a812 cmpl $0x2, 0xe8(%rsi) jne 0x29a812 addq $0x168, %r15 # imm = 0x168 cmpq $0x28, %r15 je 0x29a81d movq 0x8(%rbx,%r15), %rax testq %rax, %rax je 0x29a7dc lock incl (%rax) leaq 0x28(%rbx), %r12 movq 0x30(%rbx), %rax testq %rax, %rax je 0x29a982 lock decl (%rax) jne 0x29a982 movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x29a97a movq (%rdi), %rax callq 0x18(%rax) jmp 0x29a982 movq %rbx, %rdi movq %r14, %rsi callq 0x29acde cmpb $0x1, (%r14) jne 0x29a50d movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x29a942 lock decl (%rax) jne 0x29a942 movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x29a93a movq (%rdi), %rax callq 0x18(%rax) jmp 0x29a942 xorl %r15d, %r15d jmp 0x29a6c9 addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x8 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x29a81d lock decl (%rax) jne 0x29a81d movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x29a96d movq (%rdi), %rax callq 0x18(%rax) jmp 0x29a81d addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x4 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x29a81d lock decl (%rax) jne 0x29a81d movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x29a96d movq (%rdi), %rax callq 0x18(%rax) jmp 0x29a81d movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rbx) vmovups %xmm0, (%rbx) vmovups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x29a50d movq %rsi, %rdi callq 0x5f3e0 jmp 0x29a81d movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%r12) vmovups %xmm0, (%r12) andl $0x0, 0x60(%rbx) vmovups %xmm0, 0x50(%rbx) vmovups (%rbx,%r15), %xmm0 vmovups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r15), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r15), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r15), %rax movq %rax, 0x48(%rbx) vmovups 0x28(%rbx,%r15), %xmm0 vmovups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r15), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r15), %rax movq %rax, 0x68(%rbx) jmp 0x29a81d jmp 0x29aa4a jmp 0x29aa4a movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x29aa69 lock decl (%rax) jne 0x29aa69 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x29aa38 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29aa69 jmp 0x29aa4a movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x29aa69 lock decl (%rax) jne 0x29aa69 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi jne 0x29aa42 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29aa69 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29aa69 movq %rax, %rdi callq 0x61d68 jmp 0x29aa5c jmp 0x29aa5c jmp 0x29aa5c jmp 0x29aa5c jmp 0x29aa5c movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq %rbx, %rdi callq 0x5f340 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
ncnn::ConvolutionDepthWise_x86_fma::create_group_ops(ncnn::Option const&)	int ConvolutionDepthWise_x86_fma::create_group_ops(const Option& opt) { // create Convolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Convolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[5]; weights[0] = weight_data_g; weights[1] = bias_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[2] = weight_data_int8_scales_g; weights[3] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[4] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[4]; weights[0] = weight_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[1] = weight_data_int8_scales_g; weights[2] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[3] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x278, %rsp # imm = 0x278 movq %rsi, 0x250(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x104(%rdi,%rdx), %eax movl 0x108(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x10(%rdi), %rax movq %rax, 0x1e0(%rsp) xorl %r12d, %r12d movq 0x10(%r14), %rax movq 0x18(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x29ad7b movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x29ad76 movq (%rdi), %rax callq 0x8(%rax) incq %r12 jmp 0x29ad4d imull %r15d, %ebx cmpq %rax, %rcx je 0x29ad88 movq %rax, 0x18(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x108(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0x1e0(%rsp), %rdi callq 0x6fbc2 leaq 0x118(%r14), %rax movq %rax, 0x258(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0x6c(%rsp) movslq %ebx, %rax movq %rax, 0x260(%rsp) movslq %r15d, %rax movq %rax, 0x248(%rsp) pushq $0x1 popq %rbx xorl %edx, %edx movl %r15d, 0x14(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rcx cmpq %rcx, %rdx jge 0x29baf2 movq %rdx, %rcx movq %rdx, 0x8(%rsp) movq 0x260(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdx, 0x80(%rsp) movl %esi, 0x88(%rsp) movq %rax, 0x90(%rsp) movl %ebx, 0x98(%rsp) movl %edi, 0x9c(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa0(%rsp) movl %ebx, 0xa8(%rsp) movq %rdi, 0xb0(%rsp) leaq 0x200(%rsp), %rdi leaq 0x70(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x29aed9 lock decl (%rax) jne 0x29aed9 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29aed1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29aed9 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x29af49 movq 0x8(%rsp), %r15 movq 0x248(%rsp), %rcx imulq %rcx, %r15 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0x1f8(%rsp) imulq %rsi, %r15 addq 0x1b0(%r14,%rax), %r15 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x1c(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0x1f0(%rsp) movl %ebx, %r13d movl 0x14(%rsp), %eax movl %eax, 0x18(%rsp) movq %rcx, 0x1e8(%rsp) jmp 0x29af83 xorl %r15d, %r15d movq $0x0, 0x1f8(%rsp) movl $0x0, 0x1c(%rsp) movq $0x0, 0x1f0(%rsp) xorl %r13d, %r13d movl $0x0, 0x18(%rsp) movq $0x0, 0x1e8(%rsp) pushq $0x6 popq %rdi callq 0x782bf movq %rax, %rbp leaq 0x268(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq %r12, %rdi xorl %esi, %esi movl 0x14(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r12, %rdi movl %ebx, %esi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r12, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r12, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r12, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r12, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r12, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r12, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r12, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x6 popq %rsi movl 0x6c(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r12, %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x110(%r14,%rax), %edx movq %r12, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x258(%rsp), %rdx movq %r12, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r12, %rsi callq 0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x29b170 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x1a8, %rax # imm = 0x1A8 jne 0x29b0fe movq 0x208(%rsp), %rax testq %rax, %rax je 0x29b138 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29b1f1 lock decl (%rax) jne 0x29b1f1 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29b1e9 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b1f1 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x160, %rax # imm = 0x160 jne 0x29b177 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29b1b1 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29b5a8 lock decl (%rax) jne 0x29b5a8 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29b5a0 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b5a8 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x29b295 lock decl (%rax) jne 0x29b295 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x29b28d movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b295 movq %rsi, %rdi callq 0x5f3e0 movq %r15, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq 0x1f8(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x1c(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x1f0(%rsp), %rax movq %rax, 0xd8(%rsp) movl %r13d, 0xe0(%rsp) movl 0x18(%rsp), %eax movl %eax, 0xe4(%rsp) movl %r13d, 0xe8(%rsp) movl %r13d, 0xec(%rsp) movl %r13d, 0xf0(%rsp) movq 0x1e8(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x29b52a andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x29b39a vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x29b38c movq 0x28(%rsp), %rax testq %rax, %rax je 0x29b3a7 lock incl (%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x29b3de lock decl (%rax) jne 0x29b3de movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x29b3d6 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b3de movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovaps %xmm0, 0x100(%rsp) movq 0x30(%rsp), %rax movq %rax, 0x110(%rsp) movl 0x38(%rsp), %eax movl %eax, 0x118(%rsp) movq 0x40(%rsp), %rax movq %rax, 0x120(%rsp) vmovups 0x48(%rsp), %xmm0 vmovups %xmm0, 0x128(%rsp) movl 0x58(%rsp), %eax movl %eax, 0x138(%rsp) movq 0x60(%rsp), %rax movq %rax, 0x140(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %r12d movq 0x260(%r14,%rax), %rbx movq 0x150(%rsp), %rax testq %rax, %rax je 0x29b4a0 lock decl (%rax) jne 0x29b4a0 movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x29b498 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b4a0 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %r12d, 0x160(%rsp) movq %rbx, 0x168(%rsp) vbroadcastss 0x15d12e(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x29b523 lock decl (%rax) jne 0x29b523 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x29b51b movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b523 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x29b74a movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x198(%rsp), %rax testq %rax, %rax je 0x29b6f6 lock decl (%rax) jne 0x29b6f6 movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x29b6ee movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b6f6 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x29b92d andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x29b6a3 vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x29b695 movq 0x28(%rsp), %rax testq %rax, %rax je 0x29b6b0 lock incl (%rax) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x29b7e2 lock decl (%rax) jne 0x29b7e2 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x29b7da movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b7e2 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x190(%rsp) andq $0x0, 0x198(%rsp) movq %r15, 0x1a0(%rsp) movl %ebx, 0x1a8(%rsp) movq %r12, 0x1b0(%rsp) vbroadcastss 0x15ced9(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x1b8(%rsp) movl $0x1, 0x1c8(%rsp) movq $0x1, 0x1d0(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0x120, %r15d # imm = 0x120 movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x29b7a8 lock decl (%rax) jne 0x29b7a8 movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x29b7a0 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b7a8 movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x29b777 jmp 0x29ba7b movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovups %xmm0, 0xb8(%rsp) movq 0x30(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x38(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x40(%rsp), %rax movq %rax, 0xd8(%rsp) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movl 0x58(%rsp), %eax movl %eax, 0xf0(%rsp) movq 0x60(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %ebx movq 0x260(%r14,%rax), %r12 movq 0x108(%rsp), %rax testq %rax, %rax je 0x29b8a4 lock decl (%rax) jne 0x29b8a4 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x29b89c movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b8a4 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x100(%rsp) andq $0x0, 0x108(%rsp) movq %r15, 0x110(%rsp) movl %ebx, 0x118(%rsp) movq %r12, 0x120(%rsp) vbroadcastss 0x15cd2b(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x128(%rsp) movl $0x1, 0x138(%rsp) movq $0x1, 0x140(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x29b926 lock decl (%rax) jne 0x29b926 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x29b91e movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b926 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x29b9f0 movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x150(%rsp), %rax testq %rax, %rax je 0x29b99c lock decl (%rax) jne 0x29b99c movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x29b994 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29b99c movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %ebx, 0x160(%rsp) movq %r12, 0x168(%rsp) vbroadcastss 0x15cc33(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0xd8, %r15d movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x29ba4e lock decl (%rax) jne 0x29ba4e movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x29ba46 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29ba4e movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x29ba1d movq (%rbp), %rax movq %rbp, %rdi movq 0x250(%rsp), %rsi callq 0x20(%rax) leaq 0x268(%rsp), %rdi movq 0x1e0(%rsp), %rax movq (%rax), %rax movq 0x8(%rsp), %rcx movq %rbp, (%rax,%rcx,8) callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29bae5 lock decl (%rax) jne 0x29bae5 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi je 0x29badd movq (%rdi), %rax callq 0x18(%rax) jmp 0x29bae5 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rdx incq %rdx jmp 0x29adf3 xorl %eax, %eax addq $0x278, %rsp # imm = 0x278 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x29bbba jmp 0x29bd51 jmp 0x29bb12 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x29bbcc lock decl (%rax) jne 0x29bbcc movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x29bb48 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29bbcc movq (%rdi), %rax callq 0x18(%rax) jmp 0x29bbcc jmp 0x29bd51 jmp 0x29bc44 jmp 0x29bbba jmp 0x29bb5e movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x29bc56 lock decl (%rax) jne 0x29bc56 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x29bb94 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29bc56 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29bc56 jmp 0x29bd51 jmp 0x29bd51 jmp 0x29bd51 jmp 0x29bd51 jmp 0x29bbba jmp 0x29bc44 movq %rax, %rbx jmp 0x29bbcc movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0xd8, %r14d vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x29bc0f lock decl (%rax) jne 0x29bc0f movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x29bc03 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x29bc0f movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x29bbd6 jmp 0x29bd05 jmp 0x29bd51 jmp 0x29bc44 movq %rax, %rbx jmp 0x29bc56 movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0x120, %r14d # imm = 0x120 vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x29bc99 lock decl (%rax) jne 0x29bc99 movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x29bc8d movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x29bc99 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x29bc60 jmp 0x29bd05 jmp 0x29bd51 jmp 0x29bcf9 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x29bd49 lock decl (%rax) jne 0x29bd49 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29bd39 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29bd49 jmp 0x29bd51 movq %rax, %rbx jmp 0x29bd12 jmp 0x29bd51 jmp 0x29bd51 movq %rax, %rbx leaq 0x268(%rsp), %rdi callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29bd49 lock decl (%rax) jne 0x29bd49 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi jne 0x29bd43 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29bd49 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_fma::create_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_fma::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %rax movq (%rdi), %rax addq -0x30(%rax), %rdi callq 0x29a4ee xorl %eax, %eax popq %rcx retq nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
ncnn::ConvolutionDepthWise_x86_fma::destroy_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_fma::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }	pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x29bd9d movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x29bd98 movq (%rdi), %rax callq 0x8(%rax) andq $0x0, 0x8(%rbx) xorl %r15d, %r15d movq 0x10(%rbx), %rax movq 0x18(%rbx), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r15 jge 0x29bddf movq (%rax,%r15,8), %rdi movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x10(%rbx), %rax movq (%rax,%r15,8), %rdi testq %rdi, %rdi je 0x29bdda movq (%rdi), %rax callq 0x8(%rax) incq %r15 jmp 0x29bda0 cmpq %rax, %rcx je 0x29bde8 movq %rax, 0x18(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %r15 retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_fma::destroy_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_fma::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }	pushq %rax movq (%rdi), %rax addq -0x38(%rax), %rdi callq 0x29bd6c xorl %eax, %eax popq %rcx retq nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
ncnn::ConvolutionDepthWise_x86_fma::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_fma::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float)bias_data) + g 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float)bias_data) + g 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float)bias_data) + g 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x2e8, %rsp # imm = 0x2E8 movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdi movq %rcx, 0x58(%rsp) cmpb $0x1, 0x1e(%rcx) jne 0x29be51 cmpl $0x0, 0x10c(%r14,%rdi) je 0x29be51 movq %r14, %rdi movq 0x58(%rsp), %rcx addq $0x2e8, %rsp # imm = 0x2E8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x29ff92 movq %rdx, 0x28(%rsp) movl 0x38(%rsi), %ecx movq %rcx, 0x60(%rsp) movq 0x10(%rsi), %r13 movslq 0x18(%rsi), %r15 movl 0xd4(%r14,%rdi), %ebx decl %ebx imull 0xdc(%r14,%rdi), %ebx movl 0xd8(%r14,%rdi), %ebp decl %ebp imull 0xe0(%r14,%rdi), %ebp leaq 0x80(%rsp), %rdx andq $0x0, 0x40(%rdx) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rdx) vmovups %xmm0, 0xc(%rdx) vmovaps %xmm0, 0x20(%rdx) vmovups %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %r14, %rdi movq 0x58(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0x80(%rsp) je 0x29fd9f movslq 0xb8(%rsp), %rax imulq 0xc0(%rsp), %rax testq %rax, %rax je 0x29fd9f notl %ebx notl %ebp movl 0xac(%rsp), %edi addl %edi, %ebx movq (%r14), %rax movq -0x18(%rax), %rcx movl %ebx, %eax cltd idivl 0xe4(%r14,%rcx) movl %eax, %ebx leal 0x1(%rbx), %esi addl 0xb0(%rsp), %ebp movl %ebp, %eax cltd idivl 0xe8(%r14,%rcx) movq %rax, 0x30(%rsp) leal 0x1(%rax), %r8d movl 0xd0(%r14,%rcx), %ecx movq 0x58(%rsp), %rax cmpb $0x1, 0x27(%rax) pushq $0x8 popq %rbp movl %edi, 0x10(%rsp) jne 0x29bf5d xorl %eax, %eax testb $0x3, %cl sete %al testb $0x7, %cl leal 0x1(%rax,%rax,2), %r9d cmovel %ebp, %r9d jmp 0x29bf61 pushq $0x1 popq %r9 movq %r13, %rax xorl %edx, %edx divq %r15 movq %rax, %r10 movl %r9d, %eax movq %rax, 0x40(%rsp) imulq %rax, %r10 movl %ecx, %eax cltd idivl %r9d movq 0x58(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq 0x28(%rsp), %r13 movq %r13, %rdi movl %esi, 0x70(%rsp) movl %r8d, 0xd0(%rsp) movl %r8d, %edx movl %eax, %ecx movq %r10, 0xe0(%rsp) movq %r10, %r8 callq 0x628f2 movq (%r13), %r9 testq %r9, %r9 je 0x29fd9f movq 0x40(%r13), %r11 movslq 0x38(%r13), %rax imulq %r11, %rax testq %rax, %rax je 0x29fd9f movq %r15, %r10 movl %r10d, %eax imull 0x60(%rsp), %eax movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xd0(%r14,%rdi), %ecx movl 0x108(%r14,%rdi), %esi cmpl %esi, %eax jne 0x29e16e cmpl %ecx, %eax jne 0x29e16e addq %r14, %rdi cmpl $0x1, %r10d je 0x29d338 movslq %ebx, %rbx cmpl $0x4, %r10d movq %r14, 0x8(%rsp) je 0x29ca8e movl %eax, %ecx cmpl $0x8, %r10d jne 0x29e16e movl 0xd4(%rdi), %eax movl 0xd8(%rdi), %r15d cmpl $0x5, %eax je 0x29dca2 cmpl $0x3, %eax jne 0x29fa97 cmpl $0x3, %r15d jne 0x29fa97 cmpl $0x1, 0xdc(%rdi) jne 0x29f37c cmpl $0x1, 0xe0(%rdi) jne 0x29f37c cmpl $0x1, 0xe4(%rdi) jne 0x29f37c cmpl $0x1, 0xe8(%rdi) jne 0x29f37c movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xac(%rsp), %edx movl 0xb8(%rsp), %esi movq 0x1b0(%rdi), %r8 leal 0x10(,%rdx,8), %edx movslq %edx, %rdi leal (,%rax,8), %edx movslq %edx, %rdx xorl %r9d, %r9d testl %esi, %esi cmovlel %r9d, %esi movq %rsi, 0x40(%rsp) shlq $0x2, %rdx movq %rdx, 0x10(%rsp) shlq $0x2, %rdi movq %r8, 0x60(%rsp) cmpq 0x40(%rsp), %r9 je 0x29fdf4 testq %r8, %r8 je 0x29c100 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r8,%rdx), %ymm0 jmp 0x29c104 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r10 imulq %r9, %r10 movq 0x10(%r13), %rsi imulq %rsi, %r10 addq (%r13), %r10 movslq 0x2c(%r13), %rdx movslq 0x54(%r14), %r11 imulq %r9, %r11 imulq 0x38(%r14), %r11 movq 0x28(%r14), %rbx imulq %rsi, %rdx addq %r10, %rdx movslq 0xac(%rsp), %rsi movq 0xc0(%rsp), %r15 movq %r9, 0x30(%rsp) imulq %r9, %r15 movq 0x90(%rsp), %r8 imulq %r8, %r15 addq 0x80(%rsp), %r15 imulq %r8, %rsi leaq (%r15,%rsi), %r12 leaq (%r15,%rsi,2), %r13 leaq (%rsi,%rsi,2), %rsi addq %r15, %rsi xorl %ebp, %ebp movl %ebp, %r8d orl $0x1, %r8d cmpl %ecx, %r8d jge 0x29ca6f xorl %r9d, %r9d xorl %r8d, %r8d leal 0x3(%r8), %r14d cmpl %eax, %r14d jge 0x29c5c0 vmovaps 0x20(%r15,%r9), %ymm3 vmovaps 0x40(%r15,%r9), %ymm4 vmovaps 0x60(%r15,%r9), %ymm5 vmovaps 0x80(%r15,%r9), %ymm6 vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm13 vmovaps 0x40(%rbx,%r11), %ymm12 vmovaps 0x60(%rbx,%r11), %ymm9 vmovaps (%r15,%r9), %ymm2 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vfmadd231ps %ymm3, %ymm13, %ymm2 # ymm2 = (ymm13 * ymm3) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm3) + ymm0 vfmadd231ps %ymm4, %ymm13, %ymm3 # ymm3 = (ymm13 * ymm4) + ymm3 vfmadd231ps %ymm4, %ymm12, %ymm2 # ymm2 = (ymm12 * ymm4) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm4 # ymm4 = (ymm1 * ymm4) + ymm0 vfmadd231ps %ymm5, %ymm13, %ymm4 # ymm4 = (ymm13 * ymm5) + ymm4 vfmadd231ps %ymm5, %ymm12, %ymm3 # ymm3 = (ymm12 * ymm5) + ymm3 vfmadd213ps %ymm0, %ymm1, %ymm5 # ymm5 = (ymm1 * ymm5) + ymm0 vfmadd231ps %ymm6, %ymm13, %ymm5 # ymm5 = (ymm13 * ymm6) + ymm5 vfmadd231ps %ymm6, %ymm12, %ymm4 # ymm4 = (ymm12 * ymm6) + ymm4 vfmadd231ps 0xa0(%r15,%r9), %ymm12, %ymm5 # ymm5 = (ymm12 * mem) + ymm5 vmovaps 0x80(%rbx,%r11), %ymm11 vmovaps 0xa0(%rbx,%r11), %ymm8 vmovaps (%r12,%r9), %ymm6 vmovaps 0x20(%r12,%r9), %ymm7 vmovaps 0x40(%r12,%r9), %ymm10 vmovaps 0x60(%r12,%r9), %ymm15 vmovaps 0x80(%r12,%r9), %ymm14 vfmadd231ps %ymm6, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm6) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm6 # ymm6 = (ymm1 * ymm6) + ymm0 vfmadd231ps %ymm7, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm7) + ymm3 vfmadd231ps %ymm7, %ymm13, %ymm6 # ymm6 = (ymm13 * ymm7) + ymm6 vfmadd231ps %ymm7, %ymm11, %ymm2 # ymm2 = (ymm11 * ymm7) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm7 # ymm7 = (ymm1 * ymm7) + ymm0 vfmadd231ps %ymm10, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm10) + ymm4 vfmadd231ps %ymm10, %ymm13, %ymm7 # ymm7 = (ymm13 * ymm10) + ymm7 vfmadd231ps %ymm10, %ymm11, %ymm3 # ymm3 = (ymm11 * ymm10) + ymm3 vfmadd231ps %ymm10, %ymm12, %ymm6 # ymm6 = (ymm12 * ymm10) + ymm6 vfmadd231ps %ymm10, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm10) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm10 # ymm10 = (ymm1 * ymm10) + ymm0 vfmadd213ps %ymm0, %ymm15, %ymm1 # ymm1 = (ymm15 * ymm1) + ymm0 vfmadd231ps %ymm15, %ymm13, %ymm10 # ymm10 = (ymm13 * ymm15) + ymm10 vfmadd231ps %ymm13, %ymm14, %ymm1 # ymm1 = (ymm14 * ymm13) + ymm1 vmovaps 0xa0(%r12,%r9), %ymm13 vfmadd231ps %ymm15, %ymm9, %ymm5 # ymm5 = (ymm9 * ymm15) + ymm5 vfmadd231ps %ymm15, %ymm11, %ymm4 # ymm4 = (ymm11 * ymm15) + ymm4 vfmadd231ps %ymm14, %ymm11, %ymm5 # ymm5 = (ymm11 * ymm14) + ymm5 vfmadd231ps %ymm15, %ymm12, %ymm7 # ymm7 = (ymm12 * ymm15) + ymm7 vfmadd231ps %ymm14, %ymm12, %ymm10 # ymm10 = (ymm12 * ymm14) + ymm10 vfmadd231ps %ymm12, %ymm13, %ymm1 # ymm1 = (ymm13 * ymm12) + ymm1 vfmadd231ps %ymm15, %ymm8, %ymm3 # ymm3 = (ymm8 * ymm15) + ymm3 vfmadd231ps %ymm14, %ymm8, %ymm4 # ymm4 = (ymm8 * ymm14) + ymm4 vfmadd231ps %ymm13, %ymm8, %ymm5 # ymm5 = (ymm8 * ymm13) + ymm5 vmovaps (%r13,%r9), %ymm14 vmovaps 0x20(%r13,%r9), %ymm15 vmovaps 0x40(%r13,%r9), %ymm13 vmovaps 0x60(%r13,%r9), %ymm12 vfmadd231ps %ymm14, %ymm9, %ymm6 # ymm6 = (ymm9 * ymm14) + ymm6 vfmadd231ps %ymm15, %ymm9, %ymm7 # ymm7 = (ymm9 * ymm15) + ymm7 vfmadd231ps %ymm13, %ymm9, %ymm10 # ymm10 = (ymm9 * ymm13) + ymm10 vfmadd231ps %ymm9, %ymm12, %ymm1 # ymm1 = (ymm12 * ymm9) + ymm1 vmovaps 0xc0(%rbx,%r11), %ymm9 vfmadd231ps %ymm14, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm14) + ymm2 vmovaps 0x80(%r13,%r9), %ymm14 vfmadd231ps %ymm15, %ymm11, %ymm6 # ymm6 = (ymm11 * ymm15) + ymm6 vfmadd231ps %ymm13, %ymm11, %ymm7 # ymm7 = (ymm11 * ymm13) + ymm7 vfmadd231ps %ymm12, %ymm11, %ymm10 # ymm10 = (ymm11 * ymm12) + ymm10 vfmadd231ps %ymm11, %ymm14, %ymm1 # ymm1 = (ymm14 * ymm11) + ymm1 vmovaps 0xe0(%rbx,%r11), %ymm11 vfmadd231ps %ymm15, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm15) + ymm3 vfmadd231ps %ymm15, %ymm11, %ymm2 # ymm2 = (ymm11 * ymm15) + ymm2 vmovaps 0xa0(%r13,%r9), %ymm15 vfmadd231ps %ymm13, %ymm8, %ymm6 # ymm6 = (ymm8 * ymm13) + ymm6 vfmadd231ps %ymm12, %ymm8, %ymm7 # ymm7 = (ymm8 * ymm12) + ymm7 vfmadd231ps %ymm14, %ymm8, %ymm10 # ymm10 = (ymm8 * ymm14) + ymm10 vfmadd231ps %ymm8, %ymm15, %ymm1 # ymm1 = (ymm15 * ymm8) + ymm1 vmovaps 0x100(%rbx,%r11), %ymm8 vfmadd231ps %ymm13, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm13) + ymm4 vfmadd231ps %ymm13, %ymm11, %ymm3 # ymm3 = (ymm11 * ymm13) + ymm3 vfmadd231ps %ymm13, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm13) + ymm2 vfmadd231ps %ymm12, %ymm9, %ymm5 # ymm5 = (ymm9 * ymm12) + ymm5 vfmadd231ps %ymm12, %ymm11, %ymm4 # ymm4 = (ymm11 * ymm12) + ymm4 vfmadd231ps %ymm14, %ymm11, %ymm5 # ymm5 = (ymm11 * ymm14) + ymm5 vfmadd231ps %ymm12, %ymm8, %ymm3 # ymm3 = (ymm8 * ymm12) + ymm3 vfmadd231ps %ymm14, %ymm8, %ymm4 # ymm4 = (ymm8 * ymm14) + ymm4 vfmadd231ps %ymm15, %ymm8, %ymm5 # ymm5 = (ymm8 * ymm15) + ymm5 vmovaps 0x20(%rsi,%r9), %ymm12 vmovaps 0x40(%rsi,%r9), %ymm13 vfmadd231ps (%rsi,%r9), %ymm9, %ymm6 # ymm6 = (ymm9 * mem) + ymm6 vmovaps 0x60(%rsi,%r9), %ymm14 vfmadd231ps %ymm12, %ymm9, %ymm7 # ymm7 = (ymm9 * ymm12) + ymm7 vfmadd231ps %ymm13, %ymm9, %ymm10 # ymm10 = (ymm9 * ymm13) + ymm10 vfmadd231ps %ymm9, %ymm14, %ymm1 # ymm1 = (ymm14 * ymm9) + ymm1 vfmadd231ps %ymm12, %ymm11, %ymm6 # ymm6 = (ymm11 * ymm12) + ymm6 vmovaps 0x80(%rsi,%r9), %ymm9 vfmadd231ps %ymm13, %ymm11, %ymm7 # ymm7 = (ymm11 * ymm13) + ymm7 vfmadd231ps %ymm14, %ymm11, %ymm10 # ymm10 = (ymm11 * ymm14) + ymm10 vfmadd231ps %ymm11, %ymm9, %ymm1 # ymm1 = (ymm9 * ymm11) + ymm1 vfmadd231ps %ymm13, %ymm8, %ymm6 # ymm6 = (ymm8 * ymm13) + ymm6 vfmadd231ps %ymm14, %ymm8, %ymm7 # ymm7 = (ymm8 * ymm14) + ymm7 vfmadd231ps %ymm9, %ymm8, %ymm10 # ymm10 = (ymm8 * ymm9) + ymm10 vfmadd231ps 0xa0(%rsi,%r9), %ymm8, %ymm1 # ymm1 = (ymm8 * mem) + ymm1 vmovaps %ymm2, (%r10,%r9) vmovaps %ymm3, 0x20(%r10,%r9) vmovaps %ymm4, 0x40(%r10,%r9) vmovaps %ymm5, 0x60(%r10,%r9) vmovaps %ymm6, (%rdx,%r9) vmovaps %ymm7, 0x20(%rdx,%r9) vmovaps %ymm10, 0x40(%rdx,%r9) vmovaps %ymm1, 0x60(%rdx,%r9) addl $0x4, %r8d subq $-0x80, %r9 jmp 0x29c18c vmovaps 0x20(%r15,%r9), %ymm3 vmovaps 0x40(%r15,%r9), %ymm4 vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm9 vmovaps 0x40(%rbx,%r11), %ymm8 vmovaps 0x60(%rbx,%r11), %ymm5 vmovaps (%r15,%r9), %ymm2 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vfmadd231ps %ymm3, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm3) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm3) + ymm0 vfmadd231ps %ymm4, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm4) + ymm3 vfmadd231ps %ymm4, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm4) + ymm2 vfmadd231ps 0x60(%r15,%r9), %ymm8, %ymm3 # ymm3 = (ymm8 * mem) + ymm3 vmovaps 0x80(%rbx,%r11), %ymm7 vmovaps 0xa0(%rbx,%r11), %ymm6 vmovaps (%r12,%r9), %ymm4 vmovaps 0x20(%r12,%r9), %ymm10 vmovaps 0x40(%r12,%r9), %ymm11 vmovaps 0x60(%r12,%r9), %ymm12 vfmadd231ps %ymm4, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm4) + ymm2 vfmadd231ps %ymm10, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm10) + ymm3 vfmadd213ps %ymm0, %ymm1, %ymm4 # ymm4 = (ymm1 * ymm4) + ymm0 vfmadd213ps %ymm0, %ymm10, %ymm1 # ymm1 = (ymm10 * ymm1) + ymm0 vfmadd231ps %ymm10, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm10) + ymm2 vfmadd231ps %ymm11, %ymm7, %ymm3 # ymm3 = (ymm7 * ymm11) + ymm3 vfmadd231ps %ymm10, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm10) + ymm4 vfmadd231ps %ymm9, %ymm11, %ymm1 # ymm1 = (ymm11 * ymm9) + ymm1 vfmadd231ps %ymm11, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm11) + ymm2 vfmadd231ps %ymm12, %ymm6, %ymm3 # ymm3 = (ymm6 * ymm12) + ymm3 vfmadd231ps %ymm11, %ymm8, %ymm4 # ymm4 = (ymm8 * ymm11) + ymm4 vfmadd231ps %ymm12, %ymm8, %ymm1 # ymm1 = (ymm8 * ymm12) + ymm1 vmovaps 0xc0(%rbx,%r11), %ymm10 vmovaps 0xe0(%rbx,%r11), %ymm9 vmovaps 0x100(%rbx,%r11), %ymm8 vmovaps (%r13,%r9), %ymm11 vmovaps 0x20(%r13,%r9), %ymm12 vmovaps 0x40(%r13,%r9), %ymm13 vmovaps 0x60(%r13,%r9), %ymm14 vfmadd231ps %ymm11, %ymm10, %ymm2 # ymm2 = (ymm10 * ymm11) + ymm2 vfmadd231ps %ymm12, %ymm10, %ymm3 # ymm3 = (ymm10 * ymm12) + ymm3 vfmadd231ps %ymm11, %ymm5, %ymm4 # ymm4 = (ymm5 * ymm11) + ymm4 vfmadd231ps %ymm5, %ymm12, %ymm1 # ymm1 = (ymm12 * ymm5) + ymm1 vfmadd231ps %ymm12, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm12) + ymm2 vfmadd231ps %ymm13, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm13) + ymm3 vfmadd231ps %ymm12, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm12) + ymm4 vfmadd231ps %ymm7, %ymm13, %ymm1 # ymm1 = (ymm13 * ymm7) + ymm1 vfmadd231ps %ymm13, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm13) + ymm2 vfmadd231ps %ymm14, %ymm8, %ymm3 # ymm3 = (ymm8 * ymm14) + ymm3 vfmadd231ps %ymm13, %ymm6, %ymm4 # ymm4 = (ymm6 * ymm13) + ymm4 vfmadd231ps %ymm14, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm14) + ymm1 vmovaps 0x20(%rsi,%r9), %ymm5 vmovaps 0x40(%rsi,%r9), %ymm6 vfmadd231ps (%rsi,%r9), %ymm10, %ymm4 # ymm4 = (ymm10 * mem) + ymm4 vfmadd231ps %ymm10, %ymm5, %ymm1 # ymm1 = (ymm5 * ymm10) + ymm1 vfmadd231ps %ymm5, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm5) + ymm4 vfmadd231ps %ymm9, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm9) + ymm1 vfmadd231ps %ymm6, %ymm8, %ymm4 # ymm4 = (ymm8 * ymm6) + ymm4 vfmadd231ps 0x60(%rsi,%r9), %ymm8, %ymm1 # ymm1 = (ymm8 * mem) + ymm1 vmovaps %ymm2, (%r10,%r9) vmovaps %ymm3, 0x20(%r10,%r9) vmovaps %ymm4, (%rdx,%r9) vmovaps %ymm1, 0x20(%rdx,%r9) addl $0x2, %r8d addq $0x40, %r9 leal 0x1(%r8), %r14d cmpl %eax, %r14d jl 0x29c43f movq 0x8(%rsp), %r14 cmpl %eax, %r8d jge 0x29c6d2 vmovaps (%rbx,%r11), %ymm4 vmovaps 0x20(%rbx,%r11), %ymm5 vmovaps 0x40(%rbx,%r11), %ymm6 vmovaps (%r15,%r9), %ymm1 vfmadd213ps %ymm0, %ymm4, %ymm1 # ymm1 = (ymm4 * ymm1) + ymm0 vfmadd231ps 0x20(%r15,%r9), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vmovaps 0x60(%rbx,%r11), %ymm7 vfmadd231ps 0x40(%r15,%r9), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps 0x80(%rbx,%r11), %ymm8 vmovaps 0xa0(%rbx,%r11), %ymm3 vmovaps (%r12,%r9), %ymm2 vmovaps 0x20(%r12,%r9), %ymm9 vmovaps 0x40(%r12,%r9), %ymm10 vfmadd231ps %ymm2, %ymm7, %ymm1 # ymm1 = (ymm7 * ymm2) + ymm1 vfmadd213ps %ymm0, %ymm4, %ymm2 # ymm2 = (ymm4 * ymm2) + ymm0 vfmadd231ps %ymm9, %ymm8, %ymm1 # ymm1 = (ymm8 * ymm9) + ymm1 vfmadd231ps %ymm9, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm9) + ymm2 vfmadd231ps %ymm10, %ymm3, %ymm1 # ymm1 = (ymm3 * ymm10) + ymm1 vfmadd231ps %ymm10, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm10) + ymm2 vmovaps 0xc0(%rbx,%r11), %ymm4 vmovaps 0xe0(%rbx,%r11), %ymm5 vmovaps 0x100(%rbx,%r11), %ymm6 vmovaps (%r13,%r9), %ymm9 vmovaps 0x20(%r13,%r9), %ymm10 vmovaps 0x40(%r13,%r9), %ymm11 vfmadd231ps %ymm9, %ymm4, %ymm1 # ymm1 = (ymm4 * ymm9) + ymm1 vfmadd231ps %ymm9, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm9) + ymm2 vfmadd231ps %ymm10, %ymm5, %ymm1 # ymm1 = (ymm5 * ymm10) + ymm1 vfmadd231ps %ymm10, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm10) + ymm2 vfmadd231ps %ymm11, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm11) + ymm1 vfmadd231ps %ymm11, %ymm3, %ymm2 # ymm2 = (ymm3 * ymm11) + ymm2 vfmadd231ps (%rsi,%r9), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x20(%rsi,%r9), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps 0x40(%rsi,%r9), %ymm6, %ymm2 # ymm2 = (ymm6 * mem) + ymm2 vmovaps %ymm1, (%r10,%r9) vmovaps %ymm2, (%rdx,%r9) incl %r8d addq $0x20, %r9 jmp 0x29c5d2 addq %rdi, %r15 addq %r9, %r15 addq %rdi, %r12 addq %r9, %r12 addq %rdi, %r13 addq %r9, %r13 addq %rdi, %rsi addq %r9, %rsi movq 0x10(%rsp), %r8 addq %r8, %r10 addq %r9, %r10 addq %r8, %rdx addq %r9, %rdx addl $0x2, %ebp jmp 0x29c176 xorl %edx, %edx xorl %esi, %esi leal 0x3(%rsi), %r8d cmpl %eax, %r8d jge 0x29c9a5 vmovaps 0x20(%r15,%rdx), %ymm3 vmovaps 0x40(%r15,%rdx), %ymm4 vmovaps 0x60(%r15,%rdx), %ymm8 vmovaps 0x80(%r15,%rdx), %ymm6 vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm9 vmovaps 0x40(%rbx,%r11), %ymm7 vmovaps 0x60(%rbx,%r11), %ymm5 vmovaps (%r15,%rdx), %ymm2 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vfmadd231ps %ymm3, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm3) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm3) + ymm0 vfmadd231ps %ymm4, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm4) + ymm3 vfmadd231ps %ymm4, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm4) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm4 # ymm4 = (ymm1 * ymm4) + ymm0 vfmadd213ps %ymm0, %ymm8, %ymm1 # ymm1 = (ymm8 * ymm1) + ymm0 vfmadd231ps %ymm8, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm8) + ymm4 vfmadd231ps %ymm9, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm9) + ymm1 vfmadd231ps %ymm8, %ymm7, %ymm3 # ymm3 = (ymm7 * ymm8) + ymm3 vfmadd231ps 0xa0(%r15,%rdx), %ymm7, %ymm1 # ymm1 = (ymm7 * mem) + ymm1 vfmadd231ps %ymm6, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm6) + ymm4 vmovaps 0x80(%rbx,%r11), %ymm6 vmovaps 0xa0(%rbx,%r11), %ymm7 vmovaps 0x20(%r12,%rdx), %ymm8 vmovaps 0x40(%r12,%rdx), %ymm9 vmovaps 0x60(%r12,%rdx), %ymm10 vfmadd231ps (%r12,%rdx), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vmovaps 0x80(%r12,%rdx), %ymm11 vfmadd231ps %ymm8, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm8) + ymm3 vfmadd231ps %ymm9, %ymm5, %ymm4 # ymm4 = (ymm5 * ymm9) + ymm4 vfmadd231ps %ymm5, %ymm10, %ymm1 # ymm1 = (ymm10 * ymm5) + ymm1 vfmadd231ps %ymm8, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm8) + ymm2 vfmadd231ps %ymm9, %ymm6, %ymm3 # ymm3 = (ymm6 * ymm9) + ymm3 vfmadd231ps %ymm10, %ymm6, %ymm4 # ymm4 = (ymm6 * ymm10) + ymm4 vfmadd231ps %ymm6, %ymm11, %ymm1 # ymm1 = (ymm11 * ymm6) + ymm1 vfmadd231ps %ymm9, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm9) + ymm2 vfmadd231ps %ymm10, %ymm7, %ymm3 # ymm3 = (ymm7 * ymm10) + ymm3 vfmadd231ps 0xa0(%r12,%rdx), %ymm7, %ymm1 # ymm1 = (ymm7 * mem) + ymm1 vfmadd231ps %ymm11, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm11) + ymm4 vmovaps 0xc0(%rbx,%r11), %ymm5 vmovaps 0xe0(%rbx,%r11), %ymm6 vmovaps 0x100(%rbx,%r11), %ymm7 vmovaps 0x20(%r13,%rdx), %ymm8 vmovaps 0x40(%r13,%rdx), %ymm9 vmovaps 0x60(%r13,%rdx), %ymm10 vmovaps 0x80(%r13,%rdx), %ymm11 vfmadd231ps (%r13,%rdx), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps %ymm8, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm8) + ymm3 vfmadd231ps %ymm9, %ymm5, %ymm4 # ymm4 = (ymm5 * ymm9) + ymm4 vfmadd231ps %ymm5, %ymm10, %ymm1 # ymm1 = (ymm10 * ymm5) + ymm1 vfmadd231ps %ymm8, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm8) + ymm2 vfmadd231ps %ymm9, %ymm6, %ymm3 # ymm3 = (ymm6 * ymm9) + ymm3 vfmadd231ps %ymm10, %ymm6, %ymm4 # ymm4 = (ymm6 * ymm10) + ymm4 vfmadd231ps %ymm6, %ymm11, %ymm1 # ymm1 = (ymm11 * ymm6) + ymm1 vfmadd231ps %ymm9, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm9) + ymm2 vfmadd231ps %ymm10, %ymm7, %ymm3 # ymm3 = (ymm7 * ymm10) + ymm3 vfmadd231ps %ymm11, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm11) + ymm4 vfmadd231ps 0xa0(%r13,%rdx), %ymm7, %ymm1 # ymm1 = (ymm7 * mem) + ymm1 vmovaps %ymm2, (%r10,%rdx) vmovaps %ymm3, 0x20(%r10,%rdx) vmovaps %ymm4, 0x40(%r10,%rdx) vmovaps %ymm1, 0x60(%r10,%rdx) addl $0x4, %esi subq $-0x80, %rdx jmp 0x29c707 vmovaps 0x20(%r15,%rdx), %ymm3 vmovaps 0x40(%r15,%rdx), %ymm4 vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm5 vmovaps 0x40(%rbx,%r11), %ymm6 vmovaps 0x60(%rbx,%r11), %ymm7 vmovaps (%r15,%rdx), %ymm2 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vfmadd213ps %ymm0, %ymm3, %ymm1 # ymm1 = (ymm3 * ymm1) + ymm0 vfmadd231ps %ymm3, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm3) + ymm2 vfmadd231ps %ymm5, %ymm4, %ymm1 # ymm1 = (ymm4 * ymm5) + ymm1 vfmadd231ps %ymm4, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm4) + ymm2 vfmadd231ps 0x60(%r15,%rdx), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps 0x80(%rbx,%r11), %ymm3 vmovaps 0xa0(%rbx,%r11), %ymm4 vmovaps 0x20(%r12,%rdx), %ymm5 vfmadd231ps (%r12,%rdx), %ymm7, %ymm2 # ymm2 = (ymm7 * mem) + ymm2 vmovaps 0x40(%r12,%rdx), %ymm6 vfmadd231ps %ymm7, %ymm5, %ymm1 # ymm1 = (ymm5 * ymm7) + ymm1 vfmadd231ps %ymm5, %ymm3, %ymm2 # ymm2 = (ymm3 * ymm5) + ymm2 vfmadd231ps %ymm3, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm3) + ymm1 vfmadd231ps %ymm6, %ymm4, %ymm2 # ymm2 = (ymm4 * ymm6) + ymm2 vfmadd231ps 0x60(%r12,%rdx), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vmovaps 0xc0(%rbx,%r11), %ymm3 vmovaps 0xe0(%rbx,%r11), %ymm4 vmovaps 0x100(%rbx,%r11), %ymm5 vmovaps 0x20(%r13,%rdx), %ymm6 vmovaps 0x40(%r13,%rdx), %ymm7 vfmadd231ps (%r13,%rdx), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps %ymm3, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm3) + ymm1 vfmadd231ps %ymm6, %ymm4, %ymm2 # ymm2 = (ymm4 * ymm6) + ymm2 vfmadd231ps %ymm4, %ymm7, %ymm1 # ymm1 = (ymm7 * ymm4) + ymm1 vfmadd231ps %ymm7, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm7) + ymm2 vfmadd231ps 0x60(%r13,%rdx), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vmovaps %ymm2, (%r10,%rdx) vmovaps %ymm1, 0x20(%r10,%rdx) addl $0x2, %esi addq $0x40, %rdx leal 0x1(%rsi), %r8d cmpl %eax, %r8d jl 0x29c8b1 jmp 0x29ca4d vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm2 vmovaps 0x40(%rbx,%r11), %ymm3 vfmadd132ps (%r15,%rdx), %ymm0, %ymm1 # ymm1 = (ymm1 * mem) + ymm0 vfmadd231ps 0x20(%r15,%rdx), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vmovaps 0x60(%rbx,%r11), %ymm2 vfmadd231ps 0x40(%r15,%rdx), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vmovaps 0x80(%rbx,%r11), %ymm3 vfmadd231ps (%r12,%rdx), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r12,%rdx), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vmovaps 0xa0(%rbx,%r11), %ymm2 vfmadd231ps 0x40(%r12,%rdx), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vmovaps 0xc0(%rbx,%r11), %ymm2 vmovaps 0xe0(%rbx,%r11), %ymm3 vmovaps 0x100(%rbx,%r11), %ymm4 vfmadd132ps (%r13,%rdx), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r13,%rdx), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r13,%rdx), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vmovaps %ymm2, (%r10,%rdx) incl %esi addq $0x20, %rdx cmpl %eax, %esi jl 0x29c9b7 addq %rdx, %r15 addq $0x40, %r15 addq %rdx, %r12 addq $0x40, %r12 addq %rdx, %r13 addq $0x40, %r13 incl %ebp addq %rdx, %r10 cmpl %ecx, %ebp jl 0x29c703 movq 0x30(%rsp), %r9 incq %r9 movq 0x28(%rsp), %r13 movq 0x60(%rsp), %r8 jmp 0x29c0e1 movl 0xd4(%rdi), %eax movl 0xd8(%rdi), %r15d cmpl $0x5, %eax je 0x29d684 cmpl $0x3, %eax jne 0x29ed47 cmpl $0x3, %r15d jne 0x29ed47 cmpl $0x1, 0xdc(%rdi) jne 0x29e688 cmpl $0x1, 0xe0(%rdi) jne 0x29e688 cmpl $0x1, 0xe4(%rdi) jne 0x29e688 cmpl $0x1, 0xe8(%rdi) jne 0x29e688 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xb8(%rsp), %edx movq 0x1b0(%rdi), %rsi xorl %edi, %edi testl %ecx, %ecx cmovlel %edi, %ecx testl %edx, %edx cmovlel %edi, %edx movq %rdx, 0x1d8(%rsp) movl $0x80, %r8d movq %rsi, 0x1d0(%rsp) cmpq 0x1d8(%rsp), %rdi je 0x29fdf4 testq %rsi, %rsi je 0x29cb44 movq %rdi, %rdx shlq $0x4, %rdx vmovups (%rsi,%rdx), %xmm0 jmp 0x29cb48 vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x60(%rsp) movq 0x40(%r13), %r9 imulq %rdi, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %rdx movslq 0x54(%r14), %rsi imulq %rdi, %rsi imulq 0x38(%r14), %rsi movslq 0xac(%rsp), %r10 movq 0xc0(%rsp), %r13 imulq %rdi, %r13 movq 0x90(%rsp), %r11 imulq %r11, %r13 addq 0x80(%rsp), %r13 imulq %r11, %r10 leaq (%r10,%r13), %rbp leaq (,%r10,2), %r11 addq %r13, %r11 vmovaps (%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x70(%rsp) vmovaps 0x10(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x10(%rsp) vmovaps 0x20(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0xd0(%rsp) vmovaps 0x30(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0xe0(%rsp) vmovaps 0x40(%rdx,%rsi), %xmm15 vmovaps 0x50(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x170(%rsp) vmovaps 0x60(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x30(%rsp) vmovaps 0x70(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x160(%rsp) vmovaps 0x80(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x150(%rsp) xorl %r10d, %r10d vmovaps %xmm15, 0x40(%rsp) cmpl %ecx, %r10d je 0x29d323 leaq 0x20(%r13), %r15 leaq 0x20(%rbp), %rbx leaq 0x20(%r11), %rdx xorl %esi, %esi xorl %r12d, %r12d leal 0x7(%r12), %r14d cmpl %eax, %r14d jge 0x29d152 vmovaps (%r13,%rsi), %xmm0 vmovaps 0x60(%rsp), %xmm14 vmovaps 0x70(%rsp), %xmm5 vfmadd213ps %xmm14, %xmm5, %xmm0 # xmm0 = (xmm5 * xmm0) + xmm14 vmovaps 0x10(%r13,%rsi), %xmm1 vmovaps 0x20(%r13,%rsi), %xmm11 vmovaps 0x30(%r13,%rsi), %xmm10 vfmadd231ps 0x10(%rsp), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0xd0(%rsp), %xmm6 vfmadd231ps %xmm11, %xmm6, %xmm0 # xmm0 = (xmm6 * xmm11) + xmm0 vmovaps 0xe0(%rsp), %xmm7 vfmadd231ps (%rbp,%rsi), %xmm7, %xmm0 # xmm0 = (xmm7 * mem) + xmm0 vmovaps 0x10(%rbp,%rsi), %xmm2 vmovaps 0x20(%rbp,%rsi), %xmm3 vfmadd231ps %xmm2, %xmm15, %xmm0 # xmm0 = (xmm15 * xmm2) + xmm0 vmovaps 0x170(%rsp), %xmm8 vfmadd231ps %xmm3, %xmm8, %xmm0 # xmm0 = (xmm8 * xmm3) + xmm0 vmovaps 0x30(%rsp), %xmm4 vfmadd231ps (%r11,%rsi), %xmm4, %xmm0 # xmm0 = (xmm4 * mem) + xmm0 vmovaps 0x30(%rbp,%rsi), %xmm15 vmovaps %xmm15, 0x280(%rsp) vmovaps 0x10(%r11,%rsi), %xmm12 vmovaps 0x20(%r11,%rsi), %xmm13 vmovaps 0x30(%r11,%rsi), %xmm4 vmovaps %xmm4, 0xf0(%rsp) vmovaps 0x160(%rsp), %xmm9 vfmadd231ps %xmm12, %xmm9, %xmm0 # xmm0 = (xmm9 * xmm12) + xmm0 vmovaps 0x150(%rsp), %xmm4 vfmadd231ps %xmm13, %xmm4, %xmm0 # xmm0 = (xmm4 * xmm13) + xmm0 vmovaps %xmm0, (%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm1 # xmm1 = (xmm5 * xmm1) + xmm14 vfmadd231ps 0x10(%rsp), %xmm11, %xmm1 # xmm1 = (xmm11 * mem) + xmm1 vfmadd231ps %xmm10, %xmm6, %xmm1 # xmm1 = (xmm6 * xmm10) + xmm1 vfmadd231ps %xmm2, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm2) + xmm1 vfmadd231ps 0x40(%rsp), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps %xmm15, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm15) + xmm1 vfmadd231ps 0x30(%rsp), %xmm12, %xmm1 # xmm1 = (xmm12 * mem) + xmm1 vfmadd231ps %xmm13, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm13) + xmm1 vmovaps 0xf0(%rsp), %xmm0 vfmadd231ps %xmm0, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm0) + xmm1 vmovaps 0x40(%r13,%rsi), %xmm12 vmovaps 0x40(%rbp,%rsi), %xmm2 vmovaps 0x40(%r11,%rsi), %xmm15 vmovaps %xmm15, 0x270(%rsp) vmovaps %xmm1, 0x10(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm11 # xmm11 = (xmm5 * xmm11) + xmm14 vfmadd231ps 0x10(%rsp), %xmm10, %xmm11 # xmm11 = (xmm10 * mem) + xmm11 vfmadd231ps %xmm12, %xmm6, %xmm11 # xmm11 = (xmm6 * xmm12) + xmm11 vfmadd231ps %xmm3, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm3) + xmm11 vmovaps 0x280(%rsp), %xmm15 vfmadd231ps 0x40(%rsp), %xmm15, %xmm11 # xmm11 = (xmm15 * mem) + xmm11 vfmadd231ps %xmm2, %xmm8, %xmm11 # xmm11 = (xmm8 * xmm2) + xmm11 vfmadd231ps 0x30(%rsp), %xmm13, %xmm11 # xmm11 = (xmm13 * mem) + xmm11 vfmadd231ps %xmm0, %xmm9, %xmm11 # xmm11 = (xmm9 * xmm0) + xmm11 vmovaps 0x270(%rsp), %xmm0 vfmadd231ps %xmm0, %xmm4, %xmm11 # xmm11 = (xmm4 * xmm0) + xmm11 vmovaps 0x50(%r13,%rsi), %xmm13 vmovaps 0x50(%rbp,%rsi), %xmm3 vmovaps 0x50(%r11,%rsi), %xmm1 vmovaps %xmm11, 0x20(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm10 # xmm10 = (xmm5 * xmm10) + xmm14 vfmadd231ps 0x10(%rsp), %xmm12, %xmm10 # xmm10 = (xmm12 * mem) + xmm10 vfmadd231ps %xmm13, %xmm6, %xmm10 # xmm10 = (xmm6 * xmm13) + xmm10 vfmadd231ps %xmm15, %xmm7, %xmm10 # xmm10 = (xmm7 * xmm15) + xmm10 vfmadd231ps 0x40(%rsp), %xmm2, %xmm10 # xmm10 = (xmm2 * mem) + xmm10 vfmadd231ps %xmm3, %xmm8, %xmm10 # xmm10 = (xmm8 * xmm3) + xmm10 vmovaps 0xf0(%rsp), %xmm11 vfmadd231ps 0x30(%rsp), %xmm11, %xmm10 # xmm10 = (xmm11 * mem) + xmm10 vfmadd231ps %xmm0, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm0) + xmm10 vfmadd231ps %xmm1, %xmm4, %xmm10 # xmm10 = (xmm4 * xmm1) + xmm10 vmovaps %xmm1, 0x290(%rsp) vmovaps 0x60(%r13,%rsi), %xmm11 vmovaps 0x60(%rbp,%rsi), %xmm15 vmovaps 0x60(%r11,%rsi), %xmm0 vmovaps %xmm10, 0x30(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm14 vfmadd231ps 0x10(%rsp), %xmm13, %xmm12 # xmm12 = (xmm13 * mem) + xmm12 vfmadd231ps %xmm11, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm11) + xmm12 vfmadd231ps %xmm2, %xmm7, %xmm12 # xmm12 = (xmm7 * xmm2) + xmm12 vfmadd231ps 0x40(%rsp), %xmm3, %xmm12 # xmm12 = (xmm3 * mem) + xmm12 vfmadd231ps %xmm15, %xmm8, %xmm12 # xmm12 = (xmm8 * xmm15) + xmm12 vmovaps 0x270(%rsp), %xmm2 vfmadd231ps 0x30(%rsp), %xmm2, %xmm12 # xmm12 = (xmm2 * mem) + xmm12 vfmadd231ps %xmm1, %xmm9, %xmm12 # xmm12 = (xmm9 * xmm1) + xmm12 vfmadd231ps %xmm0, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm0) + xmm12 vmovaps %xmm0, 0xf0(%rsp) vmovaps 0x70(%r13,%rsi), %xmm10 vmovaps 0x70(%rbp,%rsi), %xmm2 vmovaps 0x70(%r11,%rsi), %xmm1 vmovaps %xmm12, 0x40(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm13 # xmm13 = (xmm5 * xmm13) + xmm14 vfmadd231ps 0x10(%rsp), %xmm11, %xmm13 # xmm13 = (xmm11 * mem) + xmm13 vfmadd231ps %xmm10, %xmm6, %xmm13 # xmm13 = (xmm6 * xmm10) + xmm13 vfmadd231ps %xmm3, %xmm7, %xmm13 # xmm13 = (xmm7 * xmm3) + xmm13 vfmadd231ps 0x40(%rsp), %xmm15, %xmm13 # xmm13 = (xmm15 * mem) + xmm13 vfmadd231ps %xmm2, %xmm8, %xmm13 # xmm13 = (xmm8 * xmm2) + xmm13 vmovaps 0x290(%rsp), %xmm3 vfmadd231ps 0x30(%rsp), %xmm3, %xmm13 # xmm13 = (xmm3 * mem) + xmm13 vfmadd231ps %xmm0, %xmm9, %xmm13 # xmm13 = (xmm9 * xmm0) + xmm13 vfmadd231ps %xmm1, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm1) + xmm13 vmovaps 0x80(%r13,%rsi), %xmm3 vmovaps 0x80(%rbp,%rsi), %xmm12 vmovaps 0x80(%r11,%rsi), %xmm0 vmovaps %xmm13, 0x50(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm11 # xmm11 = (xmm5 * xmm11) + xmm14 vfmadd231ps 0x10(%rsp), %xmm10, %xmm11 # xmm11 = (xmm10 * mem) + xmm11 vfmadd231ps %xmm3, %xmm6, %xmm11 # xmm11 = (xmm6 * xmm3) + xmm11 vfmadd231ps %xmm15, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm15) + xmm11 vmovaps 0x40(%rsp), %xmm15 vfmadd231ps %xmm2, %xmm15, %xmm11 # xmm11 = (xmm15 * xmm2) + xmm11 vfmadd231ps %xmm12, %xmm8, %xmm11 # xmm11 = (xmm8 * xmm12) + xmm11 vmovaps 0xf0(%rsp), %xmm13 vfmadd231ps 0x30(%rsp), %xmm13, %xmm11 # xmm11 = (xmm13 * mem) + xmm11 vfmadd213ps %xmm14, %xmm5, %xmm10 # xmm10 = (xmm5 * xmm10) + xmm14 vfmadd231ps 0x10(%rsp), %xmm3, %xmm10 # xmm10 = (xmm3 * mem) + xmm10 vfmadd231ps 0x90(%r13,%rsi), %xmm6, %xmm10 # xmm10 = (xmm6 * mem) + xmm10 vfmadd231ps %xmm2, %xmm7, %xmm10 # xmm10 = (xmm7 * xmm2) + xmm10 vfmadd231ps %xmm12, %xmm15, %xmm10 # xmm10 = (xmm15 * xmm12) + xmm10 vfmadd231ps %xmm1, %xmm9, %xmm11 # xmm11 = (xmm9 * xmm1) + xmm11 vfmadd231ps 0x90(%rbp,%rsi), %xmm8, %xmm10 # xmm10 = (xmm8 * mem) + xmm10 vfmadd231ps 0x30(%rsp), %xmm1, %xmm10 # xmm10 = (xmm1 * mem) + xmm10 vfmadd231ps %xmm0, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm0) + xmm10 vfmadd231ps 0x90(%r11,%rsi), %xmm4, %xmm10 # xmm10 = (xmm4 * mem) + xmm10 vfmadd231ps %xmm0, %xmm4, %xmm11 # xmm11 = (xmm4 * xmm0) + xmm11 vmovaps %xmm11, 0x60(%r9,%rsi) vmovaps %xmm10, 0x70(%r9,%rsi) addl $0x8, %r12d addq %r8, %rsi addq %r8, %r15 addq %r8, %rbx addq %r8, %rdx jmp 0x29cc45 vmovaps (%r13,%rsi), %xmm13 vmovaps 0x70(%rsp), %xmm12 vfmadd213ps 0x60(%rsp), %xmm12, %xmm13 # xmm13 = (xmm12 * xmm13) + mem vmovaps 0x10(%r13,%rsi), %xmm1 vmovaps 0x20(%r13,%rsi), %xmm10 vmovaps 0x30(%r13,%rsi), %xmm11 vfmadd231ps 0x10(%rsp), %xmm1, %xmm13 # xmm13 = (xmm1 * mem) + xmm13 vmovaps 0xd0(%rsp), %xmm6 vfmadd231ps %xmm10, %xmm6, %xmm13 # xmm13 = (xmm6 * xmm10) + xmm13 vmovaps 0xe0(%rsp), %xmm7 vfmadd231ps (%rbp,%rsi), %xmm7, %xmm13 # xmm13 = (xmm7 * mem) + xmm13 vmovaps 0x10(%rbp,%rsi), %xmm14 vmovaps 0x20(%rbp,%rsi), %xmm5 vfmadd231ps %xmm14, %xmm15, %xmm13 # xmm13 = (xmm15 * xmm14) + xmm13 vmovaps 0x170(%rsp), %xmm8 vfmadd231ps %xmm5, %xmm8, %xmm13 # xmm13 = (xmm8 * xmm5) + xmm13 vmovaps 0x30(%rsp), %xmm0 vfmadd231ps (%r11,%rsi), %xmm0, %xmm13 # xmm13 = (xmm0 * mem) + xmm13 vmovaps 0x30(%rbp,%rsi), %xmm0 vmovaps 0x10(%r11,%rsi), %xmm15 vmovaps 0x20(%r11,%rsi), %xmm3 vmovaps 0x30(%r11,%rsi), %xmm2 vmovaps %xmm2, 0xf0(%rsp) vmovaps 0x160(%rsp), %xmm9 vfmadd231ps %xmm15, %xmm9, %xmm13 # xmm13 = (xmm9 * xmm15) + xmm13 vmovaps 0x150(%rsp), %xmm4 vfmadd231ps %xmm3, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm3) + xmm13 vmovaps %xmm13, (%r9,%rsi) vmovaps 0x60(%rsp), %xmm13 vfmadd213ps %xmm13, %xmm12, %xmm1 # xmm1 = (xmm12 * xmm1) + xmm13 vfmadd231ps 0x10(%rsp), %xmm10, %xmm1 # xmm1 = (xmm10 * mem) + xmm1 vfmadd231ps %xmm11, %xmm6, %xmm1 # xmm1 = (xmm6 * xmm11) + xmm1 vfmadd231ps %xmm14, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm14) + xmm1 vfmadd231ps 0x40(%rsp), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps %xmm0, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm0) + xmm1 vfmadd231ps 0x30(%rsp), %xmm15, %xmm1 # xmm1 = (xmm15 * mem) + xmm1 vfmadd231ps %xmm3, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm3) + xmm1 vfmadd231ps %xmm2, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm2) + xmm1 vmovaps 0x40(%r13,%rsi), %xmm14 vmovaps 0x40(%rbp,%rsi), %xmm15 vmovaps 0x40(%r11,%rsi), %xmm2 vmovaps %xmm1, 0x10(%r9,%rsi) vfmadd213ps %xmm13, %xmm12, %xmm10 # xmm10 = (xmm12 * xmm10) + xmm13 vfmadd231ps 0x10(%rsp), %xmm11, %xmm10 # xmm10 = (xmm11 * mem) + xmm10 vfmadd231ps %xmm14, %xmm6, %xmm10 # xmm10 = (xmm6 * xmm14) + xmm10 vfmadd231ps %xmm5, %xmm7, %xmm10 # xmm10 = (xmm7 * xmm5) + xmm10 vfmadd231ps 0x40(%rsp), %xmm0, %xmm10 # xmm10 = (xmm0 * mem) + xmm10 vfmadd231ps %xmm15, %xmm8, %xmm10 # xmm10 = (xmm8 * xmm15) + xmm10 vfmadd231ps 0x30(%rsp), %xmm3, %xmm10 # xmm10 = (xmm3 * mem) + xmm10 vfmadd213ps %xmm13, %xmm12, %xmm11 # xmm11 = (xmm12 * xmm11) + xmm13 vfmadd231ps 0x10(%rsp), %xmm14, %xmm11 # xmm11 = (xmm14 * mem) + xmm11 vfmadd231ps 0x50(%r13,%rsi), %xmm6, %xmm11 # xmm11 = (xmm6 * mem) + xmm11 vfmadd231ps %xmm0, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm0) + xmm11 vfmadd231ps 0x40(%rsp), %xmm15, %xmm11 # xmm11 = (xmm15 * mem) + xmm11 vmovaps 0x40(%rsp), %xmm15 vmovaps 0xf0(%rsp), %xmm1 vfmadd231ps %xmm1, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm1) + xmm10 vfmadd231ps 0x50(%rbp,%rsi), %xmm8, %xmm11 # xmm11 = (xmm8 * mem) + xmm11 vfmadd231ps 0x30(%rsp), %xmm1, %xmm11 # xmm11 = (xmm1 * mem) + xmm11 vfmadd231ps %xmm2, %xmm9, %xmm11 # xmm11 = (xmm9 * xmm2) + xmm11 vfmadd231ps 0x50(%r11,%rsi), %xmm4, %xmm11 # xmm11 = (xmm4 * mem) + xmm11 vfmadd231ps %xmm2, %xmm4, %xmm10 # xmm10 = (xmm4 * xmm2) + xmm10 vmovaps %xmm10, 0x20(%r9,%rsi) vmovaps %xmm11, 0x30(%r9,%rsi) addl $0x4, %r12d addq $0x40, %rsi addq $0x40, %r15 addq $0x40, %rbx addq $0x40, %rdx leal 0x3(%r12), %r14d cmpl %eax, %r14d jl 0x29cfa2 jmp 0x29d25d vmovaps 0x10(%r13,%rsi), %xmm0 vmovaps 0x20(%r13,%rsi), %xmm2 vmovaps (%r13,%rsi), %xmm1 vmovaps 0x60(%rsp), %xmm3 vmovaps 0x70(%rsp), %xmm4 vfmadd213ps %xmm3, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm1) + xmm3 vmovaps 0x10(%rsp), %xmm6 vfmadd231ps %xmm0, %xmm6, %xmm1 # xmm1 = (xmm6 * xmm0) + xmm1 vmovaps 0xd0(%rsp), %xmm5 vfmadd231ps %xmm2, %xmm5, %xmm1 # xmm1 = (xmm5 * xmm2) + xmm1 vfmadd213ps %xmm3, %xmm4, %xmm0 # xmm0 = (xmm4 * xmm0) + xmm3 vfmadd231ps %xmm2, %xmm6, %xmm0 # xmm0 = (xmm6 * xmm2) + xmm0 vmovaps 0x10(%rbp,%rsi), %xmm2 vmovaps 0xe0(%rsp), %xmm3 vfmadd231ps (%rbp,%rsi), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x30(%r13,%rsi), %xmm5, %xmm0 # xmm0 = (xmm5 * mem) + xmm0 vfmadd231ps %xmm2, %xmm15, %xmm1 # xmm1 = (xmm15 * xmm2) + xmm1 vfmadd231ps %xmm2, %xmm3, %xmm0 # xmm0 = (xmm3 * xmm2) + xmm0 vmovaps 0x20(%rbp,%rsi), %xmm2 vmovaps 0x170(%rsp), %xmm3 vfmadd231ps %xmm2, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm2) + xmm1 vfmadd231ps %xmm2, %xmm15, %xmm0 # xmm0 = (xmm15 * xmm2) + xmm0 vmovaps 0x10(%r11,%rsi), %xmm2 vmovaps 0x30(%rsp), %xmm5 vfmadd231ps (%r11,%rsi), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vmovaps 0x160(%rsp), %xmm4 vfmadd231ps %xmm2, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm2) + xmm1 vfmadd231ps 0x30(%rbp,%rsi), %xmm3, %xmm0 # xmm0 = (xmm3 * mem) + xmm0 vfmadd231ps %xmm2, %xmm5, %xmm0 # xmm0 = (xmm5 * xmm2) + xmm0 vmovaps 0x20(%r11,%rsi), %xmm2 vfmadd231ps %xmm2, %xmm4, %xmm0 # xmm0 = (xmm4 * xmm2) + xmm0 vmovaps 0x150(%rsp), %xmm3 vfmadd231ps 0x30(%r11,%rsi), %xmm3, %xmm0 # xmm0 = (xmm3 * mem) + xmm0 vfmadd231ps %xmm2, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm2) + xmm1 vmovaps %xmm1, (%r9,%rsi) vmovaps %xmm0, 0x10(%r9,%rsi) addl $0x2, %r12d addq $0x20, %rsi addq $0x20, %r15 addq $0x20, %rbx addq $0x20, %rdx leal 0x1(%r12), %r14d cmpl %eax, %r14d jl 0x29d165 addq %rsi, %r9 movq 0x8(%rsp), %r14 cmpl %eax, %r12d jge 0x29d312 vmovaps -0x20(%r15), %xmm0 vmovaps 0x70(%rsp), %xmm1 vfmadd213ps 0x60(%rsp), %xmm1, %xmm0 # xmm0 = (xmm1 * xmm0) + mem vmovaps 0x10(%rsp), %xmm1 vfmadd231ps -0x10(%r15), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0xd0(%rsp), %xmm1 vfmadd231ps (%r15), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0xe0(%rsp), %xmm1 vfmadd231ps -0x20(%rbx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vfmadd231ps -0x10(%rbx), %xmm15, %xmm0 # xmm0 = (xmm15 * mem) + xmm0 vmovaps 0x170(%rsp), %xmm1 vfmadd231ps (%rbx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0x30(%rsp), %xmm1 vfmadd231ps -0x20(%rdx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0x160(%rsp), %xmm1 vfmadd231ps -0x10(%rdx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0x150(%rsp), %xmm1 vfmadd231ps (%rdx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps %xmm0, (%r9) addq $0x10, %r9 incl %r12d addq $0x10, %r15 addq $0x10, %rbx addq $0x10, %rdx jmp 0x29d273 incl %r10d movq %r15, %r13 movq %rbx, %rbp movq %rdx, %r11 jmp 0x29cc2b incq %rdi movq 0x28(%rsp), %r13 movq 0x1d0(%rsp), %rsi jmp 0x29cb23 cmpl $0x3, 0xd4(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x3, 0xd8(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x1, 0xdc(%rdi) movq %r9, 0x30(%rsp) jne 0x29dfac cmpl $0x1, 0xe0(%rdi) jne 0x29dfac cmpl $0x1, 0xe4(%rdi) jne 0x29dfac cmpl $0x1, 0xe8(%rdi) jne 0x29dfac movslq 0xac(%rsp), %rax movl 0x2c(%r13), %edx movl 0x30(%r13), %r8d movl 0xb8(%rsp), %ecx movq 0x28(%r14), %r10 imulq 0x10(%r13), %r11 movq 0x1b0(%rdi), %rsi movslq %edx, %rdi movq 0x80(%rsp), %rbx movq %rbx, 0x170(%rsp) movq 0x90(%rsp), %rbx imulq 0xc0(%rsp), %rbx movq %rbx, 0x160(%rsp) leaq (%rax,%rax), %rbx movq %rbx, 0x150(%rsp) leaq (%rax,%rax,2), %rbx movq %rbx, 0xf0(%rsp) movq %rax, 0xd0(%rsp) addl $0x2, %eax movslq %eax, %rbp xorl %eax, %eax testl %ecx, %ecx cmovlel %eax, %ecx movq %rcx, 0x60(%rsp) shlq $0x2, %rbp movq %rdi, 0x8(%rsp) leaq (,%rdi,4), %rcx movq %rcx, 0x10(%rsp) movq %rsi, 0xe0(%rsp) movq %r11, 0x70(%rsp) cmpq 0x60(%rsp), %rax je 0x29fdf4 testq %rsi, %rsi je 0x29d44e vmovss (%rsi,%rax,4), %xmm0 jmp 0x29d452 vxorps %xmm0, %xmm0, %xmm0 movq %r11, %rsi imulq %rax, %rsi addq %r9, %rsi imulq $0x24, %rax, %r13 movq 0x160(%rsp), %r15 movq %rax, 0x40(%rsp) imulq %rax, %r15 addq 0x170(%rsp), %r15 movq 0xd0(%rsp), %rax leaq (%r15,%rax,4), %rax movq 0x150(%rsp), %rcx leaq (%r15,%rcx,4), %r9 movq 0xf0(%rsp), %rcx leaq (%r15,%rcx,4), %r11 xorl %edi, %edi movq %rsi, %rbx movl %edi, %ecx orl $0x1, %ecx cmpl %r8d, %ecx jge 0x29d5c7 movq 0x8(%rsp), %rcx leaq (%rsi,%rcx,4), %rsi xorl %r12d, %r12d movl %edx, %ecx testl %ecx, %ecx jle 0x29d59c vmovsd 0x4(%r15,%r12), %xmm1 vmovss (%r15,%r12), %xmm2 vmovlhps %xmm1, %xmm2, %xmm2 # xmm2 = xmm2[0],xmm1[0] vshufps $0xd8, %xmm1, %xmm2, %xmm1 # xmm1 = xmm2[0,2],xmm1[1,3] vmovss (%rax,%r12), %xmm2 vinsertps $0x30, %xmm2, %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],xmm2[0] vmovups (%r10,%r13), %xmm3 vmovups 0x10(%r10,%r13), %xmm4 vmovsd 0x4(%rax,%r12), %xmm5 vmovss (%r9,%r12), %xmm6 vinsertps $0x20, %xmm6, %xmm5, %xmm7 # xmm7 = xmm5[0,1],xmm6[0],xmm5[3] vinsertps $0x30, 0x4(%r9,%r12), %xmm7, %xmm7 # xmm7 = xmm7[0,1,2],mem[0] vmulps %xmm7, %xmm4, %xmm7 vhaddps %xmm7, %xmm7, %xmm7 vmulps %xmm1, %xmm3, %xmm1 vhaddps %xmm7, %xmm7, %xmm7 vmovss 0x20(%r10,%r13), %xmm8 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm1, %xmm7, %xmm1 vfmadd231ss 0x8(%r9,%r12), %xmm8, %xmm1 # xmm1 = (xmm8 * mem) + xmm1 vmovlhps %xmm5, %xmm3, %xmm7 # xmm7 = xmm3[0],xmm5[0] vshufps $0xd8, %xmm5, %xmm7, %xmm5 # xmm5 = xmm7[0,2],xmm5[1,3] vinsertps $0x30, %xmm6, %xmm5, %xmm5 # xmm5 = xmm5[0,1,2],xmm6[0] vmovsd 0x4(%r9,%r12), %xmm6 vmovhps (%r11,%r12), %xmm6, %xmm6 # xmm6 = xmm6[0,1],mem[0,1] vmulps %xmm4, %xmm6, %xmm4 vhaddps %xmm4, %xmm4, %xmm4 vmovss %xmm2, %xmm3, %xmm2 # xmm2 = xmm2[0],xmm3[1,2,3] vhaddps %xmm4, %xmm4, %xmm3 vmulps %xmm2, %xmm5, %xmm2 vhaddps %xmm2, %xmm2, %xmm2 vaddss %xmm0, %xmm1, %xmm1 vhaddps %xmm2, %xmm2, %xmm2 vaddss %xmm2, %xmm3, %xmm2 vfmadd231ss 0x8(%r11,%r12), %xmm8, %xmm2 # xmm2 = (xmm8 * mem) + xmm2 vaddss %xmm0, %xmm2, %xmm2 vmovss %xmm1, (%rbx,%r12) vmovss %xmm2, (%rsi,%r12) decl %ecx addq $0x4, %r12 jmp 0x29d4be addq %rbp, %r15 addq %r12, %r15 addq %rbp, %rax addq %r12, %rax addq %rbp, %r9 addq %r12, %r9 addq %rbp, %r11 addq %r12, %r11 addq 0x10(%rsp), %rbx addq %r12, %rbx addl $0x2, %edi addq %r12, %rsi jmp 0x29d4a2 movq 0x70(%rsp), %r11 cmpl %r8d, %edi jge 0x29d665 xorl %ecx, %ecx movl %edx, %esi testl %esi, %esi jle 0x29d646 vmovsd (%r15,%rcx), %xmm1 vinsertps $0x20, 0x8(%r15,%rcx), %xmm1, %xmm1 # xmm1 = xmm1[0,1],mem[0],xmm1[3] vinsertps $0x30, (%rax,%rcx), %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],mem[0] vmovsd 0x4(%rax,%rcx), %xmm2 vinsertf128 $0x1, %xmm2, %ymm1, %ymm2 vmovsd (%r9,%rcx), %xmm3 vinsertf128 $0x1, %xmm3, %ymm1, %ymm1 vshufpd $0x2, %ymm1, %ymm2, %ymm1 # ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[2] vmulps (%r10,%r13), %ymm1, %ymm1 vextractf128 $0x1, %ymm1, %xmm2 vhaddps %xmm1, %xmm2, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vmovss 0x20(%r10,%r13), %xmm2 vfmadd231ss 0x8(%r9,%rcx), %xmm2, %xmm1 # xmm1 = (xmm2 * mem) + xmm1 vaddss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%rbx,%rcx) decl %esi addq $0x4, %rcx jmp 0x29d5d9 addq %rcx, %r15 addq $0x8, %r15 addq %rcx, %rax addq $0x8, %rax addq %rcx, %r9 addq $0x8, %r9 incl %edi addq %rcx, %rbx jmp 0x29d5cc movq 0x40(%rsp), %rax incq %rax movq 0x28(%rsp), %r13 movq 0x30(%rsp), %r9 movq 0xe0(%rsp), %rsi jmp 0x29d437 cmpl $0x5, %r15d jne 0x29ed47 cmpl $0x1, 0xdc(%rdi) jne 0x29ea4a cmpl $0x1, 0xe0(%rdi) jne 0x29ea4a cmpl $0x1, 0xe4(%rdi) jne 0x29ea4a cmpl $0x1, 0xe8(%rdi) jne 0x29ea4a movl 0x2c(%r13), %r8d movl 0x30(%r13), %eax movl %eax, 0x8(%rsp) movl 0xac(%rsp), %eax movl 0xb8(%rsp), %ecx movq 0x1b0(%rdi), %rdx leal 0x10(,%rax,4), %eax movslq %eax, %rdi leal (,%r8,4), %eax cltq xorl %esi, %esi testl %r8d, %r8d cmovlel %esi, %r8d testl %ecx, %ecx cmovlel %esi, %ecx movq %rcx, 0x60(%rsp) shlq $0x2, %rax movq %rax, 0x30(%rsp) shlq $0x2, %rdi movq %rdx, 0x70(%rsp) movq %r8, 0x10(%rsp) cmpq 0x60(%rsp), %rsi je 0x29fdf4 testq %rdx, %rdx je 0x29d73f movq %rsi, %rax shlq $0x4, %rax vmovups (%rdx,%rax), %xmm0 jmp 0x29d743 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r10 imulq %rsi, %r10 movq 0x10(%r13), %rax imulq %rax, %r10 addq (%r13), %r10 movslq 0x2c(%r13), %r9 movq 0x28(%r14), %r11 movslq 0x54(%r14), %rbx imulq %rsi, %rbx imulq 0x38(%r14), %rbx imulq %rax, %r9 addq %r10, %r9 movslq 0xac(%rsp), %rax movq 0xc0(%rsp), %r15 movq %rsi, 0x40(%rsp) imulq %rsi, %r15 movq 0x90(%rsp), %rcx imulq %rcx, %r15 addq 0x80(%rsp), %r15 imulq %rcx, %rax leaq (%r15,%rax), %r12 leaq (%r15,%rax,2), %r13 leaq (%rax,%rax,2), %rbp addq %r15, %rbp leaq (%r15,%rax,4), %rdx leaq (%rax,%rax,4), %rax addq %r15, %rax xorl %esi, %esi movl %esi, %ecx orl $0x1, %ecx cmpl 0x8(%rsp), %ecx jge 0x29dc81 movl %r8d, %ecx xorl %r8d, %r8d subl $0x1, %ecx jb 0x29da6f vmovaps (%r11,%rbx), %xmm1 vmovaps 0x10(%r11,%rbx), %xmm3 vmovaps 0x20(%r11,%rbx), %xmm4 vmovaps 0x30(%r11,%rbx), %xmm5 vmovaps (%r15,%r8), %xmm2 vfmadd213ps %xmm0, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm0 vfmadd231ps 0x10(%r15,%r8), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%r15,%r8), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%r15,%r8), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vmovaps 0x40(%r11,%rbx), %xmm6 vfmadd231ps 0x40(%r15,%r8), %xmm6, %xmm2 # xmm2 = (xmm6 * mem) + xmm2 vmovaps (%r12,%r8), %xmm7 vmovaps 0x10(%r12,%r8), %xmm8 vmovaps 0x20(%r12,%r8), %xmm9 vmovaps 0x30(%r12,%r8), %xmm10 vmovaps 0x40(%r12,%r8), %xmm11 vfmadd213ps %xmm0, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm1) + xmm0 vfmadd231ps %xmm3, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm3) + xmm1 vfmadd231ps %xmm4, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm4) + xmm1 vfmadd231ps %xmm5, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm5) + xmm1 vfmadd231ps %xmm6, %xmm11, %xmm1 # xmm1 = (xmm11 * xmm6) + xmm1 vmovaps 0x50(%r11,%rbx), %xmm3 vmovaps 0x60(%r11,%rbx), %xmm4 vmovaps 0x70(%r11,%rbx), %xmm5 vmovaps 0x80(%r11,%rbx), %xmm6 vmovaps 0x90(%r11,%rbx), %xmm12 vfmadd231ps %xmm7, %xmm3, %xmm2 # xmm2 = (xmm3 * xmm7) + xmm2 vfmadd231ps %xmm8, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm8) + xmm2 vfmadd231ps %xmm9, %xmm5, %xmm2 # xmm2 = (xmm5 * xmm9) + xmm2 vfmadd231ps %xmm10, %xmm6, %xmm2 # xmm2 = (xmm6 * xmm10) + xmm2 vfmadd231ps %xmm11, %xmm12, %xmm2 # xmm2 = (xmm12 * xmm11) + xmm2 vmovaps (%r13,%r8), %xmm7 vmovaps 0x10(%r13,%r8), %xmm8 vmovaps 0x20(%r13,%r8), %xmm9 vmovaps 0x30(%r13,%r8), %xmm10 vmovaps 0x40(%r13,%r8), %xmm11 vfmadd231ps %xmm3, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm3) + xmm1 vfmadd231ps %xmm4, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm4) + xmm1 vfmadd231ps %xmm5, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm5) + xmm1 vfmadd231ps %xmm6, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm6) + xmm1 vfmadd231ps %xmm12, %xmm11, %xmm1 # xmm1 = (xmm11 * xmm12) + xmm1 vmovaps 0xa0(%r11,%rbx), %xmm3 vmovaps 0xb0(%r11,%rbx), %xmm4 vmovaps 0xc0(%r11,%rbx), %xmm5 vmovaps 0xd0(%r11,%rbx), %xmm6 vmovaps 0xe0(%r11,%rbx), %xmm12 vfmadd231ps %xmm7, %xmm3, %xmm2 # xmm2 = (xmm3 * xmm7) + xmm2 vfmadd231ps %xmm8, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm8) + xmm2 vfmadd231ps %xmm9, %xmm5, %xmm2 # xmm2 = (xmm5 * xmm9) + xmm2 vfmadd231ps %xmm10, %xmm6, %xmm2 # xmm2 = (xmm6 * xmm10) + xmm2 vfmadd231ps %xmm11, %xmm12, %xmm2 # xmm2 = (xmm12 * xmm11) + xmm2 vmovaps (%rbp,%r8), %xmm7 vmovaps 0x10(%rbp,%r8), %xmm8 vmovaps 0x20(%rbp,%r8), %xmm9 vmovaps 0x30(%rbp,%r8), %xmm10 vmovaps 0x40(%rbp,%r8), %xmm11 vfmadd231ps %xmm3, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm3) + xmm1 vfmadd231ps %xmm4, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm4) + xmm1 vfmadd231ps %xmm5, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm5) + xmm1 vfmadd231ps %xmm6, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm6) + xmm1 vfmadd231ps %xmm12, %xmm11, %xmm1 # xmm1 = (xmm11 * xmm12) + xmm1 vmovaps 0xf0(%r11,%rbx), %xmm4 vmovaps 0x100(%r11,%rbx), %xmm5 vmovaps 0x110(%r11,%rbx), %xmm6 vmovaps 0x120(%r11,%rbx), %xmm12 vmovaps 0x130(%r11,%rbx), %xmm13 vfmadd231ps %xmm7, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm7) + xmm2 vfmadd231ps %xmm8, %xmm5, %xmm2 # xmm2 = (xmm5 * xmm8) + xmm2 vfmadd231ps %xmm9, %xmm6, %xmm2 # xmm2 = (xmm6 * xmm9) + xmm2 vfmadd231ps %xmm10, %xmm12, %xmm2 # xmm2 = (xmm12 * xmm10) + xmm2 vfmadd231ps %xmm11, %xmm13, %xmm2 # xmm2 = (xmm13 * xmm11) + xmm2 vmovaps (%rdx,%r8), %xmm7 vmovaps 0x10(%rdx,%r8), %xmm8 vmovaps 0x20(%rdx,%r8), %xmm9 vmovaps 0x30(%rdx,%r8), %xmm10 vmovaps 0x40(%rdx,%r8), %xmm3 vfmadd231ps %xmm4, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm4) + xmm1 vfmadd231ps %xmm5, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm5) + xmm1 vfmadd231ps %xmm6, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm6) + xmm1 vfmadd231ps %xmm12, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm12) + xmm1 vfmadd231ps %xmm13, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm13) + xmm1 vmovaps 0x140(%r11,%rbx), %xmm4 vmovaps 0x150(%r11,%rbx), %xmm5 vmovaps 0x160(%r11,%rbx), %xmm6 vmovaps 0x170(%r11,%rbx), %xmm11 vmovaps 0x180(%r11,%rbx), %xmm12 vfmadd231ps %xmm7, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm7) + xmm2 vfmadd231ps %xmm8, %xmm5, %xmm2 # xmm2 = (xmm5 * xmm8) + xmm2 vfmadd231ps %xmm9, %xmm6, %xmm2 # xmm2 = (xmm6 * xmm9) + xmm2 vfmadd231ps %xmm10, %xmm11, %xmm2 # xmm2 = (xmm11 * xmm10) + xmm2 vfmadd231ps (%rax,%r8), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x10(%rax,%r8), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x20(%rax,%r8), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vfmadd231ps 0x30(%rax,%r8), %xmm11, %xmm1 # xmm1 = (xmm11 * mem) + xmm1 vfmadd231ps 0x40(%rax,%r8), %xmm12, %xmm1 # xmm1 = (xmm12 * mem) + xmm1 vfmadd231ps %xmm3, %xmm12, %xmm2 # xmm2 = (xmm12 * xmm3) + xmm2 vmovaps %xmm2, (%r10,%r8) vmovaps %xmm1, (%r9,%r8) addq $0x10, %r8 jmp 0x29d7d5 addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp addq %rdi, %rdx addq %r8, %rdx addq %rdi, %rax addq %r8, %rax movq 0x30(%rsp), %rcx addq %rcx, %r10 addq %r8, %r10 addq %rcx, %r9 addq %r8, %r9 addl $0x2, %esi movq 0x10(%rsp), %r8 jmp 0x29d7c0 movl %r8d, %ecx xorl %eax, %eax subl $0x1, %ecx jb 0x29dc5b vmovaps (%r11,%rbx), %xmm1 vmovaps 0x10(%r11,%rbx), %xmm2 vmovaps 0x20(%r11,%rbx), %xmm3 vmovaps 0x30(%r11,%rbx), %xmm4 vmovaps 0x40(%r11,%rbx), %xmm5 vfmadd132ps (%r15,%rax), %xmm0, %xmm1 # xmm1 = (xmm1 * mem) + xmm0 vfmadd231ps 0x10(%r15,%rax), %xmm2, %xmm1 # xmm1 = (xmm2 * mem) + xmm1 vfmadd231ps 0x20(%r15,%rax), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x30(%r15,%rax), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x40(%r15,%rax), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vmovaps 0x50(%r11,%rbx), %xmm2 vmovaps 0x60(%r11,%rbx), %xmm3 vmovaps 0x70(%r11,%rbx), %xmm4 vmovaps 0x80(%r11,%rbx), %xmm5 vmovaps 0x90(%r11,%rbx), %xmm6 vfmadd132ps (%r12,%rax), %xmm1, %xmm2 # xmm2 = (xmm2 * mem) + xmm1 vfmadd231ps 0x10(%r12,%rax), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%r12,%rax), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%r12,%rax), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vfmadd231ps 0x40(%r12,%rax), %xmm6, %xmm2 # xmm2 = (xmm6 * mem) + xmm2 vmovaps 0xa0(%r11,%rbx), %xmm1 vmovaps 0xb0(%r11,%rbx), %xmm3 vmovaps 0xc0(%r11,%rbx), %xmm4 vmovaps 0xd0(%r11,%rbx), %xmm5 vmovaps 0xe0(%r11,%rbx), %xmm6 vfmadd132ps (%r13,%rax), %xmm2, %xmm1 # xmm1 = (xmm1 * mem) + xmm2 vfmadd231ps 0x10(%r13,%rax), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x20(%r13,%rax), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x30(%r13,%rax), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x40(%r13,%rax), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vmovaps 0xf0(%r11,%rbx), %xmm2 vmovaps 0x100(%r11,%rbx), %xmm3 vmovaps 0x110(%r11,%rbx), %xmm4 vmovaps 0x120(%r11,%rbx), %xmm5 vfmadd132ps (%rbp,%rax), %xmm1, %xmm2 # xmm2 = (xmm2 * mem) + xmm1 vfmadd231ps 0x10(%rbp,%rax), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%rbp,%rax), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%rbp,%rax), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vmovaps 0x130(%r11,%rbx), %xmm1 vfmadd231ps 0x40(%rbp,%rax), %xmm1, %xmm2 # xmm2 = (xmm1 * mem) + xmm2 vmovaps 0x140(%r11,%rbx), %xmm1 vmovaps 0x150(%r11,%rbx), %xmm3 vmovaps 0x160(%r11,%rbx), %xmm4 vmovaps 0x170(%r11,%rbx), %xmm5 vmovaps 0x180(%r11,%rbx), %xmm6 vfmadd132ps (%rdx,%rax), %xmm2, %xmm1 # xmm1 = (xmm1 * mem) + xmm2 vfmadd231ps 0x10(%rdx,%rax), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x20(%rdx,%rax), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x30(%rdx,%rax), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x40(%rdx,%rax), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vmovaps %xmm1, (%r10,%rax) addq $0x10, %rax jmp 0x29dab6 addq %rax, %r15 addq $0x40, %r15 leaq 0x40(%r12,%rax), %r12 addq %rax, %r13 addq $0x40, %r13 addq %rax, %rbp addq $0x40, %rbp addq %rax, %rdx addq $0x40, %rdx incl %esi addq %rax, %r10 cmpl 0x8(%rsp), %esi jl 0x29dab1 movq 0x40(%rsp), %rsi incq %rsi movq 0x28(%rsp), %r13 movq 0x70(%rsp), %rdx jmp 0x29d721 cmpl $0x5, %r15d jne 0x29fa97 cmpl $0x1, 0xdc(%rdi) jne 0x29f77f cmpl $0x1, 0xe0(%rdi) jne 0x29f77f cmpl $0x1, 0xe4(%rdi) jne 0x29f77f cmpl $0x1, 0xe8(%rdi) jne 0x29f77f movl 0x2c(%r13), %esi movl 0x30(%r13), %ecx movl 0xb8(%rsp), %eax xorl %edx, %edx testl %esi, %esi cmovlel %edx, %esi movq 0x1b0(%rdi), %rdi testl %ecx, %ecx cmovlel %edx, %ecx testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x10(%rsp) movl $0x80, %r8d movq %rdi, 0x30(%rsp) cmpq 0x10(%rsp), %rdx je 0x29fdf4 testq %rdi, %rdi je 0x29dd35 movq %rdx, %rax shlq $0x5, %rax vmovups (%rdi,%rax), %ymm0 jmp 0x29dd39 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %rdx, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %rdx, %r11 imulq 0x38(%r14), %r11 movq 0xc0(%rsp), %rbx movq %rdx, 0x8(%rsp) imulq %rdx, %rbx movq 0x90(%rsp), %rax imulq %rax, %rbx addq 0x80(%rsp), %rbx movslq 0xac(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x29df95 movl %esi, %eax xorl %edi, %edi subl $0x1, %eax jb 0x29df6d vmovaps (%r10,%r11), %ymm1 vmovaps 0x20(%r10,%r11), %ymm2 vmovaps 0x40(%r10,%r11), %ymm3 vmovaps 0x60(%r10,%r11), %ymm4 vmovaps 0x80(%r10,%r11), %ymm5 vfmadd132ps (%rbx,%rdi), %ymm0, %ymm1 # ymm1 = (ymm1 * mem) + ymm0 vfmadd231ps 0x20(%rbx,%rdi), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vfmadd231ps 0x40(%rbx,%rdi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x60(%rbx,%rdi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x80(%rbx,%rdi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vmovaps 0xa0(%r10,%r11), %ymm2 vmovaps 0xc0(%r10,%r11), %ymm3 vmovaps 0xe0(%r10,%r11), %ymm4 vmovaps 0x100(%r10,%r11), %ymm5 vmovaps 0x120(%r10,%r11), %ymm6 vfmadd132ps (%r15,%rdi), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r15,%rdi), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r15,%rdi), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x60(%r15,%rdi), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps 0x80(%r15,%rdi), %ymm6, %ymm2 # ymm2 = (ymm6 * mem) + ymm2 vmovaps 0x140(%r10,%r11), %ymm1 vmovaps 0x160(%r10,%r11), %ymm3 vmovaps 0x180(%r10,%r11), %ymm4 vmovaps 0x1a0(%r10,%r11), %ymm5 vmovaps 0x1c0(%r10,%r11), %ymm6 vfmadd132ps (%r12,%rdi), %ymm2, %ymm1 # ymm1 = (ymm1 * mem) + ymm2 vfmadd231ps 0x20(%r12,%rdi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x40(%r12,%rdi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x60(%r12,%rdi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vfmadd231ps 0x80(%r12,%rdi), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps 0x1e0(%r10,%r11), %ymm2 vmovaps 0x200(%r10,%r11), %ymm3 vmovaps 0x220(%r10,%r11), %ymm4 vmovaps 0x240(%r10,%r11), %ymm5 vfmadd132ps (%r13,%rdi), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r13,%rdi), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r13,%rdi), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x60(%r13,%rdi), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vmovaps 0x260(%r10,%r11), %ymm1 vfmadd231ps 0x80(%r13,%rdi), %ymm1, %ymm2 # ymm2 = (ymm1 * mem) + ymm2 vmovaps 0x280(%r10,%r11), %ymm1 vmovaps 0x2a0(%r10,%r11), %ymm3 vmovaps 0x2c0(%r10,%r11), %ymm4 vmovaps 0x2e0(%r10,%r11), %ymm5 vmovaps 0x300(%r10,%r11), %ymm6 vfmadd132ps (%rbp,%rdi), %ymm2, %ymm1 # ymm1 = (ymm1 * mem) + ymm2 vfmadd231ps 0x20(%rbp,%rdi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x40(%rbp,%rdi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x60(%rbp,%rdi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vfmadd231ps 0x80(%rbp,%rdi), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps %ymm1, (%r9,%rdi) addq $0x20, %rdi jmp 0x29ddad addq %rdi, %rbx addq %r8, %rbx addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp incl %edx addq %rdi, %r9 jmp 0x29dda1 movq 0x8(%rsp), %rdx incq %rdx movq 0x28(%rsp), %r13 movq 0x30(%rsp), %rdi jmp 0x29dd17 cmpl $0x1, 0xdc(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x1, 0xe0(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x2, 0xe4(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x2, 0xe8(%rdi) movl %eax, %ecx jne 0x29e16e movslq 0xac(%rsp), %rdx movl 0x2c(%r13), %eax movl 0x30(%r13), %r8d movl 0xb8(%rsp), %esi movl %edx, %ecx subl %eax, %ecx addl %ecx, %ecx movq 0x28(%r14), %r10 movq 0x1b0(%rdi), %rdi imulq 0x10(%r13), %r11 movq 0x80(%rsp), %rbx movq %rbx, 0x60(%rsp) movq 0x90(%rsp), %rbx imulq 0xc0(%rsp), %rbx movq %rbx, 0x70(%rsp) movq %rdx, 0x10(%rsp) addq %rdx, %rdx movq %rdx, 0xd0(%rsp) movslq %ecx, %r15 xorl %r12d, %r12d testl %r8d, %r8d cmovlel %r12d, %r8d testl %esi, %esi cmovlel %r12d, %esi movq %rsi, 0x8(%rsp) shlq $0x2, %r15 movq %rdi, 0x40(%rsp) cmpq 0x8(%rsp), %r12 je 0x29fdf4 testq %rdi, %rdi je 0x29e081 vmovss (%rdi,%r12,4), %xmm0 jmp 0x29e085 vxorps %xmm0, %xmm0, %xmm0 movq %r11, %rdi movq %r11, %r13 imulq %r12, %r13 addq %r9, %r13 imulq $0x24, %r12, %rbp movq 0x70(%rsp), %rbx imulq %r12, %rbx addq 0x60(%rsp), %rbx movq 0x10(%rsp), %rcx leaq (%rbx,%rcx,4), %r11 movq 0xd0(%rsp), %rcx leaq (%rbx,%rcx,4), %rsi xorl %r9d, %r9d cmpl %r8d, %r9d je 0x29e154 xorl %edx, %edx movl %eax, %ecx testl %ecx, %ecx jle 0x29e13a vmovsd (%rbx,%rdx), %xmm1 vinsertps $0x20, 0x8(%rbx,%rdx), %xmm1, %xmm1 # xmm1 = xmm1[0,1],mem[0],xmm1[3] vinsertps $0x30, (%r11,%rdx), %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],mem[0] vmovsd 0x4(%r11,%rdx), %xmm2 vinsertf128 $0x1, %xmm2, %ymm1, %ymm2 vmovsd (%rsi,%rdx), %xmm3 vinsertf128 $0x1, %xmm3, %ymm1, %ymm1 vshufpd $0x2, %ymm1, %ymm2, %ymm1 # ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[2] vmulps (%r10,%rbp), %ymm1, %ymm1 vextractf128 $0x1, %ymm1, %xmm2 vhaddps %xmm1, %xmm2, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vmovss 0x20(%r10,%rbp), %xmm2 vfmadd231ss 0x8(%rsi,%rdx), %xmm2, %xmm1 # xmm1 = (xmm2 * mem) + xmm1 vaddss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%r13) addq $0x4, %r13 decl %ecx addq $0x8, %rdx jmp 0x29e0c9 addq %r15, %rbx addq %rdx, %rbx addq %r15, %r11 addq %rdx, %r11 addq %r15, %rsi addq %rdx, %rsi incl %r9d jmp 0x29e0bc incq %r12 movq 0x28(%rsp), %r13 movq 0x30(%rsp), %r9 movq %rdi, %r11 movq 0x40(%rsp), %rdi jmp 0x29e069 cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %r8d movq 0x58(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x29e1b4 xorl %eax, %eax testb $0x3, %dil sete %al testb $0x7, %dil leal 0x1(%rax,%rax,2), %eax cmovel %ebp, %eax movl %eax, 0x10(%rsp) xorl %eax, %eax testb $0x3, %r8b sete %al testb $0x7, %r8b leal 0x1(%rax,%rax,2), %eax cmovel %ebp, %eax jmp 0x29e1bb pushq $0x1 popq %rax movl %eax, 0x10(%rsp) movl %eax, 0x8(%rsp) movq 0x40(%rsp), %r15 movq 0x88(%rsp), %rax vmovaps 0x80(%rsp), %xmm0 vmovaps %xmm0, 0x180(%rsp) movq 0x90(%rsp), %rcx movq %rcx, 0x190(%rsp) movl 0x98(%rsp), %ecx movl %ecx, 0x198(%rsp) movq 0xa0(%rsp), %rcx movq %rcx, 0x1a0(%rsp) vmovups 0xa8(%rsp), %xmm0 vmovups %xmm0, 0x1a8(%rsp) movl 0xb8(%rsp), %ecx movl %ecx, 0x1b8(%rsp) movq 0xc0(%rsp), %rcx movq %rcx, 0x1c0(%rsp) testq %rax, %rax je 0x29e244 lock incl (%rax) cmpl 0x10(%rsp), %r10d jle 0x29e298 movl %r8d, %ebx movl %edi, %ebp movq 0x58(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0x100(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x80(%rsp), %rdi leaq 0x180(%rsp), %rsi movl 0x10(%rsp), %edx vzeroupper callq 0x64e3b movl %ebp, %edi movl %ebx, %r8d movq 0x28(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm0 vmovaps %xmm0, 0x100(%rsp) movq 0x10(%rdx), %rcx movq %rcx, 0x110(%rsp) movl 0x18(%rdx), %ecx movl %ecx, 0x118(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x120(%rsp) vmovups 0x28(%rdx), %xmm0 vmovups %xmm0, 0x128(%rsp) movl 0x38(%rdx), %ecx movl %ecx, 0x138(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x140(%rsp) testq %rax, %rax je 0x29e2fc lock incl (%rax) cmpl %r15d, 0x8(%rsp) jae 0x29e391 movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd0(%r14,%rax), %eax cltd movl 0x8(%rsp), %r9d idivl %r9d movl %eax, %ecx movq 0xe0(%rsp), %rax xorl %edx, %edx divq %r15 movl %edi, %ebp movl %r8d, %ebx movl %r9d, %r8d imulq %rax, %r8 movq 0x58(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0x100(%rsp), %rdi movl 0x70(%rsp), %esi movl 0xd0(%rsp), %edx callq 0x628f2 pushq $-0x64 popq %r12 cmpq $0x0, 0x100(%rsp) je 0x29fd31 movl %ebx, %r8d movl %ebp, %edi movslq 0x138(%rsp), %rax imulq 0x140(%rsp), %rax testq %rax, %rax je 0x29fd31 xorl %r15d, %r15d xorl %r12d, %r12d xorl %ebp, %ebp movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rax cmpq %rax, %rbp jge 0x29e5ff movl %r15d, %eax cltd movl 0x10(%rsp), %esi idivl %esi movl %eax, %ecx movl %edi, %r13d movl %edi, %eax cltd idivl %esi movslq %ecx, %rdx imulq 0x1c0(%rsp), %rdx movq 0x190(%rsp), %rcx imulq %rcx, %rdx addq 0x180(%rsp), %rdx movl 0x198(%rsp), %esi movq 0x1a0(%rsp), %rdi movq %rdx, 0x228(%rsp) andq $0x0, 0x230(%rsp) movq %rcx, 0x238(%rsp) movl %esi, 0x240(%rsp) movq %rdi, 0x248(%rsp) movl %eax, 0x260(%rsp) vmovups 0x1a8(%rsp), %xmm0 movslq 0x1b4(%rsp), %rax movslq 0x1ac(%rsp), %rdx movslq 0x1b0(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x268(%rsp) vmovups %xmm0, 0x250(%rsp) movl %r12d, %eax cltd movl 0x8(%rsp), %esi idivl %esi movl %eax, %ecx movl %r8d, %ebx movl %r8d, %eax cltd idivl %esi movslq %ecx, %rdx imulq 0x140(%rsp), %rdx movq 0x110(%rsp), %rsi imulq %rsi, %rdx addq 0x100(%rsp), %rdx movl 0x118(%rsp), %edi movq 0x120(%rsp), %rcx movq %rdx, 0x1e0(%rsp) andq $0x0, 0x1e8(%rsp) movq %rsi, 0x1f0(%rsp) movl %edi, 0x1f8(%rsp) movq %rcx, 0x200(%rsp) movl %eax, 0x218(%rsp) vmovups 0x128(%rsp), %xmm0 movslq 0x134(%rsp), %rax movslq 0x12c(%rsp), %rdx movslq 0x130(%rsp), %rdi imulq %rdx, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x220(%rsp) vmovups %xmm0, 0x208(%rsp) movq 0x10(%r14), %rax movq (%rax,%rbp,8), %rdi movq 0x58(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 vmovups %ymm0, 0x2a0(%rsp) vmovups %ymm1, 0x2c0(%rsp) movq %rcx, 0x2a8(%rsp) movq (%rdi), %rax leaq 0x228(%rsp), %rsi leaq 0x1e0(%rsp), %rdx leaq 0x2a0(%rsp), %rcx vzeroupper callq 0x38(%rax) movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x29e5b4 lock decl (%rax) jne 0x29e5b4 movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi je 0x29e5ac movq (%rdi), %rax callq 0x18(%rax) jmp 0x29e5b4 movq %rsi, %rdi callq 0x5f3e0 movq 0x230(%rsp), %rax testq %rax, %rax je 0x29e5eb lock decl (%rax) jne 0x29e5eb movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi je 0x29e5e3 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29e5eb movq %rsi, %rdi callq 0x5f3e0 incq %rbp movl %ebx, %r8d addl %ebx, %r12d movl %r13d, %edi addl %r13d, %r15d jmp 0x29e399 movq 0x40(%rsp), %rdx cmpl %edx, 0x8(%rsp) jae 0x29e629 xorl %r12d, %r12d leaq 0x100(%rsp), %rdi movq 0x28(%rsp), %rsi movq 0x58(%rsp), %rcx callq 0x64e3b jmp 0x29fd31 xorl %r12d, %r12d leaq 0x100(%rsp), %rax movq 0x28(%rsp), %rcx cmpq %rcx, %rax je 0x29fd31 movq 0x108(%rsp), %rax testq %rax, %rax je 0x29e652 lock incl (%rax) movq 0x8(%rcx), %rax testq %rax, %rax je 0x29fcd9 lock decl (%rax) jne 0x29fcd9 movq 0x28(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x29fcd1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29fcd9 pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x29ed47 cmpl $0x1, 0xe0(%rdi) jne 0x29ed47 cmpl $0x2, 0xe4(%rdi) jne 0x29ed47 cmpl $0x2, 0xe8(%rdi) jne 0x29ed47 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xac(%rsp), %r8d movl 0xb8(%rsp), %edx subl %eax, %r8d shll $0x3, %r8d movq 0x1b0(%rdi), %rsi movslq %r8d, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x29fdf4 testq %rsi, %rsi je 0x29e718 movq %r8, %r9 shlq $0x4, %r9 vmovups (%rsi,%r9), %xmm0 jmp 0x29e71c vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r8, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq %r14, %r10 movq 0x28(%r14), %r14 movslq 0x54(%r10), %r15 imulq %r8, %r15 imulq 0x38(%r10), %r15 movslq 0xac(%rsp), %rbx movq 0xc0(%rsp), %r10 imulq %r8, %r10 movq 0x90(%rsp), %r11 imulq %r11, %r10 addq 0x80(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx vmovaps (%r14,%r15), %xmm1 vmovaps 0x10(%r14,%r15), %xmm2 vmovaps 0x20(%r14,%r15), %xmm3 vmovaps 0x30(%r14,%r15), %xmm4 vmovaps 0x40(%r14,%r15), %xmm5 vmovaps 0x50(%r14,%r15), %xmm6 vmovaps 0x60(%r14,%r15), %xmm7 vmovaps 0x70(%r14,%r15), %xmm8 vmovaps 0x80(%r14,%r15), %xmm9 xorl %ebp, %ebp cmpl %ecx, %ebp je 0x29ea3d xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r14d cmpl %eax, %r14d jge 0x29e9be vmovaps (%r10,%r15), %xmm12 vfmadd213ps %xmm0, %xmm1, %xmm12 # xmm12 = (xmm1 * xmm12) + xmm0 vfmadd231ps 0x10(%r10,%r15), %xmm2, %xmm12 # xmm12 = (xmm2 * mem) + xmm12 vmovaps 0x20(%r10,%r15), %xmm11 vfmadd231ps %xmm11, %xmm3, %xmm12 # xmm12 = (xmm3 * xmm11) + xmm12 vfmadd231ps (%r11,%r15), %xmm4, %xmm12 # xmm12 = (xmm4 * mem) + xmm12 vfmadd231ps 0x10(%r11,%r15), %xmm5, %xmm12 # xmm12 = (xmm5 * mem) + xmm12 vmovaps 0x20(%r11,%r15), %xmm13 vfmadd231ps %xmm13, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm13) + xmm12 vfmadd231ps (%rbx,%r15), %xmm7, %xmm12 # xmm12 = (xmm7 * mem) + xmm12 vfmadd231ps 0x10(%rbx,%r15), %xmm8, %xmm12 # xmm12 = (xmm8 * mem) + xmm12 vmovaps 0x40(%r10,%r15), %xmm10 vfmadd213ps %xmm0, %xmm1, %xmm11 # xmm11 = (xmm1 * xmm11) + xmm0 vfmadd231ps 0x30(%r10,%r15), %xmm2, %xmm11 # xmm11 = (xmm2 * mem) + xmm11 vmovaps 0x40(%r11,%r15), %xmm14 vfmadd231ps %xmm10, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm10) + xmm11 vfmadd231ps %xmm13, %xmm4, %xmm11 # xmm11 = (xmm4 * xmm13) + xmm11 vmovaps 0x20(%rbx,%r15), %xmm13 vfmadd231ps %xmm13, %xmm9, %xmm12 # xmm12 = (xmm9 * xmm13) + xmm12 vfmadd231ps 0x30(%r11,%r15), %xmm5, %xmm11 # xmm11 = (xmm5 * mem) + xmm11 vfmadd231ps %xmm14, %xmm6, %xmm11 # xmm11 = (xmm6 * xmm14) + xmm11 vfmadd231ps %xmm13, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm13) + xmm11 vmovaps 0x40(%rbx,%r15), %xmm13 vfmadd231ps 0x30(%rbx,%r15), %xmm8, %xmm11 # xmm11 = (xmm8 * mem) + xmm11 vmovaps %xmm12, (%r9) vfmadd231ps %xmm13, %xmm9, %xmm11 # xmm11 = (xmm9 * xmm13) + xmm11 vmovaps 0x60(%r10,%r15), %xmm12 vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vfmadd231ps 0x50(%r10,%r15), %xmm2, %xmm10 # xmm10 = (xmm2 * mem) + xmm10 vfmadd231ps %xmm12, %xmm3, %xmm10 # xmm10 = (xmm3 * xmm12) + xmm10 vfmadd231ps %xmm14, %xmm4, %xmm10 # xmm10 = (xmm4 * xmm14) + xmm10 vmovaps 0x60(%r11,%r15), %xmm14 vfmadd231ps 0x50(%r11,%r15), %xmm5, %xmm10 # xmm10 = (xmm5 * mem) + xmm10 vfmadd231ps %xmm14, %xmm6, %xmm10 # xmm10 = (xmm6 * xmm14) + xmm10 vfmadd231ps %xmm13, %xmm7, %xmm10 # xmm10 = (xmm7 * xmm13) + xmm10 vmovaps 0x60(%rbx,%r15), %xmm13 vfmadd231ps 0x50(%rbx,%r15), %xmm8, %xmm10 # xmm10 = (xmm8 * mem) + xmm10 vmovaps %xmm11, 0x10(%r9) vfmadd213ps %xmm0, %xmm1, %xmm12 # xmm12 = (xmm1 * xmm12) + xmm0 vfmadd231ps 0x70(%r10,%r15), %xmm2, %xmm12 # xmm12 = (xmm2 * mem) + xmm12 vfmadd231ps 0x80(%r10,%r15), %xmm3, %xmm12 # xmm12 = (xmm3 * mem) + xmm12 vfmadd231ps %xmm14, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm14) + xmm12 vfmadd231ps 0x70(%r11,%r15), %xmm5, %xmm12 # xmm12 = (xmm5 * mem) + xmm12 vfmadd231ps 0x80(%r11,%r15), %xmm6, %xmm12 # xmm12 = (xmm6 * mem) + xmm12 vfmadd231ps %xmm13, %xmm7, %xmm12 # xmm12 = (xmm7 * xmm13) + xmm12 vfmadd231ps 0x70(%rbx,%r15), %xmm8, %xmm12 # xmm12 = (xmm8 * mem) + xmm12 vfmadd231ps %xmm13, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm13) + xmm10 vfmadd231ps 0x80(%rbx,%r15), %xmm9, %xmm12 # xmm12 = (xmm9 * mem) + xmm12 vmovaps %xmm10, 0x20(%r9) vmovaps %xmm12, 0x30(%r9) addq $0x40, %r9 addl $0x4, %r12d subq $-0x80, %r15 jmp 0x29e7c6 vmovaps (%r10,%r15), %xmm10 vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vfmadd231ps 0x10(%r10,%r15), %xmm2, %xmm10 # xmm10 = (xmm2 * mem) + xmm10 vmovaps 0x20(%r10,%r15), %xmm11 vfmadd231ps %xmm11, %xmm3, %xmm10 # xmm10 = (xmm3 * xmm11) + xmm10 vfmadd231ps (%r11,%r15), %xmm4, %xmm10 # xmm10 = (xmm4 * mem) + xmm10 vmovaps 0x20(%r11,%r15), %xmm12 vfmadd231ps 0x10(%r11,%r15), %xmm5, %xmm10 # xmm10 = (xmm5 * mem) + xmm10 vfmadd231ps %xmm12, %xmm6, %xmm10 # xmm10 = (xmm6 * xmm12) + xmm10 vfmadd231ps (%rbx,%r15), %xmm7, %xmm10 # xmm10 = (xmm7 * mem) + xmm10 vfmadd231ps 0x10(%rbx,%r15), %xmm8, %xmm10 # xmm10 = (xmm8 * mem) + xmm10 vfmadd213ps %xmm0, %xmm1, %xmm11 # xmm11 = (xmm1 * xmm11) + xmm0 vfmadd231ps 0x30(%r10,%r15), %xmm2, %xmm11 # xmm11 = (xmm2 * mem) + xmm11 vfmadd231ps 0x40(%r10,%r15), %xmm3, %xmm11 # xmm11 = (xmm3 * mem) + xmm11 vfmadd231ps %xmm12, %xmm4, %xmm11 # xmm11 = (xmm4 * xmm12) + xmm11 vfmadd231ps 0x30(%r11,%r15), %xmm5, %xmm11 # xmm11 = (xmm5 * mem) + xmm11 vfmadd231ps 0x40(%r11,%r15), %xmm6, %xmm11 # xmm11 = (xmm6 * mem) + xmm11 vmovaps 0x20(%rbx,%r15), %xmm12 vfmadd231ps %xmm12, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm12) + xmm11 vfmadd231ps 0x30(%rbx,%r15), %xmm8, %xmm11 # xmm11 = (xmm8 * mem) + xmm11 vfmadd231ps %xmm12, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm12) + xmm10 vfmadd231ps 0x40(%rbx,%r15), %xmm9, %xmm11 # xmm11 = (xmm9 * mem) + xmm11 vmovaps %xmm10, (%r9) vmovaps %xmm11, 0x10(%r9) addq $0x20, %r9 addl $0x2, %r12d addq $0x40, %r15 leal 0x1(%r12), %r14d cmpl %eax, %r14d jl 0x29e91e jmp 0x29ea1f vmovaps (%r10,%r15), %xmm10 vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vfmadd231ps 0x10(%r10,%r15), %xmm2, %xmm10 # xmm10 = (xmm2 * mem) + xmm10 vfmadd231ps 0x20(%r10,%r15), %xmm3, %xmm10 # xmm10 = (xmm3 * mem) + xmm10 vfmadd231ps (%r11,%r15), %xmm4, %xmm10 # xmm10 = (xmm4 * mem) + xmm10 vfmadd231ps 0x10(%r11,%r15), %xmm5, %xmm10 # xmm10 = (xmm5 * mem) + xmm10 vfmadd231ps 0x20(%r11,%r15), %xmm6, %xmm10 # xmm10 = (xmm6 * mem) + xmm10 vfmadd231ps (%rbx,%r15), %xmm7, %xmm10 # xmm10 = (xmm7 * mem) + xmm10 vfmadd231ps 0x10(%rbx,%r15), %xmm8, %xmm10 # xmm10 = (xmm8 * mem) + xmm10 vfmadd231ps 0x20(%rbx,%r15), %xmm9, %xmm10 # xmm10 = (xmm9 * mem) + xmm10 vmovaps %xmm10, (%r9) addq $0x10, %r9 incl %r12d addq $0x20, %r15 cmpl %eax, %r12d jl 0x29e9ce addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x29e7b8 incq %r8 movq 0x8(%rsp), %r14 jmp 0x29e6fb pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x29ed47 cmpl $0x1, 0xe0(%rdi) jne 0x29ed47 cmpl $0x2, 0xe4(%rdi) jne 0x29ed47 cmpl $0x2, 0xe8(%rdi) jne 0x29ed47 movl 0x2c(%r13), %edx movl 0x30(%r13), %ecx movl 0xac(%rsp), %eax movl 0xb8(%rsp), %esi subl %edx, %eax shll $0x3, %eax movq 0x1b0(%rdi), %r9 xorl %r15d, %r15d testl %edx, %edx cmovlel %r15d, %edx movl %edx, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %esi, %esi cmovlel %r15d, %esi movq %rsi, 0x10(%rsp) shlq $0x2, %r8 movq %r9, 0x30(%rsp) cmpq 0x10(%rsp), %r15 je 0x29fdf4 testq %r9, %r9 je 0x29eaeb movq %r15, %rax shlq $0x4, %rax vmovups (%r9,%rax), %xmm0 jmp 0x29eaef vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r15, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %r15, %r11 imulq 0x38(%r14), %r11 movq 0xc0(%rsp), %rbx movq %r15, 0x8(%rsp) imulq %r15, %rbx movq 0x90(%rsp), %rax imulq %rax, %rbx addq 0x80(%rsp), %rbx movslq 0xac(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x29ed30 movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x29ed0b vmovaps (%r10,%r11), %xmm1 vmovaps 0x10(%r10,%r11), %xmm2 vmovaps 0x20(%r10,%r11), %xmm3 vmovaps 0x30(%r10,%r11), %xmm4 vmovaps 0x40(%r10,%r11), %xmm5 vfmadd132ps (%rbx,%rsi), %xmm0, %xmm1 # xmm1 = (xmm1 * mem) + xmm0 vfmadd231ps 0x10(%rbx,%rsi), %xmm2, %xmm1 # xmm1 = (xmm2 * mem) + xmm1 vfmadd231ps 0x20(%rbx,%rsi), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x30(%rbx,%rsi), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x40(%rbx,%rsi), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vmovaps 0x50(%r10,%r11), %xmm2 vmovaps 0x60(%r10,%r11), %xmm3 vmovaps 0x70(%r10,%r11), %xmm4 vmovaps 0x80(%r10,%r11), %xmm5 vmovaps 0x90(%r10,%r11), %xmm6 vfmadd132ps (%r15,%rsi), %xmm1, %xmm2 # xmm2 = (xmm2 * mem) + xmm1 vfmadd231ps 0x10(%r15,%rsi), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%r15,%rsi), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%r15,%rsi), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vfmadd231ps 0x40(%r15,%rsi), %xmm6, %xmm2 # xmm2 = (xmm6 * mem) + xmm2 vmovaps 0xa0(%r10,%r11), %xmm1 vmovaps 0xb0(%r10,%r11), %xmm3 vmovaps 0xc0(%r10,%r11), %xmm4 vmovaps 0xd0(%r10,%r11), %xmm5 vmovaps 0xe0(%r10,%r11), %xmm6 vfmadd132ps (%r12,%rsi), %xmm2, %xmm1 # xmm1 = (xmm1 * mem) + xmm2 vfmadd231ps 0x10(%r12,%rsi), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x20(%r12,%rsi), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x30(%r12,%rsi), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x40(%r12,%rsi), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vmovaps 0xf0(%r10,%r11), %xmm2 vmovaps 0x100(%r10,%r11), %xmm3 vmovaps 0x110(%r10,%r11), %xmm4 vmovaps 0x120(%r10,%r11), %xmm5 vmovaps 0x130(%r10,%r11), %xmm6 vfmadd132ps (%r13,%rsi), %xmm1, %xmm2 # xmm2 = (xmm2 * mem) + xmm1 vfmadd231ps 0x10(%r13,%rsi), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%r13,%rsi), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%r13,%rsi), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vfmadd231ps 0x40(%r13,%rsi), %xmm6, %xmm2 # xmm2 = (xmm6 * mem) + xmm2 vmovaps 0x140(%r10,%r11), %xmm1 vmovaps 0x150(%r10,%r11), %xmm3 vmovaps 0x160(%r10,%r11), %xmm4 vmovaps 0x170(%r10,%r11), %xmm5 vmovaps 0x180(%r10,%r11), %xmm6 vfmadd132ps (%rbp,%rsi), %xmm2, %xmm1 # xmm1 = (xmm1 * mem) + xmm2 vfmadd231ps 0x10(%rbp,%rsi), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x20(%rbp,%rsi), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x30(%rbp,%rsi), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x40(%rbp,%rsi), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vmovaps %xmm1, (%r9) addq $0x10, %r9 addq $0x20, %rsi jmp 0x29eb63 addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x29eb57 movq 0x8(%rsp), %r15 incq %r15 movq 0x28(%rsp), %r13 movq 0x30(%rsp), %r9 jmp 0x29eacc imull %eax, %r15d movslq %r15d, %rsi leaq 0x180(%rsp), %rdi leaq 0x100(%rsp), %rdx callq 0x73bbe movq (%r14), %rcx movq -0x18(%rcx), %rdx movl 0x10(%rsp), %r11d imull 0xe0(%r14,%rdx), %r11d movq 0x180(%rsp), %rax movl 0xdc(%r14,%rdx), %esi imull 0xd4(%r14,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%r14,%rdx), %r8d jge 0x29edd9 movslq %esi, %rsi leaq (%rax,%rsi,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%r14,%rdx), %r9d jge 0x29edce movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%r14,%rdx), %edi incq %r9 jmp 0x29edaf addl %r11d, %edi incl %r8d addq %r9, %rsi jmp 0x29ed9b leal (,%r15,4), %ecx movl %ecx, 0xe0(%rsp) movl 0x70(%rsp), %ecx shll $0x2, %ecx movslq %ecx, %rcx movq %rcx, 0x40(%rsp) xorl %ecx, %ecx testl %r15d, %r15d cmovlel %ecx, %r15d movq 0x60(%rsp), %r12 testl %r12d, %r12d cmovlel %ecx, %r12d movq %r12, 0x60(%rsp) shlq $0x2, %r15 leaq 0x15981c(%rip), %rdi # 0x3f8638 vxorps %xmm10, %xmm10, %xmm10 vbroadcastss 0x14f1ea(%rip), %xmm5 # 0x3ee014 vbroadcastss 0x14fe55(%rip), %xmm14 # 0x3eec88 xorl %edx, %edx vbroadcastss 0x15237e(%rip), %xmm12 # 0x3f11bc vbroadcastss 0x152381(%rip), %xmm6 # 0x3f11c8 vbroadcastss 0x152380(%rip), %xmm3 # 0x3f11d0 cmpq 0x60(%rsp), %rdx je 0x29fcb9 movq %rcx, 0x70(%rsp) movslq %ecx, %r9 movq 0x40(%r13), %r10 imulq %rdx, %r10 imulq 0x10(%r13), %r10 shlq $0x2, %r9 addq (%r13), %r10 movslq 0xac(%rsp), %r11 movq 0xc0(%rsp), %r13 imulq %rdx, %r13 movq 0x90(%rsp), %rcx imulq %rcx, %r13 addq 0x80(%rsp), %r13 imulq %rcx, %r11 movq %rdx, 0xd0(%rsp) shlq $0x4, %rdx movq %rdx, 0x10(%rsp) addq 0x28(%r14), %r9 xorl %ecx, %ecx cmpl 0x30(%rsp), %ecx jg 0x29f35b movq (%r14), %rsi xorl %r8d, %r8d cmpq %rbx, %r8 jg 0x29f34b movq -0x18(%rsi), %r12 cmpl $0x0, 0x100(%r14,%r12) je 0x29eef7 movq 0x1b0(%r14,%r12), %rdx movq 0x10(%rsp), %rbp vmovups (%rdx,%rbp), %xmm8 jmp 0x29eefc vxorps %xmm8, %xmm8, %xmm8 movslq 0xe8(%r14,%r12), %rdx movq %r14, %rbp movslq %ecx, %r14 imulq %rdx, %r14 imulq %r11, %r14 addq %r13, %r14 movl 0xe4(%rbp,%r12), %edx imull %r8d, %edx shll $0x2, %edx movslq %edx, %rdx leaq (%r14,%rdx,4), %rdx xorl %r14d, %r14d cmpq %r14, %r15 je 0x29ef4d movslq (%rax,%r14), %rbp shlq $0x4, %rbp vmovups (%r9,%r14,4), %xmm1 vfmadd231ps (%rdx,%rbp), %xmm1, %xmm8 # xmm8 = (xmm1 * mem) + xmm8 addq $0x4, %r14 jmp 0x29ef2e movq 0x8(%rsp), %r14 movl 0x110(%r14,%r12), %edx decl %edx cmpl $0x5, %edx ja 0x29f332 movslq (%rdi,%rdx,4), %rdx addq %rdi, %rdx vmaxps %xmm10, %xmm8, %xmm1 jmpq %rdx movq 0x118(%r14,%r12), %rdx vminps %xmm10, %xmm8, %xmm8 vbroadcastss (%rdx), %xmm15 vfmadd231ps %xmm15, %xmm8, %xmm1 # xmm1 = (xmm8 xmm15) + xmm1 jmp 0x29f336 vbroadcastss 0x152220(%rip), %xmm11 # 0x3f11b8 vminps %xmm11, %xmm8, %xmm1 vmaxps %xmm1, %xmm12, %xmm1 vbroadcastss 0x152216(%rip), %xmm13 # 0x3f11c0 vmovaps %xmm13, %xmm10 vfmadd213ps %xmm5, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm5 vcvttps2dq %xmm10, %xmm15 vcvtdq2ps %xmm15, %xmm15 vcmpltps %xmm15, %xmm10, %xmm10 vandps %xmm14, %xmm10, %xmm10 vsubps %xmm10, %xmm15, %xmm10 vbroadcastss 0x1521ed(%rip), %xmm11 # 0x3f11c4 vfmsub231ps %xmm11, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm11) - xmm1 vfnmsub231ps %xmm6, %xmm10, %xmm1 # xmm1 = -(xmm10 * xmm6) - xmm1 vmulps %xmm1, %xmm1, %xmm15 vbroadcastss 0x1521de(%rip), %xmm2 # 0x3f11cc vfmadd213ps %xmm3, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm3 vbroadcastss 0x1521d8(%rip), %xmm7 # 0x3f11d4 vfmadd213ps %xmm7, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm7 vbroadcastss 0x1521ce(%rip), %xmm7 # 0x3f11d8 vfmadd213ps %xmm7, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm7 vbroadcastss 0x1521c4(%rip), %xmm4 # 0x3f11dc vfmadd213ps %xmm4, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm4 vfmadd213ps %xmm5, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm5 vfmadd213ps %xmm1, %xmm15, %xmm2 # xmm2 = (xmm15 * xmm2) + xmm1 vaddps %xmm2, %xmm14, %xmm2 vcvttps2dq %xmm10, %xmm1 vpslld $0x17, %xmm1, %xmm1 vpaddd %xmm1, %xmm14, %xmm1 vfmadd213ps %xmm14, %xmm2, %xmm1 # xmm1 = (xmm2 * xmm1) + xmm14 vbroadcastss 0x152199(%rip), %xmm2 # 0x3f11e0 vmaxps %xmm2, %xmm1, %xmm2 vpsrld $0x17, %xmm2, %xmm10 vbroadcastss 0x15218f(%rip), %xmm4 # 0x3f11e8 vpaddd %xmm4, %xmm10, %xmm10 vbroadcastss 0x15217e(%rip), %xmm4 # 0x3f11e4 vandps %xmm4, %xmm2, %xmm2 vorps %xmm5, %xmm2, %xmm2 vcvtdq2ps %xmm10, %xmm10 vbroadcastss 0x152170(%rip), %xmm4 # 0x3f11ec vcmpltps %xmm4, %xmm2, %xmm15 vandps %xmm2, %xmm15, %xmm4 vbroadcastss 0x152162(%rip), %xmm9 # 0x3f11f0 vaddps %xmm2, %xmm9, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vandps %xmm14, %xmm15, %xmm4 vsubps %xmm4, %xmm10, %xmm4 vmulps %xmm2, %xmm2, %xmm10 vbroadcastss 0x152148(%rip), %xmm15 # 0x3f11f4 vbroadcastss 0x152143(%rip), %xmm9 # 0x3f11f8 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x152139(%rip), %xmm9 # 0x3f11fc vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x15212f(%rip), %xmm9 # 0x3f1200 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x152125(%rip), %xmm9 # 0x3f1204 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x15211b(%rip), %xmm9 # 0x3f1208 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x152111(%rip), %xmm9 # 0x3f120c vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x152107(%rip), %xmm9 # 0x3f1210 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x1520fd(%rip), %xmm9 # 0x3f1214 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vmulps %xmm2, %xmm10, %xmm9 vmulps %xmm15, %xmm9, %xmm9 vfmadd231ps %xmm6, %xmm4, %xmm9 # xmm9 = (xmm4 * xmm6) + xmm9 vfmsub231ps %xmm10, %xmm5, %xmm9 # xmm9 = (xmm5 * xmm10) - xmm9 vxorps %xmm10, %xmm10, %xmm10 vcmpleps %xmm10, %xmm1, %xmm1 vsubps %xmm2, %xmm9, %xmm2 vfnmadd231ps %xmm4, %xmm11, %xmm2 # xmm2 = -(xmm11 * xmm4) + xmm2 vaddps %xmm2, %xmm2, %xmm2 vbroadcastss 0x1527b0(%rip), %xmm4 # 0x3f1900 vblendvps %xmm1, %xmm4, %xmm2, %xmm1 vbroadcastss 0x152059(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm1, %xmm1 vmaxps %xmm1, %xmm12, %xmm1 vfmadd213ps %xmm5, %xmm1, %xmm13 # xmm13 = (xmm1 * xmm13) + xmm5 vcvttps2dq %xmm13, %xmm4 vcvtdq2ps %xmm4, %xmm4 vcmpltps %xmm4, %xmm13, %xmm2 vandps %xmm2, %xmm14, %xmm2 vsubps %xmm2, %xmm4, %xmm2 vfmsub231ps %xmm11, %xmm2, %xmm1 # xmm1 = (xmm2 * xmm11) - xmm1 vfnmsub231ps %xmm6, %xmm2, %xmm1 # xmm1 = -(xmm2 * xmm6) - xmm1 vmulps %xmm1, %xmm1, %xmm4 vbroadcastss 0x152033(%rip), %xmm9 # 0x3f11cc vfmadd213ps %xmm3, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm3 vbroadcastss 0x15202d(%rip), %xmm0 # 0x3f11d4 vfmadd213ps %xmm0, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm0 vfmadd213ps %xmm7, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm7 vbroadcastss 0x152022(%rip), %xmm0 # 0x3f11dc vfmadd213ps %xmm0, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm0 vfmadd213ps %xmm5, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm5 vfmadd213ps %xmm1, %xmm4, %xmm9 # xmm9 = (xmm4 * xmm9) + xmm1 vaddps %xmm14, %xmm9, %xmm4 vcvttps2dq %xmm2, %xmm1 vpslld $0x17, %xmm1, %xmm1 vpaddd %xmm1, %xmm14, %xmm1 vfmadd213ps %xmm14, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm1) + xmm14 vrcpps %xmm1, %xmm2 vaddps %xmm2, %xmm2, %xmm4 vbroadcastss 0x153517(%rip), %xmm0 # 0x3f2708 vfmsub213ps %xmm0, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm1) - xmm0 vfnmadd213ps %xmm4, %xmm2, %xmm1 # xmm1 = -(xmm2 * xmm1) + xmm4 vfmsub213ps %xmm8, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm1) - xmm8 jmp 0x29f336 movq 0x118(%r14,%r12), %rdx vbroadcastss (%rdx), %xmm1 vbroadcastss 0x4(%rdx), %xmm15 vmaxps %xmm1, %xmm8, %xmm1 vminps %xmm1, %xmm15, %xmm1 jmp 0x29f336 vbroadcastss 0x151f86(%rip), %xmm1 # 0x3f11b4 vxorps %xmm1, %xmm8, %xmm1 vbroadcastss 0x151f7d(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm1, %xmm1 vbroadcastss 0x151f74(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm1, %xmm1 vbroadcastss 0x151f6b(%rip), %xmm8 # 0x3f11c0 vfmadd213ps %xmm5, %xmm1, %xmm8 # xmm8 = (xmm1 * xmm8) + xmm5 vcvttps2dq %xmm8, %xmm15 vcvtdq2ps %xmm15, %xmm15 vcmpltps %xmm15, %xmm8, %xmm8 vandps %xmm14, %xmm8, %xmm8 vsubps %xmm8, %xmm15, %xmm8 vbroadcastss 0x151f47(%rip), %xmm2 # 0x3f11c4 vfmsub231ps %xmm2, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm2) - xmm1 vbroadcastss 0x153491(%rip), %xmm2 # 0x3f271c vfmsub231ps %xmm2, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm2) - xmm1 vmulps %xmm1, %xmm1, %xmm15 vbroadcastss 0x151f2f(%rip), %xmm10 # 0x3f11cc vbroadcastss 0x151f2a(%rip), %xmm2 # 0x3f11d0 vfmadd213ps %xmm2, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm2 vbroadcastss 0x151f20(%rip), %xmm2 # 0x3f11d4 vfmadd213ps %xmm2, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm2 vbroadcastss 0x151f16(%rip), %xmm0 # 0x3f11d8 vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vbroadcastss 0x151f0c(%rip), %xmm0 # 0x3f11dc vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vfmadd213ps %xmm5, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm5 vfmadd213ps %xmm1, %xmm15, %xmm10 # xmm10 = (xmm15 * xmm10) + xmm1 vaddps %xmm14, %xmm10, %xmm10 vcvttps2dq %xmm8, %xmm1 vpslld $0x17, %xmm1, %xmm1 vpaddd %xmm1, %xmm14, %xmm1 vfmadd213ps %xmm14, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm1) + xmm14 vxorps %xmm10, %xmm10, %xmm10 vrcpps %xmm1, %xmm8 vfmsub213ps %xmm14, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm1) - xmm14 vfnmadd132ps %xmm8, %xmm8, %xmm1 # xmm1 = -(xmm1 * xmm8) + xmm8 jmp 0x29f336 movq 0x118(%r14,%r12), %rdx vbroadcastss (%rdx), %xmm1 vbroadcastss 0x4(%rdx), %xmm2 vfmadd231ps %xmm1, %xmm8, %xmm2 # xmm2 = (xmm8 * xmm1) + xmm2 vmaxps %xmm2, %xmm10, %xmm1 vminps %xmm1, %xmm14, %xmm1 vmulps %xmm1, %xmm8, %xmm1 jmp 0x29f336 vmovaps %xmm8, %xmm1 movq %r8, %rdx shlq $0x4, %rdx vmovups %xmm1, (%r10,%rdx) incq %r8 jmp 0x29eecb movq 0x40(%rsp), %rdx leaq (%r10,%rdx,4), %r10 incl %ecx jmp 0x29eebb movq 0xd0(%rsp), %rdx incq %rdx movq 0x70(%rsp), %rcx addl 0xe0(%rsp), %ecx movq 0x28(%rsp), %r13 jmp 0x29ee50 pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x29fa97 cmpl $0x1, 0xe0(%rdi) jne 0x29fa97 cmpl $0x2, 0xe4(%rdi) jne 0x29fa97 cmpl $0x2, 0xe8(%rdi) jne 0x29fa97 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xac(%rsp), %r8d movl 0xb8(%rsp), %edx subl %eax, %r8d shll $0x4, %r8d movq 0x1b0(%rdi), %rsi movslq %r8d, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x29fdf4 testq %rsi, %rsi je 0x29f40c movq %r8, %r9 shlq $0x5, %r9 vmovups (%rsi,%r9), %ymm0 jmp 0x29f410 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r8, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq %r14, %r10 movq 0x28(%r14), %r14 movslq 0x54(%r10), %r15 imulq %r8, %r15 imulq 0x38(%r10), %r15 movslq 0xac(%rsp), %rbx movq 0xc0(%rsp), %r10 imulq %r8, %r10 movq 0x90(%rsp), %r11 imulq %r11, %r10 addq 0x80(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx vmovaps (%r14,%r15), %ymm1 vmovaps 0x20(%r14,%r15), %ymm2 vmovaps 0x40(%r14,%r15), %ymm3 vmovaps 0x60(%r14,%r15), %ymm4 vmovaps 0x80(%r14,%r15), %ymm5 vmovaps 0xa0(%r14,%r15), %ymm6 vmovaps 0xc0(%r14,%r15), %ymm7 vmovaps 0xe0(%r14,%r15), %ymm8 vmovaps 0x100(%r14,%r15), %ymm9 xorl %ebp, %ebp cmpl %ecx, %ebp je 0x29f772 xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r14d cmpl %eax, %r14d jge 0x29f6f3 vmovaps 0x40(%r10,%r15), %ymm11 vmovaps 0x80(%r10,%r15), %ymm12 vmovaps 0xc0(%r10,%r15), %ymm14 vmovaps (%r10,%r15), %ymm10 vfmadd213ps %ymm0, %ymm1, %ymm10 # ymm10 = (ymm1 * ymm10) + ymm0 vfmadd231ps 0x20(%r10,%r15), %ymm2, %ymm10 # ymm10 = (ymm2 * mem) + ymm10 vfmadd231ps %ymm11, %ymm3, %ymm10 # ymm10 = (ymm3 * ymm11) + ymm10 vfmadd213ps %ymm0, %ymm1, %ymm11 # ymm11 = (ymm1 * ymm11) + ymm0 vfmadd231ps 0x60(%r10,%r15), %ymm2, %ymm11 # ymm11 = (ymm2 * mem) + ymm11 vfmadd231ps %ymm12, %ymm3, %ymm11 # ymm11 = (ymm3 * ymm12) + ymm11 vfmadd213ps %ymm0, %ymm1, %ymm12 # ymm12 = (ymm1 * ymm12) + ymm0 vmovaps %ymm14, %ymm13 vfmadd213ps %ymm0, %ymm1, %ymm13 # ymm13 = (ymm1 * ymm13) + ymm0 vfmadd231ps 0xa0(%r10,%r15), %ymm2, %ymm12 # ymm12 = (ymm2 * mem) + ymm12 vfmadd231ps 0xe0(%r10,%r15), %ymm2, %ymm13 # ymm13 = (ymm2 * mem) + ymm13 vfmadd231ps %ymm14, %ymm3, %ymm12 # ymm12 = (ymm3 * ymm14) + ymm12 vfmadd231ps 0x100(%r10,%r15), %ymm3, %ymm13 # ymm13 = (ymm3 * mem) + ymm13 vmovaps 0x40(%r11,%r15), %ymm14 vmovaps 0x80(%r11,%r15), %ymm15 vfmadd231ps (%r11,%r15), %ymm4, %ymm10 # ymm10 = (ymm4 * mem) + ymm10 vfmadd231ps %ymm14, %ymm4, %ymm11 # ymm11 = (ymm4 * ymm14) + ymm11 vfmadd231ps 0x20(%r11,%r15), %ymm5, %ymm10 # ymm10 = (ymm5 * mem) + ymm10 vfmadd231ps %ymm14, %ymm6, %ymm10 # ymm10 = (ymm6 * ymm14) + ymm10 vmovaps 0xc0(%r11,%r15), %ymm14 vfmadd231ps %ymm15, %ymm4, %ymm12 # ymm12 = (ymm4 * ymm15) + ymm12 vfmadd231ps %ymm14, %ymm4, %ymm13 # ymm13 = (ymm4 * ymm14) + ymm13 vfmadd231ps 0x60(%r11,%r15), %ymm5, %ymm11 # ymm11 = (ymm5 * mem) + ymm11 vfmadd231ps 0xa0(%r11,%r15), %ymm5, %ymm12 # ymm12 = (ymm5 * mem) + ymm12 vfmadd231ps 0xe0(%r11,%r15), %ymm5, %ymm13 # ymm13 = (ymm5 * mem) + ymm13 vfmadd231ps %ymm15, %ymm6, %ymm11 # ymm11 = (ymm6 * ymm15) + ymm11 vfmadd231ps %ymm14, %ymm6, %ymm12 # ymm12 = (ymm6 * ymm14) + ymm12 vfmadd231ps 0x100(%r11,%r15), %ymm6, %ymm13 # ymm13 = (ymm6 * mem) + ymm13 vmovaps 0x40(%rbx,%r15), %ymm14 vfmadd231ps (%rbx,%r15), %ymm7, %ymm10 # ymm10 = (ymm7 * mem) + ymm10 vfmadd231ps %ymm14, %ymm7, %ymm11 # ymm11 = (ymm7 * ymm14) + ymm11 vfmadd231ps 0x20(%rbx,%r15), %ymm8, %ymm10 # ymm10 = (ymm8 * mem) + ymm10 vfmadd231ps %ymm14, %ymm9, %ymm10 # ymm10 = (ymm9 * ymm14) + ymm10 vmovaps 0x80(%rbx,%r15), %ymm14 vfmadd231ps %ymm14, %ymm7, %ymm12 # ymm12 = (ymm7 * ymm14) + ymm12 vfmadd231ps 0x60(%rbx,%r15), %ymm8, %ymm11 # ymm11 = (ymm8 * mem) + ymm11 vfmadd231ps %ymm14, %ymm9, %ymm11 # ymm11 = (ymm9 * ymm14) + ymm11 vmovaps 0xc0(%rbx,%r15), %ymm14 vfmadd231ps %ymm14, %ymm7, %ymm13 # ymm13 = (ymm7 * ymm14) + ymm13 vfmadd231ps 0xa0(%rbx,%r15), %ymm8, %ymm12 # ymm12 = (ymm8 * mem) + ymm12 vfmadd231ps 0xe0(%rbx,%r15), %ymm8, %ymm13 # ymm13 = (ymm8 * mem) + ymm13 vfmadd231ps 0x100(%rbx,%r15), %ymm9, %ymm13 # ymm13 = (ymm9 * mem) + ymm13 vfmadd231ps %ymm14, %ymm9, %ymm12 # ymm12 = (ymm9 * ymm14) + ymm12 vmovaps %ymm10, (%r9) vmovaps %ymm11, 0x20(%r9) vmovaps %ymm12, 0x40(%r9) vmovaps %ymm13, 0x60(%r9) subq $-0x80, %r9 addl $0x4, %r12d addq $0x100, %r15 # imm = 0x100 jmp 0x29f4c6 vmovaps (%r10,%r15), %ymm10 vfmadd213ps %ymm0, %ymm1, %ymm10 # ymm10 = (ymm1 * ymm10) + ymm0 vfmadd231ps 0x20(%r10,%r15), %ymm2, %ymm10 # ymm10 = (ymm2 * mem) + ymm10 vmovaps 0x40(%r10,%r15), %ymm11 vfmadd231ps %ymm11, %ymm3, %ymm10 # ymm10 = (ymm3 * ymm11) + ymm10 vfmadd213ps %ymm0, %ymm1, %ymm11 # ymm11 = (ymm1 * ymm11) + ymm0 vfmadd231ps 0x60(%r10,%r15), %ymm2, %ymm11 # ymm11 = (ymm2 * mem) + ymm11 vfmadd231ps 0x80(%r10,%r15), %ymm3, %ymm11 # ymm11 = (ymm3 * mem) + ymm11 vmovaps 0x40(%r11,%r15), %ymm12 vfmadd231ps (%r11,%r15), %ymm4, %ymm10 # ymm10 = (ymm4 * mem) + ymm10 vfmadd231ps %ymm12, %ymm4, %ymm11 # ymm11 = (ymm4 * ymm12) + ymm11 vfmadd231ps 0x20(%r11,%r15), %ymm5, %ymm10 # ymm10 = (ymm5 * mem) + ymm10 vfmadd231ps 0x60(%r11,%r15), %ymm5, %ymm11 # ymm11 = (ymm5 * mem) + ymm11 vfmadd231ps %ymm12, %ymm6, %ymm10 # ymm10 = (ymm6 * ymm12) + ymm10 vfmadd231ps 0x80(%r11,%r15), %ymm6, %ymm11 # ymm11 = (ymm6 * mem) + ymm11 vmovaps 0x40(%rbx,%r15), %ymm12 vfmadd231ps (%rbx,%r15), %ymm7, %ymm10 # ymm10 = (ymm7 * mem) + ymm10 vfmadd231ps %ymm12, %ymm7, %ymm11 # ymm11 = (ymm7 * ymm12) + ymm11 vfmadd231ps 0x20(%rbx,%r15), %ymm8, %ymm10 # ymm10 = (ymm8 * mem) + ymm10 vfmadd231ps 0x60(%rbx,%r15), %ymm8, %ymm11 # ymm11 = (ymm8 * mem) + ymm11 vfmadd231ps %ymm12, %ymm9, %ymm10 # ymm10 = (ymm9 * ymm12) + ymm10 vfmadd231ps 0x80(%rbx,%r15), %ymm9, %ymm11 # ymm11 = (ymm9 * mem) + ymm11 vmovaps %ymm10, (%r9) vmovaps %ymm11, 0x20(%r9) addq $0x40, %r9 addl $0x2, %r12d subq $-0x80, %r15 leal 0x1(%r12), %r14d cmpl %eax, %r14d jl 0x29f64a jmp 0x29f754 vmovaps (%r10,%r15), %ymm10 vfmadd213ps %ymm0, %ymm1, %ymm10 # ymm10 = (ymm1 * ymm10) + ymm0 vfmadd231ps 0x20(%r10,%r15), %ymm2, %ymm10 # ymm10 = (ymm2 * mem) + ymm10 vfmadd231ps 0x40(%r10,%r15), %ymm3, %ymm10 # ymm10 = (ymm3 * mem) + ymm10 vfmadd231ps (%r11,%r15), %ymm4, %ymm10 # ymm10 = (ymm4 * mem) + ymm10 vfmadd231ps 0x20(%r11,%r15), %ymm5, %ymm10 # ymm10 = (ymm5 * mem) + ymm10 vfmadd231ps 0x40(%r11,%r15), %ymm6, %ymm10 # ymm10 = (ymm6 * mem) + ymm10 vfmadd231ps (%rbx,%r15), %ymm7, %ymm10 # ymm10 = (ymm7 * mem) + ymm10 vfmadd231ps 0x20(%rbx,%r15), %ymm8, %ymm10 # ymm10 = (ymm8 * mem) + ymm10 vfmadd231ps 0x40(%rbx,%r15), %ymm9, %ymm10 # ymm10 = (ymm9 * mem) + ymm10 vmovaps %ymm10, (%r9) addq $0x20, %r9 incl %r12d addq $0x40, %r15 cmpl %eax, %r12d jl 0x29f703 addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x29f4b8 incq %r8 movq 0x8(%rsp), %r14 jmp 0x29f3ef pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x29fa97 cmpl $0x1, 0xe0(%rdi) jne 0x29fa97 cmpl $0x2, 0xe4(%rdi) jne 0x29fa97 cmpl $0x2, 0xe8(%rdi) jne 0x29fa97 movl 0x2c(%r13), %edx movl 0x30(%r13), %ecx movl 0xac(%rsp), %eax movl 0xb8(%rsp), %esi subl %edx, %eax shll $0x4, %eax movq 0x1b0(%rdi), %r9 xorl %r15d, %r15d testl %edx, %edx cmovlel %r15d, %edx movl %edx, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %esi, %esi cmovlel %r15d, %esi movq %rsi, 0x10(%rsp) shlq $0x2, %r8 movq %r9, 0x30(%rsp) cmpq 0x10(%rsp), %r15 je 0x29fdf4 testq %r9, %r9 je 0x29f820 movq %r15, %rax shlq $0x5, %rax vmovups (%r9,%rax), %ymm0 jmp 0x29f824 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r15, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %r15, %r11 imulq 0x38(%r14), %r11 movq 0xc0(%rsp), %rbx movq %r15, 0x8(%rsp) imulq %r15, %rbx movq 0x90(%rsp), %rax imulq %rax, %rbx addq 0x80(%rsp), %rbx movslq 0xac(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x29fa80 movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x29fa5b vmovaps (%r10,%r11), %ymm1 vmovaps 0x20(%r10,%r11), %ymm2 vmovaps 0x40(%r10,%r11), %ymm3 vmovaps 0x60(%r10,%r11), %ymm4 vmovaps 0x80(%r10,%r11), %ymm5 vfmadd132ps (%rbx,%rsi), %ymm0, %ymm1 # ymm1 = (ymm1 * mem) + ymm0 vfmadd231ps 0x20(%rbx,%rsi), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vfmadd231ps 0x40(%rbx,%rsi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x60(%rbx,%rsi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x80(%rbx,%rsi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vmovaps 0xa0(%r10,%r11), %ymm2 vmovaps 0xc0(%r10,%r11), %ymm3 vmovaps 0xe0(%r10,%r11), %ymm4 vmovaps 0x100(%r10,%r11), %ymm5 vmovaps 0x120(%r10,%r11), %ymm6 vfmadd132ps (%r15,%rsi), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r15,%rsi), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r15,%rsi), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x60(%r15,%rsi), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps 0x80(%r15,%rsi), %ymm6, %ymm2 # ymm2 = (ymm6 * mem) + ymm2 vmovaps 0x140(%r10,%r11), %ymm1 vmovaps 0x160(%r10,%r11), %ymm3 vmovaps 0x180(%r10,%r11), %ymm4 vmovaps 0x1a0(%r10,%r11), %ymm5 vmovaps 0x1c0(%r10,%r11), %ymm6 vfmadd132ps (%r12,%rsi), %ymm2, %ymm1 # ymm1 = (ymm1 * mem) + ymm2 vfmadd231ps 0x20(%r12,%rsi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x40(%r12,%rsi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x60(%r12,%rsi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vfmadd231ps 0x80(%r12,%rsi), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps 0x1e0(%r10,%r11), %ymm2 vmovaps 0x200(%r10,%r11), %ymm3 vmovaps 0x220(%r10,%r11), %ymm4 vmovaps 0x240(%r10,%r11), %ymm5 vmovaps 0x260(%r10,%r11), %ymm6 vfmadd132ps (%r13,%rsi), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r13,%rsi), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r13,%rsi), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x60(%r13,%rsi), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps 0x80(%r13,%rsi), %ymm6, %ymm2 # ymm2 = (ymm6 * mem) + ymm2 vmovaps 0x280(%r10,%r11), %ymm1 vmovaps 0x2a0(%r10,%r11), %ymm3 vmovaps 0x2c0(%r10,%r11), %ymm4 vmovaps 0x2e0(%r10,%r11), %ymm5 vmovaps 0x300(%r10,%r11), %ymm6 vfmadd132ps (%rbp,%rsi), %ymm2, %ymm1 # ymm1 = (ymm1 * mem) + ymm2 vfmadd231ps 0x20(%rbp,%rsi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x40(%rbp,%rsi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x60(%rbp,%rsi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vfmadd231ps 0x80(%rbp,%rsi), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps %ymm1, (%r9) addq $0x20, %r9 addq $0x40, %rsi jmp 0x29f898 addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x29f88c movq 0x8(%rsp), %r15 incq %r15 movq 0x28(%rsp), %r13 movq 0x30(%rsp), %r9 jmp 0x29f801 imull %eax, %r15d movslq %r15d, %rsi leaq 0x180(%rsp), %rdi leaq 0x100(%rsp), %rdx callq 0x73bbe movq (%r14), %rcx movq -0x18(%rcx), %rdx movl 0x10(%rsp), %r11d imull 0xe0(%r14,%rdx), %r11d movq 0x180(%rsp), %rax movl 0xdc(%r14,%rdx), %esi imull 0xd4(%r14,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%r14,%rdx), %esi jge 0x29fb28 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%r14,%rdx), %r9d jge 0x29fb1e movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%r14,%rdx), %edi incq %r9 jmp 0x29faff addl %r11d, %edi incl %esi addq %r9, %r8 jmp 0x29faeb leal (,%r15,8), %ecx movl %ecx, 0xd0(%rsp) movl 0x70(%rsp), %ecx shll $0x3, %ecx movslq %ecx, %rcx movq %rcx, 0x10(%rsp) xorl %ecx, %ecx testl %r15d, %r15d cmovlel %ecx, %r15d movq 0x60(%rsp), %r12 testl %r12d, %r12d cmovlel %ecx, %r12d movq %r12, 0x60(%rsp) shlq $0x2, %r15 xorl %edx, %edx cmpq 0x60(%rsp), %rdx je 0x29fc9f movq %rcx, 0x40(%rsp) movslq %ecx, %r8 movq 0x40(%r13), %r9 imulq %rdx, %r9 imulq 0x10(%r13), %r9 shlq $0x2, %r8 addq (%r13), %r9 movslq 0xac(%rsp), %r10 movq 0xc0(%rsp), %r11 imulq %rdx, %r11 movq 0x90(%rsp), %rcx imulq %rcx, %r11 addq 0x80(%rsp), %r11 imulq %rcx, %r10 movq %rdx, 0x70(%rsp) movq %rdx, %r13 shlq $0x5, %r13 addq 0x28(%r14), %r8 xorl %ebp, %ebp cmpl 0x30(%rsp), %ebp jg 0x29fc81 movq (%r14), %rcx xorl %r12d, %r12d cmpq %rbx, %r12 jg 0x29fc71 movq -0x18(%rcx), %rsi cmpl $0x0, 0x100(%r14,%rsi) je 0x29fc04 movq 0x1b0(%r14,%rsi), %rdi vmovups (%rdi,%r13), %ymm0 jmp 0x29fc08 vxorps %xmm0, %xmm0, %xmm0 movslq 0xe8(%r14,%rsi), %rdi movq %r14, %rdx movslq %ebp, %r14 imulq %rdi, %r14 imulq %r10, %r14 addq %r11, %r14 movl 0xe4(%rdx,%rsi), %esi imull %r12d, %esi shll $0x3, %esi movslq %esi, %rsi leaq (%r14,%rsi,4), %rsi xorl %edi, %edi cmpq %rdi, %r15 je 0x29fc57 movslq (%rax,%rdi), %r14 shlq $0x5, %r14 vmovups (%rsi,%r14), %ymm1 vfmadd231ps (%r8,%rdi,8), %ymm1, %ymm0 # ymm0 = (ymm1 * mem) + ymm0 addq $0x4, %rdi jmp 0x29fc38 movq %r12, %rsi shlq $0x5, %rsi vmovups %ymm0, (%r9,%rsi) incq %r12 movq 0x8(%rsp), %r14 jmp 0x29fbdc movq 0x10(%rsp), %rcx leaq (%r9,%rcx,4), %r9 incl %ebp jmp 0x29fbcc movq 0x70(%rsp), %rdx incq %rdx movq 0x40(%rsp), %rcx addl 0xd0(%rsp), %ecx movq 0x28(%rsp), %r13 jmp 0x29fb66 movq 0x8(%r14), %rdi testq %rdi, %rdi je 0x29fcb9 movq (%rdi), %rax movq %r13, %rsi movq 0x58(%rsp), %rdx vzeroupper callq 0x48(%rax) leaq 0x180(%rsp), %rdi vzeroupper callq 0x624be xorl %r12d, %r12d jmp 0x29fd9f movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x100(%rsp), %xmm0 movq 0x28(%rsp), %rcx vmovups %xmm0, (%rcx) movq 0x110(%rsp), %rax movq %rax, 0x10(%rcx) movl 0x118(%rsp), %eax movl %eax, 0x18(%rcx) movq 0x120(%rsp), %rax movq %rax, 0x20(%rcx) vmovups 0x128(%rsp), %xmm0 vmovups %xmm0, 0x28(%rcx) movl 0x138(%rsp), %eax movl %eax, 0x38(%rcx) movq 0x140(%rsp), %rax movq %rax, 0x40(%rcx) movq 0x108(%rsp), %rax testq %rax, %rax je 0x29fd68 lock decl (%rax) jne 0x29fd68 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x29fd60 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29fd68 movq %rsi, %rdi callq 0x5f3e0 movq 0x188(%rsp), %rax testq %rax, %rax je 0x29fd9f lock decl (%rax) jne 0x29fd9f movq 0x180(%rsp), %rsi movq 0x1a0(%rsp), %rdi testq %rdi, %rdi je 0x29fd97 movq (%rdi), %rax callq 0x18(%rax) jmp 0x29fd9f movq %rsi, %rdi callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax je 0x29fddc lock decl (%rax) jne 0x29fddc movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x29fdd1 movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x29fddc movq %rsi, %rdi vzeroupper callq 0x5f3e0 movl %r12d, %eax addq $0x2e8, %rsp # imm = 0x2E8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movq 0x8(%r14), %rdi testq %rdi, %rdi je 0x29fcc9 movq (%rdi), %rax movq %r13, %rsi movq 0x58(%rsp), %rdx vzeroupper callq 0x48(%rax) jmp 0x29fcc9 movq %rax, %rbx leaq 0x180(%rsp), %rdi callq 0x624be jmp 0x29ff42 jmp 0x29fe5b jmp 0x29fe5b jmp 0x29ff89 jmp 0x29ff89 movq %rax, %rbx jmp 0x29ff0b movq %rax, %rbx jmp 0x29fed4 jmp 0x29ff89 jmp 0x29fe5b jmp 0x29ff89 jmp 0x29ff89 movq %rax, %rbx jmp 0x29ff42 movq %rax, %rbx movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x29fe9d lock decl (%rax) jne 0x29fe9d movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi jne 0x29fe97 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29fe9d movq (%rdi), %rax callq 0x18(%rax) movq 0x230(%rsp), %rax testq %rax, %rax je 0x29fed4 lock decl (%rax) jne 0x29fed4 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi jne 0x29fece movq %rsi, %rdi callq 0x5f3e0 jmp 0x29fed4 movq (%rdi), %rax callq 0x18(%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x29ff0b lock decl (%rax) jne 0x29ff0b movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi jne 0x29ff05 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29ff0b movq (%rdi), %rax callq 0x18(%rax) movq 0x188(%rsp), %rax testq %rax, %rax je 0x29ff42 lock decl (%rax) jne 0x29ff42 movq 0x180(%rsp), %rsi movq 0x1a0(%rsp), %rdi testq %rdi, %rdi jne 0x29ff3c movq %rsi, %rdi callq 0x5f3e0 jmp 0x29ff42 movq (%rdi), %rax callq 0x18(%rax) movq 0x88(%rsp), %rax testq %rax, %rax je 0x29ff79 lock decl (%rax) jne 0x29ff79 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi jne 0x29ff73 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29ff79 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x29ff89 jmp 0x29ff89 jmp 0x29ff89 jmp 0x29ff89 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
ncnn::ConvolutionDepthWise_x86_fma::forward_int8_x86(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_fma::forward_int8_x86(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; int elempack = bottom_blob.elempack; int elembits = bottom_blob.elembits(); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_int8 = bottom_blob; if (elembits != 8) { const int channels_g = channels * elempack / group; Mat scales(channels * elempack); { float* ps = scales; for (int g = 0; g < group; g++) { float scale = bottom_blob_int8_scales[g]; for (int q = 0; q < channels_g; q++) { ps++ = scale; } } } Option opt_q = opt; opt_q.blob_allocator = opt.workspace_allocator; quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q); } Mat bottom_blob_bordered; make_padding(bottom_blob_int8, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; channels = bottom_blob_bordered.c; elempack = bottom_blob_bordered.elempack; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; // depth-wise if (channels elempack == group && group == num_output) { int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { out_elempack = num_output % 8 == 0 ? 8 : 1; } #endif // __SSE2__ bool use_int8_requantize = int8_scale_term > 100; size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #if __SSE2__ if (elempack == 8) { { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128i _sum0 = _mm_setzero_si128(); __m128i _sum1 = _mm_setzero_si128(); const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { // TODO use _mm_cvtepi8_epi16 on sse4.1 __m128i _val = _mm_loadl_epi64((const __m128i)(sptr + space_ofs[k] 8)); _val = _mm_unpacklo_epi8(_val, _mm_cmpgt_epi8(_mm_setzero_si128(), _val)); __m128i _w = _mm_loadl_epi64((const __m128i)(kptr + k 8)); _w = _mm_unpacklo_epi8(_w, _mm_cmpgt_epi8(_mm_setzero_si128(), _w)); __m128i _sl = _mm_mullo_epi16(_val, _w); __m128i _sh = _mm_mulhi_epi16(_val, _w); __m128i _s0 = _mm_unpacklo_epi16(_sl, _sh); __m128i _s1 = _mm_unpackhi_epi16(_sl, _sh); _sum0 = _mm_add_epi32(_sum0, _s0); _sum1 = _mm_add_epi32(_sum1, _s1); } __m128 _scale_in0; __m128 _scale_in1; { __m128 _bottom_blob_int8_scales0 = _mm_loadu_ps((const float)bottom_blob_int8_scales + g 8); __m128 _bottom_blob_int8_scales1 = _mm_loadu_ps((const float)bottom_blob_int8_scales + g 8 + 4); __m128 _weight_data_int8_scales0 = _mm_loadu_ps((const float)weight_data_int8_scales + g 8); __m128 _weight_data_int8_scales1 = _mm_loadu_ps((const float)weight_data_int8_scales + g 8 + 4); _scale_in0 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales0, _weight_data_int8_scales0)); _scale_in1 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales1, _weight_data_int8_scales1)); __m128 _m0 = _mm_cmpneq_ps(_weight_data_int8_scales0, _mm_setzero_ps()); __m128 _m1 = _mm_cmpneq_ps(_weight_data_int8_scales1, _mm_setzero_ps()); _scale_in0 = _mm_and_ps(_scale_in0, _m0); _scale_in1 = _mm_and_ps(_scale_in1, _m1); } __m128 _sumfp32_0 = _mm_mul_ps(_mm_cvtepi32_ps(_sum0), _scale_in0); __m128 _sumfp32_1 = _mm_mul_ps(_mm_cvtepi32_ps(_sum1), _scale_in1); if (bias_term) { __m128 _bias0 = _mm_loadu_ps((const float)bias_data + g 8); __m128 _bias1 = _mm_loadu_ps((const float)bias_data + g 8 + 4); _sumfp32_0 = _mm_add_ps(_sumfp32_0, _bias0); _sumfp32_1 = _mm_add_ps(_sumfp32_1, _bias1); } _sumfp32_0 = activation_sse(_sumfp32_0, activation_type, activation_params); _sumfp32_1 = activation_sse(_sumfp32_1, activation_type, activation_params); if (use_int8_requantize) { // requantize and relu __m128 _scale_out0 = _mm_loadu_ps((const float)top_blob_int8_scales + g 8); __m128 _scale_out1 = _mm_loadu_ps((const float)top_blob_int8_scales + g 8 + 4); _sumfp32_0 = _mm_mul_ps(_sumfp32_0, _scale_out0); _sumfp32_1 = _mm_mul_ps(_sumfp32_1, _scale_out1); int64_t _sum8 = float2int8_sse(_sumfp32_0, _sumfp32_1); (int64_t)outptr_s8 = _sum8; outptr_s8 += 8; } else { // dequantize and relu _mm_storeu_ps(outptr_f32, _sumfp32_0); _mm_storeu_ps(outptr_f32 + 4, _sumfp32_1); outptr_f32 += 8; } } } } } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && stride_w == 1 && stride_h == 1 && dilation_w == 1 && dilation_h == 1 && (activation_type == 0 \|\| activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s1_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s1_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2 && (activation_type == 0 \|\| activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s2_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s2_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char)weight_data_tm + maxk g; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr_s8[0] = sums8; outptr_s8 += 1; } else { // dequantize outptr_f32[0] = sumfp32; outptr_f32 += 1; } } } } } } return 0; } bool use_int8_requantize = int8_scale_term > 100; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { if (use_int8_requantize) out_elempack = num_output % 8 == 0 ? 8 : 1; else out_elempack = num_output % 4 == 0 ? 4 : 1; } #endif // __SSE2__ size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { g_elempack = channels_g % 8 == 0 ? 8 : 1; if (use_int8_requantize) out_g_elempack = num_output_g % 8 == 0 ? 8 : 1; else out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x318, %rsp # imm = 0x318 movq %rcx, 0x30(%rsp) movq %rdx, 0x20(%rsp) movq %rsi, %r15 movl 0x38(%rsi), %ecx movl 0x18(%rsi), %r12d movq 0x10(%rsi), %rsi testl %r12d, %r12d je 0x29ffd3 leal (,%rsi,8), %eax cltd idivl %r12d cmpl $0x8, %eax sete %dl jmp 0x29ffd5 xorl %edx, %edx movq (%rdi), %rax movq -0x18(%rax), %r9 movq %rdi, %r8 movl 0xd4(%rdi,%r9), %ebp movl 0xd8(%rdi,%r9), %ebx decl %ebp imull 0xdc(%rdi,%r9), %ebp decl %ebx imull 0xe0(%rdi,%r9), %ebx movq 0x8(%r15), %rdi vmovups (%r15), %xmm0 vmovaps %xmm0, 0x1e0(%rsp) movq %rsi, 0x1f0(%rsp) movl %r12d, 0x1f8(%rsp) movq 0x20(%r15), %rsi movq %rsi, 0x200(%rsp) vmovdqu 0x28(%r15), %xmm0 vmovdqu %xmm0, 0x208(%rsp) movl %ecx, 0x218(%rsp) movq 0x40(%r15), %rsi movq %rsi, 0x220(%rsp) testq %rdi, %rdi je 0x2a0060 lock incl (%rdi) movq (%r8), %rax movq %r8, 0x8(%rsp) testb %dl, %dl je 0x2a0073 movq 0x8(%rsp), %rcx jmp 0x2a0186 imull %ecx, %r12d movq -0x18(%rax), %rax movq 0x8(%rsp), %r13 movl 0x108(%r13,%rax), %eax movl %eax, 0x10(%rsp) leaq 0x70(%rsp), %rdi andq $0x0, 0x40(%rdi) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdi) vmovdqu %xmm0, 0xc(%rdi) vmovdqa %xmm0, 0x20(%rdi) vmovdqu %xmm0, 0x2c(%rdi) xorl %r14d, %r14d pushq $0x4 popq %rdx movl %r12d, %esi xorl %ecx, %ecx callq 0x635fa movl %r12d, %eax cltd idivl 0x10(%rsp) movq 0x70(%rsp), %rcx movq (%r13), %rdx testl %eax, %eax cmovlel %r14d, %eax movq 0x8(%rsp), %r8 movq -0x18(%rdx), %rsi movslq 0x108(%r8,%rsi), %rdi cmpq %rdi, %r14 jge 0x2a010e movq 0x240(%r8,%rsi), %rsi vmovd (%rsi,%r14,4), %xmm0 movl %eax, %esi subl $0x1, %esi jb 0x2a0109 vmovd %xmm0, (%rcx) addq $0x4, %rcx jmp 0x2a00fa incq %r14 jmp 0x2a00d9 movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0x130(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x1e0(%rsp), %rsi leaq 0x70(%rsp), %rdx movq %r15, %rdi vzeroupper callq 0x652e3 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a017e lock decl (%rax) jne 0x2a017e movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a0176 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a017e movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx movq (%rcx), %rax leaq 0x70(%rsp), %rdx andq $0x0, 0x40(%rdx) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdx) vmovdqu %xmm0, 0xc(%rdx) vmovdqa %xmm0, 0x20(%rdx) vmovdqu %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %rcx, %rdi leaq 0x1e0(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2a1ab2 movslq 0xa8(%rsp), %r14 movq 0xb0(%rsp), %rax imulq %r14, %rax testq %rax, %rax je 0x2a1ab2 notl %ebp movl 0x88(%rsp), %r13d movl 0x9c(%rsp), %edi addl %edi, %ebp movq 0x8(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx movl %ebp, %eax movq %rdx, %rbp cltd idivl 0xe4(%rbp,%rcx) movl %eax, %esi notl %ebx addl 0xa0(%rsp), %ebx movl %ebx, %eax cltd idivl 0xe8(%rbp,%rcx) movl %eax, %edx movq %rsi, 0x10(%rsp) incl %esi movq %rdx, 0x68(%rsp) leal 0x1(%rdx), %r10d movl %r13d, %ebx imull %r14d, %ebx pushq $0x8 popq %r15 cmpl 0x108(%rbp,%rcx), %ebx jne 0x2a039a cmpl 0xd0(%rbp,%rcx), %ebx jne 0x2a039a movl %edi, 0x18(%rsp) movq 0x30(%rsp), %rdi cmpb $0x0, 0x27(%rdi) pushq $0x1 popq %rax cmovel %eax, %r15d testb $0x7, %bl cmovnel %eax, %r15d movl 0x10c(%rbp,%rcx), %eax leal (,%r15,4), %r8d movl %eax, 0x38(%rsp) cmpl $0x65, %eax cmovgel %r15d, %r8d movl %ebx, %eax cltd idivl %r15d movq 0x8(%rdi), %rcx movq %rcx, (%rsp) movq 0x20(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %eax, %ecx movl %r15d, %r9d callq 0x628f2 cmpq $0x0, (%rbx) je 0x2a1ab2 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2a1ab2 xorl %r12d, %r12d cmpl $0x1, %r13d je 0x2a0977 cmpl $0x8, %r13d jne 0x2a1ab2 movq (%rbp), %rax movq -0x18(%rax), %rax movslq 0xd4(%rbp,%rax), %rcx movslq 0xd8(%rbp,%rax), %r15 imulq %rcx, %r15 leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %r15, %rsi callq 0x73bbe movq (%rbp), %rcx movq -0x18(%rcx), %rdx movl 0x18(%rsp), %r11d imull 0xe0(%rbp,%rdx), %r11d movq 0x130(%rsp), %rax movl 0xdc(%rbp,%rdx), %esi imull 0xd4(%rbp,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%rbp,%rdx), %esi jge 0x2a0aa8 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%rbp,%rdx), %r9d jge 0x2a0390 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%rbp,%rdx), %edi incq %r9 jmp 0x2a0372 addl %r11d, %edi incl %esi addq %r9, %r8 jmp 0x2a035b movl 0xd0(%rbp,%rcx), %eax movl 0x10c(%rbp,%rcx), %r14d movq 0x30(%rsp), %rcx cmpb $0x1, 0x27(%rcx) movl %r13d, 0x40(%rsp) movl %r12d, %r13d jne 0x2a03ce cmpl $0x65, %r14d jl 0x2a03d4 testb $0x7, %al pushq $0x1 popq %r9 cmovel %r15d, %r9d jmp 0x2a03e2 pushq $0x1 popq %r9 jmp 0x2a03e2 xorl %ecx, %ecx testb $0x3, %al sete %cl leal (%rcx,%rcx,2), %r9d incl %r9d leal (,%r9,4), %r8d cmpl $0x65, %r14d cmovgel %r9d, %r8d cltd idivl %r9d movq 0x30(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq 0x20(%rsp), %r12 movq %r12, %rdi movl %esi, 0x18(%rsp) movl %r10d, 0x28(%rsp) movl %r10d, %edx movl %eax, %ecx movq %r8, 0x48(%rsp) movq %r9, 0x38(%rsp) callq 0x628f2 cmpq $0x0, (%r12) je 0x2a04a1 movslq 0x38(%r12), %rax imulq 0x40(%r12), %rax testq %rax, %rax movl %r13d, %r12d movl 0x40(%rsp), %r8d je 0x2a1ab2 movq (%rbp), %rax movq -0x18(%rax), %rax movl 0xd0(%rbp,%rax), %ecx movl 0x108(%rbp,%rax), %esi movl %ebx, %eax cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %ebp movq 0x30(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x2a04a9 testb $0x7, %dil pushq $0x1 popq %rbx movl %ebx, %r12d cmovel %r15d, %r12d cmpl $0x65, %r14d movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d jl 0x2a04bb testb $0x7, %bpl cmovel %r15d, %ebx jmp 0x2a04c9 movl %r13d, %r12d jmp 0x2a1ab2 pushq $0x1 popq %r12 movl %r12d, %ebx movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d jmp 0x2a04c9 xorl %eax, %eax testb $0x3, %bpl sete %al leal (%rax,%rax,2), %ebx incl %ebx movq 0x78(%rsp), %rax vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0x130(%rsp) movq 0x80(%rsp), %rcx movq %rcx, 0x140(%rsp) movl 0x88(%rsp), %ecx movl %ecx, 0x148(%rsp) movq 0x90(%rsp), %rcx movq %rcx, 0x150(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0x158(%rsp) movl 0xa8(%rsp), %ecx movl %ecx, 0x168(%rsp) movq 0xb0(%rsp), %rcx movq %rcx, 0x170(%rsp) testq %rax, %rax movq 0x20(%rsp), %r14 je 0x2a0548 lock incl (%rax) cmpl %r12d, %r8d movl %edi, 0x10(%rsp) jle 0x2a059d movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0xe0(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x70(%rsp), %rdi leaq 0x130(%rsp), %rsi movl %r12d, %edx vzeroupper callq 0x64e3b movl 0x10(%rsp), %edi movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d movq 0x8(%r14), %rax vmovups (%r14), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movq 0x10(%r14), %rcx movq %rcx, 0xf0(%rsp) movl 0x18(%r14), %ecx movl %ecx, 0xf8(%rsp) movq 0x20(%r14), %rcx movq %rcx, 0x100(%rsp) vmovups 0x28(%r14), %xmm0 vmovups %xmm0, 0x108(%rsp) movl 0x38(%r14), %ecx movl %ecx, 0x118(%rsp) movq 0x40(%r14), %rcx movq %rcx, 0x120(%rsp) testq %rax, %rax je 0x2a0600 lock incl (%rax) cmpl 0x38(%rsp), %ebx jae 0x2a0680 movq 0x8(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xd0(%rcx,%rax), %eax cltd idivl %ebx movl %eax, %ecx movzbl 0x48(%rsp), %eax divb 0x38(%rsp) movzbl %al, %r8d imull %ebx, %r8d movq 0x30(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0xe0(%rsp), %rdi movl %r9d, %edx movl %ebx, %r9d callq 0x628f2 movl 0x10(%rsp), %edi pushq $-0x64 popq %r14 cmpq $0x0, 0xe0(%rsp) je 0x2a1a41 movslq 0x118(%rsp), %rax imulq 0x120(%rsp), %rax testq %rax, %rax je 0x2a1a41 xorl %r15d, %r15d xorl %r13d, %r13d xorl %r14d, %r14d movq 0x8(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rax movslq 0x108(%r8,%rax), %rax cmpq %rax, %r14 jge 0x2a08f5 movl %r15d, %eax cltd idivl %r12d movl %eax, %ecx movl %edi, %eax cltd idivl %r12d movslq %ecx, %rdx imulq 0x170(%rsp), %rdx movq 0x140(%rsp), %rcx imulq %rcx, %rdx addq 0x130(%rsp), %rdx movl 0x148(%rsp), %esi movq 0x150(%rsp), %rdi movq %rdx, 0x198(%rsp) andq $0x0, 0x1a0(%rsp) movq %rcx, 0x1a8(%rsp) movl %esi, 0x1b0(%rsp) movq %rdi, 0x1b8(%rsp) movl %eax, 0x1d0(%rsp) vmovups 0x158(%rsp), %xmm0 movslq 0x164(%rsp), %rax movslq 0x15c(%rsp), %rdx movslq 0x160(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x1d8(%rsp) vmovups %xmm0, 0x1c0(%rsp) movl %r13d, %eax cltd idivl %ebx movl %eax, %ecx movl %ebp, %eax cltd idivl %ebx movslq %ecx, %rdx imulq 0x120(%rsp), %rdx movq 0xf0(%rsp), %rsi imulq %rsi, %rdx addq 0xe0(%rsp), %rdx movl 0xf8(%rsp), %edi movq 0x100(%rsp), %rcx movq %rdx, 0x228(%rsp) andq $0x0, 0x230(%rsp) movq %rsi, 0x238(%rsp) movl %edi, 0x240(%rsp) movq %rcx, 0x248(%rsp) movl %eax, 0x260(%rsp) vmovups 0x108(%rsp), %xmm0 movslq 0x114(%rsp), %rax movslq 0x10c(%rsp), %rdx movslq 0x110(%rsp), %rdi imulq %rdx, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x268(%rsp) vmovups %xmm0, 0x250(%rsp) movq 0x10(%r8), %rax movq (%rax,%r14,8), %rdi movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 vmovups %ymm0, 0x2d0(%rsp) vmovups %ymm1, 0x2f0(%rsp) movq %rcx, 0x2d8(%rsp) movq (%rdi), %rax leaq 0x198(%rsp), %rsi leaq 0x228(%rsp), %rdx leaq 0x2d0(%rsp), %rcx vzeroupper callq 0x38(%rax) movq 0x230(%rsp), %rax testq %rax, %rax movl 0x10(%rsp), %edi je 0x2a08a8 lock decl (%rax) jne 0x2a08a8 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi je 0x2a089c movq (%rdi), %rax callq 0x18(%rax) movl 0x10(%rsp), %edi jmp 0x2a08a8 movq %rsi, %rdi callq 0x5f3e0 movl 0x10(%rsp), %edi movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x2a08e7 lock decl (%rax) jne 0x2a08e7 movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi je 0x2a08db movq (%rdi), %rax callq 0x18(%rax) movl 0x10(%rsp), %edi jmp 0x2a08e7 movq %rsi, %rdi callq 0x5f3e0 movl 0x10(%rsp), %edi incq %r14 addl %ebp, %r13d addl %edi, %r15d jmp 0x2a0689 movq 0x38(%rsp), %rdx cmpl %edx, %ebx jae 0x2a091d xorl %r14d, %r14d leaq 0xe0(%rsp), %rdi movq 0x20(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x64e3b jmp 0x2a1a41 leaq 0xe0(%rsp), %rax movq 0x20(%rsp), %rbx xorl %r14d, %r14d cmpq %rbx, %rax je 0x2a1a41 movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2a0946 lock incl (%rax) movq 0x8(%rbx), %rax testq %rax, %rax je 0x2a19ee lock decl (%rax) jne 0x2a19ee movq (%rbx), %rsi movq 0x20(%rbx), %rdi testq %rdi, %rdi je 0x2a19e6 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a19ee movq (%rbp), %rax movq -0x18(%rax), %rdx movl 0xd4(%rbp,%rdx), %ecx movl 0xd8(%rbp,%rdx), %r13d movl %ecx, %esi xorl $0x3, %esi movl %r13d, %edi xorl $0x3, %edi orl %esi, %edi movq 0x20(%rsp), %r14 jne 0x2a15a1 cmpl $0x1, 0xe4(%rbp,%rdx) jne 0x2a1496 cmpl $0x1, 0xe8(%rbp,%rdx) jne 0x2a1496 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x2a1496 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x2a1496 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x2a1496 cmpl $0x65, 0x38(%rsp) jl 0x2a1b2f vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx vmovaps %xmm0, (%rbx) andq $0x0, 0x10(%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 leaq 0x198(%rsp), %r12 movq -0x18(%rax), %r13 movslq 0x108(%rbp,%r13), %rax cmpq %rax, %r14 jge 0x2a1e6c movq 0x1f8(%rbp,%r13), %rax vmovss (%rax,%r14,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2a0a66 movq 0x240(%rbp,%r13), %rax vmulss (%rax,%r14,4), %xmm1, %xmm0 vmovss 0x14e226(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rbp,%r13), %rax vmovss (%rax,%r14,4), %xmm0 vmovss %xmm0, 0x198(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c movq %rbx, %rdi movq %r12, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2a0a1f leal (,%r15,8), %ecx movl %ecx, 0x28(%rsp) xorl %edx, %edx testl %r15d, %r15d cmovlel %edx, %r15d testl %r14d, %r14d cmovlel %edx, %r14d vpxor %xmm0, %xmm0, %xmm0 xorl %esi, %esi movq %r14, 0x18(%rsp) cmpq %r14, %rsi je 0x2a19d1 movq 0x20(%rsp), %rcx movq 0x10(%rcx), %r8 imulq %rsi, %r8 imulq 0x40(%rcx), %r8 movq %rdx, 0x30(%rsp) movslq %edx, %r9 addq (%rcx), %r8 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rsi, %r11 movq 0x80(%rsp), %rcx imulq %rcx, %r11 addq 0x70(%rsp), %r11 imulq %rcx, %r10 movq %rsi, 0x40(%rsp) leaq (,%rsi,8), %rbx addq 0x28(%rbp), %r9 xorl %ecx, %ecx movq %r8, %rdx cmpl 0x68(%rsp), %ecx jg 0x2a147b movq (%rbp), %rdi xorl %r13d, %r13d cmpl 0x10(%rsp), %r13d jg 0x2a1474 movq -0x18(%rdi), %rsi movslq 0xe8(%rbp,%rsi), %r14 movslq %ecx, %r12 imulq %r14, %r12 imulq %r10, %r12 movl 0xe4(%rbp,%rsi), %ebp imull %r13d, %ebp shll $0x3, %ebp movslq %ebp, %r14 addq %r11, %r14 addq %r12, %r14 vpxor %xmm3, %xmm3, %xmm3 xorl %r12d, %r12d vpxor %xmm4, %xmm4, %xmm4 cmpq %r12, %r15 je 0x2a0bcc movslq (%rax,%r12,4), %rbp vmovq (%r14,%rbp,8), %xmm5 vpcmpgtb %xmm5, %xmm0, %xmm6 vpunpcklbw %xmm6, %xmm5, %xmm5 # xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7] vmovq (%r9,%r12,8), %xmm6 vpcmpgtb %xmm6, %xmm0, %xmm11 vpunpcklbw %xmm11, %xmm6, %xmm6 # xmm6 = xmm6[0],xmm11[0],xmm6[1],xmm11[1],xmm6[2],xmm11[2],xmm6[3],xmm11[3],xmm6[4],xmm11[4],xmm6[5],xmm11[5],xmm6[6],xmm11[6],xmm6[7],xmm11[7] vpmullw %xmm5, %xmm6, %xmm11 vpmulhw %xmm6, %xmm5, %xmm5 vpunpcklwd %xmm5, %xmm11, %xmm6 # xmm6 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3] vpunpckhwd %xmm5, %xmm11, %xmm5 # xmm5 = xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7] vpaddd %xmm6, %xmm4, %xmm4 vpaddd %xmm5, %xmm3, %xmm3 incq %r12 jmp 0x2a0b89 movq 0x8(%rsp), %rbp movq 0x1f8(%rbp,%rsi), %r14 movq 0x240(%rbp,%rsi), %r12 vmovups (%r14,%rbx,4), %xmm5 vmovups 0x10(%r14,%rbx,4), %xmm6 vmulps (%r12,%rbx,4), %xmm5, %xmm11 vmulps 0x10(%r12,%rbx,4), %xmm6, %xmm13 vrcpps %xmm11, %xmm11 vrcpps %xmm13, %xmm13 vcmpneqps %xmm0, %xmm5, %xmm5 vandps %xmm5, %xmm11, %xmm5 vcmpneqps %xmm0, %xmm6, %xmm6 vandps %xmm6, %xmm13, %xmm6 vcvtdq2ps %xmm4, %xmm4 vmulps %xmm4, %xmm5, %xmm5 vcvtdq2ps %xmm3, %xmm3 vmulps %xmm3, %xmm6, %xmm6 cmpl $0x0, 0x100(%rbp,%rsi) je 0x2a0c46 movq 0x1b0(%rbp,%rsi), %r14 vaddps (%r14,%rbx,4), %xmm5, %xmm5 vaddps 0x10(%r14,%rbx,4), %xmm6, %xmm6 movl 0x110(%rbp,%rsi), %r14d decl %r14d cmpl $0x5, %r14d ja 0x2a13e7 leaq 0x157a06(%rip), %r12 # 0x3f8668 movslq (%r12,%r14,4), %r14 addq %r12, %r14 vmaxps %xmm0, %xmm5, %xmm4 vmaxps %xmm0, %xmm6, %xmm3 jmpq %r14 movq 0x118(%rbp,%rsi), %r14 vminps %xmm0, %xmm5, %xmm5 vbroadcastss (%r14), %xmm11 vfmadd231ps %xmm5, %xmm11, %xmm4 # xmm4 = (xmm11 * xmm5) + xmm4 vminps %xmm0, %xmm6, %xmm5 vfmadd231ps %xmm5, %xmm11, %xmm3 # xmm3 = (xmm11 * xmm5) + xmm3 jmp 0x2a13ef vbroadcastss 0x150517(%rip), %xmm1 # 0x3f11b8 vminps %xmm1, %xmm5, %xmm3 vbroadcastss 0x15050e(%rip), %xmm1 # 0x3f11bc vmaxps %xmm1, %xmm3, %xmm3 vbroadcastss 0x150505(%rip), %xmm15 # 0x3f11c0 vmovaps %xmm15, %xmm4 vbroadcastss 0x14d34c(%rip), %xmm8 # 0x3ee014 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm11 vcvtdq2ps %xmm11, %xmm11 vcmpltps %xmm11, %xmm4, %xmm4 vbroadcastss 0x14dfa3(%rip), %xmm1 # 0x3eec88 vandps %xmm1, %xmm4, %xmm4 vsubps %xmm4, %xmm11, %xmm4 vbroadcastss 0x1504ce(%rip), %xmm7 # 0x3f11c4 vfmsub231ps %xmm7, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm7) - xmm3 vmovaps %xmm7, %xmm9 vbroadcastss 0x1504c0(%rip), %xmm7 # 0x3f11c8 vfnmsub231ps %xmm7, %xmm4, %xmm3 # xmm3 = -(xmm4 * xmm7) - xmm3 vmovaps %xmm7, %xmm2 vmulps %xmm3, %xmm3, %xmm11 vbroadcastss 0x1504ae(%rip), %xmm13 # 0x3f11cc vbroadcastss 0x1504a9(%rip), %xmm7 # 0x3f11d0 vfmadd213ps %xmm7, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm7 vbroadcastss 0x15049f(%rip), %xmm12 # 0x3f11d4 vfmadd213ps %xmm12, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm12 vbroadcastss 0x150495(%rip), %xmm10 # 0x3f11d8 vfmadd213ps %xmm10, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm10 vbroadcastss 0x15048b(%rip), %xmm7 # 0x3f11dc vfmadd213ps %xmm7, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm7 vfmadd213ps %xmm8, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm8 vfmadd213ps %xmm3, %xmm11, %xmm13 # xmm13 = (xmm11 * xmm13) + xmm3 vaddps %xmm1, %xmm13, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm1, %xmm4, %xmm4 vfmadd213ps %xmm1, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm1 vcmpleps %xmm0, %xmm4, %xmm3 vbroadcastss 0x15045c(%rip), %xmm7 # 0x3f11e0 vmaxps %xmm7, %xmm4, %xmm4 vpsrld $0x17, %xmm4, %xmm11 vbroadcastss 0x15044e(%rip), %xmm7 # 0x3f11e4 vandps %xmm7, %xmm4, %xmm4 vorps %xmm4, %xmm8, %xmm4 vbroadcastss 0x150441(%rip), %xmm7 # 0x3f11e8 vpaddd %xmm7, %xmm11, %xmm11 vcvtdq2ps %xmm11, %xmm11 vbroadcastss 0x150433(%rip), %xmm7 # 0x3f11ec vcmpltps %xmm7, %xmm4, %xmm13 vandps %xmm4, %xmm13, %xmm7 vbroadcastss 0x150425(%rip), %xmm12 # 0x3f11f0 vaddps %xmm4, %xmm12, %xmm4 vaddps %xmm7, %xmm4, %xmm4 vandps %xmm1, %xmm13, %xmm7 vsubps %xmm7, %xmm11, %xmm7 vmulps %xmm4, %xmm4, %xmm11 vbroadcastss 0x15040c(%rip), %xmm13 # 0x3f11f4 vbroadcastss 0x150407(%rip), %xmm12 # 0x3f11f8 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503fd(%rip), %xmm12 # 0x3f11fc vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503f3(%rip), %xmm12 # 0x3f1200 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503e9(%rip), %xmm12 # 0x3f1204 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503df(%rip), %xmm12 # 0x3f1208 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503d5(%rip), %xmm12 # 0x3f120c vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503cb(%rip), %xmm12 # 0x3f1210 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503c1(%rip), %xmm12 # 0x3f1214 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vmulps %xmm4, %xmm11, %xmm14 vmulps %xmm13, %xmm14, %xmm13 vmovaps %xmm2, %xmm14 vfmadd231ps %xmm2, %xmm7, %xmm13 # xmm13 = (xmm7 * xmm2) + xmm13 vfmsub231ps %xmm11, %xmm8, %xmm13 # xmm13 = (xmm8 * xmm11) - xmm13 vsubps %xmm4, %xmm13, %xmm4 vfmsub231ps %xmm7, %xmm9, %xmm4 # xmm4 = (xmm9 * xmm7) - xmm4 vbroadcastss 0x151897(%rip), %xmm2 # 0x3f2718 vmulps %xmm2, %xmm4, %xmm4 vbroadcastss 0x150a72(%rip), %xmm2 # 0x3f1900 vblendvps %xmm3, %xmm2, %xmm4, %xmm3 vbroadcastss 0x15031b(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm3, %xmm3 vbroadcastss 0x150312(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm3, %xmm3 vmovaps %xmm15, %xmm4 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm7 vcvtdq2ps %xmm7, %xmm7 vcmpltps %xmm7, %xmm4, %xmm4 vandps %xmm1, %xmm4, %xmm4 vsubps %xmm4, %xmm7, %xmm4 vfmsub231ps %xmm9, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm9) - xmm3 vfnmsub231ps %xmm14, %xmm4, %xmm3 # xmm3 = -(xmm4 * xmm14) - xmm3 vmovaps %xmm14, %xmm13 vmulps %xmm3, %xmm3, %xmm7 vbroadcastss 0x1502e4(%rip), %xmm9 # 0x3f11cc vmovaps %xmm9, %xmm11 vbroadcastss 0x1502da(%rip), %xmm14 # 0x3f11d0 vfmadd213ps %xmm14, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm14 vbroadcastss 0x1502d0(%rip), %xmm10 # 0x3f11d4 vfmadd213ps %xmm10, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm10 vbroadcastss 0x1502c6(%rip), %xmm2 # 0x3f11d8 vfmadd213ps %xmm2, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm2 vbroadcastss 0x1502bc(%rip), %xmm12 # 0x3f11dc vfmadd213ps %xmm12, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm12 vfmadd213ps %xmm8, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm8 vfmadd213ps %xmm3, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm11) + xmm3 vaddps %xmm1, %xmm11, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm1, %xmm4, %xmm4 vfmadd213ps %xmm1, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm1 vrcpps %xmm4, %xmm3 vaddps %xmm3, %xmm3, %xmm7 vbroadcastss 0x1517b2(%rip), %xmm2 # 0x3f2708 vfmsub213ps %xmm2, %xmm7, %xmm4 # xmm4 = (xmm7 * xmm4) - xmm2 vfnmadd213ps %xmm7, %xmm3, %xmm4 # xmm4 = -(xmm3 * xmm4) + xmm7 vbroadcastss 0x15024f(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm6, %xmm3 vbroadcastss 0x150246(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm3, %xmm3 vmovaps %xmm15, %xmm7 vfmadd213ps %xmm8, %xmm3, %xmm7 # xmm7 = (xmm3 * xmm7) + xmm8 vcvttps2dq %xmm7, %xmm11 vcvtdq2ps %xmm11, %xmm11 vcmpltps %xmm11, %xmm7, %xmm7 vandps %xmm1, %xmm7, %xmm7 vsubps %xmm7, %xmm11, %xmm7 vbroadcastss 0x150221(%rip), %xmm2 # 0x3f11c4 vfmsub231ps %xmm2, %xmm7, %xmm3 # xmm3 = (xmm7 * xmm2) - xmm3 vfnmsub231ps %xmm13, %xmm7, %xmm3 # xmm3 = -(xmm7 * xmm13) - xmm3 vmulps %xmm3, %xmm3, %xmm11 vmovaps %xmm9, %xmm13 vfmadd213ps %xmm14, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm14 vfmadd213ps %xmm10, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm10 vbroadcastss 0x15020f(%rip), %xmm10 # 0x3f11d8 vfmadd213ps %xmm10, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm10 vfmadd213ps %xmm12, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm12 vfmadd213ps %xmm8, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm8 vfmadd213ps %xmm3, %xmm11, %xmm13 # xmm13 = (xmm11 * xmm13) + xmm3 vaddps %xmm1, %xmm13, %xmm11 vcvttps2dq %xmm7, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm1, %xmm3, %xmm3 vfmadd213ps %xmm1, %xmm11, %xmm3 # xmm3 = (xmm11 * xmm3) + xmm1 vbroadcastss 0x1501e4(%rip), %xmm2 # 0x3f11e0 vmaxps %xmm2, %xmm3, %xmm7 vpsrld $0x17, %xmm7, %xmm11 vbroadcastss 0x1501d6(%rip), %xmm2 # 0x3f11e4 vandps %xmm2, %xmm7, %xmm7 vorps %xmm7, %xmm8, %xmm7 vbroadcastss 0x1501c9(%rip), %xmm2 # 0x3f11e8 vpaddd %xmm2, %xmm11, %xmm11 vcvtdq2ps %xmm11, %xmm11 vbroadcastss 0x1501bb(%rip), %xmm2 # 0x3f11ec vcmpltps %xmm2, %xmm7, %xmm13 vandps %xmm7, %xmm13, %xmm14 vbroadcastss 0x1501ad(%rip), %xmm2 # 0x3f11f0 vaddps %xmm2, %xmm7, %xmm7 vaddps %xmm7, %xmm14, %xmm7 vandps %xmm1, %xmm13, %xmm13 vsubps %xmm13, %xmm11, %xmm11 vmulps %xmm7, %xmm7, %xmm13 vbroadcastss 0x150193(%rip), %xmm14 # 0x3f11f4 vbroadcastss 0x15018e(%rip), %xmm2 # 0x3f11f8 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150184(%rip), %xmm2 # 0x3f11fc vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x15017a(%rip), %xmm2 # 0x3f1200 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150170(%rip), %xmm2 # 0x3f1204 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150166(%rip), %xmm2 # 0x3f1208 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x15015c(%rip), %xmm2 # 0x3f120c vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150152(%rip), %xmm2 # 0x3f1210 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150148(%rip), %xmm2 # 0x3f1214 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vmovaps %xmm15, %xmm2 vmulps %xmm7, %xmm13, %xmm15 vmulps %xmm14, %xmm15, %xmm14 vmovaps %xmm2, %xmm15 vbroadcastss 0x1500dd(%rip), %xmm9 # 0x3f11c8 vfmadd231ps %xmm9, %xmm11, %xmm14 # xmm14 = (xmm11 * xmm9) + xmm14 vfmsub231ps %xmm13, %xmm8, %xmm14 # xmm14 = (xmm8 * xmm13) - xmm14 vcmpleps %xmm0, %xmm3, %xmm3 vsubps %xmm7, %xmm14, %xmm7 vbroadcastss 0x1500bd(%rip), %xmm12 # 0x3f11c4 vfmsub231ps %xmm11, %xmm12, %xmm7 # xmm7 = (xmm12 * xmm11) - xmm7 vbroadcastss 0x151603(%rip), %xmm2 # 0x3f2718 vmulps %xmm2, %xmm7, %xmm7 vbroadcastss 0x1507de(%rip), %xmm2 # 0x3f1900 vblendvps %xmm3, %xmm2, %xmm7, %xmm3 vbroadcastss 0x150087(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm3, %xmm3 vbroadcastss 0x15007e(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm3, %xmm3 vmovaps %xmm15, %xmm7 vfmadd213ps %xmm8, %xmm3, %xmm7 # xmm7 = (xmm3 * xmm7) + xmm8 vcvttps2dq %xmm7, %xmm11 vcvtdq2ps %xmm11, %xmm11 vcmpltps %xmm11, %xmm7, %xmm7 vandps %xmm1, %xmm7, %xmm7 vsubps %xmm7, %xmm11, %xmm7 vfmsub231ps %xmm12, %xmm7, %xmm3 # xmm3 = (xmm7 * xmm12) - xmm3 vfnmsub231ps %xmm9, %xmm7, %xmm3 # xmm3 = -(xmm7 * xmm9) - xmm3 vmulps %xmm3, %xmm3, %xmm11 vbroadcastss 0x150053(%rip), %xmm13 # 0x3f11cc vbroadcastss 0x15004e(%rip), %xmm2 # 0x3f11d0 vfmadd213ps %xmm2, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm2 vbroadcastss 0x150044(%rip), %xmm2 # 0x3f11d4 vfmadd213ps %xmm2, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm2 vfmadd213ps %xmm10, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm10 vbroadcastss 0x150039(%rip), %xmm2 # 0x3f11dc vfmadd213ps %xmm2, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm2 vfmadd213ps %xmm8, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm8 vfmadd213ps %xmm3, %xmm11, %xmm13 # xmm13 = (xmm11 * xmm13) + xmm3 vaddps %xmm1, %xmm13, %xmm11 vcvttps2dq %xmm7, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm1, %xmm3, %xmm3 vfmadd213ps %xmm1, %xmm11, %xmm3 # xmm3 = (xmm11 * xmm3) + xmm1 vrcpps %xmm3, %xmm7 vaddps %xmm7, %xmm7, %xmm11 vbroadcastss 0x15152f(%rip), %xmm1 # 0x3f2708 vfmsub213ps %xmm1, %xmm11, %xmm3 # xmm3 = (xmm11 * xmm3) - xmm1 vfnmadd213ps %xmm11, %xmm7, %xmm3 # xmm3 = -(xmm7 * xmm3) + xmm11 vfmsub213ps %xmm5, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm5 vfmsub213ps %xmm6, %xmm6, %xmm3 # xmm3 = (xmm6 * xmm3) - xmm6 jmp 0x2a13ef movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm3 vbroadcastss 0x4(%r14), %xmm11 vmaxps %xmm3, %xmm5, %xmm4 vminps %xmm4, %xmm11, %xmm4 vmaxps %xmm3, %xmm6, %xmm3 vminps %xmm3, %xmm11, %xmm3 jmp 0x2a13ef vbroadcastss 0x14ff91(%rip), %xmm1 # 0x3f11b4 vxorps %xmm1, %xmm5, %xmm3 vbroadcastss 0x14ff88(%rip), %xmm9 # 0x3f11b8 vminps %xmm3, %xmm9, %xmm3 vbroadcastss 0x14ff7f(%rip), %xmm10 # 0x3f11bc vmaxps %xmm3, %xmm10, %xmm5 vbroadcastss 0x14ff76(%rip), %xmm8 # 0x3f11c0 vmovaps %xmm8, %xmm3 vbroadcastss 0x14cdbd(%rip), %xmm1 # 0x3ee014 vfmadd213ps %xmm1, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm3) + xmm1 vcvttps2dq %xmm3, %xmm4 vcvtdq2ps %xmm4, %xmm4 vcmpltps %xmm4, %xmm3, %xmm3 vbroadcastss 0x14da16(%rip), %xmm2 # 0x3eec88 vandps %xmm2, %xmm3, %xmm3 vsubps %xmm3, %xmm4, %xmm3 vbroadcastss 0x14ff41(%rip), %xmm7 # 0x3f11c4 vfmsub231ps %xmm7, %xmm3, %xmm5 # xmm5 = (xmm3 * xmm7) - xmm5 vbroadcastss 0x14ff37(%rip), %xmm13 # 0x3f11c8 vfnmsub231ps %xmm13, %xmm3, %xmm5 # xmm5 = -(xmm3 * xmm13) - xmm5 vmulps %xmm5, %xmm5, %xmm11 vbroadcastss 0x14ff29(%rip), %xmm12 # 0x3f11cc vmovaps %xmm12, %xmm4 vbroadcastss 0x14ff20(%rip), %xmm14 # 0x3f11d0 vfmadd213ps %xmm14, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm14 vbroadcastss 0x14ff16(%rip), %xmm15 # 0x3f11d4 vfmadd213ps %xmm15, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm15 vbroadcastss 0x14ff0c(%rip), %xmm7 # 0x3f11d8 vfmadd213ps %xmm7, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm7 vbroadcastss 0x14ff02(%rip), %xmm7 # 0x3f11dc vfmadd213ps %xmm7, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm7 vfmadd213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm1 vfmadd213ps %xmm5, %xmm11, %xmm4 # xmm4 = (xmm11 * xmm4) + xmm5 vbroadcastss 0x14fec2(%rip), %xmm5 # 0x3f11b4 vxorps %xmm5, %xmm6, %xmm5 vminps %xmm5, %xmm9, %xmm5 vmaxps %xmm5, %xmm10, %xmm5 vfmadd213ps %xmm1, %xmm5, %xmm8 # xmm8 = (xmm5 * xmm8) + xmm1 vcvttps2dq %xmm8, %xmm11 vcvtdq2ps %xmm11, %xmm11 vcmpltps %xmm11, %xmm8, %xmm6 vandps %xmm2, %xmm6, %xmm6 vsubps %xmm6, %xmm11, %xmm6 vbroadcastss 0x14fea0(%rip), %xmm8 # 0x3f11c4 vfmsub231ps %xmm8, %xmm6, %xmm5 # xmm5 = (xmm6 * xmm8) - xmm5 vfnmsub231ps %xmm13, %xmm6, %xmm5 # xmm5 = -(xmm6 * xmm13) - xmm5 vmulps %xmm5, %xmm5, %xmm11 vfmadd213ps %xmm14, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm14 vfmadd213ps %xmm15, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm15 vbroadcastss 0x14fe93(%rip), %xmm8 # 0x3f11d8 vfmadd213ps %xmm8, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm8 vfmadd213ps %xmm7, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm7 vfmadd213ps %xmm1, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm1 vfmadd213ps %xmm5, %xmm11, %xmm12 # xmm12 = (xmm11 * xmm12) + xmm5 vaddps %xmm2, %xmm4, %xmm5 vcvttps2dq %xmm3, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm2, %xmm3, %xmm4 vfmadd213ps %xmm2, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm2 vrcpps %xmm4, %xmm3 vfmsub213ps %xmm2, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) - xmm2 vfnmadd132ps %xmm3, %xmm3, %xmm4 # xmm4 = -(xmm4 * xmm3) + xmm3 vaddps %xmm2, %xmm12, %xmm5 vcvttps2dq %xmm6, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm2, %xmm3, %xmm3 vfmadd213ps %xmm2, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm3) + xmm2 vrcpps %xmm3, %xmm5 vfmsub213ps %xmm2, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm3) - xmm2 vfnmadd132ps %xmm5, %xmm5, %xmm3 # xmm3 = -(xmm3 * xmm5) + xmm5 jmp 0x2a13ef movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm3 vbroadcastss 0x4(%r14), %xmm7 vmovaps %xmm5, %xmm4 vfmadd213ps %xmm7, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm7 vmaxps %xmm0, %xmm4, %xmm4 vbroadcastss 0x14d8bc(%rip), %xmm1 # 0x3eec88 vminps %xmm1, %xmm4, %xmm4 vmulps %xmm5, %xmm4, %xmm4 vfmadd213ps %xmm7, %xmm6, %xmm3 # xmm3 = (xmm6 * xmm3) + xmm7 vmaxps %xmm0, %xmm3, %xmm3 vminps %xmm1, %xmm3, %xmm3 vmulps %xmm6, %xmm3, %xmm3 jmp 0x2a13ef vmovaps %xmm5, %xmm4 vmovaps %xmm6, %xmm3 cmpl $0x65, 0x38(%rsp) jl 0x2a145d movq 0x288(%rbp,%rsi), %rsi vmulps (%rsi,%rbx,4), %xmm4, %xmm4 vmulps 0x10(%rsi,%rbx,4), %xmm3, %xmm3 vbroadcastss 0x14fda2(%rip), %xmm1 # 0x3f11b4 vandps %xmm1, %xmm4, %xmm5 vandps %xmm1, %xmm3, %xmm6 vbroadcastss 0x14cbf1(%rip), %xmm1 # 0x3ee014 vorps %xmm1, %xmm5, %xmm5 vorps %xmm1, %xmm6, %xmm6 vaddps %xmm5, %xmm4, %xmm4 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vcvttps2dq %xmm3, %xmm3 vpackssdw %xmm3, %xmm4, %xmm3 vpminsw 0x157109(%rip), %xmm3, %xmm3 # 0x3f8550 vpmaxsw 0x157111(%rip), %xmm3, %xmm3 # 0x3f8560 vpacksswb %xmm3, %xmm3, %xmm3 vmovq %xmm3, (%rdx) addq $0x8, %rdx jmp 0x2a146c vmovups %xmm4, (%r8) vmovups %xmm3, 0x10(%r8) addq $0x20, %r8 incl %r13d jmp 0x2a0b45 incl %ecx jmp 0x2a0b34 movq 0x40(%rsp), %rsi incq %rsi movq 0x30(%rsp), %rdx addl 0x28(%rsp), %edx movq 0x18(%rsp), %r14 jmp 0x2a0acf pushq $0x3 popq %r13 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x2a15a1 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x2a15a1 cmpl $0x2, 0xe4(%rbp,%rdx) jne 0x2a15a1 cmpl $0x2, 0xe8(%rbp,%rdx) jne 0x2a15a1 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x2a15a1 cmpl $0x65, 0x38(%rsp) jl 0x2a20a9 vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %r13 vmovaps %xmm0, (%r13) andq $0x0, 0x10(%r13) xorl %ebx, %ebx movq %rbp, %rcx leaq 0xe0(%rsp), %rbp leaq 0x198(%rsp), %r12 movq -0x18(%rax), %r14 movslq 0x108(%rcx,%r14), %rax cmpq %rax, %rbx jge 0x2a2400 movq 0x1f8(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2a155c movq 0x240(%rcx,%r14), %rax vmulss (%rax,%rbx,4), %xmm1, %xmm0 vmovss 0x14d730(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm0 vmovss %xmm0, 0x198(%rsp) movq %r13, %rdi movq %rbp, %rsi callq 0x1ea12c movq %r13, %rdi movq %r12, %rsi callq 0x1ea12c incq %rbx movq 0x8(%rsp), %rcx movq (%rcx), %rax jmp 0x2a1517 imull %ecx, %r13d movslq %r13d, %rsi leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %rsi, 0xc8(%rsp) callq 0x73bbe movq (%rbp), %r9 movq -0x18(%r9), %rax movl 0x18(%rsp), %r10d imull 0xe0(%rbp,%rax), %r10d movq 0x130(%rsp), %rbx movl 0xdc(%rbp,%rax), %ecx imull 0xd4(%rbp,%rax), %ecx subl %ecx, %r10d xorl %ecx, %ecx xorl %edx, %edx xorl %esi, %esi cmpl 0xd8(%rbp,%rax), %esi jge 0x2a1634 movslq %ecx, %rcx leaq (%rbx,%rcx,4), %r8 xorl %edi, %edi cmpl 0xd4(%rbp,%rax), %edi jge 0x2a162a movl %edx, (%r8,%rdi,4) movq -0x18(%r9), %rax addl 0xdc(%rbp,%rax), %edx incq %rdi jmp 0x2a160d addl %r10d, %edx incl %esi addq %rdi, %rcx jmp 0x2a15fb xorl %edx, %edx testl %r13d, %r13d movl $0x0, %ecx movq %rcx, 0x50(%rsp) cmovlel %edx, %r13d vbroadcastss 0x14fb64(%rip), %xmm4 # 0x3f11b4 leaq 0x156ff9(%rip), %r8 # 0x3f8650 xorl %ebp, %ebp movq %r9, 0xc0(%rsp) movq 0x8(%rsp), %rcx movslq 0x108(%rcx,%rax), %rax cmpq %rax, %rbp jge 0x2a19d1 movq 0x10(%r14), %r9 imulq %rbp, %r9 imulq 0x40(%r14), %r9 addq (%r14), %r9 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rbp, %r11 movq 0x80(%rsp), %rax imulq %rax, %r11 addq 0x70(%rsp), %r11 imulq %rax, %r10 movq 0x28(%rcx), %r15 addq 0x50(%rsp), %r15 xorl %eax, %eax movq %r9, 0x18(%rsp) movq %r10, 0x28(%rsp) movq %r11, 0x48(%rsp) movq %rbp, 0xd0(%rsp) cmpl 0x68(%rsp), %eax jg 0x2a19a6 movl %eax, 0x30(%rsp) movq 0x8(%rsp), %rax movq (%rax), %rax movq %rax, 0x40(%rsp) xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d jg 0x2a199b movq 0x40(%rsp), %rax movq -0x18(%rax), %r12 movq 0x8(%rsp), %rcx movslq 0xe8(%rcx,%r12), %rax movslq 0x30(%rsp), %rdx imulq %rax, %rdx imulq %r10, %rdx movslq 0xe4(%rcx,%r12), %rax movslq %r14d, %rcx imulq %rax, %rcx addq %r11, %rcx addq %rdx, %rcx xorl %edx, %edx xorl %eax, %eax cmpq %rdx, %r13 je 0x2a1753 movslq (%rbx,%rdx,4), %rsi movsbl (%rcx,%rsi), %esi movsbl (%r15,%rdx), %edi imull %esi, %edi addl %edi, %eax incq %rdx jmp 0x2a1737 movq 0x8(%rsp), %rdx movq 0x1f8(%rdx,%r12), %rcx vmovss (%rcx,%rbp,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2a1788 movq 0x240(%rdx,%r12), %rcx vmulss (%rcx,%rbp,4), %xmm1, %xmm0 vmovss 0x14d504(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vcvtsi2ss %eax, %xmm6, %xmm1 vmulss %xmm1, %xmm0, %xmm5 cmpl $0x0, 0x100(%rdx,%r12) je 0x2a17a8 movq 0x1b0(%rdx,%r12), %rax vaddss (%rax,%rbp,4), %xmm5, %xmm5 movl 0x110(%rdx,%r12), %eax decl %eax cmpl $0x5, %eax ja 0x2a1920 movslq (%r8,%rax,4), %rax addq %r8, %rax jmpq %rax vmaxss 0x14c844(%rip), %xmm5, %xmm0 # 0x3ee010 jmp 0x2a1924 vmovaps %xmm5, %xmm0 movq %r9, 0x58(%rsp) vmovss %xmm5, 0xd8(%rsp) callq 0x5f410 vaddss 0x14d498(%rip), %xmm0, %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x48(%rsp), %r11 movq 0x28(%rsp), %r10 movq 0x58(%rsp), %r9 leaq 0x156e40(%rip), %r8 # 0x3f8650 vbroadcastss 0x14f99b(%rip), %xmm4 # 0x3f11b4 vmulss 0xd8(%rsp), %xmm0, %xmm0 jmp 0x2a1924 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss 0x4(%rax), %xmm1 vmaxss (%rax), %xmm5, %xmm0 vucomiss %xmm1, %xmm0 jbe 0x2a1924 vmovaps %xmm1, %xmm0 jmp 0x2a1924 vminss 0x14f960(%rip), %xmm5, %xmm0 # 0x3f11b8 vxorps %xmm4, %xmm0, %xmm1 vcmpltss 0x14f957(%rip), %xmm0, %xmm0 # 0x3f11bc vbroadcastss 0x14f94a(%rip), %xmm2 # 0x3f11b8 vblendvps %xmm0, %xmm2, %xmm1, %xmm0 movq %r9, %rbp callq 0x5f410 movq 0x48(%rsp), %r11 movq 0x28(%rsp), %r10 movq %rbp, %r9 movq 0xd0(%rsp), %rbp leaq 0x156db8(%rip), %r8 # 0x3f8650 vbroadcastss 0x14f913(%rip), %xmm4 # 0x3f11b4 vmovss 0x14d3df(%rip), %xmm1 # 0x3eec88 vaddss %xmm1, %xmm0, %xmm0 vdivss %xmm0, %xmm1, %xmm0 jmp 0x2a1924 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vxorps %xmm0, %xmm0, %xmm0 vcmpltss %xmm5, %xmm0, %xmm0 vmovss (%rax), %xmm1 vbroadcastss 0x14d3b2(%rip), %xmm2 # 0x3eec88 vblendvps %xmm0, %xmm2, %xmm1, %xmm0 vmulss %xmm5, %xmm0, %xmm0 jmp 0x2a1924 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss (%rax), %xmm1 vmovss 0x4(%rax), %xmm2 vxorps %xmm4, %xmm2, %xmm0 vdivss %xmm1, %xmm0, %xmm3 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm3, %xmm5 jb 0x2a1924 vmovss 0x14d376(%rip), %xmm0 # 0x3eec88 vdivss %xmm1, %xmm0, %xmm0 vaddss %xmm0, %xmm3, %xmm0 vucomiss %xmm0, %xmm5 jbe 0x2a1990 vmovaps %xmm5, %xmm0 cmpl $0x65, 0x38(%rsp) jl 0x2a1976 movq 0x8(%rsp), %rax movq 0x288(%rax,%r12), %rax vmulss (%rax,%rbp,4), %xmm0, %xmm0 vandps %xmm4, %xmm0, %xmm1 vbroadcastss 0x156cba(%rip), %xmm2 # 0x3f8604 vorps %xmm2, %xmm1, %xmm1 vaddss %xmm1, %xmm0, %xmm0 vroundss $0xb, %xmm0, %xmm0, %xmm0 vcvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movb %al, (%r9) incq %r9 jmp 0x2a1988 movq 0x18(%rsp), %rax vmovss %xmm0, (%rax) addq $0x4, %rax movq %rax, 0x18(%rsp) incl %r14d jmp 0x2a16f0 vfmadd213ss %xmm2, %xmm5, %xmm1 # xmm1 = (xmm5 xmm1) + xmm2 vmulss %xmm5, %xmm1, %xmm0 jmp 0x2a1924 movl 0x30(%rsp), %eax incl %eax jmp 0x2a16d2 incq %rbp movq 0xc0(%rsp), %r9 movq -0x18(%r9), %rax movq 0x50(%rsp), %rcx addq 0xc8(%rsp), %rcx movq %rcx, 0x50(%rsp) movq 0x20(%rsp), %r14 jmp 0x2a1661 leaq 0x130(%rsp), %rdi callq 0x624be xorl %r12d, %r12d jmp 0x2a1ab2 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0xe0(%rsp), %xmm0 vmovups %xmm0, (%rbx) movq 0xf0(%rsp), %rax movq %rax, 0x10(%rbx) movl 0xf8(%rsp), %eax movl %eax, 0x18(%rbx) movq 0x100(%rsp), %rax movq %rax, 0x20(%rbx) vmovups 0x108(%rsp), %xmm0 vmovups %xmm0, 0x28(%rbx) movl 0x118(%rsp), %eax movl %eax, 0x38(%rbx) movq 0x120(%rsp), %rax movq %rax, 0x40(%rbx) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2a1a78 lock decl (%rax) jne 0x2a1a78 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x2a1a70 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a1a78 movq %rsi, %rdi callq 0x5f3e0 movq 0x138(%rsp), %rax testq %rax, %rax movl %r14d, %r12d je 0x2a1ab2 lock decl (%rax) jne 0x2a1ab2 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi je 0x2a1aaa movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a1ab2 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a1ae3 lock decl (%rax) jne 0x2a1ae3 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a1adb movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a1ae3 movq %rsi, %rdi callq 0x5f3e0 movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x2a1b1a lock decl (%rax) jne 0x2a1b1a movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi je 0x2a1b12 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a1b1a movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x318, %rsp # imm = 0x318 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovaps %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2a1ba5 movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x14d104(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2a1b4f leaq 0x2a0(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdx movq 0x20(%rsp), %rcx movslq 0x2c(%rcx), %rsi movslq 0x30(%rcx), %rax movl 0x38(%rcx), %edi movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movl 0x34(%rcx), %r8d movl %r8d, 0x58(%rsp) movq (%rcx), %r8 movq 0x10(%rcx), %r11 movq 0x40(%rcx), %r15 imulq %r11, %r15 movq %rax, %rbx movq %rsi, 0x38(%rsp) imulq %rsi, %rbx movq %r11, 0xd0(%rsp) movq %rbx, 0x50(%rsp) imulq %rbx, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0xc8(%rsp) movq 0x80(%rsp), %rsi imulq 0xb0(%rsp), %rsi movq %rsi, 0xc0(%rsp) xorl %esi, %esi testl %eax, %eax movl $0x0, %r11d cmovgl %eax, %r11d movl %r11d, 0x10(%rsp) movl 0x28(%rcx), %eax movl %eax, 0x64(%rsp) testl %edi, %edi cmovlel %esi, %edi movq %rdi, 0x18(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x190(%rsp) movq %rdx, 0x28(%rsp) leaq (%rdx,%rdx), %rax movq %rax, 0x188(%rsp) movq 0x2a0(%rsp), %rax movq %rax, 0x180(%rsp) movq %r8, 0xd8(%rsp) xorl %esi, %esi movq %r10, 0x48(%rsp) movq %r15, 0x68(%rsp) cmpq 0x18(%rsp), %rsi je 0x2a1e5f movq 0xc8(%rsp), %rax xorl %edx, %edx divq 0xd0(%rsp) cmpl $0x4, 0x64(%rsp) cmoveq 0x50(%rsp), %rax testq %r10, %r10 je 0x2a1cd7 vmovss (%r10,%rsi,4), %xmm1 jmp 0x2a1cdb vxorps %xmm1, %xmm1, %xmm1 imulq %rsi, %r15 addq 0xd8(%rsp), %r15 movq 0x180(%rsp), %rcx vmovss (%rcx,%rsi,4), %xmm0 imull 0x58(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2a1d14 vmovss %xmm1, (%r8,%rcx,4) incq %rcx jmp 0x2a1d05 leaq (%rsi,%rsi,8), %rax movq 0xc0(%rsp), %r13 movq %rsi, 0x40(%rsp) imulq %rsi, %r13 addq 0x190(%rsp), %r13 movq 0x28(%rsp), %rcx leaq (%rcx,%r13), %r11 movq 0x188(%rsp), %rcx leaq (%rcx,%r13), %rbx xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d je 0x2a1e45 xorl %ecx, %ecx movq %r15, %rdx movq 0x38(%rsp), %rsi movl %esi, %r10d testl %r10d, %r10d jle 0x2a1e25 movsbl (%r13,%rcx), %r12d movsbl (%r9,%rax), %esi imull %r12d, %esi movsbl 0x1(%r13,%rcx), %r12d movsbl 0x1(%r9,%rax), %edi imull %r12d, %edi addl %esi, %edi movsbl 0x2(%r13,%rcx), %esi movsbl 0x2(%r9,%rax), %r12d imull %esi, %r12d movsbl (%r11,%rcx), %esi movsbl 0x3(%r9,%rax), %ebp imull %esi, %ebp addl %r12d, %ebp addl %edi, %ebp movsbl 0x1(%r11,%rcx), %esi movsbl 0x4(%r9,%rax), %edi imull %esi, %edi movsbl 0x2(%r11,%rcx), %esi movsbl 0x5(%r9,%rax), %r12d imull %esi, %r12d addl %edi, %r12d movsbl (%rbx,%rcx), %esi movsbl 0x6(%r9,%rax), %edi imull %esi, %edi addl %r12d, %edi addl %ebp, %edi movsbl 0x1(%rbx,%rcx), %esi movsbl 0x7(%r9,%rax), %ebp imull %esi, %ebp movsbl 0x2(%rbx,%rcx), %esi movsbl 0x8(%r9,%rax), %r12d imull %esi, %r12d addl %ebp, %r12d addl %edi, %r12d vcvtsi2ss %r12d, %xmm2, %xmm1 vfmadd213ss (%r15,%rcx,4), %xmm0, %xmm1 # xmm1 = (xmm0 * xmm1) + mem vmovss %xmm1, (%r15,%rcx,4) addq $0x4, %rdx decl %r10d incq %rcx jmp 0x2a1d61 addq %rcx, %r13 addq $0x2, %r13 addq %rcx, %r11 addq $0x2, %r11 addq %rcx, %rbx addq $0x2, %rbx incl %r14d movq %rdx, %r15 jmp 0x2a1d49 movq 0x40(%rsp), %rsi incq %rsi movq 0x68(%rsp), %r15 addq %r15, %r8 movq 0x48(%rsp), %r10 jmp 0x2a1ca2 leaq 0x2a0(%rsp), %rdi jmp 0x2a23e4 leaq 0x2b8(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rax movq 0x20(%rsp), %r15 movl 0x2c(%r15), %ecx movl %ecx, 0x38(%rsp) movl 0x30(%r15), %ecx movl 0x38(%r15), %edx movq 0x28(%rbp), %rdi movq 0x1b0(%rbp,%r13), %rsi xorl %r9d, %r9d testl %ecx, %ecx cmovlel %r9d, %ecx movl %ecx, 0x10(%rsp) movq %rax, 0x28(%rsp) addq %rax, %rax movq %rax, 0x58(%rsp) testl %edx, %edx cmovlel %r9d, %edx movq %rdx, 0x18(%rsp) vbroadcastss 0x14f2dc(%rip), %xmm0 # 0x3f11b4 vbroadcastss 0x156723(%rip), %xmm1 # 0x3f8604 pushq $-0x7f popq %r8 pushq $0x7f popq %rbx movq %rsi, 0x48(%rsp) cmpq 0x18(%rsp), %r9 je 0x2a209c testq %rsi, %rsi je 0x2a1f05 vmovss (%rsi,%r9,4), %xmm2 jmp 0x2a1f09 vxorps %xmm2, %xmm2, %xmm2 movq 0x40(%r15), %r14 imulq %r9, %r14 imulq 0x10(%r15), %r14 addq (%r15), %r14 movq 0x2b8(%rsp), %rax vmovss (%rax,%r9,8), %xmm3 vmovss 0x4(%rax,%r9,8), %xmm4 movq 0xb0(%rsp), %r15 imulq %r9, %r15 imulq 0x80(%rsp), %r15 addq 0x70(%rsp), %r15 movq %r9, 0x40(%rsp) leaq (%r9,%r9,8), %r12 movq 0x28(%rsp), %rax leaq (%r15,%rax), %r13 movq 0x58(%rsp), %rax leaq (%r15,%rax), %rbp xorl %eax, %eax cmpl 0x10(%rsp), %eax je 0x2a2085 xorl %r10d, %r10d movl 0x38(%rsp), %ecx testl %ecx, %ecx jle 0x2a2066 movsbl (%r15,%r10), %r11d movsbl (%rdi,%r12), %r9d imull %r11d, %r9d movsbl 0x1(%r15,%r10), %r11d movsbl 0x1(%rdi,%r12), %esi imull %r11d, %esi addl %r9d, %esi movsbl 0x2(%r15,%r10), %r9d movsbl 0x2(%rdi,%r12), %r11d imull %r9d, %r11d movsbl (%r13,%r10), %r9d movsbl 0x3(%rdi,%r12), %edx imull %r9d, %edx addl %r11d, %edx addl %esi, %edx movsbl 0x1(%r13,%r10), %esi movsbl 0x4(%rdi,%r12), %r9d imull %esi, %r9d movsbl 0x2(%r13,%r10), %esi movsbl 0x5(%rdi,%r12), %r11d imull %esi, %r11d addl %r9d, %r11d movsbl (%rbp,%r10), %esi movsbl 0x6(%rdi,%r12), %r9d imull %esi, %r9d addl %r11d, %r9d addl %edx, %r9d movsbl 0x1(%rbp,%r10), %edx movsbl 0x7(%rdi,%r12), %esi imull %edx, %esi movsbl 0x2(%rbp,%r10), %edx movsbl 0x8(%rdi,%r12), %r11d imull %edx, %r11d addl %esi, %r11d addl %r9d, %r11d vcvtsi2ss %r11d, %xmm7, %xmm5 vfmadd213ss %xmm2, %xmm3, %xmm5 # xmm5 = (xmm3 * xmm5) + xmm2 vmulss %xmm4, %xmm5, %xmm5 vandps %xmm0, %xmm5, %xmm6 vorps %xmm1, %xmm6, %xmm6 vaddss %xmm6, %xmm5, %xmm5 vroundss $0xb, %xmm5, %xmm5, %xmm5 vcvttss2si %xmm5, %r11d cmpl $-0x7e, %r11d jge 0x2a204f movl %r8d, %r11d cmpl $0x7f, %r11d jl 0x2a2058 movl %ebx, %r11d movb %r11b, (%r14,%r10) decl %ecx incq %r10 jmp 0x2a1f76 addq %r10, %r15 addq $0x2, %r15 addq %r10, %r13 addq $0x2, %r13 addq %r10, %rbp addq $0x2, %rbp incl %eax addq %r10, %r14 jmp 0x2a1f65 movq 0x40(%rsp), %r9 incq %r9 movq 0x20(%rsp), %r15 movq 0x48(%rsp), %rsi jmp 0x2a1eed leaq 0x2b8(%rsp), %rdi jmp 0x2a262f vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovaps %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2a211f movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x14cb8a(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2a20c9 leaq 0x270(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdi movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movslq 0x30(%rdx), %rax movl 0x38(%rdx), %r8d movl %edi, %ecx subl %esi, %ecx addl %ecx, %ecx movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movq 0x10(%rdx), %r11 movq 0x40(%rdx), %rbx imulq %r11, %rbx movq %rax, %r14 movq %rsi, 0x38(%rsp) imulq %rsi, %r14 movq %r11, 0x58(%rsp) movq %r14, 0xd0(%rsp) imulq %r14, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0x68(%rsp) movq 0x80(%rsp), %rsi imulq 0xb0(%rsp), %rsi movq %rsi, 0x50(%rsp) xorl %r11d, %r11d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi movl %esi, 0x10(%rsp) movl 0x34(%rdx), %eax movl %eax, 0xc8(%rsp) movq (%rdx), %r12 movslq %ecx, %rsi testl %r8d, %r8d cmovlel %r11d, %r8d movq %r8, 0x18(%rsp) movl 0x28(%rdx), %eax movl %eax, 0x64(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x190(%rsp) movq %rdi, 0x28(%rsp) leaq (%rdi,%rdi), %rax movq %rax, 0x188(%rsp) movq 0x270(%rsp), %rax movq %rax, 0x180(%rsp) movq %r12, 0xc0(%rsp) xorl %edi, %edi movq %r10, 0x48(%rsp) movq %rbx, 0xd8(%rsp) cmpq 0x18(%rsp), %rdi je 0x2a23dc movq 0x68(%rsp), %rax xorl %edx, %edx divq 0x58(%rsp) cmpl $0x4, 0x64(%rsp) cmoveq 0xd0(%rsp), %rax testq %r10, %r10 je 0x2a2256 vmovss (%r10,%rdi,4), %xmm1 jmp 0x2a225a vxorps %xmm1, %xmm1, %xmm1 movq %rbx, %rdx imulq %rdi, %rdx addq 0xc0(%rsp), %rdx movq 0x180(%rsp), %rcx vmovss (%rcx,%rdi,4), %xmm0 imull 0xc8(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2a2299 vmovss %xmm1, (%r12,%rcx,4) incq %rcx jmp 0x2a228a leaq (%rdi,%rdi,8), %rax movq 0x50(%rsp), %r8 movq %rdi, 0x40(%rsp) imulq %rdi, %r8 addq 0x190(%rsp), %r8 movq 0x28(%rsp), %rcx leaq (%r8,%rcx), %r13 movq 0x188(%rsp), %rcx leaq (%r8,%rcx), %r11 xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d je 0x2a23bf xorl %ecx, %ecx movq 0x38(%rsp), %rdi movl %edi, %r10d testl %r10d, %r10d jle 0x2a23a5 movsbl (%r8,%rcx), %ebx movsbl (%r9,%rax), %r15d imull %ebx, %r15d movsbl 0x1(%r8,%rcx), %ebx movsbl 0x1(%r9,%rax), %edi imull %ebx, %edi addl %r15d, %edi movsbl 0x2(%r8,%rcx), %ebx movsbl 0x2(%r9,%rax), %r15d imull %ebx, %r15d movsbl (%r13,%rcx), %ebx movsbl 0x3(%r9,%rax), %ebp imull %ebx, %ebp addl %r15d, %ebp addl %edi, %ebp movsbl 0x1(%r13,%rcx), %edi movsbl 0x4(%r9,%rax), %ebx imull %edi, %ebx movsbl 0x2(%r13,%rcx), %edi movsbl 0x5(%r9,%rax), %r15d imull %edi, %r15d addl %ebx, %r15d movsbl (%r11,%rcx), %edi movsbl 0x6(%r9,%rax), %ebx imull %edi, %ebx addl %r15d, %ebx addl %ebp, %ebx movsbl 0x1(%r11,%rcx), %edi movsbl 0x7(%r9,%rax), %ebp imull %edi, %ebp movsbl 0x2(%r11,%rcx), %edi movsbl 0x8(%r9,%rax), %r15d imull %edi, %r15d addl %ebp, %r15d addl %ebx, %r15d vcvtsi2ss %r15d, %xmm2, %xmm1 vfmadd213ss (%rdx), %xmm0, %xmm1 # xmm1 = (xmm0 * xmm1) + mem vmovss %xmm1, (%rdx) addq $0x4, %rdx decl %r10d addq $0x2, %rcx jmp 0x2a22e0 addq %rsi, %r8 addq %rcx, %r8 addq %rsi, %r13 addq %rcx, %r13 addq %rsi, %r11 addq %rcx, %r11 incl %r14d jmp 0x2a22cb movq 0x40(%rsp), %rdi incq %rdi movq 0xd8(%rsp), %rbx addq %rbx, %r12 movq 0x48(%rsp), %r10 jmp 0x2a2224 leaq 0x270(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x20(%rsp), %r15 jmp 0x2a2641 leaq 0x288(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rsi movq 0x20(%rsp), %r15 movl 0x2c(%r15), %eax movl 0x30(%r15), %edx movl 0x38(%r15), %r8d movl %esi, %ecx movl %eax, 0x10(%rsp) subl %eax, %ecx addl %ecx, %ecx movq 0x8(%rsp), %rax movq 0x28(%rax), %rdi movq 0x1b0(%rax,%r14), %rax movq %rsi, 0x18(%rsp) addq %rsi, %rsi movq %rsi, 0x48(%rsp) movslq %ecx, %r10 xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %r8d, %r8d cmovlel %esi, %r8d movq %r8, 0x40(%rsp) vbroadcastss 0x14ed3f(%rip), %xmm0 # 0x3f11b4 vbroadcastss 0x156186(%rip), %xmm1 # 0x3f8604 movq %rax, 0x28(%rsp) cmpq 0x40(%rsp), %rsi je 0x2a2627 testq %rax, %rax je 0x2a249a vmovss (%rax,%rsi,4), %xmm2 jmp 0x2a249e vxorps %xmm2, %xmm2, %xmm2 movq 0x40(%r15), %rbx imulq %rsi, %rbx imulq 0x10(%r15), %rbx addq (%r15), %rbx movq 0x288(%rsp), %rcx vmovss (%rcx,%rsi,8), %xmm3 vmovss 0x4(%rcx,%rsi,8), %xmm4 movq 0xb0(%rsp), %r14 imulq %rsi, %r14 imulq 0x80(%rsp), %r14 addq 0x70(%rsp), %r14 movq %rsi, 0x38(%rsp) leaq (%rsi,%rsi,8), %r15 movq 0x18(%rsp), %rax leaq (%r14,%rax), %r12 movq 0x48(%rsp), %rax leaq (%r14,%rax), %r13 xorl %ebp, %ebp cmpl %edx, %ebp je 0x2a2610 xorl %r9d, %r9d movl 0x10(%rsp), %ecx testl %ecx, %ecx jle 0x2a25f7 movsbl (%r14,%r9), %esi movsbl (%rdi,%r15), %r8d imull %esi, %r8d movsbl 0x1(%r14,%r9), %esi movsbl 0x1(%rdi,%r15), %r11d imull %esi, %r11d addl %r8d, %r11d movsbl 0x2(%r14,%r9), %esi movsbl 0x2(%rdi,%r15), %r8d imull %esi, %r8d movsbl (%r12,%r9), %esi movsbl 0x3(%rdi,%r15), %eax imull %esi, %eax addl %r8d, %eax addl %r11d, %eax movsbl 0x1(%r12,%r9), %esi movsbl 0x4(%rdi,%r15), %r8d imull %esi, %r8d movsbl 0x2(%r12,%r9), %esi movsbl 0x5(%rdi,%r15), %r11d imull %esi, %r11d addl %r8d, %r11d movsbl (%r13,%r9), %esi movsbl 0x6(%rdi,%r15), %r8d imull %esi, %r8d addl %r11d, %r8d addl %eax, %r8d movsbl 0x1(%r13,%r9), %eax movsbl 0x7(%rdi,%r15), %esi imull %eax, %esi movsbl 0x2(%r13,%r9), %eax movsbl 0x8(%rdi,%r15), %r11d imull %eax, %r11d addl %esi, %r11d addl %r8d, %r11d vcvtsi2ss %r11d, %xmm7, %xmm5 vfmadd213ss %xmm2, %xmm3, %xmm5 # xmm5 = (xmm3 * xmm5) + xmm2 vmulss %xmm4, %xmm5, %xmm5 vandps %xmm0, %xmm5, %xmm6 vorps %xmm1, %xmm6, %xmm6 vaddss %xmm6, %xmm5, %xmm5 vroundss $0xb, %xmm5, %xmm5, %xmm5 vcvttss2si %xmm5, %esi cmpl $-0x7e, %esi jge 0x2a25de pushq $-0x7f popq %rsi cmpl $0x7f, %esi jl 0x2a25e6 pushq $0x7f popq %rsi movb %sil, (%rbx) incq %rbx decl %ecx addq $0x2, %r9 jmp 0x2a2507 addq %r10, %r14 addq %r9, %r14 addq %r10, %r12 addq %r9, %r12 addq %r10, %r13 addq %r9, %r13 incl %ebp jmp 0x2a24f8 movq 0x38(%rsp), %rsi incq %rsi movq 0x20(%rsp), %r15 movq 0x28(%rsp), %rax jmp 0x2a2483 leaq 0x288(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x8(%rsp), %rax xorl %r12d, %r12d movq 0x8(%rax), %rdi testq %rdi, %rdi je 0x2a1ab2 movq (%rdi), %rax movq %r15, %rsi movq 0x30(%rsp), %rdx callq 0x48(%rax) jmp 0x2a1ab2 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 movq %rax, %rbx leaq 0x130(%rsp), %rdi callq 0x621c2 jmp 0x2a27f9 jmp 0x2a2712 jmp 0x2a2712 jmp 0x2a2873 jmp 0x2a2873 movq %rax, %rbx jmp 0x2a27c2 jmp 0x2a2873 jmp 0x2a2712 movq %rax, %rbx jmp 0x2a278b jmp 0x2a2873 jmp 0x2a2873 jmp 0x2a2712 jmp 0x2a2873 jmp 0x2a2873 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a282a lock decl (%rax) jne 0x2a282a movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a281a movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a282a jmp 0x2a2873 movq %rax, %rbx jmp 0x2a282a movq %rax, %rbx jmp 0x2a27f9 movq %rax, %rbx movq 0x230(%rsp), %rax testq %rax, %rax je 0x2a2754 lock decl (%rax) jne 0x2a2754 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi jne 0x2a274e movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a2754 movq (%rdi), %rax callq 0x18(%rax) movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x2a278b lock decl (%rax) jne 0x2a278b movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi jne 0x2a2785 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a278b movq (%rdi), %rax callq 0x18(%rax) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2a27c2 lock decl (%rax) jne 0x2a27c2 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi jne 0x2a27bc movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a27c2 movq (%rdi), %rax callq 0x18(%rax) movq 0x138(%rsp), %rax testq %rax, %rax je 0x2a27f9 lock decl (%rax) jne 0x2a27f9 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi jne 0x2a27f3 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a27f9 movq (%rdi), %rax callq 0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a282a lock decl (%rax) jne 0x2a282a movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2a2824 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a282a movq (%rdi), %rax callq 0x18(%rax) movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x2a2861 lock decl (%rax) jne 0x2a2861 movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi jne 0x2a285b movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a2861 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2a2873 jmp 0x2a2873 jmp 0x2a2873 jmp 0x2a2873 jmp 0x2a2873 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_fma::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_fma::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float)bias_data) + g 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float)bias_data) + g 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float)bias_data) + g 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x29be02	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
ncnn::ConvolutionDepthWise_x86_fma::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_fma::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w = weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w = bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x168, %rsp # imm = 0x168 movq %rsi, %rbp movq %rdi, %r13 movq (%rsi), %r14 leaq 0x48(%r14), %rdi movq (%rdx), %rax movq %rax, 0xb8(%rsp) movl 0x60(%r14), %ebx movl 0x74(%r14), %eax movl %eax, 0x1c(%rsp) movl 0x78(%r14), %eax movl %eax, 0x18(%rsp) imull 0x80(%r14), %ebx leaq 0x70(%rsp), %rsi andq $0x0, 0x40(%rsi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rsi) vmovups %xmm0, 0xc(%rsi) vmovaps %xmm0, 0x20(%rsi) vmovups %xmm0, 0x2c(%rsi) movq %rcx, %r15 movq %rcx, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2a2ea9 movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x2a2ea9 movslq 0x88(%rsp), %rcx movl 0x9c(%rsp), %eax imull %ecx, %eax movl %eax, 0x9c(%rsp) movq 0x80(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x80(%rsp) movl $0x1, 0x88(%rsp) andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) vmovaps %xmm0, 0x40(%rsp) vmovups %xmm0, 0x4c(%rsp) movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r13,%rax) je 0x2a29f2 movl $0x90, %edi addq (%rbp), %rdi leaq 0x20(%rsp), %rsi movq %r15, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x20(%rsp) je 0x2a2e7b movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x2a2e7b movslq 0x38(%rsp), %rcx movl 0x4c(%rsp), %eax imull %ecx, %eax movl %eax, 0x4c(%rsp) movq 0x30(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x30(%rsp) movl $0x1, 0x38(%rsp) pushq $0x2a popq %rdi callq 0x782bf movq %rax, %r12 leaq 0x8(%rsp), %rdi callq 0x71548 leaq 0x8(%rsp), %rdi xorl %esi, %esi movl %ebx, %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi movl 0x1c(%rsp), %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0xb popq %rsi movl 0x18(%rsp), %edx callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xdc(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xd popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xec(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x4 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xf popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xe popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x10 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax vmovss 0xfc(%r13,%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x12 popq %rsi callq 0x71952 movq (%r13), %rax movq -0x18(%rax), %rax movl 0x100(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x5 popq %rsi callq 0x7193a movl 0x9c(%rsp), %edx leaq 0x8(%rsp), %rdi pushq $0x6 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x108(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x7 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x10c(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x110(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax leaq (%rax,%r13), %rdx addq $0x118, %rdx # imm = 0x118 leaq 0x8(%rsp), %rdi pushq $0xa popq %rsi callq 0x7196c movq (%r12), %rax leaq 0x8(%rsp), %rsi movq %r12, %rdi callq 0x10(%rax) andq $0x0, 0x110(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0xd0(%rsp) vmovups %xmm0, 0xdc(%rsp) vmovaps %xmm0, 0xf0(%rsp) vmovups %xmm0, 0xfc(%rsp) andq $0x0, 0x158(%rsp) vmovups %xmm0, 0x118(%rsp) vmovups %xmm0, 0x124(%rsp) vmovups %xmm0, 0x138(%rsp) vmovups %xmm0, 0x144(%rsp) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a2c60 lock incl (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2a2c97 lock decl (%rax) jne 0x2a2c97 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x2a2c8f movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a2c97 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0xd0(%rsp) movq 0x80(%rsp), %rax movq %rax, 0xe0(%rsp) movl 0x88(%rsp), %eax movl %eax, 0xe8(%rsp) movq 0x90(%rsp), %rax movq %rax, 0xf0(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0xf8(%rsp) movl 0xa8(%rsp), %eax movl %eax, 0x108(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x110(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a2d11 lock incl (%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x2a2d48 lock decl (%rax) jne 0x2a2d48 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x2a2d40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a2d48 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 leaq 0xd0(%rsp), %rsi vmovups %xmm0, 0x48(%rsi) movq 0x30(%rsp), %rax movq %rax, 0x58(%rsi) movl 0x38(%rsp), %eax movl %eax, 0x60(%rsi) movq 0x40(%rsp), %rax movq %rax, 0x68(%rsi) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsi) movl 0x58(%rsp), %eax movl %eax, 0x80(%rsi) movq 0x60(%rsp), %rax movq %rax, 0x88(%rsi) leaq 0xc0(%rsp), %rdi callq 0x6b00e movq (%r12), %rax leaq 0xc0(%rsp), %rsi movq %r12, %rdi callq 0x18(%rax) leaq 0xc0(%rsp), %rdi callq 0x6b03a movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq 0x20(%rax) movq (%r12), %rax movq %r12, %rdi movq %r14, %rsi movq 0xb8(%rsp), %rdx movq %r15, %rcx callq 0x38(%rax) movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq 0x28(%rax) movq (%r12), %rax movq %r12, %rdi callq 0x8(%rax) pushq $0x48 popq %rbx vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%rbx), %rax testq %rax, %rax je 0x2a2e43 lock decl (%rax) jne 0x2a2e43 movq 0xd0(%rsp,%rbx), %rsi movq 0xf0(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x2a2e37 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2a2e43 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%rbx), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x2a2e04 leaq 0x8(%rsp), %rdi callq 0x71614 xorl %r12d, %r12d movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a2ea9 lock decl (%rax) jne 0x2a2ea9 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2a2ea1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a2ea9 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a2eda lock decl (%rax) jne 0x2a2eda movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a2ed2 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a2eda movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x168, %rsp # imm = 0x168 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2a300f jmp 0x2a300f movq %rax, %rbx leaq 0xc0(%rsp), %rdi callq 0x6b03a jmp 0x2a2f1c jmp 0x2a2f19 jmp 0x2a2f11 jmp 0x2a2f11 movq %rax, %rbx jmp 0x2a2fa6 movq %rax, %rbx pushq $0x48 popq %r14 vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%r14), %rax testq %rax, %rax je 0x2a2f63 lock decl (%rax) jne 0x2a2f63 movq 0xd0(%rsp,%r14), %rsi movq 0xf0(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2a2f57 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2a2f63 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2a2f24 jmp 0x2a2f9c jmp 0x2a300f movq %rax, %rbx jmp 0x2a2fd4 jmp 0x2a300f movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a2fd4 lock decl (%rax) jne 0x2a2fd4 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2a2fce movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a2fd4 movq (%rdi), %rax callq 0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a3005 lock decl (%rax) jne 0x2a3005 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2a2fff movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a3005 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2a300f movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_fma::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_fma::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w = weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w = bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }	movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x2a2888	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp
ncnn::ConvolutionDepthWise_x86_avx::create_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x50, %rsp movq (%rdi), %rax movq -0x18(%rax), %r13 cmpl $0x0, 0x160(%rdi,%r13) je 0x2a30ed xorl %eax, %eax addq $0x50, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 retq movq %rsi, %r14 movq %rdi, %rbx movl 0x110(%rdi,%r13), %ecx decl %ecx cmpl $0x5, %ecx ja 0x2a3444 leaq 0x15559b(%rip), %rax # 0x3f86a8 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx jmpq %rcx pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x2a3280 pushq $0x47 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x2a3280 pushq $0x36 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax vmovss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) jmp 0x2a3280 pushq $0x1e popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq 0x10(%rax) jmp 0x2a3280 pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) jmp 0x2a3280 pushq $0x43 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax vmovss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq 0x10(%rax) leaq 0x8(%rsp), %rdi callq 0x71614 movq (%r15), %rax movq %r15, %rdi movq %r14, %rsi callq 0x20(%rax) movq (%rbx), %rax movq %r15, 0x8(%rbx) movq -0x18(%rax), %r15 cmpb $0x1, 0x1e(%r14) jne 0x2a32c3 cmpq $0x1, 0x178(%rbx,%r15) jne 0x2a32c3 movq %rbx, %rdi movq %r14, %rsi callq 0x2a3642 jmp 0x2a30dd movl 0xd0(%rbx,%r15), %esi movl 0xd4(%rbx,%r15), %r10d movl 0xd8(%rbx,%r15), %r11d movl %r11d, %r8d imull %r10d, %r8d movl 0x104(%rbx,%r15), %eax movl 0x108(%rbx,%r15), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x2a33e2 imull %ecx, %eax cmpl %ecx, %eax jne 0x2a33e2 leaq (%rbx,%r15), %rsi cmpb $0x1, 0x27(%r14) jne 0x2a3339 testb $0x7, %cl je 0x2a344c testb $0x3, %cl je 0x2a34ad xorl $0x3, %r10d xorl $0x3, %r11d orl %r10d, %r11d jne 0x2a33e2 cmpl $0x1, 0xdc(%rsi) jne 0x2a336e cmpl $0x1, 0xe0(%rsi) jne 0x2a336e cmpl $0x1, 0xe4(%rsi) jne 0x2a336e cmpl $0x1, 0xe8(%rsi) je 0x2a3392 cmpl $0x1, 0xdc(%rsi) jne 0x2a33e2 cmpl $0x1, 0xe0(%rsi) jne 0x2a33e2 cmpl $0x2, 0xe4(%rsi) jne 0x2a33e2 cmpl $0x2, 0xe8(%rsi) jne 0x2a33e2 addq $0x168, %r15 # imm = 0x168 cmpq $0x28, %r15 je 0x2a33ed movq 0x8(%rbx,%r15), %rax testq %rax, %rax je 0x2a33ac lock incl (%rax) leaq 0x28(%rbx), %r12 movq 0x30(%rbx), %rax testq %rax, %rax je 0x2a3552 lock decl (%rax) jne 0x2a3552 movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x2a354a movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a3552 movq %rbx, %rdi movq %r14, %rsi callq 0x2a38ae cmpb $0x1, (%r14) jne 0x2a30dd movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x2a3512 lock decl (%rax) jne 0x2a3512 movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x2a350a movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a3512 xorl %r15d, %r15d jmp 0x2a3299 addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x8 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x2a33ed lock decl (%rax) jne 0x2a33ed movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x2a353d movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a33ed addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x4 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x2a33ed lock decl (%rax) jne 0x2a33ed movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x2a353d movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a33ed movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rbx) vmovups %xmm0, (%rbx) vmovups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x2a30dd movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a33ed movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%r12) vmovups %xmm0, (%r12) andl $0x0, 0x60(%rbx) vmovups %xmm0, 0x50(%rbx) vmovups (%rbx,%r15), %xmm0 vmovups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r15), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r15), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r15), %rax movq %rax, 0x48(%rbx) vmovups 0x28(%rbx,%r15), %xmm0 vmovups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r15), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r15), %rax movq %rax, 0x68(%rbx) jmp 0x2a33ed jmp 0x2a361a jmp 0x2a361a movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x2a3639 lock decl (%rax) jne 0x2a3639 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x2a3608 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a3639 jmp 0x2a361a movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x2a3639 lock decl (%rax) jne 0x2a3639 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi jne 0x2a3612 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a3639 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a3639 movq %rax, %rdi callq 0x61d68 jmp 0x2a362c jmp 0x2a362c jmp 0x2a362c jmp 0x2a362c jmp 0x2a362c movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq %rbx, %rdi callq 0x5f340 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
ncnn::ConvolutionDepthWise_x86_avx::create_pipeline_int8_x86(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx::create_pipeline_int8_x86(const Option& opt) { const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { elempack = channels % 8 == 0 ? 8 : 1; } #endif // __SSE2__ if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } if (elempack == 1) { weight_data_tm = weight_data; } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %r15 pushq %r14 pushq %r12 pushq %rbx subq $0x48, %rsp movq %rsi, %r15 movq %rdi, %rbx movq (%rdi), %rax movq -0x18(%rax), %r14 movl 0xd8(%rdi,%r14), %r8d imull 0xd4(%rdi,%r14), %r8d movl 0xd0(%rdi,%r14), %esi movl 0x104(%rdi,%r14), %eax movl 0x108(%rdi,%r14), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x2a3724 imull %ecx, %eax cmpl %ecx, %eax jne 0x2a3724 testb $0x7, %cl jne 0x2a3777 cmpb $0x0, 0x27(%r15) je 0x2a3777 addq %rbx, %r14 addq $0x168, %r14 # imm = 0x168 movq %rsp, %r12 movq %r12, %rdi movq %r14, %rsi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e addq $0x28, %rbx pushq $0x8 popq %rdx movq %r12, %rdi movq %rbx, %rsi movq %r15, %rcx callq 0x64e3b movq 0x8(%rsp), %rax testq %rax, %rax je 0x2a385d lock decl (%rax) jne 0x2a385d movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2a37ec movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a385d movq %rbx, %rdi movq %r15, %rsi callq 0x2a38ae cmpb $0x1, (%r15) jne 0x2a385d movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x2a37c4 lock decl (%rax) jne 0x2a37c4 movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x2a37bc movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a37c4 addq $0x168, %r14 # imm = 0x168 cmpq $0x28, %r14 je 0x2a385d movq 0x8(%rbx,%r14), %rax testq %rax, %rax je 0x2a3795 lock incl (%rax) leaq 0x28(%rbx), %r15 movq 0x30(%rbx), %rax testq %rax, %rax je 0x2a37fe lock decl (%rax) jne 0x2a37fe movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x2a37f6 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a37fe movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rbx) vmovups %xmm0, (%rbx) vmovups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x2a385d movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a385d movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%r15) vmovups %xmm0, (%r15) andl $0x0, 0x60(%rbx) vmovups %xmm0, 0x50(%rbx) vmovups (%rbx,%r14), %xmm0 vmovups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r14), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r14), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r14), %rax movq %rax, 0x48(%rbx) vmovups 0x28(%rbx,%r14), %xmm0 vmovups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r14), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r14), %rax movq %rax, 0x68(%rbx) xorl %eax, %eax addq $0x48, %rsp popq %rbx popq %r12 popq %r14 popq %r15 retq jmp 0x2a38a5 movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2a389d lock decl (%rax) jne 0x2a389d movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2a3897 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a389d movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
ncnn::ConvolutionDepthWise_x86_avx::create_group_ops(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx::create_group_ops(const Option& opt) { // create Convolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Convolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[5]; weights[0] = weight_data_g; weights[1] = bias_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[2] = weight_data_int8_scales_g; weights[3] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[4] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[4]; weights[0] = weight_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[1] = weight_data_int8_scales_g; weights[2] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[3] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x278, %rsp # imm = 0x278 movq %rsi, 0x250(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x104(%rdi,%rdx), %eax movl 0x108(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x10(%rdi), %rax movq %rax, 0x1e0(%rsp) xorl %r12d, %r12d movq 0x10(%r14), %rax movq 0x18(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x2a394b movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x2a3946 movq (%rdi), %rax callq 0x8(%rax) incq %r12 jmp 0x2a391d imull %r15d, %ebx cmpq %rax, %rcx je 0x2a3958 movq %rax, 0x18(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x108(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0x1e0(%rsp), %rdi callq 0x6fbc2 leaq 0x118(%r14), %rax movq %rax, 0x258(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0x6c(%rsp) movslq %ebx, %rax movq %rax, 0x260(%rsp) movslq %r15d, %rax movq %rax, 0x248(%rsp) pushq $0x1 popq %rbx xorl %edx, %edx movl %r15d, 0x14(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rcx cmpq %rcx, %rdx jge 0x2a46c2 movq %rdx, %rcx movq %rdx, 0x8(%rsp) movq 0x260(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdx, 0x80(%rsp) movl %esi, 0x88(%rsp) movq %rax, 0x90(%rsp) movl %ebx, 0x98(%rsp) movl %edi, 0x9c(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa0(%rsp) movl %ebx, 0xa8(%rsp) movq %rdi, 0xb0(%rsp) leaq 0x200(%rsp), %rdi leaq 0x70(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a3aa9 lock decl (%rax) jne 0x2a3aa9 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a3aa1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a3aa9 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x2a3b19 movq 0x8(%rsp), %r15 movq 0x248(%rsp), %rcx imulq %rcx, %r15 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0x1f8(%rsp) imulq %rsi, %r15 addq 0x1b0(%r14,%rax), %r15 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x1c(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0x1f0(%rsp) movl %ebx, %r13d movl 0x14(%rsp), %eax movl %eax, 0x18(%rsp) movq %rcx, 0x1e8(%rsp) jmp 0x2a3b53 xorl %r15d, %r15d movq $0x0, 0x1f8(%rsp) movl $0x0, 0x1c(%rsp) movq $0x0, 0x1f0(%rsp) xorl %r13d, %r13d movl $0x0, 0x18(%rsp) movq $0x0, 0x1e8(%rsp) pushq $0x6 popq %rdi callq 0x782bf movq %rax, %rbp leaq 0x268(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq %r12, %rdi xorl %esi, %esi movl 0x14(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r12, %rdi movl %ebx, %esi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r12, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r12, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r12, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r12, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r12, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r12, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r12, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x6 popq %rsi movl 0x6c(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r12, %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x110(%r14,%rax), %edx movq %r12, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x258(%rsp), %rdx movq %r12, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r12, %rsi callq 0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x2a3d40 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x1a8, %rax # imm = 0x1A8 jne 0x2a3cce movq 0x208(%rsp), %rax testq %rax, %rax je 0x2a3d08 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a3dc1 lock decl (%rax) jne 0x2a3dc1 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a3db9 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a3dc1 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x160, %rax # imm = 0x160 jne 0x2a3d47 movq 0x208(%rsp), %rax testq %rax, %rax je 0x2a3d81 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a4178 lock decl (%rax) jne 0x2a4178 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a4170 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a4178 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x2a3e65 lock decl (%rax) jne 0x2a3e65 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x2a3e5d movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a3e65 movq %rsi, %rdi callq 0x5f3e0 movq %r15, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq 0x1f8(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x1c(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x1f0(%rsp), %rax movq %rax, 0xd8(%rsp) movl %r13d, 0xe0(%rsp) movl 0x18(%rsp), %eax movl %eax, 0xe4(%rsp) movl %r13d, 0xe8(%rsp) movl %r13d, 0xec(%rsp) movl %r13d, 0xf0(%rsp) movq 0x1e8(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2a40fa andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x2a3f6a vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x2a3f5c movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a3f77 lock incl (%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x2a3fae lock decl (%rax) jne 0x2a3fae movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x2a3fa6 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a3fae movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovaps %xmm0, 0x100(%rsp) movq 0x30(%rsp), %rax movq %rax, 0x110(%rsp) movl 0x38(%rsp), %eax movl %eax, 0x118(%rsp) movq 0x40(%rsp), %rax movq %rax, 0x120(%rsp) vmovups 0x48(%rsp), %xmm0 vmovups %xmm0, 0x128(%rsp) movl 0x58(%rsp), %eax movl %eax, 0x138(%rsp) movq 0x60(%rsp), %rax movq %rax, 0x140(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %r12d movq 0x260(%r14,%rax), %rbx movq 0x150(%rsp), %rax testq %rax, %rax je 0x2a4070 lock decl (%rax) jne 0x2a4070 movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x2a4068 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a4070 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %r12d, 0x160(%rsp) movq %rbx, 0x168(%rsp) vbroadcastss 0x15455e(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x2a40f3 lock decl (%rax) jne 0x2a40f3 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2a40eb movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a40f3 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x2a431a movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x198(%rsp), %rax testq %rax, %rax je 0x2a42c6 lock decl (%rax) jne 0x2a42c6 movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x2a42be movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a42c6 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2a44fd andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x2a4273 vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x2a4265 movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a4280 lock incl (%rax) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x2a43b2 lock decl (%rax) jne 0x2a43b2 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x2a43aa movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a43b2 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x190(%rsp) andq $0x0, 0x198(%rsp) movq %r15, 0x1a0(%rsp) movl %ebx, 0x1a8(%rsp) movq %r12, 0x1b0(%rsp) vbroadcastss 0x154309(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x1b8(%rsp) movl $0x1, 0x1c8(%rsp) movq $0x1, 0x1d0(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0x120, %r15d # imm = 0x120 movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x2a4378 lock decl (%rax) jne 0x2a4378 movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x2a4370 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a4378 movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x2a4347 jmp 0x2a464b movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovups %xmm0, 0xb8(%rsp) movq 0x30(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x38(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x40(%rsp), %rax movq %rax, 0xd8(%rsp) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movl 0x58(%rsp), %eax movl %eax, 0xf0(%rsp) movq 0x60(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %ebx movq 0x260(%r14,%rax), %r12 movq 0x108(%rsp), %rax testq %rax, %rax je 0x2a4474 lock decl (%rax) jne 0x2a4474 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x2a446c movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a4474 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x100(%rsp) andq $0x0, 0x108(%rsp) movq %r15, 0x110(%rsp) movl %ebx, 0x118(%rsp) movq %r12, 0x120(%rsp) vbroadcastss 0x15415b(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x128(%rsp) movl $0x1, 0x138(%rsp) movq $0x1, 0x140(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x2a44f6 lock decl (%rax) jne 0x2a44f6 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2a44ee movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a44f6 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x2a45c0 movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x150(%rsp), %rax testq %rax, %rax je 0x2a456c lock decl (%rax) jne 0x2a456c movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x2a4564 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a456c movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %ebx, 0x160(%rsp) movq %r12, 0x168(%rsp) vbroadcastss 0x154063(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0xd8, %r15d movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x2a461e lock decl (%rax) jne 0x2a461e movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x2a4616 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a461e movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x2a45ed movq (%rbp), %rax movq %rbp, %rdi movq 0x250(%rsp), %rsi callq 0x20(%rax) leaq 0x268(%rsp), %rdi movq 0x1e0(%rsp), %rax movq (%rax), %rax movq 0x8(%rsp), %rcx movq %rbp, (%rax,%rcx,8) callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x2a46b5 lock decl (%rax) jne 0x2a46b5 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi je 0x2a46ad movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a46b5 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rdx incq %rdx jmp 0x2a39c3 xorl %eax, %eax addq $0x278, %rsp # imm = 0x278 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2a478a jmp 0x2a4921 jmp 0x2a46e2 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a479c lock decl (%rax) jne 0x2a479c movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2a4718 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a479c movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a479c jmp 0x2a4921 jmp 0x2a4814 jmp 0x2a478a jmp 0x2a472e movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a4826 lock decl (%rax) jne 0x2a4826 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2a4764 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a4826 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a4826 jmp 0x2a4921 jmp 0x2a4921 jmp 0x2a4921 jmp 0x2a4921 jmp 0x2a478a jmp 0x2a4814 movq %rax, %rbx jmp 0x2a479c movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0xd8, %r14d vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x2a47df lock decl (%rax) jne 0x2a47df movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2a47d3 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2a47df movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2a47a6 jmp 0x2a48d5 jmp 0x2a4921 jmp 0x2a4814 movq %rax, %rbx jmp 0x2a4826 movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0x120, %r14d # imm = 0x120 vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x2a4869 lock decl (%rax) jne 0x2a4869 movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2a485d movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2a4869 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2a4830 jmp 0x2a48d5 jmp 0x2a4921 jmp 0x2a48c9 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a4919 lock decl (%rax) jne 0x2a4919 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a4909 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a4919 jmp 0x2a4921 movq %rax, %rbx jmp 0x2a48e2 jmp 0x2a4921 jmp 0x2a4921 movq %rax, %rbx leaq 0x268(%rsp), %rdi callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x2a4919 lock decl (%rax) jne 0x2a4919 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi jne 0x2a4913 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a4919 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_avx::create_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }	pushq %rax movq (%rdi), %rax addq -0x30(%rax), %rdi callq 0x2a30be xorl %eax, %eax popq %rcx retq nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
ncnn::ConvolutionDepthWise_x86_avx::destroy_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }	pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x2a496d movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x2a4968 movq (%rdi), %rax callq 0x8(%rax) andq $0x0, 0x8(%rbx) xorl %r15d, %r15d movq 0x10(%rbx), %rax movq 0x18(%rbx), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r15 jge 0x2a49af movq (%rax,%r15,8), %rdi movq (%rdi), %rax movq %r14, %rsi callq 0x28(%rax) movq 0x10(%rbx), %rax movq (%rax,%r15,8), %rdi testq %rdi, %rdi je 0x2a49aa movq (%rdi), %rax callq 0x8(%rax) incq %r15 jmp 0x2a4970 cmpq %rax, %rcx je 0x2a49b8 movq %rax, 0x18(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %r15 retq	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_avx::destroy_pipeline(ncnn::Option const&)	int ConvolutionDepthWise_x86_avx::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }	pushq %rax movq (%rdi), %rax addq -0x38(%rax), %rdi callq 0x2a493c xorl %eax, %eax popq %rcx retq nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
ncnn::ConvolutionDepthWise_x86_avx::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float)bias_data) + g 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float)bias_data) + g 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float)bias_data) + g 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x3b8, %rsp # imm = 0x3B8 movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdi movq %rcx, 0x48(%rsp) cmpb $0x1, 0x1e(%rcx) jne 0x2a4a21 cmpl $0x0, 0x10c(%r14,%rdi) je 0x2a4a21 movq %r14, %rdi movq 0x48(%rsp), %rcx addq $0x3b8, %rsp # imm = 0x3B8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x2a98c2 movq %rdx, 0x18(%rsp) movl 0x38(%rsi), %ecx movq %rcx, 0xf0(%rsp) movq 0x10(%rsi), %r13 movslq 0x18(%rsi), %r15 movl 0xd4(%r14,%rdi), %ebx decl %ebx imull 0xdc(%r14,%rdi), %ebx movl 0xd8(%r14,%rdi), %ebp decl %ebp imull 0xe0(%r14,%rdi), %ebp leaq 0x90(%rsp), %rdx andq $0x0, 0x40(%rdx) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rdx) vmovups %xmm0, 0xc(%rdx) vmovaps %xmm0, 0x20(%rdx) vmovups %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %r14, %rdi movq 0x48(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0x90(%rsp) je 0x2a96d0 movslq 0xc8(%rsp), %rax imulq 0xd0(%rsp), %rax testq %rax, %rax je 0x2a96d0 notl %ebx notl %ebp movl 0xbc(%rsp), %edi addl %edi, %ebx movq (%r14), %rax movq -0x18(%rax), %rcx movl %ebx, %eax cltd idivl 0xe4(%r14,%rcx) movl %eax, %ebx leal 0x1(%rbx), %esi addl 0xc0(%rsp), %ebp movl %ebp, %eax cltd idivl 0xe8(%r14,%rcx) movq %rax, 0x50(%rsp) leal 0x1(%rax), %r8d movl 0xd0(%r14,%rcx), %ecx movq 0x48(%rsp), %rax cmpb $0x1, 0x27(%rax) pushq $0x8 popq %rbp movl %edi, 0x20(%rsp) jne 0x2a4b30 xorl %eax, %eax testb $0x3, %cl sete %al testb $0x7, %cl leal 0x1(%rax,%rax,2), %r9d cmovel %ebp, %r9d jmp 0x2a4b34 pushq $0x1 popq %r9 movq %r13, %rax xorl %edx, %edx divq %r15 movq %rax, %r10 movl %r9d, %eax movq %rax, 0x70(%rsp) imulq %rax, %r10 movl %ecx, %eax cltd idivl %r9d movq 0x48(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq 0x18(%rsp), %r13 movq %r13, %rdi movl %esi, 0x110(%rsp) movl %r8d, 0x130(%rsp) movl %r8d, %edx movl %eax, %ecx movq %r10, 0x170(%rsp) movq %r10, %r8 callq 0x628f2 movq (%r13), %r9 testq %r9, %r9 je 0x2a96d0 movq 0x40(%r13), %r11 movslq 0x38(%r13), %rax imulq %r11, %rax testq %rax, %rax je 0x2a96d0 movq %r15, %r10 movl %r10d, %eax imull 0xf0(%rsp), %eax movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xd0(%r14,%rdi), %ecx movl 0x108(%r14,%rdi), %esi cmpl %esi, %eax jne 0x2a75fb cmpl %ecx, %eax jne 0x2a75fb addq %r14, %rdi cmpl $0x1, %r10d je 0x2a6634 movslq %ebx, %rbx cmpl $0x4, %r10d movq %r14, 0x10(%rsp) je 0x2a59a4 movl %eax, %ecx cmpl $0x8, %r10d jne 0x2a75fb movl 0xd4(%rdi), %eax movl 0xd8(%rdi), %r15d cmpl $0x5, %eax je 0x2a70c6 cmpl $0x3, %eax jne 0x2a93b2 cmpl $0x3, %r15d jne 0x2a93b2 cmpl $0x1, 0xdc(%rdi) jne 0x2a8abf cmpl $0x1, 0xe0(%rdi) jne 0x2a8abf cmpl $0x1, 0xe4(%rdi) jne 0x2a8abf cmpl $0x1, 0xe8(%rdi) jne 0x2a8abf movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xbc(%rsp), %edx movl 0xc8(%rsp), %esi movq 0x1b0(%rdi), %r8 leal 0x10(,%rdx,8), %edx movslq %edx, %rdi leal (,%rax,8), %edx movslq %edx, %rdx xorl %r9d, %r9d testl %esi, %esi cmovlel %r9d, %esi movq %rsi, 0x160(%rsp) shlq $0x2, %rdx movq %rdx, 0x190(%rsp) shlq $0x2, %rdi movq %r8, 0x150(%rsp) cmpq 0x160(%rsp), %r9 je 0x2a9725 testq %r8, %r8 je 0x2a4ce5 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r8,%rdx), %ymm15 jmp 0x2a4cea vxorps %xmm15, %xmm15, %xmm15 movq 0x40(%r13), %r10 imulq %r9, %r10 movq 0x10(%r13), %rsi imulq %rsi, %r10 addq (%r13), %r10 movslq 0x2c(%r13), %rdx movslq 0x54(%r14), %r11 imulq %r9, %r11 imulq 0x38(%r14), %r11 movq 0x28(%r14), %rbx imulq %rsi, %rdx addq %r10, %rdx movslq 0xbc(%rsp), %rsi movq 0xd0(%rsp), %r15 movq %r9, 0xe0(%rsp) imulq %r9, %r15 movq 0xa0(%rsp), %r8 imulq %r8, %r15 addq 0x90(%rsp), %r15 imulq %r8, %rsi leaq (%r15,%rsi), %r12 leaq (%r15,%rsi,2), %r13 leaq (%rsi,%rsi,2), %rsi addq %r15, %rsi xorl %ebp, %ebp vmovups %ymm15, 0x20(%rsp) movl %ebp, %r8d orl $0x1, %r8d cmpl %ecx, %r8d jge 0x2a597f xorl %r9d, %r9d xorl %r8d, %r8d leal 0x3(%r8), %r14d cmpl %eax, %r14d jge 0x2a53c6 vmovaps 0x20(%r15,%r9), %ymm0 vmovaps 0x40(%r15,%r9), %ymm2 vmovaps 0x60(%r15,%r9), %ymm6 vmovaps 0x80(%r15,%r9), %ymm7 vmovaps (%rbx,%r11), %ymm3 vmovaps 0x20(%rbx,%r11), %ymm5 vmovaps 0x40(%rbx,%r11), %ymm4 vmovaps 0x60(%rbx,%r11), %ymm1 vmulps (%r15,%r9), %ymm3, %ymm8 vaddps %ymm15, %ymm8, %ymm8 vmulps %ymm3, %ymm0, %ymm9 vaddps %ymm15, %ymm9, %ymm9 vmulps %ymm3, %ymm2, %ymm10 vaddps %ymm15, %ymm10, %ymm10 vmulps %ymm3, %ymm6, %ymm11 vaddps %ymm15, %ymm11, %ymm11 vmulps %ymm5, %ymm0, %ymm0 vmulps %ymm5, %ymm2, %ymm12 vmulps %ymm5, %ymm6, %ymm13 vmulps %ymm5, %ymm7, %ymm14 vmulps %ymm4, %ymm2, %ymm2 vaddps %ymm2, %ymm0, %ymm0 vaddps %ymm0, %ymm8, %ymm0 vmovups %ymm0, 0xf0(%rsp) vmulps %ymm4, %ymm6, %ymm0 vaddps %ymm0, %ymm12, %ymm0 vaddps %ymm0, %ymm9, %ymm0 vmovups %ymm0, 0x110(%rsp) vmulps %ymm4, %ymm7, %ymm0 vaddps %ymm0, %ymm13, %ymm0 vmulps 0xa0(%r15,%r9), %ymm4, %ymm2 vaddps %ymm0, %ymm10, %ymm0 vmovups %ymm0, 0x1f0(%rsp) vaddps %ymm2, %ymm14, %ymm0 vaddps %ymm0, %ymm11, %ymm0 vmovups %ymm0, 0x130(%rsp) vmovaps 0x80(%rbx,%r11), %ymm2 vmovaps (%r12,%r9), %ymm11 vmovaps 0x20(%r12,%r9), %ymm7 vmovaps 0x40(%r12,%r9), %ymm15 vmovaps 0x60(%r12,%r9), %ymm14 vmulps %ymm1, %ymm11, %ymm0 vmulps %ymm1, %ymm7, %ymm6 vmulps %ymm2, %ymm7, %ymm8 vaddps %ymm0, %ymm8, %ymm9 vmulps %ymm2, %ymm15, %ymm0 vaddps %ymm0, %ymm6, %ymm8 vmulps %ymm1, %ymm15, %ymm0 vmulps %ymm2, %ymm14, %ymm6 vaddps %ymm6, %ymm0, %ymm10 vmovaps 0x80(%r12,%r9), %ymm0 vmulps %ymm1, %ymm14, %ymm6 vmulps %ymm2, %ymm0, %ymm13 vaddps %ymm6, %ymm13, %ymm6 vmulps %ymm5, %ymm7, %ymm13 vmulps %ymm4, %ymm15, %ymm12 vaddps %ymm12, %ymm13, %ymm12 vmulps %ymm3, %ymm11, %ymm11 vaddps 0x20(%rsp), %ymm11, %ymm11 vaddps %ymm12, %ymm11, %ymm11 vmovups %ymm11, 0x50(%rsp) vmulps %ymm5, %ymm15, %ymm11 vmulps %ymm4, %ymm14, %ymm12 vaddps %ymm12, %ymm11, %ymm11 vmulps %ymm3, %ymm7, %ymm7 vaddps 0x20(%rsp), %ymm7, %ymm7 vaddps %ymm7, %ymm11, %ymm7 vmovups %ymm7, 0x70(%rsp) vmulps %ymm5, %ymm14, %ymm11 vmulps %ymm4, %ymm0, %ymm12 vaddps %ymm12, %ymm11, %ymm11 vmulps %ymm3, %ymm15, %ymm12 vaddps 0x20(%rsp), %ymm12, %ymm12 vaddps %ymm11, %ymm12, %ymm7 vmovups %ymm7, 0x170(%rsp) vmulps %ymm5, %ymm0, %ymm5 vmovaps 0xa0(%r12,%r9), %ymm12 vmulps %ymm4, %ymm12, %ymm4 vaddps %ymm4, %ymm5, %ymm4 vmulps %ymm3, %ymm14, %ymm3 vaddps 0x20(%rsp), %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vmovups %ymm3, 0x270(%rsp) vmovaps 0xa0(%rbx,%r11), %ymm13 vmulps %ymm13, %ymm15, %ymm4 vaddps %ymm4, %ymm9, %ymm4 vaddps 0xf0(%rsp), %ymm4, %ymm3 vmovups %ymm3, 0xf0(%rsp) vmulps %ymm13, %ymm14, %ymm5 vaddps %ymm5, %ymm8, %ymm5 vaddps 0x110(%rsp), %ymm5, %ymm3 vmovups %ymm3, 0x110(%rsp) vmulps %ymm0, %ymm13, %ymm0 vaddps %ymm0, %ymm10, %ymm0 vaddps 0x1f0(%rsp), %ymm0, %ymm0 vmovups %ymm0, 0x1f0(%rsp) vmulps %ymm13, %ymm12, %ymm0 vaddps %ymm0, %ymm6, %ymm0 vaddps 0x130(%rsp), %ymm0, %ymm0 vmovups %ymm0, 0x130(%rsp) vmovaps (%r13,%r9), %ymm6 vmovaps 0x20(%r13,%r9), %ymm9 vmovaps 0x40(%r13,%r9), %ymm10 vmulps %ymm1, %ymm6, %ymm12 vmulps %ymm2, %ymm9, %ymm14 vaddps %ymm14, %ymm12, %ymm12 vmulps %ymm1, %ymm9, %ymm14 vmulps %ymm2, %ymm10, %ymm15 vaddps %ymm15, %ymm14, %ymm8 vmovaps 0x60(%r13,%r9), %ymm14 vmulps %ymm1, %ymm10, %ymm15 vmulps %ymm2, %ymm14, %ymm7 vaddps %ymm7, %ymm15, %ymm7 vmulps %ymm1, %ymm14, %ymm1 vmovaps 0x80(%r13,%r9), %ymm15 vmulps %ymm2, %ymm15, %ymm2 vaddps %ymm2, %ymm1, %ymm5 vmovaps 0xc0(%rbx,%r11), %ymm1 vmulps %ymm1, %ymm6, %ymm6 vmovaps 0xe0(%rbx,%r11), %ymm2 vmulps %ymm2, %ymm9, %ymm11 vaddps %ymm6, %ymm11, %ymm11 vmulps %ymm1, %ymm9, %ymm6 vmulps %ymm2, %ymm10, %ymm9 vaddps %ymm6, %ymm9, %ymm0 vmulps %ymm1, %ymm10, %ymm6 vmulps %ymm2, %ymm14, %ymm9 vaddps %ymm6, %ymm9, %ymm3 vmulps %ymm1, %ymm14, %ymm6 vmulps %ymm2, %ymm15, %ymm9 vaddps %ymm6, %ymm9, %ymm4 vmulps %ymm13, %ymm10, %ymm6 vaddps %ymm6, %ymm12, %ymm6 vaddps 0x50(%rsp), %ymm6, %ymm6 vmovups %ymm6, 0x50(%rsp) vmulps %ymm13, %ymm14, %ymm9 vaddps %ymm9, %ymm8, %ymm8 vaddps 0x70(%rsp), %ymm8, %ymm9 vmulps %ymm13, %ymm15, %ymm8 vaddps %ymm7, %ymm8, %ymm7 vaddps 0x170(%rsp), %ymm7, %ymm7 vmovaps 0xa0(%r13,%r9), %ymm8 vmulps %ymm13, %ymm8, %ymm12 vaddps %ymm5, %ymm12, %ymm5 vaddps 0x270(%rsp), %ymm5, %ymm13 vmovaps 0x100(%rbx,%r11), %ymm12 vmulps %ymm12, %ymm10, %ymm5 vaddps %ymm5, %ymm11, %ymm5 vaddps 0xf0(%rsp), %ymm5, %ymm10 vmulps %ymm12, %ymm14, %ymm5 vaddps %ymm5, %ymm0, %ymm0 vaddps 0x110(%rsp), %ymm0, %ymm5 vmulps %ymm12, %ymm15, %ymm0 vaddps %ymm0, %ymm3, %ymm0 vaddps 0x1f0(%rsp), %ymm0, %ymm3 vmulps %ymm12, %ymm8, %ymm0 vaddps %ymm0, %ymm4, %ymm0 vaddps 0x130(%rsp), %ymm0, %ymm0 vmovaps 0x20(%rsi,%r9), %ymm4 vmovaps 0x40(%rsi,%r9), %ymm8 vmulps (%rsi,%r9), %ymm1, %ymm11 vmovaps 0x60(%rsi,%r9), %ymm14 vmulps %ymm1, %ymm4, %ymm15 vmulps %ymm2, %ymm4, %ymm4 vaddps %ymm4, %ymm11, %ymm4 vmulps %ymm2, %ymm8, %ymm11 vaddps %ymm11, %ymm15, %ymm11 vmulps %ymm1, %ymm8, %ymm15 vmulps %ymm2, %ymm14, %ymm6 vaddps %ymm6, %ymm15, %ymm6 vmulps %ymm1, %ymm14, %ymm1 vmovaps 0x80(%rsi,%r9), %ymm15 vmulps %ymm2, %ymm15, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps %ymm12, %ymm8, %ymm2 vaddps %ymm2, %ymm4, %ymm2 vaddps 0x50(%rsp), %ymm2, %ymm2 vmulps %ymm12, %ymm14, %ymm4 vaddps %ymm4, %ymm11, %ymm4 vaddps %ymm4, %ymm9, %ymm4 vmulps %ymm12, %ymm15, %ymm8 vmovups 0x20(%rsp), %ymm15 vaddps %ymm6, %ymm8, %ymm6 vaddps %ymm6, %ymm7, %ymm6 vmulps 0xa0(%rsi,%r9), %ymm12, %ymm7 vaddps %ymm7, %ymm1, %ymm1 vaddps %ymm1, %ymm13, %ymm1 vmovaps %ymm10, (%r10,%r9) vmovaps %ymm5, 0x20(%r10,%r9) vmovaps %ymm3, 0x40(%r10,%r9) vmovaps %ymm0, 0x60(%r10,%r9) vmovaps %ymm2, (%rdx,%r9) vmovaps %ymm4, 0x20(%rdx,%r9) vmovaps %ymm6, 0x40(%rdx,%r9) vmovaps %ymm1, 0x60(%rdx,%r9) addl $0x4, %r8d subq $-0x80, %r9 jmp 0x2a4d7b vmovaps 0x20(%r15,%r9), %ymm1 vmovaps 0x40(%r15,%r9), %ymm3 vmovaps (%rbx,%r11), %ymm5 vmovaps 0x20(%rbx,%r11), %ymm0 vmovaps 0x40(%rbx,%r11), %ymm4 vmulps (%r15,%r9), %ymm5, %ymm6 vmovaps 0x60(%rbx,%r11), %ymm2 vaddps %ymm6, %ymm15, %ymm6 vmulps %ymm5, %ymm1, %ymm7 vaddps %ymm7, %ymm15, %ymm7 vmulps %ymm0, %ymm1, %ymm1 vmulps %ymm0, %ymm3, %ymm8 vmulps %ymm4, %ymm3, %ymm3 vaddps %ymm3, %ymm1, %ymm1 vaddps %ymm1, %ymm6, %ymm6 vmulps 0x60(%r15,%r9), %ymm4, %ymm1 vaddps %ymm1, %ymm8, %ymm1 vaddps %ymm1, %ymm7, %ymm7 vmovaps 0x80(%rbx,%r11), %ymm3 vmovaps 0xa0(%rbx,%r11), %ymm1 vmovaps (%r12,%r9), %ymm8 vmovaps 0x20(%r12,%r9), %ymm9 vmovaps 0x40(%r12,%r9), %ymm10 vmovaps 0x60(%r12,%r9), %ymm11 vmulps %ymm2, %ymm8, %ymm12 vmulps %ymm2, %ymm9, %ymm13 vmulps %ymm5, %ymm8, %ymm8 vaddps %ymm15, %ymm8, %ymm8 vmulps %ymm5, %ymm9, %ymm5 vaddps %ymm5, %ymm15, %ymm5 vmulps %ymm3, %ymm9, %ymm14 vaddps %ymm14, %ymm12, %ymm12 vmulps %ymm3, %ymm10, %ymm14 vaddps %ymm14, %ymm13, %ymm13 vmulps %ymm0, %ymm9, %ymm9 vmulps %ymm0, %ymm10, %ymm14 vmulps %ymm1, %ymm10, %ymm0 vaddps %ymm0, %ymm12, %ymm0 vaddps %ymm0, %ymm6, %ymm0 vmulps %ymm1, %ymm11, %ymm6 vaddps %ymm6, %ymm13, %ymm6 vaddps %ymm6, %ymm7, %ymm6 vmulps %ymm4, %ymm10, %ymm7 vaddps %ymm7, %ymm9, %ymm7 vaddps %ymm7, %ymm8, %ymm7 vmulps %ymm4, %ymm11, %ymm4 vaddps %ymm4, %ymm14, %ymm4 vaddps %ymm4, %ymm5, %ymm8 vmovaps 0xc0(%rbx,%r11), %ymm5 vmovaps 0xe0(%rbx,%r11), %ymm4 vmovaps (%r13,%r9), %ymm9 vmovaps 0x20(%r13,%r9), %ymm10 vmovaps 0x40(%r13,%r9), %ymm11 vmulps %ymm5, %ymm9, %ymm12 vmulps %ymm5, %ymm10, %ymm13 vmulps %ymm4, %ymm10, %ymm14 vaddps %ymm14, %ymm12, %ymm12 vmulps %ymm4, %ymm11, %ymm14 vaddps %ymm14, %ymm13, %ymm13 vmulps %ymm2, %ymm9, %ymm9 vmulps %ymm2, %ymm10, %ymm2 vmulps %ymm3, %ymm10, %ymm10 vaddps %ymm10, %ymm9, %ymm9 vmovaps 0x100(%rbx,%r11), %ymm10 vmulps %ymm3, %ymm11, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps %ymm10, %ymm11, %ymm3 vaddps %ymm3, %ymm12, %ymm3 vmovaps 0x60(%r13,%r9), %ymm12 vaddps %ymm3, %ymm0, %ymm0 vmulps %ymm10, %ymm12, %ymm3 vaddps %ymm3, %ymm13, %ymm3 vaddps %ymm3, %ymm6, %ymm3 vmulps %ymm1, %ymm11, %ymm6 vaddps %ymm6, %ymm9, %ymm6 vaddps %ymm6, %ymm7, %ymm6 vmulps %ymm1, %ymm12, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vaddps %ymm1, %ymm8, %ymm1 vmulps (%rsi,%r9), %ymm5, %ymm2 vmovaps 0x20(%rsi,%r9), %ymm7 vmulps %ymm5, %ymm7, %ymm5 vmulps %ymm4, %ymm7, %ymm7 vaddps %ymm7, %ymm2, %ymm2 vmovaps 0x40(%rsi,%r9), %ymm7 vmulps %ymm4, %ymm7, %ymm4 vaddps %ymm4, %ymm5, %ymm4 vmulps %ymm7, %ymm10, %ymm5 vaddps %ymm5, %ymm2, %ymm2 vaddps %ymm2, %ymm6, %ymm2 vmulps 0x60(%rsi,%r9), %ymm10, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vaddps %ymm4, %ymm1, %ymm1 vmovaps %ymm0, (%r10,%r9) vmovaps %ymm3, 0x20(%r10,%r9) vmovaps %ymm2, (%rdx,%r9) vmovaps %ymm1, 0x20(%rdx,%r9) addl $0x2, %r8d addq $0x40, %r9 leal 0x1(%r8), %r14d cmpl %eax, %r14d jl 0x2a51d2 movq 0x10(%rsp), %r14 cmpl %eax, %r8d jge 0x2a5510 vmovaps (%rbx,%r11), %ymm0 vmovaps 0x20(%rbx,%r11), %ymm1 vmovaps 0x40(%rbx,%r11), %ymm2 vmulps (%r15,%r9), %ymm0, %ymm3 vmovaps 0x60(%rbx,%r11), %ymm4 vaddps %ymm3, %ymm15, %ymm3 vmulps 0x20(%r15,%r9), %ymm1, %ymm5 vmulps 0x40(%r15,%r9), %ymm2, %ymm6 vaddps %ymm6, %ymm5, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmovaps 0x80(%rbx,%r11), %ymm5 vmovaps 0xa0(%rbx,%r11), %ymm6 vmovaps (%r12,%r9), %ymm7 vmovaps 0x20(%r12,%r9), %ymm8 vmovaps 0x40(%r12,%r9), %ymm9 vmulps %ymm4, %ymm7, %ymm10 vmulps %ymm0, %ymm7, %ymm0 vaddps %ymm0, %ymm15, %ymm0 vmulps %ymm5, %ymm8, %ymm7 vaddps %ymm7, %ymm10, %ymm7 vmulps %ymm1, %ymm8, %ymm1 vmulps %ymm6, %ymm9, %ymm8 vaddps %ymm7, %ymm8, %ymm7 vaddps %ymm7, %ymm3, %ymm3 vmulps %ymm2, %ymm9, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps 0xc0(%rbx,%r11), %ymm1 vmovaps 0xe0(%rbx,%r11), %ymm2 vmovaps 0x100(%rbx,%r11), %ymm7 vmovaps (%r13,%r9), %ymm8 vmovaps 0x20(%r13,%r9), %ymm9 vmovaps 0x40(%r13,%r9), %ymm10 vmulps %ymm1, %ymm8, %ymm11 vmulps %ymm4, %ymm8, %ymm4 vmulps %ymm2, %ymm9, %ymm8 vaddps %ymm8, %ymm11, %ymm8 vmulps %ymm5, %ymm9, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps %ymm7, %ymm10, %ymm5 vaddps %ymm5, %ymm8, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmulps %ymm6, %ymm10, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps (%rsi,%r9), %ymm1, %ymm1 vaddps %ymm4, %ymm0, %ymm0 vmulps 0x20(%rsi,%r9), %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x40(%rsi,%r9), %ymm7, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps %ymm3, (%r10,%r9) vmovaps %ymm0, (%rdx,%r9) incl %r8d addq $0x20, %r9 jmp 0x2a53d8 addq %rdi, %r15 addq %r9, %r15 addq %rdi, %r12 addq %r9, %r12 addq %rdi, %r13 addq %r9, %r13 addq %rdi, %rsi addq %r9, %rsi movq 0x190(%rsp), %r8 addq %r8, %r10 addq %r9, %r10 addq %r8, %rdx addq %r9, %rdx addl $0x2, %ebp jmp 0x2a4d65 xorl %edx, %edx xorl %esi, %esi leal 0x3(%rsi), %r8d cmpl %eax, %r8d jge 0x2a5891 vmovaps 0x20(%r15,%rdx), %ymm1 vmovaps 0x40(%r15,%rdx), %ymm2 vmovaps 0x60(%r15,%rdx), %ymm3 vmovaps 0x80(%r15,%rdx), %ymm4 vmovaps (%rbx,%r11), %ymm5 vmovaps 0x20(%rbx,%r11), %ymm6 vmovaps 0x40(%rbx,%r11), %ymm7 vmulps (%r15,%rdx), %ymm5, %ymm8 vmovaps 0x60(%rbx,%r11), %ymm0 vaddps %ymm15, %ymm8, %ymm8 vmulps %ymm5, %ymm1, %ymm9 vaddps %ymm15, %ymm9, %ymm9 vmulps %ymm5, %ymm2, %ymm10 vaddps %ymm15, %ymm10, %ymm10 vmulps %ymm5, %ymm3, %ymm5 vaddps %ymm5, %ymm15, %ymm5 vmulps %ymm6, %ymm1, %ymm1 vmulps %ymm6, %ymm2, %ymm11 vmulps %ymm6, %ymm3, %ymm12 vmulps %ymm6, %ymm4, %ymm6 vmulps %ymm7, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm8, %ymm1 vmulps %ymm7, %ymm3, %ymm2 vaddps %ymm2, %ymm11, %ymm2 vaddps %ymm2, %ymm9, %ymm2 vmulps %ymm7, %ymm4, %ymm3 vaddps %ymm3, %ymm12, %ymm3 vaddps %ymm3, %ymm10, %ymm3 vmulps 0xa0(%r15,%rdx), %ymm7, %ymm4 vaddps %ymm4, %ymm6, %ymm4 vaddps %ymm4, %ymm5, %ymm4 vmovaps 0x80(%rbx,%r11), %ymm5 vmovaps 0xa0(%rbx,%r11), %ymm6 vmovaps 0x20(%r12,%rdx), %ymm7 vmovaps 0x40(%r12,%rdx), %ymm8 vmovaps 0x60(%r12,%rdx), %ymm9 vmulps (%r12,%rdx), %ymm0, %ymm10 vmovaps 0x80(%r12,%rdx), %ymm11 vmulps %ymm0, %ymm7, %ymm12 vmulps %ymm0, %ymm8, %ymm13 vmulps %ymm0, %ymm9, %ymm0 vmulps %ymm5, %ymm7, %ymm7 vaddps %ymm7, %ymm10, %ymm7 vmulps %ymm5, %ymm8, %ymm10 vaddps %ymm10, %ymm12, %ymm10 vmulps %ymm5, %ymm9, %ymm12 vaddps %ymm12, %ymm13, %ymm12 vmulps %ymm5, %ymm11, %ymm5 vaddps %ymm5, %ymm0, %ymm0 vmulps %ymm6, %ymm8, %ymm5 vaddps %ymm5, %ymm7, %ymm5 vaddps %ymm5, %ymm1, %ymm1 vmulps %ymm6, %ymm9, %ymm5 vaddps %ymm5, %ymm10, %ymm5 vaddps %ymm5, %ymm2, %ymm2 vmulps %ymm6, %ymm11, %ymm5 vaddps %ymm5, %ymm12, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmulps 0xa0(%r12,%rdx), %ymm6, %ymm5 vaddps %ymm5, %ymm0, %ymm0 vaddps %ymm0, %ymm4, %ymm0 vmovaps 0xc0(%rbx,%r11), %ymm4 vmovaps 0xe0(%rbx,%r11), %ymm5 vmovaps 0x100(%rbx,%r11), %ymm6 vmovaps 0x20(%r13,%rdx), %ymm7 vmovaps 0x40(%r13,%rdx), %ymm8 vmovaps 0x60(%r13,%rdx), %ymm9 vmovaps 0x80(%r13,%rdx), %ymm10 vmulps (%r13,%rdx), %ymm4, %ymm11 vmulps %ymm4, %ymm7, %ymm12 vmulps %ymm4, %ymm8, %ymm13 vmulps %ymm4, %ymm9, %ymm4 vmulps %ymm5, %ymm7, %ymm7 vaddps %ymm7, %ymm11, %ymm7 vmulps %ymm5, %ymm8, %ymm11 vaddps %ymm11, %ymm12, %ymm11 vmulps %ymm5, %ymm9, %ymm12 vaddps %ymm12, %ymm13, %ymm12 vmulps %ymm5, %ymm10, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps %ymm6, %ymm8, %ymm5 vaddps %ymm5, %ymm7, %ymm5 vaddps %ymm5, %ymm1, %ymm1 vmulps %ymm6, %ymm9, %ymm5 vaddps %ymm5, %ymm11, %ymm5 vaddps %ymm5, %ymm2, %ymm2 vmulps %ymm6, %ymm10, %ymm5 vaddps %ymm5, %ymm12, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmulps 0xa0(%r13,%rdx), %ymm6, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmovaps %ymm1, (%r10,%rdx) vmovaps %ymm2, 0x20(%r10,%rdx) vmovaps %ymm3, 0x40(%r10,%rdx) vmovaps %ymm0, 0x60(%r10,%rdx) addl $0x4, %esi subq $-0x80, %rdx jmp 0x2a5548 vmovaps 0x20(%r15,%rdx), %ymm0 vmovaps 0x40(%r15,%rdx), %ymm1 vmovaps (%rbx,%r11), %ymm2 vmovaps 0x20(%rbx,%r11), %ymm3 vmovaps 0x40(%rbx,%r11), %ymm4 vmovaps 0x60(%rbx,%r11), %ymm5 vmulps (%r15,%rdx), %ymm2, %ymm6 vaddps %ymm6, %ymm15, %ymm6 vmulps %ymm2, %ymm0, %ymm2 vaddps %ymm2, %ymm15, %ymm2 vmulps %ymm3, %ymm0, %ymm0 vmulps %ymm3, %ymm1, %ymm3 vmulps %ymm4, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmulps 0x60(%r15,%rdx), %ymm4, %ymm1 vaddps %ymm0, %ymm6, %ymm0 vaddps %ymm1, %ymm3, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vmovaps 0x80(%rbx,%r11), %ymm2 vmovaps 0xa0(%rbx,%r11), %ymm3 vmovaps 0x20(%r12,%rdx), %ymm4 vmulps (%r12,%rdx), %ymm5, %ymm6 vmovaps 0x40(%r12,%rdx), %ymm7 vmulps %ymm5, %ymm4, %ymm5 vmulps %ymm2, %ymm4, %ymm4 vaddps %ymm4, %ymm6, %ymm4 vmulps %ymm2, %ymm7, %ymm2 vaddps %ymm2, %ymm5, %ymm2 vmulps %ymm3, %ymm7, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmulps 0x60(%r12,%rdx), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0xc0(%rbx,%r11), %ymm2 vmovaps 0xe0(%rbx,%r11), %ymm3 vmovaps 0x100(%rbx,%r11), %ymm4 vmovaps 0x20(%r13,%rdx), %ymm5 vmovaps 0x40(%r13,%rdx), %ymm6 vmulps (%r13,%rdx), %ymm2, %ymm7 vmulps %ymm2, %ymm5, %ymm2 vmulps %ymm3, %ymm5, %ymm5 vaddps %ymm5, %ymm7, %ymm5 vmulps %ymm3, %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps %ymm4, %ymm6, %ymm3 vaddps %ymm3, %ymm5, %ymm3 vaddps %ymm3, %ymm0, %ymm0 vmulps 0x60(%r13,%rdx), %ymm4, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps %ymm0, (%r10,%rdx) vmovaps %ymm1, 0x20(%r10,%rdx) addl $0x2, %esi addq $0x40, %rdx leal 0x1(%rsi), %r8d cmpl %eax, %r8d jl 0x2a5766 jmp 0x2a595d vmovaps (%r15,%rdx), %ymm0 vmovaps 0x20(%r15,%rdx), %ymm1 vmovaps 0x40(%r15,%rdx), %ymm2 vmulps (%rbx,%r11), %ymm0, %ymm0 vmulps 0x20(%rbx,%r11), %ymm1, %ymm1 vaddps %ymm0, %ymm15, %ymm0 vmulps 0x40(%rbx,%r11), %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps (%r12,%rdx), %ymm1 vmovaps 0x20(%r12,%rdx), %ymm2 vmovaps 0x40(%r12,%rdx), %ymm3 vmulps 0x60(%rbx,%r11), %ymm1, %ymm1 vmulps 0x80(%rbx,%r11), %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0xa0(%rbx,%r11), %ymm3, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps (%r13,%rdx), %ymm1 vmovaps 0x20(%r13,%rdx), %ymm2 vmovaps 0x40(%r13,%rdx), %ymm3 vmulps 0xc0(%rbx,%r11), %ymm1, %ymm1 vmulps 0xe0(%rbx,%r11), %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x100(%rbx,%r11), %ymm3, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps %ymm0, (%r10,%rdx) incl %esi addq $0x20, %rdx cmpl %eax, %esi jl 0x2a58a3 addq %rdx, %r15 addq $0x40, %r15 addq %rdx, %r12 addq $0x40, %r12 addq %rdx, %r13 addq $0x40, %r13 incl %ebp addq %rdx, %r10 cmpl %ecx, %ebp jl 0x2a5544 movq 0xe0(%rsp), %r9 incq %r9 movq 0x18(%rsp), %r13 movq 0x150(%rsp), %r8 jmp 0x2a4cc3 movl 0xd4(%rdi), %eax movl 0xd8(%rdi), %r15d cmpl $0x5, %eax je 0x2a6998 cmpl $0x3, %eax jne 0x2a841c cmpl $0x3, %r15d jne 0x2a841c cmpl $0x1, 0xdc(%rdi) jne 0x2a7b18 cmpl $0x1, 0xe0(%rdi) jne 0x2a7b18 cmpl $0x1, 0xe4(%rdi) jne 0x2a7b18 cmpl $0x1, 0xe8(%rdi) jne 0x2a7b18 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xc8(%rsp), %edx movq 0x1b0(%rdi), %rsi xorl %edi, %edi testl %ecx, %ecx cmovlel %edi, %ecx testl %edx, %edx cmovlel %edi, %edx movq %rdx, 0x2b8(%rsp) movl $0x80, %r8d movq %rsi, 0x2b0(%rsp) cmpq 0x2b8(%rsp), %rdi je 0x2a9725 testq %rsi, %rsi je 0x2a5a5a movq %rdi, %rdx shlq $0x4, %rdx vmovups (%rsi,%rdx), %xmm0 jmp 0x2a5a5e vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) movq 0x40(%r13), %r9 imulq %rdi, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %rdx movslq 0x54(%r14), %rsi imulq %rdi, %rsi imulq 0x38(%r14), %rsi movslq 0xbc(%rsp), %r10 movq 0xd0(%rsp), %r13 imulq %rdi, %r13 movq 0xa0(%rsp), %r11 imulq %r11, %r13 addq 0x90(%rsp), %r13 imulq %r11, %r10 leaq (%r10,%r13), %rbp leaq (,%r10,2), %r11 addq %r13, %r11 vmovaps (%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x50(%rsp) vmovaps 0x10(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x70(%rsp) vmovaps 0x20(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x110(%rsp) vmovaps 0x30(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0xf0(%rsp) vmovaps 0x40(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x130(%rsp) vmovaps 0x50(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x170(%rsp) vmovaps 0x60(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x190(%rsp) vmovaps 0x70(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x270(%rsp) vmovaps 0x80(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x1f0(%rsp) xorl %r10d, %r10d cmpl %ecx, %r10d je 0x2a661f leaq 0x20(%r13), %r15 leaq 0x20(%rbp), %rbx leaq 0x20(%r11), %rdx xorl %esi, %esi xorl %r12d, %r12d leal 0x7(%r12), %r14d cmpl %eax, %r14d jge 0x2a63ad vmovaps 0x50(%rsp), %xmm0 vmulps (%r13,%rsi), %xmm0, %xmm0 vaddps 0x20(%rsp), %xmm0, %xmm1 vmovaps 0x10(%r13,%rsi), %xmm3 vmovaps 0x20(%r13,%rsi), %xmm0 vmovaps %xmm0, 0xe0(%rsp) vmovaps 0x30(%r13,%rsi), %xmm10 vmulps 0x70(%rsp), %xmm3, %xmm2 vmulps 0x110(%rsp), %xmm0, %xmm4 vaddps %xmm4, %xmm2, %xmm2 vmovaps 0xf0(%rsp), %xmm0 vmulps (%rbp,%rsi), %xmm0, %xmm4 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0x10(%rbp,%rsi), %xmm13 vmovaps 0x20(%rbp,%rsi), %xmm0 vmovaps %xmm0, 0x260(%rsp) vmovaps 0x30(%rbp,%rsi), %xmm2 vmovaps %xmm2, 0x160(%rsp) vmovaps 0x130(%rsp), %xmm5 vmulps %xmm5, %xmm13, %xmm12 vmovaps %xmm5, %xmm9 vaddps %xmm4, %xmm12, %xmm4 vmulps 0x170(%rsp), %xmm0, %xmm12 vaddps %xmm4, %xmm12, %xmm4 vaddps %xmm4, %xmm1, %xmm1 vmovaps 0x190(%rsp), %xmm11 vmulps (%r11,%rsi), %xmm11, %xmm14 vmovaps 0x10(%r11,%rsi), %xmm15 vmovaps 0x20(%r11,%rsi), %xmm0 vmovaps 0x30(%r11,%rsi), %xmm12 vmovaps 0x270(%rsp), %xmm7 vmulps %xmm7, %xmm15, %xmm5 vaddps %xmm5, %xmm14, %xmm5 vmovaps 0x1f0(%rsp), %xmm8 vmulps %xmm0, %xmm8, %xmm14 vaddps %xmm5, %xmm14, %xmm5 vaddps %xmm5, %xmm1, %xmm1 vmovaps %xmm1, (%r9,%rsi) vmulps 0x50(%rsp), %xmm3, %xmm1 vaddps 0x20(%rsp), %xmm1, %xmm1 vmovaps 0xe0(%rsp), %xmm2 vmulps 0x70(%rsp), %xmm2, %xmm3 vmovaps 0x110(%rsp), %xmm2 vmovaps %xmm10, %xmm14 vmovaps %xmm10, 0x2a0(%rsp) vmulps %xmm2, %xmm10, %xmm5 vmovaps 0xf0(%rsp), %xmm4 vmulps %xmm4, %xmm13, %xmm13 vaddps %xmm3, %xmm13, %xmm3 vaddps %xmm3, %xmm1, %xmm1 vmovaps %xmm1, 0x150(%rsp) vmovaps %xmm9, %xmm10 vmovaps 0x260(%rsp), %xmm9 vmulps %xmm10, %xmm9, %xmm3 vmovaps 0x160(%rsp), %xmm1 vmovaps 0x170(%rsp), %xmm6 vmulps %xmm6, %xmm1, %xmm13 vaddps %xmm5, %xmm13, %xmm5 vmulps %xmm11, %xmm15, %xmm13 vaddps %xmm3, %xmm13, %xmm3 vmulps %xmm7, %xmm0, %xmm13 vmovaps %xmm0, %xmm11 vaddps %xmm3, %xmm13, %xmm3 vaddps 0x150(%rsp), %xmm3, %xmm7 vmulps %xmm8, %xmm12, %xmm3 vmovaps %xmm12, 0x290(%rsp) vaddps %xmm3, %xmm5, %xmm3 vaddps %xmm3, %xmm7, %xmm5 vmovaps 0x40(%r13,%rsi), %xmm3 vmovaps 0x40(%rbp,%rsi), %xmm7 vmovaps 0x40(%r11,%rsi), %xmm8 vmovaps %xmm5, 0x10(%r9,%rsi) vmovaps 0xe0(%rsp), %xmm0 vmulps 0x50(%rsp), %xmm0, %xmm0 vaddps 0x20(%rsp), %xmm0, %xmm0 vmulps 0x70(%rsp), %xmm14, %xmm5 vmulps %xmm2, %xmm3, %xmm14 vmulps %xmm4, %xmm9, %xmm2 vmovaps %xmm10, %xmm13 vmulps %xmm1, %xmm10, %xmm15 vaddps %xmm5, %xmm15, %xmm5 vmulps %xmm6, %xmm7, %xmm15 vmovaps %xmm7, 0x150(%rsp) vaddps %xmm15, %xmm14, %xmm14 vmovaps 0x190(%rsp), %xmm1 vmulps %xmm1, %xmm11, %xmm4 vmovaps %xmm1, %xmm9 vaddps %xmm4, %xmm2, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x270(%rsp), %xmm10 vmulps %xmm10, %xmm12, %xmm2 vaddps %xmm2, %xmm5, %xmm2 vaddps %xmm2, %xmm0, %xmm4 vmovaps %xmm8, %xmm0 vmovaps %xmm8, 0x350(%rsp) vmovaps 0x1f0(%rsp), %xmm6 vmulps %xmm6, %xmm8, %xmm2 vaddps %xmm2, %xmm14, %xmm2 vaddps %xmm2, %xmm4, %xmm5 vmovaps 0x50(%r13,%rsi), %xmm1 vmovaps %xmm1, 0x260(%rsp) vmovaps 0x50(%rbp,%rsi), %xmm4 vmovaps %xmm4, 0x360(%rsp) vmovaps 0x50(%r11,%rsi), %xmm2 vmovaps %xmm2, 0xe0(%rsp) vmovaps %xmm5, 0x20(%r9,%rsi) vmovaps 0x2a0(%rsp), %xmm2 vmulps 0x50(%rsp), %xmm2, %xmm5 vaddps 0x20(%rsp), %xmm5, %xmm5 vmulps 0x70(%rsp), %xmm3, %xmm12 vmovaps 0x110(%rsp), %xmm2 vmulps %xmm2, %xmm1, %xmm14 vmovaps 0x160(%rsp), %xmm1 vmulps 0xf0(%rsp), %xmm1, %xmm11 vmulps %xmm7, %xmm13, %xmm15 vaddps %xmm15, %xmm12, %xmm8 vmovaps 0x170(%rsp), %xmm13 vmulps %xmm4, %xmm13, %xmm15 vaddps %xmm15, %xmm14, %xmm14 vmulps 0x290(%rsp), %xmm9, %xmm12 vmovaps %xmm9, %xmm15 vaddps %xmm12, %xmm11, %xmm11 vaddps %xmm5, %xmm11, %xmm5 vmulps %xmm0, %xmm10, %xmm11 vaddps %xmm11, %xmm8, %xmm10 vaddps %xmm5, %xmm10, %xmm5 vmovaps 0xe0(%rsp), %xmm0 vmulps %xmm6, %xmm0, %xmm10 vaddps %xmm10, %xmm14, %xmm10 vaddps %xmm5, %xmm10, %xmm5 vmovaps 0x60(%r13,%rsi), %xmm8 vmovaps 0x60(%rbp,%rsi), %xmm1 vmovaps %xmm1, 0x290(%rsp) vmovaps 0x60(%r11,%rsi), %xmm14 vmovaps %xmm5, 0x30(%r9,%rsi) vmulps 0x50(%rsp), %xmm3, %xmm3 vaddps 0x20(%rsp), %xmm3, %xmm3 vmovaps 0x260(%rsp), %xmm7 vmulps 0x70(%rsp), %xmm7, %xmm5 vmulps %xmm2, %xmm8, %xmm10 vmovaps %xmm8, 0x160(%rsp) vmovaps %xmm2, %xmm11 vmovaps 0xf0(%rsp), %xmm4 vmulps 0x150(%rsp), %xmm4, %xmm2 vmovaps 0x130(%rsp), %xmm6 vmovaps 0x360(%rsp), %xmm9 vmulps %xmm6, %xmm9, %xmm12 vaddps %xmm5, %xmm12, %xmm5 vmulps %xmm1, %xmm13, %xmm12 vaddps %xmm12, %xmm10, %xmm10 vmulps 0x350(%rsp), %xmm15, %xmm12 vaddps %xmm2, %xmm12, %xmm1 vaddps %xmm1, %xmm3, %xmm1 vmovaps 0x270(%rsp), %xmm12 vmulps %xmm0, %xmm12, %xmm3 vaddps %xmm3, %xmm5, %xmm3 vaddps %xmm3, %xmm1, %xmm1 vmovaps 0x1f0(%rsp), %xmm2 vmulps %xmm2, %xmm14, %xmm3 vaddps %xmm3, %xmm10, %xmm3 vaddps %xmm3, %xmm1, %xmm1 vmovaps 0x70(%r13,%rsi), %xmm5 vmovaps %xmm5, 0x2a0(%rsp) vmovaps 0x70(%rbp,%rsi), %xmm0 vmovaps 0x70(%r11,%rsi), %xmm10 vmovaps %xmm1, 0x40(%r9,%rsi) vmulps 0x50(%rsp), %xmm7, %xmm1 vaddps 0x20(%rsp), %xmm1, %xmm1 vmulps 0x70(%rsp), %xmm8, %xmm3 vmulps %xmm5, %xmm11, %xmm7 vmulps %xmm4, %xmm9, %xmm13 vmovaps %xmm4, %xmm11 vmovaps 0x290(%rsp), %xmm9 vmulps %xmm6, %xmm9, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmovaps 0x170(%rsp), %xmm8 vmulps %xmm0, %xmm8, %xmm5 vmovaps %xmm0, %xmm4 vmovaps %xmm0, 0x150(%rsp) vaddps %xmm5, %xmm7, %xmm5 vmulps 0xe0(%rsp), %xmm15, %xmm0 vaddps %xmm0, %xmm13, %xmm0 vaddps %xmm0, %xmm1, %xmm0 vmulps %xmm12, %xmm14, %xmm1 vaddps %xmm1, %xmm3, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmulps %xmm2, %xmm10, %xmm1 vmovaps %xmm10, %xmm2 vaddps %xmm1, %xmm5, %xmm1 vaddps %xmm1, %xmm0, %xmm3 vmovaps 0x80(%r13,%rsi), %xmm1 vmovaps 0x80(%rbp,%rsi), %xmm10 vmovaps 0x80(%r11,%rsi), %xmm0 vmovaps %xmm0, 0xe0(%rsp) vmovaps %xmm3, 0x50(%r9,%rsi) vmovaps 0x2a0(%rsp), %xmm7 vmulps 0x70(%rsp), %xmm7, %xmm3 vmovaps 0x110(%rsp), %xmm13 vmulps %xmm1, %xmm13, %xmm0 vmulps %xmm6, %xmm4, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmulps %xmm8, %xmm10, %xmm5 vaddps %xmm5, %xmm0, %xmm4 vmovaps %xmm11, %xmm0 vmulps %xmm11, %xmm9, %xmm5 vmulps %xmm15, %xmm14, %xmm14 vmovaps %xmm15, %xmm6 vaddps %xmm5, %xmm14, %xmm5 vmovaps 0x50(%rsp), %xmm14 vmulps 0x160(%rsp), %xmm14, %xmm11 vaddps 0x20(%rsp), %xmm11, %xmm11 vaddps %xmm5, %xmm11, %xmm5 vmulps %xmm2, %xmm12, %xmm11 vmovaps %xmm2, %xmm15 vaddps %xmm3, %xmm11, %xmm3 vaddps %xmm3, %xmm5, %xmm3 vmovaps 0x1f0(%rsp), %xmm2 vmovaps 0xe0(%rsp), %xmm9 vmulps %xmm2, %xmm9, %xmm5 vaddps %xmm5, %xmm4, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps 0x90(%r13,%rsi), %xmm13, %xmm4 vmulps 0x90(%rbp,%rsi), %xmm8, %xmm5 vaddps %xmm5, %xmm4, %xmm4 vmulps 0x90(%r11,%rsi), %xmm2, %xmm5 vaddps %xmm5, %xmm4, %xmm4 vmovaps %xmm3, 0x60(%r9,%rsi) vmulps 0x70(%rsp), %xmm1, %xmm2 vmulps 0x130(%rsp), %xmm10, %xmm1 vaddps %xmm1, %xmm2, %xmm1 vmulps 0x150(%rsp), %xmm0, %xmm2 vmulps %xmm6, %xmm15, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps %xmm7, %xmm14, %xmm3 vaddps 0x20(%rsp), %xmm3, %xmm3 vaddps %xmm2, %xmm3, %xmm2 vmulps %xmm12, %xmm9, %xmm0 vaddps %xmm0, %xmm1, %xmm0 vaddps %xmm0, %xmm2, %xmm0 vaddps %xmm4, %xmm0, %xmm0 vmovaps %xmm0, 0x70(%r9,%rsi) addl $0x8, %r12d addq %r8, %rsi addq %r8, %r15 addq %r8, %rbx addq %r8, %rdx jmp 0x2a5b61 vmovaps 0x50(%rsp), %xmm0 vmulps (%r13,%rsi), %xmm0, %xmm0 vaddps 0x20(%rsp), %xmm0, %xmm5 vmovaps 0x10(%r13,%rsi), %xmm1 vmovaps 0x20(%r13,%rsi), %xmm0 vmovaps 0x30(%r13,%rsi), %xmm2 vmovaps %xmm2, 0xe0(%rsp) vmulps 0x70(%rsp), %xmm1, %xmm2 vmovaps 0x110(%rsp), %xmm12 vmulps %xmm0, %xmm12, %xmm3 vmovaps %xmm0, %xmm10 vmovaps %xmm0, 0x150(%rsp) vaddps %xmm3, %xmm2, %xmm2 vmovaps 0xf0(%rsp), %xmm0 vmulps (%rbp,%rsi), %xmm0, %xmm3 vaddps %xmm2, %xmm5, %xmm2 vmovaps 0x10(%rbp,%rsi), %xmm4 vmovaps 0x20(%rbp,%rsi), %xmm13 vmovaps 0x30(%rbp,%rsi), %xmm5 vmovaps %xmm5, 0x160(%rsp) vmulps 0x130(%rsp), %xmm4, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmovaps 0x170(%rsp), %xmm9 vmulps %xmm9, %xmm13, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm3 vmovaps 0x190(%rsp), %xmm11 vmulps (%r11,%rsi), %xmm11, %xmm5 vmovaps 0x10(%r11,%rsi), %xmm14 vmovaps 0x20(%r11,%rsi), %xmm8 vmovaps 0x30(%r11,%rsi), %xmm2 vmovaps 0x270(%rsp), %xmm6 vmulps %xmm6, %xmm14, %xmm15 vaddps %xmm5, %xmm15, %xmm5 vmovaps 0x1f0(%rsp), %xmm7 vmulps %xmm7, %xmm8, %xmm15 vaddps %xmm5, %xmm15, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmovaps %xmm3, (%r9,%rsi) vmulps 0x50(%rsp), %xmm1, %xmm1 vaddps 0x20(%rsp), %xmm1, %xmm1 vmulps 0x70(%rsp), %xmm10, %xmm3 vmovaps 0xe0(%rsp), %xmm15 vmulps %xmm12, %xmm15, %xmm5 vmulps %xmm0, %xmm4, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm1, %xmm10 vmovaps 0x130(%rsp), %xmm1 vmulps %xmm1, %xmm13, %xmm3 vmovaps 0x160(%rsp), %xmm0 vmulps %xmm0, %xmm9, %xmm4 vaddps %xmm4, %xmm5, %xmm4 vmulps %xmm11, %xmm14, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmulps %xmm6, %xmm8, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vaddps %xmm3, %xmm10, %xmm5 vmulps %xmm7, %xmm2, %xmm3 vmovaps %xmm2, %xmm10 vaddps %xmm3, %xmm4, %xmm3 vaddps %xmm3, %xmm5, %xmm5 vmovaps 0x40(%r13,%rsi), %xmm4 vmovaps 0x40(%rbp,%rsi), %xmm3 vmovaps 0x40(%r11,%rsi), %xmm2 vmovaps %xmm2, 0x260(%rsp) vmovaps %xmm5, 0x10(%r9,%rsi) vmulps 0x70(%rsp), %xmm15, %xmm5 vmulps %xmm4, %xmm12, %xmm14 vmulps %xmm1, %xmm0, %xmm15 vaddps %xmm5, %xmm15, %xmm5 vmulps %xmm3, %xmm9, %xmm15 vaddps %xmm15, %xmm14, %xmm14 vmovaps 0xf0(%rsp), %xmm15 vmulps %xmm15, %xmm13, %xmm0 vmulps %xmm11, %xmm8, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x150(%rsp), %xmm2 vmulps 0x50(%rsp), %xmm2, %xmm2 vmovaps 0x20(%rsp), %xmm13 vaddps %xmm2, %xmm13, %xmm2 vaddps %xmm0, %xmm2, %xmm0 vmulps %xmm6, %xmm10, %xmm2 vaddps %xmm2, %xmm5, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x260(%rsp), %xmm8 vmulps %xmm7, %xmm8, %xmm2 vaddps %xmm2, %xmm14, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmulps 0x50(%r13,%rsi), %xmm12, %xmm2 vmulps 0x50(%rbp,%rsi), %xmm9, %xmm5 vaddps %xmm5, %xmm2, %xmm2 vmulps 0x50(%r11,%rsi), %xmm7, %xmm5 vaddps %xmm5, %xmm2, %xmm2 vmovaps %xmm0, 0x20(%r9,%rsi) vmulps 0x70(%rsp), %xmm4, %xmm0 vmulps %xmm1, %xmm3, %xmm3 vaddps %xmm3, %xmm0, %xmm0 vmulps 0x160(%rsp), %xmm15, %xmm3 vmulps %xmm11, %xmm10, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmovaps 0xe0(%rsp), %xmm1 vmulps 0x50(%rsp), %xmm1, %xmm4 vmovaps %xmm13, 0x20(%rsp) vaddps %xmm4, %xmm13, %xmm4 vaddps %xmm3, %xmm4, %xmm3 vmulps %xmm6, %xmm8, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm0, %xmm3, %xmm0 vaddps %xmm2, %xmm0, %xmm0 vmovaps %xmm0, 0x30(%r9,%rsi) addl $0x4, %r12d addq $0x40, %rsi addq $0x40, %r15 addq $0x40, %rbx addq $0x40, %rdx leal 0x3(%r12), %r14d cmpl %eax, %r14d jl 0x2a6129 jmp 0x2a6534 vmovaps 0x10(%r13,%rsi), %xmm0 vmovaps %xmm0, 0xe0(%rsp) vmovaps 0x20(%r13,%rsi), %xmm4 vmovaps 0x10(%rbp,%rsi), %xmm5 vmovaps 0x20(%rbp,%rsi), %xmm3 vmovaps %xmm3, 0x160(%rsp) vmovaps 0x10(%r11,%rsi), %xmm2 vmovaps 0x50(%rsp), %xmm1 vmulps (%r13,%rsi), %xmm1, %xmm10 vmovaps 0x20(%r11,%rsi), %xmm1 vmovaps %xmm1, 0x150(%rsp) vmovaps 0x20(%rsp), %xmm14 vaddps %xmm14, %xmm10, %xmm10 vmovaps 0x70(%rsp), %xmm13 vmulps %xmm0, %xmm13, %xmm11 vmovaps 0x110(%rsp), %xmm6 vmulps %xmm6, %xmm4, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vaddps %xmm11, %xmm10, %xmm10 vmovaps 0xf0(%rsp), %xmm7 vmulps (%rbp,%rsi), %xmm7, %xmm11 vmovaps 0x130(%rsp), %xmm8 vmulps %xmm5, %xmm8, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vmovaps 0x170(%rsp), %xmm9 vmulps %xmm3, %xmm9, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vaddps %xmm11, %xmm10, %xmm10 vmovaps 0x190(%rsp), %xmm15 vmulps (%r11,%rsi), %xmm15, %xmm11 vmovaps 0x270(%rsp), %xmm3 vmulps %xmm3, %xmm2, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vmovaps 0x1f0(%rsp), %xmm0 vmulps %xmm0, %xmm1, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vaddps %xmm11, %xmm10, %xmm10 vmulps 0x30(%r13,%rsi), %xmm6, %xmm11 vmulps 0x30(%rbp,%rsi), %xmm9, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vmulps 0x30(%r11,%rsi), %xmm0, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vmovaps %xmm10, (%r9,%rsi) vmulps %xmm4, %xmm13, %xmm4 vmulps %xmm7, %xmm5, %xmm5 vaddps %xmm5, %xmm4, %xmm4 vmovaps 0xe0(%rsp), %xmm0 vmulps 0x50(%rsp), %xmm0, %xmm0 vaddps %xmm0, %xmm14, %xmm0 vaddps %xmm4, %xmm0, %xmm0 vmulps 0x160(%rsp), %xmm8, %xmm1 vmulps %xmm2, %xmm15, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x150(%rsp), %xmm3, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm0, %xmm11, %xmm0 vmovaps %xmm0, 0x10(%r9,%rsi) addl $0x2, %r12d addq $0x20, %rsi addq $0x20, %r15 addq $0x20, %rbx addq $0x20, %rdx leal 0x1(%r12), %r14d cmpl %eax, %r14d jl 0x2a63c0 addq %rsi, %r9 movq 0x10(%rsp), %r14 cmpl %eax, %r12d jge 0x2a660e vmovaps 0x50(%rsp), %xmm0 vmulps -0x20(%r15), %xmm0, %xmm0 vaddps 0x20(%rsp), %xmm0, %xmm0 vmovaps 0x70(%rsp), %xmm1 vmulps -0x10(%r15), %xmm1, %xmm1 vmovaps 0x110(%rsp), %xmm2 vmulps (%r15), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmovaps 0xf0(%rsp), %xmm1 vmulps -0x20(%rbx), %xmm1, %xmm1 vmovaps 0x130(%rsp), %xmm2 vmulps -0x10(%rbx), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0x170(%rsp), %xmm2 vmulps (%rbx), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0x190(%rsp), %xmm2 vmulps -0x20(%rdx), %xmm2, %xmm2 vmovaps 0x270(%rsp), %xmm3 vmulps -0x10(%rdx), %xmm3, %xmm3 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm3, %xmm2, %xmm1 vmovaps 0x1f0(%rsp), %xmm2 vmulps (%rdx), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmovaps %xmm0, (%r9) addq $0x10, %r9 incl %r12d addq $0x10, %r15 addq $0x10, %rbx addq $0x10, %rdx jmp 0x2a654a incl %r10d movq %r15, %r13 movq %rbx, %rbp movq %rdx, %r11 jmp 0x2a5b47 incq %rdi movq 0x18(%rsp), %r13 movq 0x2b0(%rsp), %rsi jmp 0x2a5a39 cmpl $0x3, 0xd4(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x3, 0xd8(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x1, 0xdc(%rdi) movq %r9, 0x50(%rsp) jne 0x2a742a cmpl $0x1, 0xe0(%rdi) jne 0x2a742a cmpl $0x1, 0xe4(%rdi) jne 0x2a742a cmpl $0x1, 0xe8(%rdi) jne 0x2a742a movslq 0xbc(%rsp), %rax movl 0x2c(%r13), %edx movl 0x30(%r13), %r8d movl 0xc8(%rsp), %ecx movq 0x28(%r14), %r10 imulq 0x10(%r13), %r11 movq 0x1b0(%rdi), %rsi movslq %edx, %rdi movq 0x90(%rsp), %rbx movq %rbx, 0x1f0(%rsp) movq 0xa0(%rsp), %rbx imulq 0xd0(%rsp), %rbx movq %rbx, 0x270(%rsp) leaq (%rax,%rax), %rbx movq %rbx, 0x190(%rsp) leaq (%rax,%rax,2), %rbx movq %rbx, 0xe0(%rsp) movq %rax, 0x130(%rsp) addl $0x2, %eax movslq %eax, %rbp xorl %eax, %eax testl %ecx, %ecx cmovlel %eax, %ecx movq %rcx, 0xf0(%rsp) shlq $0x2, %rbp movq %rdi, 0x10(%rsp) leaq (,%rdi,4), %rcx movq %rcx, 0x20(%rsp) movq %rsi, 0x170(%rsp) movq %r11, 0x110(%rsp) cmpq 0xf0(%rsp), %rax je 0x2a9725 testq %rsi, %rsi je 0x2a6753 vmovss (%rsi,%rax,4), %xmm0 jmp 0x2a6757 vxorps %xmm0, %xmm0, %xmm0 movq %r11, %rsi imulq %rax, %rsi addq %r9, %rsi imulq $0x24, %rax, %r13 movq 0x270(%rsp), %r15 movq %rax, 0x70(%rsp) imulq %rax, %r15 addq 0x1f0(%rsp), %r15 movq 0x130(%rsp), %rax leaq (%r15,%rax,4), %rax movq 0x190(%rsp), %rcx leaq (%r15,%rcx,4), %r9 movq 0xe0(%rsp), %rcx leaq (%r15,%rcx,4), %r11 xorl %edi, %edi movq %rsi, %rbx movl %edi, %ecx orl $0x1, %ecx cmpl %r8d, %ecx jge 0x2a68d4 movq 0x10(%rsp), %rcx leaq (%rsi,%rcx,4), %rsi xorl %r12d, %r12d movl %edx, %ecx testl %ecx, %ecx jle 0x2a68a9 vmovsd 0x4(%r15,%r12), %xmm1 vmovss (%r15,%r12), %xmm2 vmovlhps %xmm1, %xmm2, %xmm2 # xmm2 = xmm2[0],xmm1[0] vshufps $0xd8, %xmm1, %xmm2, %xmm2 # xmm2 = xmm2[0,2],xmm1[1,3] vmovss (%rax,%r12), %xmm1 vinsertps $0x30, %xmm1, %xmm2, %xmm2 # xmm2 = xmm2[0,1,2],xmm1[0] vmovsd 0x4(%rax,%r12), %xmm3 vmovss (%r9,%r12), %xmm4 vinsertps $0x20, %xmm4, %xmm3, %xmm5 # xmm5 = xmm3[0,1],xmm4[0],xmm3[3] vinsertps $0x30, 0x4(%r9,%r12), %xmm5, %xmm5 # xmm5 = xmm5[0,1,2],mem[0] vmovups 0x10(%r10,%r13), %xmm6 vmovss 0x20(%r10,%r13), %xmm7 vmulss 0x8(%r9,%r12), %xmm7, %xmm8 vmulps %xmm5, %xmm6, %xmm5 vhaddps %xmm5, %xmm5, %xmm5 vmovups (%r10,%r13), %xmm9 vhaddps %xmm5, %xmm5, %xmm5 vmulps %xmm2, %xmm9, %xmm2 vhaddps %xmm2, %xmm2, %xmm2 vhaddps %xmm2, %xmm2, %xmm2 vaddss %xmm2, %xmm8, %xmm2 vmovlhps %xmm3, %xmm9, %xmm8 # xmm8 = xmm9[0],xmm3[0] vshufps $0xd8, %xmm3, %xmm8, %xmm3 # xmm3 = xmm8[0,2],xmm3[1,3] vinsertps $0x30, %xmm4, %xmm3, %xmm3 # xmm3 = xmm3[0,1,2],xmm4[0] vmovsd 0x4(%r9,%r12), %xmm4 vmovhps (%r11,%r12), %xmm4, %xmm4 # xmm4 = xmm4[0,1],mem[0,1] vmulss 0x8(%r11,%r12), %xmm7, %xmm7 vmulps %xmm6, %xmm4, %xmm4 vhaddps %xmm4, %xmm4, %xmm4 vmovss %xmm1, %xmm9, %xmm1 # xmm1 = xmm1[0],xmm9[1,2,3] vhaddps %xmm4, %xmm4, %xmm4 vmulps %xmm1, %xmm3, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm0, %xmm2, %xmm2 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm2, %xmm5, %xmm2 vaddss %xmm7, %xmm4, %xmm3 vaddss %xmm0, %xmm1, %xmm1 vaddss %xmm1, %xmm3, %xmm1 vmovss %xmm2, (%rbx,%r12) vmovss %xmm1, (%rsi,%r12) decl %ecx addq $0x4, %r12 jmp 0x2a67c3 addq %rbp, %r15 addq %r12, %r15 addq %rbp, %rax addq %r12, %rax addq %rbp, %r9 addq %r12, %r9 addq %rbp, %r11 addq %r12, %r11 addq 0x20(%rsp), %rbx addq %r12, %rbx addl $0x2, %edi addq %r12, %rsi jmp 0x2a67a7 movq 0x110(%rsp), %r11 cmpl %r8d, %edi jge 0x2a6979 xorl %ecx, %ecx movl %edx, %esi testl %esi, %esi jle 0x2a695a vmovsd (%r15,%rcx), %xmm1 vinsertps $0x20, 0x8(%r15,%rcx), %xmm1, %xmm1 # xmm1 = xmm1[0,1],mem[0],xmm1[3] vinsertps $0x30, (%rax,%rcx), %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],mem[0] vmovsd 0x4(%rax,%rcx), %xmm2 vinsertf128 $0x1, %xmm2, %ymm1, %ymm2 vmovsd (%r9,%rcx), %xmm3 vinsertf128 $0x1, %xmm3, %ymm1, %ymm1 vshufpd $0x2, %ymm1, %ymm2, %ymm1 # ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[2] vmulps (%r10,%r13), %ymm1, %ymm1 vmovss 0x20(%r10,%r13), %xmm2 vmulss 0x8(%r9,%rcx), %xmm2, %xmm2 vextractf128 $0x1, %ymm1, %xmm3 vhaddps %xmm1, %xmm3, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm0, %xmm2, %xmm2 vaddss %xmm1, %xmm2, %xmm1 vmovss %xmm1, (%rbx,%rcx) decl %esi addq $0x4, %rcx jmp 0x2a68e9 addq %rcx, %r15 addq $0x8, %r15 addq %rcx, %rax addq $0x8, %rax addq %rcx, %r9 addq $0x8, %r9 incl %edi addq %rcx, %rbx jmp 0x2a68dc movq 0x70(%rsp), %rax incq %rax movq 0x18(%rsp), %r13 movq 0x50(%rsp), %r9 movq 0x170(%rsp), %rsi jmp 0x2a6739 cmpl $0x5, %r15d jne 0x2a841c cmpl $0x1, 0xdc(%rdi) jne 0x2a80c5 cmpl $0x1, 0xe0(%rdi) jne 0x2a80c5 cmpl $0x1, 0xe4(%rdi) jne 0x2a80c5 cmpl $0x1, 0xe8(%rdi) jne 0x2a80c5 movl 0x2c(%r13), %r8d movl 0x30(%r13), %eax movl %eax, 0x10(%rsp) movl 0xbc(%rsp), %eax movl 0xc8(%rsp), %ecx movq 0x1b0(%rdi), %rdx leal 0x10(,%rax,4), %eax movslq %eax, %rdi leal (,%r8,4), %eax cltq xorl %esi, %esi testl %r8d, %r8d cmovlel %esi, %r8d testl %ecx, %ecx cmovlel %esi, %ecx movq %rcx, 0xf0(%rsp) shlq $0x2, %rax movq %rax, 0x50(%rsp) shlq $0x2, %rdi movq %rdx, 0x110(%rsp) movq %r8, 0x20(%rsp) cmpq 0xf0(%rsp), %rsi je 0x2a9725 testq %rdx, %rdx je 0x2a6a5c movq %rsi, %rax shlq $0x4, %rax vmovups (%rdx,%rax), %xmm0 jmp 0x2a6a60 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r10 imulq %rsi, %r10 movq 0x10(%r13), %rax imulq %rax, %r10 addq (%r13), %r10 movslq 0x2c(%r13), %r9 movq 0x28(%r14), %r11 movslq 0x54(%r14), %rbx imulq %rsi, %rbx imulq 0x38(%r14), %rbx imulq %rax, %r9 addq %r10, %r9 movslq 0xbc(%rsp), %rax movq 0xd0(%rsp), %r15 movq %rsi, 0x70(%rsp) imulq %rsi, %r15 movq 0xa0(%rsp), %rcx imulq %rcx, %r15 addq 0x90(%rsp), %r15 imulq %rcx, %rax leaq (%r15,%rax), %r12 leaq (%r15,%rax,2), %r13 leaq (%rax,%rax,2), %rbp addq %r15, %rbp leaq (%r15,%rax,4), %rdx leaq (%rax,%rax,4), %rax addq %r15, %rax xorl %esi, %esi movl %esi, %ecx orl $0x1, %ecx cmpl 0x10(%rsp), %ecx jge 0x2a70a2 movl %r8d, %ecx xorl %r8d, %r8d subl $0x1, %ecx jb 0x2a6e36 vmovaps (%r11,%rbx), %xmm1 vmovaps 0x10(%r11,%rbx), %xmm2 vmovaps 0x20(%r11,%rbx), %xmm3 vmovaps 0x30(%r11,%rbx), %xmm4 vmulps (%r15,%r8), %xmm1, %xmm5 vmovaps 0x40(%r11,%rbx), %xmm6 vaddps %xmm0, %xmm5, %xmm5 vmulps 0x10(%r15,%r8), %xmm2, %xmm7 vmulps 0x20(%r15,%r8), %xmm3, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps 0x30(%r15,%r8), %xmm4, %xmm7 vmulps 0x40(%r15,%r8), %xmm6, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vmovaps (%r12,%r8), %xmm8 vmovaps 0x10(%r12,%r8), %xmm9 vmovaps 0x20(%r12,%r8), %xmm10 vmovaps 0x30(%r12,%r8), %xmm11 vmovaps 0x40(%r12,%r8), %xmm12 vmulps %xmm1, %xmm8, %xmm1 vaddps %xmm0, %xmm1, %xmm1 vmulps %xmm2, %xmm9, %xmm2 vmulps %xmm3, %xmm10, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps %xmm4, %xmm11, %xmm2 vmulps %xmm6, %xmm12, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0x50(%r11,%rbx), %xmm3 vmovaps 0x60(%r11,%rbx), %xmm4 vmovaps 0x70(%r11,%rbx), %xmm6 vmovaps 0x80(%r11,%rbx), %xmm13 vmovaps 0x90(%r11,%rbx), %xmm14 vmulps %xmm3, %xmm8, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm4, %xmm9, %xmm7 vmulps %xmm6, %xmm10, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vmulps %xmm11, %xmm13, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vmulps %xmm12, %xmm14, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmovaps (%r13,%r8), %xmm7 vmovaps 0x10(%r13,%r8), %xmm8 vmovaps 0x20(%r13,%r8), %xmm9 vmovaps 0x30(%r13,%r8), %xmm10 vmovaps 0x40(%r13,%r8), %xmm11 vmulps %xmm3, %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps %xmm4, %xmm8, %xmm2 vmulps %xmm6, %xmm9, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps %xmm13, %xmm10, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps %xmm14, %xmm11, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm2 vmovaps 0xa0(%r11,%rbx), %xmm3 vmovaps 0xb0(%r11,%rbx), %xmm4 vmovaps 0xc0(%r11,%rbx), %xmm6 vmovaps 0xd0(%r11,%rbx), %xmm12 vmovaps 0xe0(%r11,%rbx), %xmm13 vmulps %xmm7, %xmm3, %xmm1 vmulps %xmm4, %xmm8, %xmm7 vaddps %xmm7, %xmm1, %xmm1 vmulps %xmm6, %xmm9, %xmm7 vaddps %xmm7, %xmm1, %xmm1 vmulps %xmm10, %xmm12, %xmm7 vaddps %xmm7, %xmm1, %xmm1 vmulps %xmm11, %xmm13, %xmm7 vaddps %xmm7, %xmm1, %xmm1 vaddps %xmm1, %xmm5, %xmm1 vmovaps (%rbp,%r8), %xmm5 vmovaps 0x10(%rbp,%r8), %xmm7 vmovaps 0x20(%rbp,%r8), %xmm8 vmovaps 0x30(%rbp,%r8), %xmm9 vmovaps 0x40(%rbp,%r8), %xmm10 vmulps %xmm3, %xmm5, %xmm3 vmulps %xmm4, %xmm7, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm6, %xmm8, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm12, %xmm9, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm13, %xmm10, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0xf0(%r11,%rbx), %xmm3 vmovaps 0x100(%r11,%rbx), %xmm4 vmovaps 0x110(%r11,%rbx), %xmm6 vmovaps 0x120(%r11,%rbx), %xmm11 vmovaps 0x130(%r11,%rbx), %xmm12 vmulps %xmm5, %xmm3, %xmm5 vmulps %xmm7, %xmm4, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm6, %xmm8, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm9, %xmm11, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm10, %xmm12, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmovaps (%rdx,%r8), %xmm7 vmovaps 0x10(%rdx,%r8), %xmm8 vmovaps 0x20(%rdx,%r8), %xmm9 vmovaps 0x30(%rdx,%r8), %xmm10 vmovaps 0x40(%rdx,%r8), %xmm13 vmulps %xmm3, %xmm7, %xmm3 vmulps %xmm4, %xmm8, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm6, %xmm9, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm11, %xmm10, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm12, %xmm13, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmovaps 0x140(%r11,%rbx), %xmm4 vmovaps 0x150(%r11,%rbx), %xmm6 vmovaps 0x160(%r11,%rbx), %xmm11 vmovaps 0x170(%r11,%rbx), %xmm12 vmovaps 0x180(%r11,%rbx), %xmm14 vmulps %xmm7, %xmm4, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vaddps %xmm5, %xmm1, %xmm1 vmulps %xmm6, %xmm8, %xmm5 vmulps %xmm9, %xmm11, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm10, %xmm12, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm13, %xmm14, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vaddps %xmm5, %xmm1, %xmm1 vmulps (%rax,%r8), %xmm4, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps 0x10(%rax,%r8), %xmm6, %xmm4 vmulps 0x20(%rax,%r8), %xmm11, %xmm5 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm5, %xmm4, %xmm3 vmulps 0x30(%rax,%r8), %xmm12, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps 0x40(%rax,%r8), %xmm14, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps %xmm1, (%r10,%r8) vmovaps %xmm2, (%r9,%r8) addq $0x10, %r8 jmp 0x2a6af2 addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp addq %rdi, %rdx addq %r8, %rdx addq %rdi, %rax addq %r8, %rax movq 0x50(%rsp), %rcx addq %rcx, %r10 addq %r8, %r10 addq %rcx, %r9 addq %r8, %r9 addl $0x2, %esi movq 0x20(%rsp), %r8 jmp 0x2a6add movl %r8d, %ecx xorl %eax, %eax subl $0x1, %ecx jb 0x2a707c vmovaps (%r11,%rbx), %xmm1 vmovaps 0x10(%r11,%rbx), %xmm2 vmovaps 0x20(%r11,%rbx), %xmm3 vmovaps 0x30(%r11,%rbx), %xmm4 vmulps (%r15,%rax), %xmm1, %xmm1 vmovaps 0x40(%r11,%rbx), %xmm5 vaddps %xmm0, %xmm1, %xmm1 vmulps 0x10(%r15,%rax), %xmm2, %xmm2 vmulps 0x20(%r15,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x30(%r15,%rax), %xmm4, %xmm2 vmulps 0x40(%r15,%rax), %xmm5, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0x50(%r11,%rbx), %xmm3 vmovaps 0x60(%r11,%rbx), %xmm4 vmovaps 0x70(%r11,%rbx), %xmm5 vmovaps 0x80(%r11,%rbx), %xmm6 vmovaps 0x90(%r11,%rbx), %xmm7 vmulps (%r12,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x10(%r12,%rax), %xmm4, %xmm2 vmulps 0x20(%r12,%rax), %xmm5, %xmm3 vmulps 0x30(%r12,%rax), %xmm6, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x40(%r12,%rax), %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0xa0(%r11,%rbx), %xmm2 vmovaps 0xb0(%r11,%rbx), %xmm3 vmovaps 0xc0(%r11,%rbx), %xmm4 vmovaps 0xd0(%r11,%rbx), %xmm5 vmulps (%r13,%rax), %xmm2, %xmm2 vmovaps 0xe0(%r11,%rbx), %xmm6 vmulps 0x10(%r13,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x20(%r13,%rax), %xmm4, %xmm3 vmulps 0x30(%r13,%rax), %xmm5, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x40(%r13,%rax), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0xf0(%r11,%rbx), %xmm2 vmovaps 0x100(%r11,%rbx), %xmm3 vmovaps 0x110(%r11,%rbx), %xmm4 vmovaps 0x120(%r11,%rbx), %xmm5 vmulps (%rbp,%rax), %xmm2, %xmm2 vmovaps 0x130(%r11,%rbx), %xmm6 vmulps 0x10(%rbp,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x20(%rbp,%rax), %xmm4, %xmm3 vmulps 0x30(%rbp,%rax), %xmm5, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x40(%rbp,%rax), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0x140(%r11,%rbx), %xmm3 vmovaps 0x150(%r11,%rbx), %xmm4 vmovaps 0x160(%r11,%rbx), %xmm5 vmovaps 0x170(%r11,%rbx), %xmm6 vmovaps 0x180(%r11,%rbx), %xmm7 vmulps (%rdx,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x10(%rdx,%rax), %xmm4, %xmm2 vmulps 0x20(%rdx,%rax), %xmm5, %xmm3 vmulps 0x30(%rdx,%rax), %xmm6, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x40(%rdx,%rax), %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps %xmm1, (%r10,%rax) addq $0x10, %rax jmp 0x2a6e7d addq %rax, %r15 addq $0x40, %r15 leaq 0x40(%r12,%rax), %r12 addq %rax, %r13 addq $0x40, %r13 addq %rax, %rbp addq $0x40, %rbp addq %rax, %rdx addq $0x40, %rdx incl %esi addq %rax, %r10 cmpl 0x10(%rsp), %esi jl 0x2a6e78 movq 0x70(%rsp), %rsi incq %rsi movq 0x18(%rsp), %r13 movq 0x110(%rsp), %rdx jmp 0x2a6a3b cmpl $0x5, %r15d jne 0x2a93b2 cmpl $0x1, 0xdc(%rdi) jne 0x2a9040 cmpl $0x1, 0xe0(%rdi) jne 0x2a9040 cmpl $0x1, 0xe4(%rdi) jne 0x2a9040 cmpl $0x1, 0xe8(%rdi) jne 0x2a9040 movl 0x2c(%r13), %esi movl 0x30(%r13), %ecx movl 0xc8(%rsp), %eax xorl %edx, %edx testl %esi, %esi cmovlel %edx, %esi movq 0x1b0(%rdi), %rdi testl %ecx, %ecx cmovlel %edx, %ecx testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x20(%rsp) movl $0x80, %r8d movq %rdi, 0x50(%rsp) cmpq 0x20(%rsp), %rdx je 0x2a9725 testq %rdi, %rdi je 0x2a7159 movq %rdx, %rax shlq $0x5, %rax vmovups (%rdi,%rax), %ymm0 jmp 0x2a715d vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %rdx, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %rdx, %r11 imulq 0x38(%r14), %r11 movq 0xd0(%rsp), %rbx movq %rdx, 0x10(%rsp) imulq %rdx, %rbx movq 0xa0(%rsp), %rax imulq %rax, %rbx addq 0x90(%rsp), %rbx movslq 0xbc(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x2a7413 movl %esi, %eax xorl %edi, %edi subl $0x1, %eax jb 0x2a73eb vmovaps (%r10,%r11), %ymm1 vmovaps 0x20(%r10,%r11), %ymm2 vmovaps 0x40(%r10,%r11), %ymm3 vmovaps 0x60(%r10,%r11), %ymm4 vmulps (%rbx,%rdi), %ymm1, %ymm1 vmovaps 0x80(%r10,%r11), %ymm5 vaddps %ymm0, %ymm1, %ymm1 vmulps 0x20(%rbx,%rdi), %ymm2, %ymm2 vmulps 0x40(%rbx,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x60(%rbx,%rdi), %ymm4, %ymm2 vmulps 0x80(%rbx,%rdi), %ymm5, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmovaps 0xa0(%r10,%r11), %ymm3 vmovaps 0xc0(%r10,%r11), %ymm4 vmovaps 0xe0(%r10,%r11), %ymm5 vmovaps 0x100(%r10,%r11), %ymm6 vmovaps 0x120(%r10,%r11), %ymm7 vmulps (%r15,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x20(%r15,%rdi), %ymm4, %ymm2 vmulps 0x40(%r15,%rdi), %ymm5, %ymm3 vmulps 0x60(%r15,%rdi), %ymm6, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x80(%r15,%rdi), %ymm7, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0x140(%r10,%r11), %ymm2 vmovaps 0x160(%r10,%r11), %ymm3 vmovaps 0x180(%r10,%r11), %ymm4 vmovaps 0x1a0(%r10,%r11), %ymm5 vmulps (%r12,%rdi), %ymm2, %ymm2 vmovaps 0x1c0(%r10,%r11), %ymm6 vmulps 0x20(%r12,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x40(%r12,%rdi), %ymm4, %ymm3 vmulps 0x60(%r12,%rdi), %ymm5, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x80(%r12,%rdi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0x1e0(%r10,%r11), %ymm2 vmovaps 0x200(%r10,%r11), %ymm3 vmovaps 0x220(%r10,%r11), %ymm4 vmovaps 0x240(%r10,%r11), %ymm5 vmulps (%r13,%rdi), %ymm2, %ymm2 vmovaps 0x260(%r10,%r11), %ymm6 vmulps 0x20(%r13,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x40(%r13,%rdi), %ymm4, %ymm3 vmulps 0x60(%r13,%rdi), %ymm5, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x80(%r13,%rdi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmovaps 0x280(%r10,%r11), %ymm3 vmovaps 0x2a0(%r10,%r11), %ymm4 vmovaps 0x2c0(%r10,%r11), %ymm5 vmovaps 0x2e0(%r10,%r11), %ymm6 vmovaps 0x300(%r10,%r11), %ymm7 vmulps (%rbp,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x20(%rbp,%rdi), %ymm4, %ymm2 vmulps 0x40(%rbp,%rdi), %ymm5, %ymm3 vmulps 0x60(%rbp,%rdi), %ymm6, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x80(%rbp,%rdi), %ymm7, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps %ymm1, (%r9,%rdi) addq $0x20, %rdi jmp 0x2a71d1 addq %rdi, %rbx addq %r8, %rbx addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp incl %edx addq %rdi, %r9 jmp 0x2a71c5 movq 0x10(%rsp), %rdx incq %rdx movq 0x18(%rsp), %r13 movq 0x50(%rsp), %rdi jmp 0x2a713b cmpl $0x1, 0xdc(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x1, 0xe0(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x2, 0xe4(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x2, 0xe8(%rdi) movl %eax, %ecx jne 0x2a75fb movslq 0xbc(%rsp), %rdx movl 0x2c(%r13), %eax movl 0x30(%r13), %r8d movl 0xc8(%rsp), %esi movl %edx, %ecx subl %eax, %ecx addl %ecx, %ecx movq 0x28(%r14), %r10 movq 0x1b0(%rdi), %rdi imulq 0x10(%r13), %r11 movq 0x90(%rsp), %rbx movq %rbx, 0xf0(%rsp) movq 0xa0(%rsp), %rbx imulq 0xd0(%rsp), %rbx movq %rbx, 0x110(%rsp) movq %rdx, 0x20(%rsp) addq %rdx, %rdx movq %rdx, 0x130(%rsp) movslq %ecx, %r15 xorl %r12d, %r12d testl %r8d, %r8d cmovlel %r12d, %r8d testl %esi, %esi cmovlel %r12d, %esi movq %rsi, 0x10(%rsp) shlq $0x2, %r15 movq %rdi, 0x70(%rsp) cmpq 0x10(%rsp), %r12 je 0x2a9725 testq %rdi, %rdi je 0x2a7505 vmovss (%rdi,%r12,4), %xmm0 jmp 0x2a7509 vxorps %xmm0, %xmm0, %xmm0 movq %r11, %rdi movq %r11, %r13 imulq %r12, %r13 addq %r9, %r13 imulq $0x24, %r12, %rbp movq 0x110(%rsp), %r11 imulq %r12, %r11 addq 0xf0(%rsp), %r11 movq 0x20(%rsp), %rcx leaq (%r11,%rcx,4), %rsi movq 0x130(%rsp), %rcx leaq (%r11,%rcx,4), %rbx xorl %r9d, %r9d cmpl %r8d, %r9d je 0x2a75e1 xorl %edx, %edx movl %eax, %ecx testl %ecx, %ecx jle 0x2a75c7 vmovsd (%r11,%rdx), %xmm1 vinsertps $0x20, 0x8(%r11,%rdx), %xmm1, %xmm1 # xmm1 = xmm1[0,1],mem[0],xmm1[3] vinsertps $0x30, (%rsi,%rdx), %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],mem[0] vmovsd 0x4(%rsi,%rdx), %xmm2 vinsertf128 $0x1, %xmm2, %ymm1, %ymm2 vmovsd (%rbx,%rdx), %xmm3 vinsertf128 $0x1, %xmm3, %ymm1, %ymm1 vshufpd $0x2, %ymm1, %ymm2, %ymm1 # ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[2] vmulps (%r10,%rbp), %ymm1, %ymm1 vmovss 0x20(%r10,%rbp), %xmm2 vmulss 0x8(%rbx,%rdx), %xmm2, %xmm2 vextractf128 $0x1, %ymm1, %xmm3 vhaddps %xmm1, %xmm3, %xmm1 vaddss %xmm0, %xmm2, %xmm2 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm1, %xmm2, %xmm1 vmovss %xmm1, (%r13) addq $0x4, %r13 decl %ecx addq $0x8, %rdx jmp 0x2a7553 addq %r15, %r11 addq %rdx, %r11 addq %r15, %rsi addq %rdx, %rsi addq %r15, %rbx addq %rdx, %rbx incl %r9d jmp 0x2a7546 incq %r12 movq 0x18(%rsp), %r13 movq 0x50(%rsp), %r9 movq %rdi, %r11 movq 0x70(%rsp), %rdi jmp 0x2a74ed cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %r8d movq 0x48(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x2a7641 xorl %eax, %eax testb $0x3, %dil sete %al testb $0x7, %dil leal 0x1(%rax,%rax,2), %eax cmovel %ebp, %eax movl %eax, 0x20(%rsp) xorl %eax, %eax testb $0x3, %r8b sete %al testb $0x7, %r8b leal 0x1(%rax,%rax,2), %eax cmovel %ebp, %eax jmp 0x2a7648 pushq $0x1 popq %rax movl %eax, 0x20(%rsp) movl %eax, 0x10(%rsp) movq 0x70(%rsp), %r15 movq 0x98(%rsp), %rax vmovaps 0x90(%rsp), %xmm0 vmovaps %xmm0, 0x210(%rsp) movq 0xa0(%rsp), %rcx movq %rcx, 0x220(%rsp) movl 0xa8(%rsp), %ecx movl %ecx, 0x228(%rsp) movq 0xb0(%rsp), %rcx movq %rcx, 0x230(%rsp) vmovups 0xb8(%rsp), %xmm0 vmovups %xmm0, 0x238(%rsp) movl 0xc8(%rsp), %ecx movl %ecx, 0x248(%rsp) movq 0xd0(%rsp), %rcx movq %rcx, 0x250(%rsp) testq %rax, %rax je 0x2a76d1 lock incl (%rax) cmpl 0x20(%rsp), %r10d jle 0x2a7725 movl %r8d, %ebx movl %edi, %ebp movq 0x48(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0x1a0(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x90(%rsp), %rdi leaq 0x210(%rsp), %rsi movl 0x20(%rsp), %edx vzeroupper callq 0x64e3b movl %ebp, %edi movl %ebx, %r8d movq 0x18(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm0 vmovaps %xmm0, 0x1a0(%rsp) movq 0x10(%rdx), %rcx movq %rcx, 0x1b0(%rsp) movl 0x18(%rdx), %ecx movl %ecx, 0x1b8(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x1c0(%rsp) vmovups 0x28(%rdx), %xmm0 vmovups %xmm0, 0x1c8(%rsp) movl 0x38(%rdx), %ecx movl %ecx, 0x1d8(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x1e0(%rsp) testq %rax, %rax je 0x2a7789 lock incl (%rax) cmpl %r15d, 0x10(%rsp) jae 0x2a7821 movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd0(%r14,%rax), %eax cltd movl 0x10(%rsp), %r9d idivl %r9d movl %eax, %ecx movq 0x170(%rsp), %rax xorl %edx, %edx divq %r15 movl %edi, %ebp movl %r8d, %ebx movl %r9d, %r8d imulq %rax, %r8 movq 0x48(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0x1a0(%rsp), %rdi movl 0x110(%rsp), %esi movl 0x130(%rsp), %edx callq 0x628f2 pushq $-0x64 popq %r12 cmpq $0x0, 0x1a0(%rsp) je 0x2a9662 movl %ebx, %r8d movl %ebp, %edi movslq 0x1d8(%rsp), %rax imulq 0x1e0(%rsp), %rax testq %rax, %rax je 0x2a9662 xorl %r15d, %r15d xorl %r12d, %r12d xorl %ebp, %ebp movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rax cmpq %rax, %rbp jge 0x2a7a8f movl %r15d, %eax cltd movl 0x20(%rsp), %esi idivl %esi movl %eax, %ecx movl %edi, %r13d movl %edi, %eax cltd idivl %esi movslq %ecx, %rdx imulq 0x250(%rsp), %rdx movq 0x220(%rsp), %rcx imulq %rcx, %rdx addq 0x210(%rsp), %rdx movl 0x228(%rsp), %esi movq 0x230(%rsp), %rdi movq %rdx, 0x308(%rsp) andq $0x0, 0x310(%rsp) movq %rcx, 0x318(%rsp) movl %esi, 0x320(%rsp) movq %rdi, 0x328(%rsp) movl %eax, 0x340(%rsp) vmovups 0x238(%rsp), %xmm0 movslq 0x244(%rsp), %rax movslq 0x23c(%rsp), %rdx movslq 0x240(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x348(%rsp) vmovups %xmm0, 0x330(%rsp) movl %r12d, %eax cltd movl 0x10(%rsp), %esi idivl %esi movl %eax, %ecx movl %r8d, %ebx movl %r8d, %eax cltd idivl %esi movslq %ecx, %rdx imulq 0x1e0(%rsp), %rdx movq 0x1b0(%rsp), %rsi imulq %rsi, %rdx addq 0x1a0(%rsp), %rdx movl 0x1b8(%rsp), %edi movq 0x1c0(%rsp), %rcx movq %rdx, 0x2c0(%rsp) andq $0x0, 0x2c8(%rsp) movq %rsi, 0x2d0(%rsp) movl %edi, 0x2d8(%rsp) movq %rcx, 0x2e0(%rsp) movl %eax, 0x2f8(%rsp) vmovups 0x1c8(%rsp), %xmm0 movslq 0x1d4(%rsp), %rax movslq 0x1cc(%rsp), %rdx movslq 0x1d0(%rsp), %rdi imulq %rdx, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x300(%rsp) vmovups %xmm0, 0x2e8(%rsp) movq 0x10(%r14), %rax movq (%rax,%rbp,8), %rdi movq 0x48(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 vmovups %ymm0, 0x370(%rsp) vmovups %ymm1, 0x390(%rsp) movq %rcx, 0x378(%rsp) movq (%rdi), %rax leaq 0x308(%rsp), %rsi leaq 0x2c0(%rsp), %rdx leaq 0x370(%rsp), %rcx vzeroupper callq 0x38(%rax) movq 0x2c8(%rsp), %rax testq %rax, %rax je 0x2a7a44 lock decl (%rax) jne 0x2a7a44 movq 0x2c0(%rsp), %rsi movq 0x2e0(%rsp), %rdi testq %rdi, %rdi je 0x2a7a3c movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a7a44 movq %rsi, %rdi callq 0x5f3e0 movq 0x310(%rsp), %rax testq %rax, %rax je 0x2a7a7b lock decl (%rax) jne 0x2a7a7b movq 0x308(%rsp), %rsi movq 0x328(%rsp), %rdi testq %rdi, %rdi je 0x2a7a73 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a7a7b movq %rsi, %rdi callq 0x5f3e0 incq %rbp movl %ebx, %r8d addl %ebx, %r12d movl %r13d, %edi addl %r13d, %r15d jmp 0x2a7829 movq 0x70(%rsp), %rdx cmpl %edx, 0x10(%rsp) jae 0x2a7ab9 xorl %r12d, %r12d leaq 0x1a0(%rsp), %rdi movq 0x18(%rsp), %rsi movq 0x48(%rsp), %rcx callq 0x64e3b jmp 0x2a9662 xorl %r12d, %r12d leaq 0x1a0(%rsp), %rax movq 0x18(%rsp), %rcx cmpq %rcx, %rax je 0x2a9662 movq 0x1a8(%rsp), %rax testq %rax, %rax je 0x2a7ae2 lock incl (%rax) movq 0x8(%rcx), %rax testq %rax, %rax je 0x2a960a lock decl (%rax) jne 0x2a960a movq 0x18(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x2a9602 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a960a pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x2a841c cmpl $0x1, 0xe0(%rdi) jne 0x2a841c cmpl $0x2, 0xe4(%rdi) jne 0x2a841c cmpl $0x2, 0xe8(%rdi) jne 0x2a841c movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xbc(%rsp), %r8d movl 0xc8(%rsp), %edx subl %eax, %r8d shll $0x3, %r8d movq 0x1b0(%rdi), %rsi movslq %r8d, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x2a9725 testq %rsi, %rsi je 0x2a7ba8 movq %r8, %r9 shlq $0x4, %r9 vmovups (%rsi,%r9), %xmm15 jmp 0x2a7bad vxorps %xmm15, %xmm15, %xmm15 movq 0x40(%r13), %r9 imulq %r8, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq %r14, %r10 movq 0x28(%r14), %r14 movslq 0x54(%r10), %r15 imulq %r8, %r15 imulq 0x38(%r10), %r15 movslq 0xbc(%rsp), %rbx movq 0xd0(%rsp), %r10 imulq %r8, %r10 movq 0xa0(%rsp), %r11 imulq %r11, %r10 addq 0x90(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx vmovaps (%r14,%r15), %xmm3 vmovaps 0x10(%r14,%r15), %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovaps 0x20(%r14,%r15), %xmm0 vmovaps %xmm0, 0x170(%rsp) vmovaps 0x30(%r14,%r15), %xmm6 vmovaps 0x40(%r14,%r15), %xmm7 vmovaps 0x50(%r14,%r15), %xmm14 vmovaps 0x60(%r14,%r15), %xmm0 vmovaps %xmm0, 0x50(%rsp) vmovaps 0x70(%r14,%r15), %xmm0 vmovaps %xmm0, 0xf0(%rsp) vmovaps 0x80(%r14,%r15), %xmm0 vmovaps %xmm0, 0x130(%rsp) xorl %ebp, %ebp vmovaps %xmm14, 0x1f0(%rsp) vmovaps %xmm3, 0x70(%rsp) vmovaps %xmm15, 0x110(%rsp) cmpl %ecx, %ebp je 0x2a80b8 xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r14d cmpl %eax, %r14d jge 0x2a7ed4 vmulps (%r10,%r15), %xmm3, %xmm10 vaddps %xmm15, %xmm10, %xmm11 vmovaps 0x20(%rsp), %xmm0 vmulps 0x10(%r10,%r15), %xmm0, %xmm12 vmovaps 0x20(%r10,%r15), %xmm14 vmovaps 0x40(%r10,%r15), %xmm10 vmovaps 0x170(%rsp), %xmm8 vmulps %xmm8, %xmm14, %xmm13 vaddps %xmm13, %xmm12, %xmm12 vaddps %xmm12, %xmm11, %xmm12 vmulps (%r11,%r15), %xmm6, %xmm11 vmulps 0x10(%r11,%r15), %xmm7, %xmm13 vaddps %xmm13, %xmm11, %xmm13 vmovaps 0x20(%r11,%r15), %xmm15 vmovaps 0x40(%r11,%r15), %xmm11 vmovaps 0x1f0(%rsp), %xmm9 vmulps %xmm9, %xmm15, %xmm0 vaddps %xmm0, %xmm13, %xmm0 vaddps %xmm0, %xmm12, %xmm0 vmovaps 0x50(%rsp), %xmm5 vmulps (%rbx,%r15), %xmm5, %xmm12 vmovaps 0xf0(%rsp), %xmm4 vmulps 0x10(%rbx,%r15), %xmm4, %xmm13 vaddps %xmm13, %xmm12, %xmm12 vmovaps 0x20(%rbx,%r15), %xmm1 vmovaps 0x40(%rbx,%r15), %xmm13 vmovaps 0x130(%rsp), %xmm3 vmulps %xmm3, %xmm1, %xmm2 vaddps %xmm2, %xmm12, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x20(%rsp), %xmm2 vmulps 0x30(%r10,%r15), %xmm2, %xmm2 vmulps 0x30(%r11,%r15), %xmm7, %xmm12 vaddps %xmm2, %xmm12, %xmm2 vmulps 0x30(%rbx,%r15), %xmm4, %xmm12 vaddps %xmm2, %xmm12, %xmm2 vmovaps %xmm0, (%r9) vmulps 0x70(%rsp), %xmm14, %xmm0 vaddps 0x110(%rsp), %xmm0, %xmm0 vmulps %xmm8, %xmm10, %xmm12 vmulps %xmm6, %xmm15, %xmm14 vmulps %xmm9, %xmm11, %xmm15 vaddps %xmm15, %xmm12, %xmm12 vmulps %xmm5, %xmm1, %xmm1 vaddps %xmm1, %xmm14, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm2, %xmm0, %xmm0 vmulps %xmm3, %xmm13, %xmm1 vaddps %xmm1, %xmm12, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmovaps 0x60(%r10,%r15), %xmm12 vmovaps 0x60(%r11,%r15), %xmm14 vmovaps 0x20(%rsp), %xmm1 vmulps 0x50(%r10,%r15), %xmm1, %xmm1 vmulps 0x50(%r11,%r15), %xmm7, %xmm2 vmovaps 0x60(%rbx,%r15), %xmm15 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x50(%rbx,%r15), %xmm4, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps %xmm0, 0x10(%r9) vmulps %xmm8, %xmm12, %xmm0 vmulps %xmm9, %xmm14, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmulps 0x70(%rsp), %xmm10, %xmm2 vaddps 0x110(%rsp), %xmm2, %xmm2 vmulps %xmm6, %xmm11, %xmm10 vmulps %xmm5, %xmm13, %xmm11 vaddps %xmm11, %xmm10, %xmm10 vaddps %xmm2, %xmm10, %xmm2 vaddps %xmm1, %xmm2, %xmm1 vmulps %xmm3, %xmm15, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x20(%rsp), %xmm2 vmulps 0x70(%r10,%r15), %xmm2, %xmm2 vmulps 0x80(%r10,%r15), %xmm8, %xmm10 vmulps 0x70(%r11,%r15), %xmm7, %xmm11 vaddps %xmm0, %xmm1, %xmm0 vaddps %xmm2, %xmm11, %xmm1 vmulps 0x80(%r11,%r15), %xmm9, %xmm2 vaddps %xmm2, %xmm10, %xmm2 vmulps 0x70(%rbx,%r15), %xmm4, %xmm10 vaddps %xmm1, %xmm10, %xmm1 vmulps 0x80(%rbx,%r15), %xmm3, %xmm10 vaddps %xmm2, %xmm10, %xmm2 vmovaps %xmm0, 0x20(%r9) vmulps %xmm6, %xmm14, %xmm0 vmulps %xmm5, %xmm15, %xmm10 vmovaps 0x110(%rsp), %xmm15 vmovaps 0x70(%rsp), %xmm3 vaddps %xmm0, %xmm10, %xmm0 vmulps %xmm3, %xmm12, %xmm10 vaddps %xmm15, %xmm10, %xmm10 vaddps %xmm0, %xmm10, %xmm0 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm2, %xmm0, %xmm0 vmovaps %xmm0, 0x30(%r9) addq $0x40, %r9 addl $0x4, %r12d subq $-0x80, %r15 jmp 0x2a7c96 vmovaps 0x1f0(%rsp), %xmm14 leal 0x1(%r12), %r14d cmpl %eax, %r14d jge 0x2a8096 vmovaps 0x20(%r10,%r15), %xmm10 vmovaps 0x20(%r11,%r15), %xmm11 vmulps (%r10,%r15), %xmm3, %xmm0 vmovaps 0x20(%rbx,%r15), %xmm12 vaddps %xmm0, %xmm15, %xmm0 vmovaps 0x20(%rsp), %xmm13 vmulps 0x10(%r10,%r15), %xmm13, %xmm1 vmovaps 0x170(%rsp), %xmm5 vmulps %xmm5, %xmm10, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps (%r11,%r15), %xmm6, %xmm2 vaddps %xmm1, %xmm0, %xmm0 vmulps 0x10(%r11,%r15), %xmm7, %xmm1 vaddps %xmm1, %xmm2, %xmm1 vmulps %xmm14, %xmm11, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0x50(%rsp), %xmm9 vmulps (%rbx,%r15), %xmm9, %xmm2 vaddps %xmm1, %xmm0, %xmm0 vmovaps 0xf0(%rsp), %xmm8 vmulps 0x10(%rbx,%r15), %xmm8, %xmm1 vaddps %xmm1, %xmm2, %xmm1 vmovaps 0x130(%rsp), %xmm4 vmulps %xmm4, %xmm12, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x30(%r10,%r15), %xmm13, %xmm2 vmulps 0x40(%r10,%r15), %xmm5, %xmm13 vmovaps %xmm14, %xmm5 vmulps 0x30(%r11,%r15), %xmm7, %xmm14 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm2, %xmm14, %xmm1 vmovaps %xmm5, %xmm14 vmulps 0x40(%r11,%r15), %xmm5, %xmm2 vaddps %xmm2, %xmm13, %xmm2 vmulps 0x30(%rbx,%r15), %xmm8, %xmm13 vaddps %xmm1, %xmm13, %xmm1 vmulps 0x40(%rbx,%r15), %xmm4, %xmm13 vaddps %xmm2, %xmm13, %xmm2 vmovaps %xmm0, (%r9) vmulps %xmm6, %xmm11, %xmm0 vmulps %xmm9, %xmm12, %xmm11 vaddps %xmm0, %xmm11, %xmm0 vmulps %xmm3, %xmm10, %xmm10 vaddps %xmm15, %xmm10, %xmm10 vaddps %xmm0, %xmm10, %xmm0 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm2, %xmm0, %xmm0 vmovaps %xmm0, 0x10(%r9) addq $0x20, %r9 addl $0x2, %r12d addq $0x40, %r15 jmp 0x2a7edd vmulps (%r10,%r15), %xmm3, %xmm0 vaddps %xmm0, %xmm15, %xmm0 vmovaps 0x20(%rsp), %xmm1 vmulps 0x10(%r10,%r15), %xmm1, %xmm1 vmovaps 0x170(%rsp), %xmm2 vmulps 0x20(%r10,%r15), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmulps (%r11,%r15), %xmm6, %xmm1 vmulps 0x10(%r11,%r15), %xmm7, %xmm2 vmulps 0x20(%r11,%r15), %xmm14, %xmm10 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm10, %xmm1 vmovaps 0x50(%rsp), %xmm2 vmulps (%rbx,%r15), %xmm2, %xmm2 vaddps %xmm1, %xmm0, %xmm0 vmovaps 0xf0(%rsp), %xmm1 vmulps 0x10(%rbx,%r15), %xmm1, %xmm1 vaddps %xmm1, %xmm2, %xmm1 vmovaps 0x130(%rsp), %xmm2 vmulps 0x20(%rbx,%r15), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmovaps %xmm0, (%r9) addq $0x10, %r9 incl %r12d addq $0x20, %r15 cmpl %eax, %r12d jl 0x2a7fff addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x2a7c88 incq %r8 movq 0x10(%rsp), %r14 jmp 0x2a7b8b pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x2a841c cmpl $0x1, 0xe0(%rdi) jne 0x2a841c cmpl $0x2, 0xe4(%rdi) jne 0x2a841c cmpl $0x2, 0xe8(%rdi) jne 0x2a841c movl 0x2c(%r13), %edx movl 0x30(%r13), %ecx movl 0xbc(%rsp), %eax movl 0xc8(%rsp), %esi subl %edx, %eax shll $0x3, %eax movq 0x1b0(%rdi), %r9 xorl %r15d, %r15d testl %edx, %edx cmovlel %r15d, %edx movl %edx, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %esi, %esi cmovlel %r15d, %esi movq %rsi, 0x20(%rsp) shlq $0x2, %r8 movq %r9, 0x50(%rsp) cmpq 0x20(%rsp), %r15 je 0x2a9725 testq %r9, %r9 je 0x2a8166 movq %r15, %rax shlq $0x4, %rax vmovups (%r9,%rax), %xmm0 jmp 0x2a816a vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r15, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %r15, %r11 imulq 0x38(%r14), %r11 movq 0xd0(%rsp), %rbx movq %r15, 0x10(%rsp) imulq %r15, %rbx movq 0xa0(%rsp), %rax imulq %rax, %rbx addq 0x90(%rsp), %rbx movslq 0xbc(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x2a8405 movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x2a83e0 vmovaps (%r10,%r11), %xmm1 vmovaps 0x10(%r10,%r11), %xmm2 vmovaps 0x20(%r10,%r11), %xmm3 vmovaps 0x30(%r10,%r11), %xmm4 vmovaps 0x40(%r10,%r11), %xmm5 vmulps (%rbx,%rsi), %xmm1, %xmm1 vmulps 0x10(%rbx,%rsi), %xmm2, %xmm2 vmulps 0x20(%rbx,%rsi), %xmm3, %xmm3 vaddps %xmm0, %xmm1, %xmm1 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x30(%rbx,%rsi), %xmm4, %xmm3 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x40(%rbx,%rsi), %xmm5, %xmm2 vaddps %xmm2, %xmm3, %xmm2 vmovaps 0x50(%r10,%r11), %xmm3 vmovaps 0x60(%r10,%r11), %xmm4 vmovaps 0x70(%r10,%r11), %xmm5 vmovaps 0x80(%r10,%r11), %xmm6 vmovaps 0x90(%r10,%r11), %xmm7 vmulps (%r15,%rsi), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x10(%r15,%rsi), %xmm4, %xmm3 vmulps 0x20(%r15,%rsi), %xmm5, %xmm4 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm4, %xmm3, %xmm2 vmulps 0x30(%r15,%rsi), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x40(%r15,%rsi), %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0xa0(%r10,%r11), %xmm2 vmovaps 0xb0(%r10,%r11), %xmm3 vmovaps 0xc0(%r10,%r11), %xmm4 vmovaps 0xd0(%r10,%r11), %xmm5 vmovaps 0xe0(%r10,%r11), %xmm6 vmulps (%r12,%rsi), %xmm2, %xmm2 vmulps 0x10(%r12,%rsi), %xmm3, %xmm3 vmulps 0x20(%r12,%rsi), %xmm4, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x30(%r12,%rsi), %xmm5, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x40(%r12,%rsi), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0xf0(%r10,%r11), %xmm2 vmovaps 0x100(%r10,%r11), %xmm3 vmovaps 0x110(%r10,%r11), %xmm4 vmovaps 0x120(%r10,%r11), %xmm5 vmovaps 0x130(%r10,%r11), %xmm6 vmulps (%r13,%rsi), %xmm2, %xmm2 vmulps 0x10(%r13,%rsi), %xmm3, %xmm3 vmulps 0x20(%r13,%rsi), %xmm4, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x30(%r13,%rsi), %xmm5, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x40(%r13,%rsi), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0x140(%r10,%r11), %xmm3 vmovaps 0x150(%r10,%r11), %xmm4 vmovaps 0x160(%r10,%r11), %xmm5 vmovaps 0x170(%r10,%r11), %xmm6 vmovaps 0x180(%r10,%r11), %xmm7 vmulps (%rbp,%rsi), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x10(%rbp,%rsi), %xmm4, %xmm3 vmulps 0x20(%rbp,%rsi), %xmm5, %xmm4 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm4, %xmm3, %xmm2 vmulps 0x30(%rbp,%rsi), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x40(%rbp,%rsi), %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps %xmm1, (%r9) addq $0x10, %r9 addq $0x20, %rsi jmp 0x2a81de addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x2a81d2 movq 0x10(%rsp), %r15 incq %r15 movq 0x18(%rsp), %r13 movq 0x50(%rsp), %r9 jmp 0x2a8147 imull %eax, %r15d movslq %r15d, %rsi leaq 0x210(%rsp), %rdi leaq 0x1a0(%rsp), %rdx callq 0x73bbe movq (%r14), %rcx movq -0x18(%rcx), %rdx movl 0x20(%rsp), %r11d imull 0xe0(%r14,%rdx), %r11d movq 0x210(%rsp), %rax movl 0xdc(%r14,%rdx), %esi imull 0xd4(%r14,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%r14,%rdx), %r8d jge 0x2a84ae movslq %esi, %rsi leaq (%rax,%rsi,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%r14,%rdx), %r9d jge 0x2a84a3 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%r14,%rdx), %edi incq %r9 jmp 0x2a8484 addl %r11d, %edi incl %r8d addq %r9, %rsi jmp 0x2a8470 leal (,%r15,4), %ecx movl %ecx, 0x170(%rsp) movl 0x110(%rsp), %ecx shll $0x2, %ecx xorl %edx, %edx testl %r15d, %r15d cmovlel %edx, %r15d movq 0xf0(%rsp), %r12 testl %r12d, %r12d cmovlel %edx, %r12d movq %r12, 0xf0(%rsp) movslq %ecx, %rcx movq %rcx, 0x70(%rsp) shlq $0x2, %r15 leaq 0x1501c6(%rip), %rdi # 0x3f86c0 vxorps %xmm0, %xmm0, %xmm0 vbroadcastss 0x145b0d(%rip), %xmm5 # 0x3ee014 vbroadcastss 0x146778(%rip), %xmm6 # 0x3eec88 xorl %ecx, %ecx vbroadcastss 0x148ca5(%rip), %xmm9 # 0x3f11c0 vbroadcastss 0x148ca8(%rip), %xmm10 # 0x3f11cc vbroadcastss 0x148ca3(%rip), %xmm11 # 0x3f11d0 vbroadcastss 0x148c9e(%rip), %xmm12 # 0x3f11d4 vbroadcastss 0x148c99(%rip), %xmm13 # 0x3f11d8 cmpq 0xf0(%rsp), %rcx je 0x2a95ea movq %rdx, 0x110(%rsp) movslq %edx, %r9 movq 0x40(%r13), %r10 imulq %rcx, %r10 imulq 0x10(%r13), %r10 shlq $0x2, %r9 addq (%r13), %r10 movslq 0xbc(%rsp), %r11 movq 0xd0(%rsp), %r13 imulq %rcx, %r13 movq %rcx, %rdx movq 0xa0(%rsp), %rcx imulq %rcx, %r13 addq 0x90(%rsp), %r13 imulq %rcx, %r11 movq %rdx, 0x130(%rsp) shlq $0x4, %rdx movq %rdx, 0x20(%rsp) addq 0x28(%r14), %r9 xorl %ecx, %ecx cmpl 0x50(%rsp), %ecx jg 0x2a8a9b movq (%r14), %rdx xorl %r8d, %r8d cmpq %rbx, %r8 jg 0x2a8a8b movq -0x18(%rdx), %r12 cmpl $0x0, 0x100(%r14,%r12) je 0x2a85ef movq 0x1b0(%r14,%r12), %rsi movq 0x20(%rsp), %rbp vmovups (%rsi,%rbp), %xmm1 jmp 0x2a85f3 vxorps %xmm1, %xmm1, %xmm1 movslq 0xe8(%r14,%r12), %rsi movq %r14, %rbp movslq %ecx, %r14 imulq %rsi, %r14 imulq %r11, %r14 addq %r13, %r14 movl 0xe4(%rbp,%r12), %esi imull %r8d, %esi shll $0x2, %esi movslq %esi, %rsi leaq (%r14,%rsi,4), %rsi xorl %r14d, %r14d cmpq %r14, %r15 je 0x2a8647 movslq (%rax,%r14), %rbp shlq $0x4, %rbp vmovups (%r9,%r14,4), %xmm2 vmulps (%rsi,%rbp), %xmm2, %xmm2 vaddps %xmm1, %xmm2, %xmm1 addq $0x4, %r14 jmp 0x2a8625 movq 0x10(%rsp), %r14 movl 0x110(%r14,%r12), %esi decl %esi cmpl $0x5, %esi ja 0x2a8a72 movslq (%rdi,%rsi,4), %rsi addq %rdi, %rsi vmaxps %xmm0, %xmm1, %xmm7 jmpq %rsi movq 0x118(%r14,%r12), %rsi vminps %xmm0, %xmm1, %xmm1 vbroadcastss (%rsi), %xmm15 vmulps %xmm1, %xmm15, %xmm1 vaddps %xmm7, %xmm1, %xmm7 jmp 0x2a8a76 vbroadcastss 0x148b25(%rip), %xmm4 # 0x3f11b8 vminps %xmm4, %xmm1, %xmm7 vbroadcastss 0x148b1c(%rip), %xmm8 # 0x3f11bc vmaxps %xmm7, %xmm8, %xmm7 vmulps %xmm7, %xmm9, %xmm14 vaddps %xmm5, %xmm14, %xmm14 vcvttps2dq %xmm14, %xmm15 vcvtdq2ps %xmm15, %xmm15 vcmpltps %xmm15, %xmm14, %xmm14 vandps %xmm6, %xmm14, %xmm14 vsubps %xmm14, %xmm15, %xmm14 vbroadcastss 0x149252(%rip), %xmm4 # 0x3f1920 vmulps %xmm4, %xmm14, %xmm15 vsubps %xmm15, %xmm7, %xmm7 vmulps %xmm7, %xmm7, %xmm15 vmulps %xmm7, %xmm10, %xmm2 vaddps %xmm2, %xmm11, %xmm2 vmulps %xmm7, %xmm2, %xmm2 vaddps %xmm2, %xmm12, %xmm2 vmulps %xmm7, %xmm2, %xmm2 vaddps %xmm2, %xmm13, %xmm2 vmulps %xmm7, %xmm2, %xmm2 vmovaps %xmm5, %xmm0 vbroadcastss 0x148ad8(%rip), %xmm5 # 0x3f11dc vaddps %xmm5, %xmm2, %xmm2 vmulps %xmm7, %xmm2, %xmm2 vaddps %xmm0, %xmm2, %xmm2 vmulps %xmm2, %xmm15, %xmm2 vaddps %xmm6, %xmm7, %xmm7 vaddps %xmm2, %xmm7, %xmm2 vcvttps2dq %xmm14, %xmm7 vpslld $0x17, %xmm7, %xmm7 vpaddd %xmm6, %xmm7, %xmm7 vmulps %xmm7, %xmm2, %xmm2 vaddps %xmm6, %xmm2, %xmm7 vbroadcastss 0x148aa5(%rip), %xmm2 # 0x3f11e0 vmaxps %xmm2, %xmm7, %xmm2 vpsrld $0x17, %xmm2, %xmm14 vbroadcastss 0x148a9b(%rip), %xmm3 # 0x3f11e8 vpaddd %xmm3, %xmm14, %xmm14 vbroadcastss 0x148a8a(%rip), %xmm3 # 0x3f11e4 vandps %xmm3, %xmm2, %xmm2 vorps %xmm0, %xmm2, %xmm2 vcvtdq2ps %xmm14, %xmm14 vbroadcastss 0x148a7c(%rip), %xmm3 # 0x3f11ec vcmpltps %xmm3, %xmm2, %xmm15 vandps %xmm2, %xmm15, %xmm3 vbroadcastss 0x148a6e(%rip), %xmm8 # 0x3f11f0 vaddps %xmm2, %xmm8, %xmm2 vaddps %xmm3, %xmm2, %xmm2 vandps %xmm6, %xmm15, %xmm3 vsubps %xmm3, %xmm14, %xmm3 vmulps %xmm2, %xmm2, %xmm14 vbroadcastss 0x148a55(%rip), %xmm15 # 0x3f11f4 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a4c(%rip), %xmm5 # 0x3f11f8 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a3f(%rip), %xmm5 # 0x3f11fc vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a32(%rip), %xmm5 # 0x3f1200 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a25(%rip), %xmm5 # 0x3f1204 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a18(%rip), %xmm5 # 0x3f1208 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a0b(%rip), %xmm5 # 0x3f120c vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x1489fe(%rip), %xmm5 # 0x3f1210 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x1489f1(%rip), %xmm5 # 0x3f1214 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x1490f0(%rip), %xmm5 # 0x3f1924 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm15, %xmm14, %xmm14 vcmpleps 0x14583a(%rip), %xmm7, %xmm7 # 0x3ee080 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm2, %xmm3, %xmm2 vaddps %xmm2, %xmm14, %xmm2 vbroadcastss 0x149ebd(%rip), %xmm3 # 0x3f2718 vmulps %xmm3, %xmm2, %xmm2 vbroadcastss 0x149098(%rip), %xmm3 # 0x3f1900 vblendvps %xmm7, %xmm3, %xmm2, %xmm2 vbroadcastss 0x148941(%rip), %xmm3 # 0x3f11b8 vminps %xmm3, %xmm2, %xmm2 vbroadcastss 0x148938(%rip), %xmm3 # 0x3f11bc vmaxps %xmm3, %xmm2, %xmm2 vmulps %xmm2, %xmm9, %xmm3 vaddps %xmm0, %xmm3, %xmm3 vcvttps2dq %xmm3, %xmm7 vcvtdq2ps %xmm7, %xmm7 vcmpltps %xmm7, %xmm3, %xmm3 vandps %xmm6, %xmm3, %xmm3 vsubps %xmm3, %xmm7, %xmm3 vmulps %xmm4, %xmm3, %xmm7 vsubps %xmm7, %xmm2, %xmm2 vmulps %xmm2, %xmm2, %xmm7 vmulps %xmm2, %xmm10, %xmm14 vaddps %xmm11, %xmm14, %xmm14 vmulps %xmm2, %xmm14, %xmm14 vaddps %xmm12, %xmm14, %xmm14 vmulps %xmm2, %xmm14, %xmm14 vaddps %xmm13, %xmm14, %xmm14 vmulps %xmm2, %xmm14, %xmm14 vbroadcastss 0x148903(%rip), %xmm4 # 0x3f11dc vaddps %xmm4, %xmm14, %xmm14 vmovaps %xmm0, %xmm5 vxorps %xmm0, %xmm0, %xmm0 vmulps %xmm2, %xmm14, %xmm14 vaddps %xmm5, %xmm14, %xmm14 vmulps %xmm7, %xmm14, %xmm7 vaddps %xmm6, %xmm2, %xmm2 vaddps %xmm7, %xmm2, %xmm2 vcvttps2dq %xmm3, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm6, %xmm3, %xmm3 vmulps %xmm3, %xmm2, %xmm2 vaddps %xmm6, %xmm2, %xmm2 vrcpps %xmm2, %xmm3 vaddps %xmm3, %xmm3, %xmm7 vmulps %xmm7, %xmm2, %xmm2 vbroadcastss 0x149de5(%rip), %xmm4 # 0x3f2708 vsubps %xmm2, %xmm4, %xmm2 vmulps %xmm2, %xmm3, %xmm2 vaddps %xmm7, %xmm8, %xmm3 vaddps %xmm2, %xmm3, %xmm2 jmp 0x2a8a6c movq 0x118(%r14,%r12), %rsi vbroadcastss (%rsi), %xmm7 vbroadcastss 0x4(%rsi), %xmm15 vmaxps %xmm7, %xmm1, %xmm1 vminps %xmm1, %xmm15, %xmm7 jmp 0x2a8a76 vbroadcastss 0x148853(%rip), %xmm2 # 0x3f11b4 vxorps %xmm2, %xmm1, %xmm1 vbroadcastss 0x14884a(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm1, %xmm1 vbroadcastss 0x148841(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm1, %xmm1 vbroadcastss 0x148838(%rip), %xmm2 # 0x3f11c0 vmulps %xmm2, %xmm1, %xmm7 vaddps %xmm5, %xmm7, %xmm7 vcvttps2dq %xmm7, %xmm15 vcvtdq2ps %xmm15, %xmm15 vcmpltps %xmm15, %xmm7, %xmm7 vandps %xmm6, %xmm7, %xmm7 vsubps %xmm7, %xmm15, %xmm7 vbroadcastss 0x14a030(%rip), %xmm2 # 0x3f29e0 vmulps %xmm2, %xmm7, %xmm15 vaddps %xmm1, %xmm15, %xmm1 vmulps %xmm1, %xmm1, %xmm15 vbroadcastss 0x148807(%rip), %xmm2 # 0x3f11cc vmulps %xmm2, %xmm1, %xmm14 vbroadcastss 0x1487fe(%rip), %xmm2 # 0x3f11d0 vaddps %xmm2, %xmm14, %xmm14 vmulps %xmm1, %xmm14, %xmm14 vbroadcastss 0x1487f1(%rip), %xmm2 # 0x3f11d4 vaddps %xmm2, %xmm14, %xmm14 vmulps %xmm1, %xmm14, %xmm14 vbroadcastss 0x1487e4(%rip), %xmm2 # 0x3f11d8 vaddps %xmm2, %xmm14, %xmm14 vmulps %xmm1, %xmm14, %xmm14 vbroadcastss 0x1487d7(%rip), %xmm2 # 0x3f11dc vaddps %xmm2, %xmm14, %xmm14 vmulps %xmm1, %xmm14, %xmm14 vaddps %xmm5, %xmm14, %xmm14 vmulps %xmm14, %xmm15, %xmm14 vaddps %xmm6, %xmm1, %xmm1 vaddps %xmm1, %xmm14, %xmm1 vcvttps2dq %xmm7, %xmm7 vpslld $0x17, %xmm7, %xmm7 vpaddd %xmm6, %xmm7, %xmm7 vmulps %xmm7, %xmm1, %xmm1 vaddps %xmm6, %xmm1, %xmm1 vrcpps %xmm1, %xmm7 vmulps %xmm7, %xmm1, %xmm1 vsubps %xmm1, %xmm6, %xmm1 vmulps %xmm1, %xmm7, %xmm1 vaddps %xmm1, %xmm7, %xmm7 jmp 0x2a8a76 movq 0x118(%r14,%r12), %rsi vbroadcastss (%rsi), %xmm2 vbroadcastss 0x4(%rsi), %xmm3 vmulps %xmm1, %xmm2, %xmm2 vaddps %xmm3, %xmm2, %xmm2 vmaxps %xmm0, %xmm2, %xmm2 vminps %xmm6, %xmm2, %xmm2 vmulps %xmm1, %xmm2, %xmm7 jmp 0x2a8a76 vmovaps %xmm1, %xmm7 movq %r8, %rsi shlq $0x4, %rsi vmovups %xmm7, (%r10,%rsi) incq %r8 jmp 0x2a85c3 movq 0x70(%rsp), %rdx leaq (%r10,%rdx,4), %r10 incl %ecx jmp 0x2a85b3 movq 0x130(%rsp), %rcx incq %rcx movq 0x110(%rsp), %rdx addl 0x170(%rsp), %edx movq 0x18(%rsp), %r13 jmp 0x2a853f pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x2a93b2 cmpl $0x1, 0xe0(%rdi) jne 0x2a93b2 cmpl $0x2, 0xe4(%rdi) jne 0x2a93b2 cmpl $0x2, 0xe8(%rdi) jne 0x2a93b2 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xbc(%rsp), %r8d movl 0xc8(%rsp), %edx subl %eax, %r8d shll $0x4, %r8d movq 0x1b0(%rdi), %rsi movslq %r8d, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x2a9725 testq %rsi, %rsi je 0x2a8b4f movq %r8, %r9 shlq $0x5, %r9 vmovups (%rsi,%r9), %ymm10 jmp 0x2a8b54 vxorps %xmm10, %xmm10, %xmm10 movq 0x40(%r13), %r9 imulq %r8, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq %r14, %r10 movq 0x28(%r14), %r14 movslq 0x54(%r10), %r15 imulq %r8, %r15 imulq 0x38(%r10), %r15 movslq 0xbc(%rsp), %rbx movq 0xd0(%rsp), %r10 imulq %r8, %r10 movq 0xa0(%rsp), %r11 imulq %r11, %r10 addq 0x90(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx vmovaps (%r14,%r15), %ymm11 vmovaps 0x20(%r14,%r15), %ymm12 vmovaps 0x40(%r14,%r15), %ymm13 vmovaps 0x60(%r14,%r15), %ymm14 vmovaps 0x80(%r14,%r15), %ymm15 vmovaps 0xa0(%r14,%r15), %ymm6 vmovaps 0xc0(%r14,%r15), %ymm7 vmovaps 0xe0(%r14,%r15), %ymm8 vmovaps 0x100(%r14,%r15), %ymm9 xorl %ebp, %ebp vmovups %ymm10, 0x20(%rsp) vmovups %ymm11, 0x50(%rsp) vmovups %ymm12, 0x70(%rsp) vmovups %ymm13, 0xf0(%rsp) vmovups %ymm14, 0x110(%rsp) vmovups %ymm15, 0x130(%rsp) vmovups %ymm6, 0x170(%rsp) cmpl %ecx, %ebp je 0x2a9033 xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r14d cmpl %eax, %r14d jge 0x2a8e76 vmovaps 0x40(%r10,%r15), %ymm10 vmovaps 0x80(%r10,%r15), %ymm11 vmovaps 0xc0(%r10,%r15), %ymm12 vmovups 0x50(%rsp), %ymm0 vmulps (%r10,%r15), %ymm0, %ymm13 vmovups 0x20(%rsp), %ymm1 vaddps %ymm1, %ymm13, %ymm13 vmulps %ymm0, %ymm10, %ymm14 vaddps %ymm1, %ymm14, %ymm14 vmulps %ymm0, %ymm11, %ymm15 vaddps %ymm1, %ymm15, %ymm15 vmulps %ymm0, %ymm12, %ymm0 vaddps %ymm1, %ymm0, %ymm0 vmovups 0x70(%rsp), %ymm4 vmulps 0x20(%r10,%r15), %ymm4, %ymm1 vmulps 0x60(%r10,%r15), %ymm4, %ymm2 vmulps 0xa0(%r10,%r15), %ymm4, %ymm3 vmulps 0xe0(%r10,%r15), %ymm4, %ymm4 vmovups 0xf0(%rsp), %ymm5 vmulps %ymm5, %ymm10, %ymm10 vaddps %ymm1, %ymm10, %ymm1 vaddps %ymm1, %ymm13, %ymm1 vmulps %ymm5, %ymm11, %ymm10 vaddps %ymm2, %ymm10, %ymm2 vaddps %ymm2, %ymm14, %ymm2 vmulps %ymm5, %ymm12, %ymm10 vaddps %ymm3, %ymm10, %ymm3 vmulps 0x100(%r10,%r15), %ymm5, %ymm10 vaddps %ymm3, %ymm15, %ymm3 vaddps %ymm4, %ymm10, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmovaps 0x40(%r11,%r15), %ymm4 vmovaps 0x80(%r11,%r15), %ymm10 vmovaps 0xc0(%r11,%r15), %ymm11 vmovups 0x110(%rsp), %ymm5 vmulps (%r11,%r15), %ymm5, %ymm12 vmulps %ymm5, %ymm4, %ymm13 vmulps %ymm5, %ymm10, %ymm14 vmovups 0x130(%rsp), %ymm6 vmulps 0x20(%r11,%r15), %ymm6, %ymm15 vmulps %ymm5, %ymm11, %ymm5 vaddps %ymm15, %ymm12, %ymm12 vmulps 0x60(%r11,%r15), %ymm6, %ymm15 vaddps %ymm15, %ymm13, %ymm13 vmulps 0xa0(%r11,%r15), %ymm6, %ymm15 vaddps %ymm15, %ymm14, %ymm14 vmulps 0xe0(%r11,%r15), %ymm6, %ymm15 vaddps %ymm5, %ymm15, %ymm5 vmovups 0x170(%rsp), %ymm6 vmulps %ymm6, %ymm4, %ymm4 vaddps %ymm4, %ymm12, %ymm4 vaddps %ymm4, %ymm1, %ymm1 vmulps %ymm6, %ymm10, %ymm4 vaddps %ymm4, %ymm13, %ymm4 vaddps %ymm4, %ymm2, %ymm2 vmulps %ymm6, %ymm11, %ymm4 vaddps %ymm4, %ymm14, %ymm4 vaddps %ymm4, %ymm3, %ymm3 vmulps 0x100(%r11,%r15), %ymm6, %ymm4 vaddps %ymm4, %ymm5, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmovaps 0x40(%rbx,%r15), %ymm4 vmovaps 0x80(%rbx,%r15), %ymm5 vmulps (%rbx,%r15), %ymm7, %ymm10 vmovaps 0xc0(%rbx,%r15), %ymm11 vmulps %ymm7, %ymm4, %ymm12 vmulps 0x20(%rbx,%r15), %ymm8, %ymm13 vmulps %ymm7, %ymm5, %ymm14 vaddps %ymm13, %ymm10, %ymm10 vmulps 0x60(%rbx,%r15), %ymm8, %ymm13 vmulps %ymm7, %ymm11, %ymm15 vaddps %ymm13, %ymm12, %ymm12 vmulps 0xa0(%rbx,%r15), %ymm8, %ymm13 vaddps %ymm13, %ymm14, %ymm13 vmulps 0xe0(%rbx,%r15), %ymm8, %ymm14 vaddps %ymm14, %ymm15, %ymm14 vmulps %ymm4, %ymm9, %ymm4 vaddps %ymm4, %ymm10, %ymm4 vaddps %ymm4, %ymm1, %ymm1 vmulps %ymm5, %ymm9, %ymm4 vaddps %ymm4, %ymm12, %ymm4 vaddps %ymm4, %ymm2, %ymm2 vmulps %ymm9, %ymm11, %ymm4 vaddps %ymm4, %ymm13, %ymm4 vaddps %ymm4, %ymm3, %ymm3 vmulps 0x100(%rbx,%r15), %ymm9, %ymm4 vaddps %ymm4, %ymm14, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmovaps %ymm1, (%r9) vmovaps %ymm2, 0x20(%r9) vmovaps %ymm3, 0x40(%r9) vmovaps %ymm0, 0x60(%r9) subq $-0x80, %r9 addl $0x4, %r12d addq $0x100, %r15 # imm = 0x100 jmp 0x2a8c40 vmovups 0x20(%rsp), %ymm10 vmovups 0x50(%rsp), %ymm11 vmovups 0x70(%rsp), %ymm12 vmovups 0xf0(%rsp), %ymm13 vmovups 0x110(%rsp), %ymm14 vmovups 0x130(%rsp), %ymm15 vmovups 0x170(%rsp), %ymm6 leal 0x1(%r12), %r14d cmpl %eax, %r14d jge 0x2a9015 vmovaps 0x40(%r10,%r15), %ymm0 vmulps (%r10,%r15), %ymm11, %ymm1 vaddps %ymm1, %ymm10, %ymm1 vmulps %ymm0, %ymm11, %ymm2 vaddps %ymm2, %ymm10, %ymm2 vmulps 0x20(%r10,%r15), %ymm12, %ymm3 vmulps 0x60(%r10,%r15), %ymm12, %ymm4 vmulps %ymm0, %ymm13, %ymm0 vaddps %ymm0, %ymm3, %ymm0 vaddps %ymm0, %ymm1, %ymm0 vmulps 0x80(%r10,%r15), %ymm13, %ymm1 vaddps %ymm1, %ymm4, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vmovaps 0x40(%r11,%r15), %ymm2 vmulps (%r11,%r15), %ymm14, %ymm3 vmulps %ymm2, %ymm14, %ymm4 vmulps 0x20(%r11,%r15), %ymm15, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmulps 0x60(%r11,%r15), %ymm15, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps %ymm6, %ymm2, %ymm2 vaddps %ymm2, %ymm3, %ymm2 vaddps %ymm2, %ymm0, %ymm0 vmulps 0x80(%r11,%r15), %ymm6, %ymm2 vaddps %ymm2, %ymm4, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps (%rbx,%r15), %ymm7, %ymm2 vmovaps 0x40(%rbx,%r15), %ymm3 vmulps %ymm7, %ymm3, %ymm4 vmulps 0x20(%rbx,%r15), %ymm8, %ymm5 vaddps %ymm5, %ymm2, %ymm2 vmulps 0x60(%rbx,%r15), %ymm8, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps %ymm3, %ymm9, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%rbx,%r15), %ymm9, %ymm3 vaddps %ymm2, %ymm0, %ymm0 vaddps %ymm3, %ymm4, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps %ymm0, (%r9) vmovaps %ymm1, 0x20(%r9) addq $0x40, %r9 addl $0x2, %r12d subq $-0x80, %r15 jmp 0x2a8eac vmulps (%r10,%r15), %ymm11, %ymm0 vaddps %ymm0, %ymm10, %ymm0 vmulps 0x20(%r10,%r15), %ymm12, %ymm1 vmulps 0x40(%r10,%r15), %ymm13, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmulps (%r11,%r15), %ymm14, %ymm1 vmulps 0x20(%r11,%r15), %ymm15, %ymm2 vmulps 0x40(%r11,%r15), %ymm6, %ymm3 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm3, %ymm1, %ymm1 vmulps (%rbx,%r15), %ymm7, %ymm2 vaddps %ymm1, %ymm0, %ymm0 vmulps 0x20(%rbx,%r15), %ymm8, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vmulps 0x40(%rbx,%r15), %ymm9, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps %ymm0, (%r9) addq $0x20, %r9 incl %r12d addq $0x40, %r15 cmpl %eax, %r12d jl 0x2a8fa5 addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x2a8c32 incq %r8 movq 0x10(%rsp), %r14 jmp 0x2a8b32 pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x2a93b2 cmpl $0x1, 0xe0(%rdi) jne 0x2a93b2 cmpl $0x2, 0xe4(%rdi) jne 0x2a93b2 cmpl $0x2, 0xe8(%rdi) jne 0x2a93b2 movl 0x2c(%r13), %edx movl 0x30(%r13), %ecx movl 0xbc(%rsp), %eax movl 0xc8(%rsp), %esi subl %edx, %eax shll $0x4, %eax movq 0x1b0(%rdi), %r9 xorl %r15d, %r15d testl %edx, %edx cmovlel %r15d, %edx movl %edx, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %esi, %esi cmovlel %r15d, %esi movq %rsi, 0x20(%rsp) shlq $0x2, %r8 movq %r9, 0x50(%rsp) cmpq 0x20(%rsp), %r15 je 0x2a9725 testq %r9, %r9 je 0x2a90e1 movq %r15, %rax shlq $0x5, %rax vmovups (%r9,%rax), %ymm0 jmp 0x2a90e5 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r15, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %r15, %r11 imulq 0x38(%r14), %r11 movq 0xd0(%rsp), %rbx movq %r15, 0x10(%rsp) imulq %r15, %rbx movq 0xa0(%rsp), %rax imulq %rax, %rbx addq 0x90(%rsp), %rbx movslq 0xbc(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x2a939b movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x2a9376 vmovaps (%r10,%r11), %ymm1 vmovaps 0x20(%r10,%r11), %ymm2 vmovaps 0x40(%r10,%r11), %ymm3 vmovaps 0x60(%r10,%r11), %ymm4 vmovaps 0x80(%r10,%r11), %ymm5 vmulps (%rbx,%rsi), %ymm1, %ymm1 vmulps 0x20(%rbx,%rsi), %ymm2, %ymm2 vmulps 0x40(%rbx,%rsi), %ymm3, %ymm3 vaddps %ymm0, %ymm1, %ymm1 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x60(%rbx,%rsi), %ymm4, %ymm3 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x80(%rbx,%rsi), %ymm5, %ymm2 vaddps %ymm2, %ymm3, %ymm2 vmovaps 0xa0(%r10,%r11), %ymm3 vmovaps 0xc0(%r10,%r11), %ymm4 vmovaps 0xe0(%r10,%r11), %ymm5 vmovaps 0x100(%r10,%r11), %ymm6 vmovaps 0x120(%r10,%r11), %ymm7 vmulps (%r15,%rsi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x20(%r15,%rsi), %ymm4, %ymm3 vmulps 0x40(%r15,%rsi), %ymm5, %ymm4 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm4, %ymm3, %ymm2 vmulps 0x60(%r15,%rsi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%r15,%rsi), %ymm7, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0x140(%r10,%r11), %ymm2 vmovaps 0x160(%r10,%r11), %ymm3 vmovaps 0x180(%r10,%r11), %ymm4 vmovaps 0x1a0(%r10,%r11), %ymm5 vmovaps 0x1c0(%r10,%r11), %ymm6 vmulps (%r12,%rsi), %ymm2, %ymm2 vmulps 0x20(%r12,%rsi), %ymm3, %ymm3 vmulps 0x40(%r12,%rsi), %ymm4, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x60(%r12,%rsi), %ymm5, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%r12,%rsi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0x1e0(%r10,%r11), %ymm2 vmovaps 0x200(%r10,%r11), %ymm3 vmovaps 0x220(%r10,%r11), %ymm4 vmovaps 0x240(%r10,%r11), %ymm5 vmovaps 0x260(%r10,%r11), %ymm6 vmulps (%r13,%rsi), %ymm2, %ymm2 vmulps 0x20(%r13,%rsi), %ymm3, %ymm3 vmulps 0x40(%r13,%rsi), %ymm4, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x60(%r13,%rsi), %ymm5, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%r13,%rsi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmovaps 0x280(%r10,%r11), %ymm3 vmovaps 0x2a0(%r10,%r11), %ymm4 vmovaps 0x2c0(%r10,%r11), %ymm5 vmovaps 0x2e0(%r10,%r11), %ymm6 vmovaps 0x300(%r10,%r11), %ymm7 vmulps (%rbp,%rsi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x20(%rbp,%rsi), %ymm4, %ymm3 vmulps 0x40(%rbp,%rsi), %ymm5, %ymm4 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm4, %ymm3, %ymm2 vmulps 0x60(%rbp,%rsi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%rbp,%rsi), %ymm7, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps %ymm1, (%r9) addq $0x20, %r9 addq $0x40, %rsi jmp 0x2a9159 addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x2a914d movq 0x10(%rsp), %r15 incq %r15 movq 0x18(%rsp), %r13 movq 0x50(%rsp), %r9 jmp 0x2a90c2 imull %eax, %r15d movslq %r15d, %rsi leaq 0x210(%rsp), %rdi leaq 0x1a0(%rsp), %rdx callq 0x73bbe movq (%r14), %rcx movq -0x18(%rcx), %rdx movl 0x20(%rsp), %r11d imull 0xe0(%r14,%rdx), %r11d movq 0x210(%rsp), %rax movl 0xdc(%r14,%rdx), %esi imull 0xd4(%r14,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%r14,%rdx), %esi jge 0x2a9443 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%r14,%rdx), %r9d jge 0x2a9439 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%r14,%rdx), %edi incq %r9 jmp 0x2a941a addl %r11d, %edi incl %esi addq %r9, %r8 jmp 0x2a9406 leal (,%r15,8), %ecx movl %ecx, 0x130(%rsp) movl 0x110(%rsp), %ecx shll $0x3, %ecx movslq %ecx, %rcx movq %rcx, 0x20(%rsp) xorl %ecx, %ecx testl %r15d, %r15d cmovlel %ecx, %r15d movq 0xf0(%rsp), %r12 testl %r12d, %r12d cmovlel %ecx, %r12d movq %r12, 0xf0(%rsp) shlq $0x2, %r15 xorl %edx, %edx cmpq 0xf0(%rsp), %rdx je 0x2a95d0 movq %rcx, 0x70(%rsp) movslq %ecx, %r8 movq 0x40(%r13), %r9 imulq %rdx, %r9 imulq 0x10(%r13), %r9 shlq $0x2, %r8 addq (%r13), %r9 movslq 0xbc(%rsp), %r10 movq 0xd0(%rsp), %r11 imulq %rdx, %r11 movq 0xa0(%rsp), %rcx imulq %rcx, %r11 addq 0x90(%rsp), %r11 imulq %rcx, %r10 movq %rdx, 0x110(%rsp) movq %rdx, %r13 shlq $0x5, %r13 addq 0x28(%r14), %r8 xorl %ebp, %ebp cmpl 0x50(%rsp), %ebp jg 0x2a95af movq (%r14), %rcx xorl %r12d, %r12d cmpq %rbx, %r12 jg 0x2a959f movq -0x18(%rcx), %rsi cmpl $0x0, 0x100(%r14,%rsi) je 0x2a952e movq 0x1b0(%r14,%rsi), %rdi vmovups (%rdi,%r13), %ymm0 jmp 0x2a9532 vxorps %xmm0, %xmm0, %xmm0 movslq 0xe8(%r14,%rsi), %rdi movq %r14, %rdx movslq %ebp, %r14 imulq %rdi, %r14 imulq %r10, %r14 addq %r11, %r14 movl 0xe4(%rdx,%rsi), %esi imull %r12d, %esi shll $0x3, %esi movslq %esi, %rsi leaq (%r14,%rsi,4), %rsi xorl %edi, %edi cmpq %rdi, %r15 je 0x2a9585 movslq (%rax,%rdi), %r14 shlq $0x5, %r14 vmovups (%r8,%rdi,8), %ymm1 vmulps (%rsi,%r14), %ymm1, %ymm1 vaddps %ymm0, %ymm1, %ymm0 addq $0x4, %rdi jmp 0x2a9562 movq %r12, %rsi shlq $0x5, %rsi vmovups %ymm0, (%r9,%rsi) incq %r12 movq 0x10(%rsp), %r14 jmp 0x2a9506 movq 0x20(%rsp), %rcx leaq (%r9,%rcx,4), %r9 incl %ebp jmp 0x2a94f6 movq 0x110(%rsp), %rdx incq %rdx movq 0x70(%rsp), %rcx addl 0x130(%rsp), %ecx movq 0x18(%rsp), %r13 jmp 0x2a948a movq 0x8(%r14), %rdi testq %rdi, %rdi je 0x2a95ea movq (%rdi), %rax movq %r13, %rsi movq 0x48(%rsp), %rdx vzeroupper callq 0x48(%rax) leaq 0x210(%rsp), %rdi vzeroupper callq 0x624be xorl %r12d, %r12d jmp 0x2a96d0 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x1a0(%rsp), %xmm0 movq 0x18(%rsp), %rcx vmovups %xmm0, (%rcx) movq 0x1b0(%rsp), %rax movq %rax, 0x10(%rcx) movl 0x1b8(%rsp), %eax movl %eax, 0x18(%rcx) movq 0x1c0(%rsp), %rax movq %rax, 0x20(%rcx) vmovups 0x1c8(%rsp), %xmm0 vmovups %xmm0, 0x28(%rcx) movl 0x1d8(%rsp), %eax movl %eax, 0x38(%rcx) movq 0x1e0(%rsp), %rax movq %rax, 0x40(%rcx) movq 0x1a8(%rsp), %rax testq %rax, %rax je 0x2a9699 lock decl (%rax) jne 0x2a9699 movq 0x1a0(%rsp), %rsi movq 0x1c0(%rsp), %rdi testq %rdi, %rdi je 0x2a9691 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a9699 movq %rsi, %rdi callq 0x5f3e0 movq 0x218(%rsp), %rax testq %rax, %rax je 0x2a96d0 lock decl (%rax) jne 0x2a96d0 movq 0x210(%rsp), %rsi movq 0x230(%rsp), %rdi testq %rdi, %rdi je 0x2a96c8 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a96d0 movq %rsi, %rdi callq 0x5f3e0 movq 0x98(%rsp), %rax testq %rax, %rax je 0x2a970d lock decl (%rax) jne 0x2a970d movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi je 0x2a9702 movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x2a970d movq %rsi, %rdi vzeroupper callq 0x5f3e0 movl %r12d, %eax addq $0x3b8, %rsp # imm = 0x3B8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movq 0x8(%r14), %rdi testq %rdi, %rdi je 0x2a95fa movq (%rdi), %rax movq %r13, %rsi movq 0x48(%rsp), %rdx vzeroupper callq 0x48(%rax) jmp 0x2a95fa movq %rax, %rbx leaq 0x210(%rsp), %rdi callq 0x624be jmp 0x2a9873 jmp 0x2a978c jmp 0x2a978c jmp 0x2a98ba jmp 0x2a98ba movq %rax, %rbx jmp 0x2a983c movq %rax, %rbx jmp 0x2a9805 jmp 0x2a98ba jmp 0x2a978c jmp 0x2a98ba jmp 0x2a98ba movq %rax, %rbx jmp 0x2a9873 movq %rax, %rbx movq 0x2c8(%rsp), %rax testq %rax, %rax je 0x2a97ce lock decl (%rax) jne 0x2a97ce movq 0x2c0(%rsp), %rsi movq 0x2e0(%rsp), %rdi testq %rdi, %rdi jne 0x2a97c8 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a97ce movq (%rdi), %rax callq 0x18(%rax) movq 0x310(%rsp), %rax testq %rax, %rax je 0x2a9805 lock decl (%rax) jne 0x2a9805 movq 0x308(%rsp), %rsi movq 0x328(%rsp), %rdi testq %rdi, %rdi jne 0x2a97ff movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a9805 movq (%rdi), %rax callq 0x18(%rax) movq 0x1a8(%rsp), %rax testq %rax, %rax je 0x2a983c lock decl (%rax) jne 0x2a983c movq 0x1a0(%rsp), %rsi movq 0x1c0(%rsp), %rdi testq %rdi, %rdi jne 0x2a9836 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a983c movq (%rdi), %rax callq 0x18(%rax) movq 0x218(%rsp), %rax testq %rax, %rax je 0x2a9873 lock decl (%rax) jne 0x2a9873 movq 0x210(%rsp), %rsi movq 0x230(%rsp), %rdi testq %rdi, %rdi jne 0x2a986d movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a9873 movq (%rdi), %rax callq 0x18(%rax) movq 0x98(%rsp), %rax testq %rax, %rax je 0x2a98aa lock decl (%rax) jne 0x2a98aa movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi jne 0x2a98a4 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a98aa movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2a98ba jmp 0x2a98ba jmp 0x2a98ba jmp 0x2a98ba movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
ncnn::ConvolutionDepthWise_x86_avx::forward_int8_x86(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx::forward_int8_x86(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; int elempack = bottom_blob.elempack; int elembits = bottom_blob.elembits(); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_int8 = bottom_blob; if (elembits != 8) { const int channels_g = channels * elempack / group; Mat scales(channels * elempack); { float* ps = scales; for (int g = 0; g < group; g++) { float scale = bottom_blob_int8_scales[g]; for (int q = 0; q < channels_g; q++) { ps++ = scale; } } } Option opt_q = opt; opt_q.blob_allocator = opt.workspace_allocator; quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q); } Mat bottom_blob_bordered; make_padding(bottom_blob_int8, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; channels = bottom_blob_bordered.c; elempack = bottom_blob_bordered.elempack; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; // depth-wise if (channels elempack == group && group == num_output) { int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { out_elempack = num_output % 8 == 0 ? 8 : 1; } #endif // __SSE2__ bool use_int8_requantize = int8_scale_term > 100; size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #if __SSE2__ if (elempack == 8) { { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128i _sum0 = _mm_setzero_si128(); __m128i _sum1 = _mm_setzero_si128(); const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { // TODO use _mm_cvtepi8_epi16 on sse4.1 __m128i _val = _mm_loadl_epi64((const __m128i)(sptr + space_ofs[k] 8)); _val = _mm_unpacklo_epi8(_val, _mm_cmpgt_epi8(_mm_setzero_si128(), _val)); __m128i _w = _mm_loadl_epi64((const __m128i)(kptr + k 8)); _w = _mm_unpacklo_epi8(_w, _mm_cmpgt_epi8(_mm_setzero_si128(), _w)); __m128i _sl = _mm_mullo_epi16(_val, _w); __m128i _sh = _mm_mulhi_epi16(_val, _w); __m128i _s0 = _mm_unpacklo_epi16(_sl, _sh); __m128i _s1 = _mm_unpackhi_epi16(_sl, _sh); _sum0 = _mm_add_epi32(_sum0, _s0); _sum1 = _mm_add_epi32(_sum1, _s1); } __m128 _scale_in0; __m128 _scale_in1; { __m128 _bottom_blob_int8_scales0 = _mm_loadu_ps((const float)bottom_blob_int8_scales + g 8); __m128 _bottom_blob_int8_scales1 = _mm_loadu_ps((const float)bottom_blob_int8_scales + g 8 + 4); __m128 _weight_data_int8_scales0 = _mm_loadu_ps((const float)weight_data_int8_scales + g 8); __m128 _weight_data_int8_scales1 = _mm_loadu_ps((const float)weight_data_int8_scales + g 8 + 4); _scale_in0 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales0, _weight_data_int8_scales0)); _scale_in1 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales1, _weight_data_int8_scales1)); __m128 _m0 = _mm_cmpneq_ps(_weight_data_int8_scales0, _mm_setzero_ps()); __m128 _m1 = _mm_cmpneq_ps(_weight_data_int8_scales1, _mm_setzero_ps()); _scale_in0 = _mm_and_ps(_scale_in0, _m0); _scale_in1 = _mm_and_ps(_scale_in1, _m1); } __m128 _sumfp32_0 = _mm_mul_ps(_mm_cvtepi32_ps(_sum0), _scale_in0); __m128 _sumfp32_1 = _mm_mul_ps(_mm_cvtepi32_ps(_sum1), _scale_in1); if (bias_term) { __m128 _bias0 = _mm_loadu_ps((const float)bias_data + g 8); __m128 _bias1 = _mm_loadu_ps((const float)bias_data + g 8 + 4); _sumfp32_0 = _mm_add_ps(_sumfp32_0, _bias0); _sumfp32_1 = _mm_add_ps(_sumfp32_1, _bias1); } _sumfp32_0 = activation_sse(_sumfp32_0, activation_type, activation_params); _sumfp32_1 = activation_sse(_sumfp32_1, activation_type, activation_params); if (use_int8_requantize) { // requantize and relu __m128 _scale_out0 = _mm_loadu_ps((const float)top_blob_int8_scales + g 8); __m128 _scale_out1 = _mm_loadu_ps((const float)top_blob_int8_scales + g 8 + 4); _sumfp32_0 = _mm_mul_ps(_sumfp32_0, _scale_out0); _sumfp32_1 = _mm_mul_ps(_sumfp32_1, _scale_out1); int64_t _sum8 = float2int8_sse(_sumfp32_0, _sumfp32_1); (int64_t)outptr_s8 = _sum8; outptr_s8 += 8; } else { // dequantize and relu _mm_storeu_ps(outptr_f32, _sumfp32_0); _mm_storeu_ps(outptr_f32 + 4, _sumfp32_1); outptr_f32 += 8; } } } } } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && stride_w == 1 && stride_h == 1 && dilation_w == 1 && dilation_h == 1 && (activation_type == 0 \|\| activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s1_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s1_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2 && (activation_type == 0 \|\| activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s2_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s2_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char)weight_data_tm + maxk g; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr_s8[0] = sums8; outptr_s8 += 1; } else { // dequantize outptr_f32[0] = sumfp32; outptr_f32 += 1; } } } } } } return 0; } bool use_int8_requantize = int8_scale_term > 100; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { if (use_int8_requantize) out_elempack = num_output % 8 == 0 ? 8 : 1; else out_elempack = num_output % 4 == 0 ? 4 : 1; } #endif // __SSE2__ size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { g_elempack = channels_g % 8 == 0 ? 8 : 1; if (use_int8_requantize) out_g_elempack = num_output_g % 8 == 0 ? 8 : 1; else out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x318, %rsp # imm = 0x318 movq %rcx, 0x30(%rsp) movq %rdx, 0x20(%rsp) movq %rsi, %r15 movl 0x38(%rsi), %ecx movl 0x18(%rsi), %r12d movq 0x10(%rsi), %rsi testl %r12d, %r12d je 0x2a9903 leal (,%rsi,8), %eax cltd idivl %r12d cmpl $0x8, %eax sete %dl jmp 0x2a9905 xorl %edx, %edx movq (%rdi), %rax movq -0x18(%rax), %r9 movq %rdi, %r8 movl 0xd4(%rdi,%r9), %ebp movl 0xd8(%rdi,%r9), %ebx decl %ebp imull 0xdc(%rdi,%r9), %ebp decl %ebx imull 0xe0(%rdi,%r9), %ebx movq 0x8(%r15), %rdi vmovups (%r15), %xmm0 vmovaps %xmm0, 0x1e0(%rsp) movq %rsi, 0x1f0(%rsp) movl %r12d, 0x1f8(%rsp) movq 0x20(%r15), %rsi movq %rsi, 0x200(%rsp) vmovdqu 0x28(%r15), %xmm0 vmovdqu %xmm0, 0x208(%rsp) movl %ecx, 0x218(%rsp) movq 0x40(%r15), %rsi movq %rsi, 0x220(%rsp) testq %rdi, %rdi je 0x2a9990 lock incl (%rdi) movq (%r8), %rax movq %r8, 0x8(%rsp) testb %dl, %dl je 0x2a99a3 movq 0x8(%rsp), %rcx jmp 0x2a9ab6 imull %ecx, %r12d movq -0x18(%rax), %rax movq 0x8(%rsp), %r13 movl 0x108(%r13,%rax), %eax movl %eax, 0x10(%rsp) leaq 0x70(%rsp), %rdi andq $0x0, 0x40(%rdi) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdi) vmovdqu %xmm0, 0xc(%rdi) vmovdqa %xmm0, 0x20(%rdi) vmovdqu %xmm0, 0x2c(%rdi) xorl %r14d, %r14d pushq $0x4 popq %rdx movl %r12d, %esi xorl %ecx, %ecx callq 0x635fa movl %r12d, %eax cltd idivl 0x10(%rsp) movq 0x70(%rsp), %rcx movq (%r13), %rdx testl %eax, %eax cmovlel %r14d, %eax movq 0x8(%rsp), %r8 movq -0x18(%rdx), %rsi movslq 0x108(%r8,%rsi), %rdi cmpq %rdi, %r14 jge 0x2a9a3e movq 0x240(%r8,%rsi), %rsi vmovd (%rsi,%r14,4), %xmm0 movl %eax, %esi subl $0x1, %esi jb 0x2a9a39 vmovd %xmm0, (%rcx) addq $0x4, %rcx jmp 0x2a9a2a incq %r14 jmp 0x2a9a09 movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0x130(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x1e0(%rsp), %rsi leaq 0x70(%rsp), %rdx movq %r15, %rdi vzeroupper callq 0x652e3 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a9aae lock decl (%rax) jne 0x2a9aae movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a9aa6 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2a9aae movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx movq (%rcx), %rax leaq 0x70(%rsp), %rdx andq $0x0, 0x40(%rdx) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdx) vmovdqu %xmm0, 0xc(%rdx) vmovdqa %xmm0, 0x20(%rdx) vmovdqu %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %rcx, %rdi leaq 0x1e0(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2ab463 movslq 0xa8(%rsp), %r14 movq 0xb0(%rsp), %rax imulq %r14, %rax testq %rax, %rax je 0x2ab463 notl %ebp movl 0x88(%rsp), %r13d movl 0x9c(%rsp), %edi addl %edi, %ebp movq 0x8(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx movl %ebp, %eax movq %rdx, %rbp cltd idivl 0xe4(%rbp,%rcx) movl %eax, %esi notl %ebx addl 0xa0(%rsp), %ebx movl %ebx, %eax cltd idivl 0xe8(%rbp,%rcx) movl %eax, %edx movq %rsi, 0x10(%rsp) incl %esi movq %rdx, 0x68(%rsp) leal 0x1(%rdx), %r10d movl %r13d, %ebx imull %r14d, %ebx pushq $0x8 popq %r15 cmpl 0x108(%rbp,%rcx), %ebx jne 0x2a9cca cmpl 0xd0(%rbp,%rcx), %ebx jne 0x2a9cca movl %edi, 0x18(%rsp) movq 0x30(%rsp), %rdi cmpb $0x0, 0x27(%rdi) pushq $0x1 popq %rax cmovel %eax, %r15d testb $0x7, %bl cmovnel %eax, %r15d movl 0x10c(%rbp,%rcx), %eax leal (,%r15,4), %r8d movl %eax, 0x38(%rsp) cmpl $0x65, %eax cmovgel %r15d, %r8d movl %ebx, %eax cltd idivl %r15d movq 0x8(%rdi), %rcx movq %rcx, (%rsp) movq 0x20(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %eax, %ecx movl %r15d, %r9d callq 0x628f2 cmpq $0x0, (%rbx) je 0x2ab463 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ab463 xorl %r12d, %r12d cmpl $0x1, %r13d je 0x2aa2a7 cmpl $0x8, %r13d jne 0x2ab463 movq (%rbp), %rax movq -0x18(%rax), %rax movslq 0xd4(%rbp,%rax), %rcx movslq 0xd8(%rbp,%rax), %r15 imulq %rcx, %r15 leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %r15, %rsi callq 0x73bbe movq (%rbp), %rcx movq -0x18(%rcx), %rdx movl 0x18(%rsp), %r11d imull 0xe0(%rbp,%rdx), %r11d movq 0x130(%rsp), %rax movl 0xdc(%rbp,%rdx), %esi imull 0xd4(%rbp,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%rbp,%rdx), %esi jge 0x2aa3d8 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%rbp,%rdx), %r9d jge 0x2a9cc0 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%rbp,%rdx), %edi incq %r9 jmp 0x2a9ca2 addl %r11d, %edi incl %esi addq %r9, %r8 jmp 0x2a9c8b movl 0xd0(%rbp,%rcx), %eax movl 0x10c(%rbp,%rcx), %r14d movq 0x30(%rsp), %rcx cmpb $0x1, 0x27(%rcx) movl %r13d, 0x40(%rsp) movl %r12d, %r13d jne 0x2a9cfe cmpl $0x65, %r14d jl 0x2a9d04 testb $0x7, %al pushq $0x1 popq %r9 cmovel %r15d, %r9d jmp 0x2a9d12 pushq $0x1 popq %r9 jmp 0x2a9d12 xorl %ecx, %ecx testb $0x3, %al sete %cl leal (%rcx,%rcx,2), %r9d incl %r9d leal (,%r9,4), %r8d cmpl $0x65, %r14d cmovgel %r9d, %r8d cltd idivl %r9d movq 0x30(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq 0x20(%rsp), %r12 movq %r12, %rdi movl %esi, 0x18(%rsp) movl %r10d, 0x28(%rsp) movl %r10d, %edx movl %eax, %ecx movq %r8, 0x48(%rsp) movq %r9, 0x38(%rsp) callq 0x628f2 cmpq $0x0, (%r12) je 0x2a9dd1 movslq 0x38(%r12), %rax imulq 0x40(%r12), %rax testq %rax, %rax movl %r13d, %r12d movl 0x40(%rsp), %r8d je 0x2ab463 movq (%rbp), %rax movq -0x18(%rax), %rax movl 0xd0(%rbp,%rax), %ecx movl 0x108(%rbp,%rax), %esi movl %ebx, %eax cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %ebp movq 0x30(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x2a9dd9 testb $0x7, %dil pushq $0x1 popq %rbx movl %ebx, %r12d cmovel %r15d, %r12d cmpl $0x65, %r14d movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d jl 0x2a9deb testb $0x7, %bpl cmovel %r15d, %ebx jmp 0x2a9df9 movl %r13d, %r12d jmp 0x2ab463 pushq $0x1 popq %r12 movl %r12d, %ebx movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d jmp 0x2a9df9 xorl %eax, %eax testb $0x3, %bpl sete %al leal (%rax,%rax,2), %ebx incl %ebx movq 0x78(%rsp), %rax vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0x130(%rsp) movq 0x80(%rsp), %rcx movq %rcx, 0x140(%rsp) movl 0x88(%rsp), %ecx movl %ecx, 0x148(%rsp) movq 0x90(%rsp), %rcx movq %rcx, 0x150(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0x158(%rsp) movl 0xa8(%rsp), %ecx movl %ecx, 0x168(%rsp) movq 0xb0(%rsp), %rcx movq %rcx, 0x170(%rsp) testq %rax, %rax movq 0x20(%rsp), %r14 je 0x2a9e78 lock incl (%rax) cmpl %r12d, %r8d movl %edi, 0x10(%rsp) jle 0x2a9ecd movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0xe0(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x70(%rsp), %rdi leaq 0x130(%rsp), %rsi movl %r12d, %edx vzeroupper callq 0x64e3b movl 0x10(%rsp), %edi movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d movq 0x8(%r14), %rax vmovups (%r14), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movq 0x10(%r14), %rcx movq %rcx, 0xf0(%rsp) movl 0x18(%r14), %ecx movl %ecx, 0xf8(%rsp) movq 0x20(%r14), %rcx movq %rcx, 0x100(%rsp) vmovups 0x28(%r14), %xmm0 vmovups %xmm0, 0x108(%rsp) movl 0x38(%r14), %ecx movl %ecx, 0x118(%rsp) movq 0x40(%r14), %rcx movq %rcx, 0x120(%rsp) testq %rax, %rax je 0x2a9f30 lock incl (%rax) cmpl 0x38(%rsp), %ebx jae 0x2a9fb0 movq 0x8(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xd0(%rcx,%rax), %eax cltd idivl %ebx movl %eax, %ecx movzbl 0x48(%rsp), %eax divb 0x38(%rsp) movzbl %al, %r8d imull %ebx, %r8d movq 0x30(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0xe0(%rsp), %rdi movl %r9d, %edx movl %ebx, %r9d callq 0x628f2 movl 0x10(%rsp), %edi pushq $-0x64 popq %r14 cmpq $0x0, 0xe0(%rsp) je 0x2ab3f2 movslq 0x118(%rsp), %rax imulq 0x120(%rsp), %rax testq %rax, %rax je 0x2ab3f2 xorl %r15d, %r15d xorl %r13d, %r13d xorl %r14d, %r14d movq 0x8(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rax movslq 0x108(%r8,%rax), %rax cmpq %rax, %r14 jge 0x2aa225 movl %r15d, %eax cltd idivl %r12d movl %eax, %ecx movl %edi, %eax cltd idivl %r12d movslq %ecx, %rdx imulq 0x170(%rsp), %rdx movq 0x140(%rsp), %rcx imulq %rcx, %rdx addq 0x130(%rsp), %rdx movl 0x148(%rsp), %esi movq 0x150(%rsp), %rdi movq %rdx, 0x198(%rsp) andq $0x0, 0x1a0(%rsp) movq %rcx, 0x1a8(%rsp) movl %esi, 0x1b0(%rsp) movq %rdi, 0x1b8(%rsp) movl %eax, 0x1d0(%rsp) vmovups 0x158(%rsp), %xmm0 movslq 0x164(%rsp), %rax movslq 0x15c(%rsp), %rdx movslq 0x160(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x1d8(%rsp) vmovups %xmm0, 0x1c0(%rsp) movl %r13d, %eax cltd idivl %ebx movl %eax, %ecx movl %ebp, %eax cltd idivl %ebx movslq %ecx, %rdx imulq 0x120(%rsp), %rdx movq 0xf0(%rsp), %rsi imulq %rsi, %rdx addq 0xe0(%rsp), %rdx movl 0xf8(%rsp), %edi movq 0x100(%rsp), %rcx movq %rdx, 0x228(%rsp) andq $0x0, 0x230(%rsp) movq %rsi, 0x238(%rsp) movl %edi, 0x240(%rsp) movq %rcx, 0x248(%rsp) movl %eax, 0x260(%rsp) vmovups 0x108(%rsp), %xmm0 movslq 0x114(%rsp), %rax movslq 0x10c(%rsp), %rdx movslq 0x110(%rsp), %rdi imulq %rdx, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x268(%rsp) vmovups %xmm0, 0x250(%rsp) movq 0x10(%r8), %rax movq (%rax,%r14,8), %rdi movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 vmovups %ymm0, 0x2d0(%rsp) vmovups %ymm1, 0x2f0(%rsp) movq %rcx, 0x2d8(%rsp) movq (%rdi), %rax leaq 0x198(%rsp), %rsi leaq 0x228(%rsp), %rdx leaq 0x2d0(%rsp), %rcx vzeroupper callq 0x38(%rax) movq 0x230(%rsp), %rax testq %rax, %rax movl 0x10(%rsp), %edi je 0x2aa1d8 lock decl (%rax) jne 0x2aa1d8 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi je 0x2aa1cc movq (%rdi), %rax callq 0x18(%rax) movl 0x10(%rsp), %edi jmp 0x2aa1d8 movq %rsi, %rdi callq 0x5f3e0 movl 0x10(%rsp), %edi movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x2aa217 lock decl (%rax) jne 0x2aa217 movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi je 0x2aa20b movq (%rdi), %rax callq 0x18(%rax) movl 0x10(%rsp), %edi jmp 0x2aa217 movq %rsi, %rdi callq 0x5f3e0 movl 0x10(%rsp), %edi incq %r14 addl %ebp, %r13d addl %edi, %r15d jmp 0x2a9fb9 movq 0x38(%rsp), %rdx cmpl %edx, %ebx jae 0x2aa24d xorl %r14d, %r14d leaq 0xe0(%rsp), %rdi movq 0x20(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x64e3b jmp 0x2ab3f2 leaq 0xe0(%rsp), %rax movq 0x20(%rsp), %rbx xorl %r14d, %r14d cmpq %rbx, %rax je 0x2ab3f2 movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2aa276 lock incl (%rax) movq 0x8(%rbx), %rax testq %rax, %rax je 0x2ab39f lock decl (%rax) jne 0x2ab39f movq (%rbx), %rsi movq 0x20(%rbx), %rdi testq %rdi, %rdi je 0x2ab397 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ab39f movq (%rbp), %rax movq -0x18(%rax), %rdx movl 0xd4(%rbp,%rdx), %ecx movl 0xd8(%rbp,%rdx), %r13d movl %ecx, %esi xorl $0x3, %esi movl %r13d, %edi xorl $0x3, %edi orl %esi, %edi movq 0x20(%rsp), %r14 jne 0x2aaf50 cmpl $0x1, 0xe4(%rbp,%rdx) jne 0x2aae45 cmpl $0x1, 0xe8(%rbp,%rdx) jne 0x2aae45 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x2aae45 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x2aae45 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x2aae45 cmpl $0x65, 0x38(%rsp) jl 0x2ab4e0 vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx vmovaps %xmm0, (%rbx) andq $0x0, 0x10(%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 leaq 0x198(%rsp), %r12 movq -0x18(%rax), %r13 movslq 0x108(%rbp,%r13), %rax cmpq %rax, %r14 jge 0x2ab821 movq 0x1f8(%rbp,%r13), %rax vmovss (%rax,%r14,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2aa396 movq 0x240(%rbp,%r13), %rax vmulss (%rax,%r14,4), %xmm1, %xmm0 vmovss 0x1448f6(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rbp,%r13), %rax vmovss (%rax,%r14,4), %xmm0 vmovss %xmm0, 0x198(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c movq %rbx, %rdi movq %r12, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2aa34f leal (,%r15,8), %ecx movl %ecx, 0x28(%rsp) xorl %edx, %edx testl %r15d, %r15d cmovlel %edx, %r15d testl %r14d, %r14d cmovlel %edx, %r14d vpxor %xmm0, %xmm0, %xmm0 vbroadcastss 0x143c13(%rip), %xmm2 # 0x3ee014 vbroadcastss 0x14487e(%rip), %xmm8 # 0x3eec88 xorl %esi, %esi movq %r14, 0x18(%rsp) cmpq %r14, %rsi je 0x2ab382 movq 0x20(%rsp), %rcx movq 0x10(%rcx), %r8 imulq %rsi, %r8 imulq 0x40(%rcx), %r8 movq %rdx, 0x30(%rsp) movslq %edx, %r9 addq (%rcx), %r8 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rsi, %r11 movq 0x80(%rsp), %rcx imulq %rcx, %r11 addq 0x70(%rsp), %r11 imulq %rcx, %r10 movq %rsi, 0x40(%rsp) leaq (,%rsi,8), %rbx addq 0x28(%rbp), %r9 xorl %ecx, %ecx movq %r8, %rdx cmpl 0x68(%rsp), %ecx jg 0x2aae2a movq (%rbp), %rdi xorl %r13d, %r13d cmpl 0x10(%rsp), %r13d jg 0x2aae23 movq -0x18(%rdi), %rsi movslq 0xe8(%rbp,%rsi), %r14 movslq %ecx, %r12 imulq %r14, %r12 imulq %r10, %r12 movl 0xe4(%rbp,%rsi), %ebp imull %r13d, %ebp shll $0x3, %ebp movslq %ebp, %r14 addq %r11, %r14 addq %r12, %r14 vpxor %xmm3, %xmm3, %xmm3 xorl %r12d, %r12d vpxor %xmm4, %xmm4, %xmm4 cmpq %r12, %r15 je 0x2aa50d movslq (%rax,%r12,4), %rbp vmovq (%r14,%rbp,8), %xmm5 vpcmpgtb %xmm5, %xmm0, %xmm6 vpunpcklbw %xmm6, %xmm5, %xmm5 # xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7] vmovq (%r9,%r12,8), %xmm6 vpcmpgtb %xmm6, %xmm0, %xmm7 vpunpcklbw %xmm7, %xmm6, %xmm6 # xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3],xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7] vpmullw %xmm5, %xmm6, %xmm7 vpmulhw %xmm6, %xmm5, %xmm5 vpunpcklwd %xmm5, %xmm7, %xmm6 # xmm6 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3] vpunpckhwd %xmm5, %xmm7, %xmm5 # xmm5 = xmm7[4],xmm5[4],xmm7[5],xmm5[5],xmm7[6],xmm5[6],xmm7[7],xmm5[7] vpaddd %xmm6, %xmm4, %xmm4 vpaddd %xmm5, %xmm3, %xmm3 incq %r12 jmp 0x2aa4cb movq 0x8(%rsp), %rbp movq 0x1f8(%rbp,%rsi), %r14 movq 0x240(%rbp,%rsi), %r12 vmovups (%r14,%rbx,4), %xmm5 vmovups 0x10(%r14,%rbx,4), %xmm15 vmulps (%r12,%rbx,4), %xmm5, %xmm6 vmulps 0x10(%r12,%rbx,4), %xmm15, %xmm7 vrcpps %xmm6, %xmm6 vrcpps %xmm7, %xmm7 vcmpneqps %xmm0, %xmm5, %xmm5 vandps %xmm6, %xmm5, %xmm5 vcmpneqps %xmm0, %xmm15, %xmm6 vandps %xmm7, %xmm6, %xmm6 vcvtdq2ps %xmm4, %xmm4 vmulps %xmm4, %xmm5, %xmm4 vcvtdq2ps %xmm3, %xmm3 vmulps %xmm3, %xmm6, %xmm5 cmpl $0x0, 0x100(%rbp,%rsi) je 0x2aa585 movq 0x1b0(%rbp,%rsi), %r14 vaddps (%r14,%rbx,4), %xmm4, %xmm4 vaddps 0x10(%r14,%rbx,4), %xmm5, %xmm5 movl 0x110(%rbp,%rsi), %r14d decl %r14d cmpl $0x5, %r14d ja 0x2aad9f leaq 0x14e14f(%rip), %r12 # 0x3f86f0 movslq (%r12,%r14,4), %r14 addq %r12, %r14 vmaxps %xmm0, %xmm4, %xmm15 vmaxps %xmm0, %xmm5, %xmm3 jmpq %r14 movq 0x118(%rbp,%rsi), %r14 vminps %xmm0, %xmm4, %xmm4 vbroadcastss (%r14), %xmm6 vmulps %xmm4, %xmm6, %xmm4 vaddps %xmm4, %xmm15, %xmm15 vminps %xmm0, %xmm5, %xmm4 vmulps %xmm4, %xmm6, %xmm4 jmp 0x2aad5c vbroadcastss 0x146bd6(%rip), %xmm1 # 0x3f11b8 vminps %xmm1, %xmm4, %xmm3 vbroadcastss 0x146bcd(%rip), %xmm1 # 0x3f11bc vmaxps %xmm1, %xmm3, %xmm3 vbroadcastss 0x146bc4(%rip), %xmm1 # 0x3f11c0 vmulps %xmm1, %xmm3, %xmm6 vaddps %xmm2, %xmm6, %xmm6 vcvttps2dq %xmm6, %xmm7 vcvtdq2ps %xmm7, %xmm7 vcmpltps %xmm7, %xmm6, %xmm6 vandps %xmm6, %xmm8, %xmm6 vsubps %xmm6, %xmm7, %xmm6 vbroadcastss 0x1472fe(%rip), %xmm9 # 0x3f1920 vmulps %xmm6, %xmm9, %xmm7 vsubps %xmm7, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm7 vbroadcastss 0x146b95(%rip), %xmm10 # 0x3f11cc vmulps %xmm3, %xmm10, %xmm15 vmovaps %xmm10, %xmm12 vbroadcastss 0x146b87(%rip), %xmm10 # 0x3f11d0 vaddps %xmm10, %xmm15, %xmm15 vmovaps %xmm10, %xmm13 vmulps %xmm3, %xmm15, %xmm15 vbroadcastss 0x146b74(%rip), %xmm11 # 0x3f11d4 vaddps %xmm11, %xmm15, %xmm15 vmovaps %xmm11, %xmm14 vmulps %xmm3, %xmm15, %xmm15 vbroadcastss 0x146b61(%rip), %xmm1 # 0x3f11d8 vaddps %xmm1, %xmm15, %xmm15 vmulps %xmm3, %xmm15, %xmm15 vbroadcastss 0x146b54(%rip), %xmm10 # 0x3f11dc vaddps %xmm10, %xmm15, %xmm15 vmulps %xmm3, %xmm15, %xmm15 vaddps %xmm2, %xmm15, %xmm15 vmulps %xmm7, %xmm15, %xmm7 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm7, %xmm3, %xmm3 vcvttps2dq %xmm6, %xmm6 vpslld $0x17, %xmm6, %xmm6 vpaddd %xmm6, %xmm8, %xmm6 vmulps %xmm6, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm6 vcmpleps %xmm0, %xmm6, %xmm3 vbroadcastss 0x146b1c(%rip), %xmm7 # 0x3f11e0 vmaxps %xmm7, %xmm6, %xmm6 vpsrld $0x17, %xmm6, %xmm7 vbroadcastss 0x146b0e(%rip), %xmm10 # 0x3f11e4 vandps %xmm6, %xmm10, %xmm6 vorps %xmm2, %xmm6, %xmm6 vbroadcastss 0x146b01(%rip), %xmm10 # 0x3f11e8 vpaddd %xmm7, %xmm10, %xmm7 vcvtdq2ps %xmm7, %xmm7 vbroadcastss 0x146af4(%rip), %xmm10 # 0x3f11ec vcmpltps %xmm10, %xmm6, %xmm15 vandps %xmm6, %xmm15, %xmm10 vbroadcastss 0x146ae5(%rip), %xmm1 # 0x3f11f0 vaddps %xmm1, %xmm6, %xmm6 vaddps %xmm6, %xmm10, %xmm6 vandps %xmm8, %xmm15, %xmm10 vsubps %xmm10, %xmm7, %xmm7 vmulps %xmm6, %xmm6, %xmm10 vbroadcastss 0x146aca(%rip), %xmm15 # 0x3f11f4 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146ac1(%rip), %xmm11 # 0x3f11f8 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146ab3(%rip), %xmm11 # 0x3f11fc vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146aa5(%rip), %xmm11 # 0x3f1200 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a97(%rip), %xmm11 # 0x3f1204 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a89(%rip), %xmm11 # 0x3f1208 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a7b(%rip), %xmm11 # 0x3f120c vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a6d(%rip), %xmm11 # 0x3f1210 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a5f(%rip), %xmm11 # 0x3f1214 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vmulps %xmm7, %xmm9, %xmm7 vbroadcastss 0x147159(%rip), %xmm11 # 0x3f1924 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm15, %xmm10, %xmm10 vaddps %xmm6, %xmm7, %xmm6 vaddps %xmm6, %xmm10, %xmm6 vbroadcastss 0x147f32(%rip), %xmm11 # 0x3f2718 vmulps %xmm6, %xmm11, %xmm6 vbroadcastss 0x14710d(%rip), %xmm11 # 0x3f1900 vblendvps %xmm3, %xmm11, %xmm6, %xmm3 vbroadcastss 0x1469b6(%rip), %xmm1 # 0x3f11b8 vminps %xmm1, %xmm3, %xmm3 vbroadcastss 0x1469ad(%rip), %xmm1 # 0x3f11bc vmaxps %xmm1, %xmm3, %xmm3 vbroadcastss 0x1469a4(%rip), %xmm1 # 0x3f11c0 vmulps %xmm1, %xmm3, %xmm6 vaddps %xmm2, %xmm6, %xmm6 vcvttps2dq %xmm6, %xmm7 vcvtdq2ps %xmm7, %xmm7 vcmpltps %xmm7, %xmm6, %xmm6 vandps %xmm6, %xmm8, %xmm6 vsubps %xmm6, %xmm7, %xmm6 vmulps %xmm6, %xmm9, %xmm7 vsubps %xmm7, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm7 vmovaps %xmm12, %xmm1 vmulps %xmm3, %xmm12, %xmm10 vmovaps %xmm13, %xmm12 vaddps %xmm13, %xmm10, %xmm10 vmulps %xmm3, %xmm10, %xmm10 vmovaps %xmm14, %xmm13 vaddps %xmm14, %xmm10, %xmm10 vmulps %xmm3, %xmm10, %xmm10 vbroadcastss 0x146966(%rip), %xmm14 # 0x3f11d8 vaddps %xmm14, %xmm10, %xmm10 vmulps %xmm3, %xmm10, %xmm10 vbroadcastss 0x146958(%rip), %xmm11 # 0x3f11dc vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm3, %xmm10, %xmm10 vaddps %xmm2, %xmm10, %xmm10 vmulps %xmm7, %xmm10, %xmm7 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm7, %xmm3, %xmm3 vcvttps2dq %xmm6, %xmm6 vpslld $0x17, %xmm6, %xmm6 vpaddd %xmm6, %xmm8, %xmm6 vmulps %xmm6, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vrcpps %xmm3, %xmm6 vaddps %xmm6, %xmm6, %xmm7 vmulps %xmm7, %xmm3, %xmm3 vbroadcastss 0x147e41(%rip), %xmm11 # 0x3f2708 vsubps %xmm3, %xmm11, %xmm3 vmulps %xmm3, %xmm6, %xmm3 vbroadcastss 0x146918(%rip), %xmm11 # 0x3f11f0 vaddps %xmm7, %xmm11, %xmm6 vaddps %xmm3, %xmm6, %xmm3 vmulps %xmm4, %xmm3, %xmm15 vbroadcastss 0x1468cb(%rip), %xmm3 # 0x3f11b8 vminps %xmm3, %xmm5, %xmm3 vbroadcastss 0x1468c2(%rip), %xmm4 # 0x3f11bc vmaxps %xmm4, %xmm3, %xmm3 vbroadcastss 0x1468b9(%rip), %xmm4 # 0x3f11c0 vmulps %xmm4, %xmm3, %xmm4 vaddps %xmm2, %xmm4, %xmm4 vcvttps2dq %xmm4, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm4, %xmm4 vandps %xmm4, %xmm8, %xmm4 vsubps %xmm4, %xmm6, %xmm4 vmulps %xmm4, %xmm9, %xmm6 vsubps %xmm6, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm6 vmulps %xmm1, %xmm3, %xmm7 vaddps %xmm7, %xmm12, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm13, %xmm7 vmovaps %xmm13, %xmm12 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm14, %xmm7 vmovaps %xmm14, %xmm13 vmulps %xmm3, %xmm7, %xmm7 vbroadcastss 0x14687d(%rip), %xmm14 # 0x3f11dc vaddps %xmm7, %xmm14, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm2, %xmm7, %xmm7 vmulps %xmm7, %xmm6, %xmm6 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm4, %xmm8, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vbroadcastss 0x14684b(%rip), %xmm1 # 0x3f11e0 vmaxps %xmm1, %xmm3, %xmm4 vpsrld $0x17, %xmm4, %xmm6 vbroadcastss 0x14683d(%rip), %xmm1 # 0x3f11e4 vandps %xmm1, %xmm4, %xmm4 vorps %xmm2, %xmm4, %xmm4 vbroadcastss 0x146830(%rip), %xmm1 # 0x3f11e8 vpaddd %xmm1, %xmm6, %xmm6 vcvtdq2ps %xmm6, %xmm6 vbroadcastss 0x146823(%rip), %xmm1 # 0x3f11ec vcmpltps %xmm1, %xmm4, %xmm7 vandps %xmm4, %xmm7, %xmm10 vaddps %xmm4, %xmm11, %xmm4 vmovaps %xmm11, %xmm1 vaddps %xmm4, %xmm10, %xmm4 vandps %xmm7, %xmm8, %xmm7 vsubps %xmm7, %xmm6, %xmm6 vmulps %xmm4, %xmm4, %xmm7 vbroadcastss 0x146801(%rip), %xmm10 # 0x3f11f4 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467f8(%rip), %xmm11 # 0x3f11f8 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467ea(%rip), %xmm11 # 0x3f11fc vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467dc(%rip), %xmm11 # 0x3f1200 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467ce(%rip), %xmm11 # 0x3f1204 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467c0(%rip), %xmm11 # 0x3f1208 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467b2(%rip), %xmm11 # 0x3f120c vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467a4(%rip), %xmm11 # 0x3f1210 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x146796(%rip), %xmm11 # 0x3f1214 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x146e94(%rip), %xmm11 # 0x3f1924 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm7, %xmm10, %xmm7 vcmpleps %xmm0, %xmm3, %xmm3 vmulps %xmm6, %xmm9, %xmm6 vaddps %xmm4, %xmm6, %xmm4 vaddps %xmm7, %xmm4, %xmm4 vbroadcastss 0x147c65(%rip), %xmm6 # 0x3f2718 vmulps %xmm6, %xmm4, %xmm4 vbroadcastss 0x146e40(%rip), %xmm6 # 0x3f1900 vblendvps %xmm3, %xmm6, %xmm4, %xmm3 vbroadcastss 0x1466e9(%rip), %xmm4 # 0x3f11b8 vminps %xmm4, %xmm3, %xmm3 vbroadcastss 0x1466e0(%rip), %xmm4 # 0x3f11bc vmaxps %xmm4, %xmm3, %xmm3 vbroadcastss 0x1466d7(%rip), %xmm4 # 0x3f11c0 vmulps %xmm4, %xmm3, %xmm4 vaddps %xmm2, %xmm4, %xmm4 vcvttps2dq %xmm4, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm4, %xmm4 vandps %xmm4, %xmm8, %xmm4 vsubps %xmm4, %xmm6, %xmm4 vmulps %xmm4, %xmm9, %xmm6 vsubps %xmm6, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm6 vbroadcastss 0x1466b1(%rip), %xmm7 # 0x3f11cc vmulps %xmm7, %xmm3, %xmm7 vbroadcastss 0x1466a8(%rip), %xmm9 # 0x3f11d0 vaddps %xmm7, %xmm9, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm12, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm13, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm14, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm2, %xmm7, %xmm7 vmulps %xmm7, %xmm6, %xmm6 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm4, %xmm8, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vrcpps %xmm3, %xmm4 vaddps %xmm4, %xmm4, %xmm6 vmulps %xmm6, %xmm3, %xmm3 vbroadcastss 0x147b86(%rip), %xmm7 # 0x3f2708 vsubps %xmm3, %xmm7, %xmm3 vmulps %xmm3, %xmm4, %xmm3 vaddps %xmm1, %xmm6, %xmm4 vaddps %xmm3, %xmm4, %xmm3 jmp 0x2aad99 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm3 vbroadcastss 0x4(%r14), %xmm6 vmaxps %xmm3, %xmm4, %xmm4 vminps %xmm6, %xmm4, %xmm15 vmaxps %xmm3, %xmm5, %xmm3 vminps %xmm6, %xmm3, %xmm3 jmp 0x2aada7 vbroadcastss 0x1465ec(%rip), %xmm12 # 0x3f11b4 vxorps %xmm4, %xmm12, %xmm3 vbroadcastss 0x1465e3(%rip), %xmm13 # 0x3f11b8 vminps %xmm3, %xmm13, %xmm3 vbroadcastss 0x1465da(%rip), %xmm14 # 0x3f11bc vmaxps %xmm3, %xmm14, %xmm3 vbroadcastss 0x1465d1(%rip), %xmm9 # 0x3f11c0 vmulps %xmm3, %xmm9, %xmm4 vaddps %xmm2, %xmm4, %xmm4 vcvttps2dq %xmm4, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm4, %xmm4 vandps %xmm4, %xmm8, %xmm4 vsubps %xmm4, %xmm6, %xmm4 vbroadcastss 0x146d0b(%rip), %xmm1 # 0x3f1920 vmulps %xmm1, %xmm4, %xmm6 vsubps %xmm6, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm6 vbroadcastss 0x1465a2(%rip), %xmm10 # 0x3f11cc vmulps %xmm3, %xmm10, %xmm7 vbroadcastss 0x146599(%rip), %xmm11 # 0x3f11d0 vaddps %xmm7, %xmm11, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vmovaps %xmm2, %xmm1 vbroadcastss 0x146588(%rip), %xmm2 # 0x3f11d4 vaddps %xmm2, %xmm7, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vbroadcastss 0x14657b(%rip), %xmm10 # 0x3f11d8 vaddps %xmm7, %xmm10, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vbroadcastss 0x14656e(%rip), %xmm11 # 0x3f11dc vaddps %xmm7, %xmm11, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm1, %xmm7, %xmm7 vmulps %xmm7, %xmm6, %xmm6 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm4, %xmm8, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vrcpps %xmm3, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vsubps %xmm3, %xmm8, %xmm3 vmulps %xmm3, %xmm4, %xmm3 vaddps %xmm3, %xmm4, %xmm15 vxorps %xmm5, %xmm12, %xmm3 vminps %xmm3, %xmm13, %xmm3 vmaxps %xmm3, %xmm14, %xmm3 vmulps %xmm3, %xmm9, %xmm4 vaddps %xmm1, %xmm4, %xmm4 vcvttps2dq %xmm4, %xmm5 vcvtdq2ps %xmm5, %xmm5 vcmpltps %xmm5, %xmm4, %xmm4 vandps %xmm4, %xmm8, %xmm4 vsubps %xmm4, %xmm5, %xmm4 vbroadcastss 0x146c3f(%rip), %xmm5 # 0x3f1920 vmulps %xmm5, %xmm4, %xmm5 vsubps %xmm5, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm5 vbroadcastss 0x1464d6(%rip), %xmm6 # 0x3f11cc vmulps %xmm6, %xmm3, %xmm6 vbroadcastss 0x1464cd(%rip), %xmm7 # 0x3f11d0 vaddps %xmm7, %xmm6, %xmm6 vmulps %xmm3, %xmm6, %xmm6 vaddps %xmm2, %xmm6, %xmm6 vmovaps %xmm1, %xmm2 vmulps %xmm3, %xmm6, %xmm6 vaddps %xmm6, %xmm10, %xmm6 vmulps %xmm3, %xmm6, %xmm6 vaddps %xmm6, %xmm11, %xmm6 vmulps %xmm3, %xmm6, %xmm6 vaddps %xmm1, %xmm6, %xmm6 vmulps %xmm6, %xmm5, %xmm5 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm5, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm4, %xmm8, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vrcpps %xmm3, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vsubps %xmm3, %xmm8, %xmm3 vmulps %xmm3, %xmm4, %xmm3 vaddps %xmm3, %xmm4, %xmm3 jmp 0x2aada7 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm3 vbroadcastss 0x4(%r14), %xmm6 vmulps %xmm4, %xmm3, %xmm7 vaddps %xmm6, %xmm7, %xmm7 vmaxps %xmm0, %xmm7, %xmm7 vminps %xmm7, %xmm8, %xmm7 vmulps %xmm4, %xmm7, %xmm15 vmulps %xmm5, %xmm3, %xmm3 vaddps %xmm6, %xmm3, %xmm3 vmaxps %xmm0, %xmm3, %xmm3 vminps %xmm3, %xmm8, %xmm3 vmulps %xmm5, %xmm3, %xmm3 jmp 0x2aada7 vmovaps %xmm4, %xmm15 vmovaps %xmm5, %xmm3 cmpl $0x65, 0x38(%rsp) jl 0x2aae0c movq 0x288(%rbp,%rsi), %rsi vmulps (%rsi,%rbx,4), %xmm15, %xmm4 vmulps 0x10(%rsi,%rbx,4), %xmm3, %xmm3 vbroadcastss 0x1463ea(%rip), %xmm1 # 0x3f11b4 vandps %xmm1, %xmm4, %xmm5 vandps %xmm1, %xmm3, %xmm6 vorps %xmm2, %xmm5, %xmm5 vorps %xmm2, %xmm6, %xmm6 vaddps %xmm5, %xmm4, %xmm4 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vcvttps2dq %xmm3, %xmm3 vpackssdw %xmm3, %xmm4, %xmm3 vpminsw 0x14d75a(%rip), %xmm3, %xmm3 # 0x3f8550 vpmaxsw 0x14d762(%rip), %xmm3, %xmm3 # 0x3f8560 vpacksswb %xmm3, %xmm3, %xmm3 vmovq %xmm3, (%rdx) addq $0x8, %rdx jmp 0x2aae1b vmovups %xmm15, (%r8) vmovups %xmm3, 0x10(%r8) addq $0x20, %r8 incl %r13d jmp 0x2aa487 incl %ecx jmp 0x2aa476 movq 0x40(%rsp), %rsi incq %rsi movq 0x30(%rsp), %rdx addl 0x28(%rsp), %edx movq 0x18(%rsp), %r14 jmp 0x2aa411 pushq $0x3 popq %r13 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x2aaf50 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x2aaf50 cmpl $0x2, 0xe4(%rbp,%rdx) jne 0x2aaf50 cmpl $0x2, 0xe8(%rbp,%rdx) jne 0x2aaf50 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x2aaf50 cmpl $0x65, 0x38(%rsp) jl 0x2aba61 vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %r13 vmovaps %xmm0, (%r13) andq $0x0, 0x10(%r13) xorl %ebx, %ebx movq %rbp, %rcx leaq 0xe0(%rsp), %rbp leaq 0x198(%rsp), %r12 movq -0x18(%rax), %r14 movslq 0x108(%rcx,%r14), %rax cmpq %rax, %rbx jge 0x2abdbb movq 0x1f8(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2aaf0b movq 0x240(%rcx,%r14), %rax vmulss (%rax,%rbx,4), %xmm1, %xmm0 vmovss 0x143d81(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm0 vmovss %xmm0, 0x198(%rsp) movq %r13, %rdi movq %rbp, %rsi callq 0x1ea12c movq %r13, %rdi movq %r12, %rsi callq 0x1ea12c incq %rbx movq 0x8(%rsp), %rcx movq (%rcx), %rax jmp 0x2aaec6 imull %ecx, %r13d movslq %r13d, %rsi leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %rsi, 0xc8(%rsp) callq 0x73bbe movq (%rbp), %r9 movq -0x18(%r9), %rax movl 0x18(%rsp), %r10d imull 0xe0(%rbp,%rax), %r10d movq 0x130(%rsp), %rbx movl 0xdc(%rbp,%rax), %ecx imull 0xd4(%rbp,%rax), %ecx subl %ecx, %r10d xorl %ecx, %ecx xorl %edx, %edx xorl %esi, %esi cmpl 0xd8(%rbp,%rax), %esi jge 0x2aafe3 movslq %ecx, %rcx leaq (%rbx,%rcx,4), %r8 xorl %edi, %edi cmpl 0xd4(%rbp,%rax), %edi jge 0x2aafd9 movl %edx, (%r8,%rdi,4) movq -0x18(%r9), %rax addl 0xdc(%rbp,%rax), %edx incq %rdi jmp 0x2aafbc addl %r10d, %edx incl %esi addq %rdi, %rcx jmp 0x2aafaa xorl %edx, %edx testl %r13d, %r13d movl $0x0, %ecx movq %rcx, 0x50(%rsp) cmovlel %edx, %r13d vbroadcastss 0x1461b5(%rip), %xmm4 # 0x3f11b4 leaq 0x14d6d2(%rip), %r8 # 0x3f86d8 xorl %ebp, %ebp movq %r9, 0xc0(%rsp) movq 0x8(%rsp), %rcx movslq 0x108(%rcx,%rax), %rax cmpq %rax, %rbp jge 0x2ab382 movq 0x10(%r14), %r9 imulq %rbp, %r9 imulq 0x40(%r14), %r9 addq (%r14), %r9 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rbp, %r11 movq 0x80(%rsp), %rax imulq %rax, %r11 addq 0x70(%rsp), %r11 imulq %rax, %r10 movq 0x28(%rcx), %r15 addq 0x50(%rsp), %r15 xorl %eax, %eax movq %r9, 0x18(%rsp) movq %r10, 0x28(%rsp) movq %r11, 0x48(%rsp) movq %rbp, 0xd0(%rsp) cmpl 0x68(%rsp), %eax jg 0x2ab357 movl %eax, 0x30(%rsp) movq 0x8(%rsp), %rax movq (%rax), %rax movq %rax, 0x40(%rsp) xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d jg 0x2ab34c movq 0x40(%rsp), %rax movq -0x18(%rax), %r12 movq 0x8(%rsp), %rcx movslq 0xe8(%rcx,%r12), %rax movslq 0x30(%rsp), %rdx imulq %rax, %rdx imulq %r10, %rdx movslq 0xe4(%rcx,%r12), %rax movslq %r14d, %rcx imulq %rax, %rcx addq %r11, %rcx addq %rdx, %rcx xorl %edx, %edx xorl %eax, %eax cmpq %rdx, %r13 je 0x2ab102 movslq (%rbx,%rdx,4), %rsi movsbl (%rcx,%rsi), %esi movsbl (%r15,%rdx), %edi imull %esi, %edi addl %edi, %eax incq %rdx jmp 0x2ab0e6 movq 0x8(%rsp), %rdx movq 0x1f8(%rdx,%r12), %rcx vmovss (%rcx,%rbp,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2ab137 movq 0x240(%rdx,%r12), %rcx vmulss (%rcx,%rbp,4), %xmm1, %xmm0 vmovss 0x143b55(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vcvtsi2ss %eax, %xmm6, %xmm1 vmulss %xmm1, %xmm0, %xmm5 cmpl $0x0, 0x100(%rdx,%r12) je 0x2ab157 movq 0x1b0(%rdx,%r12), %rax vaddss (%rax,%rbp,4), %xmm5, %xmm5 movl 0x110(%rdx,%r12), %eax decl %eax cmpl $0x5, %eax ja 0x2ab2cf movslq (%r8,%rax,4), %rax addq %r8, %rax jmpq %rax vmaxss 0x142e95(%rip), %xmm5, %xmm0 # 0x3ee010 jmp 0x2ab2d3 vmovaps %xmm5, %xmm0 movq %r9, 0x58(%rsp) vmovss %xmm5, 0xd8(%rsp) callq 0x5f410 vaddss 0x143ae9(%rip), %xmm0, %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x48(%rsp), %r11 movq 0x28(%rsp), %r10 movq 0x58(%rsp), %r9 leaq 0x14d519(%rip), %r8 # 0x3f86d8 vbroadcastss 0x145fec(%rip), %xmm4 # 0x3f11b4 vmulss 0xd8(%rsp), %xmm0, %xmm0 jmp 0x2ab2d3 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss 0x4(%rax), %xmm1 vmaxss (%rax), %xmm5, %xmm0 vucomiss %xmm1, %xmm0 jbe 0x2ab2d3 vmovaps %xmm1, %xmm0 jmp 0x2ab2d3 vminss 0x145fb1(%rip), %xmm5, %xmm0 # 0x3f11b8 vxorps %xmm4, %xmm0, %xmm1 vcmpltss 0x145fa8(%rip), %xmm0, %xmm0 # 0x3f11bc vbroadcastss 0x145f9b(%rip), %xmm2 # 0x3f11b8 vblendvps %xmm0, %xmm2, %xmm1, %xmm0 movq %r9, %rbp callq 0x5f410 movq 0x48(%rsp), %r11 movq 0x28(%rsp), %r10 movq %rbp, %r9 movq 0xd0(%rsp), %rbp leaq 0x14d491(%rip), %r8 # 0x3f86d8 vbroadcastss 0x145f64(%rip), %xmm4 # 0x3f11b4 vmovss 0x143a30(%rip), %xmm1 # 0x3eec88 vaddss %xmm1, %xmm0, %xmm0 vdivss %xmm0, %xmm1, %xmm0 jmp 0x2ab2d3 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vxorps %xmm0, %xmm0, %xmm0 vcmpltss %xmm5, %xmm0, %xmm0 vmovss (%rax), %xmm1 vbroadcastss 0x143a03(%rip), %xmm2 # 0x3eec88 vblendvps %xmm0, %xmm2, %xmm1, %xmm0 vmulss %xmm5, %xmm0, %xmm0 jmp 0x2ab2d3 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss (%rax), %xmm1 vmovss 0x4(%rax), %xmm2 vxorps %xmm4, %xmm2, %xmm0 vdivss %xmm1, %xmm0, %xmm3 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm3, %xmm5 jb 0x2ab2d3 vmovss 0x1439c7(%rip), %xmm0 # 0x3eec88 vdivss %xmm1, %xmm0, %xmm0 vaddss %xmm0, %xmm3, %xmm0 vucomiss %xmm0, %xmm5 jbe 0x2ab33f vmovaps %xmm5, %xmm0 cmpl $0x65, 0x38(%rsp) jl 0x2ab325 movq 0x8(%rsp), %rax movq 0x288(%rax,%r12), %rax vmulss (%rax,%rbp,4), %xmm0, %xmm0 vandps %xmm4, %xmm0, %xmm1 vbroadcastss 0x14d30b(%rip), %xmm2 # 0x3f8604 vorps %xmm2, %xmm1, %xmm1 vaddss %xmm1, %xmm0, %xmm0 vroundss $0xb, %xmm0, %xmm0, %xmm0 vcvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movb %al, (%r9) incq %r9 jmp 0x2ab337 movq 0x18(%rsp), %rax vmovss %xmm0, (%rax) addq $0x4, %rax movq %rax, 0x18(%rsp) incl %r14d jmp 0x2ab09f vmulss %xmm5, %xmm1, %xmm0 vaddss %xmm2, %xmm0, %xmm0 jmp 0x2ab28b movl 0x30(%rsp), %eax incl %eax jmp 0x2ab081 incq %rbp movq 0xc0(%rsp), %r9 movq -0x18(%r9), %rax movq 0x50(%rsp), %rcx addq 0xc8(%rsp), %rcx movq %rcx, 0x50(%rsp) movq 0x20(%rsp), %r14 jmp 0x2ab010 leaq 0x130(%rsp), %rdi callq 0x624be xorl %r12d, %r12d jmp 0x2ab463 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0xe0(%rsp), %xmm0 vmovups %xmm0, (%rbx) movq 0xf0(%rsp), %rax movq %rax, 0x10(%rbx) movl 0xf8(%rsp), %eax movl %eax, 0x18(%rbx) movq 0x100(%rsp), %rax movq %rax, 0x20(%rbx) vmovups 0x108(%rsp), %xmm0 vmovups %xmm0, 0x28(%rbx) movl 0x118(%rsp), %eax movl %eax, 0x38(%rbx) movq 0x120(%rsp), %rax movq %rax, 0x40(%rbx) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2ab429 lock decl (%rax) jne 0x2ab429 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x2ab421 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ab429 movq %rsi, %rdi callq 0x5f3e0 movq 0x138(%rsp), %rax testq %rax, %rax movl %r14d, %r12d je 0x2ab463 lock decl (%rax) jne 0x2ab463 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi je 0x2ab45b movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ab463 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ab494 lock decl (%rax) jne 0x2ab494 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2ab48c movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ab494 movq %rsi, %rdi callq 0x5f3e0 movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x2ab4cb lock decl (%rax) jne 0x2ab4cb movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi je 0x2ab4c3 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ab4cb movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x318, %rsp # imm = 0x318 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovaps %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2ab556 movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x143753(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2ab500 leaq 0x2a0(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdx movq 0x20(%rsp), %rcx movslq 0x2c(%rcx), %rsi movslq 0x30(%rcx), %rax movl 0x38(%rcx), %edi movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movl 0x34(%rcx), %r8d movl %r8d, 0x58(%rsp) movq (%rcx), %r8 movq 0x10(%rcx), %r11 movq 0x40(%rcx), %r15 imulq %r11, %r15 movq %rax, %rbx movq %rsi, 0x38(%rsp) imulq %rsi, %rbx movq %r11, 0xd0(%rsp) movq %rbx, 0x50(%rsp) imulq %rbx, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0xc8(%rsp) movq 0x80(%rsp), %rsi imulq 0xb0(%rsp), %rsi movq %rsi, 0xc0(%rsp) xorl %esi, %esi testl %eax, %eax movl $0x0, %r11d cmovgl %eax, %r11d movl %r11d, 0x10(%rsp) movl 0x28(%rcx), %eax movl %eax, 0x64(%rsp) testl %edi, %edi cmovlel %esi, %edi movq %rdi, 0x18(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x190(%rsp) movq %rdx, 0x28(%rsp) leaq (%rdx,%rdx), %rax movq %rax, 0x188(%rsp) movq 0x2a0(%rsp), %rax movq %rax, 0x180(%rsp) movq %r8, 0xd8(%rsp) xorl %esi, %esi movq %r10, 0x48(%rsp) movq %r15, 0x68(%rsp) cmpq 0x18(%rsp), %rsi je 0x2ab814 movq 0xc8(%rsp), %rax xorl %edx, %edx divq 0xd0(%rsp) cmpl $0x4, 0x64(%rsp) cmoveq 0x50(%rsp), %rax testq %r10, %r10 je 0x2ab688 vmovss (%r10,%rsi,4), %xmm1 jmp 0x2ab68c vxorps %xmm1, %xmm1, %xmm1 imulq %rsi, %r15 addq 0xd8(%rsp), %r15 movq 0x180(%rsp), %rcx vmovss (%rcx,%rsi,4), %xmm0 imull 0x58(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2ab6c5 vmovss %xmm1, (%r8,%rcx,4) incq %rcx jmp 0x2ab6b6 leaq (%rsi,%rsi,8), %rax movq 0xc0(%rsp), %r13 movq %rsi, 0x40(%rsp) imulq %rsi, %r13 addq 0x190(%rsp), %r13 movq 0x28(%rsp), %rcx leaq (%rcx,%r13), %r11 movq 0x188(%rsp), %rcx leaq (%rcx,%r13), %rbx xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d je 0x2ab7fa xorl %ecx, %ecx movq %r15, %rdx movq 0x38(%rsp), %rsi movl %esi, %r10d testl %r10d, %r10d jle 0x2ab7da movsbl (%r13,%rcx), %r12d movsbl (%r9,%rax), %esi imull %r12d, %esi movsbl 0x1(%r13,%rcx), %r12d movsbl 0x1(%r9,%rax), %edi imull %r12d, %edi addl %esi, %edi movsbl 0x2(%r13,%rcx), %esi movsbl 0x2(%r9,%rax), %r12d imull %esi, %r12d movsbl (%r11,%rcx), %esi movsbl 0x3(%r9,%rax), %ebp imull %esi, %ebp addl %r12d, %ebp addl %edi, %ebp movsbl 0x1(%r11,%rcx), %esi movsbl 0x4(%r9,%rax), %edi imull %esi, %edi movsbl 0x2(%r11,%rcx), %esi movsbl 0x5(%r9,%rax), %r12d imull %esi, %r12d addl %edi, %r12d movsbl (%rbx,%rcx), %esi movsbl 0x6(%r9,%rax), %edi imull %esi, %edi addl %r12d, %edi addl %ebp, %edi movsbl 0x1(%rbx,%rcx), %esi movsbl 0x7(%r9,%rax), %ebp imull %esi, %ebp movsbl 0x2(%rbx,%rcx), %esi movsbl 0x8(%r9,%rax), %r12d imull %esi, %r12d addl %ebp, %r12d addl %edi, %r12d vcvtsi2ss %r12d, %xmm2, %xmm1 vmulss %xmm1, %xmm0, %xmm1 vaddss (%r15,%rcx,4), %xmm1, %xmm1 vmovss %xmm1, (%r15,%rcx,4) addq $0x4, %rdx decl %r10d incq %rcx jmp 0x2ab712 addq %rcx, %r13 addq $0x2, %r13 addq %rcx, %r11 addq $0x2, %r11 addq %rcx, %rbx addq $0x2, %rbx incl %r14d movq %rdx, %r15 jmp 0x2ab6fa movq 0x40(%rsp), %rsi incq %rsi movq 0x68(%rsp), %r15 addq %r15, %r8 movq 0x48(%rsp), %r10 jmp 0x2ab653 leaq 0x2a0(%rsp), %rdi jmp 0x2abd9f leaq 0x2b8(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rax movq 0x20(%rsp), %r15 movl 0x2c(%r15), %ecx movl %ecx, 0x38(%rsp) movl 0x30(%r15), %ecx movl 0x38(%r15), %edx movq 0x28(%rbp), %rdi movq 0x1b0(%rbp,%r13), %rsi xorl %r9d, %r9d testl %ecx, %ecx cmovlel %r9d, %ecx movl %ecx, 0x10(%rsp) movq %rax, 0x28(%rsp) addq %rax, %rax movq %rax, 0x58(%rsp) testl %edx, %edx cmovlel %r9d, %edx movq %rdx, 0x18(%rsp) vbroadcastss 0x145927(%rip), %xmm0 # 0x3f11b4 vbroadcastss 0x14cd6e(%rip), %xmm1 # 0x3f8604 pushq $-0x7f popq %r8 pushq $0x7f popq %rbx movq %rsi, 0x48(%rsp) cmpq 0x18(%rsp), %r9 je 0x2aba54 testq %rsi, %rsi je 0x2ab8ba vmovss (%rsi,%r9,4), %xmm2 jmp 0x2ab8be vxorps %xmm2, %xmm2, %xmm2 movq 0x40(%r15), %r14 imulq %r9, %r14 imulq 0x10(%r15), %r14 addq (%r15), %r14 movq 0x2b8(%rsp), %rax vmovss (%rax,%r9,8), %xmm3 vmovss 0x4(%rax,%r9,8), %xmm4 movq 0xb0(%rsp), %r15 imulq %r9, %r15 imulq 0x80(%rsp), %r15 addq 0x70(%rsp), %r15 movq %r9, 0x40(%rsp) leaq (%r9,%r9,8), %r12 movq 0x28(%rsp), %rax leaq (%r15,%rax), %r13 movq 0x58(%rsp), %rax leaq (%r15,%rax), %rbp xorl %eax, %eax cmpl 0x10(%rsp), %eax je 0x2aba3d xorl %r10d, %r10d movl 0x38(%rsp), %ecx testl %ecx, %ecx jle 0x2aba1e movsbl (%r15,%r10), %r11d movsbl (%rdi,%r12), %r9d imull %r11d, %r9d movsbl 0x1(%r15,%r10), %r11d movsbl 0x1(%rdi,%r12), %esi imull %r11d, %esi addl %r9d, %esi movsbl 0x2(%r15,%r10), %r9d movsbl 0x2(%rdi,%r12), %r11d imull %r9d, %r11d movsbl (%r13,%r10), %r9d movsbl 0x3(%rdi,%r12), %edx imull %r9d, %edx addl %r11d, %edx addl %esi, %edx movsbl 0x1(%r13,%r10), %esi movsbl 0x4(%rdi,%r12), %r9d imull %esi, %r9d movsbl 0x2(%r13,%r10), %esi movsbl 0x5(%rdi,%r12), %r11d imull %esi, %r11d addl %r9d, %r11d movsbl (%rbp,%r10), %esi movsbl 0x6(%rdi,%r12), %r9d imull %esi, %r9d addl %r11d, %r9d addl %edx, %r9d movsbl 0x1(%rbp,%r10), %edx movsbl 0x7(%rdi,%r12), %esi imull %edx, %esi movsbl 0x2(%rbp,%r10), %edx movsbl 0x8(%rdi,%r12), %r11d imull %edx, %r11d addl %esi, %r11d addl %r9d, %r11d vcvtsi2ss %r11d, %xmm7, %xmm5 vmulss %xmm5, %xmm3, %xmm5 vaddss %xmm2, %xmm5, %xmm5 vmulss %xmm4, %xmm5, %xmm5 vandps %xmm0, %xmm5, %xmm6 vorps %xmm1, %xmm6, %xmm6 vaddss %xmm6, %xmm5, %xmm5 vroundss $0xb, %xmm5, %xmm5, %xmm5 vcvttss2si %xmm5, %r11d cmpl $-0x7e, %r11d jge 0x2aba07 movl %r8d, %r11d cmpl $0x7f, %r11d jl 0x2aba10 movl %ebx, %r11d movb %r11b, (%r14,%r10) decl %ecx incq %r10 jmp 0x2ab92b addq %r10, %r15 addq $0x2, %r15 addq %r10, %r13 addq $0x2, %r13 addq %r10, %rbp addq $0x2, %rbp incl %eax addq %r10, %r14 jmp 0x2ab91a movq 0x40(%rsp), %r9 incq %r9 movq 0x20(%rsp), %r15 movq 0x48(%rsp), %rsi jmp 0x2ab8a2 leaq 0x2b8(%rsp), %rdi jmp 0x2abfed vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovaps %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2abad7 movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x1431d2(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2aba81 leaq 0x270(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdi movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movslq 0x30(%rdx), %rax movl 0x38(%rdx), %r8d movl %edi, %ecx subl %esi, %ecx addl %ecx, %ecx movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movq 0x10(%rdx), %r11 movq 0x40(%rdx), %rbx imulq %r11, %rbx movq %rax, %r14 movq %rsi, 0x38(%rsp) imulq %rsi, %r14 movq %r11, 0x58(%rsp) movq %r14, 0xd0(%rsp) imulq %r14, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0x68(%rsp) movq 0x80(%rsp), %rsi imulq 0xb0(%rsp), %rsi movq %rsi, 0x50(%rsp) xorl %r11d, %r11d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi movl %esi, 0x10(%rsp) movl 0x34(%rdx), %eax movl %eax, 0xc8(%rsp) movq (%rdx), %r12 movslq %ecx, %rsi testl %r8d, %r8d cmovlel %r11d, %r8d movq %r8, 0x18(%rsp) movl 0x28(%rdx), %eax movl %eax, 0x64(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x190(%rsp) movq %rdi, 0x28(%rsp) leaq (%rdi,%rdi), %rax movq %rax, 0x188(%rsp) movq 0x270(%rsp), %rax movq %rax, 0x180(%rsp) movq %r12, 0xc0(%rsp) xorl %edi, %edi movq %r10, 0x48(%rsp) movq %rbx, 0xd8(%rsp) cmpq 0x18(%rsp), %rdi je 0x2abd97 movq 0x68(%rsp), %rax xorl %edx, %edx divq 0x58(%rsp) cmpl $0x4, 0x64(%rsp) cmoveq 0xd0(%rsp), %rax testq %r10, %r10 je 0x2abc0e vmovss (%r10,%rdi,4), %xmm1 jmp 0x2abc12 vxorps %xmm1, %xmm1, %xmm1 movq %rbx, %rdx imulq %rdi, %rdx addq 0xc0(%rsp), %rdx movq 0x180(%rsp), %rcx vmovss (%rcx,%rdi,4), %xmm0 imull 0xc8(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2abc51 vmovss %xmm1, (%r12,%rcx,4) incq %rcx jmp 0x2abc42 leaq (%rdi,%rdi,8), %rax movq 0x50(%rsp), %r8 movq %rdi, 0x40(%rsp) imulq %rdi, %r8 addq 0x190(%rsp), %r8 movq 0x28(%rsp), %rcx leaq (%r8,%rcx), %r13 movq 0x188(%rsp), %rcx leaq (%r8,%rcx), %r11 xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d je 0x2abd7a xorl %ecx, %ecx movq 0x38(%rsp), %rdi movl %edi, %r10d testl %r10d, %r10d jle 0x2abd60 movsbl (%r8,%rcx), %ebx movsbl (%r9,%rax), %r15d imull %ebx, %r15d movsbl 0x1(%r8,%rcx), %ebx movsbl 0x1(%r9,%rax), %edi imull %ebx, %edi addl %r15d, %edi movsbl 0x2(%r8,%rcx), %ebx movsbl 0x2(%r9,%rax), %r15d imull %ebx, %r15d movsbl (%r13,%rcx), %ebx movsbl 0x3(%r9,%rax), %ebp imull %ebx, %ebp addl %r15d, %ebp addl %edi, %ebp movsbl 0x1(%r13,%rcx), %edi movsbl 0x4(%r9,%rax), %ebx imull %edi, %ebx movsbl 0x2(%r13,%rcx), %edi movsbl 0x5(%r9,%rax), %r15d imull %edi, %r15d addl %ebx, %r15d movsbl (%r11,%rcx), %edi movsbl 0x6(%r9,%rax), %ebx imull %edi, %ebx addl %r15d, %ebx addl %ebp, %ebx movsbl 0x1(%r11,%rcx), %edi movsbl 0x7(%r9,%rax), %ebp imull %edi, %ebp movsbl 0x2(%r11,%rcx), %edi movsbl 0x8(%r9,%rax), %r15d imull %edi, %r15d addl %ebp, %r15d addl %ebx, %r15d vcvtsi2ss %r15d, %xmm2, %xmm1 vmulss %xmm1, %xmm0, %xmm1 vaddss (%rdx), %xmm1, %xmm1 vmovss %xmm1, (%rdx) addq $0x4, %rdx decl %r10d addq $0x2, %rcx jmp 0x2abc98 addq %rsi, %r8 addq %rcx, %r8 addq %rsi, %r13 addq %rcx, %r13 addq %rsi, %r11 addq %rcx, %r11 incl %r14d jmp 0x2abc83 movq 0x40(%rsp), %rdi incq %rdi movq 0xd8(%rsp), %rbx addq %rbx, %r12 movq 0x48(%rsp), %r10 jmp 0x2abbdc leaq 0x270(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x20(%rsp), %r15 jmp 0x2abfff leaq 0x288(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rsi movq 0x20(%rsp), %r15 movl 0x2c(%r15), %eax movl 0x30(%r15), %edx movl 0x38(%r15), %r8d movl %esi, %ecx movl %eax, 0x10(%rsp) subl %eax, %ecx addl %ecx, %ecx movq 0x8(%rsp), %rax movq 0x28(%rax), %rdi movq 0x1b0(%rax,%r14), %rax movq %rsi, 0x18(%rsp) addq %rsi, %rsi movq %rsi, 0x48(%rsp) movslq %ecx, %r10 xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %r8d, %r8d cmovlel %esi, %r8d movq %r8, 0x40(%rsp) vbroadcastss 0x145384(%rip), %xmm0 # 0x3f11b4 vbroadcastss 0x14c7cb(%rip), %xmm1 # 0x3f8604 movq %rax, 0x28(%rsp) cmpq 0x40(%rsp), %rsi je 0x2abfe5 testq %rax, %rax je 0x2abe55 vmovss (%rax,%rsi,4), %xmm2 jmp 0x2abe59 vxorps %xmm2, %xmm2, %xmm2 movq 0x40(%r15), %rbx imulq %rsi, %rbx imulq 0x10(%r15), %rbx addq (%r15), %rbx movq 0x288(%rsp), %rcx vmovss (%rcx,%rsi,8), %xmm3 vmovss 0x4(%rcx,%rsi,8), %xmm4 movq 0xb0(%rsp), %r14 imulq %rsi, %r14 imulq 0x80(%rsp), %r14 addq 0x70(%rsp), %r14 movq %rsi, 0x38(%rsp) leaq (%rsi,%rsi,8), %r15 movq 0x18(%rsp), %rax leaq (%r14,%rax), %r12 movq 0x48(%rsp), %rax leaq (%r14,%rax), %r13 xorl %ebp, %ebp cmpl %edx, %ebp je 0x2abfce xorl %r9d, %r9d movl 0x10(%rsp), %ecx testl %ecx, %ecx jle 0x2abfb5 movsbl (%r14,%r9), %esi movsbl (%rdi,%r15), %r8d imull %esi, %r8d movsbl 0x1(%r14,%r9), %esi movsbl 0x1(%rdi,%r15), %r11d imull %esi, %r11d addl %r8d, %r11d movsbl 0x2(%r14,%r9), %esi movsbl 0x2(%rdi,%r15), %r8d imull %esi, %r8d movsbl (%r12,%r9), %esi movsbl 0x3(%rdi,%r15), %eax imull %esi, %eax addl %r8d, %eax addl %r11d, %eax movsbl 0x1(%r12,%r9), %esi movsbl 0x4(%rdi,%r15), %r8d imull %esi, %r8d movsbl 0x2(%r12,%r9), %esi movsbl 0x5(%rdi,%r15), %r11d imull %esi, %r11d addl %r8d, %r11d movsbl (%r13,%r9), %esi movsbl 0x6(%rdi,%r15), %r8d imull %esi, %r8d addl %r11d, %r8d addl %eax, %r8d movsbl 0x1(%r13,%r9), %eax movsbl 0x7(%rdi,%r15), %esi imull %eax, %esi movsbl 0x2(%r13,%r9), %eax movsbl 0x8(%rdi,%r15), %r11d imull %eax, %r11d addl %esi, %r11d addl %r8d, %r11d vcvtsi2ss %r11d, %xmm7, %xmm5 vmulss %xmm5, %xmm3, %xmm5 vaddss %xmm2, %xmm5, %xmm5 vmulss %xmm4, %xmm5, %xmm5 vandps %xmm0, %xmm5, %xmm6 vorps %xmm1, %xmm6, %xmm6 vaddss %xmm6, %xmm5, %xmm5 vroundss $0xb, %xmm5, %xmm5, %xmm5 vcvttss2si %xmm5, %esi cmpl $-0x7e, %esi jge 0x2abf9c pushq $-0x7f popq %rsi cmpl $0x7f, %esi jl 0x2abfa4 pushq $0x7f popq %rsi movb %sil, (%rbx) incq %rbx decl %ecx addq $0x2, %r9 jmp 0x2abec2 addq %r10, %r14 addq %r9, %r14 addq %r10, %r12 addq %r9, %r12 addq %r10, %r13 addq %r9, %r13 incl %ebp jmp 0x2abeb3 movq 0x38(%rsp), %rsi incq %rsi movq 0x20(%rsp), %r15 movq 0x28(%rsp), %rax jmp 0x2abe3e leaq 0x288(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x8(%rsp), %rax xorl %r12d, %r12d movq 0x8(%rax), %rdi testq %rdi, %rdi je 0x2ab463 movq (%rdi), %rax movq %r15, %rsi movq 0x30(%rsp), %rdx callq 0x48(%rax) jmp 0x2ab463 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 movq %rax, %rbx leaq 0x130(%rsp), %rdi callq 0x621c2 jmp 0x2ac1b7 jmp 0x2ac0d0 jmp 0x2ac0d0 jmp 0x2ac231 jmp 0x2ac231 movq %rax, %rbx jmp 0x2ac180 jmp 0x2ac231 jmp 0x2ac0d0 movq %rax, %rbx jmp 0x2ac149 jmp 0x2ac231 jmp 0x2ac231 jmp 0x2ac0d0 jmp 0x2ac231 jmp 0x2ac231 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac1e8 lock decl (%rax) jne 0x2ac1e8 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2ac1d8 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ac1e8 jmp 0x2ac231 movq %rax, %rbx jmp 0x2ac1e8 movq %rax, %rbx jmp 0x2ac1b7 movq %rax, %rbx movq 0x230(%rsp), %rax testq %rax, %rax je 0x2ac112 lock decl (%rax) jne 0x2ac112 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi jne 0x2ac10c movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac112 movq (%rdi), %rax callq 0x18(%rax) movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x2ac149 lock decl (%rax) jne 0x2ac149 movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi jne 0x2ac143 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac149 movq (%rdi), %rax callq 0x18(%rax) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2ac180 lock decl (%rax) jne 0x2ac180 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi jne 0x2ac17a movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac180 movq (%rdi), %rax callq 0x18(%rax) movq 0x138(%rsp), %rax testq %rax, %rax je 0x2ac1b7 lock decl (%rax) jne 0x2ac1b7 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi jne 0x2ac1b1 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac1b7 movq (%rdi), %rax callq 0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac1e8 lock decl (%rax) jne 0x2ac1e8 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2ac1e2 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac1e8 movq (%rdi), %rax callq 0x18(%rax) movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x2ac21f lock decl (%rax) jne 0x2ac21f movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi jne 0x2ac219 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac21f movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2ac231 jmp 0x2ac231 jmp 0x2ac231 jmp 0x2ac231 jmp 0x2ac231 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_avx::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float)bias_data) + g 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float)bias_data) + g 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float)weight_data_tm + maxk g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float)bias_data) + g 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }	movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2a49d2	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
ncnn::ConvolutionDepthWise_x86_avx::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w = weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w = bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x168, %rsp # imm = 0x168 movq %rsi, %rbp movq %rdi, %r13 movq (%rsi), %r14 leaq 0x48(%r14), %rdi movq (%rdx), %rax movq %rax, 0xb8(%rsp) movl 0x60(%r14), %ebx movl 0x74(%r14), %eax movl %eax, 0x1c(%rsp) movl 0x78(%r14), %eax movl %eax, 0x18(%rsp) imull 0x80(%r14), %ebx leaq 0x70(%rsp), %rsi andq $0x0, 0x40(%rsi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rsi) vmovups %xmm0, 0xc(%rsi) vmovaps %xmm0, 0x20(%rsi) vmovups %xmm0, 0x2c(%rsi) movq %rcx, %r15 movq %rcx, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2ac867 movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x2ac867 movslq 0x88(%rsp), %rcx movl 0x9c(%rsp), %eax imull %ecx, %eax movl %eax, 0x9c(%rsp) movq 0x80(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x80(%rsp) movl $0x1, 0x88(%rsp) andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) vmovaps %xmm0, 0x40(%rsp) vmovups %xmm0, 0x4c(%rsp) movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r13,%rax) je 0x2ac3b0 movl $0x90, %edi addq (%rbp), %rdi leaq 0x20(%rsp), %rsi movq %r15, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x20(%rsp) je 0x2ac839 movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x2ac839 movslq 0x38(%rsp), %rcx movl 0x4c(%rsp), %eax imull %ecx, %eax movl %eax, 0x4c(%rsp) movq 0x30(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x30(%rsp) movl $0x1, 0x38(%rsp) pushq $0x2a popq %rdi callq 0x782bf movq %rax, %r12 leaq 0x8(%rsp), %rdi callq 0x71548 leaq 0x8(%rsp), %rdi xorl %esi, %esi movl %ebx, %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi movl 0x1c(%rsp), %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0xb popq %rsi movl 0x18(%rsp), %edx callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xdc(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xd popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xec(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x4 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xf popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xe popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x10 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax vmovss 0xfc(%r13,%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x12 popq %rsi callq 0x71952 movq (%r13), %rax movq -0x18(%rax), %rax movl 0x100(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x5 popq %rsi callq 0x7193a movl 0x9c(%rsp), %edx leaq 0x8(%rsp), %rdi pushq $0x6 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x108(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x7 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x10c(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x110(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax leaq (%rax,%r13), %rdx addq $0x118, %rdx # imm = 0x118 leaq 0x8(%rsp), %rdi pushq $0xa popq %rsi callq 0x7196c movq (%r12), %rax leaq 0x8(%rsp), %rsi movq %r12, %rdi callq 0x10(%rax) andq $0x0, 0x110(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0xd0(%rsp) vmovups %xmm0, 0xdc(%rsp) vmovaps %xmm0, 0xf0(%rsp) vmovups %xmm0, 0xfc(%rsp) andq $0x0, 0x158(%rsp) vmovups %xmm0, 0x118(%rsp) vmovups %xmm0, 0x124(%rsp) vmovups %xmm0, 0x138(%rsp) vmovups %xmm0, 0x144(%rsp) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac61e lock incl (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2ac655 lock decl (%rax) jne 0x2ac655 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x2ac64d movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ac655 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0xd0(%rsp) movq 0x80(%rsp), %rax movq %rax, 0xe0(%rsp) movl 0x88(%rsp), %eax movl %eax, 0xe8(%rsp) movq 0x90(%rsp), %rax movq %rax, 0xf0(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0xf8(%rsp) movl 0xa8(%rsp), %eax movl %eax, 0x108(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x110(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax je 0x2ac6cf lock incl (%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x2ac706 lock decl (%rax) jne 0x2ac706 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x2ac6fe movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ac706 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 leaq 0xd0(%rsp), %rsi vmovups %xmm0, 0x48(%rsi) movq 0x30(%rsp), %rax movq %rax, 0x58(%rsi) movl 0x38(%rsp), %eax movl %eax, 0x60(%rsi) movq 0x40(%rsp), %rax movq %rax, 0x68(%rsi) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsi) movl 0x58(%rsp), %eax movl %eax, 0x80(%rsi) movq 0x60(%rsp), %rax movq %rax, 0x88(%rsi) leaq 0xc0(%rsp), %rdi callq 0x6b00e movq (%r12), %rax leaq 0xc0(%rsp), %rsi movq %r12, %rdi callq 0x18(%rax) leaq 0xc0(%rsp), %rdi callq 0x6b03a movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq 0x20(%rax) movq (%r12), %rax movq %r12, %rdi movq %r14, %rsi movq 0xb8(%rsp), %rdx movq %r15, %rcx callq 0x38(%rax) movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq 0x28(%rax) movq (%r12), %rax movq %r12, %rdi callq 0x8(%rax) pushq $0x48 popq %rbx vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%rbx), %rax testq %rax, %rax je 0x2ac801 lock decl (%rax) jne 0x2ac801 movq 0xd0(%rsp,%rbx), %rsi movq 0xf0(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x2ac7f5 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2ac801 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%rbx), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x2ac7c2 leaq 0x8(%rsp), %rdi callq 0x71614 xorl %r12d, %r12d movq 0x28(%rsp), %rax testq %rax, %rax je 0x2ac867 lock decl (%rax) jne 0x2ac867 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2ac85f movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ac867 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac898 lock decl (%rax) jne 0x2ac898 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2ac890 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2ac898 movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x168, %rsp # imm = 0x168 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2ac9cd jmp 0x2ac9cd movq %rax, %rbx leaq 0xc0(%rsp), %rdi callq 0x6b03a jmp 0x2ac8da jmp 0x2ac8d7 jmp 0x2ac8cf jmp 0x2ac8cf movq %rax, %rbx jmp 0x2ac964 movq %rax, %rbx pushq $0x48 popq %r14 vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%r14), %rax testq %rax, %rax je 0x2ac921 lock decl (%rax) jne 0x2ac921 movq 0xd0(%rsp,%r14), %rsi movq 0xf0(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2ac915 movq (%rdi), %rax callq 0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2ac921 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2ac8e2 jmp 0x2ac95a jmp 0x2ac9cd movq %rax, %rbx jmp 0x2ac992 jmp 0x2ac9cd movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq 0x28(%rsp), %rax testq %rax, %rax je 0x2ac992 lock decl (%rax) jne 0x2ac992 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2ac98c movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac992 movq (%rdi), %rax callq 0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac9c3 lock decl (%rax) jne 0x2ac9c3 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2ac9bd movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac9c3 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2ac9cd movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
virtual thunk to ncnn::ConvolutionDepthWise_x86_avx::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const	int ConvolutionDepthWise_x86_avx::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w = weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w = bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }	movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x2ac246 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp
ncnn::Padding::load_param(ncnn::ParamDict const&)	int Padding::load_param(const ParamDict& pd) { top = pd.get(0, 0); bottom = pd.get(1, 0); left = pd.get(2, 0); right = pd.get(3, 0); type = pd.get(4, 0); value = pd.get(5, 0.f); per_channel_pad_data_size = pd.get(6, 0); front = pd.get(7, 0); behind = pd.get(8, 0); return 0; }	pushq %r14 pushq %rbx pushq %rax movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd4(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x3 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xdc(%rbx) pushq $0x4 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe0(%rbx) pushq $0x5 popq %rsi xorps %xmm0, %xmm0 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xe4(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xf0(%rbx) pushq $0x7 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe8(%rbx) pushq $0x8 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xec(%rbx) xorl %eax, %eax addq $0x8, %rsp popq %rbx popq %r14 retq nop	/csukuangfj[P]ncnn/src/layer/padding.cpp
ncnn::Padding::load_model(ncnn::ModelBin const&)	int Padding::load_model(const ModelBin& mb) { if (per_channel_pad_data_size) { per_channel_pad_data = mb.load(per_channel_pad_data_size, 1); } return 0; }	pushq %r14 pushq %rbx subq $0x48, %rsp movl 0xf0(%rdi), %edx testl %edx, %edx je 0x2acbe2 movq %rdi, %rbx movq (%rsi), %rax movq %rsp, %r14 pushq $0x1 popq %rcx movq %r14, %rdi callq 0x10(%rax) leaq 0xf8(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x2acbba testq %rax, %rax je 0x2acb2b lock incl (%rax) movq 0x100(%rbx), %rax testq %rax, %rax je 0x2acb5f lock decl (%rax) jne 0x2acb5f movq 0xf8(%rbx), %rsi movq 0x118(%rbx), %rdi testq %rdi, %rdi je 0x2acb57 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2acb5f movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0xf8(%rbx) movq 0x8(%rsp), %rax movq %rax, 0x100(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0x108(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0x110(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x118(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x120(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x130(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x138(%rbx) testq %rax, %rax je 0x2acbe2 lock decl (%rax) jne 0x2acbe2 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2acbda movq (%rdi), %rax callq 0x18(%rax) jmp 0x2acbe2 movq %rsi, %rdi callq 0x5f3e0 xorl %eax, %eax addq $0x48, %rsp popq %rbx popq %r14 retq movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2acc1c lock decl (%rax) jne 0x2acc1c movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2acc16 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2acc1c movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2acc26 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/padding.cpp
ncnn::Padding::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int outw = w + left + right; if (dims == 1) { top_blob.create(outw, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; if (elemsize == 1) copy_make_border_image<signed char>(bottom_blob, top_blob, 0, left, type, static_cast<signed char>(value)); if (elemsize == 2) copy_make_border_image<unsigned short>(bottom_blob, top_blob, 0, left, type, support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(value) : float32_to_bfloat16(value)); if (elemsize == 4) copy_make_border_image<float>(bottom_blob, top_blob, 0, left, type, value); return 0; } int outh = h + top + bottom; if (dims == 2) { top_blob.create(outw, outh, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; if (elemsize == 1) copy_make_border_image<signed char>(bottom_blob, top_blob, top, left, type, static_cast<signed char>(value)); if (elemsize == 2) copy_make_border_image<unsigned short>(bottom_blob, top_blob, top, left, type, support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(value) : float32_to_bfloat16(value)); if (elemsize == 4) copy_make_border_image<float>(bottom_blob, top_blob, top, left, type, value); return 0; } if (dims == 3) { int outc = channels + front + behind; top_blob.create(outw, outh, outc, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc; q++) { Mat borderm = top_blob.channel(q); float pad_value = per_channel_pad_data_size ? per_channel_pad_data[q] : value; //Channel padding if (((q < front) \|\| (q >= (channels + front))) && type == 0) { if (elemsize == 1) { borderm.fill(static_cast<signed char>(pad_value)); } if (elemsize == 2) { borderm.fill(support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(pad_value) : float32_to_bfloat16(pad_value)); } if (elemsize == 4) { borderm.fill(pad_value); } } else { int q_ = q - front; if (type == 1) { q_ = q_ <= 0 ? 0 : q_; q_ = q_ >= channels - 1 ? channels - 1 : q_; } if (type == 2) { q_ = abs(q_); q_ = (channels - 1) - abs(q_ - (channels - 1)); } const Mat m = bottom_blob.channel(q_); if (elemsize == 1) copy_make_border_image<signed char>(m, borderm, top, left, type, static_cast<signed char>(pad_value)); if (elemsize == 2) copy_make_border_image<unsigned short>(m, borderm, top, left, type, support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(pad_value) : float32_to_bfloat16(pad_value)); if (elemsize == 4) copy_make_border_image<float>(m, borderm, top, left, type, pad_value); } } return 0; } if (dims == 4) { int outd = d + front + behind; top_blob.create(outw, outh, outd, channels, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float pad_value = per_channel_pad_data_size ? per_channel_pad_data[q] : value; for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if (((z < front) \|\| (z >= (d + front))) && type == 0) { if (elemsize == 1) { borderm.fill(static_cast<signed char>(pad_value)); } if (elemsize == 2) { borderm.fill(support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(pad_value) : float32_to_bfloat16(pad_value)); } if (elemsize == 4) { borderm.fill(pad_value); } } else { int z_ = z - front; if (type == 1) { z_ = z_ <= 0 ? 0 : z_; z_ = z_ >= d - 1 ? d - 1 : z_; } if (type == 2) { z_ = abs(z_); z_ = (d - 1) - abs(z_ - (d - 1)); } const Mat m = bottom_blob.channel(q).depth(z_); if (elemsize == 1) copy_make_border_image<signed char>(m, borderm, top, left, type, static_cast<signed char>(pad_value)); if (elemsize == 2) copy_make_border_image<unsigned short>(m, borderm, top, left, type, support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(pad_value) : float32_to_bfloat16(pad_value)); if (elemsize == 4) copy_make_border_image<float>(m, borderm, top, left, type, pad_value); } } } return 0; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x118, %rsp # imm = 0x118 movq %rcx, 0x20(%rsp) movq %rdx, %rbx movq %rsi, %r14 movl 0xd0(%rdi), %eax movl 0xd4(%rdi), %edx testl %eax, %eax sete %cl testl %edx, %edx sete %sil movq 0xd8(%rdi), %xmm1 movq %rdi, 0x8(%rsp) movq 0xe8(%rdi), %xmm0 movdqa %xmm1, %xmm2 punpcklqdq %xmm0, %xmm2 # xmm2 = xmm2[0],xmm0[0] pxor %xmm3, %xmm3 pcmpeqd %xmm2, %xmm3 movmskps %xmm3, %edi xorl $0xf, %edi sete %dil andb %cl, %sil andb %dil, %sil cmpb $0x1, %sil jne 0x2acceb movq $0x0, 0x10(%rsp) cmpq %r14, %rbx je 0x2ad8eb movq 0x8(%r14), %rax testq %rax, %rax je 0x2accba lock incl (%rax) movq 0x8(%rbx), %rax testq %rax, %rax je 0x2ad760 lock decl (%rax) jne 0x2ad760 movq (%rbx), %rsi movq 0x20(%rbx), %rdi testq %rdi, %rdi je 0x2ad758 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ad760 movl 0x28(%r14), %ecx movq 0x10(%r14), %r8 pshufd $0x55, %xmm1, %xmm2 # xmm2 = xmm1[1,1,1,1] paddd %xmm1, %xmm2 movd %xmm2, %esi addl 0x2c(%r14), %esi pushq $-0x64 popq %rdi movq %rdi, 0x10(%rsp) cmpl $0x1, %ecx movq %r8, 0x28(%rsp) jne 0x2acdaf movq 0x20(%rsp), %rax movq 0x8(%rax), %rcx movq %rbx, %rdi movq %r8, %rdx callq 0x635fa cmpq $0x0, (%rbx) je 0x2ad8eb movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ad8eb movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad7ef cmpq $0x2, %rax je 0x2ad7b8 movq $0x0, 0x10(%rsp) cmpq $0x1, %rax jne 0x2ad8eb movq 0x8(%rsp), %rax movl 0xd8(%rax), %ecx movl 0xe0(%rax), %r8d cvttss2si 0xe4(%rax), %eax movsbl %al, %r9d movq $0x0, 0x10(%rsp) movq %r14, %rdi movq %rbx, %rsi xorl %edx, %edx callq 0x2ad910 jmp 0x2ad8eb movl 0x38(%r14), %edi movq %rdi, 0x40(%rsp) addl %eax, %edx addl 0x30(%r14), %edx pshufd $0x55, %xmm0, %xmm1 # xmm1 = xmm0[1,1,1,1] cmpl $0x4, %ecx movq %r14, 0x60(%rsp) je 0x2ad2bc cmpl $0x3, %ecx je 0x2ace6c cmpl $0x2, %ecx jne 0x2ad8e2 movq 0x20(%rsp), %rax movq 0x8(%rax), %r8 movq %rbx, %rdi movq 0x28(%rsp), %rcx callq 0x636fa cmpq $0x0, (%rbx) je 0x2ad8eb movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ad8eb movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad892 cmpq $0x2, %rax je 0x2ad854 cmpq $0x1, %rax jne 0x2ad8e2 movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx movl 0xe0(%rax), %r8d cvttss2si 0xe4(%rax), %eax movsbl %al, %r9d movq %r14, %rdi movq %rbx, %rsi callq 0x2ad910 jmp 0x2ad8e2 paddd %xmm0, %xmm1 movd %xmm1, %ecx addl 0x40(%rsp), %ecx movq 0x20(%rsp), %rax movq 0x8(%rax), %r9 movq %rbx, %rdi movq %rcx, %r14 movq 0x28(%rsp), %r8 callq 0x63810 cmpq $0x0, (%rbx) je 0x2ad8eb movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ad8eb movq 0x8(%rsp), %rax addq $0xe4, %rax movq %rax, 0x50(%rsp) xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x10(%rsp) cmovlel %ecx, %r14d movq %r14, 0x48(%rsp) movq 0x40(%rsp), %rax decl %eax movl %eax, 0x68(%rsp) xorl %esi, %esi cmpq 0x48(%rsp), %rsi je 0x2ad8eb movslq 0x2c(%rbx), %rax movslq 0x30(%rbx), %rcx movl 0x34(%rbx), %r14d movq (%rbx), %r12 movq 0x10(%rbx), %rbp movq 0x40(%rbx), %r13 movq %r13, %rdx movq %rsi, 0x38(%rsp) imulq %rsi, %rdx imulq %rbp, %rdx addq %r12, %rdx movl 0x18(%rbx), %esi movq 0x20(%rbx), %rdi movq %rdx, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq %rbp, 0xc8(%rsp) movl %esi, 0xd0(%rsp) movq %rdi, 0xd8(%rsp) movl %eax, 0xe4(%rsp) movl %ecx, 0xe8(%rsp) movl $0x1, 0xec(%rsp) movl %r14d, 0xf0(%rsp) imulq %rax, %rcx movq %rbp, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rbp movq %rax, %r15 movq %rax, 0xf8(%rsp) movl 0x28(%rbx), %eax leal -0x1(%rax), %edx movl %edx, 0xe0(%rsp) cmpl $0x4, %eax jne 0x2acfa7 movq %rcx, 0xf8(%rsp) movq %rcx, %r15 movq 0x38(%rsp), %rdi leaq (,%rdi,4), %rax movq 0x8(%rsp), %rcx addq 0xf8(%rcx), %rax cmpl $0x0, 0xf0(%rcx) cmoveq 0x50(%rsp), %rax movd (%rax), %xmm0 movl 0xe0(%rcx), %r8d movl 0xe8(%rcx), %edx movslq %edx, %rax cmpq %rax, %rdi setl %cl addl 0x40(%rsp), %eax cltq cmpq %rax, %rdi setge %al orb %cl, %al testl %r8d, %r8d sete %sil andb %al, %sil movd %xmm0, %ecx cmpb $0x1, %sil jne 0x2ad061 movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad201 cmpq $0x2, %rax je 0x2ad19f cmpq $0x1, %rax jne 0x2ad2af cvttss2si %xmm0, %eax imull %r15d, %r14d testl %r14d, %r14d movl $0x0, %ecx cmovlel %ecx, %r14d imulq %rbp, %r13 imulq 0x38(%rsp), %r13 addq %r13, %r12 xorl %ecx, %ecx cmpq %rcx, %r14 je 0x2ad2af movb %al, (%r12,%rcx) incq %rcx jmp 0x2ad04f movl %edi, %eax subl %edx, %eax cmpl $0x2, %r8d movl %r8d, %ebp je 0x2ad08d cmpl $0x1, %ebp movq 0x60(%rsp), %r11 jne 0x2ad0a8 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax movl 0x68(%rsp), %edx cmpl %edx, %eax cmovgel %edx, %eax jmp 0x2ad0a8 movl %eax, %edx negl %edx cmovsl %eax, %edx movl 0x68(%rsp), %esi subl %esi, %edx movl %edx, %eax negl %eax cmovnsl %edx, %eax addl %esi, %eax movq 0x60(%rsp), %r11 movslq 0x2c(%r11), %rdx movslq 0x30(%r11), %rsi cltq imulq 0x40(%r11), %rax movq 0x10(%r11), %rdi imulq %rdi, %rax addq (%r11), %rax movl 0x34(%r11), %r8d movl 0x18(%r11), %r9d movq 0x20(%r11), %r10 movq %rax, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdi, 0x80(%rsp) movl %r9d, 0x88(%rsp) movq %r10, 0x90(%rsp) movl %edx, 0x9c(%rsp) movl %esi, 0xa0(%rsp) movl $0x1, 0xa4(%rsp) movl %r8d, 0xa8(%rsp) imulq %rdx, %rsi movq %rdi, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xb0(%rsp) movl 0x28(%r11), %eax leal -0x1(%rax), %edx movl %edx, 0x98(%rsp) cmpl $0x4, %eax jne 0x2ad14d movq %rsi, 0xb0(%rsp) movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad234 cmpq $0x2, %rax je 0x2ad1c9 cmpq $0x1, %rax jne 0x2ad2af movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx cvttss2si %xmm0, %eax movsbl %al, %r9d leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %ebp, %r8d callq 0x2ad910 jmp 0x2ad2af movq 0x8(%rsp), %rax cmpb $0x1, 0xd(%rax) jne 0x2ad25c movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) jne 0x2ad25c callq 0x6460e movl %eax, %ecx jmp 0x2ad25f movq 0x8(%rsp), %rax movl 0xd0(%rax), %r15d movl 0xd8(%rax), %r14d cmpb $0x1, 0xd(%rax) jne 0x2ad28d movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) jne 0x2ad28d callq 0x6460e movl %eax, %ecx jmp 0x2ad290 imull %r15d, %r14d testl %r14d, %r14d movl $0x0, %eax cmovlel %eax, %r14d imulq %rbp, %r13 imulq 0x38(%rsp), %r13 addq %r13, %r12 xorl %eax, %eax cmpl %eax, %r14d je 0x2ad2af movss %xmm0, (%r12,%rax,4) incq %rax jmp 0x2ad220 movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %ebp, %r8d callq 0x2ae697 jmp 0x2ad2af shrl $0x10, %ecx imull %r15d, %r14d testl %r14d, %r14d movl $0x0, %eax cmovlel %eax, %r14d imulq %rbp, %r13 imulq 0x38(%rsp), %r13 addq %r13, %r12 xorl %eax, %eax cmpq %rax, %r14 je 0x2ad2af movw %cx, (%r12,%rax,2) incq %rax jmp 0x2ad27e shrl $0x10, %ecx movzwl %cx, %r9d leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %r15d, %edx movl %r14d, %ecx movl %ebp, %r8d callq 0x2adf5b movq 0x38(%rsp), %rsi incq %rsi jmp 0x2acee2 movl 0x34(%r14), %eax paddd %xmm0, %xmm1 movd %xmm1, %ecx movq %rax, 0x100(%rsp) addl %eax, %ecx movq 0x20(%rsp), %rax movq 0x8(%rax), %rax movq %rax, (%rsp) movq %rbx, %rdi movq %rcx, %r14 movq 0x40(%rsp), %r8 movq 0x28(%rsp), %r9 callq 0x6393c cmpq $0x0, (%rbx) je 0x2ad8eb movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ad8eb movq 0x8(%rsp), %rax addq $0xe4, %rax movq %rax, 0x108(%rsp) xorl %ecx, %ecx testl %r14d, %r14d cmovlel %ecx, %r14d movq %r14, 0x68(%rsp) movq 0x100(%rsp), %rax decl %eax movl %eax, 0x5c(%rsp) movq 0x40(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x10(%rsp) cmovlel %ecx, %edx movq %rdx, 0x40(%rsp) xorl %edx, %edx movq %rbx, 0x110(%rsp) cmpq 0x40(%rsp), %rdx je 0x2ad8eb leaq (,%rdx,4), %rax movq 0x8(%rsp), %rcx addq 0xf8(%rcx), %rax cmpl $0x0, 0xf0(%rcx) cmoveq 0x108(%rsp), %rax movd (%rax), %xmm0 movd %xmm0, %eax shrl $0x10, %eax movl %eax, 0x34(%rsp) cvttss2si %xmm0, %eax movl %eax, 0x38(%rsp) xorl %edi, %edi movss %xmm0, 0x1c(%rsp) movq %rdx, 0x50(%rsp) cmpq 0x68(%rsp), %rdi je 0x2ad750 movslq 0x2c(%rbx), %r15 movslq 0x30(%rbx), %r14 movq (%rbx), %r13 movq %rbx, %rsi movq 0x10(%rbx), %rbx movq 0x40(%rsi), %r12 imulq %rdx, %r12 movq %r12, %rax imulq %rbx, %rax addq %r13, %rax movl 0x18(%rsi), %ecx movq %r14, %rbp imulq %r15, %rbp movq %rdi, %rdx imulq %rbx, %rdx imulq %rbp, %rdx addq %rax, %rdx movq 0x20(%rsi), %rax movq %rdx, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq %rbx, 0xc8(%rsp) movl %ecx, 0xd0(%rsp) movq %rax, 0xd8(%rsp) movl $0x2, 0xe0(%rsp) movl %r15d, 0xe4(%rsp) movl %r14d, 0xe8(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xec(%rsp) movq %rbp, 0xf8(%rsp) movq 0x8(%rsp), %rsi movl 0xe8(%rsi), %ecx movslq %ecx, %rax cmpq %rax, %rdi setl %dl addl 0x100(%rsp), %eax cltq cmpq %rax, %rdi movl 0xe0(%rsi), %esi setge %al orb %dl, %al testl %esi, %esi sete %dl andb %al, %dl cmpb $0x1, %dl movq %rdi, 0x48(%rsp) jne 0x2ad4f3 movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad6d4 cmpq $0x2, %rax je 0x2ad627 cmpq $0x1, %rax jne 0x2ad736 testl %ebp, %ebp movl $0x0, %eax cmovlel %eax, %ebp imulq 0x48(%rsp), %r14 imulq %r15, %r14 addq %r12, %r14 imulq %r14, %rbx addq %rbx, %r13 xorl %eax, %eax cmpq %rax, %rbp je 0x2ad736 movl 0x38(%rsp), %ecx movb %cl, (%r13,%rax) incq %rax jmp 0x2ad4dc movl %edi, %eax subl %ecx, %eax cmpl $0x2, %esi je 0x2ad523 cmpl $0x1, %esi movq 0x60(%rsp), %r9 movq 0x50(%rsp), %rdi movl %esi, %r14d jne 0x2ad546 testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax movl 0x5c(%rsp), %ecx cmpl %ecx, %eax cmovgel %ecx, %eax jmp 0x2ad546 movl %esi, %r14d movl %eax, %ecx negl %ecx cmovsl %eax, %ecx movl 0x5c(%rsp), %edx subl %edx, %ecx movl %ecx, %eax negl %eax cmovnsl %ecx, %eax addl %edx, %eax movq 0x60(%rsp), %r9 movq 0x50(%rsp), %rdi movslq 0x2c(%r9), %rcx movslq 0x30(%r9), %rdx movq 0x40(%r9), %rsi imulq %rdi, %rsi movq 0x10(%r9), %rdi imulq %rdi, %rsi addq (%r9), %rsi movl 0x18(%r9), %r8d movq 0x20(%r9), %r9 movq %rdx, %r10 imulq %rcx, %r10 cltq imulq %rdi, %rax imulq %r10, %rax addq %rsi, %rax movq %rax, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdi, 0x80(%rsp) movl %r8d, 0x88(%rsp) movq %r9, 0x90(%rsp) movl $0x2, 0x98(%rsp) movl %ecx, 0x9c(%rsp) movl %edx, 0xa0(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa4(%rsp) movq %r10, 0xb0(%rsp) movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad704 cmpq $0x2, %rax je 0x2ad684 cmpq $0x1, %rax jne 0x2ad736 movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx movsbl 0x38(%rsp), %r9d leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %r14d, %r8d callq 0x2ad910 jmp 0x2ad730 movq 0x8(%rsp), %rax cmpb $0x1, 0xd(%rax) movl 0x34(%rsp), %eax jne 0x2ad650 movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) movl 0x34(%rsp), %eax jne 0x2ad650 callq 0x6460e movss 0x1c(%rsp), %xmm0 testl %ebp, %ebp movl $0x0, %ecx cmovlel %ecx, %ebp imulq 0x48(%rsp), %r14 imulq %r15, %r14 addq %r12, %r14 imulq %r14, %rbx addq %rbx, %r13 xorl %ecx, %ecx cmpq %rcx, %rbp je 0x2ad736 movw %ax, (%r13,%rcx,2) incq %rcx jmp 0x2ad670 movq 0x8(%rsp), %rax movl 0xd0(%rax), %r13d movl 0xd8(%rax), %ebp cmpb $0x1, 0xd(%rax) movl 0x34(%rsp), %eax jne 0x2ad6b4 movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) movl 0x34(%rsp), %eax jne 0x2ad6b4 callq 0x6460e movzwl %ax, %r9d leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %r13d, %edx movl %ebp, %ecx movl %r14d, %r8d callq 0x2adf5b jmp 0x2ad730 testl %ebp, %ebp movl $0x0, %eax cmovlel %eax, %ebp imulq 0x48(%rsp), %r14 imulq %r15, %r14 addq %r12, %r14 imulq %r14, %rbx addq %rbx, %r13 xorl %eax, %eax cmpl %eax, %ebp je 0x2ad736 movss %xmm0, (%r13,%rax,4) incq %rax jmp 0x2ad6f4 movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %r14d, %r8d movss 0x1c(%rsp), %xmm0 callq 0x2ae697 movss 0x1c(%rsp), %xmm0 movq 0x48(%rsp), %rdi incq %rdi movq 0x110(%rsp), %rbx movq 0x50(%rsp), %rdx jmp 0x2ad3b5 incq %rdx jmp 0x2ad362 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rbx) pxor %xmm0, %xmm0 movdqu %xmm0, (%rbx) movdqu %xmm0, 0xc(%rbx) andl $0x0, 0x38(%rbx) movdqu %xmm0, 0x28(%rbx) movups (%r14), %xmm0 movups %xmm0, (%rbx) movq 0x10(%r14), %rax movq %rax, 0x10(%rbx) movl 0x18(%r14), %eax movl %eax, 0x18(%rbx) movq 0x20(%r14), %rax movq %rax, 0x20(%rbx) movdqu 0x28(%r14), %xmm0 movdqu %xmm0, 0x28(%rbx) movl 0x38(%r14), %eax movl %eax, 0x38(%rbx) movq 0x40(%r14), %rax movq %rax, 0x40(%rbx) jmp 0x2ad8eb movq 0x8(%rsp), %rax movl 0xd8(%rax), %ebp movl 0xe0(%rax), %r15d cmpb $0x1, 0xd(%rax) jne 0x2ad824 movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) jne 0x2ad824 movq 0x8(%rsp), %rax movd 0xe4(%rax), %xmm0 callq 0x6460e jmp 0x2ad830 movq 0x8(%rsp), %rax movl 0xd8(%rax), %ecx movl 0xe0(%rax), %r8d movd 0xe4(%rax), %xmm0 movq $0x0, 0x10(%rsp) movq %r14, %rdi movq %rbx, %rsi xorl %edx, %edx callq 0x2ae697 jmp 0x2ad8eb movq 0x8(%rsp), %rax movzwl 0xe6(%rax), %eax movzwl %ax, %r9d movq $0x0, 0x10(%rsp) movq %r14, %rdi movq %rbx, %rsi xorl %edx, %edx movl %ebp, %ecx movl %r15d, %r8d callq 0x2adf5b jmp 0x2ad8eb movq 0x8(%rsp), %rax movl 0xd0(%rax), %ebp movl 0xd8(%rax), %r15d movl 0xe0(%rax), %r13d cmpb $0x1, 0xd(%rax) jne 0x2ad8bf movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) jne 0x2ad8bf movq 0x8(%rsp), %rax movd 0xe4(%rax), %xmm0 callq 0x6460e jmp 0x2ad8cb movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx movl 0xe0(%rax), %r8d movd 0xe4(%rax), %xmm0 movq %r14, %rdi movq %rbx, %rsi callq 0x2ae697 jmp 0x2ad8e2 movq 0x8(%rsp), %rax movzwl 0xe6(%rax), %eax movzwl %ax, %r9d movq %r14, %rdi movq %rbx, %rsi movl %ebp, %edx movl %r15d, %ecx movl %r13d, %r8d callq 0x2adf5b movq $0x0, 0x10(%rsp) movq 0x10(%rsp), %rax addq $0x118, %rsp # imm = 0x118 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2ad908 jmp 0x2ad908 jmp 0x2ad908 movq %rax, %rdi callq 0x5f340	/csukuangfj[P]ncnn/src/layer/padding.cpp
void ncnn::copy_make_border_image<signed char>(ncnn::Mat const&, ncnn::Mat&, int, int, int, signed char)	static void copy_make_border_image(const Mat& src, Mat& dst, int top, int left, int type, T v) { int w = dst.w; int h = dst.h; const T* ptr = src; T* outptr = dst; if (type == 0) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = v; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = v; } ptr += src.w; outptr += w; } // fill bottom for (; y < h; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } } if (type == 1) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } ptr += src.w; outptr += w; } // fill bottom ptr -= src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } } if (type == 2) { int y = 0; // fill top ptr += top * src.w; for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } ptr += src.w; outptr += w; } // fill bottom ptr -= 2 * src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } } }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x38, %rsp movl %r9d, %r13d movl %ecx, %r9d movq %rdi, %r14 movslq 0x2c(%rsi), %r12 movl 0x30(%rsi), %eax movl %eax, 0x2c(%rsp) movq (%rdi), %r15 movq (%rsi), %rbp movslq %ecx, %rax movq %rax, (%rsp) testl %r8d, %r8d movl %ecx, 0xc(%rsp) movl %edx, 0x8(%rsp) jne 0x2ada5f movl %r8d, 0x30(%rsp) xorl %eax, %eax testl %r12d, %r12d movl $0x0, %ebx cmovgl %r12d, %ebx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi cmpl %esi, %eax je 0x2ad989 xorl %ecx, %ecx cmpq %rcx, %rbx je 0x2ad982 movb %r13b, (%rbp,%rcx) incq %rcx jmp 0x2ad973 addq %r12, %rbp incl %eax jmp 0x2ad96d xorl %edi, %edi movq (%rsp), %rax testl %eax, %eax cmovgl %r9d, %edi negq %rax movq %rax, 0x18(%rsp) movq %rdi, 0x20(%rsp) movl 0x30(%r14), %eax addl 0x8(%rsp), %eax cmpl %eax, %esi jge 0x2ada54 xorl %eax, %eax cmpq %rax, %rdi je 0x2ad9c3 movb %r13b, (%rbp,%rax) incq %rax jmp 0x2ad9b4 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2ad9f4 movq 0x18(%rsp), %rax leaq (%r15,%rax), %rcx movq %rdi, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2ada1c movb (%rcx,%rax), %dl movb %dl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2ad9d9 movq (%rsp), %rax leaq (%rax,%rbp), %rdi movl %esi, 0x10(%rsp) movq %r15, %rsi callq 0x5f3c0 movq 0x20(%rsp), %rdi movl 0x10(%rsp), %esi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %edi, %eax cltq cmpq %r12, %rax jge 0x2ada2d movb %r13b, (%rbp,%rax) incq %rax jmp 0x2ada1e movslq 0x2c(%r14), %rax addq %rax, %r15 addq %r12, %rbp incl %esi jmp 0x2ad9a2 xorl %eax, %eax cmpq %rax, %rbx je 0x2ada4f movb %r13b, (%rbp,%rax) incq %rax jmp 0x2ada40 addq %r12, %rbp incl %esi cmpl 0x2c(%rsp), %esi jl 0x2ada3e movl 0x30(%rsp), %r8d cmpl $0x1, %r8d jne 0x2adb3c movl %r8d, 0x30(%rsp) xorl %r8d, %r8d movq (%rsp), %rax testl %eax, %eax movl $0x0, %ebx cmovgl %r9d, %ebx movl 0x8(%rsp), %ecx testl %ecx, %ecx movl $0x0, %esi cmovgl %ecx, %esi leaq (%rbx,%rbp), %rdi movq %r15, %r13 subq %rax, %r13 movl %esi, 0x20(%rsp) cmpl %esi, %r8d je 0x2adb45 xorl %eax, %eax cmpq %rax, %rbx je 0x2adab8 movb (%r15), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adaa7 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2adae2 movq %rbx, %rax addl %r9d, %edx movslq %edx, %rcx cmpq %rcx, %rax jge 0x2adb15 movb (%r13,%rax), %cl movb %cl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2adac5 movq (%rsp), %rax movq %rdi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movq %r15, %rsi movq %r8, 0x18(%rsp) callq 0x5f3c0 movq 0x18(%rsp), %r8 movq 0x10(%rsp), %rdi movl 0x20(%rsp), %esi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %ebx, %eax cltq cmpq %r12, %rax jge 0x2adb2e movslq 0x2c(%r14), %rcx movb -0x1(%rcx,%r15), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adb17 addq %r12, %rbp incl %r8d addq %r12, %rdi jmp 0x2ada9c movl 0x8(%rsp), %edx jmp 0x2adcb4 movq (%rsp), %rax negq %rax movq %rax, 0x18(%rsp) movl 0x30(%r14), %eax addl 0x8(%rsp), %eax cmpl %eax, %esi jge 0x2adbf9 xorl %eax, %eax cmpq %rax, %rbx je 0x2adb74 movb (%r15), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adb63 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2adba5 movq 0x18(%rsp), %rax leaq (%r15,%rax), %rcx movq %rbx, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2adbd0 movb (%rcx,%rax), %dl movb %dl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2adb8a movq (%rsp), %rax movq %rdi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movl %esi, %r13d movq %r15, %rsi callq 0x5f3c0 movq 0x10(%rsp), %rdi movl %r13d, %esi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %ebx, %eax cltq movslq 0x2c(%r14), %rcx cmpq %r12, %rax jge 0x2adbe9 movb -0x1(%rcx,%r15), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adbd2 addq %rcx, %r15 addq %r12, %rbp incl %esi addq %r12, %rdi jmp 0x2adb51 movslq 0x2c(%r14), %rax movq %r15, %r13 subq %rax, %r13 addq (%rsp), %rax subq %rax, %r15 cmpl 0x2c(%rsp), %esi jge 0x2adca8 xorl %eax, %eax cmpq %rax, %rbx je 0x2adc28 movb (%r13), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adc16 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2adc55 movq %rdi, %rcx movq %rbx, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2adc82 movb (%r15,%rax), %dl movb %dl, (%rcx) incq %rax movl 0x2c(%r14), %edx incq %rcx jmp 0x2adc38 movq (%rsp), %rax movq %rdi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movl %esi, 0x20(%rsp) movq %r13, %rsi callq 0x5f3c0 movq 0x10(%rsp), %rdi movl 0x20(%rsp), %esi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %ebx, %eax cltq cmpq %r12, %rax jge 0x2adc9b movslq 0x2c(%r14), %rcx movb -0x1(%r13,%rcx), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adc84 addq %r12, %rbp incl %esi addq %r12, %rdi jmp 0x2adc0a movq %r13, %r15 movl 0x8(%rsp), %edx movl 0x30(%rsp), %r8d cmpl $0x2, %r8d jne 0x2adf4c movl 0x2c(%r14), %eax movl %eax, %ecx imull %edx, %ecx movslq %ecx, %rcx addq %rcx, %r15 xorl %esi, %esi movq (%rsp), %rcx testl %ecx, %ecx movl $0x0, %r13d cmovgl %ecx, %r13d testl %edx, %edx movl $0x0, %edi cmovgl %edx, %edi movq %rbp, %rbx addq %r13, %rbx movq %rcx, %rdx negq %rdx movq %rdx, 0x20(%rsp) leal -0x2(%rcx), %r10d movl %r10d, 0x18(%rsp) movl %edi, 0x10(%rsp) cmpl %edi, %esi je 0x2addc4 movq (%rsp), %rax addq %r15, %rax xorl %ecx, %ecx cmpq %rcx, %r13 je 0x2add2b movb (%rax), %dl movb %dl, (%rbp,%rcx) incq %rcx decq %rax jmp 0x2add18 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2add5c movq 0x20(%rsp), %rax leaq (%r15,%rax), %rcx movq %r13, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2add8b movb (%rcx,%rax), %dl movb %dl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2add41 movq (%rsp), %rax leaq (%rax,%rbp), %rdi movq %rsi, 0x30(%rsp) movq %r15, %rsi callq 0x5f3c0 movq 0x30(%rsp), %rsi movl 0x18(%rsp), %r10d movl 0x10(%rsp), %edi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %r13d, %eax cltq movl %r10d, %ecx subl %eax, %ecx cmpq %r12, %rax jge 0x2addb0 movl 0x2c(%r14), %edx leal (%rcx,%rdx,2), %edx movslq %edx, %rdx movb (%r15,%rdx), %dl movb %dl, (%rbp,%rax) incq %rax decl %ecx jmp 0x2add92 addq %r12, %rbp movslq 0x2c(%r14), %rax subq %rax, %r15 incl %esi addq %r12, %rbx jmp 0x2add07 movl 0x8(%rsp), %esi movl 0x30(%r14), %ecx addl %esi, %ecx cmpl %ecx, %edi jge 0x2ade88 movq (%rsp), %rax addq %r15, %rax xorl %ecx, %ecx cmpq %rcx, %r13 je 0x2addf2 movb (%rax), %dl movb %dl, (%rbp,%rcx) incq %rcx decq %rax jmp 0x2adddf movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2ade23 movq 0x20(%rsp), %rax leaq (%r15,%rax), %rcx movq %r13, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2ade50 movb (%rcx,%rax), %dl movb %dl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2ade08 movq (%rsp), %rax movl %edi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movq %r15, %rsi callq 0x5f3c0 movl 0x8(%rsp), %esi movl 0x18(%rsp), %r10d movl 0x10(%rsp), %edi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %r13d, %eax movslq %eax, %rcx movl %r10d, %edx subl %ecx, %edx movl 0x2c(%r14), %eax cmpq %r12, %rcx jge 0x2ade75 leal (%rdx,%rax,2), %eax cltq movb (%r15,%rax), %al movb %al, (%rbp,%rcx) incq %rcx decl %edx jmp 0x2ade58 movslq %eax, %rcx addq %rcx, %r15 addq %r12, %rbp incl %edi addq %r12, %rbx jmp 0x2addc8 addl %eax, %eax cltq subq %rax, %r15 cmpl 0x2c(%rsp), %edi jge 0x2adf4c movq (%rsp), %rax addq %r15, %rax xorl %ecx, %ecx cmpq %rcx, %r13 je 0x2adeb5 movb (%rax), %dl movb %dl, (%rbp,%rcx) incq %rcx decq %rax jmp 0x2adea2 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2adeea movq 0x20(%rsp), %rax leaq (%r15,%rax), %rcx movq %rbx, %rsi movq %r13, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2adf13 movb (%rcx,%rax), %dl movb %dl, (%rsi) incq %rax movl 0x2c(%r14), %edx incq %rsi jmp 0x2adece movq (%rsp), %rax movl %edi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movq %r15, %rsi callq 0x5f3c0 movl 0x18(%rsp), %r10d movl 0x10(%rsp), %edi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %r13d, %eax cltq movl %r10d, %ecx subl %eax, %ecx cmpq %r12, %rax jge 0x2adf38 movl 0x2c(%r14), %edx leal (%rcx,%rdx,2), %edx movslq %edx, %rdx movb (%r15,%rdx), %dl movb %dl, (%rbp,%rax) incq %rax decl %ecx jmp 0x2adf1a addq %r12, %rbp movslq 0x2c(%r14), %rax subq %rax, %r15 incl %edi addq %r12, %rbx jmp 0x2ade8f addq $0x38, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq	/csukuangfj[P]ncnn/src/layer/padding.cpp
void ncnn::copy_make_border_image<unsigned short>(ncnn::Mat const&, ncnn::Mat&, int, int, int, unsigned short)	static void copy_make_border_image(const Mat& src, Mat& dst, int top, int left, int type, T v) { int w = dst.w; int h = dst.h; const T* ptr = src; T* outptr = dst; if (type == 0) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = v; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = v; } ptr += src.w; outptr += w; } // fill bottom for (; y < h; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } } if (type == 1) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } ptr += src.w; outptr += w; } // fill bottom ptr -= src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } } if (type == 2) { int y = 0; // fill top ptr += top * src.w; for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } ptr += src.w; outptr += w; } // fill bottom ptr -= 2 * src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } } }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x58, %rsp movslq 0x2c(%rsi), %rbx movl 0x30(%rsi), %eax movl %eax, 0x4c(%rsp) movq (%rdi), %r12 movq (%rsi), %rbp movq %rcx, 0x28(%rsp) movslq %ecx, %rax movq %rax, 0x18(%rsp) testl %r8d, %r8d movq %rdi, 0x40(%rsp) movl %edx, 0xc(%rsp) jne 0x2ae0d4 movl %r9d, %r13d movl %r8d, 0x30(%rsp) xorl %eax, %eax testl %ebx, %ebx movl $0x0, %r14d cmovgl %ebx, %r14d testl %edx, %edx movl $0x0, %r9d cmovgl %edx, %r9d leaq (%rbx,%rbx), %r10 cmpl %r9d, %eax je 0x2adfdd xorl %ecx, %ecx cmpq %rcx, %r14 je 0x2adfd6 movw %r13w, (%rbp,%rcx,2) incq %rcx jmp 0x2adfc6 incl %eax addq %r10, %rbp jmp 0x2adfbf xorl %r15d, %r15d movq 0x18(%rsp), %rax testl %eax, %eax cmovgl 0x28(%rsp), %r15d addq %rax, %rax negq %rax movq %rax, 0x38(%rsp) movq %r10, 0x20(%rsp) movl 0x30(%rdi), %eax addl 0xc(%rsp), %eax cmpl %eax, %r9d jge 0x2ae0ac xorl %eax, %eax cmpq %rax, %r15 je 0x2ae01f movw %r13w, (%rbp,%rax,2) incq %rax jmp 0x2ae00f movl 0x2c(%rdi), %edx cmpl $0xb, %edx jg 0x2ae052 movq 0x28(%rsp), %rax addl %edx, %eax movslq %eax, %rcx movq 0x38(%rsp), %rax leaq (%r12,%rax), %rsi movq %r15, %rax cmpq %rcx, %rax jge 0x2ae088 movzwl (%rsi,%rax,2), %r8d movw %r8w, (%rbp,%rax,2) incq %rax jmp 0x2ae03d movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addq %rdx, %rdx movq %r12, %rsi movl %r9d, 0x10(%rsp) callq 0x5f3c0 movq 0x20(%rsp), %r10 movl 0x10(%rsp), %r9d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r15), %eax cltq cmpq %rbx, %rax jge 0x2ae09a movw %r13w, (%rbp,%rax,2) incq %rax jmp 0x2ae08a movslq %edx, %rax leaq (%r12,%rax,2), %r12 addq %r10, %rbp incl %r9d jmp 0x2adffd movl 0x4c(%rsp), %ecx cmpl %ecx, %r9d jge 0x2ae0cf xorl %eax, %eax cmpq %rax, %r14 je 0x2ae0c7 movw %r13w, (%rbp,%rax,2) incq %rax jmp 0x2ae0b7 addq %r10, %rbp incl %r9d jmp 0x2ae0b0 movl 0x30(%rsp), %r8d cmpl $0x1, %r8d jne 0x2ae1dc movl %r8d, 0x30(%rsp) xorl %esi, %esi movq 0x18(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl 0x28(%rsp), %r14d movl 0xc(%rsp), %ecx testl %ecx, %ecx movl $0x0, %r8d cmovgl %ecx, %r8d leaq (%rbx,%rbx), %r9 leaq (%rbp,%r14,2), %r10 addq %rax, %rax movq %r12, %r15 movq %rax, 0x50(%rsp) subq %rax, %r15 movq %r9, 0x20(%rsp) cmpl %r8d, %esi je 0x2ae1e5 xorl %eax, %eax cmpq %rax, %r14 je 0x2ae143 movzwl (%r12), %ecx movw %cx, (%rbp,%rax,2) incq %rax jmp 0x2ae12f movl 0x2c(%rdi), %edx cmpl $0xb, %edx jg 0x2ae169 addl 0x28(%rsp), %edx movslq %edx, %rcx movq %r14, %rax cmpq %rcx, %rax jge 0x2ae1b1 movzwl (%r15,%rax,2), %edx movw %dx, (%rbp,%rax,2) incq %rax jmp 0x2ae155 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addq %rdx, %rdx movq %rsi, 0x38(%rsp) movq %r12, %rsi movl %r8d, %r13d movq %r10, 0x10(%rsp) callq 0x5f3c0 movq 0x38(%rsp), %rsi movq 0x10(%rsp), %r10 movq 0x20(%rsp), %r9 movl %r13d, %r8d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %ecx leal (%rcx,%r14), %eax jmp 0x2ae1b4 movl 0x2c(%rdi), %ecx decl %ecx movslq %ecx, %rcx cltq cmpq %rbx, %rax jge 0x2ae1cf movzwl (%r12,%rcx,2), %edx movw %dx, (%rbp,%rax,2) incq %rax jmp 0x2ae1bb addq %r9, %rbp incl %esi addq %r9, %r10 jmp 0x2ae124 movl 0xc(%rsp), %ecx jmp 0x2ae38a movq 0x50(%rsp), %rax negq %rax movq %rax, 0x10(%rsp) movl 0x30(%rdi), %eax addl 0xc(%rsp), %eax cmpl %eax, %r8d jge 0x2ae2b4 xorl %eax, %eax cmpq %rax, %r14 je 0x2ae218 movzwl (%r12), %ecx movw %cx, (%rbp,%rax,2) incq %rax jmp 0x2ae204 movl 0x2c(%rdi), %edx cmpl $0xb, %edx jg 0x2ae246 addl 0x28(%rsp), %edx movslq %edx, %rax movq 0x10(%rsp), %rcx leaq (%r12,%rcx), %rdx movq %r14, %rcx cmpq %rax, %rcx jge 0x2ae280 movzwl (%rdx,%rcx,2), %esi movw %si, (%rbp,%rcx,2) incq %rcx jmp 0x2ae233 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addq %rdx, %rdx movq %r12, %rsi movl %r8d, %r13d movq %r10, %r15 callq 0x5f3c0 movq %r15, %r10 movq 0x20(%rsp), %r9 movl %r13d, %r8d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r14), %ecx jmp 0x2ae283 movl 0x2c(%rdi), %eax leal -0x1(%rax), %edx movslq %edx, %rdx movslq %ecx, %rcx cmpq %rbx, %rcx jge 0x2ae2a0 movzwl (%r12,%rdx,2), %esi movw %si, (%rbp,%rcx,2) incq %rcx jmp 0x2ae28c cltq leaq (%r12,%rax,2), %r12 addq %r9, %rbp incl %r8d addq %r9, %r10 jmp 0x2ae1f2 movslq 0x2c(%rdi), %rdx movq %r12, %r13 subq %rdx, %r13 subq %rdx, %r13 movq 0x50(%rsp), %rax leaq (%rax,%rdx,2), %rax subq %rax, %r12 cmpl 0x4c(%rsp), %r8d jge 0x2ae37e xorl %eax, %eax cmpq %rax, %r14 je 0x2ae2ee movzwl (%r13), %ecx movw %cx, (%rbp,%rax,2) incq %rax jmp 0x2ae2da cmpl $0xb, %edx jg 0x2ae316 addl 0x28(%rsp), %edx movslq %edx, %rcx movq %r10, %rdx movq %r14, %rax cmpq %rcx, %rax jge 0x2ae353 movzwl (%r12,%rax,2), %esi movw %si, (%rdx) incq %rax addq $0x2, %rdx jmp 0x2ae300 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addl %edx, %edx movq %r13, %rsi movl %r8d, %r15d movq %r10, 0x10(%rsp) callq 0x5f3c0 movq 0x10(%rsp), %r10 movq 0x20(%rsp), %r9 movl %r15d, %r8d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r14), %eax jmp 0x2ae356 movl 0x2c(%rdi), %edx movslq %edx, %rcx cltq cmpq %rbx, %rax jge 0x2ae370 movzwl -0x2(%r13,%rcx,2), %esi movw %si, (%rbp,%rax,2) incq %rax jmp 0x2ae35b addq %r9, %rbp incl %r8d addq %r9, %r10 jmp 0x2ae2cd movq %r13, %r12 movl 0xc(%rsp), %ecx movl 0x30(%rsp), %r8d cmpl $0x2, %r8d jne 0x2ae688 movl 0x2c(%rdi), %edx movl %edx, %eax imull %ecx, %eax cltq leaq (%r12,%rax,2), %r15 xorl %r8d, %r8d movq 0x18(%rsp), %rax testl %eax, %eax movl $0x0, %r12d cmovgl %eax, %r12d testl %ecx, %ecx movl $0x0, %r13d cmovgl %ecx, %r13d leaq (%rbx,%rbx), %r9 leaq (,%r12,2), %r10 addq %rbp, %r10 leaq (%rax,%rax), %r11 movq %r11, %rcx negq %rcx movq %rcx, 0x38(%rsp) leal -0x2(%rax), %r14d movq %r9, 0x10(%rsp) movq %r11, 0x20(%rsp) movq %r14, 0x30(%rsp) movl %r13d, 0x50(%rsp) cmpl %r13d, %r8d je 0x2ae4cb leaq (%r15,%r11), %rax xorl %ecx, %ecx cmpq %rcx, %r12 je 0x2ae41d movzwl (%rax), %esi movw %si, (%rbp,%rcx,2) incq %rcx addq $-0x2, %rax jmp 0x2ae407 cmpl $0xb, %edx jg 0x2ae448 addl 0x28(%rsp), %edx movslq %edx, %rcx movq 0x38(%rsp), %rax leaq (%r15,%rax), %rdx movq %r12, %rax cmpq %rcx, %rax jge 0x2ae490 movzwl (%rdx,%rax,2), %esi movw %si, (%rbp,%rax,2) incq %rax jmp 0x2ae435 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addl %edx, %edx movq %r15, %rsi movq %r10, %r13 movq %r8, %r14 callq 0x5f3c0 movq %r14, %r8 movq 0x20(%rsp), %r11 movq %r13, %r10 movl 0x50(%rsp), %r13d movq 0x30(%rsp), %r14 movq 0x10(%rsp), %r9 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r12), %eax jmp 0x2ae493 movl 0x2c(%rdi), %edx cltq leal (%r14,%rdx,2), %ecx subl %eax, %ecx cmpq %rbx, %rax jge 0x2ae4b4 movslq %ecx, %rcx movzwl (%r15,%rcx,2), %esi movw %si, (%rbp,%rax,2) incq %rax decl %ecx jmp 0x2ae49b addq %r9, %rbp movslq %edx, %rax addq %rax, %rax subq %rax, %r15 incl %r8d addq %r9, %r10 jmp 0x2ae3f8 movl 0xc(%rsp), %r8d movl 0x30(%rdi), %eax addl %r8d, %eax cmpl %eax, %r13d jge 0x2ae5ac leaq (%r15,%r11), %rax xorl %ecx, %ecx cmpq %rcx, %r12 je 0x2ae4fb movzwl (%rax), %esi movw %si, (%rbp,%rcx,2) incq %rcx addq $-0x2, %rax jmp 0x2ae4e5 cmpl $0xb, %edx jg 0x2ae526 addl 0x28(%rsp), %edx movslq %edx, %rcx movq 0x38(%rsp), %rax leaq (%r15,%rax), %rdx movq %r12, %rax cmpq %rcx, %rax jge 0x2ae573 movzwl (%rdx,%rax,2), %esi movw %si, (%rbp,%rax,2) incq %rax jmp 0x2ae513 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addl %edx, %edx movq %r15, %rsi movl %r13d, 0x50(%rsp) movq %r10, %r13 movl %r8d, %r14d callq 0x5f3c0 movl %r14d, %r8d movq 0x20(%rsp), %r11 movq %r13, %r10 movl 0x50(%rsp), %r13d movq 0x30(%rsp), %r14 movq 0x10(%rsp), %r9 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r12), %eax jmp 0x2ae576 movl 0x2c(%rdi), %edx cltq leal (%r14,%rdx,2), %ecx subl %eax, %ecx cmpq %rbx, %rax jge 0x2ae597 movslq %ecx, %rcx movzwl (%r15,%rcx,2), %esi movw %si, (%rbp,%rax,2) incq %rax decl %ecx jmp 0x2ae57e movslq %edx, %rax leaq (%r15,%rax,2), %r15 addq %r9, %rbp incl %r13d addq %r9, %r10 jmp 0x2ae4d0 leal (%rdx,%rdx), %eax cltq addq %rax, %rax subq %rax, %r15 cmpl 0x4c(%rsp), %r13d jge 0x2ae688 leaq (%r15,%r11), %rax xorl %ecx, %ecx cmpq %rcx, %r12 je 0x2ae5de movzwl (%rax), %esi movw %si, (%rbp,%rcx,2) incq %rcx addq $-0x2, %rax jmp 0x2ae5c8 cmpl $0xb, %edx jg 0x2ae610 addl 0x28(%rsp), %edx movslq %edx, %rcx movq 0x38(%rsp), %rax leaq (%r15,%rax), %rdx movq %r10, %rsi movq %r12, %rax cmpq %rcx, %rax jge 0x2ae64d movzwl (%rdx,%rax,2), %r8d movw %r8w, (%rsi) incq %rax addq $0x2, %rsi jmp 0x2ae5f9 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addl %edx, %edx movq %r15, %rsi movq %r10, %r14 callq 0x5f3c0 movq 0x20(%rsp), %r11 movq %r14, %r10 movq 0x30(%rsp), %r14 movq 0x10(%rsp), %r9 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r12), %eax jmp 0x2ae650 movl 0x2c(%rdi), %edx cltq leal (%r14,%rdx,2), %ecx subl %eax, %ecx cmpq %rbx, %rax jge 0x2ae671 movslq %ecx, %rcx movzwl (%r15,%rcx,2), %esi movw %si, (%rbp,%rax,2) incq %rax decl %ecx jmp 0x2ae658 addq %r9, %rbp movslq %edx, %rax addq %rax, %rax subq %rax, %r15 incl %r13d addq %r9, %r10 jmp 0x2ae5b7 addq $0x58, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq	/csukuangfj[P]ncnn/src/layer/padding.cpp
void ncnn::copy_make_border_image<float>(ncnn::Mat const&, ncnn::Mat&, int, int, int, float)	static void copy_make_border_image(const Mat& src, Mat& dst, int top, int left, int type, T v) { int w = dst.w; int h = dst.h; const T* ptr = src; T* outptr = dst; if (type == 0) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = v; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = v; } ptr += src.w; outptr += w; } // fill bottom for (; y < h; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } } if (type == 1) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } ptr += src.w; outptr += w; } // fill bottom ptr -= src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } } if (type == 2) { int y = 0; // fill top ptr += top * src.w; for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } ptr += src.w; outptr += w; } // fill bottom ptr -= 2 * src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } } }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x58, %rsp movslq 0x2c(%rsi), %rbx movl 0x30(%rsi), %eax movl %eax, 0x4c(%rsp) movq (%rdi), %r12 movq (%rsi), %rbp movq %rcx, 0x30(%rsp) movslq %ecx, %rax movq %rax, 0x18(%rsp) testl %r8d, %r8d movq %rdi, 0x40(%rsp) movl %edx, 0xc(%rsp) jne 0x2ae81b movl %r8d, 0x38(%rsp) xorl %eax, %eax testl %ebx, %ebx movl $0x0, %r14d cmovgl %ebx, %r14d testl %edx, %edx movl $0x0, %r9d cmovgl %edx, %r9d leaq (,%rbx,4), %r15 cmpl %r9d, %eax je 0x2ae71a xorl %ecx, %ecx cmpq %rcx, %r14 je 0x2ae713 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2ae703 incl %eax addq %r15, %rbp jmp 0x2ae6fc xorl %r13d, %r13d movq 0x18(%rsp), %rax testl %eax, %eax cmovgl 0x30(%rsp), %r13d leaq (,%rax,4), %rax negq %rax movq %rax, 0x20(%rsp) movss %xmm0, 0x28(%rsp) movl 0x30(%rdi), %eax addl 0xc(%rsp), %eax cmpl %eax, %r9d jge 0x2ae80b xorl %eax, %eax cmpq %rax, %r13 je 0x2ae762 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae752 movslq 0x2c(%rdi), %rdx cmpq $0xb, %rdx jg 0x2ae797 movq 0x30(%rsp), %rax addl %edx, %eax movslq %eax, %rcx movq 0x20(%rsp), %rax leaq (%r12,%rax), %rsi movq %r13, %rax cmpq %rcx, %rax jge 0x2ae7cf movss (%rsi,%rax,4), %xmm1 movss %xmm1, (%rbp,%rax,4) incq %rax jmp 0x2ae782 movq 0x18(%rsp), %rax leaq (,%rax,4), %rdi addq %rbp, %rdi shlq $0x2, %rdx movq %r12, %rsi movl %r9d, 0x10(%rsp) callq 0x5f3c0 movl 0x10(%rsp), %r9d movss 0x28(%rsp), %xmm0 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r13), %eax cltq cmpq %rbx, %rax jge 0x2ae7e1 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae7d1 movslq %edx, %rax leaq (%r12,%rax,4), %r12 addq %r15, %rbp incl %r9d jmp 0x2ae740 xorl %eax, %eax cmpq %rax, %r14 je 0x2ae805 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae7f5 addq %r15, %rbp incl %r9d cmpl 0x4c(%rsp), %r9d jl 0x2ae7f3 movl 0xc(%rsp), %edx movl 0x38(%rsp), %r8d cmpl $0x1, %r8d jne 0x2aeaef movl %r8d, 0x38(%rsp) xorl %r10d, %r10d movq 0x18(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl 0x30(%rsp), %r14d testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi leaq (,%rbx,4), %r8 leaq (%rbp,%r14,4), %r9 leaq (,%rax,4), %rax movq %r12, %r15 movq %rax, 0x50(%rsp) subq %rax, %r15 movq %r8, 0x10(%rsp) movl %esi, 0x28(%rsp) cmpl %esi, %r10d je 0x2ae92f xorl %eax, %eax cmpq %rax, %r14 je 0x2ae894 movss (%r12), %xmm0 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae87e movslq 0x2c(%rdi), %rdx cmpq $0xb, %rdx jg 0x2ae8be addl 0x30(%rsp), %edx movslq %edx, %rcx movq %r14, %rax cmpq %rcx, %rax jge 0x2ae901 movss (%r15,%rax,4), %xmm0 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae8a8 movq 0x18(%rsp), %rax leaq (,%rax,4), %rdi addq %rbp, %rdi shlq $0x2, %rdx movq %r12, %rsi movq %r9, %r13 movq %r10, 0x20(%rsp) callq 0x5f3c0 movq 0x20(%rsp), %r10 movq %r13, %r9 movq 0x10(%rsp), %r8 movl 0x28(%rsp), %esi movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %ecx leal (%rcx,%r14), %eax jmp 0x2ae904 movl 0x2c(%rdi), %ecx decl %ecx movslq %ecx, %rcx cltq cmpq %rbx, %rax jge 0x2ae921 movss (%r12,%rcx,4), %xmm0 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae90b addq %r8, %rbp incl %r10d addq %r8, %r9 jmp 0x2ae873 movq 0x50(%rsp), %rax negq %rax movq %rax, 0x20(%rsp) movl 0xc(%rsp), %r10d movl 0x30(%rdi), %eax addl %r10d, %eax cmpl %eax, %esi jge 0x2aea0e xorl %eax, %eax cmpq %rax, %r14 je 0x2ae967 movss (%r12), %xmm0 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae951 movslq 0x2c(%rdi), %rdx cmpq $0xb, %rdx jg 0x2ae999 addl 0x30(%rsp), %edx movslq %edx, %rax movq 0x20(%rsp), %rcx leaq (%r12,%rcx), %rdx movq %r14, %rcx cmpq %rax, %rcx jge 0x2ae9d9 movss (%rdx,%rcx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2ae984 movq 0x18(%rsp), %rax leaq (,%rax,4), %rdi addq %rbp, %rdi shlq $0x2, %rdx movl %esi, %r13d movq %r12, %rsi movq %r9, %r15 callq 0x5f3c0 movq %r15, %r9 movq 0x10(%rsp), %r8 movl %r13d, %esi movl 0xc(%rsp), %r10d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r14), %ecx jmp 0x2ae9dc movl 0x2c(%rdi), %eax leal -0x1(%rax), %edx movslq %edx, %rdx movslq %ecx, %rcx cmpq %rbx, %rcx jge 0x2ae9fb movss (%r12,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2ae9e5 cltq leaq (%r12,%rax,4), %r12 addq %r8, %rbp incl %esi addq %r8, %r9 jmp 0x2ae941 movslq 0x2c(%rdi), %rax leaq (,%rax,4), %rcx movq %r12, %r13 subq %rcx, %r13 movq 0x50(%rsp), %rcx leaq (%rcx,%rax,4), %rcx subq %rcx, %r12 cmpl 0x4c(%rsp), %esi jge 0x2aeae3 xorl %ecx, %ecx cmpq %rcx, %r14 je 0x2aea4e movss (%r13), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2aea38 cmpl $0xb, %eax jg 0x2aea77 addl 0x30(%rsp), %eax cltq movq %r9, %rdx movq %r14, %rcx cmpq %rax, %rcx jge 0x2aeab6 movss (%r12,%rcx,4), %xmm0 movss %xmm0, (%rdx) incq %rcx addq $0x4, %rdx jmp 0x2aea5f movq 0x18(%rsp), %rcx leaq (,%rcx,4), %rdi addq %rbp, %rdi movl %eax, %edx shlq $0x2, %rdx movl %esi, 0x28(%rsp) movq %r13, %rsi movq %r9, %r15 callq 0x5f3c0 movq %r15, %r9 movq 0x10(%rsp), %r8 movl 0x28(%rsp), %esi movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r14), %ecx jmp 0x2aeab9 movl 0x2c(%rdi), %eax movslq %eax, %rdx movslq %ecx, %rcx cmpq %rbx, %rcx jge 0x2aead6 movss -0x4(%r13,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2aeabf addq %r8, %rbp incl %esi addq %r8, %r9 jmp 0x2aea2c movq %r13, %r12 movl 0x38(%rsp), %r8d movl 0xc(%rsp), %edx cmpl $0x2, %r8d jne 0x2aedff movl 0x2c(%rdi), %eax movl %eax, %ecx imull %edx, %ecx movslq %ecx, %rcx leaq (%r12,%rcx,4), %r15 xorl %esi, %esi movq 0x18(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r12d cmovgl %ecx, %r12d testl %edx, %edx movl $0x0, %r9d cmovgl %edx, %r9d leaq (,%rbx,4), %rdx movq %rdx, 0x10(%rsp) leaq (,%r12,4), %r13 addq %rbp, %r13 leaq (,%rcx,4), %r10 movq %r10, %rdx negq %rdx movq %rdx, 0x20(%rsp) leal -0x2(%rcx), %r11d movq %r10, 0x28(%rsp) movq %r11, 0x38(%rsp) cmpl %r9d, %esi je 0x2aec3c leaq (%r15,%r10), %rcx xorl %edx, %edx cmpq %rdx, %r12 je 0x2aeb87 movss (%rcx), %xmm0 movss %xmm0, (%rbp,%rdx,4) incq %rdx addq $-0x4, %rcx jmp 0x2aeb6f cmpl $0xb, %eax jg 0x2aebb3 addl 0x30(%rsp), %eax cltq movq 0x20(%rsp), %rcx leaq (%r15,%rcx), %rdx movq %r12, %rcx cmpq %rax, %rcx jge 0x2aebf9 movss (%rdx,%rcx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2aeb9e movq 0x18(%rsp), %rcx leaq (,%rcx,4), %rdi addq %rbp, %rdi movl %eax, %edx shlq $0x2, %rdx movq %rsi, 0x50(%rsp) movq %r15, %rsi movl %r9d, %r14d callq 0x5f3c0 movq 0x50(%rsp), %rsi movq 0x38(%rsp), %r11 movq 0x28(%rsp), %r10 movl %r14d, %r9d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r12), %ecx jmp 0x2aebfc movl 0x2c(%rdi), %eax movslq %ecx, %rcx leal (%r11,%rax,2), %edx subl %ecx, %edx cmpq %rbx, %rcx jge 0x2aec20 movslq %edx, %rdx movss (%r15,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx decl %edx jmp 0x2aec05 movq 0x10(%rsp), %rdx addq %rdx, %rbp movslq %eax, %rcx shlq $0x2, %rcx subq %rcx, %r15 incl %esi addq %rdx, %r13 jmp 0x2aeb60 movq 0x10(%rsp), %r14 movl 0x30(%rdi), %ecx addl 0xc(%rsp), %ecx cmpl %ecx, %r9d jge 0x2aed18 leaq (%r15,%r10), %rcx xorl %edx, %edx cmpq %rdx, %r12 je 0x2aec6f movss (%rcx), %xmm0 movss %xmm0, (%rbp,%rdx,4) incq %rdx addq $-0x4, %rcx jmp 0x2aec57 cmpl $0xb, %eax jg 0x2aec9b addl 0x30(%rsp), %eax cltq movq 0x20(%rsp), %rcx leaq (%r15,%rcx), %rdx movq %r12, %rcx cmpq %rax, %rcx jge 0x2aecdc movss (%rdx,%rcx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2aec86 movq 0x18(%rsp), %rcx leaq (,%rcx,4), %rdi addq %rbp, %rdi movl %eax, %edx shlq $0x2, %rdx movq %r15, %rsi movl %r9d, %r14d callq 0x5f3c0 movq 0x38(%rsp), %r11 movq 0x28(%rsp), %r10 movl %r14d, %r9d movq 0x10(%rsp), %r14 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r12), %ecx jmp 0x2aecdf movl 0x2c(%rdi), %eax movslq %ecx, %rcx leal (%r11,%rax,2), %edx subl %ecx, %edx cmpq %rbx, %rcx jge 0x2aed03 movslq %edx, %rdx movss (%r15,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx decl %edx jmp 0x2aece8 movslq %eax, %rcx leaq (%r15,%rcx,4), %r15 addq %r14, %rbp incl %r9d addq %r14, %r13 jmp 0x2aec41 leal (%rax,%rax), %ecx movslq %ecx, %rcx shlq $0x2, %rcx subq %rcx, %r15 cmpl 0x4c(%rsp), %r9d jge 0x2aedff leaq (%r15,%r10), %rcx xorl %edx, %edx cmpq %rdx, %r12 je 0x2aed4e movss (%rcx), %xmm0 movss %xmm0, (%rbp,%rdx,4) incq %rdx addq $-0x4, %rcx jmp 0x2aed36 cmpl $0xb, %eax jg 0x2aed7f addl 0x30(%rsp), %eax cltq movq 0x20(%rsp), %rcx leaq (%r15,%rcx), %rdx movq %r13, %rsi movq %r12, %rcx cmpq %rax, %rcx jge 0x2aedc0 movss (%rdx,%rcx,4), %xmm0 movss %xmm0, (%rsi) incq %rcx addq $0x4, %rsi jmp 0x2aed68 movq 0x18(%rsp), %rcx leaq (,%rcx,4), %rdi addq %rbp, %rdi movl %eax, %edx shlq $0x2, %rdx movq %r15, %rsi movl %r9d, %r14d callq 0x5f3c0 movq 0x38(%rsp), %r11 movq 0x28(%rsp), %r10 movl %r14d, %r9d movq 0x10(%rsp), %r14 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r12), %ecx jmp 0x2aedc3 movl 0x2c(%rdi), %eax movslq %ecx, %rcx leal (%r11,%rax,2), %edx subl %ecx, %edx cmpq %rbx, %rcx jge 0x2aede7 movslq %edx, %rdx movss (%r15,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx decl %edx jmp 0x2aedcc addq %r14, %rbp movslq %eax, %rcx shlq $0x2, %rcx subq %rcx, %r15 incl %r9d addq %r14, %r13 jmp 0x2aed25 addq $0x58, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq	/csukuangfj[P]ncnn/src/layer/padding.cpp
ncnn::Padding_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0xe8, %rsp movq %rcx, %rbx movq %rdx, %r15 movq %rsi, %rcx movq (%rdi), %rax movq -0x18(%rax), %rbp movq %rdi, (%rsp) movl 0xd0(%rdi,%rbp), %r10d testl %r10d, %r10d movq %rsi, 0x8(%rsp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xd4(%rax,%rbp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xd8(%rax,%rbp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xdc(%rax,%rbp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xe8(%rax,%rbp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xec(%rax,%rbp) je 0x2afc57 movl 0x18(%rcx), %r11d movq 0x10(%rcx), %r9 testl %r11d, %r11d je 0x2aef32 leal (,%r9,8), %eax cltd idivl %r11d cmpl $0x8, %eax jne 0x2aef32 movq (%rsp), %rdi movq %rcx, %rsi movq %r15, %rdx movq %rbx, %rcx addq $0xe8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x2afd44 movl 0x28(%rcx), %r13d movl 0x2c(%rcx), %esi movl 0x30(%rcx), %r14d movl 0x34(%rcx), %eax movl %eax, 0x18(%rsp) movl 0x38(%rcx), %eax movq %rax, 0x10(%rsp) cmpl $0x4, %r11d jne 0x2afab9 leal -0x1(%r13), %eax cmpl $0x3, %eax ja 0x2afab9 leaq 0x1497d6(%rip), %rcx # 0x3f8740 movslq (%rcx,%rax,4), %rax addq %rcx, %rax pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) jmpq %rax movq (%rsp), %rcx movl 0xd8(%rcx,%rbp), %r8d leal (%r8,%rsi,4), %eax addl 0xdc(%rcx,%rbp), %eax testb $0x3, %al sete %r10b movq %r9, %rdx shrq $0x2, %rdx leal (%r10,%r10), %ecx shlq %cl, %rdx testb $0x3, %r8b sete %cl andb %r10b, %cl cmpb $0x1, %cl movq 0x8(%rsp), %rcx jne 0x2afab9 movq (%rsp), %rdi cmpl $0x0, 0xe0(%rdi,%rbp) jne 0x2afab9 sarl $0x2, %eax movq 0x8(%rbx), %r8 pushq $0x4 popq %rbp movq %r15, %rdi movl %eax, %esi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%r15) je 0x2afba1 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2afba1 movq (%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx movss 0xe4(%rdx,%rcx), %xmm0 shufps $0x0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq $0x0, 0x20(%rsp) movq 0x8(%rsp), %rdi movq %r15, %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b0c1b jmp 0x2afba1 movq (%rsp), %rdx movl 0xe8(%rdx,%rbp), %eax movq 0x10(%rsp), %rcx leal (%rax,%rcx,4), %r12d addl 0xec(%rdx,%rbp), %r12d testb $0x3, %r12b sete %dl movq %r9, %r8 shrq $0x2, %r8 leal (%rdx,%rdx), %ecx shlq %cl, %r8 testb $0x3, %al sete %al andb %dl, %al cmpb $0x1, %al movq 0x8(%rsp), %rcx jne 0x2afab9 movq (%rsp), %rcx movl 0xd8(%rcx,%rbp), %eax addl %esi, %eax addl 0xdc(%rcx,%rbp), %eax addl %r14d, %r10d addl 0xd4(%rcx,%rbp), %r10d movq 0x10(%rsp), %rcx leal (,%rcx,4), %ecx cmpl %ecx, %r12d movq 0x8(%rsp), %rcx je 0x2af0e0 movq (%rsp), %rdx cmpl $0x0, 0xe0(%rdx,%rbp) jne 0x2afab9 sarl $0x2, %r12d subq $0x8, %rsp pushq $0x4 popq %r9 movq %r15, %rdi movl %eax, %esi movl %r10d, %edx movl %r12d, %ecx pushq 0x8(%rbx) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%r15) je 0x2afba1 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2afba1 movq (%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xe8(%rcx,%rax), %eax pushq $-0x4 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r12d, %r12d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r12d movl %edx, 0xbc(%rsp) movl %edx, %r14d xorl %r9d, %r9d movq %r12, 0xe0(%rsp) movq (%rsp), %rbx cmpq %r12, %r9 je 0x2afba1 movslq 0x2c(%r15), %rax movslq 0x30(%r15), %rcx movq 0x40(%r15), %r13 imulq %r9, %r13 movq 0x10(%r15), %rdi imulq %rdi, %r13 addq (%r15), %r13 movl 0x34(%r15), %esi movl 0x18(%r15), %edx movq 0x20(%r15), %r8 movq %r13, 0x28(%rsp) andq $0x0, 0x30(%rsp) movq %rdi, 0x38(%rsp) movl %edx, 0x40(%rsp) movq %r8, 0x48(%rsp) movl %eax, 0x54(%rsp) movl %ecx, 0x58(%rsp) movl $0x1, 0x5c(%rsp) movl %esi, 0x60(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0x68(%rsp) movl 0x28(%r15), %edx leal -0x1(%rdx), %edi movl %edi, 0x50(%rsp) cmpl $0x4, %edx jne 0x2af1f7 movq %rcx, 0x68(%rsp) movq %rcx, %rax movq (%rsp), %rdx movq (%rdx), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%rdx,%rdi) je 0x2af226 movq (%rsp), %rdx movq 0xf8(%rdx,%rdi), %rdx movq %r9, %r8 shlq $0x4, %r8 movups (%rdx,%r8), %xmm0 jmp 0x2af237 movq (%rsp), %rdx movss 0xe4(%rdx,%rdi), %xmm0 shufps $0x0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] movl %r14d, %r14d movl 0xbc(%rsp), %edx movq %r9, 0x18(%rsp) addl %r9d, %edx setns %r8b cmpl 0x10(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2af417 movq %r14, 0xc0(%rsp) movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r14 movq 0x40(%r10), %r11 movl %edx, %ebp imulq %r11, %rbp imulq %r14, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %r14, 0x80(%rsp) movl %edx, 0x88(%rsp) movq %r9, 0x90(%rsp) movl %eax, 0x9c(%rsp) movl %esi, 0xa0(%rsp) movl $0x1, 0xa4(%rsp) movl %r8d, 0xa8(%rsp) imulq %rax, %rsi movq %r14, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r14 movq %rax, 0xb0(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x98(%rsp) cmpl $0x4, %eax jne 0x2af315 movq %rsi, 0xb0(%rsp) movq (%rsp), %rax movl 0xe0(%rax,%rdi), %eax testl %eax, %eax jne 0x2af36f movl 0xd0(%rbx,%rdi), %edx movl 0xd4(%rbx,%rdi), %ecx movl 0xd8(%rbx,%rdi), %r8d movl 0xdc(%rbx,%rdi), %r9d leaq 0x70(%rsp), %rdi leaq 0x28(%rsp), %rsi movq %r11, 0xd8(%rsp) callq 0x2b0c1b movq 0xd8(%rsp), %r11 movq (%rbx), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%rbx,%rdi), %eax cmpl $0x1, %eax jne 0x2af521 movq (%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r14, %r11 imulq 0xc0(%rsp), %r11 addq %r11, %r12 cmpl %r9d, %r8d je 0x2af43d movaps (%rbp), %xmm0 movl %eax, %edi subl $0x1, %edi jb 0x2af3da movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af3ca xorl %edi, %edi xorl %r10d, %r10d cmpl 0x9c(%rsp), %r10d jge 0x2af3fd movaps (%r12,%rdi), %xmm0 movaps %xmm0, (%r13,%rdi) incl %r10d addq $0x10, %rdi jmp 0x2af3df addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2af412 movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af402 incl %r8d jmp 0x2af3bf imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0x18(%rsp), %r9 subl $0x1, %esi jb 0x2af7be movups %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af429 xorl %edi, %edi cmpl 0xa0(%rsp), %edi jge 0x2af49b movaps (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2af460 movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af44f xorl %r8d, %r8d cmpl 0x9c(%rsp), %r8d jge 0x2af483 movaps (%rbp), %xmm0 movaps %xmm0, (%r13) addq $0x10, %rbp addq $0x10, %r13 incl %r8d jmp 0x2af463 movl %edx, %r8d subl $0x1, %r8d jb 0x2af497 movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af486 incl %edi jmp 0x2af43f movl 0x9c(%rsp), %edi shll $0x2, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2af512 movaps (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2af4d2 movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af4c1 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x9c(%rsp), %r9d jge 0x2af4f7 movaps (%rbp,%r8), %xmm0 movaps %xmm0, (%r13,%r8) incl %r9d addq $0x10, %r8 jmp 0x2af4d8 addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2af50e movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af4fd incl %edi jmp 0x2af4b6 movq -0x18(%rcx), %rdi movq (%rsp), %rax movl 0xe0(%rax,%rdi), %eax cmpl $0x2, %eax pushq $-0x20 popq %r13 jne 0x2af748 movq (%rsp), %r8 movl 0xd0(%r8,%rdi), %r9d movslq 0xd8(%r8,%rdi), %rax movl 0x9c(%rsp), %ecx imull %r9d, %ecx shll $0x2, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x70(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r8,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r8,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x28(%rsp), %r8 shlq $0x4, %rax shlq $0x4, %rdx shlq $0x4, %rsi cmpl %r9d, %r10d je 0x2af626 leaq (%rcx,%rax), %r11 xorl %r12d, %r12d xorl %r14d, %r14d movq %rdx, %r13 addq %r14, %r13 je 0x2af5c6 movaps (%r11,%r14), %xmm0 movaps %xmm0, (%r8,%r12) addq $-0x10, %r14 addq $0x10, %r12 jmp 0x2af5aa subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 pushq $-0x20 popq %r13 cmpl 0x9c(%rsp), %ebp jge 0x2af5ef movaps (%r11), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r11 addq $0x10, %r8 incl %ebp jmp 0x2af5d2 movq %r13, %r14 leaq (%rsi,%r14), %r12 cmpq $-0x20, %r12 je 0x2af60f movaps (%r11,%r14), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r14 jmp 0x2af5f2 movslq 0x9c(%rsp), %r11 shlq $0x4, %r11 subq %r11, %rcx incl %r10d jmp 0x2af597 xorl %r9d, %r9d cmpl 0xa0(%rsp), %r9d jge 0x2af6a2 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r12 addq %r11, %r12 je 0x2af659 movaps (%r10,%r11), %xmm0 movaps %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2af63d subq %r11, %r8 xorl %r10d, %r10d cmpl 0x9c(%rsp), %r10d jge 0x2af67d movaps (%rcx), %xmm0 movaps %xmm0, (%r8) addq $0x10, %rcx addq $0x10, %r8 incl %r10d jmp 0x2af65f movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x20, %r11 je 0x2af69d movaps (%rcx,%r10), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r10 jmp 0x2af680 incl %r9d jmp 0x2af629 movslq 0x9c(%rsp), %r9 shlq $0x5, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2af748 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r12 addq %r11, %r12 je 0x2af6e9 movaps (%r10,%r11), %xmm0 movaps %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2af6cd subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x9c(%rsp), %r11d jge 0x2af711 movaps (%r10), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r10 addq $0x10, %r8 incl %r11d jmp 0x2af6f2 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x20, %r14 je 0x2af731 movaps (%r10,%r11), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r11 jmp 0x2af714 movslq 0x9c(%rsp), %r10 shlq $0x4, %r10 subq %r10, %rcx incl %r9d jmp 0x2af6ba movq 0x78(%rsp), %rax testq %rax, %rax movq 0xc0(%rsp), %r14 je 0x2af781 lock decl (%rax) jne 0x2af781 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2af779 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2af781 movq %rsi, %rdi callq 0x5f3e0 movq 0x30(%rsp), %rax testq %rax, %rax movq 0x18(%rsp), %r9 je 0x2af7be lock decl (%rax) jne 0x2af7be movq 0x28(%rsp), %rsi movq 0x48(%rsp), %rdi testq %rdi, %rdi je 0x2af7b1 movq (%rdi), %rax callq 0x18(%rax) movq 0x18(%rsp), %r9 jmp 0x2af7be movq %rsi, %rdi callq 0x5f3e0 movq 0x18(%rsp), %r9 incq %r9 incl %r14d movq 0xe0(%rsp), %r12 jmp 0x2af165 movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%rbp) movq 0x8(%rsp), %rcx jne 0x2afab9 movq (%rsp), %rax movl 0xe8(%rax,%rbp), %r12d addl 0x18(%rsp), %r12d addl 0xec(%rax,%rbp), %r12d addl %r14d, %r10d addl 0xd4(%rax,%rbp), %r10d addl 0xd8(%rax,%rbp), %esi addl 0xdc(%rax,%rbp), %esi movq %r15, %rdi movl %r10d, %edx movl %r12d, %ecx movq 0x10(%rsp), %r8 pushq 0x8(%rbx) pushq $0x4 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%r15) je 0x2afba1 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2afba1 xorl %ecx, %ecx testl %r12d, %r12d cmovlel %ecx, %r12d movq 0x10(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x10(%rsp) movabsq $0x100000001, %r13 # imm = 0x100000001 xorl %ebx, %ebx cmpq 0x10(%rsp), %rbx je 0x2afba1 movq (%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rcx,%rax) je 0x2af8b9 movq (%rsp), %rcx movq 0xf8(%rcx,%rax), %rax movq %rbx, %rcx shlq $0x4, %rcx movups (%rax,%rcx), %xmm0 jmp 0x2af8ca movq (%rsp), %rcx movss 0xe4(%rcx,%rax), %xmm0 shufps $0x0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] xorl %ebp, %ebp movaps %xmm0, 0xc0(%rsp) cmpq %r12, %rbp je 0x2afa67 movslq 0x2c(%r15), %rdx movslq 0x30(%r15), %rsi movq 0x40(%r15), %rdi imulq %rbx, %rdi movq 0x10(%r15), %r8 imulq %r8, %rdi addq (%r15), %rdi movl 0x18(%r15), %r9d movq 0x20(%r15), %r10 movq %rsi, %rax imulq %rdx, %rax movq %rbp, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x28(%rsp) andq $0x0, 0x30(%rsp) movq %r8, 0x38(%rsp) movl %r9d, 0x40(%rsp) movq %r10, 0x48(%rsp) movl $0x2, 0x50(%rsp) movl %edx, 0x54(%rsp) movl %esi, 0x58(%rsp) movq %r13, 0x5c(%rsp) movq %rax, 0x68(%rsp) movq (%rsp), %rdi movq (%rdi), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rdi,%rsi), %edi movl %ebp, %edx subl %edi, %edx setns %dil cmpl 0x18(%rsp), %edx setl %r8b testb %r8b, %dil je 0x2afa4f movq 0x8(%rsp), %rcx movslq 0x2c(%rcx), %rax movq %r15, %r14 movslq 0x30(%rcx), %r15 movq 0x40(%rcx), %rdi imulq %rbx, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %r15, %rcx imulq %rax, %rcx movl %edx, %edx movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %r8, 0x80(%rsp) movl %r9d, 0x88(%rsp) movq %r10, 0x90(%rsp) movl $0x2, 0x98(%rsp) movl %eax, 0x9c(%rsp) movl %r15d, 0xa0(%rsp) movq (%rsp), %rax movq %r14, %r15 movq %r13, 0xa4(%rsp) movq %rcx, 0xb0(%rsp) movl 0xd0(%rax,%rsi), %edx movl 0xd4(%rax,%rsi), %ecx movl 0xd8(%rax,%rsi), %r8d movl 0xdc(%rax,%rsi), %r9d leaq 0x70(%rsp), %rdi leaq 0x28(%rsp), %rsi movaps 0xc0(%rsp), %xmm0 callq 0x2b0c1b movaps 0xc0(%rsp), %xmm0 incq %rbp jmp 0x2af8d4 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2afa47 movups %xmm0, (%rcx) addq $0x10, %rcx jmp 0x2afa59 incq %rbx jmp 0x2af880 leal (%r10,%r14,4), %edx movq (%rsp), %rax addl 0xd4(%rax,%rbp), %edx testb $0x3, %dl sete %r8b movq %r9, %rax shrq $0x2, %rax leal (%r8,%r8), %ecx shlq %cl, %rax testb $0x3, %r10b sete %cl andb %r8b, %cl cmpb $0x1, %cl movq 0x8(%rsp), %rcx jne 0x2afab9 movq (%rsp), %rdi cmpl $0x0, 0xe0(%rdi,%rbp) je 0x2afbb8 movq (%rcx), %rax movq %rax, 0x28(%rsp) movq 0x8(%rcx), %rax movq %rax, 0x30(%rsp) movq %r9, 0x38(%rsp) movl %r11d, 0x40(%rsp) movq %rcx, %rdx movq 0x20(%rcx), %rcx movq %rcx, 0x48(%rsp) movl %r13d, 0x50(%rsp) movl %esi, 0x54(%rsp) movl %r14d, 0x58(%rsp) movl 0x18(%rsp), %ecx movl %ecx, 0x5c(%rsp) movq 0x10(%rsp), %rcx movl %ecx, 0x60(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x68(%rsp) testq %rax, %rax je 0x2afb10 lock incl (%rax) cmpl $0x1, %r11d je 0x2afb53 movups (%rbx), %xmm0 movups 0x10(%rbx), %xmm1 movups 0x20(%rbx), %xmm2 movups 0x30(%rbx), %xmm3 leaq 0x70(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x10(%rbx), %rax movq %rax, 0x8(%rcx) leaq 0x28(%rsp), %rsi pushq $0x1 popq %rdx movq 0x8(%rsp), %rdi callq 0x64e3b movq (%rsp), %rdi movq (%rdi), %rax addq -0x18(%rax), %rdi leaq 0x28(%rsp), %rsi movq %r15, %rdx movq %rbx, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x30(%rsp), %rax testq %rax, %rax je 0x2afba1 lock decl (%rax) jne 0x2afba1 movq 0x28(%rsp), %rsi movq 0x48(%rsp), %rdi testq %rdi, %rdi je 0x2afb99 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2afba1 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0xe8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq movq (%rsp), %rcx addl 0xd8(%rcx,%rbp), %esi addl 0xdc(%rcx,%rbp), %esi sarl $0x2, %edx movq 0x8(%rbx), %r9 pushq $0x4 popq %rbp movq %r15, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%r15) je 0x2afba1 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2afba1 movq (%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi movss 0xe4(%rdi,%rsi), %xmm0 shufps $0x0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq 0x8(%rsp), %rdi movq %r15, %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b0c1b movq $0x0, 0x20(%rsp) jmp 0x2afba1 movq $0x0, 0x20(%rsp) cmpq %rcx, %r15 je 0x2afba1 movq 0x8(%rcx), %rax testq %rax, %rax je 0x2afc75 lock incl (%rax) movq 0x8(%r15), %rax testq %rax, %rax je 0x2afca4 lock decl (%rax) jne 0x2afca4 movq (%r15), %rsi movq 0x20(%r15), %rdi testq %rdi, %rdi je 0x2afc97 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2afc9f movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx andq $0x0, 0x40(%r15) xorps %xmm0, %xmm0 movups %xmm0, (%r15) movups %xmm0, 0xc(%r15) andl $0x0, 0x38(%r15) movups %xmm0, 0x28(%r15) movups (%rcx), %xmm0 movups %xmm0, (%r15) movq 0x10(%rcx), %rax movq %rax, 0x10(%r15) movl 0x18(%rcx), %eax movl %eax, 0x18(%r15) movq 0x20(%rcx), %rax movq %rax, 0x20(%r15) movups 0x28(%rcx), %xmm0 movups %xmm0, 0x28(%r15) movl 0x38(%rcx), %eax movl %eax, 0x38(%r15) movq 0x40(%rcx), %rax movq %rax, 0x40(%r15) jmp 0x2afba1 jmp 0x2afd3b jmp 0x2afd3b jmp 0x2afd3b jmp 0x2afd02 movq %rax, %rbx movq 0x30(%rsp), %rax testq %rax, %rax je 0x2afd33 lock decl (%rax) jne 0x2afd33 movq 0x28(%rsp), %rsi movq 0x48(%rsp), %rdi testq %rdi, %rdi jne 0x2afd2d movq %rsi, %rdi callq 0x5f3e0 jmp 0x2afd33 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/x86/padding_x86.cpp
ncnn::Padding_x86::forward_int8(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_int8_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_int8_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x118, %rsp # imm = 0x118 movq %rcx, %r12 movq %rdx, %rbx movq %rsi, %rax movq %rdi, %r13 movl 0x2c(%rsi), %esi movl 0x30(%rax), %edx movl 0x34(%rax), %ecx movl %ecx, 0x18(%rsp) movl 0x38(%rax), %ebp movq 0x10(%rax), %r9 movl 0x18(%rax), %edi movq %rax, 0x28(%rsp) movl 0x28(%rax), %r11d cmpl $0x8, %edi jne 0x2b09db leal -0x1(%r11), %eax cmpl $0x3, %eax ja 0x2b09db leaq 0x1489b2(%rip), %rcx # 0x3f8750 movslq (%rcx,%rax,4), %rax addq %rcx, %rax pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) movq %r13, 0x10(%rsp) jmpq %rax movq (%r13), %rax movq -0x18(%rax), %r8 movl 0xd8(%r13,%r8), %r14d leal (%r14,%rsi,8), %eax addl 0xdc(%r13,%r8), %eax xorl %r15d, %r15d testb $0x7, %al sete %r15b movq %r9, %r10 shrq $0x3, %r10 leal (%r15,%r15,2), %ecx shlq %cl, %r10 testb $0x7, %r14b sete %cl andb %cl, %r15b cmpb $0x1, %r15b jne 0x2b09db cmpl $0x0, 0xe0(%r13,%r8) jne 0x2b09db sarl $0x3, %eax movq 0x8(%r12), %r8 pushq $0x8 popq %rbp movq %rbx, %rdi movl %eax, %esi movq %r10, %rdx movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b0ad5 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b0ad5 movq 0x10(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rcx cvttss2si 0xe4(%r8,%rcx), %rax movq %rax, %rdx shlq $0x8, %rdx movq %rax, %rsi shlq $0x10, %rsi orq %rdx, %rsi movq %rax, %rdx shlq $0x18, %rdx movq %rax, %rdi shlq $0x20, %rdi orq %rdx, %rdi orq %rsi, %rdi movq %rax, %rdx shlq $0x28, %rdx movq %rax, %rsi shlq $0x30, %rsi orq %rdx, %rsi movq %rax, %r9 shlq $0x38, %r9 orq %rsi, %r9 orq %rdi, %r9 orq %rax, %r9 movl 0xd8(%r8,%rcx), %eax movl 0xdc(%r8,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq %r9, (%rsp) movq $0x0, 0x20(%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b0cbe jmp 0x2b0ad5 movq (%r13), %rax movq -0x18(%rax), %r10 movl 0xe8(%r13,%r10), %eax leal (%rax,%rbp,8), %r15d addl 0xec(%r13,%r10), %r15d xorl %r14d, %r14d testb $0x7, %r15b sete %r14b movq %r9, %r8 shrq $0x3, %r8 leal (%r14,%r14,2), %ecx shlq %cl, %r8 testb $0x7, %al sete %al andb %al, %r14b cmpb $0x1, %r14b jne 0x2b09db movl 0xd8(%r13,%r10), %eax addl %esi, %eax addl 0xdc(%r13,%r10), %eax movl 0xd0(%r13,%r10), %ecx addl %edx, %ecx addl 0xd4(%r13,%r10), %ecx movq %rbp, 0x30(%rsp) movq 0x30(%rsp), %r14 leal (,%r14,8), %ebp cmpl %ebp, %r15d movq 0x30(%rsp), %rbp je 0x2aff73 cmpl $0x0, 0xe0(%r13,%r10) jne 0x2b09db sarl $0x3, %r15d movq 0x8(%r12), %rdx movq %rdx, (%rsp) pushq $0x8 popq %r9 movq %rbx, %rdi movl %eax, %esi movl %ecx, %edx movl %r15d, %ecx callq 0x628f2 cmpq $0x0, (%rbx) je 0x2b0ad5 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b0ad5 movq %r15, %r13 movq 0x10(%rsp), %r12 movq (%r12), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r13d, %r13d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r13d movl %edx, 0xa4(%rsp) movl %edx, %r10d xorl %ecx, %ecx movq %r13, 0x110(%rsp) cmpq %r13, %rcx je 0x2b0ad5 movslq 0x2c(%rbx), %rax movslq 0x30(%rbx), %r9 movl 0x34(%rbx), %esi movq (%rbx), %rdi movq 0x10(%rbx), %r11 movq 0x40(%rbx), %r8 movq %r8, %rbp movq %rcx, 0x48(%rsp) imulq %rcx, %rbp imulq %r11, %rbp addq %rdi, %rbp movl 0x18(%rbx), %ecx movq 0x20(%rbx), %rdx movq %rbp, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %r11, 0x60(%rsp) movl %ecx, 0x68(%rsp) movq %rdx, 0x70(%rsp) movl %eax, 0x7c(%rsp) movl %r9d, 0x80(%rsp) movl $0x1, 0x84(%rsp) movl %esi, 0x88(%rsp) imulq %rax, %r9 movq %r11, %rax imulq %r9, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r11 movq %rax, 0x90(%rsp) movl 0x28(%rbx), %ecx leal -0x1(%rcx), %edx movl %edx, 0x78(%rsp) cmpl $0x4, %ecx jne 0x2b00a8 movq %r9, 0x90(%rsp) movq %r9, %rax movq 0x98(%rsp), %rcx movq -0x18(%rcx), %r9 cvttss2si 0xe4(%r12,%r9), %rcx movl %r10d, %r10d movq %r10, 0x18(%rsp) movq %rcx, %rdx shlq $0x8, %rdx movq %rcx, %r10 shlq $0x10, %r10 orq %rdx, %r10 movq %rcx, %rdx shlq $0x18, %rdx movq %rcx, %r14 shlq $0x20, %r14 orq %rdx, %r14 orq %r10, %r14 movq %rcx, %rdx shlq $0x28, %rdx movq %rcx, %r15 shlq $0x30, %r15 orq %rdx, %r15 movq %rcx, %r10 shlq $0x38, %r10 orq %r15, %r10 orq %r14, %r10 orq %rcx, %r10 movl 0xa4(%rsp), %ecx movq 0x48(%rsp), %r15 addl %r15d, %ecx setns %dl cmpl 0x30(%rsp), %ecx setl %r14b testb %r14b, %dl je 0x2b031e movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rdi movslq 0x30(%rsi), %r8 movl 0x34(%rsi), %eax movq (%rsi), %r13 movq 0x10(%rsi), %r11 movq 0x40(%rsi), %rdx movl %ecx, %r12d movq %rdx, 0x38(%rsp) imulq %rdx, %r12 imulq %r11, %r12 addq %r13, %r12 movl 0x18(%rsi), %ecx movq 0x20(%rsi), %rdx movq %r12, 0xc0(%rsp) andq $0x0, 0xc8(%rsp) movq %r11, 0xd0(%rsp) movl %ecx, 0xd8(%rsp) movq %rdx, 0xe0(%rsp) movl %edi, 0xec(%rsp) movl %r8d, 0xf0(%rsp) movl $0x1, 0xf4(%rsp) movl %eax, 0xf8(%rsp) movq %r8, 0xb0(%rsp) movq %r8, %rcx movq %rdi, 0x40(%rsp) imulq %rdi, %rcx movq %r11, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx movq %r11, 0xa8(%rsp) divq %r11 movq %rax, 0x100(%rsp) movl 0x28(%rsi), %eax leal -0x1(%rax), %edx movl %edx, 0xe8(%rsp) cmpl $0x4, %eax jne 0x2b01ff movq %rcx, 0x100(%rsp) movq 0x10(%rsp), %r14 movl 0xe0(%r14,%r9), %eax testl %eax, %eax jne 0x2b025d movl 0xd0(%r14,%r9), %edx movl 0xd4(%r14,%r9), %ecx movl 0xd8(%r14,%r9), %r8d movl 0xdc(%r14,%r9), %r9d movq %r10, (%rsp) leaq 0xc0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b0cbe movq (%r14), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %r9 movl 0xe0(%r14,%r9), %eax cmpl $0x1, %eax jne 0x2b0463 movq 0x10(%rsp), %rax movl 0xd0(%rax,%r9), %r8d movl 0xd4(%rax,%r9), %ecx movl %ecx, 0xa0(%rsp) movl 0xd8(%rax,%r9), %edx xorl %r10d, %r10d testl %edx, %edx cmovlel %r10d, %edx movq 0x40(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r11d cmovgl %ecx, %r11d movl 0xdc(%rax,%r9), %esi testl %esi, %esi cmovlel %r10d, %esi testl %r8d, %r8d cmovlel %r10d, %r8d movq 0x38(%rsp), %r9 imulq 0xa8(%rsp), %r9 imulq 0x18(%rsp), %r9 addq %r13, %r9 movq %rbp, %rdi cmpl %r8d, %r10d je 0x2b034d movl %edx, %ecx subl $0x1, %ecx jb 0x2b02ec movq (%r12), %r14 movq %r14, (%rdi) addq $0x8, %rdi jmp 0x2b02da xorl %r14d, %r14d cmpl %r14d, %r11d je 0x2b0304 movq (%r9,%r14,8), %rcx movq %rcx, (%rdi) addq $0x8, %rdi incq %r14 jmp 0x2b02ef movl %esi, %ecx subl $0x1, %ecx jb 0x2b0319 movq -0x8(%r9,%r14,8), %r15 movq %r15, (%rdi) addq $0x8, %rdi jmp 0x2b0306 incl %r10d jmp 0x2b02d3 imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi imulq %r11, %r8 imulq %r15, %r8 addq %r8, %rdi xorl %eax, %eax movq %r15, %rcx cmpq %rax, %rsi je 0x2b067e movq %r10, (%rdi,%rax,8) incq %rax jmp 0x2b033b movslq 0xec(%rsp), %rax xorl %r11d, %r11d testl %eax, %eax movl $0x0, %r8d movq %rax, 0x108(%rsp) cmovgl %eax, %r8d movq 0xb0(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl %eax, %r14d movq %r12, %r9 cmpl %r14d, %r11d je 0x2b03df movl %edx, %ecx subl $0x1, %ecx jb 0x2b039b movq (%r9), %r10 movq %r10, (%rdi) addq $0x8, %rdi jmp 0x2b038a xorl %r15d, %r15d xorl %r10d, %r10d cmpl %r15d, %r8d je 0x2b03b7 movq (%r9,%r15,8), %rcx movq %rcx, (%rdi,%r15,8) addq $-0x8, %r10 incq %r15 jmp 0x2b03a1 movq %r9, %rcx subq %r10, %rcx subq %r10, %rdi movl %esi, %r10d subl $0x1, %r10d jb 0x2b03d7 movq -0x8(%r9,%r15,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b03c3 incl %r11d movq %rcx, %r9 jmp 0x2b0383 movq 0x108(%rsp), %rax shlq $0x3, %rax subq %rax, %r9 xorl %r10d, %r10d movl 0xa0(%rsp), %r14d testl %r14d, %r14d cmovlel %r10d, %r14d cmpl %r14d, %r10d je 0x2b044a movl %edx, %ecx subl $0x1, %ecx jb 0x2b0418 movq (%r9), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b0407 xorl %r11d, %r11d cmpl %r11d, %r8d je 0x2b0430 movq (%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi incq %r11 jmp 0x2b041b movl %esi, %ecx subl $0x1, %ecx jb 0x2b0445 movq -0x8(%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b0432 incl %r10d jmp 0x2b0400 movq 0x98(%rsp), %rax movq -0x18(%rax), %r9 movq 0x10(%rsp), %rax movl 0xe0(%rax,%r9), %eax cmpl $0x2, %eax jne 0x2b066c movq 0x10(%rsp), %rcx movl 0xd0(%rcx,%r9), %r14d movslq 0xd8(%rcx,%r9), %rax movq %rcx, %r8 movl %r14d, %r11d movq 0x40(%rsp), %rcx imull %ecx, %r11d movq %rcx, %rdx negq %rdx xorl %r15d, %r15d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi testl %ecx, %ecx movl $0x0, %edi cmovgl %ecx, %edi movl 0xdc(%r8,%r9), %r8d testl %r8d, %r8d cmovlel %r15d, %r8d testl %r14d, %r14d cmovlel %r15d, %r14d movq 0x38(%rsp), %r10 imulq 0xa8(%rsp), %r10 movslq %r11d, %rcx leaq (%r12,%rcx,8), %r11 imulq 0x18(%rsp), %r10 leaq (%r10,%rax,8), %r10 leaq (%r10,%rcx,8), %r10 movq 0x10(%rsp), %r12 movl 0xd4(%r12,%r9), %r9d movl %r9d, 0x38(%rsp) shlq $0x3, %rax addq %r10, %r13 movq 0x40(%rsp), %rcx leaq (,%rcx,8), %r10 negq %r10 negq %rsi negq %r8 cmpl %r14d, %r15d je 0x2b0575 xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b0533 movq (%r13,%rcx,8), %r12 movq %r12, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b051c xorl %ecx, %ecx movq %r11, %r12 cmpl %edi, %ecx je 0x2b0550 movq (%r12), %r9 addq $0x8, %r12 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b0538 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b0569 movq -0x10(%r12,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b0552 leaq (%r11,%rdx,8), %r11 incl %r15d addq %r10, %r13 jmp 0x2b0515 xorl %r14d, %r14d movq 0xb0(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r15d cmovgl %ecx, %r15d movl 0x38(%rsp), %r13d cmpl %r15d, %r14d je 0x2b05ea leaq (%r11,%rax), %rcx xorl %r12d, %r12d cmpq %r12, %rsi je 0x2b05b3 movq (%rcx,%r12,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %r12 jmp 0x2b059d xorl %ecx, %ecx cmpl %edi, %ecx je 0x2b05cc movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b05b5 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b05e5 movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b05ce incl %r14d jmp 0x2b0591 movq 0x40(%rsp), %rcx addl %ecx, %ecx movslq %ecx, %rcx shlq $0x3, %rcx movq %r11, %r14 subq %rcx, %r14 xorl %r15d, %r15d testl %r13d, %r13d cmovlel %r15d, %r13d subq %rcx, %rax addq %r11, %rax cmpl %r13d, %r15d je 0x2b066c xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b062b movq (%rax,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b0615 xorl %ecx, %ecx movq %r14, %r11 cmpl %edi, %ecx je 0x2b0647 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b0630 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b0660 movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b0649 leaq (%r14,%rdx,8), %r14 incl %r15d addq %r10, %rax jmp 0x2b060e movq 0x10(%rsp), %r12 movq 0x110(%rsp), %r13 movq 0x48(%rsp), %rcx incq %rcx movq 0x18(%rsp), %r10 incl %r10d jmp 0x2afffe movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xe0(%r13,%rax) jne 0x2b09db movl 0xe8(%r13,%rax), %ecx addl 0x18(%rsp), %ecx addl 0xec(%r13,%rax), %ecx addl 0xd0(%r13,%rax), %edx addl 0xd4(%r13,%rax), %edx addl 0xd8(%r13,%rax), %esi addl 0xdc(%r13,%rax), %esi movq 0x8(%r12), %rax movq %rax, 0x8(%rsp) movl $0x8, (%rsp) movq %rbx, %rdi movq %rcx, %r14 movl %ebp, %r8d callq 0x62a26 cmpq $0x0, (%rbx) je 0x2b0ad5 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b0ad5 xorl %ecx, %ecx testl %r14d, %r14d cmovlel %ecx, %r14d movq %r14, 0x48(%rsp) testl %ebp, %ebp movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %ebp movq %rbp, 0x30(%rsp) movabsq $0x100000001, %r12 # imm = 0x100000001 xorl %ebp, %ebp movq %rbx, 0xb8(%rsp) movq 0x10(%rsp), %r13 cmpq 0x30(%rsp), %rbp je 0x2b0ad5 movq 0x10(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax cvttss2si 0xe4(%rcx,%rax), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rsi shlq $0x20, %rsi orq %rcx, %rsi orq %rdx, %rsi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r14 shlq $0x38, %r14 orq %rdx, %r14 orq %rsi, %r14 orq %rax, %r14 xorl %r15d, %r15d cmpq 0x48(%rsp), %r15 je 0x2b0981 movslq 0x2c(%rbx), %rdi movslq 0x30(%rbx), %rdx movq (%rbx), %rax movq 0x10(%rbx), %rsi movq 0x40(%rbx), %r8 imulq %rbp, %r8 movq %r8, %r9 imulq %rsi, %r9 addq %rax, %r9 movl 0x18(%rbx), %r10d movq %rdx, %rcx imulq %rdi, %rcx movq %r15, %r11 imulq %rsi, %r11 imulq %rcx, %r11 addq %r9, %r11 movq 0x20(%rbx), %r9 movq %r11, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %rsi, 0x60(%rsp) movl %r10d, 0x68(%rsp) movq %r9, 0x70(%rsp) movl $0x2, 0x78(%rsp) movl %edi, 0x7c(%rsp) movl %edx, 0x80(%rsp) movq %r12, 0x84(%rsp) movq %rcx, 0x90(%rsp) movq (%r13), %r9 movq -0x18(%r9), %r9 movl 0xe8(%r13,%r9), %r11d movl %r15d, %r10d subl %r11d, %r10d setns %r11b cmpl 0x18(%rsp), %r10d movq %r12, %rbx setl %r12b testb %r12b, %r11b je 0x2b0948 movq 0x28(%rsp), %r8 movslq 0x2c(%r8), %rax movslq 0x30(%r8), %rcx movq 0x40(%r8), %rdx imulq %rbp, %rdx movq 0x10(%r8), %rsi imulq %rsi, %rdx addq (%r8), %rdx movl 0x18(%r8), %edi movq 0x20(%r8), %r8 movq %rcx, %r11 imulq %rax, %r11 movl %r10d, %r10d movq %r11, %r12 imulq %rsi, %r12 imulq %r10, %r12 addq %rdx, %r12 movq %r12, 0xc0(%rsp) andq $0x0, 0xc8(%rsp) movq %rsi, 0xd0(%rsp) movl %edi, 0xd8(%rsp) movq %r8, 0xe0(%rsp) movl $0x2, 0xe8(%rsp) movl %eax, 0xec(%rsp) movl %ecx, 0xf0(%rsp) movq %rbx, 0xf4(%rsp) movq %r11, 0x100(%rsp) movl 0xd0(%r13,%r9), %edx movl 0xd4(%r13,%r9), %ecx movl 0xd8(%r13,%r9), %r8d movl 0xdc(%r13,%r9), %r9d movq %r14, (%rsp) leaq 0xc0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b0cbe movq %rbx, %r12 movq 0xb8(%rsp), %rbx incq %r15 jmp 0x2b07b9 testl %ecx, %ecx movl $0x0, %r9d cmovlel %r9d, %ecx imulq %r15, %rdx imulq %rdi, %rdx addq %r8, %rdx imulq %rdx, %rsi addq %rsi, %rax xorl %edx, %edx movq %rbx, %r12 movq 0xb8(%rsp), %rbx cmpq %rdx, %rcx je 0x2b0940 movq %r14, (%rax,%rdx,8) incq %rdx jmp 0x2b0973 incq %rbp jmp 0x2b074f movq (%r13), %rax movq -0x18(%rax), %r10 movl 0xd0(%r13,%r10), %r14d leal (%r14,%rdx,8), %eax addl 0xd4(%r13,%r10), %eax xorl %r15d, %r15d testb $0x7, %al sete %r15b movq %r9, %r8 shrq $0x3, %r8 leal (%r15,%r15,2), %ecx shlq %cl, %r8 testb $0x7, %r14b sete %cl andb %cl, %r15b cmpb $0x1, %r15b jne 0x2b09db cmpl $0x0, 0xe0(%r13,%r10) je 0x2b0aec movq 0x28(%rsp), %r8 movq (%r8), %rax movq %rax, 0x50(%rsp) movq 0x8(%r8), %rax movq %rax, 0x58(%rsp) movq %r9, 0x60(%rsp) movl %edi, 0x68(%rsp) movq 0x20(%r8), %rcx movq %rcx, 0x70(%rsp) movl %r11d, 0x78(%rsp) movl %esi, 0x7c(%rsp) movl %edx, 0x80(%rsp) movl 0x18(%rsp), %ecx movl %ecx, 0x84(%rsp) movl %ebp, 0x88(%rsp) movq 0x40(%r8), %rcx movq %rcx, 0x90(%rsp) testq %rax, %rax je 0x2b0a39 lock incl (%rax) cmpl $0x1, %edi je 0x2b0a87 movups (%r12), %xmm0 movups 0x10(%r12), %xmm1 movups 0x20(%r12), %xmm2 movups 0x30(%r12), %xmm3 leaq 0xc0(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x10(%r12), %rax movq %rax, 0x8(%rcx) leaq 0x50(%rsp), %rsi pushq $0x1 popq %rdx movq 0x28(%rsp), %rdi callq 0x64e3b movq (%r13), %rax addq -0x18(%rax), %r13 leaq 0x50(%rsp), %rsi movq %r13, %rdi movq %rbx, %rdx movq %r12, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b0ad5 lock decl (%rax) jne 0x2b0ad5 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2b0acd movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b0ad5 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0x118, %rsp # imm = 0x118 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq addl 0xd8(%r13,%r10), %esi addl 0xdc(%r13,%r10), %esi sarl $0x3, %eax movq 0x8(%r12), %r9 pushq $0x8 popq %rbp movq %rbx, %rdi movl %eax, %edx movq %r8, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b0ad5 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b0ad5 movq 0x10(%rsp), %r9 movq (%r9), %rax movq -0x18(%rax), %rsi cvttss2si 0xe4(%r9,%rsi), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rdi shlq $0x20, %rdi orq %rcx, %rdi orq %rdx, %rdi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r11 shlq $0x38, %r11 orq %rdx, %r11 orq %rdi, %r11 orq %rax, %r11 movl 0xd0(%r9,%rsi), %eax movl 0xd4(%r9,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%r9,%rsi), %r8d movl 0xdc(%r9,%rsi), %r9d movq %r11, (%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b0cbe movq $0x0, 0x20(%rsp) jmp 0x2b0ad5 jmp 0x2b0c13 jmp 0x2b0bda movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b0c0b lock decl (%rax) jne 0x2b0c0b movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2b0c05 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b0c0b movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/x86/padding_x86.cpp
virtual thunk to ncnn::Padding_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2aee70	/csukuangfj[P]ncnn/src/layer/x86/padding_x86.cpp
ncnn::padding_constant_pack8_int8_sse(ncnn::Mat const&, ncnn::Mat&, int, int, int, int, long)	static void padding_constant_pack8_int8_sse(const Mat& src, Mat& dst, int top, int bottom, int left, int right, int64_t _v) { const int64_t* ptr = src; int64_t* outptr = dst; // fill top for (int y = 0; y < top; y++) { for (int x = 0; x < dst.w; x++) { outptr++ = _v; } } // fill center for (int y = 0; y < src.h; y++) { for (int x = 0; x < left; x++) { outptr++ = _v; } for (int x = 0; x < src.w; x++) { outptr++ = ptr++; } for (int x = 0; x < right; x++) { outptr++ = _v; } } // fill bottom for (int y = 0; y < bottom; y++) { for (int x = 0; x < dst.w; x++) { outptr++ = _v; } } }	pushq %rbp pushq %r15 pushq %r14 pushq %r12 pushq %rbx movq 0x30(%rsp), %rax movq (%rdi), %r11 movl 0x2c(%rsi), %ebx xorl %ebp, %ebp testl %ebx, %ebx cmovlel %ebp, %ebx movq (%rsi), %r10 testl %edx, %edx cmovlel %ebp, %edx cmpl %edx, %ebp je 0x2b0cfa movl %ebx, %r14d subl $0x1, %r14d jb 0x2b0cf6 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b0ce7 incl %ebp jmp 0x2b0ce0 xorl %edx, %edx testl %r8d, %r8d cmovlel %edx, %r8d movl 0x2c(%rdi), %ebx testl %ebx, %ebx cmovlel %edx, %ebx testl %r9d, %r9d cmovlel %edx, %r9d movl 0x30(%rdi), %edi testl %edi, %edi cmovlel %edx, %edi cmpl %edi, %edx je 0x2b0d66 movl %r8d, %ebp subl $0x1, %ebp jb 0x2b0d2f movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b0d21 xorl %r15d, %r15d xorl %r14d, %r14d cmpl %r15d, %ebx je 0x2b0d4b movq (%r11,%r15,8), %r12 movq %r12, (%r10,%r15,8) addq $-0x8, %r14 incq %r15 jmp 0x2b0d35 subq %r14, %r11 subq %r14, %r10 movl %r9d, %ebp subl $0x1, %ebp jb 0x2b0d62 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b0d54 incl %edx jmp 0x2b0d1a movl 0x2c(%rsi), %edx xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %ecx, %ecx cmovlel %esi, %ecx movl %edx, %edi cmpl %ecx, %esi je 0x2b0d8d subl $0x1, %edi jb 0x2b0d89 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b0d7b incl %esi jmp 0x2b0d75 popq %rbx popq %r12 popq %r14 popq %r15 popq %rbp retq nop	/csukuangfj[P]ncnn/src/layer/x86/padding_pack8_int8.h
ncnn::Padding_x86_avx512::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86_avx512::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x118, %rsp # imm = 0x118 movq %rcx, %rbp movq %rdx, 0x8(%rsp) movq %rsi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rax leaq (%rdi,%rax), %r11 movq %rdi, 0x10(%rsp) movl 0xd0(%rdi,%rax), %r10d testl %r10d, %r10d jne 0x2b0e4f cmpl $0x0, 0xd4(%r11) jne 0x2b0e4f cmpl $0x0, 0xd8(%r11) jne 0x2b0e4f cmpl $0x0, 0xdc(%r11) jne 0x2b0e4f cmpl $0x0, 0xe8(%r11) jne 0x2b0e4f cmpl $0x0, 0xec(%r11) je 0x2b351e movl 0x18(%r14), %r15d movq 0x10(%r14), %r9 testl %r15d, %r15d je 0x2b0e93 leal (,%r9,8), %eax cltd idivl %r15d cmpl $0x8, %eax jne 0x2b0ea5 movq 0x10(%rsp), %rdi movq %r14, %rsi movq 0x8(%rsp), %rdx movq %rbp, %rcx addq $0x118, %rsp # imm = 0x118 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x2b362c vmovdqu 0x2c(%r14), %xmm0 movq %r14, %rdi movl 0x28(%r14), %r13d jmp 0x2b1157 vmovdqu 0x2c(%r14), %xmm0 movq %r14, (%rsp) movl 0x28(%r14), %ecx vpextrd $0x2, %xmm0, %r12d pushq $0x4 popq %r13 vpextrd $0x1, %xmm0, %eax vmovd %xmm0, %esi vpextrd $0x3, %xmm0, %edx movq %rdx, 0x18(%rsp) pushq $-0x64 popq %rdx movq %rdx, 0x20(%rsp) cmpl $0x4, %r15d je 0x2b104a cmpl $0x8, %r15d je 0x2b0f4a cmpl $0x10, %r15d jne 0x2b1150 leal -0x1(%rcx), %edx cmpl $0x3, %edx ja 0x2b1150 leaq 0x147891(%rip), %rcx # 0x3f8798 movslq (%rcx,%rdx,4), %rdx addq %rcx, %rdx jmpq %rdx shll $0x4, %esi movl 0xd8(%r11), %eax addl %eax, %esi addl 0xdc(%r11), %esi testb $0xf, %sil je 0x2b2bce testb $0x7, %sil je 0x2b3371 xorl %ecx, %ecx testb $0x3, %sil sete %cl leal (%rcx,%rcx,2), %ecx incl %ecx jmp 0x2b3374 leal -0x1(%rcx), %edx cmpl $0x3, %edx ja 0x2b1150 leaq 0x14782b(%rip), %rcx # 0x3f8788 movslq (%rcx,%rdx,4), %rdx addq %rcx, %rdx jmpq %rdx movl 0xd8(%r11), %edx leal (%rdx,%rsi,8), %eax addl 0xdc(%r11), %eax testb $0x3, %al sete %cl movq %r9, %r10 shrq $0x3, %r10 addb %cl, %cl testb $0x7, %al movzbl %cl, %esi pushq $0x3 popq %rcx cmovnel %esi, %ecx shlq %cl, %r10 orl %eax, %edx pushq $0x1 popq %r13 testb $0x7, %dl jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 pushq $0x8 popq %r12 cltd idivl %r12d movq 0x8(%rbp), %r8 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %esi movq %r10, %rdx movl %r12d, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vpbroadcastd 0xe4(%rdx,%rcx), %ymm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %r12d movl %eax, %r8d movl %ecx, %eax cltd idivl %r12d movq $0x0, 0x20(%rsp) movq (%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b4597 jmp 0x2b1239 leal -0x1(%rcx), %edx cmpl $0x3, %edx ja 0x2b1150 leaq 0x14771b(%rip), %rcx # 0x3f8778 movslq (%rcx,%rdx,4), %rdx addq %rcx, %rdx jmpq %rdx movl 0xd8(%r11), %r8d leal (%r8,%rsi,4), %esi addl 0xdc(%r11), %esi testb $0x3, %sil sete %cl movq %r9, %rdx shrq $0x2, %rdx addb %cl, %cl movl %esi, %eax andl $0x7, %eax movzbl %cl, %r10d pushq $0x3 popq %rcx cmovnel %r10d, %ecx shlq %cl, %rdx pushq $0x1 popq %r13 testb $0x3, %r8b jne 0x2b1153 cmpl $0x4, %eax jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 sarl $0x2, %esi movq 0x8(%rbp), %r8 pushq $0x4 popq %rbp movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vpbroadcastd 0xe4(%rdx,%rcx), %xmm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq $0x0, 0x20(%rsp) movq (%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b4637 jmp 0x2b1239 movl %ecx, %r13d movq (%rsp), %rdi movq 0x8(%rdi), %rax vmovups (%rdi), %xmm1 vmovaps %xmm1, 0x80(%rsp) movq %r9, 0x90(%rsp) movl %r15d, 0x98(%rsp) movq 0x20(%rdi), %rcx movq %rcx, 0xa0(%rsp) movl %r13d, 0xa8(%rsp) vmovdqu %xmm0, 0xac(%rsp) movq 0x40(%rdi), %rcx movq %rcx, 0xc0(%rsp) testq %rax, %rax je 0x2b11a9 lock incl (%rax) cmpl $0x1, %r15d je 0x2b11dc vmovdqu64 (%rbp), %zmm0 leaq 0x30(%rsp), %rcx vmovdqu64 %zmm0, (%rcx) movq 0x10(%rbp), %rax movq %rax, 0x8(%rcx) leaq 0x80(%rsp), %rsi pushq $0x1 popq %rdx vzeroupper callq 0x64e3b movq 0x10(%rsp), %rdi movq (%rdi), %rax addq -0x18(%rax), %rdi leaq 0x80(%rsp), %rsi movq 0x8(%rsp), %rdx movq %rbp, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x88(%rsp), %rax testq %rax, %rax je 0x2b1239 lock decl (%rax) jne 0x2b1239 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b1231 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b1239 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0x118, %rsp # imm = 0x118 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movl 0xe8(%r11), %r14d movq 0x18(%rsp), %rdi leal (%r14,%rdi,8), %edx addl 0xec(%r11), %edx testb $0x3, %dl sete %cl movq %r9, %r8 shrq $0x3, %r8 addb %cl, %cl testb $0x7, %dl movzbl %cl, %ecx pushq $0x3 popq %r13 cmovel %r13d, %ecx shlq %cl, %r8 orl %edx, %r14d testb $0x7, %r14b jne 0x2b1153 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi addl %eax, %r10d addl 0xd4(%r11), %r10d leal (,%rdi,8), %eax cmpl %eax, %edx movq (%rsp), %rdi je 0x2b12cc cmpl $0x0, 0xe0(%r11) jne 0x2b1157 pushq $0x8 popq %r9 movl %edx, %eax cltd idivl %r9d subq $0x8, %rsp movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movq %rax, %r14 movl %eax, %ecx pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xc8(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xc8(%rsp), %r9 je 0x2b1239 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b1419 movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b1443 movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x5, %r8 vmovdqu (%rdx,%r8), %ymm0 jmp 0x2b144d vpbroadcastd 0xe4(%r12,%rdi), %ymm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b160a movq (%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b1504 movq %rsi, 0x70(%rsp) movq 0x10(%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b1560 movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b4597 movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b1634 movq 0x10(%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b163e vmovdqa (%rbp), %ymm0 movl %eax, %edi subl $0x1, %edi jb 0x2b15cd vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b15bc xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b15ef vmovdqa (%r12,%rdi), %ymm0 vmovdqa %ymm0, (%r13,%rdi) incl %r10d addq $0x20, %rdi jmp 0x2b15d2 addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b1605 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b15f4 incl %r8d jmp 0x2b15ac imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b19db vmovdqu %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b161f movq 0x10(%rsp), %r12 jmp 0x2b1722 xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b169b vmovdqa (%rbp), %ymm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b1660 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b164e xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b1682 vmovdqa (%rbp), %ymm0 vmovdqa %ymm0, (%r13) addq $0x20, %rbp addq $0x20, %r13 incl %r8d jmp 0x2b1663 movl %edx, %r8d subl $0x1, %r8d jb 0x2b1697 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b1685 incl %edi jmp 0x2b1640 movl 0x5c(%rsp), %edi shll $0x3, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b1711 vmovdqa (%rbp), %ymm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b16d1 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b16bf xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b16f5 vmovdqa (%rbp,%r8), %ymm0 vmovdqa %ymm0, (%r13,%r8) incl %r9d addq $0x20, %r8 jmp 0x2b16d7 addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b170d vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b16fb incl %edi jmp 0x2b16b3 movq -0x18(%rcx), %rdi movq 0x10(%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x40 popq %r13 jne 0x2b1943 movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x3, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x5, %rax shlq $0x5, %rdx shlq $0x5, %rsi cmpl %r9d, %r10d je 0x2b1824 leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b17c5 vmovdqa (%r11,%r14), %ymm0 vmovdqa %ymm0, (%r8,%r15) addq $-0x20, %r14 addq $0x20, %r15 jmp 0x2b17a7 subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b17e9 vmovdqa (%r11), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r11 addq $0x20, %r8 incl %ebp jmp 0x2b17cd movq %r13, %r14 movq 0x10(%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x40, %r15 je 0x2b1810 vmovdqa (%r11,%r14), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r14 jmp 0x2b17f1 movslq 0x5c(%rsp), %r11 shlq $0x5, %r11 subq %r11, %rcx incl %r10d jmp 0x2b1794 xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b18a0 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b1856 vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8,%r14) addq $-0x20, %r11 addq $0x20, %r14 jmp 0x2b1838 subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b1879 vmovdqa (%rcx), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %rcx addq $0x20, %r8 incl %r10d jmp 0x2b185c movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x40, %r11 je 0x2b189b vmovdqa (%rcx,%r10), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r10 jmp 0x2b187c incl %r9d jmp 0x2b1827 movslq 0x5c(%rsp), %r9 shlq $0x6, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b1943 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b18e6 vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8,%r14) addq $-0x20, %r11 addq $0x20, %r14 jmp 0x2b18c8 subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b190d vmovdqa (%r10), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r10 addq $0x20, %r8 incl %r11d jmp 0x2b18ef movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x40, %r14 je 0x2b192f vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r11 jmp 0x2b1910 movslq 0x5c(%rsp), %r10 shlq $0x5, %r10 subq %r10, %rcx incl %r9d jmp 0x2b18b5 movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b1977 lock decl (%rax) jne 0x2b1977 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b196c movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x2b1977 movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b19db lock decl (%rax) jne 0x2b19db movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b19c3 movq (%rdi), %rax vzeroupper callq 0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b19db movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b1359 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 movl 0xe8(%r11), %r13d addl %r12d, %r13d addl 0xec(%r11), %r13d addl %eax, %r10d addl 0xd4(%r11), %r10d addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x8 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %r11 # imm = 0x100000001 xorl %r14d, %r14d movq 0x10(%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b1239 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b1aca movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x5, %rcx vmovdqu (%rax,%rcx), %ymm0 jmp 0x2b1ad4 vpbroadcastd 0xe4(%rbp,%rax), %ymm0 xorl %r15d, %r15d vmovdqu %ymm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b1c86 movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %r11, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b1c6d movq (%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %r13, %rbx movq %r11, %r13 movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq 0x10(%rsp), %rbp movq %r13, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovups 0xd0(%rsp), %ymm0 callq 0x2b4597 vmovdqu 0xd0(%rsp), %ymm0 movq %r13, %r11 movq %rbx, %r13 incq %r15 jmp 0x2b1ae0 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b1c65 vmovdqu %ymm0, (%rcx) addq $0x20, %rcx jmp 0x2b1c77 incq %r14 jmp 0x2b1a97 movl 0xe8(%r11), %r14d movq 0x18(%rsp), %rdi leal (%r14,%rdi,4), %edx addl 0xec(%r11), %edx testb $0x3, %dl sete %cl movq %r9, %r8 shrq $0x2, %r8 addb %cl, %cl movq %rdx, %r12 andl $0x7, %edx movzbl %cl, %ecx pushq $0x3 popq %r13 cmovel %r13d, %ecx shlq %cl, %r8 testb $0x3, %r14b jne 0x2b1153 cmpl $0x4, %edx jne 0x2b1153 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi addl %eax, %r10d addl 0xd4(%r11), %r10d leal (,%rdi,4), %eax cmpl %eax, %r12d movq (%rsp), %rdi je 0x2b1d11 cmpl $0x0, 0xe0(%r11) jne 0x2b1157 movq %r12, %rcx sarl $0x2, %ecx subq $0x8, %rsp pushq $0x4 popq %r9 movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movq %rcx, %r14 pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x4 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xc8(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xc8(%rsp), %r9 je 0x2b1239 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b1e5c movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b1e86 movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x4, %r8 vmovdqu (%rdx,%r8), %xmm0 jmp 0x2b1e90 vpbroadcastd 0xe4(%r12,%rdi), %xmm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b204d movq (%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b1f47 movq %rsi, 0x70(%rsp) movq 0x10(%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b1fa3 movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b4637 movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b2077 movq 0x10(%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b2081 vmovdqa (%rbp), %xmm0 movl %eax, %edi subl $0x1, %edi jb 0x2b2010 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b1fff xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b2032 vmovdqa (%r12,%rdi), %xmm0 vmovdqa %xmm0, (%r13,%rdi) incl %r10d addq $0x10, %rdi jmp 0x2b2015 addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b2048 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b2037 incl %r8d jmp 0x2b1fef imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b2412 vmovdqu %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b2062 movq 0x10(%rsp), %r12 jmp 0x2b2165 xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b20de vmovdqa (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b20a3 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b2091 xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b20c5 vmovdqa (%rbp), %xmm0 vmovdqa %xmm0, (%r13) addq $0x10, %rbp addq $0x10, %r13 incl %r8d jmp 0x2b20a6 movl %edx, %r8d subl $0x1, %r8d jb 0x2b20da vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b20c8 incl %edi jmp 0x2b2083 movl 0x5c(%rsp), %edi shll $0x2, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b2154 vmovdqa (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b2114 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b2102 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b2138 vmovdqa (%rbp,%r8), %xmm0 vmovdqa %xmm0, (%r13,%r8) incl %r9d addq $0x10, %r8 jmp 0x2b211a addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b2150 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b213e incl %edi jmp 0x2b20f6 movq -0x18(%rcx), %rdi movq 0x10(%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x20 popq %r13 jne 0x2b2386 movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x2, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x4, %rax shlq $0x4, %rdx shlq $0x4, %rsi cmpl %r9d, %r10d je 0x2b2267 leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b2208 vmovdqa (%r11,%r14), %xmm0 vmovdqa %xmm0, (%r8,%r15) addq $-0x10, %r14 addq $0x10, %r15 jmp 0x2b21ea subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b222c vmovdqa (%r11), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r11 addq $0x10, %r8 incl %ebp jmp 0x2b2210 movq %r13, %r14 movq 0x10(%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x20, %r15 je 0x2b2253 vmovdqa (%r11,%r14), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r14 jmp 0x2b2234 movslq 0x5c(%rsp), %r11 shlq $0x4, %r11 subq %r11, %rcx incl %r10d jmp 0x2b21d7 xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b22e3 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b2299 vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2b227b subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b22bc vmovdqa (%rcx), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %rcx addq $0x10, %r8 incl %r10d jmp 0x2b229f movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x20, %r11 je 0x2b22de vmovdqa (%rcx,%r10), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r10 jmp 0x2b22bf incl %r9d jmp 0x2b226a movslq 0x5c(%rsp), %r9 shlq $0x5, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b2386 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b2329 vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2b230b subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b2350 vmovdqa (%r10), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r10 addq $0x10, %r8 incl %r11d jmp 0x2b2332 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x20, %r14 je 0x2b2372 vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r11 jmp 0x2b2353 movslq 0x5c(%rsp), %r10 shlq $0x4, %r10 subq %r10, %rcx incl %r9d jmp 0x2b22f8 movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b23b4 lock decl (%rax) jne 0x2b23b4 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b23ac movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b23b4 movq %rsi, %rdi callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b2412 lock decl (%rax) jne 0x2b2412 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b23fd movq (%rdi), %rax callq 0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b2412 movq %rsi, %rdi callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b1d9c movq 0x18(%rsp), %rdx shll $0x4, %edx movl 0xe8(%r11), %r14d leal (%r14,%rdx), %ecx addl 0xec(%r11), %ecx testb $0xf, %cl je 0x2b2bca testb $0x7, %cl je 0x2b2bdc xorl %r8d, %r8d testb $0x3, %cl sete %r8b leal (%r8,%r8,2), %r8d incl %r8d jmp 0x2b2be0 leal (%r10,%rax,8), %eax addl 0xd4(%r11), %eax testb $0x3, %al sete %cl movq %r9, %r8 shrq $0x3, %r8 addb %cl, %cl testb $0x7, %al movzbl %cl, %edx pushq $0x3 popq %rcx cmovnel %edx, %ecx shlq %cl, %r8 orl %eax, %r10d pushq $0x2 popq %r13 testb $0x7, %r10b jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi pushq $0x8 popq %r12 cltd idivl %r12d movq 0x8(%rbp), %r9 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %edx movq %r8, %rcx movl %r12d, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vpbroadcastd 0xe4(%rdi,%rsi), %ymm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %r12d movl %eax, %r10d movl %ecx, %eax cltd idivl %r12d movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq (%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b4597 jmp 0x2b3510 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 movl 0xe8(%r11), %r13d addl %r12d, %r13d addl 0xec(%r11), %r13d addl %eax, %r10d addl 0xd4(%r11), %r10d addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x4 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %r11 # imm = 0x100000001 xorl %r14d, %r14d movq 0x10(%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b1239 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b2636 movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x4, %rcx vmovdqu (%rax,%rcx), %xmm0 jmp 0x2b2640 vpbroadcastd 0xe4(%rbp,%rax), %xmm0 xorl %r15d, %r15d vmovdqa %xmm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b27f8 movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %r11, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b27df movq (%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %r13, %rbx movl %r12d, %r13d movq %r11, %r12 movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq 0x10(%rsp), %rbp movq %r12, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovaps 0xd0(%rsp), %xmm0 callq 0x2b4637 vmovdqa 0xd0(%rsp), %xmm0 movq %r12, %r11 movl %r13d, %r12d movq %rbx, %r13 incq %r15 jmp 0x2b264c testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b27d7 vmovdqu %xmm0, (%rcx) addq $0x10, %rcx jmp 0x2b27e9 incq %r14 jmp 0x2b2603 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 movl 0xe8(%r11), %r13d addl %r12d, %r13d addl 0xec(%r11), %r13d addl %eax, %r10d addl 0xd4(%r11), %r10d addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x10 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %rbx # imm = 0x100000001 xorl %r14d, %r14d movq 0x10(%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b1239 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b28e6 movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x6, %rcx vmovdqu64 (%rax,%rcx), %zmm0 jmp 0x2b28ee vpbroadcastd 0xe4(%rbp,%rax), %zmm0 xorl %r15d, %r15d vmovdqu64 %zmm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b2a9c movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %rbx, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b2a81 movq (%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq 0x10(%rsp), %rbp movq %rbx, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovups 0xd0(%rsp), %zmm0 callq 0x2b44ec vmovdqu64 0xd0(%rsp), %zmm0 incq %r15 jmp 0x2b28fc testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b2a79 vmovdqu64 %zmm0, (%rcx) addq $0x40, %rcx jmp 0x2b2a8b incq %r14 jmp 0x2b28b1 leal (%r10,%rax,4), %edx addl 0xd4(%r11), %edx testb $0x3, %dl sete %cl movq %r9, %rax shrq $0x2, %rax addb %cl, %cl movl %edx, %r8d andl $0x7, %r8d movzbl %cl, %r14d pushq $0x3 popq %rcx cmovnel %r14d, %ecx shlq %cl, %rax pushq $0x2 popq %r13 testb $0x3, %r10b jne 0x2b1153 cmpl $0x4, %r8d jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi sarl $0x2, %edx movq 0x8(%rbp), %r9 pushq $0x4 popq %rbp movq 0x8(%rsp), %rbx movq %rbx, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vpbroadcastd 0xe4(%rdi,%rsi), %xmm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq (%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b4637 jmp 0x2b3510 shll $0x4, %eax addl %r10d, %eax addl 0xd4(%r11), %eax testb $0xf, %al je 0x2b2bd5 testb $0x7, %al je 0x2b343b xorl %ecx, %ecx testb $0x3, %al sete %cl leal (%rcx,%rcx,2), %ecx incl %ecx jmp 0x2b343e pushq $0x10 jmp 0x2b2bde pushq $0x10 jmp 0x2b3373 pushq $0x10 jmp 0x2b343d pushq $0x8 popq %r8 pushq $0x3 popq %r13 testb $0xf, %r14b jne 0x2b1153 cmpl $0x10, %r8d jne 0x2b1153 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi addl %eax, %r10d addl 0xd4(%r11), %r10d cmpl %edx, %ecx movq (%rsp), %rdi je 0x2b2c26 cmpl $0x0, 0xe0(%r11) jne 0x2b1157 shrq $0x4, %r9 movl %r8d, %r8d imulq %r9, %r8 pushq $0x10 popq %r9 movl %ecx, %eax cltd idivl %r9d subq $0x8, %rsp movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movq %rax, %r14 movl %eax, %ecx pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x10 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xc8(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xc8(%rsp), %r9 je 0x2b1239 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b2d7e movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b2da9 movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x6, %r8 vmovdqu64 (%rdx,%r8), %zmm0 jmp 0x2b2db1 vpbroadcastd 0xe4(%r12,%rdi), %zmm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b2f74 movq (%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b2e68 movq %rsi, 0x70(%rsp) movq 0x10(%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b2ec4 movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b44ec movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b2f9f movq 0x10(%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b2fa9 vmovdqa64 (%rbp), %zmm0 movl %eax, %edi subl $0x1, %edi jb 0x2b2f34 vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2f22 xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b2f58 vmovdqa64 (%r12,%rdi), %zmm0 vmovdqa64 %zmm0, (%r13,%rdi) incl %r10d addq $0x40, %rdi jmp 0x2b2f39 addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b2f6f vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2f5d incl %r8d jmp 0x2b2f10 imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b3366 vmovdqu64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2f89 movq 0x10(%rsp), %r12 jmp 0x2b309a xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b300d vmovdqa64 (%rbp), %zmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b2fce vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2fbb xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b2ff3 vmovdqa64 (%rbp), %zmm0 vmovdqa64 %zmm0, (%r13) addq $0x40, %rbp addq $0x40, %r13 incl %r8d jmp 0x2b2fd1 movl %edx, %r8d subl $0x1, %r8d jb 0x2b3009 vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2ff6 incl %edi jmp 0x2b2fab movl 0x5c(%rsp), %edi shll $0x4, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b3089 vmovdqa64 (%rbp), %zmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b3046 vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b3033 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b306c vmovdqa64 (%rbp,%r8), %zmm0 vmovdqa64 %zmm0, (%r13,%r8) incl %r9d addq $0x40, %r8 jmp 0x2b304c addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b3085 vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b3072 incl %edi jmp 0x2b3025 movq -0x18(%rcx), %rdi movq 0x10(%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x80 popq %r13 jne 0x2b32ce movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x4, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x6, %rax shlq $0x6, %rdx shlq $0x6, %rsi cmpl %r9d, %r10d je 0x2b31a2 leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b313f vmovdqa64 (%r11,%r14), %zmm0 vmovdqa64 %zmm0, (%r8,%r15) addq $-0x40, %r14 addq $0x40, %r15 jmp 0x2b311f subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b3165 vmovdqa64 (%r11), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r11 addq $0x40, %r8 incl %ebp jmp 0x2b3147 movq %r13, %r14 movq 0x10(%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x80, %r15 je 0x2b318e vmovdqa64 (%r11,%r14), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r8 addq $-0x40, %r14 jmp 0x2b316d movslq 0x5c(%rsp), %r11 shlq $0x6, %r11 subq %r11, %rcx incl %r10d jmp 0x2b310c xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b3225 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b31d6 vmovdqa64 (%r10,%r11), %zmm0 vmovdqa64 %zmm0, (%r8,%r14) addq $-0x40, %r11 addq $0x40, %r14 jmp 0x2b31b6 subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b31fc vmovdqa64 (%rcx), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %rcx addq $0x40, %r8 incl %r10d jmp 0x2b31dc movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x80, %r11 je 0x2b3220 vmovdqa64 (%rcx,%r10), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r8 addq $-0x40, %r10 jmp 0x2b31ff incl %r9d jmp 0x2b31a5 movslq 0x5c(%rsp), %r9 shlq $0x7, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b32ce leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b326d vmovdqa64 (%r10,%r11), %zmm0 vmovdqa64 %zmm0, (%r8,%r14) addq $-0x40, %r11 addq $0x40, %r14 jmp 0x2b324d subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b3296 vmovdqa64 (%r10), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r10 addq $0x40, %r8 incl %r11d jmp 0x2b3276 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x80, %r14 je 0x2b32ba vmovdqa64 (%r10,%r11), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r8 addq $-0x40, %r11 jmp 0x2b3299 movslq 0x5c(%rsp), %r10 shlq $0x6, %r10 subq %r10, %rcx incl %r9d jmp 0x2b323a movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b3302 lock decl (%rax) jne 0x2b3302 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b32f7 movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x2b3302 movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b3366 lock decl (%rax) jne 0x2b3366 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b334e movq (%rdi), %rax vzeroupper callq 0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b3366 movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b2cbe pushq $0x8 popq %rcx pushq $0x1 popq %r13 testb $0xf, %al jne 0x2b1153 cmpl $0x10, %ecx jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 shrq $0x4, %r9 movl %ecx, %ecx imulq %r9, %rcx pushq $0x10 popq %r12 movl %esi, %eax cltd idivl %r12d movq 0x8(%rbp), %r8 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %esi movq %rcx, %rdx movl %r12d, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vpbroadcastd 0xe4(%rdx,%rcx), %zmm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %r12d movl %eax, %r8d movl %ecx, %eax cltd idivl %r12d movq $0x0, 0x20(%rsp) movq (%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b44ec jmp 0x2b1239 pushq $0x8 popq %rcx pushq $0x2 popq %r13 testb $0xf, %r10b jne 0x2b1153 cmpl $0x10, %ecx jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi shrq $0x4, %r9 movl %ecx, %ecx imulq %r9, %rcx pushq $0x10 popq %r12 cltd idivl %r12d movq 0x8(%rbp), %r9 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %edx movl %r12d, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vpbroadcastd 0xe4(%rdi,%rsi), %zmm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %r12d movl %eax, %r10d movl %ecx, %eax cltd idivl %r12d movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq (%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b44ec movq $0x0, 0x20(%rsp) jmp 0x2b1239 movq $0x0, 0x20(%rsp) cmpq %r14, 0x8(%rsp) je 0x2b1239 movq 0x8(%r14), %rax testq %rax, %rax je 0x2b353e lock incl (%rax) movq 0x8(%rsp), %rax movq 0x8(%rax), %rax testq %rax, %rax je 0x2b3572 lock decl (%rax) jne 0x2b3572 movq 0x8(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x2b356a movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b3572 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx andq $0x0, 0x40(%rcx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, (%rcx) vmovups %xmm0, 0xc(%rcx) andl $0x0, 0x38(%rcx) vmovups %xmm0, 0x28(%rcx) vmovups (%r14), %xmm0 vmovups %xmm0, (%rcx) movq 0x10(%r14), %rax movq %rax, 0x10(%rcx) movl 0x18(%r14), %eax movl %eax, 0x18(%rcx) movq 0x20(%r14), %rax movq %rax, 0x20(%rcx) vmovdqu 0x28(%r14), %xmm0 vmovdqu %xmm0, 0x28(%rcx) movl 0x38(%r14), %eax movl %eax, 0x38(%rcx) movq 0x40(%r14), %rax movq %rax, 0x40(%rcx) jmp 0x2b1239 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b35e1 movq %rax, %rbx movq 0x88(%rsp), %rax testq %rax, %rax je 0x2b361b lock decl (%rax) jne 0x2b361b movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi jne 0x2b3615 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b361b movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_avx512.cpp
ncnn::Padding_x86_avx512::forward_int8(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86_avx512::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_int8_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_int8_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x128, %rsp # imm = 0x128 movq %rcx, %r12 movq %rdx, %rbx movq %rdi, %r14 movl 0x38(%rsi), %r15d movq 0x10(%rsi), %r9 movl 0x18(%rsi), %edi movq %rsi, 0x28(%rsp) vmovdqu 0x28(%rsi), %xmm0 cmpl $0x8, %edi jne 0x2b42dc vmovd %xmm0, %eax decl %eax cmpl $0x3, %eax ja 0x2b42dc leaq 0x14512e(%rip), %rcx # 0x3f87a8 movslq (%rcx,%rax,4), %rax vpextrd $0x1, %xmm0, %esi addq %rcx, %rax vpextrd $0x2, %xmm0, %edx pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) movq %r14, 0x18(%rsp) jmpq %rax movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %r8d leal (%r8,%rsi,8), %esi addl 0xdc(%r14,%rax), %esi xorl %r10d, %r10d testb $0x7, %sil sete %r10b movq %r9, %rdx shrq $0x3, %rdx leal (%r10,%r10,2), %ecx shlq %cl, %rdx testb $0x7, %r8b sete %cl andb %cl, %r10b cmpb $0x1, %r10b jne 0x2b42dc cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b42dc sarl $0x3, %esi movq 0x8(%r12), %r8 pushq $0x8 popq %rbp movq %rbx, %rdi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b43a8 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b43a8 movq 0x18(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rcx vcvttss2si 0xe4(%r8,%rcx), %rax movq %rax, %rdx shlq $0x8, %rdx movq %rax, %rsi shlq $0x10, %rsi orq %rdx, %rsi movq %rax, %rdx shlq $0x18, %rdx movq %rax, %rdi shlq $0x20, %rdi orq %rdx, %rdi orq %rsi, %rdi movq %rax, %rdx shlq $0x28, %rdx movq %rax, %rsi shlq $0x30, %rsi orq %rdx, %rsi movq %rax, %r9 shlq $0x38, %r9 orq %rsi, %r9 orq %rdi, %r9 orq %rax, %r9 movl 0xd8(%r8,%rcx), %eax movl 0xdc(%r8,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq %r9, (%rsp) movq $0x0, 0x20(%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b46e0 jmp 0x2b43a8 movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %r10d leal (%r10,%r15,8), %r13d addl 0xec(%r14,%rax), %r13d xorl %r11d, %r11d testb $0x7, %r13b sete %r11b movq %r9, %r8 shrq $0x3, %r8 leal (%r11,%r11,2), %ecx shlq %cl, %r8 testb $0x7, %r10b sete %cl andb %cl, %r11b cmpb $0x1, %r11b jne 0x2b42dc addl 0xd8(%r14,%rax), %esi addl 0xdc(%r14,%rax), %esi addl 0xd0(%r14,%rax), %edx addl 0xd4(%r14,%rax), %edx leal (,%r15,8), %ecx cmpl %ecx, %r13d je 0x2b3845 cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b42dc sarl $0x3, %r13d movq 0x8(%r12), %rax movq %rax, (%rsp) pushq $0x8 popq %r9 movq %rbx, %rdi movl %r13d, %ecx callq 0x628f2 cmpq $0x0, (%rbx) je 0x2b43a8 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b43a8 movq 0x18(%rsp), %rcx movq (%rcx), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %rax movl 0xe8(%rcx,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %r12d xorl %ecx, %ecx testl %r13d, %r13d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r13d movl %r12d, %r10d xorl %ecx, %ecx movq %r15, 0xa0(%rsp) movq %r13, 0xc0(%rsp) movl %r12d, 0xac(%rsp) cmpq %r13, %rcx je 0x2b43a8 movslq 0x2c(%rbx), %rax movslq 0x30(%rbx), %r9 movl 0x34(%rbx), %esi movq (%rbx), %rdi movq 0x10(%rbx), %r11 movq 0x40(%rbx), %r8 movq %r8, %rbp movq %rcx, 0x48(%rsp) imulq %rcx, %rbp imulq %r11, %rbp addq %rdi, %rbp movl 0x18(%rbx), %ecx movq 0x20(%rbx), %rdx movq %rbp, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %r11, 0x60(%rsp) movl %ecx, 0x68(%rsp) movq %rdx, 0x70(%rsp) movl %eax, 0x7c(%rsp) movl %r9d, 0x80(%rsp) movl $0x1, 0x84(%rsp) movl %esi, 0x88(%rsp) imulq %rax, %r9 movq %r11, %rax imulq %r9, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r11 movq %rax, 0x90(%rsp) movl 0x28(%rbx), %ecx leal -0x1(%rcx), %edx movl %edx, 0x78(%rsp) cmpl $0x4, %ecx jne 0x2b397b movq %r9, 0x90(%rsp) movq %r9, %rax movq 0x98(%rsp), %rcx movq -0x18(%rcx), %r9 movq 0x18(%rsp), %rcx vcvttss2si 0xe4(%rcx,%r9), %rcx movl %r10d, %r10d movq %r10, 0x30(%rsp) movq %rcx, %rdx shlq $0x8, %rdx movq %rcx, %r10 shlq $0x10, %r10 orq %rdx, %r10 movq %rcx, %rdx shlq $0x18, %rdx movq %rcx, %r14 shlq $0x20, %r14 orq %rdx, %r14 orq %r10, %r14 movq %rcx, %rdx shlq $0x28, %rdx movq %rcx, %r15 shlq $0x30, %r15 orq %rdx, %r15 movq %rcx, %r10 shlq $0x38, %r10 orq %r15, %r10 orq %r14, %r10 orq %rcx, %r10 movl %r12d, %ecx movl %r12d, %r13d movq 0x48(%rsp), %r12 addl %r12d, %ecx setns %dl cmpl 0xa0(%rsp), %ecx setl %r14b testb %r14b, %dl je 0x2b3bfc movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rdi movslq 0x30(%rsi), %r8 movl 0x34(%rsi), %eax movq (%rsi), %r13 movq 0x10(%rsi), %r11 movq 0x40(%rsi), %rdx movl %ecx, %r12d movq %rdx, 0x38(%rsp) imulq %rdx, %r12 imulq %r11, %r12 addq %r13, %r12 movl 0x18(%rsi), %ecx movq 0x20(%rsi), %rdx movq %r12, 0xd0(%rsp) andq $0x0, 0xd8(%rsp) movq %r11, 0xe0(%rsp) movl %ecx, 0xe8(%rsp) movq %rdx, 0xf0(%rsp) movl %edi, 0xfc(%rsp) movl %r8d, 0x100(%rsp) movl $0x1, 0x104(%rsp) movl %eax, 0x108(%rsp) movq %r8, 0xb8(%rsp) movq %r8, %rcx movq %rdi, 0x40(%rsp) imulq %rdi, %rcx movq %r11, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx movq %r11, 0xb0(%rsp) divq %r11 movq %rax, 0x110(%rsp) movl 0x28(%rsi), %eax leal -0x1(%rax), %edx movl %edx, 0xf8(%rsp) cmpl $0x4, %eax jne 0x2b3ad9 movq %rcx, 0x110(%rsp) movq 0x18(%rsp), %r14 movl 0xe0(%r14,%r9), %eax testl %eax, %eax jne 0x2b3b37 movl 0xd0(%r14,%r9), %edx movl 0xd4(%r14,%r9), %ecx movl 0xd8(%r14,%r9), %r8d movl 0xdc(%r14,%r9), %r9d movq %r10, (%rsp) leaq 0xd0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b46e0 movq (%r14), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %r9 movl 0xe0(%r14,%r9), %eax cmpl $0x1, %eax jne 0x2b3d4c movq 0x18(%rsp), %rax movl 0xd0(%rax,%r9), %r8d movl 0xd4(%rax,%r9), %ecx movl %ecx, 0xa8(%rsp) movl 0xd8(%rax,%r9), %edx xorl %r10d, %r10d testl %edx, %edx cmovlel %r10d, %edx movq 0x40(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r11d cmovgl %ecx, %r11d movl 0xdc(%rax,%r9), %esi testl %esi, %esi cmovlel %r10d, %esi testl %r8d, %r8d cmovlel %r10d, %r8d movq 0x38(%rsp), %r9 imulq 0xb0(%rsp), %r9 imulq 0x30(%rsp), %r9 addq %r13, %r9 movq %rbp, %rdi cmpl %r8d, %r10d je 0x2b3c36 movl %edx, %ecx subl $0x1, %ecx jb 0x2b3bca movq (%r12), %r14 movq %r14, (%rdi) addq $0x8, %rdi jmp 0x2b3bb8 xorl %r14d, %r14d cmpl %r14d, %r11d je 0x2b3be2 movq (%r9,%r14,8), %rcx movq %rcx, (%rdi) addq $0x8, %rdi incq %r14 jmp 0x2b3bcd movl %esi, %ecx subl $0x1, %ecx jb 0x2b3bf7 movq -0x8(%r9,%r14,8), %r15 movq %r15, (%rdi) addq $0x8, %rdi jmp 0x2b3be4 incl %r10d jmp 0x2b3bad imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi imulq %r11, %r8 imulq %r12, %r8 addq %r8, %rdi xorl %eax, %eax movq %r12, %rcx movl %r13d, %r12d movq 0xc0(%rsp), %r13 cmpq %rax, %rsi je 0x2b3f6a movq %r10, (%rdi,%rax,8) incq %rax jmp 0x2b3c24 movslq 0xfc(%rsp), %rax xorl %r11d, %r11d testl %eax, %eax movl $0x0, %r8d movq %rax, 0x120(%rsp) cmovgl %eax, %r8d movq 0xb8(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl %eax, %r14d movq %r12, %r9 cmpl %r14d, %r11d je 0x2b3cc8 movl %edx, %ecx subl $0x1, %ecx jb 0x2b3c84 movq (%r9), %r10 movq %r10, (%rdi) addq $0x8, %rdi jmp 0x2b3c73 xorl %r15d, %r15d xorl %r10d, %r10d cmpl %r15d, %r8d je 0x2b3ca0 movq (%r9,%r15,8), %rcx movq %rcx, (%rdi,%r15,8) addq $-0x8, %r10 incq %r15 jmp 0x2b3c8a movq %r9, %rcx subq %r10, %rcx subq %r10, %rdi movl %esi, %r10d subl $0x1, %r10d jb 0x2b3cc0 movq -0x8(%r9,%r15,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b3cac incl %r11d movq %rcx, %r9 jmp 0x2b3c6c movq 0x120(%rsp), %rax shlq $0x3, %rax subq %rax, %r9 xorl %r10d, %r10d movl 0xa8(%rsp), %r14d testl %r14d, %r14d cmovlel %r10d, %r14d cmpl %r14d, %r10d je 0x2b3d33 movl %edx, %ecx subl $0x1, %ecx jb 0x2b3d01 movq (%r9), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b3cf0 xorl %r11d, %r11d cmpl %r11d, %r8d je 0x2b3d19 movq (%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi incq %r11 jmp 0x2b3d04 movl %esi, %ecx subl $0x1, %ecx jb 0x2b3d2e movq -0x8(%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b3d1b incl %r10d jmp 0x2b3ce9 movq 0x98(%rsp), %rax movq -0x18(%rax), %r9 movq 0x18(%rsp), %rax movl 0xe0(%rax,%r9), %eax cmpl $0x2, %eax jne 0x2b3f55 movq 0x18(%rsp), %rcx movl 0xd0(%rcx,%r9), %r14d movslq 0xd8(%rcx,%r9), %rax movq %rcx, %r8 movl %r14d, %r11d movq 0x40(%rsp), %rcx imull %ecx, %r11d movq %rcx, %rdx negq %rdx xorl %r15d, %r15d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi testl %ecx, %ecx movl $0x0, %edi cmovgl %ecx, %edi movl 0xdc(%r8,%r9), %r8d testl %r8d, %r8d cmovlel %r15d, %r8d testl %r14d, %r14d cmovlel %r15d, %r14d movq 0x38(%rsp), %r10 imulq 0xb0(%rsp), %r10 movslq %r11d, %rcx leaq (%r12,%rcx,8), %r11 imulq 0x30(%rsp), %r10 leaq (%r10,%rax,8), %r10 leaq (%r10,%rcx,8), %r10 movq 0x18(%rsp), %r12 movl 0xd4(%r12,%r9), %r9d movl %r9d, 0x38(%rsp) shlq $0x3, %rax addq %r10, %r13 movq 0x40(%rsp), %rcx leaq (,%rcx,8), %r10 negq %r10 negq %rsi negq %r8 cmpl %r14d, %r15d je 0x2b3e5e xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b3e1c movq (%r13,%rcx,8), %r12 movq %r12, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3e05 xorl %ecx, %ecx movq %r11, %r12 cmpl %edi, %ecx je 0x2b3e39 movq (%r12), %r9 addq $0x8, %r12 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b3e21 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b3e52 movq -0x10(%r12,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3e3b leaq (%r11,%rdx,8), %r11 incl %r15d addq %r10, %r13 jmp 0x2b3dfe xorl %r14d, %r14d movq 0xb8(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r15d cmovgl %ecx, %r15d movl 0x38(%rsp), %r13d cmpl %r15d, %r14d je 0x2b3ed3 leaq (%r11,%rax), %rcx xorl %r12d, %r12d cmpq %r12, %rsi je 0x2b3e9c movq (%rcx,%r12,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %r12 jmp 0x2b3e86 xorl %ecx, %ecx cmpl %edi, %ecx je 0x2b3eb5 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b3e9e xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b3ece movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3eb7 incl %r14d jmp 0x2b3e7a movq 0x40(%rsp), %rcx addl %ecx, %ecx movslq %ecx, %rcx shlq $0x3, %rcx movq %r11, %r14 subq %rcx, %r14 xorl %r15d, %r15d testl %r13d, %r13d cmovlel %r15d, %r13d subq %rcx, %rax addq %r11, %rax cmpl %r13d, %r15d je 0x2b3f55 xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b3f14 movq (%rax,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3efe xorl %ecx, %ecx movq %r14, %r11 cmpl %edi, %ecx je 0x2b3f30 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b3f19 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b3f49 movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3f32 leaq (%r14,%rdx,8), %r14 incl %r15d addq %r10, %rax jmp 0x2b3ef7 movq 0xc0(%rsp), %r13 movl 0xac(%rsp), %r12d movq 0x48(%rsp), %rcx incq %rcx movq 0x30(%rsp), %r10 incl %r10d jmp 0x2b38d1 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b42dc vpextrd $0x3, %xmm0, %edi movl 0xe8(%r14,%rax), %ecx movl %edi, 0x48(%rsp) addl %edi, %ecx addl 0xec(%r14,%rax), %ecx addl 0xd0(%r14,%rax), %edx addl 0xd4(%r14,%rax), %edx addl 0xd8(%r14,%rax), %esi addl 0xdc(%r14,%rax), %esi movq 0x8(%r12), %rax movq %rax, 0x8(%rsp) movl $0x8, (%rsp) movq %rbx, %rdi movq %rcx, %r14 movl %r15d, %r8d callq 0x62a26 cmpq $0x0, (%rbx) je 0x2b43a8 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b43a8 xorl %ecx, %ecx testl %r14d, %r14d cmovlel %ecx, %r14d movq %r14, 0x30(%rsp) testl %r15d, %r15d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r15d movq %r15, 0xa0(%rsp) movabsq $0x100000001, %r13 # imm = 0x100000001 xorl %r14d, %r14d movq 0x18(%rsp), %rbp movq %rbx, 0xc8(%rsp) cmpq 0xa0(%rsp), %r14 je 0x2b43a8 movq (%rbp), %rax movq -0x18(%rax), %rax vcvttss2si 0xe4(%rbp,%rax), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rsi shlq $0x20, %rsi orq %rcx, %rsi orq %rdx, %rsi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r15 shlq $0x38, %r15 orq %rdx, %r15 orq %rsi, %r15 orq %rax, %r15 xorl %r12d, %r12d cmpq 0x30(%rsp), %r12 je 0x2b4282 movslq 0x2c(%rbx), %rdi movslq 0x30(%rbx), %rdx movq (%rbx), %rax movq 0x10(%rbx), %rsi movq 0x40(%rbx), %r8 imulq %r14, %r8 movq %r8, %r9 imulq %rsi, %r9 addq %rax, %r9 movl 0x18(%rbx), %r10d movq %rdx, %rcx imulq %rdi, %rcx movq %r12, %r11 imulq %rsi, %r11 imulq %rcx, %r11 addq %r9, %r11 movq 0x20(%rbx), %r9 movq %r11, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %rsi, 0x60(%rsp) movl %r10d, 0x68(%rsp) movq %r9, 0x70(%rsp) movl $0x2, 0x78(%rsp) movl %edi, 0x7c(%rsp) movl %edx, 0x80(%rsp) movq %r13, 0x84(%rsp) movq %rcx, 0x90(%rsp) movq (%rbp), %r9 movq -0x18(%r9), %r9 movl 0xe8(%rbp,%r9), %r11d movl %r12d, %r10d subl %r11d, %r10d setns %r11b cmpl 0x48(%rsp), %r10d movq %r13, %rbx movq %rbp, %r13 setl %bpl testb %bpl, %r11b je 0x2b4246 movq 0x28(%rsp), %r8 movslq 0x2c(%r8), %rax movslq 0x30(%r8), %rcx movq 0x40(%r8), %rdx imulq %r14, %rdx movq 0x10(%r8), %rsi imulq %rsi, %rdx addq (%r8), %rdx movl 0x18(%r8), %edi movq 0x20(%r8), %r8 movq %rcx, %r11 imulq %rax, %r11 movl %r10d, %r10d movq %r11, %rbp imulq %rsi, %rbp imulq %r10, %rbp addq %rdx, %rbp movq %rbp, 0xd0(%rsp) andq $0x0, 0xd8(%rsp) movq %rsi, 0xe0(%rsp) movl %edi, 0xe8(%rsp) movq %r8, 0xf0(%rsp) movl $0x2, 0xf8(%rsp) movl %eax, 0xfc(%rsp) movl %ecx, 0x100(%rsp) movq %rbx, 0x104(%rsp) movq %r11, 0x110(%rsp) movl 0xd0(%r13,%r9), %edx movl 0xd4(%r13,%r9), %ecx movl 0xd8(%r13,%r9), %r8d movl 0xdc(%r13,%r9), %r9d movq %r15, (%rsp) leaq 0xd0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b46e0 movq %r13, %rbp movq %rbx, %r13 movq 0xc8(%rsp), %rbx incq %r12 jmp 0x2b40b1 testl %ecx, %ecx movl $0x0, %r9d cmovlel %r9d, %ecx imulq %r12, %rdx imulq %rdi, %rdx addq %r8, %rdx imulq %rdx, %rsi addq %rsi, %rax xorl %edx, %edx movq %r13, %rbp movq %rbx, %r13 movq 0xc8(%rsp), %rbx cmpq %rdx, %rcx je 0x2b423e movq %r15, (%rax,%rdx,8) incq %rdx jmp 0x2b4274 incq %r14 jmp 0x2b4048 movq (%r14), %rax movq -0x18(%rax), %r8 movl 0xd0(%r14,%r8), %r10d leal (%r10,%rdx,8), %edx addl 0xd4(%r14,%r8), %edx xorl %r11d, %r11d testb $0x7, %dl sete %r11b movq %r9, %rax shrq $0x3, %rax leal (%r11,%r11,2), %ecx shlq %cl, %rax testb $0x7, %r10b sete %cl andb %cl, %r11b cmpb $0x1, %r11b jne 0x2b42dc cmpl $0x0, 0xe0(%r14,%r8) je 0x2b43bf movq 0x28(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm1 vmovaps %xmm1, 0x50(%rsp) movq %r9, 0x60(%rsp) movl %edi, 0x68(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x70(%rsp) vmovdqu %xmm0, 0x78(%rsp) movl %r15d, 0x88(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x90(%rsp) testq %rax, %rax je 0x2b4323 lock incl (%rax) cmpl $0x1, %edi je 0x2b435b vmovdqu64 (%r12), %zmm0 leaq 0xd0(%rsp), %rcx vmovdqu64 %zmm0, (%rcx) movq 0x10(%r12), %rax movq %rax, 0x8(%rcx) leaq 0x50(%rsp), %rsi pushq $0x1 popq %rdx movq 0x28(%rsp), %rdi vzeroupper callq 0x64e3b movq (%r14), %rax addq -0x18(%rax), %r14 leaq 0x50(%rsp), %rsi movq %r14, %rdi movq %rbx, %rdx movq %r12, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b43a8 lock decl (%rax) jne 0x2b43a8 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2b43a0 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b43a8 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0x128, %rsp # imm = 0x128 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq addl 0xd8(%r14,%r8), %esi addl 0xdc(%r14,%r8), %esi sarl $0x3, %edx movq 0x8(%r12), %r9 pushq $0x8 popq %rbp movq %rbx, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b43a8 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b43a8 movq 0x18(%rsp), %r9 movq (%r9), %rax movq -0x18(%rax), %rsi vcvttss2si 0xe4(%r9,%rsi), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rdi shlq $0x20, %rdi orq %rcx, %rdi orq %rdx, %rdi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r11 shlq $0x38, %r11 orq %rdx, %r11 orq %rdi, %r11 orq %rax, %r11 movl 0xd0(%r9,%rsi), %eax movl 0xd4(%r9,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%r9,%rsi), %r8d movl 0xdc(%r9,%rsi), %r9d movq %r11, (%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b46e0 movq $0x0, 0x20(%rsp) jmp 0x2b43a8 jmp 0x2b44e4 jmp 0x2b44ab movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b44dc lock decl (%rax) jne 0x2b44dc movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2b44d6 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b44dc movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_avx512.cpp
virtual thunk to ncnn::Padding_x86_avx512::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86_avx512::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2b0de0	/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_avx512.cpp
ncnn::padding_constant_pack8_int8_sse(ncnn::Mat const&, ncnn::Mat&, int, int, int, int, long)	static void padding_constant_pack8_int8_sse(const Mat& src, Mat& dst, int top, int bottom, int left, int right, int64_t _v) { const int64_t* ptr = src; int64_t* outptr = dst; // fill top for (int y = 0; y < top; y++) { for (int x = 0; x < dst.w; x++) { outptr++ = _v; } } // fill center for (int y = 0; y < src.h; y++) { for (int x = 0; x < left; x++) { outptr++ = _v; } for (int x = 0; x < src.w; x++) { outptr++ = ptr++; } for (int x = 0; x < right; x++) { outptr++ = _v; } } // fill bottom for (int y = 0; y < bottom; y++) { for (int x = 0; x < dst.w; x++) { outptr++ = _v; } } }	pushq %rbp pushq %r15 pushq %r14 pushq %r12 pushq %rbx movq 0x30(%rsp), %rax movq (%rdi), %r11 movl 0x2c(%rsi), %ebx xorl %ebp, %ebp testl %ebx, %ebx cmovlel %ebp, %ebx movq (%rsi), %r10 testl %edx, %edx cmovlel %ebp, %edx cmpl %edx, %ebp je 0x2b471c movl %ebx, %r14d subl $0x1, %r14d jb 0x2b4718 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b4709 incl %ebp jmp 0x2b4702 xorl %edx, %edx testl %r8d, %r8d cmovlel %edx, %r8d movl 0x2c(%rdi), %ebx testl %ebx, %ebx cmovlel %edx, %ebx testl %r9d, %r9d cmovlel %edx, %r9d movl 0x30(%rdi), %edi testl %edi, %edi cmovlel %edx, %edi cmpl %edi, %edx je 0x2b4788 movl %r8d, %ebp subl $0x1, %ebp jb 0x2b4751 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b4743 xorl %r15d, %r15d xorl %r14d, %r14d cmpl %r15d, %ebx je 0x2b476d movq (%r11,%r15,8), %r12 movq %r12, (%r10,%r15,8) addq $-0x8, %r14 incq %r15 jmp 0x2b4757 subq %r14, %r11 subq %r14, %r10 movl %r9d, %ebp subl $0x1, %ebp jb 0x2b4784 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b4776 incl %edx jmp 0x2b473c movl 0x2c(%rsi), %edx xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %ecx, %ecx cmovlel %esi, %ecx movl %edx, %edi cmpl %ecx, %esi je 0x2b47af subl $0x1, %edi jb 0x2b47ab movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b479d incl %esi jmp 0x2b4797 popq %rbx popq %r12 popq %r14 popq %r15 popq %rbp retq	/csukuangfj[P]ncnn/src/layer/x86/padding_pack8_int8.h
ncnn::Padding_x86_fma::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86_fma::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0xf8, %rsp movq %rcx, %rbp movq %rdx, 0x8(%rsp) movq %rsi, %r14 movq (%rdi), %rax movq -0x18(%rax), %r11 movl 0xd0(%rdi,%r11), %r10d testl %r10d, %r10d jne 0x2b486b cmpl $0x0, 0xd4(%rdi,%r11) jne 0x2b486b cmpl $0x0, 0xd8(%rdi,%r11) jne 0x2b486b cmpl $0x0, 0xdc(%rdi,%r11) jne 0x2b486b cmpl $0x0, 0xe8(%rdi,%r11) jne 0x2b486b cmpl $0x0, 0xec(%rdi,%r11) je 0x2b62f3 movq %rdi, (%rsp) movl 0x18(%r14), %r15d movq 0x10(%r14), %r9 testl %r15d, %r15d je 0x2b48b2 leal (,%r9,8), %eax cltd idivl %r15d cmpl $0x8, %eax jne 0x2b48c4 movq (%rsp), %rdi movq %r14, %rsi movq 0x8(%rsp), %rdx movq %rbp, %rcx addq $0xf8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x2b63fc vmovdqu 0x2c(%r14), %xmm0 movq %r14, %rdi movl 0x28(%r14), %r13d jmp 0x2b4b30 vmovdqu 0x2c(%r14), %xmm0 movq %r14, 0x10(%rsp) movl 0x28(%r14), %eax vpextrd $0x2, %xmm0, %r12d pushq $0x4 popq %r13 vpextrd $0x1, %xmm0, %edx vmovd %xmm0, %esi vpextrd $0x3, %xmm0, %ecx movq %rcx, 0x18(%rsp) pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) cmpl $0x4, %r15d je 0x2b4a16 cmpl $0x8, %r15d jne 0x2b4b28 leal -0x1(%rax), %ecx cmpl $0x3, %ecx ja 0x2b4b28 leaq 0x143ec3(%rip), %rax # 0x3f87e4 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx pushq $0x8 popq %r14 jmpq %rcx movq (%rsp), %rdi movl 0xd8(%rdi,%r11), %edx leal (%rdx,%rsi,8), %eax addl 0xdc(%rdi,%r11), %eax testb $0x3, %al sete %cl movq %r9, %r10 shrq $0x3, %r10 addb %cl, %cl testb $0x7, %al movzbl %cl, %esi pushq $0x3 popq %rcx cmovnel %esi, %ecx shlq %cl, %r10 orl %eax, %edx pushq $0x1 popq %r13 testb $0x7, %dl jne 0x2b4b2b cmpl $0x0, 0xe0(%rdi,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 cltd idivl %r14d movq 0x8(%rbp), %r8 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %esi movq %r10, %rdx movl %r14d, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vbroadcastss 0xe4(%rdx,%rcx), %ymm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %r14d movl %eax, %r8d movl %ecx, %eax cltd idivl %r14d movq $0x0, 0x20(%rsp) movq 0x10(%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b72c5 jmp 0x2b4c17 leal -0x1(%rax), %ecx cmpl $0x3, %ecx ja 0x2b4b28 leaq 0x143dab(%rip), %rax # 0x3f87d4 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx jmpq %rcx movq (%rsp), %rax movl 0xd8(%rax,%r11), %r8d leal (%r8,%rsi,4), %esi addl 0xdc(%rax,%r11), %esi testb $0x3, %sil sete %cl movq %r9, %rdx shrq $0x2, %rdx addb %cl, %cl movl %esi, %eax andl $0x7, %eax movzbl %cl, %r10d pushq $0x3 popq %rcx cmovnel %r10d, %ecx shlq %cl, %rdx pushq $0x1 popq %r13 testb $0x3, %r8b jne 0x2b4b2b cmpl $0x4, %eax jne 0x2b4b2b movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 sarl $0x2, %esi movq 0x8(%rbp), %r8 pushq $0x4 popq %rbp movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vbroadcastss 0xe4(%rdx,%rcx), %xmm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq $0x0, 0x20(%rsp) movq 0x10(%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b7365 jmp 0x2b4c17 movl %eax, %r13d movq 0x10(%rsp), %rdi movq 0x8(%rdi), %rax vmovups (%rdi), %xmm1 vmovaps %xmm1, 0x80(%rsp) movq %r9, 0x90(%rsp) movl %r15d, 0x98(%rsp) movq 0x20(%rdi), %rcx movq %rcx, 0xa0(%rsp) movl %r13d, 0xa8(%rsp) vmovdqu %xmm0, 0xac(%rsp) movq 0x40(%rdi), %rcx movq %rcx, 0xc0(%rsp) testq %rax, %rax je 0x2b4b82 lock incl (%rax) cmpl $0x1, %r15d je 0x2b4bbb vmovdqu (%rbp), %ymm0 vmovups 0x20(%rbp), %ymm1 leaq 0x30(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovdqu %ymm0, (%rcx) movq 0x10(%rbp), %rax movq %rax, 0x8(%rcx) leaq 0x80(%rsp), %rsi pushq $0x1 popq %rdx vzeroupper callq 0x64e3b movq (%rsp), %rdi movq (%rdi), %rax addq -0x18(%rax), %rdi leaq 0x80(%rsp), %rsi movq 0x8(%rsp), %rdx movq %rbp, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x88(%rsp), %rax testq %rax, %rax je 0x2b4c17 lock decl (%rax) jne 0x2b4c17 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b4c0f movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b4c17 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0xf8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq pushq $0x3 popq %r13 movq (%rsp), %rdi movl 0xe8(%rdi,%r11), %r14d movq 0x18(%rsp), %r12 leal (%r14,%r12,8), %eax addl 0xec(%rdi,%r11), %eax testb $0x3, %al sete %cl movq %r9, %r8 shrq $0x3, %r8 addb %cl, %cl testb $0x7, %al movzbl %cl, %ecx cmovel %r13d, %ecx shlq %cl, %r8 orl %eax, %r14d testb $0x7, %r14b jne 0x2b4b2b addl 0xd8(%rdi,%r11), %esi addl 0xdc(%rdi,%r11), %esi addl %edx, %r10d addl 0xd4(%rdi,%r11), %r10d leal (,%r12,8), %ecx cmpl %ecx, %eax movq 0x10(%rsp), %rdi pushq $0x8 popq %r14 je 0x2b4cbc movq (%rsp), %rcx cmpl $0x0, 0xe0(%rcx,%r11) jne 0x2b4b30 cltd idivl %r14d subq $0x8, %rsp movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r14d, %r9d movq %rax, %r14 movl %eax, %ecx pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xf0(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xf0(%rsp), %r9 je 0x2b4c17 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b4e05 movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b4e2f movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x5, %r8 vmovups (%rdx,%r8), %ymm0 jmp 0x2b4e39 vbroadcastss 0xe4(%r12,%rdi), %ymm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b4ff5 movq 0x10(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b4ef1 movq %rsi, 0x70(%rsp) movq (%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b4f4c movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b72c5 movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b501f movq (%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b5028 vmovdqa (%rbp), %ymm0 movl %eax, %edi subl $0x1, %edi jb 0x2b4fb8 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b4fa7 xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b4fda vmovdqa (%r12,%rdi), %ymm0 vmovdqa %ymm0, (%r13,%rdi) incl %r10d addq $0x20, %rdi jmp 0x2b4fbd addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b4ff0 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b4fdf incl %r8d jmp 0x2b4f97 imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b53c3 vmovups %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b500a movq (%rsp), %r12 jmp 0x2b510b xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b5085 vmovdqa (%rbp), %ymm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b504a vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b5038 xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b506c vmovdqa (%rbp), %ymm0 vmovdqa %ymm0, (%r13) addq $0x20, %rbp addq $0x20, %r13 incl %r8d jmp 0x2b504d movl %edx, %r8d subl $0x1, %r8d jb 0x2b5081 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b506f incl %edi jmp 0x2b502a movl 0x5c(%rsp), %edi shll $0x3, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b50fb vmovdqa (%rbp), %ymm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b50bb vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b50a9 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b50df vmovdqa (%rbp,%r8), %ymm0 vmovdqa %ymm0, (%r13,%r8) incl %r9d addq $0x20, %r8 jmp 0x2b50c1 addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b50f7 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b50e5 incl %edi jmp 0x2b509d movq -0x18(%rcx), %rdi movq (%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x40 popq %r13 jne 0x2b532b movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x3, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x5, %rax shlq $0x5, %rdx shlq $0x5, %rsi cmpl %r9d, %r10d je 0x2b520c leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b51ae vmovdqa (%r11,%r14), %ymm0 vmovdqa %ymm0, (%r8,%r15) addq $-0x20, %r14 addq $0x20, %r15 jmp 0x2b5190 subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b51d2 vmovdqa (%r11), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r11 addq $0x20, %r8 incl %ebp jmp 0x2b51b6 movq %r13, %r14 movq (%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x40, %r15 je 0x2b51f8 vmovdqa (%r11,%r14), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r14 jmp 0x2b51d9 movslq 0x5c(%rsp), %r11 shlq $0x5, %r11 subq %r11, %rcx incl %r10d jmp 0x2b517d xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b5288 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b523e vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8,%r14) addq $-0x20, %r11 addq $0x20, %r14 jmp 0x2b5220 subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b5261 vmovdqa (%rcx), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %rcx addq $0x20, %r8 incl %r10d jmp 0x2b5244 movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x40, %r11 je 0x2b5283 vmovdqa (%rcx,%r10), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r10 jmp 0x2b5264 incl %r9d jmp 0x2b520f movslq 0x5c(%rsp), %r9 shlq $0x6, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b532b leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b52ce vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8,%r14) addq $-0x20, %r11 addq $0x20, %r14 jmp 0x2b52b0 subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b52f5 vmovdqa (%r10), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r10 addq $0x20, %r8 incl %r11d jmp 0x2b52d7 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x40, %r14 je 0x2b5317 vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r11 jmp 0x2b52f8 movslq 0x5c(%rsp), %r10 shlq $0x5, %r10 subq %r10, %rcx incl %r9d jmp 0x2b529d movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b535f lock decl (%rax) jne 0x2b535f movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b5354 movq (%rdi), %rax vzeroupper callq 0x18(%rax) jmp 0x2b535f movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b53c3 lock decl (%rax) jne 0x2b53c3 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b53ab movq (%rdi), %rax vzeroupper callq 0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b53c3 movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b4d45 movq (%rsp), %rax movl 0xe8(%rax,%r11), %r14d movq 0x18(%rsp), %rdi leal (%r14,%rdi,4), %r13d addl 0xec(%rax,%r11), %r13d testb $0x3, %r13b sete %cl movq %r9, %r8 shrq $0x2, %r8 addb %cl, %cl movq %r13, %r12 movl %r13d, %eax andl $0x7, %eax movzbl %cl, %ecx pushq $0x3 popq %r13 cmovel %r13d, %ecx shlq %cl, %r8 testb $0x3, %r14b jne 0x2b4b2b cmpl $0x4, %eax jne 0x2b4b2b movq (%rsp), %rax addl 0xd8(%rax,%r11), %esi addl 0xdc(%rax,%r11), %esi addl %edx, %r10d addl 0xd4(%rax,%r11), %r10d leal (,%rdi,4), %eax cmpl %eax, %r12d movq 0x10(%rsp), %rdi je 0x2b5468 movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%r11) jne 0x2b4b30 movq %r12, %rcx sarl $0x2, %ecx subq $0x8, %rsp pushq $0x4 popq %r9 movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movq %rcx, %r14 pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x4 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xf0(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xf0(%rsp), %r9 je 0x2b4c17 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b55b2 movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b55dc movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x4, %r8 vmovups (%rdx,%r8), %xmm0 jmp 0x2b55e6 vbroadcastss 0xe4(%r12,%rdi), %xmm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b57a2 movq 0x10(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b569e movq %rsi, 0x70(%rsp) movq (%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b56f9 movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b7365 movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b57cc movq (%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b57d5 vmovdqa (%rbp), %xmm0 movl %eax, %edi subl $0x1, %edi jb 0x2b5765 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b5754 xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b5787 vmovdqa (%r12,%rdi), %xmm0 vmovdqa %xmm0, (%r13,%rdi) incl %r10d addq $0x10, %rdi jmp 0x2b576a addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b579d vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b578c incl %r8d jmp 0x2b5744 imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b5b64 vmovups %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b57b7 movq (%rsp), %r12 jmp 0x2b58b8 xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b5832 vmovdqa (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b57f7 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b57e5 xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b5819 vmovdqa (%rbp), %xmm0 vmovdqa %xmm0, (%r13) addq $0x10, %rbp addq $0x10, %r13 incl %r8d jmp 0x2b57fa movl %edx, %r8d subl $0x1, %r8d jb 0x2b582e vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b581c incl %edi jmp 0x2b57d7 movl 0x5c(%rsp), %edi shll $0x2, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b58a8 vmovdqa (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b5868 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b5856 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b588c vmovdqa (%rbp,%r8), %xmm0 vmovdqa %xmm0, (%r13,%r8) incl %r9d addq $0x10, %r8 jmp 0x2b586e addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b58a4 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b5892 incl %edi jmp 0x2b584a movq -0x18(%rcx), %rdi movq (%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x20 popq %r13 jne 0x2b5ad8 movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x2, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x4, %rax shlq $0x4, %rdx shlq $0x4, %rsi cmpl %r9d, %r10d je 0x2b59b9 leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b595b vmovdqa (%r11,%r14), %xmm0 vmovdqa %xmm0, (%r8,%r15) addq $-0x10, %r14 addq $0x10, %r15 jmp 0x2b593d subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b597f vmovdqa (%r11), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r11 addq $0x10, %r8 incl %ebp jmp 0x2b5963 movq %r13, %r14 movq (%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x20, %r15 je 0x2b59a5 vmovdqa (%r11,%r14), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r14 jmp 0x2b5986 movslq 0x5c(%rsp), %r11 shlq $0x4, %r11 subq %r11, %rcx incl %r10d jmp 0x2b592a xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b5a35 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b59eb vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2b59cd subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b5a0e vmovdqa (%rcx), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %rcx addq $0x10, %r8 incl %r10d jmp 0x2b59f1 movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x20, %r11 je 0x2b5a30 vmovdqa (%rcx,%r10), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r10 jmp 0x2b5a11 incl %r9d jmp 0x2b59bc movslq 0x5c(%rsp), %r9 shlq $0x5, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b5ad8 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b5a7b vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2b5a5d subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b5aa2 vmovdqa (%r10), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r10 addq $0x10, %r8 incl %r11d jmp 0x2b5a84 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x20, %r14 je 0x2b5ac4 vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r11 jmp 0x2b5aa5 movslq 0x5c(%rsp), %r10 shlq $0x4, %r10 subq %r10, %rcx incl %r9d jmp 0x2b5a4a movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b5b06 lock decl (%rax) jne 0x2b5b06 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b5afe movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b5b06 movq %rsi, %rdi callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b5b64 lock decl (%rax) jne 0x2b5b64 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b5b4f movq (%rdi), %rax callq 0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b5b64 movq %rsi, %rdi callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b54f2 movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 movq (%rsp), %rax movl 0xe8(%rax,%r11), %r13d addl %r12d, %r13d addl 0xec(%rax,%r11), %r13d addl %edx, %r10d addl 0xd4(%rax,%r11), %r10d addl 0xd8(%rax,%r11), %esi addl 0xdc(%rax,%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x8 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %r11 # imm = 0x100000001 xorl %r14d, %r14d movq (%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b4c17 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b5c61 movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x5, %rcx vmovups (%rax,%rcx), %ymm0 jmp 0x2b5c6b vbroadcastss 0xe4(%rbp,%rax), %ymm0 xorl %r15d, %r15d vmovups %ymm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b5e1d movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %r11, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b5e04 movq 0x10(%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %r13, %rbx movq %r11, %r13 movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq (%rsp), %rbp movq %r13, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovups 0xd0(%rsp), %ymm0 callq 0x2b72c5 vmovups 0xd0(%rsp), %ymm0 movq %r13, %r11 movq %rbx, %r13 incq %r15 jmp 0x2b5c77 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b5dfc vmovups %ymm0, (%rcx) addq $0x20, %rcx jmp 0x2b5e0e incq %r14 jmp 0x2b5c2e leal (%r10,%rdx,8), %eax movq (%rsp), %rdi addl 0xd4(%rdi,%r11), %eax testb $0x3, %al sete %cl movq %r9, %r8 shrq $0x3, %r8 addb %cl, %cl testb $0x7, %al movzbl %cl, %edx pushq $0x3 popq %rcx cmovnel %edx, %ecx shlq %cl, %r8 orl %eax, %r10d pushq $0x2 popq %r13 testb $0x7, %r10b jne 0x2b4b2b cmpl $0x0, 0xe0(%rdi,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 movq (%rsp), %rcx addl 0xd8(%rcx,%r11), %esi addl 0xdc(%rcx,%r11), %esi cltd idivl %r14d movq 0x8(%rbp), %r9 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %edx movq %r8, %rcx movl %r14d, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vbroadcastss 0xe4(%rdi,%rsi), %ymm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %r14d movl %eax, %r10d movl %ecx, %eax cltd idivl %r14d movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq 0x10(%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b72c5 jmp 0x2b62e5 movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 movq (%rsp), %rax movl 0xe8(%rax,%r11), %r13d addl %r12d, %r13d addl 0xec(%rax,%r11), %r13d addl %edx, %r10d addl 0xd4(%rax,%r11), %r10d addl 0xd8(%rax,%r11), %esi addl 0xdc(%rax,%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x4 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %r11 # imm = 0x100000001 xorl %r14d, %r14d movq (%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b4c17 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b6013 movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x4, %rcx vmovups (%rax,%rcx), %xmm0 jmp 0x2b601d vbroadcastss 0xe4(%rbp,%rax), %xmm0 xorl %r15d, %r15d vmovaps %xmm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b61d5 movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %r11, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b61bc movq 0x10(%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %r13, %rbx movl %r12d, %r13d movq %r11, %r12 movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq (%rsp), %rbp movq %r12, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovaps 0xd0(%rsp), %xmm0 callq 0x2b7365 vmovaps 0xd0(%rsp), %xmm0 movq %r12, %r11 movl %r13d, %r12d movq %rbx, %r13 incq %r15 jmp 0x2b6029 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b61b4 vmovups %xmm0, (%rcx) addq $0x10, %rcx jmp 0x2b61c6 incq %r14 jmp 0x2b5fe0 leal (%r10,%rdx,4), %edx movq (%rsp), %rax addl 0xd4(%rax,%r11), %edx testb $0x3, %dl sete %cl movq %r9, %rax shrq $0x2, %rax addb %cl, %cl movl %edx, %r8d andl $0x7, %r8d movzbl %cl, %r14d pushq $0x3 popq %rcx cmovnel %r14d, %ecx shlq %cl, %rax pushq $0x2 popq %r13 testb $0x3, %r10b jne 0x2b4b2b cmpl $0x4, %r8d jne 0x2b4b2b movq (%rsp), %rcx cmpl $0x0, 0xe0(%rcx,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 movq (%rsp), %rcx addl 0xd8(%rcx,%r11), %esi addl 0xdc(%rcx,%r11), %esi sarl $0x2, %edx movq 0x8(%rbp), %r9 pushq $0x4 popq %rbp movq 0x8(%rsp), %rbx movq %rbx, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vbroadcastss 0xe4(%rdi,%rsi), %xmm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq 0x10(%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b7365 movq $0x0, 0x20(%rsp) jmp 0x2b4c17 movq $0x0, 0x20(%rsp) cmpq %r14, 0x8(%rsp) je 0x2b4c17 movq 0x8(%r14), %rax testq %rax, %rax je 0x2b6313 lock incl (%rax) movq 0x8(%rsp), %rax movq 0x8(%rax), %rax testq %rax, %rax je 0x2b6347 lock decl (%rax) jne 0x2b6347 movq 0x8(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x2b633f movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b6347 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx andq $0x0, 0x40(%rcx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, (%rcx) vmovups %xmm0, 0xc(%rcx) andl $0x0, 0x38(%rcx) vmovups %xmm0, 0x28(%rcx) vmovups (%r14), %xmm0 vmovups %xmm0, (%rcx) movq 0x10(%r14), %rax movq %rax, 0x10(%rcx) movl 0x18(%r14), %eax movl %eax, 0x18(%rcx) movq 0x20(%r14), %rax movq %rax, 0x20(%rcx) vmovdqu 0x28(%r14), %xmm0 vmovdqu %xmm0, 0x28(%rcx) movl 0x38(%r14), %eax movl %eax, 0x38(%rcx) movq 0x40(%r14), %rax movq %rax, 0x40(%rcx) jmp 0x2b4c17 jmp 0x2b63f4 jmp 0x2b63f4 jmp 0x2b63f4 jmp 0x2b63f4 jmp 0x2b63f4 jmp 0x2b63b2 movq %rax, %rbx movq 0x88(%rsp), %rax testq %rax, %rax je 0x2b63ec lock decl (%rax) jne 0x2b63ec movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi jne 0x2b63e6 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b63ec movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_fma.cpp
ncnn::Padding_x86_fma::forward_int8(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86_fma::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_int8_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_int8_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 \| (v8 << 8) \| (v8 << 16) \| (v8 << 24) \| (v8 << 32) \| (v8 << 40) \| (v8 << 48) \| (v8 << 56); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x128, %rsp # imm = 0x128 movq %rcx, %r12 movq %rdx, %rbx movq %rdi, %r14 movl 0x38(%rsi), %r15d movq 0x10(%rsi), %r9 movl 0x18(%rsi), %edi movq %rsi, 0x28(%rsp) vmovdqu 0x28(%rsi), %xmm0 cmpl $0x8, %edi jne 0x2b70ac vmovd %xmm0, %eax decl %eax cmpl $0x3, %eax ja 0x2b70ac leaq 0x1423aa(%rip), %rcx # 0x3f87f4 movslq (%rcx,%rax,4), %rax vpextrd $0x1, %xmm0, %esi addq %rcx, %rax vpextrd $0x2, %xmm0, %edx pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) movq %r14, 0x18(%rsp) jmpq %rax movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %r8d leal (%r8,%rsi,8), %esi addl 0xdc(%r14,%rax), %esi xorl %r10d, %r10d testb $0x7, %sil sete %r10b movq %r9, %rdx shrq $0x3, %rdx leal (%r10,%r10,2), %ecx shlq %cl, %rdx testb $0x7, %r8b sete %cl andb %cl, %r10b cmpb $0x1, %r10b jne 0x2b70ac cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b70ac sarl $0x3, %esi movq 0x8(%r12), %r8 pushq $0x8 popq %rbp movq %rbx, %rdi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b7181 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b7181 movq 0x18(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rcx vcvttss2si 0xe4(%r8,%rcx), %rax movq %rax, %rdx shlq $0x8, %rdx movq %rax, %rsi shlq $0x10, %rsi orq %rdx, %rsi movq %rax, %rdx shlq $0x18, %rdx movq %rax, %rdi shlq $0x20, %rdi orq %rdx, %rdi orq %rsi, %rdi movq %rax, %rdx shlq $0x28, %rdx movq %rax, %rsi shlq $0x30, %rsi orq %rdx, %rsi movq %rax, %r9 shlq $0x38, %r9 orq %rsi, %r9 orq %rdi, %r9 orq %rax, %r9 movl 0xd8(%r8,%rcx), %eax movl 0xdc(%r8,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq %r9, (%rsp) movq $0x0, 0x20(%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b740e jmp 0x2b7181 movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %r10d leal (%r10,%r15,8), %r13d addl 0xec(%r14,%rax), %r13d xorl %r11d, %r11d testb $0x7, %r13b sete %r11b movq %r9, %r8 shrq $0x3, %r8 leal (%r11,%r11,2), %ecx shlq %cl, %r8 testb $0x7, %r10b sete %cl andb %cl, %r11b cmpb $0x1, %r11b jne 0x2b70ac addl 0xd8(%r14,%rax), %esi addl 0xdc(%r14,%rax), %esi addl 0xd0(%r14,%rax), %edx addl 0xd4(%r14,%rax), %edx leal (,%r15,8), %ecx cmpl %ecx, %r13d je 0x2b6615 cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b70ac sarl $0x3, %r13d movq 0x8(%r12), %rax movq %rax, (%rsp) pushq $0x8 popq %r9 movq %rbx, %rdi movl %r13d, %ecx callq 0x628f2 cmpq $0x0, (%rbx) je 0x2b7181 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b7181 movq 0x18(%rsp), %rcx movq (%rcx), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %rax movl 0xe8(%rcx,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %r12d xorl %ecx, %ecx testl %r13d, %r13d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r13d movl %r12d, %r10d xorl %ecx, %ecx movq %r15, 0xa0(%rsp) movq %r13, 0xc0(%rsp) movl %r12d, 0xac(%rsp) cmpq %r13, %rcx je 0x2b7181 movslq 0x2c(%rbx), %rax movslq 0x30(%rbx), %r9 movl 0x34(%rbx), %esi movq (%rbx), %rdi movq 0x10(%rbx), %r11 movq 0x40(%rbx), %r8 movq %r8, %rbp movq %rcx, 0x48(%rsp) imulq %rcx, %rbp imulq %r11, %rbp addq %rdi, %rbp movl 0x18(%rbx), %ecx movq 0x20(%rbx), %rdx movq %rbp, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %r11, 0x60(%rsp) movl %ecx, 0x68(%rsp) movq %rdx, 0x70(%rsp) movl %eax, 0x7c(%rsp) movl %r9d, 0x80(%rsp) movl $0x1, 0x84(%rsp) movl %esi, 0x88(%rsp) imulq %rax, %r9 movq %r11, %rax imulq %r9, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r11 movq %rax, 0x90(%rsp) movl 0x28(%rbx), %ecx leal -0x1(%rcx), %edx movl %edx, 0x78(%rsp) cmpl $0x4, %ecx jne 0x2b674b movq %r9, 0x90(%rsp) movq %r9, %rax movq 0x98(%rsp), %rcx movq -0x18(%rcx), %r9 movq 0x18(%rsp), %rcx vcvttss2si 0xe4(%rcx,%r9), %rcx movl %r10d, %r10d movq %r10, 0x30(%rsp) movq %rcx, %rdx shlq $0x8, %rdx movq %rcx, %r10 shlq $0x10, %r10 orq %rdx, %r10 movq %rcx, %rdx shlq $0x18, %rdx movq %rcx, %r14 shlq $0x20, %r14 orq %rdx, %r14 orq %r10, %r14 movq %rcx, %rdx shlq $0x28, %rdx movq %rcx, %r15 shlq $0x30, %r15 orq %rdx, %r15 movq %rcx, %r10 shlq $0x38, %r10 orq %r15, %r10 orq %r14, %r10 orq %rcx, %r10 movl %r12d, %ecx movl %r12d, %r13d movq 0x48(%rsp), %r12 addl %r12d, %ecx setns %dl cmpl 0xa0(%rsp), %ecx setl %r14b testb %r14b, %dl je 0x2b69cc movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rdi movslq 0x30(%rsi), %r8 movl 0x34(%rsi), %eax movq (%rsi), %r13 movq 0x10(%rsi), %r11 movq 0x40(%rsi), %rdx movl %ecx, %r12d movq %rdx, 0x38(%rsp) imulq %rdx, %r12 imulq %r11, %r12 addq %r13, %r12 movl 0x18(%rsi), %ecx movq 0x20(%rsi), %rdx movq %r12, 0xd0(%rsp) andq $0x0, 0xd8(%rsp) movq %r11, 0xe0(%rsp) movl %ecx, 0xe8(%rsp) movq %rdx, 0xf0(%rsp) movl %edi, 0xfc(%rsp) movl %r8d, 0x100(%rsp) movl $0x1, 0x104(%rsp) movl %eax, 0x108(%rsp) movq %r8, 0xb8(%rsp) movq %r8, %rcx movq %rdi, 0x40(%rsp) imulq %rdi, %rcx movq %r11, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx movq %r11, 0xb0(%rsp) divq %r11 movq %rax, 0x110(%rsp) movl 0x28(%rsi), %eax leal -0x1(%rax), %edx movl %edx, 0xf8(%rsp) cmpl $0x4, %eax jne 0x2b68a9 movq %rcx, 0x110(%rsp) movq 0x18(%rsp), %r14 movl 0xe0(%r14,%r9), %eax testl %eax, %eax jne 0x2b6907 movl 0xd0(%r14,%r9), %edx movl 0xd4(%r14,%r9), %ecx movl 0xd8(%r14,%r9), %r8d movl 0xdc(%r14,%r9), %r9d movq %r10, (%rsp) leaq 0xd0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b740e movq (%r14), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %r9 movl 0xe0(%r14,%r9), %eax cmpl $0x1, %eax jne 0x2b6b1c movq 0x18(%rsp), %rax movl 0xd0(%rax,%r9), %r8d movl 0xd4(%rax,%r9), %ecx movl %ecx, 0xa8(%rsp) movl 0xd8(%rax,%r9), %edx xorl %r10d, %r10d testl %edx, %edx cmovlel %r10d, %edx movq 0x40(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r11d cmovgl %ecx, %r11d movl 0xdc(%rax,%r9), %esi testl %esi, %esi cmovlel %r10d, %esi testl %r8d, %r8d cmovlel %r10d, %r8d movq 0x38(%rsp), %r9 imulq 0xb0(%rsp), %r9 imulq 0x30(%rsp), %r9 addq %r13, %r9 movq %rbp, %rdi cmpl %r8d, %r10d je 0x2b6a06 movl %edx, %ecx subl $0x1, %ecx jb 0x2b699a movq (%r12), %r14 movq %r14, (%rdi) addq $0x8, %rdi jmp 0x2b6988 xorl %r14d, %r14d cmpl %r14d, %r11d je 0x2b69b2 movq (%r9,%r14,8), %rcx movq %rcx, (%rdi) addq $0x8, %rdi incq %r14 jmp 0x2b699d movl %esi, %ecx subl $0x1, %ecx jb 0x2b69c7 movq -0x8(%r9,%r14,8), %r15 movq %r15, (%rdi) addq $0x8, %rdi jmp 0x2b69b4 incl %r10d jmp 0x2b697d imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi imulq %r11, %r8 imulq %r12, %r8 addq %r8, %rdi xorl %eax, %eax movq %r12, %rcx movl %r13d, %r12d movq 0xc0(%rsp), %r13 cmpq %rax, %rsi je 0x2b6d3a movq %r10, (%rdi,%rax,8) incq %rax jmp 0x2b69f4 movslq 0xfc(%rsp), %rax xorl %r11d, %r11d testl %eax, %eax movl $0x0, %r8d movq %rax, 0x120(%rsp) cmovgl %eax, %r8d movq 0xb8(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl %eax, %r14d movq %r12, %r9 cmpl %r14d, %r11d je 0x2b6a98 movl %edx, %ecx subl $0x1, %ecx jb 0x2b6a54 movq (%r9), %r10 movq %r10, (%rdi) addq $0x8, %rdi jmp 0x2b6a43 xorl %r15d, %r15d xorl %r10d, %r10d cmpl %r15d, %r8d je 0x2b6a70 movq (%r9,%r15,8), %rcx movq %rcx, (%rdi,%r15,8) addq $-0x8, %r10 incq %r15 jmp 0x2b6a5a movq %r9, %rcx subq %r10, %rcx subq %r10, %rdi movl %esi, %r10d subl $0x1, %r10d jb 0x2b6a90 movq -0x8(%r9,%r15,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b6a7c incl %r11d movq %rcx, %r9 jmp 0x2b6a3c movq 0x120(%rsp), %rax shlq $0x3, %rax subq %rax, %r9 xorl %r10d, %r10d movl 0xa8(%rsp), %r14d testl %r14d, %r14d cmovlel %r10d, %r14d cmpl %r14d, %r10d je 0x2b6b03 movl %edx, %ecx subl $0x1, %ecx jb 0x2b6ad1 movq (%r9), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b6ac0 xorl %r11d, %r11d cmpl %r11d, %r8d je 0x2b6ae9 movq (%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi incq %r11 jmp 0x2b6ad4 movl %esi, %ecx subl $0x1, %ecx jb 0x2b6afe movq -0x8(%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b6aeb incl %r10d jmp 0x2b6ab9 movq 0x98(%rsp), %rax movq -0x18(%rax), %r9 movq 0x18(%rsp), %rax movl 0xe0(%rax,%r9), %eax cmpl $0x2, %eax jne 0x2b6d25 movq 0x18(%rsp), %rcx movl 0xd0(%rcx,%r9), %r14d movslq 0xd8(%rcx,%r9), %rax movq %rcx, %r8 movl %r14d, %r11d movq 0x40(%rsp), %rcx imull %ecx, %r11d movq %rcx, %rdx negq %rdx xorl %r15d, %r15d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi testl %ecx, %ecx movl $0x0, %edi cmovgl %ecx, %edi movl 0xdc(%r8,%r9), %r8d testl %r8d, %r8d cmovlel %r15d, %r8d testl %r14d, %r14d cmovlel %r15d, %r14d movq 0x38(%rsp), %r10 imulq 0xb0(%rsp), %r10 movslq %r11d, %rcx leaq (%r12,%rcx,8), %r11 imulq 0x30(%rsp), %r10 leaq (%r10,%rax,8), %r10 leaq (%r10,%rcx,8), %r10 movq 0x18(%rsp), %r12 movl 0xd4(%r12,%r9), %r9d movl %r9d, 0x38(%rsp) shlq $0x3, %rax addq %r10, %r13 movq 0x40(%rsp), %rcx leaq (,%rcx,8), %r10 negq %r10 negq %rsi negq %r8 cmpl %r14d, %r15d je 0x2b6c2e xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b6bec movq (%r13,%rcx,8), %r12 movq %r12, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6bd5 xorl %ecx, %ecx movq %r11, %r12 cmpl %edi, %ecx je 0x2b6c09 movq (%r12), %r9 addq $0x8, %r12 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b6bf1 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b6c22 movq -0x10(%r12,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6c0b leaq (%r11,%rdx,8), %r11 incl %r15d addq %r10, %r13 jmp 0x2b6bce xorl %r14d, %r14d movq 0xb8(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r15d cmovgl %ecx, %r15d movl 0x38(%rsp), %r13d cmpl %r15d, %r14d je 0x2b6ca3 leaq (%r11,%rax), %rcx xorl %r12d, %r12d cmpq %r12, %rsi je 0x2b6c6c movq (%rcx,%r12,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %r12 jmp 0x2b6c56 xorl %ecx, %ecx cmpl %edi, %ecx je 0x2b6c85 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b6c6e xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b6c9e movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6c87 incl %r14d jmp 0x2b6c4a movq 0x40(%rsp), %rcx addl %ecx, %ecx movslq %ecx, %rcx shlq $0x3, %rcx movq %r11, %r14 subq %rcx, %r14 xorl %r15d, %r15d testl %r13d, %r13d cmovlel %r15d, %r13d subq %rcx, %rax addq %r11, %rax cmpl %r13d, %r15d je 0x2b6d25 xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b6ce4 movq (%rax,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6cce xorl %ecx, %ecx movq %r14, %r11 cmpl %edi, %ecx je 0x2b6d00 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b6ce9 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b6d19 movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6d02 leaq (%r14,%rdx,8), %r14 incl %r15d addq %r10, %rax jmp 0x2b6cc7 movq 0xc0(%rsp), %r13 movl 0xac(%rsp), %r12d movq 0x48(%rsp), %rcx incq %rcx movq 0x30(%rsp), %r10 incl %r10d jmp 0x2b66a1 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b70ac vpextrd $0x3, %xmm0, %edi movl 0xe8(%r14,%rax), %ecx movl %edi, 0x48(%rsp) addl %edi, %ecx addl 0xec(%r14,%rax), %ecx addl 0xd0(%r14,%rax), %edx addl 0xd4(%r14,%rax), %edx addl 0xd8(%r14,%rax), %esi addl 0xdc(%r14,%rax), %esi movq 0x8(%r12), %rax movq %rax, 0x8(%rsp) movl $0x8, (%rsp) movq %rbx, %rdi movq %rcx, %r14 movl %r15d, %r8d callq 0x62a26 cmpq $0x0, (%rbx) je 0x2b7181 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b7181 xorl %ecx, %ecx testl %r14d, %r14d cmovlel %ecx, %r14d movq %r14, 0x30(%rsp) testl %r15d, %r15d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r15d movq %r15, 0xa0(%rsp) movabsq $0x100000001, %r13 # imm = 0x100000001 xorl %r14d, %r14d movq 0x18(%rsp), %rbp movq %rbx, 0xc8(%rsp) cmpq 0xa0(%rsp), %r14 je 0x2b7181 movq (%rbp), %rax movq -0x18(%rax), %rax vcvttss2si 0xe4(%rbp,%rax), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rsi shlq $0x20, %rsi orq %rcx, %rsi orq %rdx, %rsi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r15 shlq $0x38, %r15 orq %rdx, %r15 orq %rsi, %r15 orq %rax, %r15 xorl %r12d, %r12d cmpq 0x30(%rsp), %r12 je 0x2b7052 movslq 0x2c(%rbx), %rdi movslq 0x30(%rbx), %rdx movq (%rbx), %rax movq 0x10(%rbx), %rsi movq 0x40(%rbx), %r8 imulq %r14, %r8 movq %r8, %r9 imulq %rsi, %r9 addq %rax, %r9 movl 0x18(%rbx), %r10d movq %rdx, %rcx imulq %rdi, %rcx movq %r12, %r11 imulq %rsi, %r11 imulq %rcx, %r11 addq %r9, %r11 movq 0x20(%rbx), %r9 movq %r11, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %rsi, 0x60(%rsp) movl %r10d, 0x68(%rsp) movq %r9, 0x70(%rsp) movl $0x2, 0x78(%rsp) movl %edi, 0x7c(%rsp) movl %edx, 0x80(%rsp) movq %r13, 0x84(%rsp) movq %rcx, 0x90(%rsp) movq (%rbp), %r9 movq -0x18(%r9), %r9 movl 0xe8(%rbp,%r9), %r11d movl %r12d, %r10d subl %r11d, %r10d setns %r11b cmpl 0x48(%rsp), %r10d movq %r13, %rbx movq %rbp, %r13 setl %bpl testb %bpl, %r11b je 0x2b7016 movq 0x28(%rsp), %r8 movslq 0x2c(%r8), %rax movslq 0x30(%r8), %rcx movq 0x40(%r8), %rdx imulq %r14, %rdx movq 0x10(%r8), %rsi imulq %rsi, %rdx addq (%r8), %rdx movl 0x18(%r8), %edi movq 0x20(%r8), %r8 movq %rcx, %r11 imulq %rax, %r11 movl %r10d, %r10d movq %r11, %rbp imulq %rsi, %rbp imulq %r10, %rbp addq %rdx, %rbp movq %rbp, 0xd0(%rsp) andq $0x0, 0xd8(%rsp) movq %rsi, 0xe0(%rsp) movl %edi, 0xe8(%rsp) movq %r8, 0xf0(%rsp) movl $0x2, 0xf8(%rsp) movl %eax, 0xfc(%rsp) movl %ecx, 0x100(%rsp) movq %rbx, 0x104(%rsp) movq %r11, 0x110(%rsp) movl 0xd0(%r13,%r9), %edx movl 0xd4(%r13,%r9), %ecx movl 0xd8(%r13,%r9), %r8d movl 0xdc(%r13,%r9), %r9d movq %r15, (%rsp) leaq 0xd0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b740e movq %r13, %rbp movq %rbx, %r13 movq 0xc8(%rsp), %rbx incq %r12 jmp 0x2b6e81 testl %ecx, %ecx movl $0x0, %r9d cmovlel %r9d, %ecx imulq %r12, %rdx imulq %rdi, %rdx addq %r8, %rdx imulq %rdx, %rsi addq %rsi, %rax xorl %edx, %edx movq %r13, %rbp movq %rbx, %r13 movq 0xc8(%rsp), %rbx cmpq %rdx, %rcx je 0x2b700e movq %r15, (%rax,%rdx,8) incq %rdx jmp 0x2b7044 incq %r14 jmp 0x2b6e18 movq (%r14), %rax movq -0x18(%rax), %r8 movl 0xd0(%r14,%r8), %r10d leal (%r10,%rdx,8), %edx addl 0xd4(%r14,%r8), %edx xorl %r11d, %r11d testb $0x7, %dl sete %r11b movq %r9, %rax shrq $0x3, %rax leal (%r11,%r11,2), %ecx shlq %cl, %rax testb $0x7, %r10b sete %cl andb %cl, %r11b cmpb $0x1, %r11b jne 0x2b70ac cmpl $0x0, 0xe0(%r14,%r8) je 0x2b7198 movq 0x28(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm1 vmovaps %xmm1, 0x50(%rsp) movq %r9, 0x60(%rsp) movl %edi, 0x68(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x70(%rsp) vmovdqu %xmm0, 0x78(%rsp) movl %r15d, 0x88(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x90(%rsp) testq %rax, %rax je 0x2b70f3 lock incl (%rax) cmpl $0x1, %edi je 0x2b7134 vmovdqu (%r12), %ymm0 vmovups 0x20(%r12), %ymm1 leaq 0xd0(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovdqu %ymm0, (%rcx) movq 0x10(%r12), %rax movq %rax, 0x8(%rcx) leaq 0x50(%rsp), %rsi pushq $0x1 popq %rdx movq 0x28(%rsp), %rdi vzeroupper callq 0x64e3b movq (%r14), %rax addq -0x18(%rax), %r14 leaq 0x50(%rsp), %rsi movq %r14, %rdi movq %rbx, %rdx movq %r12, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b7181 lock decl (%rax) jne 0x2b7181 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2b7179 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2b7181 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0x128, %rsp # imm = 0x128 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq addl 0xd8(%r14,%r8), %esi addl 0xdc(%r14,%r8), %esi sarl $0x3, %edx movq 0x8(%r12), %r9 pushq $0x8 popq %rbp movq %rbx, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b7181 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b7181 movq 0x18(%rsp), %r9 movq (%r9), %rax movq -0x18(%rax), %rsi vcvttss2si 0xe4(%r9,%rsi), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rdi shlq $0x20, %rdi orq %rcx, %rdi orq %rdx, %rdi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r11 shlq $0x38, %r11 orq %rdx, %r11 orq %rdi, %r11 orq %rax, %r11 movl 0xd0(%r9,%rsi), %eax movl 0xd4(%r9,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%r9,%rsi), %r8d movl 0xdc(%r9,%rsi), %r9d movq %r11, (%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b740e movq $0x0, 0x20(%rsp) jmp 0x2b7181 jmp 0x2b72bd jmp 0x2b7284 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b72b5 lock decl (%rax) jne 0x2b72b5 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2b72af movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b72b5 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_fma.cpp
virtual thunk to ncnn::Padding_x86_fma::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Padding_x86_fma::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float)per_channel_pad_data + q 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float)per_channel_pad_data + q 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 \|\| (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float)per_channel_pad_data + q 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 \|\| (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }	movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2b4800	/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_fma.cpp
ncnn::padding_constant_pack8_int8_sse(ncnn::Mat const&, ncnn::Mat&, int, int, int, int, long)	static void padding_constant_pack8_int8_sse(const Mat& src, Mat& dst, int top, int bottom, int left, int right, int64_t _v) { const int64_t* ptr = src; int64_t* outptr = dst; // fill top for (int y = 0; y < top; y++) { for (int x = 0; x < dst.w; x++) { outptr++ = _v; } } // fill center for (int y = 0; y < src.h; y++) { for (int x = 0; x < left; x++) { outptr++ = _v; } for (int x = 0; x < src.w; x++) { outptr++ = ptr++; } for (int x = 0; x < right; x++) { outptr++ = _v; } } // fill bottom for (int y = 0; y < bottom; y++) { for (int x = 0; x < dst.w; x++) { outptr++ = _v; } } }	pushq %rbp pushq %r15 pushq %r14 pushq %r12 pushq %rbx movq 0x30(%rsp), %rax movq (%rdi), %r11 movl 0x2c(%rsi), %ebx xorl %ebp, %ebp testl %ebx, %ebx cmovlel %ebp, %ebx movq (%rsi), %r10 testl %edx, %edx cmovlel %ebp, %edx cmpl %edx, %ebp je 0x2b744a movl %ebx, %r14d subl $0x1, %r14d jb 0x2b7446 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b7437 incl %ebp jmp 0x2b7430 xorl %edx, %edx testl %r8d, %r8d cmovlel %edx, %r8d movl 0x2c(%rdi), %ebx testl %ebx, %ebx cmovlel %edx, %ebx testl %r9d, %r9d cmovlel %edx, %r9d movl 0x30(%rdi), %edi testl %edi, %edi cmovlel %edx, %edi cmpl %edi, %edx je 0x2b74b6 movl %r8d, %ebp subl $0x1, %ebp jb 0x2b747f movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b7471 xorl %r15d, %r15d xorl %r14d, %r14d cmpl %r15d, %ebx je 0x2b749b movq (%r11,%r15,8), %r12 movq %r12, (%r10,%r15,8) addq $-0x8, %r14 incq %r15 jmp 0x2b7485 subq %r14, %r11 subq %r14, %r10 movl %r9d, %ebp subl $0x1, %ebp jb 0x2b74b2 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b74a4 incl %edx jmp 0x2b746a movl 0x2c(%rsi), %edx xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %ecx, %ecx cmovlel %esi, %ecx movl %edx, %edi cmpl %ecx, %esi je 0x2b74dd subl $0x1, %edi jb 0x2b74d9 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b74cb incl %esi jmp 0x2b74c5 popq %rbx popq %r12 popq %r14 popq %r15 popq %rbp retq nop	/csukuangfj[P]ncnn/src/layer/x86/padding_pack8_int8.h
ncnn::ExpandDims::load_param(ncnn::ParamDict const&)	int ExpandDims::load_param(const ParamDict& pd) { expand_w = pd.get(0, 0); expand_h = pd.get(1, 0); expand_d = pd.get(11, 0); expand_c = pd.get(2, 0); axes = pd.get(3, Mat()); return 0; }	pushq %r15 pushq %r14 pushq %rbx subq $0xa0, %rsp movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd4(%rbx) pushq $0xb popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xdc(%rbx) leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) xorps %xmm0, %xmm0 movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 pushq $0x3 popq %rdx movq %r15, %rdi movq %r14, %rsi callq 0x718da leaq 0xe0(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2bc8e4 testq %rax, %rax je 0x2bc855 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x2bc889 lock decl (%rax) jne 0x2bc889 movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x2bc881 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2bc889 movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x8(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x2bc90c lock decl (%rax) jne 0x2bc90c movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2bc904 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2bc90c movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2bc956 lock decl (%rax) jne 0x2bc956 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2bc94e movq (%rdi), %rax callq 0x18(%rax) jmp 0x2bc956 movq %rsi, %rdi callq 0x5f3e0 xorl %eax, %eax addq $0xa0, %rsp popq %rbx popq %r14 popq %r15 retq movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2bc995 lock decl (%rax) jne 0x2bc995 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2bc98f movq %rsi, %rdi callq 0x5f3e0 jmp 0x2bc995 movq (%rdi), %rax callq 0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2bc9bc jmp 0x2bc9f2 jmp 0x2bc9f2 jmp 0x2bc9f2 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2bc9ea lock decl (%rax) jne 0x2bc9ea movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2bc9e4 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2bc9ea movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68	/csukuangfj[P]ncnn/src/layer/expanddims.cpp
ncnn::PriorBox::load_param(ncnn::ParamDict const&)	int PriorBox::load_param(const ParamDict& pd) { min_sizes = pd.get(0, Mat()); max_sizes = pd.get(1, Mat()); aspect_ratios = pd.get(2, Mat()); variances[0] = pd.get(3, 0.1f); variances[1] = pd.get(4, 0.1f); variances[2] = pd.get(5, 0.2f); variances[3] = pd.get(6, 0.2f); flip = pd.get(7, 1); clip = pd.get(8, 0); image_width = pd.get(9, 0); image_height = pd.get(10, 0); step_width = pd.get(11, -233.f); step_height = pd.get(12, -233.f); offset = pd.get(13, 0.f); step_mmdetection = pd.get(14, 0); center_mmdetection = pd.get(15, 0); return 0; }	pushq %r15 pushq %r14 pushq %rbx subq $0xa0, %rsp movq %rsi, %r14 movq %rdi, %rbx leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) xorps %xmm0, %xmm0 movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 movq %r15, %rdi xorl %edx, %edx callq 0x718da leaq 0xd0(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2c0907 testq %rax, %rax je 0x2c0878 lock incl (%rax) movq 0xd8(%rbx), %rax testq %rax, %rax je 0x2c08ac lock decl (%rax) jne 0x2c08ac movq 0xd0(%rbx), %rsi movq 0xf0(%rbx), %rdi testq %rdi, %rdi je 0x2c08a4 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c08ac movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0xd0(%rbx) movq 0x8(%rsp), %rax movq %rax, 0xd8(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0xe0(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0xf8(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x108(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x110(%rbx) testq %rax, %rax je 0x2c092f lock decl (%rax) jne 0x2c092f movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2c0927 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c092f movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c097f lock decl (%rax) jne 0x2c097f movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0974 movq (%rdi), %rax callq 0x18(%rax) xorps %xmm0, %xmm0 jmp 0x2c097f movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 pushq $0x1 popq %rdx movq %r15, %rdi movq %r14, %rsi callq 0x718da leaq 0x118(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2c0a55 testq %rax, %rax je 0x2c09c6 lock incl (%rax) movq 0x120(%rbx), %rax testq %rax, %rax je 0x2c09fa lock decl (%rax) jne 0x2c09fa movq 0x118(%rbx), %rsi movq 0x138(%rbx), %rdi testq %rdi, %rdi je 0x2c09f2 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c09fa movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0x118(%rbx) movq 0x8(%rsp), %rax movq %rax, 0x120(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0x128(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0x130(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x138(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x140(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x150(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x158(%rbx) testq %rax, %rax je 0x2c0a7d lock decl (%rax) jne 0x2c0a7d movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2c0a75 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c0a7d movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0acd lock decl (%rax) jne 0x2c0acd movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0ac2 movq (%rdi), %rax callq 0x18(%rax) xorps %xmm0, %xmm0 jmp 0x2c0acd movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 pushq $0x2 popq %rdx movq %r15, %rdi movq %r14, %rsi callq 0x718da leaq 0x160(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2c0ba3 testq %rax, %rax je 0x2c0b14 lock incl (%rax) movq 0x168(%rbx), %rax testq %rax, %rax je 0x2c0b48 lock decl (%rax) jne 0x2c0b48 movq 0x160(%rbx), %rsi movq 0x180(%rbx), %rdi testq %rdi, %rdi je 0x2c0b40 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c0b48 movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0x160(%rbx) movq 0x8(%rsp), %rax movq %rax, 0x168(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0x170(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0x178(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x180(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x188(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x198(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x1a0(%rbx) testq %rax, %rax je 0x2c0bcb lock decl (%rax) jne 0x2c0bcb movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2c0bc3 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c0bcb movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0c15 lock decl (%rax) jne 0x2c0c15 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0c0d movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c0c15 movq %rsi, %rdi callq 0x5f3e0 pushq $0x3 popq %rsi movss 0x132fe8(%rip), %xmm0 # 0x3f3c08 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0x1a8(%rbx) pushq $0x4 popq %rsi movq %r14, %rdi movss 0x132fca(%rip), %xmm0 # 0x3f3c08 callq 0x718c0 movss %xmm0, 0x1ac(%rbx) pushq $0x5 popq %rsi movss 0x137c7e(%rip), %xmm0 # 0x3f88d4 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0x1b0(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi movss 0x137c60(%rip), %xmm0 # 0x3f88d4 callq 0x718c0 movss %xmm0, 0x1b4(%rbx) pushq $0x7 popq %rsi pushq $0x1 popq %rdx movq %r14, %rdi callq 0x718a6 movl %eax, 0x1b8(%rbx) pushq $0x8 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x1bc(%rbx) pushq $0x9 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x1c0(%rbx) pushq $0xa popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x1c4(%rbx) pushq $0xb popq %rsi movss 0x137bff(%rip), %xmm0 # 0x3f88d8 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0x1c8(%rbx) pushq $0xc popq %rsi movq %r14, %rdi movss 0x137be1(%rip), %xmm0 # 0x3f88d8 callq 0x718c0 movss %xmm0, 0x1cc(%rbx) pushq $0xd popq %rsi xorps %xmm0, %xmm0 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0x1d0(%rbx) pushq $0xe popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 testl %eax, %eax setne 0x1d4(%rbx) pushq $0xf popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 testl %eax, %eax setne 0x1d5(%rbx) xorl %eax, %eax addq $0xa0, %rsp popq %rbx popq %r14 popq %r15 retq movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2c0d85 lock decl (%rax) jne 0x2c0d85 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2c0d7f movq %rsi, %rdi callq 0x5f3e0 jmp 0x2c0d85 movq (%rdi), %rax callq 0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2c0e75 jmp 0x2c0f01 movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2c0ddb lock decl (%rax) jne 0x2c0ddb movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2c0dd5 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2c0ddb movq (%rdi), %rax callq 0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2c0ea0 jmp 0x2c0f01 movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2c0e31 lock decl (%rax) jne 0x2c0e31 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2c0e2b movq %rsi, %rdi callq 0x5f3e0 jmp 0x2c0e31 movq (%rdi), %rax callq 0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2c0ecb jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0ef1 lock decl (%rax) jne 0x2c0ef1 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0ee9 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c0ef1 jmp 0x2c0f01 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0ef1 lock decl (%rax) jne 0x2c0ef1 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0ee9 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c0ef1 jmp 0x2c0f01 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0ef1 lock decl (%rax) jne 0x2c0ef1 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2c0ef9 movq %rsi, %rdi callq 0x5f3e0 movq %rbx, %rdi callq 0x5f340 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c0ef1 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/priorbox.cpp
ncnn::DetectionOutput::load_param(ncnn::ParamDict const&)	int DetectionOutput::load_param(const ParamDict& pd) { num_class = pd.get(0, 0); nms_threshold = pd.get(1, 0.05f); nms_top_k = pd.get(2, 300); keep_top_k = pd.get(3, 100); confidence_threshold = pd.get(4, 0.5f); variances[0] = pd.get(5, 0.1f); variances[1] = pd.get(6, 0.1f); variances[2] = pd.get(7, 0.2f); variances[3] = pd.get(8, 0.2f); return 0; }	pushq %r14 pushq %rbx pushq %rax movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rsi movss 0x1371b7(%rip), %xmm0 # 0x3f88f0 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xd4(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi movl $0x12c, %edx # imm = 0x12C callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x3 popq %rsi pushq $0x64 popq %rdx movq %r14, %rdi callq 0x718a6 movl %eax, 0xdc(%rbx) pushq $0x4 popq %rsi movss 0x12c896(%rip), %xmm0 # 0x3ee014 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xe0(%rbx) pushq $0x5 popq %rsi movss 0x13246f(%rip), %xmm0 # 0x3f3c08 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xe4(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi movss 0x132451(%rip), %xmm0 # 0x3f3c08 callq 0x718c0 movss %xmm0, 0xe8(%rbx) pushq $0x7 popq %rsi movss 0x137105(%rip), %xmm0 # 0x3f88d4 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xec(%rbx) pushq $0x8 popq %rsi movq %r14, %rdi movss 0x1370e7(%rip), %xmm0 # 0x3f88d4 callq 0x718c0 movss %xmm0, 0xf0(%rbx) xorl %eax, %eax addq $0x8, %rsp popq %rbx popq %r14 retq	/csukuangfj[P]ncnn/src/layer/detectionoutput.cpp
ncnn::Interp::load_param(ncnn::ParamDict const&)	int Interp::load_param(const ParamDict& pd) { resize_type = pd.get(0, 0); height_scale = pd.get(1, 1.f); width_scale = pd.get(2, 1.f); output_height = pd.get(3, 0); output_width = pd.get(4, 0); dynamic_target_size = pd.get(5, 0); align_corner = pd.get(6, 0); if (resize_type < 0 \|\| resize_type > 3) { NCNN_LOGE("unsupported resize type %d", resize_type); return -1; } if (dynamic_target_size == 1) { one_blob_only = false; } return 0; }	pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r15 movq %rdi, %r14 xorl %ebx, %ebx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%r14) pushq $0x1 popq %rsi movss 0x12bc91(%rip), %xmm0 # 0x3eec88 movq %r15, %rdi callq 0x718c0 movss %xmm0, 0xd8(%r14) pushq $0x2 popq %rsi movq %r15, %rdi movss 0x12bc72(%rip), %xmm0 # 0x3eec88 callq 0x718c0 movss %xmm0, 0xd4(%r14) pushq $0x3 popq %rsi movq %r15, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe0(%r14) pushq $0x4 popq %rsi movq %r15, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xdc(%r14) pushq $0x5 popq %rsi movq %r15, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe4(%r14) pushq $0x6 popq %rsi movq %r15, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe8(%r14) movl 0xd0(%r14), %edx cmpl $0x4, %edx jae 0x2c3097 cmpl $0x1, 0xe4(%r14) jne 0x2c308f movb $0x0, 0x8(%r14) movl %ebx, %eax popq %rbx popq %r14 popq %r15 retq movq 0x1ccf3a(%rip), %rbx # 0x48ffd8 movq (%rbx), %rdi leaq 0x1358a8(%rip), %rsi # 0x3f8950 xorl %eax, %eax callq 0x5f150 movq (%rbx), %rsi pushq $0xa popq %rdi callq 0x5f1c0 pushq $-0x1 popq %rbx jmp 0x2c308f nop	/csukuangfj[P]ncnn/src/layer/interp.cpp
ncnn::Interp::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int Interp::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int outw = output_width; int outh = output_height; if (bottom_blob.dims == 1) { w = 1; h = 1; } if (outw == 0 \|\| outh == 0) { outw = static_cast<int>(w * width_scale); outh = static_cast<int>(h * height_scale); } Mat reference_blob; reference_blob.w = outw; reference_blob.h = outh; std::vector<Mat> bottom_blobs(2); bottom_blobs[0] = bottom_blob; bottom_blobs[1] = reference_blob; std::vector<Mat> top_blobs(1); int ret = forward(bottom_blobs, top_blobs, opt); top_blob = top_blobs[0]; return ret; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x58, %rsp movq %rcx, %r14 movq %rdx, %rbx movq %rsi, %r12 movq %rdi, %r15 movq 0xdc(%rdi), %xmm0 movdqa %xmm0, 0x40(%rsp) pshufd $0x50, %xmm0, %xmm0 # xmm0 = xmm0[0,0,1,1] pxor %xmm1, %xmm1 pcmpeqd %xmm0, %xmm1 movmskpd %xmm1, %eax testl %eax, %eax je 0x2c314f xorl %eax, %eax cmpl $0x1, 0x28(%r12) sete %al movsd 0x2c(%r12), %xmm0 movd %eax, %xmm1 pshufd $0x50, %xmm1, %xmm1 # xmm1 = xmm1[0,0,1,1] pslld $0x1f, %xmm1 psrad $0x1f, %xmm1 cvtdq2ps %xmm0, %xmm0 movdqa %xmm1, %xmm2 pandn %xmm0, %xmm2 pand 0x12af7b(%rip), %xmm1 # 0x3ee0b0 por %xmm2, %xmm1 movsd 0xd4(%r15), %xmm0 mulps %xmm1, %xmm0 cvttps2dq %xmm0, %xmm0 movdqa %xmm0, 0x40(%rsp) leaq 0x28(%rsp), %r13 pushq $0x2 popq %rsi leaq 0x10(%rsp), %rdx movq %r13, %rdi callq 0x6fa72 movq (%r13), %r13 cmpq %r12, %r13 je 0x2c3212 movq 0x8(%r12), %rax testq %rax, %rax je 0x2c317e lock incl (%rax) movq 0x8(%r13), %rax testq %rax, %rax je 0x2c31a9 lock decl (%rax) jne 0x2c31a9 movq (%r13), %rsi movq 0x20(%r13), %rdi testq %rdi, %rdi je 0x2c31a1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c31a9 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%r13) pxor %xmm0, %xmm0 movdqu %xmm0, (%r13) movdqu %xmm0, 0xc(%r13) andl $0x0, 0x38(%r13) movdqu %xmm0, 0x28(%r13) movups (%r12), %xmm0 movups %xmm0, (%r13) movq 0x10(%r12), %rax movq %rax, 0x10(%r13) movl 0x18(%r12), %eax movl %eax, 0x18(%r13) movq 0x20(%r12), %rax movq %rax, 0x20(%r13) movdqu 0x28(%r12), %xmm0 movdqu %xmm0, 0x28(%r13) movl 0x38(%r12), %eax movl %eax, 0x38(%r13) movq 0x40(%r12), %rax movq %rax, 0x40(%r13) movq 0x28(%rsp), %r13 movq 0x50(%r13), %rax testq %rax, %rax je 0x2c323d lock decl (%rax) jne 0x2c323d movq 0x48(%r13), %rsi movq 0x68(%r13), %rdi testq %rdi, %rdi je 0x2c3235 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c323d movq %rsi, %rdi callq 0x5f3e0 leaq 0x48(%r13), %rax pxor %xmm0, %xmm0 movdqu %xmm0, 0xc(%rax) movdqu %xmm0, (%rax) movdqu %xmm0, 0x68(%r13) movdqa 0x40(%rsp), %xmm0 movq %xmm0, 0x74(%r13) andq $0x0, 0x7c(%r13) andq $0x0, 0x88(%r13) leaq 0x10(%rsp), %rdi pushq $0x1 popq %rsi leaq 0xf(%rsp), %rdx callq 0x6fa72 movq (%r15), %rax leaq 0x28(%rsp), %rsi leaq 0x10(%rsp), %rdx movq %r15, %rdi movq %r14, %rcx callq 0x30(%rax) movl %eax, %ebp movq 0x10(%rsp), %r14 cmpq %rbx, %r14 je 0x2c332e movq 0x8(%r14), %rax testq %rax, %rax je 0x2c32b1 lock incl (%rax) movq 0x8(%rbx), %rax testq %rax, %rax je 0x2c32db lock decl (%rax) jne 0x2c32db movq (%rbx), %rsi movq 0x20(%rbx), %rdi testq %rdi, %rdi je 0x2c32d3 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2c32db movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rbx) pxor %xmm0, %xmm0 movdqu %xmm0, (%rbx) movdqu %xmm0, 0xc(%rbx) andl $0x0, 0x38(%rbx) movdqu %xmm0, 0x28(%rbx) movups (%r14), %xmm0 movups %xmm0, (%rbx) movq 0x10(%r14), %rax movq %rax, 0x10(%rbx) movl 0x18(%r14), %eax movl %eax, 0x18(%rbx) movq 0x20(%r14), %rax movq %rax, 0x20(%rbx) movdqu 0x28(%r14), %xmm0 movdqu %xmm0, 0x28(%rbx) movl 0x38(%r14), %eax movl %eax, 0x38(%rbx) movq 0x40(%r14), %rax movq %rax, 0x40(%rbx) leaq 0x10(%rsp), %rdi callq 0x6fac4 leaq 0x28(%rsp), %rdi callq 0x6fac4 movl %ebp, %eax addq $0x58, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2c3355 movq %rax, %rbx jmp 0x2c3367 movq %rax, %rbx leaq 0x10(%rsp), %rdi callq 0x6fac4 leaq 0x28(%rsp), %rdi callq 0x6fac4 movq %rbx, %rdi callq 0x5f340 nop	/csukuangfj[P]ncnn/src/layer/interp.cpp
ncnn::linear_coeffs(int, int, int, float, int)	static void linear_coeffs(int w, int outw, int* xofs, float* alpha, int align_corner) { double scale = (double)w / outw; if (align_corner) { scale = (double)(w - 1) / (outw - 1); } for (int dx = 0; dx < outw; dx++) { float fx = (float)((dx + 0.5) * scale - 0.5); if (align_corner) { fx = (float)(dx * scale); } int sx = (int)floorf(fx); fx -= sx; if (sx < 0) { sx = 0; fx = 0.f; } if (sx >= w - 1) { sx = w - 2; fx = 1.f; } xofs[dx] = sx; alpha[dx * 2] = 1.f - fx; alpha[dx * 2 + 1] = fx; } }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x18, %rsp movl %r8d, %ebx movq %rcx, %r14 movq %rdx, 0x10(%rsp) movl %esi, %r12d movl %edi, %r13d testl %r8d, %r8d je 0x2c7078 leal -0x1(%r13), %ebp cvtsi2sd %ebp, %xmm1 leal -0x1(%r12), %eax cvtsi2sd %eax, %xmm0 divsd %xmm0, %xmm1 movsd %xmm1, (%rsp) jmp 0x2c708f cvtsi2sd %r13d, %xmm1 cvtsi2sd %r12d, %xmm0 divsd %xmm0, %xmm1 movsd %xmm1, (%rsp) leal -0x1(%r13), %ebp addl $-0x2, %r13d xorl %eax, %eax testl %r12d, %r12d cmovlel %eax, %r12d xorl %r15d, %r15d cmpq %r15, %r12 je 0x2c7140 cvtsi2sd %r15d, %xmm0 movapd %xmm0, %xmm1 addsd 0x12700f(%rip), %xmm1 # 0x3ee0c8 mulsd (%rsp), %xmm1 addsd 0x127342(%rip), %xmm1 # 0x3ee408 testl %ebx, %ebx je 0x2c70d3 mulsd (%rsp), %xmm0 movapd %xmm0, %xmm1 cvtsd2ss %xmm1, %xmm0 movss %xmm0, 0xc(%rsp) callq 0x5f360 cvttss2si %xmm0, %eax xorps %xmm1, %xmm1 testl %eax, %eax js 0x2c70fe cvttps2dq %xmm0, %xmm0 cvtdq2ps %xmm0, %xmm0 movss 0xc(%rsp), %xmm1 subss %xmm0, %xmm1 movl $0x0, %ecx cmovlel %ecx, %eax cmpl %ebp, %eax jl 0x2c7112 movss 0x127b76(%rip), %xmm1 # 0x3eec88 cmovgel %r13d, %eax movq 0x10(%rsp), %rcx movl %eax, (%rcx,%r15,4) movss 0x127b61(%rip), %xmm0 # 0x3eec88 subss %xmm1, %xmm0 movss %xmm0, (%r14,%r15,8) movss %xmm1, 0x4(%r14,%r15,8) incq %r15 jmp 0x2c709f addq $0x18, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq	/csukuangfj[P]ncnn/src/layer/x86/interp_bilinear.h
ncnn::cubic_coeffs(int, int, int, float, int)	static void cubic_coeffs(int w, int outw, int* xofs, float* alpha, int align_corner) { double scale = (double)w / outw; if (align_corner) { scale = (double)(w - 1) / (outw - 1); } for (int dx = 0; dx < outw; dx++) { float fx = (float)((dx + 0.5) * scale - 0.5); if (align_corner) { fx = (float)(dx * scale); } int sx = static_cast<int>(floor(fx)); fx -= sx; interpolate_cubic(fx, alpha + dx * 4); if (sx <= -1) { sx = 1; alpha[dx * 4 + 0] = 1.f - alpha[dx * 4 + 3]; alpha[dx * 4 + 1] = alpha[dx * 4 + 3]; alpha[dx * 4 + 2] = 0.f; alpha[dx * 4 + 3] = 0.f; } if (sx == 0) { sx = 1; alpha[dx * 4 + 0] = alpha[dx * 4 + 0] + alpha[dx * 4 + 1]; alpha[dx * 4 + 1] = alpha[dx * 4 + 2]; alpha[dx * 4 + 2] = alpha[dx * 4 + 3]; alpha[dx * 4 + 3] = 0.f; } if (sx == w - 2) { sx = w - 3; alpha[dx * 4 + 3] = alpha[dx * 4 + 2] + alpha[dx * 4 + 3]; alpha[dx * 4 + 2] = alpha[dx * 4 + 1]; alpha[dx * 4 + 1] = alpha[dx * 4 + 0]; alpha[dx * 4 + 0] = 0.f; } if (sx >= w - 1) { sx = w - 3; alpha[dx * 4 + 3] = 1.f - alpha[dx * 4 + 0]; alpha[dx * 4 + 2] = alpha[dx * 4 + 0]; alpha[dx * 4 + 1] = 0.f; alpha[dx * 4 + 0] = 0.f; } xofs[dx] = sx; } }	testl %r8d, %r8d je 0x2d3720 leal -0x1(%rdi), %eax vcvtsi2sd %eax, %xmm0, %xmm0 leal -0x1(%rsi), %r9d vcvtsi2sd %r9d, %xmm1, %xmm1 vdivsd %xmm1, %xmm0, %xmm0 jmp 0x2d372f vcvtsi2sd %edi, %xmm0, %xmm0 vcvtsi2sd %esi, %xmm1, %xmm1 vdivsd %xmm1, %xmm0, %xmm0 leal -0x1(%rdi), %eax pushq %rbp pushq %r14 pushq %rbx leal -0x2(%rdi), %r9d addl $-0x3, %edi xorl %r10d, %r10d testl %esi, %esi cmovlel %r10d, %esi leaq 0xc(%rcx), %r11 vmovss 0x11b539(%rip), %xmm3 # 0x3eec88 vmovss 0x1251b9(%rip), %xmm4 # 0x3f8910 vmovss 0x124129(%rip), %xmm5 # 0x3f7888 vmovsd 0x1251b9(%rip), %xmm6 # 0x3f8920 vmovss 0x1251a5(%rip), %xmm7 # 0x3f8914 vbroadcastss 0x1251a0(%rip), %xmm8 # 0x3f8918 vbroadcastss 0x11b507(%rip), %xmm9 # 0x3eec88 vmovss 0x11ef9b(%rip), %xmm10 # 0x3f2724 vmovss 0x125187(%rip), %xmm11 # 0x3f8918 vbroadcastss 0x11da1a(%rip), %xmm12 # 0x3f11b4 movl $0x1, %ebx cmpq %r10, %rsi je 0x2d3901 vcvtsi2sd %r10d, %xmm15, %xmm13 vaddsd 0x11a913(%rip), %xmm13, %xmm14 # 0x3ee0c8 vmulsd %xmm0, %xmm14, %xmm14 vaddsd 0x11ac47(%rip), %xmm14, %xmm14 # 0x3ee408 testl %r8d, %r8d je 0x2d37ca vmulsd %xmm0, %xmm13, %xmm14 vcvtsd2ss %xmm14, %xmm14, %xmm13 vroundss $0x9, %xmm13, %xmm13, %xmm14 vcvttss2si %xmm14, %ebp vsubss %xmm14, %xmm13, %xmm13 vaddss %xmm3, %xmm13, %xmm14 vsubss %xmm13, %xmm3, %xmm15 vmulss %xmm4, %xmm14, %xmm1 vaddss %xmm5, %xmm1, %xmm1 vblendps $0xe, %xmm9, %xmm1, %xmm1 # xmm1 = xmm1[0],xmm9[1,2,3] vinsertps $0x10, %xmm13, %xmm14, %xmm13 # xmm13 = xmm14[0],xmm13[0],xmm14[2,3] vmulps %xmm13, %xmm13, %xmm14 vmulps %xmm6, %xmm13, %xmm13 vsubps %xmm13, %xmm7, %xmm2 vaddps %xmm8, %xmm13, %xmm13 vblendps $0x2, %xmm13, %xmm2, %xmm2 # xmm2 = xmm2[0],xmm13[1],xmm2[2,3] vmulps %xmm2, %xmm14, %xmm2 vaddps %xmm1, %xmm2, %xmm13 vmovlps %xmm13, -0xc(%r11) vmulss %xmm15, %xmm15, %xmm1 vmulss %xmm10, %xmm15, %xmm2 vaddss %xmm2, %xmm11, %xmm2 vmulss %xmm2, %xmm1, %xmm1 vaddss %xmm3, %xmm1, %xmm2 vmovss %xmm2, -0x4(%r11) vmovshdup %xmm13, %xmm2 # xmm2 = xmm13[1,1,3,3] vaddss %xmm1, %xmm13, %xmm1 vaddss %xmm2, %xmm1, %xmm1 vxorps %xmm1, %xmm12, %xmm1 vmovss %xmm1, (%r11) testl %ebp, %ebp js 0x2d3878 jne 0x2d38a8 vaddss -0x8(%r11), %xmm13, %xmm13 vmovss %xmm13, -0xc(%r11) vmovsd -0x4(%r11), %xmm1 vmovsd %xmm1, -0x8(%r11) movq %r11, %r14 jmp 0x2d38a2 leaq (,%r10,4), %r14 leaq (%rcx,%r14,4), %r14 addq $0xc, %r14 vmovss (%r11), %xmm1 vsubss %xmm1, %xmm3, %xmm13 vmovss %xmm13, -0xc(%r11) vmovss %xmm1, -0x8(%r11) andl $0x0, -0x4(%r11) andl $0x0, (%r14) movl %ebx, %ebp cmpl %r9d, %ebp jne 0x2d38d1 vmovss (%r11), %xmm1 vaddss -0x4(%r11), %xmm1, %xmm1 vmovss -0x8(%r11), %xmm2 vmovss %xmm1, (%r11) vmovss %xmm2, -0x4(%r11) vmovss %xmm13, -0x8(%r11) jmp 0x2d38ea cmpl %eax, %ebp jl 0x2d38f1 vsubss %xmm13, %xmm3, %xmm1 vmovss %xmm1, (%r11) vmovss %xmm13, -0x4(%r11) andl $0x0, -0x8(%r11) andl $0x0, -0xc(%r11) movl %edi, %ebp movl %ebp, (%rdx,%r10,4) incq %r10 addq $0x10, %r11 jmp 0x2d379f popq %rbx popq %r14 popq %rbp retq	/csukuangfj[P]ncnn/src/layer/x86/interp_bicubic.h
ncnn::DeconvolutionDepthWise::load_param(ncnn::ParamDict const&)	int DeconvolutionDepthWise::load_param(const ParamDict& pd) { num_output = pd.get(0, 0); kernel_w = pd.get(1, 0); kernel_h = pd.get(11, kernel_w); dilation_w = pd.get(2, 1); dilation_h = pd.get(12, dilation_w); stride_w = pd.get(3, 1); stride_h = pd.get(13, stride_w); pad_left = pd.get(4, 0); pad_right = pd.get(15, pad_left); pad_top = pd.get(14, pad_left); pad_bottom = pd.get(16, pad_top); output_pad_right = pd.get(18, 0); output_pad_bottom = pd.get(19, output_pad_right); output_w = pd.get(20, 0); output_h = pd.get(21, output_w); bias_term = pd.get(5, 0); weight_data_size = pd.get(6, 0); group = pd.get(7, 1); activation_type = pd.get(9, 0); activation_params = pd.get(10, Mat()); return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %rbx subq $0x98, %rsp movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rbp movq %r14, %rdi movl %ebp, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd4(%rbx) pushq $0xb popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi movl %ebp, %edx callq 0x718a6 movl %eax, 0xdc(%rbx) pushq $0xc popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xe0(%rbx) pushq $0x3 popq %rsi movq %r14, %rdi movl %ebp, %edx callq 0x718a6 movl %eax, 0xe4(%rbx) pushq $0xd popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xe8(%rbx) pushq $0x4 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xec(%rbx) pushq $0xf popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xf0(%rbx) movl 0xec(%rbx), %edx pushq $0xe popq %rsi movq %r14, %rdi callq 0x718a6 movl %eax, 0xf4(%rbx) pushq $0x10 popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xf8(%rbx) pushq $0x12 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xfc(%rbx) pushq $0x13 popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0x100(%rbx) pushq $0x14 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x104(%rbx) pushq $0x15 popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0x108(%rbx) pushq $0x5 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x10c(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x110(%rbx) pushq $0x7 popq %rsi movq %r14, %rdi movl %ebp, %edx callq 0x718a6 movl %eax, 0x114(%rbx) pushq $0x9 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x118(%rbx) leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) xorps %xmm0, %xmm0 movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 pushq $0xa popq %rdx movq %r15, %rdi movq %r14, %rsi callq 0x718da leaq 0x120(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2d3bf6 testq %rax, %rax je 0x2d3b67 lock incl (%rax) movq 0x128(%rbx), %rax testq %rax, %rax je 0x2d3b9b lock decl (%rax) jne 0x2d3b9b movq 0x120(%rbx), %rsi movq 0x140(%rbx), %rdi testq %rdi, %rdi je 0x2d3b93 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d3b9b movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0x120(%rbx) movq 0x8(%rsp), %rax movq %rax, 0x128(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0x130(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0x138(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x140(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x148(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x158(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x160(%rbx) testq %rax, %rax je 0x2d3c1e lock decl (%rax) jne 0x2d3c1e movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2d3c16 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d3c1e movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2d3c68 lock decl (%rax) jne 0x2d3c68 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2d3c60 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d3c68 movq %rsi, %rdi callq 0x5f3e0 xorl %eax, %eax addq $0x98, %rsp popq %rbx popq %r14 popq %r15 popq %rbp retq movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2d3ca8 lock decl (%rax) jne 0x2d3ca8 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2d3ca2 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d3ca8 movq (%rdi), %rax callq 0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2d3ccf jmp 0x2d3d05 jmp 0x2d3d05 jmp 0x2d3d05 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2d3cfd lock decl (%rax) jne 0x2d3cfd movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2d3cf7 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d3cfd movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/deconvolutiondepthwise.cpp
ncnn::DeconvolutionDepthWise::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const	int DeconvolutionDepthWise::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; size_t elemsize = bottom_blob.elemsize; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; int outw = (w - 1) * stride_w + kernel_extent_w + output_pad_right; int outh = (h - 1) * stride_h + kernel_extent_h + output_pad_bottom; Mat top_blob_bordered; if (pad_left > 0 \|\| pad_right > 0 \|\| pad_top > 0 \|\| pad_bottom > 0 \|\| (output_w > 0 && output_h > 0)) { top_blob_bordered.create(outw, outh, num_output, elemsize, opt.workspace_allocator); } else { top_blob_bordered = top_blob; top_blob_bordered.create(outw, outh, num_output, elemsize, opt.blob_allocator); } if (top_blob_bordered.empty()) return -100; int ret = deconvolutiondepthwise(bottom_blob, top_blob_bordered, weight_data, bias_data, kernel_w, kernel_h, stride_w, stride_h, dilation_w, dilation_h, group, activation_type, activation_params, opt); if (ret != 0) return ret; cut_padding(top_blob_bordered, top_blob, opt); if (top_blob.empty()) return -100; return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x198, %rsp # imm = 0x198 movl 0x2c(%rsi), %r14d movl 0x30(%rsi), %ebp movl 0xd4(%rdi), %ebx decl %ebx imull 0xdc(%rdi), %ebx movq %rsi, 0x20(%rsp) movq 0x10(%rsi), %r8 decl %r14d imull 0xe4(%rdi), %r14d movl 0xd8(%rdi), %r12d decl %r12d imull 0xe0(%rdi), %r12d decl %ebp imull 0xe8(%rdi), %ebp movl 0xfc(%rdi), %esi movl 0x100(%rdi), %r10d andq $0x0, 0xb0(%rsp) movq %rcx, %r9 movq %rdx, %r13 pxor %xmm0, %xmm0 movdqa %xmm0, 0x70(%rsp) movdqu %xmm0, 0x7c(%rsp) movdqa %xmm0, 0x90(%rsp) movdqu %xmm0, 0x9c(%rsp) movdqu 0xec(%rdi), %xmm1 pcmpgtd %xmm0, %xmm1 movmskps %xmm1, %eax pushq $0x10 popq %r15 testl %eax, %eax movq %rdi, 0x18(%rsp) movq %rcx, 0x118(%rsp) jne 0x2d4172 cmpl $0x0, 0x104(%rdi) jle 0x2d409b movq 0x18(%rsp), %rax cmpl $0x0, 0x108(%rax) jg 0x2d4172 pushq $0x8 popq %r15 leaq 0x70(%rsp), %rax cmpq %r13, %rax je 0x2d4172 movq 0x8(%r13), %rax testq %rax, %rax je 0x2d4117 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2d4117 lock decl (%rax) jne 0x2d4117 movq %rsi, 0x8(%rsp) movq %r10, 0x58(%rsp) movq %r8, 0x10(%rsp) movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2d4100 movq (%rdi), %rax callq 0x18(%rax) movq 0x10(%rsp), %r8 movq 0x58(%rsp), %r10 movq 0x8(%rsp), %rsi jmp 0x2d4117 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %r8 movq 0x58(%rsp), %r10 movq 0x8(%rsp), %rsi movups (%r13), %xmm0 movaps %xmm0, 0x70(%rsp) movq 0x10(%r13), %rax movq %rax, 0x80(%rsp) movl 0x18(%r13), %eax movl %eax, 0x88(%rsp) movq 0x20(%r13), %rax movq %rax, 0x90(%rsp) movdqu 0x28(%r13), %xmm0 movdqu %xmm0, 0x98(%rsp) movl 0x38(%r13), %eax movl %eax, 0xa8(%rsp) movq 0x40(%r13), %rax movq %rax, 0xb0(%rsp) movq 0x118(%rsp), %r9 addl %r14d, %ebx addl %ebp, %r12d addl %ebx, %esi incl %esi leal 0x1(%r10,%r12), %edx movq 0x18(%rsp), %rax movl 0xd0(%rax), %ecx movq (%r9,%r15), %r9 leaq 0x70(%rsp), %rdi callq 0x63810 pushq $-0x64 popq %rbx cmpq $0x0, 0x70(%rsp) je 0x2d4c36 movslq 0xa8(%rsp), %rcx movq 0xb0(%rsp), %rax imulq %rcx, %rax testq %rax, %rax je 0x2d4c36 movq %r13, 0x150(%rsp) movq %rcx, 0x40(%rsp) movq 0x18(%rsp), %rax movl 0xd4(%rax), %r14d movl 0xd8(%rax), %r13d movslq 0xe4(%rax), %rcx movq %rcx, 0x58(%rsp) movslq 0xe8(%rax), %rcx movq %rcx, 0x60(%rsp) movl 0xdc(%rax), %ebp movl 0xe0(%rax), %ebx movl 0x114(%rax), %ecx movq %rcx, 0xc0(%rsp) movl 0x118(%rax), %eax movq %rax, 0x128(%rsp) movq 0x20(%rsp), %rax movl 0x38(%rax), %eax movl %eax, 0x10(%rsp) movl 0x9c(%rsp), %r15d movl %r13d, %eax imull %r14d, %eax movslq %eax, %r12 leaq 0x180(%rsp), %rdi leaq 0x37(%rsp), %rdx movq %r12, %rsi callq 0x73bbe imull %ebx, %r15d movl %ebp, %eax imull %r14d, %eax subl %eax, %r15d xorl %eax, %eax testl %r14d, %r14d cmovlel %eax, %r14d testl %r13d, %r13d cmovlel %eax, %r13d movq 0x180(%rsp), %rbx movl %r14d, %ecx xorl %edx, %edx xorl %esi, %esi cmpl %r13d, %eax je 0x2d42ae movslq %ecx, %rdi movslq %esi, %r8 leal (%r8,%r14), %esi cmpq %r8, %rdi je 0x2d42a4 movl %edx, (%rbx,%r8,4) incq %r8 addl %ebp, %edx jmp 0x2d4294 addl %r15d, %edx incl %eax addl %r14d, %ecx jmp 0x2d4285 movl 0x10(%rsp), %eax movl %eax, %edx movq 0xc0(%rsp), %rsi xorl %esi, %edx movq 0x40(%rsp), %r8 movl %r8d, %ecx xorl %esi, %ecx orl %edx, %ecx jne 0x2d46d2 movq 0x20(%rsp), %rdi movq 0x10(%rdi), %rax imulq 0x40(%rdi), %rax movq %rax, 0x38(%rsp) movq 0x18(%rsp), %rsi movq 0x1b0(%rsi), %rcx movslq 0x9c(%rsp), %rax movslq 0xa0(%rsp), %r9 movq 0x80(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %r10, %r11 movq %r9, %rdx imulq %rax, %rdx movq %r10, %r14 movq %rdx, 0x100(%rsp) imulq %rdx, %r14 addq $0xf, %r14 andq $-0x10, %r14 movq %r14, 0xf8(%rsp) movq %rcx, 0xc8(%rsp) testq %rcx, %rcx sete %cl movslq 0x1e8(%rsi), %rdx imulq 0x1f0(%rsi), %rdx testq %rdx, %rdx sete %bpl movslq 0x2c(%rdi), %rdx imull %eax, %r9d xorl %r13d, %r13d testl %r12d, %r12d movl $0x0, %r14d cmovgl %r12d, %r14d testl %edx, %edx movl $0x0, %r15d movq %rdx, 0x50(%rsp) cmovgl %edx, %r15d movl 0x30(%rdi), %edx testl %edx, %edx cmovlel %r13d, %edx movq %rdx, 0x8(%rsp) movq (%rdi), %rdx movq %rdx, 0xe8(%rsp) testl %r9d, %r9d cmovlel %r13d, %r9d testl %r8d, %r8d cmovlel %r13d, %r8d movq 0x60(%rsp), %rdx imulq %rax, %rdx movq 0x168(%rsi), %r13 orb %cl, %bpl movb %bpl, 0xf0(%rsp) movq %r10, 0x130(%rsp) imulq %r10, %rdx movq %rdx, 0x60(%rsp) shlq $0x2, %r12 movl 0xa4(%rsp), %eax movl %eax, 0xe0(%rsp) movq 0x70(%rsp), %rax movl 0x98(%rsp), %edx movl %edx, 0xd0(%rsp) movq %rax, 0xd8(%rsp) movq %rax, %rdi xorl %esi, %esi movq %r8, 0x40(%rsp) movq %r9, 0x48(%rsp) movq %r11, 0x68(%rsp) cmpq %r8, %rsi je 0x2d4bee movq 0xf8(%rsp), %rax xorl %edx, %edx divq 0x130(%rsp) cmpl $0x4, 0xd0(%rsp) cmoveq 0x100(%rsp), %rax pxor %xmm0, %xmm0 cmpb $0x0, 0xf0(%rsp) jne 0x2d4461 movq 0xc8(%rsp), %rcx movd (%rcx,%rsi,4), %xmm0 movq 0x38(%rsp), %rcx imulq %rsi, %rcx addq 0xe8(%rsp), %rcx movq %r11, %rdx movq %rsi, 0x28(%rsp) imulq %rsi, %rdx addq 0xd8(%rsp), %rdx imull 0xe0(%rsp), %eax testl %eax, %eax movl $0x0, %esi cmovlel %esi, %eax xorl %esi, %esi movq 0x18(%rsp), %rbp cmpl %esi, %eax je 0x2d44ad movd %xmm0, (%rdi,%rsi,4) incq %rsi jmp 0x2d449f xorl %eax, %eax movq %rdi, 0x10(%rsp) cmpq 0x8(%rsp), %rax je 0x2d4526 movq 0x60(%rsp), %rsi imulq %rax, %rsi addq %rdx, %rsi movq %rax, %rdi imulq 0x50(%rsp), %rdi leaq (%rcx,%rdi,4), %rdi xorl %r8d, %r8d cmpq %r15, %r8 je 0x2d451c movq %r8, %r9 imulq 0x58(%rsp), %r9 leaq (%rsi,%r9,4), %r9 movss (%rdi,%r8,4), %xmm0 xorl %r10d, %r10d cmpq %r10, %r14 je 0x2d4517 movss (%r13,%r10,4), %xmm1 mulss %xmm0, %xmm1 movslq (%rbx,%r10,4), %r11 addss (%r9,%r11,4), %xmm1 movss %xmm1, (%r9,%r11,4) incq %r10 jmp 0x2d44f2 incq %r8 jmp 0x2d44d7 incq %rax movq 0x10(%rsp), %rdi jmp 0x2d44b4 movq 0x120(%rbp), %rax movq %rax, 0x20(%rsp) xorl %ebp, %ebp movq 0x48(%rsp), %rcx leaq 0x1244b4(%rip), %rdx # 0x3f89f4 cmpq %rbp, %rcx je 0x2d46b5 movss (%rdi,%rbp,4), %xmm4 movq 0x128(%rsp), %rax decl %eax cmpl $0x5, %eax ja 0x2d456e movslq (%rdx,%rax,4), %rax addq %rdx, %rax jmpq %rax maxss 0x119aa2(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2d46a8 movaps %xmm4, %xmm0 movaps %xmm4, 0x140(%rsp) callq 0x5f410 addss 0x11a6fa(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x10(%rsp), %rdi leaq 0x124450(%rip), %rdx # 0x3f89f4 movq 0x48(%rsp), %rcx mulss 0x140(%rsp), %xmm0 jmp 0x2d46a8 movq 0x20(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2d46a5 jmp 0x2d46a8 movss 0x11cbda(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x119aa4(%rip), %xmm0 # 0x3ee090 cmpltss 0x11cbc7(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x10(%rsp), %rdi leaq 0x1243df(%rip), %rdx # 0x3f89f4 movq 0x48(%rsp), %rcx movaps %xmm0, %xmm1 movss 0x11a663(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2d46a8 pxor %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x11a645(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x20(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2d46a8 movq 0x20(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x119a1d(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 pxor %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2d46a8 movss 0x11a600(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x2d456e mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 movd %xmm0, (%rdi,%rbp,4) incq %rbp jmp 0x2d4540 movq 0x28(%rsp), %rsi incq %rsi movq 0x68(%rsp), %r11 addq %r11, %rdi addq %r12, %r13 movq 0x40(%rsp), %r8 jmp 0x2d441a cltd idivl %esi movl %eax, %ecx movl %r8d, %eax cltd idivl %esi movl %ecx, %edx imull %r12d, %edx xorl %edi, %edi testl %r12d, %r12d movl $0x0, %ebp cmovgl %r12d, %ebp testl %ecx, %ecx movl $0x0, %r15d cmovgl %ecx, %r15d testl %eax, %eax movl $0x0, %r8d cmovgl %eax, %r8d movq %r8, 0x100(%rsp) movslq %edx, %rdx movslq %ecx, %rcx movq %rcx, 0x160(%rsp) cltq testl %esi, %esi cmovlel %edi, %esi movl %eax, %ecx imull %edx, %ecx movl %ecx, 0x10c(%rsp) shlq $0x2, %rdx movq %rdx, 0x178(%rsp) shlq $0x2, %r12 xorl %ecx, %ecx movq $0x0, 0x120(%rsp) xorl %r14d, %r14d movq %rsi, 0xc0(%rsp) movq %rax, 0x158(%rsp) cmpq %rsi, %r14 je 0x2d4bee movl %ecx, 0x110(%rsp) movslq %ecx, %rdi shlq $0x2, %rdi movq %r14, %rcx imulq %rax, %rcx movq %rcx, 0xf8(%rsp) movslq 0x9c(%rsp), %r8 movslq 0xa0(%rsp), %r9 movq 0x80(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %r10, %r11 movq %r9, %rax imulq %r8, %rax movq %r10, %rcx movq %rax, 0xe8(%rsp) imulq %rax, %rcx addq $0xf, %rcx andq $-0x10, %rcx movq %rcx, 0xe0(%rsp) movq 0x18(%rsp), %rdx movq 0x1b0(%rdx), %rax movq %rax, 0x170(%rsp) testq %rax, %rax sete 0x8(%rsp) movslq 0x1e8(%rdx), %rcx imulq 0x1f0(%rdx), %rcx testq %rcx, %rcx sete %r13b movq 0x20(%rsp), %rcx movl 0x2c(%rcx), %eax movq %r14, 0x168(%rsp) imulq 0x160(%rsp), %r14 imull %r8d, %r9d testl %eax, %eax movl $0x0, %esi cmovlel %esi, %eax movq %rax, 0x10(%rsp) movl 0x30(%rcx), %eax testl %eax, %eax cmovlel %esi, %eax movq %rax, 0x140(%rsp) testl %r9d, %r9d cmovlel %esi, %r9d movq %r9, 0x48(%rsp) movq 0x70(%rsp), %rax orb 0x8(%rsp), %r13b movb %r13b, 0xd8(%rsp) imulq 0x60(%rsp), %r8 movq %r10, 0xf0(%rsp) imulq %r10, %r8 movq %r8, 0x40(%rsp) movq %r11, 0x68(%rsp) imulq 0x120(%rsp), %r11 addq 0x168(%rdx), %rdi movq %rdi, 0x8(%rsp) movq %rax, 0xd0(%rsp) addq %rax, %r11 movl 0xa4(%rsp), %eax movl %eax, 0xc8(%rsp) movl 0x98(%rsp), %eax movl %eax, 0x114(%rsp) xorl %edx, %edx cmpq 0x100(%rsp), %rdx je 0x2d4bb8 movq 0xf8(%rsp), %rax movq %rdx, 0x38(%rsp) leaq (%rdx,%rax), %rsi movq 0xe0(%rsp), %rax xorl %edx, %edx divq 0xf0(%rsp) cmpl $0x4, 0x114(%rsp) cmoveq 0xe8(%rsp), %rax pxor %xmm0, %xmm0 cmpb $0x0, 0xd8(%rsp) jne 0x2d4916 movq 0x170(%rsp), %rcx movd (%rcx,%rsi,4), %xmm0 imulq 0x68(%rsp), %rsi addq 0xd0(%rsp), %rsi imull 0xc8(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2d4947 movd %xmm0, (%r11,%rcx,4) incq %rcx jmp 0x2d4938 xorl %eax, %eax movq %r11, 0x50(%rsp) movq %rsi, 0x28(%rsp) cmpq 0x140(%rsp), %rax je 0x2d4a0e movq 0x40(%rsp), %rdx imulq %rax, %rdx addq %rsi, %rdx xorl %esi, %esi cmpq 0x10(%rsp), %rsi je 0x2d49fc movq %rsi, %rcx imulq 0x58(%rsp), %rcx leaq (%rdx,%rcx,4), %rdi movq 0x20(%rsp), %r10 movslq 0x2c(%r10), %rcx movq 0x10(%r10), %r9 movq 0x40(%r10), %r8 imulq %rax, %rcx imulq %r9, %rcx addq (%r10), %rcx imulq %r9, %r8 leaq (%rcx,%rsi,4), %r9 movq 0x8(%rsp), %r10 xorl %r11d, %r11d cmpq %r15, %r11 je 0x2d49f4 leaq (%r11,%r14), %rcx imulq %r8, %rcx movss (%r9,%rcx), %xmm0 xorl %ecx, %ecx cmpq %rcx, %rbp je 0x2d49ec movss (%r10,%rcx,4), %xmm1 mulss %xmm0, %xmm1 movslq (%rbx,%rcx,4), %r13 addss (%rdi,%r13,4), %xmm1 movss %xmm1, (%rdi,%r13,4) incq %rcx jmp 0x2d49c8 incq %r11 addq %r12, %r10 jmp 0x2d49b3 incq %rsi jmp 0x2d496f incq %rax movq 0x50(%rsp), %r11 movq 0x28(%rsp), %rsi jmp 0x2d4953 movq 0x18(%rsp), %rax movq 0x120(%rax), %rax movq %rax, 0x28(%rsp) xorl %r13d, %r13d movq 0x38(%rsp), %rdx cmpq %r13, 0x48(%rsp) je 0x2d4b99 movss (%r11,%r13,4), %xmm4 movq 0x128(%rsp), %rax decl %eax cmpl $0x5, %eax ja 0x2d4a5f leaq 0x123f8e(%rip), %rcx # 0x3f89dc movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq %rax maxss 0x1195b1(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2d4b8b movaps %xmm4, %xmm0 movaps %xmm4, 0x130(%rsp) callq 0x5f410 addss 0x11a209(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x38(%rsp), %rdx movq 0x50(%rsp), %r11 mulss 0x130(%rsp), %xmm0 jmp 0x2d4b8b movq 0x28(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2d4b88 jmp 0x2d4b8b movss 0x11c6f0(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x1195ba(%rip), %xmm0 # 0x3ee090 cmpltss 0x11c6dd(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x38(%rsp), %rdx movq 0x50(%rsp), %r11 movaps %xmm0, %xmm1 movss 0x11a180(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2d4b8b pxor %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x11a162(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x28(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2d4b8b movq 0x28(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x11953a(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 pxor %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2d4b8b movss 0x11a11d(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x2d4a5f mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 movd %xmm0, (%r11,%r13,4) incq %r13 jmp 0x2d4a27 incq %rdx addq 0x68(%rsp), %r11 movq 0x8(%rsp), %rax addq 0x178(%rsp), %rax movq %rax, 0x8(%rsp) jmp 0x2d48b9 movq 0x168(%rsp), %r14 incq %r14 movq 0x158(%rsp), %rax addq %rax, 0x120(%rsp) movl 0x110(%rsp), %ecx addl 0x10c(%rsp), %ecx movq 0xc0(%rsp), %rsi jmp 0x2d4763 leaq 0x180(%rsp), %rdi callq 0x624be leaq 0x70(%rsp), %rsi movq 0x18(%rsp), %rdi movq 0x150(%rsp), %rbx movq %rbx, %rdx movq 0x118(%rsp), %rcx callq 0x2d4cc4 cmpq $0x0, (%rbx) je 0x2d4c33 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax xorl %ebx, %ebx testq %rax, %rax jne 0x2d4c36 pushq $-0x64 popq %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2d4c67 lock decl (%rax) jne 0x2d4c67 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2d4c5f movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d4c67 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0x198, %rsp # imm = 0x198 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2d4cbb jmp 0x2d4c7f movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2d4cb3 lock decl (%rax) jne 0x2d4cb3 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2d4cad movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d4cb3 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/deconvolutiondepthwise.cpp
ncnn::DeconvolutionDepthWise_x86::create_group_ops(ncnn::Option const&)	int DeconvolutionDepthWise_x86::create_group_ops(const Option& opt) { // create Deconvolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Deconvolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(18, output_pad_right); pd.set(19, output_pad_bottom); pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[2]; weights[0] = weight_data_g; weights[1] = bias_data_g; op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[1]; weights[0] = weight_data_g; op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }	pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x158, %rsp # imm = 0x158 movq %rsi, 0x120(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x110(%rdi,%rdx), %eax movl 0x114(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x8(%rdi), %rax movq %rax, 0xa8(%rsp) xorl %r12d, %r12d movq 0x8(%r14), %rax movq 0x10(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x2d5365 movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x2d5360 movq (%rdi), %rax callq 0x8(%rax) incq %r12 jmp 0x2d5337 imull %r15d, %ebx cmpq %rax, %rcx je 0x2d5372 movq %rax, 0x10(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x114(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0xa8(%rsp), %rdi callq 0x6fbc2 leaq 0x120(%r14), %rax movq %rax, 0x128(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0xa4(%rsp) movslq %ebx, %rax movq %rax, 0x130(%rsp) movl %r15d, 0xc(%rsp) movslq %r15d, %rax movq %rax, 0x118(%rsp) xorl %r12d, %r12d leaq 0x148(%rsp), %r15 movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x114(%r14,%rax), %rcx cmpq %rcx, %r12 jge 0x2d5a6e movq %r12, %rcx movq 0x130(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x10(%rsp) andq $0x0, 0x18(%rsp) movq %rdx, 0x20(%rsp) movl %esi, 0x28(%rsp) movq %rax, 0x30(%rsp) pushq $0x1 popq %rax movl %eax, 0x38(%rsp) movl %edi, 0x3c(%rsp) movabsq $0x100000001, %rcx # imm = 0x100000001 movq %rcx, 0x40(%rsp) movl %eax, 0x48(%rsp) movq %rdi, 0x50(%rsp) leaq 0xd0(%rsp), %rdi leaq 0x10(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d54af lock decl (%rax) jne 0x2d54af movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d54a7 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d54af movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2d5521 movq %r12, %r13 movq 0x118(%rsp), %rcx imulq %rcx, %r13 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0xc0(%rsp) imulq %rsi, %r13 addq 0x1b0(%r14,%rax), %r13 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x8(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0xb8(%rsp) movl 0xc(%rsp), %eax movq %rax, 0xc8(%rsp) pushq $0x1 popq %rbx movq %rcx, 0xb0(%rsp) jmp 0x2d555e movq $0x0, 0xc8(%rsp) xorl %ebx, %ebx movq $0x0, 0xb8(%rsp) movl $0x0, 0x8(%rsp) movq $0x0, 0xc0(%rsp) xorl %r13d, %r13d movq $0x0, 0xb0(%rsp) pushq $0x8 popq %rdi callq 0x782bf movq %rax, %rbp movq %r15, %rdi callq 0x71548 movq %r15, %rdi xorl %esi, %esi movl 0xc(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r15, %rdi pushq $0x1 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r15, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r15, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r15, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r15, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r15, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r15, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r15, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xfc(%r14,%rax), %edx movq %r15, %rdi pushq $0x12 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r15, %rdi pushq $0x13 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r15, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r15, %rdi pushq $0x6 popq %rsi movl 0xa4(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x118(%r14,%rax), %edx movq %r15, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x128(%rsp), %rdx movq %r15, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq 0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2d575e andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) leaq 0x30(%rsp), %rax movups %xmm0, 0xc(%rax) movaps %xmm0, (%rax) andq $0x0, 0x98(%rsp) movups %xmm0, 0x34(%rax) movups %xmm0, 0x28(%rax) movups %xmm0, 0x54(%rax) movups %xmm0, 0x48(%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2d5730 lock incl (%rax) movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d57d6 lock decl (%rax) jne 0x2d57d6 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d57ce movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d57d6 andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) leaq 0x30(%rsp), %rax movups %xmm0, 0xc(%rax) movaps %xmm0, (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax leaq 0x138(%rsp), %rbx je 0x2d59b7 lock incl (%rax) movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d59b7 lock decl (%rax) jne 0x2d59b7 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d59af movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d59b7 movq %rsi, %rdi callq 0x5f3e0 movaps 0xd0(%rsp), %xmm0 movaps %xmm0, 0x10(%rsp) movq 0xe0(%rsp), %rax movq %rax, 0x20(%rsp) movl 0xe8(%rsp), %eax movl %eax, 0x28(%rsp) movq 0xf0(%rsp), %rax movq %rax, 0x30(%rsp) movups 0xf8(%rsp), %xmm0 movups %xmm0, 0x38(%rsp) movl 0x108(%rsp), %eax movl %eax, 0x48(%rsp) movq 0x110(%rsp), %rax movq %rax, 0x50(%rsp) movq 0x60(%rsp), %rax testq %rax, %rax je 0x2d585b lock decl (%rax) jne 0x2d585b movq 0x58(%rsp), %rsi movq 0x78(%rsp), %rdi testq %rdi, %rdi je 0x2d5853 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d585b movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x58(%rsp) andq $0x0, 0x60(%rsp) movq 0xc0(%rsp), %rax movq %rax, 0x68(%rsp) movl 0x8(%rsp), %eax movl %eax, 0x70(%rsp) movq 0xb8(%rsp), %rax movq %rax, 0x78(%rsp) movl %ebx, 0x80(%rsp) movq 0xc8(%rsp), %rax movl %eax, 0x84(%rsp) movl %ebx, 0x88(%rsp) movl %ebx, 0x8c(%rsp) movl %ebx, 0x90(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x98(%rsp) leaq 0x138(%rsp), %rbx movq %rbx, %rdi leaq 0x10(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %rbx, %rsi callq 0x18(%rax) movq %rbx, %rdi callq 0x6b03a pushq $0x48 popq %rbx movq 0x18(%rsp,%rbx), %rax testq %rax, %rax je 0x2d591e lock decl (%rax) jne 0x2d591e movq 0x10(%rsp,%rbx), %rsi movq 0x30(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x2d5916 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d591e movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%rbx), %rax addq $0x10, %rax andq $0x0, 0x40(%rax) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x2d58f0 movq (%rbp), %rax movq %rbp, %rdi movq 0x120(%rsp), %rsi callq 0x20(%rax) movq 0xa8(%rsp), %rax movq (%rax), %rax movq %rbp, (%rax,%r12,8) movq %r15, %rdi callq 0x71614 movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2d59a7 lock decl (%rax) jne 0x2d59a7 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x2d599f movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d59a7 movq %rsi, %rdi callq 0x5f3e0 incq %r12 jmp 0x2d53e6 movq %rsi, %rdi callq 0x5f3e0 movaps 0xd0(%rsp), %xmm0 movaps %xmm0, 0x10(%rsp) movq 0xe0(%rsp), %rax movq %rax, 0x20(%rsp) movl 0xe8(%rsp), %eax movl %eax, 0x28(%rsp) movq 0xf0(%rsp), %rax movq %rax, 0x30(%rsp) movups 0xf8(%rsp), %xmm0 movups %xmm0, 0x38(%rsp) movl 0x108(%rsp), %eax movl %eax, 0x48(%rsp) movq 0x110(%rsp), %rax movq %rax, 0x50(%rsp) movq %rbx, %rdi leaq 0x10(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %rbx, %rsi callq 0x18(%rax) movq %rbx, %rdi callq 0x6b03a movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d5947 lock decl (%rax) jne 0x2d5947 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d5a61 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d5947 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d5947 xorl %eax, %eax addq $0x158, %rsp # imm = 0x158 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2d5a95 jmp 0x2d5bed jmp 0x2d5bed jmp 0x2d5bed jmp 0x2d5aff movq %rax, %rbx jmp 0x2d5aaa movq %rax, %rbx leaq 0x138(%rsp), %rdi callq 0x6b03a movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d5ba1 lock decl (%rax) jne 0x2d5ba1 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi jne 0x2d5add movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d5ba1 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d5ba1 jmp 0x2d5bed movq %rax, %rbx leaq 0x138(%rsp), %rdi callq 0x6b03a jmp 0x2d5b02 movq %rax, %rbx pushq $0x48 popq %r14 xorps %xmm0, %xmm0 movq 0x18(%rsp,%r14), %rax testq %rax, %rax je 0x2d5b3d lock decl (%rax) jne 0x2d5b3d movq 0x10(%rsp,%r14), %rsi movq 0x30(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2d5b32 movq (%rdi), %rax callq 0x18(%rax) xorps %xmm0, %xmm0 jmp 0x2d5b3d movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x10, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2d5b09 jmp 0x2d5ba1 jmp 0x2d5bed jmp 0x2d5b97 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d5be5 lock decl (%rax) jne 0x2d5be5 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d5bd5 movq (%rdi), %rax callq 0x18(%rax) jmp 0x2d5be5 jmp 0x2d5bed movq %rax, %rbx jmp 0x2d5bae jmp 0x2d5bed movq %rax, %rbx leaq 0x148(%rsp), %rdi callq 0x71614 movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2d5be5 lock decl (%rax) jne 0x2d5be5 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi jne 0x2d5bdf movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d5be5 movq (%rdi), %rax callq 0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop	/csukuangfj[P]ncnn/src/layer/x86/deconvolutiondepthwise_x86.cpp

ncnn::im2col_sgemm_pack8to1_int8_sse_xop(ncnn::Mat const&, ncnn::Mat&, ncnn::Mat const&, ncnn::Option const&)

void im2col_sgemm_pack8to1_int8_sse_xop(const Mat& bottom_im2col, Mat& top_blob, const Mat& kernel, const Option& opt) { im2col_sgemm_pack8to1_int8_sse(bottom_im2col, top_blob, kernel, opt); }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x88, %rsp movq %rcx, %rax movq %rdx, 0x10(%rsp) movq %rdi, %r12 movslq 0x2c(%rdi), %r13 movl 0x30(%rdi), %ebx movl 0x38(%rdi), %r15d movq %rsi, 0x28(%rsp) movslq 0x38(%rsi), %rcx movq %rcx, 0x18(%rsp) leaq 0x30(%rsp), %rdi andq $0x0, 0x40(%rdi) movl %r13d, %ecx shrl %ecx movl %r13d, %r10d andl $0x1, %r10d addl %ecx, %r10d cmpq $0x2, %r13 setge %cl vpxor %xmm0, %xmm0, %xmm0 vmovdqu %xmm0, 0xc(%rdi) vmovdqa %xmm0, (%rdi) vmovdqu %xmm0, 0x2c(%rdi) vmovdqa %xmm0, 0x20(%rdi) cmovll %r13d, %r10d movl %ebx, %esi shll %cl, %esi movq 0x10(%rax), %rax movq %rax, (%rsp) pushq $0x8 popq %r8 pushq $0x8 popq %r9 movl %r15d, %edx movl %r10d, %ecx callq 0x628f2 movl %r13d, %eax sarl %eax xorl %ecx, %ecx testl %ebx, %ebx movl %ebx, %edx movl $0x0, %ebx movl %edx, 0x24(%rsp) cmovgl %edx, %ebx testl %r15d, %r15d movl $0x0, %edi cmovgl %r15d, %edi testl %eax, %eax cmovlel %ecx, %eax leaq (,%r13,8), %r8 xorl %edx, %edx cmpq %rax, %rdx je 0x138d13 movq 0x70(%rsp), %r9 imulq %rdx, %r9 imulq 0x40(%rsp), %r9 addq 0x30(%rsp), %r9 xorl %r10d, %r10d cmpq %rdi, %r10 je 0x138d0a movq 0x40(%r12), %r11 movq (%r12), %rsi imulq 0x10(%r12), %r11 addq %rcx, %rsi imulq %r10, %r11 addq %rsi, %r11 movl %ebx, %ebp subl $0x1, %ebp jb 0x138d05 vmovdqu (%r11), %xmm0 vmovdqu %xmm0, (%r9) addq $0x10, %r9 addq %r8, %r11 jmp 0x138ced incq %r10 jmp 0x138ccd incq %rdx addq $0x10, %rcx jmp 0x138cb1 movq %r13, %rcx andq $-0x2, %rcx movq 0x30(%rsp), %r9 leaq (,%rcx,8), %r10 cmpq %r13, %rcx jge 0x138d8c movl %ecx, %eax cltd pushq $0x2 popq %rsi idivl %esi addl %eax, %edx movslq %edx, %rax imulq 0x70(%rsp), %rax imulq 0x40(%rsp), %rax addq %r9, %rax movq (%r12), %rdx addq %r10, %rdx xorl %ebp, %ebp cmpq %rdi, %rbp je 0x138d83 movq 0x40(%r12), %r14 imulq 0x10(%r12), %r14 imulq %rbp, %r14 addq %rdx, %r14 movl %ebx, %esi subl $0x1, %esi jb 0x138d7e movq (%r14), %r11 movq %r11, (%rax) addq $0x8, %rax addq %r8, %r14 jmp 0x138d6a incq %rbp jmp 0x138d51 incq %rcx addq $0x8, %r10 jmp 0x138d27 movq 0x18(%rsp), %rax movl %eax, %ebx sarl $0x2, %ebx movq 0x28(%rsp), %rcx movq (%rcx), %rdx movq %rdx, 0x80(%rsp) movq 0x10(%rcx), %rdx imulq 0x40(%rcx), %rdx movq %rdx, 0x78(%rsp) imull 0x24(%rsp), %r15d xorl %esi, %esi testl %r15d, %r15d cmovlel %esi, %r15d testl %ebx, %ebx cmovlel %esi, %ebx vpxor %xmm0, %xmm0, %xmm0 cmpq %rbx, %rsi je 0x1390aa leaq (,%rsi,4), %rdi movq 0x78(%rsp), %rcx imulq %rcx, %rdi movq 0x80(%rsp), %rax addq %rax, %rdi leaq 0x1(,%rsi,4), %r8 imulq %rcx, %r8 addq %rax, %r8 leaq 0x2(,%rsi,4), %r9 imulq %rcx, %r9 addq %rax, %r9 leaq 0x3(,%rsi,4), %r10 imulq %rcx, %r10 addq %rax, %r10 movq 0x10(%rsp), %rax movq 0x40(%rax), %r14 imulq 0x10(%rax), %r14 movq 0x30(%rsp), %r12 imulq %rsi, %r14 addq (%rax), %r14 movq 0x40(%rsp), %rbp imulq 0x70(%rsp), %rbp xorl %r11d, %r11d movq %r11, %rax orq $0x1, %rax cmpq %r13, %rax jge 0x138f9f vpxor %xmm1, %xmm1, %xmm1 xorl %eax, %eax movl %r15d, %ecx vpxor %xmm2, %xmm2, %xmm2 vpxor %xmm3, %xmm3, %xmm3 vpxor %xmm4, %xmm4, %xmm4 vpxor %xmm5, %xmm5, %xmm5 vpxor %xmm6, %xmm6, %xmm6 vpxor %xmm7, %xmm7, %xmm7 vpxor %xmm8, %xmm8, %xmm8 subl $0x1, %ecx jb 0x138ef7 vmovdqu (%r12,%rax), %xmm9 vpcmpgtb %xmm9, %xmm0, %xmm10 vpunpcklbw %xmm10, %xmm9, %xmm11 # xmm11 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3],xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7] vpunpckhbw %xmm10, %xmm9, %xmm9 # xmm9 = xmm9[8],xmm10[8],xmm9[9],xmm10[9],xmm9[10],xmm10[10],xmm9[11],xmm10[11],xmm9[12],xmm10[12],xmm9[13],xmm10[13],xmm9[14],xmm10[14],xmm9[15],xmm10[15] vmovdqu (%r14,%rax,2), %xmm10 vmovdqu 0x10(%r14,%rax,2), %xmm12 vpcmpgtb %xmm10, %xmm0, %xmm13 vpcmpgtb %xmm12, %xmm0, %xmm14 vpunpcklbw %xmm13, %xmm10, %xmm15 # xmm15 = xmm10[0],xmm13[0],xmm10[1],xmm13[1],xmm10[2],xmm13[2],xmm10[3],xmm13[3],xmm10[4],xmm13[4],xmm10[5],xmm13[5],xmm10[6],xmm13[6],xmm10[7],xmm13[7] vpunpckhbw %xmm13, %xmm10, %xmm10 # xmm10 = xmm10[8],xmm13[8],xmm10[9],xmm13[9],xmm10[10],xmm13[10],xmm10[11],xmm13[11],xmm10[12],xmm13[12],xmm10[13],xmm13[13],xmm10[14],xmm13[14],xmm10[15],xmm13[15] vpunpcklbw %xmm14, %xmm12, %xmm13 # xmm13 = xmm12[0],xmm14[0],xmm12[1],xmm14[1],xmm12[2],xmm14[2],xmm12[3],xmm14[3],xmm12[4],xmm14[4],xmm12[5],xmm14[5],xmm12[6],xmm14[6],xmm12[7],xmm14[7] vpunpckhbw %xmm14, %xmm12, %xmm12 # xmm12 = xmm12[8],xmm14[8],xmm12[9],xmm14[9],xmm12[10],xmm14[10],xmm12[11],xmm14[11],xmm12[12],xmm14[12],xmm12[13],xmm14[13],xmm12[14],xmm14[14],xmm12[15],xmm14[15] vpmadcswd %xmm8, %xmm15, %xmm11, %xmm8 vpmadcswd %xmm7, %xmm10, %xmm11, %xmm7 vpmadcswd %xmm6, %xmm13, %xmm11, %xmm6 vpmadcswd %xmm5, %xmm12, %xmm11, %xmm5 vpmadcswd %xmm4, %xmm15, %xmm9, %xmm4 vpmadcswd %xmm3, %xmm10, %xmm9, %xmm3 vpmadcswd %xmm2, %xmm13, %xmm9, %xmm2 vpmadcswd %xmm1, %xmm12, %xmm9, %xmm1 addq $0x10, %rax jmp 0x138e7c vpunpckldq %xmm7, %xmm8, %xmm9 # xmm9 = xmm8[0],xmm7[0],xmm8[1],xmm7[1] vpunpckldq %xmm5, %xmm6, %xmm10 # xmm10 = xmm6[0],xmm5[0],xmm6[1],xmm5[1] vpunpckhdq %xmm7, %xmm8, %xmm7 # xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3] vpunpckhdq %xmm5, %xmm6, %xmm5 # xmm5 = xmm6[2],xmm5[2],xmm6[3],xmm5[3] vpunpckldq %xmm3, %xmm4, %xmm6 # xmm6 = xmm4[0],xmm3[0],xmm4[1],xmm3[1] vpunpckldq %xmm1, %xmm2, %xmm8 # xmm8 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] vpunpckhdq %xmm3, %xmm4, %xmm3 # xmm3 = xmm4[2],xmm3[2],xmm4[3],xmm3[3] vpunpckhdq %xmm1, %xmm2, %xmm1 # xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3] vpunpcklqdq %xmm10, %xmm9, %xmm2 # xmm2 = xmm9[0],xmm10[0] vpunpckhqdq %xmm10, %xmm9, %xmm4 # xmm4 = xmm9[1],xmm10[1] vpaddd %xmm4, %xmm2, %xmm2 vpunpcklqdq %xmm5, %xmm7, %xmm4 # xmm4 = xmm7[0],xmm5[0] vpunpckhqdq %xmm5, %xmm7, %xmm5 # xmm5 = xmm7[1],xmm5[1] vpaddd %xmm4, %xmm5, %xmm4 vpaddd %xmm4, %xmm2, %xmm2 vpunpcklqdq %xmm8, %xmm6, %xmm4 # xmm4 = xmm6[0],xmm8[0] vpunpckhqdq %xmm8, %xmm6, %xmm5 # xmm5 = xmm6[1],xmm8[1] vpaddd %xmm5, %xmm4, %xmm4 vpunpcklqdq %xmm1, %xmm3, %xmm5 # xmm5 = xmm3[0],xmm1[0] vmovd %xmm2, (%rdi) vpextrd $0x1, %xmm2, (%r8) vpextrd $0x2, %xmm2, (%r9) vpextrd $0x3, %xmm2, (%r10) vpunpckhqdq %xmm1, %xmm3, %xmm1 # xmm1 = xmm3[1],xmm1[1] vpaddd %xmm5, %xmm1, %xmm1 vpaddd %xmm1, %xmm4, %xmm1 vmovd %xmm1, 0x4(%rdi) vpextrd $0x1, %xmm1, 0x4(%r8) vpextrd $0x2, %xmm1, 0x4(%r9) vpextrd $0x3, %xmm1, 0x4(%r10) addq $0x8, %rdi addq $0x8, %r8 addq $0x8, %r9 addq $0x8, %r10 addq $0x2, %r11 addq %rbp, %r12 jmp 0x138e46 movq 0x30(%rsp), %r14 movq 0x40(%rsp), %r12 movq 0x10(%rsp), %rax movq 0x40(%rax), %rdx imulq %rsi, %rdx imulq 0x10(%rax), %rdx addq (%rax), %rdx imulq 0x70(%rsp), %r12 cmpl %r13d, %r11d jge 0x1390a2 movl %r11d, %ecx shrl %ecx movl %r11d, %eax andl $0x1, %eax addl %ecx, %eax imulq %r12, %rax addq %r14, %rax vpxor %xmm1, %xmm1, %xmm1 xorl %ecx, %ecx movq %rdx, %rbp vpxor %xmm2, %xmm2, %xmm2 vpxor %xmm3, %xmm3, %xmm3 vpxor %xmm4, %xmm4, %xmm4 cmpl %ecx, %r15d je 0x139048 vpmovsxbw (%rax,%rcx,8), %xmm5 vmovdqu (%rbp), %xmm6 vmovdqu 0x10(%rbp), %xmm7 vpcmpgtb %xmm6, %xmm0, %xmm8 vpcmpgtb %xmm7, %xmm0, %xmm9 vpunpcklbw %xmm8, %xmm6, %xmm10 # xmm10 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3],xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7] vpunpckhbw %xmm8, %xmm6, %xmm6 # xmm6 = xmm6[8],xmm8[8],xmm6[9],xmm8[9],xmm6[10],xmm8[10],xmm6[11],xmm8[11],xmm6[12],xmm8[12],xmm6[13],xmm8[13],xmm6[14],xmm8[14],xmm6[15],xmm8[15] vpunpcklbw %xmm9, %xmm7, %xmm8 # xmm8 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3],xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7] vpunpckhbw %xmm9, %xmm7, %xmm7 # xmm7 = xmm7[8],xmm9[8],xmm7[9],xmm9[9],xmm7[10],xmm9[10],xmm7[11],xmm9[11],xmm7[12],xmm9[12],xmm7[13],xmm9[13],xmm7[14],xmm9[14],xmm7[15],xmm9[15] vpmadcswd %xmm1, %xmm10, %xmm5, %xmm1 vpmadcswd %xmm2, %xmm6, %xmm5, %xmm2 vpmadcswd %xmm3, %xmm8, %xmm5, %xmm3 vpmadcswd %xmm4, %xmm7, %xmm5, %xmm4 addq $0x20, %rbp incq %rcx jmp 0x138ff6 vpunpckldq %xmm2, %xmm1, %xmm5 # xmm5 = xmm1[0],xmm2[0],xmm1[1],xmm2[1] vpunpckldq %xmm4, %xmm3, %xmm6 # xmm6 = xmm3[0],xmm4[0],xmm3[1],xmm4[1] vpunpckhdq %xmm2, %xmm1, %xmm1 # xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3] vpunpckhdq %xmm4, %xmm3, %xmm2 # xmm2 = xmm3[2],xmm4[2],xmm3[3],xmm4[3] vpunpcklqdq %xmm6, %xmm5, %xmm3 # xmm3 = xmm5[0],xmm6[0] vpunpckhqdq %xmm6, %xmm5, %xmm4 # xmm4 = xmm5[1],xmm6[1] vpaddd %xmm4, %xmm3, %xmm3 vpunpcklqdq %xmm2, %xmm1, %xmm4 # xmm4 = xmm1[0],xmm2[0] vpunpckhqdq %xmm2, %xmm1, %xmm1 # xmm1 = xmm1[1],xmm2[1] vpaddd %xmm4, %xmm1, %xmm1 vpaddd %xmm1, %xmm3, %xmm1 vmovd %xmm1, (%rdi) vpextrd $0x1, %xmm1, (%r8) vpextrd $0x2, %xmm1, (%r9) vpextrd $0x3, %xmm1, (%r10) addq $0x4, %rdi addq $0x4, %r8 addq $0x4, %r9 addq $0x4, %r10 incl %r11d jmp 0x138fc4 incq %rsi jmp 0x138dcc movq 0x18(%rsp), %rcx andq $-0x4, %rcx movq 0x28(%rsp), %rax movq 0x10(%rax), %rsi imulq 0x40(%rax), %rsi movq (%rax), %rdi movl %r15d, %r8d pushq $0x4 popq %r9 vpxor %xmm0, %xmm0, %xmm0 cmpq 0x18(%rsp), %rcx jge 0x139213 movq 0x30(%rsp), %r11 movq 0x40(%rsp), %r14 imulq 0x70(%rsp), %r14 movl %ecx, %eax cltd idivl %r9d movq %rsi, %r10 imulq %rcx, %r10 addl %eax, %edx movslq %edx, %rax movq 0x10(%rsp), %rdx movq 0x40(%rdx), %r15 imulq %rax, %r15 imulq 0x10(%rdx), %r15 addq %rdi, %r10 addq (%rdx), %r15 xorl %edx, %edx movq %rdx, %rbx orq $0x1, %rbx cmpq %r13, %rbx jge 0x139196 movq %rdx, %r12 shrq %r12 imulq %r14, %r12 addq %r11, %r12 vpxor %xmm1, %xmm1, %xmm1 xorl %ebp, %ebp vpxor %xmm2, %xmm2, %xmm2 cmpl %ebp, %r8d je 0x13916b vmovdqu (%r12), %xmm3 vpcmpgtb %xmm3, %xmm0, %xmm4 vpunpcklbw %xmm4, %xmm3, %xmm5 # xmm5 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7] vpunpckhbw %xmm4, %xmm3, %xmm3 # xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15] vpmovsxbw (%r15,%rbp,8), %xmm4 vpmadcswd %xmm1, %xmm4, %xmm5, %xmm1 vpmadcswd %xmm2, %xmm4, %xmm3, %xmm2 addq $0x10, %r12 incq %rbp jmp 0x139139 vpshufd $0xee, %xmm1, %xmm3 # xmm3 = xmm1[2,3,2,3] vpaddd %xmm1, %xmm3, %xmm1 vpshufd $0xee, %xmm2, %xmm3 # xmm3 = xmm2[2,3,2,3] vpaddd %xmm2, %xmm3, %xmm2 vphaddd %xmm2, %xmm1, %xmm1 vpshufd $0xe8, %xmm1, %xmm1 # xmm1 = xmm1[0,2,2,3] vmovq %xmm1, (%r10) addq $0x8, %r10 addq $0x2, %rdx jmp 0x139116 movq 0x30(%rsp), %r11 movq 0x40(%rsp), %r14 movq 0x10(%rsp), %rbx imulq 0x40(%rbx), %rax imulq 0x10(%rbx), %rax addq (%rbx), %rax imulq 0x70(%rsp), %r14 cmpl %r13d, %edx jge 0x13920b movl %edx, %ebx shrl %ebx movl %edx, %r15d andl $0x1, %r15d addl %ebx, %r15d imulq %r14, %r15 addq %r11, %r15 vpxor %xmm1, %xmm1, %xmm1 xorl %ebx, %ebx cmpl %ebx, %r8d je 0x1391f4 vpmovsxbw (%r15,%rbx,8), %xmm2 vpmovsxbw (%rax,%rbx,8), %xmm3 vpmadcswd %xmm1, %xmm3, %xmm2, %xmm1 incq %rbx jmp 0x1391d8 vphaddd %xmm1, %xmm1, %xmm1 vphaddd %xmm1, %xmm1, %xmm1 vmovd %xmm1, (%r10) addq $0x4, %r10 incl %edx jmp 0x1391b8 incq %rcx jmp 0x1390cf movq 0x38(%rsp), %rax testq %rax, %rax je 0x139241 lock decl (%rax) jne 0x139241 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x139239 movq (%rdi), %rax callq *0x18(%rax) jmp 0x139241 movq %rsi, %rdi callq 0x5f3e0 addq $0x88, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x13928e movq %rax, %rbx movq 0x38(%rsp), %rax testq %rax, %rax je 0x139286 lock decl (%rax) jne 0x139286 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi jne 0x139280 movq %rsi, %rdi callq 0x5f3e0 jmp 0x139286 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/x86/convolution_x86_xop.cpp

ncnn::crop_pack4_sse(ncnn::Mat const&, ncnn::Mat&, int, int)

static void crop_pack4_sse(const Mat& src, Mat& dst, int top, int left) { int w = dst.w; int h = dst.h; int right = src.w - dst.w - left; const float* ptr = src.row(top) + left * 4; float* outptr = dst; for (int y = 0; y < h; y++) { for (int x = 0; x < w; x++) { __m128 _p = _mm_loadu_ps(ptr); _mm_storeu_ps(outptr, _p); ptr += 4; outptr += 4; } ptr += (left + right) * 4; } }

movl 0x2c(%rsi), %eax movl 0x30(%rsi), %r8d movslq 0x2c(%rdi), %r9 movslq %edx, %rdx imulq %r9, %rdx subl %eax, %r9d imulq 0x10(%rdi), %rdx addq (%rdi), %rdx shll $0x2, %ecx movslq %ecx, %rcx leaq (%rdx,%rcx,4), %rcx movq (%rsi), %rdx shll $0x2, %r9d movslq %r9d, %rsi xorl %edi, %edi testl %eax, %eax cmovlel %edi, %eax testl %r8d, %r8d cmovlel %edi, %r8d shlq $0x2, %rsi movl %eax, %r9d cmpl %r8d, %edi je 0x1409f4 subl $0x1, %r9d jb 0x1409ed movups (%rcx), %xmm0 movups %xmm0, (%rdx) addq $0x10, %rcx addq $0x10, %rdx jmp 0x1409d7 addq %rsi, %rcx incl %edi jmp 0x1409cf retq nop

/csukuangfj[P]ncnn/src/layer/x86/crop_x86.cpp

ncnn::Deconvolution_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Deconvolution_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { // deconvolv with NxN kernel // value = value + bias int w = bottom_blob.w; int h = bottom_blob.h; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; // NCNN_LOGE("Deconvolution input %d x %d pad = %d %d ksize=%d %d stride=%d %d", w, h, pad_w, pad_h, kernel_w, kernel_h, stride_w, stride_h); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; int outw = (w - 1) * stride_w + kernel_extent_w + output_pad_right; int outh = (h - 1) * stride_h + kernel_extent_h + output_pad_bottom; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; int out_channels = num_output / out_elempack; Mat top_blob_bordered; if (pad_left > 0 || pad_right > 0 || pad_top > 0 || pad_bottom > 0 || (output_w > 0 && output_h > 0)) { top_blob_bordered.create(outw, outh, out_channels, out_elemsize, out_elempack, opt.workspace_allocator); } else { top_blob_bordered = top_blob; top_blob_bordered.create(outw, outh, out_channels, out_elemsize, out_elempack, opt.blob_allocator); } if (top_blob_bordered.empty()) return -100; const int maxk = kernel_w * kernel_h; if (opt.use_sgemm_convolution) { // sgemm Mat bottom_blob_2 = bottom_blob; { bottom_blob_2.w = bottom_blob.w * bottom_blob.h; bottom_blob_2.h = 1; } Mat top_col2im; Option opt_b = opt; opt_b.blob_allocator = top_blob_bordered.allocator; gemm->forward(bottom_blob_2, top_col2im, opt_b); { // col2im const int gap = (outw * stride_h - w * stride_w) * out_elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (out_elempack == 16) { #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < out_channels; p++) { const float* sptr = top_col2im.row(p * maxk); Mat outm = top_blob_bordered.channel(p); if (bias_data.empty()) { outm.fill(_mm512_setzero_ps()); } else { outm.fill(_mm512_load_ps((const float*)bias_data + p * 16)); } for (int u = 0; u < kernel_h; u++) { for (int v = 0; v < kernel_w; v++) { float* ptr = outm.row(dilation_h * u) + dilation_w * v * 16; for (int i = 0; i < h; i++) { for (int j = 0; j < w; j++) { __m512 _val = _mm512_load_ps(ptr); __m512 _s = _mm512_load_ps(sptr); _val = _mm512_add_ps(_val, _s); _mm512_store_ps(ptr, _val); ptr += stride_w * 16; sptr += 16; } ptr += gap; } } } } } #endif // __AVX512F__ if (out_elempack == 8) { #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < out_channels; p++) { const float* sptr = top_col2im.row(p * maxk); Mat outm = top_blob_bordered.channel(p); if (bias_data.empty()) { outm.fill(_mm256_setzero_ps()); } else { outm.fill(_mm256_load_ps((const float*)bias_data + p * 8)); } for (int u = 0; u < kernel_h; u++) { for (int v = 0; v < kernel_w; v++) { float* ptr = outm.row(dilation_h * u) + dilation_w * v * 8; for (int i = 0; i < h; i++) { for (int j = 0; j < w; j++) { __m256 _val = _mm256_load_ps(ptr); __m256 _s = _mm256_load_ps(sptr); _val = _mm256_add_ps(_val, _s); _mm256_store_ps(ptr, _val); ptr += stride_w * 8; sptr += 8; } ptr += gap; } } } } } #endif // __AVX__ if (out_elempack == 4) { #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < out_channels; p++) { const float* sptr = top_col2im.row(p * maxk); Mat outm = top_blob_bordered.channel(p); if (bias_data.empty()) { outm.fill(_mm_setzero_ps()); } else { outm.fill(_mm_load_ps((const float*)bias_data + p * 4)); } for (int u = 0; u < kernel_h; u++) { for (int v = 0; v < kernel_w; v++) { float* ptr = outm.row(dilation_h * u) + dilation_w * v * 4; for (int i = 0; i < h; i++) { for (int j = 0; j < w; j++) { __m128 _val = _mm_load_ps(ptr); __m128 _s = _mm_load_ps(sptr); _val = _mm_add_ps(_val, _s); _mm_store_ps(ptr, _val); ptr += stride_w * 4; sptr += 4; } ptr += gap; } } } } } #endif // __SSE2__ if (out_elempack == 1) { #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < out_channels; p++) { const float* sptr = top_col2im.row(p * maxk); Mat outm = top_blob_bordered.channel(p); const float bias = bias_data.empty() ? 0.f : bias_data[p]; outm.fill(bias); for (int u = 0; u < kernel_h; u++) { for (int v = 0; v < kernel_w; v++) { float* ptr = outm.row(dilation_h * u) + dilation_w * v; for (int i = 0; i < h; i++) { for (int j = 0; j < w; j++) { ptr[0] += sptr[0]; ptr += stride_w; sptr += 1; } ptr += gap; } } } } } } if (activation) { activation->forward_inplace(top_blob_bordered, opt); } } else { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16 && out_elempack == 16) { deconvolution_pack16_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 8 && out_elempack == 16) { deconvolution_pack8to16_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 16 && out_elempack == 8) { deconvolution_pack16to8_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 4 && out_elempack == 16) { deconvolution_pack4to16_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 16 && out_elempack == 4) { deconvolution_pack16to4_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 1 && out_elempack == 16) { deconvolution_pack1to16_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 16 && out_elempack == 1) { deconvolution_pack16to1_avx512(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } #endif // __AVX512F__ if (elempack == 8 && out_elempack == 8) { deconvolution_pack8_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 4 && out_elempack == 8) { deconvolution_pack4to8_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 8 && out_elempack == 4) { deconvolution_pack8to4_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 1 && out_elempack == 8) { deconvolution_pack1to8_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 8 && out_elempack == 1) { deconvolution_pack8to1_avx(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } #endif // __AVX__ if (elempack == 4 && out_elempack == 4) { deconvolution_pack4_sse(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 1 && out_elempack == 4) { deconvolution_pack1to4_sse(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } if (elempack == 4 && out_elempack == 1) { deconvolution_pack4to1_sse(bottom_blob, top_blob_bordered, weight_data_tm, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, activation_type, activation_params, opt); } #endif // __SSE2__ if (elempack == 1 && out_elempack == 1) { // num_output #pragma omp parallel for num_threads(opt.num_threads) for (int p = 0; p < num_output; p++) { float* outptr = top_blob_bordered.channel(p); // shadowed variable for less openmp task args const int w = bottom_blob.w; const int h = bottom_blob.h; const int channels = bottom_blob.c; const int outw = top_blob_bordered.w; const int outh = top_blob_bordered.h; for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { float sum = 0.f; if (bias_term) { sum = bias_data[p]; } const float* kptr = (const float*)weight_data_tm.channel(p); // channels for (int q = 0; q < channels; q++) { const Mat m = bottom_blob.channel(q); for (int y = 0; y < kernel_h; y++) { int sys = (i + y * dilation_h - (kernel_extent_h - 1)); if (sys < 0 || sys % stride_h != 0) continue; int sy = sys / stride_h; if (sy >= h) continue; const float* sptr = m.row(sy); for (int x = 0; x < kernel_w; x++) { int sxs = (j + x * dilation_w - (kernel_extent_w - 1)); if (sxs < 0 || sxs % stride_w != 0) continue; int sx = sxs / stride_w; if (sx >= w) continue; float val = sptr[sx]; int k = y * kernel_w + x; float w = kptr[k]; sum += val * w; } } kptr += maxk; } sum = activation_ss(sum, activation_type, activation_params); outptr[j] = sum; } outptr += outw; } } } } cut_padding(top_blob_bordered, top_blob, opt); if (top_blob.empty()) return -100; return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x258, %rsp # imm = 0x258 movq %rdx, 0x160(%rsp) movq %rdi, %r15 movq (%rdi), %rdi movq -0x18(%rdi), %rax movl 0xd4(%r15,%rax), %edx decl %edx imull 0xdc(%r15,%rax), %edx movl %edx, 0x144(%rsp) movl 0x2c(%rsi), %r13d leal -0x1(%r13), %r12d imull 0xe4(%r15,%rax), %r12d movl 0xd8(%r15,%rax), %edx decl %edx imull 0xe0(%r15,%rax), %edx movl %edx, 0x140(%rsp) movl 0x30(%rsi), %r14d leal -0x1(%r14), %ebp imull 0xe8(%r15,%rax), %ebp movl 0xd0(%r15,%rax), %r8d movl 0xfc(%r15,%rax), %r11d movl 0x100(%r15,%rax), %r10d testb $0x3, %r8b sete %r9b movq %rcx, 0x130(%rsp) andb 0x27(%rcx), %r9b movq 0x10(%rsi), %rax movq %rsi, 0x70(%rsp) movslq 0x18(%rsi), %rsi xorl %edx, %edx movq %rsi, 0x148(%rsp) divq %rsi movq %rax, %rbx movb %r9b, 0xf(%rsp) movzbl %r9b, %ecx leal (%rcx,%rcx,2), %r9d incl %r9d addb %cl, %cl shlq %cl, %rbx movl %r8d, %eax cltd idivl %r9d movq %rax, 0xa0(%rsp) andq $0x0, 0x120(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0xe0(%rsp) movups %xmm0, 0xec(%rsp) movaps %xmm0, 0x100(%rsp) movups %xmm0, 0x10c(%rsp) movq -0x18(%rdi), %rax cmpl $0x0, 0xec(%r15,%rax) pushq $0x10 popq %rsi movq %rcx, 0x10(%rsp) jg 0x154c2e cmpl $0x0, 0xf0(%r15,%rax) jg 0x154c2e cmpl $0x0, 0xf4(%r15,%rax) jg 0x154c2e cmpl $0x0, 0xf8(%r15,%rax) jle 0x156a97 addl 0x144(%rsp), %r12d addl 0x140(%rsp), %ebp addl %r11d, %r12d incl %r12d leal 0x1(%r10,%rbp), %edx movq 0x130(%rsp), %rax movq (%rax,%rsi), %rax movq %rax, (%rsp) leaq 0xe0(%rsp), %rdi movl %r12d, %esi movq 0xa0(%rsp), %rcx movq %rbx, %r8 callq 0x628f2 pushq $-0x64 popq %rbx cmpq $0x0, 0xe0(%rsp) je 0x156a4c movslq 0x118(%rsp), %rcx movq 0x120(%rsp), %rax imulq %rcx, %rax testq %rax, %rax je 0x156a4c movq %rcx, %rsi movq (%r15), %rdx movq -0x18(%rdx), %rax movslq 0xd4(%r15,%rax), %rcx movq %rax, 0x40(%rsp) movl 0xd8(%r15,%rax), %eax movl %eax, 0x48(%rsp) movq %rcx, %rdi imull %ecx, %eax movq 0x130(%rsp), %rcx cmpb $0x1, 0x1d(%rcx) cltq movq %rax, 0x158(%rsp) jne 0x154fda movq 0x70(%rsp), %rdi movq (%rdi), %rax movq %rax, 0x178(%rsp) movq 0x8(%rdi), %rcx movq %rcx, 0x180(%rsp) movq 0x10(%rdi), %rax movq %rax, 0x188(%rsp) movl 0x18(%rdi), %eax movl %eax, 0x190(%rsp) movq 0x20(%rdi), %rax movq %rax, 0x198(%rsp) movups 0x28(%rdi), %xmm0 movl 0x2c(%rdi), %edx movl 0x30(%rdi), %eax movups %xmm0, 0x1a0(%rsp) movl 0x38(%rdi), %esi movl %esi, 0x1b0(%rsp) movq 0x40(%rdi), %rsi movq %rsi, 0x1b8(%rsp) testq %rcx, %rcx je 0x154d5b lock incl (%rcx) movl 0x2c(%rdi), %edx movl 0x30(%rdi), %eax imull %edx, %eax leaq 0x178(%rsp), %rsi leaq 0x1c0(%rsp), %rdx andq $0x0, 0x40(%rdx) movl %eax, 0x2c(%rsi) movl $0x1, 0x30(%rsi) xorps %xmm0, %xmm0 movaps %xmm0, (%rdx) movups %xmm0, 0xc(%rdx) movaps %xmm0, 0x20(%rdx) movups %xmm0, 0x2c(%rdx) movq 0x130(%rsp), %rax movups (%rax), %xmm0 movups 0x10(%rax), %xmm1 movups 0x20(%rax), %xmm2 movups 0x30(%rax), %xmm3 leaq 0x210(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x100(%rsp), %rax movq %rax, 0x8(%rcx) movq 0x10(%r15), %rdi movq (%rdi), %rax callq *0x38(%rax) movq (%r15), %rsi movq -0x18(%rsi), %r8 imull 0xe8(%r15,%r8), %r12d movl 0xe4(%r15,%r8), %eax imull %r13d, %eax subl %eax, %r12d movq 0x10(%rsp), %rcx shll %cl, %r12d movslq %r12d, %rdi cmpb $0x0, 0xf(%rsp) je 0x156751 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d testl %r14d, %r14d cmovlel %ecx, %r14d movq 0xa0(%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, 0xa0(%rsp) xorps %xmm0, %xmm0 xorl %edx, %edx cmpq 0xa0(%rsp), %rdx je 0x15697a movq %rdx, %rax imulq 0x158(%rsp), %rax movslq 0x1ec(%rsp), %r10 imulq %rax, %r10 imulq 0x1d0(%rsp), %r10 addq 0x1c0(%rsp), %r10 movslq 0x10c(%rsp), %r12 movslq 0x110(%rsp), %rcx movq 0x120(%rsp), %r11 movq %rdx, 0x10(%rsp) imulq %rdx, %r11 movq 0xf0(%rsp), %rbx imulq %rbx, %r11 addq 0xe0(%rsp), %r11 movl 0x114(%rsp), %ebp imulq %r12, %rcx movq %rbx, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rbx cmpl $0x4, 0x108(%rsp) cmoveq %rcx, %rax movq 0x1a8(%r15,%r8), %rcx testq %rcx, %rcx je 0x154f17 movslq 0x1e0(%r15,%r8), %rdx imulq 0x1e8(%r15,%r8), %rdx testq %rdx, %rdx je 0x154f17 movq 0x10(%rsp), %rdx shlq $0x4, %rdx movaps (%rcx,%rdx), %xmm1 imull %eax, %ebp testl %ebp, %ebp movl $0x0, %eax cmovlel %eax, %ebp movq %r11, %rax subl $0x1, %ebp jb 0x154f35 movups %xmm1, (%rax) addq $0x10, %rax jmp 0x154f09 imull %eax, %ebp testl %ebp, %ebp movl $0x0, %eax cmovlel %eax, %ebp movq %r11, %rax subl $0x1, %ebp jb 0x154f35 movups %xmm0, (%rax) addq $0x10, %rax jmp 0x154f27 imulq %r12, %rbx movq -0x18(%rsi), %r8 xorl %eax, %eax cmpl 0xd8(%r15,%r8), %eax jge 0x154fcd xorl %edx, %edx cmpl 0xd4(%r15,%r8), %edx jge 0x154fc6 movslq 0xe0(%r15,%r8), %rcx movslq %eax, %r12 imulq %rcx, %r12 imulq %rbx, %r12 addq %r11, %r12 movl 0xdc(%r15,%r8), %ecx imull %edx, %ecx shll $0x2, %ecx movslq %ecx, %rcx leaq (%r12,%rcx,4), %r12 xorl %ebp, %ebp cmpl %r14d, %ebp je 0x154fc2 movl %r13d, %ecx subl $0x1, %ecx jb 0x154fba movaps (%r10), %xmm1 addps (%r12), %xmm1 movaps %xmm1, (%r12) movq -0x18(%rsi), %r8 movslq 0xe4(%r15,%r8), %r9 shlq $0x4, %r9 addq %r9, %r12 addq $0x10, %r10 jmp 0x154f8e leaq (%r12,%rdi,4), %r12 incl %ebp jmp 0x154f86 incl %edx jmp 0x154f4f incl %eax jmp 0x154f3f movq 0x10(%rsp), %rdx incq %rdx jmp 0x154e32 movq %rdx, 0x150(%rsp) cmpl $0x4, 0x148(%rsp) setne %al movb 0xf(%rsp), %cl xorb $0x1, %cl movb %cl, 0xf(%rsp) orb %cl, %al movq 0x70(%rsp), %rdx movl 0x48(%rsp), %r13d jne 0x15572e movq %rdi, %r14 movl %r14d, %eax movq 0x40(%rsp), %rcx movl 0xdc(%r15,%rcx), %r11d movl 0xe0(%r15,%rcx), %edi movl %edi, 0x50(%rsp) movl 0xe4(%r15,%rcx), %edi movl 0xe8(%r15,%rcx), %r8d movl %r8d, 0xa0(%rsp) movl 0x114(%r15,%rcx), %r9d movq 0x1a8(%r15,%rcx), %r10 movl %r14d, %ecx imull %r13d, %ecx shll $0x4, %ecx xorl %ebx, %ebx testl %r14d, %r14d movl $0x0, %r8d cmovgl %r14d, %r8d testl %r13d, %r13d movl $0x0, %r12d cmovgl %r13d, %r12d movq %r12, 0x30(%rsp) movslq %ecx, %rcx testl %esi, %esi cmovlel %ebx, %esi movq %rsi, 0x138(%rsp) xorl %esi, %esi shlq $0x2, %rcx movq %rcx, 0xc8(%rsp) shlq $0x6, %r14 pushq $0x1 popq %rcx subl %eax, %ecx imull %r11d, %ecx movl %ecx, 0x60(%rsp) shlq $0x6, %r8 decl %r9d movq %r9, 0xc0(%rsp) xorps %xmm0, %xmm0 movaps 0x29bf4b(%rip), %xmm15 # 0x3f1010 movaps 0x29bf53(%rip), %xmm9 # 0x3f1020 movaps 0x29bf5b(%rip), %xmm8 # 0x3f1030 movaps 0x298fc4(%rip), %xmm5 # 0x3ee0a0 movaps 0x29bf5d(%rip), %xmm6 # 0x3f1040 movq %r10, 0xb8(%rsp) movaps 0x29bf7d(%rip), %xmm11 # 0x3f1070 movaps 0x29bf85(%rip), %xmm12 # 0x3f1080 movaps 0x29bf8d(%rip), %xmm13 # 0x3f1090 cmpq 0x138(%rsp), %rsi je 0x15572e movq 0x120(%rsp), %rax imulq %rsi, %rax imulq 0xf0(%rsp), %rax addq 0xe0(%rsp), %rax movl 0x2c(%rdx), %r12d movl 0x30(%rdx), %ecx movl %ecx, 0x38(%rsp) movl 0x38(%rdx), %ecx movq %rsi, 0x80(%rsp) shlq $0x4, %rsi movq %rsi, 0x78(%rsp) xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx movq %rcx, 0x58(%rsp) movl 0x10c(%rsp), %ecx testl %ecx, %ecx cmovlel %edx, %ecx movl %ecx, 0x68(%rsp) movl 0x110(%rsp), %ecx testl %ecx, %ecx cmovlel %edx, %ecx movl %ecx, 0x64(%rsp) movl 0x50(%rsp), %esi cmpl 0x64(%rsp), %edx je 0x155719 xorl %ecx, %ecx movl 0x60(%rsp), %r9d movl %r9d, 0x20(%rsp) movl %edx, 0x10(%rsp) cmpl 0x68(%rsp), %ecx je 0x155712 testq %r10, %r10 movq %rax, 0x18(%rsp) movl %ecx, 0x28(%rsp) je 0x1551bb movq 0x78(%rsp), %rax movups (%r10,%rax), %xmm1 jmp 0x1551be xorps %xmm1, %xmm1 movq 0x18(%r15), %rax movq 0x70(%rsp), %rdx movslq 0x2c(%rdx), %rbx movq (%rdx), %rcx movq %rcx, 0x90(%rsp) movq 0x10(%rdx), %rcx movq 0x40(%rdx), %rdx imulq %rcx, %rdx movq %rdx, 0x88(%rsp) imulq %rcx, %rbx movq 0x58(%r15), %rcx imulq 0x80(%rsp), %rcx imulq 0x28(%r15), %rcx leaq (%rax,%rcx), %r9 addq $0x30, %r9 xorl %eax, %eax movq 0x30(%rsp), %rcx cmpq 0x58(%rsp), %rax je 0x155338 movq 0x88(%rsp), %rbp movq %rax, 0x98(%rsp) imulq %rax, %rbp addq 0x90(%rsp), %rbp movq %r9, 0xd0(%rsp) xorl %r10d, %r10d cmpq %rcx, %r10 je 0x155318 movl %r10d, %eax subl %r13d, %eax incl %eax imull %esi, %eax addl 0x10(%rsp), %eax js 0x15530d cltd idivl 0xa0(%rsp) testl %edx, %edx jne 0x15530d cmpl 0x38(%rsp), %eax jge 0x15530d movq %r14, %r13 movslq %eax, %rsi imulq %rbx, %rsi addq %rbp, %rsi movl 0x20(%rsp), %ecx xorl %r14d, %r14d cmpq %r14, %r8 je 0x1552fc testl %ecx, %ecx js 0x1552f3 movl %ecx, %eax cltd idivl %edi testl %edx, %edx jne 0x1552f3 cmpl %r12d, %eax jge 0x1552f3 shll $0x2, %eax cltq movss (%rsi,%rax,4), %xmm2 movss 0x4(%rsi,%rax,4), %xmm3 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps -0x30(%r9,%r14), %xmm2 movss 0x8(%rsi,%rax,4), %xmm7 addps %xmm1, %xmm2 movss 0xc(%rsi,%rax,4), %xmm1 shufps $0x0, %xmm3, %xmm3 # xmm3 = xmm3[0,0,0,0] shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] shufps $0x0, %xmm1, %xmm1 # xmm1 = xmm1[0,0,0,0] mulps -0x20(%r9,%r14), %xmm3 mulps -0x10(%r9,%r14), %xmm7 addps %xmm3, %xmm7 addps %xmm2, %xmm7 mulps (%r9,%r14), %xmm1 addps %xmm7, %xmm1 addq $0x40, %r14 addl %r11d, %ecx jmp 0x15528d movq %r13, %r14 movl 0x48(%rsp), %r13d movl 0x50(%rsp), %esi movq 0x30(%rsp), %rcx incq %r10 addq %r14, %r9 jmp 0x155241 movq 0x98(%rsp), %rax incq %rax movq 0xd0(%rsp), %r9 addq 0xc8(%rsp), %r9 jmp 0x15520f movq 0xc0(%rsp), %rax cmpl $0x5, %eax ja 0x1556fc leaq 0x29dd18(%rip), %rcx # 0x3f3068 movslq (%rcx,%rax,4), %rax addq %rcx, %rax movaps %xmm1, %xmm7 maxps %xmm0, %xmm7 movq 0xb8(%rsp), %r10 movl 0x10(%rsp), %edx jmpq *%rax movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax minps %xmm0, %xmm1 movss (%rax), %xmm2 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm2 addps %xmm2, %xmm7 jmp 0x1556e1 movaps %xmm1, %xmm2 minps %xmm15, %xmm2 maxps %xmm9, %xmm2 movaps %xmm2, %xmm14 mulps %xmm8, %xmm14 addps %xmm5, %xmm14 cvttps2dq %xmm14, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm14 andps %xmm6, %xmm14 subps %xmm14, %xmm7 movaps %xmm15, %xmm8 cvttps2dq %xmm7, %xmm15 movaps 0x29bc86(%rip), %xmm0 # 0x3f1050 mulps %xmm0, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm7 mulps %xmm2, %xmm7 movaps %xmm2, %xmm14 movaps 0x29bc7e(%rip), %xmm10 # 0x3f1060 mulps %xmm10, %xmm14 addps %xmm11, %xmm14 mulps %xmm2, %xmm14 addps %xmm12, %xmm14 mulps %xmm2, %xmm14 addps %xmm13, %xmm14 mulps %xmm2, %xmm14 xorps %xmm4, %xmm4 movaps %xmm5, %xmm0 movaps 0x29bc95(%rip), %xmm5 # 0x3f10a0 addps %xmm5, %xmm14 mulps %xmm2, %xmm14 addps %xmm0, %xmm14 mulps %xmm7, %xmm14 addps %xmm6, %xmm2 addps %xmm14, %xmm2 pslld $0x17, %xmm15 paddd %xmm6, %xmm15 mulps %xmm2, %xmm15 addps %xmm6, %xmm15 movaps %xmm15, %xmm2 maxps 0x29bc6f(%rip), %xmm15 # 0x3f10b0 movaps %xmm15, %xmm7 andps 0x29bc73(%rip), %xmm15 # 0x3f10c0 orps %xmm0, %xmm15 movaps %xmm15, %xmm3 cmpltps 0x29bc83(%rip), %xmm3 # 0x3f10e0 movaps %xmm3, %xmm14 andps %xmm15, %xmm14 movaps 0x29bc83(%rip), %xmm10 # 0x3f10f0 addps %xmm10, %xmm15 addps %xmm14, %xmm15 cmpleps %xmm4, %xmm2 psrld $0x17, %xmm7 paddd 0x29bc4a(%rip), %xmm7 # 0x3f10d0 cvtdq2ps %xmm7, %xmm14 andps %xmm6, %xmm3 subps %xmm3, %xmm14 movaps %xmm15, %xmm7 mulps 0x29bc64(%rip), %xmm7 # 0x3f1100 addps 0x29bc6d(%rip), %xmm7 # 0x3f1110 mulps %xmm15, %xmm7 addps 0x29bc72(%rip), %xmm7 # 0x3f1120 mulps %xmm15, %xmm7 addps 0x29bc77(%rip), %xmm7 # 0x3f1130 mulps %xmm15, %xmm7 addps 0x29bc7c(%rip), %xmm7 # 0x3f1140 mulps %xmm15, %xmm7 addps 0x29bc81(%rip), %xmm7 # 0x3f1150 mulps %xmm15, %xmm7 addps 0x29bc86(%rip), %xmm7 # 0x3f1160 mulps %xmm15, %xmm7 addps 0x29bc8b(%rip), %xmm7 # 0x3f1170 mulps %xmm15, %xmm7 addps 0x29bc90(%rip), %xmm7 # 0x3f1180 mulps %xmm15, %xmm7 movaps 0x29bb55(%rip), %xmm4 # 0x3f1050 mulps %xmm4, %xmm14 addps %xmm15, %xmm14 mulps %xmm15, %xmm15 addps 0x29bc82(%rip), %xmm7 # 0x3f1190 mulps %xmm15, %xmm7 movaps %xmm8, %xmm15 movaps 0x29bb12(%rip), %xmm8 # 0x3f1030 addps %xmm7, %xmm14 mulps 0x29c796(%rip), %xmm14 # 0x3f1cc0 movaps %xmm2, %xmm3 andnps %xmm14, %xmm3 andps 0x29c798(%rip), %xmm2 # 0x3f1cd0 orps %xmm3, %xmm2 minps %xmm15, %xmm2 maxps %xmm9, %xmm2 movaps %xmm2, %xmm3 mulps %xmm8, %xmm3 addps %xmm0, %xmm3 cvttps2dq %xmm3, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm3 andps %xmm6, %xmm3 subps %xmm3, %xmm7 cvttps2dq %xmm7, %xmm14 mulps %xmm4, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm3 mulps %xmm2, %xmm3 movaps %xmm2, %xmm7 mulps 0x29bae7(%rip), %xmm7 # 0x3f1060 addps %xmm11, %xmm7 mulps %xmm2, %xmm7 addps %xmm12, %xmm7 mulps %xmm2, %xmm7 addps %xmm13, %xmm7 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 movaps %xmm0, %xmm5 xorps %xmm0, %xmm0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm3, %xmm7 addps %xmm6, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm6, %xmm14 mulps %xmm2, %xmm14 addps %xmm6, %xmm14 rcpps %xmm14, %xmm2 movaps %xmm2, %xmm7 addps %xmm2, %xmm7 mulps %xmm7, %xmm14 movaps 0x29c712(%rip), %xmm3 # 0x3f1ce0 subps %xmm14, %xmm3 mulps %xmm2, %xmm3 addps %xmm10, %xmm7 addps %xmm3, %xmm7 jmp 0x1556de movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax movss (%rax), %xmm2 movss 0x4(%rax), %xmm7 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] maxps %xmm2, %xmm1 minps %xmm7, %xmm1 movaps %xmm1, %xmm7 jmp 0x1556e1 xorps 0x298a7c(%rip), %xmm1 # 0x3ee090 minps %xmm15, %xmm1 maxps %xmm9, %xmm1 movaps %xmm1, %xmm7 mulps %xmm8, %xmm7 addps %xmm5, %xmm7 cvttps2dq %xmm7, %xmm2 cvtdq2ps %xmm2, %xmm2 cmpltps %xmm2, %xmm7 andps %xmm6, %xmm7 subps %xmm7, %xmm2 cvttps2dq %xmm2, %xmm14 mulps 0x29c6ad(%rip), %xmm2 # 0x3f1cf0 addps %xmm1, %xmm2 movaps %xmm2, %xmm1 mulps %xmm2, %xmm1 movaps %xmm2, %xmm7 mulps 0x29ba0a(%rip), %xmm7 # 0x3f1060 addps 0x29ba13(%rip), %xmm7 # 0x3f1070 mulps %xmm2, %xmm7 addps 0x29ba19(%rip), %xmm7 # 0x3f1080 mulps %xmm2, %xmm7 addps 0x29ba1f(%rip), %xmm7 # 0x3f1090 mulps %xmm2, %xmm7 addps 0x29ba25(%rip), %xmm7 # 0x3f10a0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm1, %xmm7 addps %xmm6, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm6, %xmm14 mulps %xmm2, %xmm14 addps %xmm6, %xmm14 rcpps %xmm14, %xmm1 mulps %xmm1, %xmm14 movaps %xmm6, %xmm7 subps %xmm14, %xmm7 mulps %xmm1, %xmm7 addps %xmm1, %xmm7 jmp 0x1556e1 movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax movss (%rax), %xmm7 movss 0x4(%rax), %xmm2 shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm7 addps %xmm2, %xmm7 maxps %xmm0, %xmm7 minps %xmm6, %xmm7 mulps %xmm1, %xmm7 movq 0x18(%rsp), %rax movl 0x28(%rsp), %ecx movups %xmm7, (%rax) addq $0x10, %rax incl %ecx incl 0x20(%rsp) jmp 0x155197 movaps %xmm1, %xmm7 movq 0xb8(%rsp), %r10 movq 0x18(%rsp), %rax movl 0x10(%rsp), %edx jmp 0x1556e6 incl %edx jmp 0x15517d movq 0x80(%rsp), %rsi incq %rsi movq 0x70(%rsp), %rdx jmp 0x155103 cmpl $0x1, 0x148(%rsp) setne %al orb 0xf(%rsp), %al jne 0x155e33 movq 0x150(%rsp), %rax movq -0x18(%rax), %rcx movslq 0xd4(%r15,%rcx), %rax movl 0xd8(%r15,%rcx), %ebp movl 0xdc(%r15,%rcx), %r8d movl 0xe0(%r15,%rcx), %r14d movl 0xe4(%r15,%rcx), %r10d movl 0xe8(%r15,%rcx), %ebx movl 0x114(%r15,%rcx), %edi movl 0x118(%rsp), %r9d movq %rcx, 0x40(%rsp) movq 0x1a8(%r15,%rcx), %rcx movq %rcx, 0xc0(%rsp) movl %eax, %ecx imull %ebp, %ecx shll $0x2, %ecx movslq %ecx, %rcx xorl %r11d, %r11d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi testl %ebp, %ebp movl $0x0, %r13d cmovgl %ebp, %r13d testl %r9d, %r9d cmovlel %r11d, %r9d movq %r9, 0x138(%rsp) shlq $0x2, %rcx movq %rcx, 0x18(%rsp) pushq $0x1 popq %rcx subl %eax, %ecx movq %rax, %r11 xorl %eax, %eax shlq $0x4, %r11 imull %r8d, %ecx movl %ecx, 0x60(%rsp) shlq $0x4, %rsi decl %edi xorps %xmm0, %xmm0 movaps 0x29b809(%rip), %xmm15 # 0x3f1010 movaps 0x29b811(%rip), %xmm9 # 0x3f1020 movaps 0x29b819(%rip), %xmm8 # 0x3f1030 movaps 0x298882(%rip), %xmm5 # 0x3ee0a0 movaps 0x29b81b(%rip), %xmm6 # 0x3f1040 movq %rdi, 0xb8(%rsp) movq %r13, 0x50(%rsp) movq %r11, 0x48(%rsp) movaps 0x29b831(%rip), %xmm11 # 0x3f1070 movaps 0x29b839(%rip), %xmm12 # 0x3f1080 movaps 0x29b841(%rip), %xmm13 # 0x3f1090 cmpq 0x138(%rsp), %rax je 0x155e33 movq 0x120(%rsp), %rcx imulq %rax, %rcx imulq 0xf0(%rsp), %rcx addq 0xe0(%rsp), %rcx movq %rcx, 0x28(%rsp) movl 0x2c(%rdx), %ecx movl %ecx, 0x10(%rsp) movl 0x30(%rdx), %ecx movl %ecx, 0xa0(%rsp) movl 0x38(%rdx), %ecx movq %rax, 0x80(%rsp) shlq $0x4, %rax movq %rax, 0x78(%rsp) xorl %r12d, %r12d testl %ecx, %ecx cmovlel %r12d, %ecx movq %rcx, 0x58(%rsp) movl 0x10c(%rsp), %eax testl %eax, %eax cmovlel %r12d, %eax movl %eax, 0x68(%rsp) movl 0x110(%rsp), %eax testl %eax, %eax cmovlel %r12d, %eax movl %eax, 0x64(%rsp) movq 0xc0(%rsp), %rax cmpl 0x64(%rsp), %r12d je 0x155e1e xorl %edx, %edx movl 0x60(%rsp), %ecx movl %ecx, 0x20(%rsp) cmpl 0x68(%rsp), %edx je 0x155e16 testq %rax, %rax je 0x15590b movq 0x78(%rsp), %rcx movups (%rax,%rcx), %xmm1 jmp 0x15590e xorps %xmm1, %xmm1 movq 0x70(%rsp), %rcx movslq 0x2c(%rcx), %rdi movq (%rcx), %rax movq %rax, 0x88(%rsp) movq 0x10(%rcx), %rax movq 0x40(%rcx), %rcx imulq %rax, %rcx movq %rcx, 0xc8(%rsp) movq 0x58(%r15), %rcx imulq 0x80(%rsp), %rcx imulq 0x28(%r15), %rcx imulq %rax, %rdi movq %rdi, 0x30(%rsp) addq 0x18(%r15), %rcx xorl %eax, %eax movl %edx, 0x90(%rsp) cmpq 0x58(%rsp), %rax je 0x155a55 movq 0xc8(%rsp), %rdx movq %rax, 0x98(%rsp) imulq %rax, %rdx addq 0x88(%rsp), %rdx movq %rdx, 0x38(%rsp) movq %rcx, 0xd0(%rsp) movq %rcx, %rdi xorl %r9d, %r9d cmpq %r13, %r9 je 0x155a31 movl %r9d, %eax subl %ebp, %eax incl %eax imull %r14d, %eax addl %r12d, %eax js 0x155a26 cltd idivl %ebx testl %edx, %edx jne 0x155a26 cmpl 0xa0(%rsp), %eax jge 0x155a26 movl %ebx, %r11d movl %r14d, %ebx movl %ebp, %r13d movslq %eax, %r14 imulq 0x30(%rsp), %r14 addq 0x38(%rsp), %r14 movl 0x20(%rsp), %ecx xorl %ebp, %ebp cmpq %rbp, %rsi je 0x155a13 testl %ecx, %ecx js 0x155a0a movl %ecx, %eax cltd idivl %r10d testl %edx, %edx jne 0x155a0a cmpl 0x10(%rsp), %eax jge 0x155a0a cltq movss (%r14,%rax,4), %xmm2 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps (%rdi,%rbp), %xmm2 addps %xmm2, %xmm1 addq $0x10, %rbp addl %r8d, %ecx jmp 0x1559de movl %r13d, %ebp movl %ebx, %r14d movl %r11d, %ebx movq 0x50(%rsp), %r13 movq 0x48(%rsp), %r11 incq %r9 addq %r11, %rdi jmp 0x155998 movq 0x98(%rsp), %rax incq %rax movq 0xd0(%rsp), %rcx addq 0x18(%rsp), %rcx movl 0x90(%rsp), %edx jmp 0x15595e movq 0xb8(%rsp), %rax cmpl $0x5, %eax ja 0x155d16 leaq 0x29d613(%rip), %rcx # 0x3f3080 movslq (%rcx,%rax,4), %rax addq %rcx, %rax movaps %xmm1, %xmm7 maxps %xmm0, %xmm7 jmpq *%rax movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax minps %xmm0, %xmm1 movss (%rax), %xmm2 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm2 addps %xmm2, %xmm7 jmp 0x155df2 movaps %xmm1, %xmm2 minps %xmm15, %xmm2 maxps %xmm9, %xmm2 movaps %xmm2, %xmm14 mulps %xmm8, %xmm14 addps %xmm5, %xmm14 cvttps2dq %xmm14, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm14 andps %xmm6, %xmm14 subps %xmm14, %xmm7 movaps %xmm15, %xmm8 cvttps2dq %xmm7, %xmm15 movaps 0x29b575(%rip), %xmm0 # 0x3f1050 mulps %xmm0, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm7 mulps %xmm2, %xmm7 movaps %xmm2, %xmm14 movaps 0x29b56d(%rip), %xmm10 # 0x3f1060 mulps %xmm10, %xmm14 addps %xmm11, %xmm14 mulps %xmm2, %xmm14 addps %xmm12, %xmm14 mulps %xmm2, %xmm14 addps %xmm13, %xmm14 mulps %xmm2, %xmm14 xorps %xmm4, %xmm4 movaps %xmm5, %xmm0 movaps 0x29b584(%rip), %xmm5 # 0x3f10a0 addps %xmm5, %xmm14 mulps %xmm2, %xmm14 addps %xmm0, %xmm14 mulps %xmm7, %xmm14 addps %xmm6, %xmm2 addps %xmm14, %xmm2 pslld $0x17, %xmm15 paddd %xmm6, %xmm15 mulps %xmm2, %xmm15 addps %xmm6, %xmm15 movaps %xmm15, %xmm2 maxps 0x29b55e(%rip), %xmm15 # 0x3f10b0 movaps %xmm15, %xmm7 andps 0x29b562(%rip), %xmm15 # 0x3f10c0 orps %xmm0, %xmm15 movaps %xmm15, %xmm3 cmpltps 0x29b572(%rip), %xmm3 # 0x3f10e0 movaps %xmm3, %xmm14 andps %xmm15, %xmm14 movaps 0x29b572(%rip), %xmm10 # 0x3f10f0 addps %xmm10, %xmm15 addps %xmm14, %xmm15 cmpleps %xmm4, %xmm2 psrld $0x17, %xmm7 paddd 0x29b539(%rip), %xmm7 # 0x3f10d0 cvtdq2ps %xmm7, %xmm14 andps %xmm6, %xmm3 subps %xmm3, %xmm14 movaps %xmm15, %xmm7 mulps 0x29b553(%rip), %xmm7 # 0x3f1100 addps 0x29b55c(%rip), %xmm7 # 0x3f1110 mulps %xmm15, %xmm7 addps 0x29b561(%rip), %xmm7 # 0x3f1120 mulps %xmm15, %xmm7 addps 0x29b566(%rip), %xmm7 # 0x3f1130 mulps %xmm15, %xmm7 addps 0x29b56b(%rip), %xmm7 # 0x3f1140 mulps %xmm15, %xmm7 addps 0x29b570(%rip), %xmm7 # 0x3f1150 mulps %xmm15, %xmm7 addps 0x29b575(%rip), %xmm7 # 0x3f1160 mulps %xmm15, %xmm7 addps 0x29b57a(%rip), %xmm7 # 0x3f1170 mulps %xmm15, %xmm7 addps 0x29b57f(%rip), %xmm7 # 0x3f1180 mulps %xmm15, %xmm7 movaps 0x29b444(%rip), %xmm4 # 0x3f1050 mulps %xmm4, %xmm14 addps %xmm15, %xmm14 mulps %xmm15, %xmm15 addps 0x29b571(%rip), %xmm7 # 0x3f1190 mulps %xmm15, %xmm7 movaps %xmm8, %xmm15 movaps 0x29b401(%rip), %xmm8 # 0x3f1030 addps %xmm7, %xmm14 mulps 0x29c085(%rip), %xmm14 # 0x3f1cc0 movaps %xmm2, %xmm3 andnps %xmm14, %xmm3 andps 0x29c087(%rip), %xmm2 # 0x3f1cd0 orps %xmm3, %xmm2 minps %xmm15, %xmm2 maxps %xmm9, %xmm2 movaps %xmm2, %xmm3 mulps %xmm8, %xmm3 addps %xmm0, %xmm3 cvttps2dq %xmm3, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm3 andps %xmm6, %xmm3 subps %xmm3, %xmm7 cvttps2dq %xmm7, %xmm14 mulps %xmm4, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm3 mulps %xmm2, %xmm3 movaps %xmm2, %xmm7 mulps 0x29b3d6(%rip), %xmm7 # 0x3f1060 addps %xmm11, %xmm7 mulps %xmm2, %xmm7 addps %xmm12, %xmm7 mulps %xmm2, %xmm7 addps %xmm13, %xmm7 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 movaps %xmm0, %xmm5 xorps %xmm0, %xmm0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm3, %xmm7 addps %xmm6, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm6, %xmm14 mulps %xmm2, %xmm14 addps %xmm6, %xmm14 rcpps %xmm14, %xmm2 movaps %xmm2, %xmm7 addps %xmm2, %xmm7 mulps %xmm7, %xmm14 movaps 0x29c001(%rip), %xmm3 # 0x3f1ce0 subps %xmm14, %xmm3 mulps %xmm2, %xmm3 addps %xmm10, %xmm7 addps %xmm3, %xmm7 jmp 0x155def movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax movss (%rax), %xmm2 movss 0x4(%rax), %xmm7 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] maxps %xmm2, %xmm1 minps %xmm7, %xmm1 movaps %xmm1, %xmm7 jmp 0x155df2 xorps 0x29836b(%rip), %xmm1 # 0x3ee090 minps %xmm15, %xmm1 maxps %xmm9, %xmm1 movaps %xmm1, %xmm7 mulps %xmm8, %xmm7 addps %xmm5, %xmm7 cvttps2dq %xmm7, %xmm2 cvtdq2ps %xmm2, %xmm2 cmpltps %xmm2, %xmm7 andps %xmm6, %xmm7 subps %xmm7, %xmm2 cvttps2dq %xmm2, %xmm14 mulps 0x29bf9c(%rip), %xmm2 # 0x3f1cf0 addps %xmm1, %xmm2 movaps %xmm2, %xmm1 mulps %xmm2, %xmm1 movaps %xmm2, %xmm7 mulps 0x29b2f9(%rip), %xmm7 # 0x3f1060 addps 0x29b302(%rip), %xmm7 # 0x3f1070 mulps %xmm2, %xmm7 addps 0x29b308(%rip), %xmm7 # 0x3f1080 mulps %xmm2, %xmm7 addps 0x29b30e(%rip), %xmm7 # 0x3f1090 mulps %xmm2, %xmm7 addps 0x29b314(%rip), %xmm7 # 0x3f10a0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm1, %xmm7 addps %xmm6, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm6, %xmm14 mulps %xmm2, %xmm14 addps %xmm6, %xmm14 rcpps %xmm14, %xmm1 mulps %xmm1, %xmm14 movaps %xmm6, %xmm7 subps %xmm14, %xmm7 mulps %xmm1, %xmm7 addps %xmm1, %xmm7 jmp 0x155df2 movq 0x40(%rsp), %rax movq 0x118(%r15,%rax), %rax movss (%rax), %xmm7 movss 0x4(%rax), %xmm2 shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm7 addps %xmm2, %xmm7 maxps %xmm0, %xmm7 minps %xmm6, %xmm7 mulps %xmm1, %xmm7 movq 0x28(%rsp), %rax movups %xmm7, (%rax) addq $0x10, %rax movq %rax, 0x28(%rsp) incl %edx incl 0x20(%rsp) movq 0xc0(%rsp), %rax jmp 0x1558f1 incl %r12d jmp 0x1558dc movq 0x80(%rsp), %rax incq %rax movq 0x70(%rsp), %rdx jmp 0x15584f cmpl $0x4, 0x148(%rsp) jne 0x1562f5 cmpb $0x0, 0xf(%rsp) je 0x1562f5 movq 0x150(%rsp), %rax movq -0x18(%rax), %rsi movslq 0xd4(%r15,%rsi), %rax movl 0xd8(%r15,%rsi), %edi movl 0xdc(%r15,%rsi), %r14d movl 0xe0(%r15,%rsi), %r10d movl 0x114(%r15,%rsi), %r8d movl 0x118(%rsp), %r9d movq 0xf0(%rsp), %rcx imulq 0x120(%rsp), %rcx movq %rcx, 0x170(%rsp) movl %eax, %ecx imull %edi, %ecx shll $0x2, %ecx xorl %r11d, %r11d testl %eax, %eax movl $0x0, %ebx cmovgl %eax, %ebx testl %edi, %edi movl $0x0, %edx cmovgl %edi, %edx movq %rdx, 0x30(%rsp) movq 0x70(%rsp), %rdx movl 0x38(%rdx), %r12d testl %r12d, %r12d cmovlel %r11d, %r12d movq %r12, 0x58(%rsp) movl 0x10c(%rsp), %ebp testl %ebp, %ebp cmovlel %r11d, %ebp movl %ebp, 0x80(%rsp) movl 0x110(%rsp), %ebp testl %ebp, %ebp cmovlel %r11d, %ebp movl %ebp, 0x64(%rsp) movslq %ecx, %rcx testl %r9d, %r9d cmovlel %r11d, %r9d movq %r9, 0x150(%rsp) shlq $0x2, %rcx movq %rcx, 0xc8(%rsp) pushq $0x1 popq %rcx subl %eax, %ecx shlq $0x4, %rax movq %rax, 0x10(%rsp) imull %r14d, %ecx movl %ecx, 0x60(%rsp) shlq $0x4, %rbx decl %r8d movq %r8, 0xb8(%rsp) movl 0xe4(%r15,%rsi), %r12d movl 0xe8(%r15,%rsi), %eax movl %eax, 0xa0(%rsp) movq %rsi, 0x138(%rsp) movq 0x1a8(%r15,%rsi), %rcx movq 0xe0(%rsp), %rax movq %rax, 0x168(%rsp) movl 0x2c(%rdx), %r13d movl 0x30(%rdx), %eax movl %eax, 0x38(%rsp) movl %edi, 0x50(%rsp) movl %r10d, 0x48(%rsp) movq %rcx, 0x68(%rsp) cmpq 0x150(%rsp), %r11 je 0x1562f5 movq 0x170(%rsp), %rdx movq %r11, 0x78(%rsp) imulq %r11, %rdx addq 0x168(%rsp), %rdx movq %rdx, 0x18(%rsp) xorl %ebp, %ebp cmpl 0x64(%rsp), %ebp je 0x1562e8 movq 0x138(%rsp), %rax movq 0x118(%r15,%rax), %rax movq %rax, 0x40(%rsp) movq 0x58(%r15), %rax imulq 0x78(%rsp), %rax imulq 0x28(%r15), %rax addq 0x18(%r15), %rax movq %rax, 0xc0(%rsp) xorl %eax, %eax movl 0x60(%rsp), %edx movl %edx, 0x20(%rsp) cmpl 0x80(%rsp), %eax je 0x1562e1 movl %eax, 0x28(%rsp) testq %rcx, %rcx je 0x156026 movq 0x78(%rsp), %rax movss (%rcx,%rax,4), %xmm0 jmp 0x156029 xorps %xmm0, %xmm0 movq 0x70(%rsp), %rcx movslq 0x2c(%rcx), %rsi movq (%rcx), %rax movq %rax, 0x90(%rsp) movq 0x10(%rcx), %rax movq 0x40(%rcx), %rcx imulq %rax, %rcx movq %rcx, 0x88(%rsp) imulq %rax, %rsi xorps %xmm1, %xmm1 movq 0xc0(%rsp), %r9 xorl %eax, %eax movq 0x30(%rsp), %rcx cmpq 0x58(%rsp), %rax je 0x15613d movq 0x88(%rsp), %r11 movq %rax, 0x98(%rsp) imulq %rax, %r11 addq 0x90(%rsp), %r11 movq %r9, 0xd0(%rsp) xorl %r8d, %r8d cmpq %rcx, %r8 je 0x15611d movl %r8d, %eax subl %edi, %eax incl %eax imull %r10d, %eax addl %ebp, %eax js 0x156110 cltd idivl 0xa0(%rsp) testl %edx, %edx jne 0x156110 cmpl 0x38(%rsp), %eax jge 0x156110 movslq %eax, %rdi imulq %rsi, %rdi addq %r11, %rdi movl 0x20(%rsp), %ecx xorl %r10d, %r10d cmpq %r10, %rbx je 0x156102 testl %ecx, %ecx js 0x1560f9 movl %ecx, %eax cltd idivl %r12d testl %edx, %edx jne 0x1560f9 cmpl %r13d, %eax jge 0x1560f9 shll $0x2, %eax cltq movaps (%r9,%r10), %xmm2 mulps (%rdi,%rax,4), %xmm2 addps %xmm2, %xmm1 addq $0x10, %r10 addl %r14d, %ecx jmp 0x1560d0 movl 0x50(%rsp), %edi movl 0x48(%rsp), %r10d movq 0x30(%rsp), %rcx incq %r8 addq 0x10(%rsp), %r9 jmp 0x156099 movq 0x98(%rsp), %rax incq %rax movq 0xd0(%rsp), %r9 addq 0xc8(%rsp), %r9 jmp 0x156067 movaps %xmm1, %xmm4 unpckhpd %xmm1, %xmm4 # xmm4 = xmm4[1],xmm1[1] addps %xmm1, %xmm4 addss %xmm4, %xmm0 shufps $0x55, %xmm4, %xmm4 # xmm4 = xmm4[1,1,1,1] addss %xmm0, %xmm4 movq 0xb8(%rsp), %rax cmpl $0x5, %eax ja 0x1562b8 leaq 0x29cf2d(%rip), %rcx # 0x3f3098 movslq (%rcx,%rax,4), %rax addq %rcx, %rax movq 0x68(%rsp), %rcx jmpq *%rax maxss 0x297e8f(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x1562c0 movaps %xmm4, %xmm0 movaps %xmm4, 0xd0(%rsp) callq 0x5f410 addss 0x298ae7(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x68(%rsp), %rcx movl 0x48(%rsp), %r10d movl 0x50(%rsp), %edi mulss 0xd0(%rsp), %xmm0 jmp 0x1562c0 movq 0x40(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x1562b3 jmp 0x1562c0 movss 0x29afca(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x297e94(%rip), %xmm0 # 0x3ee090 cmpltss 0x29afb7(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x68(%rsp), %rcx movl 0x48(%rsp), %r10d movl 0x50(%rsp), %edi movaps %xmm0, %xmm1 movss 0x298a56(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x1562c0 xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x298a36(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x40(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x1562c0 movq 0x40(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x297e0e(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x1562c0 movss 0x2989f2(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x156181 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 jmp 0x1562c0 movaps %xmm4, %xmm0 movq 0x68(%rsp), %rcx movq 0x18(%rsp), %rax movss %xmm0, (%rax) addq $0x4, %rax movq %rax, 0x18(%rsp) movl 0x28(%rsp), %eax incl %eax incl 0x20(%rsp) jmp 0x156004 incl %ebp jmp 0x155fc0 movq 0x78(%rsp), %r11 incq %r11 jmp 0x155f92 cmpl $0x1, 0x148(%rsp) setne %al movb 0xf(%rsp), %cl xorb $0x1, %cl orb %al, %cl movq 0x70(%rsp), %rax jne 0x156a07 movq (%r15), %rcx movq %rcx, 0xc0(%rsp) movq 0xe0(%rsp), %rcx movq %rcx, 0xb8(%rsp) movq 0x120(%rsp), %rcx imulq 0xf0(%rsp), %rcx movq %rcx, 0x78(%rsp) movl 0x2c(%rax), %ebp movl 0x30(%rax), %ecx movl %ecx, 0x10(%rsp) movl 0x38(%rax), %eax movslq 0x10c(%rsp), %rcx xorl %edx, %edx testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x20(%rsp) testl %ecx, %ecx movl $0x0, %esi movq %rcx, 0x80(%rsp) cmovgl %ecx, %esi movq %rsi, 0x90(%rsp) movl 0x110(%rsp), %eax testl %eax, %eax cmovlel %edx, %eax movl %eax, 0x68(%rsp) shlq $0x2, 0x158(%rsp) movq 0xc0(%rsp), %rax movq -0x18(%rax), %rax movslq 0xd0(%r15,%rax), %rax cmpq %rax, %rdx jge 0x156a07 movq 0x78(%rsp), %rax imulq %rdx, %rax addq 0xb8(%rsp), %rax movq %rax, 0x58(%rsp) xorl %r12d, %r12d movq %rdx, 0x18(%rsp) cmpl 0x68(%rsp), %r12d je 0x156749 movq (%r15), %rax movq %rax, 0x88(%rsp) movl %r12d, 0x28(%rsp) movq 0x58(%r15), %rcx imulq %rdx, %rcx imulq 0x28(%r15), %rcx subl 0x140(%rsp), %r12d addq 0x18(%r15), %rcx movq %rcx, 0xc8(%rsp) xorl %edx, %edx cmpq 0x90(%rsp), %rdx je 0x156721 movq 0x88(%rsp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r15,%rax) movq %rax, 0x98(%rsp) je 0x156453 movq 0x1a8(%r15,%rax), %rax movq 0x18(%rsp), %rcx movss (%rax,%rcx,4), %xmm4 jmp 0x156456 xorps %xmm4, %xmm4 movq 0x70(%rsp), %rcx movslq 0x2c(%rcx), %rsi movq (%rcx), %rax movq %rax, 0x30(%rsp) movq 0x10(%rcx), %rax movq 0x40(%rcx), %rcx imulq %rax, %rcx movq %rcx, 0x50(%rsp) movq (%r15), %rcx movq %rcx, 0x48(%rsp) imulq %rax, %rsi movq %rsi, 0xa0(%rsp) movq %rdx, 0xd0(%rsp) movl %edx, %r10d subl 0x144(%rsp), %r10d movq 0xc8(%rsp), %r11 xorl %eax, %eax cmpq 0x20(%rsp), %rax je 0x156595 movq 0x50(%rsp), %r8 movq %rax, 0x38(%rsp) imulq %rax, %r8 addq 0x30(%rsp), %r8 movq 0x48(%rsp), %rax movq -0x18(%rax), %r14 movl 0xd8(%r15,%r14), %edi xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x156580 movl 0xe0(%r15,%r14), %eax imull %r9d, %eax addl %r12d, %eax js 0x156578 cltd idivl 0xe8(%r15,%r14) testl %edx, %edx jne 0x156578 cmpl 0x10(%rsp), %eax jge 0x156578 movslq %eax, %rcx imulq 0xa0(%rsp), %rcx addq %r8, %rcx movl 0xd4(%r15,%r14), %eax testl %eax, %eax movl $0x0, %ebx cmovgl %eax, %ebx imull %r9d, %eax cltq leaq (%r11,%rax,4), %r13 xorl %esi, %esi cmpq %rsi, %rbx je 0x156578 movl 0xdc(%r15,%r14), %eax imull %esi, %eax addl %r10d, %eax js 0x156573 cltd idivl 0xe4(%r15,%r14) testl %edx, %edx jne 0x156573 cmpl %ebp, %eax jge 0x156573 cltq movss (%r13,%rsi,4), %xmm0 mulss (%rcx,%rax,4), %xmm0 addss %xmm0, %xmm4 incq %rsi jmp 0x15653b incl %r9d jmp 0x1564e1 movq 0x38(%rsp), %rax incq %rax addq 0x158(%rsp), %r11 jmp 0x1564a9 movq 0x98(%rsp), %rsi movl 0x114(%r15,%rsi), %eax decl %eax cmpl $0x5, %eax ja 0x156704 leaq 0x29caf9(%rip), %rcx # 0x3f30b0 movslq (%rcx,%rax,4), %rax addq %rcx, %rax movq 0xd0(%rsp), %rdx jmpq *%rax maxss 0x297a40(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x15670f movaps %xmm4, %xmm0 movaps %xmm4, 0xa0(%rsp) callq 0x5f410 addss 0x298698(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0xd0(%rsp), %rdx mulss 0xa0(%rsp), %xmm0 jmp 0x15670f movq 0x118(%r15,%rsi), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x1566ff jmp 0x15670f movss 0x29ab7e(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x297a48(%rip), %xmm0 # 0x3ee090 cmpltss 0x29ab6b(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0xd0(%rsp), %rdx movaps %xmm0, %xmm1 movss 0x298610(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x15670f movq 0x118(%r15,%rsi), %rax movss (%rax), %xmm1 xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm2 andnps %xmm1, %xmm2 movss 0x2985e1(%rip), %xmm1 # 0x3eec88 andps %xmm1, %xmm0 orps %xmm2, %xmm0 mulss %xmm4, %xmm0 jmp 0x15670f movq 0x118(%r15,%rsi), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x2979c2(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x15670f movss 0x2985a6(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x1565d0 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 jmp 0x15670f movaps %xmm4, %xmm0 movq 0xd0(%rsp), %rdx movq 0x58(%rsp), %rax movss %xmm0, (%rax,%rdx,4) incq %rdx jmp 0x156412 movq 0x80(%rsp), %rax movq 0x58(%rsp), %rcx leaq (%rcx,%rax,4), %rcx movq %rcx, 0x58(%rsp) movl 0x28(%rsp), %r12d incl %r12d movq 0x18(%rsp), %rdx jmp 0x1563d4 incq %rdx jmp 0x156399 movslq 0x1ec(%rsp), %rcx imulq 0x1d0(%rsp), %rcx movq 0x1c0(%rsp), %rax movq %rax, 0x30(%rsp) movslq 0x10c(%rsp), %r11 movslq 0x110(%rsp), %rax movq 0xf0(%rsp), %rdx movq 0x120(%rsp), %rbp imulq %rdx, %rbp imulq %r11, %rax movq %rdx, %r9 movq %rax, 0x50(%rsp) imulq %rax, %r9 addq $0xf, %r9 andq $-0x10, %r9 movq %r9, 0x98(%rsp) movq %rdx, 0x48(%rsp) imulq %rdx, %r11 xorl %edx, %edx testl %r13d, %r13d cmovlel %edx, %r13d movl 0x114(%rsp), %eax movl %eax, 0x58(%rsp) testl %r14d, %r14d cmovlel %edx, %r14d movq 0xe0(%rsp), %r9 movq 0xa0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xa0(%rsp) movl 0x108(%rsp), %eax movl %eax, 0x88(%rsp) imulq 0x158(%rsp), %rcx movq %rcx, 0x20(%rsp) movq %r9, 0x90(%rsp) movq %r9, 0x38(%rsp) xorl %eax, %eax movq %rbp, 0xd0(%rsp) cmpq 0xa0(%rsp), %rax je 0x15697a movq %rax, 0x10(%rsp) movq 0x98(%rsp), %rax xorl %edx, %edx divq 0x48(%rsp) cmpl $0x4, 0x88(%rsp) cmoveq 0x50(%rsp), %rax movq 0x1a8(%r15,%r8), %rcx xorps %xmm0, %xmm0 testq %rcx, %rcx je 0x15688e movslq 0x1e0(%r15,%r8), %rdx imulq 0x1e8(%r15,%r8), %rdx testq %rdx, %rdx movq 0x38(%rsp), %r8 je 0x156893 movq 0x10(%rsp), %rdx movss (%rcx,%rdx,4), %xmm0 jmp 0x156893 movq 0x38(%rsp), %r8 movq 0x10(%rsp), %rcx movq %rcx, %rdx imulq 0x20(%rsp), %rdx addq 0x30(%rsp), %rdx imulq %rcx, %rbp addq 0x90(%rsp), %rbp imull 0x58(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x1568d2 movss %xmm0, (%r8,%rcx,4) incq %rcx jmp 0x1568c3 movq -0x18(%rsi), %r8 xorl %eax, %eax cmpl 0xd8(%r15,%r8), %eax jge 0x156960 xorl %r9d, %r9d cmpl 0xd4(%r15,%r8), %r9d jge 0x156959 movslq 0xe0(%r15,%r8), %rcx movslq %eax, %r10 imulq %rcx, %r10 imulq %r11, %r10 addq %rbp, %r10 movslq 0xdc(%r15,%r8), %rcx movslq %r9d, %r12 imulq %rcx, %r12 leaq (%r10,%r12,4), %rcx xorl %r10d, %r10d cmpl %r14d, %r10d je 0x156954 movl %r13d, %r12d subl $0x1, %r12d jb 0x15694b movss (%rcx), %xmm0 addss (%rdx), %xmm0 movss %xmm0, (%rcx) movq -0x18(%rsi), %r8 movslq 0xe4(%r15,%r8), %rbx leaq (%rcx,%rbx,4), %rcx addq $0x4, %rdx jmp 0x156923 leaq (%rcx,%rdi,4), %rcx incl %r10d jmp 0x15691b incl %r9d jmp 0x1568e5 incl %eax jmp 0x1568d8 movq 0x10(%rsp), %rax incq %rax movq 0xd0(%rsp), %rbp addq %rbp, 0x38(%rsp) jmp 0x156827 movq 0x8(%r15), %rdi testq %rdi, %rdi je 0x156999 movq (%rdi), %rax leaq 0xe0(%rsp), %rsi movq 0x130(%rsp), %rdx callq *0x48(%rax) movq 0x1c8(%rsp), %rax testq %rax, %rax je 0x1569d0 lock decl (%rax) jne 0x1569d0 movq 0x1c0(%rsp), %rsi movq 0x1e0(%rsp), %rdi testq %rdi, %rdi je 0x1569c8 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1569d0 movq %rsi, %rdi callq 0x5f3e0 movq 0x180(%rsp), %rax testq %rax, %rax je 0x156a07 lock decl (%rax) jne 0x156a07 movq 0x178(%rsp), %rsi movq 0x198(%rsp), %rdi testq %rdi, %rdi je 0x1569ff movq (%rdi), %rax callq *0x18(%rax) jmp 0x156a07 movq %rsi, %rdi callq 0x5f3e0 movq (%r15), %rax addq -0x18(%rax), %r15 leaq 0xe0(%rsp), %rsi movq %r15, %rdi movq 0x160(%rsp), %rbx movq %rbx, %rdx movq 0x130(%rsp), %rcx callq 0x153cf6 cmpq $0x0, (%rbx) pushq $-0x64 popq %rcx je 0x156a4a movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax xorl %ebx, %ebx testq %rax, %rax jne 0x156a4c movl %ecx, %ebx movq 0xe8(%rsp), %rax testq %rax, %rax je 0x156a83 lock decl (%rax) jne 0x156a83 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x156a7b movq (%rdi), %rax callq *0x18(%rax) jmp 0x156a83 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0x258, %rsp # imm = 0x258 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq cmpl $0x0, 0x104(%r15,%rax) jle 0x156ab1 cmpl $0x0, 0x108(%r15,%rax) jg 0x154c2e pushq $0x8 popq %rsi leaq 0xe0(%rsp), %rax movq 0x160(%rsp), %rcx cmpq %rcx, %rax je 0x154c2e movq 0x8(%rcx), %rax testq %rax, %rax je 0x156b43 lock incl (%rax) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x156b43 lock decl (%rax) jne 0x156b43 movl %r9d, 0x30(%rsp) movq %r11, 0x20(%rsp) movq %r10, 0x38(%rsp) movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x156b29 movq (%rdi), %rax callq *0x18(%rax) movq 0x38(%rsp), %r10 movq 0x20(%rsp), %r11 movl 0x30(%rsp), %r9d pushq $0x8 popq %rsi jmp 0x156b43 movq %rsi, %rdi callq 0x5f3e0 movq 0x38(%rsp), %r10 movq 0x20(%rsp), %r11 movl 0x30(%rsp), %r9d pushq $0x8 popq %rsi movq 0x160(%rsp), %rcx movups (%rcx), %xmm0 movaps %xmm0, 0xe0(%rsp) movq 0x10(%rcx), %rax movq %rax, 0xf0(%rsp) movl 0x18(%rcx), %eax movl %eax, 0xf8(%rsp) movq 0x20(%rcx), %rax movq %rax, 0x100(%rsp) movups 0x28(%rcx), %xmm0 movups %xmm0, 0x108(%rsp) movl 0x38(%rcx), %eax movl %eax, 0x118(%rsp) movq 0x40(%rcx), %rax movq %rax, 0x120(%rsp) jmp 0x154c2e jmp 0x156c69 jmp 0x156c69 jmp 0x156c69 movq %rax, %rbx movq 0x1c8(%rsp), %rax testq %rax, %rax je 0x156be8 lock decl (%rax) jne 0x156be8 movq 0x1c0(%rsp), %rsi movq 0x1e0(%rsp), %rdi testq %rdi, %rdi jne 0x156be2 movq %rsi, %rdi callq 0x5f3e0 jmp 0x156be8 movq (%rdi), %rax callq *0x18(%rax) movq 0x180(%rsp), %rax testq %rax, %rax je 0x156c2a lock decl (%rax) jne 0x156c2a movq 0x178(%rsp), %rsi movq 0x198(%rsp), %rdi testq %rdi, %rdi jne 0x156c19 movq %rsi, %rdi callq 0x5f3e0 jmp 0x156c2a movq (%rdi), %rax callq *0x18(%rax) jmp 0x156c2a jmp 0x156c69 jmp 0x156c69 jmp 0x156c27 movq %rax, %rbx movq 0xe8(%rsp), %rax testq %rax, %rax je 0x156c61 lock decl (%rax) jne 0x156c61 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi jne 0x156c5b movq %rsi, %rdi callq 0x5f3e0 jmp 0x156c61 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/x86/deconvolution_x86.cpp

virtual thunk to ncnn::Deconvolution_x86_avx::create_pipeline(ncnn::Option const&)

int Deconvolution_x86_avx::create_pipeline(const Option& opt) { activation = create_activation_layer(activation_type, activation_params, opt); const int maxk = kernel_w * kernel_h; int num_input = weight_data_size / maxk / num_output; int elempack = 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = num_input % 16 == 0 ? 16 : num_input % 8 == 0 ? 8 : num_input % 4 == 0 ? 4 : 1; out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = num_input % 8 == 0 ? 8 : num_input % 4 == 0 ? 4 : 1; out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else elempack = num_input % 4 == 0 ? 4 : 1; out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ if (opt.use_sgemm_convolution) { const int maxk = kernel_w * kernel_h; gemm = ncnn::create_layer(ncnn::LayerType::Gemm); ncnn::ParamDict pd; pd.set(2, 1); // transA pd.set(3, 0); // transB pd.set(4, 1); // constantA pd.set(5, 0); // constantB pd.set(6, 1); // constantC pd.set(7, maxk * num_output); // M = maxk*num_output pd.set(8, 0); // N = size pd.set(9, num_input); // K = inch pd.set(10, -1); // constant_broadcast_type_C = null pd.set(11, 0); // output_N1M pd.set(12, out_elempack); gemm->load_param(pd); // maxk-inch-outch to pa-maxk-outch/pa-inch Mat tmp; { Mat weight_data_r2 = weight_data.reshape(maxk, num_input, num_output); tmp.create(maxk * num_output, num_input); for (int p = 0; p < num_input; p += 1) { float* g00 = tmp.row(p); for (int q = 0; q + (out_elempack - 1) < num_output; q += out_elempack) { for (int k = 0; k < maxk; k++) { for (int i = 0; i < out_elempack; i++) { const float* k00 = weight_data_r2.channel(q + i).row(p); g00[0] = k00[k]; g00++; } } } } } ncnn::Mat weights[1]; weights[0] = tmp; gemm->load_model(ModelBinFromMatArray(weights)); gemm->create_pipeline(opt); } else { Mat weight_data_transposed(weight_data.w); { float* pt = weight_data_transposed; const float* p = weight_data; for (int i = 0; i < num_input * num_output; i++) { for (int k = 0; k < maxk; k++) { pt[maxk - 1 - k] = p[k]; } p += maxk; pt += maxk; } } // src = kw-kh-inch-outch // dst = pb-pa-kw-kh-inch/pa-outch/pb Mat weight_data_r2 = weight_data_transposed.reshape(maxk, num_input, num_output); weight_data_tm.create(maxk, num_input / elempack, num_output / out_elempack, (size_t)4u * elempack * out_elempack, elempack * out_elempack); for (int q = 0; q + (out_elempack - 1) < num_output; q += out_elempack) { float* g00 = weight_data_tm.channel(q / out_elempack); for (int p = 0; p + (elempack - 1) < num_input; p += elempack) { for (int k = 0; k < maxk; k++) { for (int i = 0; i < elempack; i++) { for (int j = 0; j < out_elempack; j++) { const float* k00 = weight_data_r2.channel(q + j).row(p + i); g00[0] = k00[k]; g00++; } } } } } } if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %rax movq (%rdi), %rax addq -0x30(%rax), %rdi callq 0x164cde xorl %eax, %eax popq %rcx retq nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/deconvolution_x86_avx.cpp

ncnn::Deconvolution_x86_avx::destroy_pipeline(ncnn::Option const&)

int Deconvolution_x86_avx::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } if (gemm) { gemm->destroy_pipeline(opt); delete gemm; gemm = 0; } return 0; }

pushq %r14 pushq %rbx pushq %rax movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x165840 movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x16583b movq (%rdi), %rax callq *0x8(%rax) andq $0x0, 0x8(%rbx) movq 0x10(%rbx), %rdi testq %rdi, %rdi je 0x165866 movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x10(%rbx), %rdi testq %rdi, %rdi je 0x165861 movq (%rdi), %rax callq *0x8(%rax) andq $0x0, 0x10(%rbx) xorl %eax, %eax addq $0x8, %rsp popq %rbx popq %r14 retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/deconvolution_x86_avx.cpp

ncnn::InnerProduct_x86::create_pipeline_int8_x86(ncnn::Option const&)

int InnerProduct_x86::create_pipeline_int8_x86(const Option& opt) { const int num_input = weight_data_size / num_output; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { out_elempack = num_output % 8 == 0 ? 8 : 1; } #endif // __SSE2__ // src = inch-outch // dst = pb-inch-outch/pb { Mat weight_data_r2 = weight_data.reshape(num_input, num_output); weight_data_tm.create(num_input, num_output / out_elempack, (size_t)out_elempack, out_elempack); for (int q = 0; q + (out_elempack - 1) < num_output; q += out_elempack) { signed char* g0 = weight_data_tm.row<signed char>(q / out_elempack); for (int p = 0; p < num_input; p++) { for (int j = 0; j < out_elempack; j++) { *g0++ = weight_data_r2.row<signed char>(q + j)[p]; } } } } scale_in_data.create(num_output); for (int p = 0; p < num_output; p++) { // dequantize float scale_in; if (weight_data_int8_scales[p] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[0] * weight_data_int8_scales[p]); scale_in_data[p] = scale_in; } if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x48, %rsp movq %rsi, %rbx movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rax leaq 0x130(%rdi,%rax), %rsi movl -0x60(%rsi), %ecx movl -0x58(%rsi), %eax cltd idivl %ecx movl %eax, %r15d testb $0x7, %cl sete %bpl andb 0x27(%rbx), %bpl pushq $0x8 popq %rax pushq $0x1 popq %r12 cmovnel %eax, %r12d xorl %r13d, %r13d movq %rsp, %rdi movl %r15d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x10(%r14), %rdi movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd0(%r14,%rax), %eax cltd idivl %r12d movl %r15d, %esi movl %eax, %edx movq %r12, %rcx movl %r12d, %r8d xorl %r9d, %r9d callq 0x627de leal -0x1(%r12), %eax movq (%r14), %rdx movzbl %bpl, %ecx leal (%rcx,%rcx,2), %ecx testl %r15d, %r15d cmovlel %r13d, %r15d leaq (%rax,%r13), %rsi movq -0x18(%rdx), %rdi movslq 0xd0(%r14,%rdi), %rdi cmpq %rdi, %rsi jge 0x175b27 movl %r13d, %esi shrl %cl, %esi movslq 0x3c(%r14), %rdi imulq %rdi, %rsi imulq 0x20(%r14), %rsi addq 0x10(%r14), %rsi xorl %edi, %edi cmpq %r15, %rdi je 0x175b22 xorl %r8d, %r8d cmpq %r8, %r12 je 0x175b1a leaq (%r8,%r13), %r9 movslq 0x2c(%rsp), %r10 imulq %r9, %r10 imulq 0x10(%rsp), %r10 addq (%rsp), %r10 movb (%rdi,%r10), %r9b movb %r9b, (%rsi,%r8) incq %r8 jmp 0x175af1 incq %rdi addq %r8, %rsi jmp 0x175ae9 addq %r12, %r13 jmp 0x175abc movq 0x8(%rsp), %rax testq %rax, %rax je 0x175b54 lock decl (%rax) jne 0x175b54 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x175b4c movq (%rdi), %rax callq *0x18(%rax) jmp 0x175b54 movq %rsi, %rdi callq 0x5f3e0 leaq 0x58(%r14), %rdi movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd0(%r14,%rax), %esi xorl %r12d, %r12d pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq 0x58(%r14), %rcx movss 0x279105(%rip), %xmm0 # 0x3eec88 movq -0x18(%rax), %rdx leaq (%r14,%rdx), %r15 movslq 0xd0(%r14,%rdx), %rdx cmpq %rdx, %r12 jge 0x175bca movq 0x1c0(%r15), %rdx movss (%rdx,%r12,4), %xmm1 xorps %xmm2, %xmm2 ucomiss %xmm1, %xmm2 je 0x175bbf movq 0x208(%r15), %rdx mulss (%rdx), %xmm1 movaps %xmm0, %xmm2 divss %xmm1, %xmm2 movss %xmm2, (%rcx,%r12,4) incq %r12 jmp 0x175b83 cmpb $0x0, (%rbx) je 0x175c2c leaq 0x130(%r15), %rbx movq 0x138(%r15), %rax testq %rax, %rax je 0x175c0a lock decl (%rax) jne 0x175c0a movq 0x130(%r15), %rsi movq 0x150(%r15), %rdi testq %rdi, %rdi je 0x175c02 movq (%rdi), %rax callq *0x18(%rax) jmp 0x175c0a movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x170(%r15) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rbx) movups %xmm0, (%rbx) movups %xmm0, 0x158(%r15) andl $0x0, 0x168(%r15) xorl %eax, %eax addq $0x48, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x175c77 movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x175c6f lock decl (%rax) jne 0x175c6f movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x175c69 movq %rsi, %rdi callq 0x5f3e0 jmp 0x175c6f movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/x86/innerproduct_x86.cpp

ncnn::Input::Input()

Input::Input() { one_blob_only = true; support_inplace = true; support_vulkan = true; support_packing = true; support_bf16_storage = true; support_image_storage = true; }

pushq %rbx movq %rdi, %rbx callq 0x77e84 leaq 0x2a7e68(%rip), %rax # 0x4836d0 movq %rax, (%rbx) movb $0x1, %al movb %al, 0xf(%rbx) movb %al, 0xc(%rbx) movl $0x1010101, 0x8(%rbx) # imm = 0x1010101 popq %rbx retq

/csukuangfj[P]ncnn/src/layer/input.cpp

ncnn::Log::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int Log::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; int size = w * h; if (base == -1.f) { #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); for (int i = 0; i < size; i++) { ptr[i] = logf(shift + ptr[i] * scale); } } } else { float log_base_inv = 1.f / logf(base); #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); for (int i = 0; i < size; i++) { ptr[i] = logf(shift + ptr[i] * scale) * log_base_inv; } } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x18, %rsp movq %rsi, %r13 movq %rdi, %rbx movl 0x30(%rsi), %r15d movl 0x38(%rsi), %r12d imull 0x2c(%rsi), %r15d movss 0xd0(%rdi), %xmm0 movss 0x21584b(%rip), %xmm1 # 0x3f11f0 ucomiss %xmm0, %xmm1 jne 0x1dba13 movq (%r13), %r14 movq 0x40(%r13), %rax xorl %ebp, %ebp testl %r15d, %r15d cmovlel %ebp, %r15d testl %r12d, %r12d cmovlel %ebp, %r12d imulq 0x10(%r13), %rax movq %rax, 0x8(%rsp) cmpq %r12, %rbp je 0x1dba94 movq %r12, %r13 xorl %r12d, %r12d cmpq %r12, %r15 je 0x1dba06 movss 0xd4(%rbx), %xmm0 mulss (%r14,%r12,4), %xmm0 addss 0xd8(%rbx), %xmm0 callq 0x5f200 movss %xmm0, (%r14,%r12,4) incq %r12 jmp 0x1db9db incq %rbp addq 0x8(%rsp), %r14 movq %r13, %r12 jmp 0x1db9cc callq 0x5f200 movq %r13, %rcx movq (%r13), %r13 xorl %ebp, %ebp testl %r15d, %r15d cmovlel %ebp, %r15d movq 0x40(%rcx), %rax testl %r12d, %r12d cmovlel %ebp, %r12d movss 0x21324d(%rip), %xmm1 # 0x3eec88 divss %xmm0, %xmm1 movss %xmm1, 0x8(%rsp) imulq 0x10(%rcx), %rax movq %rax, 0x10(%rsp) cmpq %r12, %rbp je 0x1dba94 xorl %r14d, %r14d cmpq %r14, %r15 je 0x1dba8a movss 0xd4(%rbx), %xmm0 mulss (%r13,%r14,4), %xmm0 addss 0xd8(%rbx), %xmm0 callq 0x5f200 mulss 0x8(%rsp), %xmm0 movss %xmm0, (%r13,%r14,4) incq %r14 jmp 0x1dba57 incq %rbp addq 0x10(%rsp), %r13 jmp 0x1dba4f xorl %eax, %eax addq $0x18, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq nop

/csukuangfj[P]ncnn/src/layer/log.cpp

ncnn::LRN::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int LRN::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); for (int i = 0; i < size; i++) { outptr[i] = ptr[i] * ptr[i]; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum float* ssptr = square_sum.channel(q); for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 || p >= channels) continue; const float* sptr = square_blob.channel(p); for (int i = 0; i < size; i++) { ssptr[i] += sptr[i]; } } float* ptr = bottom_top_blob.channel(q); for (int i = 0; i < size; i++) { ptr[i] = ptr[i] * powf(bias + alpha_div_size * ssptr[i], -beta); } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; opt_b.use_packing_layout = false; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size * local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x158, %rsp # imm = 0x158 movq %rsi, %rbp movl 0x2c(%rsi), %esi movl 0x30(%rbp), %ecx movl 0x38(%rbp), %r13d movq 0x10(%rbp), %r15 leaq 0x90(%rsp), %rax andq $0x0, 0x40(%rax) movq %rdx, %r14 movq %rdi, %rbx xorps %xmm0, %xmm0 movaps %xmm0, (%rax) movups %xmm0, 0xc(%rax) movaps %xmm0, 0x20(%rax) movups %xmm0, 0x2c(%rax) movq 0x10(%rdx), %r9 movq %rax, %rdi movq %rsi, %r12 movq %rcx, (%rsp) movl %ecx, %edx movl %r13d, %ecx movq %r15, %r8 callq 0x63810 movq 0x90(%rsp), %rcx pushq $-0x64 popq %r10 testq %rcx, %rcx je 0x1dbc70 movq 0xd0(%rsp), %rdx movslq 0xc8(%rsp), %rax movq %rdx, %rsi imulq %rax, %rsi testq %rsi, %rsi je 0x1dbc70 movq %r15, %r11 movq %r14, 0x18(%rsp) movq %r12, %rsi movslq %esi, %rdi movq %rdi, 0x78(%rsp) movq (%rsp), %r14 imull %esi, %r14d movq (%rbp), %rsi movq 0x40(%rbp), %rdi imulq 0xa0(%rsp), %rdx xorl %r8d, %r8d testl %r14d, %r14d cmovlel %r8d, %r14d testl %r13d, %r13d movl $0x0, %r15d cmovgl %r13d, %r15d imulq 0x10(%rbp), %rdi cmpq %r15, %r8 je 0x1dbcb1 xorl %r9d, %r9d cmpq %r9, %r14 je 0x1dbc65 movss (%rsi,%r9,4), %xmm0 mulss %xmm0, %xmm0 movss %xmm0, (%rcx,%r9,4) incq %r9 jmp 0x1dbc4b incq %r8 addq %rdx, %rcx addq %rdi, %rsi jmp 0x1dbc43 movl %r10d, %r14d movq 0x98(%rsp), %rax testq %rax, %rax je 0x1dbe5c lock decl (%rax) jne 0x1dbe5c movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi je 0x1dbe54 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dbe5c movl 0xd0(%rbx), %ecx testl %ecx, %ecx je 0x1dbddf xorl %r14d, %r14d cmpl $0x1, %ecx jne 0x1dbc73 movq 0x90(%rsp), %rdx movq 0x98(%rsp), %rcx movq %rdx, 0x20(%rsp) movq %rcx, 0x28(%rsp) movq 0xa0(%rsp), %rdx movq %rdx, 0x30(%rsp) movl 0xa8(%rsp), %edx movl %edx, 0x38(%rsp) movq 0xb0(%rsp), %rdx movq %rdx, 0x40(%rsp) movups 0xb8(%rsp), %xmm0 movups %xmm0, 0x48(%rsp) movl %eax, 0x58(%rsp) movq 0xd0(%rsp), %rax movq %rax, 0x60(%rsp) testq %rcx, %rcx je 0x1dbd2c lock incl (%rcx) movl 0xd4(%rbx), %r13d cmpl $0x2, %r13d jl 0x1dbea9 movl %r10d, %r14d movl %r13d, %edx shrl %edx movq 0x18(%rsp), %r11 movups (%r11), %xmm0 movups 0x10(%r11), %xmm1 movups 0x20(%r11), %xmm2 movups 0x30(%r11), %xmm3 leaq 0x110(%rsp), %rax movaps %xmm3, 0x30(%rax) movaps %xmm2, 0x20(%rax) movaps %xmm1, 0x10(%rax) movaps %xmm0, (%rax) movq 0x10(%r11), %rcx movq %rcx, 0x8(%rax) movb $0x0, 0x27(%rax) movl %edx, %ecx notl %ecx addl %r13d, %ecx leaq 0x90(%rsp), %rdi leaq 0x20(%rsp), %rsi xorps %xmm0, %xmm0 movl %edx, %r8d movl %ecx, %r9d pushq %rax pushq $0x0 callq 0x6466c popq %rax popq %rcx cmpq $0x0, 0x20(%rsp) movl %r14d, %ecx je 0x1dc271 movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x1dc271 movq %rbp, 0x10(%rsp) movl 0x4c(%rsp), %ebp movl 0xd4(%rbx), %r13d jmp 0x1dbeb1 leaq 0x20(%rsp), %rdi andq $0x0, 0x40(%rdi) xorps %xmm0, %xmm0 movaps %xmm0, (%rdi) movups %xmm0, 0xc(%rdi) movaps %xmm0, 0x20(%rdi) movups %xmm0, 0x2c(%rdi) movq 0x18(%rsp), %rax movq 0x10(%rax), %r9 movl %r12d, %esi movq (%rsp), %rdx movl %r13d, %ecx movq %r11, %r8 callq 0x63810 movq 0x20(%rsp), %rax testq %rax, %rax pushq $-0x64 popq %rdi je 0x1dbe71 movq 0x60(%rsp), %rcx movslq 0x58(%rsp), %rdx movq %rcx, %rsi imulq %rdx, %rsi testq %rsi, %rsi je 0x1dbe71 imull %edx, %ecx xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpl %edx, %ecx je 0x1dc0f2 andl $0x0, (%rax,%rdx,4) incq %rdx jmp 0x1dbe43 movq %rsi, %rdi callq 0x5f3e0 movl %r14d, %eax addq $0x158, %rsp # imm = 0x158 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc2a7 lock decl (%rax) jne 0x1dc2a7 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dc29c movq (%rdi), %rax callq *0x18(%rax) pushq $-0x64 popq %rdi jmp 0x1dc2a7 movq %rbp, 0x10(%rsp) movl %r12d, %ebp imull %r13d, %r13d movss 0xd8(%rbx), %xmm0 movss %xmm0, 0x8(%rsp) leaq 0x110(%rsp), %rdi leaq 0xf(%rsp), %rdx movq %r13, %rsi callq 0x73bbe cvtsi2ss %r13d, %xmm0 movq 0x110(%rsp), %r14 movl 0xd4(%rbx), %ecx subl %ecx, %ebp xorl %eax, %eax xorl %edx, %edx xorl %esi, %esi cmpl %ecx, %esi jge 0x1dbf24 cltq leaq (%r14,%rax,4), %r8 movl %edx, %r9d xorl %edi, %edi cmpl %ecx, %edi jge 0x1dbf19 leaq (%r9,%rdi), %rcx movl %ecx, (%r8,%rdi,4) movl 0xd4(%rbx), %ecx incq %rdi jmp 0x1dbf02 addl %ebp, %edx addl %edi, %edx incl %esi addq %rdi, %rax jmp 0x1dbef3 movq %rbx, 0x18(%rsp) movq 0x10(%rsp), %rcx movq 0x40(%rcx), %rax imulq 0x10(%rcx), %rax movq %rax, 0xf8(%rsp) movslq 0x4c(%rsp), %rdx movq 0x30(%rsp), %rax movq 0x60(%rsp), %rsi imulq %rax, %rsi movq %rsi, 0xf0(%rsp) imulq %rax, %rdx movq %rdx, 0x108(%rsp) xorl %eax, %eax movq %r12, %rbx testl %ebx, %ebx cmovlel %eax, %ebx movq (%rcx), %rcx movq %rcx, 0xe8(%rsp) movq (%rsp), %r12 testl %r12d, %r12d cmovlel %eax, %r12d movq %r12, (%rsp) xorl %eax, %eax movq 0x20(%rsp), %rcx movq %rcx, 0xe0(%rsp) movss 0x212ce7(%rip), %xmm1 # 0x3eec88 divss %xmm0, %xmm1 movss %xmm1, 0x10(%rsp) movq %r15, 0x70(%rsp) cmpq %r15, %rax je 0x1dc0b0 movq 0xf8(%rsp), %r12 imulq %rax, %r12 addq 0xe8(%rsp), %r12 movq 0xf0(%rsp), %rcx movq %rax, 0x100(%rsp) imulq %rax, %rcx addq 0xe0(%rsp), %rcx movq %rcx, 0x80(%rsp) xorl %eax, %eax cmpq (%rsp), %rax je 0x1dc09b movq 0x108(%rsp), %rbp movq %rax, 0x88(%rsp) imulq %rax, %rbp addq 0x80(%rsp), %rbp xorl %r15d, %r15d cmpq %rbx, %r15 je 0x1dc082 leaq (,%r15,4), %rax addq %rbp, %rax xorps %xmm0, %xmm0 xorl %ecx, %ecx cmpq %rcx, %r13 je 0x1dc044 movslq (%r14,%rcx,4), %rdx addss (%rax,%rdx,4), %xmm0 incq %rcx jmp 0x1dc031 mulss 0x8(%rsp), %xmm0 mulss 0x10(%rsp), %xmm0 movq 0x18(%rsp), %rax addss 0xe0(%rax), %xmm0 movss 0xdc(%rax), %xmm1 xorps 0x212024(%rip), %xmm1 # 0x3ee090 callq 0x5f0e0 mulss (%r12,%r15,4), %xmm0 movss %xmm0, (%r12,%r15,4) incq %r15 jmp 0x1dc01c movq 0x78(%rsp), %rax leaq (%r12,%rax,4), %r12 movq 0x88(%rsp), %rax incq %rax jmp 0x1dbff3 movq 0x100(%rsp), %rax incq %rax movq 0x70(%rsp), %r15 jmp 0x1dbfb0 leaq 0x110(%rsp), %rdi callq 0x624be movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc2b7 lock decl (%rax) jne 0x1dc2b7 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dc29a movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dc2b7 movq %r15, 0x70(%rsp) movl 0xd4(%rbx), %ecx cvtsi2ss %ecx, %xmm0 movss 0xd8(%rbx), %xmm1 movss %xmm1, 0x18(%rsp) movq 0x20(%rsp), %r12 movq 0x60(%rsp), %rdi pushq $-0x2 popq %rsi movl %ecx, %eax cltd idivl %esi movq %rax, (%rsp) pushq $0x2 popq %rsi movl %ecx, %eax cltd idivl %esi movq (%rbp), %r15 movq 0x40(%rbp), %rcx cltq movq %rax, 0x78(%rsp) movss 0x212b44(%rip), %xmm1 # 0x3eec88 divss %xmm0, %xmm1 movss %xmm1, 0x8(%rsp) imulq 0x30(%rsp), %rdi movq %rdi, 0x88(%rsp) imulq 0x10(%rbp), %rcx movq %rcx, 0x80(%rsp) xorl %edx, %edx cmpq 0x70(%rsp), %rdx je 0x1dc24b movq (%rsp), %rax addl %edx, %eax movq 0x78(%rsp), %rcx movq %rdx, 0x10(%rsp) addq %rdx, %rcx movq 0x90(%rsp), %rdx movq 0xd0(%rsp), %rsi imulq 0xa0(%rsp), %rsi movslq %eax, %rdi cmpq %rdi, %rcx jl 0x1dc1e8 testl %eax, %eax sets %dil cmpl %r13d, %eax setge %r8b orb %dil, %r8b jne 0x1dc1e4 movl %eax, %edi imulq %rsi, %rdi addq %rdx, %rdi xorl %r8d, %r8d cmpq %r8, %r14 je 0x1dc1e4 movss (%r12,%r8,4), %xmm0 addss (%rdi,%r8,4), %xmm0 movss %xmm0, (%r12,%r8,4) incq %r8 jmp 0x1dc1c8 incl %eax jmp 0x1dc1a2 xorl %ebp, %ebp cmpq %rbp, %r14 je 0x1dc22e movss (%r12,%rbp,4), %xmm0 mulss 0x18(%rsp), %xmm0 mulss 0x8(%rsp), %xmm0 addss 0xe0(%rbx), %xmm0 movss 0xdc(%rbx), %xmm1 xorps 0x211e78(%rip), %xmm1 # 0x3ee090 callq 0x5f0e0 mulss (%r15,%rbp,4), %xmm0 movss %xmm0, (%r15,%rbp,4) incq %rbp jmp 0x1dc1ea movq 0x10(%rsp), %rdx incq %rdx addq 0x88(%rsp), %r12 addq 0x80(%rsp), %r15 jmp 0x1dc16b movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc2b7 lock decl (%rax) jne 0x1dc2b7 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dc2af movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dc2b7 movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc2ca lock decl (%rax) jne 0x1dc2ca movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dc2bf movq (%rdi), %rax callq *0x18(%rax) movl %r14d, %ecx jmp 0x1dc2ca jmp 0x1dc2af movq %rsi, %rdi callq 0x5f3e0 pushq $-0x64 popq %rdi movl %edi, %r14d jmp 0x1dbc73 movq %rsi, %rdi callq 0x5f3e0 xorl %r14d, %r14d jmp 0x1dbc73 movq %rsi, %rdi callq 0x5f3e0 movl %r14d, %ecx movl %ecx, %r14d jmp 0x1dbc73 jmp 0x1dc38e jmp 0x1dc38e jmp 0x1dc38e jmp 0x1dc38e jmp 0x1dc2e8 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc34f lock decl (%rax) jne 0x1dc34f movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1dc30b jmp 0x1dc336 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dc34f jmp 0x1dc38e movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dc34f lock decl (%rax) jne 0x1dc34f movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1dc340 movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dc34f movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dc34f jmp 0x1dc38e jmp 0x1dc38e movq %rax, %rbx movq 0x98(%rsp), %rax testq %rax, %rax je 0x1dc386 lock decl (%rax) jne 0x1dc386 movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi jne 0x1dc380 movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dc386 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/lrn.cpp

virtual thunk to ncnn::LRN_x86::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int LRN_x86::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { *outptr = *ptr * *ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 || p >= channels) continue; const float* sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ssptr += *sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ptr = *ptr * powf(bias + alpha_div_size * *ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size * local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }

movq (%rdi), %rax addq -0x58(%rax), %rdi jmp 0x1dc3b0

/csukuangfj[P]ncnn/src/layer/x86/lrn_x86.cpp

ncnn::LRN_x86_avx512::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int LRN_x86_avx512::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { *outptr = *ptr * *ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 || p >= channels) continue; const float* sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ssptr += *sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ptr = *ptr * powf(bias + alpha_div_size * *ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size * local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x178, %rsp # imm = 0x178 movq %rsi, %r15 movl 0x2c(%rsi), %esi movl 0x30(%r15), %ecx movl 0x38(%r15), %r12d movq 0x10(%r15), %r13 leaq 0x90(%rsp), %rax andq $0x0, 0x40(%rax) movq %rdx, %r14 movq %rdi, %rbx vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rax) vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, 0x20(%rax) vmovups %xmm0, 0x2c(%rax) movq 0x10(%rdx), %r9 movq %rax, %rdi movq %rsi, (%rsp) movq %rcx, 0x10(%rsp) movl %ecx, %edx movl %r12d, %ecx movq %r13, %r8 callq 0x63810 pushq $-0x64 popq %rbp cmpq $0x0, 0x90(%rsp) je 0x1dd6c2 movslq 0xc8(%rsp), %rax imulq 0xd0(%rsp), %rax testq %rax, %rax je 0x1dd6c2 movq %r13, 0x110(%rsp) movq %r14, 0x80(%rsp) movq (%rsp), %rcx movslq %ecx, %rax movq %rax, 0x108(%rsp) movq 0x10(%rsp), %rax movl %eax, %r13d imull %ecx, %r13d xorl %eax, %eax testl %r12d, %r12d movl $0x0, %ecx cmovgl %r12d, %ecx movq %rcx, 0x68(%rsp) movq %rbx, 0x18(%rsp) cmpq 0x68(%rsp), %rax je 0x1dcdff movq (%r15), %rcx movq 0x10(%r15), %r8 movq 0x40(%r15), %rdi movq %rdi, %r9 imulq %rax, %r9 imulq %r8, %r9 addq %rcx, %r9 movq 0x90(%rsp), %rdx movq 0xa0(%rsp), %r11 movq 0xd0(%rsp), %r10 movq %r10, %rbx imulq %rax, %rbx imulq %r11, %rbx addq %rdx, %rbx xorl %esi, %esi xorl %r14d, %r14d leal 0x7(%r14), %ebp cmpl %r13d, %ebp jge 0x1dcdc4 vmovups (%r9), %ymm0 vmulps %ymm0, %ymm0, %ymm0 vmovups %ymm0, (%rbx) addq $0x20, %r9 addq $0x20, %rbx addl $0x8, %r14d addq $0x8, %rsi jmp 0x1dcd9c imulq %r11, %r10 imulq %rax, %r10 addq %r10, %rdx imulq %r8, %rdi imulq %rax, %rdi addq %rdi, %rcx movq 0x18(%rsp), %rbx cmpl %r13d, %esi jge 0x1dcdf7 vmovss (%rcx,%rsi,4), %xmm0 vmulss %xmm0, %xmm0, %xmm0 vmovss %xmm0, (%rdx,%rsi,4) incq %rsi jmp 0x1dcddf incq %rax jmp 0x1dcd4d movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd0(%rbx,%rcx), %ecx testl %ecx, %ecx je 0x1dcf42 xorl %ebp, %ebp cmpl $0x1, %ecx jne 0x1dd6c2 movq 0x98(%rsp), %rcx vmovaps 0x90(%rsp), %xmm0 vmovaps %xmm0, 0x20(%rsp) movq 0xa0(%rsp), %rdx movq %rdx, 0x30(%rsp) movl 0xa8(%rsp), %edx movl %edx, 0x38(%rsp) movq 0xb0(%rsp), %rdx movq %rdx, 0x40(%rsp) vmovups 0xb8(%rsp), %xmm0 vmovups %xmm0, 0x48(%rsp) movl 0xc8(%rsp), %edx movl %edx, 0x58(%rsp) movq 0xd0(%rsp), %rdx movq %rdx, 0x60(%rsp) testq %rcx, %rcx je 0x1dce8e lock incl (%rcx) movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %r13d cmpl $0x2, %r13d jl 0x1dcffe shrl %r13d movq 0x80(%rsp), %rcx vmovups (%rcx), %zmm0 leaq 0x130(%rsp), %r10 vmovups %zmm0, (%r10) movq 0x10(%rcx), %rcx movq %rcx, 0x8(%r10) movq -0x18(%rax), %rax movl %r13d, %ecx notl %ecx addl 0xd4(%rbx,%rax), %ecx leaq 0x90(%rsp), %rdi leaq 0x20(%rsp), %rsi vxorps %xmm0, %xmm0, %xmm0 movl %r13d, %edx movl %r13d, %r8d movl %ecx, %r9d pushq %r10 pushq $0x0 vzeroupper callq 0x6466c popq %rax popq %rcx cmpq $0x0, 0x20(%rsp) pushq $-0x64 popq %rbp je 0x1dd68b movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x1dd68b movq %r15, %r14 movl 0x4c(%rsp), %r15d movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %r13d jmp 0x1dd008 leaq 0x20(%rsp), %rdi andq $0x0, 0x40(%rdi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rdi) vmovups %xmm0, 0xc(%rdi) vmovaps %xmm0, 0x20(%rdi) vmovups %xmm0, 0x2c(%rdi) movq 0x80(%rsp), %rax movq 0x10(%rax), %r9 movq (%rsp), %rsi movq 0x10(%rsp), %rdx movl %r12d, %ecx movq 0x110(%rsp), %r8 vzeroupper callq 0x63810 movq 0x20(%rsp), %rax testq %rax, %rax je 0x1dcfc6 movq 0x60(%rsp), %rcx movslq 0x58(%rsp), %rdx movq %rcx, %rsi imulq %rdx, %rsi testq %rsi, %rsi je 0x1dcfc6 imull %edx, %ecx xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpl %edx, %ecx je 0x1dd26a andl $0x0, (%rax,%rdx,4) incq %rdx jmp 0x1dcfb5 movq 0x28(%rsp), %rax testq %rax, %rax pushq $-0x64 popq %rbp je 0x1dd6c2 lock decl (%rax) jne 0x1dd6c2 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dd6b3 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dd6c2 movq %r15, %r14 movq (%rsp), %rax movl %eax, %r15d imull %r13d, %r13d vmovss 0xd8(%rbx,%rcx), %xmm0 vmovss %xmm0, 0x80(%rsp) leaq 0x130(%rsp), %rdi leaq 0xf(%rsp), %rdx movq %r13, %rsi vzeroupper callq 0x73bbe vcvtsi2ss %r13d, %xmm1, %xmm0 movq 0x130(%rsp), %rbp movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %edx subl %edx, %r15d xorl %ecx, %ecx xorl %esi, %esi xorl %edi, %edi cmpl %edx, %edi jge 0x1dd09c movslq %ecx, %rcx leaq (,%rcx,4), %r9 addq %rbp, %r9 movl %esi, %r10d xorl %r8d, %r8d cmpl %edx, %r8d jge 0x1dd08f leaq (%r10,%r8), %rdx movl %edx, (%r9,%r8,4) movq -0x18(%rax), %rdx movl 0xd4(%rbx,%rdx), %edx incq %r8 jmp 0x1dd072 addl %r15d, %esi addl %r8d, %esi incl %edi addq %r8, %rcx jmp 0x1dd05a movq (%r14), %rax movq %rax, 0xf0(%rsp) movq 0x40(%r14), %rax imulq 0x10(%r14), %rax movq %rax, 0xe8(%rsp) movslq 0x4c(%rsp), %rcx movq 0x30(%rsp), %rax movq 0x60(%rsp), %rdx imulq %rax, %rdx movq %rdx, 0xe0(%rsp) imulq %rax, %rcx movq %rcx, 0x100(%rsp) xorl %ecx, %ecx movq (%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, (%rsp) movq 0x20(%rsp), %rax movq %rax, 0xd8(%rsp) movq 0x10(%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, 0x10(%rsp) vmovss 0x211b76(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0x110(%rsp) cmpq 0x68(%rsp), %rcx je 0x1dd228 movq 0xe8(%rsp), %r14 imulq %rcx, %r14 addq 0xf0(%rsp), %r14 movq 0xe0(%rsp), %rax movq %rcx, 0xf8(%rsp) imulq %rcx, %rax addq 0xd8(%rsp), %rax movq %rax, 0x70(%rsp) xorl %eax, %eax cmpq 0x10(%rsp), %rax je 0x1dd218 movq 0x100(%rsp), %r15 movq %rax, 0x78(%rsp) imulq %rax, %r15 addq 0x70(%rsp), %r15 movq (%rbx), %r12 xorl %ebx, %ebx cmpq (%rsp), %rbx je 0x1dd1fa leaq (%r15,%rbx,4), %rax vxorps %xmm0, %xmm0, %xmm0 xorl %ecx, %ecx cmpq %rcx, %r13 je 0x1dd1ab movslq (%rbp,%rcx,4), %rdx vaddss (%rax,%rdx,4), %xmm0, %xmm0 incq %rcx jmp 0x1dd197 movq -0x18(%r12), %rax vmulss 0x80(%rsp), %xmm0, %xmm0 movq 0x18(%rsp), %rcx vmovss 0x110(%rsp), %xmm1 vfmadd213ss 0xe0(%rcx,%rax), %xmm1, %xmm0 # xmm0 = (xmm1 * xmm0) + mem vmovss 0xdc(%rcx,%rax), %xmm1 vxorps 0x213fd0(%rip){1to4}, %xmm1, %xmm1 # 0x3f11b4 callq 0x5f0e0 vmulss (%r14,%rbx,4), %xmm0, %xmm0 vmovss %xmm0, (%r14,%rbx,4) incq %rbx jmp 0x1dd187 movq 0x108(%rsp), %rax leaq (%r14,%rax,4), %r14 movq 0x78(%rsp), %rax incq %rax movq 0x18(%rsp), %rbx jmp 0x1dd161 movq 0xf8(%rsp), %rcx incq %rcx jmp 0x1dd11f leaq 0x130(%rsp), %rdi callq 0x624be movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd6c0 lock decl (%rax) jne 0x1dd6c0 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dd6b1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dd6c0 movq %r15, 0x78(%rsp) movq (%rbx), %rax movq %rax, (%rsp) movq -0x18(%rax), %rax vmovss 0xd8(%rbx,%rax), %xmm0 vcvtsi2ssl 0xd4(%rbx,%rax), %xmm1, %xmm1 vdivss %xmm1, %xmm0, %xmm0 vmovaps %xmm0, 0x80(%rsp) vbroadcastss %xmm0, %ymm0 vmovups %ymm0, 0x110(%rsp) xorl %r15d, %r15d pushq $0x2 popq %r14 movl %r12d, 0x70(%rsp) cmpq 0x68(%rsp), %r15 je 0x1dd662 movq (%rsp), %rax movq -0x18(%rax), %r9 movl 0xd4(%rbx,%r9), %esi movl %esi, %eax cltd pushq $-0x2 popq %rcx idivl %ecx movl %eax, %ecx addl %r15d, %ecx movl %esi, %eax cltd idivl %r14d addl %r15d, %eax cmpl %eax, %ecx jg 0x1dd3ae movl %ecx, %ecx testl %ecx, %ecx sets %al cmpl %r12d, %ecx setge %dl orb %al, %dl jne 0x1dd3a7 movq 0x90(%rsp), %rax movq 0xa0(%rsp), %r8 movq 0xd0(%rsp), %rdi movq %rdi, %r9 imulq %rcx, %r9 imulq %r8, %r9 addq %rax, %r9 movq 0x60(%rsp), %r10 imulq %r15, %r10 imulq 0x30(%rsp), %r10 movq 0x20(%rsp), %rdx leaq (%rdx,%r10), %r11 xorl %esi, %esi xorl %ebx, %ebx leal 0x7(%rbx), %ebp cmpl %r13d, %ebp jge 0x1dd36b vmovups (%r11), %ymm0 vaddps (%r9), %ymm0, %ymm0 vmovups %ymm0, (%r11) addq $0x20, %r9 addq $0x20, %r11 addl $0x8, %ebx addq $0x8, %rsi jmp 0x1dd343 addq %r10, %rdx imulq %r8, %rdi imulq %rcx, %rdi addq %rdi, %rax movq 0x18(%rsp), %rbx cmpl %r13d, %esi jge 0x1dd397 vmovss (%rdx,%rsi,4), %xmm0 vaddss (%rax,%rsi,4), %xmm0, %xmm0 vmovss %xmm0, (%rdx,%rsi,4) incq %rsi jmp 0x1dd37e movq (%rsp), %rax movq -0x18(%rax), %r9 movl 0xd4(%rbx,%r9), %esi incl %ecx jmp 0x1dd2db movq 0x78(%rsp), %rax movq (%rax), %r12 movq 0x10(%rax), %rcx movq 0x40(%rax), %rax movq %rax, %rdi imulq %r15, %rdi imulq %rcx, %rdi addq %r12, %rdi movq 0x20(%rsp), %rbp movq 0x30(%rsp), %rsi movq 0x60(%rsp), %rdx movq %rdx, %r8 imulq %r15, %r8 imulq %rsi, %r8 addq %rbp, %r8 vbroadcastss 0xe0(%rbx,%r9), %ymm0 vmovss 0xdc(%rbx,%r9), %xmm1 vxorps 0x213dad(%rip){1to4}, %xmm1, %xmm1 # 0x3f11b4 vbroadcastss %xmm1, %ymm1 xorl %r14d, %r14d xorl %r9d, %r9d leal 0x7(%r9), %r10d cmpl %r13d, %r10d jge 0x1dd5e8 vmovups (%r8), %ymm2 vfmadd132ps 0x110(%rsp), %ymm0, %ymm2 # ymm2 = (ymm2 * mem) + ymm0 vcmpleps 0x215587(%rip), %ymm2, %k1 # 0x3f29c0 vmaxps 0x213d9d(%rip){1to8}, %ymm2, %ymm2 # 0x3f11e0 vpsrld $0x17, %ymm2, %ymm3 vpbroadcastd 0x213d93(%rip), %ymm4 # 0x3f11e4 vpternlogd $0xea, 0x210bb8(%rip){1to8}, %ymm4, %ymm2 # 0x3ee014 vcmpltps 0x213d85(%rip){1to8}, %ymm2, %k2 # 0x3f11ec vaddps 0x213d7f(%rip){1to8}, %ymm2, %ymm4 # 0x3f11f0 vaddps %ymm2, %ymm4, %ymm4 {%k2} vmulps %ymm4, %ymm4, %ymm2 vbroadcastss 0x213d70(%rip), %ymm5 # 0x3f11f4 vfmadd213ps 0x213d6a(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d64(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d5e(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d58(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d52(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d4c(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d46(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vfmadd213ps 0x213d40(%rip){1to8}, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + mem vmulps %ymm4, %ymm2, %ymm6 vmulps %ymm5, %ymm6, %ymm5 vpaddd 0x213d02(%rip){1to8}, %ymm3, %ymm3 # 0x3f11e8 vcvtdq2ps %ymm3, %ymm3 vbroadcastss 0x211795(%rip), %ymm6 # 0x3eec88 vsubps %ymm6, %ymm3, %ymm3 {%k2} vbroadcastss 0x213cc6(%rip), %ymm7 # 0x3f11c8 vfmadd231ps %ymm7, %ymm3, %ymm5 # ymm5 = (ymm3 * ymm7) + ymm5 vbroadcastss 0x210b04(%rip), %ymm8 # 0x3ee014 vfmsub231ps %ymm2, %ymm8, %ymm5 # ymm5 = (ymm8 * ymm2) - ymm5 vsubps %ymm4, %ymm5, %ymm2 vbroadcastss 0x213ca2(%rip), %ymm5 # 0x3f11c4 vfmsub231ps %ymm3, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm3) - ymm2 vpcmpeqd %ymm3, %ymm3, %ymm3 vmovaps %ymm3, %ymm2 {%k1} vmulps %ymm1, %ymm2, %ymm2 vminps 0x213c79(%rip){1to8}, %ymm2, %ymm2 # 0x3f11b8 vmaxps 0x213c73(%rip){1to8}, %ymm2, %ymm2 # 0x3f11bc vmovaps %ymm8, %ymm3 vfmadd231ps 0x213c69(%rip){1to8}, %ymm2, %ymm3 # ymm3 = (ymm2 * mem) + ymm3 vroundps $0x1, %ymm3, %ymm4 vcmpltps %ymm4, %ymm3, %k1 vsubps %ymm6, %ymm4, %ymm4 {%k1} vfmsub231ps %ymm5, %ymm4, %ymm2 # ymm2 = (ymm4 * ymm5) - ymm2 vfnmsub231ps %ymm7, %ymm4, %ymm2 # ymm2 = -(ymm4 * ymm7) - ymm2 vmulps %ymm2, %ymm2, %ymm3 vbroadcastss 0x213c4b(%rip), %ymm5 # 0x3f11cc vfmadd213ps 0x213c45(%rip){1to8}, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + mem vfmadd213ps 0x213c3f(%rip){1to8}, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + mem vfmadd213ps 0x213c39(%rip){1to8}, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + mem vfmadd213ps 0x213c33(%rip){1to8}, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + mem vfmadd213ps %ymm8, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm8 vfmadd213ps %ymm2, %ymm3, %ymm5 # ymm5 = (ymm3 * ymm5) + ymm2 vcvttps2dq %ymm4, %ymm2 vpslld $0x17, %ymm2, %ymm2 vpaddd 0x2116c2(%rip){1to8}, %ymm2, %ymm2 # 0x3eec88 vmulps (%rdi), %ymm2, %ymm2 vfmadd213ps %ymm2, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm2) + ymm2 vmovups %ymm2, (%rdi) addq $0x20, %r8 addq $0x20, %rdi addl $0x8, %r9d addq $0x8, %r14 jmp 0x1dd412 imulq %rcx, %rax imulq %r15, %rax addq %rax, %r12 imulq %rsi, %rdx imulq %r15, %rdx addq %rdx, %rbp cmpl %r13d, %r14d jge 0x1dd651 movq (%rsp), %rax movq -0x18(%rax), %rax vmovss (%rbp,%r14,4), %xmm0 vmovaps 0x80(%rsp), %xmm1 vfmadd213ss 0xe0(%rbx,%rax), %xmm1, %xmm0 # xmm0 = (xmm1 * xmm0) + mem vmovss 0xdc(%rbx,%rax), %xmm1 vxorps 0x213b7c(%rip){1to4}, %xmm1, %xmm1 # 0x3f11b4 vzeroupper callq 0x5f0e0 vmulss (%r12,%r14,4), %xmm0, %xmm0 vmovss %xmm0, (%r12,%r14,4) incq %r14 jmp 0x1dd5fe incq %r15 movl 0x70(%rsp), %r12d pushq $0x2 popq %r14 jmp 0x1dd2b3 movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd6c0 lock decl (%rax) jne 0x1dd6c0 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dd6b5 movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x1dd6c0 movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd6c2 lock decl (%rax) jne 0x1dd6c2 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dd716 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dd6c2 jmp 0x1dd6b5 jmp 0x1dd716 movq %rsi, %rdi vzeroupper callq 0x5f3e0 xorl %ebp, %ebp movq 0x98(%rsp), %rax testq %rax, %rax je 0x1dd6ff lock decl (%rax) jne 0x1dd6ff movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi je 0x1dd6f4 movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x1dd6ff movq %rsi, %rdi vzeroupper callq 0x5f3e0 movl %ebp, %eax addq $0x178, %rsp # imm = 0x178 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dd6c2 jmp 0x1dd7dc jmp 0x1dd7dc jmp 0x1dd7dc jmp 0x1dd7dc jmp 0x1dd736 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd79d lock decl (%rax) jne 0x1dd79d movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1dd759 jmp 0x1dd784 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dd79d jmp 0x1dd7dc movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dd79d lock decl (%rax) jne 0x1dd79d movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1dd78e movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dd79d movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dd79d jmp 0x1dd7dc jmp 0x1dd7dc movq %rax, %rbx movq 0x98(%rsp), %rax testq %rax, %rax je 0x1dd7d4 lock decl (%rax) jne 0x1dd7d4 movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi jne 0x1dd7ce movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dd7d4 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/build_O2/src/layer/x86/lrn_x86_avx512.cpp

virtual thunk to ncnn::LRN_x86_avx512::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int LRN_x86_avx512::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { *outptr = *ptr * *ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 || p >= channels) continue; const float* sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ssptr += *sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ptr = *ptr * powf(bias + alpha_div_size * *ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size * local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }

movq (%rdi), %rax addq -0x58(%rax), %rdi jmp 0x1dcc74

/csukuangfj[P]ncnn/build_O2/src/layer/x86/lrn_x86_avx512.cpp

ncnn::LRN_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int LRN_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { *outptr = *ptr * *ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 || p >= channels) continue; const float* sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ssptr += *sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ptr = *ptr * powf(bias + alpha_div_size * *ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size * local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x178, %rsp # imm = 0x178 movq %rsi, %r15 movl 0x2c(%rsi), %esi movl 0x30(%r15), %ecx movl 0x38(%r15), %r12d movq 0x10(%r15), %r13 leaq 0x90(%rsp), %rax andq $0x0, 0x40(%rax) movq %rdx, %r14 movq %rdi, %rbx vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rax) vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, 0x20(%rax) vmovups %xmm0, 0x2c(%rax) movq 0x10(%rdx), %r9 movq %rax, %rdi movq %rsi, (%rsp) movq %rcx, 0x10(%rsp) movl %ecx, %edx movl %r12d, %ecx movq %r13, %r8 callq 0x63810 pushq $-0x64 popq %rbp cmpq $0x0, 0x90(%rsp) je 0x1de30e movslq 0xc8(%rsp), %rax imulq 0xd0(%rsp), %rax testq %rax, %rax je 0x1de30e movq %r13, 0x110(%rsp) movq %r14, 0x80(%rsp) movq (%rsp), %rcx movslq %ecx, %rax movq %rax, 0x108(%rsp) movq 0x10(%rsp), %rax movl %eax, %r13d imull %ecx, %r13d xorl %eax, %eax testl %r12d, %r12d movl $0x0, %ecx cmovgl %r12d, %ecx movq %rcx, 0x68(%rsp) movq %rbx, 0x18(%rsp) cmpq 0x68(%rsp), %rax je 0x1dd9c3 movq (%r15), %rcx movq 0x10(%r15), %r8 movq 0x40(%r15), %rdi movq %rdi, %r9 imulq %rax, %r9 imulq %r8, %r9 addq %rcx, %r9 movq 0x90(%rsp), %rdx movq 0xa0(%rsp), %r11 movq 0xd0(%rsp), %r10 movq %r10, %rbx imulq %rax, %rbx imulq %r11, %rbx addq %rdx, %rbx xorl %esi, %esi xorl %r14d, %r14d leal 0x7(%r14), %ebp cmpl %r13d, %ebp jge 0x1dd988 vmovups (%r9), %ymm0 vmulps %ymm0, %ymm0, %ymm0 vmovups %ymm0, (%rbx) addq $0x20, %r9 addq $0x20, %rbx addl $0x8, %r14d addq $0x8, %rsi jmp 0x1dd960 imulq %r11, %r10 imulq %rax, %r10 addq %r10, %rdx imulq %r8, %rdi imulq %rax, %rdi addq %rdi, %rcx movq 0x18(%rsp), %rbx cmpl %r13d, %esi jge 0x1dd9bb vmovss (%rcx,%rsi,4), %xmm0 vmulss %xmm0, %xmm0, %xmm0 vmovss %xmm0, (%rdx,%rsi,4) incq %rsi jmp 0x1dd9a3 incq %rax jmp 0x1dd911 movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd0(%rbx,%rcx), %ecx testl %ecx, %ecx je 0x1ddb0e xorl %ebp, %ebp cmpl $0x1, %ecx jne 0x1de30e movq 0x98(%rsp), %rcx vmovaps 0x90(%rsp), %xmm0 vmovaps %xmm0, 0x20(%rsp) movq 0xa0(%rsp), %rdx movq %rdx, 0x30(%rsp) movl 0xa8(%rsp), %edx movl %edx, 0x38(%rsp) movq 0xb0(%rsp), %rdx movq %rdx, 0x40(%rsp) vmovups 0xb8(%rsp), %xmm0 vmovups %xmm0, 0x48(%rsp) movl 0xc8(%rsp), %edx movl %edx, 0x58(%rsp) movq 0xd0(%rsp), %rdx movq %rdx, 0x60(%rsp) testq %rcx, %rcx je 0x1dda52 lock incl (%rcx) movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %r13d cmpl $0x2, %r13d jl 0x1ddbca shrl %r13d movq 0x80(%rsp), %rcx vmovups (%rcx), %ymm0 vmovups 0x20(%rcx), %ymm1 leaq 0x130(%rsp), %r10 vmovups %ymm1, 0x20(%r10) vmovups %ymm0, (%r10) movq 0x10(%rcx), %rcx movq %rcx, 0x8(%r10) movq -0x18(%rax), %rax movl %r13d, %ecx notl %ecx addl 0xd4(%rbx,%rax), %ecx leaq 0x90(%rsp), %rdi leaq 0x20(%rsp), %rsi vxorps %xmm0, %xmm0, %xmm0 movl %r13d, %edx movl %r13d, %r8d movl %ecx, %r9d pushq %r10 pushq $0x0 vzeroupper callq 0x6466c popq %rax popq %rcx cmpq $0x0, 0x20(%rsp) pushq $-0x64 popq %rbp je 0x1de2d7 movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x1de2d7 movq %r15, %r14 movl 0x4c(%rsp), %r15d movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %r13d jmp 0x1ddbd4 leaq 0x20(%rsp), %rdi andq $0x0, 0x40(%rdi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rdi) vmovups %xmm0, 0xc(%rdi) vmovaps %xmm0, 0x20(%rdi) vmovups %xmm0, 0x2c(%rdi) movq 0x80(%rsp), %rax movq 0x10(%rax), %r9 movq (%rsp), %rsi movq 0x10(%rsp), %rdx movl %r12d, %ecx movq 0x110(%rsp), %r8 vzeroupper callq 0x63810 movq 0x20(%rsp), %rax testq %rax, %rax je 0x1ddb92 movq 0x60(%rsp), %rcx movslq 0x58(%rsp), %rdx movq %rcx, %rsi imulq %rdx, %rsi testq %rsi, %rsi je 0x1ddb92 imull %edx, %ecx xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpl %edx, %ecx je 0x1dde3b andl $0x0, (%rax,%rdx,4) incq %rdx jmp 0x1ddb81 movq 0x28(%rsp), %rax testq %rax, %rax pushq $-0x64 popq %rbp je 0x1de30e lock decl (%rax) jne 0x1de30e movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1de2ff movq (%rdi), %rax callq *0x18(%rax) jmp 0x1de30e movq %r15, %r14 movq (%rsp), %rax movl %eax, %r15d imull %r13d, %r13d vmovss 0xd8(%rbx,%rcx), %xmm0 vmovss %xmm0, 0x80(%rsp) leaq 0x130(%rsp), %rdi leaq 0xf(%rsp), %rdx movq %r13, %rsi vzeroupper callq 0x73bbe vcvtsi2ss %r13d, %xmm2, %xmm0 movq 0x130(%rsp), %rbp movq (%rbx), %rax movq -0x18(%rax), %rcx movl 0xd4(%rbx,%rcx), %edx subl %edx, %r15d xorl %ecx, %ecx xorl %esi, %esi xorl %edi, %edi cmpl %edx, %edi jge 0x1ddc68 movslq %ecx, %rcx leaq (,%rcx,4), %r9 addq %rbp, %r9 movl %esi, %r10d xorl %r8d, %r8d cmpl %edx, %r8d jge 0x1ddc5b leaq (%r10,%r8), %rdx movl %edx, (%r9,%r8,4) movq -0x18(%rax), %rdx movl 0xd4(%rbx,%rdx), %edx incq %r8 jmp 0x1ddc3e addl %r15d, %esi addl %r8d, %esi incl %edi addq %r8, %rcx jmp 0x1ddc26 movq 0x40(%r14), %rax imulq 0x10(%r14), %rax movq %rax, 0xf0(%rsp) movslq 0x4c(%rsp), %rcx movq 0x30(%rsp), %rax movq 0x60(%rsp), %rdx imulq %rax, %rdx movq %rdx, 0xe8(%rsp) imulq %rax, %rcx movq %rcx, 0x100(%rsp) xorl %ecx, %ecx movq (%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, (%rsp) movq (%r14), %rax movq %rax, 0xe0(%rsp) xorl %ecx, %ecx movq 0x10(%rsp), %rax testl %eax, %eax cmovlel %ecx, %eax movq %rax, 0x10(%rsp) movq 0x20(%rsp), %rax movq %rax, 0xd8(%rsp) vmovss 0x210fa8(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0x110(%rsp) cmpq 0x68(%rsp), %rcx je 0x1dddf9 movq 0xf0(%rsp), %r14 imulq %rcx, %r14 addq 0xe0(%rsp), %r14 movq 0xe8(%rsp), %rax movq %rcx, 0xf8(%rsp) imulq %rcx, %rax addq 0xd8(%rsp), %rax movq %rax, 0x70(%rsp) xorl %eax, %eax cmpq 0x10(%rsp), %rax je 0x1ddde9 movq 0x100(%rsp), %r15 movq %rax, 0x78(%rsp) imulq %rax, %r15 addq 0x70(%rsp), %r15 movq (%rbx), %r12 xorl %ebx, %ebx cmpq (%rsp), %rbx je 0x1dddcb leaq (%r15,%rbx,4), %rax vxorps %xmm0, %xmm0, %xmm0 xorl %ecx, %ecx cmpq %rcx, %r13 je 0x1ddd79 movslq (%rbp,%rcx,4), %rdx vaddss (%rax,%rdx,4), %xmm0, %xmm0 incq %rcx jmp 0x1ddd65 movq -0x18(%r12), %rax vmulss 0x80(%rsp), %xmm0, %xmm0 movq 0x18(%rsp), %rcx vmovss 0x110(%rsp), %xmm1 vfmadd213ss 0xe0(%rcx,%rax), %xmm1, %xmm0 # xmm0 = (xmm1 * xmm0) + mem vmovss 0xdc(%rcx,%rax), %xmm1 vbroadcastss 0x213403(%rip), %xmm2 # 0x3f11b4 vxorps %xmm2, %xmm1, %xmm1 callq 0x5f0e0 vmulss (%r14,%rbx,4), %xmm0, %xmm0 vmovss %xmm0, (%r14,%rbx,4) incq %rbx jmp 0x1ddd55 movq 0x108(%rsp), %rax leaq (%r14,%rax,4), %r14 movq 0x78(%rsp), %rax incq %rax movq 0x18(%rsp), %rbx jmp 0x1ddd2f movq 0xf8(%rsp), %rcx incq %rcx jmp 0x1ddced leaq 0x130(%rsp), %rdi callq 0x624be movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de30c lock decl (%rax) jne 0x1de30c movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1de2fd movq (%rdi), %rax callq *0x18(%rax) jmp 0x1de30c movq %r15, 0x78(%rsp) movq (%rbx), %rax movq %rax, (%rsp) movq -0x18(%rax), %rax vmovss 0xd8(%rbx,%rax), %xmm0 vcvtsi2ssl 0xd4(%rbx,%rax), %xmm1, %xmm1 vdivss %xmm1, %xmm0, %xmm0 vmovaps %xmm0, 0x80(%rsp) vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovups %ymm0, 0x110(%rsp) xorl %r15d, %r15d pushq $0x2 popq %r14 movl %r12d, 0x70(%rsp) cmpq 0x68(%rsp), %r15 je 0x1de2ae movq (%rsp), %rax movq -0x18(%rax), %r9 movl 0xd4(%rbx,%r9), %esi movl %esi, %eax cltd pushq $-0x2 popq %rcx idivl %ecx movl %eax, %ecx addl %r15d, %ecx movl %esi, %eax cltd idivl %r14d addl %r15d, %eax cmpl %eax, %ecx jg 0x1ddf85 movl %ecx, %ecx testl %ecx, %ecx sets %al cmpl %r12d, %ecx setge %dl orb %al, %dl jne 0x1ddf7e movq 0x90(%rsp), %rax movq 0xa0(%rsp), %r8 movq 0xd0(%rsp), %rdi movq %rdi, %r9 imulq %rcx, %r9 imulq %r8, %r9 addq %rax, %r9 movq 0x60(%rsp), %r10 imulq %r15, %r10 imulq 0x30(%rsp), %r10 movq 0x20(%rsp), %rdx leaq (%rdx,%r10), %r11 xorl %esi, %esi xorl %ebx, %ebx leal 0x7(%rbx), %ebp cmpl %r13d, %ebp jge 0x1ddf42 vmovups (%r11), %ymm0 vaddps (%r9), %ymm0, %ymm0 vmovups %ymm0, (%r11) addq $0x20, %r9 addq $0x20, %r11 addl $0x8, %ebx addq $0x8, %rsi jmp 0x1ddf1a addq %r10, %rdx imulq %r8, %rdi imulq %rcx, %rdi addq %rdi, %rax movq 0x18(%rsp), %rbx cmpl %r13d, %esi jge 0x1ddf6e vmovss (%rdx,%rsi,4), %xmm0 vaddss (%rax,%rsi,4), %xmm0, %xmm0 vmovss %xmm0, (%rdx,%rsi,4) incq %rsi jmp 0x1ddf55 movq (%rsp), %rax movq -0x18(%rax), %r9 movl 0xd4(%rbx,%r9), %esi incl %ecx jmp 0x1ddeb2 movq 0x78(%rsp), %rax movq (%rax), %r12 movq 0x10(%rax), %rcx movq 0x40(%rax), %rax movq %rax, %rdi imulq %r15, %rdi imulq %rcx, %rdi addq %r12, %rdi movq 0x20(%rsp), %rbp movq 0x30(%rsp), %rsi movq 0x60(%rsp), %rdx movq %rdx, %r8 imulq %r15, %r8 imulq %rsi, %r8 addq %rbp, %r8 vbroadcastss 0xe0(%rbx,%r9), %ymm0 vbroadcastss 0xdc(%rbx,%r9), %xmm1 vbroadcastss 0x2131d7(%rip), %xmm2 # 0x3f11b4 vxorps %xmm2, %xmm1, %xmm1 vinsertf128 $0x1, %xmm1, %ymm1, %ymm1 xorl %r14d, %r14d xorl %r9d, %r9d leal 0x7(%r9), %r10d cmpl %r13d, %r10d jge 0x1de231 vmovups (%r8), %ymm2 vfmadd132ps 0x110(%rsp), %ymm0, %ymm2 # ymm2 = (ymm2 * mem) + ymm0 vbroadcastss 0x2131ce(%rip), %ymm3 # 0x3f11e0 vmaxps %ymm3, %ymm2, %ymm3 vpsrld $0x17, %xmm3, %xmm4 vextractf128 $0x1, %ymm3, %xmm5 vpsrld $0x17, %xmm5, %xmm5 vbroadcastss 0x2131b5(%rip), %ymm6 # 0x3f11e4 vandps %ymm6, %ymm3, %ymm3 vbroadcastss 0x20ffd8(%rip), %ymm8 # 0x3ee014 vorps %ymm3, %ymm8, %ymm3 vbroadcastss 0x2131a3(%rip), %ymm6 # 0x3f11ec vcmpleps %ymm3, %ymm6, %ymm6 vandnps %ymm3, %ymm6, %ymm7 vbroadcastss 0x213195(%rip), %ymm9 # 0x3f11f0 vaddps %ymm3, %ymm9, %ymm3 vaddps %ymm7, %ymm3, %ymm3 vextractf128 $0x1, %ymm6, %xmm7 vpsubd %xmm7, %xmm5, %xmm5 vbroadcastss 0x21388e(%rip), %xmm7 # 0x3f1904 vpaddd %xmm7, %xmm5, %xmm5 vpsubd %xmm6, %xmm4, %xmm4 vpaddd %xmm7, %xmm4, %xmm4 vinsertf128 $0x1, %xmm5, %ymm4, %ymm4 vmulps %ymm3, %ymm3, %ymm5 vbroadcastss 0x21315f(%rip), %ymm6 # 0x3f11f4 vbroadcastss 0x21315a(%rip), %ymm7 # 0x3f11f8 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213150(%rip), %ymm7 # 0x3f11fc vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213146(%rip), %ymm7 # 0x3f1200 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x21313c(%rip), %ymm7 # 0x3f1204 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213132(%rip), %ymm7 # 0x3f1208 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213128(%rip), %ymm7 # 0x3f120c vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x21311e(%rip), %ymm7 # 0x3f1210 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vbroadcastss 0x213114(%rip), %ymm7 # 0x3f1214 vfmadd213ps %ymm7, %ymm3, %ymm6 # ymm6 = (ymm3 * ymm6) + ymm7 vmulps %ymm3, %ymm5, %ymm7 vmulps %ymm6, %ymm7, %ymm6 vcvtdq2ps %ymm4, %ymm4 vbroadcastss 0x2130ae(%rip), %ymm7 # 0x3f11c8 vfmadd231ps %ymm7, %ymm4, %ymm6 # ymm6 = (ymm4 * ymm7) + ymm6 vfmsub231ps %ymm5, %ymm8, %ymm6 # ymm6 = (ymm8 * ymm5) - ymm6 vcmpleps 0x214893(%rip), %ymm2, %ymm2 # 0x3f29c0 vsubps %ymm3, %ymm6, %ymm3 vbroadcastss 0x21308a(%rip), %ymm5 # 0x3f11c4 vfmsub231ps %ymm4, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm4) - ymm3 vorps %ymm3, %ymm2, %ymm2 vmulps %ymm1, %ymm2, %ymm2 vbroadcastss 0x213068(%rip), %ymm3 # 0x3f11b8 vminps %ymm3, %ymm2, %ymm2 vbroadcastss 0x21305f(%rip), %ymm3 # 0x3f11bc vmaxps %ymm3, %ymm2, %ymm2 vbroadcastss 0x213056(%rip), %ymm3 # 0x3f11c0 vfmadd213ps %ymm8, %ymm2, %ymm3 # ymm3 = (ymm2 * ymm3) + ymm8 vroundps $0x1, %ymm3, %ymm4 vcmpltps %ymm4, %ymm3, %ymm3 vbroadcastss 0x210b05(%rip), %ymm6 # 0x3eec88 vandps %ymm6, %ymm3, %ymm3 vsubps %ymm3, %ymm4, %ymm3 vfmsub231ps %ymm5, %ymm3, %ymm2 # ymm2 = (ymm3 * ymm5) - ymm2 vfnmsub231ps %ymm7, %ymm3, %ymm2 # ymm2 = -(ymm3 * ymm7) - ymm2 vmulps %ymm2, %ymm2, %ymm4 vbroadcastss 0x21302a(%rip), %ymm5 # 0x3f11cc vbroadcastss 0x213025(%rip), %ymm6 # 0x3f11d0 vfmadd213ps %ymm6, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm6 vbroadcastss 0x21301b(%rip), %ymm6 # 0x3f11d4 vfmadd213ps %ymm6, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm6 vbroadcastss 0x213011(%rip), %ymm6 # 0x3f11d8 vfmadd213ps %ymm6, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm6 vbroadcastss 0x213007(%rip), %ymm6 # 0x3f11dc vfmadd213ps %ymm6, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm6 vfmadd213ps %ymm8, %ymm2, %ymm5 # ymm5 = (ymm2 * ymm5) + ymm8 vfmadd213ps %ymm2, %ymm4, %ymm5 # ymm5 = (ymm4 * ymm5) + ymm2 vcvttps2dq %ymm3, %ymm2 vpslld $0x17, %xmm2, %xmm3 vextractf128 $0x1, %ymm2, %xmm2 vpslld $0x17, %xmm2, %xmm2 vbroadcastss 0x210a87(%rip), %xmm4 # 0x3eec88 vpaddd %xmm4, %xmm2, %xmm2 vpaddd %xmm4, %xmm3, %xmm3 vinsertf128 $0x1, %xmm2, %ymm3, %ymm2 vmulps (%rdi), %ymm2, %ymm2 vfmadd213ps %ymm2, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm2) + ymm2 vmovups %ymm2, (%rdi) addq $0x20, %r8 addq $0x20, %rdi addl $0x8, %r9d addq $0x8, %r14 jmp 0x1ddfed imulq %rcx, %rax imulq %r15, %rax addq %rax, %r12 imulq %rsi, %rdx imulq %r15, %rdx addq %rdx, %rbp cmpl %r13d, %r14d jge 0x1de29d movq (%rsp), %rax movq -0x18(%rax), %rax vmovss (%rbp,%r14,4), %xmm0 vmovaps 0x80(%rsp), %xmm1 vfmadd213ss 0xe0(%rbx,%rax), %xmm1, %xmm0 # xmm0 = (xmm1 * xmm0) + mem vmovss 0xdc(%rbx,%rax), %xmm1 vbroadcastss 0x212f34(%rip), %xmm2 # 0x3f11b4 vxorps %xmm2, %xmm1, %xmm1 vzeroupper callq 0x5f0e0 vmulss (%r12,%r14,4), %xmm0, %xmm0 vmovss %xmm0, (%r12,%r14,4) incq %r14 jmp 0x1de247 incq %r15 movl 0x70(%rsp), %r12d pushq $0x2 popq %r14 jmp 0x1dde8a movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de30c lock decl (%rax) jne 0x1de30c movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1de301 movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x1de30c movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de30e lock decl (%rax) jne 0x1de30e movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1de362 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1de30e jmp 0x1de301 jmp 0x1de362 movq %rsi, %rdi vzeroupper callq 0x5f3e0 xorl %ebp, %ebp movq 0x98(%rsp), %rax testq %rax, %rax je 0x1de34b lock decl (%rax) jne 0x1de34b movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi je 0x1de340 movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x1de34b movq %rsi, %rdi vzeroupper callq 0x5f3e0 movl %ebp, %eax addq $0x178, %rsp # imm = 0x178 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movq %rsi, %rdi callq 0x5f3e0 jmp 0x1de30e jmp 0x1de428 jmp 0x1de428 jmp 0x1de428 jmp 0x1de428 jmp 0x1de382 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de3e9 lock decl (%rax) jne 0x1de3e9 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1de3a5 jmp 0x1de3d0 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1de3e9 jmp 0x1de428 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x1de3e9 lock decl (%rax) jne 0x1de3e9 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x1de3da movq %rsi, %rdi callq 0x5f3e0 jmp 0x1de3e9 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1de3e9 jmp 0x1de428 jmp 0x1de428 movq %rax, %rbx movq 0x98(%rsp), %rax testq %rax, %rax je 0x1de420 lock decl (%rax) jne 0x1de420 movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi jne 0x1de41a movq %rsi, %rdi callq 0x5f3e0 jmp 0x1de420 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/build_O2/src/layer/x86/lrn_x86_fma.cpp

virtual thunk to ncnn::LRN_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int LRN_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { int w = bottom_top_blob.w; int h = bottom_top_blob.h; int channels = bottom_top_blob.c; size_t elemsize = bottom_top_blob.elemsize; int size = w * h; // squared values with local_size padding Mat square_blob; square_blob.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_top_blob.channel(q); float* outptr = square_blob.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _outp = _mm256_mul_ps(_p, _p); _mm256_storeu_ps(outptr, _outp); ptr += 8; outptr += 8; } #endif // __AVX__ for (; i < size; i++) { *outptr = *ptr * *ptr; ptr++; outptr++; } } if (region_type == NormRegion_ACROSS_CHANNELS) { Mat square_sum; square_sum.create(w, h, channels, elemsize, opt.workspace_allocator); if (square_sum.empty()) return -100; square_sum.fill(0.f); const float alpha_div_size = alpha / local_size; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // square sum for (int p = q - local_size / 2; p <= q + local_size / 2; p++) { if (p < 0 || p >= channels) continue; const float* sptr = square_blob.channel(p); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ for (; i + 7 < size; i += 8) { __m256 _sp = _mm256_loadu_ps(sptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_add_ps(_ssp, _sp); _mm256_storeu_ps(ssptr, _ssp); sptr += 8; ssptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ssptr += *sptr; sptr++; ssptr++; } } float* ptr = bottom_top_blob.channel(q); float* ssptr = square_sum.channel(q); int i = 0; #if __AVX__ __m256 _bias = _mm256_set1_ps(bias); __m256 _ads = _mm256_set1_ps(alpha_div_size); __m256 _mb = _mm256_set1_ps(-beta); for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _ssp = _mm256_loadu_ps(ssptr); _ssp = _mm256_mul_ps(_ssp, _ads); _ssp = _mm256_add_ps(_ssp, _bias); _ssp = pow256_ps(_ssp, _mb); _p = _mm256_mul_ps(_p, _ssp); _mm256_storeu_ps(ptr, _p); ssptr += 8; ptr += 8; } #endif // __AVX__ for (; i < size; i++) { *ptr = *ptr * powf(bias + alpha_div_size * *ssptr, -beta); ssptr++; ptr++; } } } else if (region_type == NormRegion_WITHIN_CHANNEL) { int outw = w; int outh = h; Mat square_blob_bordered = square_blob; int pad = local_size / 2; if (pad > 0) { Option opt_b = opt; opt_b.blob_allocator = opt.workspace_allocator; copy_make_border(square_blob, square_blob_bordered, pad, local_size - pad - 1, pad, local_size - pad - 1, BORDER_CONSTANT, 0.f, opt_b); if (square_blob_bordered.empty()) return -100; w = square_blob_bordered.w; h = square_blob_bordered.h; } const int maxk = local_size * local_size; const float alpha_div_size = alpha / maxk; // norm window offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w - local_size; for (int i = 0; i < local_size; i++) { for (int j = 0; j < local_size; j++) { space_ofs[p1] = p2; p1++; p2++; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* ptr = bottom_top_blob.channel(q); const Mat m = square_blob_bordered.channel(q); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { const float* sptr = m.row(i) + j; float ss = 0.f; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; ss += val; } ptr[j] = ptr[j] * powf(bias + alpha_div_size * ss, -beta); } ptr += outw; } } } return 0; }

movq (%rdi), %rax addq -0x58(%rax), %rdi jmp 0x1dd838

/csukuangfj[P]ncnn/build_O2/src/layer/x86/lrn_x86_fma.cpp

ncnn::MemoryData::load_model(ncnn::ModelBin const&)

int MemoryData::load_model(const ModelBin& mb) { if (d != 0) { data = mb.load(w, h, d, c, 1); } else if (c != 0) { data = mb.load(w, h, c, 1); } else if (h != 0) { data = mb.load(w, h, 1); } else if (w != 0) { data = mb.load(w, 1); } else // 0 0 0 { data.create(1); } if (data.empty()) return -100; return 0; }

pushq %r14 pushq %rbx subq $0x58, %rsp movq %rdi, %rbx movl 0xd8(%rdi), %r8d testl %r8d, %r8d je 0x1df233 movl 0xd0(%rbx), %edx movl 0xd4(%rbx), %ecx movl 0xdc(%rbx), %r9d movq (%rsi), %rax movl $0x1, (%rsp) leaq 0x10(%rsp), %r14 movq %r14, %rdi callq *0x28(%rax) leaq 0xe0(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x1df407 testq %rax, %rax je 0x1df1f8 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x1df3ab lock decl (%rax) jne 0x1df3ab movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x1df3a3 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df3ab movl 0xdc(%rbx), %r8d testl %r8d, %r8d je 0x1df2b4 movl 0xd0(%rbx), %edx movl 0xd4(%rbx), %ecx movq (%rsi), %rax leaq 0x10(%rsp), %r14 pushq $0x1 popq %r9 movq %r14, %rdi callq *0x20(%rax) leaq 0xe0(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x1df4b4 testq %rax, %rax je 0x1df279 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x1df458 lock decl (%rax) jne 0x1df458 movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x1df450 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df458 movl 0xd4(%rbx), %ecx testl %ecx, %ecx je 0x1df32d movl 0xd0(%rbx), %edx movq (%rsi), %rax leaq 0x10(%rsp), %r14 pushq $0x1 popq %r8 movq %r14, %rdi callq *0x18(%rax) leaq 0xe0(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x1df548 testq %rax, %rax je 0x1df2f2 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x1df4ec lock decl (%rax) jne 0x1df4ec movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x1df4e4 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df4ec movl 0xd0(%rbx), %edx testl %edx, %edx je 0x1df437 movq (%rsi), %rax leaq 0x10(%rsp), %r14 pushq $0x1 popq %rcx movq %r14, %rdi callq *0x10(%rax) leaq 0xe0(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x1df5dc testq %rax, %rax je 0x1df368 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x1df580 lock decl (%rax) jne 0x1df580 movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x1df578 movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df580 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x1df605 lock decl (%rax) jne 0x1df605 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df5fd movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df605 leaq 0xe0(%rbx), %rdi pushq $0x1 popq %rsi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa jmp 0x1df605 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x1df605 lock decl (%rax) jne 0x1df605 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df5fd movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df605 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x1df605 lock decl (%rax) jne 0x1df605 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df5fd movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df605 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x1df605 lock decl (%rax) jne 0x1df605 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df5fd movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df605 movq %rsi, %rdi callq 0x5f3e0 cmpq $0x0, 0xe0(%rbx) je 0x1df625 movslq 0x118(%rbx), %rcx imulq 0x120(%rbx), %rcx xorl %eax, %eax testq %rcx, %rcx jne 0x1df628 pushq $-0x64 popq %rax addq $0x58, %rsp popq %rbx popq %r14 retq movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x1df6fd lock decl (%rax) jne 0x1df6fd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df6ed movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df6fd jmp 0x1df707 jmp 0x1df707 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x1df6fd lock decl (%rax) jne 0x1df6fd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df6ed movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df6fd jmp 0x1df707 jmp 0x1df707 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x1df6fd lock decl (%rax) jne 0x1df6fd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x1df6ed movq (%rdi), %rax callq *0x18(%rax) jmp 0x1df6fd jmp 0x1df707 jmp 0x1df707 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x1df6fd lock decl (%rax) jne 0x1df6fd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi jne 0x1df6f7 movq %rsi, %rdi callq 0x5f3e0 jmp 0x1df6fd movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x1df707 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/memorydata.cpp

ncnn::MVN::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int MVN::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int size = w * h; top_blob.create(w, h, channels, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; // prepare sum per channel Mat sum(channels, elemsize, opt.workspace_allocator); if (sum.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_blob.channel(q); float s = 0.f; for (int i = 0; i < size; i++) { s += ptr[i]; } sum[q] = s; } if (across_channels) { // compute mean across channels float mean = 0.f; for (int q = 0; q < channels; q++) { mean += sum[q]; } mean = mean / (channels * size); // subtract mean #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_blob.channel(q); float* outptr = top_blob.channel(q); for (int i = 0; i < size; i++) { outptr[i] = ptr[i] - mean; } } } else { // subtract mean #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = bottom_blob.channel(q); float* outptr = top_blob.channel(q); float mean = sum[q] / size; for (int i = 0; i < size; i++) { outptr[i] = ptr[i] - mean; } } } if (normalize_variance) { // prepare squared sum per channel Mat sqsum(channels, elemsize, opt.workspace_allocator); if (sqsum.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = top_blob.channel(q); float s = 0.f; for (int i = 0; i < size; i++) { s += ptr[i] * ptr[i]; } sqsum[q] = s; } if (across_channels) { // compute squared mean across channels float sqmean = 0.f; for (int q = 0; q < channels; q++) { sqmean += sqsum[q]; } sqmean = sqmean / (channels * size); // normalize variance float norm_var = sqrtf(sqmean) + eps; float norm_var_inv = 1.f / norm_var; // apply normalize_variance #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* outptr = top_blob.channel(q); for (int i = 0; i < size; i++) { outptr[i] = outptr[i] * norm_var_inv; } } } else { // apply normalize_variance #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float* outptr = top_blob.channel(q); float sqmean = sqsum[q] / size; float norm_var = sqrt(sqmean) + eps; float norm_var_inv = 1.f / norm_var; for (int i = 0; i < size; i++) { outptr[i] = outptr[i] * norm_var_inv; } } } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0xb8, %rsp movq %rdx, %r14 movq %rsi, %r12 movq %rdi, 0x8(%rsp) movl 0x2c(%rsi), %ebp movl 0x30(%rsi), %r15d movl 0x38(%rsi), %r13d movq 0x10(%rsi), %rbx movq %rcx, 0x10(%rsp) movq 0x8(%rcx), %r9 movq %rdx, %rdi movl %ebp, %esi movl %r15d, %edx movl %r13d, %ecx movq %rbx, %r8 callq 0x63810 pushq $-0x64 popq %rax cmpq $0x0, (%r14) je 0x1dfd1f movslq 0x38(%r14), %rcx imulq 0x40(%r14), %rcx testq %rcx, %rcx je 0x1dfd1f movl %r13d, %esi movl %r15d, %r13d movq 0x10(%rsp), %rax movq 0x10(%rax), %rcx leaq 0x70(%rsp), %r15 andq $0x0, 0x40(%r15) xorps %xmm0, %xmm0 movaps %xmm0, (%r15) movups %xmm0, 0xc(%r15) movaps %xmm0, 0x20(%r15) movups %xmm0, 0x2c(%r15) movq %r15, %rdi movq %rbx, %rdx movl %esi, %ebx movq %rdx, 0x18(%rsp) callq 0x635fa movq (%r15), %r8 pushq $-0x64 popq %rax testq %r8, %r8 je 0x1dfce8 movslq 0xa8(%rsp), %rcx imulq 0xb0(%rsp), %rcx testq %rcx, %rcx je 0x1dfce8 imull %ebp, %r13d movq (%r12), %rcx movq 0x40(%r12), %rdx imulq 0x10(%r12), %rdx xorl %esi, %esi testl %r13d, %r13d movl $0x0, %r15d movl %r13d, 0x4(%rsp) cmovgl %r13d, %r15d movl %ebx, %r13d testl %ebx, %ebx movl $0x0, %ebp cmovgl %ebx, %ebp movq 0x8(%rsp), %r10 cmpq %rbp, %rsi je 0x1df9f8 xorps %xmm0, %xmm0 xorl %edi, %edi cmpq %rdi, %r15 je 0x1df9ea addss (%rcx,%rdi,4), %xmm0 incq %rdi jmp 0x1df9db movss %xmm0, (%r8,%rsi,4) incq %rsi addq %rdx, %rcx jmp 0x1df9d1 cmpl $0x0, 0xd4(%r10) je 0x1dfa79 xorps %xmm0, %xmm0 movq 0x70(%rsp), %rdx xorl %ecx, %ecx cmpq %rcx, %rbp je 0x1dfa1b addss (%rdx,%rcx,4), %xmm0 incq %rcx jmp 0x1dfa0c movl %r13d, %ecx imull 0x4(%rsp), %ecx cvtsi2ss %ecx, %xmm1 divss %xmm1, %xmm0 movq (%r12), %r9 movq 0x40(%r12), %rcx movq 0x40(%r14), %rdx imulq 0x10(%r14), %rdx movq (%r14), %rsi imulq 0x10(%r12), %rcx xorl %edi, %edi cmpq %rbp, %rdi je 0x1dfae5 xorl %r8d, %r8d cmpq %r8, %r15 je 0x1dfa6e movss (%r9,%r8,4), %xmm1 subss %xmm0, %xmm1 movss %xmm1, (%rsi,%r8,4) incq %r8 jmp 0x1dfa54 incq %rdi addq %rdx, %rsi addq %rcx, %r9 jmp 0x1dfa48 movq (%r12), %r11 movq 0x40(%r12), %rcx movq (%r14), %rdx movq 0x40(%r14), %rsi imulq 0x10(%r14), %rsi movq 0x70(%rsp), %rdi cvtsi2ssl 0x4(%rsp), %xmm1 movss 0x20f1e7(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 imulq 0x10(%r12), %rcx xorl %r8d, %r8d cmpq %rbp, %r8 je 0x1dfae5 movss (%rdi,%r8,4), %xmm1 mulss %xmm0, %xmm1 xorl %r9d, %r9d cmpq %r9, %r15 je 0x1dfada movss (%r11,%r9,4), %xmm2 subss %xmm1, %xmm2 movss %xmm2, (%rdx,%r9,4) incq %r9 jmp 0x1dfac0 incq %r8 addq %rsi, %rdx addq %rcx, %r11 jmp 0x1dfaae cmpl $0x0, 0xd0(%r10) je 0x1dfce6 movl %eax, %ebx movq 0x10(%rsp), %rax movq 0x10(%rax), %rcx leaq 0x20(%rsp), %rdi andq $0x0, 0x40(%rdi) xorps %xmm0, %xmm0 movaps %xmm0, (%rdi) movups %xmm0, 0xc(%rdi) movaps %xmm0, 0x20(%rdi) movups %xmm0, 0x2c(%rdi) movl %r13d, %esi movq 0x18(%rsp), %rdx callq 0x635fa movq 0x20(%rsp), %r9 testq %r9, %r9 movl %ebx, %eax je 0x1dfb85 movslq 0x58(%rsp), %rcx imulq 0x60(%rsp), %rcx testq %rcx, %rcx je 0x1dfb85 movq (%r14), %rcx movq 0x40(%r14), %rdx imulq 0x10(%r14), %rdx xorl %esi, %esi movq 0x8(%rsp), %r8 cmpq %rbp, %rsi je 0x1dfbbc xorps %xmm0, %xmm0 xorl %edi, %edi cmpq %rdi, %r15 je 0x1dfb77 movss (%rcx,%rdi,4), %xmm1 mulss %xmm1, %xmm1 addss %xmm1, %xmm0 incq %rdi jmp 0x1dfb60 movss %xmm0, (%r9,%rsi,4) incq %rsi addq %rdx, %rcx jmp 0x1dfb56 movq 0x28(%rsp), %rcx testq %rcx, %rcx je 0x1dfce8 lock decl (%rcx) jne 0x1dfce8 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dfcd2 movq (%rdi), %rax callq *0x18(%rax) movl %ebx, %eax jmp 0x1dfce8 cmpl $0x0, 0xd4(%r8) je 0x1dfc40 xorps %xmm0, %xmm0 movq 0x20(%rsp), %rax xorl %ecx, %ecx cmpq %rcx, %rbp je 0x1dfbdf addss (%rax,%rcx,4), %xmm0 incq %rcx jmp 0x1dfbd0 imull 0x4(%rsp), %r13d cvtsi2ss %r13d, %xmm1 divss %xmm1, %xmm0 sqrtss %xmm0, %xmm1 addss 0xd8(%r8), %xmm1 movq 0x40(%r14), %rax imulq 0x10(%r14), %rax movq (%r14), %rcx movss 0x20f079(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 xorl %edx, %edx cmpq %rbp, %rdx je 0x1dfcac xorl %esi, %esi cmpq %rsi, %r15 je 0x1dfc38 movss (%rcx,%rsi,4), %xmm1 mulss %xmm0, %xmm1 movss %xmm1, (%rcx,%rsi,4) incq %rsi jmp 0x1dfc20 incq %rdx addq %rax, %rcx jmp 0x1dfc15 movq (%r14), %rax movq 0x40(%r14), %rcx imulq 0x10(%r14), %rcx movq 0x20(%rsp), %rdx cvtsi2ssl 0x4(%rsp), %xmm2 movss 0x20f029(%rip), %xmm0 # 0x3eec88 movaps %xmm0, %xmm1 divss %xmm2, %xmm1 xorl %esi, %esi cmpq %rbp, %rsi je 0x1dfcac movss (%rdx,%rsi,4), %xmm2 mulss %xmm1, %xmm2 sqrtss %xmm2, %xmm3 addss 0xd8(%r8), %xmm3 movaps %xmm0, %xmm2 divss %xmm3, %xmm2 xorl %edi, %edi cmpq %rdi, %r15 je 0x1dfca4 movss (%rax,%rdi,4), %xmm3 mulss %xmm2, %xmm3 movss %xmm3, (%rax,%rdi,4) incq %rdi jmp 0x1dfc8c incq %rsi addq %rcx, %rax jmp 0x1dfc68 movq 0x28(%rsp), %rax testq %rax, %rax je 0x1dfce6 lock decl (%rax) jne 0x1dfce6 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x1dfcde movq (%rdi), %rax callq *0x18(%rax) jmp 0x1dfce6 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax jmp 0x1dfce8 movq %rsi, %rdi callq 0x5f3e0 xorl %eax, %eax movq 0x78(%rsp), %rcx testq %rcx, %rcx je 0x1dfd1f lock decl (%rcx) jne 0x1dfd1f movl %eax, %ebx movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x1dfd15 movq (%rdi), %rax callq *0x18(%rax) movl %ebx, %eax jmp 0x1dfd1f movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0xb8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x1dfd35 jmp 0x1dfd35 movq %rax, %rdi callq 0x61d68 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x1dfd71 lock decl (%rax) jne 0x1dfd71 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x1dfd6b movq %rsi, %rdi callq 0x5f3e0 jmp 0x1dfd71 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x1dfd35 nop

/csukuangfj[P]ncnn/src/layer/mvn.cpp

ncnn::Pooling::load_param(ncnn::ParamDict const&)

int Pooling::load_param(const ParamDict& pd) { pooling_type = pd.get(0, 0); kernel_w = pd.get(1, 0); kernel_h = pd.get(11, kernel_w); stride_w = pd.get(2, 1); stride_h = pd.get(12, stride_w); pad_left = pd.get(3, 0); pad_right = pd.get(14, pad_left); pad_top = pd.get(13, pad_left); pad_bottom = pd.get(15, pad_top); global_pooling = pd.get(4, 0); pad_mode = pd.get(5, 0); avgpool_count_include_pad = pd.get(6, 0); adaptive_pooling = pd.get(7, 0); out_w = pd.get(8, 0); out_h = pd.get(18, out_w); return 0; }

pushq %rbp pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rbp movq %r14, %rdi movl %ebp, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd4(%rbx) pushq $0xb popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi movl %ebp, %edx callq 0x718a6 movl %eax, 0xdc(%rbx) pushq $0xc popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xe0(%rbx) pushq $0x3 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe4(%rbx) pushq $0xe popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xe8(%rbx) movl 0xe4(%rbx), %edx pushq $0xd popq %rsi movq %r14, %rdi callq 0x718a6 movl %eax, 0xec(%rbx) pushq $0xf popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xf0(%rbx) pushq $0x4 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xf4(%rbx) pushq $0x5 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xf8(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xfc(%rbx) pushq $0x7 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x100(%rbx) pushq $0x8 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x104(%rbx) pushq $0x12 popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0x108(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %rbp retq nop

/csukuangfj[P]ncnn/src/layer/pooling.cpp

ncnn::Pooling_x86_fma::create_pipeline(ncnn::Option const&)

int Pooling_x86_fma::create_pipeline(const Option& /*opt*/) { if (adaptive_pooling) { support_packing = false; support_bf16_storage = false; support_fp16_storage = false; support_int8_storage = false; support_tensor_storage = false; } return 0; }

movq (%rdi), %rax movq -0x18(%rax), %rcx cmpl $0x0, 0x100(%rdi,%rcx) je 0x1e431f xorl %edx, %edx movb %dl, 0xb(%rdi,%rcx) movq -0x18(%rax), %rcx movb %dl, 0xc(%rdi,%rcx) movq -0x18(%rax), %rcx movb %dl, 0xd(%rdi,%rcx) movq -0x18(%rax), %rcx movb %dl, 0xe(%rdi,%rcx) movq -0x18(%rax), %rax movb %dl, 0x10(%rdi,%rax) xorl %eax, %eax retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/pooling_x86_fma.cpp

ncnn::binary_op_broadcast_inner(ncnn::Mat const&, ncnn::Mat const&, ncnn::Mat&, int, ncnn::Option const&)

static int binary_op_broadcast_inner(const Mat& a, const Mat& b, Mat& c, int op_type, const Option& opt) { // squeeze inner axes Mat b2 = b; if (b.dims == 2 && b.w == 1) b2 = b.reshape(b.h); else if (b.dims == 3 && b.h == 1) b2 = b.reshape(b.c); else if (b.dims == 3 && b.w == 1) b2 = b.reshape(b.h, b.c); else if (b.dims == 4 && b.d == 1) b2 = b.reshape(b.c); else if (b.dims == 4 && b.h == 1) b2 = b.reshape(b.d, b.c); else if (b.dims == 4 && b.w == 1) b2 = b.reshape(b.h, b.d, b.c); using namespace BinaryOp_x86_fma_functor; if (op_type == BinaryOp::Operation_ADD) return binary_op_broadcast_inner<binary_op_add>(a, b2, c, opt); if (op_type == BinaryOp::Operation_SUB) return binary_op_broadcast_inner<binary_op_sub>(a, b2, c, opt); if (op_type == BinaryOp::Operation_MUL) return binary_op_broadcast_inner<binary_op_mul>(a, b2, c, opt); if (op_type == BinaryOp::Operation_DIV) return binary_op_broadcast_inner<binary_op_div>(a, b2, c, opt); if (op_type == BinaryOp::Operation_MAX) return binary_op_broadcast_inner<binary_op_max>(a, b2, c, opt); if (op_type == BinaryOp::Operation_MIN) return binary_op_broadcast_inner<binary_op_min>(a, b2, c, opt); if (op_type == BinaryOp::Operation_POW) return binary_op_broadcast_inner<binary_op_pow>(a, b2, c, opt); if (op_type == BinaryOp::Operation_RSUB) return binary_op_broadcast_inner<binary_op_rsub>(a, b2, c, opt); if (op_type == BinaryOp::Operation_RDIV) return binary_op_broadcast_inner<binary_op_rdiv>(a, b2, c, opt); if (op_type == BinaryOp::Operation_RPOW) return binary_op_broadcast_inner<binary_op_rpow>(a, b2, c, opt); if (op_type == BinaryOp::Operation_ATAN2) return binary_op_broadcast_inner<binary_op_atan2>(a, b2, c, opt); if (op_type == BinaryOp::Operation_RATAN2) return binary_op_broadcast_inner<binary_op_ratan2>(a, b2, c, opt); // should never reach here return 0; }

pushq %rbp movq %rsp, %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx andq $-0x20, %rsp subq $0x1a0, %rsp # imm = 0x1A0 movl %ecx, %r13d movq %rdx, 0x48(%rsp) movq %rdi, 0x20(%rsp) movq (%rsi), %rbx movq 0x8(%rsi), %rcx movq 0x10(%rsi), %rax movq %rax, 0x78(%rsp) movq 0x20(%rsi), %rax movq %rax, 0x50(%rsp) vmovq 0x28(%rsi), %xmm1 movq 0x40(%rsi), %rax movq %rax, 0x98(%rsp) vmovdqa %xmm1, %xmm0 testq %rcx, %rcx je 0x264d73 lock incl (%rcx) vmovq 0x28(%rsi), %xmm0 movq %rcx, 0x58(%rsp) vmovd %xmm0, %ecx movl %ecx, %eax xorl $0x2, %eax vpextrd $0x1, %xmm0, %edi xorl $0x1, %edi orl %edi, %eax jne 0x264de3 movl 0x30(%rsi), %edx leaq 0x140(%rsp), %rdi xorl %ecx, %ecx callq 0x62c8a movq 0x148(%rsp), %rax testq %rax, %rax je 0x264db0 lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x264f79 lock decl (%rax) jne 0x264f79 movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x264f71 movq (%rdi), %rax movq %rbx, %rsi callq *0x18(%rax) jmp 0x264f79 movl %ecx, %eax xorl $0x3, %eax movl 0x30(%rsi), %edx movl %edx, %r8d xorl $0x1, %r8d movl %eax, %r9d orl %r8d, %r9d jne 0x264e4f movl 0x38(%rsi), %edx leaq 0x140(%rsp), %rdi xorl %ecx, %ecx callq 0x62c8a movq 0x148(%rsp), %rax testq %rax, %rax je 0x264e1c lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x26506e lock decl (%rax) jne 0x26506e movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x265066 movq (%rdi), %rax movq %rbx, %rsi callq *0x18(%rax) jmp 0x26506e orl %edi, %eax jne 0x264ea9 movl 0x38(%rsi), %ecx leaq 0x140(%rsp), %rdi xorl %r8d, %r8d callq 0x62e4e movq 0x148(%rsp), %rax testq %rax, %rax je 0x264e76 lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x265102 lock decl (%rax) jne 0x265102 movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2650fa movq (%rdi), %rax movq %rbx, %rsi callq *0x18(%rax) jmp 0x265102 xorl $0x4, %ecx movl 0x34(%rsi), %eax movl %eax, %r9d xorl $0x1, %r9d orl %ecx, %r9d jne 0x264f10 movl 0x38(%rsi), %edx leaq 0x140(%rsp), %rdi xorl %ecx, %ecx callq 0x62c8a movq 0x148(%rsp), %rax testq %rax, %rax je 0x264edd lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x265196 lock decl (%rax) jne 0x265196 movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x26518e movq (%rdi), %rax movq %rbx, %rsi callq *0x18(%rax) jmp 0x265196 orl %ecx, %r8d jne 0x265005 movl 0x38(%rsi), %ecx leaq 0x140(%rsp), %rdi movl %eax, %edx xorl %r8d, %r8d callq 0x62e4e movq 0x148(%rsp), %rax testq %rax, %rax je 0x264f3e lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x26522a lock decl (%rax) jne 0x26522a movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x265222 movq (%rdi), %rax movq %rbx, %rsi callq *0x18(%rax) jmp 0x26522a movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq *0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c orl %edi, %ecx jne 0x265351 movl 0x38(%rsi), %r8d leaq 0x140(%rsp), %rdi movl %eax, %ecx xorl %r9d, %r9d callq 0x63020 movq 0x148(%rsp), %rax testq %rax, %rax je 0x265033 lock incl (%rax) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2652be lock decl (%rax) jne 0x2652be movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2652b6 movq (%rdi), %rax movq %rbx, %rsi callq *0x18(%rax) jmp 0x2652be movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq *0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq *0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq *0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq *0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rbx, %rdi callq 0x5f3e0 movq 0x140(%rsp), %rbx movq 0x148(%rsp), %r14 movq 0x150(%rsp), %rax movq %rax, 0x78(%rsp) movq 0x160(%rsp), %rax movq %rax, 0x50(%rsp) vmovq 0x168(%rsp), %xmm1 movq 0x180(%rsp), %rax movq %rax, 0x98(%rsp) testq %r14, %r14 je 0x26534c lock decl (%r14) jne 0x26534c vmovdqa %xmm1, 0x80(%rsp) movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26533b movq (%rdi), %rax callq *0x18(%rax) vmovdqa 0x80(%rsp), %xmm1 jmp 0x26534c movq %rsi, %rdi callq 0x5f3e0 vmovdqa 0x80(%rsp), %xmm1 movq %r14, 0x58(%rsp) movq %rbx, 0x18(%rsp) cmpl $0xb, %r13d ja 0x26bab3 movl %r13d, %eax leaq 0x192102(%rip), %rcx # 0x3f746c movslq (%rcx,%rax,4), %rax addq %rcx, %rax vmovd %xmm1, 0x38(%rsp) vpextrd $0x1, %xmm1, 0x34(%rsp) jmpq *%rax movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x268001 cmpl $0x1, 0x38(%rsp) jne 0x268001 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x267ff9 vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x2653fd movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x265416 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x265416 movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x26541c vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x2654a8 vaddps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x26546a vaddps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x26548e imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2654eb vaddss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2654d5 incq %r14 movq 0x18(%rsp), %r10 jmp 0x2653d8 movq 0x20(%rsp), %rax movl 0x2c(%rax), %ecx movl %ecx, 0x80(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x40(%rsp) movl 0x34(%rax), %ecx movq %rcx, 0xf8(%rsp) movl 0x38(%rax), %ecx movq %rcx, 0xf0(%rsp) movl 0x18(%rax), %r12d movl 0x28(%rax), %eax cmpl $0x2, %eax movq 0x18(%rsp), %rbx movl %r12d, 0xdc(%rsp) jne 0x26966d cmpl $0x1, 0x38(%rsp) jne 0x26966d movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %r13d, %r13d movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x269665 vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0xb0(%rsp) jne 0x2655ad movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x2655cf vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x2655cf movq %r13, %rax shlq $0x5, %rax vmovups (%rbx,%rax), %ymm0 jmp 0x2655d5 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, %r12 movq %rcx, 0xa0(%rsp) imulq %rcx, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0x28(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, 0xa8(%rsp) movq %rcx, 0xc8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x60(%rsp) addq %rcx, %r13 vmovaps %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x2656db vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x26564e vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x26569a movq 0xa8(%rsp), %r13 movq 0x138(%rsp), %rax imulq %r13, %rax imulq 0xc8(%rsp), %rax addq %rax, 0x60(%rsp) movq 0xd0(%rsp), %rax imulq %r13, %rax imulq 0xa0(%rsp), %rax addq %rax, 0x28(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d vmovdqa 0xb0(%rsp), %xmm0 jge 0x26575d movq 0x28(%rsp), %rax vmovd (%rax,%r14,4), %xmm1 vzeroupper callq 0x5f170 movq 0x60(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x26572c incq %r13 jmp 0x265573 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x266967 cmpl $0x1, 0x38(%rsp) jne 0x266967 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x26695f vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x2657e1 movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x2657fa vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2657fa movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x265800 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x26588c vmaxps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x26584e vmaxps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x265872 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2658cf vmaxss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2658b9 incq %r14 movq 0x18(%rsp), %r10 jmp 0x2657bc movq 0x20(%rsp), %rax movl 0x2c(%rax), %ecx movl %ecx, 0x80(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x40(%rsp) movl 0x34(%rax), %ecx movq %rcx, 0xf8(%rsp) movl 0x38(%rax), %ecx movq %rcx, 0xf0(%rsp) movl 0x18(%rax), %r12d movl 0x28(%rax), %eax cmpl $0x2, %eax movq 0x18(%rsp), %rbx movl %r12d, 0xdc(%rsp) jne 0x2698b9 cmpl $0x1, 0x38(%rsp) jne 0x2698b9 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %r13d, %r13d movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x2698b1 vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0xb0(%rsp) jne 0x265991 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x2659b3 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x2659b3 movq %r13, %rax shlq $0x5, %rax vmovups (%rbx,%rax), %ymm0 jmp 0x2659b9 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, %r12 movq %rcx, 0xa0(%rsp) imulq %rcx, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0x28(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, 0xa8(%rsp) movq %rcx, 0xc8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x60(%rsp) addq %rcx, %r13 vmovaps %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x265abf vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x265a32 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x265a7e movq 0xa8(%rsp), %r13 movq 0x138(%rsp), %rax imulq %r13, %rax imulq 0xc8(%rsp), %rax addq %rax, 0x60(%rsp) movq 0xd0(%rsp), %rax imulq %r13, %rax imulq 0xa0(%rsp), %rax addq %rax, 0x28(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d vmovdqa 0xb0(%rsp), %xmm0 jge 0x265b41 movq 0x28(%rsp), %rax vmovd (%rax,%r14,4), %xmm1 vzeroupper callq 0x5f0e0 movq 0x60(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x265b10 incq %r13 jmp 0x265957 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x266ed2 cmpl $0x1, 0x38(%rsp) jne 0x266ed2 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x266eca vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x265bc5 movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x265bde vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x265bde movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x265be4 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x265c70 vmulps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x265c32 vmulps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x265c56 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x265cb3 vmulss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x265c9d incq %r14 movq 0x18(%rsp), %r10 jmp 0x265ba0 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x26743d cmpl $0x1, 0x38(%rsp) jne 0x26743d movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) vbroadcastss 0x188f68(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x188f5f(%rip), %xmm1 # 0x3eec88 vmovss 0x188f57(%rip), %xmm2 # 0x3eec88 cmpq 0x40(%rsp), %r14 je 0x267435 vmovss (%r10,%r14,4), %xmm3 cmpl $0x4, %edi jne 0x265d56 movq %r14, %rsi shlq $0x4, %rsi vmovups (%r10,%rsi), %xmm4 jmp 0x265d6f vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x265d6f movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm5 jmp 0x265d75 vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r15 imulq %r15, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %r12d, %r12d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x265df5 vmulps (%r13), %ymm5, %ymm6 vmovups %ymm6, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r12 jmp 0x265dd1 vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%rsi), %edx cmpl %r8d, %edx jge 0x265e27 vmulps (%r13), %xmm4, %xmm5 vmovups %xmm5, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r12 jmp 0x265e03 vdivss %xmm3, %xmm2, %xmm3 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r15 imulq 0xb0(%rsp), %r15 addq %r15, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r12d jge 0x265e66 vmulss (%r11,%r12,4), %xmm3, %xmm4 vmovss %xmm4, (%rbx,%r12,4) incq %r12 jmp 0x265e50 incq %r14 movq 0x18(%rsp), %r10 jmp 0x265d31 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x267a96 cmpl $0x1, 0x38(%rsp) jne 0x267a96 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x267a8e vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x265eef movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x265f08 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x265f08 movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x265f0e vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x265f9a vsubps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x265f5c vsubps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x265f80 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x265fdd vsubss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x265fc7 incq %r14 movq 0x18(%rsp), %r10 jmp 0x265eca movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x26816b cmpl $0x1, 0x38(%rsp) jne 0x26816b movl %edi, %r8d imull %eax, %r8d xorl %r15d, %r15d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r15 je 0x268163 vmovd (%r10,%r15,4), %xmm0 cmpl $0x4, %edi jne 0x266066 movq %r15, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x26607f vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26607f movq %r15, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x266085 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r15, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r15, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r14d, %r14d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x266119 vmovups (%r13), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r14 jmp 0x2660d3 vmovups (%r13), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r14 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x2660fb imulq %r15, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r15, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r14d jge 0x266160 vmovss (%r11,%r14,4), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%rbx,%r14,4) incq %r14 jmp 0x266146 incq %r15 movq 0x18(%rsp), %r10 jmp 0x266041 movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x268b07 cmpl $0x1, 0x38(%rsp) jne 0x268b07 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x268aff vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x2661e9 movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x266202 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266202 movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x266208 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x266294 vminps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x266256 vminps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x26627a imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2662d7 vminss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2662c1 incq %r14 movq 0x18(%rsp), %r10 jmp 0x2661c4 movq 0x20(%rsp), %rax movl 0x2c(%rax), %ecx movl %ecx, 0x80(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x40(%rsp) movl 0x34(%rax), %ecx movq %rcx, 0xf8(%rsp) movl 0x38(%rax), %ecx movq %rcx, 0xf0(%rsp) movl 0x18(%rax), %r12d movl 0x28(%rax), %eax cmpl $0x2, %eax movq 0x18(%rsp), %rbx movl %r12d, 0xdc(%rsp) jne 0x269b05 cmpl $0x1, 0x38(%rsp) jne 0x269b05 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %r13d, %r13d movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x269afd vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0x60(%rsp) jne 0x266396 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x2663b8 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x2663b8 movq %r13, %rax shlq $0x5, %rax vmovdqu (%rbx,%rax), %ymm0 jmp 0x2663be vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, %r12 movq %rcx, 0xa0(%rsp) imulq %rcx, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0xb0(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, 0xa8(%rsp) movq %rcx, 0xc8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x28(%rsp) addq %rcx, %r13 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x2664c7 vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x26643a vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x266486 movq 0xa8(%rsp), %r13 movq 0x138(%rsp), %rax imulq %r13, %rax imulq 0xc8(%rsp), %rax addq %rax, 0x28(%rsp) movq 0xd0(%rsp), %rax imulq %r13, %rax imulq 0xa0(%rsp), %rax addq %rax, 0xb0(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d jge 0x26654c movq 0xb0(%rsp), %rax vmovd (%rax,%r14,4), %xmm0 vmovdqa 0x60(%rsp), %xmm1 vzeroupper callq 0x5f0e0 movq 0x28(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x26651b incq %r13 jmp 0x26635f movq 0x20(%rsp), %rdx movl 0x2c(%rdx), %eax movl 0x30(%rdx), %ecx movl 0x34(%rdx), %r13d movl 0x38(%rdx), %esi movq %rsi, 0x60(%rsp) movl 0x18(%rdx), %edi movl 0x28(%rdx), %esi cmpl $0x2, %esi movq 0x18(%rsp), %r10 movq %r13, 0x80(%rsp) jne 0x269072 cmpl $0x1, 0x38(%rsp) jne 0x269072 movl %edi, %r8d imull %eax, %r8d xorl %r14d, %r14d testl %ecx, %ecx movl $0x0, %edx cmovgl %ecx, %edx movq %rdx, 0x40(%rsp) cmpq 0x40(%rsp), %r14 je 0x26906a vmovd (%r10,%r14,4), %xmm0 cmpl $0x4, %edi jne 0x2665d0 movq %r14, %rsi shlq $0x4, %rsi vmovdqu (%r10,%rsi), %xmm1 jmp 0x2665e9 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2665e9 movq %r14, %rsi shlq $0x5, %rsi vmovups (%r10,%rsi), %ymm2 jmp 0x2665ef vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r13 movq %rsi, 0xb0(%rsp) imulq %rsi, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %r12 imulq %r12, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movslq 0x2c(%rdx), %rsi movq %r14, %r9 movq %rsi, 0x28(%rsp) imulq %rsi, %r9 movq (%rdx), %rbx movq 0x10(%rdx), %r10 imulq %r10, %r9 addq %rbx, %r9 xorl %r15d, %r15d xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x26669f vmovups (%r13), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%r9) addq $0x20, %r13 addq $0x20, %r9 addl $0x8, %esi addq $0x8, %r15 jmp 0x26663d vmovups (%r13), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%r9) addq $0x10, %r13 addq $0x10, %r9 addl $0x4, %esi addq $0x4, %r15 leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x266673 imulq %r14, %r10 imulq 0x28(%rsp), %r10 addq %r10, %rbx imulq %r14, %r12 imulq 0xb0(%rsp), %r12 addq %r12, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2666e2 vdivss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2666cc incq %r14 movq 0x18(%rsp), %r10 jmp 0x2665ab movq 0x20(%rsp), %rax movl 0x2c(%rax), %ecx movl %ecx, 0x80(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x40(%rsp) movl 0x34(%rax), %ecx movq %rcx, 0xf8(%rsp) movl 0x38(%rax), %ecx movq %rcx, 0xf0(%rsp) movl 0x18(%rax), %r12d movl 0x28(%rax), %eax cmpl $0x2, %eax movq 0x18(%rsp), %rbx movl %r12d, 0xdc(%rsp) jne 0x269d54 cmpl $0x1, 0x38(%rsp) jne 0x269d54 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %r13d, %r13d movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x269d4c vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0x60(%rsp) jne 0x2667a1 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x2667c3 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x2667c3 movq %r13, %rax shlq $0x5, %rax vmovdqu (%rbx,%rax), %ymm0 jmp 0x2667c9 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, %r12 movq %rcx, 0xa0(%rsp) imulq %rcx, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0xb0(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movslq 0x2c(%rax), %rcx movq %r13, 0xa8(%rsp) movq %rcx, 0xc8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x28(%rsp) addq %rcx, %r13 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x2668d2 vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x266845 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x266891 movq 0xa8(%rsp), %r13 movq 0x138(%rsp), %rax imulq %r13, %rax imulq 0xc8(%rsp), %rax addq %rax, 0x28(%rsp) movq 0xd0(%rsp), %rax imulq %r13, %rax imulq 0xa0(%rsp), %rax addq %rax, 0xb0(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d jge 0x266957 movq 0xb0(%rsp), %rax vmovd (%rax,%r14,4), %xmm0 vmovdqa 0x60(%rsp), %xmm1 vzeroupper callq 0x5f170 movq 0x28(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x266926 incq %r13 jmp 0x26676a movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x266ad1 cmpl $0x1, 0x38(%rsp) jne 0x266ad1 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x266ac9 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x2669c9 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x2669e2 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2669e2 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x2669e8 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x266a77 vmaxps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x266a37 vmaxps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x266a5d imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x266abc vmaxss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x266aa6 incq %r9 movq 0x18(%rsp), %r10 jmp 0x2669a4 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x266c17 cmpl $0x2, 0x38(%rsp) jne 0x266c17 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x266c0f movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x266bff vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x266b7c movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x266b95 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266b95 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x266b9b vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x266bd3 vmaxps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x266b9d vmaxps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x266bbd jmp 0x266bf2 vmaxss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x266bdd incq %r13 jmp 0x266b59 incq %r9 movq 0x80(%rsp), %r13 jmp 0x266b17 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x266d62 cmpl $0x2, 0x38(%rsp) jne 0x266d62 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x266d5a movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x266d4a vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x266cc7 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x266ce0 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266ce0 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x266ce6 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x266d1e vmaxps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x266ce8 vmaxps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x266d08 jmp 0x266d3d vmaxss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x266d28 incq %r13 jmp 0x266ca4 incq %r9 movq 0x80(%rsp), %r13 jmp 0x266c62 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x266ec2 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x266eba vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x266e3d movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x266e56 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266e56 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x266e5c vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x266e91 vmaxps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x266e5e vmaxps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x266e7c jmp 0x266eae vmaxss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x266e9a incq %r12 jmp 0x266e1a incq %r14 jmp 0x266e04 incq %r8 jmp 0x266dbb movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x26703c cmpl $0x1, 0x38(%rsp) jne 0x26703c movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x267034 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x266f34 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x266f4d vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x266f4d movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x266f53 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x266fe2 vmulps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x266fa2 vmulps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x266fc8 imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x267027 vmulss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x267011 incq %r9 movq 0x18(%rsp), %r10 jmp 0x266f0f movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x267182 cmpl $0x2, 0x38(%rsp) jne 0x267182 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x26717a movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x26716a vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x2670e7 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x267100 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267100 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267106 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x26713e vmulps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x267108 vmulps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x267128 jmp 0x26715d vmulss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x267148 incq %r13 jmp 0x2670c4 incq %r9 movq 0x80(%rsp), %r13 jmp 0x267082 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x2672cd cmpl $0x2, 0x38(%rsp) jne 0x2672cd movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x2672c5 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x2672b5 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x267232 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x26724b vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26724b movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267251 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x267289 vmulps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x267253 vmulps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x267273 jmp 0x2672a8 vmulss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x267293 incq %r13 jmp 0x26720f incq %r9 movq 0x80(%rsp), %r13 jmp 0x2671cd movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x26742d movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x267425 vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x2673a8 movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2673c1 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2673c1 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2673c7 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x2673fc vmulps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x2673c9 vmulps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x2673e7 jmp 0x267419 vmulss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x267405 incq %r12 jmp 0x267385 incq %r14 jmp 0x26736f incq %r8 jmp 0x267326 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x2675e1 cmpl $0x1, 0x38(%rsp) jne 0x2675e1 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) vbroadcastss 0x187805(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x1877fc(%rip), %xmm1 # 0x3eec88 vmovss 0x1877f4(%rip), %xmm2 # 0x3eec88 cmpq 0x40(%rsp), %r9 je 0x2675d9 vmovss (%r10,%r9,4), %xmm3 cmpl $0x4, %edi jne 0x2674b9 movq %r9, %rdx shlq $0x4, %rdx vmovups (%r10,%rdx), %xmm4 jmp 0x2674d2 vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x2674d2 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm5 jmp 0x2674d8 vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 movq %r12, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %r15d, %r15d xorl %r10d, %r10d leal 0x7(%r10), %edx cmpl %r8d, %edx jge 0x26755a vmulps (%r13), %ymm5, %ymm6 vmovups %ymm6, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r10d addq $0x8, %r15 jmp 0x267535 vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%r10), %edx cmpl %r8d, %edx jge 0x26758d vmulps (%r13), %xmm4, %xmm5 vmovups %xmm5, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r10d addq $0x4, %r15 jmp 0x267568 vdivss %xmm3, %xmm2, %xmm3 imulq 0x28(%rsp), %r12 imulq %r9, %r12 addq %r12, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2675cc vmulss (%r11,%r15,4), %xmm3, %xmm4 vmovss %xmm4, (%rbx,%r15,4) incq %r15 jmp 0x2675b6 incq %r9 movq 0x18(%rsp), %r10 jmp 0x267494 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x267763 cmpl $0x2, 0x38(%rsp) jne 0x267763 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx vbroadcastss 0x187658(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x18764f(%rip), %xmm1 # 0x3eec88 vmovss 0x187647(%rip), %xmm2 # 0x3eec88 cmpq %r11, %r9 je 0x26775b movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x26774b vmovss (%r15,%r13,4), %xmm3 cmpl $0x4, %edi jne 0x2676a6 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm4 jmp 0x2676bf vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x2676bf movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm5 jmp 0x2676c5 vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x2676f5 vmulps (%r14), %ymm5, %ymm6 vmovups %ymm6, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x2676d5 vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%rsi), %edx cmpl %r8d, %edx jge 0x267723 vmulps (%r14), %xmm4, %xmm5 vmovups %xmm5, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi jmp 0x267703 vdivss %xmm3, %xmm2, %xmm3 cmpl %r8d, %esi jge 0x267743 vmulss (%r14), %xmm3, %xmm4 vmovss %xmm4, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi jmp 0x267727 incq %r13 jmp 0x267683 incq %r9 movq 0x80(%rsp), %r13 jmp 0x267641 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x2678ea cmpl $0x2, 0x38(%rsp) jne 0x2678ea movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx vbroadcastss 0x1874d1(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x1874c8(%rip), %xmm1 # 0x3eec88 vmovss 0x1874c0(%rip), %xmm2 # 0x3eec88 cmpq %r11, %r9 je 0x2678e2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x2678d2 vmovss (%r15,%r13,4), %xmm3 cmpl $0x4, %edi jne 0x26782d movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm4 jmp 0x267846 vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x267846 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm5 jmp 0x26784c vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x26787c vmulps (%r14), %ymm5, %ymm6 vmovups %ymm6, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x26785c vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%rsi), %edx cmpl %r8d, %edx jge 0x2678aa vmulps (%r14), %xmm4, %xmm5 vmovups %xmm5, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi jmp 0x26788a vdivss %xmm3, %xmm2, %xmm3 cmpl %r8d, %esi jge 0x2678ca vmulss (%r14), %xmm3, %xmm4 vmovss %xmm4, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi jmp 0x2678ae incq %r13 jmp 0x26780a incq %r9 movq 0x80(%rsp), %r13 jmp 0x2677c8 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 vbroadcastss 0x18733c(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x187333(%rip), %xmm1 # 0x3eec88 vmovss 0x18732b(%rip), %xmm2 # 0x3eec88 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x267a86 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x267a7e vmovss (%r15,%r12,4), %xmm3 cmpl $0x4, %edi jne 0x2679df movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm4 jmp 0x2679f8 vshufps $0x0, %xmm3, %xmm3, %xmm4 # xmm4 = xmm3[0,0,0,0] cmpl $0x8, %edi jne 0x2679f8 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm5 jmp 0x2679fe vinsertf128 $0x1, %xmm4, %ymm4, %ymm5 vrcpps %ymm5, %ymm6 vfmsub213ps %ymm0, %ymm6, %ymm5 # ymm5 = (ymm6 * ymm5) - ymm0 vfnmadd132ps %ymm6, %ymm6, %ymm5 # ymm5 = -(ymm5 * ymm6) + ymm6 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x267a2c vmulps (%r10), %ymm5, %ymm6 vmovups %ymm6, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x267a0e vrcpps %xmm4, %xmm5 vfmsub213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm1 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 leal 0x3(%rsi), %edx cmpl %eax, %edx jge 0x267a58 vmulps (%r10), %xmm4, %xmm5 vmovups %xmm5, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi jmp 0x267a3a vdivss %xmm3, %xmm2, %xmm3 cmpl %eax, %esi jge 0x267a76 vmulss (%r10), %xmm3, %xmm4 vmovss %xmm4, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi jmp 0x267a5c incq %r12 jmp 0x2679bc incq %r14 jmp 0x2679a6 incq %r8 jmp 0x26795d movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x267c00 cmpl $0x1, 0x38(%rsp) jne 0x267c00 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x267bf8 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x267af8 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x267b11 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267b11 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x267b17 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x267ba6 vsubps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x267b66 vsubps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x267b8c imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x267beb vsubss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x267bd5 incq %r9 movq 0x18(%rsp), %r10 jmp 0x267ad3 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x267d46 cmpl $0x2, 0x38(%rsp) jne 0x267d46 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x267d3e movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x267d2e vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x267cab movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x267cc4 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267cc4 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267cca vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x267d02 vsubps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x267ccc vsubps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x267cec jmp 0x267d21 vsubss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x267d0c incq %r13 jmp 0x267c88 incq %r9 movq 0x80(%rsp), %r13 jmp 0x267c46 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x267e91 cmpl $0x2, 0x38(%rsp) jne 0x267e91 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x267e89 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x267e79 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x267df6 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x267e0f vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267e0f movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267e15 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x267e4d vsubps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x267e17 vsubps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x267e37 jmp 0x267e6c vsubss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x267e57 incq %r13 jmp 0x267dd3 incq %r9 movq 0x80(%rsp), %r13 jmp 0x267d91 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x267ff1 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x267fe9 vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x267f6c movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x267f85 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x267f85 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x267f8b vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x267fc0 vsubps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x267f8d vsubps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x267fab jmp 0x267fdd vsubss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x267fc9 incq %r12 jmp 0x267f49 incq %r14 jmp 0x267f33 incq %r8 jmp 0x267eea movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x2682e1 cmpl $0x1, 0x38(%rsp) jne 0x2682e1 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x2682d9 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x268063 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x26807c vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26807c movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x268082 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x268111 vaddps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x2680d1 vaddps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x2680f7 imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x268156 vaddss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x268140 incq %r9 movq 0x18(%rsp), %r10 jmp 0x26803e movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x268427 cmpl $0x1, 0x38(%rsp) jne 0x268427 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x26841f vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x2681cd movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x2681e6 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2681e6 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x2681ec vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x268283 vmovups (%r13), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x26823b vmovups (%r13), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x268265 imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2682cc vmovss (%r11,%r15,4), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2682b2 incq %r9 movq 0x18(%rsp), %r10 jmp 0x2681a8 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x268579 cmpl $0x2, 0x38(%rsp) jne 0x268579 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268571 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x26840f vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x26838c movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2683a5 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2683a5 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2683ab vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x2683e3 vaddps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x2683ad vaddps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x2683cd jmp 0x268402 vaddss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x2683ed incq %r13 jmp 0x268369 incq %r9 movq 0x80(%rsp), %r13 jmp 0x268327 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x2686c4 cmpl $0x2, 0x38(%rsp) jne 0x2686c4 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x2686bc movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x268561 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x2684d2 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2684eb vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2684eb movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2684f1 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x268531 vmovups (%r14), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x2684f3 vmovups (%r14), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x268517 jmp 0x268554 vmovss (%r14), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x26853b incq %r13 jmp 0x2684af incq %r9 movq 0x80(%rsp), %r13 jmp 0x26846d movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26881b cmpl $0x2, 0x38(%rsp) jne 0x26881b movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268813 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x2686ac vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x268629 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268642 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268642 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268648 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x268680 vaddps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x26864a vaddps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x26866a jmp 0x26869f vaddss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x26868a incq %r13 jmp 0x268606 incq %r9 movq 0x80(%rsp), %r13 jmp 0x2685c4 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26898b cmpl $0x2, 0x38(%rsp) jne 0x26898b movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268983 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x268803 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x268774 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x26878d vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26878d movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268793 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x2687d3 vmovups (%r14), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x268795 vmovups (%r14), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x2687b9 jmp 0x2687f6 vmovss (%r14), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x2687dd incq %r13 jmp 0x268751 incq %r9 movq 0x80(%rsp), %r13 jmp 0x26870f movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x26897b movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x268973 vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x2688f6 movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x26890f vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x26890f movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268915 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x26894a vaddps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x268917 vaddps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x268935 jmp 0x268967 vaddss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x268953 incq %r12 jmp 0x2688d3 incq %r14 jmp 0x2688bd incq %r8 jmp 0x268874 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x268af7 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x268aef vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x268a66 movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268a7f vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268a7f movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268a85 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x268ac2 vmovups (%r10), %ymm3 vsubps %ymm2, %ymm3, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x268a87 vmovups (%r10), %xmm2 vsubps %xmm1, %xmm2, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x268aa9 jmp 0x268ae3 vmovss (%r10), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x268acb incq %r12 jmp 0x268a43 incq %r14 jmp 0x268a2d incq %r8 jmp 0x2689e4 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x268c71 cmpl $0x1, 0x38(%rsp) jne 0x268c71 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x268c69 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x268b69 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x268b82 vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268b82 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x268b88 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x268c17 vminps (%r13), %ymm2, %ymm3 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x268bd7 vminps (%r13), %xmm1, %xmm2 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x268bfd imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x268c5c vminss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x268c46 incq %r9 movq 0x18(%rsp), %r10 jmp 0x268b44 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x268db7 cmpl $0x2, 0x38(%rsp) jne 0x268db7 movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268daf movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x268d9f vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x268d1c movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268d35 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268d35 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268d3b vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x268d73 vminps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x268d3d vminps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x268d5d jmp 0x268d92 vminss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x268d7d incq %r13 jmp 0x268cf9 incq %r9 movq 0x80(%rsp), %r13 jmp 0x268cb7 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x268f02 cmpl $0x2, 0x38(%rsp) jne 0x268f02 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x268efa movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x268eea vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x268e67 movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268e80 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268e80 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268e86 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x268ebe vminps (%r14), %ymm2, %ymm3 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x268e88 vminps (%r14), %xmm1, %xmm2 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x268ea8 jmp 0x268edd vminss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x268ec8 incq %r13 jmp 0x268e44 incq %r9 movq 0x80(%rsp), %r13 jmp 0x268e02 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x269062 movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x26905a vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x268fdd movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x268ff6 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x268ff6 movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x268ffc vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x269031 vminps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x268ffe vminps (%r10), %xmm1, %xmm2 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x26901c jmp 0x26904e vminss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x26903a incq %r12 jmp 0x268fba incq %r14 jmp 0x268fa4 incq %r8 jmp 0x268f5b movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi leal -0x3(%rsi), %edx cmpl $0x1, %edx ja 0x269200 cmpl $0x1, 0x38(%rsp) jne 0x269200 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d imull %r13d, %r8d xorl %r9d, %r9d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi movq %rsi, 0x40(%rsp) cmpq 0x40(%rsp), %r9 je 0x2691f8 vmovd (%r10,%r9,4), %xmm0 cmpl $0x4, %edi jne 0x2690d4 movq %r9, %rdx shlq $0x4, %rdx vmovdqu (%r10,%rdx), %xmm1 jmp 0x2690ed vpshufd $0x0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2690ed movq %r9, %rdx shlq $0x5, %rdx vmovups (%r10,%rdx), %ymm2 jmp 0x2690f3 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 movq %r14, %r13 imulq %r9, %r13 movq (%rdx), %r11 movq 0x10(%rdx), %rdx movq %rdx, 0xb0(%rsp) imulq %rdx, %r13 addq %r11, %r13 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r10 movq %r10, %rsi imulq %r9, %rsi movq (%rdx), %rbx movq 0x10(%rdx), %rdx movq %rdx, 0x28(%rsp) imulq %rdx, %rsi addq %rbx, %rsi xorl %r15d, %r15d xorl %r12d, %r12d leal 0x7(%r12), %edx cmpl %r8d, %edx jge 0x2691a6 vmovups (%r13), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%rsi) addq $0x20, %r13 addq $0x20, %rsi addl $0x8, %r12d addq $0x8, %r15 jmp 0x269142 vmovups (%r13), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%rsi) addq $0x10, %r13 addq $0x10, %rsi addl $0x4, %r12d addq $0x4, %r15 leal 0x3(%r12), %edx cmpl %r8d, %edx jl 0x26917a imulq 0x28(%rsp), %r10 imulq %r9, %r10 addq %r10, %rbx imulq 0xb0(%rsp), %r14 imulq %r9, %r14 addq %r14, %r11 movq 0x80(%rsp), %r13 cmpl %r8d, %r15d jge 0x2691eb vdivss (%r11,%r15,4), %xmm0, %xmm1 vmovss %xmm1, (%rbx,%r15,4) incq %r15 jmp 0x2691d5 incq %r9 movq 0x18(%rsp), %r10 jmp 0x2690af movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x3, %esi jne 0x26936a cmpl $0x2, 0x38(%rsp) jne 0x26936a movl %edi, %r8d imull %eax, %r8d xorl %r9d, %r9d testl %ecx, %ecx movl $0x0, %r10d cmovgl %ecx, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x269362 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x269352 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x2692ab movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2692c4 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2692c4 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2692ca vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x269326 vmovups (%r14), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x2692cc vmovups (%r14), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x2692fe jmp 0x269345 vdivss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x269330 incq %r13 jmp 0x269288 incq %r9 movq 0x80(%rsp), %r13 jmp 0x269246 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x2694d9 cmpl $0x2, 0x38(%rsp) jne 0x2694d9 movl %edi, %r8d imull %eax, %r8d imull %ecx, %r8d xorl %r9d, %r9d testl %r13d, %r13d movl $0x0, %r10d cmovgl %r13d, %r10d movq 0x60(%rsp), %rdx testl %edx, %edx movl $0x0, %r11d cmovgl %edx, %r11d movslq 0x34(%rsp), %rbx imulq 0x78(%rsp), %rbx cmpq %r11, %r9 je 0x2694d1 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r14 imulq %r9, %r14 imulq 0x10(%rdx), %r14 addq (%rdx), %r14 movq %rbx, %r15 imulq %r9, %r15 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r12 imulq %r9, %r12 imulq 0x10(%rdx), %r12 addq 0x18(%rsp), %r15 addq (%rdx), %r12 xorl %r13d, %r13d cmpq %r10, %r13 je 0x2694c1 vmovss (%r15,%r13,4), %xmm0 cmpl $0x4, %edi jne 0x26941a movq %r13, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x269433 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x269433 movq %r13, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x269439 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %r8d, %edx jge 0x269495 vmovups (%r14), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%r12) addq $0x20, %r14 addq $0x20, %r12 addl $0x8, %esi jmp 0x26943b vmovups (%r14), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%r12) addq $0x10, %r14 addq $0x10, %r12 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %r8d, %edx jl 0x26946d jmp 0x2694b4 vdivss (%r14), %xmm0, %xmm1 vmovss %xmm1, (%r12) addq $0x4, %r14 addq $0x4, %r12 incl %esi cmpl %r8d, %esi jl 0x26949f incq %r13 jmp 0x2693f7 incq %r9 movq 0x80(%rsp), %r13 jmp 0x2693b5 movq 0x20(%rsp), %rdx movl 0x28(%rdx), %esi cmpl $0x4, %esi jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 imull %edi, %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %r13d, %r13d cmovlel %r8d, %r13d movq 0x60(%rsp), %rdx testl %edx, %edx cmovlel %r8d, %edx movq %rdx, 0x60(%rsp) movq 0x98(%rsp), %rdx movq 0x78(%rsp), %rsi imulq %rsi, %rdx movq %rdx, 0x98(%rsp) movslq 0x34(%rsp), %r9 imulq %rsi, %r9 cmpq 0x60(%rsp), %r8 je 0x26bab3 movq 0x20(%rsp), %rdx movq 0x40(%rdx), %r10 imulq %r8, %r10 imulq 0x10(%rdx), %r10 addq (%rdx), %r10 movq 0x48(%rsp), %rdx movq 0x40(%rdx), %r11 imulq %r8, %r11 imulq 0x10(%rdx), %r11 addq (%rdx), %r11 movq 0x98(%rsp), %rbx imulq %r8, %rbx addq 0x18(%rsp), %rbx xorl %r14d, %r14d cmpq %r13, %r14 je 0x26965d movq %r9, %r15 imulq %r14, %r15 addq %rbx, %r15 xorl %r12d, %r12d cmpq %rcx, %r12 je 0x269655 vmovss (%r15,%r12,4), %xmm0 cmpl $0x4, %edi jne 0x2695b4 movq %r12, %rdx shlq $0x4, %rdx vmovups (%r15,%rdx), %xmm1 jmp 0x2695cd vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] cmpl $0x8, %edi jne 0x2695cd movq %r12, %rdx shlq $0x5, %rdx vmovups (%r15,%rdx), %ymm2 jmp 0x2695d3 vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %esi, %esi leal 0x7(%rsi), %edx cmpl %eax, %edx jge 0x26962c vmovups (%r10), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm2, %ymm5 vfmsub213ps %ymm2, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm3) - ymm2 vfnmadd213ps %ymm5, %ymm4, %ymm3 # ymm3 = -(ymm4 * ymm3) + ymm5 vmovups %ymm3, (%r11) addq $0x20, %r10 addq $0x20, %r11 addl $0x8, %esi jmp 0x2695d5 vmovups (%r10), %xmm2 vrcpps %xmm2, %xmm3 vmulps %xmm3, %xmm1, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm2) - xmm1 vfnmadd213ps %xmm4, %xmm3, %xmm2 # xmm2 = -(xmm3 * xmm2) + xmm4 vmovups %xmm2, (%r11) addq $0x10, %r10 addq $0x10, %r11 addl $0x4, %esi leal 0x3(%rsi), %edx cmpl %eax, %edx jl 0x269605 jmp 0x269649 vdivss (%r10), %xmm0, %xmm1 vmovss %xmm1, (%r11) addq $0x4, %r10 addq $0x4, %r11 incl %esi cmpl %eax, %esi jl 0x269635 incq %r12 jmp 0x269591 incq %r14 jmp 0x26957b incq %r8 jmp 0x269532 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax leal -0x3(%rax), %ecx cmpl $0x1, %ecx ja 0x269fa3 cmpl $0x1, 0x38(%rsp) jne 0x269fa3 movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d imull 0xf8(%rsp), %r15d xorl %r13d, %r13d movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x269f9b vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0xb0(%rsp) jne 0x2696f6 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x269718 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x269718 movq %r13, %rax shlq $0x5, %rax vmovups (%rbx,%rax), %ymm0 jmp 0x26971e vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movq 0x40(%rax), %r12 movq %r12, 0xa0(%rsp) imulq %r13, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0x28(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movq 0x40(%rax), %rdx movq %r13, %rcx movq %rdx, 0xc8(%rsp) movq %rdx, %r13 movq %rcx, 0xa8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x60(%rsp) addq %rcx, %r13 vmovaps %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x269827 vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x26979a vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x2697e6 movq 0xc8(%rsp), %rax imulq 0x138(%rsp), %rax movq 0xa8(%rsp), %r13 imulq %r13, %rax addq %rax, 0x60(%rsp) movq 0xa0(%rsp), %rax imulq 0xd0(%rsp), %rax imulq %r13, %rax addq %rax, 0x28(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d vmovdqa 0xb0(%rsp), %xmm0 jge 0x2698a9 movq 0x28(%rsp), %rax vmovd (%rax,%r14,4), %xmm1 vzeroupper callq 0x5f170 movq 0x60(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x269878 incq %r13 jmp 0x2696bc movq 0x20(%rsp), %rax movl 0x28(%rax), %eax leal -0x3(%rax), %ecx cmpl $0x1, %ecx ja 0x26a1bc cmpl $0x1, 0x38(%rsp) jne 0x26a1bc movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d imull 0xf8(%rsp), %r15d xorl %r13d, %r13d movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x26a1b4 vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0xb0(%rsp) jne 0x269942 movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x269964 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x269964 movq %r13, %rax shlq $0x5, %rax vmovups (%rbx,%rax), %ymm0 jmp 0x26996a vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movq 0x40(%rax), %r12 movq %r12, 0xa0(%rsp) imulq %r13, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0x28(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movq 0x40(%rax), %rdx movq %r13, %rcx movq %rdx, 0xc8(%rsp) movq %rdx, %r13 movq %rcx, 0xa8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x60(%rsp) addq %rcx, %r13 vmovaps %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x269a73 vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x2699e6 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x269a32 movq 0xc8(%rsp), %rax imulq 0x138(%rsp), %rax movq 0xa8(%rsp), %r13 imulq %r13, %rax addq %rax, 0x60(%rsp) movq 0xa0(%rsp), %rax imulq 0xd0(%rsp), %rax imulq %r13, %rax addq %rax, 0x28(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d vmovdqa 0xb0(%rsp), %xmm0 jge 0x269af5 movq 0x28(%rsp), %rax vmovd (%rax,%r14,4), %xmm1 vzeroupper callq 0x5f0e0 movq 0x60(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x269ac4 incq %r13 jmp 0x269908 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax leal -0x3(%rax), %ecx cmpl $0x1, %ecx ja 0x26a3d5 cmpl $0x1, 0x38(%rsp) jne 0x26a3d5 movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d imull 0xf8(%rsp), %r15d xorl %r13d, %r13d movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x26a3cd vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0x60(%rsp) jne 0x269b8b movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x269bad vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x269bad movq %r13, %rax shlq $0x5, %rax vmovdqu (%rbx,%rax), %ymm0 jmp 0x269bb3 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movq 0x40(%rax), %r12 movq %r12, 0xa0(%rsp) imulq %r13, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0xb0(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movq 0x40(%rax), %rdx movq %r13, %rcx movq %rdx, 0xc8(%rsp) movq %rdx, %r13 movq %rcx, 0xa8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x28(%rsp) addq %rcx, %r13 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x269cbf vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x269c32 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x269c7e movq 0xc8(%rsp), %rax imulq 0x138(%rsp), %rax movq 0xa8(%rsp), %r13 imulq %r13, %rax addq %rax, 0x28(%rsp) movq 0xa0(%rsp), %rax imulq 0xd0(%rsp), %rax imulq %r13, %rax addq %rax, 0xb0(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d jge 0x269d44 movq 0xb0(%rsp), %rax vmovd (%rax,%r14,4), %xmm0 vmovdqa 0x60(%rsp), %xmm1 vzeroupper callq 0x5f0e0 movq 0x28(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x269d13 incq %r13 jmp 0x269b54 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax leal -0x3(%rax), %ecx cmpl $0x1, %ecx ja 0x26a5ee cmpl $0x1, 0x38(%rsp) jne 0x26a5ee movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d imull 0xf8(%rsp), %r15d xorl %r13d, %r13d movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x130(%rsp) cmpq 0x130(%rsp), %r13 je 0x26a5e6 vmovss (%rbx,%r13,4), %xmm0 cmpl $0x4, %r12d vmovaps %xmm0, 0x60(%rsp) jne 0x269dda movq %r13, %rax shlq $0x4, %rax vmovups (%rbx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x269dfc vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x269dfc movq %r13, %rax shlq $0x5, %rax vmovdqu (%rbx,%rax), %ymm0 jmp 0x269e02 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 movq 0x20(%rsp), %rax movq 0x40(%rax), %r12 movq %r12, 0xa0(%rsp) imulq %r13, %r12 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0xd0(%rsp) imulq %rax, %r12 movq %rcx, 0xb0(%rsp) addq %rcx, %r12 movq 0x48(%rsp), %rax movq 0x40(%rax), %rdx movq %r13, %rcx movq %rdx, 0xc8(%rsp) movq %rdx, %r13 movq %rcx, 0xa8(%rsp) imulq %rcx, %r13 movq (%rax), %rcx movq 0x10(%rax), %rax movq %rax, 0x138(%rsp) imulq %rax, %r13 movq %rcx, 0x28(%rsp) addq %rcx, %r13 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d xorl %ebx, %ebx leal 0x7(%rbx), %eax cmpl %r15d, %eax jge 0x269f0e vmovdqu (%r12), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%r13) addq $0x20, %r12 addq $0x20, %r13 addl $0x8, %ebx addq $0x8, %r14 jmp 0x269e81 vmovdqu (%r12), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%r13) addq $0x10, %r12 addq $0x10, %r13 addl $0x4, %ebx addq $0x4, %r14 leal 0x3(%rbx), %eax cmpl %r15d, %eax jl 0x269ecd movq 0xc8(%rsp), %rax imulq 0x138(%rsp), %rax movq 0xa8(%rsp), %r13 imulq %r13, %rax addq %rax, 0x28(%rsp) movq 0xa0(%rsp), %rax imulq 0xd0(%rsp), %rax imulq %r13, %rax addq %rax, 0xb0(%rsp) movq 0x18(%rsp), %rbx movl 0xdc(%rsp), %r12d cmpl %r15d, %r14d jge 0x269f93 movq 0xb0(%rsp), %rax vmovd (%rax,%r14,4), %xmm0 vmovdqa 0x60(%rsp), %xmm1 vzeroupper callq 0x5f170 movq 0x28(%rsp), %rax vmovd %xmm0, (%rax,%r14,4) incq %r14 jmp 0x269f62 incq %r13 jmp 0x269da3 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x3, %eax jne 0x26a807 cmpl $0x2, 0x38(%rsp) jne 0x26a807 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %ecx, %ecx movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26a7ff movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26a19f vmovd (%rsi,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26a0a0 movq %rcx, %rax shlq $0x4, %rax vmovdqu (%rsi,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26a0c2 vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a0c2 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a0c8 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a154 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a0d4 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a118 jmp 0x26a185 vmovd (%r13), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f170 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a15f movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a065 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a006 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x3, %eax jne 0x26aa29 cmpl $0x2, 0x38(%rsp) jne 0x26aa29 movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %ecx, %ecx movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26aa21 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26a3b8 vmovd (%rsi,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26a2b9 movq %rcx, %rax shlq $0x4, %rax vmovdqu (%rsi,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26a2db vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a2db movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a2e1 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a36d vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a2ed vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a331 jmp 0x26a39e vmovd (%r13), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a378 movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a27e movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a21f movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x3, %eax jne 0x26ac4b cmpl $0x2, 0x38(%rsp) jne 0x26ac4b movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %ecx, %ecx movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26ac43 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26a5d1 vmovss (%rsi,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26a4d2 movq %rcx, %rax shlq $0x4, %rax vmovups (%rsi,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26a4f4 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a4f4 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a4fa vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a586 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a506 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a54a jmp 0x26a5b7 vmovd (%r13), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a591 movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a497 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a438 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x3, %eax jne 0x26ae6d cmpl $0x2, 0x38(%rsp) jne 0x26ae6d movl %r12d, %r15d imull 0x80(%rsp), %r15d xorl %ecx, %ecx movq 0x40(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26ae65 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26a7ea vmovss (%rsi,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26a6eb movq %rcx, %rax shlq $0x4, %rax vmovups (%rsi,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26a70d vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a70d movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a713 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a79f vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a71f vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a763 jmp 0x26a7d0 vmovd (%r13), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f170 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a7aa movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a6b0 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a651 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26b08f cmpl $0x2, 0x38(%rsp) jne 0x26b08f movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d xorl %ecx, %ecx movq 0xf8(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26b087 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26aa0c vmovd (%rsi,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26a90d movq %rcx, %rax shlq $0x4, %rax vmovdqu (%rsi,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26a92f vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26a92f movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26a935 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26a9c1 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26a941 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26a985 jmp 0x26a9f2 vmovd (%r13), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f170 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26a9cc movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26a8d2 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26a873 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26b31a cmpl $0x2, 0x38(%rsp) jne 0x26b31a movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d xorl %ecx, %ecx movq 0xf8(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26b312 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26ac2e vmovd (%rsi,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26ab2f movq %rcx, %rax shlq $0x4, %rax vmovdqu (%rsi,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26ab51 vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26ab51 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26ab57 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26abe3 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26ab63 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26aba7 jmp 0x26ac14 vmovd (%r13), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26abee movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26aaf4 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26aa95 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26b5a5 cmpl $0x2, 0x38(%rsp) jne 0x26b5a5 movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d xorl %ecx, %ecx movq 0xf8(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26b59d movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26ae50 vmovss (%rsi,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26ad51 movq %rcx, %rax shlq $0x4, %rax vmovups (%rsi,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26ad73 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26ad73 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26ad79 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26ae05 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26ad85 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26adc9 jmp 0x26ae36 vmovd (%r13), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26ae10 movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26ad16 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26acb7 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26b830 cmpl $0x2, 0x38(%rsp) jne 0x26b830 movl %r12d, %r15d imull 0x80(%rsp), %r15d imull 0x40(%rsp), %r15d xorl %ecx, %ecx movq 0xf8(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0x60(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movq %rdx, 0xd0(%rsp) movslq 0x34(%rsp), %rax imulq 0x78(%rsp), %rax movq %rax, 0xc8(%rsp) cmpq 0xd0(%rsp), %rcx je 0x26b828 movq 0x20(%rsp), %rax movq 0x40(%rax), %r13 imulq %rcx, %r13 imulq 0x10(%rax), %r13 addq (%rax), %r13 movq 0xc8(%rsp), %rsi imulq %rcx, %rsi movq %rbx, %rax movq %rcx, %rdx movq 0x48(%rsp), %rcx movq 0x40(%rcx), %rbx movq %rdx, 0xa0(%rsp) imulq %rdx, %rbx imulq 0x10(%rcx), %rbx addq %rax, %rsi addq (%rcx), %rbx xorl %ecx, %ecx movq %rsi, 0xa8(%rsp) cmpq 0x60(%rsp), %rcx je 0x26b072 vmovss (%rsi,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26af73 movq %rcx, %rax shlq $0x4, %rax vmovups (%rsi,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26af95 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26af95 movq %rcx, %rax shlq $0x5, %rax vmovdqu (%rsi,%rax), %ymm0 jmp 0x26af9b vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r14d, %r14d leal 0x7(%r14), %eax cmpl %r15d, %eax jge 0x26b027 vmovdqu (%r13), %ymm0 vmovdqa %ymm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%rbx) addq $0x20, %r13 addq $0x20, %rbx addl $0x8, %r14d jmp 0x26afa7 vmovdqu (%r13), %xmm0 vmovdqa %xmm0, 0x100(%rsp) leaq 0x17(%rsp), %rdi leaq 0x100(%rsp), %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%rbx) addq $0x10, %r13 addq $0x10, %rbx addl $0x4, %r14d leal 0x3(%r14), %eax cmpl %r15d, %eax jl 0x26afeb jmp 0x26b058 vmovd (%r13), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f170 vmovd %xmm0, (%rbx) addq $0x4, %r13 addq $0x4, %rbx incl %r14d cmpl %r15d, %r14d jl 0x26b032 movq 0x28(%rsp), %rcx incq %rcx movq 0xa8(%rsp), %rsi jmp 0x26af38 movq 0xa0(%rsp), %rcx incq %rcx movq 0x18(%rsp), %rbx jmp 0x26aed9 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 movl 0x80(%rsp), %eax imull %r12d, %eax movl %eax, 0x80(%rsp) xorl %edx, %edx movq 0x40(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x40(%rsp) movq 0xf8(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf8(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf0(%rsp) movq 0x98(%rsp), %rax movq 0x78(%rsp), %rcx imulq %rcx, %rax movq %rax, 0x98(%rsp) movslq 0x34(%rsp), %rax imulq %rcx, %rax movq %rax, 0x78(%rsp) leaq 0x17(%rsp), %r13 leaq 0x100(%rsp), %r14 cmpq 0xf0(%rsp), %rdx je 0x26bab3 movq 0x20(%rsp), %rax movq 0x40(%rax), %rbx imulq %rdx, %rbx imulq 0x10(%rax), %rbx addq (%rax), %rbx movq 0x48(%rsp), %rax movq 0x40(%rax), %r15 imulq %rdx, %r15 imulq 0x10(%rax), %r15 addq (%rax), %r15 movq 0x98(%rsp), %rax movq %rdx, 0xa0(%rsp) imulq %rdx, %rax addq 0x18(%rsp), %rax movq %rax, 0x38(%rsp) xorl %eax, %eax cmpq 0xf8(%rsp), %rax je 0x26b302 movq 0x78(%rsp), %rcx imulq %rax, %rcx addq 0x38(%rsp), %rcx movq %rcx, 0x60(%rsp) xorl %ecx, %ecx movq %rax, 0xa8(%rsp) cmpq 0x40(%rsp), %rcx je 0x26b2fa movq 0x60(%rsp), %rax vmovd (%rax,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26b1ee movq 0x28(%rsp), %rax shlq $0x4, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26b217 vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26b217 movq 0x28(%rsp), %rax shlq $0x5, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %ymm0 jmp 0x26b21d vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r12d, %r12d leal 0x7(%r12), %eax cmpl 0x80(%rsp), %eax jge 0x26b29e vmovdqu (%rbx), %ymm0 vmovdqa %ymm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0x140(%rsp), %rdx callq 0x26d916 vmovdqu %ymm0, (%r15) addq $0x20, %rbx addq $0x20, %r15 addl $0x8, %r12d jmp 0x26b229 vmovdqu (%rbx), %xmm0 vmovdqa %xmm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26daa0 vmovdqu %xmm0, (%r15) addq $0x10, %rbx addq $0x10, %r15 addl $0x4, %r12d leal 0x3(%r12), %eax cmpl 0x80(%rsp), %eax jl 0x26b26a jmp 0x26b2d3 vmovd (%rbx), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f170 vmovd %xmm0, (%r15) addq $0x4, %rbx addq $0x4, %r15 incl %r12d cmpl 0x80(%rsp), %r12d jl 0x26b2ae movq 0x28(%rsp), %rcx incq %rcx movl 0xdc(%rsp), %r12d movq 0xa8(%rsp), %rax jmp 0x26b1a7 incq %rax jmp 0x26b17c movq 0xa0(%rsp), %rdx incq %rdx jmp 0x26b124 movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 movl 0x80(%rsp), %eax imull %r12d, %eax movl %eax, 0x80(%rsp) xorl %edx, %edx movq 0x40(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x40(%rsp) movq 0xf8(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf8(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf0(%rsp) movq 0x98(%rsp), %rax movq 0x78(%rsp), %rcx imulq %rcx, %rax movq %rax, 0x98(%rsp) movslq 0x34(%rsp), %rax imulq %rcx, %rax movq %rax, 0x78(%rsp) leaq 0x17(%rsp), %r13 leaq 0x100(%rsp), %r14 cmpq 0xf0(%rsp), %rdx je 0x26bab3 movq 0x20(%rsp), %rax movq 0x40(%rax), %rbx imulq %rdx, %rbx imulq 0x10(%rax), %rbx addq (%rax), %rbx movq 0x48(%rsp), %rax movq 0x40(%rax), %r15 imulq %rdx, %r15 imulq 0x10(%rax), %r15 addq (%rax), %r15 movq 0x98(%rsp), %rax movq %rdx, 0xa0(%rsp) imulq %rdx, %rax addq 0x18(%rsp), %rax movq %rax, 0x38(%rsp) xorl %eax, %eax cmpq 0xf8(%rsp), %rax je 0x26b58d movq 0x78(%rsp), %rcx imulq %rax, %rcx addq 0x38(%rsp), %rcx movq %rcx, 0x60(%rsp) xorl %ecx, %ecx movq %rax, 0xa8(%rsp) cmpq 0x40(%rsp), %rcx je 0x26b585 movq 0x60(%rsp), %rax vmovd (%rax,%rcx,4), %xmm1 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovdqa %xmm1, 0xb0(%rsp) jne 0x26b479 movq 0x28(%rsp), %rax shlq $0x4, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %xmm0 vmovdqa %xmm0, 0xe0(%rsp) jmp 0x26b4a2 vpshufd $0x0, %xmm1, %xmm0 # xmm0 = xmm1[0,0,0,0] vmovdqa %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26b4a2 movq 0x28(%rsp), %rax shlq $0x5, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %ymm0 jmp 0x26b4a8 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r12d, %r12d leal 0x7(%r12), %eax cmpl 0x80(%rsp), %eax jge 0x26b529 vmovdqu (%rbx), %ymm0 vmovdqa %ymm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0x140(%rsp), %rdx callq 0x26d220 vmovdqu %ymm0, (%r15) addq $0x20, %rbx addq $0x20, %r15 addl $0x8, %r12d jmp 0x26b4b4 vmovdqu (%rbx), %xmm0 vmovdqa %xmm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d430 vmovdqu %xmm0, (%r15) addq $0x10, %rbx addq $0x10, %r15 addl $0x4, %r12d leal 0x3(%r12), %eax cmpl 0x80(%rsp), %eax jl 0x26b4f5 jmp 0x26b55e vmovd (%rbx), %xmm1 vmovdqa 0xb0(%rsp), %xmm0 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%r15) addq $0x4, %rbx addq $0x4, %r15 incl %r12d cmpl 0x80(%rsp), %r12d jl 0x26b539 movq 0x28(%rsp), %rcx incq %rcx movl 0xdc(%rsp), %r12d movq 0xa8(%rsp), %rax jmp 0x26b432 incq %rax jmp 0x26b407 movq 0xa0(%rsp), %rdx incq %rdx jmp 0x26b3af movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 movl 0x80(%rsp), %eax imull %r12d, %eax movl %eax, 0x80(%rsp) xorl %edx, %edx movq 0x40(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x40(%rsp) movq 0xf8(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf8(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf0(%rsp) movq 0x98(%rsp), %rax movq 0x78(%rsp), %rcx imulq %rcx, %rax movq %rax, 0x98(%rsp) movslq 0x34(%rsp), %rax imulq %rcx, %rax movq %rax, 0x78(%rsp) leaq 0x17(%rsp), %r13 leaq 0x100(%rsp), %r14 cmpq 0xf0(%rsp), %rdx je 0x26bab3 movq 0x20(%rsp), %rax movq 0x40(%rax), %rbx imulq %rdx, %rbx imulq 0x10(%rax), %rbx addq (%rax), %rbx movq 0x48(%rsp), %rax movq 0x40(%rax), %r15 imulq %rdx, %r15 imulq 0x10(%rax), %r15 addq (%rax), %r15 movq 0x98(%rsp), %rax movq %rdx, 0xa0(%rsp) imulq %rdx, %rax addq 0x18(%rsp), %rax movq %rax, 0x38(%rsp) xorl %eax, %eax cmpq 0xf8(%rsp), %rax je 0x26b818 movq 0x78(%rsp), %rcx imulq %rax, %rcx addq 0x38(%rsp), %rcx movq %rcx, 0x60(%rsp) xorl %ecx, %ecx movq %rax, 0xa8(%rsp) cmpq 0x40(%rsp), %rcx je 0x26b810 movq 0x60(%rsp), %rax vmovss (%rax,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26b704 movq 0x28(%rsp), %rax shlq $0x4, %rax movq 0x60(%rsp), %rcx vmovups (%rcx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26b72d vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26b72d movq 0x28(%rsp), %rax shlq $0x5, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %ymm0 jmp 0x26b733 vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r12d, %r12d leal 0x7(%r12), %eax cmpl 0x80(%rsp), %eax jge 0x26b7b4 vmovdqu (%rbx), %ymm0 vmovdqa %ymm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0x140(%rsp), %rdx callq 0x26ce36 vmovdqu %ymm0, (%r15) addq $0x20, %rbx addq $0x20, %r15 addl $0x8, %r12d jmp 0x26b73f vmovdqu (%rbx), %xmm0 vmovdqa %xmm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d046 vmovdqu %xmm0, (%r15) addq $0x10, %rbx addq $0x10, %r15 addl $0x4, %r12d leal 0x3(%r12), %eax cmpl 0x80(%rsp), %eax jl 0x26b780 jmp 0x26b7e9 vmovd (%rbx), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f0e0 vmovd %xmm0, (%r15) addq $0x4, %rbx addq $0x4, %r15 incl %r12d cmpl 0x80(%rsp), %r12d jl 0x26b7c4 movq 0x28(%rsp), %rcx incq %rcx movl 0xdc(%rsp), %r12d movq 0xa8(%rsp), %rax jmp 0x26b6bd incq %rax jmp 0x26b692 movq 0xa0(%rsp), %rdx incq %rdx jmp 0x26b63a movq 0x20(%rsp), %rax movl 0x28(%rax), %eax cmpl $0x4, %eax jne 0x26bab3 cmpl $0x3, 0x38(%rsp) jne 0x26bab3 movl 0x80(%rsp), %eax imull %r12d, %eax movl %eax, 0x80(%rsp) xorl %edx, %edx movq 0x40(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x40(%rsp) movq 0xf8(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf8(%rsp) movq 0xf0(%rsp), %rax testl %eax, %eax cmovlel %edx, %eax movq %rax, 0xf0(%rsp) movq 0x98(%rsp), %rax movq 0x78(%rsp), %rcx imulq %rcx, %rax movq %rax, 0x98(%rsp) movslq 0x34(%rsp), %rax imulq %rcx, %rax movq %rax, 0x78(%rsp) leaq 0x17(%rsp), %r13 leaq 0x100(%rsp), %r14 cmpq 0xf0(%rsp), %rdx je 0x26bab3 movq 0x20(%rsp), %rax movq 0x40(%rax), %rbx imulq %rdx, %rbx imulq 0x10(%rax), %rbx addq (%rax), %rbx movq 0x48(%rsp), %rax movq 0x40(%rax), %r15 imulq %rdx, %r15 imulq 0x10(%rax), %r15 addq (%rax), %r15 movq 0x98(%rsp), %rax movq %rdx, 0xa0(%rsp) imulq %rdx, %rax addq 0x18(%rsp), %rax movq %rax, 0x38(%rsp) xorl %eax, %eax cmpq 0xf8(%rsp), %rax je 0x26baa3 movq 0x78(%rsp), %rcx imulq %rax, %rcx addq 0x38(%rsp), %rcx movq %rcx, 0x60(%rsp) xorl %ecx, %ecx movq %rax, 0xa8(%rsp) cmpq 0x40(%rsp), %rcx je 0x26ba9b movq 0x60(%rsp), %rax vmovss (%rax,%rcx,4), %xmm0 cmpl $0x4, %r12d movq %rcx, 0x28(%rsp) vmovaps %xmm0, 0xb0(%rsp) jne 0x26b98f movq 0x28(%rsp), %rax shlq $0x4, %rax movq 0x60(%rsp), %rcx vmovups (%rcx,%rax), %xmm0 vmovaps %xmm0, 0xe0(%rsp) jmp 0x26b9b8 vshufps $0x0, %xmm0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] vmovaps %xmm0, 0xe0(%rsp) cmpl $0x8, %r12d jne 0x26b9b8 movq 0x28(%rsp), %rax shlq $0x5, %rax movq 0x60(%rsp), %rcx vmovdqu (%rcx,%rax), %ymm0 jmp 0x26b9be vinsertf128 $0x1, %xmm0, %ymm0, %ymm0 vmovdqa %ymm0, 0x140(%rsp) xorl %r12d, %r12d leal 0x7(%r12), %eax cmpl 0x80(%rsp), %eax jge 0x26ba3f vmovdqu (%rbx), %ymm0 vmovdqa %ymm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0x140(%rsp), %rdx callq 0x26d60a vmovdqu %ymm0, (%r15) addq $0x20, %rbx addq $0x20, %r15 addl $0x8, %r12d jmp 0x26b9ca vmovdqu (%rbx), %xmm0 vmovdqa %xmm0, 0x100(%rsp) movq %r13, %rdi movq %r14, %rsi leaq 0xe0(%rsp), %rdx vzeroupper callq 0x26d794 vmovdqu %xmm0, (%r15) addq $0x10, %rbx addq $0x10, %r15 addl $0x4, %r12d leal 0x3(%r12), %eax cmpl 0x80(%rsp), %eax jl 0x26ba0b jmp 0x26ba74 vmovd (%rbx), %xmm0 vmovdqa 0xb0(%rsp), %xmm1 vzeroupper callq 0x5f170 vmovd %xmm0, (%r15) addq $0x4, %rbx addq $0x4, %r15 incl %r12d cmpl 0x80(%rsp), %r12d jl 0x26ba4f movq 0x28(%rsp), %rcx incq %rcx movl 0xdc(%rsp), %r12d movq 0xa8(%rsp), %rax jmp 0x26b948 incq %rax jmp 0x26b91d movq 0xa0(%rsp), %rdx incq %rdx jmp 0x26b8c5 movq 0x58(%rsp), %rax testq %rax, %rax je 0x26bae9 lock decl (%rax) jne 0x26bae9 movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x26badc movq (%rdi), %rax movq 0x18(%rsp), %rsi vzeroupper callq *0x18(%rax) jmp 0x26bae9 movq 0x18(%rsp), %rdi vzeroupper callq 0x5f3e0 leaq -0x28(%rbp), %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq *0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq *0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq *0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bcbe jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq *0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bcbe jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi je 0x26bc9e movq (%rdi), %rax callq *0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bcbe jmp 0x26bd56 movq %rax, %r15 movq 0x148(%rsp), %rax testq %rax, %rax je 0x26bcc1 lock decl (%rax) jne 0x26bcc1 movq 0x140(%rsp), %rsi movq 0x160(%rsp), %rdi testq %rdi, %rdi jne 0x26bca8 movq %rsi, %rdi callq 0x5f3e0 jmp 0x26bcc1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x26bcc1 jmp 0x26bd56 jmp 0x26bcbe jmp 0x26bd56 jmp 0x26bcbe movq %rax, %r15 movq %rbx, %rsi jmp 0x26bd1c jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 jmp 0x26bd14 movq %rax, %r15 movq 0x18(%rsp), %rsi movq 0x50(%rsp), %rdi movq 0x58(%rsp), %rax testq %rax, %rax je 0x26bd4b lock decl (%rax) jne 0x26bd4b testq %rdi, %rdi jne 0x26bd42 movq %rsi, %rdi vzeroupper callq 0x5f3e0 jmp 0x26bd4b movq (%rdi), %rax vzeroupper callq *0x18(%rax) movq %r15, %rdi vzeroupper callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp

virtual thunk to ncnn::BinaryOp_x86_fma::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int BinaryOp_x86_fma::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const bool b_is_scalar = bottom_blobs[1].w * bottom_blobs[1].h * bottom_blobs[1].d * bottom_blobs[1].c * bottom_blobs[1].elempack == 1; const bool a_rank_is_lower = bottom_blobs[0].dims < bottom_blobs[1].dims && !b_is_scalar; const bool a_size_is_lower = bottom_blobs[0].w * bottom_blobs[0].h * bottom_blobs[0].d * bottom_blobs[0].c * bottom_blobs[0].elempack < bottom_blobs[1].w * bottom_blobs[1].h * bottom_blobs[1].d * bottom_blobs[1].c * bottom_blobs[1].elempack; const bool a_is_lower = a_rank_is_lower || (!a_rank_is_lower && a_size_is_lower); const Mat& A = a_is_lower ? bottom_blobs[1] : bottom_blobs[0]; const Mat& B = a_is_lower ? bottom_blobs[0] : bottom_blobs[1]; const int op_type_r = a_is_lower ? get_reverse_op_type(op_type) : op_type; Mat& top_blob = top_blobs[0]; top_blob.create_like(A, opt.blob_allocator); if (top_blob.empty()) return -100; // B is a scalar if (B.w * B.h * B.d * B.c * B.elempack == 1) { return binary_op_scalar(A, B[0], top_blob, op_type_r, opt); } // no broadcast if (A.dims == B.dims && A.w == B.w && A.h == B.h && A.d == B.d && A.c == B.c && A.elempack == B.elempack) { return binary_op_no_broadcast(A, B, top_blob, op_type_r, opt); } // broadcast B for inner axis if ((B.dims < A.dims) || (A.dims == 2 && B.w == 1 && B.h == A.h) || (A.dims == 3 && B.w == 1 && B.h == 1 && B.c == A.c) || (A.dims == 3 && B.w == 1 && B.h == A.h && B.c == A.c) || (A.dims == 4 && B.w == 1 && B.h == 1 && B.d == 1 && B.c == A.c) || (A.dims == 4 && B.w == 1 && B.h == 1 && B.d == A.d && B.c == A.c) || (A.dims == 4 && B.w == 1 && B.h == A.h && B.d == A.d && B.c == A.c)) { return binary_op_broadcast_inner(A, B, top_blob, op_type_r, opt); } // broadcast B for outer axis if (B.elempack == 1 && ((A.dims == 2 && B.w == A.w && B.h == 1) || (A.dims == 3 && B.w == A.w && B.h == 1 && B.c == 1) || (A.dims == 3 && B.w == A.w && B.h == A.h && B.c == 1) || (A.dims == 4 && B.w == A.w && B.h == 1 && B.d == 1 && B.c == 1) || (A.dims == 4 && B.w == A.w && B.h == A.h && B.d == 1 && B.c == 1) || (A.dims == 4 && B.w == A.w && B.h == A.h && B.d == A.d && B.c == 1))) { return binary_op_broadcast_outer(A, B, top_blob, op_type_r, opt); } // some special broadcast rule here if (A.dims == 3 && B.dims == 3 && A.w == B.w && B.h == 1 && A.c == B.c) { return binary_op_broadcast_20(A, B, top_blob, op_type_r, opt); } return 0; }

movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x25fe94

/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp

ncnn::BinaryOp_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int BinaryOp_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { using namespace BinaryOp_x86_fma_functor; if (op_type == Operation_ADD) return binary_op_scalar_inplace<binary_op_add>(bottom_top_blob, b, opt); if (op_type == Operation_SUB) return binary_op_scalar_inplace<binary_op_sub>(bottom_top_blob, b, opt); if (op_type == Operation_MUL) return binary_op_scalar_inplace<binary_op_mul>(bottom_top_blob, b, opt); if (op_type == Operation_DIV) return binary_op_scalar_inplace<binary_op_div>(bottom_top_blob, b, opt); if (op_type == Operation_MAX) return binary_op_scalar_inplace<binary_op_max>(bottom_top_blob, b, opt); if (op_type == Operation_MIN) return binary_op_scalar_inplace<binary_op_min>(bottom_top_blob, b, opt); if (op_type == Operation_POW) return binary_op_scalar_inplace<binary_op_pow>(bottom_top_blob, b, opt); if (op_type == Operation_RSUB) return binary_op_scalar_inplace<binary_op_rsub>(bottom_top_blob, b, opt); if (op_type == Operation_RDIV) return binary_op_scalar_inplace<binary_op_rdiv>(bottom_top_blob, b, opt); if (op_type == Operation_RPOW) return binary_op_scalar_inplace<binary_op_rpow>(bottom_top_blob, b, opt); if (op_type == Operation_ATAN2) return binary_op_scalar_inplace<binary_op_atan2>(bottom_top_blob, b, opt); if (op_type == Operation_RATAN2) return binary_op_scalar_inplace<binary_op_ratan2>(bottom_top_blob, b, opt); return 0; }

pushq %rbp movq %rsp, %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx andq $-0x20, %rsp subq $0xe0, %rsp movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %eax cmpq $0xb, %rax ja 0x26c8c4 movq %rsi, %r14 leaq 0x18b6f8(%rip), %rcx # 0x3f749c movslq (%rcx,%rax,4), %rax addq %rcx, %rax movq %rsi, 0x38(%rsp) jmpq *%rax vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26be37 vaddps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26be01 vaddps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26be21 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26be5d vaddss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26be4a incq %rdx jmp 0x26bdda vbroadcastss 0xd8(%rdi,%rdx), %ymm0 vmovaps %ymm0, 0x40(%rsp) movl 0x30(%r14), %ebx movl 0x38(%r14), %eax imull 0x2c(%r14), %ebx imull 0x34(%r14), %ebx imull 0x18(%r14), %ebx xorl %r15d, %r15d testl %eax, %eax cmovlel %r15d, %eax movq %rax, 0x18(%rsp) cmpq 0x18(%rsp), %r15 je 0x26c8c4 movq (%r14), %r12 movq 0x10(%r14), %rax movq 0x40(%r14), %r14 movq %r14, 0x28(%rsp) movq %r15, 0x30(%rsp) imulq %r15, %r14 movq %rax, 0x20(%rsp) imulq %rax, %r14 addq %r12, %r14 vmovaps 0x40(%rsp), %ymm0 vmovaps %ymm0, 0xa0(%rsp) xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %ebx, %eax jge 0x26bf1c vmovups (%r14), %ymm0 vmovaps %ymm0, 0x60(%rsp) leaq 0x17(%rsp), %rdi leaq 0x60(%rsp), %rsi leaq 0xa0(%rsp), %rdx callq 0x26d916 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x26bedf vmovaps 0x40(%rsp), %ymm0 vmovaps %xmm0, 0x60(%rsp) leal 0x3(%r15), %eax cmpl %ebx, %eax jge 0x26bf6b vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x90(%rsp) leaq 0x17(%rsp), %rdi leaq 0x90(%rsp), %rsi leaq 0x60(%rsp), %rdx vzeroupper callq 0x26daa0 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 jmp 0x26bf28 movq 0x28(%rsp), %rax imulq 0x20(%rsp), %rax movq 0x30(%rsp), %r15 imulq %r15, %rax addq %rax, %r12 movq 0x38(%rsp), %r14 cmpl %ebx, %r13d jge 0x26bfab vmovss (%r12,%r13,4), %xmm1 vmovaps 0x40(%rsp), %ymm0 vzeroupper callq 0x5f170 vmovss %xmm0, (%r12,%r13,4) incq %r13 jmp 0x26bf87 incq %r15 jmp 0x26be9a vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c038 vmaxps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c002 vmaxps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c022 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c05e vmaxss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c04b incq %rdx jmp 0x26bfdb vbroadcastss 0xd8(%rdi,%rdx), %ymm0 vmovaps %ymm0, 0x40(%rsp) movl 0x30(%r14), %ebx movl 0x38(%r14), %eax imull 0x2c(%r14), %ebx imull 0x34(%r14), %ebx imull 0x18(%r14), %ebx xorl %r15d, %r15d testl %eax, %eax cmovlel %r15d, %eax movq %rax, 0x18(%rsp) cmpq 0x18(%rsp), %r15 je 0x26c8c4 movq (%r14), %r12 movq 0x10(%r14), %rax movq 0x40(%r14), %r14 movq %r14, 0x28(%rsp) movq %r15, 0x30(%rsp) imulq %r15, %r14 movq %rax, 0x20(%rsp) imulq %rax, %r14 addq %r12, %r14 vmovaps 0x40(%rsp), %ymm0 vmovaps %ymm0, 0xa0(%rsp) xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %ebx, %eax jge 0x26c11d vmovups (%r14), %ymm0 vmovaps %ymm0, 0x60(%rsp) leaq 0x17(%rsp), %rdi leaq 0x60(%rsp), %rsi leaq 0xa0(%rsp), %rdx callq 0x26d220 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x26c0e0 vmovaps 0x40(%rsp), %ymm0 vmovaps %xmm0, 0x60(%rsp) leal 0x3(%r15), %eax cmpl %ebx, %eax jge 0x26c16c vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x90(%rsp) leaq 0x17(%rsp), %rdi leaq 0x90(%rsp), %rsi leaq 0x60(%rsp), %rdx vzeroupper callq 0x26d430 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 jmp 0x26c129 movq 0x28(%rsp), %rax imulq 0x20(%rsp), %rax movq 0x30(%rsp), %r15 imulq %r15, %rax addq %rax, %r12 movq 0x38(%rsp), %r14 cmpl %ebx, %r13d jge 0x26c1ac vmovss (%r12,%r13,4), %xmm1 vmovaps 0x40(%rsp), %ymm0 vzeroupper callq 0x5f0e0 vmovss %xmm0, (%r12,%r13,4) incq %r13 jmp 0x26c188 incq %r15 jmp 0x26c09b vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c239 vmulps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c203 vmulps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c223 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c25f vmulss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c24c incq %rdx jmp 0x26c1dc movl 0x30(%r14), %eax imull 0x2c(%r14), %eax imull 0x34(%r14), %eax movl 0x38(%r14), %ecx imull 0x18(%r14), %eax vmovss 0x182a02(%rip), %xmm0 # 0x3eec88 vdivss 0xd8(%rdi,%rdx), %xmm0, %xmm0 vshufps $0x0, %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[0,0,0,0] vinsertf128 $0x1, %xmm1, %ymm1, %ymm2 xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c2fe vmulps (%r10), %ymm2, %ymm3 vmovups %ymm3, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c2c8 vmulps (%r10), %xmm1, %xmm3 vmovaps %xmm3, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c2e8 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c324 vmulss (%rsi,%rdi,4), %xmm0, %xmm3 vmovss %xmm3, (%rsi,%rdi,4) incq %rdi jmp 0x26c311 incq %rdx jmp 0x26c2a1 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c3b1 vsubps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c37b vsubps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c39b imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c3d7 vsubss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c3c4 incq %rdx jmp 0x26c354 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c46c vmovups (%r10), %ymm1 vsubps %ymm0, %ymm1, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c42e vmovaps (%r10), %xmm1 vsubps %xmm0, %xmm1, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c452 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c496 vmovss (%rsi,%rdi,4), %xmm1 vsubss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c47f incq %rdx jmp 0x26c407 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c523 vminps (%r10), %ymm0, %ymm1 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c4ed vminps (%r10), %xmm0, %xmm1 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c50d imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c549 vminss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c536 incq %rdx jmp 0x26c4c6 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 vmovaps %ymm0, 0x40(%rsp) movl 0x30(%r14), %ebx movl 0x38(%r14), %eax imull 0x2c(%r14), %ebx imull 0x34(%r14), %ebx imull 0x18(%r14), %ebx xorl %r15d, %r15d testl %eax, %eax cmovlel %r15d, %eax movq %rax, 0x18(%rsp) cmpq 0x18(%rsp), %r15 je 0x26c8c4 movq (%r14), %r12 movq 0x10(%r14), %rax movq 0x40(%r14), %r14 movq %r14, 0x28(%rsp) movq %r15, 0x30(%rsp) imulq %r15, %r14 movq %rax, 0x20(%rsp) imulq %rax, %r14 addq %r12, %r14 vmovaps 0x40(%rsp), %ymm0 vmovaps %ymm0, 0xa0(%rsp) xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %ebx, %eax jge 0x26c608 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x60(%rsp) leaq 0x17(%rsp), %rdi leaq 0x60(%rsp), %rsi leaq 0xa0(%rsp), %rdx callq 0x26ce36 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x26c5cb vmovaps 0x40(%rsp), %ymm0 vmovaps %xmm0, 0x60(%rsp) leal 0x3(%r15), %eax cmpl %ebx, %eax jge 0x26c657 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x90(%rsp) leaq 0x17(%rsp), %rdi leaq 0x90(%rsp), %rsi leaq 0x60(%rsp), %rdx vzeroupper callq 0x26d046 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 jmp 0x26c614 movq 0x28(%rsp), %rax imulq 0x20(%rsp), %rax movq 0x30(%rsp), %r15 imulq %r15, %rax addq %rax, %r12 movq 0x38(%rsp), %r14 cmpl %ebx, %r13d jge 0x26c697 vmovss (%r12,%r13,4), %xmm0 vmovaps 0x40(%rsp), %ymm1 vzeroupper callq 0x5f0e0 vmovss %xmm0, (%r12,%r13,4) incq %r13 jmp 0x26c673 incq %r15 jmp 0x26c586 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 vmovaps %ymm0, 0x40(%rsp) movl 0x30(%r14), %ebx movl 0x38(%r14), %eax imull 0x2c(%r14), %ebx imull 0x34(%r14), %ebx imull 0x18(%r14), %ebx xorl %r15d, %r15d testl %eax, %eax cmovlel %r15d, %eax movq %rax, 0x18(%rsp) cmpq 0x18(%rsp), %r15 je 0x26c8c4 movq (%r14), %r12 movq 0x10(%r14), %rax movq 0x40(%r14), %r14 movq %r14, 0x28(%rsp) movq %r15, 0x30(%rsp) imulq %r15, %r14 movq %rax, 0x20(%rsp) imulq %rax, %r14 addq %r12, %r14 vmovaps 0x40(%rsp), %ymm0 vmovaps %ymm0, 0xa0(%rsp) xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %ebx, %eax jge 0x26c756 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x60(%rsp) leaq 0x17(%rsp), %rdi leaq 0x60(%rsp), %rsi leaq 0xa0(%rsp), %rdx callq 0x26d60a vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x26c719 vmovaps 0x40(%rsp), %ymm0 vmovaps %xmm0, 0x60(%rsp) leal 0x3(%r15), %eax cmpl %ebx, %eax jge 0x26c7a5 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x90(%rsp) leaq 0x17(%rsp), %rdi leaq 0x90(%rsp), %rsi leaq 0x60(%rsp), %rdx vzeroupper callq 0x26d794 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 jmp 0x26c762 movq 0x28(%rsp), %rax imulq 0x20(%rsp), %rax movq 0x30(%rsp), %r15 imulq %r15, %rax addq %rax, %r12 movq 0x38(%rsp), %r14 cmpl %ebx, %r13d jge 0x26c7e5 vmovss (%r12,%r13,4), %xmm0 vmovaps 0x40(%rsp), %ymm1 vzeroupper callq 0x5f170 vmovss %xmm0, (%r12,%r13,4) incq %r13 jmp 0x26c7c1 incq %r15 jmp 0x26c6d4 vbroadcastss 0xd8(%rdi,%rdx), %ymm0 movl 0x30(%r14), %eax movl 0x38(%r14), %ecx imull 0x2c(%r14), %eax imull 0x34(%r14), %eax imull 0x18(%r14), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x26c8c4 movq (%r14), %rsi movq 0x10(%r14), %r9 movq 0x40(%r14), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x26c896 vmovups (%r10), %ymm1 vrcpps %ymm1, %ymm2 vmulps %ymm2, %ymm0, %ymm3 vfmsub213ps %ymm0, %ymm3, %ymm1 # ymm1 = (ymm3 * ymm1) - ymm0 vfnmadd213ps %ymm3, %ymm2, %ymm1 # ymm1 = -(ymm2 * ymm1) + ymm3 vmovups %ymm1, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x26c83c vmovaps (%r10), %xmm1 vrcpps %xmm1, %xmm2 vmulps %xmm2, %xmm0, %xmm3 vfmsub213ps %xmm0, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm1) - xmm0 vfnmadd213ps %xmm3, %xmm2, %xmm1 # xmm1 = -(xmm2 * xmm1) + xmm3 vmovaps %xmm1, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x26c86e imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x26c8bc vdivss (%rsi,%rdi,4), %xmm0, %xmm1 vmovss %xmm1, (%rsi,%rdi,4) incq %rdi jmp 0x26c8a9 incq %rdx jmp 0x26c815 xorl %eax, %eax leaq -0x28(%rbp), %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp

virtual thunk to ncnn::BinaryOp_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int BinaryOp_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { using namespace BinaryOp_x86_fma_functor; if (op_type == Operation_ADD) return binary_op_scalar_inplace<binary_op_add>(bottom_top_blob, b, opt); if (op_type == Operation_SUB) return binary_op_scalar_inplace<binary_op_sub>(bottom_top_blob, b, opt); if (op_type == Operation_MUL) return binary_op_scalar_inplace<binary_op_mul>(bottom_top_blob, b, opt); if (op_type == Operation_DIV) return binary_op_scalar_inplace<binary_op_div>(bottom_top_blob, b, opt); if (op_type == Operation_MAX) return binary_op_scalar_inplace<binary_op_max>(bottom_top_blob, b, opt); if (op_type == Operation_MIN) return binary_op_scalar_inplace<binary_op_min>(bottom_top_blob, b, opt); if (op_type == Operation_POW) return binary_op_scalar_inplace<binary_op_pow>(bottom_top_blob, b, opt); if (op_type == Operation_RSUB) return binary_op_scalar_inplace<binary_op_rsub>(bottom_top_blob, b, opt); if (op_type == Operation_RDIV) return binary_op_scalar_inplace<binary_op_rdiv>(bottom_top_blob, b, opt); if (op_type == Operation_RPOW) return binary_op_scalar_inplace<binary_op_rpow>(bottom_top_blob, b, opt); if (op_type == Operation_ATAN2) return binary_op_scalar_inplace<binary_op_atan2>(bottom_top_blob, b, opt); if (op_type == Operation_RATAN2) return binary_op_scalar_inplace<binary_op_ratan2>(bottom_top_blob, b, opt); return 0; }

pushq %rax movq (%rdi), %rax addq -0x58(%rax), %rdi callq 0x26bd6a xorl %eax, %eax popq %rcx retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp

int ncnn::binary_op_no_broadcast<ncnn::BinaryOp_x86_fma_functor::binary_op_sub>(ncnn::Mat const&, ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&)

static int binary_op_no_broadcast(const Mat& a, const Mat& b, Mat& c, const Option& opt) { Op op; const int channels = a.c; const int size = a.w * a.h * a.d * a.elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = a.channel(q); const float* ptr1 = b.channel(q); float* outptr = c.channel(q); int i = 0; #if __SSE2__ #if __AVX__ #if __AVX512F__ for (; i + 15 < size; i += 16) { __m512 _p = _mm512_loadu_ps(ptr); __m512 _p1 = _mm512_loadu_ps(ptr1); __m512 _outp = op.func_pack16(_p, _p1); _mm512_storeu_ps(outptr, _outp); ptr += 16; ptr1 += 16; outptr += 16; } #endif // __AVX512F__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _p1 = _mm256_loadu_ps(ptr1); __m256 _outp = op.func_pack8(_p, _p1); _mm256_storeu_ps(outptr, _outp); ptr += 8; ptr1 += 8; outptr += 8; } #endif // __AVX__ for (; i + 3 < size; i += 4) { __m128 _p = _mm_load_ps(ptr); __m128 _p1 = _mm_load_ps(ptr1); __m128 _outp = op.func_pack4(_p, _p1); _mm_store_ps(outptr, _outp); ptr += 4; ptr1 += 4; outptr += 4; } #endif // __SSE2__ for (; i < size; i++) { *outptr = op.func(*ptr, *ptr1); ptr += 1; ptr1 += 1; outptr += 1; } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %rbx movl 0x30(%rdi), %eax movl 0x38(%rdi), %ecx imull 0x2c(%rdi), %eax imull 0x34(%rdi), %eax imull 0x18(%rdi), %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx cmpq %rcx, %r8 je 0x26c9bc movq 0x10(%rsi), %r9 imulq 0x40(%rsi), %r9 movq 0x10(%rdi), %r10 imulq %r8, %r9 addq (%rsi), %r9 movq 0x10(%rdx), %r11 imulq 0x40(%rdx), %r11 imulq %r8, %r11 addq (%rdx), %r11 imulq 0x40(%rdi), %r10 imulq %r8, %r10 addq (%rdi), %r10 xorl %ebx, %ebx xorl %r14d, %r14d xorl %r15d, %r15d leal 0x7(%r15), %ebp cmpl %eax, %ebp jge 0x26c991 vmovups (%r10,%r14), %ymm0 vsubps (%r9,%r14), %ymm0, %ymm0 vmovups %ymm0, (%r11,%r14) addl $0x8, %r15d addq $0x20, %r14 addq $0x8, %rbx jmp 0x26c94b vmovaps (%r10,%r14), %xmm0 vsubps (%r9,%r14), %xmm0, %xmm0 vmovaps %xmm0, (%r11,%r14) addl $0x4, %r15d addq $0x10, %r14 addq $0x4, %rbx leal 0x3(%r15), %ebp cmpl %eax, %ebp jl 0x26c973 jmp 0x26c9b0 vmovss (%r10,%rbx,4), %xmm0 vsubss (%r9,%rbx,4), %xmm0, %xmm0 vmovss %xmm0, (%r11,%rbx,4) incq %rbx cmpl %eax, %ebx jl 0x26c99b incq %r8 jmp 0x26c90a popq %rbx popq %r14 popq %r15 popq %rbp vzeroupper retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp

int ncnn::binary_op_no_broadcast<ncnn::BinaryOp_x86_fma_functor::binary_op_div>(ncnn::Mat const&, ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&)

static int binary_op_no_broadcast(const Mat& a, const Mat& b, Mat& c, const Option& opt) { Op op; const int channels = a.c; const int size = a.w * a.h * a.d * a.elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { const float* ptr = a.channel(q); const float* ptr1 = b.channel(q); float* outptr = c.channel(q); int i = 0; #if __SSE2__ #if __AVX__ #if __AVX512F__ for (; i + 15 < size; i += 16) { __m512 _p = _mm512_loadu_ps(ptr); __m512 _p1 = _mm512_loadu_ps(ptr1); __m512 _outp = op.func_pack16(_p, _p1); _mm512_storeu_ps(outptr, _outp); ptr += 16; ptr1 += 16; outptr += 16; } #endif // __AVX512F__ for (; i + 7 < size; i += 8) { __m256 _p = _mm256_loadu_ps(ptr); __m256 _p1 = _mm256_loadu_ps(ptr1); __m256 _outp = op.func_pack8(_p, _p1); _mm256_storeu_ps(outptr, _outp); ptr += 8; ptr1 += 8; outptr += 8; } #endif // __AVX__ for (; i + 3 < size; i += 4) { __m128 _p = _mm_load_ps(ptr); __m128 _p1 = _mm_load_ps(ptr1); __m128 _outp = op.func_pack4(_p, _p1); _mm_store_ps(outptr, _outp); ptr += 4; ptr1 += 4; outptr += 4; } #endif // __SSE2__ for (; i < size; i++) { *outptr = op.func(*ptr, *ptr1); ptr += 1; ptr1 += 1; outptr += 1; } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %rbx movl 0x30(%rdi), %eax movl 0x38(%rdi), %ecx imull 0x2c(%rdi), %eax imull 0x34(%rdi), %eax imull 0x18(%rdi), %eax xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx cmpq %rcx, %r8 je 0x26cabd movq 0x10(%rsi), %r9 imulq 0x40(%rsi), %r9 movq 0x10(%rdi), %r10 imulq %r8, %r9 addq (%rsi), %r9 movq 0x10(%rdx), %r11 imulq 0x40(%rdx), %r11 imulq %r8, %r11 addq (%rdx), %r11 imulq 0x40(%rdi), %r10 imulq %r8, %r10 addq (%rdi), %r10 xorl %ebx, %ebx xorl %r14d, %r14d xorl %r15d, %r15d leal 0x7(%r15), %ebp cmpl %eax, %ebp jge 0x26ca92 vmovups (%r9,%r14), %ymm0 vrcpps %ymm0, %ymm1 vmovups (%r10,%r14), %ymm2 vmulps %ymm1, %ymm2, %ymm3 vfmsub213ps %ymm2, %ymm3, %ymm0 # ymm0 = (ymm3 * ymm0) - ymm2 vfnmadd213ps %ymm3, %ymm1, %ymm0 # ymm0 = -(ymm1 * ymm0) + ymm3 vmovups %ymm0, (%r11,%r14) addl $0x8, %r15d addq $0x20, %r14 addq $0x8, %rbx jmp 0x26ca28 vmovaps (%r10,%r14), %xmm0 vmovaps (%r9,%r14), %xmm1 vrcpps %xmm1, %xmm2 vmulps %xmm2, %xmm0, %xmm3 vfmsub213ps %xmm0, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm1) - xmm0 vfnmadd213ps %xmm3, %xmm2, %xmm1 # xmm1 = -(xmm2 * xmm1) + xmm3 vmovaps %xmm1, (%r11,%r14) addl $0x4, %r15d addq $0x10, %r14 addq $0x4, %rbx leal 0x3(%r15), %ebp cmpl %eax, %ebp jl 0x26ca62 jmp 0x26cab1 vmovss (%r10,%rbx,4), %xmm0 vdivss (%r9,%rbx,4), %xmm0, %xmm0 vmovss %xmm0, (%r11,%rbx,4) incq %rbx cmpl %eax, %ebx jl 0x26ca9c incq %r8 jmp 0x26c9e7 popq %rbx popq %r14 popq %r15 popq %rbp vzeroupper retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/binaryop_x86_fma.cpp

ncnn::UnaryOp_x86_fma::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int UnaryOp_x86_fma::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { using namespace UnaryOp_x86_fma_functor; if (op_type == Operation_ABS) return unary_op_inplace<unary_op_abs>(bottom_top_blob, opt); if (op_type == Operation_NEG) return unary_op_inplace<unary_op_neg>(bottom_top_blob, opt); if (op_type == Operation_FLOOR) return unary_op_inplace<unary_op_floor>(bottom_top_blob, opt); if (op_type == Operation_CEIL) return unary_op_inplace<unary_op_ceil>(bottom_top_blob, opt); if (op_type == Operation_SQUARE) return unary_op_inplace<unary_op_square>(bottom_top_blob, opt); if (op_type == Operation_SQRT) return unary_op_inplace<unary_op_sqrt>(bottom_top_blob, opt); if (op_type == Operation_RSQRT) return unary_op_inplace<unary_op_rsqrt>(bottom_top_blob, opt); if (op_type == Operation_EXP) return unary_op_inplace<unary_op_exp>(bottom_top_blob, opt); if (op_type == Operation_LOG) return unary_op_inplace<unary_op_log>(bottom_top_blob, opt); if (op_type == Operation_SIN) return unary_op_inplace<unary_op_sin>(bottom_top_blob, opt); if (op_type == Operation_COS) return unary_op_inplace<unary_op_cos>(bottom_top_blob, opt); if (op_type == Operation_TAN) return unary_op_inplace<unary_op_tan>(bottom_top_blob, opt); if (op_type == Operation_ASIN) return unary_op_inplace<unary_op_asin>(bottom_top_blob, opt); if (op_type == Operation_ACOS) return unary_op_inplace<unary_op_acos>(bottom_top_blob, opt); if (op_type == Operation_ATAN) return unary_op_inplace<unary_op_atan>(bottom_top_blob, opt); if (op_type == Operation_RECIPROCAL) return unary_op_inplace<unary_op_reciprocal>(bottom_top_blob, opt); if (op_type == Operation_TANH) return unary_op_inplace<unary_op_tanh>(bottom_top_blob, opt); if (op_type == Operation_LOG10) return unary_op_inplace<unary_op_log10>(bottom_top_blob, opt); return 0; }

pushq %rbp movq %rsp, %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx andq $-0x20, %rsp subq $0x80, %rsp movq %rsi, 0x18(%rsp) movq (%rdi), %rax movq -0x18(%rax), %rax movl 0xd0(%rdi,%rax), %eax cmpq $0x11, %rax ja 0x281716 leaq 0x177d68(%rip), %rcx # 0x3f8348 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq *%rax movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x1712f0(%rip), %ymm0 # 0x3f1900 vbroadcastss 0x1712e7(%rip), %xmm1 # 0x3f1900 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x28067b vandps (%r10), %ymm0, %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280645 vandps (%r10), %xmm1, %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280665 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2806a5 vmovss (%rsi,%rdi,4), %xmm2 vandps %xmm1, %xmm2, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x28068e incq %rdx jmp 0x280619 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280775 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2828e6 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x280711 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2829f4 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280747 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2807b3 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f160 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x280795 incq %r15 jmp 0x2806d6 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280883 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282726 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28081f vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x282806 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280855 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2808c1 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f540 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2808a3 incq %r15 jmp 0x2807e4 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280991 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282016 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28092d vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2821ce vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280963 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2809cf vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f4c0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2809b1 incq %r15 jmp 0x2808f2 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x280a5f vmovups (%r10), %ymm0 vmulps %ymm0, %ymm0, %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280a21 vmovaps (%r10), %xmm0 vmulps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280a45 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x280a89 vmovss (%rsi,%rdi,4), %xmm0 vmulss %xmm0, %xmm0, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x280a72 incq %rdx jmp 0x2809f5 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280b59 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282330 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x280af5 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x282422 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280b2b movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x280b97 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f240 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x280b79 incq %r15 jmp 0x280aba movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280c67 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x281b3e vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x280c03 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x281c96 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280c39 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x280ca5 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f3b0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x280c87 incq %r15 jmp 0x280bc8 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x280d2f vroundps $0x1, (%r10), %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280cf7 vroundps $0x1, (%r10), %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280d18 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x280d57 vroundss $0x9, (%rsi,%rdi,4), %xmm1, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x280d42 incq %rdx jmp 0x280ccb movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16df02(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x16def9(%rip), %xmm1 # 0x3eec88 vmovss 0x16def1(%rip), %xmm2 # 0x3eec88 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x280e15 vmovups (%r10), %ymm3 vrcpps %ymm3, %ymm4 vfmsub213ps %ymm0, %ymm4, %ymm3 # ymm3 = (ymm4 * ymm3) - ymm0 vfnmadd132ps %ymm4, %ymm4, %ymm3 # ymm3 = -(ymm3 * ymm4) + ymm4 vmovups %ymm3, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280dc3 vmovaps (%r10), %xmm3 vrcpps %xmm3, %xmm4 vfmsub213ps %xmm1, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) - xmm1 vfnmadd132ps %xmm4, %xmm4, %xmm3 # xmm3 = -(xmm3 * xmm4) + xmm4 vmovaps %xmm3, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280df1 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x280e3b vdivss (%rsi,%rdi,4), %xmm2, %xmm3 vmovss %xmm3, (%rsi,%rdi,4) incq %rdi jmp 0x280e28 incq %rdx jmp 0x280d97 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x280ec5 vroundps $0x2, (%r10), %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x280e8d vroundps $0x2, (%r10), %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x280eae imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x280eed vroundss $0xa, (%rsi,%rdi,4), %xmm1, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x280ed8 incq %rdx jmp 0x280e61 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x280fbd vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x28173c vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x280f59 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x281826 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x280f8f movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x280ffb vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f410 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x280fdd incq %r15 jmp 0x280f1e movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x17018a(%rip), %ymm0 # 0x3f11b4 vbroadcastss 0x170181(%rip), %xmm1 # 0x3f11b4 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x281095 vxorps (%r10), %ymm0, %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x28105f vxorps (%r10), %xmm1, %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x28107f imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2810bf vmovss (%rsi,%rdi,4), %xmm2 vxorps %xmm1, %xmm2, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x2810a8 incq %rdx jmp 0x281033 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x28118f vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282afa vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28112b vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x282c36 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x281161 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2811cd vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f560 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2811af incq %r15 jmp 0x2810f0 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x1745c8(%rip), %ymm0 # 0x3f57c4 vbroadcastss 0x17071f(%rip), %ymm1 # 0x3f1924 vbroadcastss 0x1706f2(%rip), %ymm2 # 0x3f1900 vbroadcastss 0x16ffc9(%rip), %ymm3 # 0x3f11e0 vbroadcastss 0x1745a4(%rip), %xmm4 # 0x3f57c4 vbroadcastss 0x1706fb(%rip), %xmm5 # 0x3f1924 vbroadcastss 0x1706ce(%rip), %xmm6 # 0x3f1900 vbroadcastss 0x16ffa5(%rip), %xmm7 # 0x3f11e0 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x2812eb vmovups (%r10), %ymm8 vrsqrtps %ymm8, %ymm9 vmulps %ymm9, %ymm8, %ymm10 vfmadd213ps %ymm0, %ymm10, %ymm9 # ymm9 = (ymm10 * ymm9) + ymm0 vmulps %ymm1, %ymm10, %ymm10 vmulps %ymm9, %ymm10, %ymm9 vandps %ymm2, %ymm8, %ymm8 vcmpleps %ymm8, %ymm3, %ymm8 vandps %ymm9, %ymm8, %ymm8 vmovups %ymm8, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x281267 vmovaps (%r10), %xmm8 vrsqrtps %xmm8, %xmm9 vmulps %xmm9, %xmm8, %xmm10 vfmadd213ps %xmm4, %xmm10, %xmm9 # xmm9 = (xmm10 * xmm9) + xmm4 vmulps %xmm5, %xmm10, %xmm10 vmulps %xmm9, %xmm10, %xmm9 vandps %xmm6, %xmm8, %xmm8 vcmpleps %xmm8, %xmm7, %xmm8 vandps %xmm9, %xmm8, %xmm8 vmovaps %xmm8, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x2812ae imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x281311 vsqrtss (%rsi,%rdi,4), %xmm11, %xmm8 vmovss %xmm8, (%rsi,%rdi,4) incq %rdi jmp 0x2812fe incq %rdx jmp 0x28123b movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x2813e1 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x282512 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28137d vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x28261c vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x2813b3 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x28141f vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f0a0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x281401 incq %r15 jmp 0x281342 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vmovss 0x174377(%rip), %xmm0 # 0x3f57c4 vmovss 0x1704cf(%rip), %xmm1 # 0x3f1924 cmpq %rcx, %rdx je 0x281716 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x2814b7 vrsqrtps (%r10), %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x281481 vrsqrtps (%r10), %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x2814a1 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2814f2 vmovss (%rsi,%rdi,4), %xmm2 vrsqrtss %xmm2, %xmm2, %xmm3 vmulss %xmm3, %xmm2, %xmm2 vfmadd213ss %xmm0, %xmm3, %xmm2 # xmm2 = (xmm3 * xmm2) + xmm0 vmulss %xmm1, %xmm3, %xmm3 vmulss %xmm2, %xmm3, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x2814ca incq %rdx jmp 0x281455 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x2815c2 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x281dac vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28155e vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x281f00 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x281594 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x281600 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f4a0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2815e2 incq %r15 jmp 0x281523 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x281716 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x2816d0 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2818f4 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28166c vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x281a22 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x2816a2 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x28170e vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f200 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2816f0 incq %r15 jmp 0x281631 xorl %eax, %eax leaq -0x28(%rbp), %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/unaryop_x86_fma.cpp

ncnn::UnaryOp_x86_fma_functor::unary_op_exp::func_pack8(float vector[8] const&) const

static NCNN_FORCEINLINE __m256 exp256_ps(__m256 x) { __m256 tmp = _mm256_setzero_ps(), fx; __m256i imm0; __m256 one = *(__m256*)_ps256_1; x = _mm256_min_ps(x, *(__m256*)_ps256_exp_hi); x = _mm256_max_ps(x, *(__m256*)_ps256_exp_lo); /* express exp(x) as exp(g + n*log(2)) */ fx = _mm256_comp_fmadd_ps(x, *(__m256*)_ps256_cephes_LOG2EF, *(__m256*)_ps256_0p5); /* how to perform a floorf with SSE: just below */ //imm0 = _mm256_cvttps_epi32(fx); //tmp = _mm256_cvtepi32_ps(imm0); tmp = _mm256_floor_ps(fx); /* if greater, subtract 1 */ //__m256 mask = _mm256_cmpgt_ps(tmp, fx); __m256 mask = _mm256_cmp_ps(tmp, fx, _CMP_GT_OS); mask = _mm256_and_ps(mask, one); fx = _mm256_sub_ps(tmp, mask); // x = x - fx * exp_C1 x = _mm256_comp_fnmadd_ps(fx, *(__m256*)_ps256_cephes_exp_C1, x); // x = x - fx * exp_C2 x = _mm256_comp_fnmadd_ps(fx, *(__m256*)_ps256_cephes_exp_C2, x); tmp = _mm256_mul_ps(x, x); __m256 y = *(__m256*)_ps256_cephes_exp_p0; y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p1); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p2); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p3); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p4); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p5); y = _mm256_comp_fmadd_ps(y, tmp, x); y = _mm256_add_ps(y, one); /* build 2^n */ imm0 = _mm256_cvttps_epi32(fx); // another two AVX2 instructions imm0 = _mm256_comp_add_epi32(imm0, *(__m256i*)_pi32_256_0x7f); imm0 = _mm256_comp_slli_epi32(imm0, 23); __m256 pow2n = _mm256_castsi256_ps(imm0); y = _mm256_mul_ps(y, pow2n); return y; }

vbroadcastss 0x16fa73(%rip), %ymm0 # 0x3f11b8 vminps (%rsi), %ymm0, %ymm0 vbroadcastss 0x16fa6a(%rip), %ymm1 # 0x3f11bc vmaxps %ymm1, %ymm0, %ymm1 vbroadcastss 0x16c8b5(%rip), %ymm0 # 0x3ee014 vbroadcastss 0x16fa58(%rip), %ymm2 # 0x3f11c0 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vroundps $0x1, %ymm2, %ymm3 vbroadcastss 0x16d50c(%rip), %ymm4 # 0x3eec88 vcmpltps %ymm3, %ymm2, %ymm2 vandps %ymm4, %ymm2, %ymm2 vbroadcastss 0x16fa36(%rip), %ymm4 # 0x3f11c4 vsubps %ymm2, %ymm3, %ymm2 vfmsub213ps %ymm1, %ymm2, %ymm4 # ymm4 = (ymm2 * ymm4) - ymm1 vbroadcastss 0x170f7c(%rip), %ymm1 # 0x3f271c vbroadcastss 0x16fa27(%rip), %ymm3 # 0x3f11d0 vbroadcastss 0x16fa1a(%rip), %ymm5 # 0x3f11cc vfmsub213ps %ymm4, %ymm2, %ymm1 # ymm1 = (ymm2 * ymm1) - ymm4 vfmadd213ps %ymm3, %ymm1, %ymm5 # ymm5 = (ymm1 * ymm5) + ymm3 vbroadcastss 0x16fa0f(%rip), %ymm3 # 0x3f11d4 vmulps %ymm1, %ymm1, %ymm4 vfmadd231ps %ymm5, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm5) + ymm3 vbroadcastss 0x16fa01(%rip), %ymm5 # 0x3f11d8 vfmadd231ps %ymm3, %ymm1, %ymm5 # ymm5 = (ymm1 * ymm3) + ymm5 vbroadcastss 0x16f9f7(%rip), %ymm3 # 0x3f11dc vfmadd231ps %ymm5, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm5) + ymm3 vfmadd213ps %ymm0, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm3) + ymm0 vfmadd213ps %ymm1, %ymm4, %ymm3 # ymm3 = (ymm4 * ymm3) + ymm1 vcvttps2dq %ymm2, %ymm0 vpslld $0x17, %xmm0, %xmm1 vextractf128 $0x1, %ymm0, %xmm0 vpslld $0x17, %xmm0, %xmm0 vbroadcastss 0x16d477(%rip), %xmm2 # 0x3eec88 vpaddd %xmm2, %xmm0, %xmm0 vpaddd %xmm2, %xmm1, %xmm1 vinsertf128 $0x1, %xmm0, %ymm1, %ymm0 vfmadd213ps %ymm0, %ymm3, %ymm0 # ymm0 = (ymm3 * ymm0) + ymm0 retq nop

/csukuangfj[P]ncnn/src/layer/x86/avx_mathfun.h

ncnn::UnaryOp_x86_avx::forward_inplace(ncnn::Mat&, ncnn::Option const&) const

int UnaryOp_x86_avx::forward_inplace(Mat& bottom_top_blob, const Option& opt) const { using namespace UnaryOp_x86_avx_functor; if (op_type == Operation_ABS) return unary_op_inplace<unary_op_abs>(bottom_top_blob, opt); if (op_type == Operation_NEG) return unary_op_inplace<unary_op_neg>(bottom_top_blob, opt); if (op_type == Operation_FLOOR) return unary_op_inplace<unary_op_floor>(bottom_top_blob, opt); if (op_type == Operation_CEIL) return unary_op_inplace<unary_op_ceil>(bottom_top_blob, opt); if (op_type == Operation_SQUARE) return unary_op_inplace<unary_op_square>(bottom_top_blob, opt); if (op_type == Operation_SQRT) return unary_op_inplace<unary_op_sqrt>(bottom_top_blob, opt); if (op_type == Operation_RSQRT) return unary_op_inplace<unary_op_rsqrt>(bottom_top_blob, opt); if (op_type == Operation_EXP) return unary_op_inplace<unary_op_exp>(bottom_top_blob, opt); if (op_type == Operation_LOG) return unary_op_inplace<unary_op_log>(bottom_top_blob, opt); if (op_type == Operation_SIN) return unary_op_inplace<unary_op_sin>(bottom_top_blob, opt); if (op_type == Operation_COS) return unary_op_inplace<unary_op_cos>(bottom_top_blob, opt); if (op_type == Operation_TAN) return unary_op_inplace<unary_op_tan>(bottom_top_blob, opt); if (op_type == Operation_ASIN) return unary_op_inplace<unary_op_asin>(bottom_top_blob, opt); if (op_type == Operation_ACOS) return unary_op_inplace<unary_op_acos>(bottom_top_blob, opt); if (op_type == Operation_ATAN) return unary_op_inplace<unary_op_atan>(bottom_top_blob, opt); if (op_type == Operation_RECIPROCAL) return unary_op_inplace<unary_op_reciprocal>(bottom_top_blob, opt); if (op_type == Operation_TANH) return unary_op_inplace<unary_op_tanh>(bottom_top_blob, opt); if (op_type == Operation_LOG10) return unary_op_inplace<unary_op_log10>(bottom_top_blob, opt); return 0; }

pushq %rbp movq %rsp, %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx andq $-0x20, %rsp subq $0x80, %rsp movq %rsi, 0x18(%rsp) movq (%rdi), %rax movq -0x18(%rax), %rax movl 0xd0(%rdi,%rax), %eax cmpq $0x11, %rax ja 0x283f35 leaq 0x1755e4(%rip), %rcx # 0x3f83c8 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq *%rax movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16eaec(%rip), %ymm0 # 0x3f1900 vbroadcastss 0x16eae3(%rip), %xmm1 # 0x3f1900 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x282e7f vandps (%r10), %ymm0, %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x282e49 vandps (%r10), %xmm1, %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x282e69 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x282ea9 vmovss (%rsi,%rdi,4), %xmm2 vandps %xmm1, %xmm2, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x282e92 incq %rdx jmp 0x282e1d movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x282f79 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2851c4 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x282f15 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2852fa vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x282f4b movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x282fb7 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f160 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x282f99 incq %r15 jmp 0x282eda movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283087 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x284fc0 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283023 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2850c2 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283059 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2830c5 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f540 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2830a7 incq %r15 jmp 0x282fe8 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283195 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x28484e vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283131 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284a08 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283167 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2831d3 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f4c0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2831b5 incq %r15 jmp 0x2830f6 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283263 vmovups (%r10), %ymm0 vmulps %ymm0, %ymm0, %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x283225 vmovaps (%r10), %xmm0 vmulps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x283249 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x28328d vmovss (%rsi,%rdi,4), %xmm0 vmulss %xmm0, %xmm0, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x283276 incq %rdx jmp 0x2831f9 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x28335d vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x284b6c vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x2832f9 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284c78 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x28332f movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x28339b vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f240 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x28337d incq %r15 jmp 0x2832be movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x28346b vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2843a2 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283407 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x2844f0 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x28343d movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2834a9 vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f3b0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x28348b incq %r15 jmp 0x2833cc movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283533 vroundps $0x1, (%r10), %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x2834fb vroundps $0x1, (%r10), %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x28351c imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x28355b vroundss $0x9, (%rsi,%rdi,4), %xmm1, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x283546 incq %rdx jmp 0x2834cf movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16b6fe(%rip), %ymm0 # 0x3eec88 vbroadcastss 0x16b6f5(%rip), %xmm1 # 0x3eec88 vmovss 0x16b6ed(%rip), %xmm2 # 0x3eec88 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283625 vmovups (%r10), %ymm3 vrcpps %ymm3, %ymm4 vmulps %ymm4, %ymm3, %ymm3 vsubps %ymm3, %ymm0, %ymm3 vmulps %ymm3, %ymm4, %ymm3 vaddps %ymm3, %ymm4, %ymm3 vmovups %ymm3, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x2835c7 vmovaps (%r10), %xmm3 vrcpps %xmm3, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vsubps %xmm3, %xmm1, %xmm3 vmulps %xmm3, %xmm4, %xmm3 vaddps %xmm3, %xmm4, %xmm3 vmovaps %xmm3, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x2835fb imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x28364b vdivss (%rsi,%rdi,4), %xmm2, %xmm3 vmovss %xmm3, (%rsi,%rdi,4) incq %rdi jmp 0x283638 incq %rdx jmp 0x28359b movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x2836d5 vroundps $0x2, (%r10), %ymm0 vmovups %ymm0, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x28369d vroundps $0x2, (%r10), %xmm0 vmovaps %xmm0, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x2836be imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2836fd vroundss $0xa, (%rsi,%rdi,4), %xmm1, %xmm0 vmovss %xmm0, (%rsi,%rdi,4) incq %rdi jmp 0x2836e8 incq %rdx jmp 0x283671 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x2837cd vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x283f5c vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283769 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284052 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x28379f movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x28380b vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f410 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2837ed incq %r15 jmp 0x28372e movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16d97a(%rip), %ymm0 # 0x3f11b4 vbroadcastss 0x16d971(%rip), %xmm1 # 0x3f11b4 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x2838a5 vxorps (%r10), %ymm0, %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x28386f vxorps (%r10), %xmm1, %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x28388f imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x2838cf vmovss (%rsi,%rdi,4), %xmm2 vxorps %xmm1, %xmm2, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x2838b8 incq %rdx jmp 0x283843 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x28399f vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x285414 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x28393b vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x285564 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283971 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x2839dd vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f560 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x2839bf incq %r15 jmp 0x283900 movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vbroadcastss 0x16df18(%rip), %ymm0 # 0x3f1924 vbroadcastss 0x171daf(%rip), %ymm1 # 0x3f57c4 vbroadcastss 0x16dee2(%rip), %ymm2 # 0x3f1900 vbroadcastss 0x16d7b9(%rip), %ymm3 # 0x3f11e0 vbroadcastss 0x16def4(%rip), %xmm4 # 0x3f1924 vbroadcastss 0x171d8b(%rip), %xmm5 # 0x3f57c4 vbroadcastss 0x16debe(%rip), %xmm6 # 0x3f1900 vbroadcastss 0x16d795(%rip), %xmm7 # 0x3f11e0 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283b07 vmovups (%r10), %ymm8 vrsqrtps %ymm8, %ymm9 vmulps %ymm9, %ymm8, %ymm10 vmulps %ymm0, %ymm10, %ymm11 vmulps %ymm9, %ymm10, %ymm9 vaddps %ymm1, %ymm9, %ymm9 vmulps %ymm9, %ymm11, %ymm9 vandps %ymm2, %ymm8, %ymm8 vcmpleps %ymm8, %ymm3, %ymm8 vandps %ymm9, %ymm8, %ymm8 vmovups %ymm8, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x283a77 vmovaps (%r10), %xmm8 vrsqrtps %xmm8, %xmm9 vmulps %xmm9, %xmm8, %xmm10 vmulps %xmm4, %xmm10, %xmm11 vmulps %xmm9, %xmm10, %xmm9 vaddps %xmm5, %xmm9, %xmm9 vmulps %xmm9, %xmm11, %xmm9 vandps %xmm6, %xmm8, %xmm8 vcmpleps %xmm8, %xmm7, %xmm8 vandps %xmm9, %xmm8, %xmm8 vmovaps %xmm8, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x283ac6 imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x283b2d vsqrtss (%rsi,%rdi,4), %xmm12, %xmm8 vmovss %xmm8, (%rsi,%rdi,4) incq %rdi jmp 0x283b1a incq %rdx jmp 0x283a4b movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283bfd vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x284d84 vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283b99 vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284ea2 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283bcf movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x283c3b vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f0a0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x283c1d incq %r15 jmp 0x283b5e movq 0x18(%rsp), %rdx movl 0x30(%rdx), %eax movl 0x38(%rdx), %ecx imull 0x2c(%rdx), %eax imull 0x34(%rdx), %eax imull 0x18(%rdx), %eax xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx vmovss 0x171b5b(%rip), %xmm0 # 0x3f57c4 vmovss 0x16dcb3(%rip), %xmm1 # 0x3f1924 cmpq %rcx, %rdx je 0x283f35 movq 0x18(%rsp), %rdi movq (%rdi), %rsi movq 0x10(%rdi), %r9 movq 0x40(%rdi), %r8 movq %r8, %r10 imulq %rdx, %r10 imulq %r9, %r10 addq %rsi, %r10 xorl %edi, %edi xorl %r11d, %r11d leal 0x7(%r11), %ebx cmpl %eax, %ebx jge 0x283cd3 vrsqrtps (%r10), %ymm2 vmovups %ymm2, (%r10) addq $0x20, %r10 addl $0x8, %r11d addq $0x8, %rdi jmp 0x283c9d vrsqrtps (%r10), %xmm2 vmovaps %xmm2, (%r10) addq $0x10, %r10 addl $0x4, %r11d addq $0x4, %rdi leal 0x3(%r11), %ebx cmpl %eax, %ebx jl 0x283cbd imulq %r9, %r8 imulq %rdx, %r8 addq %r8, %rsi cmpl %eax, %edi jge 0x283d11 vmovss (%rsi,%rdi,4), %xmm2 vrsqrtss %xmm2, %xmm2, %xmm3 vmulss %xmm3, %xmm2, %xmm2 vmulss %xmm3, %xmm2, %xmm2 vaddss %xmm0, %xmm2, %xmm2 vmulss %xmm1, %xmm3, %xmm3 vmulss %xmm2, %xmm3, %xmm2 vmovss %xmm2, (%rsi,%rdi,4) incq %rdi jmp 0x283ce6 incq %rdx jmp 0x283c71 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283de1 vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x2845fa vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283d7d vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284744 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283db3 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x283e1f vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f4a0 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x283e01 incq %r15 jmp 0x283d42 movq 0x18(%rsp), %rax movl 0x30(%rax), %r12d movl 0x38(%rax), %ecx imull 0x2c(%rax), %r12d imull 0x34(%rax), %r12d imull 0x18(%rax), %r12d xorl %r15d, %r15d testl %ecx, %ecx cmovlel %r15d, %ecx movq %rcx, 0x20(%rsp) cmpq 0x20(%rsp), %r15 je 0x283f35 movq 0x18(%rsp), %rax movq (%rax), %rbx movq 0x10(%rax), %rcx movq 0x40(%rax), %r14 movq %r14, 0x30(%rsp) movq %r15, 0x38(%rsp) imulq %r15, %r14 movq %rcx, 0x28(%rsp) imulq %rcx, %r14 addq %rbx, %r14 xorl %r13d, %r13d xorl %r15d, %r15d leal 0x7(%r15), %eax cmpl %r12d, %eax jge 0x283eef vmovups (%r14), %ymm0 vmovaps %ymm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi callq 0x28412c vmovups %ymm0, (%r14) addq $0x20, %r14 addl $0x8, %r15d addq $0x8, %r13 jmp 0x283e8b vmovaps (%r14), %xmm0 vmovaps %xmm0, 0x40(%rsp) leaq 0x17(%rsp), %rdi leaq 0x40(%rsp), %rsi vzeroupper callq 0x284270 vmovaps %xmm0, (%r14) addq $0x10, %r14 addl $0x4, %r15d addq $0x4, %r13 leal 0x3(%r15), %eax cmpl %r12d, %eax jl 0x283ec1 movq 0x30(%rsp), %rax imulq 0x28(%rsp), %rax movq 0x38(%rsp), %r15 imulq %r15, %rax addq %rax, %rbx cmpl %r12d, %r13d jge 0x283f2d vmovss (%rbx,%r13,4), %xmm0 vzeroupper callq 0x5f200 vmovss %xmm0, (%rbx,%r13,4) incq %r13 jmp 0x283f0f incq %r15 jmp 0x283e50 xorl %eax, %eax leaq -0x28(%rbp), %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/unaryop_x86_avx.cpp

ncnn::UnaryOp_x86_avx_functor::unary_op_exp::func_pack8(float vector[8] const&) const

static NCNN_FORCEINLINE __m256 exp256_ps(__m256 x) { __m256 tmp = _mm256_setzero_ps(), fx; __m256i imm0; __m256 one = *(__m256*)_ps256_1; x = _mm256_min_ps(x, *(__m256*)_ps256_exp_hi); x = _mm256_max_ps(x, *(__m256*)_ps256_exp_lo); /* express exp(x) as exp(g + n*log(2)) */ fx = _mm256_comp_fmadd_ps(x, *(__m256*)_ps256_cephes_LOG2EF, *(__m256*)_ps256_0p5); /* how to perform a floorf with SSE: just below */ //imm0 = _mm256_cvttps_epi32(fx); //tmp = _mm256_cvtepi32_ps(imm0); tmp = _mm256_floor_ps(fx); /* if greater, subtract 1 */ //__m256 mask = _mm256_cmpgt_ps(tmp, fx); __m256 mask = _mm256_cmp_ps(tmp, fx, _CMP_GT_OS); mask = _mm256_and_ps(mask, one); fx = _mm256_sub_ps(tmp, mask); // x = x - fx * exp_C1 x = _mm256_comp_fnmadd_ps(fx, *(__m256*)_ps256_cephes_exp_C1, x); // x = x - fx * exp_C2 x = _mm256_comp_fnmadd_ps(fx, *(__m256*)_ps256_cephes_exp_C2, x); tmp = _mm256_mul_ps(x, x); __m256 y = *(__m256*)_ps256_cephes_exp_p0; y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p1); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p2); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p3); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p4); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_exp_p5); y = _mm256_comp_fmadd_ps(y, tmp, x); y = _mm256_add_ps(y, one); /* build 2^n */ imm0 = _mm256_cvttps_epi32(fx); // another two AVX2 instructions imm0 = _mm256_comp_add_epi32(imm0, *(__m256i*)_pi32_256_0x7f); imm0 = _mm256_comp_slli_epi32(imm0, 23); __m256 pow2n = _mm256_castsi256_ps(imm0); y = _mm256_mul_ps(y, pow2n); return y; }

vbroadcastss 0x16d253(%rip), %ymm0 # 0x3f11b8 vminps (%rsi), %ymm0, %ymm0 vbroadcastss 0x16d24a(%rip), %ymm1 # 0x3f11bc vmaxps %ymm1, %ymm0, %ymm0 vbroadcastss 0x16d241(%rip), %ymm1 # 0x3f11c0 vmulps %ymm1, %ymm0, %ymm1 vbroadcastss 0x16a088(%rip), %ymm2 # 0x3ee014 vaddps %ymm2, %ymm1, %ymm1 vroundps $0x1, %ymm1, %ymm3 vcmpltps %ymm3, %ymm1, %ymm1 vbroadcastss 0x16ace4(%rip), %ymm4 # 0x3eec88 vandps %ymm4, %ymm1, %ymm1 vsubps %ymm1, %ymm3, %ymm1 vbroadcastss 0x16ea2b(%rip), %ymm3 # 0x3f29e0 vmulps %ymm3, %ymm1, %ymm3 vaddps %ymm3, %ymm0, %ymm0 vbroadcastss 0x16d206(%rip), %ymm3 # 0x3f11cc vmulps %ymm0, %ymm0, %ymm5 vmulps %ymm3, %ymm0, %ymm3 vbroadcastss 0x16d1f9(%rip), %ymm6 # 0x3f11d0 vaddps %ymm6, %ymm3, %ymm3 vmulps %ymm0, %ymm3, %ymm3 vbroadcastss 0x16d1ec(%rip), %ymm6 # 0x3f11d4 vaddps %ymm6, %ymm3, %ymm3 vmulps %ymm0, %ymm3, %ymm3 vbroadcastss 0x16d1df(%rip), %ymm6 # 0x3f11d8 vaddps %ymm6, %ymm3, %ymm3 vmulps %ymm0, %ymm3, %ymm3 vbroadcastss 0x16d1d2(%rip), %ymm6 # 0x3f11dc vaddps %ymm6, %ymm3, %ymm3 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm2, %ymm3, %ymm2 vmulps %ymm2, %ymm5, %ymm2 vaddps %ymm4, %ymm0, %ymm0 vaddps %ymm2, %ymm0, %ymm0 vcvttps2dq %ymm1, %ymm1 vpslld $0x17, %xmm1, %xmm2 vextractf128 $0x1, %ymm1, %xmm1 vpslld $0x17, %xmm1, %xmm1 vbroadcastss 0x16ac49(%rip), %xmm3 # 0x3eec88 vpaddd %xmm3, %xmm1, %xmm1 vpaddd %xmm3, %xmm2, %xmm2 vinsertf128 $0x1, %xmm1, %ymm2, %ymm1 vmulps %ymm1, %ymm0, %ymm0 retq

/csukuangfj[P]ncnn/src/layer/x86/avx_mathfun.h

ncnn::UnaryOp_x86_avx_functor::unary_op_cos::func_pack8(float vector[8] const&) const

static NCNN_FORCEINLINE __m256 cos256_ps(__m256 x) { // any x __m256 xmm1, xmm2 = _mm256_setzero_ps(), xmm3, y; __m256i imm0, imm2; #ifndef __AVX2__ __m128i imm0_1, imm0_2; __m128i imm2_1, imm2_2; #endif /* take the absolute value */ x = _mm256_and_ps(x, *(__m256*)_ps256_inv_sign_mask); /* scale by 4/Pi */ y = _mm256_mul_ps(x, *(__m256*)_ps256_cephes_FOPI); #ifdef __AVX2__ /* store the integer part of y in mm0 */ imm2 = _mm256_cvttps_epi32(y); /* j=(j+1) & (~1) (see the cephes sources) */ imm2 = _mm256_comp_add_epi32(imm2, *(__m256i*)_pi32_256_1); imm2 = _mm256_and_si256(imm2, *(__m256i*)_pi32_256_inv1); y = _mm256_cvtepi32_ps(imm2); imm2 = _mm256_comp_sub_epi32(imm2, *(__m256i*)_pi32_256_2); /* get the swap sign flag */ imm0 = _mm256_andnot_si256(imm2, *(__m256i*)_pi32_256_4); imm0 = _mm256_comp_slli_epi32(imm0, 29); /* get the polynom selection mask */ imm2 = _mm256_and_si256(imm2, *(__m256i*)_pi32_256_2); imm2 = _mm256_cmpeq_epi32(imm2, *(__m256i*)_pi32_256_0); #else /* we use SSE2 routines to perform the integer ops */ COPY_IMM_TO_XMM(_mm256_cvttps_epi32(y), imm2_1, imm2_2); imm2_1 = _mm_add_epi32(imm2_1, *(__m128i*)_pi32avx_1); imm2_2 = _mm_add_epi32(imm2_2, *(__m128i*)_pi32avx_1); imm2_1 = _mm_and_si128(imm2_1, *(__m128i*)_pi32avx_inv1); imm2_2 = _mm_and_si128(imm2_2, *(__m128i*)_pi32avx_inv1); COPY_XMM_TO_IMM(imm2_1, imm2_2, imm2); y = _mm256_cvtepi32_ps(imm2); imm2_1 = _mm_sub_epi32(imm2_1, *(__m128i*)_pi32avx_2); imm2_2 = _mm_sub_epi32(imm2_2, *(__m128i*)_pi32avx_2); imm0_1 = _mm_andnot_si128(imm2_1, *(__m128i*)_pi32avx_4); imm0_2 = _mm_andnot_si128(imm2_2, *(__m128i*)_pi32avx_4); imm0_1 = _mm_slli_epi32(imm0_1, 29); imm0_2 = _mm_slli_epi32(imm0_2, 29); COPY_XMM_TO_IMM(imm0_1, imm0_2, imm0); imm2_1 = _mm_and_si128(imm2_1, *(__m128i*)_pi32avx_2); imm2_2 = _mm_and_si128(imm2_2, *(__m128i*)_pi32avx_2); imm2_1 = _mm_cmpeq_epi32(imm2_1, _mm_setzero_si128()); imm2_2 = _mm_cmpeq_epi32(imm2_2, _mm_setzero_si128()); COPY_XMM_TO_IMM(imm2_1, imm2_2, imm2); #endif __m256 sign_bit = _mm256_castsi256_ps(imm0); __m256 poly_mask = _mm256_castsi256_ps(imm2); /* The magic pass: "Extended precision modular arithmetic" x = ((x - y * DP1) - y * DP2) - y * DP3; */ xmm1 = *(__m256*)_ps256_minus_cephes_DP1; xmm2 = *(__m256*)_ps256_minus_cephes_DP2; xmm3 = *(__m256*)_ps256_minus_cephes_DP3; x = _mm256_comp_fmadd_ps(y, xmm1, x); x = _mm256_comp_fmadd_ps(y, xmm2, x); x = _mm256_comp_fmadd_ps(y, xmm3, x); /* Evaluate the first polynom (0 <= x <= Pi/4) */ y = *(__m256*)_ps256_coscof_p0; __m256 z = _mm256_mul_ps(x, x); y = _mm256_comp_fmadd_ps(y, z, *(__m256*)_ps256_coscof_p1); y = _mm256_comp_fmadd_ps(y, z, *(__m256*)_ps256_coscof_p2); y = _mm256_mul_ps(y, z); y = _mm256_mul_ps(y, z); // y = y - z * 0.5 y = _mm256_comp_fnmadd_ps(z, *(__m256*)_ps256_0p5, y); y = _mm256_add_ps(y, *(__m256*)_ps256_1); /* Evaluate the second polynom (Pi/4 <= x <= 0) */ __m256 y2 = *(__m256*)_ps256_sincof_p0; y2 = _mm256_mul_ps(y2, z); y2 = _mm256_add_ps(y2, *(__m256*)_ps256_sincof_p1); y2 = _mm256_mul_ps(y2, z); y2 = _mm256_add_ps(y2, *(__m256*)_ps256_sincof_p2); y2 = _mm256_mul_ps(y2, z); y2 = _mm256_mul_ps(y2, x); y2 = _mm256_add_ps(y2, x); /* select the correct result from the two polynoms */ xmm3 = poly_mask; y2 = _mm256_and_ps(xmm3, y2); //, xmm3); y = _mm256_andnot_ps(xmm3, y); y = _mm256_add_ps(y, y2); /* update the sign */ y = _mm256_xor_ps(y, sign_bit); return y; }

vbroadcastss 0x16d2fd(%rip), %ymm0 # 0x3f1900 vandps (%rsi), %ymm0, %ymm2 vbroadcastss 0x173240(%rip), %ymm0 # 0x3f7850 vmulps %ymm0, %ymm2, %ymm0 vcvttps2dq %ymm0, %ymm0 vpcmpeqd %xmm1, %xmm1, %xmm1 vpsubd %xmm1, %xmm0, %xmm3 vextractf128 $0x1, %ymm0, %xmm0 vpsubd %xmm1, %xmm0, %xmm0 vmovddup 0x173d7e(%rip), %xmm1 # xmm1 = mem[0,0] vpand %xmm1, %xmm3, %xmm3 vpand %xmm1, %xmm0, %xmm0 vinsertf128 $0x1, %xmm0, %ymm3, %ymm1 vcvtdq2ps %ymm1, %ymm4 vbroadcastss 0x173207(%rip), %xmm1 # 0x3f7854 vpaddd %xmm1, %xmm3, %xmm3 vpaddd %xmm1, %xmm0, %xmm5 vpslld $0x1d, %xmm3, %xmm0 vmovddup 0x173d56(%rip), %xmm1 # xmm1 = mem[0,0] vpandn %xmm1, %xmm0, %xmm0 vpslld $0x1d, %xmm5, %xmm6 vpandn %xmm1, %xmm6, %xmm1 vinsertf128 $0x1, %xmm1, %ymm0, %ymm0 vbroadcastss 0x1731da(%rip), %xmm6 # 0x3f7858 vpand %xmm6, %xmm3, %xmm1 vbroadcastss 0x173d35(%rip), %ymm3 # 0x3f83c0 vpand %xmm6, %xmm5, %xmm5 vmulps %ymm3, %ymm4, %ymm3 vaddps %ymm2, %ymm3, %ymm2 vmulps %ymm2, %ymm2, %ymm3 vbroadcastss 0x1731c4(%rip), %ymm4 # 0x3f7868 vbroadcastss 0x1731bf(%rip), %ymm6 # 0x3f786c vmulps %ymm4, %ymm3, %ymm4 vaddps %ymm6, %ymm4, %ymm4 vbroadcastss 0x1731b2(%rip), %ymm6 # 0x3f7870 vmulps %ymm3, %ymm4, %ymm4 vaddps %ymm6, %ymm4, %ymm4 vbroadcastss 0x16d255(%rip), %ymm6 # 0x3f1924 vmulps %ymm3, %ymm4, %ymm4 vaddps %ymm6, %ymm4, %ymm4 vbroadcastss 0x16a5a8(%rip), %ymm6 # 0x3eec88 vmulps %ymm3, %ymm4, %ymm4 vaddps %ymm6, %ymm4, %ymm4 vbroadcastss 0x173183(%rip), %ymm6 # 0x3f7874 vmulps %ymm6, %ymm3, %ymm6 vbroadcastss 0x17317a(%rip), %ymm7 # 0x3f7878 vaddps %ymm7, %ymm6, %ymm6 vmulps %ymm3, %ymm6, %ymm6 vbroadcastss 0x17316d(%rip), %ymm7 # 0x3f787c vaddps %ymm7, %ymm6, %ymm6 vmulps %ymm2, %ymm3, %ymm3 vmulps %ymm6, %ymm3, %ymm3 vaddps %ymm2, %ymm3, %ymm2 vinsertf128 $0x1, %xmm5, %ymm1, %ymm1 vcvtdq2ps %ymm1, %ymm1 vxorps %xmm3, %xmm3, %xmm3 vcmpeqps %ymm3, %ymm1, %ymm1 vandps %ymm2, %ymm1, %ymm2 vandnps %ymm4, %ymm1, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vxorps %ymm1, %ymm0, %ymm0 retq nop

/csukuangfj[P]ncnn/src/layer/x86/avx_mathfun.h

ncnn::UnaryOp_x86_avx_functor::unary_op_log10::func_pack8(float vector[8] const&) const

static NCNN_FORCEINLINE __m256 log256_ps(__m256 x) { __m256i imm0; __m256 one = *(__m256*)_ps256_1; //__m256 invalid_mask = _mm256_cmple_ps(x, _mm256_setzero_ps()); __m256 invalid_mask = _mm256_cmp_ps(x, _mm256_setzero_ps(), _CMP_LE_OS); x = _mm256_max_ps(x, *(__m256*)_ps256_min_norm_pos); /* cut off denormalized stuff */ // can be done with AVX2 imm0 = _mm256_comp_srli_epi32(_mm256_castps_si256(x), 23); /* keep only the fractional part */ x = _mm256_and_ps(x, *(__m256*)_ps256_inv_mant_mask); x = _mm256_or_ps(x, *(__m256*)_ps256_0p5); // this is again another AVX2 instruction imm0 = _mm256_comp_sub_epi32(imm0, *(__m256i*)_pi32_256_0x7f); __m256 e = _mm256_cvtepi32_ps(imm0); e = _mm256_add_ps(e, one); /* part2: if( x < SQRTHF ) { e -= 1; x = x + x - 1.0; } else { x = x - 1.0; } */ //__m256 mask = _mm256_cmplt_ps(x, *(__m256*)_ps256_cephes_SQRTHF); __m256 mask = _mm256_cmp_ps(x, *(__m256*)_ps256_cephes_SQRTHF, _CMP_LT_OS); __m256 tmp = _mm256_and_ps(x, mask); x = _mm256_sub_ps(x, one); e = _mm256_sub_ps(e, _mm256_and_ps(one, mask)); x = _mm256_add_ps(x, tmp); __m256 z = _mm256_mul_ps(x, x); __m256 y = *(__m256*)_ps256_cephes_log_p0; y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_log_p1); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_log_p2); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_log_p3); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_log_p4); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_log_p5); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_log_p6); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_log_p7); y = _mm256_comp_fmadd_ps(y, x, *(__m256*)_ps256_cephes_log_p8); y = _mm256_mul_ps(y, x); y = _mm256_mul_ps(y, z); y = _mm256_comp_fmadd_ps(e, *(__m256*)_ps256_cephes_log_q1, y); //y = -z * 0.5 + y y = _mm256_comp_fnmadd_ps(z, *(__m256*)_ps256_0p5, y); x = _mm256_add_ps(x, y); x = _mm256_comp_fmadd_ps(e, *(__m256*)_ps256_cephes_log_q2, x); y = _mm256_or_ps(x, invalid_mask); // negative arg will be NAN return y; }

vbroadcastss 0x16bdc3(%rip), %ymm0 # 0x3f11e0 vmaxps (%rsi), %ymm0, %ymm0 vpsrld $0x17, %xmm0, %xmm1 vextractf128 $0x1, %ymm0, %xmm2 vpsrld $0x17, %xmm2, %xmm2 vbroadcastss 0x16bdaa(%rip), %ymm3 # 0x3f11e4 vandps %ymm3, %ymm0, %ymm0 vbroadcastss 0x168bcd(%rip), %ymm3 # 0x3ee014 vorps %ymm3, %ymm0, %ymm0 vbroadcastss 0x16c4b0(%rip), %xmm3 # 0x3f1904 vbroadcastss 0x16bd8f(%rip), %ymm4 # 0x3f11ec vcmpleps %ymm0, %ymm4, %ymm4 vandnps %ymm0, %ymm4, %ymm5 vbroadcastss 0x16bd81(%rip), %ymm6 # 0x3f11f0 vaddps %ymm6, %ymm0, %ymm0 vaddps %ymm5, %ymm0, %ymm0 vextractf128 $0x1, %ymm4, %xmm5 vpsubd %xmm5, %xmm2, %xmm2 vpaddd %xmm3, %xmm2, %xmm2 vpsubd %xmm4, %xmm1, %xmm1 vpaddd %xmm3, %xmm1, %xmm1 vinsertf128 $0x1, %xmm2, %ymm1, %ymm1 vcvtdq2ps %ymm1, %ymm2 vmulps %ymm0, %ymm0, %ymm1 vbroadcastss 0x16bd50(%rip), %ymm3 # 0x3f11f4 vbroadcastss 0x16bd4b(%rip), %ymm4 # 0x3f11f8 vmulps %ymm3, %ymm0, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd3e(%rip), %ymm4 # 0x3f11fc vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd31(%rip), %ymm4 # 0x3f1200 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd24(%rip), %ymm4 # 0x3f1204 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd17(%rip), %ymm4 # 0x3f1208 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bd0a(%rip), %ymm4 # 0x3f120c vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bcfd(%rip), %ymm4 # 0x3f1210 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16bcf0(%rip), %ymm4 # 0x3f1214 vmulps %ymm0, %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vbroadcastss 0x16c3eb(%rip), %ymm4 # 0x3f1920 vmulps %ymm0, %ymm3, %ymm3 vmulps %ymm4, %ymm2, %ymm2 vbroadcastss 0x16c3de(%rip), %ymm4 # 0x3f1924 vaddps %ymm4, %ymm3, %ymm3 vmulps %ymm3, %ymm1, %ymm1 vaddps %ymm0, %ymm2, %ymm0 vaddps %ymm1, %ymm0, %ymm0 vbroadcastss 0x172e65(%rip), %ymm1 # 0x3f83c4 vmulps %ymm1, %ymm0, %ymm0 retq

/csukuangfj[P]ncnn/src/layer/x86/avx_mathfun.h

ncnn::ConvolutionDepthWise::load_model(ncnn::ModelBin const&)

int ConvolutionDepthWise::load_model(const ModelBin& mb) { if (dynamic_weight) return 0; weight_data = mb.load(weight_data_size, 0); if (weight_data.empty()) return -100; if (bias_term) { bias_data = mb.load(num_output, 1); if (bias_data.empty()) return -100; } #if NCNN_INT8 if (int8_scale_term == 1 || int8_scale_term == 101) { weight_data_int8_scales = mb.load(group, 1); bottom_blob_int8_scales = mb.load(1, 1); float bottom_blob_int8_scale = bottom_blob_int8_scales[0]; bottom_blob_int8_scales = Mat(group); bottom_blob_int8_scales.fill(bottom_blob_int8_scale); } else if (int8_scale_term == 2 || int8_scale_term == 102) { weight_data_int8_scales = mb.load(1, 1); bottom_blob_int8_scales = mb.load(1, 1); // extend group if only one provided float weight_data_int8_scale = weight_data_int8_scales[0]; weight_data_int8_scales = Mat(group); weight_data_int8_scales.fill(weight_data_int8_scale); float bottom_blob_int8_scale = bottom_blob_int8_scales[0]; bottom_blob_int8_scales = Mat(group); bottom_blob_int8_scales.fill(bottom_blob_int8_scale); } if (int8_scale_term > 100) { top_blob_int8_scales = mb.load(1, 1); float top_blob_int8_scale = top_blob_int8_scales[0]; top_blob_int8_scales = Mat(group); top_blob_int8_scales.fill(top_blob_int8_scale); } #endif // NCNN_INT8 return 0; }

pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x60, %rsp cmpl $0x0, 0x160(%rdi) je 0x285b4c xorl %r12d, %r12d movl %r12d, %eax addq $0x60, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 retq movq %rsi, %r15 movq %rdi, %rbx movl 0x104(%rdi), %edx movq (%rsi), %rax leaq 0x10(%rsp), %r14 movq %r14, %rdi xorl %ecx, %ecx callq *0x10(%rax) leaq 0x168(%rbx), %r13 movq 0x8(%r14), %rax cmpq %r14, %r13 je 0x285c14 testq %rax, %rax je 0x285b84 lock incl (%rax) movq 0x170(%rbx), %rax testq %rax, %rax je 0x285bb8 lock decl (%rax) jne 0x285bb8 movq 0x168(%rbx), %rsi movq 0x188(%rbx), %rdi testq %rdi, %rdi je 0x285bb0 movq (%rdi), %rax callq *0x18(%rax) jmp 0x285bb8 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x168(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x170(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x178(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x180(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x188(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x190(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x1a0(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x1a8(%rbx) testq %rax, %rax je 0x285c3d lock decl (%rax) jne 0x285c3d movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x285c35 movq (%rdi), %rax callq *0x18(%rax) jmp 0x285c3d movq %rsi, %rdi callq 0x5f3e0 pushq $-0x64 popq %r12 cmpq $0x0, (%r13) je 0x285b3b movslq 0x1a0(%rbx), %rax imulq 0x1a8(%rbx), %rax testq %rax, %rax je 0x285b3b cmpl $0x0, 0x100(%rbx) je 0x285d7f movl 0xd0(%rbx), %edx movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi callq *0x10(%rax) leaq 0x1b0(%rbx), %r13 movq 0x18(%rsp), %rax cmpq %r14, %r13 je 0x285d33 testq %rax, %rax je 0x285ca3 lock incl (%rax) movq 0x1b8(%rbx), %rax testq %rax, %rax je 0x285cd7 lock decl (%rax) jne 0x285cd7 movq 0x1b0(%rbx), %rsi movq 0x1d0(%rbx), %rdi testq %rdi, %rdi je 0x285ccf movq (%rdi), %rax callq *0x18(%rax) jmp 0x285cd7 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x1b0(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x1b8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x1c0(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x1c8(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x1d0(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x1d8(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x1e8(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x1f0(%rbx) testq %rax, %rax je 0x285d5c lock decl (%rax) jne 0x285d5c movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x285d54 movq (%rdi), %rax callq *0x18(%rax) jmp 0x285d5c movq %rsi, %rdi callq 0x5f3e0 cmpq $0x0, (%r13) je 0x285b3b movslq 0x1e8(%rbx), %rax imulq 0x1f0(%rbx), %rax testq %rax, %rax je 0x285b3b movl 0x10c(%rbx), %eax cmpl $0x1, %eax je 0x285d9d cmpl $0x2, %eax je 0x285e0a cmpl $0x66, %eax je 0x285e0a cmpl $0x65, %eax jne 0x286582 movl 0x108(%rbx), %edx movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi callq *0x10(%rax) leaq 0x1f8(%rbx), %rcx movq 0x18(%rsp), %rax cmpq %r14, %rcx je 0x285fbe testq %rax, %rax je 0x285dcf lock incl (%rax) movq 0x200(%rbx), %rax testq %rax, %rax je 0x285f62 lock decl (%rax) jne 0x285f62 movq 0x1f8(%rbx), %rsi movq 0x218(%rbx), %rdi testq %rdi, %rdi je 0x285f5a movq (%rdi), %rax callq *0x18(%rax) jmp 0x285f62 movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi movl %ecx, %edx callq *0x10(%rax) leaq 0x1f8(%rbx), %r13 movq 0x18(%rsp), %rax cmpq %r14, %r13 je 0x285ec8 testq %rax, %rax je 0x285e38 lock incl (%rax) movq 0x200(%rbx), %rax testq %rax, %rax je 0x285e6c lock decl (%rax) jne 0x285e6c movq 0x1f8(%rbx), %rsi movq 0x218(%rbx), %rdi testq %rdi, %rdi je 0x285e64 movq (%rdi), %rax callq *0x18(%rax) jmp 0x285e6c movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x1f8(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x200(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x208(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x210(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x218(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x220(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x230(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x238(%rbx) testq %rax, %rax je 0x285ef1 lock decl (%rax) jne 0x285ef1 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x285ee9 movq (%rdi), %rax callq *0x18(%rax) jmp 0x285ef1 movq %rsi, %rdi callq 0x5f3e0 movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi movl %ecx, %edx callq *0x10(%rax) leaq 0x240(%rbx), %r12 movq 0x18(%rsp), %rax cmpq %r14, %r12 je 0x2860b4 testq %rax, %rax je 0x285f1f lock incl (%rax) movq 0x248(%rbx), %rax testq %rax, %rax je 0x286058 lock decl (%rax) jne 0x286058 movq 0x240(%rbx), %rsi movq 0x260(%rbx), %rdi testq %rdi, %rdi je 0x286050 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286058 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x1f8(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x200(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x208(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x210(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x218(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x220(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x230(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x238(%rbx) testq %rax, %rax je 0x285fe7 lock decl (%rax) jne 0x285fe7 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x285fdf movq (%rdi), %rax callq *0x18(%rax) jmp 0x285fe7 movq %rsi, %rdi callq 0x5f3e0 movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi movl %ecx, %edx callq *0x10(%rax) leaq 0x240(%rbx), %r12 movq 0x18(%rsp), %rax cmpq %r14, %r12 je 0x2861d3 testq %rax, %rax je 0x286015 lock incl (%rax) movq 0x248(%rbx), %rax testq %rax, %rax je 0x286177 lock decl (%rax) jne 0x286177 movq 0x240(%rbx), %rsi movq 0x260(%rbx), %rdi testq %rdi, %rdi je 0x28616f movq (%rdi), %rax callq *0x18(%rax) jmp 0x286177 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x240(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x248(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x250(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x258(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x260(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x268(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x278(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x280(%rbx) testq %rax, %rax je 0x2860dd lock decl (%rax) jne 0x2860dd movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2860d5 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2860dd movq %rsi, %rdi callq 0x5f3e0 movq 0x1f8(%rbx), %rax movss (%rax), %xmm0 movss %xmm0, 0xc(%rsp) movl 0x108(%rbx), %esi andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) movaps %xmm0, 0x30(%rsp) movups %xmm0, 0x3c(%rsp) pushq $0x4 popq %rdx movq %r14, %rdi xorl %ecx, %ecx callq 0x635fa movq 0x18(%rsp), %rax cmpq %r14, %r13 je 0x2862f2 testq %rax, %rax je 0x286134 lock incl (%rax) movq 0x200(%rbx), %rax testq %rax, %rax je 0x286296 lock decl (%rax) jne 0x286296 movq 0x1f8(%rbx), %rsi movq 0x218(%rbx), %rdi testq %rdi, %rdi je 0x28628e movq (%rdi), %rax callq *0x18(%rax) jmp 0x286296 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x240(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x248(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x250(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x258(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x260(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x268(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x278(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x280(%rbx) testq %rax, %rax je 0x2861fc lock decl (%rax) jne 0x2861fc movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2861f4 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2861fc movq %rsi, %rdi callq 0x5f3e0 movq 0x240(%rbx), %rax movss (%rax), %xmm0 movss %xmm0, 0xc(%rsp) movl 0x108(%rbx), %esi andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) movaps %xmm0, 0x30(%rsp) movups %xmm0, 0x3c(%rsp) pushq $0x4 popq %rdx movq %r14, %rdi xorl %ecx, %ecx callq 0x635fa movq 0x18(%rsp), %rax cmpq %r14, %r12 je 0x28644c testq %rax, %rax je 0x286253 lock incl (%rax) movq 0x248(%rbx), %rax testq %rax, %rax je 0x2863f0 lock decl (%rax) jne 0x2863f0 movq 0x240(%rbx), %rsi movq 0x260(%rbx), %rdi testq %rdi, %rdi je 0x2863e8 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2863f0 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x1f8(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x200(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x208(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x210(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x218(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x220(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x230(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x238(%rbx) testq %rax, %rax movss 0xc(%rsp), %xmm0 je 0x28632d lock decl (%rax) jne 0x28632d movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x28631f movq (%rdi), %rax callq *0x18(%rax) movss 0xc(%rsp), %xmm0 jmp 0x28632d movq %rsi, %rdi callq 0x5f3e0 movss 0xc(%rsp), %xmm0 movl 0x230(%rbx), %eax imull 0x238(%rbx), %eax movq 0x1f8(%rbx), %rcx xorl %edx, %edx testl %eax, %eax cmovlel %edx, %eax cmpl %edx, %eax je 0x286356 movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x286348 movq 0x240(%rbx), %rax movss (%rax), %xmm0 movss %xmm0, 0xc(%rsp) movl 0x108(%rbx), %esi andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) movaps %xmm0, 0x30(%rsp) movups %xmm0, 0x3c(%rsp) pushq $0x4 popq %rdx movq %r14, %rdi xorl %ecx, %ecx callq 0x635fa movq 0x18(%rsp), %rax cmpq %r14, %r12 je 0x286518 testq %rax, %rax je 0x2863ad lock incl (%rax) movq 0x248(%rbx), %rax testq %rax, %rax je 0x2864bc lock decl (%rax) jne 0x2864bc movq 0x240(%rbx), %rsi movq 0x260(%rbx), %rdi testq %rdi, %rdi je 0x2864b4 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2864bc movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x240(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x248(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x250(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x258(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x260(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x268(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x278(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x280(%rbx) testq %rax, %rax movss 0xc(%rsp), %xmm0 je 0x286487 lock decl (%rax) jne 0x286487 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286479 movq (%rdi), %rax callq *0x18(%rax) movss 0xc(%rsp), %xmm0 jmp 0x286487 movq %rsi, %rdi callq 0x5f3e0 movss 0xc(%rsp), %xmm0 movl 0x278(%rbx), %eax imull 0x280(%rbx), %eax movq 0x240(%rbx), %rcx xorl %edx, %edx testl %eax, %eax cmovlel %edx, %eax cmpl %edx, %eax je 0x28657c movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x2864a2 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x240(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x248(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x250(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x258(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x260(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x268(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x278(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x280(%rbx) testq %rax, %rax movss 0xc(%rsp), %xmm0 je 0x286553 lock decl (%rax) jne 0x286553 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286545 movq (%rdi), %rax callq *0x18(%rax) movss 0xc(%rsp), %xmm0 jmp 0x286553 movq %rsi, %rdi callq 0x5f3e0 movss 0xc(%rsp), %xmm0 movl 0x278(%rbx), %eax imull 0x280(%rbx), %eax movq 0x240(%rbx), %rcx xorl %edx, %edx testl %eax, %eax cmovlel %edx, %eax cmpl %edx, %eax je 0x28657c movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x28656e movl 0x10c(%rbx), %eax cmpl $0x65, %eax jl 0x285b38 movq (%r15), %rax pushq $0x1 popq %rcx movq %r14, %rdi movq %r15, %rsi movl %ecx, %edx callq *0x10(%rax) leaq 0x288(%rbx), %r15 movq 0x18(%rsp), %rax cmpq %r14, %r15 je 0x286649 testq %rax, %rax je 0x2865b9 lock incl (%rax) movq 0x290(%rbx), %rax testq %rax, %rax je 0x2865ed lock decl (%rax) jne 0x2865ed movq 0x288(%rbx), %rsi movq 0x2a8(%rbx), %rdi testq %rdi, %rdi je 0x2865e5 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2865ed movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x288(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x290(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x298(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x2a0(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x2a8(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x2b0(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x2c0(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x2c8(%rbx) testq %rax, %rax je 0x286672 lock decl (%rax) jne 0x286672 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x28666a movq (%rdi), %rax callq *0x18(%rax) jmp 0x286672 movq %rsi, %rdi callq 0x5f3e0 movq 0x288(%rbx), %rax movss (%rax), %xmm0 movss %xmm0, 0xc(%rsp) movl 0x108(%rbx), %esi andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) movaps %xmm0, 0x30(%rsp) movups %xmm0, 0x3c(%rsp) pushq $0x4 popq %rdx movq %r14, %rdi xorl %ecx, %ecx callq 0x635fa movq 0x18(%rsp), %rax cmpq %r14, %r15 je 0x286759 testq %rax, %rax je 0x2866c9 lock incl (%rax) movq 0x290(%rbx), %rax testq %rax, %rax je 0x2866fd lock decl (%rax) jne 0x2866fd movq 0x288(%rbx), %rsi movq 0x2a8(%rbx), %rdi testq %rdi, %rdi je 0x2866f5 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2866fd movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rax movq %rax, 0x288(%rbx) movq 0x18(%rsp), %rax movq %rax, 0x290(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x298(%rbx) movl 0x28(%rsp), %ecx movl %ecx, 0x2a0(%rbx) movq 0x30(%rsp), %rcx movq %rcx, 0x2a8(%rbx) movups 0x38(%rsp), %xmm0 movups %xmm0, 0x2b0(%rbx) movl 0x48(%rsp), %ecx movl %ecx, 0x2c0(%rbx) movq 0x50(%rsp), %rcx movq %rcx, 0x2c8(%rbx) testq %rax, %rax movss 0xc(%rsp), %xmm0 je 0x286794 lock decl (%rax) jne 0x286794 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286786 movq (%rdi), %rax callq *0x18(%rax) movss 0xc(%rsp), %xmm0 jmp 0x286794 movq %rsi, %rdi callq 0x5f3e0 movss 0xc(%rsp), %xmm0 movl 0x2c0(%rbx), %eax imull 0x2c8(%rbx), %eax movq 0x288(%rbx), %rcx xorl %r12d, %r12d testl %eax, %eax cmovlel %r12d, %eax xorl %edx, %edx cmpl %edx, %eax je 0x285b3b movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x2867b3 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x286a40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a jmp 0x286a5a movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x286a48 lock decl (%rax) jne 0x286a48 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi jne 0x286a50 movq %rsi, %rdi callq 0x5f3e0 movq %rbx, %rdi callq 0x5f340 movq (%rdi), %rax callq *0x18(%rax) jmp 0x286a48 jmp 0x286a5a movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/convolutiondepthwise.cpp

ncnn::ConvolutionDepthWise::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { // convolv with NxN kernel // value = value + bias #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return forward_int8(bottom_blob, top_blob, opt); } #endif // NCNN_LOGE("ConvolutionDepthWise input %d x %d pad = %d %d ksize=%d %d stride=%d %d", w, h, pad_w, pad_h, kernel_w, kernel_h, stride_w, stride_h); Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; const int w = bottom_blob_bordered.w; const int h = bottom_blob_bordered.h; const size_t elemsize = bottom_blob_bordered.elemsize; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; const int outw = (w - kernel_extent_w) / stride_w + 1; const int outh = (h - kernel_extent_h) / stride_h + 1; top_blob.create(outw, outh, num_output, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; int ret = convolutiondepthwise(bottom_blob_bordered, top_blob, weight_data, bias_data, kernel_w, kernel_h, stride_w, stride_h, dilation_w, dilation_h, group, activation_type, activation_params, opt); if (ret != 0) return ret; return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x48, %rsp movq %rcx, %r12 movq %rdx, %r15 movq %rdi, %r14 cmpb $0x1, 0x1e(%rcx) jne 0x286fbf cmpq $0x1, 0x178(%r14) jne 0x286fbf movq %r14, %rdi movq %r15, %rdx movq %r12, %rcx addq $0x48, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x287170 movq %rsp, %rdx andq $0x0, 0x40(%rdx) xorps %xmm0, %xmm0 movaps %xmm0, (%rdx) movups %xmm0, 0xc(%rdx) movaps %xmm0, 0x20(%rdx) movups %xmm0, 0x2c(%rdx) movq %r14, %rdi movq %r12, %rcx callq 0x287daa pushq $-0x64 popq %rbx cmpq $0x0, (%rsp) je 0x2870eb movslq 0x38(%rsp), %rax imulq 0x40(%rsp), %rax testq %rax, %rax je 0x2870eb movq 0x10(%rsp), %r8 movl 0xd4(%r14), %eax decl %eax imull 0xdc(%r14), %eax movl 0xd0(%r14), %ecx notl %eax movl 0xd8(%r14), %esi decl %esi imull 0xe0(%r14), %esi addl 0x2c(%rsp), %eax cltd idivl 0xe4(%r14) movl %eax, %edi notl %esi addl 0x30(%rsp), %esi movl %esi, %eax cltd idivl 0xe8(%r14) leal 0x1(%rdi), %esi leal 0x1(%rax), %edx movq 0x8(%r12), %r9 movq %r15, %rdi callq 0x63810 cmpq $0x0, (%r15) je 0x2870eb movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2870eb leaq 0x168(%r14), %rdx leaq 0x1b0(%r14), %rcx movl 0xd4(%r14), %r8d movl 0xd8(%r14), %r9d movl 0xe4(%r14), %eax movl 0xe8(%r14), %r10d movl 0xdc(%r14), %r11d movl 0xe0(%r14), %r12d movl 0x108(%r14), %r13d movl 0x110(%r14), %ebp addq $0x118, %r14 # imm = 0x118 xorl %ebx, %ebx subq $0x8, %rsp leaq 0x8(%rsp), %rdi movq %r15, %rsi pushq %r14 pushq %rbp pushq %r13 pushq %r12 pushq %r11 pushq %r10 pushq %rax callq 0x287dbf addq $0x40, %rsp movq 0x8(%rsp), %rax testq %rax, %rax je 0x287118 lock decl (%rax) jne 0x287118 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x287110 movq (%rdi), %rax callq *0x18(%rax) jmp 0x287118 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0x48, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x28712f jmp 0x287167 jmp 0x28712f movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x28715f lock decl (%rax) jne 0x28715f movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x287159 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28715f movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/convolutiondepthwise.cpp

ncnn::ConvolutionDepthWise::forward_int8(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { // convolv with NxN kernel // value = value + bias int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; if (channels % group != 0 || num_output % group != 0) { // reject invalid group return -100; } // NCNN_LOGE("ConvolutionDepthWise input %d x %d pad = %d %d ksize=%d %d stride=%d %d", w, h, pad_w, pad_h, kernel_w, kernel_h, stride_w, stride_h); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_int8 = bottom_blob; if (elemsize != 1) { const int channels_g = channels / group; Mat scales(channels); { float* ps = scales; for (int g = 0; g < group; g++) { float scale = bottom_blob_int8_scales[g]; for (int q = 0; q < channels_g; q++) { *ps++ = scale; } } } Option opt_q = opt; opt_q.blob_allocator = opt.workspace_allocator; quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q); } Mat bottom_blob_bordered; make_padding(bottom_blob_int8, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } // int8 bool use_int8_requantize = int8_scale_term > 100; size_t out_elemsize = use_int8_requantize ? 1u : 4u; top_blob.create(outw, outh, num_output, out_elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels == group && group == num_output) { #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { signed char* outptr = top_blob.channel(g); const signed char* kptr = (const signed char*)weight_data + maxk * g; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* sptr = m.row<signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr[0] = sums8; outptr += 1; } else { // dequantize ((float*)outptr)[0] = sumfp32; outptr += 4; } } } } } else { // group convolution const int channels_g = channels / group; const int num_output_g = num_output / group; #ifdef _WIN32 #pragma omp parallel for num_threads(opt.num_threads) #else // _WIN32 #pragma omp parallel for collapse(2) num_threads(opt.num_threads) #endif // _WIN32 for (int g = 0; g < group; g++) { for (int p = 0; p < num_output_g; p++) { signed char* outptr = top_blob.channel(g * num_output_g + p); const signed char* weight_data_ptr = (const signed char*)weight_data + maxk * channels_g * num_output_g * g; for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* kptr = weight_data_ptr + maxk * channels_g * p; // channels_g for (int q = 0; q < channels_g; q++) { const Mat m = bottom_blob_bordered.channel(channels_g * g + q); const signed char* sptr = m.row<signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } kptr += maxk; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g * num_output_g + p]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr[0] = sums8; outptr += 1; } else { // dequantize ((float*)outptr)[0] = sumfp32; outptr += 4; } } } } } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x1a8, %rsp # imm = 0x1A8 movq %rcx, %r14 movq %rdx, %r8 movl 0x38(%rsi), %ebp movl 0x108(%rdi), %ecx movl %ebp, %eax cltd idivl %ecx pushq $-0x64 popq %rbx testl %edx, %edx jne 0x2871a9 movl 0xd0(%rdi), %eax cltd idivl %ecx testl %edx, %edx je 0x2871bd movl %ebx, %eax addq $0x1a8, %rsp # imm = 0x1A8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq movq %rsi, %r13 movq %r8, 0x58(%rsp) movl 0xd4(%rdi), %r15d decl %r15d imull 0xdc(%rdi), %r15d movl 0xd8(%rdi), %r12d decl %r12d movq %rdi, 0x8(%rsp) imull 0xe0(%rdi), %r12d movq 0x10(%rsi), %rax movq (%rsi), %rcx movq %rcx, 0x118(%rsp) movq 0x8(%rsi), %rcx movq %rcx, 0x120(%rsp) movq %rax, 0x128(%rsp) movl 0x18(%rsi), %edx movl %edx, 0x130(%rsp) movq 0x20(%rsi), %rdx movq %rdx, 0x138(%rsp) movups 0x28(%rsi), %xmm0 movups %xmm0, 0x140(%rsp) movl %ebp, 0x150(%rsp) movq 0x40(%rsi), %rdx movq %rdx, 0x158(%rsp) testq %rcx, %rcx je 0x28724e lock incl (%rcx) cmpq $0x1, %rax movl %ebp, 0x10(%rsp) je 0x28735c movq %r14, 0x18(%rsp) movq 0x8(%rsp), %r14 movl 0x108(%r14), %eax movl %eax, 0x30(%rsp) leaq 0x70(%rsp), %rdi andq $0x0, 0x40(%rdi) xorps %xmm0, %xmm0 movaps %xmm0, (%rdi) movups %xmm0, 0xc(%rdi) movaps %xmm0, 0x20(%rdi) movups %xmm0, 0x2c(%rdi) xorl %ebx, %ebx pushq $0x4 popq %rdx movl %ebp, %esi xorl %ecx, %ecx callq 0x635fa movl %ebp, %eax cltd idivl 0x30(%rsp) movq 0x70(%rsp), %rcx movl 0x108(%r14), %edx testl %eax, %eax cmovlel %ebx, %eax movq 0x240(%r14), %rsi testl %edx, %edx cmovlel %ebx, %edx movq 0x18(%rsp), %r14 cmpq %rdx, %rbx je 0x2872e4 movss (%rsi,%rbx,4), %xmm0 movl %eax, %edi subl $0x1, %edi jb 0x2872df movss %xmm0, (%rcx) addq $0x4, %rcx jmp 0x2872d0 incq %rbx jmp 0x2872c4 movups (%r14), %xmm0 movups 0x10(%r14), %xmm1 movups 0x20(%r14), %xmm2 movups 0x30(%r14), %xmm3 leaq 0x160(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x10(%r14), %rax movq %rax, 0x8(%rcx) leaq 0x118(%rsp), %rsi leaq 0x70(%rsp), %rdx movq %r13, %rdi callq 0x652e3 movq 0x78(%rsp), %rax testq %rax, %rax je 0x28735c lock decl (%rax) jne 0x28735c movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x287354 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28735c movq %rsi, %rdi callq 0x5f3e0 leaq 0x70(%rsp), %rdx andq $0x0, 0x40(%rdx) xorps %xmm0, %xmm0 movaps %xmm0, (%rdx) movups %xmm0, 0xc(%rdx) movaps %xmm0, 0x20(%rdx) movups %xmm0, 0x2c(%rdx) leaq 0x118(%rsp), %rsi movq 0x8(%rsp), %rdi movq %r14, %rcx callq 0x287daa pushq $-0x64 popq %rbx cmpq $0x0, 0x70(%rsp) je 0x287c5f movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x287c5f movq %r14, 0x18(%rsp) notl %r15d movl 0x9c(%rsp), %ebx addl %ebx, %r15d movl %r15d, %eax cltd movq 0x8(%rsp), %r14 idivl 0xe4(%r14) movl %eax, %r15d notl %r12d addl 0xa0(%rsp), %r12d movl %r12d, %eax cltd idivl 0xe8(%r14) movl %eax, %r12d movslq 0xd4(%r14), %rax movslq 0xd8(%r14), %r13 imulq %rax, %r13 leaq 0x160(%rsp), %rdi leaq 0x27(%rsp), %rdx movq %r13, %rsi callq 0x73bbe movq %r15, 0x28(%rsp) leal 0x1(%r15), %esi movq %r12, 0x60(%rsp) leal 0x1(%r12), %edx imull 0xe0(%r14), %ebx movq 0x160(%rsp), %rbp movl 0xd4(%r14), %eax movl 0xdc(%r14), %ecx imull %eax, %ecx subl %ecx, %ebx xorl %ecx, %ecx xorl %edi, %edi xorl %r8d, %r8d movq 0x58(%rsp), %r14 movq 0x8(%rsp), %r9 cmpl 0xd8(%r9), %ecx jge 0x2874a4 movslq %r8d, %r8 leaq (,%r8,4), %r10 addq %rbp, %r10 xorl %r9d, %r9d cmpl %eax, %r9d jge 0x28749b movl %edi, (%r10,%r9,4) movq 0x8(%rsp), %rax addl 0xdc(%rax), %edi movl 0xd4(%rax), %eax incq %r9 jmp 0x28747c addl %ebx, %edi incl %ecx addq %r9, %r8 jmp 0x28745d movq 0x8(%rsp), %rax movl 0xd0(%rax), %ecx movl 0x10c(%rax), %edi xorl %eax, %eax movl %edi, 0x44(%rsp) cmpl $0x65, %edi setl %al leaq 0x1(%rax,%rax,2), %r8 movq 0x18(%rsp), %rax movq 0x8(%rax), %r9 movq %r14, %rdi callq 0x63810 pushq $-0x64 popq %rax movq %rax, 0x50(%rsp) cmpq $0x0, (%r14) movq 0x28(%rsp), %r8 movl 0x10(%rsp), %eax je 0x287c4d movslq 0x38(%r14), %rcx imulq 0x40(%r14), %rcx testq %rcx, %rcx je 0x287c4d movq 0x8(%rsp), %rdx movl 0xd0(%rdx), %ecx movl 0x108(%rdx), %edi movl %eax, %esi xorl %edi, %esi movl %edi, %edx xorl %ecx, %edx orl %esi, %edx jne 0x28784a movq $0x0, 0x50(%rsp) testl %r13d, %r13d movl $0x0, %ebx cmovgl %r13d, %ebx movq $0x0, 0x48(%rsp) xorl %r9d, %r9d movq 0x8(%rsp), %rdx cltq cmpq %rax, %r9 jge 0x287c4d movq 0x40(%r14), %r12 imulq %r9, %r12 imulq 0x10(%r14), %r12 addq (%r14), %r12 movslq 0x9c(%rsp), %rcx movq 0xb0(%rsp), %rsi imulq %r9, %rsi movq 0x80(%rsp), %rax imulq %rax, %rsi addq 0x70(%rsp), %rsi movq %rsi, 0x30(%rsp) imulq %rax, %rcx movq %rcx, 0x10(%rsp) movq 0x168(%rdx), %r15 addq 0x48(%rsp), %r15 xorl %eax, %eax movq %r9, 0x38(%rsp) cmpl 0x60(%rsp), %eax jg 0x287832 xorl %r14d, %r14d movl %eax, 0xc0(%rsp) cltq movq %rax, 0x18(%rsp) cmpl %r8d, %r14d jg 0x287824 movslq 0xe8(%rdx), %rax imulq 0x18(%rsp), %rax imulq 0x10(%rsp), %rax movslq 0xe4(%rdx), %rdx movslq %r14d, %rcx imulq %rdx, %rcx addq 0x30(%rsp), %rcx addq %rax, %rcx xorl %edx, %edx xorl %eax, %eax cmpq %rdx, %rbx je 0x28761a movslq (%rbp,%rdx,4), %rsi movsbl (%rcx,%rsi), %esi movsbl (%r15,%rdx), %edi imull %esi, %edi addl %edi, %eax incq %rdx jmp 0x2875fd movq 0x8(%rsp), %rdx movq 0x1f8(%rdx), %rcx movss (%rcx,%r9,4), %xmm1 xorps %xmm0, %xmm0 ucomiss %xmm1, %xmm0 je 0x28764d movq 0x240(%rdx), %rcx mulss (%rcx,%r9,4), %xmm1 movss 0x16763f(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 cvtsi2ss %eax, %xmm4 mulss %xmm0, %xmm4 cmpl $0x0, 0x100(%rdx) je 0x28766b movq 0x1b0(%rdx), %rax addss (%rax,%r9,4), %xmm4 movl 0x110(%rdx), %eax decl %eax cmpl $0x5, %eax ja 0x287690 leaq 0x170dd9(%rip), %rcx # 0x3f8458 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq *%rax maxss 0x166980(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2877ca movaps %xmm4, %xmm0 movaps %xmm4, 0xd0(%rsp) callq 0x5f410 addss 0x1675d8(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x38(%rsp), %r9 movq 0x28(%rsp), %r8 movq 0x8(%rsp), %rdx mulss 0xd0(%rsp), %xmm0 jmp 0x2877ca movq 0x118(%rdx), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2877c7 jmp 0x2877ca movss 0x169ab8(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x166982(%rip), %xmm0 # 0x3ee090 cmpltss 0x169aa5(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x38(%rsp), %r9 movq 0x28(%rsp), %r8 movq 0x8(%rsp), %rdx movaps %xmm0, %xmm1 movss 0x167543(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2877ca movq 0x118(%rdx), %rax xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x16751f(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2877ca movq 0x118(%rdx), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x1668fa(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2877ca movss 0x1674de(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x287690 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 cmpl $0x65, 0x44(%rsp) jl 0x287810 movq 0x288(%rdx), %rax mulss (%rax,%r9,4), %xmm0 callq 0x5f2d0 movq 0x38(%rsp), %r9 movq 0x28(%rsp), %r8 movq 0x8(%rsp), %rdx cvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movb %al, (%r12) pushq $0x1 jmp 0x287818 movss %xmm0, (%r12) pushq $0x4 popq %rax addq %rax, %r12 incl %r14d jmp 0x2875c7 movl 0xc0(%rsp), %eax incl %eax jmp 0x2875ac incq %r9 movl 0x108(%rdx), %eax addq %r13, 0x48(%rsp) movq 0x58(%rsp), %r14 jmp 0x28754b cltd idivl %edi movl %eax, %esi movl %ecx, %eax cltd idivl %edi movl %esi, %ecx imull %r13d, %ecx movl %eax, %edx imull %ecx, %edx movl %edx, 0x6c(%rsp) movq $0x0, 0x50(%rsp) testl %r13d, %r13d movl $0x0, %r12d cmovgl %r13d, %r12d testl %esi, %esi movl $0x0, %r15d cmovgl %esi, %r15d movslq %ecx, %rcx movq %rcx, 0xf8(%rsp) testl %eax, %eax movl $0x0, %ecx cmovgl %eax, %ecx movq %rcx, 0x110(%rsp) movslq %esi, %rcx movq %rcx, 0xf0(%rsp) cltq movq %rax, 0xe8(%rsp) xorl %ebx, %ebx movq 0x8(%rsp), %rdx movq 0x60(%rsp), %rcx movslq %edi, %rax cmpq %rax, %rbx jge 0x287c4d movl 0x6c(%rsp), %eax imull %ebx, %eax movq %rbx, %rsi imulq 0xe8(%rsp), %rsi movq %rsi, 0x108(%rsp) cltq movq %rax, 0x100(%rsp) movq %rbx, 0x10(%rsp) imulq 0xf0(%rsp), %rbx xorl %eax, %eax cmpq 0x110(%rsp), %rax je 0x287c3a movq %rax, %rsi movq 0x108(%rsp), %rax leaq (%rsi,%rax), %r9 movq 0x58(%rsp), %rax movq 0x40(%rax), %rdi movq %r9, 0xd0(%rsp) imulq %r9, %rdi imulq 0x10(%rax), %rdi addq (%rax), %rdi movq 0x168(%rdx), %rax addq 0x100(%rsp), %rax movq %rsi, 0x48(%rsp) imulq 0xf8(%rsp), %rsi addq %rax, %rsi movq %rsi, 0x38(%rsp) xorl %eax, %eax cmpl %ecx, %eax jg 0x287c2d movl %eax, 0x30(%rsp) xorl %r14d, %r14d cmpl %r8d, %r14d jg 0x287c1d movq %rdi, 0x18(%rsp) movslq 0x9c(%rsp), %rcx movq %rdx, %r8 movq 0x80(%rsp), %rdx movq 0xb0(%rsp), %rax movslq 0xe8(%r8), %rsi movslq 0x30(%rsp), %rdi imulq %rsi, %rdi imulq %rdx, %rcx imulq %rdi, %rcx addq 0x70(%rsp), %rcx imulq %rdx, %rax movslq 0xe4(%r8), %rsi movslq %r14d, %rdx imulq %rsi, %rdx addq %rcx, %rdx xorl %esi, %esi xorl %ecx, %ecx movq 0x38(%rsp), %rdi cmpq %r15, %rsi je 0x287a13 leaq (%rsi,%rbx), %r8 imulq %rax, %r8 addq %rdx, %r8 xorl %r9d, %r9d cmpq %r9, %r12 je 0x287a0b movslq (%rbp,%r9,4), %r10 movsbl (%r8,%r10), %r10d movsbl (%rdi,%r9), %r11d imull %r10d, %r11d addl %r11d, %ecx incq %r9 jmp 0x2879eb addq %r13, %rdi incq %rsi jmp 0x2879d8 movq 0x8(%rsp), %rdx movq 0x1f8(%rdx), %rax movq 0x10(%rsp), %rsi movss (%rax,%rsi,4), %xmm1 xorps %xmm0, %xmm0 ucomiss %xmm1, %xmm0 je 0x287a49 movq 0x240(%rdx), %rax mulss (%rax,%rsi,4), %xmm1 movss 0x167243(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 cvtsi2ss %ecx, %xmm4 mulss %xmm0, %xmm4 cmpl $0x0, 0x100(%rdx) je 0x287a6e movq 0x1b0(%rdx), %rax movq 0xd0(%rsp), %rcx addss (%rax,%rcx,4), %xmm4 movl 0x110(%rdx), %eax decl %eax cmpl $0x5, %eax ja 0x287a93 leaq 0x1709be(%rip), %rcx # 0x3f8440 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq *%rax maxss 0x16657d(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x287bc3 movaps %xmm4, %xmm0 movaps %xmm4, 0xc0(%rsp) callq 0x5f410 addss 0x1671d5(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x10(%rsp), %rsi movq 0x8(%rsp), %rdx mulss 0xc0(%rsp), %xmm0 jmp 0x287bc3 movq 0x118(%rdx), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x287bc0 jmp 0x287bc3 movss 0x1696ba(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x166584(%rip), %xmm0 # 0x3ee090 cmpltss 0x1696a7(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x10(%rsp), %rsi movq 0x8(%rsp), %rdx movaps %xmm0, %xmm1 movss 0x16714a(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x287bc3 movq 0x118(%rdx), %rax xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x167126(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x287bc3 movq 0x118(%rdx), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x166501(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x287bc3 movss 0x1670e5(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x287a93 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 cmpl $0x65, 0x44(%rsp) jl 0x287c01 movq 0x288(%rdx), %rax mulss (%rax,%rsi,4), %xmm0 callq 0x5f2d0 movq 0x8(%rsp), %rdx cvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movq 0x18(%rsp), %rdi movb %al, (%rdi) pushq $0x1 jmp 0x287c0c movq 0x18(%rsp), %rdi movss %xmm0, (%rdi) pushq $0x4 popq %rax movq 0x28(%rsp), %r8 addq %rax, %rdi incl %r14d jmp 0x287974 movl 0x30(%rsp), %eax incl %eax movq 0x60(%rsp), %rcx jmp 0x287965 movq 0x48(%rsp), %rax incq %rax jmp 0x287904 movq 0x10(%rsp), %rbx incq %rbx movl 0x108(%rdx), %edi jmp 0x2878c3 leaq 0x160(%rsp), %rdi callq 0x624be movq 0x50(%rsp), %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x287c90 lock decl (%rax) jne 0x287c90 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x287c88 movq (%rdi), %rax callq *0x18(%rax) jmp 0x287c90 movq %rsi, %rdi callq 0x5f3e0 movq 0x120(%rsp), %rax testq %rax, %rax je 0x2871a9 lock decl (%rax) jne 0x2871a9 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x287cca movq (%rdi), %rax callq *0x18(%rax) jmp 0x2871a9 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2871a9 jmp 0x287da2 jmp 0x287da2 jmp 0x287da2 movq %rax, %rbx leaq 0x160(%rsp), %rdi callq 0x624be jmp 0x287d30 jmp 0x287d2d movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x287d61 lock decl (%rax) jne 0x287d61 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x287d51 movq (%rdi), %rax callq *0x18(%rax) jmp 0x287d61 jmp 0x287da2 movq %rax, %rbx jmp 0x287d61 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x287d61 lock decl (%rax) jne 0x287d61 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x287d5b movq %rsi, %rdi callq 0x5f3e0 jmp 0x287d61 movq (%rdi), %rax callq *0x18(%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x287d98 lock decl (%rax) jne 0x287d98 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi jne 0x287d92 movq %rsi, %rdi callq 0x5f3e0 jmp 0x287d98 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x287da2 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/convolutiondepthwise.cpp

ncnn::convolutiondepthwise(ncnn::Mat const&, ncnn::Mat&, ncnn::Mat const&, ncnn::Mat const&, int, int, int, int, int, int, int, int, ncnn::Mat const&, ncnn::Option const&)

static int convolutiondepthwise(const Mat& bottom_blob, Mat& top_blob, const Mat& weight_data, const Mat& bias_data, int kernel_w, int kernel_h, int stride_w, int stride_h, int dilation_w, int dilation_h, int group, int activation_type, const Mat& activation_params, const Option& opt) { const int w = bottom_blob.w; const int inch = bottom_blob.c; const int outw = top_blob.w; const int outh = top_blob.h; const int outch = top_blob.c; const int bias_term = bias_data.empty() ? 0 : 1; const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } // depth-wise if (inch == group && group == outch) { #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data + maxk * g; const Mat m = bottom_blob.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { float sum = 0.f; if (bias_term) sum = bias_data[g]; const float* sptr = m.row(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; float w = kptr[k]; sum += val * w; } outptr[j] = activation_ss(sum, activation_type, activation_params); } outptr += outw; } } } else { // group convolution const int inch_g = inch / group; const int outch_g = outch / group; #ifdef _WIN32 #pragma omp parallel for num_threads(opt.num_threads) #else #pragma omp parallel for collapse(2) num_threads(opt.num_threads) #endif for (int g = 0; g < group; g++) { for (int p = 0; p < outch_g; p++) { float* outptr = top_blob.channel(g * outch_g + p); const float* weight_data_ptr = (const float*)weight_data + maxk * inch_g * outch_g * g; // shadowed variable for less openmp task args const int outw = top_blob.w; const int outh = top_blob.h; for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { float sum = 0.f; if (bias_term) sum = bias_data[outch_g * g + p]; const float* kptr = weight_data_ptr + maxk * inch_g * p; for (int q = 0; q < inch_g; q++) { const Mat m = bottom_blob.channel(inch_g * g + q); const float* sptr = m.row(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { float val = sptr[space_ofs[k]]; float w = kptr[k]; sum += val * w; } kptr += maxk; } outptr[j] = activation_ss(sum, activation_type, activation_params); } outptr += outw; } } } } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x128, %rsp # imm = 0x128 movl %r9d, %ebp movl %r8d, %r15d movq %rdx, 0xa0(%rsp) movq %rsi, 0x58(%rsp) movl 0x170(%rsp), %ebx movq %rdi, 0x28(%rsp) movl 0x2c(%rdi), %r12d cmpq $0x0, (%rcx) movq %rcx, 0xa8(%rsp) je 0x287e14 movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax sete 0xf(%rsp) jmp 0x287e19 movb $0x1, 0xf(%rsp) movl 0x180(%rsp), %r13d movq 0x28(%rsp), %rax movl 0x38(%rax), %eax movl %eax, 0x40(%rsp) movq 0x58(%rsp), %rax movslq 0x2c(%rax), %rcx movq %rcx, 0x48(%rsp) movl 0x30(%rax), %ecx movq %rcx, 0x18(%rsp) movl 0x38(%rax), %eax movl %eax, 0x10(%rsp) movl %ebp, %eax imull %r15d, %eax movslq %eax, %r14 leaq 0x110(%rsp), %rdi leaq 0x27(%rsp), %rdx movq %r14, %rsi callq 0x73bbe movl %ebp, %r9d imull 0x178(%rsp), %r12d movl %ebx, %eax imull %r15d, %eax subl %eax, %r12d xorl %eax, %eax testl %r15d, %r15d cmovlel %eax, %r15d testl %ebp, %ebp cmovlel %eax, %r9d leaq 0x110(%rsp), %rcx movq (%rcx), %rbp movl %r15d, %ecx xorl %edx, %edx xorl %esi, %esi cmpl %r9d, %eax je 0x287ec8 movslq %ecx, %rdi movslq %esi, %r8 leal (%r15,%r8), %esi cmpq %r8, %rdi je 0x287ebe movl %edx, (%rbp,%r8,4) incq %r8 addl %ebx, %edx jmp 0x287ead addl %r12d, %edx incl %eax addl %r15d, %ecx jmp 0x287e9e movl 0x40(%rsp), %eax movl %eax, %edx xorl %r13d, %edx movl 0x10(%rsp), %esi movl %esi, %ecx xorl %r13d, %ecx orl %edx, %ecx movb 0xf(%rsp), %r10b movl 0x188(%rsp), %r11d jne 0x288230 movq 0x58(%rsp), %rax movq (%rax), %rcx movq %rcx, 0x80(%rsp) movq 0x40(%rax), %rcx imulq 0x10(%rax), %rcx movq %rcx, 0x78(%rsp) movq %r13, %rdx movq 0xa0(%rsp), %rax movq (%rax), %r13 movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rax movq (%rsi), %rcx movq %rcx, 0x70(%rsp) movq 0x10(%rsi), %rcx movq 0x40(%rsi), %rsi imulq %rcx, %rsi movq %rsi, 0x68(%rsp) imulq %rcx, %rax xorl %esi, %esi testl %r14d, %r14d movl $0x0, %r12d cmovgl %r14d, %r12d movslq 0x160(%rsp), %rdi movq 0x48(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r8d cmovgl %ecx, %r8d movslq 0x168(%rsp), %rcx movq 0x18(%rsp), %r15 testl %r15d, %r15d cmovlel %esi, %r15d movq %r15, 0x18(%rsp) testl %edx, %edx cmovlel %esi, %edx movq %rdx, 0x60(%rsp) imulq %rax, %rcx movq %rcx, 0x88(%rsp) shlq $0x2, %r14 leaq 0x1704e5(%rip), %r9 # 0x3f8488 movq %rdi, 0x28(%rsp) movq %r8, 0x50(%rsp) cmpq 0x60(%rsp), %rsi je 0x288651 movq 0x78(%rsp), %rax imulq %rsi, %rax addq 0x80(%rsp), %rax movq %rax, 0x10(%rsp) movq 0x68(%rsp), %rax imulq %rsi, %rax addq 0x70(%rsp), %rax movq %rax, 0xb0(%rsp) xorl %ebx, %ebx movq %rsi, 0x30(%rsp) cmpq 0x18(%rsp), %rbx je 0x288225 movq 0xa8(%rsp), %rax movq (%rax), %rax movq %rax, 0x40(%rsp) movq %rbx, 0x90(%rsp) imulq 0x88(%rsp), %rbx addq 0xb0(%rsp), %rbx movq 0x190(%rsp), %rax movq (%rax), %rax movq %rax, 0x38(%rsp) xorl %r15d, %r15d cmpq %r8, %r15 je 0x288202 xorps %xmm4, %xmm4 testb %r10b, %r10b jne 0x28804d movq 0x40(%rsp), %rax movss (%rax,%rsi,4), %xmm4 movq %r15, %rax imulq %rdi, %rax leaq (%rbx,%rax,4), %rax xorl %ecx, %ecx cmpq %rcx, %r12 je 0x288079 movslq (%rbp,%rcx,4), %rdx movss (%r13,%rcx,4), %xmm0 mulss (%rax,%rdx,4), %xmm0 addss %xmm0, %xmm4 incq %rcx jmp 0x28805a leal -0x1(%r11), %eax cmpl $0x5, %eax ja 0x288093 movslq (%r9,%rax,4), %rax addq %r9, %rax jmpq *%rax maxss 0x165f7d(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2881ef movaps %xmm4, %xmm0 movaps %xmm4, 0xc0(%rsp) callq 0x5f410 addss 0x166bd5(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 leaq 0x1703c4(%rip), %r9 # 0x3f8488 movq 0x50(%rsp), %r8 movq 0x28(%rsp), %rdi movq 0x30(%rsp), %rsi movl 0x188(%rsp), %r11d movb 0xf(%rsp), %r10b mulss 0xc0(%rsp), %xmm0 jmp 0x2881ef movq 0x38(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2881ec jmp 0x2881ef movss 0x1690a3(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x165f6d(%rip), %xmm0 # 0x3ee090 cmpltss 0x169090(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 leaq 0x170341(%rip), %r9 # 0x3f8488 movq 0x50(%rsp), %r8 movq 0x28(%rsp), %rdi movq 0x30(%rsp), %rsi movl 0x188(%rsp), %r11d movb 0xf(%rsp), %r10b movaps %xmm0, %xmm1 movss 0x166b1a(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2881ef xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x166afd(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x38(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2881ef movq 0x38(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x165ed5(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2881ef movss 0x166ab9(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x288093 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 movq 0x10(%rsp), %rax movss %xmm0, (%rax,%r15,4) incq %r15 jmp 0x288032 movq 0x48(%rsp), %rax movq 0x10(%rsp), %rcx leaq (%rcx,%rax,4), %rcx movq %rcx, 0x10(%rsp) movq 0x90(%rsp), %rbx incq %rbx jmp 0x287feb incq %rsi addq %r14, %r13 jmp 0x287fad cltd idivl %r13d movl %eax, %ecx movl %esi, %eax cltd idivl %r13d movl %ecx, %edx imull %r14d, %edx xorl %esi, %esi testl %r14d, %r14d movl $0x0, %ebx cmovgl %r14d, %ebx testl %ecx, %ecx movl $0x0, %r12d cmovgl %ecx, %r12d movslq 0x160(%rsp), %rdi movq %rdi, 0x50(%rsp) movslq 0x168(%rsp), %rdi movq %rdi, 0x70(%rsp) testl %eax, %eax movl $0x0, %edi cmovgl %eax, %edi movq %rdi, 0x68(%rsp) movslq %edx, %rdx movslq %ecx, %rcx cltq testl %r13d, %r13d cmovlel %esi, %r13d movq %rax, 0xd8(%rsp) movq %rcx, 0xe0(%rsp) imull %ecx, %eax imull %r14d, %eax movl %eax, 0x98(%rsp) shlq $0x2, %rdx movq %rdx, 0xf0(%rsp) shlq $0x2, %r14 xorl %eax, %eax xorl %ecx, %ecx movq %r13, 0x60(%rsp) cmpq %r13, %rcx je 0x288651 movl %eax, 0x9c(%rsp) movslq %eax, %rdx shlq $0x2, %rdx movq %rcx, %rsi movq 0x58(%rsp), %rax movq 0x40(%rax), %rdi imulq 0x10(%rax), %rdi movq %rdi, 0x100(%rsp) imulq 0xd8(%rsp), %rsi movq %rsi, 0x108(%rsp) movslq 0x2c(%rax), %rsi movq %rcx, 0xe8(%rsp) movq %rcx, %r13 imulq 0xe0(%rsp), %r13 testl %esi, %esi movl $0x0, %edi movq %rsi, 0x78(%rsp) cmovgl %esi, %edi movq %rdi, 0x40(%rsp) movl 0x30(%rax), %esi testl %esi, %esi movl $0x0, %ecx cmovlel %ecx, %esi movq %rsi, 0x88(%rsp) movq (%rax), %rax movq %rax, 0xf8(%rsp) movq 0xa0(%rsp), %rax addq (%rax), %rdx movq %rdx, 0x30(%rsp) xorl %ecx, %ecx cmpq 0x68(%rsp), %rcx je 0x28862e movq 0x108(%rsp), %rax leaq (%rcx,%rax), %rdx movq 0x100(%rsp), %rsi movq %rdx, 0x90(%rsp) imulq %rdx, %rsi addq 0xf8(%rsp), %rsi movq %rsi, 0x10(%rsp) xorl %edx, %edx movq %rcx, 0x80(%rsp) cmpq 0x88(%rsp), %rdx je 0x288614 movq 0xa8(%rsp), %rax movq (%rax), %rax movq %rax, 0xc0(%rsp) movq %rdx, 0x48(%rsp) imulq 0x70(%rsp), %rdx movq %rdx, 0x38(%rsp) movq 0x190(%rsp), %rax movq (%rax), %rax movq %rax, 0x18(%rsp) xorl %r15d, %r15d cmpq 0x40(%rsp), %r15 je 0x2885ec xorps %xmm4, %xmm4 testb %r10b, %r10b jne 0x288418 movq 0x90(%rsp), %rax movq 0xc0(%rsp), %rcx movss (%rcx,%rax,4), %xmm4 movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rcx imulq 0x38(%rsp), %rcx movq 0x10(%rsi), %rdx movq 0x40(%rsi), %rax imulq %rdx, %rax imulq %rdx, %rcx addq (%rsi), %rcx movq %r15, %rdx imulq 0x50(%rsp), %rdx leaq (%rcx,%rdx,4), %rcx movq 0x30(%rsp), %rdx xorl %esi, %esi cmpq %r12, %rsi je 0x288488 leaq (%rsi,%r13), %rdi imulq %rax, %rdi addq %rcx, %rdi xorl %r8d, %r8d cmpq %r8, %rbx je 0x288480 movslq (%rbp,%r8,4), %r9 movss (%rdx,%r8,4), %xmm0 mulss (%rdi,%r9,4), %xmm0 addss %xmm0, %xmm4 incq %r8 jmp 0x288461 incq %rsi addq %r14, %rdx jmp 0x28844e leal -0x1(%r11), %eax cmpl $0x5, %eax ja 0x2884a9 leaq 0x16ffd8(%rip), %rcx # 0x3f8470 movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq *%rax maxss 0x165b67(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2885d9 movaps %xmm4, %xmm0 movaps %xmm4, 0xb0(%rsp) callq 0x5f410 addss 0x1667bf(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movl 0x188(%rsp), %r11d movb 0xf(%rsp), %r10b mulss 0xb0(%rsp), %xmm0 jmp 0x2885d9 movq 0x18(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2885d6 jmp 0x2885d9 movss 0x168ca3(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x165b6d(%rip), %xmm0 # 0x3ee090 cmpltss 0x168c90(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movl 0x188(%rsp), %r11d movb 0xf(%rsp), %r10b movaps %xmm0, %xmm1 movss 0x166730(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2885d9 xorps %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x166713(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x18(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2885d9 movq 0x18(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x165aeb(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 xorps %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2885d9 movss 0x1666cf(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x2884a9 mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 movq 0x10(%rsp), %rax movss %xmm0, (%rax,%r15,4) incq %r15 jmp 0x2883f0 movq 0x78(%rsp), %rax movq 0x10(%rsp), %rcx leaq (%rcx,%rax,4), %rcx movq %rcx, 0x10(%rsp) movq 0x48(%rsp), %rdx incq %rdx movq 0x80(%rsp), %rcx jmp 0x2883ac incq %rcx movq 0x30(%rsp), %rax addq 0xf0(%rsp), %rax movq %rax, 0x30(%rsp) jmp 0x28836a movq 0xe8(%rsp), %rcx incq %rcx movl 0x9c(%rsp), %eax addl 0x98(%rsp), %eax movq 0x60(%rsp), %r13 jmp 0x2882cb leaq 0x110(%rsp), %rdi callq 0x624be addq $0x128, %rsp # imm = 0x128 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq

/csukuangfj[P]ncnn/src/layer/convolutiondepthwise.cpp

ncnn::ConvolutionDepthWise_x86::create_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x50, %rsp movq (%rdi), %rax movq -0x18(%rax), %r13 cmpl $0x0, 0x160(%rdi,%r13) je 0x288dd5 xorl %eax, %eax addq $0x50, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 retq movq %rsi, %r14 movq %rdi, %rbx movl 0x110(%rdi,%r13), %ecx decl %ecx cmpl $0x5, %ecx ja 0x289171 leaq 0x16f6cb(%rip), %rax # 0x3f84c0 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx jmpq *%rcx pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x288f68 pushq $0x47 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x288f68 pushq $0x36 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax movss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax movss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) jmp 0x288f68 pushq $0x1e popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x288f68 pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax movss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) jmp 0x288f68 pushq $0x43 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax movss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax movss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) leaq 0x8(%rsp), %rdi callq 0x71614 movq (%r15), %rax movq %r15, %rdi movq %r14, %rsi callq *0x20(%rax) movq (%rbx), %rax movq %r15, 0x8(%rbx) movq -0x18(%rax), %r15 cmpb $0x1, 0x1e(%r14) jne 0x288fab cmpq $0x1, 0x178(%rbx,%r15) jne 0x288fab movq %rbx, %rdi movq %r14, %rsi callq 0x289278 jmp 0x288dc5 movl 0xd0(%rbx,%r15), %esi movl 0xd4(%rbx,%r15), %r10d movl 0xd8(%rbx,%r15), %r11d movl %r11d, %r8d imull %r10d, %r8d movl 0x104(%rbx,%r15), %eax movl 0x108(%rbx,%r15), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x28911e imull %ecx, %eax cmpl %ecx, %eax jne 0x28911e leaq (%rbx,%r15), %rsi testb $0x3, %cl jne 0x289075 cmpb $0x0, 0x27(%r14) je 0x289075 addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x4 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x289129 lock decl (%rax) jne 0x289129 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x2891a8 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289129 xorl $0x3, %r10d xorl $0x3, %r11d orl %r10d, %r11d jne 0x28911e cmpl $0x1, 0xdc(%rsi) jne 0x2890aa cmpl $0x1, 0xe0(%rsi) jne 0x2890aa cmpl $0x1, 0xe4(%rsi) jne 0x2890aa cmpl $0x1, 0xe8(%rsi) je 0x2890ce cmpl $0x1, 0xdc(%rsi) jne 0x28911e cmpl $0x1, 0xe0(%rsi) jne 0x28911e cmpl $0x2, 0xe4(%rsi) jne 0x28911e cmpl $0x2, 0xe8(%rsi) jne 0x28911e addq $0x168, %r15 # imm = 0x168 cmpq $0x28, %r15 je 0x289129 movq 0x8(%rbx,%r15), %rax testq %rax, %rax je 0x2890e8 lock incl (%rax) leaq 0x28(%rbx), %r12 movq 0x30(%rbx), %rax testq %rax, %rax je 0x2891bd lock decl (%rax) jne 0x2891bd movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x2891b5 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2891bd movq %rbx, %rdi movq %r14, %rsi callq 0x2894d8 cmpb $0x1, (%r14) jne 0x288dc5 movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x289181 lock decl (%rax) jne 0x289181 movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x289179 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289181 xorl %r15d, %r15d jmp 0x288f81 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rbx) movups %xmm0, (%rbx) movups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x288dc5 movq %rsi, %rdi callq 0x5f3e0 jmp 0x289129 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%r12) movups %xmm0, (%r12) andl $0x0, 0x60(%rbx) movups %xmm0, 0x50(%rbx) movups (%rbx,%r15), %xmm0 movups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r15), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r15), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r15), %rax movq %rax, 0x48(%rbx) movups 0x28(%rbx,%r15), %xmm0 movups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r15), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r15), %rax movq %rax, 0x68(%rbx) jmp 0x289129 jmp 0x289256 jmp 0x289262 jmp 0x289262 jmp 0x289262 movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x28926f lock decl (%rax) jne 0x28926f movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi jne 0x28924e movq %rsi, %rdi callq 0x5f3e0 jmp 0x28926f movq (%rdi), %rax callq *0x18(%rax) jmp 0x28926f movq %rax, %rdi callq 0x61d68 jmp 0x289262 jmp 0x289262 movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq %rbx, %rdi callq 0x5f340 nop

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

ncnn::ConvolutionDepthWise_x86::create_pipeline_int8_x86(ncnn::Option const&)

int ConvolutionDepthWise_x86::create_pipeline_int8_x86(const Option& opt) { const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { elempack = channels % 8 == 0 ? 8 : 1; } #endif // __SSE2__ if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } if (elempack == 1) { weight_data_tm = weight_data; } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %r15 pushq %r14 pushq %r12 pushq %rbx subq $0x48, %rsp movq %rsi, %r15 movq %rdi, %rbx movq (%rdi), %rax movq -0x18(%rax), %r14 movl 0xd8(%rdi,%r14), %r8d imull 0xd4(%rdi,%r14), %r8d movl 0xd0(%rdi,%r14), %esi movl 0x104(%rdi,%r14), %eax movl 0x108(%rdi,%r14), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x28935a imull %ecx, %eax cmpl %ecx, %eax jne 0x28935a testb $0x7, %cl jne 0x2893ad cmpb $0x0, 0x27(%r15) je 0x2893ad addq %rbx, %r14 addq $0x168, %r14 # imm = 0x168 movq %rsp, %r12 movq %r12, %rdi movq %r14, %rsi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e addq $0x28, %rbx pushq $0x8 popq %rdx movq %r12, %rdi movq %rbx, %rsi movq %r15, %rcx callq 0x64e3b movq 0x8(%rsp), %rax testq %rax, %rax je 0x289487 lock decl (%rax) jne 0x289487 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x28941e movq (%rdi), %rax callq *0x18(%rax) jmp 0x289487 movq %rbx, %rdi movq %r15, %rsi callq 0x2894d8 cmpb $0x1, (%r15) jne 0x289487 movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x2893fa lock decl (%rax) jne 0x2893fa movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x2893f2 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2893fa addq $0x168, %r14 # imm = 0x168 cmpq $0x28, %r14 je 0x289487 movq 0x8(%rbx,%r14), %rax testq %rax, %rax je 0x2893cb lock incl (%rax) leaq 0x28(%rbx), %r15 movq 0x30(%rbx), %rax testq %rax, %rax je 0x289430 lock decl (%rax) jne 0x289430 movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x289428 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289430 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rbx) movups %xmm0, (%rbx) movups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x289487 movq %rsi, %rdi callq 0x5f3e0 jmp 0x289487 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%r15) movups %xmm0, (%r15) andl $0x0, 0x60(%rbx) movups %xmm0, 0x50(%rbx) movups (%rbx,%r14), %xmm0 movups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r14), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r14), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r14), %rax movq %rax, 0x48(%rbx) movups 0x28(%rbx,%r14), %xmm0 movups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r14), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r14), %rax movq %rax, 0x68(%rbx) xorl %eax, %eax addq $0x48, %rsp popq %rbx popq %r12 popq %r14 popq %r15 retq jmp 0x2894cf movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2894c7 lock decl (%rax) jne 0x2894c7 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2894c1 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2894c7 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

ncnn::ConvolutionDepthWise_x86::create_group_ops(ncnn::Option const&)

int ConvolutionDepthWise_x86::create_group_ops(const Option& opt) { // create Convolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Convolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[5]; weights[0] = weight_data_g; weights[1] = bias_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[2] = weight_data_int8_scales_g; weights[3] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[4] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[4]; weights[0] = weight_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[1] = weight_data_int8_scales_g; weights[2] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[3] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x278, %rsp # imm = 0x278 movq %rsi, 0x250(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x104(%rdi,%rdx), %eax movl 0x108(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x10(%rdi), %rax movq %rax, 0x1e0(%rsp) xorl %r12d, %r12d movq 0x10(%r14), %rax movq 0x18(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x289575 movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x289570 movq (%rdi), %rax callq *0x8(%rax) incq %r12 jmp 0x289547 imull %r15d, %ebx cmpq %rax, %rcx je 0x289582 movq %rax, 0x18(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x108(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0x1e0(%rsp), %rdi callq 0x6fbc2 leaq 0x118(%r14), %rax movq %rax, 0x258(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0x6c(%rsp) movslq %ebx, %rax movq %rax, 0x260(%rsp) movslq %r15d, %rax movq %rax, 0x248(%rsp) pushq $0x1 popq %rbx xorl %edx, %edx movl %r15d, 0x14(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rcx cmpq %rcx, %rdx jge 0x28a2ad movq %rdx, %rcx movq %rdx, 0x8(%rsp) movq 0x260(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdx, 0x80(%rsp) movl %esi, 0x88(%rsp) movq %rax, 0x90(%rsp) movl %ebx, 0x98(%rsp) movl %edi, 0x9c(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa0(%rsp) movl %ebx, 0xa8(%rsp) movq %rdi, 0xb0(%rsp) leaq 0x200(%rsp), %rdi leaq 0x70(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2896d3 lock decl (%rax) jne 0x2896d3 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2896cb movq (%rdi), %rax callq *0x18(%rax) jmp 0x2896d3 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x289743 movq 0x8(%rsp), %r15 movq 0x248(%rsp), %rcx imulq %rcx, %r15 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0x1f8(%rsp) imulq %rsi, %r15 addq 0x1b0(%r14,%rax), %r15 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x1c(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0x1f0(%rsp) movl %ebx, %r13d movl 0x14(%rsp), %eax movl %eax, 0x18(%rsp) movq %rcx, 0x1e8(%rsp) jmp 0x28977d xorl %r15d, %r15d movq $0x0, 0x1f8(%rsp) movl $0x0, 0x1c(%rsp) movq $0x0, 0x1f0(%rsp) xorl %r13d, %r13d movl $0x0, 0x18(%rsp) movq $0x0, 0x1e8(%rsp) pushq $0x6 popq %rdi callq 0x782bf movq %rax, %rbp leaq 0x268(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq %r12, %rdi xorl %esi, %esi movl 0x14(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r12, %rdi movl %ebx, %esi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r12, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r12, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r12, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r12, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r12, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r12, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r12, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x6 popq %rsi movl 0x6c(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r12, %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x110(%r14,%rax), %edx movq %r12, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x258(%rsp), %rdx movq %r12, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r12, %rsi callq *0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x28995e pushq $0x40 popq %rax xorps %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) movups %xmm0, 0x30(%rsp,%rax) movups %xmm0, 0x3c(%rsp,%rax) movups %xmm0, 0x50(%rsp,%rax) movups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x1a8, %rax # imm = 0x1A8 jne 0x2898f7 movq 0x208(%rsp), %rax testq %rax, %rax je 0x28992d lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2899da lock decl (%rax) jne 0x2899da movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2899d2 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2899da pushq $0x40 popq %rax xorps %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) movups %xmm0, 0x30(%rsp,%rax) movups %xmm0, 0x3c(%rsp,%rax) movups %xmm0, 0x50(%rsp,%rax) movups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x160, %rax # imm = 0x160 jne 0x289964 movq 0x208(%rsp), %rax testq %rax, %rax je 0x28999a lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x289d81 lock decl (%rax) jne 0x289d81 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x289d79 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289d81 movq %rsi, %rdi callq 0x5f3e0 movaps 0x200(%rsp), %xmm0 movaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) movups 0x228(%rsp), %xmm0 movups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x289a7a lock decl (%rax) jne 0x289a7a movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x289a72 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289a7a movq %rsi, %rdi callq 0x5f3e0 movq %r15, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq 0x1f8(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x1c(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x1f0(%rsp), %rax movq %rax, 0xd8(%rsp) movl %r13d, 0xe0(%rsp) movl 0x18(%rsp), %eax movl %eax, 0xe4(%rsp) movl %r13d, 0xe8(%rsp) movl %r13d, 0xec(%rsp) movl %r13d, 0xf0(%rsp) movq 0x1e8(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x289d03 andq $0x0, 0x60(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x20(%rsp) movups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax movups %xmm0, 0xc(%rax) movaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx movss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x289b7a movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x289b6c movq 0x28(%rsp), %rax testq %rax, %rax je 0x289b87 lock incl (%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x289bbe lock decl (%rax) jne 0x289bbe movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x289bb6 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289bbe movq %rsi, %rdi callq 0x5f3e0 movaps 0x20(%rsp), %xmm0 movaps %xmm0, 0x100(%rsp) movq 0x30(%rsp), %rax movq %rax, 0x110(%rsp) movl 0x38(%rsp), %eax movl %eax, 0x118(%rsp) movq 0x40(%rsp), %rax movq %rax, 0x120(%rsp) movups 0x48(%rsp), %xmm0 movups %xmm0, 0x128(%rsp) movl 0x58(%rsp), %eax movl %eax, 0x138(%rsp) movq 0x60(%rsp), %rax movq %rax, 0x140(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %r12d movq 0x260(%r14,%rax), %rbx movq 0x150(%rsp), %rax testq %rax, %rax je 0x289c7c lock decl (%rax) jne 0x289c7c movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x289c74 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289c7c movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %r12d, 0x160(%rsp) movq %rbx, 0x168(%rsp) movaps 0x16e784(%rip), %xmm0 # 0x3f8430 movaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x289cfc lock decl (%rax) jne 0x289cfc movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x289cf4 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289cfc movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x289f17 movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x198(%rsp), %rax testq %rax, %rax je 0x289ec6 lock decl (%rax) jne 0x289ec6 movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x289ebe movq (%rdi), %rax callq *0x18(%rax) jmp 0x289ec6 movq %rsi, %rdi callq 0x5f3e0 movaps 0x200(%rsp), %xmm0 movaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) movups 0x228(%rsp), %xmm0 movups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x28a0ef andq $0x0, 0x60(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x20(%rsp) movups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax movups %xmm0, 0xc(%rax) movaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx movss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x289e73 movss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x289e65 movq 0x28(%rsp), %rax testq %rax, %rax je 0x289e80 lock incl (%rax) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x289fab lock decl (%rax) jne 0x289fab movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x289fa3 movq (%rdi), %rax callq *0x18(%rax) jmp 0x289fab movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x190(%rsp) andq $0x0, 0x198(%rsp) movq %r15, 0x1a0(%rsp) movl %ebx, 0x1a8(%rsp) movq %r12, 0x1b0(%rsp) movaps 0x16e53b(%rip), %xmm0 # 0x3f8430 movups %xmm0, 0x1b8(%rsp) movl $0x1, 0x1c8(%rsp) movq $0x1, 0x1d0(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0x120, %r15d # imm = 0x120 movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x289f75 lock decl (%rax) jne 0x289f75 movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x289f6d movq (%rdi), %rax callq *0x18(%rax) jmp 0x289f75 movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x289f44 jmp 0x28a236 movq %rsi, %rdi callq 0x5f3e0 movaps 0x20(%rsp), %xmm0 movups %xmm0, 0xb8(%rsp) movq 0x30(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x38(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x40(%rsp), %rax movq %rax, 0xd8(%rsp) movups 0x48(%rsp), %xmm0 movaps %xmm0, 0xe0(%rsp) movl 0x58(%rsp), %eax movl %eax, 0xf0(%rsp) movq 0x60(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %ebx movq 0x260(%r14,%rax), %r12 movq 0x108(%rsp), %rax testq %rax, %rax je 0x28a069 lock decl (%rax) jne 0x28a069 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x28a061 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28a069 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x100(%rsp) andq $0x0, 0x108(%rsp) movq %r15, 0x110(%rsp) movl %ebx, 0x118(%rsp) movq %r12, 0x120(%rsp) movaps 0x16e398(%rip), %xmm0 # 0x3f8430 movups %xmm0, 0x128(%rsp) movl $0x1, 0x138(%rsp) movq $0x1, 0x140(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x28a0e8 lock decl (%rax) jne 0x28a0e8 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x28a0e0 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28a0e8 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x28a1af movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x150(%rsp), %rax testq %rax, %rax je 0x28a15e lock decl (%rax) jne 0x28a15e movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x28a156 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28a15e movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %ebx, 0x160(%rsp) movq %r12, 0x168(%rsp) movaps 0x16e2a3(%rip), %xmm0 # 0x3f8430 movaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0xd8, %r15d movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x28a20d lock decl (%rax) jne 0x28a20d movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x28a205 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28a20d movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x28a1dc movq (%rbp), %rax movq %rbp, %rdi movq 0x250(%rsp), %rsi callq *0x20(%rax) leaq 0x268(%rsp), %rdi movq 0x1e0(%rsp), %rax movq (%rax), %rax movq 0x8(%rsp), %rcx movq %rbp, (%rax,%rcx,8) callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x28a2a0 lock decl (%rax) jne 0x28a2a0 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi je 0x28a298 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28a2a0 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rdx incq %rdx jmp 0x2895ed xorl %eax, %eax addq $0x278, %rsp # imm = 0x278 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x28a375 jmp 0x28a500 jmp 0x28a2cd movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x28a387 lock decl (%rax) jne 0x28a387 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x28a303 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28a387 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28a387 jmp 0x28a500 jmp 0x28a3f9 jmp 0x28a375 jmp 0x28a319 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x28a40b lock decl (%rax) jne 0x28a40b movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x28a34f movq %rsi, %rdi callq 0x5f3e0 jmp 0x28a40b movq (%rdi), %rax callq *0x18(%rax) jmp 0x28a40b jmp 0x28a500 jmp 0x28a500 jmp 0x28a500 jmp 0x28a500 jmp 0x28a375 jmp 0x28a3f9 movq %rax, %rbx jmp 0x28a387 movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0xd8, %r14d xorps %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x28a3c7 lock decl (%rax) jne 0x28a3c7 movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x28a3bc movq (%rdi), %rax callq *0x18(%rax) xorps %xmm0, %xmm0 jmp 0x28a3c7 movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x28a390 jmp 0x28a4b4 jmp 0x28a500 jmp 0x28a3f9 movq %rax, %rbx jmp 0x28a40b movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0x120, %r14d # imm = 0x120 xorps %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x28a44b lock decl (%rax) jne 0x28a44b movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x28a440 movq (%rdi), %rax callq *0x18(%rax) xorps %xmm0, %xmm0 jmp 0x28a44b movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x28a414 jmp 0x28a4b4 jmp 0x28a500 jmp 0x28a4a8 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x28a4f8 lock decl (%rax) jne 0x28a4f8 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x28a4e8 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28a4f8 jmp 0x28a500 movq %rax, %rbx jmp 0x28a4c1 jmp 0x28a500 jmp 0x28a500 movq %rax, %rbx leaq 0x268(%rsp), %rdi callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x28a4f8 lock decl (%rax) jne 0x28a4f8 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi jne 0x28a4f2 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28a4f8 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86::create_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %rax movq (%rdi), %rax addq -0x30(%rax), %rdi callq 0x288da6 xorl %eax, %eax popq %rcx retq nop

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

ncnn::ConvolutionDepthWise_x86::destroy_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }

pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x28a54b movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x28a546 movq (%rdi), %rax callq *0x8(%rax) andq $0x0, 0x8(%rbx) xorl %r15d, %r15d movq 0x10(%rbx), %rax movq 0x18(%rbx), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r15 jge 0x28a58d movq (%rax,%r15,8), %rdi movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x10(%rbx), %rax movq (%rax,%r15,8), %rdi testq %rdi, %rdi je 0x28a588 movq (%rdi), %rax callq *0x8(%rax) incq %r15 jmp 0x28a54e cmpq %rax, %rcx je 0x28a596 movq %rax, 0x18(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %r15 retq

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86::destroy_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }

pushq %rax movq (%rdi), %rax addq -0x38(%rax), %rdi callq 0x28a51a xorl %eax, %eax popq %rcx retq nop

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

ncnn::ConvolutionDepthWise_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float*)bias_data) + g * 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float*)bias_data) + g * 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float*)bias_data) + g * 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x2c8, %rsp # imm = 0x2C8 movq %rcx, %r14 movq (%rdi), %rax movq -0x18(%rax), %rcx cmpb $0x1, 0x1e(%r14) jne 0x28a5f5 cmpl $0x0, 0x10c(%rdi,%rcx) je 0x28a5f5 movq %r14, %rcx addq $0x2c8, %rsp # imm = 0x2C8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x28d5aa movq %rdx, 0x48(%rsp) movl 0x38(%rsi), %edx movq %rdx, 0x50(%rsp) movq 0x10(%rsi), %r13 movslq 0x18(%rsi), %r15 movl 0xd4(%rdi,%rcx), %ebx decl %ebx imull 0xdc(%rdi,%rcx), %ebx movq %rdi, %r8 movl 0xd8(%rdi,%rcx), %ebp decl %ebp imull 0xe0(%rdi,%rcx), %ebp leaq 0xb0(%rsp), %rdx andq $0x0, 0x40(%rdx) xorps %xmm0, %xmm0 movaps %xmm0, (%rdx) movups %xmm0, 0xc(%rdx) movaps %xmm0, 0x20(%rdx) movups %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi movq %r8, 0x88(%rsp) addq %r8, %rdi movq %r14, %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0xb0(%rsp) je 0x28d3c9 movslq 0xe8(%rsp), %rax imulq 0xf0(%rsp), %rax testq %rax, %rax je 0x28d3c9 movq %r15, %r10 notl %ebx movl 0xdc(%rsp), %eax movl %eax, 0x10(%rsp) addl %eax, %ebx movq 0x88(%rsp), %rsi movq (%rsi), %rax movq -0x18(%rax), %rcx movl %ebx, %eax cltd idivl 0xe4(%rsi,%rcx) movl %eax, %edi notl %ebp addl 0xe0(%rsp), %ebp movl %ebp, %eax movq %rsi, %rbp cltd idivl 0xe8(%rsi,%rcx) movq %rdi, 0xa0(%rsp) leal 0x1(%rdi), %esi movq %rax, 0x20(%rsp) leal 0x1(%rax), %r8d movl 0xd0(%rbp,%rcx), %edi testb $0x3, %dil sete %al andb 0x27(%r14), %al movzbl %al, %r15d leal (%r15,%r15,2), %r9d incl %r9d movq %r13, %rax xorl %edx, %edx movq %r10, %rbx divq %r10 movq %rax, %r13 addb %r15b, %r15b movl %r15d, %ecx shlq %cl, %r13 movl %edi, %eax cltd idivl %r9d movq %r14, 0x190(%rsp) movq 0x8(%r14), %rcx movq %rcx, (%rsp) movq 0x48(%rsp), %r14 movq %r14, %rdi movl %esi, 0x60(%rsp) movl %r8d, 0x70(%rsp) movl %r8d, %edx movl %eax, %ecx movq %r13, %r8 movl %r9d, 0x30(%rsp) callq 0x628f2 movq (%r14), %r10 testq %r10, %r10 je 0x28d3c9 movq %r13, %r11 movq 0x40(%r14), %r13 movslq 0x38(%r14), %rax imulq %r13, %rax testq %rax, %rax je 0x28d3c9 movq %rbx, %r9 movl %r9d, %eax imull 0x50(%rsp), %eax movq (%rbp), %rcx movq -0x18(%rcx), %r8 movl 0xd0(%rbp,%r8), %ecx movl 0x108(%rbp,%r8), %esi cmpl %esi, %eax jne 0x28beda cmpl %ecx, %eax jne 0x28beda cmpl $0x1, %r9d je 0x28b2ef movl %eax, %ecx cmpl $0x4, %r9d jne 0x28beda movl 0xd4(%rbp,%r8), %eax movl 0xd8(%rbp,%r8), %r15d cmpl $0x5, %eax je 0x28b672 cmpl $0x3, %eax jne 0x28ccf0 cmpl $0x3, %r15d jne 0x28ccf0 cmpl $0x1, 0xdc(%rbp,%r8) jne 0x28c434 cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28c434 cmpl $0x1, 0xe4(%rbp,%r8) jne 0x28c434 cmpl $0x1, 0xe8(%rbp,%r8) jne 0x28c434 movl 0x2c(%r14), %eax movl 0x30(%r14), %ecx movl 0xe8(%rsp), %edx movq 0x1b0(%rbp,%r8), %rsi xorl %edi, %edi testl %ecx, %ecx cmovlel %edi, %ecx testl %edx, %edx cmovlel %edi, %edx movq %rdx, 0x1e8(%rsp) movl $0x80, %r8d movq %rsi, 0x1e0(%rsp) cmpq 0x1e8(%rsp), %rdi je 0x28d415 testq %rsi, %rsi je 0x28a884 movq %rdi, %rdx shlq $0x4, %rdx movups (%rsi,%rdx), %xmm0 jmp 0x28a887 xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movq 0x40(%r14), %r9 imulq %rdi, %r9 imulq 0x10(%r14), %r9 addq (%r14), %r9 movq 0x28(%rbp), %rdx movslq 0x54(%rbp), %rsi imulq %rdi, %rsi imulq 0x38(%rbp), %rsi movslq 0xdc(%rsp), %r10 movq 0xf0(%rsp), %r13 movq %rdi, %r14 imulq %rdi, %r13 movq 0xc0(%rsp), %r11 imulq %r11, %r13 addq 0xb0(%rsp), %r13 imulq %r11, %r10 leaq (%r10,%r13), %rbp leaq (,%r10,2), %r11 addq %r13, %r11 movaps (%rdx,%rsi), %xmm0 movaps %xmm0, 0x20(%rsp) movaps 0x10(%rdx,%rsi), %xmm0 movaps %xmm0, 0x50(%rsp) movaps 0x20(%rdx,%rsi), %xmm0 movaps %xmm0, 0x110(%rsp) movaps 0x30(%rdx,%rsi), %xmm0 movaps %xmm0, 0x90(%rsp) movaps 0x40(%rdx,%rsi), %xmm0 movaps %xmm0, 0x30(%rsp) movaps 0x50(%rdx,%rsi), %xmm0 movaps %xmm0, 0xa0(%rsp) movaps 0x60(%rdx,%rsi), %xmm0 movaps %xmm0, 0x60(%rsp) movaps 0x70(%rdx,%rsi), %xmm0 movaps %xmm0, 0x100(%rsp) movaps 0x80(%rdx,%rsi), %xmm0 movaps %xmm0, 0x70(%rsp) xorl %r10d, %r10d cmpl %ecx, %r10d je 0x28b2cf leaq 0x20(%r13), %r15 leaq 0x20(%rbp), %rbx leaq 0x20(%r11), %rdx xorl %esi, %esi xorl %r12d, %r12d leal 0x7(%r12), %edi cmpl %eax, %edi jge 0x28b0bf movaps (%r13,%rsi), %xmm0 mulps 0x20(%rsp), %xmm0 addps 0x10(%rsp), %xmm0 movaps 0x10(%r13,%rsi), %xmm3 movaps 0x20(%r13,%rsi), %xmm1 movaps 0x30(%r13,%rsi), %xmm9 movaps %xmm9, 0x180(%rsp) movaps %xmm3, %xmm2 mulps 0x50(%rsp), %xmm2 movaps %xmm1, %xmm4 movaps 0x110(%rsp), %xmm8 mulps %xmm8, %xmm4 addps %xmm2, %xmm4 addps %xmm0, %xmm4 movaps (%rbp,%rsi), %xmm12 movaps 0x90(%rsp), %xmm7 mulps %xmm7, %xmm12 movaps 0x10(%rbp,%rsi), %xmm0 movaps 0x20(%rbp,%rsi), %xmm2 movaps 0x30(%rbp,%rsi), %xmm5 movaps %xmm5, 0x120(%rsp) movaps %xmm0, %xmm13 mulps 0x30(%rsp), %xmm13 addps %xmm12, %xmm13 movaps %xmm2, %xmm14 mulps 0xa0(%rsp), %xmm14 addps %xmm13, %xmm14 addps %xmm4, %xmm14 movaps (%r11,%rsi), %xmm15 movaps 0x60(%rsp), %xmm11 mulps %xmm11, %xmm15 movaps 0x10(%r11,%rsi), %xmm5 movaps 0x20(%r11,%rsi), %xmm4 movaps 0x30(%r11,%rsi), %xmm12 movaps %xmm5, %xmm13 movaps 0x100(%rsp), %xmm10 mulps %xmm10, %xmm13 addps %xmm15, %xmm13 movaps %xmm4, %xmm15 mulps 0x70(%rsp), %xmm15 addps %xmm13, %xmm15 addps %xmm14, %xmm15 movaps %xmm15, (%r9,%rsi) mulps 0x20(%rsp), %xmm3 addps 0x10(%rsp), %xmm3 movaps %xmm1, %xmm13 movaps 0x50(%rsp), %xmm15 mulps %xmm15, %xmm13 movaps %xmm9, %xmm14 mulps %xmm8, %xmm14 mulps %xmm7, %xmm0 addps %xmm13, %xmm0 addps %xmm3, %xmm0 movaps %xmm2, %xmm3 mulps 0x30(%rsp), %xmm3 movaps 0x120(%rsp), %xmm9 movaps %xmm9, %xmm13 movaps 0xa0(%rsp), %xmm6 mulps %xmm6, %xmm13 addps %xmm14, %xmm13 mulps %xmm11, %xmm5 addps %xmm3, %xmm5 movaps %xmm4, %xmm3 mulps %xmm10, %xmm3 addps %xmm5, %xmm3 addps %xmm0, %xmm3 movaps %xmm12, %xmm5 movaps 0x70(%rsp), %xmm10 mulps %xmm10, %xmm5 addps %xmm13, %xmm5 addps %xmm3, %xmm5 movaps 0x40(%r13,%rsi), %xmm0 movaps 0x40(%rbp,%rsi), %xmm3 movaps 0x40(%r11,%rsi), %xmm13 movaps %xmm5, 0x10(%r9,%rsi) mulps 0x20(%rsp), %xmm1 addps 0x10(%rsp), %xmm1 movaps 0x180(%rsp), %xmm11 movaps %xmm11, %xmm5 mulps %xmm15, %xmm5 movaps %xmm0, %xmm14 mulps %xmm8, %xmm14 mulps %xmm7, %xmm2 movaps 0x30(%rsp), %xmm8 mulps %xmm8, %xmm9 addps %xmm5, %xmm9 movaps %xmm3, %xmm5 mulps %xmm6, %xmm5 addps %xmm14, %xmm5 mulps 0x60(%rsp), %xmm4 addps %xmm2, %xmm4 addps %xmm1, %xmm4 movaps %xmm12, %xmm1 movaps 0x100(%rsp), %xmm6 mulps %xmm6, %xmm1 addps %xmm9, %xmm1 addps %xmm4, %xmm1 movaps %xmm13, %xmm2 mulps %xmm10, %xmm2 addps %xmm5, %xmm2 addps %xmm1, %xmm2 movaps 0x50(%r13,%rsi), %xmm4 movaps 0x50(%rbp,%rsi), %xmm1 movaps 0x50(%r11,%rsi), %xmm14 movaps %xmm2, 0x20(%r9,%rsi) movaps %xmm11, %xmm2 mulps 0x20(%rsp), %xmm2 addps 0x10(%rsp), %xmm2 movaps %xmm2, %xmm9 movaps %xmm0, %xmm2 mulps 0x50(%rsp), %xmm2 movaps %xmm4, %xmm5 mulps 0x110(%rsp), %xmm5 movaps 0x120(%rsp), %xmm10 mulps %xmm7, %xmm10 movaps %xmm3, %xmm15 mulps %xmm8, %xmm15 addps %xmm2, %xmm15 movaps %xmm1, %xmm2 movaps %xmm1, %xmm8 mulps 0xa0(%rsp), %xmm2 addps %xmm5, %xmm2 movaps 0x60(%rsp), %xmm7 mulps %xmm7, %xmm12 addps %xmm10, %xmm12 addps %xmm9, %xmm12 movaps %xmm13, %xmm5 mulps %xmm6, %xmm5 addps %xmm15, %xmm5 addps %xmm12, %xmm5 movaps %xmm14, %xmm10 movaps 0x70(%rsp), %xmm6 mulps %xmm6, %xmm10 addps %xmm2, %xmm10 addps %xmm5, %xmm10 movaps 0x60(%r13,%rsi), %xmm15 movaps 0x60(%rbp,%rsi), %xmm1 movaps %xmm1, 0x120(%rsp) movaps 0x60(%r11,%rsi), %xmm12 movaps %xmm10, 0x30(%r9,%rsi) mulps 0x20(%rsp), %xmm0 addps 0x10(%rsp), %xmm0 movaps %xmm4, %xmm5 movaps 0x50(%rsp), %xmm9 mulps %xmm9, %xmm5 movaps %xmm15, %xmm10 movaps 0x110(%rsp), %xmm2 mulps %xmm2, %xmm10 mulps 0x90(%rsp), %xmm3 movaps %xmm8, %xmm11 mulps 0x30(%rsp), %xmm11 addps %xmm5, %xmm11 movaps %xmm1, %xmm5 mulps 0xa0(%rsp), %xmm5 addps %xmm10, %xmm5 mulps %xmm7, %xmm13 addps %xmm3, %xmm13 addps %xmm0, %xmm13 movaps %xmm14, %xmm0 mulps 0x100(%rsp), %xmm0 addps %xmm11, %xmm0 addps %xmm13, %xmm0 movaps %xmm12, %xmm3 mulps %xmm6, %xmm3 addps %xmm5, %xmm3 addps %xmm0, %xmm3 movaps 0x70(%r13,%rsi), %xmm11 movaps 0x70(%rbp,%rsi), %xmm13 movaps 0x70(%r11,%rsi), %xmm10 movaps %xmm3, 0x40(%r9,%rsi) mulps 0x20(%rsp), %xmm4 addps 0x10(%rsp), %xmm4 movaps %xmm15, %xmm0 mulps %xmm9, %xmm0 movaps %xmm9, %xmm7 movaps %xmm11, %xmm3 mulps %xmm2, %xmm3 movaps 0x90(%rsp), %xmm6 mulps %xmm6, %xmm8 movaps 0x120(%rsp), %xmm5 movaps 0x30(%rsp), %xmm2 mulps %xmm2, %xmm5 addps %xmm0, %xmm5 movaps %xmm13, %xmm0 movaps %xmm13, 0x180(%rsp) movaps 0xa0(%rsp), %xmm9 mulps %xmm9, %xmm0 addps %xmm3, %xmm0 mulps 0x60(%rsp), %xmm14 addps %xmm8, %xmm14 addps %xmm4, %xmm14 movaps %xmm12, %xmm1 movaps 0x100(%rsp), %xmm8 mulps %xmm8, %xmm1 addps %xmm5, %xmm1 addps %xmm14, %xmm1 movaps %xmm10, %xmm4 mulps 0x70(%rsp), %xmm4 addps %xmm0, %xmm4 addps %xmm1, %xmm4 movaps 0x80(%r13,%rsi), %xmm3 movaps 0x80(%rbp,%rsi), %xmm1 movaps 0x80(%r11,%rsi), %xmm0 movaps %xmm4, 0x50(%r9,%rsi) movaps %xmm11, %xmm4 mulps %xmm7, %xmm4 movaps %xmm3, %xmm5 movaps 0x110(%rsp), %xmm7 mulps %xmm7, %xmm5 movaps %xmm13, %xmm14 mulps %xmm2, %xmm14 addps %xmm4, %xmm14 movaps %xmm1, %xmm4 mulps %xmm9, %xmm4 addps %xmm5, %xmm4 movaps 0x120(%rsp), %xmm2 mulps %xmm6, %xmm2 movaps 0x60(%rsp), %xmm13 mulps %xmm13, %xmm12 addps %xmm2, %xmm12 mulps 0x20(%rsp), %xmm15 addps 0x10(%rsp), %xmm15 addps %xmm15, %xmm12 movaps %xmm10, %xmm2 mulps %xmm8, %xmm2 addps %xmm14, %xmm2 addps %xmm12, %xmm2 movaps %xmm0, %xmm5 movaps 0x70(%rsp), %xmm6 mulps %xmm6, %xmm5 addps %xmm4, %xmm5 addps %xmm2, %xmm5 movaps 0x90(%r13,%rsi), %xmm2 mulps %xmm7, %xmm2 movaps 0x90(%rbp,%rsi), %xmm4 mulps %xmm9, %xmm4 addps %xmm2, %xmm4 movaps 0x90(%r11,%rsi), %xmm2 mulps %xmm6, %xmm2 addps %xmm4, %xmm2 movaps %xmm5, 0x60(%r9,%rsi) mulps 0x50(%rsp), %xmm3 mulps 0x30(%rsp), %xmm1 addps %xmm3, %xmm1 movaps 0x180(%rsp), %xmm3 mulps 0x90(%rsp), %xmm3 mulps %xmm13, %xmm10 addps %xmm3, %xmm10 mulps 0x20(%rsp), %xmm11 addps 0x10(%rsp), %xmm11 addps %xmm11, %xmm10 mulps %xmm8, %xmm0 addps %xmm1, %xmm0 addps %xmm10, %xmm0 addps %xmm2, %xmm0 movaps %xmm0, 0x70(%r9,%rsi) addl $0x8, %r12d addq %r8, %rsi addq %r8, %r15 addq %r8, %rbx addq %r8, %rdx jmp 0x28a970 movaps (%r13,%rsi), %xmm0 mulps 0x20(%rsp), %xmm0 addps 0x10(%rsp), %xmm0 movaps 0x10(%r13,%rsi), %xmm3 movaps 0x20(%r13,%rsi), %xmm13 movaps 0x30(%r13,%rsi), %xmm11 movaps %xmm3, %xmm1 mulps 0x50(%rsp), %xmm1 movaps %xmm13, %xmm4 movaps 0x110(%rsp), %xmm2 mulps %xmm2, %xmm4 movaps %xmm2, %xmm7 addps %xmm1, %xmm4 addps %xmm0, %xmm4 movaps (%rbp,%rsi), %xmm1 mulps 0x90(%rsp), %xmm1 movaps 0x10(%rbp,%rsi), %xmm2 movaps 0x20(%rbp,%rsi), %xmm8 movaps 0x30(%rbp,%rsi), %xmm9 movaps %xmm2, %xmm5 mulps 0x30(%rsp), %xmm5 addps %xmm1, %xmm5 movaps %xmm8, %xmm14 movaps 0xa0(%rsp), %xmm6 mulps %xmm6, %xmm14 addps %xmm5, %xmm14 addps %xmm4, %xmm14 movaps (%r11,%rsi), %xmm5 mulps 0x60(%rsp), %xmm5 movaps 0x10(%r11,%rsi), %xmm4 movaps 0x20(%r11,%rsi), %xmm1 movaps 0x30(%r11,%rsi), %xmm12 movaps %xmm4, %xmm15 movaps 0x100(%rsp), %xmm10 mulps %xmm10, %xmm15 addps %xmm5, %xmm15 movaps %xmm1, %xmm5 movaps 0x70(%rsp), %xmm0 mulps %xmm0, %xmm5 addps %xmm15, %xmm5 addps %xmm14, %xmm5 movaps %xmm5, (%r9,%rsi) mulps 0x20(%rsp), %xmm3 addps 0x10(%rsp), %xmm3 movaps %xmm13, %xmm5 movaps 0x50(%rsp), %xmm15 mulps %xmm15, %xmm5 movaps %xmm11, %xmm14 mulps %xmm7, %xmm14 movaps 0x90(%rsp), %xmm7 mulps %xmm7, %xmm2 addps %xmm5, %xmm2 addps %xmm3, %xmm2 movaps %xmm8, %xmm3 movaps 0x30(%rsp), %xmm0 mulps %xmm0, %xmm3 movaps %xmm9, %xmm5 movaps %xmm9, 0x120(%rsp) mulps %xmm6, %xmm5 addps %xmm14, %xmm5 movaps 0x60(%rsp), %xmm6 mulps %xmm6, %xmm4 addps %xmm3, %xmm4 movaps %xmm1, %xmm3 mulps %xmm10, %xmm3 addps %xmm4, %xmm3 addps %xmm2, %xmm3 movaps %xmm12, %xmm14 mulps 0x70(%rsp), %xmm14 addps %xmm5, %xmm14 addps %xmm3, %xmm14 movaps 0x40(%r13,%rsi), %xmm4 movaps 0x40(%rbp,%rsi), %xmm3 movaps 0x40(%r11,%rsi), %xmm2 movaps %xmm14, 0x10(%r9,%rsi) movaps %xmm11, %xmm5 mulps %xmm15, %xmm5 movaps %xmm4, %xmm14 movaps 0x110(%rsp), %xmm10 mulps %xmm10, %xmm14 movaps %xmm9, %xmm15 mulps %xmm0, %xmm15 addps %xmm5, %xmm15 movaps %xmm3, %xmm5 movaps 0xa0(%rsp), %xmm9 mulps %xmm9, %xmm5 addps %xmm14, %xmm5 movaps 0x10(%rsp), %xmm14 mulps %xmm7, %xmm8 mulps %xmm6, %xmm1 addps %xmm8, %xmm1 mulps 0x20(%rsp), %xmm13 addps %xmm14, %xmm13 addps %xmm13, %xmm1 movaps %xmm12, %xmm0 movaps 0x100(%rsp), %xmm6 mulps %xmm6, %xmm0 addps %xmm15, %xmm0 addps %xmm1, %xmm0 movaps %xmm2, %xmm1 movaps 0x70(%rsp), %xmm7 mulps %xmm7, %xmm1 addps %xmm5, %xmm1 addps %xmm0, %xmm1 movaps 0x50(%r13,%rsi), %xmm0 mulps %xmm10, %xmm0 movaps 0x50(%rbp,%rsi), %xmm5 mulps %xmm9, %xmm5 addps %xmm0, %xmm5 movaps 0x50(%r11,%rsi), %xmm0 mulps %xmm7, %xmm0 addps %xmm5, %xmm0 movaps %xmm1, 0x20(%r9,%rsi) mulps 0x50(%rsp), %xmm4 mulps 0x30(%rsp), %xmm3 addps %xmm4, %xmm3 movaps 0x120(%rsp), %xmm1 mulps 0x90(%rsp), %xmm1 mulps 0x60(%rsp), %xmm12 addps %xmm1, %xmm12 mulps 0x20(%rsp), %xmm11 movaps %xmm14, 0x10(%rsp) addps %xmm14, %xmm11 addps %xmm11, %xmm12 mulps %xmm6, %xmm2 addps %xmm3, %xmm2 addps %xmm12, %xmm2 addps %xmm0, %xmm2 movaps %xmm2, 0x30(%r9,%rsi) addl $0x4, %r12d addq $0x40, %rsi addq $0x40, %r15 addq $0x40, %rbx addq $0x40, %rdx leal 0x3(%r12), %edi cmpl %eax, %edi jl 0x28ae58 jmp 0x28b20f movaps 0x10(%r13,%rsi), %xmm4 movaps 0x20(%r13,%rsi), %xmm10 movaps 0x10(%rbp,%rsi), %xmm0 movaps 0x20(%rbp,%rsi), %xmm6 movaps 0x10(%r11,%rsi), %xmm3 movaps 0x20(%r11,%rsi), %xmm1 movaps (%r13,%rsi), %xmm5 mulps 0x20(%rsp), %xmm5 addps 0x10(%rsp), %xmm5 movaps %xmm4, %xmm11 mulps 0x50(%rsp), %xmm11 movaps %xmm10, %xmm12 movaps 0x110(%rsp), %xmm7 mulps %xmm7, %xmm12 addps %xmm11, %xmm12 addps %xmm5, %xmm12 movaps (%rbp,%rsi), %xmm5 movaps 0x90(%rsp), %xmm14 mulps %xmm14, %xmm5 movaps %xmm0, %xmm11 mulps 0x30(%rsp), %xmm11 addps %xmm5, %xmm11 movaps %xmm6, %xmm5 movaps 0xa0(%rsp), %xmm15 mulps %xmm15, %xmm5 addps %xmm11, %xmm5 addps %xmm12, %xmm5 movaps (%r11,%rsi), %xmm11 movaps 0x60(%rsp), %xmm8 mulps %xmm8, %xmm11 movaps %xmm3, %xmm12 movaps 0x100(%rsp), %xmm9 mulps %xmm9, %xmm12 addps %xmm11, %xmm12 movaps %xmm1, %xmm13 movaps 0x70(%rsp), %xmm2 mulps %xmm2, %xmm13 addps %xmm12, %xmm13 addps %xmm5, %xmm13 movaps 0x30(%r13,%rsi), %xmm5 mulps %xmm7, %xmm5 movaps 0x30(%rbp,%rsi), %xmm12 mulps %xmm15, %xmm12 addps %xmm5, %xmm12 movaps 0x30(%r11,%rsi), %xmm11 mulps %xmm2, %xmm11 addps %xmm12, %xmm11 movaps %xmm13, (%r9,%rsi) mulps 0x50(%rsp), %xmm10 mulps %xmm14, %xmm0 addps %xmm10, %xmm0 mulps 0x20(%rsp), %xmm4 addps 0x10(%rsp), %xmm4 addps %xmm4, %xmm0 mulps 0x30(%rsp), %xmm6 mulps %xmm8, %xmm3 addps %xmm6, %xmm3 mulps %xmm9, %xmm1 addps %xmm3, %xmm1 addps %xmm0, %xmm1 addps %xmm11, %xmm1 movaps %xmm1, 0x10(%r9,%rsi) addl $0x2, %r12d addq $0x20, %rsi addq $0x20, %r15 addq $0x20, %rbx addq $0x20, %rdx leal 0x1(%r12), %edi cmpl %eax, %edi jl 0x28b0d1 addq %rsi, %r9 cmpl %eax, %r12d jge 0x28b2be movaps -0x20(%r15), %xmm0 mulps 0x20(%rsp), %xmm0 addps 0x10(%rsp), %xmm0 movaps -0x10(%r15), %xmm1 mulps 0x50(%rsp), %xmm1 movaps (%r15), %xmm2 mulps 0x110(%rsp), %xmm2 addps %xmm1, %xmm2 addps %xmm0, %xmm2 movaps -0x20(%rbx), %xmm0 mulps 0x90(%rsp), %xmm0 movaps -0x10(%rbx), %xmm1 mulps 0x30(%rsp), %xmm1 addps %xmm0, %xmm1 movaps (%rbx), %xmm0 mulps 0xa0(%rsp), %xmm0 addps %xmm1, %xmm0 addps %xmm2, %xmm0 movaps -0x20(%rdx), %xmm1 mulps 0x60(%rsp), %xmm1 movaps -0x10(%rdx), %xmm2 mulps 0x100(%rsp), %xmm2 addps %xmm1, %xmm2 movaps (%rdx), %xmm1 mulps 0x70(%rsp), %xmm1 addps %xmm2, %xmm1 addps %xmm0, %xmm1 movaps %xmm1, (%r9) addq $0x10, %r9 incl %r12d addq $0x10, %r15 addq $0x10, %rbx addq $0x10, %rdx jmp 0x28b21f incl %r10d movq %r15, %r13 movq %rbx, %rbp movq %rdx, %r11 jmp 0x28a956 movq %r14, %rdi incq %rdi movq 0x48(%rsp), %r14 movq 0x88(%rsp), %rbp movq 0x1e0(%rsp), %rsi jmp 0x28a864 cmpl $0x3, 0xd4(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x3, 0xd8(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x1, 0xdc(%rbp,%r8) movq %r10, 0x20(%rsp) jne 0x28bd03 cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28bd03 cmpl $0x1, 0xe4(%rbp,%r8) jne 0x28bd03 cmpl $0x1, 0xe8(%rbp,%r8) jne 0x28bd03 movslq 0xdc(%rsp), %rax movl 0x2c(%r14), %edx movl 0x30(%r14), %edi movl 0xe8(%rsp), %ecx movq 0x28(%rbp), %r9 imulq 0x10(%r14), %r13 movq 0x1b0(%rbp,%r8), %rsi movslq %edx, %r8 movq 0xb0(%rsp), %r11 movq %r11, 0x110(%rsp) movq 0xc0(%rsp), %r11 imulq 0xf0(%rsp), %r11 movq %r11, 0x100(%rsp) leaq (%rax,%rax), %r11 movq %r11, 0x120(%rsp) leaq (%rax,%rax,2), %r11 movq %r11, 0x180(%rsp) movq %rax, 0x70(%rsp) addl $0x2, %eax movslq %eax, %rbp xorl %eax, %eax testl %ecx, %ecx cmovlel %eax, %ecx movq %rcx, 0x50(%rsp) shlq $0x2, %rbp movq %r8, 0xa0(%rsp) leaq (,%r8,4), %rcx movq %rcx, 0x10(%rsp) movq %rsi, 0x90(%rsp) movq %r13, 0x60(%rsp) cmpq 0x50(%rsp), %rax je 0x28d41b testq %rsi, %rsi je 0x28b412 movss (%rsi,%rax,4), %xmm0 jmp 0x28b415 xorps %xmm0, %xmm0 movq 0xa0(%rsp), %r14 movq %r13, %rsi imulq %rax, %rsi addq %r10, %rsi imulq $0x24, %rax, %r13 movq 0x100(%rsp), %r15 movq %rax, 0x30(%rsp) imulq %rax, %r15 addq 0x110(%rsp), %r15 movq 0x70(%rsp), %rax leaq (%r15,%rax,4), %rax movq 0x120(%rsp), %rcx leaq (%r15,%rcx,4), %r8 movq 0x180(%rsp), %rcx leaq (%r15,%rcx,4), %r11 xorl %r10d, %r10d movq %rsi, %rbx movl %r10d, %ecx orl $0x1, %ecx cmpl %edi, %ecx jge 0x28b645 leaq (%rsi,%r14,4), %rsi xorl %r12d, %r12d movl %edx, %ecx testl %ecx, %ecx jle 0x28b582 movsd 0x4(%r15,%r12), %xmm3 movsd 0x4(%rax,%r12), %xmm6 movsd (%r8,%r12), %xmm5 movups (%r9,%r13), %xmm1 movups 0x10(%r9,%r13), %xmm2 movss (%r15,%r12), %xmm4 unpcklps %xmm3, %xmm4 # xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1] movss (%rax,%r12), %xmm7 movaps %xmm7, %xmm8 shufps $0xd4, %xmm3, %xmm8 # xmm8 = xmm8[0,1],xmm3[1,3] shufps $0x24, %xmm8, %xmm4 # xmm4 = xmm4[0,1],xmm8[2,0] movaps %xmm6, %xmm9 movlhps %xmm5, %xmm9 # xmm9 = xmm9[0],xmm5[0] mulps %xmm1, %xmm4 mulps %xmm2, %xmm9 addps %xmm4, %xmm9 movss 0x8(%r8,%r12), %xmm8 movss 0x20(%r9,%r13), %xmm3 movaps %xmm3, %xmm4 mulss %xmm8, %xmm4 movaps %xmm9, %xmm10 unpckhpd %xmm9, %xmm10 # xmm10 = xmm10[1],xmm9[1] addps %xmm9, %xmm10 movaps %xmm10, %xmm9 shufps $0x55, %xmm10, %xmm9 # xmm9 = xmm9[1,1],xmm10[1,1] addss %xmm10, %xmm9 addss %xmm0, %xmm4 addss %xmm9, %xmm4 movsd (%r11,%r12), %xmm9 movaps %xmm1, %xmm10 unpcklps %xmm6, %xmm10 # xmm10 = xmm10[0],xmm6[0],xmm10[1],xmm6[1] shufps $0xd4, %xmm5, %xmm8 # xmm8 = xmm8[0,1],xmm5[1,3] shufps $0xd4, %xmm6, %xmm5 # xmm5 = xmm5[0,1],xmm6[1,3] shufps $0x24, %xmm5, %xmm10 # xmm10 = xmm10[0,1],xmm5[2,0] shufps $0x42, %xmm9, %xmm8 # xmm8 = xmm8[2,0],xmm9[0,1] movss %xmm7, %xmm1 # xmm1 = xmm7[0],xmm1[1,2,3] mulps %xmm10, %xmm1 mulps %xmm2, %xmm8 addps %xmm1, %xmm8 mulss 0x8(%r11,%r12), %xmm3 movaps %xmm8, %xmm1 unpckhpd %xmm8, %xmm1 # xmm1 = xmm1[1],xmm8[1] addps %xmm8, %xmm1 movaps %xmm1, %xmm2 shufps $0x55, %xmm1, %xmm2 # xmm2 = xmm2[1,1],xmm1[1,1] addss %xmm1, %xmm2 addss %xmm0, %xmm3 addss %xmm2, %xmm3 movss %xmm4, (%rbx,%r12) movss %xmm3, (%rsi,%r12) decl %ecx addq $0x4, %r12 jmp 0x28b482 addq %rbp, %r15 addq %r12, %r15 addq %rbp, %rax addq %r12, %rax addq %rbp, %r8 addq %r12, %r8 addq %rbp, %r11 addq %r12, %r11 addq 0x10(%rsp), %rbx addq %r12, %rbx addl $0x2, %r10d addq %r12, %rsi jmp 0x28b46b xorl %ecx, %ecx movl %edx, %esi testl %esi, %esi jle 0x28b62a movsd 0x4(%r15,%rcx), %xmm1 movups (%r9,%r13), %xmm2 movups 0x10(%r9,%r13), %xmm3 movss (%r15,%rcx), %xmm4 unpcklps %xmm1, %xmm4 # xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1] movss (%rax,%rcx), %xmm5 shufps $0xd4, %xmm1, %xmm5 # xmm5 = xmm5[0,1],xmm1[1,3] shufps $0x24, %xmm5, %xmm4 # xmm4 = xmm4[0,1],xmm5[2,0] mulps %xmm2, %xmm4 movsd 0x4(%rax,%rcx), %xmm1 movhps (%r8,%rcx), %xmm1 # xmm1 = xmm1[0,1],mem[0,1] mulps %xmm3, %xmm1 movss 0x20(%r9,%r13), %xmm2 mulss 0x8(%r8,%rcx), %xmm2 addps %xmm4, %xmm1 movaps %xmm1, %xmm3 unpckhpd %xmm1, %xmm3 # xmm3 = xmm3[1],xmm1[1] addps %xmm1, %xmm3 movaps %xmm3, %xmm1 shufps $0x55, %xmm3, %xmm1 # xmm1 = xmm1[1,1],xmm3[1,1] addss %xmm3, %xmm1 addss %xmm0, %xmm2 addss %xmm1, %xmm2 movss %xmm2, (%rbx,%rcx) decl %esi addq $0x4, %rcx jmp 0x28b5b2 addq %rcx, %r15 addq $0x8, %r15 addq %rcx, %rax addq $0x8, %rax addq %rcx, %r8 addq $0x8, %r8 incl %r10d addq %rcx, %rbx cmpl %edi, %r10d jl 0x28b5ae movq 0x30(%rsp), %rax incq %rax movq 0x48(%rsp), %r14 movq 0x20(%rsp), %r10 movq 0x60(%rsp), %r13 movq 0x90(%rsp), %rsi jmp 0x28b3fb cmpl $0x5, %r15d jne 0x28ccf0 cmpl $0x1, 0xdc(%rbp,%r8) jne 0x28c9de cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28c9de cmpl $0x1, 0xe4(%rbp,%r8) jne 0x28c9de cmpl $0x1, 0xe8(%rbp,%r8) jne 0x28c9de movl 0x2c(%r14), %ecx movl 0x30(%r14), %edx movl 0xdc(%rsp), %eax movl 0xe8(%rsp), %esi movq 0x1b0(%rbp,%r8), %r9 movl %edx, %r8d leal 0x10(,%rax,4), %eax movslq %eax, %rdi leal (,%rcx,4), %eax cltq xorl %edx, %edx testl %ecx, %ecx cmovlel %edx, %ecx movq %rcx, 0x10(%rsp) testl %esi, %esi cmovlel %edx, %esi movq %rsi, 0x50(%rsp) shlq $0x2, %rax movq %rax, 0x20(%rsp) shlq $0x2, %rdi movq %r9, 0x60(%rsp) cmpq 0x50(%rsp), %rdx je 0x28d415 testq %r9, %r9 je 0x28b732 movq %rdx, %rax shlq $0x4, %rax movups (%r9,%rax), %xmm0 jmp 0x28b735 xorps %xmm0, %xmm0 movq 0x40(%r14), %r10 imulq %rdx, %r10 movq 0x10(%r14), %rax imulq %rax, %r10 addq (%r14), %r10 movslq 0x2c(%r14), %r9 movq 0x28(%rbp), %r11 movslq 0x54(%rbp), %rbx imulq %rdx, %rbx imulq 0x38(%rbp), %rbx imulq %rax, %r9 addq %r10, %r9 movslq 0xdc(%rsp), %rax movq 0xf0(%rsp), %r15 movq %rdx, 0x30(%rsp) imulq %rdx, %r15 movq 0xc0(%rsp), %rcx imulq %rcx, %r15 addq 0xb0(%rsp), %r15 imulq %rcx, %rax leaq (%r15,%rax), %r12 leaq (%r15,%rax,2), %r13 leaq (%rax,%rax,2), %rbp addq %r15, %rbp leaq (%r15,%rax,4), %rdx leaq (%rax,%rax,4), %rax addq %r15, %rax xorl %esi, %esi movl %esi, %ecx orl $0x1, %ecx cmpl %r8d, %ecx jge 0x28baf4 movl %r8d, %r14d movq 0x10(%rsp), %rcx xorl %r8d, %r8d subl $0x1, %ecx jb 0x28bab4 movaps (%r11,%rbx), %xmm6 movaps 0x10(%r11,%rbx), %xmm8 movaps 0x20(%r11,%rbx), %xmm2 movaps 0x30(%r11,%rbx), %xmm3 movaps 0x40(%r11,%rbx), %xmm5 movaps (%r15,%r8), %xmm1 mulps %xmm6, %xmm1 addps %xmm0, %xmm1 movaps 0x10(%r15,%r8), %xmm4 mulps %xmm8, %xmm4 movaps 0x20(%r15,%r8), %xmm7 mulps %xmm2, %xmm7 addps %xmm4, %xmm7 addps %xmm1, %xmm7 movaps 0x30(%r15,%r8), %xmm1 mulps %xmm3, %xmm1 movaps 0x40(%r15,%r8), %xmm14 mulps %xmm5, %xmm14 addps %xmm1, %xmm14 movaps (%r12,%r8), %xmm1 movaps 0x10(%r12,%r8), %xmm13 movaps 0x20(%r12,%r8), %xmm12 movaps 0x30(%r12,%r8), %xmm10 movaps 0x40(%r12,%r8), %xmm4 mulps %xmm1, %xmm6 addps %xmm0, %xmm6 mulps %xmm13, %xmm8 mulps %xmm12, %xmm2 addps %xmm8, %xmm2 addps %xmm6, %xmm2 mulps %xmm10, %xmm3 mulps %xmm4, %xmm5 addps %xmm3, %xmm5 movaps 0x50(%r11,%rbx), %xmm6 movaps 0x60(%r11,%rbx), %xmm11 movaps 0x70(%r11,%rbx), %xmm9 movaps 0x80(%r11,%rbx), %xmm8 movaps 0x90(%r11,%rbx), %xmm3 mulps %xmm6, %xmm1 addps %xmm14, %xmm1 addps %xmm7, %xmm1 mulps %xmm11, %xmm13 mulps %xmm9, %xmm12 addps %xmm13, %xmm12 mulps %xmm8, %xmm10 addps %xmm12, %xmm10 mulps %xmm3, %xmm4 addps %xmm10, %xmm4 addps %xmm1, %xmm4 movaps (%r13,%r8), %xmm13 movaps 0x10(%r13,%r8), %xmm12 movaps 0x20(%r13,%r8), %xmm10 movaps 0x30(%r13,%r8), %xmm7 movaps 0x40(%r13,%r8), %xmm1 mulps %xmm13, %xmm6 addps %xmm5, %xmm6 addps %xmm2, %xmm6 mulps %xmm12, %xmm11 mulps %xmm10, %xmm9 addps %xmm11, %xmm9 mulps %xmm7, %xmm8 addps %xmm9, %xmm8 mulps %xmm1, %xmm3 addps %xmm8, %xmm3 addps %xmm6, %xmm3 movaps 0xa0(%r11,%rbx), %xmm14 movaps 0xb0(%r11,%rbx), %xmm11 movaps 0xc0(%r11,%rbx), %xmm8 movaps 0xd0(%r11,%rbx), %xmm5 movaps 0xe0(%r11,%rbx), %xmm2 mulps %xmm14, %xmm13 mulps %xmm11, %xmm12 addps %xmm13, %xmm12 mulps %xmm8, %xmm10 addps %xmm12, %xmm10 mulps %xmm5, %xmm7 addps %xmm10, %xmm7 mulps %xmm2, %xmm1 addps %xmm7, %xmm1 addps %xmm4, %xmm1 movaps (%rbp,%r8), %xmm13 movaps 0x10(%rbp,%r8), %xmm10 movaps 0x20(%rbp,%r8), %xmm9 movaps 0x30(%rbp,%r8), %xmm6 movaps 0x40(%rbp,%r8), %xmm4 mulps %xmm13, %xmm14 mulps %xmm10, %xmm11 addps %xmm14, %xmm11 mulps %xmm9, %xmm8 addps %xmm11, %xmm8 mulps %xmm6, %xmm5 addps %xmm8, %xmm5 mulps %xmm4, %xmm2 addps %xmm5, %xmm2 addps %xmm3, %xmm2 movaps 0xf0(%r11,%rbx), %xmm14 movaps 0x100(%r11,%rbx), %xmm12 movaps 0x110(%r11,%rbx), %xmm8 movaps 0x120(%r11,%rbx), %xmm7 movaps 0x130(%r11,%rbx), %xmm3 mulps %xmm14, %xmm13 mulps %xmm12, %xmm10 addps %xmm13, %xmm10 mulps %xmm8, %xmm9 addps %xmm10, %xmm9 mulps %xmm7, %xmm6 addps %xmm9, %xmm6 mulps %xmm3, %xmm4 addps %xmm6, %xmm4 movaps (%rdx,%r8), %xmm6 movaps 0x10(%rdx,%r8), %xmm13 movaps 0x20(%rdx,%r8), %xmm11 movaps 0x30(%rdx,%r8), %xmm9 movaps 0x40(%rdx,%r8), %xmm5 mulps %xmm6, %xmm14 mulps %xmm13, %xmm12 addps %xmm14, %xmm12 mulps %xmm11, %xmm8 addps %xmm12, %xmm8 mulps %xmm9, %xmm7 addps %xmm8, %xmm7 mulps %xmm5, %xmm3 addps %xmm7, %xmm3 movaps 0x140(%r11,%rbx), %xmm7 movaps 0x150(%r11,%rbx), %xmm14 movaps 0x160(%r11,%rbx), %xmm12 movaps 0x170(%r11,%rbx), %xmm10 movaps 0x180(%r11,%rbx), %xmm8 mulps %xmm7, %xmm6 addps %xmm4, %xmm6 addps %xmm1, %xmm6 mulps %xmm14, %xmm13 mulps %xmm12, %xmm11 addps %xmm13, %xmm11 mulps %xmm10, %xmm9 addps %xmm11, %xmm9 mulps %xmm8, %xmm5 addps %xmm9, %xmm5 addps %xmm6, %xmm5 mulps (%rax,%r8), %xmm7 addps %xmm3, %xmm7 mulps 0x10(%rax,%r8), %xmm14 mulps 0x20(%rax,%r8), %xmm12 addps %xmm2, %xmm7 addps %xmm14, %xmm12 mulps 0x30(%rax,%r8), %xmm10 addps %xmm12, %xmm10 mulps 0x40(%rax,%r8), %xmm8 addps %xmm10, %xmm8 addps %xmm7, %xmm8 movaps %xmm5, (%r10,%r8) movaps %xmm8, (%r9,%r8) addq $0x10, %r8 jmp 0x28b7ca addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp addq %rdi, %rdx addq %r8, %rdx addq %rdi, %rax addq %r8, %rax movq 0x20(%rsp), %rcx addq %rcx, %r10 addq %r8, %r10 addq %rcx, %r9 addq %r8, %r9 addl $0x2, %esi movl %r14d, %r8d jmp 0x28b7b1 movq 0x60(%rsp), %r9 cmpl %r8d, %esi jge 0x28bce9 movq 0x10(%rsp), %rax movl %eax, %ecx xorl %eax, %eax subl $0x1, %ecx jb 0x28bcbe movaps (%r11,%rbx), %xmm1 movaps 0x10(%r11,%rbx), %xmm2 movaps 0x20(%r11,%rbx), %xmm3 movaps 0x30(%r11,%rbx), %xmm4 mulps (%r15,%rax), %xmm1 movaps 0x40(%r11,%rbx), %xmm5 addps %xmm0, %xmm1 mulps 0x10(%r15,%rax), %xmm2 mulps 0x20(%r15,%rax), %xmm3 addps %xmm2, %xmm3 addps %xmm1, %xmm3 mulps 0x30(%r15,%rax), %xmm4 mulps 0x40(%r15,%rax), %xmm5 addps %xmm4, %xmm5 movaps 0x50(%r11,%rbx), %xmm1 movaps 0x60(%r11,%rbx), %xmm4 movaps 0x70(%r11,%rbx), %xmm6 movaps 0x80(%r11,%rbx), %xmm7 movaps 0x90(%r11,%rbx), %xmm2 mulps (%r12,%rax), %xmm1 addps %xmm5, %xmm1 addps %xmm3, %xmm1 mulps 0x10(%r12,%rax), %xmm4 mulps 0x20(%r12,%rax), %xmm6 mulps 0x30(%r12,%rax), %xmm7 addps %xmm4, %xmm6 addps %xmm6, %xmm7 mulps 0x40(%r12,%rax), %xmm2 addps %xmm7, %xmm2 addps %xmm1, %xmm2 movaps 0xa0(%r11,%rbx), %xmm3 movaps 0xb0(%r11,%rbx), %xmm4 movaps 0xc0(%r11,%rbx), %xmm5 movaps 0xd0(%r11,%rbx), %xmm6 mulps (%r13,%rax), %xmm3 movaps 0xe0(%r11,%rbx), %xmm1 mulps 0x10(%r13,%rax), %xmm4 addps %xmm3, %xmm4 mulps 0x20(%r13,%rax), %xmm5 mulps 0x30(%r13,%rax), %xmm6 addps %xmm4, %xmm5 addps %xmm5, %xmm6 mulps 0x40(%r13,%rax), %xmm1 addps %xmm6, %xmm1 addps %xmm2, %xmm1 movaps 0xf0(%r11,%rbx), %xmm2 movaps 0x100(%r11,%rbx), %xmm3 movaps 0x110(%r11,%rbx), %xmm4 movaps 0x120(%r11,%rbx), %xmm5 mulps (%rbp,%rax), %xmm2 movaps 0x130(%r11,%rbx), %xmm6 mulps 0x10(%rbp,%rax), %xmm3 addps %xmm2, %xmm3 mulps 0x20(%rbp,%rax), %xmm4 mulps 0x30(%rbp,%rax), %xmm5 addps %xmm3, %xmm4 addps %xmm4, %xmm5 mulps 0x40(%rbp,%rax), %xmm6 addps %xmm5, %xmm6 movaps 0x140(%r11,%rbx), %xmm2 movaps 0x150(%r11,%rbx), %xmm3 movaps 0x160(%r11,%rbx), %xmm4 movaps 0x170(%r11,%rbx), %xmm5 movaps 0x180(%r11,%rbx), %xmm7 mulps (%rdx,%rax), %xmm2 addps %xmm6, %xmm2 addps %xmm1, %xmm2 mulps 0x10(%rdx,%rax), %xmm3 mulps 0x20(%rdx,%rax), %xmm4 mulps 0x30(%rdx,%rax), %xmm5 addps %xmm3, %xmm4 addps %xmm4, %xmm5 mulps 0x40(%rdx,%rax), %xmm7 addps %xmm5, %xmm7 addps %xmm2, %xmm7 movaps %xmm7, (%r10,%rax) addq $0x10, %rax jmp 0x28bb0b addq %rax, %r15 addq $0x40, %r15 leaq 0x40(%r12,%rax), %r12 addq %rax, %r13 addq $0x40, %r13 addq %rax, %rbp addq $0x40, %rbp addq %rax, %rdx addq $0x40, %rdx incl %esi addq %rax, %r10 jmp 0x28baf9 movq 0x30(%rsp), %rdx incq %rdx movq 0x48(%rsp), %r14 movq 0x88(%rsp), %rbp jmp 0x28b714 cmpl $0x1, 0xdc(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x1, 0xe0(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x2, 0xe4(%rbp,%r8) movl %eax, %ecx jne 0x28beda cmpl $0x2, 0xe8(%rbp,%r8) movl %eax, %ecx jne 0x28beda movslq 0xdc(%rsp), %rdx movl 0x2c(%r14), %eax movl 0x30(%r14), %edi movl 0xe8(%rsp), %esi movl %edx, %ecx subl %eax, %ecx addl %ecx, %ecx movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r8), %r8 imulq 0x10(%r14), %r13 movq 0xb0(%rsp), %r11 movq %r11, 0x60(%rsp) movq 0xc0(%rsp), %r11 imulq 0xf0(%rsp), %r11 movq %r11, 0x70(%rsp) movq %rdx, 0x30(%rsp) addq %rdx, %rdx movq %rdx, 0x90(%rsp) movslq %ecx, %r15 xorl %r12d, %r12d testl %edi, %edi cmovlel %r12d, %edi testl %esi, %esi cmovlel %r12d, %esi movq %rsi, 0x10(%rsp) shlq $0x2, %r15 movq %r8, 0x50(%rsp) cmpq 0x10(%rsp), %r12 je 0x28d415 testq %r8, %r8 je 0x28bde0 movss (%r8,%r12,4), %xmm0 jmp 0x28bde3 xorps %xmm0, %xmm0 movq %r13, %r8 imulq %r12, %r13 addq %r10, %r13 imulq $0x24, %r12, %rbp movq 0x70(%rsp), %r11 imulq %r12, %r11 addq 0x60(%rsp), %r11 movq 0x30(%rsp), %rcx leaq (%r11,%rcx,4), %r10 movq 0x90(%rsp), %rcx leaq (%r11,%rcx,4), %rbx xorl %esi, %esi cmpl %edi, %esi je 0x28beb8 xorl %edx, %edx movl %eax, %ecx testl %ecx, %ecx jle 0x28be9f movsd 0x4(%r11,%rdx), %xmm1 movups (%r9,%rbp), %xmm2 movups 0x10(%r9,%rbp), %xmm3 movss (%r11,%rdx), %xmm4 unpcklps %xmm1, %xmm4 # xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1] movss (%r10,%rdx), %xmm5 shufps $0xd4, %xmm1, %xmm5 # xmm5 = xmm5[0,1],xmm1[1,3] shufps $0x24, %xmm5, %xmm4 # xmm4 = xmm4[0,1],xmm5[2,0] mulps %xmm2, %xmm4 movsd 0x4(%r10,%rdx), %xmm1 movhps (%rbx,%rdx), %xmm1 # xmm1 = xmm1[0,1],mem[0,1] mulps %xmm3, %xmm1 addps %xmm4, %xmm1 movss 0x20(%r9,%rbp), %xmm2 mulss 0x8(%rbx,%rdx), %xmm2 addss %xmm0, %xmm2 movaps %xmm1, %xmm3 unpckhpd %xmm1, %xmm3 # xmm3 = xmm3[1],xmm1[1] addps %xmm1, %xmm3 movaps %xmm3, %xmm1 shufps $0x55, %xmm3, %xmm1 # xmm1 = xmm1[1,1],xmm3[1,1] addss %xmm3, %xmm1 addss %xmm2, %xmm1 movss %xmm1, (%r13) addq $0x4, %r13 decl %ecx addq $0x8, %rdx jmp 0x28be22 addq %r15, %r11 addq %rdx, %r11 addq %r15, %r10 addq %rdx, %r10 addq %r15, %rbx addq %rdx, %rbx incl %esi jmp 0x28be16 incq %r12 movq 0x48(%rsp), %r14 movq 0x88(%rsp), %rbp movq 0x20(%rsp), %r10 movq %r8, %r13 movq 0x50(%rsp), %r8 jmp 0x28bdc8 cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %r10d movq 0x190(%rsp), %r14 cmpb $0x1, 0x27(%r14) jne 0x28bf18 xorl %eax, %eax testb $0x3, %dil sete %al leal (%rax,%rax,2), %eax incl %eax movl %eax, 0x20(%rsp) xorl %eax, %eax testb $0x3, %r10b sete %al leal (%rax,%rax,2), %eax incl %eax jmp 0x28bf1f pushq $0x1 popq %rax movl %eax, 0x20(%rsp) movl %eax, 0x10(%rsp) movl 0x30(%rsp), %ebx movq %r11, %r13 movq 0xb0(%rsp), %rcx movq 0xb8(%rsp), %rax movq %rcx, 0x198(%rsp) movq %rax, 0x1a0(%rsp) movq 0xc0(%rsp), %rcx movq %rcx, 0x1a8(%rsp) movl 0xc8(%rsp), %ecx movl %ecx, 0x1b0(%rsp) movq 0xd0(%rsp), %rcx movq %rcx, 0x1b8(%rsp) movups 0xd8(%rsp), %xmm0 movups %xmm0, 0x1c0(%rsp) movl 0xe8(%rsp), %ecx movl %ecx, 0x1d0(%rsp) movq 0xf0(%rsp), %rcx movq %rcx, 0x1d8(%rsp) testq %rax, %rax je 0x28bfae lock incl (%rax) cmpl 0x20(%rsp), %r9d jle 0x28c00a movl %r10d, %r12d movl %edi, %ebp movups (%r14), %xmm0 movups 0x10(%r14), %xmm1 movups 0x20(%r14), %xmm2 movups 0x30(%r14), %xmm3 leaq 0x130(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x10(%r14), %rax movq %rax, 0x8(%rcx) leaq 0xb0(%rsp), %rdi leaq 0x198(%rsp), %rsi movl 0x20(%rsp), %edx callq 0x64e3b movl %ebp, %edi movl %r12d, %r10d movq 0x48(%rsp), %rdx movq (%rdx), %rax movq %rax, 0x130(%rsp) movq 0x8(%rdx), %rax movq %rax, 0x138(%rsp) movq 0x10(%rdx), %rcx movq %rcx, 0x140(%rsp) movl 0x18(%rdx), %ecx movl %ecx, 0x148(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x150(%rsp) movups 0x28(%rdx), %xmm0 movups %xmm0, 0x158(%rsp) movl 0x38(%rdx), %ecx movl %ecx, 0x168(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x170(%rsp) testq %rax, %rax je 0x28c072 lock incl (%rax) cmpl %ebx, 0x10(%rsp) jae 0x28c0fe movq 0x88(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xd0(%rcx,%rax), %eax cltd movl 0x10(%rsp), %r9d idivl %r9d movl %r15d, %ecx shrq %cl, %r13 movl %edi, %ebx movl %r9d, %r8d imulq %r13, %r8 movq 0x10(%r14), %rcx movq %rcx, (%rsp) leaq 0x130(%rsp), %rdi movl 0x60(%rsp), %esi movl 0x70(%rsp), %edx movl %eax, %ecx movl %r10d, %ebp callq 0x628f2 pushq $-0x64 popq %r12 cmpq $0x0, 0x130(%rsp) je 0x28d35b movl %ebp, %r10d movl %ebx, %edi movslq 0x168(%rsp), %rax imulq 0x170(%rsp), %rax testq %rax, %rax je 0x28d35b xorl %r15d, %r15d xorl %r12d, %r12d xorl %ebx, %ebx movq 0x88(%rsp), %r9 movq (%r9), %rax movq -0x18(%rax), %rax movslq 0x108(%r9,%rax), %rax cmpq %rax, %rbx jge 0x28c3ab movl %r15d, %eax cltd movl 0x20(%rsp), %esi idivl %esi movl %eax, %ecx movl %edi, %ebp movl %edi, %eax cltd idivl %esi movl %eax, %edx movslq 0x1c4(%rsp), %rdi movslq 0x1c8(%rsp), %rsi movslq 0x1cc(%rsp), %rax movslq %ecx, %r8 imulq 0x1d8(%rsp), %r8 movq 0x1a8(%rsp), %rcx imulq %rcx, %r8 addq 0x198(%rsp), %r8 movq %r8, 0x238(%rsp) movl 0x1b0(%rsp), %r8d andq $0x0, 0x240(%rsp) movq %rcx, 0x248(%rsp) movl %r8d, 0x250(%rsp) movq 0x1b8(%rsp), %r8 movq %r8, 0x258(%rsp) movl %edi, 0x264(%rsp) movl %esi, 0x268(%rsp) movl %eax, 0x26c(%rsp) movl %edx, 0x270(%rsp) imulq %rdi, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x278(%rsp) movl 0x1c0(%rsp), %eax movl %eax, 0x260(%rsp) movl %r12d, %eax cltd movl 0x10(%rsp), %esi idivl %esi movl %eax, %ecx movl %r10d, %r13d movl %r10d, %eax cltd idivl %esi movl %eax, %edx movslq 0x15c(%rsp), %r8 movslq 0x160(%rsp), %rdi movslq 0x164(%rsp), %rax movslq %ecx, %rcx imulq 0x170(%rsp), %rcx movq 0x140(%rsp), %rsi imulq %rsi, %rcx addq 0x130(%rsp), %rcx movq %rcx, 0x1f0(%rsp) movl 0x148(%rsp), %ecx andq $0x0, 0x1f8(%rsp) movq %rsi, 0x200(%rsp) movl %ecx, 0x208(%rsp) movq 0x150(%rsp), %rcx movq %rcx, 0x210(%rsp) movl %r8d, 0x21c(%rsp) movl %edi, 0x220(%rsp) movl %eax, 0x224(%rsp) movl %edx, 0x228(%rsp) imulq %r8, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x230(%rsp) movl 0x158(%rsp), %eax movl %eax, 0x218(%rsp) movq 0x10(%r9), %rax movq (%rax,%rbx,8), %rdi movups (%r14), %xmm0 movups 0x10(%r14), %xmm1 movups 0x20(%r14), %xmm2 movups 0x30(%r14), %xmm3 movaps %xmm0, 0x280(%rsp) movaps %xmm3, 0x2b0(%rsp) movaps %xmm2, 0x2a0(%rsp) movaps %xmm1, 0x290(%rsp) movq %rcx, 0x288(%rsp) movq (%rdi), %rax leaq 0x238(%rsp), %rsi leaq 0x1f0(%rsp), %rdx leaq 0x280(%rsp), %rcx callq *0x38(%rax) movq 0x1f8(%rsp), %rax testq %rax, %rax je 0x28c361 lock decl (%rax) jne 0x28c361 movq 0x1f0(%rsp), %rsi movq 0x210(%rsp), %rdi testq %rdi, %rdi je 0x28c359 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28c361 movq %rsi, %rdi callq 0x5f3e0 movq 0x240(%rsp), %rax testq %rax, %rax je 0x28c398 lock decl (%rax) jne 0x28c398 movq 0x238(%rsp), %rsi movq 0x258(%rsp), %rdi testq %rdi, %rdi je 0x28c390 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28c398 movq %rsi, %rdi callq 0x5f3e0 incq %rbx movl %r13d, %r10d addl %r13d, %r12d movl %ebp, %edi addl %ebp, %r15d jmp 0x28c106 movl 0x30(%rsp), %edx cmpl %edx, 0x10(%rsp) jae 0x28c3d2 leaq 0x130(%rsp), %rdi movq 0x48(%rsp), %rsi movq %r14, %rcx callq 0x64e3b xorl %r12d, %r12d jmp 0x28d35b xorl %r12d, %r12d leaq 0x130(%rsp), %rax movq 0x48(%rsp), %rcx cmpq %rcx, %rax je 0x28d35b movq 0x138(%rsp), %rax testq %rax, %rax je 0x28c3fb lock incl (%rax) movq 0x8(%rcx), %rax testq %rax, %rax je 0x28d307 lock decl (%rax) jne 0x28d307 movq 0x48(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x28d2fc movq (%rdi), %rax callq *0x18(%rax) xorl %r12d, %r12d jmp 0x28d307 pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rbp,%r8) jne 0x28ccf0 cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28ccf0 cmpl $0x2, 0xe4(%rbp,%r8) jne 0x28ccf0 cmpl $0x2, 0xe8(%rbp,%r8) jne 0x28ccf0 movl 0x2c(%r14), %eax movl 0x30(%r14), %ecx movl 0xdc(%rsp), %edi movl 0xe8(%rsp), %edx subl %eax, %edi shll $0x3, %edi movq 0x1b0(%rbp,%r8), %rsi movslq %edi, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x28d415 testq %rsi, %rsi je 0x28c4c9 movq %r8, %r9 shlq $0x4, %r9 movups (%rsi,%r9), %xmm15 jmp 0x28c4cd xorps %xmm15, %xmm15 movq 0x40(%r14), %r9 imulq %r8, %r9 imulq 0x10(%r14), %r9 addq (%r14), %r9 movq 0x28(%rbp), %r15 movslq 0x54(%rbp), %r12 imulq %r8, %r12 imulq 0x38(%rbp), %r12 movslq 0xdc(%rsp), %rbx movq 0xf0(%rsp), %r10 imulq %r8, %r10 movq 0xc0(%rsp), %r11 imulq %r11, %r10 addq 0xb0(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx movaps (%r15,%r12), %xmm3 movaps 0x10(%r15,%r12), %xmm4 movaps 0x20(%r15,%r12), %xmm5 movaps 0x30(%r15,%r12), %xmm14 movaps 0x40(%r15,%r12), %xmm7 movaps 0x50(%r15,%r12), %xmm6 movaps 0x60(%r15,%r12), %xmm8 movaps 0x70(%r15,%r12), %xmm0 movaps %xmm0, 0x30(%rsp) movaps 0x80(%r15,%r12), %xmm0 movaps %xmm0, 0x60(%rsp) xorl %ebp, %ebp movaps %xmm5, 0x70(%rsp) movaps %xmm6, 0x90(%rsp) movaps %xmm8, 0xa0(%rsp) movaps %xmm3, 0x20(%rsp) movaps %xmm15, 0x50(%rsp) movaps %xmm4, 0x10(%rsp) cmpl %ecx, %ebp je 0x28c9ce xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r13d cmpl %eax, %r13d jge 0x28c7f1 movaps (%r10,%r15), %xmm11 mulps %xmm3, %xmm11 addps %xmm15, %xmm11 movaps 0x10(%r10,%r15), %xmm12 mulps %xmm4, %xmm12 movaps 0x20(%r10,%r15), %xmm13 movaps 0x40(%r10,%r15), %xmm10 movaps %xmm13, %xmm15 movaps 0x70(%rsp), %xmm6 mulps %xmm6, %xmm15 addps %xmm12, %xmm15 addps %xmm11, %xmm15 movaps (%r11,%r15), %xmm11 mulps %xmm14, %xmm11 movaps 0x10(%r11,%r15), %xmm12 mulps %xmm7, %xmm12 addps %xmm11, %xmm12 movaps 0x20(%r11,%r15), %xmm0 movaps 0x40(%r11,%r15), %xmm11 movaps %xmm7, %xmm8 movaps %xmm14, %xmm7 movaps %xmm0, %xmm14 movaps 0x90(%rsp), %xmm9 mulps %xmm9, %xmm14 addps %xmm12, %xmm14 addps %xmm15, %xmm14 movaps (%rbx,%r15), %xmm12 movaps 0xa0(%rsp), %xmm5 mulps %xmm5, %xmm12 movaps 0x10(%rbx,%r15), %xmm1 movaps 0x30(%rsp), %xmm4 mulps %xmm4, %xmm1 addps %xmm12, %xmm1 movaps 0x20(%rbx,%r15), %xmm2 movaps 0x40(%rbx,%r15), %xmm12 movaps %xmm2, %xmm15 movaps 0x60(%rsp), %xmm3 mulps %xmm3, %xmm15 addps %xmm1, %xmm15 addps %xmm14, %xmm15 movaps 0x30(%r10,%r15), %xmm1 mulps 0x10(%rsp), %xmm1 movaps 0x30(%r11,%r15), %xmm14 mulps %xmm8, %xmm14 addps %xmm1, %xmm14 movaps 0x30(%rbx,%r15), %xmm1 mulps %xmm4, %xmm1 addps %xmm14, %xmm1 movaps %xmm15, (%r9) mulps 0x20(%rsp), %xmm13 addps 0x50(%rsp), %xmm13 movaps %xmm10, %xmm14 mulps %xmm6, %xmm14 mulps %xmm7, %xmm0 movaps %xmm11, %xmm15 mulps %xmm9, %xmm15 addps %xmm14, %xmm15 mulps %xmm5, %xmm2 addps %xmm0, %xmm2 addps %xmm13, %xmm2 addps %xmm1, %xmm2 movaps %xmm12, %xmm0 mulps %xmm3, %xmm0 addps %xmm15, %xmm0 addps %xmm2, %xmm0 movaps 0x60(%r10,%r15), %xmm14 movaps 0x60(%r11,%r15), %xmm15 movaps 0x60(%rbx,%r15), %xmm13 movaps 0x50(%r10,%r15), %xmm1 mulps 0x10(%rsp), %xmm1 movaps 0x50(%r11,%r15), %xmm2 mulps %xmm8, %xmm2 addps %xmm1, %xmm2 movaps 0x50(%rbx,%r15), %xmm1 mulps %xmm4, %xmm1 addps %xmm2, %xmm1 movaps %xmm0, 0x10(%r9) movaps %xmm14, %xmm0 mulps %xmm6, %xmm0 movaps %xmm15, %xmm2 mulps %xmm9, %xmm2 addps %xmm0, %xmm2 mulps %xmm7, %xmm11 mulps %xmm5, %xmm12 addps %xmm11, %xmm12 mulps 0x20(%rsp), %xmm10 addps 0x50(%rsp), %xmm10 addps %xmm10, %xmm12 addps %xmm1, %xmm12 movaps %xmm13, %xmm0 mulps %xmm3, %xmm0 addps %xmm2, %xmm0 addps %xmm12, %xmm0 movaps 0x70(%r10,%r15), %xmm1 mulps 0x10(%rsp), %xmm1 movaps 0x80(%r10,%r15), %xmm2 mulps %xmm6, %xmm2 movaps 0x70(%r11,%r15), %xmm10 mulps %xmm8, %xmm10 addps %xmm1, %xmm10 movaps 0x80(%r11,%r15), %xmm1 mulps %xmm9, %xmm1 addps %xmm2, %xmm1 movaps 0x70(%rbx,%r15), %xmm2 mulps %xmm4, %xmm2 addps %xmm10, %xmm2 movaps 0x80(%rbx,%r15), %xmm10 mulps %xmm3, %xmm10 addps %xmm1, %xmm10 movaps %xmm0, 0x20(%r9) mulps %xmm7, %xmm15 mulps %xmm5, %xmm13 addps %xmm15, %xmm13 movaps 0x50(%rsp), %xmm15 movaps 0x20(%rsp), %xmm3 movaps 0x10(%rsp), %xmm4 mulps %xmm3, %xmm14 addps %xmm15, %xmm14 addps %xmm14, %xmm13 movaps %xmm7, %xmm14 movaps %xmm8, %xmm7 addps %xmm2, %xmm13 addps %xmm10, %xmm13 movaps %xmm13, 0x30(%r9) addq $0x40, %r9 addl $0x4, %r12d subq $-0x80, %r15 jmp 0x28c59a movaps 0x90(%rsp), %xmm6 movaps 0xa0(%rsp), %xmm8 leal 0x1(%r12), %r13d cmpl %eax, %r13d jge 0x28c91e movaps 0x20(%r10,%r15), %xmm11 movaps 0x20(%r11,%r15), %xmm12 movaps 0x20(%rbx,%r15), %xmm10 movaps (%r10,%r15), %xmm0 mulps %xmm3, %xmm0 addps %xmm15, %xmm0 movaps 0x10(%r10,%r15), %xmm1 mulps %xmm4, %xmm1 movaps %xmm11, %xmm2 movaps 0x70(%rsp), %xmm13 mulps %xmm13, %xmm2 addps %xmm1, %xmm2 addps %xmm0, %xmm2 movaps (%r11,%r15), %xmm0 mulps %xmm14, %xmm0 movaps 0x10(%r11,%r15), %xmm1 mulps %xmm7, %xmm1 addps %xmm0, %xmm1 movaps %xmm12, %xmm0 mulps %xmm6, %xmm0 addps %xmm1, %xmm0 addps %xmm2, %xmm0 movaps (%rbx,%r15), %xmm1 mulps %xmm8, %xmm1 movaps 0x10(%rbx,%r15), %xmm2 movaps 0x30(%rsp), %xmm9 mulps %xmm9, %xmm2 addps %xmm1, %xmm2 movaps %xmm10, %xmm1 movaps 0x60(%rsp), %xmm5 mulps %xmm5, %xmm1 addps %xmm2, %xmm1 addps %xmm0, %xmm1 movaps 0x30(%r10,%r15), %xmm0 mulps %xmm4, %xmm0 movaps 0x40(%r10,%r15), %xmm2 mulps %xmm13, %xmm2 movaps 0x30(%r11,%r15), %xmm13 mulps %xmm7, %xmm13 addps %xmm0, %xmm13 movaps 0x40(%r11,%r15), %xmm0 mulps %xmm6, %xmm0 addps %xmm2, %xmm0 movaps 0x30(%rbx,%r15), %xmm2 mulps %xmm9, %xmm2 addps %xmm13, %xmm2 movaps 0x40(%rbx,%r15), %xmm13 mulps %xmm5, %xmm13 addps %xmm0, %xmm13 movaps %xmm1, (%r9) mulps %xmm14, %xmm12 mulps %xmm8, %xmm10 addps %xmm12, %xmm10 mulps %xmm3, %xmm11 addps %xmm15, %xmm11 addps %xmm11, %xmm10 addps %xmm2, %xmm10 addps %xmm13, %xmm10 movaps %xmm10, 0x10(%r9) addq $0x20, %r9 addl $0x2, %r12d addq $0x40, %r15 jmp 0x28c802 movq 0x48(%rsp), %r14 movaps 0x70(%rsp), %xmm5 cmpl %eax, %r12d jge 0x28c9b5 movaps (%r10,%r15), %xmm0 mulps %xmm3, %xmm0 addps %xmm15, %xmm0 movaps 0x10(%r10,%r15), %xmm1 mulps %xmm4, %xmm1 movaps 0x20(%r10,%r15), %xmm2 mulps %xmm5, %xmm2 addps %xmm1, %xmm2 addps %xmm0, %xmm2 movaps (%r11,%r15), %xmm0 mulps %xmm14, %xmm0 movaps 0x10(%r11,%r15), %xmm1 mulps %xmm7, %xmm1 addps %xmm0, %xmm1 movaps 0x20(%r11,%r15), %xmm0 mulps %xmm6, %xmm0 addps %xmm1, %xmm0 addps %xmm2, %xmm0 movaps (%rbx,%r15), %xmm1 mulps %xmm8, %xmm1 movaps 0x10(%rbx,%r15), %xmm2 mulps 0x30(%rsp), %xmm2 addps %xmm1, %xmm2 movaps 0x20(%rbx,%r15), %xmm1 mulps 0x60(%rsp), %xmm1 addps %xmm2, %xmm1 addps %xmm0, %xmm1 movaps %xmm1, (%r9) addq $0x10, %r9 incl %r12d addq $0x20, %r15 jmp 0x28c928 addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x28c58c incq %r8 movq 0x88(%rsp), %rbp jmp 0x28c4ad pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rbp,%r8) jne 0x28ccf0 cmpl $0x1, 0xe0(%rbp,%r8) jne 0x28ccf0 cmpl $0x2, 0xe4(%rbp,%r8) jne 0x28ccf0 cmpl $0x2, 0xe8(%rbp,%r8) jne 0x28ccf0 movl 0x2c(%r14), %edi movl 0x30(%r14), %ecx movl 0xdc(%rsp), %eax movl 0xe8(%rsp), %edx subl %edi, %eax shll $0x3, %eax movq 0x1b0(%rbp,%r8), %rsi xorl %r15d, %r15d testl %edi, %edi cmovlel %r15d, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %edx, %edx cmovlel %r15d, %edx movq %rdx, 0x10(%rsp) shlq $0x2, %r8 movq %rsi, 0x20(%rsp) cmpq 0x10(%rsp), %r15 je 0x28d415 testq %rsi, %rsi je 0x28ca84 movq %r15, %rax shlq $0x4, %rax movups (%rsi,%rax), %xmm0 jmp 0x28ca87 xorps %xmm0, %xmm0 movq 0x40(%r14), %r9 imulq %r15, %r9 imulq 0x10(%r14), %r9 addq (%r14), %r9 movq 0x28(%rbp), %r10 movslq 0x54(%rbp), %r11 imulq %r15, %r11 imulq 0x38(%rbp), %r11 movq 0xf0(%rsp), %rbx movq %r15, %r14 imulq %r15, %rbx movq 0xc0(%rsp), %rax imulq %rax, %rbx addq 0xb0(%rsp), %rbx movslq 0xdc(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x28ccd3 movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x28ccae movaps (%r10,%r11), %xmm2 movaps 0x10(%r10,%r11), %xmm3 movaps 0x20(%r10,%r11), %xmm1 movaps 0x30(%r10,%r11), %xmm4 movaps 0x40(%r10,%r11), %xmm5 mulps (%rbx,%rsi), %xmm2 mulps 0x10(%rbx,%rsi), %xmm3 mulps 0x20(%rbx,%rsi), %xmm1 addps %xmm0, %xmm2 addps %xmm3, %xmm1 mulps 0x30(%rbx,%rsi), %xmm4 addps %xmm2, %xmm1 mulps 0x40(%rbx,%rsi), %xmm5 addps %xmm4, %xmm5 movaps 0x50(%r10,%r11), %xmm3 movaps 0x60(%r10,%r11), %xmm4 movaps 0x70(%r10,%r11), %xmm6 movaps 0x80(%r10,%r11), %xmm7 movaps 0x90(%r10,%r11), %xmm2 mulps (%r15,%rsi), %xmm3 addps %xmm5, %xmm3 mulps 0x10(%r15,%rsi), %xmm4 mulps 0x20(%r15,%rsi), %xmm6 addps %xmm1, %xmm3 addps %xmm4, %xmm6 mulps 0x30(%r15,%rsi), %xmm7 addps %xmm6, %xmm7 mulps 0x40(%r15,%rsi), %xmm2 addps %xmm7, %xmm2 addps %xmm3, %xmm2 movaps 0xa0(%r10,%r11), %xmm3 movaps 0xb0(%r10,%r11), %xmm4 movaps 0xc0(%r10,%r11), %xmm5 movaps 0xd0(%r10,%r11), %xmm6 movaps 0xe0(%r10,%r11), %xmm1 mulps (%r12,%rsi), %xmm3 mulps 0x10(%r12,%rsi), %xmm4 mulps 0x20(%r12,%rsi), %xmm5 addps %xmm3, %xmm4 addps %xmm4, %xmm5 mulps 0x30(%r12,%rsi), %xmm6 addps %xmm5, %xmm6 mulps 0x40(%r12,%rsi), %xmm1 addps %xmm6, %xmm1 addps %xmm2, %xmm1 movaps 0xf0(%r10,%r11), %xmm2 movaps 0x100(%r10,%r11), %xmm3 movaps 0x110(%r10,%r11), %xmm4 movaps 0x120(%r10,%r11), %xmm5 movaps 0x130(%r10,%r11), %xmm6 mulps (%r13,%rsi), %xmm2 mulps 0x10(%r13,%rsi), %xmm3 mulps 0x20(%r13,%rsi), %xmm4 addps %xmm2, %xmm3 addps %xmm3, %xmm4 mulps 0x30(%r13,%rsi), %xmm5 addps %xmm4, %xmm5 mulps 0x40(%r13,%rsi), %xmm6 addps %xmm5, %xmm6 movaps 0x140(%r10,%r11), %xmm2 movaps 0x150(%r10,%r11), %xmm3 movaps 0x160(%r10,%r11), %xmm4 movaps 0x170(%r10,%r11), %xmm5 movaps 0x180(%r10,%r11), %xmm7 mulps (%rbp,%rsi), %xmm2 addps %xmm6, %xmm2 mulps 0x10(%rbp,%rsi), %xmm3 mulps 0x20(%rbp,%rsi), %xmm4 addps %xmm1, %xmm2 addps %xmm3, %xmm4 mulps 0x30(%rbp,%rsi), %xmm5 addps %xmm4, %xmm5 mulps 0x40(%rbp,%rsi), %xmm7 addps %xmm5, %xmm7 addps %xmm2, %xmm7 movaps %xmm7, (%r9) addq $0x10, %r9 addq $0x20, %rsi jmp 0x28caf8 addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x28caec movq %r14, %r15 incq %r15 movq 0x48(%rsp), %r14 movq 0x88(%rsp), %rbp movq 0x20(%rsp), %rsi jmp 0x28ca67 imull %eax, %r15d movslq %r15d, %rsi leaq 0x198(%rsp), %rdi leaq 0x130(%rsp), %rdx callq 0x73bbe movq (%rbp), %rcx movq -0x18(%rcx), %rdx movl 0x10(%rsp), %r11d imull 0xe0(%rbp,%rdx), %r11d movq 0x198(%rsp), %rax movl 0xdc(%rbp,%rdx), %esi imull 0xd4(%rbp,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%rbp,%rdx), %r8d jge 0x28cd80 movslq %edi, %rdi leaq (%rax,%rdi,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%rbp,%rdx), %r9d jge 0x28cd75 movl %esi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%rbp,%rdx), %esi incq %r9 jmp 0x28cd57 addl %r11d, %esi incl %r8d addq %r9, %rdi jmp 0x28cd43 leal (,%r15,4), %ecx movl %ecx, 0x90(%rsp) movl 0x60(%rsp), %edx shll $0x2, %edx xorl %ecx, %ecx testl %r15d, %r15d cmovlel %ecx, %r15d movslq %edx, %rdx movq %rdx, 0x30(%rsp) movq 0x50(%rsp), %r12 testl %r12d, %r12d cmovlel %ecx, %r12d movq %r12, 0x50(%rsp) movslq 0xa0(%rsp), %rdi shlq $0x2, %r15 leaq 0x16b70d(%rip), %r8 # 0x3f84d8 xorps %xmm0, %xmm0 movaps 0x16423a(%rip), %xmm15 # 0x3f1010 movaps 0x164243(%rip), %xmm4 # 0x3f1020 movaps 0x16424c(%rip), %xmm6 # 0x3f1030 movaps 0x1612b5(%rip), %xmm5 # 0x3ee0a0 movaps 0x16424d(%rip), %xmm8 # 0x3f1040 xorl %edx, %edx movaps 0x164253(%rip), %xmm9 # 0x3f1050 movaps 0x16426b(%rip), %xmm11 # 0x3f1070 movaps 0x164273(%rip), %xmm12 # 0x3f1080 movaps 0x16427b(%rip), %xmm13 # 0x3f1090 cmpq 0x50(%rsp), %rdx je 0x28d2e7 movq %rcx, 0x60(%rsp) movslq %ecx, %r10 movq 0x40(%r14), %r11 imulq %rdx, %r11 imulq 0x10(%r14), %r11 shlq $0x2, %r10 addq (%r14), %r11 movslq 0xdc(%rsp), %rbx movq 0xf0(%rsp), %r13 imulq %rdx, %r13 movq 0xc0(%rsp), %rcx imulq %rcx, %r13 addq 0xb0(%rsp), %r13 imulq %rcx, %rbx movq %rdx, 0x70(%rsp) shlq $0x4, %rdx movq %rdx, 0x10(%rsp) addq 0x28(%rbp), %r10 xorl %ecx, %ecx cmpl 0x20(%rsp), %ecx jg 0x28d2c9 movq (%rbp), %rdx xorl %r9d, %r9d cmpq %rdi, %r9 jg 0x28d2b9 movq -0x18(%rdx), %r12 cmpl $0x0, 0x100(%rbp,%r12) je 0x28ceb9 movq 0x1b0(%rbp,%r12), %rsi movq 0x10(%rsp), %r14 movups (%rsi,%r14), %xmm1 jmp 0x28cebc xorps %xmm1, %xmm1 movslq 0xe8(%rbp,%r12), %rsi movq %rbp, %r14 movslq %ecx, %rbp imulq %rsi, %rbp imulq %rbx, %rbp addq %r13, %rbp movl 0xe4(%r14,%r12), %esi imull %r9d, %esi shll $0x2, %esi movslq %esi, %rsi leaq (,%rsi,4), %rsi addq %rbp, %rsi xorl %ebp, %ebp cmpq %rbp, %r15 je 0x28cf17 movslq (%rax,%rbp), %r14 shlq $0x4, %r14 movups (%rsi,%r14), %xmm2 movups (%r10,%rbp,4), %xmm3 mulps %xmm2, %xmm3 addps %xmm3, %xmm1 addq $0x4, %rbp jmp 0x28cef4 movq 0x88(%rsp), %rbp movl 0x110(%rbp,%r12), %esi decl %esi cmpl $0x5, %esi ja 0x28d1cc movslq (%r8,%rsi,4), %rsi addq %r8, %rsi movaps %xmm1, %xmm7 maxps %xmm0, %xmm7 jmpq *%rsi movq 0x118(%rbp,%r12), %rsi minps %xmm0, %xmm1 movss (%rsi), %xmm2 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm2 addps %xmm2, %xmm7 jmp 0x28d2a5 movaps %xmm1, %xmm2 minps %xmm15, %xmm2 maxps %xmm4, %xmm2 movaps %xmm2, %xmm14 mulps %xmm6, %xmm14 addps %xmm5, %xmm14 cvttps2dq %xmm14, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm14 andps %xmm8, %xmm14 subps %xmm14, %xmm7 movaps %xmm15, %xmm4 cvttps2dq %xmm7, %xmm15 mulps %xmm9, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm7 mulps %xmm2, %xmm7 movaps %xmm2, %xmm14 movaps 0x1640b4(%rip), %xmm10 # 0x3f1060 mulps %xmm10, %xmm14 addps %xmm11, %xmm14 mulps %xmm2, %xmm14 addps %xmm12, %xmm14 mulps %xmm2, %xmm14 addps %xmm13, %xmm14 mulps %xmm2, %xmm14 xorps %xmm6, %xmm6 movaps %xmm5, %xmm0 movaps 0x1640cb(%rip), %xmm5 # 0x3f10a0 addps %xmm5, %xmm14 mulps %xmm2, %xmm14 addps %xmm0, %xmm14 mulps %xmm7, %xmm14 addps %xmm8, %xmm2 addps %xmm14, %xmm2 pslld $0x17, %xmm15 paddd %xmm8, %xmm15 mulps %xmm2, %xmm15 addps %xmm8, %xmm15 movaps %xmm15, %xmm2 maxps 0x1640a4(%rip), %xmm15 # 0x3f10b0 movaps %xmm15, %xmm7 andps 0x1640a8(%rip), %xmm15 # 0x3f10c0 orps %xmm0, %xmm15 movaps %xmm15, %xmm3 cmpltps 0x1640b8(%rip), %xmm3 # 0x3f10e0 movaps %xmm3, %xmm14 andps %xmm15, %xmm14 movaps 0x1640b8(%rip), %xmm10 # 0x3f10f0 addps %xmm10, %xmm15 addps %xmm14, %xmm15 cmpleps %xmm6, %xmm2 psrld $0x17, %xmm7 paddd 0x16407f(%rip), %xmm7 # 0x3f10d0 cvtdq2ps %xmm7, %xmm14 andps %xmm8, %xmm3 subps %xmm3, %xmm14 movaps %xmm15, %xmm7 mulps 0x164098(%rip), %xmm7 # 0x3f1100 addps 0x1640a1(%rip), %xmm7 # 0x3f1110 mulps %xmm15, %xmm7 addps 0x1640a6(%rip), %xmm7 # 0x3f1120 mulps %xmm15, %xmm7 addps 0x1640ab(%rip), %xmm7 # 0x3f1130 mulps %xmm15, %xmm7 addps 0x1640b0(%rip), %xmm7 # 0x3f1140 mulps %xmm15, %xmm7 addps 0x1640b5(%rip), %xmm7 # 0x3f1150 mulps %xmm15, %xmm7 addps 0x1640ba(%rip), %xmm7 # 0x3f1160 mulps %xmm15, %xmm7 addps 0x1640bf(%rip), %xmm7 # 0x3f1170 mulps %xmm15, %xmm7 addps 0x1640c4(%rip), %xmm7 # 0x3f1180 mulps %xmm15, %xmm7 mulps %xmm9, %xmm14 addps %xmm15, %xmm14 mulps %xmm15, %xmm15 addps 0x1640bd(%rip), %xmm7 # 0x3f1190 mulps %xmm15, %xmm7 movaps %xmm4, %xmm15 movaps 0x163f3e(%rip), %xmm4 # 0x3f1020 movaps 0x163f47(%rip), %xmm6 # 0x3f1030 addps %xmm7, %xmm14 mulps 0x164bcb(%rip), %xmm14 # 0x3f1cc0 movaps %xmm2, %xmm3 andnps %xmm14, %xmm3 andps 0x164bcd(%rip), %xmm2 # 0x3f1cd0 orps %xmm3, %xmm2 minps %xmm15, %xmm2 maxps %xmm4, %xmm2 movaps %xmm2, %xmm3 mulps %xmm6, %xmm3 addps %xmm0, %xmm3 cvttps2dq %xmm3, %xmm7 cvtdq2ps %xmm7, %xmm7 cmpltps %xmm7, %xmm3 andps %xmm8, %xmm3 subps %xmm3, %xmm7 cvttps2dq %xmm7, %xmm14 mulps %xmm9, %xmm7 subps %xmm7, %xmm2 movaps %xmm2, %xmm3 mulps %xmm2, %xmm3 movaps %xmm2, %xmm7 mulps 0x163f1c(%rip), %xmm7 # 0x3f1060 addps %xmm11, %xmm7 mulps %xmm2, %xmm7 addps %xmm12, %xmm7 mulps %xmm2, %xmm7 addps %xmm13, %xmm7 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 movaps %xmm0, %xmm5 xorps %xmm0, %xmm0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm3, %xmm7 addps %xmm8, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm8, %xmm14 mulps %xmm2, %xmm14 addps %xmm8, %xmm14 rcpps %xmm14, %xmm2 movaps %xmm2, %xmm7 addps %xmm2, %xmm7 mulps %xmm7, %xmm14 movaps 0x164b46(%rip), %xmm3 # 0x3f1ce0 subps %xmm14, %xmm3 mulps %xmm2, %xmm3 addps %xmm10, %xmm7 addps %xmm3, %xmm7 jmp 0x28d2a2 movq 0x118(%rbp,%r12), %rsi movss (%rsi), %xmm2 movss 0x4(%rsi), %xmm7 shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] maxps %xmm2, %xmm1 minps %xmm7, %xmm1 movaps %xmm1, %xmm7 jmp 0x28d2a5 xorps 0x160eb5(%rip), %xmm1 # 0x3ee090 minps %xmm15, %xmm1 maxps %xmm4, %xmm1 movaps %xmm1, %xmm7 mulps %xmm6, %xmm7 addps %xmm5, %xmm7 cvttps2dq %xmm7, %xmm2 cvtdq2ps %xmm2, %xmm2 cmpltps %xmm2, %xmm7 andps %xmm8, %xmm7 subps %xmm7, %xmm2 cvttps2dq %xmm2, %xmm14 mulps 0x164ae7(%rip), %xmm2 # 0x3f1cf0 addps %xmm1, %xmm2 movaps %xmm2, %xmm1 mulps %xmm2, %xmm1 movaps %xmm2, %xmm7 mulps 0x163e44(%rip), %xmm7 # 0x3f1060 addps 0x163e4d(%rip), %xmm7 # 0x3f1070 mulps %xmm2, %xmm7 addps 0x163e53(%rip), %xmm7 # 0x3f1080 mulps %xmm2, %xmm7 addps 0x163e59(%rip), %xmm7 # 0x3f1090 mulps %xmm2, %xmm7 addps 0x163e5f(%rip), %xmm7 # 0x3f10a0 mulps %xmm2, %xmm7 addps %xmm5, %xmm7 mulps %xmm1, %xmm7 addps %xmm8, %xmm2 addps %xmm7, %xmm2 pslld $0x17, %xmm14 paddd %xmm8, %xmm14 mulps %xmm2, %xmm14 addps %xmm8, %xmm14 rcpps %xmm14, %xmm1 mulps %xmm1, %xmm14 movaps %xmm8, %xmm7 subps %xmm14, %xmm7 mulps %xmm1, %xmm7 addps %xmm1, %xmm7 jmp 0x28d2a5 movq 0x118(%rbp,%r12), %rsi movss (%rsi), %xmm7 movss 0x4(%rsi), %xmm2 shufps $0x0, %xmm7, %xmm7 # xmm7 = xmm7[0,0,0,0] shufps $0x0, %xmm2, %xmm2 # xmm2 = xmm2[0,0,0,0] mulps %xmm1, %xmm7 addps %xmm2, %xmm7 maxps %xmm0, %xmm7 minps %xmm8, %xmm7 mulps %xmm1, %xmm7 movq %r9, %rsi shlq $0x4, %rsi movups %xmm7, (%r11,%rsi) incq %r9 jmp 0x28ce8d movq 0x30(%rsp), %rdx leaq (%r11,%rdx,4), %r11 incl %ecx jmp 0x28ce7c movq 0x70(%rsp), %rdx incq %rdx movq 0x60(%rsp), %rcx addl 0x90(%rsp), %ecx movq 0x48(%rsp), %r14 jmp 0x28ce15 leaq 0x198(%rsp), %rdi callq 0x624be xorl %r12d, %r12d jmp 0x28d3c9 movq %rsi, %rdi callq 0x5f3e0 xorl %r12d, %r12d movaps 0x130(%rsp), %xmm0 movq 0x48(%rsp), %rcx movups %xmm0, (%rcx) movq 0x140(%rsp), %rax movq %rax, 0x10(%rcx) movl 0x148(%rsp), %eax movl %eax, 0x18(%rcx) movq 0x150(%rsp), %rax movq %rax, 0x20(%rcx) movups 0x158(%rsp), %xmm0 movups %xmm0, 0x28(%rcx) movl 0x168(%rsp), %eax movl %eax, 0x38(%rcx) movq 0x170(%rsp), %rax movq %rax, 0x40(%rcx) movq 0x138(%rsp), %rax testq %rax, %rax je 0x28d392 lock decl (%rax) jne 0x28d392 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi je 0x28d38a movq (%rdi), %rax callq *0x18(%rax) jmp 0x28d392 movq %rsi, %rdi callq 0x5f3e0 movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x28d3c9 lock decl (%rax) jne 0x28d3c9 movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi je 0x28d3c1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28d3c9 movq %rsi, %rdi callq 0x5f3e0 movq 0xb8(%rsp), %rax testq %rax, %rax je 0x28d400 lock decl (%rax) jne 0x28d400 movq 0xb0(%rsp), %rsi movq 0xd0(%rsp), %rdi testq %rdi, %rdi je 0x28d3f8 movq (%rdi), %rax callq *0x18(%rax) jmp 0x28d400 movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x2c8, %rsp # imm = 0x2C8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq movq 0x8(%rbp), %rdi jmp 0x28d427 movq 0x88(%rsp), %rax movq 0x8(%rax), %rdi testq %rdi, %rdi je 0x28d2f4 movq (%rdi), %rax movq %r14, %rsi movq 0x190(%rsp), %rdx callq *0x48(%rax) jmp 0x28d2f4 jmp 0x28d473 jmp 0x28d5a1 jmp 0x28d5a1 movq %rax, %rbx jmp 0x28d523 movq %rax, %rbx jmp 0x28d4ec jmp 0x28d5a1 jmp 0x28d473 jmp 0x28d5a1 jmp 0x28d5a1 movq %rax, %rbx jmp 0x28d55a movq %rax, %rbx movq 0x1f8(%rsp), %rax testq %rax, %rax je 0x28d4b5 lock decl (%rax) jne 0x28d4b5 movq 0x1f0(%rsp), %rsi movq 0x210(%rsp), %rdi testq %rdi, %rdi jne 0x28d4af movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d4b5 movq (%rdi), %rax callq *0x18(%rax) movq 0x240(%rsp), %rax testq %rax, %rax je 0x28d4ec lock decl (%rax) jne 0x28d4ec movq 0x238(%rsp), %rsi movq 0x258(%rsp), %rdi testq %rdi, %rdi jne 0x28d4e6 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d4ec movq (%rdi), %rax callq *0x18(%rax) movq 0x138(%rsp), %rax testq %rax, %rax je 0x28d523 lock decl (%rax) jne 0x28d523 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi jne 0x28d51d movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d523 movq (%rdi), %rax callq *0x18(%rax) movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x28d55a lock decl (%rax) jne 0x28d55a movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi jne 0x28d554 movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d55a movq (%rdi), %rax callq *0x18(%rax) movq 0xb8(%rsp), %rax testq %rax, %rax je 0x28d591 lock decl (%rax) jne 0x28d591 movq 0xb0(%rsp), %rsi movq 0xd0(%rsp), %rdi testq %rdi, %rdi jne 0x28d58b movq %rsi, %rdi callq 0x5f3e0 jmp 0x28d591 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x28d5a1 jmp 0x28d5a1 jmp 0x28d5a1 jmp 0x28d5a1 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float*)bias_data) + g * 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float*)bias_data) + g * 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float*)bias_data) + g * 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x28a5b0

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

ncnn::ConvolutionDepthWise_x86::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int ConvolutionDepthWise_x86::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w *= weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w *= bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x168, %rsp # imm = 0x168 movq %rsi, %rbp movq %rdi, %r13 movq (%rsi), %r14 leaq 0x48(%r14), %rdi movq (%rdx), %rax movq %rax, 0xb8(%rsp) movl 0x60(%r14), %ebx movl 0x74(%r14), %eax movl %eax, 0x1c(%rsp) movl 0x78(%r14), %eax movl %eax, 0x18(%rsp) imull 0x80(%r14), %ebx leaq 0x70(%rsp), %rsi andq $0x0, 0x40(%rsi) xorps %xmm0, %xmm0 movaps %xmm0, (%rsi) movups %xmm0, 0xc(%rsi) movaps %xmm0, 0x20(%rsi) movups %xmm0, 0x2c(%rsi) movq %rcx, %r15 movq %rcx, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2904be movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x2904be movslq 0x88(%rsp), %rcx movl 0x9c(%rsp), %eax imull %ecx, %eax movl %eax, 0x9c(%rsp) movq 0x80(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x80(%rsp) movl $0x1, 0x88(%rsp) andq $0x0, 0x60(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x20(%rsp) movups %xmm0, 0x2c(%rsp) movaps %xmm0, 0x40(%rsp) movups %xmm0, 0x4c(%rsp) movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r13,%rax) je 0x29001e movl $0x90, %edi addq (%rbp), %rdi leaq 0x20(%rsp), %rsi movq %r15, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x20(%rsp) je 0x290490 movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x290490 movslq 0x38(%rsp), %rcx movl 0x4c(%rsp), %eax imull %ecx, %eax movl %eax, 0x4c(%rsp) movq 0x30(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x30(%rsp) movl $0x1, 0x38(%rsp) pushq $0x2a popq %rdi callq 0x782bf movq %rax, %r12 leaq 0x8(%rsp), %rdi callq 0x71548 leaq 0x8(%rsp), %rdi xorl %esi, %esi movl %ebx, %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi movl 0x1c(%rsp), %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0xb popq %rsi movl 0x18(%rsp), %edx callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xdc(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xd popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xec(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x4 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xf popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xe popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x10 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movss 0xfc(%r13,%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x12 popq %rsi callq 0x71952 movq (%r13), %rax movq -0x18(%rax), %rax movl 0x100(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x5 popq %rsi callq 0x7193a movl 0x9c(%rsp), %edx leaq 0x8(%rsp), %rdi pushq $0x6 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x108(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x7 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x10c(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x110(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax leaq (%rax,%r13), %rdx addq $0x118, %rdx # imm = 0x118 leaq 0x8(%rsp), %rdi pushq $0xa popq %rsi callq 0x7196c movq (%r12), %rax leaq 0x8(%rsp), %rsi movq %r12, %rdi callq *0x10(%rax) andq $0x0, 0x110(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0xd0(%rsp) movups %xmm0, 0xdc(%rsp) movaps %xmm0, 0xf0(%rsp) movups %xmm0, 0xfc(%rsp) andq $0x0, 0x158(%rsp) movups %xmm0, 0x118(%rsp) movups %xmm0, 0x124(%rsp) movups %xmm0, 0x138(%rsp) movups %xmm0, 0x144(%rsp) movq 0x78(%rsp), %rax testq %rax, %rax je 0x290283 lock incl (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2902ba lock decl (%rax) jne 0x2902ba movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x2902b2 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2902ba movq %rsi, %rdi callq 0x5f3e0 movaps 0x70(%rsp), %xmm0 movaps %xmm0, 0xd0(%rsp) movq 0x80(%rsp), %rax movq %rax, 0xe0(%rsp) movl 0x88(%rsp), %eax movl %eax, 0xe8(%rsp) movq 0x90(%rsp), %rax movq %rax, 0xf0(%rsp) movups 0x98(%rsp), %xmm0 movups %xmm0, 0xf8(%rsp) movl 0xa8(%rsp), %eax movl %eax, 0x108(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x110(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax je 0x290330 lock incl (%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x290367 lock decl (%rax) jne 0x290367 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x29035f movq (%rdi), %rax callq *0x18(%rax) jmp 0x290367 movq %rsi, %rdi callq 0x5f3e0 movaps 0x20(%rsp), %xmm0 leaq 0xd0(%rsp), %rsi movups %xmm0, 0x48(%rsi) movq 0x30(%rsp), %rax movq %rax, 0x58(%rsi) movl 0x38(%rsp), %eax movl %eax, 0x60(%rsi) movq 0x40(%rsp), %rax movq %rax, 0x68(%rsi) movups 0x48(%rsp), %xmm0 movaps %xmm0, 0x70(%rsi) movl 0x58(%rsp), %eax movl %eax, 0x80(%rsi) movq 0x60(%rsp), %rax movq %rax, 0x88(%rsi) leaq 0xc0(%rsp), %rdi callq 0x6b00e movq (%r12), %rax leaq 0xc0(%rsp), %rsi movq %r12, %rdi callq *0x18(%rax) leaq 0xc0(%rsp), %rdi callq 0x6b03a movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq *0x20(%rax) movq (%r12), %rax movq %r12, %rdi movq %r14, %rsi movq 0xb8(%rsp), %rdx movq %r15, %rcx callq *0x38(%rax) movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq *0x28(%rax) movq (%r12), %rax movq %r12, %rdi callq *0x8(%rax) pushq $0x48 popq %rbx xorps %xmm0, %xmm0 movq 0xd8(%rsp,%rbx), %rax testq %rax, %rax je 0x29045b lock decl (%rax) jne 0x29045b movq 0xd0(%rsp,%rbx), %rsi movq 0xf0(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x290450 movq (%rdi), %rax callq *0x18(%rax) xorps %xmm0, %xmm0 jmp 0x29045b movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%rbx), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x29041e leaq 0x8(%rsp), %rdi callq 0x71614 xorl %r12d, %r12d movq 0x28(%rsp), %rax testq %rax, %rax je 0x2904be lock decl (%rax) jne 0x2904be movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2904b6 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2904be movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2904ef lock decl (%rax) jne 0x2904ef movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2904e7 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2904ef movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x168, %rsp # imm = 0x168 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x29061e jmp 0x29061e movq %rax, %rbx leaq 0xc0(%rsp), %rdi callq 0x6b03a jmp 0x290531 jmp 0x29052e jmp 0x290526 jmp 0x290526 movq %rax, %rbx jmp 0x2905b5 movq %rax, %rbx pushq $0x48 popq %r14 xorps %xmm0, %xmm0 movq 0xd8(%rsp,%r14), %rax testq %rax, %rax je 0x290575 lock decl (%rax) jne 0x290575 movq 0xd0(%rsp,%r14), %rsi movq 0xf0(%rsp,%r14), %rdi testq %rdi, %rdi je 0x29056a movq (%rdi), %rax callq *0x18(%rax) xorps %xmm0, %xmm0 jmp 0x290575 movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x290538 jmp 0x2905ab jmp 0x29061e movq %rax, %rbx jmp 0x2905e3 jmp 0x29061e movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq 0x28(%rsp), %rax testq %rax, %rax je 0x2905e3 lock decl (%rax) jne 0x2905e3 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2905dd movq %rsi, %rdi callq 0x5f3e0 jmp 0x2905e3 movq (%rdi), %rax callq *0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x290614 lock decl (%rax) jne 0x290614 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x29060e movq %rsi, %rdi callq 0x5f3e0 jmp 0x290614 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x29061e movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int ConvolutionDepthWise_x86::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w *= weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w *= bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }

movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x28febe

/csukuangfj[P]ncnn/src/layer/x86/convolutiondepthwise_x86.cpp

ncnn::ConvolutionDepthWise_x86_avx512::create_group_ops(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx512::create_group_ops(const Option& opt) { // create Convolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Convolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[5]; weights[0] = weight_data_g; weights[1] = bias_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[2] = weight_data_int8_scales_g; weights[3] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[4] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[4]; weights[0] = weight_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[1] = weight_data_int8_scales_g; weights[2] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[3] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x278, %rsp # imm = 0x278 movq %rsi, 0x250(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x104(%rdi,%rdx), %eax movl 0x108(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x10(%rdi), %rax movq %rax, 0x1e0(%rsp) xorl %r12d, %r12d movq 0x10(%r14), %rax movq 0x18(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x29104f movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x29104a movq (%rdi), %rax callq *0x8(%rax) incq %r12 jmp 0x291021 imull %r15d, %ebx cmpq %rax, %rcx je 0x29105c movq %rax, 0x18(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x108(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0x1e0(%rsp), %rdi callq 0x6fbc2 leaq 0x118(%r14), %rax movq %rax, 0x258(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0x6c(%rsp) movslq %ebx, %rax movq %rax, 0x260(%rsp) movslq %r15d, %rax movq %rax, 0x248(%rsp) pushq $0x1 popq %rbx xorl %edx, %edx movl %r15d, 0x14(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rcx cmpq %rcx, %rdx jge 0x291dc6 movq %rdx, %rcx movq %rdx, 0x8(%rsp) movq 0x260(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdx, 0x80(%rsp) movl %esi, 0x88(%rsp) movq %rax, 0x90(%rsp) movl %ebx, 0x98(%rsp) movl %edi, 0x9c(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa0(%rsp) movl %ebx, 0xa8(%rsp) movq %rdi, 0xb0(%rsp) leaq 0x200(%rsp), %rdi leaq 0x70(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2911ad lock decl (%rax) jne 0x2911ad movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2911a5 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2911ad movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x29121d movq 0x8(%rsp), %r15 movq 0x248(%rsp), %rcx imulq %rcx, %r15 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0x1f8(%rsp) imulq %rsi, %r15 addq 0x1b0(%r14,%rax), %r15 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x1c(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0x1f0(%rsp) movl %ebx, %r13d movl 0x14(%rsp), %eax movl %eax, 0x18(%rsp) movq %rcx, 0x1e8(%rsp) jmp 0x291257 xorl %r15d, %r15d movq $0x0, 0x1f8(%rsp) movl $0x0, 0x1c(%rsp) movq $0x0, 0x1f0(%rsp) xorl %r13d, %r13d movl $0x0, 0x18(%rsp) movq $0x0, 0x1e8(%rsp) pushq $0x6 popq %rdi callq 0x782bf movq %rax, %rbp leaq 0x268(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq %r12, %rdi xorl %esi, %esi movl 0x14(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r12, %rdi movl %ebx, %esi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r12, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r12, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r12, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r12, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r12, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r12, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r12, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x6 popq %rsi movl 0x6c(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r12, %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x110(%r14,%rax), %edx movq %r12, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x258(%rsp), %rdx movq %r12, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r12, %rsi callq *0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x291444 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x1a8, %rax # imm = 0x1A8 jne 0x2913d2 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29140c lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2914c5 lock decl (%rax) jne 0x2914c5 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2914bd movq (%rdi), %rax callq *0x18(%rax) jmp 0x2914c5 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x160, %rax # imm = 0x160 jne 0x29144b movq 0x208(%rsp), %rax testq %rax, %rax je 0x291485 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29187c lock decl (%rax) jne 0x29187c movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x291874 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29187c movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x291569 lock decl (%rax) jne 0x291569 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x291561 movq (%rdi), %rax callq *0x18(%rax) jmp 0x291569 movq %rsi, %rdi callq 0x5f3e0 movq %r15, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq 0x1f8(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x1c(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x1f0(%rsp), %rax movq %rax, 0xd8(%rsp) movl %r13d, 0xe0(%rsp) movl 0x18(%rsp), %eax movl %eax, 0xe4(%rsp) movl %r13d, 0xe8(%rsp) movl %r13d, 0xec(%rsp) movl %r13d, 0xf0(%rsp) movq 0x1e8(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2917fe andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x29166e vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x291660 movq 0x28(%rsp), %rax testq %rax, %rax je 0x29167b lock incl (%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x2916b2 lock decl (%rax) jne 0x2916b2 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x2916aa movq (%rdi), %rax callq *0x18(%rax) jmp 0x2916b2 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovaps %xmm0, 0x100(%rsp) movq 0x30(%rsp), %rax movq %rax, 0x110(%rsp) movl 0x38(%rsp), %eax movl %eax, 0x118(%rsp) movq 0x40(%rsp), %rax movq %rax, 0x120(%rsp) vmovups 0x48(%rsp), %xmm0 vmovups %xmm0, 0x128(%rsp) movl 0x58(%rsp), %eax movl %eax, 0x138(%rsp) movq 0x60(%rsp), %rax movq %rax, 0x140(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %r12d movq 0x260(%r14,%rax), %rbx movq 0x150(%rsp), %rax testq %rax, %rax je 0x291774 lock decl (%rax) jne 0x291774 movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x29176c movq (%rdi), %rax callq *0x18(%rax) jmp 0x291774 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %r12d, 0x160(%rsp) movq %rbx, 0x168(%rsp) vbroadcastss 0x166e5a(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x2917f7 lock decl (%rax) jne 0x2917f7 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2917ef movq (%rdi), %rax callq *0x18(%rax) jmp 0x2917f7 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x291a1e movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x198(%rsp), %rax testq %rax, %rax je 0x2919ca lock decl (%rax) jne 0x2919ca movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x2919c2 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2919ca movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x291c01 andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x291977 vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x291969 movq 0x28(%rsp), %rax testq %rax, %rax je 0x291984 lock incl (%rax) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x291ab6 lock decl (%rax) jne 0x291ab6 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x291aae movq (%rdi), %rax callq *0x18(%rax) jmp 0x291ab6 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x190(%rsp) andq $0x0, 0x198(%rsp) movq %r15, 0x1a0(%rsp) movl %ebx, 0x1a8(%rsp) movq %r12, 0x1b0(%rsp) vbroadcastss 0x166c05(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x1b8(%rsp) movl $0x1, 0x1c8(%rsp) movq $0x1, 0x1d0(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0x120, %r15d # imm = 0x120 movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x291a7c lock decl (%rax) jne 0x291a7c movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x291a74 movq (%rdi), %rax callq *0x18(%rax) jmp 0x291a7c movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x291a4b jmp 0x291d4f movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovups %xmm0, 0xb8(%rsp) movq 0x30(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x38(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x40(%rsp), %rax movq %rax, 0xd8(%rsp) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movl 0x58(%rsp), %eax movl %eax, 0xf0(%rsp) movq 0x60(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %ebx movq 0x260(%r14,%rax), %r12 movq 0x108(%rsp), %rax testq %rax, %rax je 0x291b78 lock decl (%rax) jne 0x291b78 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x291b70 movq (%rdi), %rax callq *0x18(%rax) jmp 0x291b78 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x100(%rsp) andq $0x0, 0x108(%rsp) movq %r15, 0x110(%rsp) movl %ebx, 0x118(%rsp) movq %r12, 0x120(%rsp) vbroadcastss 0x166a57(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x128(%rsp) movl $0x1, 0x138(%rsp) movq $0x1, 0x140(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x291bfa lock decl (%rax) jne 0x291bfa movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x291bf2 movq (%rdi), %rax callq *0x18(%rax) jmp 0x291bfa movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x291cc4 movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x150(%rsp), %rax testq %rax, %rax je 0x291c70 lock decl (%rax) jne 0x291c70 movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x291c68 movq (%rdi), %rax callq *0x18(%rax) jmp 0x291c70 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %ebx, 0x160(%rsp) movq %r12, 0x168(%rsp) vbroadcastss 0x16695f(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0xd8, %r15d movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x291d22 lock decl (%rax) jne 0x291d22 movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x291d1a movq (%rdi), %rax callq *0x18(%rax) jmp 0x291d22 movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x291cf1 movq (%rbp), %rax movq %rbp, %rdi movq 0x250(%rsp), %rsi callq *0x20(%rax) leaq 0x268(%rsp), %rdi movq 0x1e0(%rsp), %rax movq (%rax), %rax movq 0x8(%rsp), %rcx movq %rbp, (%rax,%rcx,8) callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x291db9 lock decl (%rax) jne 0x291db9 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi je 0x291db1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x291db9 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rdx incq %rdx jmp 0x2910c7 xorl %eax, %eax addq $0x278, %rsp # imm = 0x278 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x291e8e jmp 0x292025 jmp 0x291de6 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x291ea0 lock decl (%rax) jne 0x291ea0 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x291e1c movq %rsi, %rdi callq 0x5f3e0 jmp 0x291ea0 movq (%rdi), %rax callq *0x18(%rax) jmp 0x291ea0 jmp 0x292025 jmp 0x291f18 jmp 0x291e8e jmp 0x291e32 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x291f2a lock decl (%rax) jne 0x291f2a movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x291e68 movq %rsi, %rdi callq 0x5f3e0 jmp 0x291f2a movq (%rdi), %rax callq *0x18(%rax) jmp 0x291f2a jmp 0x292025 jmp 0x292025 jmp 0x292025 jmp 0x292025 jmp 0x291e8e jmp 0x291f18 movq %rax, %rbx jmp 0x291ea0 movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0xd8, %r14d vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x291ee3 lock decl (%rax) jne 0x291ee3 movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x291ed7 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x291ee3 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x291eaa jmp 0x291fd9 jmp 0x292025 jmp 0x291f18 movq %rax, %rbx jmp 0x291f2a movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0x120, %r14d # imm = 0x120 vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x291f6d lock decl (%rax) jne 0x291f6d movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x291f61 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x291f6d movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x291f34 jmp 0x291fd9 jmp 0x292025 jmp 0x291fcd movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x29201d lock decl (%rax) jne 0x29201d movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29200d movq (%rdi), %rax callq *0x18(%rax) jmp 0x29201d jmp 0x292025 movq %rax, %rbx jmp 0x291fe6 jmp 0x292025 jmp 0x292025 movq %rax, %rbx leaq 0x268(%rsp), %rdi callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29201d lock decl (%rax) jne 0x29201d movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi jne 0x292017 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29201d movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp

ncnn::ConvolutionDepthWise_x86_avx512::destroy_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx512::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }

pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x292071 movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x29206c movq (%rdi), %rax callq *0x8(%rax) andq $0x0, 0x8(%rbx) xorl %r15d, %r15d movq 0x10(%rbx), %rax movq 0x18(%rbx), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r15 jge 0x2920b3 movq (%rax,%r15,8), %rdi movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x10(%rbx), %rax movq (%rax,%r15,8), %rdi testq %rdi, %rdi je 0x2920ae movq (%rdi), %rax callq *0x8(%rax) incq %r15 jmp 0x292074 cmpq %rax, %rcx je 0x2920bc movq %rax, 0x18(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %r15 retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_avx512::destroy_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx512::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }

pushq %rax movq (%rdi), %rax addq -0x38(%rax), %rdi callq 0x292040 xorl %eax, %eax popq %rcx retq nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp

ncnn::ConvolutionDepthWise_x86_avx512::forward_int8_x86(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx512::forward_int8_x86(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; int elempack = bottom_blob.elempack; int elembits = bottom_blob.elembits(); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_int8 = bottom_blob; if (elembits != 8) { const int channels_g = channels * elempack / group; Mat scales(channels * elempack); { float* ps = scales; for (int g = 0; g < group; g++) { float scale = bottom_blob_int8_scales[g]; for (int q = 0; q < channels_g; q++) { *ps++ = scale; } } } Option opt_q = opt; opt_q.blob_allocator = opt.workspace_allocator; quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q); } Mat bottom_blob_bordered; make_padding(bottom_blob_int8, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; channels = bottom_blob_bordered.c; elempack = bottom_blob_bordered.elempack; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; // depth-wise if (channels * elempack == group && group == num_output) { int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { out_elempack = num_output % 8 == 0 ? 8 : 1; } #endif // __SSE2__ bool use_int8_requantize = int8_scale_term > 100; size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #if __SSE2__ if (elempack == 8) { { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128i _sum0 = _mm_setzero_si128(); __m128i _sum1 = _mm_setzero_si128(); const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { // TODO use _mm_cvtepi8_epi16 on sse4.1 __m128i _val = _mm_loadl_epi64((const __m128i*)(sptr + space_ofs[k] * 8)); _val = _mm_unpacklo_epi8(_val, _mm_cmpgt_epi8(_mm_setzero_si128(), _val)); __m128i _w = _mm_loadl_epi64((const __m128i*)(kptr + k * 8)); _w = _mm_unpacklo_epi8(_w, _mm_cmpgt_epi8(_mm_setzero_si128(), _w)); __m128i _sl = _mm_mullo_epi16(_val, _w); __m128i _sh = _mm_mulhi_epi16(_val, _w); __m128i _s0 = _mm_unpacklo_epi16(_sl, _sh); __m128i _s1 = _mm_unpackhi_epi16(_sl, _sh); _sum0 = _mm_add_epi32(_sum0, _s0); _sum1 = _mm_add_epi32(_sum1, _s1); } __m128 _scale_in0; __m128 _scale_in1; { __m128 _bottom_blob_int8_scales0 = _mm_loadu_ps((const float*)bottom_blob_int8_scales + g * 8); __m128 _bottom_blob_int8_scales1 = _mm_loadu_ps((const float*)bottom_blob_int8_scales + g * 8 + 4); __m128 _weight_data_int8_scales0 = _mm_loadu_ps((const float*)weight_data_int8_scales + g * 8); __m128 _weight_data_int8_scales1 = _mm_loadu_ps((const float*)weight_data_int8_scales + g * 8 + 4); _scale_in0 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales0, _weight_data_int8_scales0)); _scale_in1 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales1, _weight_data_int8_scales1)); __m128 _m0 = _mm_cmpneq_ps(_weight_data_int8_scales0, _mm_setzero_ps()); __m128 _m1 = _mm_cmpneq_ps(_weight_data_int8_scales1, _mm_setzero_ps()); _scale_in0 = _mm_and_ps(_scale_in0, _m0); _scale_in1 = _mm_and_ps(_scale_in1, _m1); } __m128 _sumfp32_0 = _mm_mul_ps(_mm_cvtepi32_ps(_sum0), _scale_in0); __m128 _sumfp32_1 = _mm_mul_ps(_mm_cvtepi32_ps(_sum1), _scale_in1); if (bias_term) { __m128 _bias0 = _mm_loadu_ps((const float*)bias_data + g * 8); __m128 _bias1 = _mm_loadu_ps((const float*)bias_data + g * 8 + 4); _sumfp32_0 = _mm_add_ps(_sumfp32_0, _bias0); _sumfp32_1 = _mm_add_ps(_sumfp32_1, _bias1); } _sumfp32_0 = activation_sse(_sumfp32_0, activation_type, activation_params); _sumfp32_1 = activation_sse(_sumfp32_1, activation_type, activation_params); if (use_int8_requantize) { // requantize and relu __m128 _scale_out0 = _mm_loadu_ps((const float*)top_blob_int8_scales + g * 8); __m128 _scale_out1 = _mm_loadu_ps((const float*)top_blob_int8_scales + g * 8 + 4); _sumfp32_0 = _mm_mul_ps(_sumfp32_0, _scale_out0); _sumfp32_1 = _mm_mul_ps(_sumfp32_1, _scale_out1); int64_t _sum8 = float2int8_sse(_sumfp32_0, _sumfp32_1); *(int64_t*)outptr_s8 = _sum8; outptr_s8 += 8; } else { // dequantize and relu _mm_storeu_ps(outptr_f32, _sumfp32_0); _mm_storeu_ps(outptr_f32 + 4, _sumfp32_1); outptr_f32 += 8; } } } } } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && stride_w == 1 && stride_h == 1 && dilation_w == 1 && dilation_h == 1 && (activation_type == 0 || activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s1_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s1_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2 && (activation_type == 0 || activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s2_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s2_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char*)weight_data_tm + maxk * g; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr_s8[0] = sums8; outptr_s8 += 1; } else { // dequantize outptr_f32[0] = sumfp32; outptr_f32 += 1; } } } } } } return 0; } bool use_int8_requantize = int8_scale_term > 100; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { if (use_int8_requantize) out_elempack = num_output % 8 == 0 ? 8 : 1; else out_elempack = num_output % 4 == 0 ? 4 : 1; } #endif // __SSE2__ size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { g_elempack = channels_g % 8 == 0 ? 8 : 1; if (use_int8_requantize) out_g_elempack = num_output_g % 8 == 0 ? 8 : 1; else out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x318, %rsp # imm = 0x318 movq %rcx, 0x30(%rsp) movq %rdx, 0x20(%rsp) movq %rsi, %r15 movl 0x38(%rsi), %ecx movl 0x18(%rsi), %r12d movq 0x10(%rsi), %rsi testl %r12d, %r12d je 0x2975b9 leal (,%rsi,8), %eax cltd idivl %r12d cmpl $0x8, %eax sete %dl jmp 0x2975bb xorl %edx, %edx movq (%rdi), %rax movq -0x18(%rax), %r9 movq %rdi, %r8 movl 0xd4(%rdi,%r9), %ebp movl 0xd8(%rdi,%r9), %ebx decl %ebp imull 0xdc(%rdi,%r9), %ebp decl %ebx imull 0xe0(%rdi,%r9), %ebx movq 0x8(%r15), %rdi vmovups (%r15), %xmm0 vmovaps %xmm0, 0x1e0(%rsp) movq %rsi, 0x1f0(%rsp) movl %r12d, 0x1f8(%rsp) movq 0x20(%r15), %rsi movq %rsi, 0x200(%rsp) vmovdqu 0x28(%r15), %xmm0 vmovdqu %xmm0, 0x208(%rsp) movl %ecx, 0x218(%rsp) movq 0x40(%r15), %rsi movq %rsi, 0x220(%rsp) testq %rdi, %rdi je 0x297646 lock incl (%rdi) movq (%r8), %rax movq %r8, 0x10(%rsp) testb %dl, %dl je 0x29765e movq 0x20(%rsp), %r15 movq 0x10(%rsp), %rcx jmp 0x297770 imull %ecx, %r12d movq -0x18(%rax), %rax movq 0x10(%rsp), %r13 movl 0x108(%r13,%rax), %eax movl %eax, 0x18(%rsp) leaq 0x70(%rsp), %rdi andq $0x0, 0x40(%rdi) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdi) vmovdqu %xmm0, 0xc(%rdi) vmovdqa %xmm0, 0x20(%rdi) vmovdqu %xmm0, 0x2c(%rdi) xorl %r14d, %r14d pushq $0x4 popq %rdx movl %r12d, %esi xorl %ecx, %ecx callq 0x635fa movl %r12d, %eax cltd idivl 0x18(%rsp) movq 0x70(%rsp), %rcx movq (%r13), %rdx testl %eax, %eax cmovlel %r14d, %eax movq 0x10(%rsp), %r8 movq -0x18(%rdx), %rsi movslq 0x108(%r8,%rsi), %rdi cmpq %rdi, %r14 jge 0x2976f9 movq 0x240(%r8,%rsi), %rsi vmovd (%rsi,%r14,4), %xmm0 movl %eax, %esi subl $0x1, %esi jb 0x2976f4 vmovd %xmm0, (%rcx) addq $0x4, %rcx jmp 0x2976e5 incq %r14 jmp 0x2976c4 movq 0x30(%rsp), %rax vmovups (%rax), %zmm0 leaq 0x130(%rsp), %rcx vmovups %zmm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x1e0(%rsp), %rsi leaq 0x70(%rsp), %rdx movq %r15, %rdi vzeroupper callq 0x652e3 movq 0x78(%rsp), %rax testq %rax, %rax movq 0x20(%rsp), %r15 je 0x297768 lock decl (%rax) jne 0x297768 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x297760 movq (%rdi), %rax callq *0x18(%rax) jmp 0x297768 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %rcx movq (%rcx), %rax leaq 0x70(%rsp), %rdx andq $0x0, 0x40(%rdx) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdx) vmovdqu %xmm0, 0xc(%rdx) vmovdqa %xmm0, 0x20(%rdx) vmovdqu %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %rcx, %rdi leaq 0x1e0(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %rax cmpq $0x0, 0x70(%rsp) movl %eax, 0x48(%rsp) je 0x298fc6 movslq 0xa8(%rsp), %r14 movq 0xb0(%rsp), %rax imulq %r14, %rax testq %rax, %rax je 0x298fc6 notl %ebp movl 0x88(%rsp), %r13d movl 0x9c(%rsp), %r12d addl %r12d, %ebp movq 0x10(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx movl %ebp, %eax movq %rdx, %rbp cltd idivl 0xe4(%rbp,%rcx) movl %eax, %esi notl %ebx addl 0xa0(%rsp), %ebx movl %ebx, %eax cltd idivl 0xe8(%rbp,%rcx) movl %eax, %edx movq %rsi, 0x18(%rsp) incl %esi movq %rdx, 0x60(%rsp) leal 0x1(%rdx), %r10d movl %r13d, %ebx imull %r14d, %ebx pushq $0x8 popq %r11 cmpl 0x108(%rbp,%rcx), %ebx jne 0x297980 cmpl 0xd0(%rbp,%rcx), %ebx jne 0x297980 movq 0x30(%rsp), %rdi cmpb $0x0, 0x27(%rdi) pushq $0x1 popq %rax cmovel %eax, %r11d testb $0x7, %bl cmovnel %eax, %r11d movl 0x10c(%rbp,%rcx), %eax leal (,%r11,4), %r8d movl %eax, 0x38(%rsp) cmpl $0x65, %eax cmovgel %r11d, %r8d movl %ebx, %eax cltd idivl %r11d movq 0x8(%rdi), %rcx movq %rcx, (%rsp) movq %r15, %rdi movl %r10d, %edx movl %eax, %ecx movl %r11d, %r9d callq 0x628f2 cmpq $0x0, (%r15) je 0x298fc6 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x298fc6 cmpl $0x1, %r13d je 0x297f2e movl $0x0, 0x48(%rsp) cmpl $0x8, %r13d jne 0x298fc6 movq (%rbp), %rax movq -0x18(%rax), %rax movslq 0xd4(%rbp,%rax), %rcx movslq 0xd8(%rbp,%rax), %r15 imulq %rcx, %r15 leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %r15, %rsi callq 0x73bbe movq (%rbp), %rcx movq -0x18(%rcx), %rdx imull 0xe0(%rbp,%rdx), %r12d movq 0x130(%rsp), %rax movl 0xdc(%rbp,%rdx), %esi imull 0xd4(%rbp,%rdx), %esi subl %esi, %r12d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%rbp,%rdx), %esi jge 0x298067 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%rbp,%rdx), %r9d jge 0x297976 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%rbp,%rdx), %edi incq %r9 jmp 0x297958 addl %r12d, %edi incl %esi addq %r9, %r8 jmp 0x297941 movl 0xd0(%rbp,%rcx), %eax movl 0x10c(%rbp,%rcx), %r14d movq 0x30(%rsp), %rcx cmpb $0x1, 0x27(%rcx) jne 0x2979ac cmpl $0x65, %r14d jl 0x2979b2 testb $0x7, %al pushq $0x1 popq %r9 cmovel %r11d, %r9d jmp 0x2979c0 pushq $0x1 popq %r9 jmp 0x2979c0 xorl %ecx, %ecx testb $0x3, %al sete %cl leal (%rcx,%rcx,2), %r9d incl %r9d leal (,%r9,4), %r12d cmpl $0x65, %r14d cmovgel %r9d, %r12d cltd idivl %r9d movq 0x30(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq %r15, %rdi movl %esi, 0x28(%rsp) movl %r10d, 0x40(%rsp) movl %r10d, %edx movl %eax, %ecx movq %r12, %r8 movq %r9, 0x38(%rsp) callq 0x628f2 cmpq $0x0, (%r15) je 0x298fc6 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x298fc6 movq (%rbp), %rax movq -0x18(%rax), %rax movl 0xd0(%rbp,%rax), %ecx movl 0x108(%rbp,%rax), %esi movl %ebx, %eax cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %ebp movq 0x30(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x297a73 testb $0x7, %dil pushq $0x1 popq %rbx movl %ebx, %r15d pushq $0x8 popq %rax cmovel %eax, %r15d cmpl $0x65, %r14d movl 0x28(%rsp), %esi movl 0x40(%rsp), %r9d jl 0x297a85 testb $0x7, %bpl cmovel %eax, %ebx jmp 0x297a93 pushq $0x1 popq %r15 movl %r15d, %ebx movl 0x28(%rsp), %esi movl 0x40(%rsp), %r9d jmp 0x297a93 xorl %eax, %eax testb $0x3, %bpl sete %al leal (%rax,%rax,2), %ebx incl %ebx movq 0x78(%rsp), %rax vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0x130(%rsp) movq 0x80(%rsp), %rcx movq %rcx, 0x140(%rsp) movl 0x88(%rsp), %ecx movl %ecx, 0x148(%rsp) movq 0x90(%rsp), %rcx movq %rcx, 0x150(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0x158(%rsp) movl 0xa8(%rsp), %ecx movl %ecx, 0x168(%rsp) movq 0xb0(%rsp), %rcx movq %rcx, 0x170(%rsp) testq %rax, %rax je 0x297b0d lock incl (%rax) cmpl %r15d, %r13d movl %edi, 0x18(%rsp) jle 0x297b5c movq 0x30(%rsp), %rax vmovups (%rax), %zmm0 leaq 0xe0(%rsp), %rcx vmovups %zmm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x70(%rsp), %rdi leaq 0x130(%rsp), %rsi movl %r15d, %edx vzeroupper callq 0x64e3b movl 0x18(%rsp), %edi movl 0x28(%rsp), %esi movl 0x40(%rsp), %r9d movq 0x20(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movq 0x10(%rdx), %rcx movq %rcx, 0xf0(%rsp) movl 0x18(%rdx), %ecx movl %ecx, 0xf8(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x100(%rsp) vmovups 0x28(%rdx), %xmm0 vmovups %xmm0, 0x108(%rsp) movl 0x38(%rdx), %ecx movl %ecx, 0x118(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x120(%rsp) testq %rax, %rax je 0x297bc0 lock incl (%rax) cmpl 0x38(%rsp), %ebx jae 0x297c3f movq 0x10(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xd0(%rcx,%rax), %eax cltd idivl %ebx movl %eax, %ecx movzbl %r12b, %eax divb 0x38(%rsp) movzbl %al, %r8d imull %ebx, %r8d movq 0x30(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0xe0(%rsp), %rdi movl %r9d, %edx movl %ebx, %r9d callq 0x628f2 movl 0x18(%rsp), %edi pushq $-0x64 popq %r14 cmpq $0x0, 0xe0(%rsp) je 0x298f53 movslq 0x118(%rsp), %rax imulq 0x120(%rsp), %rax testq %rax, %rax je 0x298f53 xorl %r12d, %r12d xorl %r13d, %r13d xorl %r14d, %r14d movq 0x10(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rax movslq 0x108(%r8,%rax), %rax cmpq %rax, %r14 jge 0x297eaa movl %r12d, %eax cltd idivl %r15d movl %eax, %ecx movl %edi, %eax cltd idivl %r15d movslq %ecx, %rdx imulq 0x170(%rsp), %rdx movq 0x140(%rsp), %rcx imulq %rcx, %rdx addq 0x130(%rsp), %rdx movq %rdx, 0x190(%rsp) andq $0x0, 0x198(%rsp) movl 0x148(%rsp), %edx movq 0x150(%rsp), %rsi movq %rcx, 0x1a0(%rsp) movl %edx, 0x1a8(%rsp) movq %rsi, 0x1b0(%rsp) movl %eax, 0x1c8(%rsp) vmovups 0x158(%rsp), %xmm0 movslq 0x164(%rsp), %rax movslq 0x15c(%rsp), %rdx movslq 0x160(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x1d0(%rsp) vmovups %xmm0, 0x1b8(%rsp) movl %r13d, %eax cltd idivl %ebx movl %eax, %ecx movl %ebp, %eax cltd idivl %ebx movslq %ecx, %rdx imulq 0x120(%rsp), %rdx movq 0xf0(%rsp), %rcx imulq %rcx, %rdx addq 0xe0(%rsp), %rdx movq %rdx, 0x228(%rsp) andq $0x0, 0x230(%rsp) movl 0xf8(%rsp), %edx movq 0x100(%rsp), %rsi movq %rcx, 0x238(%rsp) movl %edx, 0x240(%rsp) movq %rsi, 0x248(%rsp) movl %eax, 0x260(%rsp) vmovups 0x108(%rsp), %xmm0 movslq 0x114(%rsp), %rax movslq 0x10c(%rsp), %rdx movslq 0x110(%rsp), %rdi imulq %rdx, %rdi imulq %rcx, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x268(%rsp) vmovups %xmm0, 0x250(%rsp) movq 0x10(%r8), %rax movq (%rax,%r14,8), %rdi movq 0x30(%rsp), %rax vmovups (%rax), %zmm0 vmovups %zmm0, 0x2d0(%rsp) movq %rsi, 0x2d8(%rsp) movq (%rdi), %rax leaq 0x190(%rsp), %rsi leaq 0x228(%rsp), %rdx leaq 0x2d0(%rsp), %rcx vzeroupper callq *0x38(%rax) movq 0x230(%rsp), %rax testq %rax, %rax movl 0x18(%rsp), %edi je 0x297e5d lock decl (%rax) jne 0x297e5d movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi je 0x297e51 movq (%rdi), %rax callq *0x18(%rax) movl 0x18(%rsp), %edi jmp 0x297e5d movq %rsi, %rdi callq 0x5f3e0 movl 0x18(%rsp), %edi movq 0x198(%rsp), %rax testq %rax, %rax je 0x297e9c lock decl (%rax) jne 0x297e9c movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x297e90 movq (%rdi), %rax callq *0x18(%rax) movl 0x18(%rsp), %edi jmp 0x297e9c movq %rsi, %rdi callq 0x5f3e0 movl 0x18(%rsp), %edi incq %r14 addl %ebp, %r13d addl %edi, %r12d jmp 0x297c48 movq 0x38(%rsp), %rdx xorl %r14d, %r14d cmpl %edx, %ebx jae 0x297ed2 leaq 0xe0(%rsp), %rdi movq 0x20(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x64e3b jmp 0x298f53 leaq 0xe0(%rsp), %rax movq 0x20(%rsp), %rcx cmpq %rcx, %rax je 0x298f53 movq 0xe8(%rsp), %rax testq %rax, %rax je 0x297ef8 lock incl (%rax) movq 0x8(%rcx), %rax testq %rax, %rax je 0x298efb lock decl (%rax) jne 0x298efb movq 0x20(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x298ef3 movq (%rdi), %rax callq *0x18(%rax) jmp 0x298efb movq (%rbp), %rax movq -0x18(%rax), %rdx movl 0xd4(%rbp,%rdx), %ecx movl 0xd8(%rbp,%rdx), %r13d movl %ecx, %esi xorl $0x3, %esi movl %r13d, %edi xorl $0x3, %edi orl %esi, %edi jne 0x298ab0 cmpl $0x1, 0xe4(%rbp,%rdx) jne 0x29899d cmpl $0x1, 0xe8(%rbp,%rdx) jne 0x29899d cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x29899d cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x29899d cmpl $0x1, 0x110(%rbp,%rdx) ja 0x29899d movl $0x0, 0x48(%rsp) cmpl $0x65, 0x38(%rsp) jl 0x299046 vpxor %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %r13 vmovdqa %xmm0, (%r13) andq $0x0, 0x10(%r13) xorl %ebx, %ebx movq %rbp, %rcx leaq 0xe0(%rsp), %rbp leaq 0x190(%rsp), %r12 movq -0x18(%rax), %r14 movslq 0x108(%rcx,%r14), %rax cmpq %rax, %rbx jge 0x299347 movq 0x1f8(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm1 vpxor %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x298022 movq 0x240(%rcx,%r14), %rax vmulss (%rax,%rbx,4), %xmm1, %xmm0 vmovss 0x156c6a(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rcx,%r14), %rax vmovd (%rax,%rbx,4), %xmm0 vmovd %xmm0, 0x190(%rsp) movq %r13, %rdi movq %rbp, %rsi callq 0x1ea12c movq %r13, %rdi movq %r12, %rsi callq 0x1ea12c incq %rbx movq 0x10(%rsp), %rcx movq (%rcx), %rax jmp 0x297fdd leal (,%r15,8), %ecx movl %ecx, 0x40(%rsp) xorl %edx, %edx testl %r15d, %r15d cmovlel %edx, %r15d testl %r14d, %r14d cmovlel %edx, %r14d vpxor %xmm0, %xmm0, %xmm0 vbroadcastss 0x159124(%rip), %xmm1 # 0x3f11b4 vpbroadcastd 0x155f7b(%rip), %xmm2 # 0x3ee014 vpbroadcastw 0x1604fd(%rip), %xmm25 # 0x3f85a0 vpbroadcastw 0x1604f5(%rip), %xmm26 # 0x3f85a2 vbroadcastss 0x159101(%rip), %xmm27 # 0x3f11b8 vbroadcastss 0x1590fb(%rip), %xmm28 # 0x3f11bc vbroadcastss 0x1590f6(%rip), %xmm7 # 0x3f11c0 vbroadcastss 0x155f41(%rip), %xmm8 # 0x3ee014 vbroadcastss 0x156bac(%rip), %xmm9 # 0x3eec88 vbroadcastss 0x1590df(%rip), %xmm10 # 0x3f11c4 vbroadcastss 0x1590da(%rip), %xmm11 # 0x3f11c8 vbroadcastss 0x1590d8(%rip), %xmm24 # 0x3f11d0 vbroadcastss 0x1590cb(%rip), %xmm13 # 0x3f11cc vbroadcastss 0x1590ca(%rip), %xmm14 # 0x3f11d4 vbroadcastss 0x1590c5(%rip), %xmm15 # 0x3f11d8 vbroadcastss 0x1590bf(%rip), %xmm16 # 0x3f11dc vpbroadcastd 0x156b61(%rip), %xmm17 # 0x3eec88 xorl %esi, %esi movq %r14, 0x28(%rsp) vbroadcastss 0x1590bc(%rip), %xmm20 # 0x3f11f4 vbroadcastss 0x1590ba(%rip), %xmm23 # 0x3f11fc cmpq %r14, %rsi je 0x298ed9 movq 0x20(%rsp), %rcx movq 0x10(%rcx), %r8 imulq %rsi, %r8 imulq 0x40(%rcx), %r8 movq %rdx, 0x30(%rsp) movslq %edx, %r9 addq (%rcx), %r8 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rsi, %r11 movq 0x80(%rsp), %rcx imulq %rcx, %r11 addq 0x70(%rsp), %r11 imulq %rcx, %r10 movq %rsi, 0x48(%rsp) leaq (,%rsi,8), %rbx addq 0x28(%rbp), %r9 xorl %ecx, %ecx movq %r8, %rdx cmpl 0x60(%rsp), %ecx jg 0x298982 movq (%rbp), %rdi xorl %r13d, %r13d cmpl 0x18(%rsp), %r13d jg 0x29897b movq -0x18(%rdi), %rsi movslq 0xe8(%rbp,%rsi), %r14 movslq %ecx, %r12 imulq %r14, %r12 imulq %r10, %r12 movl 0xe4(%rbp,%rsi), %ebp imull %r13d, %ebp shll $0x3, %ebp movslq %ebp, %r14 addq %r11, %r14 addq %r12, %r14 vpxor %xmm3, %xmm3, %xmm3 xorl %r12d, %r12d vpxor %xmm4, %xmm4, %xmm4 cmpq %r12, %r15 je 0x29823f movslq (%rax,%r12,4), %rbp vmovq (%r14,%rbp,8), %xmm5 vpcmpgtb %xmm5, %xmm0, %xmm6 vpunpcklbw %xmm6, %xmm5, %xmm5 # xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7] vmovq (%r9,%r12,8), %xmm6 vpcmpgtb %xmm6, %xmm0, %xmm12 vpunpcklbw %xmm12, %xmm6, %xmm6 # xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7] vpmullw %xmm5, %xmm6, %xmm12 vpmulhw %xmm6, %xmm5, %xmm5 vpunpcklwd %xmm5, %xmm12, %xmm6 # xmm6 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3] vpunpckhwd %xmm5, %xmm12, %xmm5 # xmm5 = xmm12[4],xmm5[4],xmm12[5],xmm5[5],xmm12[6],xmm5[6],xmm12[7],xmm5[7] vpaddd %xmm6, %xmm4, %xmm4 vpaddd %xmm5, %xmm3, %xmm3 incq %r12 jmp 0x2981fc movq 0x10(%rsp), %rbp movq 0x1f8(%rbp,%rsi), %r14 movq 0x240(%rbp,%rsi), %r12 vmovups (%r14,%rbx,4), %xmm21 vmovups 0x10(%r14,%rbx,4), %xmm22 vmulps (%r12,%rbx,4), %xmm21, %xmm5 vmulps 0x10(%r12,%rbx,4), %xmm22, %xmm6 vrcpps %xmm5, %xmm5 vrcpps %xmm6, %xmm6 vcmpneqps %xmm0, %xmm21, %k1 vcmpneqps %xmm0, %xmm22, %k2 vmovaps %xmm5, %xmm5 {%k1} {z} vmovaps %xmm6, %xmm6 {%k2} {z} vcvtdq2ps %xmm4, %xmm4 vmulps %xmm4, %xmm5, %xmm22 vcvtdq2ps %xmm3, %xmm3 vmulps %xmm3, %xmm6, %xmm21 cmpl $0x0, 0x100(%rbp,%rsi) je 0x2982c9 movq 0x1b0(%rbp,%rsi), %r14 vaddps (%r14,%rbx,4), %xmm22, %xmm22 vaddps 0x10(%r14,%rbx,4), %xmm21, %xmm21 movl 0x110(%rbp,%rsi), %r14d decl %r14d cmpl $0x5, %r14d ja 0x2988fa leaq 0x1602d7(%rip), %r12 # 0x3f85bc movslq (%r12,%r14,4), %r14 addq %r12, %r14 vmaxps %xmm0, %xmm22, %xmm3 vmaxps %xmm0, %xmm21, %xmm4 jmpq *%r14 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm5 vminps %xmm0, %xmm22, %xmm6 vfmadd231ps %xmm6, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm6) + xmm3 vminps %xmm0, %xmm21, %xmm6 vfmadd231ps %xmm6, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm6) + xmm4 jmp 0x298906 vminps %xmm27, %xmm22, %xmm3 vmaxps %xmm28, %xmm3, %xmm3 vmovaps %xmm7, %xmm4 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm5 vcvtdq2ps %xmm5, %xmm5 vcmpltps %xmm5, %xmm4, %k1 vsubps %xmm9, %xmm5, %xmm5 {%k1} vfmsub231ps %xmm10, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm10) - xmm3 vfnmsub231ps %xmm11, %xmm5, %xmm3 # xmm3 = -(xmm5 * xmm11) - xmm3 vmulps %xmm3, %xmm3, %xmm4 vmovaps %xmm13, %xmm6 vfmadd213ps %xmm24, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm24 vfmadd213ps %xmm14, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm14 vfmadd213ps %xmm15, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm15 vfmadd213ps %xmm16, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm16 vfmadd213ps %xmm8, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm8 vfmadd213ps %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vaddps %xmm6, %xmm9, %xmm3 vcvttps2dq %xmm5, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm17, %xmm4, %xmm4 vfmadd213ps %xmm9, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm9 vcmpleps %xmm0, %xmm4, %k1 vbroadcastss 0x158e38(%rip), %xmm29 # 0x3f11e0 vmaxps %xmm29, %xmm4, %xmm3 vpsrld $0x17, %xmm3, %xmm4 vpbroadcastd 0x158e27(%rip), %xmm30 # 0x3f11e4 vpternlogd $0xec, %xmm30, %xmm2, %xmm3 vmovaps %xmm1, %xmm30 vmovaps %xmm13, %xmm1 vmovdqa64 %xmm17, %xmm13 vmovaps %xmm16, %xmm17 vmovaps %xmm15, %xmm16 vmovaps %xmm14, %xmm15 vmovaps %xmm7, %xmm14 vmovaps %xmm24, %xmm7 vpbroadcastd 0x158def(%rip), %xmm24 # 0x3f11e8 vpaddd %xmm24, %xmm4, %xmm4 vcvtdq2ps %xmm4, %xmm4 vbroadcastss 0x158ddf(%rip), %xmm31 # 0x3f11ec vcmpltps %xmm31, %xmm3, %k2 vbroadcastss 0x158dd2(%rip), %xmm18 # 0x3f11f0 vaddps %xmm18, %xmm3, %xmm5 vsubps %xmm9, %xmm4, %xmm4 {%k2} vaddps %xmm3, %xmm5, %xmm5 {%k2} vmulps %xmm5, %xmm5, %xmm3 vmovaps %xmm20, %xmm6 vbroadcastss 0x158db4(%rip), %xmm19 # 0x3f11f8 vfmadd213ps %xmm19, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm19 vfmadd213ps %xmm23, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm23 vbroadcastss 0x158da6(%rip), %xmm18 # 0x3f1200 vfmadd213ps %xmm18, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm18 vbroadcastss 0x158d9a(%rip), %xmm19 # 0x3f1204 vfmadd213ps %xmm19, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm19 vbroadcastss 0x158d8e(%rip), %xmm24 # 0x3f1208 vfmadd213ps %xmm24, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm24 vbroadcastss 0x158d82(%rip), %xmm29 # 0x3f120c vfmadd213ps %xmm29, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm29 vbroadcastss 0x158d76(%rip), %xmm31 # 0x3f1210 vfmadd213ps %xmm31, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm31 vbroadcastss 0x158d6b(%rip), %xmm12 # 0x3f1214 vfmadd213ps %xmm12, %xmm5, %xmm6 # xmm6 = (xmm5 * xmm6) + xmm12 vmulps %xmm5, %xmm3, %xmm12 vmulps %xmm6, %xmm12, %xmm6 vfmadd231ps %xmm11, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm11) + xmm6 vfmsub231ps %xmm3, %xmm8, %xmm6 # xmm6 = (xmm8 * xmm3) - xmm6 vsubps %xmm5, %xmm6, %xmm3 vfmsub231ps %xmm4, %xmm10, %xmm3 # xmm3 = (xmm10 * xmm4) - xmm3 vbroadcastss 0x15a246(%rip), %xmm4 # 0x3f2718 vmulps %xmm4, %xmm3, %xmm3 vbroadcastss 0x159421(%rip), %xmm4 # 0x3f1900 vmovaps %xmm4, %xmm3 {%k1} vminps %xmm27, %xmm3, %xmm3 vmaxps %xmm28, %xmm3, %xmm3 vmovaps %xmm14, %xmm4 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm5 vcvtdq2ps %xmm5, %xmm5 vcmpltps %xmm5, %xmm4, %k1 vsubps %xmm9, %xmm5, %xmm5 {%k1} vfmsub231ps %xmm10, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm10) - xmm3 vfnmsub231ps %xmm11, %xmm5, %xmm3 # xmm3 = -(xmm5 * xmm11) - xmm3 vmulps %xmm3, %xmm3, %xmm4 vmovaps %xmm1, %xmm6 vfmadd213ps %xmm7, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm7 vfmadd213ps %xmm15, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm15 vfmadd213ps %xmm16, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm16 vfmadd213ps %xmm17, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm17 vfmadd213ps %xmm8, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm8 vfmadd213ps %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vaddps %xmm6, %xmm9, %xmm4 vcvttps2dq %xmm5, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm3, %xmm13, %xmm3 vfmadd213ps %xmm9, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) + xmm9 vrcpps %xmm3, %xmm4 vaddps %xmm4, %xmm4, %xmm5 vbroadcastss 0x15a1a0(%rip), %xmm6 # 0x3f2708 vfmsub213ps %xmm6, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm3) - xmm6 vfnmadd213ps %xmm5, %xmm4, %xmm3 # xmm3 = -(xmm4 * xmm3) + xmm5 vminps %xmm27, %xmm21, %xmm4 vmaxps %xmm28, %xmm4, %xmm4 vmovaps %xmm14, %xmm5 vfmadd213ps %xmm8, %xmm4, %xmm5 # xmm5 = (xmm4 * xmm5) + xmm8 vcvttps2dq %xmm5, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm5, %k1 vsubps %xmm9, %xmm6, %xmm6 {%k1} vfmsub231ps %xmm10, %xmm6, %xmm4 # xmm4 = (xmm6 * xmm10) - xmm4 vfnmsub231ps %xmm11, %xmm6, %xmm4 # xmm4 = -(xmm6 * xmm11) - xmm4 vmulps %xmm4, %xmm4, %xmm5 vmovaps %xmm1, %xmm12 vfmadd213ps %xmm7, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm7 vfmadd213ps %xmm15, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm15 vfmadd213ps %xmm16, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm16 vfmadd213ps %xmm17, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm17 vfmadd213ps %xmm8, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm8 vfmadd213ps %xmm4, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm4 vaddps %xmm9, %xmm12, %xmm4 vcvttps2dq %xmm6, %xmm5 vpslld $0x17, %xmm5, %xmm5 vpaddd %xmm5, %xmm13, %xmm5 vfmadd213ps %xmm9, %xmm4, %xmm5 # xmm5 = (xmm4 * xmm5) + xmm9 vcmpleps %xmm0, %xmm5, %k1 vmaxps 0x158bea(%rip){1to4}, %xmm5, %xmm4 # 0x3f11e0 vpsrld $0x17, %xmm4, %xmm5 vpternlogd $0xec, 0x158bde(%rip){1to4}, %xmm2, %xmm4 # 0x3f11e4 vcmpltps 0x158bdb(%rip){1to4}, %xmm4, %k2 # 0x3f11ec vaddps 0x158bd5(%rip){1to4}, %xmm4, %xmm6 # 0x3f11f0 vaddps %xmm4, %xmm6, %xmm6 {%k2} vmulps %xmm6, %xmm6, %xmm4 vmovaps %xmm20, %xmm12 vfmadd213ps 0x158bc3(%rip){1to4}, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + mem vfmadd213ps %xmm23, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm23 vfmadd213ps %xmm18, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm18 vfmadd213ps %xmm19, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm19 vfmadd213ps %xmm24, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm24 vfmadd213ps %xmm29, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm29 vfmadd213ps %xmm31, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + xmm31 vfmadd213ps 0x158bb1(%rip){1to4}, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm12) + mem vmulps %xmm6, %xmm4, %xmm29 vmulps %xmm12, %xmm29, %xmm12 vpaddd 0x158b6f(%rip){1to4}, %xmm5, %xmm5 # 0x3f11e8 vmovaps %xmm7, %xmm24 vmovaps %xmm14, %xmm7 vmovaps %xmm15, %xmm14 vmovaps %xmm16, %xmm15 vmovaps %xmm17, %xmm16 vmovdqa64 %xmm13, %xmm17 vmovaps %xmm1, %xmm13 vmovaps %xmm30, %xmm1 vcvtdq2ps %xmm5, %xmm5 vsubps %xmm9, %xmm5, %xmm5 {%k2} vfmadd231ps %xmm11, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm11) + xmm12 vfmsub231ps %xmm4, %xmm8, %xmm12 # xmm12 = (xmm8 * xmm4) - xmm12 vsubps %xmm6, %xmm12, %xmm4 vfmsub231ps %xmm5, %xmm10, %xmm4 # xmm4 = (xmm10 * xmm5) - xmm4 vmulps 0x15a04d(%rip){1to4}, %xmm4, %xmm4 # 0x3f2718 vbroadcastss 0x15922c(%rip), %xmm5 # 0x3f1900 vmovaps %xmm5, %xmm4 {%k1} vminps %xmm27, %xmm4, %xmm4 vmaxps %xmm28, %xmm4, %xmm4 vmovaps %xmm7, %xmm5 vfmadd213ps %xmm8, %xmm4, %xmm5 # xmm5 = (xmm4 * xmm5) + xmm8 vcvttps2dq %xmm5, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm5, %k1 vsubps %xmm9, %xmm6, %xmm6 {%k1} vfmsub231ps %xmm10, %xmm6, %xmm4 # xmm4 = (xmm6 * xmm10) - xmm4 vfnmsub231ps %xmm11, %xmm6, %xmm4 # xmm4 = -(xmm6 * xmm11) - xmm4 vmulps %xmm4, %xmm4, %xmm5 vmovaps %xmm13, %xmm12 vfmadd213ps %xmm24, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm24 vfmadd213ps %xmm14, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm14 vfmadd213ps %xmm15, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm15 vfmadd213ps %xmm16, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm16 vfmadd213ps %xmm8, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm8 vfmadd213ps %xmm4, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm4 vaddps %xmm9, %xmm12, %xmm5 vcvttps2dq %xmm6, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm17, %xmm4, %xmm4 vfmadd213ps %xmm9, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm9 vrcpps %xmm4, %xmm5 vaddps %xmm5, %xmm5, %xmm6 vfmsub213ps 0x159fa6(%rip){1to4}, %xmm6, %xmm4 # xmm4 = (xmm6 * xmm4) - mem vfnmadd213ps %xmm6, %xmm5, %xmm4 # xmm4 = -(xmm5 * xmm4) + xmm6 vfmsub213ps %xmm22, %xmm22, %xmm3 # xmm3 = (xmm22 * xmm3) - xmm22 vfmsub213ps %xmm21, %xmm21, %xmm4 # xmm4 = (xmm21 * xmm4) - xmm21 jmp 0x298906 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm4 vbroadcastss 0x4(%r14), %xmm5 vmaxps %xmm4, %xmm22, %xmm3 vminps %xmm5, %xmm3, %xmm3 vmaxps %xmm4, %xmm21, %xmm4 vminps %xmm5, %xmm4, %xmm4 jmp 0x298906 vxorps %xmm1, %xmm22, %xmm3 vminps %xmm27, %xmm3, %xmm3 vmaxps %xmm28, %xmm3, %xmm3 vmovaps %xmm7, %xmm4 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm5 vcvtdq2ps %xmm5, %xmm5 vcmpltps %xmm5, %xmm4, %k1 vsubps %xmm9, %xmm5, %xmm5 {%k1} vfmsub231ps %xmm10, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm10) - xmm3 vfnmsub231ps %xmm11, %xmm5, %xmm3 # xmm3 = -(xmm5 * xmm11) - xmm3 vmulps %xmm3, %xmm3, %xmm4 vmovaps %xmm13, %xmm6 vfmadd213ps %xmm24, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm24 vfmadd213ps %xmm14, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm14 vfmadd213ps %xmm15, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm15 vfmadd213ps %xmm16, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm16 vfmadd213ps %xmm8, %xmm3, %xmm6 # xmm6 = (xmm3 * xmm6) + xmm8 vfmadd213ps %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vaddps %xmm6, %xmm9, %xmm4 vcvttps2dq %xmm5, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm17, %xmm3, %xmm3 vfmadd213ps %xmm9, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) + xmm9 vrcpps %xmm3, %xmm4 vfmsub213ps %xmm9, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) - xmm9 vfnmadd132ps %xmm4, %xmm4, %xmm3 # xmm3 = -(xmm3 * xmm4) + xmm4 vxorps %xmm1, %xmm21, %xmm4 vminps %xmm27, %xmm4, %xmm4 vmaxps %xmm28, %xmm4, %xmm4 vmovaps %xmm7, %xmm5 vfmadd213ps %xmm8, %xmm4, %xmm5 # xmm5 = (xmm4 * xmm5) + xmm8 vcvttps2dq %xmm5, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm5, %k1 vsubps %xmm9, %xmm6, %xmm6 {%k1} vfmsub231ps %xmm10, %xmm6, %xmm4 # xmm4 = (xmm6 * xmm10) - xmm4 vfnmsub231ps %xmm11, %xmm6, %xmm4 # xmm4 = -(xmm6 * xmm11) - xmm4 vmulps %xmm4, %xmm4, %xmm5 vmovaps %xmm13, %xmm12 vfmadd213ps %xmm24, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm24 vfmadd213ps %xmm14, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm14 vfmadd213ps %xmm15, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm15 vfmadd213ps %xmm16, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm16 vfmadd213ps %xmm8, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm12) + xmm8 vfmadd213ps %xmm4, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm4 vaddps %xmm9, %xmm12, %xmm5 vcvttps2dq %xmm6, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm17, %xmm4, %xmm4 vfmadd213ps %xmm9, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm9 vrcpps %xmm4, %xmm5 vfmsub213ps %xmm9, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm9 vfnmadd132ps %xmm5, %xmm5, %xmm4 # xmm4 = -(xmm4 * xmm5) + xmm5 jmp 0x298906 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm4 vbroadcastss 0x4(%r14), %xmm5 vmovaps %xmm22, %xmm3 vfmadd213ps %xmm5, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm3) + xmm5 vmaxps %xmm0, %xmm3, %xmm3 vminps %xmm3, %xmm9, %xmm3 vmulps %xmm22, %xmm3, %xmm3 vfmadd213ps %xmm5, %xmm21, %xmm4 # xmm4 = (xmm21 * xmm4) + xmm5 vmaxps %xmm0, %xmm4, %xmm4 vminps %xmm4, %xmm9, %xmm4 vmulps %xmm21, %xmm4, %xmm4 jmp 0x298906 vmovaps %xmm22, %xmm3 vmovaps %xmm21, %xmm4 cmpl $0x65, 0x38(%rsp) jl 0x298964 movq 0x288(%rbp,%rsi), %rsi vmulps (%rsi,%rbx,4), %xmm3, %xmm3 vmulps 0x10(%rsi,%rbx,4), %xmm4, %xmm4 vmovdqa %xmm2, %xmm5 vpternlogd $0xf8, %xmm1, %xmm3, %xmm5 vmovdqa %xmm2, %xmm6 vpternlogd $0xf8, %xmm1, %xmm4, %xmm6 vaddps %xmm5, %xmm3, %xmm3 vaddps %xmm6, %xmm4, %xmm4 vcvttps2dq %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpackssdw %xmm4, %xmm3, %xmm3 vpminsw %xmm25, %xmm3, %xmm3 vpmaxsw %xmm26, %xmm3, %xmm3 vpacksswb %xmm3, %xmm3, %xmm3 vmovq %xmm3, (%rdx) addq $0x8, %rdx jmp 0x298973 vmovups %xmm3, (%r8) vmovups %xmm4, 0x10(%r8) addq $0x20, %r8 incl %r13d jmp 0x2981b8 incl %ecx jmp 0x2981a7 movq 0x48(%rsp), %rsi incq %rsi movq 0x30(%rsp), %rdx addl 0x40(%rsp), %edx movq 0x28(%rsp), %r14 jmp 0x298142 pushq $0x3 popq %r13 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x298ab0 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x298ab0 cmpl $0x2, 0xe4(%rbp,%rdx) jne 0x298ab0 cmpl $0x2, 0xe8(%rbp,%rdx) jne 0x298ab0 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x298ab0 movl $0x0, 0x48(%rsp) cmpl $0x65, 0x38(%rsp) jl 0x299539 vpxor %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %r13 vmovdqa %xmm0, (%r13) andq $0x0, 0x10(%r13) xorl %ebx, %ebx movq %rbp, %rcx leaq 0xe0(%rsp), %rbp leaq 0x190(%rsp), %r12 movq -0x18(%rax), %r14 movslq 0x108(%rcx,%r14), %rax cmpq %rax, %rbx jge 0x299868 movq 0x1f8(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm1 vpxor %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x298a6b movq 0x240(%rcx,%r14), %rax vmulss (%rax,%rbx,4), %xmm1, %xmm0 vmovss 0x156221(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rcx,%r14), %rax vmovd (%rax,%rbx,4), %xmm0 vmovd %xmm0, 0x190(%rsp) movq %r13, %rdi movq %rbp, %rsi callq 0x1ea12c movq %r13, %rdi movq %r12, %rsi callq 0x1ea12c incq %rbx movq 0x10(%rsp), %rcx movq (%rcx), %rax jmp 0x298a26 imull %ecx, %r13d movslq %r13d, %rsi leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %rsi, 0xc0(%rsp) callq 0x73bbe movq (%rbp), %rax movq %rax, 0x58(%rsp) movq -0x18(%rax), %rax imull 0xe0(%rbp,%rax), %r12d movq 0x130(%rsp), %rbx movl 0xdc(%rbp,%rax), %ecx imull 0xd4(%rbp,%rax), %ecx subl %ecx, %r12d xorl %ecx, %ecx xorl %edx, %edx xorl %esi, %esi cmpl 0xd8(%rbp,%rax), %esi jge 0x298b48 movslq %ecx, %rcx leaq (%rbx,%rcx,4), %r8 xorl %edi, %edi cmpl 0xd4(%rbp,%rax), %edi jge 0x298b3e movl %edx, (%r8,%rdi,4) movq 0x58(%rsp), %rax movq -0x18(%rax), %rax addl 0xdc(%rbp,%rax), %edx incq %rdi jmp 0x298b1c addl %r12d, %edx incl %esi addq %rdi, %rcx jmp 0x298b0a xorl %edx, %edx testl %r13d, %r13d movl $0x0, %ecx movq %rcx, 0x50(%rsp) cmovlel %edx, %r13d vbroadcastss 0x158650(%rip), %xmm4 # 0x3f11b4 leaq 0x15fa39(%rip), %r8 # 0x3f85a4 xorl %ebp, %ebp movq 0x10(%rsp), %rcx movslq 0x108(%rcx,%rax), %rax cmpq %rax, %rbp jge 0x298ed9 movq 0x10(%r15), %r9 imulq %rbp, %r9 imulq 0x40(%r15), %r9 addq (%r15), %r9 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rbp, %r11 movq 0x80(%rsp), %rax imulq %rax, %r11 addq 0x70(%rsp), %r11 imulq %rax, %r10 movq 0x28(%rcx), %r15 addq 0x50(%rsp), %r15 xorl %eax, %eax movq %r9, 0x28(%rsp) movq %r10, 0x40(%rsp) movq %r11, 0x68(%rsp) movq %rbp, 0xc8(%rsp) cmpl 0x60(%rsp), %eax jg 0x298eb1 movl %eax, 0x30(%rsp) movq 0x10(%rsp), %rax movq (%rax), %rax movq %rax, 0x48(%rsp) xorl %r14d, %r14d cmpl 0x18(%rsp), %r14d jg 0x298ea6 movq 0x48(%rsp), %rax movq -0x18(%rax), %r12 movq 0x10(%rsp), %rcx movslq 0xe8(%rcx,%r12), %rax movslq 0x30(%rsp), %rdx imulq %rax, %rdx imulq %r10, %rdx movslq 0xe4(%rcx,%r12), %rax movslq %r14d, %rcx imulq %rax, %rcx addq %r11, %rcx addq %rdx, %rcx xorl %edx, %edx xorl %eax, %eax cmpq %rdx, %r13 je 0x298c5f movslq (%rbx,%rdx,4), %rsi movsbl (%rcx,%rsi), %esi movsbl (%r15,%rdx), %edi imull %esi, %edi addl %edi, %eax incq %rdx jmp 0x298c43 movq 0x10(%rsp), %rdx movq 0x1f8(%rdx,%r12), %rcx vmovss (%rcx,%rbp,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x298c94 movq 0x240(%rdx,%r12), %rcx vmulss (%rcx,%rbp,4), %xmm1, %xmm0 vmovss 0x155ff8(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vcvtsi2ss %eax, %xmm6, %xmm1 vmulss %xmm1, %xmm0, %xmm5 cmpl $0x0, 0x100(%rdx,%r12) je 0x298cb4 movq 0x1b0(%rdx,%r12), %rax vaddss (%rax,%rbp,4), %xmm5, %xmm5 movl 0x110(%rdx,%r12), %eax decl %eax cmpl $0x5, %eax ja 0x298e2c movslq (%r8,%rax,4), %rax addq %r8, %rax jmpq *%rax vmaxss 0x155338(%rip), %xmm5, %xmm0 # 0x3ee010 jmp 0x298e30 vmovaps %xmm5, %xmm0 movq %r9, 0xd8(%rsp) vmovss %xmm5, 0xd0(%rsp) callq 0x5f410 vaddss 0x155f89(%rip), %xmm0, %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x68(%rsp), %r11 movq 0x40(%rsp), %r10 movq 0xd8(%rsp), %r9 leaq 0x15f882(%rip), %r8 # 0x3f85a4 vbroadcastss 0x158489(%rip), %xmm4 # 0x3f11b4 vmulss 0xd0(%rsp), %xmm0, %xmm0 jmp 0x298e30 movq 0x10(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss 0x4(%rax), %xmm1 vmaxss (%rax), %xmm5, %xmm0 vucomiss %xmm1, %xmm0 jbe 0x298e30 vmovaps %xmm1, %xmm0 jmp 0x298e30 vmovss 0x15844e(%rip), %xmm2 # 0x3f11b8 vminss %xmm2, %xmm5, %xmm1 vxorps %xmm4, %xmm1, %xmm0 vcmpltss 0x15843f(%rip), %xmm1, %k1 # 0x3f11bc vmovss %xmm2, %xmm0, %xmm0 {%k1} movq %r9, %rbp callq 0x5f410 movq 0x68(%rsp), %r11 movq 0x40(%rsp), %r10 movq %rbp, %r9 movq 0xc8(%rsp), %rbp leaq 0x15f7fd(%rip), %r8 # 0x3f85a4 vbroadcastss 0x158404(%rip), %xmm4 # 0x3f11b4 vmovss 0x155ed0(%rip), %xmm1 # 0x3eec88 vaddss %xmm1, %xmm0, %xmm0 vdivss %xmm0, %xmm1, %xmm0 jmp 0x298e30 movq 0x10(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss (%rax), %xmm0 vcmpgtss 0x155232(%rip), %xmm5, %k1 # 0x3ee010 vmovss 0x155ea0(%rip), %xmm0 {%k1} # 0x3eec88 vmulss %xmm5, %xmm0, %xmm0 jmp 0x298e30 movq 0x10(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss (%rax), %xmm1 vmovss 0x4(%rax), %xmm2 vxorps %xmm4, %xmm2, %xmm0 vdivss %xmm1, %xmm0, %xmm3 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm3, %xmm5 jb 0x298e30 vmovss 0x155e6a(%rip), %xmm0 # 0x3eec88 vdivss %xmm1, %xmm0, %xmm0 vaddss %xmm0, %xmm3, %xmm0 vucomiss %xmm0, %xmm5 jbe 0x298e9b vmovaps %xmm5, %xmm0 cmpl $0x65, 0x38(%rsp) jl 0x298e81 movq 0x10(%rsp), %rax movq 0x288(%rax,%r12), %rax vmulss (%rax,%rbp,4), %xmm0, %xmm0 vpbroadcastd 0x15f7b2(%rip), %xmm1 # 0x3f8604 vpternlogd $0xf8, %xmm4, %xmm0, %xmm1 vaddss %xmm1, %xmm0, %xmm0 vroundss $0xb, %xmm0, %xmm0, %xmm0 vcvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movb %al, (%r9) incq %r9 jmp 0x298e93 movq 0x28(%rsp), %rax vmovss %xmm0, (%rax) addq $0x4, %rax movq %rax, 0x28(%rsp) incl %r14d jmp 0x298bfc vfmadd213ss %xmm2, %xmm5, %xmm1 # xmm1 = (xmm5 * xmm1) + xmm2 vmulss %xmm5, %xmm1, %xmm0 jmp 0x298e30 movl 0x30(%rsp), %eax incl %eax jmp 0x298bde incq %rbp movq 0x58(%rsp), %rax movq -0x18(%rax), %rax movq 0x50(%rsp), %rcx addq 0xc0(%rsp), %rcx movq %rcx, 0x50(%rsp) movq 0x20(%rsp), %r15 jmp 0x298b6d leaq 0x130(%rsp), %rdi callq 0x624be movl $0x0, 0x48(%rsp) jmp 0x298fc6 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0xe0(%rsp), %xmm0 movq 0x20(%rsp), %rcx vmovups %xmm0, (%rcx) movq 0xf0(%rsp), %rax movq %rax, 0x10(%rcx) movl 0xf8(%rsp), %eax movl %eax, 0x18(%rcx) movq 0x100(%rsp), %rax movq %rax, 0x20(%rcx) vmovups 0x108(%rsp), %xmm0 vmovups %xmm0, 0x28(%rcx) movl 0x118(%rsp), %eax movl %eax, 0x38(%rcx) movq 0x120(%rsp), %rax movq %rax, 0x40(%rcx) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x298f8a lock decl (%rax) jne 0x298f8a movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x298f82 movq (%rdi), %rax callq *0x18(%rax) jmp 0x298f8a movq %rsi, %rdi callq 0x5f3e0 movl %r14d, 0x48(%rsp) movq 0x138(%rsp), %rax testq %rax, %rax je 0x298fc6 lock decl (%rax) jne 0x298fc6 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi je 0x298fbe movq (%rdi), %rax callq *0x18(%rax) jmp 0x298fc6 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x298ff7 lock decl (%rax) jne 0x298ff7 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x298fef movq (%rdi), %rax callq *0x18(%rax) jmp 0x298ff7 movq %rsi, %rdi callq 0x5f3e0 movq 0x1e8(%rsp), %rax testq %rax, %rax movl 0x48(%rsp), %ebx je 0x299032 lock decl (%rax) jne 0x299032 movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi je 0x29902a movq (%rdi), %rax callq *0x18(%rax) jmp 0x299032 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0x318, %rsp # imm = 0x318 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq vpxor %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovdqa %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2990bc movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x155bed(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x299066 leaq 0x2a0(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdx movq 0x20(%rsp), %rcx movslq 0x2c(%rcx), %rdi movslq 0x30(%rcx), %rax movl 0x38(%rcx), %esi movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movl 0x34(%rcx), %r8d movl %r8d, 0x68(%rsp) movq (%rcx), %r8 movq 0x10(%rcx), %r11 movq 0x40(%rcx), %r15 imulq %r11, %r15 movq %rax, %rbx imulq %rdi, %rbx movq %r11, 0xd0(%rsp) movq %rbx, 0x60(%rsp) imulq %rbx, %r11 addq $0xf, %r11 movq 0x80(%rsp), %rbx imulq 0xb0(%rsp), %rbx movq %rbx, 0x50(%rsp) andq $-0x10, %r11 movq %r11, 0x58(%rsp) xorl %r11d, %r11d testl %eax, %eax movl $0x0, %ebp cmovgl %eax, %ebp movl 0x28(%rcx), %eax movl %eax, 0xc0(%rsp) testl %esi, %esi cmovlel %r11d, %esi movq %rsi, 0x38(%rsp) vpxor %xmm0, %xmm0, %xmm0 movq 0x70(%rsp), %rax movq %rax, 0x188(%rsp) movq %rdx, 0x28(%rsp) leaq (%rdx,%rdx), %rax movq %rax, 0x180(%rsp) movq 0x2a0(%rsp), %rax movq %rax, 0x178(%rsp) movq %r8, 0xd8(%rsp) xorl %esi, %esi movq %r10, 0x40(%rsp) movq %r15, 0xc8(%rsp) cmpq 0x38(%rsp), %rsi je 0x29933a movq 0x58(%rsp), %rax xorl %edx, %edx divq 0xd0(%rsp) cmpl $0x4, 0xc0(%rsp) cmoveq 0x60(%rsp), %rax testq %r10, %r10 je 0x2991e8 vmovss (%r10,%rsi,4), %xmm2 jmp 0x2991ec vxorps %xmm2, %xmm2, %xmm2 imulq %rsi, %r15 addq 0xd8(%rsp), %r15 movq 0x178(%rsp), %rcx vmovss (%rcx,%rsi,4), %xmm1 imull 0x68(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x299225 vmovss %xmm2, (%r8,%rcx,4) incq %rcx jmp 0x299216 leaq (%rsi,%rsi,8), %rax movq 0x50(%rsp), %r13 movq %rsi, 0x18(%rsp) imulq %rsi, %r13 addq 0x188(%rsp), %r13 movq 0x28(%rsp), %rcx leaq (%rcx,%r13), %r12 movq 0x180(%rsp), %rcx leaq (%rcx,%r13), %r11 xorl %ebx, %ebx cmpl %ebp, %ebx je 0x29931d xorl %ecx, %ecx movq %r15, %rdx movl %edi, %r10d testl %r10d, %r10d jle 0x2992fe movzwl (%r13,%rcx), %r14d vmovd %r14d, %xmm2 vpinsrb $0x2, 0x2(%r13,%rcx), %xmm2, %xmm2 vpinsrb $0x3, (%r12,%rcx), %xmm2, %xmm2 vpinsrb $0x4, 0x1(%r12,%rcx), %xmm2, %xmm2 vpinsrb $0x5, 0x2(%r12,%rcx), %xmm2, %xmm2 vpinsrb $0x6, (%r11,%rcx), %xmm2, %xmm2 vpinsrb $0x7, 0x1(%r11,%rcx), %xmm2, %xmm2 vpmovsxbw (%r9,%rax), %xmm3 vpmovsxbw %xmm2, %xmm2 vpmaddwd %xmm2, %xmm3, %xmm2 vphaddd %xmm2, %xmm0, %xmm2 movsbl 0x2(%r11,%rcx), %r14d movsbl 0x8(%r9,%rax), %esi imull %r14d, %esi vphaddd %xmm2, %xmm2, %xmm2 vphaddd %xmm2, %xmm2, %xmm2 vmovd %xmm2, %r14d addl %esi, %r14d vcvtsi2ss %r14d, %xmm4, %xmm2 vfmadd213ss (%r15,%rcx,4), %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + mem vmovss %xmm2, (%r15,%rcx,4) addq $0x4, %rdx decl %r10d incq %rcx jmp 0x299266 addq %rcx, %r13 addq $0x2, %r13 addq %rcx, %r12 addq $0x2, %r12 addq %rcx, %r11 addq $0x2, %r11 incl %ebx movq %rdx, %r15 jmp 0x299256 movq 0x18(%rsp), %rsi incq %rsi movq 0xc8(%rsp), %r15 addq %r15, %r8 movq 0x40(%rsp), %r10 jmp 0x2991b3 leaq 0x2a0(%rsp), %rdi jmp 0x29984c leaq 0x2b8(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rsi movq 0x20(%rsp), %r15 movl 0x2c(%r15), %eax movl 0x30(%r15), %edx movl 0x38(%r15), %r9d movq 0x10(%rsp), %rcx movq 0x28(%rcx), %rdi movq 0x1b0(%rcx,%r14), %r8 movq %rsi, 0x38(%rsp) leaq (%rsi,%rsi), %rcx movq %rcx, 0x28(%rsp) xorl %r10d, %r10d testl %edx, %edx cmovlel %r10d, %edx testl %r9d, %r9d cmovlel %r10d, %r9d movq %r9, 0x18(%rsp) vpxor %xmm0, %xmm0, %xmm0 vpbroadcastd 0x15f24e(%rip), %xmm1 # 0x3f8604 vpbroadcastd 0x157df5(%rip), %xmm2 # 0x3f11b4 cmpq 0x18(%rsp), %r10 je 0x29952c testq %r8, %r8 je 0x2993d7 vmovss (%r8,%r10,4), %xmm3 jmp 0x2993db vxorps %xmm3, %xmm3, %xmm3 movq 0x40(%r15), %r11 imulq %r10, %r11 imulq 0x10(%r15), %r11 addq (%r15), %r11 movq 0x2b8(%rsp), %rcx vmovss (%rcx,%r10,8), %xmm4 vmovss 0x4(%rcx,%r10,8), %xmm5 movq 0xb0(%rsp), %rbx imulq %r10, %rbx imulq 0x80(%rsp), %rbx addq 0x70(%rsp), %rbx leaq (%r10,%r10,8), %r14 movq 0x38(%rsp), %rcx leaq (%rbx,%rcx), %r15 movq 0x28(%rsp), %rcx leaq (%rbx,%rcx), %r12 xorl %ebp, %ebp cmpl %edx, %ebp je 0x29951f xorl %r13d, %r13d movl %eax, %ecx testl %ecx, %ecx jle 0x299500 movzwl (%rbx,%r13), %r9d vmovd %r9d, %xmm6 vpinsrb $0x2, 0x2(%rbx,%r13), %xmm6, %xmm6 vpinsrb $0x3, (%r15,%r13), %xmm6, %xmm6 vpinsrb $0x4, 0x1(%r15,%r13), %xmm6, %xmm6 vpinsrb $0x5, 0x2(%r15,%r13), %xmm6, %xmm6 vpinsrb $0x6, (%r12,%r13), %xmm6, %xmm6 vpinsrb $0x7, 0x1(%r12,%r13), %xmm6, %xmm6 vpmovsxbw (%rdi,%r14), %xmm7 vpmovsxbw %xmm6, %xmm6 vpmaddwd %xmm6, %xmm7, %xmm6 vphaddd %xmm6, %xmm0, %xmm6 movsbl 0x2(%r12,%r13), %r9d movsbl 0x8(%rdi,%r14), %esi imull %r9d, %esi vphaddd %xmm6, %xmm6, %xmm6 vphaddd %xmm6, %xmm6, %xmm6 vmovd %xmm6, %r9d addl %esi, %r9d vcvtsi2ss %r9d, %xmm8, %xmm6 vfmadd213ss %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vmulss %xmm5, %xmm6, %xmm6 vmovdqa %xmm1, %xmm7 vpternlogd $0xf8, %xmm2, %xmm6, %xmm7 vaddss %xmm7, %xmm6, %xmm6 vroundss $0xb, %xmm6, %xmm6, %xmm6 vcvttss2si %xmm6, %esi cmpl $-0x7e, %esi pushq $-0x7f popq %r9 cmovll %r9d, %esi cmpl $0x7f, %esi pushq $0x7f popq %r9 cmovgel %r9d, %esi movb %sil, (%r11,%r13) decl %ecx incq %r13 jmp 0x29943f addq %r13, %rbx addq $0x2, %rbx addq %r13, %r15 addq $0x2, %r15 addq %r13, %r12 addq $0x2, %r12 incl %ebp addq %r13, %r11 jmp 0x299432 incq %r10 movq 0x20(%rsp), %r15 jmp 0x2993bf leaq 0x2b8(%rsp), %rdi jmp 0x299a63 vpxor %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovdqa %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2995af movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x1556fa(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x299559 leaq 0x270(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rsi movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rdi movslq 0x30(%rdx), %rcx movl 0x38(%rdx), %r8d movl %esi, %eax subl %edi, %eax addl %eax, %eax movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movl 0x34(%rdx), %r11d movl %r11d, 0xd8(%rsp) movq 0x10(%rdx), %r11 movq 0x40(%rdx), %r14 imulq %r11, %r14 movq %rcx, %rbx movq %rdi, 0x18(%rsp) imulq %rdi, %rbx movq %r11, 0xd0(%rsp) movq %rbx, 0x60(%rsp) imulq %rbx, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0x58(%rsp) movq 0x80(%rsp), %rdi imulq 0xb0(%rsp), %rdi movq %rdi, 0x50(%rsp) xorl %edi, %edi testl %ecx, %ecx movl $0x0, %ebp cmovgl %ecx, %ebp movq (%rdx), %r13 testl %r8d, %r8d cmovlel %edi, %r8d movq %r8, 0x28(%rsp) movslq %eax, %r11 vpxor %xmm0, %xmm0, %xmm0 movl 0x28(%rdx), %eax movl %eax, 0x188(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x180(%rsp) movq %rsi, 0x40(%rsp) leaq (%rsi,%rsi), %rax movq %rax, 0x178(%rsp) movq 0x270(%rsp), %rax movq %rax, 0x1d8(%rsp) movq %r13, 0xc0(%rsp) xorl %edi, %edi movq %r10, 0x68(%rsp) movq %r14, 0xc8(%rsp) cmpq 0x28(%rsp), %rdi je 0x299844 movq 0x58(%rsp), %rax xorl %edx, %edx divq 0xd0(%rsp) cmpl $0x4, 0x188(%rsp) cmoveq 0x60(%rsp), %rax testq %r10, %r10 je 0x2996ed vmovss (%r10,%rdi,4), %xmm2 jmp 0x2996f1 vxorps %xmm2, %xmm2, %xmm2 imulq %rdi, %r14 addq 0xc0(%rsp), %r14 movq 0x1d8(%rsp), %rcx vmovss (%rcx,%rdi,4), %xmm1 imull 0xd8(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x29972e vmovss %xmm2, (%r13,%rcx,4) incq %rcx jmp 0x29971e leaq (%rdi,%rdi,8), %rax movq 0x50(%rsp), %rsi movq %rdi, 0x38(%rsp) imulq %rdi, %rsi addq 0x180(%rsp), %rsi movq 0x40(%rsp), %rcx leaq (%rsi,%rcx), %rbx movq 0x178(%rsp), %rcx leaq (%rsi,%rcx), %r8 xorl %r15d, %r15d cmpl %ebp, %r15d je 0x299827 xorl %r10d, %r10d movq %r14, %rdx movq 0x18(%rsp), %rcx testl %ecx, %ecx jle 0x29980a movzwl (%rsi,%r10), %r12d vmovd %r12d, %xmm2 vpinsrb $0x2, 0x2(%rsi,%r10), %xmm2, %xmm2 vpinsrb $0x3, (%rbx,%r10), %xmm2, %xmm2 vpinsrb $0x4, 0x1(%rbx,%r10), %xmm2, %xmm2 vpinsrb $0x5, 0x2(%rbx,%r10), %xmm2, %xmm2 vpinsrb $0x6, (%r8,%r10), %xmm2, %xmm2 vpinsrb $0x7, 0x1(%r8,%r10), %xmm2, %xmm2 vpmovsxbw (%r9,%rax), %xmm3 vpmovsxbw %xmm2, %xmm2 vpmaddwd %xmm2, %xmm3, %xmm2 vphaddd %xmm2, %xmm0, %xmm2 movsbl 0x2(%r8,%r10), %r12d movsbl 0x8(%r9,%rax), %edi imull %r12d, %edi vphaddd %xmm2, %xmm2, %xmm2 vphaddd %xmm2, %xmm2, %xmm2 vmovd %xmm2, %r12d addl %edi, %r12d vcvtsi2ss %r12d, %xmm4, %xmm2 vfmadd213ss (%r14,%r10,2), %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + mem vmovss %xmm2, (%r14,%r10,2) addq $0x4, %rdx decl %ecx addq $0x2, %r10 jmp 0x299774 addq %r11, %rsi addq %r10, %rsi addq %r11, %rbx addq %r10, %rbx addq %r11, %r8 addq %r10, %r8 incl %r15d movq %rdx, %r14 jmp 0x299760 movq 0x38(%rsp), %rdi incq %rdi movq 0xc8(%rsp), %r14 addq %r14, %r13 movq 0x68(%rsp), %r10 jmp 0x2996b8 leaq 0x270(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x20(%rsp), %r15 jmp 0x299a75 leaq 0x288(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %r8 movq 0x20(%rsp), %r15 movl 0x2c(%r15), %eax movl 0x30(%r15), %edx movl 0x38(%r15), %r9d movl %r8d, %ecx subl %eax, %ecx addl %ecx, %ecx movq 0x10(%rsp), %rsi movq 0x28(%rsi), %rdi movq 0x1b0(%rsi,%r14), %rsi movq %r8, 0x38(%rsp) addq %r8, %r8 movq %r8, 0x40(%rsp) movslq %ecx, %r10 xorl %r11d, %r11d testl %edx, %edx cmovlel %r11d, %edx testl %r9d, %r9d cmovlel %r11d, %r9d movq %r9, 0x18(%rsp) vpxor %xmm0, %xmm0, %xmm0 vpbroadcastd 0x15ed24(%rip), %xmm1 # 0x3f8604 vpbroadcastd 0x1578cb(%rip), %xmm2 # 0x3f11b4 movq %rsi, 0x28(%rsp) cmpq 0x18(%rsp), %r11 je 0x299a5b testq %rsi, %rsi je 0x299906 vmovss (%rsi,%r11,4), %xmm3 jmp 0x29990a vxorps %xmm3, %xmm3, %xmm3 movq 0x40(%r15), %rbx imulq %r11, %rbx imulq 0x10(%r15), %rbx addq (%r15), %rbx movq 0x288(%rsp), %rcx vmovss (%rcx,%r11,8), %xmm4 vmovss 0x4(%rcx,%r11,8), %xmm5 movq 0xb0(%rsp), %r14 imulq %r11, %r14 imulq 0x80(%rsp), %r14 addq 0x70(%rsp), %r14 leaq (%r11,%r11,8), %r15 movq 0x38(%rsp), %rcx leaq (%r14,%rcx), %r12 movq 0x40(%rsp), %rcx leaq (%r14,%rcx), %r13 xorl %ebp, %ebp cmpl %edx, %ebp je 0x299a49 xorl %r9d, %r9d movl %eax, %ecx testl %ecx, %ecx jle 0x299a30 movzwl (%r14,%r9), %esi vmovd %esi, %xmm6 vpinsrb $0x2, 0x2(%r14,%r9), %xmm6, %xmm6 vpinsrb $0x3, (%r12,%r9), %xmm6, %xmm6 vpinsrb $0x4, 0x1(%r12,%r9), %xmm6, %xmm6 vpinsrb $0x5, 0x2(%r12,%r9), %xmm6, %xmm6 vpinsrb $0x6, (%r13,%r9), %xmm6, %xmm6 vpinsrb $0x7, 0x1(%r13,%r9), %xmm6, %xmm6 vpmovsxbw (%rdi,%r15), %xmm7 vpmovsxbw %xmm6, %xmm6 vpmaddwd %xmm6, %xmm7, %xmm6 vphaddd %xmm6, %xmm0, %xmm6 movsbl 0x2(%r13,%r9), %esi movsbl 0x8(%rdi,%r15), %r8d imull %esi, %r8d vphaddd %xmm6, %xmm6, %xmm6 vphaddd %xmm6, %xmm6, %xmm6 vmovd %xmm6, %esi addl %r8d, %esi vcvtsi2ss %esi, %xmm8, %xmm6 vfmadd213ss %xmm3, %xmm4, %xmm6 # xmm6 = (xmm4 * xmm6) + xmm3 vmulss %xmm5, %xmm6, %xmm6 vmovdqa %xmm1, %xmm7 vpternlogd $0xf8, %xmm2, %xmm6, %xmm7 vaddss %xmm7, %xmm6, %xmm6 vroundss $0xb, %xmm6, %xmm6, %xmm6 vcvttss2si %xmm6, %esi cmpl $-0x7e, %esi pushq $-0x7f popq %r8 cmovll %r8d, %esi cmpl $0x7f, %esi pushq $0x7f popq %r8 cmovgel %r8d, %esi movb %sil, (%rbx) incq %rbx decl %ecx addq $0x2, %r9 jmp 0x29996e addq %r10, %r14 addq %r9, %r14 addq %r10, %r12 addq %r9, %r12 addq %r10, %r13 addq %r9, %r13 incl %ebp jmp 0x299961 incq %r11 movq 0x20(%rsp), %r15 movq 0x28(%rsp), %rsi jmp 0x2998ee leaq 0x288(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x10(%rsp), %rax movq 0x8(%rax), %rdi testq %rdi, %rdi je 0x298fc6 movq (%rdi), %rax movq %r15, %rsi movq 0x30(%rsp), %rdx callq *0x48(%rax) jmp 0x298fc6 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 jmp 0x299aa8 movq %rax, %rbx leaq 0x130(%rsp), %rdi callq 0x621c2 jmp 0x299c2a jmp 0x299b43 jmp 0x299b43 jmp 0x299ca4 jmp 0x299ca4 movq %rax, %rbx jmp 0x299bf3 jmp 0x299ca4 jmp 0x299b43 movq %rax, %rbx jmp 0x299bbc jmp 0x299ca4 jmp 0x299ca4 jmp 0x299b43 jmp 0x299ca4 jmp 0x299ca4 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x299c5b lock decl (%rax) jne 0x299c5b movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x299c4b movq (%rdi), %rax callq *0x18(%rax) jmp 0x299c5b jmp 0x299ca4 movq %rax, %rbx jmp 0x299c5b movq %rax, %rbx jmp 0x299c2a movq %rax, %rbx movq 0x230(%rsp), %rax testq %rax, %rax je 0x299b85 lock decl (%rax) jne 0x299b85 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi jne 0x299b7f movq %rsi, %rdi callq 0x5f3e0 jmp 0x299b85 movq (%rdi), %rax callq *0x18(%rax) movq 0x198(%rsp), %rax testq %rax, %rax je 0x299bbc lock decl (%rax) jne 0x299bbc movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi jne 0x299bb6 movq %rsi, %rdi callq 0x5f3e0 jmp 0x299bbc movq (%rdi), %rax callq *0x18(%rax) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x299bf3 lock decl (%rax) jne 0x299bf3 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi jne 0x299bed movq %rsi, %rdi callq 0x5f3e0 jmp 0x299bf3 movq (%rdi), %rax callq *0x18(%rax) movq 0x138(%rsp), %rax testq %rax, %rax je 0x299c2a lock decl (%rax) jne 0x299c2a movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi jne 0x299c24 movq %rsi, %rdi callq 0x5f3e0 jmp 0x299c2a movq (%rdi), %rax callq *0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x299c5b lock decl (%rax) jne 0x299c5b movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x299c55 movq %rsi, %rdi callq 0x5f3e0 jmp 0x299c5b movq (%rdi), %rax callq *0x18(%rax) movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x299c92 lock decl (%rax) jne 0x299c92 movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi jne 0x299c8c movq %rsi, %rdi callq 0x5f3e0 jmp 0x299c92 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x299ca4 jmp 0x299ca4 jmp 0x299ca4 jmp 0x299ca4 jmp 0x299ca4 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_avx512::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx512::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float*)bias_data) + g * 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float*)bias_data) + g * 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float*)bias_data) + g * 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2920d6

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp

ncnn::ConvolutionDepthWise_x86_avx512::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx512::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w *= weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w *= bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x168, %rsp # imm = 0x168 movq %rsi, %rbp movq %rdi, %r13 movq (%rsi), %r14 leaq 0x48(%r14), %rdi movq (%rdx), %rax movq %rax, 0xb8(%rsp) movl 0x60(%r14), %ebx movl 0x74(%r14), %eax movl %eax, 0x1c(%rsp) movl 0x78(%r14), %eax movl %eax, 0x18(%rsp) imull 0x80(%r14), %ebx leaq 0x70(%rsp), %rsi andq $0x0, 0x40(%rsi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rsi) vmovups %xmm0, 0xc(%rsi) vmovaps %xmm0, 0x20(%rsi) vmovups %xmm0, 0x2c(%rsi) movq %rcx, %r15 movq %rcx, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x29a2d9 movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x29a2d9 movslq 0x88(%rsp), %rcx movl 0x9c(%rsp), %eax imull %ecx, %eax movl %eax, 0x9c(%rsp) movq 0x80(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x80(%rsp) movl $0x1, 0x88(%rsp) andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) vmovaps %xmm0, 0x40(%rsp) vmovups %xmm0, 0x4c(%rsp) movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r13,%rax) je 0x299e22 movl $0x90, %edi addq (%rbp), %rdi leaq 0x20(%rsp), %rsi movq %r15, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x20(%rsp) je 0x29a2ab movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x29a2ab movslq 0x38(%rsp), %rcx movl 0x4c(%rsp), %eax imull %ecx, %eax movl %eax, 0x4c(%rsp) movq 0x30(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x30(%rsp) movl $0x1, 0x38(%rsp) pushq $0x2a popq %rdi callq 0x782bf movq %rax, %r12 leaq 0x8(%rsp), %rdi callq 0x71548 leaq 0x8(%rsp), %rdi xorl %esi, %esi movl %ebx, %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi movl 0x1c(%rsp), %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0xb popq %rsi movl 0x18(%rsp), %edx callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xdc(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xd popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xec(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x4 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xf popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xe popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x10 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax vmovss 0xfc(%r13,%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x12 popq %rsi callq 0x71952 movq (%r13), %rax movq -0x18(%rax), %rax movl 0x100(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x5 popq %rsi callq 0x7193a movl 0x9c(%rsp), %edx leaq 0x8(%rsp), %rdi pushq $0x6 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x108(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x7 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x10c(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x110(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax leaq (%rax,%r13), %rdx addq $0x118, %rdx # imm = 0x118 leaq 0x8(%rsp), %rdi pushq $0xa popq %rsi callq 0x7196c movq (%r12), %rax leaq 0x8(%rsp), %rsi movq %r12, %rdi callq *0x10(%rax) andq $0x0, 0x110(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0xd0(%rsp) vmovups %xmm0, 0xdc(%rsp) vmovaps %xmm0, 0xf0(%rsp) vmovups %xmm0, 0xfc(%rsp) andq $0x0, 0x158(%rsp) vmovups %xmm0, 0x118(%rsp) vmovups %xmm0, 0x124(%rsp) vmovups %xmm0, 0x138(%rsp) vmovups %xmm0, 0x144(%rsp) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29a090 lock incl (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x29a0c7 lock decl (%rax) jne 0x29a0c7 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x29a0bf movq (%rdi), %rax callq *0x18(%rax) jmp 0x29a0c7 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0xd0(%rsp) movq 0x80(%rsp), %rax movq %rax, 0xe0(%rsp) movl 0x88(%rsp), %eax movl %eax, 0xe8(%rsp) movq 0x90(%rsp), %rax movq %rax, 0xf0(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0xf8(%rsp) movl 0xa8(%rsp), %eax movl %eax, 0x108(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x110(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax je 0x29a141 lock incl (%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x29a178 lock decl (%rax) jne 0x29a178 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x29a170 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29a178 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 leaq 0xd0(%rsp), %rsi vmovups %xmm0, 0x48(%rsi) movq 0x30(%rsp), %rax movq %rax, 0x58(%rsi) movl 0x38(%rsp), %eax movl %eax, 0x60(%rsi) movq 0x40(%rsp), %rax movq %rax, 0x68(%rsi) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsi) movl 0x58(%rsp), %eax movl %eax, 0x80(%rsi) movq 0x60(%rsp), %rax movq %rax, 0x88(%rsi) leaq 0xc0(%rsp), %rdi callq 0x6b00e movq (%r12), %rax leaq 0xc0(%rsp), %rsi movq %r12, %rdi callq *0x18(%rax) leaq 0xc0(%rsp), %rdi callq 0x6b03a movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq *0x20(%rax) movq (%r12), %rax movq %r12, %rdi movq %r14, %rsi movq 0xb8(%rsp), %rdx movq %r15, %rcx callq *0x38(%rax) movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq *0x28(%rax) movq (%r12), %rax movq %r12, %rdi callq *0x8(%rax) pushq $0x48 popq %rbx vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%rbx), %rax testq %rax, %rax je 0x29a273 lock decl (%rax) jne 0x29a273 movq 0xd0(%rsp,%rbx), %rsi movq 0xf0(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x29a267 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x29a273 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%rbx), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x29a234 leaq 0x8(%rsp), %rdi callq 0x71614 xorl %r12d, %r12d movq 0x28(%rsp), %rax testq %rax, %rax je 0x29a2d9 lock decl (%rax) jne 0x29a2d9 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x29a2d1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29a2d9 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x29a30a lock decl (%rax) jne 0x29a30a movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29a302 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29a30a movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x168, %rsp # imm = 0x168 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x29a43f jmp 0x29a43f movq %rax, %rbx leaq 0xc0(%rsp), %rdi callq 0x6b03a jmp 0x29a34c jmp 0x29a349 jmp 0x29a341 jmp 0x29a341 movq %rax, %rbx jmp 0x29a3d6 movq %rax, %rbx pushq $0x48 popq %r14 vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%r14), %rax testq %rax, %rax je 0x29a393 lock decl (%rax) jne 0x29a393 movq 0xd0(%rsp,%r14), %rsi movq 0xf0(%rsp,%r14), %rdi testq %rdi, %rdi je 0x29a387 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x29a393 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x29a354 jmp 0x29a3cc jmp 0x29a43f movq %rax, %rbx jmp 0x29a404 jmp 0x29a43f movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq 0x28(%rsp), %rax testq %rax, %rax je 0x29a404 lock decl (%rax) jne 0x29a404 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x29a3fe movq %rsi, %rdi callq 0x5f3e0 jmp 0x29a404 movq (%rdi), %rax callq *0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29a435 lock decl (%rax) jne 0x29a435 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x29a42f movq %rsi, %rdi callq 0x5f3e0 jmp 0x29a435 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x29a43f movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_avx512::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx512::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w *= weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w *= bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }

movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x299cb8

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx512.cpp

ncnn::ConvolutionDepthWise_x86_fma::create_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_fma::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x50, %rsp movq (%rdi), %rax movq -0x18(%rax), %r13 cmpl $0x0, 0x160(%rdi,%r13) je 0x29a51d xorl %eax, %eax addq $0x50, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 retq movq %rsi, %r14 movq %rdi, %rbx movl 0x110(%rdi,%r13), %ecx decl %ecx cmpl $0x5, %ecx ja 0x29a874 leaq 0x15e0e3(%rip), %rax # 0x3f8620 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx jmpq *%rcx pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x29a6b0 pushq $0x47 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x29a6b0 pushq $0x36 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax vmovss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) jmp 0x29a6b0 pushq $0x1e popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x29a6b0 pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) jmp 0x29a6b0 pushq $0x43 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax vmovss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) leaq 0x8(%rsp), %rdi callq 0x71614 movq (%r15), %rax movq %r15, %rdi movq %r14, %rsi callq *0x20(%rax) movq (%rbx), %rax movq %r15, 0x8(%rbx) movq -0x18(%rax), %r15 cmpb $0x1, 0x1e(%r14) jne 0x29a6f3 cmpq $0x1, 0x178(%rbx,%r15) jne 0x29a6f3 movq %rbx, %rdi movq %r14, %rsi callq 0x29aa72 jmp 0x29a50d movl 0xd0(%rbx,%r15), %esi movl 0xd4(%rbx,%r15), %r10d movl 0xd8(%rbx,%r15), %r11d movl %r11d, %r8d imull %r10d, %r8d movl 0x104(%rbx,%r15), %eax movl 0x108(%rbx,%r15), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x29a812 imull %ecx, %eax cmpl %ecx, %eax jne 0x29a812 leaq (%rbx,%r15), %rsi cmpb $0x1, 0x27(%r14) jne 0x29a769 testb $0x7, %cl je 0x29a87c testb $0x3, %cl je 0x29a8dd xorl $0x3, %r10d xorl $0x3, %r11d orl %r10d, %r11d jne 0x29a812 cmpl $0x1, 0xdc(%rsi) jne 0x29a79e cmpl $0x1, 0xe0(%rsi) jne 0x29a79e cmpl $0x1, 0xe4(%rsi) jne 0x29a79e cmpl $0x1, 0xe8(%rsi) je 0x29a7c2 cmpl $0x1, 0xdc(%rsi) jne 0x29a812 cmpl $0x1, 0xe0(%rsi) jne 0x29a812 cmpl $0x2, 0xe4(%rsi) jne 0x29a812 cmpl $0x2, 0xe8(%rsi) jne 0x29a812 addq $0x168, %r15 # imm = 0x168 cmpq $0x28, %r15 je 0x29a81d movq 0x8(%rbx,%r15), %rax testq %rax, %rax je 0x29a7dc lock incl (%rax) leaq 0x28(%rbx), %r12 movq 0x30(%rbx), %rax testq %rax, %rax je 0x29a982 lock decl (%rax) jne 0x29a982 movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x29a97a movq (%rdi), %rax callq *0x18(%rax) jmp 0x29a982 movq %rbx, %rdi movq %r14, %rsi callq 0x29acde cmpb $0x1, (%r14) jne 0x29a50d movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x29a942 lock decl (%rax) jne 0x29a942 movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x29a93a movq (%rdi), %rax callq *0x18(%rax) jmp 0x29a942 xorl %r15d, %r15d jmp 0x29a6c9 addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x8 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x29a81d lock decl (%rax) jne 0x29a81d movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x29a96d movq (%rdi), %rax callq *0x18(%rax) jmp 0x29a81d addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x4 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x29a81d lock decl (%rax) jne 0x29a81d movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x29a96d movq (%rdi), %rax callq *0x18(%rax) jmp 0x29a81d movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rbx) vmovups %xmm0, (%rbx) vmovups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x29a50d movq %rsi, %rdi callq 0x5f3e0 jmp 0x29a81d movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%r12) vmovups %xmm0, (%r12) andl $0x0, 0x60(%rbx) vmovups %xmm0, 0x50(%rbx) vmovups (%rbx,%r15), %xmm0 vmovups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r15), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r15), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r15), %rax movq %rax, 0x48(%rbx) vmovups 0x28(%rbx,%r15), %xmm0 vmovups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r15), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r15), %rax movq %rax, 0x68(%rbx) jmp 0x29a81d jmp 0x29aa4a jmp 0x29aa4a movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x29aa69 lock decl (%rax) jne 0x29aa69 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x29aa38 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29aa69 jmp 0x29aa4a movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x29aa69 lock decl (%rax) jne 0x29aa69 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi jne 0x29aa42 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29aa69 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29aa69 movq %rax, %rdi callq 0x61d68 jmp 0x29aa5c jmp 0x29aa5c jmp 0x29aa5c jmp 0x29aa5c jmp 0x29aa5c movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq %rbx, %rdi callq 0x5f340 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

ncnn::ConvolutionDepthWise_x86_fma::create_group_ops(ncnn::Option const&)

int ConvolutionDepthWise_x86_fma::create_group_ops(const Option& opt) { // create Convolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Convolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[5]; weights[0] = weight_data_g; weights[1] = bias_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[2] = weight_data_int8_scales_g; weights[3] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[4] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[4]; weights[0] = weight_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[1] = weight_data_int8_scales_g; weights[2] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[3] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x278, %rsp # imm = 0x278 movq %rsi, 0x250(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x104(%rdi,%rdx), %eax movl 0x108(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x10(%rdi), %rax movq %rax, 0x1e0(%rsp) xorl %r12d, %r12d movq 0x10(%r14), %rax movq 0x18(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x29ad7b movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x29ad76 movq (%rdi), %rax callq *0x8(%rax) incq %r12 jmp 0x29ad4d imull %r15d, %ebx cmpq %rax, %rcx je 0x29ad88 movq %rax, 0x18(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x108(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0x1e0(%rsp), %rdi callq 0x6fbc2 leaq 0x118(%r14), %rax movq %rax, 0x258(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0x6c(%rsp) movslq %ebx, %rax movq %rax, 0x260(%rsp) movslq %r15d, %rax movq %rax, 0x248(%rsp) pushq $0x1 popq %rbx xorl %edx, %edx movl %r15d, 0x14(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rcx cmpq %rcx, %rdx jge 0x29baf2 movq %rdx, %rcx movq %rdx, 0x8(%rsp) movq 0x260(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdx, 0x80(%rsp) movl %esi, 0x88(%rsp) movq %rax, 0x90(%rsp) movl %ebx, 0x98(%rsp) movl %edi, 0x9c(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa0(%rsp) movl %ebx, 0xa8(%rsp) movq %rdi, 0xb0(%rsp) leaq 0x200(%rsp), %rdi leaq 0x70(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x29aed9 lock decl (%rax) jne 0x29aed9 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29aed1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29aed9 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x29af49 movq 0x8(%rsp), %r15 movq 0x248(%rsp), %rcx imulq %rcx, %r15 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0x1f8(%rsp) imulq %rsi, %r15 addq 0x1b0(%r14,%rax), %r15 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x1c(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0x1f0(%rsp) movl %ebx, %r13d movl 0x14(%rsp), %eax movl %eax, 0x18(%rsp) movq %rcx, 0x1e8(%rsp) jmp 0x29af83 xorl %r15d, %r15d movq $0x0, 0x1f8(%rsp) movl $0x0, 0x1c(%rsp) movq $0x0, 0x1f0(%rsp) xorl %r13d, %r13d movl $0x0, 0x18(%rsp) movq $0x0, 0x1e8(%rsp) pushq $0x6 popq %rdi callq 0x782bf movq %rax, %rbp leaq 0x268(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq %r12, %rdi xorl %esi, %esi movl 0x14(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r12, %rdi movl %ebx, %esi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r12, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r12, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r12, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r12, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r12, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r12, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r12, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x6 popq %rsi movl 0x6c(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r12, %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x110(%r14,%rax), %edx movq %r12, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x258(%rsp), %rdx movq %r12, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r12, %rsi callq *0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x29b170 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x1a8, %rax # imm = 0x1A8 jne 0x29b0fe movq 0x208(%rsp), %rax testq %rax, %rax je 0x29b138 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29b1f1 lock decl (%rax) jne 0x29b1f1 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29b1e9 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b1f1 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x160, %rax # imm = 0x160 jne 0x29b177 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29b1b1 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x29b5a8 lock decl (%rax) jne 0x29b5a8 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29b5a0 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b5a8 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x29b295 lock decl (%rax) jne 0x29b295 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x29b28d movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b295 movq %rsi, %rdi callq 0x5f3e0 movq %r15, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq 0x1f8(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x1c(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x1f0(%rsp), %rax movq %rax, 0xd8(%rsp) movl %r13d, 0xe0(%rsp) movl 0x18(%rsp), %eax movl %eax, 0xe4(%rsp) movl %r13d, 0xe8(%rsp) movl %r13d, 0xec(%rsp) movl %r13d, 0xf0(%rsp) movq 0x1e8(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x29b52a andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x29b39a vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x29b38c movq 0x28(%rsp), %rax testq %rax, %rax je 0x29b3a7 lock incl (%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x29b3de lock decl (%rax) jne 0x29b3de movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x29b3d6 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b3de movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovaps %xmm0, 0x100(%rsp) movq 0x30(%rsp), %rax movq %rax, 0x110(%rsp) movl 0x38(%rsp), %eax movl %eax, 0x118(%rsp) movq 0x40(%rsp), %rax movq %rax, 0x120(%rsp) vmovups 0x48(%rsp), %xmm0 vmovups %xmm0, 0x128(%rsp) movl 0x58(%rsp), %eax movl %eax, 0x138(%rsp) movq 0x60(%rsp), %rax movq %rax, 0x140(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %r12d movq 0x260(%r14,%rax), %rbx movq 0x150(%rsp), %rax testq %rax, %rax je 0x29b4a0 lock decl (%rax) jne 0x29b4a0 movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x29b498 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b4a0 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %r12d, 0x160(%rsp) movq %rbx, 0x168(%rsp) vbroadcastss 0x15d12e(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x29b523 lock decl (%rax) jne 0x29b523 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x29b51b movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b523 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x29b74a movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x198(%rsp), %rax testq %rax, %rax je 0x29b6f6 lock decl (%rax) jne 0x29b6f6 movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x29b6ee movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b6f6 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x29b92d andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x29b6a3 vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x29b695 movq 0x28(%rsp), %rax testq %rax, %rax je 0x29b6b0 lock incl (%rax) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x29b7e2 lock decl (%rax) jne 0x29b7e2 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x29b7da movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b7e2 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x190(%rsp) andq $0x0, 0x198(%rsp) movq %r15, 0x1a0(%rsp) movl %ebx, 0x1a8(%rsp) movq %r12, 0x1b0(%rsp) vbroadcastss 0x15ced9(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x1b8(%rsp) movl $0x1, 0x1c8(%rsp) movq $0x1, 0x1d0(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0x120, %r15d # imm = 0x120 movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x29b7a8 lock decl (%rax) jne 0x29b7a8 movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x29b7a0 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b7a8 movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x29b777 jmp 0x29ba7b movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovups %xmm0, 0xb8(%rsp) movq 0x30(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x38(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x40(%rsp), %rax movq %rax, 0xd8(%rsp) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movl 0x58(%rsp), %eax movl %eax, 0xf0(%rsp) movq 0x60(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %ebx movq 0x260(%r14,%rax), %r12 movq 0x108(%rsp), %rax testq %rax, %rax je 0x29b8a4 lock decl (%rax) jne 0x29b8a4 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x29b89c movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b8a4 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x100(%rsp) andq $0x0, 0x108(%rsp) movq %r15, 0x110(%rsp) movl %ebx, 0x118(%rsp) movq %r12, 0x120(%rsp) vbroadcastss 0x15cd2b(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x128(%rsp) movl $0x1, 0x138(%rsp) movq $0x1, 0x140(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x29b926 lock decl (%rax) jne 0x29b926 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x29b91e movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b926 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x29b9f0 movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x150(%rsp), %rax testq %rax, %rax je 0x29b99c lock decl (%rax) jne 0x29b99c movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x29b994 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29b99c movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %ebx, 0x160(%rsp) movq %r12, 0x168(%rsp) vbroadcastss 0x15cc33(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0xd8, %r15d movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x29ba4e lock decl (%rax) jne 0x29ba4e movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x29ba46 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29ba4e movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x29ba1d movq (%rbp), %rax movq %rbp, %rdi movq 0x250(%rsp), %rsi callq *0x20(%rax) leaq 0x268(%rsp), %rdi movq 0x1e0(%rsp), %rax movq (%rax), %rax movq 0x8(%rsp), %rcx movq %rbp, (%rax,%rcx,8) callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29bae5 lock decl (%rax) jne 0x29bae5 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi je 0x29badd movq (%rdi), %rax callq *0x18(%rax) jmp 0x29bae5 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rdx incq %rdx jmp 0x29adf3 xorl %eax, %eax addq $0x278, %rsp # imm = 0x278 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x29bbba jmp 0x29bd51 jmp 0x29bb12 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x29bbcc lock decl (%rax) jne 0x29bbcc movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x29bb48 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29bbcc movq (%rdi), %rax callq *0x18(%rax) jmp 0x29bbcc jmp 0x29bd51 jmp 0x29bc44 jmp 0x29bbba jmp 0x29bb5e movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x29bc56 lock decl (%rax) jne 0x29bc56 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x29bb94 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29bc56 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29bc56 jmp 0x29bd51 jmp 0x29bd51 jmp 0x29bd51 jmp 0x29bd51 jmp 0x29bbba jmp 0x29bc44 movq %rax, %rbx jmp 0x29bbcc movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0xd8, %r14d vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x29bc0f lock decl (%rax) jne 0x29bc0f movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x29bc03 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x29bc0f movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x29bbd6 jmp 0x29bd05 jmp 0x29bd51 jmp 0x29bc44 movq %rax, %rbx jmp 0x29bc56 movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0x120, %r14d # imm = 0x120 vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x29bc99 lock decl (%rax) jne 0x29bc99 movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x29bc8d movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x29bc99 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x29bc60 jmp 0x29bd05 jmp 0x29bd51 jmp 0x29bcf9 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x29bd49 lock decl (%rax) jne 0x29bd49 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x29bd39 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29bd49 jmp 0x29bd51 movq %rax, %rbx jmp 0x29bd12 jmp 0x29bd51 jmp 0x29bd51 movq %rax, %rbx leaq 0x268(%rsp), %rdi callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x29bd49 lock decl (%rax) jne 0x29bd49 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi jne 0x29bd43 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29bd49 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_fma::create_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_fma::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %rax movq (%rdi), %rax addq -0x30(%rax), %rdi callq 0x29a4ee xorl %eax, %eax popq %rcx retq nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

ncnn::ConvolutionDepthWise_x86_fma::destroy_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_fma::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }

pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x29bd9d movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x29bd98 movq (%rdi), %rax callq *0x8(%rax) andq $0x0, 0x8(%rbx) xorl %r15d, %r15d movq 0x10(%rbx), %rax movq 0x18(%rbx), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r15 jge 0x29bddf movq (%rax,%r15,8), %rdi movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x10(%rbx), %rax movq (%rax,%r15,8), %rdi testq %rdi, %rdi je 0x29bdda movq (%rdi), %rax callq *0x8(%rax) incq %r15 jmp 0x29bda0 cmpq %rax, %rcx je 0x29bde8 movq %rax, 0x18(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %r15 retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_fma::destroy_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_fma::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }

pushq %rax movq (%rdi), %rax addq -0x38(%rax), %rdi callq 0x29bd6c xorl %eax, %eax popq %rcx retq nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

ncnn::ConvolutionDepthWise_x86_fma::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_fma::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float*)bias_data) + g * 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float*)bias_data) + g * 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float*)bias_data) + g * 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x2e8, %rsp # imm = 0x2E8 movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdi movq %rcx, 0x58(%rsp) cmpb $0x1, 0x1e(%rcx) jne 0x29be51 cmpl $0x0, 0x10c(%r14,%rdi) je 0x29be51 movq %r14, %rdi movq 0x58(%rsp), %rcx addq $0x2e8, %rsp # imm = 0x2E8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x29ff92 movq %rdx, 0x28(%rsp) movl 0x38(%rsi), %ecx movq %rcx, 0x60(%rsp) movq 0x10(%rsi), %r13 movslq 0x18(%rsi), %r15 movl 0xd4(%r14,%rdi), %ebx decl %ebx imull 0xdc(%r14,%rdi), %ebx movl 0xd8(%r14,%rdi), %ebp decl %ebp imull 0xe0(%r14,%rdi), %ebp leaq 0x80(%rsp), %rdx andq $0x0, 0x40(%rdx) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rdx) vmovups %xmm0, 0xc(%rdx) vmovaps %xmm0, 0x20(%rdx) vmovups %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %r14, %rdi movq 0x58(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0x80(%rsp) je 0x29fd9f movslq 0xb8(%rsp), %rax imulq 0xc0(%rsp), %rax testq %rax, %rax je 0x29fd9f notl %ebx notl %ebp movl 0xac(%rsp), %edi addl %edi, %ebx movq (%r14), %rax movq -0x18(%rax), %rcx movl %ebx, %eax cltd idivl 0xe4(%r14,%rcx) movl %eax, %ebx leal 0x1(%rbx), %esi addl 0xb0(%rsp), %ebp movl %ebp, %eax cltd idivl 0xe8(%r14,%rcx) movq %rax, 0x30(%rsp) leal 0x1(%rax), %r8d movl 0xd0(%r14,%rcx), %ecx movq 0x58(%rsp), %rax cmpb $0x1, 0x27(%rax) pushq $0x8 popq %rbp movl %edi, 0x10(%rsp) jne 0x29bf5d xorl %eax, %eax testb $0x3, %cl sete %al testb $0x7, %cl leal 0x1(%rax,%rax,2), %r9d cmovel %ebp, %r9d jmp 0x29bf61 pushq $0x1 popq %r9 movq %r13, %rax xorl %edx, %edx divq %r15 movq %rax, %r10 movl %r9d, %eax movq %rax, 0x40(%rsp) imulq %rax, %r10 movl %ecx, %eax cltd idivl %r9d movq 0x58(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq 0x28(%rsp), %r13 movq %r13, %rdi movl %esi, 0x70(%rsp) movl %r8d, 0xd0(%rsp) movl %r8d, %edx movl %eax, %ecx movq %r10, 0xe0(%rsp) movq %r10, %r8 callq 0x628f2 movq (%r13), %r9 testq %r9, %r9 je 0x29fd9f movq 0x40(%r13), %r11 movslq 0x38(%r13), %rax imulq %r11, %rax testq %rax, %rax je 0x29fd9f movq %r15, %r10 movl %r10d, %eax imull 0x60(%rsp), %eax movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xd0(%r14,%rdi), %ecx movl 0x108(%r14,%rdi), %esi cmpl %esi, %eax jne 0x29e16e cmpl %ecx, %eax jne 0x29e16e addq %r14, %rdi cmpl $0x1, %r10d je 0x29d338 movslq %ebx, %rbx cmpl $0x4, %r10d movq %r14, 0x8(%rsp) je 0x29ca8e movl %eax, %ecx cmpl $0x8, %r10d jne 0x29e16e movl 0xd4(%rdi), %eax movl 0xd8(%rdi), %r15d cmpl $0x5, %eax je 0x29dca2 cmpl $0x3, %eax jne 0x29fa97 cmpl $0x3, %r15d jne 0x29fa97 cmpl $0x1, 0xdc(%rdi) jne 0x29f37c cmpl $0x1, 0xe0(%rdi) jne 0x29f37c cmpl $0x1, 0xe4(%rdi) jne 0x29f37c cmpl $0x1, 0xe8(%rdi) jne 0x29f37c movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xac(%rsp), %edx movl 0xb8(%rsp), %esi movq 0x1b0(%rdi), %r8 leal 0x10(,%rdx,8), %edx movslq %edx, %rdi leal (,%rax,8), %edx movslq %edx, %rdx xorl %r9d, %r9d testl %esi, %esi cmovlel %r9d, %esi movq %rsi, 0x40(%rsp) shlq $0x2, %rdx movq %rdx, 0x10(%rsp) shlq $0x2, %rdi movq %r8, 0x60(%rsp) cmpq 0x40(%rsp), %r9 je 0x29fdf4 testq %r8, %r8 je 0x29c100 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r8,%rdx), %ymm0 jmp 0x29c104 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r10 imulq %r9, %r10 movq 0x10(%r13), %rsi imulq %rsi, %r10 addq (%r13), %r10 movslq 0x2c(%r13), %rdx movslq 0x54(%r14), %r11 imulq %r9, %r11 imulq 0x38(%r14), %r11 movq 0x28(%r14), %rbx imulq %rsi, %rdx addq %r10, %rdx movslq 0xac(%rsp), %rsi movq 0xc0(%rsp), %r15 movq %r9, 0x30(%rsp) imulq %r9, %r15 movq 0x90(%rsp), %r8 imulq %r8, %r15 addq 0x80(%rsp), %r15 imulq %r8, %rsi leaq (%r15,%rsi), %r12 leaq (%r15,%rsi,2), %r13 leaq (%rsi,%rsi,2), %rsi addq %r15, %rsi xorl %ebp, %ebp movl %ebp, %r8d orl $0x1, %r8d cmpl %ecx, %r8d jge 0x29ca6f xorl %r9d, %r9d xorl %r8d, %r8d leal 0x3(%r8), %r14d cmpl %eax, %r14d jge 0x29c5c0 vmovaps 0x20(%r15,%r9), %ymm3 vmovaps 0x40(%r15,%r9), %ymm4 vmovaps 0x60(%r15,%r9), %ymm5 vmovaps 0x80(%r15,%r9), %ymm6 vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm13 vmovaps 0x40(%rbx,%r11), %ymm12 vmovaps 0x60(%rbx,%r11), %ymm9 vmovaps (%r15,%r9), %ymm2 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vfmadd231ps %ymm3, %ymm13, %ymm2 # ymm2 = (ymm13 * ymm3) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm3) + ymm0 vfmadd231ps %ymm4, %ymm13, %ymm3 # ymm3 = (ymm13 * ymm4) + ymm3 vfmadd231ps %ymm4, %ymm12, %ymm2 # ymm2 = (ymm12 * ymm4) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm4 # ymm4 = (ymm1 * ymm4) + ymm0 vfmadd231ps %ymm5, %ymm13, %ymm4 # ymm4 = (ymm13 * ymm5) + ymm4 vfmadd231ps %ymm5, %ymm12, %ymm3 # ymm3 = (ymm12 * ymm5) + ymm3 vfmadd213ps %ymm0, %ymm1, %ymm5 # ymm5 = (ymm1 * ymm5) + ymm0 vfmadd231ps %ymm6, %ymm13, %ymm5 # ymm5 = (ymm13 * ymm6) + ymm5 vfmadd231ps %ymm6, %ymm12, %ymm4 # ymm4 = (ymm12 * ymm6) + ymm4 vfmadd231ps 0xa0(%r15,%r9), %ymm12, %ymm5 # ymm5 = (ymm12 * mem) + ymm5 vmovaps 0x80(%rbx,%r11), %ymm11 vmovaps 0xa0(%rbx,%r11), %ymm8 vmovaps (%r12,%r9), %ymm6 vmovaps 0x20(%r12,%r9), %ymm7 vmovaps 0x40(%r12,%r9), %ymm10 vmovaps 0x60(%r12,%r9), %ymm15 vmovaps 0x80(%r12,%r9), %ymm14 vfmadd231ps %ymm6, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm6) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm6 # ymm6 = (ymm1 * ymm6) + ymm0 vfmadd231ps %ymm7, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm7) + ymm3 vfmadd231ps %ymm7, %ymm13, %ymm6 # ymm6 = (ymm13 * ymm7) + ymm6 vfmadd231ps %ymm7, %ymm11, %ymm2 # ymm2 = (ymm11 * ymm7) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm7 # ymm7 = (ymm1 * ymm7) + ymm0 vfmadd231ps %ymm10, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm10) + ymm4 vfmadd231ps %ymm10, %ymm13, %ymm7 # ymm7 = (ymm13 * ymm10) + ymm7 vfmadd231ps %ymm10, %ymm11, %ymm3 # ymm3 = (ymm11 * ymm10) + ymm3 vfmadd231ps %ymm10, %ymm12, %ymm6 # ymm6 = (ymm12 * ymm10) + ymm6 vfmadd231ps %ymm10, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm10) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm10 # ymm10 = (ymm1 * ymm10) + ymm0 vfmadd213ps %ymm0, %ymm15, %ymm1 # ymm1 = (ymm15 * ymm1) + ymm0 vfmadd231ps %ymm15, %ymm13, %ymm10 # ymm10 = (ymm13 * ymm15) + ymm10 vfmadd231ps %ymm13, %ymm14, %ymm1 # ymm1 = (ymm14 * ymm13) + ymm1 vmovaps 0xa0(%r12,%r9), %ymm13 vfmadd231ps %ymm15, %ymm9, %ymm5 # ymm5 = (ymm9 * ymm15) + ymm5 vfmadd231ps %ymm15, %ymm11, %ymm4 # ymm4 = (ymm11 * ymm15) + ymm4 vfmadd231ps %ymm14, %ymm11, %ymm5 # ymm5 = (ymm11 * ymm14) + ymm5 vfmadd231ps %ymm15, %ymm12, %ymm7 # ymm7 = (ymm12 * ymm15) + ymm7 vfmadd231ps %ymm14, %ymm12, %ymm10 # ymm10 = (ymm12 * ymm14) + ymm10 vfmadd231ps %ymm12, %ymm13, %ymm1 # ymm1 = (ymm13 * ymm12) + ymm1 vfmadd231ps %ymm15, %ymm8, %ymm3 # ymm3 = (ymm8 * ymm15) + ymm3 vfmadd231ps %ymm14, %ymm8, %ymm4 # ymm4 = (ymm8 * ymm14) + ymm4 vfmadd231ps %ymm13, %ymm8, %ymm5 # ymm5 = (ymm8 * ymm13) + ymm5 vmovaps (%r13,%r9), %ymm14 vmovaps 0x20(%r13,%r9), %ymm15 vmovaps 0x40(%r13,%r9), %ymm13 vmovaps 0x60(%r13,%r9), %ymm12 vfmadd231ps %ymm14, %ymm9, %ymm6 # ymm6 = (ymm9 * ymm14) + ymm6 vfmadd231ps %ymm15, %ymm9, %ymm7 # ymm7 = (ymm9 * ymm15) + ymm7 vfmadd231ps %ymm13, %ymm9, %ymm10 # ymm10 = (ymm9 * ymm13) + ymm10 vfmadd231ps %ymm9, %ymm12, %ymm1 # ymm1 = (ymm12 * ymm9) + ymm1 vmovaps 0xc0(%rbx,%r11), %ymm9 vfmadd231ps %ymm14, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm14) + ymm2 vmovaps 0x80(%r13,%r9), %ymm14 vfmadd231ps %ymm15, %ymm11, %ymm6 # ymm6 = (ymm11 * ymm15) + ymm6 vfmadd231ps %ymm13, %ymm11, %ymm7 # ymm7 = (ymm11 * ymm13) + ymm7 vfmadd231ps %ymm12, %ymm11, %ymm10 # ymm10 = (ymm11 * ymm12) + ymm10 vfmadd231ps %ymm11, %ymm14, %ymm1 # ymm1 = (ymm14 * ymm11) + ymm1 vmovaps 0xe0(%rbx,%r11), %ymm11 vfmadd231ps %ymm15, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm15) + ymm3 vfmadd231ps %ymm15, %ymm11, %ymm2 # ymm2 = (ymm11 * ymm15) + ymm2 vmovaps 0xa0(%r13,%r9), %ymm15 vfmadd231ps %ymm13, %ymm8, %ymm6 # ymm6 = (ymm8 * ymm13) + ymm6 vfmadd231ps %ymm12, %ymm8, %ymm7 # ymm7 = (ymm8 * ymm12) + ymm7 vfmadd231ps %ymm14, %ymm8, %ymm10 # ymm10 = (ymm8 * ymm14) + ymm10 vfmadd231ps %ymm8, %ymm15, %ymm1 # ymm1 = (ymm15 * ymm8) + ymm1 vmovaps 0x100(%rbx,%r11), %ymm8 vfmadd231ps %ymm13, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm13) + ymm4 vfmadd231ps %ymm13, %ymm11, %ymm3 # ymm3 = (ymm11 * ymm13) + ymm3 vfmadd231ps %ymm13, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm13) + ymm2 vfmadd231ps %ymm12, %ymm9, %ymm5 # ymm5 = (ymm9 * ymm12) + ymm5 vfmadd231ps %ymm12, %ymm11, %ymm4 # ymm4 = (ymm11 * ymm12) + ymm4 vfmadd231ps %ymm14, %ymm11, %ymm5 # ymm5 = (ymm11 * ymm14) + ymm5 vfmadd231ps %ymm12, %ymm8, %ymm3 # ymm3 = (ymm8 * ymm12) + ymm3 vfmadd231ps %ymm14, %ymm8, %ymm4 # ymm4 = (ymm8 * ymm14) + ymm4 vfmadd231ps %ymm15, %ymm8, %ymm5 # ymm5 = (ymm8 * ymm15) + ymm5 vmovaps 0x20(%rsi,%r9), %ymm12 vmovaps 0x40(%rsi,%r9), %ymm13 vfmadd231ps (%rsi,%r9), %ymm9, %ymm6 # ymm6 = (ymm9 * mem) + ymm6 vmovaps 0x60(%rsi,%r9), %ymm14 vfmadd231ps %ymm12, %ymm9, %ymm7 # ymm7 = (ymm9 * ymm12) + ymm7 vfmadd231ps %ymm13, %ymm9, %ymm10 # ymm10 = (ymm9 * ymm13) + ymm10 vfmadd231ps %ymm9, %ymm14, %ymm1 # ymm1 = (ymm14 * ymm9) + ymm1 vfmadd231ps %ymm12, %ymm11, %ymm6 # ymm6 = (ymm11 * ymm12) + ymm6 vmovaps 0x80(%rsi,%r9), %ymm9 vfmadd231ps %ymm13, %ymm11, %ymm7 # ymm7 = (ymm11 * ymm13) + ymm7 vfmadd231ps %ymm14, %ymm11, %ymm10 # ymm10 = (ymm11 * ymm14) + ymm10 vfmadd231ps %ymm11, %ymm9, %ymm1 # ymm1 = (ymm9 * ymm11) + ymm1 vfmadd231ps %ymm13, %ymm8, %ymm6 # ymm6 = (ymm8 * ymm13) + ymm6 vfmadd231ps %ymm14, %ymm8, %ymm7 # ymm7 = (ymm8 * ymm14) + ymm7 vfmadd231ps %ymm9, %ymm8, %ymm10 # ymm10 = (ymm8 * ymm9) + ymm10 vfmadd231ps 0xa0(%rsi,%r9), %ymm8, %ymm1 # ymm1 = (ymm8 * mem) + ymm1 vmovaps %ymm2, (%r10,%r9) vmovaps %ymm3, 0x20(%r10,%r9) vmovaps %ymm4, 0x40(%r10,%r9) vmovaps %ymm5, 0x60(%r10,%r9) vmovaps %ymm6, (%rdx,%r9) vmovaps %ymm7, 0x20(%rdx,%r9) vmovaps %ymm10, 0x40(%rdx,%r9) vmovaps %ymm1, 0x60(%rdx,%r9) addl $0x4, %r8d subq $-0x80, %r9 jmp 0x29c18c vmovaps 0x20(%r15,%r9), %ymm3 vmovaps 0x40(%r15,%r9), %ymm4 vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm9 vmovaps 0x40(%rbx,%r11), %ymm8 vmovaps 0x60(%rbx,%r11), %ymm5 vmovaps (%r15,%r9), %ymm2 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vfmadd231ps %ymm3, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm3) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm3) + ymm0 vfmadd231ps %ymm4, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm4) + ymm3 vfmadd231ps %ymm4, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm4) + ymm2 vfmadd231ps 0x60(%r15,%r9), %ymm8, %ymm3 # ymm3 = (ymm8 * mem) + ymm3 vmovaps 0x80(%rbx,%r11), %ymm7 vmovaps 0xa0(%rbx,%r11), %ymm6 vmovaps (%r12,%r9), %ymm4 vmovaps 0x20(%r12,%r9), %ymm10 vmovaps 0x40(%r12,%r9), %ymm11 vmovaps 0x60(%r12,%r9), %ymm12 vfmadd231ps %ymm4, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm4) + ymm2 vfmadd231ps %ymm10, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm10) + ymm3 vfmadd213ps %ymm0, %ymm1, %ymm4 # ymm4 = (ymm1 * ymm4) + ymm0 vfmadd213ps %ymm0, %ymm10, %ymm1 # ymm1 = (ymm10 * ymm1) + ymm0 vfmadd231ps %ymm10, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm10) + ymm2 vfmadd231ps %ymm11, %ymm7, %ymm3 # ymm3 = (ymm7 * ymm11) + ymm3 vfmadd231ps %ymm10, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm10) + ymm4 vfmadd231ps %ymm9, %ymm11, %ymm1 # ymm1 = (ymm11 * ymm9) + ymm1 vfmadd231ps %ymm11, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm11) + ymm2 vfmadd231ps %ymm12, %ymm6, %ymm3 # ymm3 = (ymm6 * ymm12) + ymm3 vfmadd231ps %ymm11, %ymm8, %ymm4 # ymm4 = (ymm8 * ymm11) + ymm4 vfmadd231ps %ymm12, %ymm8, %ymm1 # ymm1 = (ymm8 * ymm12) + ymm1 vmovaps 0xc0(%rbx,%r11), %ymm10 vmovaps 0xe0(%rbx,%r11), %ymm9 vmovaps 0x100(%rbx,%r11), %ymm8 vmovaps (%r13,%r9), %ymm11 vmovaps 0x20(%r13,%r9), %ymm12 vmovaps 0x40(%r13,%r9), %ymm13 vmovaps 0x60(%r13,%r9), %ymm14 vfmadd231ps %ymm11, %ymm10, %ymm2 # ymm2 = (ymm10 * ymm11) + ymm2 vfmadd231ps %ymm12, %ymm10, %ymm3 # ymm3 = (ymm10 * ymm12) + ymm3 vfmadd231ps %ymm11, %ymm5, %ymm4 # ymm4 = (ymm5 * ymm11) + ymm4 vfmadd231ps %ymm5, %ymm12, %ymm1 # ymm1 = (ymm12 * ymm5) + ymm1 vfmadd231ps %ymm12, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm12) + ymm2 vfmadd231ps %ymm13, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm13) + ymm3 vfmadd231ps %ymm12, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm12) + ymm4 vfmadd231ps %ymm7, %ymm13, %ymm1 # ymm1 = (ymm13 * ymm7) + ymm1 vfmadd231ps %ymm13, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm13) + ymm2 vfmadd231ps %ymm14, %ymm8, %ymm3 # ymm3 = (ymm8 * ymm14) + ymm3 vfmadd231ps %ymm13, %ymm6, %ymm4 # ymm4 = (ymm6 * ymm13) + ymm4 vfmadd231ps %ymm14, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm14) + ymm1 vmovaps 0x20(%rsi,%r9), %ymm5 vmovaps 0x40(%rsi,%r9), %ymm6 vfmadd231ps (%rsi,%r9), %ymm10, %ymm4 # ymm4 = (ymm10 * mem) + ymm4 vfmadd231ps %ymm10, %ymm5, %ymm1 # ymm1 = (ymm5 * ymm10) + ymm1 vfmadd231ps %ymm5, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm5) + ymm4 vfmadd231ps %ymm9, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm9) + ymm1 vfmadd231ps %ymm6, %ymm8, %ymm4 # ymm4 = (ymm8 * ymm6) + ymm4 vfmadd231ps 0x60(%rsi,%r9), %ymm8, %ymm1 # ymm1 = (ymm8 * mem) + ymm1 vmovaps %ymm2, (%r10,%r9) vmovaps %ymm3, 0x20(%r10,%r9) vmovaps %ymm4, (%rdx,%r9) vmovaps %ymm1, 0x20(%rdx,%r9) addl $0x2, %r8d addq $0x40, %r9 leal 0x1(%r8), %r14d cmpl %eax, %r14d jl 0x29c43f movq 0x8(%rsp), %r14 cmpl %eax, %r8d jge 0x29c6d2 vmovaps (%rbx,%r11), %ymm4 vmovaps 0x20(%rbx,%r11), %ymm5 vmovaps 0x40(%rbx,%r11), %ymm6 vmovaps (%r15,%r9), %ymm1 vfmadd213ps %ymm0, %ymm4, %ymm1 # ymm1 = (ymm4 * ymm1) + ymm0 vfmadd231ps 0x20(%r15,%r9), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vmovaps 0x60(%rbx,%r11), %ymm7 vfmadd231ps 0x40(%r15,%r9), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps 0x80(%rbx,%r11), %ymm8 vmovaps 0xa0(%rbx,%r11), %ymm3 vmovaps (%r12,%r9), %ymm2 vmovaps 0x20(%r12,%r9), %ymm9 vmovaps 0x40(%r12,%r9), %ymm10 vfmadd231ps %ymm2, %ymm7, %ymm1 # ymm1 = (ymm7 * ymm2) + ymm1 vfmadd213ps %ymm0, %ymm4, %ymm2 # ymm2 = (ymm4 * ymm2) + ymm0 vfmadd231ps %ymm9, %ymm8, %ymm1 # ymm1 = (ymm8 * ymm9) + ymm1 vfmadd231ps %ymm9, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm9) + ymm2 vfmadd231ps %ymm10, %ymm3, %ymm1 # ymm1 = (ymm3 * ymm10) + ymm1 vfmadd231ps %ymm10, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm10) + ymm2 vmovaps 0xc0(%rbx,%r11), %ymm4 vmovaps 0xe0(%rbx,%r11), %ymm5 vmovaps 0x100(%rbx,%r11), %ymm6 vmovaps (%r13,%r9), %ymm9 vmovaps 0x20(%r13,%r9), %ymm10 vmovaps 0x40(%r13,%r9), %ymm11 vfmadd231ps %ymm9, %ymm4, %ymm1 # ymm1 = (ymm4 * ymm9) + ymm1 vfmadd231ps %ymm9, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm9) + ymm2 vfmadd231ps %ymm10, %ymm5, %ymm1 # ymm1 = (ymm5 * ymm10) + ymm1 vfmadd231ps %ymm10, %ymm8, %ymm2 # ymm2 = (ymm8 * ymm10) + ymm2 vfmadd231ps %ymm11, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm11) + ymm1 vfmadd231ps %ymm11, %ymm3, %ymm2 # ymm2 = (ymm3 * ymm11) + ymm2 vfmadd231ps (%rsi,%r9), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x20(%rsi,%r9), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps 0x40(%rsi,%r9), %ymm6, %ymm2 # ymm2 = (ymm6 * mem) + ymm2 vmovaps %ymm1, (%r10,%r9) vmovaps %ymm2, (%rdx,%r9) incl %r8d addq $0x20, %r9 jmp 0x29c5d2 addq %rdi, %r15 addq %r9, %r15 addq %rdi, %r12 addq %r9, %r12 addq %rdi, %r13 addq %r9, %r13 addq %rdi, %rsi addq %r9, %rsi movq 0x10(%rsp), %r8 addq %r8, %r10 addq %r9, %r10 addq %r8, %rdx addq %r9, %rdx addl $0x2, %ebp jmp 0x29c176 xorl %edx, %edx xorl %esi, %esi leal 0x3(%rsi), %r8d cmpl %eax, %r8d jge 0x29c9a5 vmovaps 0x20(%r15,%rdx), %ymm3 vmovaps 0x40(%r15,%rdx), %ymm4 vmovaps 0x60(%r15,%rdx), %ymm8 vmovaps 0x80(%r15,%rdx), %ymm6 vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm9 vmovaps 0x40(%rbx,%r11), %ymm7 vmovaps 0x60(%rbx,%r11), %ymm5 vmovaps (%r15,%rdx), %ymm2 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vfmadd231ps %ymm3, %ymm9, %ymm2 # ymm2 = (ymm9 * ymm3) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm3 # ymm3 = (ymm1 * ymm3) + ymm0 vfmadd231ps %ymm4, %ymm9, %ymm3 # ymm3 = (ymm9 * ymm4) + ymm3 vfmadd231ps %ymm4, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm4) + ymm2 vfmadd213ps %ymm0, %ymm1, %ymm4 # ymm4 = (ymm1 * ymm4) + ymm0 vfmadd213ps %ymm0, %ymm8, %ymm1 # ymm1 = (ymm8 * ymm1) + ymm0 vfmadd231ps %ymm8, %ymm9, %ymm4 # ymm4 = (ymm9 * ymm8) + ymm4 vfmadd231ps %ymm9, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm9) + ymm1 vfmadd231ps %ymm8, %ymm7, %ymm3 # ymm3 = (ymm7 * ymm8) + ymm3 vfmadd231ps 0xa0(%r15,%rdx), %ymm7, %ymm1 # ymm1 = (ymm7 * mem) + ymm1 vfmadd231ps %ymm6, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm6) + ymm4 vmovaps 0x80(%rbx,%r11), %ymm6 vmovaps 0xa0(%rbx,%r11), %ymm7 vmovaps 0x20(%r12,%rdx), %ymm8 vmovaps 0x40(%r12,%rdx), %ymm9 vmovaps 0x60(%r12,%rdx), %ymm10 vfmadd231ps (%r12,%rdx), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vmovaps 0x80(%r12,%rdx), %ymm11 vfmadd231ps %ymm8, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm8) + ymm3 vfmadd231ps %ymm9, %ymm5, %ymm4 # ymm4 = (ymm5 * ymm9) + ymm4 vfmadd231ps %ymm5, %ymm10, %ymm1 # ymm1 = (ymm10 * ymm5) + ymm1 vfmadd231ps %ymm8, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm8) + ymm2 vfmadd231ps %ymm9, %ymm6, %ymm3 # ymm3 = (ymm6 * ymm9) + ymm3 vfmadd231ps %ymm10, %ymm6, %ymm4 # ymm4 = (ymm6 * ymm10) + ymm4 vfmadd231ps %ymm6, %ymm11, %ymm1 # ymm1 = (ymm11 * ymm6) + ymm1 vfmadd231ps %ymm9, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm9) + ymm2 vfmadd231ps %ymm10, %ymm7, %ymm3 # ymm3 = (ymm7 * ymm10) + ymm3 vfmadd231ps 0xa0(%r12,%rdx), %ymm7, %ymm1 # ymm1 = (ymm7 * mem) + ymm1 vfmadd231ps %ymm11, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm11) + ymm4 vmovaps 0xc0(%rbx,%r11), %ymm5 vmovaps 0xe0(%rbx,%r11), %ymm6 vmovaps 0x100(%rbx,%r11), %ymm7 vmovaps 0x20(%r13,%rdx), %ymm8 vmovaps 0x40(%r13,%rdx), %ymm9 vmovaps 0x60(%r13,%rdx), %ymm10 vmovaps 0x80(%r13,%rdx), %ymm11 vfmadd231ps (%r13,%rdx), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps %ymm8, %ymm5, %ymm3 # ymm3 = (ymm5 * ymm8) + ymm3 vfmadd231ps %ymm9, %ymm5, %ymm4 # ymm4 = (ymm5 * ymm9) + ymm4 vfmadd231ps %ymm5, %ymm10, %ymm1 # ymm1 = (ymm10 * ymm5) + ymm1 vfmadd231ps %ymm8, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm8) + ymm2 vfmadd231ps %ymm9, %ymm6, %ymm3 # ymm3 = (ymm6 * ymm9) + ymm3 vfmadd231ps %ymm10, %ymm6, %ymm4 # ymm4 = (ymm6 * ymm10) + ymm4 vfmadd231ps %ymm6, %ymm11, %ymm1 # ymm1 = (ymm11 * ymm6) + ymm1 vfmadd231ps %ymm9, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm9) + ymm2 vfmadd231ps %ymm10, %ymm7, %ymm3 # ymm3 = (ymm7 * ymm10) + ymm3 vfmadd231ps %ymm11, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm11) + ymm4 vfmadd231ps 0xa0(%r13,%rdx), %ymm7, %ymm1 # ymm1 = (ymm7 * mem) + ymm1 vmovaps %ymm2, (%r10,%rdx) vmovaps %ymm3, 0x20(%r10,%rdx) vmovaps %ymm4, 0x40(%r10,%rdx) vmovaps %ymm1, 0x60(%r10,%rdx) addl $0x4, %esi subq $-0x80, %rdx jmp 0x29c707 vmovaps 0x20(%r15,%rdx), %ymm3 vmovaps 0x40(%r15,%rdx), %ymm4 vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm5 vmovaps 0x40(%rbx,%r11), %ymm6 vmovaps 0x60(%rbx,%r11), %ymm7 vmovaps (%r15,%rdx), %ymm2 vfmadd213ps %ymm0, %ymm1, %ymm2 # ymm2 = (ymm1 * ymm2) + ymm0 vfmadd213ps %ymm0, %ymm3, %ymm1 # ymm1 = (ymm3 * ymm1) + ymm0 vfmadd231ps %ymm3, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm3) + ymm2 vfmadd231ps %ymm5, %ymm4, %ymm1 # ymm1 = (ymm4 * ymm5) + ymm1 vfmadd231ps %ymm4, %ymm6, %ymm2 # ymm2 = (ymm6 * ymm4) + ymm2 vfmadd231ps 0x60(%r15,%rdx), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps 0x80(%rbx,%r11), %ymm3 vmovaps 0xa0(%rbx,%r11), %ymm4 vmovaps 0x20(%r12,%rdx), %ymm5 vfmadd231ps (%r12,%rdx), %ymm7, %ymm2 # ymm2 = (ymm7 * mem) + ymm2 vmovaps 0x40(%r12,%rdx), %ymm6 vfmadd231ps %ymm7, %ymm5, %ymm1 # ymm1 = (ymm5 * ymm7) + ymm1 vfmadd231ps %ymm5, %ymm3, %ymm2 # ymm2 = (ymm3 * ymm5) + ymm2 vfmadd231ps %ymm3, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm3) + ymm1 vfmadd231ps %ymm6, %ymm4, %ymm2 # ymm2 = (ymm4 * ymm6) + ymm2 vfmadd231ps 0x60(%r12,%rdx), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vmovaps 0xc0(%rbx,%r11), %ymm3 vmovaps 0xe0(%rbx,%r11), %ymm4 vmovaps 0x100(%rbx,%r11), %ymm5 vmovaps 0x20(%r13,%rdx), %ymm6 vmovaps 0x40(%r13,%rdx), %ymm7 vfmadd231ps (%r13,%rdx), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps %ymm3, %ymm6, %ymm1 # ymm1 = (ymm6 * ymm3) + ymm1 vfmadd231ps %ymm6, %ymm4, %ymm2 # ymm2 = (ymm4 * ymm6) + ymm2 vfmadd231ps %ymm4, %ymm7, %ymm1 # ymm1 = (ymm7 * ymm4) + ymm1 vfmadd231ps %ymm7, %ymm5, %ymm2 # ymm2 = (ymm5 * ymm7) + ymm2 vfmadd231ps 0x60(%r13,%rdx), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vmovaps %ymm2, (%r10,%rdx) vmovaps %ymm1, 0x20(%r10,%rdx) addl $0x2, %esi addq $0x40, %rdx leal 0x1(%rsi), %r8d cmpl %eax, %r8d jl 0x29c8b1 jmp 0x29ca4d vmovaps (%rbx,%r11), %ymm1 vmovaps 0x20(%rbx,%r11), %ymm2 vmovaps 0x40(%rbx,%r11), %ymm3 vfmadd132ps (%r15,%rdx), %ymm0, %ymm1 # ymm1 = (ymm1 * mem) + ymm0 vfmadd231ps 0x20(%r15,%rdx), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vmovaps 0x60(%rbx,%r11), %ymm2 vfmadd231ps 0x40(%r15,%rdx), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vmovaps 0x80(%rbx,%r11), %ymm3 vfmadd231ps (%r12,%rdx), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r12,%rdx), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vmovaps 0xa0(%rbx,%r11), %ymm2 vfmadd231ps 0x40(%r12,%rdx), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vmovaps 0xc0(%rbx,%r11), %ymm2 vmovaps 0xe0(%rbx,%r11), %ymm3 vmovaps 0x100(%rbx,%r11), %ymm4 vfmadd132ps (%r13,%rdx), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r13,%rdx), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r13,%rdx), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vmovaps %ymm2, (%r10,%rdx) incl %esi addq $0x20, %rdx cmpl %eax, %esi jl 0x29c9b7 addq %rdx, %r15 addq $0x40, %r15 addq %rdx, %r12 addq $0x40, %r12 addq %rdx, %r13 addq $0x40, %r13 incl %ebp addq %rdx, %r10 cmpl %ecx, %ebp jl 0x29c703 movq 0x30(%rsp), %r9 incq %r9 movq 0x28(%rsp), %r13 movq 0x60(%rsp), %r8 jmp 0x29c0e1 movl 0xd4(%rdi), %eax movl 0xd8(%rdi), %r15d cmpl $0x5, %eax je 0x29d684 cmpl $0x3, %eax jne 0x29ed47 cmpl $0x3, %r15d jne 0x29ed47 cmpl $0x1, 0xdc(%rdi) jne 0x29e688 cmpl $0x1, 0xe0(%rdi) jne 0x29e688 cmpl $0x1, 0xe4(%rdi) jne 0x29e688 cmpl $0x1, 0xe8(%rdi) jne 0x29e688 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xb8(%rsp), %edx movq 0x1b0(%rdi), %rsi xorl %edi, %edi testl %ecx, %ecx cmovlel %edi, %ecx testl %edx, %edx cmovlel %edi, %edx movq %rdx, 0x1d8(%rsp) movl $0x80, %r8d movq %rsi, 0x1d0(%rsp) cmpq 0x1d8(%rsp), %rdi je 0x29fdf4 testq %rsi, %rsi je 0x29cb44 movq %rdi, %rdx shlq $0x4, %rdx vmovups (%rsi,%rdx), %xmm0 jmp 0x29cb48 vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x60(%rsp) movq 0x40(%r13), %r9 imulq %rdi, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %rdx movslq 0x54(%r14), %rsi imulq %rdi, %rsi imulq 0x38(%r14), %rsi movslq 0xac(%rsp), %r10 movq 0xc0(%rsp), %r13 imulq %rdi, %r13 movq 0x90(%rsp), %r11 imulq %r11, %r13 addq 0x80(%rsp), %r13 imulq %r11, %r10 leaq (%r10,%r13), %rbp leaq (,%r10,2), %r11 addq %r13, %r11 vmovaps (%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x70(%rsp) vmovaps 0x10(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x10(%rsp) vmovaps 0x20(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0xd0(%rsp) vmovaps 0x30(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0xe0(%rsp) vmovaps 0x40(%rdx,%rsi), %xmm15 vmovaps 0x50(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x170(%rsp) vmovaps 0x60(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x30(%rsp) vmovaps 0x70(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x160(%rsp) vmovaps 0x80(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x150(%rsp) xorl %r10d, %r10d vmovaps %xmm15, 0x40(%rsp) cmpl %ecx, %r10d je 0x29d323 leaq 0x20(%r13), %r15 leaq 0x20(%rbp), %rbx leaq 0x20(%r11), %rdx xorl %esi, %esi xorl %r12d, %r12d leal 0x7(%r12), %r14d cmpl %eax, %r14d jge 0x29d152 vmovaps (%r13,%rsi), %xmm0 vmovaps 0x60(%rsp), %xmm14 vmovaps 0x70(%rsp), %xmm5 vfmadd213ps %xmm14, %xmm5, %xmm0 # xmm0 = (xmm5 * xmm0) + xmm14 vmovaps 0x10(%r13,%rsi), %xmm1 vmovaps 0x20(%r13,%rsi), %xmm11 vmovaps 0x30(%r13,%rsi), %xmm10 vfmadd231ps 0x10(%rsp), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0xd0(%rsp), %xmm6 vfmadd231ps %xmm11, %xmm6, %xmm0 # xmm0 = (xmm6 * xmm11) + xmm0 vmovaps 0xe0(%rsp), %xmm7 vfmadd231ps (%rbp,%rsi), %xmm7, %xmm0 # xmm0 = (xmm7 * mem) + xmm0 vmovaps 0x10(%rbp,%rsi), %xmm2 vmovaps 0x20(%rbp,%rsi), %xmm3 vfmadd231ps %xmm2, %xmm15, %xmm0 # xmm0 = (xmm15 * xmm2) + xmm0 vmovaps 0x170(%rsp), %xmm8 vfmadd231ps %xmm3, %xmm8, %xmm0 # xmm0 = (xmm8 * xmm3) + xmm0 vmovaps 0x30(%rsp), %xmm4 vfmadd231ps (%r11,%rsi), %xmm4, %xmm0 # xmm0 = (xmm4 * mem) + xmm0 vmovaps 0x30(%rbp,%rsi), %xmm15 vmovaps %xmm15, 0x280(%rsp) vmovaps 0x10(%r11,%rsi), %xmm12 vmovaps 0x20(%r11,%rsi), %xmm13 vmovaps 0x30(%r11,%rsi), %xmm4 vmovaps %xmm4, 0xf0(%rsp) vmovaps 0x160(%rsp), %xmm9 vfmadd231ps %xmm12, %xmm9, %xmm0 # xmm0 = (xmm9 * xmm12) + xmm0 vmovaps 0x150(%rsp), %xmm4 vfmadd231ps %xmm13, %xmm4, %xmm0 # xmm0 = (xmm4 * xmm13) + xmm0 vmovaps %xmm0, (%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm1 # xmm1 = (xmm5 * xmm1) + xmm14 vfmadd231ps 0x10(%rsp), %xmm11, %xmm1 # xmm1 = (xmm11 * mem) + xmm1 vfmadd231ps %xmm10, %xmm6, %xmm1 # xmm1 = (xmm6 * xmm10) + xmm1 vfmadd231ps %xmm2, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm2) + xmm1 vfmadd231ps 0x40(%rsp), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps %xmm15, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm15) + xmm1 vfmadd231ps 0x30(%rsp), %xmm12, %xmm1 # xmm1 = (xmm12 * mem) + xmm1 vfmadd231ps %xmm13, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm13) + xmm1 vmovaps 0xf0(%rsp), %xmm0 vfmadd231ps %xmm0, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm0) + xmm1 vmovaps 0x40(%r13,%rsi), %xmm12 vmovaps 0x40(%rbp,%rsi), %xmm2 vmovaps 0x40(%r11,%rsi), %xmm15 vmovaps %xmm15, 0x270(%rsp) vmovaps %xmm1, 0x10(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm11 # xmm11 = (xmm5 * xmm11) + xmm14 vfmadd231ps 0x10(%rsp), %xmm10, %xmm11 # xmm11 = (xmm10 * mem) + xmm11 vfmadd231ps %xmm12, %xmm6, %xmm11 # xmm11 = (xmm6 * xmm12) + xmm11 vfmadd231ps %xmm3, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm3) + xmm11 vmovaps 0x280(%rsp), %xmm15 vfmadd231ps 0x40(%rsp), %xmm15, %xmm11 # xmm11 = (xmm15 * mem) + xmm11 vfmadd231ps %xmm2, %xmm8, %xmm11 # xmm11 = (xmm8 * xmm2) + xmm11 vfmadd231ps 0x30(%rsp), %xmm13, %xmm11 # xmm11 = (xmm13 * mem) + xmm11 vfmadd231ps %xmm0, %xmm9, %xmm11 # xmm11 = (xmm9 * xmm0) + xmm11 vmovaps 0x270(%rsp), %xmm0 vfmadd231ps %xmm0, %xmm4, %xmm11 # xmm11 = (xmm4 * xmm0) + xmm11 vmovaps 0x50(%r13,%rsi), %xmm13 vmovaps 0x50(%rbp,%rsi), %xmm3 vmovaps 0x50(%r11,%rsi), %xmm1 vmovaps %xmm11, 0x20(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm10 # xmm10 = (xmm5 * xmm10) + xmm14 vfmadd231ps 0x10(%rsp), %xmm12, %xmm10 # xmm10 = (xmm12 * mem) + xmm10 vfmadd231ps %xmm13, %xmm6, %xmm10 # xmm10 = (xmm6 * xmm13) + xmm10 vfmadd231ps %xmm15, %xmm7, %xmm10 # xmm10 = (xmm7 * xmm15) + xmm10 vfmadd231ps 0x40(%rsp), %xmm2, %xmm10 # xmm10 = (xmm2 * mem) + xmm10 vfmadd231ps %xmm3, %xmm8, %xmm10 # xmm10 = (xmm8 * xmm3) + xmm10 vmovaps 0xf0(%rsp), %xmm11 vfmadd231ps 0x30(%rsp), %xmm11, %xmm10 # xmm10 = (xmm11 * mem) + xmm10 vfmadd231ps %xmm0, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm0) + xmm10 vfmadd231ps %xmm1, %xmm4, %xmm10 # xmm10 = (xmm4 * xmm1) + xmm10 vmovaps %xmm1, 0x290(%rsp) vmovaps 0x60(%r13,%rsi), %xmm11 vmovaps 0x60(%rbp,%rsi), %xmm15 vmovaps 0x60(%r11,%rsi), %xmm0 vmovaps %xmm10, 0x30(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm14 vfmadd231ps 0x10(%rsp), %xmm13, %xmm12 # xmm12 = (xmm13 * mem) + xmm12 vfmadd231ps %xmm11, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm11) + xmm12 vfmadd231ps %xmm2, %xmm7, %xmm12 # xmm12 = (xmm7 * xmm2) + xmm12 vfmadd231ps 0x40(%rsp), %xmm3, %xmm12 # xmm12 = (xmm3 * mem) + xmm12 vfmadd231ps %xmm15, %xmm8, %xmm12 # xmm12 = (xmm8 * xmm15) + xmm12 vmovaps 0x270(%rsp), %xmm2 vfmadd231ps 0x30(%rsp), %xmm2, %xmm12 # xmm12 = (xmm2 * mem) + xmm12 vfmadd231ps %xmm1, %xmm9, %xmm12 # xmm12 = (xmm9 * xmm1) + xmm12 vfmadd231ps %xmm0, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm0) + xmm12 vmovaps %xmm0, 0xf0(%rsp) vmovaps 0x70(%r13,%rsi), %xmm10 vmovaps 0x70(%rbp,%rsi), %xmm2 vmovaps 0x70(%r11,%rsi), %xmm1 vmovaps %xmm12, 0x40(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm13 # xmm13 = (xmm5 * xmm13) + xmm14 vfmadd231ps 0x10(%rsp), %xmm11, %xmm13 # xmm13 = (xmm11 * mem) + xmm13 vfmadd231ps %xmm10, %xmm6, %xmm13 # xmm13 = (xmm6 * xmm10) + xmm13 vfmadd231ps %xmm3, %xmm7, %xmm13 # xmm13 = (xmm7 * xmm3) + xmm13 vfmadd231ps 0x40(%rsp), %xmm15, %xmm13 # xmm13 = (xmm15 * mem) + xmm13 vfmadd231ps %xmm2, %xmm8, %xmm13 # xmm13 = (xmm8 * xmm2) + xmm13 vmovaps 0x290(%rsp), %xmm3 vfmadd231ps 0x30(%rsp), %xmm3, %xmm13 # xmm13 = (xmm3 * mem) + xmm13 vfmadd231ps %xmm0, %xmm9, %xmm13 # xmm13 = (xmm9 * xmm0) + xmm13 vfmadd231ps %xmm1, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm1) + xmm13 vmovaps 0x80(%r13,%rsi), %xmm3 vmovaps 0x80(%rbp,%rsi), %xmm12 vmovaps 0x80(%r11,%rsi), %xmm0 vmovaps %xmm13, 0x50(%r9,%rsi) vfmadd213ps %xmm14, %xmm5, %xmm11 # xmm11 = (xmm5 * xmm11) + xmm14 vfmadd231ps 0x10(%rsp), %xmm10, %xmm11 # xmm11 = (xmm10 * mem) + xmm11 vfmadd231ps %xmm3, %xmm6, %xmm11 # xmm11 = (xmm6 * xmm3) + xmm11 vfmadd231ps %xmm15, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm15) + xmm11 vmovaps 0x40(%rsp), %xmm15 vfmadd231ps %xmm2, %xmm15, %xmm11 # xmm11 = (xmm15 * xmm2) + xmm11 vfmadd231ps %xmm12, %xmm8, %xmm11 # xmm11 = (xmm8 * xmm12) + xmm11 vmovaps 0xf0(%rsp), %xmm13 vfmadd231ps 0x30(%rsp), %xmm13, %xmm11 # xmm11 = (xmm13 * mem) + xmm11 vfmadd213ps %xmm14, %xmm5, %xmm10 # xmm10 = (xmm5 * xmm10) + xmm14 vfmadd231ps 0x10(%rsp), %xmm3, %xmm10 # xmm10 = (xmm3 * mem) + xmm10 vfmadd231ps 0x90(%r13,%rsi), %xmm6, %xmm10 # xmm10 = (xmm6 * mem) + xmm10 vfmadd231ps %xmm2, %xmm7, %xmm10 # xmm10 = (xmm7 * xmm2) + xmm10 vfmadd231ps %xmm12, %xmm15, %xmm10 # xmm10 = (xmm15 * xmm12) + xmm10 vfmadd231ps %xmm1, %xmm9, %xmm11 # xmm11 = (xmm9 * xmm1) + xmm11 vfmadd231ps 0x90(%rbp,%rsi), %xmm8, %xmm10 # xmm10 = (xmm8 * mem) + xmm10 vfmadd231ps 0x30(%rsp), %xmm1, %xmm10 # xmm10 = (xmm1 * mem) + xmm10 vfmadd231ps %xmm0, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm0) + xmm10 vfmadd231ps 0x90(%r11,%rsi), %xmm4, %xmm10 # xmm10 = (xmm4 * mem) + xmm10 vfmadd231ps %xmm0, %xmm4, %xmm11 # xmm11 = (xmm4 * xmm0) + xmm11 vmovaps %xmm11, 0x60(%r9,%rsi) vmovaps %xmm10, 0x70(%r9,%rsi) addl $0x8, %r12d addq %r8, %rsi addq %r8, %r15 addq %r8, %rbx addq %r8, %rdx jmp 0x29cc45 vmovaps (%r13,%rsi), %xmm13 vmovaps 0x70(%rsp), %xmm12 vfmadd213ps 0x60(%rsp), %xmm12, %xmm13 # xmm13 = (xmm12 * xmm13) + mem vmovaps 0x10(%r13,%rsi), %xmm1 vmovaps 0x20(%r13,%rsi), %xmm10 vmovaps 0x30(%r13,%rsi), %xmm11 vfmadd231ps 0x10(%rsp), %xmm1, %xmm13 # xmm13 = (xmm1 * mem) + xmm13 vmovaps 0xd0(%rsp), %xmm6 vfmadd231ps %xmm10, %xmm6, %xmm13 # xmm13 = (xmm6 * xmm10) + xmm13 vmovaps 0xe0(%rsp), %xmm7 vfmadd231ps (%rbp,%rsi), %xmm7, %xmm13 # xmm13 = (xmm7 * mem) + xmm13 vmovaps 0x10(%rbp,%rsi), %xmm14 vmovaps 0x20(%rbp,%rsi), %xmm5 vfmadd231ps %xmm14, %xmm15, %xmm13 # xmm13 = (xmm15 * xmm14) + xmm13 vmovaps 0x170(%rsp), %xmm8 vfmadd231ps %xmm5, %xmm8, %xmm13 # xmm13 = (xmm8 * xmm5) + xmm13 vmovaps 0x30(%rsp), %xmm0 vfmadd231ps (%r11,%rsi), %xmm0, %xmm13 # xmm13 = (xmm0 * mem) + xmm13 vmovaps 0x30(%rbp,%rsi), %xmm0 vmovaps 0x10(%r11,%rsi), %xmm15 vmovaps 0x20(%r11,%rsi), %xmm3 vmovaps 0x30(%r11,%rsi), %xmm2 vmovaps %xmm2, 0xf0(%rsp) vmovaps 0x160(%rsp), %xmm9 vfmadd231ps %xmm15, %xmm9, %xmm13 # xmm13 = (xmm9 * xmm15) + xmm13 vmovaps 0x150(%rsp), %xmm4 vfmadd231ps %xmm3, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm3) + xmm13 vmovaps %xmm13, (%r9,%rsi) vmovaps 0x60(%rsp), %xmm13 vfmadd213ps %xmm13, %xmm12, %xmm1 # xmm1 = (xmm12 * xmm1) + xmm13 vfmadd231ps 0x10(%rsp), %xmm10, %xmm1 # xmm1 = (xmm10 * mem) + xmm1 vfmadd231ps %xmm11, %xmm6, %xmm1 # xmm1 = (xmm6 * xmm11) + xmm1 vfmadd231ps %xmm14, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm14) + xmm1 vfmadd231ps 0x40(%rsp), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps %xmm0, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm0) + xmm1 vfmadd231ps 0x30(%rsp), %xmm15, %xmm1 # xmm1 = (xmm15 * mem) + xmm1 vfmadd231ps %xmm3, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm3) + xmm1 vfmadd231ps %xmm2, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm2) + xmm1 vmovaps 0x40(%r13,%rsi), %xmm14 vmovaps 0x40(%rbp,%rsi), %xmm15 vmovaps 0x40(%r11,%rsi), %xmm2 vmovaps %xmm1, 0x10(%r9,%rsi) vfmadd213ps %xmm13, %xmm12, %xmm10 # xmm10 = (xmm12 * xmm10) + xmm13 vfmadd231ps 0x10(%rsp), %xmm11, %xmm10 # xmm10 = (xmm11 * mem) + xmm10 vfmadd231ps %xmm14, %xmm6, %xmm10 # xmm10 = (xmm6 * xmm14) + xmm10 vfmadd231ps %xmm5, %xmm7, %xmm10 # xmm10 = (xmm7 * xmm5) + xmm10 vfmadd231ps 0x40(%rsp), %xmm0, %xmm10 # xmm10 = (xmm0 * mem) + xmm10 vfmadd231ps %xmm15, %xmm8, %xmm10 # xmm10 = (xmm8 * xmm15) + xmm10 vfmadd231ps 0x30(%rsp), %xmm3, %xmm10 # xmm10 = (xmm3 * mem) + xmm10 vfmadd213ps %xmm13, %xmm12, %xmm11 # xmm11 = (xmm12 * xmm11) + xmm13 vfmadd231ps 0x10(%rsp), %xmm14, %xmm11 # xmm11 = (xmm14 * mem) + xmm11 vfmadd231ps 0x50(%r13,%rsi), %xmm6, %xmm11 # xmm11 = (xmm6 * mem) + xmm11 vfmadd231ps %xmm0, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm0) + xmm11 vfmadd231ps 0x40(%rsp), %xmm15, %xmm11 # xmm11 = (xmm15 * mem) + xmm11 vmovaps 0x40(%rsp), %xmm15 vmovaps 0xf0(%rsp), %xmm1 vfmadd231ps %xmm1, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm1) + xmm10 vfmadd231ps 0x50(%rbp,%rsi), %xmm8, %xmm11 # xmm11 = (xmm8 * mem) + xmm11 vfmadd231ps 0x30(%rsp), %xmm1, %xmm11 # xmm11 = (xmm1 * mem) + xmm11 vfmadd231ps %xmm2, %xmm9, %xmm11 # xmm11 = (xmm9 * xmm2) + xmm11 vfmadd231ps 0x50(%r11,%rsi), %xmm4, %xmm11 # xmm11 = (xmm4 * mem) + xmm11 vfmadd231ps %xmm2, %xmm4, %xmm10 # xmm10 = (xmm4 * xmm2) + xmm10 vmovaps %xmm10, 0x20(%r9,%rsi) vmovaps %xmm11, 0x30(%r9,%rsi) addl $0x4, %r12d addq $0x40, %rsi addq $0x40, %r15 addq $0x40, %rbx addq $0x40, %rdx leal 0x3(%r12), %r14d cmpl %eax, %r14d jl 0x29cfa2 jmp 0x29d25d vmovaps 0x10(%r13,%rsi), %xmm0 vmovaps 0x20(%r13,%rsi), %xmm2 vmovaps (%r13,%rsi), %xmm1 vmovaps 0x60(%rsp), %xmm3 vmovaps 0x70(%rsp), %xmm4 vfmadd213ps %xmm3, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm1) + xmm3 vmovaps 0x10(%rsp), %xmm6 vfmadd231ps %xmm0, %xmm6, %xmm1 # xmm1 = (xmm6 * xmm0) + xmm1 vmovaps 0xd0(%rsp), %xmm5 vfmadd231ps %xmm2, %xmm5, %xmm1 # xmm1 = (xmm5 * xmm2) + xmm1 vfmadd213ps %xmm3, %xmm4, %xmm0 # xmm0 = (xmm4 * xmm0) + xmm3 vfmadd231ps %xmm2, %xmm6, %xmm0 # xmm0 = (xmm6 * xmm2) + xmm0 vmovaps 0x10(%rbp,%rsi), %xmm2 vmovaps 0xe0(%rsp), %xmm3 vfmadd231ps (%rbp,%rsi), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x30(%r13,%rsi), %xmm5, %xmm0 # xmm0 = (xmm5 * mem) + xmm0 vfmadd231ps %xmm2, %xmm15, %xmm1 # xmm1 = (xmm15 * xmm2) + xmm1 vfmadd231ps %xmm2, %xmm3, %xmm0 # xmm0 = (xmm3 * xmm2) + xmm0 vmovaps 0x20(%rbp,%rsi), %xmm2 vmovaps 0x170(%rsp), %xmm3 vfmadd231ps %xmm2, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm2) + xmm1 vfmadd231ps %xmm2, %xmm15, %xmm0 # xmm0 = (xmm15 * xmm2) + xmm0 vmovaps 0x10(%r11,%rsi), %xmm2 vmovaps 0x30(%rsp), %xmm5 vfmadd231ps (%r11,%rsi), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vmovaps 0x160(%rsp), %xmm4 vfmadd231ps %xmm2, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm2) + xmm1 vfmadd231ps 0x30(%rbp,%rsi), %xmm3, %xmm0 # xmm0 = (xmm3 * mem) + xmm0 vfmadd231ps %xmm2, %xmm5, %xmm0 # xmm0 = (xmm5 * xmm2) + xmm0 vmovaps 0x20(%r11,%rsi), %xmm2 vfmadd231ps %xmm2, %xmm4, %xmm0 # xmm0 = (xmm4 * xmm2) + xmm0 vmovaps 0x150(%rsp), %xmm3 vfmadd231ps 0x30(%r11,%rsi), %xmm3, %xmm0 # xmm0 = (xmm3 * mem) + xmm0 vfmadd231ps %xmm2, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm2) + xmm1 vmovaps %xmm1, (%r9,%rsi) vmovaps %xmm0, 0x10(%r9,%rsi) addl $0x2, %r12d addq $0x20, %rsi addq $0x20, %r15 addq $0x20, %rbx addq $0x20, %rdx leal 0x1(%r12), %r14d cmpl %eax, %r14d jl 0x29d165 addq %rsi, %r9 movq 0x8(%rsp), %r14 cmpl %eax, %r12d jge 0x29d312 vmovaps -0x20(%r15), %xmm0 vmovaps 0x70(%rsp), %xmm1 vfmadd213ps 0x60(%rsp), %xmm1, %xmm0 # xmm0 = (xmm1 * xmm0) + mem vmovaps 0x10(%rsp), %xmm1 vfmadd231ps -0x10(%r15), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0xd0(%rsp), %xmm1 vfmadd231ps (%r15), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0xe0(%rsp), %xmm1 vfmadd231ps -0x20(%rbx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vfmadd231ps -0x10(%rbx), %xmm15, %xmm0 # xmm0 = (xmm15 * mem) + xmm0 vmovaps 0x170(%rsp), %xmm1 vfmadd231ps (%rbx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0x30(%rsp), %xmm1 vfmadd231ps -0x20(%rdx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0x160(%rsp), %xmm1 vfmadd231ps -0x10(%rdx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps 0x150(%rsp), %xmm1 vfmadd231ps (%rdx), %xmm1, %xmm0 # xmm0 = (xmm1 * mem) + xmm0 vmovaps %xmm0, (%r9) addq $0x10, %r9 incl %r12d addq $0x10, %r15 addq $0x10, %rbx addq $0x10, %rdx jmp 0x29d273 incl %r10d movq %r15, %r13 movq %rbx, %rbp movq %rdx, %r11 jmp 0x29cc2b incq %rdi movq 0x28(%rsp), %r13 movq 0x1d0(%rsp), %rsi jmp 0x29cb23 cmpl $0x3, 0xd4(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x3, 0xd8(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x1, 0xdc(%rdi) movq %r9, 0x30(%rsp) jne 0x29dfac cmpl $0x1, 0xe0(%rdi) jne 0x29dfac cmpl $0x1, 0xe4(%rdi) jne 0x29dfac cmpl $0x1, 0xe8(%rdi) jne 0x29dfac movslq 0xac(%rsp), %rax movl 0x2c(%r13), %edx movl 0x30(%r13), %r8d movl 0xb8(%rsp), %ecx movq 0x28(%r14), %r10 imulq 0x10(%r13), %r11 movq 0x1b0(%rdi), %rsi movslq %edx, %rdi movq 0x80(%rsp), %rbx movq %rbx, 0x170(%rsp) movq 0x90(%rsp), %rbx imulq 0xc0(%rsp), %rbx movq %rbx, 0x160(%rsp) leaq (%rax,%rax), %rbx movq %rbx, 0x150(%rsp) leaq (%rax,%rax,2), %rbx movq %rbx, 0xf0(%rsp) movq %rax, 0xd0(%rsp) addl $0x2, %eax movslq %eax, %rbp xorl %eax, %eax testl %ecx, %ecx cmovlel %eax, %ecx movq %rcx, 0x60(%rsp) shlq $0x2, %rbp movq %rdi, 0x8(%rsp) leaq (,%rdi,4), %rcx movq %rcx, 0x10(%rsp) movq %rsi, 0xe0(%rsp) movq %r11, 0x70(%rsp) cmpq 0x60(%rsp), %rax je 0x29fdf4 testq %rsi, %rsi je 0x29d44e vmovss (%rsi,%rax,4), %xmm0 jmp 0x29d452 vxorps %xmm0, %xmm0, %xmm0 movq %r11, %rsi imulq %rax, %rsi addq %r9, %rsi imulq $0x24, %rax, %r13 movq 0x160(%rsp), %r15 movq %rax, 0x40(%rsp) imulq %rax, %r15 addq 0x170(%rsp), %r15 movq 0xd0(%rsp), %rax leaq (%r15,%rax,4), %rax movq 0x150(%rsp), %rcx leaq (%r15,%rcx,4), %r9 movq 0xf0(%rsp), %rcx leaq (%r15,%rcx,4), %r11 xorl %edi, %edi movq %rsi, %rbx movl %edi, %ecx orl $0x1, %ecx cmpl %r8d, %ecx jge 0x29d5c7 movq 0x8(%rsp), %rcx leaq (%rsi,%rcx,4), %rsi xorl %r12d, %r12d movl %edx, %ecx testl %ecx, %ecx jle 0x29d59c vmovsd 0x4(%r15,%r12), %xmm1 vmovss (%r15,%r12), %xmm2 vmovlhps %xmm1, %xmm2, %xmm2 # xmm2 = xmm2[0],xmm1[0] vshufps $0xd8, %xmm1, %xmm2, %xmm1 # xmm1 = xmm2[0,2],xmm1[1,3] vmovss (%rax,%r12), %xmm2 vinsertps $0x30, %xmm2, %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],xmm2[0] vmovups (%r10,%r13), %xmm3 vmovups 0x10(%r10,%r13), %xmm4 vmovsd 0x4(%rax,%r12), %xmm5 vmovss (%r9,%r12), %xmm6 vinsertps $0x20, %xmm6, %xmm5, %xmm7 # xmm7 = xmm5[0,1],xmm6[0],xmm5[3] vinsertps $0x30, 0x4(%r9,%r12), %xmm7, %xmm7 # xmm7 = xmm7[0,1,2],mem[0] vmulps %xmm7, %xmm4, %xmm7 vhaddps %xmm7, %xmm7, %xmm7 vmulps %xmm1, %xmm3, %xmm1 vhaddps %xmm7, %xmm7, %xmm7 vmovss 0x20(%r10,%r13), %xmm8 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm1, %xmm7, %xmm1 vfmadd231ss 0x8(%r9,%r12), %xmm8, %xmm1 # xmm1 = (xmm8 * mem) + xmm1 vmovlhps %xmm5, %xmm3, %xmm7 # xmm7 = xmm3[0],xmm5[0] vshufps $0xd8, %xmm5, %xmm7, %xmm5 # xmm5 = xmm7[0,2],xmm5[1,3] vinsertps $0x30, %xmm6, %xmm5, %xmm5 # xmm5 = xmm5[0,1,2],xmm6[0] vmovsd 0x4(%r9,%r12), %xmm6 vmovhps (%r11,%r12), %xmm6, %xmm6 # xmm6 = xmm6[0,1],mem[0,1] vmulps %xmm4, %xmm6, %xmm4 vhaddps %xmm4, %xmm4, %xmm4 vmovss %xmm2, %xmm3, %xmm2 # xmm2 = xmm2[0],xmm3[1,2,3] vhaddps %xmm4, %xmm4, %xmm3 vmulps %xmm2, %xmm5, %xmm2 vhaddps %xmm2, %xmm2, %xmm2 vaddss %xmm0, %xmm1, %xmm1 vhaddps %xmm2, %xmm2, %xmm2 vaddss %xmm2, %xmm3, %xmm2 vfmadd231ss 0x8(%r11,%r12), %xmm8, %xmm2 # xmm2 = (xmm8 * mem) + xmm2 vaddss %xmm0, %xmm2, %xmm2 vmovss %xmm1, (%rbx,%r12) vmovss %xmm2, (%rsi,%r12) decl %ecx addq $0x4, %r12 jmp 0x29d4be addq %rbp, %r15 addq %r12, %r15 addq %rbp, %rax addq %r12, %rax addq %rbp, %r9 addq %r12, %r9 addq %rbp, %r11 addq %r12, %r11 addq 0x10(%rsp), %rbx addq %r12, %rbx addl $0x2, %edi addq %r12, %rsi jmp 0x29d4a2 movq 0x70(%rsp), %r11 cmpl %r8d, %edi jge 0x29d665 xorl %ecx, %ecx movl %edx, %esi testl %esi, %esi jle 0x29d646 vmovsd (%r15,%rcx), %xmm1 vinsertps $0x20, 0x8(%r15,%rcx), %xmm1, %xmm1 # xmm1 = xmm1[0,1],mem[0],xmm1[3] vinsertps $0x30, (%rax,%rcx), %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],mem[0] vmovsd 0x4(%rax,%rcx), %xmm2 vinsertf128 $0x1, %xmm2, %ymm1, %ymm2 vmovsd (%r9,%rcx), %xmm3 vinsertf128 $0x1, %xmm3, %ymm1, %ymm1 vshufpd $0x2, %ymm1, %ymm2, %ymm1 # ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[2] vmulps (%r10,%r13), %ymm1, %ymm1 vextractf128 $0x1, %ymm1, %xmm2 vhaddps %xmm1, %xmm2, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vmovss 0x20(%r10,%r13), %xmm2 vfmadd231ss 0x8(%r9,%rcx), %xmm2, %xmm1 # xmm1 = (xmm2 * mem) + xmm1 vaddss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%rbx,%rcx) decl %esi addq $0x4, %rcx jmp 0x29d5d9 addq %rcx, %r15 addq $0x8, %r15 addq %rcx, %rax addq $0x8, %rax addq %rcx, %r9 addq $0x8, %r9 incl %edi addq %rcx, %rbx jmp 0x29d5cc movq 0x40(%rsp), %rax incq %rax movq 0x28(%rsp), %r13 movq 0x30(%rsp), %r9 movq 0xe0(%rsp), %rsi jmp 0x29d437 cmpl $0x5, %r15d jne 0x29ed47 cmpl $0x1, 0xdc(%rdi) jne 0x29ea4a cmpl $0x1, 0xe0(%rdi) jne 0x29ea4a cmpl $0x1, 0xe4(%rdi) jne 0x29ea4a cmpl $0x1, 0xe8(%rdi) jne 0x29ea4a movl 0x2c(%r13), %r8d movl 0x30(%r13), %eax movl %eax, 0x8(%rsp) movl 0xac(%rsp), %eax movl 0xb8(%rsp), %ecx movq 0x1b0(%rdi), %rdx leal 0x10(,%rax,4), %eax movslq %eax, %rdi leal (,%r8,4), %eax cltq xorl %esi, %esi testl %r8d, %r8d cmovlel %esi, %r8d testl %ecx, %ecx cmovlel %esi, %ecx movq %rcx, 0x60(%rsp) shlq $0x2, %rax movq %rax, 0x30(%rsp) shlq $0x2, %rdi movq %rdx, 0x70(%rsp) movq %r8, 0x10(%rsp) cmpq 0x60(%rsp), %rsi je 0x29fdf4 testq %rdx, %rdx je 0x29d73f movq %rsi, %rax shlq $0x4, %rax vmovups (%rdx,%rax), %xmm0 jmp 0x29d743 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r10 imulq %rsi, %r10 movq 0x10(%r13), %rax imulq %rax, %r10 addq (%r13), %r10 movslq 0x2c(%r13), %r9 movq 0x28(%r14), %r11 movslq 0x54(%r14), %rbx imulq %rsi, %rbx imulq 0x38(%r14), %rbx imulq %rax, %r9 addq %r10, %r9 movslq 0xac(%rsp), %rax movq 0xc0(%rsp), %r15 movq %rsi, 0x40(%rsp) imulq %rsi, %r15 movq 0x90(%rsp), %rcx imulq %rcx, %r15 addq 0x80(%rsp), %r15 imulq %rcx, %rax leaq (%r15,%rax), %r12 leaq (%r15,%rax,2), %r13 leaq (%rax,%rax,2), %rbp addq %r15, %rbp leaq (%r15,%rax,4), %rdx leaq (%rax,%rax,4), %rax addq %r15, %rax xorl %esi, %esi movl %esi, %ecx orl $0x1, %ecx cmpl 0x8(%rsp), %ecx jge 0x29dc81 movl %r8d, %ecx xorl %r8d, %r8d subl $0x1, %ecx jb 0x29da6f vmovaps (%r11,%rbx), %xmm1 vmovaps 0x10(%r11,%rbx), %xmm3 vmovaps 0x20(%r11,%rbx), %xmm4 vmovaps 0x30(%r11,%rbx), %xmm5 vmovaps (%r15,%r8), %xmm2 vfmadd213ps %xmm0, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm0 vfmadd231ps 0x10(%r15,%r8), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%r15,%r8), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%r15,%r8), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vmovaps 0x40(%r11,%rbx), %xmm6 vfmadd231ps 0x40(%r15,%r8), %xmm6, %xmm2 # xmm2 = (xmm6 * mem) + xmm2 vmovaps (%r12,%r8), %xmm7 vmovaps 0x10(%r12,%r8), %xmm8 vmovaps 0x20(%r12,%r8), %xmm9 vmovaps 0x30(%r12,%r8), %xmm10 vmovaps 0x40(%r12,%r8), %xmm11 vfmadd213ps %xmm0, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm1) + xmm0 vfmadd231ps %xmm3, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm3) + xmm1 vfmadd231ps %xmm4, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm4) + xmm1 vfmadd231ps %xmm5, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm5) + xmm1 vfmadd231ps %xmm6, %xmm11, %xmm1 # xmm1 = (xmm11 * xmm6) + xmm1 vmovaps 0x50(%r11,%rbx), %xmm3 vmovaps 0x60(%r11,%rbx), %xmm4 vmovaps 0x70(%r11,%rbx), %xmm5 vmovaps 0x80(%r11,%rbx), %xmm6 vmovaps 0x90(%r11,%rbx), %xmm12 vfmadd231ps %xmm7, %xmm3, %xmm2 # xmm2 = (xmm3 * xmm7) + xmm2 vfmadd231ps %xmm8, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm8) + xmm2 vfmadd231ps %xmm9, %xmm5, %xmm2 # xmm2 = (xmm5 * xmm9) + xmm2 vfmadd231ps %xmm10, %xmm6, %xmm2 # xmm2 = (xmm6 * xmm10) + xmm2 vfmadd231ps %xmm11, %xmm12, %xmm2 # xmm2 = (xmm12 * xmm11) + xmm2 vmovaps (%r13,%r8), %xmm7 vmovaps 0x10(%r13,%r8), %xmm8 vmovaps 0x20(%r13,%r8), %xmm9 vmovaps 0x30(%r13,%r8), %xmm10 vmovaps 0x40(%r13,%r8), %xmm11 vfmadd231ps %xmm3, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm3) + xmm1 vfmadd231ps %xmm4, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm4) + xmm1 vfmadd231ps %xmm5, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm5) + xmm1 vfmadd231ps %xmm6, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm6) + xmm1 vfmadd231ps %xmm12, %xmm11, %xmm1 # xmm1 = (xmm11 * xmm12) + xmm1 vmovaps 0xa0(%r11,%rbx), %xmm3 vmovaps 0xb0(%r11,%rbx), %xmm4 vmovaps 0xc0(%r11,%rbx), %xmm5 vmovaps 0xd0(%r11,%rbx), %xmm6 vmovaps 0xe0(%r11,%rbx), %xmm12 vfmadd231ps %xmm7, %xmm3, %xmm2 # xmm2 = (xmm3 * xmm7) + xmm2 vfmadd231ps %xmm8, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm8) + xmm2 vfmadd231ps %xmm9, %xmm5, %xmm2 # xmm2 = (xmm5 * xmm9) + xmm2 vfmadd231ps %xmm10, %xmm6, %xmm2 # xmm2 = (xmm6 * xmm10) + xmm2 vfmadd231ps %xmm11, %xmm12, %xmm2 # xmm2 = (xmm12 * xmm11) + xmm2 vmovaps (%rbp,%r8), %xmm7 vmovaps 0x10(%rbp,%r8), %xmm8 vmovaps 0x20(%rbp,%r8), %xmm9 vmovaps 0x30(%rbp,%r8), %xmm10 vmovaps 0x40(%rbp,%r8), %xmm11 vfmadd231ps %xmm3, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm3) + xmm1 vfmadd231ps %xmm4, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm4) + xmm1 vfmadd231ps %xmm5, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm5) + xmm1 vfmadd231ps %xmm6, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm6) + xmm1 vfmadd231ps %xmm12, %xmm11, %xmm1 # xmm1 = (xmm11 * xmm12) + xmm1 vmovaps 0xf0(%r11,%rbx), %xmm4 vmovaps 0x100(%r11,%rbx), %xmm5 vmovaps 0x110(%r11,%rbx), %xmm6 vmovaps 0x120(%r11,%rbx), %xmm12 vmovaps 0x130(%r11,%rbx), %xmm13 vfmadd231ps %xmm7, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm7) + xmm2 vfmadd231ps %xmm8, %xmm5, %xmm2 # xmm2 = (xmm5 * xmm8) + xmm2 vfmadd231ps %xmm9, %xmm6, %xmm2 # xmm2 = (xmm6 * xmm9) + xmm2 vfmadd231ps %xmm10, %xmm12, %xmm2 # xmm2 = (xmm12 * xmm10) + xmm2 vfmadd231ps %xmm11, %xmm13, %xmm2 # xmm2 = (xmm13 * xmm11) + xmm2 vmovaps (%rdx,%r8), %xmm7 vmovaps 0x10(%rdx,%r8), %xmm8 vmovaps 0x20(%rdx,%r8), %xmm9 vmovaps 0x30(%rdx,%r8), %xmm10 vmovaps 0x40(%rdx,%r8), %xmm3 vfmadd231ps %xmm4, %xmm7, %xmm1 # xmm1 = (xmm7 * xmm4) + xmm1 vfmadd231ps %xmm5, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm5) + xmm1 vfmadd231ps %xmm6, %xmm9, %xmm1 # xmm1 = (xmm9 * xmm6) + xmm1 vfmadd231ps %xmm12, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm12) + xmm1 vfmadd231ps %xmm13, %xmm3, %xmm1 # xmm1 = (xmm3 * xmm13) + xmm1 vmovaps 0x140(%r11,%rbx), %xmm4 vmovaps 0x150(%r11,%rbx), %xmm5 vmovaps 0x160(%r11,%rbx), %xmm6 vmovaps 0x170(%r11,%rbx), %xmm11 vmovaps 0x180(%r11,%rbx), %xmm12 vfmadd231ps %xmm7, %xmm4, %xmm2 # xmm2 = (xmm4 * xmm7) + xmm2 vfmadd231ps %xmm8, %xmm5, %xmm2 # xmm2 = (xmm5 * xmm8) + xmm2 vfmadd231ps %xmm9, %xmm6, %xmm2 # xmm2 = (xmm6 * xmm9) + xmm2 vfmadd231ps %xmm10, %xmm11, %xmm2 # xmm2 = (xmm11 * xmm10) + xmm2 vfmadd231ps (%rax,%r8), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x10(%rax,%r8), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x20(%rax,%r8), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vfmadd231ps 0x30(%rax,%r8), %xmm11, %xmm1 # xmm1 = (xmm11 * mem) + xmm1 vfmadd231ps 0x40(%rax,%r8), %xmm12, %xmm1 # xmm1 = (xmm12 * mem) + xmm1 vfmadd231ps %xmm3, %xmm12, %xmm2 # xmm2 = (xmm12 * xmm3) + xmm2 vmovaps %xmm2, (%r10,%r8) vmovaps %xmm1, (%r9,%r8) addq $0x10, %r8 jmp 0x29d7d5 addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp addq %rdi, %rdx addq %r8, %rdx addq %rdi, %rax addq %r8, %rax movq 0x30(%rsp), %rcx addq %rcx, %r10 addq %r8, %r10 addq %rcx, %r9 addq %r8, %r9 addl $0x2, %esi movq 0x10(%rsp), %r8 jmp 0x29d7c0 movl %r8d, %ecx xorl %eax, %eax subl $0x1, %ecx jb 0x29dc5b vmovaps (%r11,%rbx), %xmm1 vmovaps 0x10(%r11,%rbx), %xmm2 vmovaps 0x20(%r11,%rbx), %xmm3 vmovaps 0x30(%r11,%rbx), %xmm4 vmovaps 0x40(%r11,%rbx), %xmm5 vfmadd132ps (%r15,%rax), %xmm0, %xmm1 # xmm1 = (xmm1 * mem) + xmm0 vfmadd231ps 0x10(%r15,%rax), %xmm2, %xmm1 # xmm1 = (xmm2 * mem) + xmm1 vfmadd231ps 0x20(%r15,%rax), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x30(%r15,%rax), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x40(%r15,%rax), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vmovaps 0x50(%r11,%rbx), %xmm2 vmovaps 0x60(%r11,%rbx), %xmm3 vmovaps 0x70(%r11,%rbx), %xmm4 vmovaps 0x80(%r11,%rbx), %xmm5 vmovaps 0x90(%r11,%rbx), %xmm6 vfmadd132ps (%r12,%rax), %xmm1, %xmm2 # xmm2 = (xmm2 * mem) + xmm1 vfmadd231ps 0x10(%r12,%rax), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%r12,%rax), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%r12,%rax), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vfmadd231ps 0x40(%r12,%rax), %xmm6, %xmm2 # xmm2 = (xmm6 * mem) + xmm2 vmovaps 0xa0(%r11,%rbx), %xmm1 vmovaps 0xb0(%r11,%rbx), %xmm3 vmovaps 0xc0(%r11,%rbx), %xmm4 vmovaps 0xd0(%r11,%rbx), %xmm5 vmovaps 0xe0(%r11,%rbx), %xmm6 vfmadd132ps (%r13,%rax), %xmm2, %xmm1 # xmm1 = (xmm1 * mem) + xmm2 vfmadd231ps 0x10(%r13,%rax), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x20(%r13,%rax), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x30(%r13,%rax), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x40(%r13,%rax), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vmovaps 0xf0(%r11,%rbx), %xmm2 vmovaps 0x100(%r11,%rbx), %xmm3 vmovaps 0x110(%r11,%rbx), %xmm4 vmovaps 0x120(%r11,%rbx), %xmm5 vfmadd132ps (%rbp,%rax), %xmm1, %xmm2 # xmm2 = (xmm2 * mem) + xmm1 vfmadd231ps 0x10(%rbp,%rax), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%rbp,%rax), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%rbp,%rax), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vmovaps 0x130(%r11,%rbx), %xmm1 vfmadd231ps 0x40(%rbp,%rax), %xmm1, %xmm2 # xmm2 = (xmm1 * mem) + xmm2 vmovaps 0x140(%r11,%rbx), %xmm1 vmovaps 0x150(%r11,%rbx), %xmm3 vmovaps 0x160(%r11,%rbx), %xmm4 vmovaps 0x170(%r11,%rbx), %xmm5 vmovaps 0x180(%r11,%rbx), %xmm6 vfmadd132ps (%rdx,%rax), %xmm2, %xmm1 # xmm1 = (xmm1 * mem) + xmm2 vfmadd231ps 0x10(%rdx,%rax), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x20(%rdx,%rax), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x30(%rdx,%rax), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x40(%rdx,%rax), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vmovaps %xmm1, (%r10,%rax) addq $0x10, %rax jmp 0x29dab6 addq %rax, %r15 addq $0x40, %r15 leaq 0x40(%r12,%rax), %r12 addq %rax, %r13 addq $0x40, %r13 addq %rax, %rbp addq $0x40, %rbp addq %rax, %rdx addq $0x40, %rdx incl %esi addq %rax, %r10 cmpl 0x8(%rsp), %esi jl 0x29dab1 movq 0x40(%rsp), %rsi incq %rsi movq 0x28(%rsp), %r13 movq 0x70(%rsp), %rdx jmp 0x29d721 cmpl $0x5, %r15d jne 0x29fa97 cmpl $0x1, 0xdc(%rdi) jne 0x29f77f cmpl $0x1, 0xe0(%rdi) jne 0x29f77f cmpl $0x1, 0xe4(%rdi) jne 0x29f77f cmpl $0x1, 0xe8(%rdi) jne 0x29f77f movl 0x2c(%r13), %esi movl 0x30(%r13), %ecx movl 0xb8(%rsp), %eax xorl %edx, %edx testl %esi, %esi cmovlel %edx, %esi movq 0x1b0(%rdi), %rdi testl %ecx, %ecx cmovlel %edx, %ecx testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x10(%rsp) movl $0x80, %r8d movq %rdi, 0x30(%rsp) cmpq 0x10(%rsp), %rdx je 0x29fdf4 testq %rdi, %rdi je 0x29dd35 movq %rdx, %rax shlq $0x5, %rax vmovups (%rdi,%rax), %ymm0 jmp 0x29dd39 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %rdx, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %rdx, %r11 imulq 0x38(%r14), %r11 movq 0xc0(%rsp), %rbx movq %rdx, 0x8(%rsp) imulq %rdx, %rbx movq 0x90(%rsp), %rax imulq %rax, %rbx addq 0x80(%rsp), %rbx movslq 0xac(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x29df95 movl %esi, %eax xorl %edi, %edi subl $0x1, %eax jb 0x29df6d vmovaps (%r10,%r11), %ymm1 vmovaps 0x20(%r10,%r11), %ymm2 vmovaps 0x40(%r10,%r11), %ymm3 vmovaps 0x60(%r10,%r11), %ymm4 vmovaps 0x80(%r10,%r11), %ymm5 vfmadd132ps (%rbx,%rdi), %ymm0, %ymm1 # ymm1 = (ymm1 * mem) + ymm0 vfmadd231ps 0x20(%rbx,%rdi), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vfmadd231ps 0x40(%rbx,%rdi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x60(%rbx,%rdi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x80(%rbx,%rdi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vmovaps 0xa0(%r10,%r11), %ymm2 vmovaps 0xc0(%r10,%r11), %ymm3 vmovaps 0xe0(%r10,%r11), %ymm4 vmovaps 0x100(%r10,%r11), %ymm5 vmovaps 0x120(%r10,%r11), %ymm6 vfmadd132ps (%r15,%rdi), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r15,%rdi), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r15,%rdi), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x60(%r15,%rdi), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps 0x80(%r15,%rdi), %ymm6, %ymm2 # ymm2 = (ymm6 * mem) + ymm2 vmovaps 0x140(%r10,%r11), %ymm1 vmovaps 0x160(%r10,%r11), %ymm3 vmovaps 0x180(%r10,%r11), %ymm4 vmovaps 0x1a0(%r10,%r11), %ymm5 vmovaps 0x1c0(%r10,%r11), %ymm6 vfmadd132ps (%r12,%rdi), %ymm2, %ymm1 # ymm1 = (ymm1 * mem) + ymm2 vfmadd231ps 0x20(%r12,%rdi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x40(%r12,%rdi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x60(%r12,%rdi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vfmadd231ps 0x80(%r12,%rdi), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps 0x1e0(%r10,%r11), %ymm2 vmovaps 0x200(%r10,%r11), %ymm3 vmovaps 0x220(%r10,%r11), %ymm4 vmovaps 0x240(%r10,%r11), %ymm5 vfmadd132ps (%r13,%rdi), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r13,%rdi), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r13,%rdi), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x60(%r13,%rdi), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vmovaps 0x260(%r10,%r11), %ymm1 vfmadd231ps 0x80(%r13,%rdi), %ymm1, %ymm2 # ymm2 = (ymm1 * mem) + ymm2 vmovaps 0x280(%r10,%r11), %ymm1 vmovaps 0x2a0(%r10,%r11), %ymm3 vmovaps 0x2c0(%r10,%r11), %ymm4 vmovaps 0x2e0(%r10,%r11), %ymm5 vmovaps 0x300(%r10,%r11), %ymm6 vfmadd132ps (%rbp,%rdi), %ymm2, %ymm1 # ymm1 = (ymm1 * mem) + ymm2 vfmadd231ps 0x20(%rbp,%rdi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x40(%rbp,%rdi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x60(%rbp,%rdi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vfmadd231ps 0x80(%rbp,%rdi), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps %ymm1, (%r9,%rdi) addq $0x20, %rdi jmp 0x29ddad addq %rdi, %rbx addq %r8, %rbx addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp incl %edx addq %rdi, %r9 jmp 0x29dda1 movq 0x8(%rsp), %rdx incq %rdx movq 0x28(%rsp), %r13 movq 0x30(%rsp), %rdi jmp 0x29dd17 cmpl $0x1, 0xdc(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x1, 0xe0(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x2, 0xe4(%rdi) movl %eax, %ecx jne 0x29e16e cmpl $0x2, 0xe8(%rdi) movl %eax, %ecx jne 0x29e16e movslq 0xac(%rsp), %rdx movl 0x2c(%r13), %eax movl 0x30(%r13), %r8d movl 0xb8(%rsp), %esi movl %edx, %ecx subl %eax, %ecx addl %ecx, %ecx movq 0x28(%r14), %r10 movq 0x1b0(%rdi), %rdi imulq 0x10(%r13), %r11 movq 0x80(%rsp), %rbx movq %rbx, 0x60(%rsp) movq 0x90(%rsp), %rbx imulq 0xc0(%rsp), %rbx movq %rbx, 0x70(%rsp) movq %rdx, 0x10(%rsp) addq %rdx, %rdx movq %rdx, 0xd0(%rsp) movslq %ecx, %r15 xorl %r12d, %r12d testl %r8d, %r8d cmovlel %r12d, %r8d testl %esi, %esi cmovlel %r12d, %esi movq %rsi, 0x8(%rsp) shlq $0x2, %r15 movq %rdi, 0x40(%rsp) cmpq 0x8(%rsp), %r12 je 0x29fdf4 testq %rdi, %rdi je 0x29e081 vmovss (%rdi,%r12,4), %xmm0 jmp 0x29e085 vxorps %xmm0, %xmm0, %xmm0 movq %r11, %rdi movq %r11, %r13 imulq %r12, %r13 addq %r9, %r13 imulq $0x24, %r12, %rbp movq 0x70(%rsp), %rbx imulq %r12, %rbx addq 0x60(%rsp), %rbx movq 0x10(%rsp), %rcx leaq (%rbx,%rcx,4), %r11 movq 0xd0(%rsp), %rcx leaq (%rbx,%rcx,4), %rsi xorl %r9d, %r9d cmpl %r8d, %r9d je 0x29e154 xorl %edx, %edx movl %eax, %ecx testl %ecx, %ecx jle 0x29e13a vmovsd (%rbx,%rdx), %xmm1 vinsertps $0x20, 0x8(%rbx,%rdx), %xmm1, %xmm1 # xmm1 = xmm1[0,1],mem[0],xmm1[3] vinsertps $0x30, (%r11,%rdx), %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],mem[0] vmovsd 0x4(%r11,%rdx), %xmm2 vinsertf128 $0x1, %xmm2, %ymm1, %ymm2 vmovsd (%rsi,%rdx), %xmm3 vinsertf128 $0x1, %xmm3, %ymm1, %ymm1 vshufpd $0x2, %ymm1, %ymm2, %ymm1 # ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[2] vmulps (%r10,%rbp), %ymm1, %ymm1 vextractf128 $0x1, %ymm1, %xmm2 vhaddps %xmm1, %xmm2, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vmovss 0x20(%r10,%rbp), %xmm2 vfmadd231ss 0x8(%rsi,%rdx), %xmm2, %xmm1 # xmm1 = (xmm2 * mem) + xmm1 vaddss %xmm0, %xmm1, %xmm1 vmovss %xmm1, (%r13) addq $0x4, %r13 decl %ecx addq $0x8, %rdx jmp 0x29e0c9 addq %r15, %rbx addq %rdx, %rbx addq %r15, %r11 addq %rdx, %r11 addq %r15, %rsi addq %rdx, %rsi incl %r9d jmp 0x29e0bc incq %r12 movq 0x28(%rsp), %r13 movq 0x30(%rsp), %r9 movq %rdi, %r11 movq 0x40(%rsp), %rdi jmp 0x29e069 cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %r8d movq 0x58(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x29e1b4 xorl %eax, %eax testb $0x3, %dil sete %al testb $0x7, %dil leal 0x1(%rax,%rax,2), %eax cmovel %ebp, %eax movl %eax, 0x10(%rsp) xorl %eax, %eax testb $0x3, %r8b sete %al testb $0x7, %r8b leal 0x1(%rax,%rax,2), %eax cmovel %ebp, %eax jmp 0x29e1bb pushq $0x1 popq %rax movl %eax, 0x10(%rsp) movl %eax, 0x8(%rsp) movq 0x40(%rsp), %r15 movq 0x88(%rsp), %rax vmovaps 0x80(%rsp), %xmm0 vmovaps %xmm0, 0x180(%rsp) movq 0x90(%rsp), %rcx movq %rcx, 0x190(%rsp) movl 0x98(%rsp), %ecx movl %ecx, 0x198(%rsp) movq 0xa0(%rsp), %rcx movq %rcx, 0x1a0(%rsp) vmovups 0xa8(%rsp), %xmm0 vmovups %xmm0, 0x1a8(%rsp) movl 0xb8(%rsp), %ecx movl %ecx, 0x1b8(%rsp) movq 0xc0(%rsp), %rcx movq %rcx, 0x1c0(%rsp) testq %rax, %rax je 0x29e244 lock incl (%rax) cmpl 0x10(%rsp), %r10d jle 0x29e298 movl %r8d, %ebx movl %edi, %ebp movq 0x58(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0x100(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x80(%rsp), %rdi leaq 0x180(%rsp), %rsi movl 0x10(%rsp), %edx vzeroupper callq 0x64e3b movl %ebp, %edi movl %ebx, %r8d movq 0x28(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm0 vmovaps %xmm0, 0x100(%rsp) movq 0x10(%rdx), %rcx movq %rcx, 0x110(%rsp) movl 0x18(%rdx), %ecx movl %ecx, 0x118(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x120(%rsp) vmovups 0x28(%rdx), %xmm0 vmovups %xmm0, 0x128(%rsp) movl 0x38(%rdx), %ecx movl %ecx, 0x138(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x140(%rsp) testq %rax, %rax je 0x29e2fc lock incl (%rax) cmpl %r15d, 0x8(%rsp) jae 0x29e391 movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd0(%r14,%rax), %eax cltd movl 0x8(%rsp), %r9d idivl %r9d movl %eax, %ecx movq 0xe0(%rsp), %rax xorl %edx, %edx divq %r15 movl %edi, %ebp movl %r8d, %ebx movl %r9d, %r8d imulq %rax, %r8 movq 0x58(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0x100(%rsp), %rdi movl 0x70(%rsp), %esi movl 0xd0(%rsp), %edx callq 0x628f2 pushq $-0x64 popq %r12 cmpq $0x0, 0x100(%rsp) je 0x29fd31 movl %ebx, %r8d movl %ebp, %edi movslq 0x138(%rsp), %rax imulq 0x140(%rsp), %rax testq %rax, %rax je 0x29fd31 xorl %r15d, %r15d xorl %r12d, %r12d xorl %ebp, %ebp movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rax cmpq %rax, %rbp jge 0x29e5ff movl %r15d, %eax cltd movl 0x10(%rsp), %esi idivl %esi movl %eax, %ecx movl %edi, %r13d movl %edi, %eax cltd idivl %esi movslq %ecx, %rdx imulq 0x1c0(%rsp), %rdx movq 0x190(%rsp), %rcx imulq %rcx, %rdx addq 0x180(%rsp), %rdx movl 0x198(%rsp), %esi movq 0x1a0(%rsp), %rdi movq %rdx, 0x228(%rsp) andq $0x0, 0x230(%rsp) movq %rcx, 0x238(%rsp) movl %esi, 0x240(%rsp) movq %rdi, 0x248(%rsp) movl %eax, 0x260(%rsp) vmovups 0x1a8(%rsp), %xmm0 movslq 0x1b4(%rsp), %rax movslq 0x1ac(%rsp), %rdx movslq 0x1b0(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x268(%rsp) vmovups %xmm0, 0x250(%rsp) movl %r12d, %eax cltd movl 0x8(%rsp), %esi idivl %esi movl %eax, %ecx movl %r8d, %ebx movl %r8d, %eax cltd idivl %esi movslq %ecx, %rdx imulq 0x140(%rsp), %rdx movq 0x110(%rsp), %rsi imulq %rsi, %rdx addq 0x100(%rsp), %rdx movl 0x118(%rsp), %edi movq 0x120(%rsp), %rcx movq %rdx, 0x1e0(%rsp) andq $0x0, 0x1e8(%rsp) movq %rsi, 0x1f0(%rsp) movl %edi, 0x1f8(%rsp) movq %rcx, 0x200(%rsp) movl %eax, 0x218(%rsp) vmovups 0x128(%rsp), %xmm0 movslq 0x134(%rsp), %rax movslq 0x12c(%rsp), %rdx movslq 0x130(%rsp), %rdi imulq %rdx, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x220(%rsp) vmovups %xmm0, 0x208(%rsp) movq 0x10(%r14), %rax movq (%rax,%rbp,8), %rdi movq 0x58(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 vmovups %ymm0, 0x2a0(%rsp) vmovups %ymm1, 0x2c0(%rsp) movq %rcx, 0x2a8(%rsp) movq (%rdi), %rax leaq 0x228(%rsp), %rsi leaq 0x1e0(%rsp), %rdx leaq 0x2a0(%rsp), %rcx vzeroupper callq *0x38(%rax) movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x29e5b4 lock decl (%rax) jne 0x29e5b4 movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi je 0x29e5ac movq (%rdi), %rax callq *0x18(%rax) jmp 0x29e5b4 movq %rsi, %rdi callq 0x5f3e0 movq 0x230(%rsp), %rax testq %rax, %rax je 0x29e5eb lock decl (%rax) jne 0x29e5eb movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi je 0x29e5e3 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29e5eb movq %rsi, %rdi callq 0x5f3e0 incq %rbp movl %ebx, %r8d addl %ebx, %r12d movl %r13d, %edi addl %r13d, %r15d jmp 0x29e399 movq 0x40(%rsp), %rdx cmpl %edx, 0x8(%rsp) jae 0x29e629 xorl %r12d, %r12d leaq 0x100(%rsp), %rdi movq 0x28(%rsp), %rsi movq 0x58(%rsp), %rcx callq 0x64e3b jmp 0x29fd31 xorl %r12d, %r12d leaq 0x100(%rsp), %rax movq 0x28(%rsp), %rcx cmpq %rcx, %rax je 0x29fd31 movq 0x108(%rsp), %rax testq %rax, %rax je 0x29e652 lock incl (%rax) movq 0x8(%rcx), %rax testq %rax, %rax je 0x29fcd9 lock decl (%rax) jne 0x29fcd9 movq 0x28(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x29fcd1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29fcd9 pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x29ed47 cmpl $0x1, 0xe0(%rdi) jne 0x29ed47 cmpl $0x2, 0xe4(%rdi) jne 0x29ed47 cmpl $0x2, 0xe8(%rdi) jne 0x29ed47 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xac(%rsp), %r8d movl 0xb8(%rsp), %edx subl %eax, %r8d shll $0x3, %r8d movq 0x1b0(%rdi), %rsi movslq %r8d, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x29fdf4 testq %rsi, %rsi je 0x29e718 movq %r8, %r9 shlq $0x4, %r9 vmovups (%rsi,%r9), %xmm0 jmp 0x29e71c vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r8, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq %r14, %r10 movq 0x28(%r14), %r14 movslq 0x54(%r10), %r15 imulq %r8, %r15 imulq 0x38(%r10), %r15 movslq 0xac(%rsp), %rbx movq 0xc0(%rsp), %r10 imulq %r8, %r10 movq 0x90(%rsp), %r11 imulq %r11, %r10 addq 0x80(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx vmovaps (%r14,%r15), %xmm1 vmovaps 0x10(%r14,%r15), %xmm2 vmovaps 0x20(%r14,%r15), %xmm3 vmovaps 0x30(%r14,%r15), %xmm4 vmovaps 0x40(%r14,%r15), %xmm5 vmovaps 0x50(%r14,%r15), %xmm6 vmovaps 0x60(%r14,%r15), %xmm7 vmovaps 0x70(%r14,%r15), %xmm8 vmovaps 0x80(%r14,%r15), %xmm9 xorl %ebp, %ebp cmpl %ecx, %ebp je 0x29ea3d xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r14d cmpl %eax, %r14d jge 0x29e9be vmovaps (%r10,%r15), %xmm12 vfmadd213ps %xmm0, %xmm1, %xmm12 # xmm12 = (xmm1 * xmm12) + xmm0 vfmadd231ps 0x10(%r10,%r15), %xmm2, %xmm12 # xmm12 = (xmm2 * mem) + xmm12 vmovaps 0x20(%r10,%r15), %xmm11 vfmadd231ps %xmm11, %xmm3, %xmm12 # xmm12 = (xmm3 * xmm11) + xmm12 vfmadd231ps (%r11,%r15), %xmm4, %xmm12 # xmm12 = (xmm4 * mem) + xmm12 vfmadd231ps 0x10(%r11,%r15), %xmm5, %xmm12 # xmm12 = (xmm5 * mem) + xmm12 vmovaps 0x20(%r11,%r15), %xmm13 vfmadd231ps %xmm13, %xmm6, %xmm12 # xmm12 = (xmm6 * xmm13) + xmm12 vfmadd231ps (%rbx,%r15), %xmm7, %xmm12 # xmm12 = (xmm7 * mem) + xmm12 vfmadd231ps 0x10(%rbx,%r15), %xmm8, %xmm12 # xmm12 = (xmm8 * mem) + xmm12 vmovaps 0x40(%r10,%r15), %xmm10 vfmadd213ps %xmm0, %xmm1, %xmm11 # xmm11 = (xmm1 * xmm11) + xmm0 vfmadd231ps 0x30(%r10,%r15), %xmm2, %xmm11 # xmm11 = (xmm2 * mem) + xmm11 vmovaps 0x40(%r11,%r15), %xmm14 vfmadd231ps %xmm10, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm10) + xmm11 vfmadd231ps %xmm13, %xmm4, %xmm11 # xmm11 = (xmm4 * xmm13) + xmm11 vmovaps 0x20(%rbx,%r15), %xmm13 vfmadd231ps %xmm13, %xmm9, %xmm12 # xmm12 = (xmm9 * xmm13) + xmm12 vfmadd231ps 0x30(%r11,%r15), %xmm5, %xmm11 # xmm11 = (xmm5 * mem) + xmm11 vfmadd231ps %xmm14, %xmm6, %xmm11 # xmm11 = (xmm6 * xmm14) + xmm11 vfmadd231ps %xmm13, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm13) + xmm11 vmovaps 0x40(%rbx,%r15), %xmm13 vfmadd231ps 0x30(%rbx,%r15), %xmm8, %xmm11 # xmm11 = (xmm8 * mem) + xmm11 vmovaps %xmm12, (%r9) vfmadd231ps %xmm13, %xmm9, %xmm11 # xmm11 = (xmm9 * xmm13) + xmm11 vmovaps 0x60(%r10,%r15), %xmm12 vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vfmadd231ps 0x50(%r10,%r15), %xmm2, %xmm10 # xmm10 = (xmm2 * mem) + xmm10 vfmadd231ps %xmm12, %xmm3, %xmm10 # xmm10 = (xmm3 * xmm12) + xmm10 vfmadd231ps %xmm14, %xmm4, %xmm10 # xmm10 = (xmm4 * xmm14) + xmm10 vmovaps 0x60(%r11,%r15), %xmm14 vfmadd231ps 0x50(%r11,%r15), %xmm5, %xmm10 # xmm10 = (xmm5 * mem) + xmm10 vfmadd231ps %xmm14, %xmm6, %xmm10 # xmm10 = (xmm6 * xmm14) + xmm10 vfmadd231ps %xmm13, %xmm7, %xmm10 # xmm10 = (xmm7 * xmm13) + xmm10 vmovaps 0x60(%rbx,%r15), %xmm13 vfmadd231ps 0x50(%rbx,%r15), %xmm8, %xmm10 # xmm10 = (xmm8 * mem) + xmm10 vmovaps %xmm11, 0x10(%r9) vfmadd213ps %xmm0, %xmm1, %xmm12 # xmm12 = (xmm1 * xmm12) + xmm0 vfmadd231ps 0x70(%r10,%r15), %xmm2, %xmm12 # xmm12 = (xmm2 * mem) + xmm12 vfmadd231ps 0x80(%r10,%r15), %xmm3, %xmm12 # xmm12 = (xmm3 * mem) + xmm12 vfmadd231ps %xmm14, %xmm4, %xmm12 # xmm12 = (xmm4 * xmm14) + xmm12 vfmadd231ps 0x70(%r11,%r15), %xmm5, %xmm12 # xmm12 = (xmm5 * mem) + xmm12 vfmadd231ps 0x80(%r11,%r15), %xmm6, %xmm12 # xmm12 = (xmm6 * mem) + xmm12 vfmadd231ps %xmm13, %xmm7, %xmm12 # xmm12 = (xmm7 * xmm13) + xmm12 vfmadd231ps 0x70(%rbx,%r15), %xmm8, %xmm12 # xmm12 = (xmm8 * mem) + xmm12 vfmadd231ps %xmm13, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm13) + xmm10 vfmadd231ps 0x80(%rbx,%r15), %xmm9, %xmm12 # xmm12 = (xmm9 * mem) + xmm12 vmovaps %xmm10, 0x20(%r9) vmovaps %xmm12, 0x30(%r9) addq $0x40, %r9 addl $0x4, %r12d subq $-0x80, %r15 jmp 0x29e7c6 vmovaps (%r10,%r15), %xmm10 vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vfmadd231ps 0x10(%r10,%r15), %xmm2, %xmm10 # xmm10 = (xmm2 * mem) + xmm10 vmovaps 0x20(%r10,%r15), %xmm11 vfmadd231ps %xmm11, %xmm3, %xmm10 # xmm10 = (xmm3 * xmm11) + xmm10 vfmadd231ps (%r11,%r15), %xmm4, %xmm10 # xmm10 = (xmm4 * mem) + xmm10 vmovaps 0x20(%r11,%r15), %xmm12 vfmadd231ps 0x10(%r11,%r15), %xmm5, %xmm10 # xmm10 = (xmm5 * mem) + xmm10 vfmadd231ps %xmm12, %xmm6, %xmm10 # xmm10 = (xmm6 * xmm12) + xmm10 vfmadd231ps (%rbx,%r15), %xmm7, %xmm10 # xmm10 = (xmm7 * mem) + xmm10 vfmadd231ps 0x10(%rbx,%r15), %xmm8, %xmm10 # xmm10 = (xmm8 * mem) + xmm10 vfmadd213ps %xmm0, %xmm1, %xmm11 # xmm11 = (xmm1 * xmm11) + xmm0 vfmadd231ps 0x30(%r10,%r15), %xmm2, %xmm11 # xmm11 = (xmm2 * mem) + xmm11 vfmadd231ps 0x40(%r10,%r15), %xmm3, %xmm11 # xmm11 = (xmm3 * mem) + xmm11 vfmadd231ps %xmm12, %xmm4, %xmm11 # xmm11 = (xmm4 * xmm12) + xmm11 vfmadd231ps 0x30(%r11,%r15), %xmm5, %xmm11 # xmm11 = (xmm5 * mem) + xmm11 vfmadd231ps 0x40(%r11,%r15), %xmm6, %xmm11 # xmm11 = (xmm6 * mem) + xmm11 vmovaps 0x20(%rbx,%r15), %xmm12 vfmadd231ps %xmm12, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm12) + xmm11 vfmadd231ps 0x30(%rbx,%r15), %xmm8, %xmm11 # xmm11 = (xmm8 * mem) + xmm11 vfmadd231ps %xmm12, %xmm9, %xmm10 # xmm10 = (xmm9 * xmm12) + xmm10 vfmadd231ps 0x40(%rbx,%r15), %xmm9, %xmm11 # xmm11 = (xmm9 * mem) + xmm11 vmovaps %xmm10, (%r9) vmovaps %xmm11, 0x10(%r9) addq $0x20, %r9 addl $0x2, %r12d addq $0x40, %r15 leal 0x1(%r12), %r14d cmpl %eax, %r14d jl 0x29e91e jmp 0x29ea1f vmovaps (%r10,%r15), %xmm10 vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vfmadd231ps 0x10(%r10,%r15), %xmm2, %xmm10 # xmm10 = (xmm2 * mem) + xmm10 vfmadd231ps 0x20(%r10,%r15), %xmm3, %xmm10 # xmm10 = (xmm3 * mem) + xmm10 vfmadd231ps (%r11,%r15), %xmm4, %xmm10 # xmm10 = (xmm4 * mem) + xmm10 vfmadd231ps 0x10(%r11,%r15), %xmm5, %xmm10 # xmm10 = (xmm5 * mem) + xmm10 vfmadd231ps 0x20(%r11,%r15), %xmm6, %xmm10 # xmm10 = (xmm6 * mem) + xmm10 vfmadd231ps (%rbx,%r15), %xmm7, %xmm10 # xmm10 = (xmm7 * mem) + xmm10 vfmadd231ps 0x10(%rbx,%r15), %xmm8, %xmm10 # xmm10 = (xmm8 * mem) + xmm10 vfmadd231ps 0x20(%rbx,%r15), %xmm9, %xmm10 # xmm10 = (xmm9 * mem) + xmm10 vmovaps %xmm10, (%r9) addq $0x10, %r9 incl %r12d addq $0x20, %r15 cmpl %eax, %r12d jl 0x29e9ce addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x29e7b8 incq %r8 movq 0x8(%rsp), %r14 jmp 0x29e6fb pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x29ed47 cmpl $0x1, 0xe0(%rdi) jne 0x29ed47 cmpl $0x2, 0xe4(%rdi) jne 0x29ed47 cmpl $0x2, 0xe8(%rdi) jne 0x29ed47 movl 0x2c(%r13), %edx movl 0x30(%r13), %ecx movl 0xac(%rsp), %eax movl 0xb8(%rsp), %esi subl %edx, %eax shll $0x3, %eax movq 0x1b0(%rdi), %r9 xorl %r15d, %r15d testl %edx, %edx cmovlel %r15d, %edx movl %edx, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %esi, %esi cmovlel %r15d, %esi movq %rsi, 0x10(%rsp) shlq $0x2, %r8 movq %r9, 0x30(%rsp) cmpq 0x10(%rsp), %r15 je 0x29fdf4 testq %r9, %r9 je 0x29eaeb movq %r15, %rax shlq $0x4, %rax vmovups (%r9,%rax), %xmm0 jmp 0x29eaef vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r15, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %r15, %r11 imulq 0x38(%r14), %r11 movq 0xc0(%rsp), %rbx movq %r15, 0x8(%rsp) imulq %r15, %rbx movq 0x90(%rsp), %rax imulq %rax, %rbx addq 0x80(%rsp), %rbx movslq 0xac(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x29ed30 movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x29ed0b vmovaps (%r10,%r11), %xmm1 vmovaps 0x10(%r10,%r11), %xmm2 vmovaps 0x20(%r10,%r11), %xmm3 vmovaps 0x30(%r10,%r11), %xmm4 vmovaps 0x40(%r10,%r11), %xmm5 vfmadd132ps (%rbx,%rsi), %xmm0, %xmm1 # xmm1 = (xmm1 * mem) + xmm0 vfmadd231ps 0x10(%rbx,%rsi), %xmm2, %xmm1 # xmm1 = (xmm2 * mem) + xmm1 vfmadd231ps 0x20(%rbx,%rsi), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x30(%rbx,%rsi), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x40(%rbx,%rsi), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vmovaps 0x50(%r10,%r11), %xmm2 vmovaps 0x60(%r10,%r11), %xmm3 vmovaps 0x70(%r10,%r11), %xmm4 vmovaps 0x80(%r10,%r11), %xmm5 vmovaps 0x90(%r10,%r11), %xmm6 vfmadd132ps (%r15,%rsi), %xmm1, %xmm2 # xmm2 = (xmm2 * mem) + xmm1 vfmadd231ps 0x10(%r15,%rsi), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%r15,%rsi), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%r15,%rsi), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vfmadd231ps 0x40(%r15,%rsi), %xmm6, %xmm2 # xmm2 = (xmm6 * mem) + xmm2 vmovaps 0xa0(%r10,%r11), %xmm1 vmovaps 0xb0(%r10,%r11), %xmm3 vmovaps 0xc0(%r10,%r11), %xmm4 vmovaps 0xd0(%r10,%r11), %xmm5 vmovaps 0xe0(%r10,%r11), %xmm6 vfmadd132ps (%r12,%rsi), %xmm2, %xmm1 # xmm1 = (xmm1 * mem) + xmm2 vfmadd231ps 0x10(%r12,%rsi), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x20(%r12,%rsi), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x30(%r12,%rsi), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x40(%r12,%rsi), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vmovaps 0xf0(%r10,%r11), %xmm2 vmovaps 0x100(%r10,%r11), %xmm3 vmovaps 0x110(%r10,%r11), %xmm4 vmovaps 0x120(%r10,%r11), %xmm5 vmovaps 0x130(%r10,%r11), %xmm6 vfmadd132ps (%r13,%rsi), %xmm1, %xmm2 # xmm2 = (xmm2 * mem) + xmm1 vfmadd231ps 0x10(%r13,%rsi), %xmm3, %xmm2 # xmm2 = (xmm3 * mem) + xmm2 vfmadd231ps 0x20(%r13,%rsi), %xmm4, %xmm2 # xmm2 = (xmm4 * mem) + xmm2 vfmadd231ps 0x30(%r13,%rsi), %xmm5, %xmm2 # xmm2 = (xmm5 * mem) + xmm2 vfmadd231ps 0x40(%r13,%rsi), %xmm6, %xmm2 # xmm2 = (xmm6 * mem) + xmm2 vmovaps 0x140(%r10,%r11), %xmm1 vmovaps 0x150(%r10,%r11), %xmm3 vmovaps 0x160(%r10,%r11), %xmm4 vmovaps 0x170(%r10,%r11), %xmm5 vmovaps 0x180(%r10,%r11), %xmm6 vfmadd132ps (%rbp,%rsi), %xmm2, %xmm1 # xmm1 = (xmm1 * mem) + xmm2 vfmadd231ps 0x10(%rbp,%rsi), %xmm3, %xmm1 # xmm1 = (xmm3 * mem) + xmm1 vfmadd231ps 0x20(%rbp,%rsi), %xmm4, %xmm1 # xmm1 = (xmm4 * mem) + xmm1 vfmadd231ps 0x30(%rbp,%rsi), %xmm5, %xmm1 # xmm1 = (xmm5 * mem) + xmm1 vfmadd231ps 0x40(%rbp,%rsi), %xmm6, %xmm1 # xmm1 = (xmm6 * mem) + xmm1 vmovaps %xmm1, (%r9) addq $0x10, %r9 addq $0x20, %rsi jmp 0x29eb63 addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x29eb57 movq 0x8(%rsp), %r15 incq %r15 movq 0x28(%rsp), %r13 movq 0x30(%rsp), %r9 jmp 0x29eacc imull %eax, %r15d movslq %r15d, %rsi leaq 0x180(%rsp), %rdi leaq 0x100(%rsp), %rdx callq 0x73bbe movq (%r14), %rcx movq -0x18(%rcx), %rdx movl 0x10(%rsp), %r11d imull 0xe0(%r14,%rdx), %r11d movq 0x180(%rsp), %rax movl 0xdc(%r14,%rdx), %esi imull 0xd4(%r14,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%r14,%rdx), %r8d jge 0x29edd9 movslq %esi, %rsi leaq (%rax,%rsi,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%r14,%rdx), %r9d jge 0x29edce movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%r14,%rdx), %edi incq %r9 jmp 0x29edaf addl %r11d, %edi incl %r8d addq %r9, %rsi jmp 0x29ed9b leal (,%r15,4), %ecx movl %ecx, 0xe0(%rsp) movl 0x70(%rsp), %ecx shll $0x2, %ecx movslq %ecx, %rcx movq %rcx, 0x40(%rsp) xorl %ecx, %ecx testl %r15d, %r15d cmovlel %ecx, %r15d movq 0x60(%rsp), %r12 testl %r12d, %r12d cmovlel %ecx, %r12d movq %r12, 0x60(%rsp) shlq $0x2, %r15 leaq 0x15981c(%rip), %rdi # 0x3f8638 vxorps %xmm10, %xmm10, %xmm10 vbroadcastss 0x14f1ea(%rip), %xmm5 # 0x3ee014 vbroadcastss 0x14fe55(%rip), %xmm14 # 0x3eec88 xorl %edx, %edx vbroadcastss 0x15237e(%rip), %xmm12 # 0x3f11bc vbroadcastss 0x152381(%rip), %xmm6 # 0x3f11c8 vbroadcastss 0x152380(%rip), %xmm3 # 0x3f11d0 cmpq 0x60(%rsp), %rdx je 0x29fcb9 movq %rcx, 0x70(%rsp) movslq %ecx, %r9 movq 0x40(%r13), %r10 imulq %rdx, %r10 imulq 0x10(%r13), %r10 shlq $0x2, %r9 addq (%r13), %r10 movslq 0xac(%rsp), %r11 movq 0xc0(%rsp), %r13 imulq %rdx, %r13 movq 0x90(%rsp), %rcx imulq %rcx, %r13 addq 0x80(%rsp), %r13 imulq %rcx, %r11 movq %rdx, 0xd0(%rsp) shlq $0x4, %rdx movq %rdx, 0x10(%rsp) addq 0x28(%r14), %r9 xorl %ecx, %ecx cmpl 0x30(%rsp), %ecx jg 0x29f35b movq (%r14), %rsi xorl %r8d, %r8d cmpq %rbx, %r8 jg 0x29f34b movq -0x18(%rsi), %r12 cmpl $0x0, 0x100(%r14,%r12) je 0x29eef7 movq 0x1b0(%r14,%r12), %rdx movq 0x10(%rsp), %rbp vmovups (%rdx,%rbp), %xmm8 jmp 0x29eefc vxorps %xmm8, %xmm8, %xmm8 movslq 0xe8(%r14,%r12), %rdx movq %r14, %rbp movslq %ecx, %r14 imulq %rdx, %r14 imulq %r11, %r14 addq %r13, %r14 movl 0xe4(%rbp,%r12), %edx imull %r8d, %edx shll $0x2, %edx movslq %edx, %rdx leaq (%r14,%rdx,4), %rdx xorl %r14d, %r14d cmpq %r14, %r15 je 0x29ef4d movslq (%rax,%r14), %rbp shlq $0x4, %rbp vmovups (%r9,%r14,4), %xmm1 vfmadd231ps (%rdx,%rbp), %xmm1, %xmm8 # xmm8 = (xmm1 * mem) + xmm8 addq $0x4, %r14 jmp 0x29ef2e movq 0x8(%rsp), %r14 movl 0x110(%r14,%r12), %edx decl %edx cmpl $0x5, %edx ja 0x29f332 movslq (%rdi,%rdx,4), %rdx addq %rdi, %rdx vmaxps %xmm10, %xmm8, %xmm1 jmpq *%rdx movq 0x118(%r14,%r12), %rdx vminps %xmm10, %xmm8, %xmm8 vbroadcastss (%rdx), %xmm15 vfmadd231ps %xmm15, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm15) + xmm1 jmp 0x29f336 vbroadcastss 0x152220(%rip), %xmm11 # 0x3f11b8 vminps %xmm11, %xmm8, %xmm1 vmaxps %xmm1, %xmm12, %xmm1 vbroadcastss 0x152216(%rip), %xmm13 # 0x3f11c0 vmovaps %xmm13, %xmm10 vfmadd213ps %xmm5, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm5 vcvttps2dq %xmm10, %xmm15 vcvtdq2ps %xmm15, %xmm15 vcmpltps %xmm15, %xmm10, %xmm10 vandps %xmm14, %xmm10, %xmm10 vsubps %xmm10, %xmm15, %xmm10 vbroadcastss 0x1521ed(%rip), %xmm11 # 0x3f11c4 vfmsub231ps %xmm11, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm11) - xmm1 vfnmsub231ps %xmm6, %xmm10, %xmm1 # xmm1 = -(xmm10 * xmm6) - xmm1 vmulps %xmm1, %xmm1, %xmm15 vbroadcastss 0x1521de(%rip), %xmm2 # 0x3f11cc vfmadd213ps %xmm3, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm3 vbroadcastss 0x1521d8(%rip), %xmm7 # 0x3f11d4 vfmadd213ps %xmm7, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm7 vbroadcastss 0x1521ce(%rip), %xmm7 # 0x3f11d8 vfmadd213ps %xmm7, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm7 vbroadcastss 0x1521c4(%rip), %xmm4 # 0x3f11dc vfmadd213ps %xmm4, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm4 vfmadd213ps %xmm5, %xmm1, %xmm2 # xmm2 = (xmm1 * xmm2) + xmm5 vfmadd213ps %xmm1, %xmm15, %xmm2 # xmm2 = (xmm15 * xmm2) + xmm1 vaddps %xmm2, %xmm14, %xmm2 vcvttps2dq %xmm10, %xmm1 vpslld $0x17, %xmm1, %xmm1 vpaddd %xmm1, %xmm14, %xmm1 vfmadd213ps %xmm14, %xmm2, %xmm1 # xmm1 = (xmm2 * xmm1) + xmm14 vbroadcastss 0x152199(%rip), %xmm2 # 0x3f11e0 vmaxps %xmm2, %xmm1, %xmm2 vpsrld $0x17, %xmm2, %xmm10 vbroadcastss 0x15218f(%rip), %xmm4 # 0x3f11e8 vpaddd %xmm4, %xmm10, %xmm10 vbroadcastss 0x15217e(%rip), %xmm4 # 0x3f11e4 vandps %xmm4, %xmm2, %xmm2 vorps %xmm5, %xmm2, %xmm2 vcvtdq2ps %xmm10, %xmm10 vbroadcastss 0x152170(%rip), %xmm4 # 0x3f11ec vcmpltps %xmm4, %xmm2, %xmm15 vandps %xmm2, %xmm15, %xmm4 vbroadcastss 0x152162(%rip), %xmm9 # 0x3f11f0 vaddps %xmm2, %xmm9, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vandps %xmm14, %xmm15, %xmm4 vsubps %xmm4, %xmm10, %xmm4 vmulps %xmm2, %xmm2, %xmm10 vbroadcastss 0x152148(%rip), %xmm15 # 0x3f11f4 vbroadcastss 0x152143(%rip), %xmm9 # 0x3f11f8 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x152139(%rip), %xmm9 # 0x3f11fc vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x15212f(%rip), %xmm9 # 0x3f1200 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x152125(%rip), %xmm9 # 0x3f1204 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x15211b(%rip), %xmm9 # 0x3f1208 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x152111(%rip), %xmm9 # 0x3f120c vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x152107(%rip), %xmm9 # 0x3f1210 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vbroadcastss 0x1520fd(%rip), %xmm9 # 0x3f1214 vfmadd213ps %xmm9, %xmm2, %xmm15 # xmm15 = (xmm2 * xmm15) + xmm9 vmulps %xmm2, %xmm10, %xmm9 vmulps %xmm15, %xmm9, %xmm9 vfmadd231ps %xmm6, %xmm4, %xmm9 # xmm9 = (xmm4 * xmm6) + xmm9 vfmsub231ps %xmm10, %xmm5, %xmm9 # xmm9 = (xmm5 * xmm10) - xmm9 vxorps %xmm10, %xmm10, %xmm10 vcmpleps %xmm10, %xmm1, %xmm1 vsubps %xmm2, %xmm9, %xmm2 vfnmadd231ps %xmm4, %xmm11, %xmm2 # xmm2 = -(xmm11 * xmm4) + xmm2 vaddps %xmm2, %xmm2, %xmm2 vbroadcastss 0x1527b0(%rip), %xmm4 # 0x3f1900 vblendvps %xmm1, %xmm4, %xmm2, %xmm1 vbroadcastss 0x152059(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm1, %xmm1 vmaxps %xmm1, %xmm12, %xmm1 vfmadd213ps %xmm5, %xmm1, %xmm13 # xmm13 = (xmm1 * xmm13) + xmm5 vcvttps2dq %xmm13, %xmm4 vcvtdq2ps %xmm4, %xmm4 vcmpltps %xmm4, %xmm13, %xmm2 vandps %xmm2, %xmm14, %xmm2 vsubps %xmm2, %xmm4, %xmm2 vfmsub231ps %xmm11, %xmm2, %xmm1 # xmm1 = (xmm2 * xmm11) - xmm1 vfnmsub231ps %xmm6, %xmm2, %xmm1 # xmm1 = -(xmm2 * xmm6) - xmm1 vmulps %xmm1, %xmm1, %xmm4 vbroadcastss 0x152033(%rip), %xmm9 # 0x3f11cc vfmadd213ps %xmm3, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm3 vbroadcastss 0x15202d(%rip), %xmm0 # 0x3f11d4 vfmadd213ps %xmm0, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm0 vfmadd213ps %xmm7, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm7 vbroadcastss 0x152022(%rip), %xmm0 # 0x3f11dc vfmadd213ps %xmm0, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm0 vfmadd213ps %xmm5, %xmm1, %xmm9 # xmm9 = (xmm1 * xmm9) + xmm5 vfmadd213ps %xmm1, %xmm4, %xmm9 # xmm9 = (xmm4 * xmm9) + xmm1 vaddps %xmm14, %xmm9, %xmm4 vcvttps2dq %xmm2, %xmm1 vpslld $0x17, %xmm1, %xmm1 vpaddd %xmm1, %xmm14, %xmm1 vfmadd213ps %xmm14, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm1) + xmm14 vrcpps %xmm1, %xmm2 vaddps %xmm2, %xmm2, %xmm4 vbroadcastss 0x153517(%rip), %xmm0 # 0x3f2708 vfmsub213ps %xmm0, %xmm4, %xmm1 # xmm1 = (xmm4 * xmm1) - xmm0 vfnmadd213ps %xmm4, %xmm2, %xmm1 # xmm1 = -(xmm2 * xmm1) + xmm4 vfmsub213ps %xmm8, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm1) - xmm8 jmp 0x29f336 movq 0x118(%r14,%r12), %rdx vbroadcastss (%rdx), %xmm1 vbroadcastss 0x4(%rdx), %xmm15 vmaxps %xmm1, %xmm8, %xmm1 vminps %xmm1, %xmm15, %xmm1 jmp 0x29f336 vbroadcastss 0x151f86(%rip), %xmm1 # 0x3f11b4 vxorps %xmm1, %xmm8, %xmm1 vbroadcastss 0x151f7d(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm1, %xmm1 vbroadcastss 0x151f74(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm1, %xmm1 vbroadcastss 0x151f6b(%rip), %xmm8 # 0x3f11c0 vfmadd213ps %xmm5, %xmm1, %xmm8 # xmm8 = (xmm1 * xmm8) + xmm5 vcvttps2dq %xmm8, %xmm15 vcvtdq2ps %xmm15, %xmm15 vcmpltps %xmm15, %xmm8, %xmm8 vandps %xmm14, %xmm8, %xmm8 vsubps %xmm8, %xmm15, %xmm8 vbroadcastss 0x151f47(%rip), %xmm2 # 0x3f11c4 vfmsub231ps %xmm2, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm2) - xmm1 vbroadcastss 0x153491(%rip), %xmm2 # 0x3f271c vfmsub231ps %xmm2, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm2) - xmm1 vmulps %xmm1, %xmm1, %xmm15 vbroadcastss 0x151f2f(%rip), %xmm10 # 0x3f11cc vbroadcastss 0x151f2a(%rip), %xmm2 # 0x3f11d0 vfmadd213ps %xmm2, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm2 vbroadcastss 0x151f20(%rip), %xmm2 # 0x3f11d4 vfmadd213ps %xmm2, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm2 vbroadcastss 0x151f16(%rip), %xmm0 # 0x3f11d8 vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vbroadcastss 0x151f0c(%rip), %xmm0 # 0x3f11dc vfmadd213ps %xmm0, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm0 vfmadd213ps %xmm5, %xmm1, %xmm10 # xmm10 = (xmm1 * xmm10) + xmm5 vfmadd213ps %xmm1, %xmm15, %xmm10 # xmm10 = (xmm15 * xmm10) + xmm1 vaddps %xmm14, %xmm10, %xmm10 vcvttps2dq %xmm8, %xmm1 vpslld $0x17, %xmm1, %xmm1 vpaddd %xmm1, %xmm14, %xmm1 vfmadd213ps %xmm14, %xmm10, %xmm1 # xmm1 = (xmm10 * xmm1) + xmm14 vxorps %xmm10, %xmm10, %xmm10 vrcpps %xmm1, %xmm8 vfmsub213ps %xmm14, %xmm8, %xmm1 # xmm1 = (xmm8 * xmm1) - xmm14 vfnmadd132ps %xmm8, %xmm8, %xmm1 # xmm1 = -(xmm1 * xmm8) + xmm8 jmp 0x29f336 movq 0x118(%r14,%r12), %rdx vbroadcastss (%rdx), %xmm1 vbroadcastss 0x4(%rdx), %xmm2 vfmadd231ps %xmm1, %xmm8, %xmm2 # xmm2 = (xmm8 * xmm1) + xmm2 vmaxps %xmm2, %xmm10, %xmm1 vminps %xmm1, %xmm14, %xmm1 vmulps %xmm1, %xmm8, %xmm1 jmp 0x29f336 vmovaps %xmm8, %xmm1 movq %r8, %rdx shlq $0x4, %rdx vmovups %xmm1, (%r10,%rdx) incq %r8 jmp 0x29eecb movq 0x40(%rsp), %rdx leaq (%r10,%rdx,4), %r10 incl %ecx jmp 0x29eebb movq 0xd0(%rsp), %rdx incq %rdx movq 0x70(%rsp), %rcx addl 0xe0(%rsp), %ecx movq 0x28(%rsp), %r13 jmp 0x29ee50 pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x29fa97 cmpl $0x1, 0xe0(%rdi) jne 0x29fa97 cmpl $0x2, 0xe4(%rdi) jne 0x29fa97 cmpl $0x2, 0xe8(%rdi) jne 0x29fa97 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xac(%rsp), %r8d movl 0xb8(%rsp), %edx subl %eax, %r8d shll $0x4, %r8d movq 0x1b0(%rdi), %rsi movslq %r8d, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x29fdf4 testq %rsi, %rsi je 0x29f40c movq %r8, %r9 shlq $0x5, %r9 vmovups (%rsi,%r9), %ymm0 jmp 0x29f410 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r8, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq %r14, %r10 movq 0x28(%r14), %r14 movslq 0x54(%r10), %r15 imulq %r8, %r15 imulq 0x38(%r10), %r15 movslq 0xac(%rsp), %rbx movq 0xc0(%rsp), %r10 imulq %r8, %r10 movq 0x90(%rsp), %r11 imulq %r11, %r10 addq 0x80(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx vmovaps (%r14,%r15), %ymm1 vmovaps 0x20(%r14,%r15), %ymm2 vmovaps 0x40(%r14,%r15), %ymm3 vmovaps 0x60(%r14,%r15), %ymm4 vmovaps 0x80(%r14,%r15), %ymm5 vmovaps 0xa0(%r14,%r15), %ymm6 vmovaps 0xc0(%r14,%r15), %ymm7 vmovaps 0xe0(%r14,%r15), %ymm8 vmovaps 0x100(%r14,%r15), %ymm9 xorl %ebp, %ebp cmpl %ecx, %ebp je 0x29f772 xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r14d cmpl %eax, %r14d jge 0x29f6f3 vmovaps 0x40(%r10,%r15), %ymm11 vmovaps 0x80(%r10,%r15), %ymm12 vmovaps 0xc0(%r10,%r15), %ymm14 vmovaps (%r10,%r15), %ymm10 vfmadd213ps %ymm0, %ymm1, %ymm10 # ymm10 = (ymm1 * ymm10) + ymm0 vfmadd231ps 0x20(%r10,%r15), %ymm2, %ymm10 # ymm10 = (ymm2 * mem) + ymm10 vfmadd231ps %ymm11, %ymm3, %ymm10 # ymm10 = (ymm3 * ymm11) + ymm10 vfmadd213ps %ymm0, %ymm1, %ymm11 # ymm11 = (ymm1 * ymm11) + ymm0 vfmadd231ps 0x60(%r10,%r15), %ymm2, %ymm11 # ymm11 = (ymm2 * mem) + ymm11 vfmadd231ps %ymm12, %ymm3, %ymm11 # ymm11 = (ymm3 * ymm12) + ymm11 vfmadd213ps %ymm0, %ymm1, %ymm12 # ymm12 = (ymm1 * ymm12) + ymm0 vmovaps %ymm14, %ymm13 vfmadd213ps %ymm0, %ymm1, %ymm13 # ymm13 = (ymm1 * ymm13) + ymm0 vfmadd231ps 0xa0(%r10,%r15), %ymm2, %ymm12 # ymm12 = (ymm2 * mem) + ymm12 vfmadd231ps 0xe0(%r10,%r15), %ymm2, %ymm13 # ymm13 = (ymm2 * mem) + ymm13 vfmadd231ps %ymm14, %ymm3, %ymm12 # ymm12 = (ymm3 * ymm14) + ymm12 vfmadd231ps 0x100(%r10,%r15), %ymm3, %ymm13 # ymm13 = (ymm3 * mem) + ymm13 vmovaps 0x40(%r11,%r15), %ymm14 vmovaps 0x80(%r11,%r15), %ymm15 vfmadd231ps (%r11,%r15), %ymm4, %ymm10 # ymm10 = (ymm4 * mem) + ymm10 vfmadd231ps %ymm14, %ymm4, %ymm11 # ymm11 = (ymm4 * ymm14) + ymm11 vfmadd231ps 0x20(%r11,%r15), %ymm5, %ymm10 # ymm10 = (ymm5 * mem) + ymm10 vfmadd231ps %ymm14, %ymm6, %ymm10 # ymm10 = (ymm6 * ymm14) + ymm10 vmovaps 0xc0(%r11,%r15), %ymm14 vfmadd231ps %ymm15, %ymm4, %ymm12 # ymm12 = (ymm4 * ymm15) + ymm12 vfmadd231ps %ymm14, %ymm4, %ymm13 # ymm13 = (ymm4 * ymm14) + ymm13 vfmadd231ps 0x60(%r11,%r15), %ymm5, %ymm11 # ymm11 = (ymm5 * mem) + ymm11 vfmadd231ps 0xa0(%r11,%r15), %ymm5, %ymm12 # ymm12 = (ymm5 * mem) + ymm12 vfmadd231ps 0xe0(%r11,%r15), %ymm5, %ymm13 # ymm13 = (ymm5 * mem) + ymm13 vfmadd231ps %ymm15, %ymm6, %ymm11 # ymm11 = (ymm6 * ymm15) + ymm11 vfmadd231ps %ymm14, %ymm6, %ymm12 # ymm12 = (ymm6 * ymm14) + ymm12 vfmadd231ps 0x100(%r11,%r15), %ymm6, %ymm13 # ymm13 = (ymm6 * mem) + ymm13 vmovaps 0x40(%rbx,%r15), %ymm14 vfmadd231ps (%rbx,%r15), %ymm7, %ymm10 # ymm10 = (ymm7 * mem) + ymm10 vfmadd231ps %ymm14, %ymm7, %ymm11 # ymm11 = (ymm7 * ymm14) + ymm11 vfmadd231ps 0x20(%rbx,%r15), %ymm8, %ymm10 # ymm10 = (ymm8 * mem) + ymm10 vfmadd231ps %ymm14, %ymm9, %ymm10 # ymm10 = (ymm9 * ymm14) + ymm10 vmovaps 0x80(%rbx,%r15), %ymm14 vfmadd231ps %ymm14, %ymm7, %ymm12 # ymm12 = (ymm7 * ymm14) + ymm12 vfmadd231ps 0x60(%rbx,%r15), %ymm8, %ymm11 # ymm11 = (ymm8 * mem) + ymm11 vfmadd231ps %ymm14, %ymm9, %ymm11 # ymm11 = (ymm9 * ymm14) + ymm11 vmovaps 0xc0(%rbx,%r15), %ymm14 vfmadd231ps %ymm14, %ymm7, %ymm13 # ymm13 = (ymm7 * ymm14) + ymm13 vfmadd231ps 0xa0(%rbx,%r15), %ymm8, %ymm12 # ymm12 = (ymm8 * mem) + ymm12 vfmadd231ps 0xe0(%rbx,%r15), %ymm8, %ymm13 # ymm13 = (ymm8 * mem) + ymm13 vfmadd231ps 0x100(%rbx,%r15), %ymm9, %ymm13 # ymm13 = (ymm9 * mem) + ymm13 vfmadd231ps %ymm14, %ymm9, %ymm12 # ymm12 = (ymm9 * ymm14) + ymm12 vmovaps %ymm10, (%r9) vmovaps %ymm11, 0x20(%r9) vmovaps %ymm12, 0x40(%r9) vmovaps %ymm13, 0x60(%r9) subq $-0x80, %r9 addl $0x4, %r12d addq $0x100, %r15 # imm = 0x100 jmp 0x29f4c6 vmovaps (%r10,%r15), %ymm10 vfmadd213ps %ymm0, %ymm1, %ymm10 # ymm10 = (ymm1 * ymm10) + ymm0 vfmadd231ps 0x20(%r10,%r15), %ymm2, %ymm10 # ymm10 = (ymm2 * mem) + ymm10 vmovaps 0x40(%r10,%r15), %ymm11 vfmadd231ps %ymm11, %ymm3, %ymm10 # ymm10 = (ymm3 * ymm11) + ymm10 vfmadd213ps %ymm0, %ymm1, %ymm11 # ymm11 = (ymm1 * ymm11) + ymm0 vfmadd231ps 0x60(%r10,%r15), %ymm2, %ymm11 # ymm11 = (ymm2 * mem) + ymm11 vfmadd231ps 0x80(%r10,%r15), %ymm3, %ymm11 # ymm11 = (ymm3 * mem) + ymm11 vmovaps 0x40(%r11,%r15), %ymm12 vfmadd231ps (%r11,%r15), %ymm4, %ymm10 # ymm10 = (ymm4 * mem) + ymm10 vfmadd231ps %ymm12, %ymm4, %ymm11 # ymm11 = (ymm4 * ymm12) + ymm11 vfmadd231ps 0x20(%r11,%r15), %ymm5, %ymm10 # ymm10 = (ymm5 * mem) + ymm10 vfmadd231ps 0x60(%r11,%r15), %ymm5, %ymm11 # ymm11 = (ymm5 * mem) + ymm11 vfmadd231ps %ymm12, %ymm6, %ymm10 # ymm10 = (ymm6 * ymm12) + ymm10 vfmadd231ps 0x80(%r11,%r15), %ymm6, %ymm11 # ymm11 = (ymm6 * mem) + ymm11 vmovaps 0x40(%rbx,%r15), %ymm12 vfmadd231ps (%rbx,%r15), %ymm7, %ymm10 # ymm10 = (ymm7 * mem) + ymm10 vfmadd231ps %ymm12, %ymm7, %ymm11 # ymm11 = (ymm7 * ymm12) + ymm11 vfmadd231ps 0x20(%rbx,%r15), %ymm8, %ymm10 # ymm10 = (ymm8 * mem) + ymm10 vfmadd231ps 0x60(%rbx,%r15), %ymm8, %ymm11 # ymm11 = (ymm8 * mem) + ymm11 vfmadd231ps %ymm12, %ymm9, %ymm10 # ymm10 = (ymm9 * ymm12) + ymm10 vfmadd231ps 0x80(%rbx,%r15), %ymm9, %ymm11 # ymm11 = (ymm9 * mem) + ymm11 vmovaps %ymm10, (%r9) vmovaps %ymm11, 0x20(%r9) addq $0x40, %r9 addl $0x2, %r12d subq $-0x80, %r15 leal 0x1(%r12), %r14d cmpl %eax, %r14d jl 0x29f64a jmp 0x29f754 vmovaps (%r10,%r15), %ymm10 vfmadd213ps %ymm0, %ymm1, %ymm10 # ymm10 = (ymm1 * ymm10) + ymm0 vfmadd231ps 0x20(%r10,%r15), %ymm2, %ymm10 # ymm10 = (ymm2 * mem) + ymm10 vfmadd231ps 0x40(%r10,%r15), %ymm3, %ymm10 # ymm10 = (ymm3 * mem) + ymm10 vfmadd231ps (%r11,%r15), %ymm4, %ymm10 # ymm10 = (ymm4 * mem) + ymm10 vfmadd231ps 0x20(%r11,%r15), %ymm5, %ymm10 # ymm10 = (ymm5 * mem) + ymm10 vfmadd231ps 0x40(%r11,%r15), %ymm6, %ymm10 # ymm10 = (ymm6 * mem) + ymm10 vfmadd231ps (%rbx,%r15), %ymm7, %ymm10 # ymm10 = (ymm7 * mem) + ymm10 vfmadd231ps 0x20(%rbx,%r15), %ymm8, %ymm10 # ymm10 = (ymm8 * mem) + ymm10 vfmadd231ps 0x40(%rbx,%r15), %ymm9, %ymm10 # ymm10 = (ymm9 * mem) + ymm10 vmovaps %ymm10, (%r9) addq $0x20, %r9 incl %r12d addq $0x40, %r15 cmpl %eax, %r12d jl 0x29f703 addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x29f4b8 incq %r8 movq 0x8(%rsp), %r14 jmp 0x29f3ef pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x29fa97 cmpl $0x1, 0xe0(%rdi) jne 0x29fa97 cmpl $0x2, 0xe4(%rdi) jne 0x29fa97 cmpl $0x2, 0xe8(%rdi) jne 0x29fa97 movl 0x2c(%r13), %edx movl 0x30(%r13), %ecx movl 0xac(%rsp), %eax movl 0xb8(%rsp), %esi subl %edx, %eax shll $0x4, %eax movq 0x1b0(%rdi), %r9 xorl %r15d, %r15d testl %edx, %edx cmovlel %r15d, %edx movl %edx, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %esi, %esi cmovlel %r15d, %esi movq %rsi, 0x10(%rsp) shlq $0x2, %r8 movq %r9, 0x30(%rsp) cmpq 0x10(%rsp), %r15 je 0x29fdf4 testq %r9, %r9 je 0x29f820 movq %r15, %rax shlq $0x5, %rax vmovups (%r9,%rax), %ymm0 jmp 0x29f824 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r15, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %r15, %r11 imulq 0x38(%r14), %r11 movq 0xc0(%rsp), %rbx movq %r15, 0x8(%rsp) imulq %r15, %rbx movq 0x90(%rsp), %rax imulq %rax, %rbx addq 0x80(%rsp), %rbx movslq 0xac(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x29fa80 movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x29fa5b vmovaps (%r10,%r11), %ymm1 vmovaps 0x20(%r10,%r11), %ymm2 vmovaps 0x40(%r10,%r11), %ymm3 vmovaps 0x60(%r10,%r11), %ymm4 vmovaps 0x80(%r10,%r11), %ymm5 vfmadd132ps (%rbx,%rsi), %ymm0, %ymm1 # ymm1 = (ymm1 * mem) + ymm0 vfmadd231ps 0x20(%rbx,%rsi), %ymm2, %ymm1 # ymm1 = (ymm2 * mem) + ymm1 vfmadd231ps 0x40(%rbx,%rsi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x60(%rbx,%rsi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x80(%rbx,%rsi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vmovaps 0xa0(%r10,%r11), %ymm2 vmovaps 0xc0(%r10,%r11), %ymm3 vmovaps 0xe0(%r10,%r11), %ymm4 vmovaps 0x100(%r10,%r11), %ymm5 vmovaps 0x120(%r10,%r11), %ymm6 vfmadd132ps (%r15,%rsi), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r15,%rsi), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r15,%rsi), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x60(%r15,%rsi), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps 0x80(%r15,%rsi), %ymm6, %ymm2 # ymm2 = (ymm6 * mem) + ymm2 vmovaps 0x140(%r10,%r11), %ymm1 vmovaps 0x160(%r10,%r11), %ymm3 vmovaps 0x180(%r10,%r11), %ymm4 vmovaps 0x1a0(%r10,%r11), %ymm5 vmovaps 0x1c0(%r10,%r11), %ymm6 vfmadd132ps (%r12,%rsi), %ymm2, %ymm1 # ymm1 = (ymm1 * mem) + ymm2 vfmadd231ps 0x20(%r12,%rsi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x40(%r12,%rsi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x60(%r12,%rsi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vfmadd231ps 0x80(%r12,%rsi), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps 0x1e0(%r10,%r11), %ymm2 vmovaps 0x200(%r10,%r11), %ymm3 vmovaps 0x220(%r10,%r11), %ymm4 vmovaps 0x240(%r10,%r11), %ymm5 vmovaps 0x260(%r10,%r11), %ymm6 vfmadd132ps (%r13,%rsi), %ymm1, %ymm2 # ymm2 = (ymm2 * mem) + ymm1 vfmadd231ps 0x20(%r13,%rsi), %ymm3, %ymm2 # ymm2 = (ymm3 * mem) + ymm2 vfmadd231ps 0x40(%r13,%rsi), %ymm4, %ymm2 # ymm2 = (ymm4 * mem) + ymm2 vfmadd231ps 0x60(%r13,%rsi), %ymm5, %ymm2 # ymm2 = (ymm5 * mem) + ymm2 vfmadd231ps 0x80(%r13,%rsi), %ymm6, %ymm2 # ymm2 = (ymm6 * mem) + ymm2 vmovaps 0x280(%r10,%r11), %ymm1 vmovaps 0x2a0(%r10,%r11), %ymm3 vmovaps 0x2c0(%r10,%r11), %ymm4 vmovaps 0x2e0(%r10,%r11), %ymm5 vmovaps 0x300(%r10,%r11), %ymm6 vfmadd132ps (%rbp,%rsi), %ymm2, %ymm1 # ymm1 = (ymm1 * mem) + ymm2 vfmadd231ps 0x20(%rbp,%rsi), %ymm3, %ymm1 # ymm1 = (ymm3 * mem) + ymm1 vfmadd231ps 0x40(%rbp,%rsi), %ymm4, %ymm1 # ymm1 = (ymm4 * mem) + ymm1 vfmadd231ps 0x60(%rbp,%rsi), %ymm5, %ymm1 # ymm1 = (ymm5 * mem) + ymm1 vfmadd231ps 0x80(%rbp,%rsi), %ymm6, %ymm1 # ymm1 = (ymm6 * mem) + ymm1 vmovaps %ymm1, (%r9) addq $0x20, %r9 addq $0x40, %rsi jmp 0x29f898 addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x29f88c movq 0x8(%rsp), %r15 incq %r15 movq 0x28(%rsp), %r13 movq 0x30(%rsp), %r9 jmp 0x29f801 imull %eax, %r15d movslq %r15d, %rsi leaq 0x180(%rsp), %rdi leaq 0x100(%rsp), %rdx callq 0x73bbe movq (%r14), %rcx movq -0x18(%rcx), %rdx movl 0x10(%rsp), %r11d imull 0xe0(%r14,%rdx), %r11d movq 0x180(%rsp), %rax movl 0xdc(%r14,%rdx), %esi imull 0xd4(%r14,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%r14,%rdx), %esi jge 0x29fb28 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%r14,%rdx), %r9d jge 0x29fb1e movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%r14,%rdx), %edi incq %r9 jmp 0x29faff addl %r11d, %edi incl %esi addq %r9, %r8 jmp 0x29faeb leal (,%r15,8), %ecx movl %ecx, 0xd0(%rsp) movl 0x70(%rsp), %ecx shll $0x3, %ecx movslq %ecx, %rcx movq %rcx, 0x10(%rsp) xorl %ecx, %ecx testl %r15d, %r15d cmovlel %ecx, %r15d movq 0x60(%rsp), %r12 testl %r12d, %r12d cmovlel %ecx, %r12d movq %r12, 0x60(%rsp) shlq $0x2, %r15 xorl %edx, %edx cmpq 0x60(%rsp), %rdx je 0x29fc9f movq %rcx, 0x40(%rsp) movslq %ecx, %r8 movq 0x40(%r13), %r9 imulq %rdx, %r9 imulq 0x10(%r13), %r9 shlq $0x2, %r8 addq (%r13), %r9 movslq 0xac(%rsp), %r10 movq 0xc0(%rsp), %r11 imulq %rdx, %r11 movq 0x90(%rsp), %rcx imulq %rcx, %r11 addq 0x80(%rsp), %r11 imulq %rcx, %r10 movq %rdx, 0x70(%rsp) movq %rdx, %r13 shlq $0x5, %r13 addq 0x28(%r14), %r8 xorl %ebp, %ebp cmpl 0x30(%rsp), %ebp jg 0x29fc81 movq (%r14), %rcx xorl %r12d, %r12d cmpq %rbx, %r12 jg 0x29fc71 movq -0x18(%rcx), %rsi cmpl $0x0, 0x100(%r14,%rsi) je 0x29fc04 movq 0x1b0(%r14,%rsi), %rdi vmovups (%rdi,%r13), %ymm0 jmp 0x29fc08 vxorps %xmm0, %xmm0, %xmm0 movslq 0xe8(%r14,%rsi), %rdi movq %r14, %rdx movslq %ebp, %r14 imulq %rdi, %r14 imulq %r10, %r14 addq %r11, %r14 movl 0xe4(%rdx,%rsi), %esi imull %r12d, %esi shll $0x3, %esi movslq %esi, %rsi leaq (%r14,%rsi,4), %rsi xorl %edi, %edi cmpq %rdi, %r15 je 0x29fc57 movslq (%rax,%rdi), %r14 shlq $0x5, %r14 vmovups (%rsi,%r14), %ymm1 vfmadd231ps (%r8,%rdi,8), %ymm1, %ymm0 # ymm0 = (ymm1 * mem) + ymm0 addq $0x4, %rdi jmp 0x29fc38 movq %r12, %rsi shlq $0x5, %rsi vmovups %ymm0, (%r9,%rsi) incq %r12 movq 0x8(%rsp), %r14 jmp 0x29fbdc movq 0x10(%rsp), %rcx leaq (%r9,%rcx,4), %r9 incl %ebp jmp 0x29fbcc movq 0x70(%rsp), %rdx incq %rdx movq 0x40(%rsp), %rcx addl 0xd0(%rsp), %ecx movq 0x28(%rsp), %r13 jmp 0x29fb66 movq 0x8(%r14), %rdi testq %rdi, %rdi je 0x29fcb9 movq (%rdi), %rax movq %r13, %rsi movq 0x58(%rsp), %rdx vzeroupper callq *0x48(%rax) leaq 0x180(%rsp), %rdi vzeroupper callq 0x624be xorl %r12d, %r12d jmp 0x29fd9f movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x100(%rsp), %xmm0 movq 0x28(%rsp), %rcx vmovups %xmm0, (%rcx) movq 0x110(%rsp), %rax movq %rax, 0x10(%rcx) movl 0x118(%rsp), %eax movl %eax, 0x18(%rcx) movq 0x120(%rsp), %rax movq %rax, 0x20(%rcx) vmovups 0x128(%rsp), %xmm0 vmovups %xmm0, 0x28(%rcx) movl 0x138(%rsp), %eax movl %eax, 0x38(%rcx) movq 0x140(%rsp), %rax movq %rax, 0x40(%rcx) movq 0x108(%rsp), %rax testq %rax, %rax je 0x29fd68 lock decl (%rax) jne 0x29fd68 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x29fd60 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29fd68 movq %rsi, %rdi callq 0x5f3e0 movq 0x188(%rsp), %rax testq %rax, %rax je 0x29fd9f lock decl (%rax) jne 0x29fd9f movq 0x180(%rsp), %rsi movq 0x1a0(%rsp), %rdi testq %rdi, %rdi je 0x29fd97 movq (%rdi), %rax callq *0x18(%rax) jmp 0x29fd9f movq %rsi, %rdi callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax je 0x29fddc lock decl (%rax) jne 0x29fddc movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x29fdd1 movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x29fddc movq %rsi, %rdi vzeroupper callq 0x5f3e0 movl %r12d, %eax addq $0x2e8, %rsp # imm = 0x2E8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movq 0x8(%r14), %rdi testq %rdi, %rdi je 0x29fcc9 movq (%rdi), %rax movq %r13, %rsi movq 0x58(%rsp), %rdx vzeroupper callq *0x48(%rax) jmp 0x29fcc9 movq %rax, %rbx leaq 0x180(%rsp), %rdi callq 0x624be jmp 0x29ff42 jmp 0x29fe5b jmp 0x29fe5b jmp 0x29ff89 jmp 0x29ff89 movq %rax, %rbx jmp 0x29ff0b movq %rax, %rbx jmp 0x29fed4 jmp 0x29ff89 jmp 0x29fe5b jmp 0x29ff89 jmp 0x29ff89 movq %rax, %rbx jmp 0x29ff42 movq %rax, %rbx movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x29fe9d lock decl (%rax) jne 0x29fe9d movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi jne 0x29fe97 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29fe9d movq (%rdi), %rax callq *0x18(%rax) movq 0x230(%rsp), %rax testq %rax, %rax je 0x29fed4 lock decl (%rax) jne 0x29fed4 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi jne 0x29fece movq %rsi, %rdi callq 0x5f3e0 jmp 0x29fed4 movq (%rdi), %rax callq *0x18(%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x29ff0b lock decl (%rax) jne 0x29ff0b movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi jne 0x29ff05 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29ff0b movq (%rdi), %rax callq *0x18(%rax) movq 0x188(%rsp), %rax testq %rax, %rax je 0x29ff42 lock decl (%rax) jne 0x29ff42 movq 0x180(%rsp), %rsi movq 0x1a0(%rsp), %rdi testq %rdi, %rdi jne 0x29ff3c movq %rsi, %rdi callq 0x5f3e0 jmp 0x29ff42 movq (%rdi), %rax callq *0x18(%rax) movq 0x88(%rsp), %rax testq %rax, %rax je 0x29ff79 lock decl (%rax) jne 0x29ff79 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi jne 0x29ff73 movq %rsi, %rdi callq 0x5f3e0 jmp 0x29ff79 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x29ff89 jmp 0x29ff89 jmp 0x29ff89 jmp 0x29ff89 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

ncnn::ConvolutionDepthWise_x86_fma::forward_int8_x86(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_fma::forward_int8_x86(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; int elempack = bottom_blob.elempack; int elembits = bottom_blob.elembits(); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_int8 = bottom_blob; if (elembits != 8) { const int channels_g = channels * elempack / group; Mat scales(channels * elempack); { float* ps = scales; for (int g = 0; g < group; g++) { float scale = bottom_blob_int8_scales[g]; for (int q = 0; q < channels_g; q++) { *ps++ = scale; } } } Option opt_q = opt; opt_q.blob_allocator = opt.workspace_allocator; quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q); } Mat bottom_blob_bordered; make_padding(bottom_blob_int8, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; channels = bottom_blob_bordered.c; elempack = bottom_blob_bordered.elempack; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; // depth-wise if (channels * elempack == group && group == num_output) { int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { out_elempack = num_output % 8 == 0 ? 8 : 1; } #endif // __SSE2__ bool use_int8_requantize = int8_scale_term > 100; size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #if __SSE2__ if (elempack == 8) { { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128i _sum0 = _mm_setzero_si128(); __m128i _sum1 = _mm_setzero_si128(); const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { // TODO use _mm_cvtepi8_epi16 on sse4.1 __m128i _val = _mm_loadl_epi64((const __m128i*)(sptr + space_ofs[k] * 8)); _val = _mm_unpacklo_epi8(_val, _mm_cmpgt_epi8(_mm_setzero_si128(), _val)); __m128i _w = _mm_loadl_epi64((const __m128i*)(kptr + k * 8)); _w = _mm_unpacklo_epi8(_w, _mm_cmpgt_epi8(_mm_setzero_si128(), _w)); __m128i _sl = _mm_mullo_epi16(_val, _w); __m128i _sh = _mm_mulhi_epi16(_val, _w); __m128i _s0 = _mm_unpacklo_epi16(_sl, _sh); __m128i _s1 = _mm_unpackhi_epi16(_sl, _sh); _sum0 = _mm_add_epi32(_sum0, _s0); _sum1 = _mm_add_epi32(_sum1, _s1); } __m128 _scale_in0; __m128 _scale_in1; { __m128 _bottom_blob_int8_scales0 = _mm_loadu_ps((const float*)bottom_blob_int8_scales + g * 8); __m128 _bottom_blob_int8_scales1 = _mm_loadu_ps((const float*)bottom_blob_int8_scales + g * 8 + 4); __m128 _weight_data_int8_scales0 = _mm_loadu_ps((const float*)weight_data_int8_scales + g * 8); __m128 _weight_data_int8_scales1 = _mm_loadu_ps((const float*)weight_data_int8_scales + g * 8 + 4); _scale_in0 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales0, _weight_data_int8_scales0)); _scale_in1 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales1, _weight_data_int8_scales1)); __m128 _m0 = _mm_cmpneq_ps(_weight_data_int8_scales0, _mm_setzero_ps()); __m128 _m1 = _mm_cmpneq_ps(_weight_data_int8_scales1, _mm_setzero_ps()); _scale_in0 = _mm_and_ps(_scale_in0, _m0); _scale_in1 = _mm_and_ps(_scale_in1, _m1); } __m128 _sumfp32_0 = _mm_mul_ps(_mm_cvtepi32_ps(_sum0), _scale_in0); __m128 _sumfp32_1 = _mm_mul_ps(_mm_cvtepi32_ps(_sum1), _scale_in1); if (bias_term) { __m128 _bias0 = _mm_loadu_ps((const float*)bias_data + g * 8); __m128 _bias1 = _mm_loadu_ps((const float*)bias_data + g * 8 + 4); _sumfp32_0 = _mm_add_ps(_sumfp32_0, _bias0); _sumfp32_1 = _mm_add_ps(_sumfp32_1, _bias1); } _sumfp32_0 = activation_sse(_sumfp32_0, activation_type, activation_params); _sumfp32_1 = activation_sse(_sumfp32_1, activation_type, activation_params); if (use_int8_requantize) { // requantize and relu __m128 _scale_out0 = _mm_loadu_ps((const float*)top_blob_int8_scales + g * 8); __m128 _scale_out1 = _mm_loadu_ps((const float*)top_blob_int8_scales + g * 8 + 4); _sumfp32_0 = _mm_mul_ps(_sumfp32_0, _scale_out0); _sumfp32_1 = _mm_mul_ps(_sumfp32_1, _scale_out1); int64_t _sum8 = float2int8_sse(_sumfp32_0, _sumfp32_1); *(int64_t*)outptr_s8 = _sum8; outptr_s8 += 8; } else { // dequantize and relu _mm_storeu_ps(outptr_f32, _sumfp32_0); _mm_storeu_ps(outptr_f32 + 4, _sumfp32_1); outptr_f32 += 8; } } } } } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && stride_w == 1 && stride_h == 1 && dilation_w == 1 && dilation_h == 1 && (activation_type == 0 || activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s1_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s1_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2 && (activation_type == 0 || activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s2_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s2_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char*)weight_data_tm + maxk * g; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr_s8[0] = sums8; outptr_s8 += 1; } else { // dequantize outptr_f32[0] = sumfp32; outptr_f32 += 1; } } } } } } return 0; } bool use_int8_requantize = int8_scale_term > 100; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { if (use_int8_requantize) out_elempack = num_output % 8 == 0 ? 8 : 1; else out_elempack = num_output % 4 == 0 ? 4 : 1; } #endif // __SSE2__ size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { g_elempack = channels_g % 8 == 0 ? 8 : 1; if (use_int8_requantize) out_g_elempack = num_output_g % 8 == 0 ? 8 : 1; else out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x318, %rsp # imm = 0x318 movq %rcx, 0x30(%rsp) movq %rdx, 0x20(%rsp) movq %rsi, %r15 movl 0x38(%rsi), %ecx movl 0x18(%rsi), %r12d movq 0x10(%rsi), %rsi testl %r12d, %r12d je 0x29ffd3 leal (,%rsi,8), %eax cltd idivl %r12d cmpl $0x8, %eax sete %dl jmp 0x29ffd5 xorl %edx, %edx movq (%rdi), %rax movq -0x18(%rax), %r9 movq %rdi, %r8 movl 0xd4(%rdi,%r9), %ebp movl 0xd8(%rdi,%r9), %ebx decl %ebp imull 0xdc(%rdi,%r9), %ebp decl %ebx imull 0xe0(%rdi,%r9), %ebx movq 0x8(%r15), %rdi vmovups (%r15), %xmm0 vmovaps %xmm0, 0x1e0(%rsp) movq %rsi, 0x1f0(%rsp) movl %r12d, 0x1f8(%rsp) movq 0x20(%r15), %rsi movq %rsi, 0x200(%rsp) vmovdqu 0x28(%r15), %xmm0 vmovdqu %xmm0, 0x208(%rsp) movl %ecx, 0x218(%rsp) movq 0x40(%r15), %rsi movq %rsi, 0x220(%rsp) testq %rdi, %rdi je 0x2a0060 lock incl (%rdi) movq (%r8), %rax movq %r8, 0x8(%rsp) testb %dl, %dl je 0x2a0073 movq 0x8(%rsp), %rcx jmp 0x2a0186 imull %ecx, %r12d movq -0x18(%rax), %rax movq 0x8(%rsp), %r13 movl 0x108(%r13,%rax), %eax movl %eax, 0x10(%rsp) leaq 0x70(%rsp), %rdi andq $0x0, 0x40(%rdi) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdi) vmovdqu %xmm0, 0xc(%rdi) vmovdqa %xmm0, 0x20(%rdi) vmovdqu %xmm0, 0x2c(%rdi) xorl %r14d, %r14d pushq $0x4 popq %rdx movl %r12d, %esi xorl %ecx, %ecx callq 0x635fa movl %r12d, %eax cltd idivl 0x10(%rsp) movq 0x70(%rsp), %rcx movq (%r13), %rdx testl %eax, %eax cmovlel %r14d, %eax movq 0x8(%rsp), %r8 movq -0x18(%rdx), %rsi movslq 0x108(%r8,%rsi), %rdi cmpq %rdi, %r14 jge 0x2a010e movq 0x240(%r8,%rsi), %rsi vmovd (%rsi,%r14,4), %xmm0 movl %eax, %esi subl $0x1, %esi jb 0x2a0109 vmovd %xmm0, (%rcx) addq $0x4, %rcx jmp 0x2a00fa incq %r14 jmp 0x2a00d9 movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0x130(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x1e0(%rsp), %rsi leaq 0x70(%rsp), %rdx movq %r15, %rdi vzeroupper callq 0x652e3 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a017e lock decl (%rax) jne 0x2a017e movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a0176 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a017e movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx movq (%rcx), %rax leaq 0x70(%rsp), %rdx andq $0x0, 0x40(%rdx) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdx) vmovdqu %xmm0, 0xc(%rdx) vmovdqa %xmm0, 0x20(%rdx) vmovdqu %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %rcx, %rdi leaq 0x1e0(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2a1ab2 movslq 0xa8(%rsp), %r14 movq 0xb0(%rsp), %rax imulq %r14, %rax testq %rax, %rax je 0x2a1ab2 notl %ebp movl 0x88(%rsp), %r13d movl 0x9c(%rsp), %edi addl %edi, %ebp movq 0x8(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx movl %ebp, %eax movq %rdx, %rbp cltd idivl 0xe4(%rbp,%rcx) movl %eax, %esi notl %ebx addl 0xa0(%rsp), %ebx movl %ebx, %eax cltd idivl 0xe8(%rbp,%rcx) movl %eax, %edx movq %rsi, 0x10(%rsp) incl %esi movq %rdx, 0x68(%rsp) leal 0x1(%rdx), %r10d movl %r13d, %ebx imull %r14d, %ebx pushq $0x8 popq %r15 cmpl 0x108(%rbp,%rcx), %ebx jne 0x2a039a cmpl 0xd0(%rbp,%rcx), %ebx jne 0x2a039a movl %edi, 0x18(%rsp) movq 0x30(%rsp), %rdi cmpb $0x0, 0x27(%rdi) pushq $0x1 popq %rax cmovel %eax, %r15d testb $0x7, %bl cmovnel %eax, %r15d movl 0x10c(%rbp,%rcx), %eax leal (,%r15,4), %r8d movl %eax, 0x38(%rsp) cmpl $0x65, %eax cmovgel %r15d, %r8d movl %ebx, %eax cltd idivl %r15d movq 0x8(%rdi), %rcx movq %rcx, (%rsp) movq 0x20(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %eax, %ecx movl %r15d, %r9d callq 0x628f2 cmpq $0x0, (%rbx) je 0x2a1ab2 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2a1ab2 xorl %r12d, %r12d cmpl $0x1, %r13d je 0x2a0977 cmpl $0x8, %r13d jne 0x2a1ab2 movq (%rbp), %rax movq -0x18(%rax), %rax movslq 0xd4(%rbp,%rax), %rcx movslq 0xd8(%rbp,%rax), %r15 imulq %rcx, %r15 leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %r15, %rsi callq 0x73bbe movq (%rbp), %rcx movq -0x18(%rcx), %rdx movl 0x18(%rsp), %r11d imull 0xe0(%rbp,%rdx), %r11d movq 0x130(%rsp), %rax movl 0xdc(%rbp,%rdx), %esi imull 0xd4(%rbp,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%rbp,%rdx), %esi jge 0x2a0aa8 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%rbp,%rdx), %r9d jge 0x2a0390 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%rbp,%rdx), %edi incq %r9 jmp 0x2a0372 addl %r11d, %edi incl %esi addq %r9, %r8 jmp 0x2a035b movl 0xd0(%rbp,%rcx), %eax movl 0x10c(%rbp,%rcx), %r14d movq 0x30(%rsp), %rcx cmpb $0x1, 0x27(%rcx) movl %r13d, 0x40(%rsp) movl %r12d, %r13d jne 0x2a03ce cmpl $0x65, %r14d jl 0x2a03d4 testb $0x7, %al pushq $0x1 popq %r9 cmovel %r15d, %r9d jmp 0x2a03e2 pushq $0x1 popq %r9 jmp 0x2a03e2 xorl %ecx, %ecx testb $0x3, %al sete %cl leal (%rcx,%rcx,2), %r9d incl %r9d leal (,%r9,4), %r8d cmpl $0x65, %r14d cmovgel %r9d, %r8d cltd idivl %r9d movq 0x30(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq 0x20(%rsp), %r12 movq %r12, %rdi movl %esi, 0x18(%rsp) movl %r10d, 0x28(%rsp) movl %r10d, %edx movl %eax, %ecx movq %r8, 0x48(%rsp) movq %r9, 0x38(%rsp) callq 0x628f2 cmpq $0x0, (%r12) je 0x2a04a1 movslq 0x38(%r12), %rax imulq 0x40(%r12), %rax testq %rax, %rax movl %r13d, %r12d movl 0x40(%rsp), %r8d je 0x2a1ab2 movq (%rbp), %rax movq -0x18(%rax), %rax movl 0xd0(%rbp,%rax), %ecx movl 0x108(%rbp,%rax), %esi movl %ebx, %eax cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %ebp movq 0x30(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x2a04a9 testb $0x7, %dil pushq $0x1 popq %rbx movl %ebx, %r12d cmovel %r15d, %r12d cmpl $0x65, %r14d movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d jl 0x2a04bb testb $0x7, %bpl cmovel %r15d, %ebx jmp 0x2a04c9 movl %r13d, %r12d jmp 0x2a1ab2 pushq $0x1 popq %r12 movl %r12d, %ebx movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d jmp 0x2a04c9 xorl %eax, %eax testb $0x3, %bpl sete %al leal (%rax,%rax,2), %ebx incl %ebx movq 0x78(%rsp), %rax vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0x130(%rsp) movq 0x80(%rsp), %rcx movq %rcx, 0x140(%rsp) movl 0x88(%rsp), %ecx movl %ecx, 0x148(%rsp) movq 0x90(%rsp), %rcx movq %rcx, 0x150(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0x158(%rsp) movl 0xa8(%rsp), %ecx movl %ecx, 0x168(%rsp) movq 0xb0(%rsp), %rcx movq %rcx, 0x170(%rsp) testq %rax, %rax movq 0x20(%rsp), %r14 je 0x2a0548 lock incl (%rax) cmpl %r12d, %r8d movl %edi, 0x10(%rsp) jle 0x2a059d movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0xe0(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x70(%rsp), %rdi leaq 0x130(%rsp), %rsi movl %r12d, %edx vzeroupper callq 0x64e3b movl 0x10(%rsp), %edi movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d movq 0x8(%r14), %rax vmovups (%r14), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movq 0x10(%r14), %rcx movq %rcx, 0xf0(%rsp) movl 0x18(%r14), %ecx movl %ecx, 0xf8(%rsp) movq 0x20(%r14), %rcx movq %rcx, 0x100(%rsp) vmovups 0x28(%r14), %xmm0 vmovups %xmm0, 0x108(%rsp) movl 0x38(%r14), %ecx movl %ecx, 0x118(%rsp) movq 0x40(%r14), %rcx movq %rcx, 0x120(%rsp) testq %rax, %rax je 0x2a0600 lock incl (%rax) cmpl 0x38(%rsp), %ebx jae 0x2a0680 movq 0x8(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xd0(%rcx,%rax), %eax cltd idivl %ebx movl %eax, %ecx movzbl 0x48(%rsp), %eax divb 0x38(%rsp) movzbl %al, %r8d imull %ebx, %r8d movq 0x30(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0xe0(%rsp), %rdi movl %r9d, %edx movl %ebx, %r9d callq 0x628f2 movl 0x10(%rsp), %edi pushq $-0x64 popq %r14 cmpq $0x0, 0xe0(%rsp) je 0x2a1a41 movslq 0x118(%rsp), %rax imulq 0x120(%rsp), %rax testq %rax, %rax je 0x2a1a41 xorl %r15d, %r15d xorl %r13d, %r13d xorl %r14d, %r14d movq 0x8(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rax movslq 0x108(%r8,%rax), %rax cmpq %rax, %r14 jge 0x2a08f5 movl %r15d, %eax cltd idivl %r12d movl %eax, %ecx movl %edi, %eax cltd idivl %r12d movslq %ecx, %rdx imulq 0x170(%rsp), %rdx movq 0x140(%rsp), %rcx imulq %rcx, %rdx addq 0x130(%rsp), %rdx movl 0x148(%rsp), %esi movq 0x150(%rsp), %rdi movq %rdx, 0x198(%rsp) andq $0x0, 0x1a0(%rsp) movq %rcx, 0x1a8(%rsp) movl %esi, 0x1b0(%rsp) movq %rdi, 0x1b8(%rsp) movl %eax, 0x1d0(%rsp) vmovups 0x158(%rsp), %xmm0 movslq 0x164(%rsp), %rax movslq 0x15c(%rsp), %rdx movslq 0x160(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x1d8(%rsp) vmovups %xmm0, 0x1c0(%rsp) movl %r13d, %eax cltd idivl %ebx movl %eax, %ecx movl %ebp, %eax cltd idivl %ebx movslq %ecx, %rdx imulq 0x120(%rsp), %rdx movq 0xf0(%rsp), %rsi imulq %rsi, %rdx addq 0xe0(%rsp), %rdx movl 0xf8(%rsp), %edi movq 0x100(%rsp), %rcx movq %rdx, 0x228(%rsp) andq $0x0, 0x230(%rsp) movq %rsi, 0x238(%rsp) movl %edi, 0x240(%rsp) movq %rcx, 0x248(%rsp) movl %eax, 0x260(%rsp) vmovups 0x108(%rsp), %xmm0 movslq 0x114(%rsp), %rax movslq 0x10c(%rsp), %rdx movslq 0x110(%rsp), %rdi imulq %rdx, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x268(%rsp) vmovups %xmm0, 0x250(%rsp) movq 0x10(%r8), %rax movq (%rax,%r14,8), %rdi movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 vmovups %ymm0, 0x2d0(%rsp) vmovups %ymm1, 0x2f0(%rsp) movq %rcx, 0x2d8(%rsp) movq (%rdi), %rax leaq 0x198(%rsp), %rsi leaq 0x228(%rsp), %rdx leaq 0x2d0(%rsp), %rcx vzeroupper callq *0x38(%rax) movq 0x230(%rsp), %rax testq %rax, %rax movl 0x10(%rsp), %edi je 0x2a08a8 lock decl (%rax) jne 0x2a08a8 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi je 0x2a089c movq (%rdi), %rax callq *0x18(%rax) movl 0x10(%rsp), %edi jmp 0x2a08a8 movq %rsi, %rdi callq 0x5f3e0 movl 0x10(%rsp), %edi movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x2a08e7 lock decl (%rax) jne 0x2a08e7 movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi je 0x2a08db movq (%rdi), %rax callq *0x18(%rax) movl 0x10(%rsp), %edi jmp 0x2a08e7 movq %rsi, %rdi callq 0x5f3e0 movl 0x10(%rsp), %edi incq %r14 addl %ebp, %r13d addl %edi, %r15d jmp 0x2a0689 movq 0x38(%rsp), %rdx cmpl %edx, %ebx jae 0x2a091d xorl %r14d, %r14d leaq 0xe0(%rsp), %rdi movq 0x20(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x64e3b jmp 0x2a1a41 leaq 0xe0(%rsp), %rax movq 0x20(%rsp), %rbx xorl %r14d, %r14d cmpq %rbx, %rax je 0x2a1a41 movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2a0946 lock incl (%rax) movq 0x8(%rbx), %rax testq %rax, %rax je 0x2a19ee lock decl (%rax) jne 0x2a19ee movq (%rbx), %rsi movq 0x20(%rbx), %rdi testq %rdi, %rdi je 0x2a19e6 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a19ee movq (%rbp), %rax movq -0x18(%rax), %rdx movl 0xd4(%rbp,%rdx), %ecx movl 0xd8(%rbp,%rdx), %r13d movl %ecx, %esi xorl $0x3, %esi movl %r13d, %edi xorl $0x3, %edi orl %esi, %edi movq 0x20(%rsp), %r14 jne 0x2a15a1 cmpl $0x1, 0xe4(%rbp,%rdx) jne 0x2a1496 cmpl $0x1, 0xe8(%rbp,%rdx) jne 0x2a1496 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x2a1496 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x2a1496 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x2a1496 cmpl $0x65, 0x38(%rsp) jl 0x2a1b2f vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx vmovaps %xmm0, (%rbx) andq $0x0, 0x10(%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 leaq 0x198(%rsp), %r12 movq -0x18(%rax), %r13 movslq 0x108(%rbp,%r13), %rax cmpq %rax, %r14 jge 0x2a1e6c movq 0x1f8(%rbp,%r13), %rax vmovss (%rax,%r14,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2a0a66 movq 0x240(%rbp,%r13), %rax vmulss (%rax,%r14,4), %xmm1, %xmm0 vmovss 0x14e226(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rbp,%r13), %rax vmovss (%rax,%r14,4), %xmm0 vmovss %xmm0, 0x198(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c movq %rbx, %rdi movq %r12, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2a0a1f leal (,%r15,8), %ecx movl %ecx, 0x28(%rsp) xorl %edx, %edx testl %r15d, %r15d cmovlel %edx, %r15d testl %r14d, %r14d cmovlel %edx, %r14d vpxor %xmm0, %xmm0, %xmm0 xorl %esi, %esi movq %r14, 0x18(%rsp) cmpq %r14, %rsi je 0x2a19d1 movq 0x20(%rsp), %rcx movq 0x10(%rcx), %r8 imulq %rsi, %r8 imulq 0x40(%rcx), %r8 movq %rdx, 0x30(%rsp) movslq %edx, %r9 addq (%rcx), %r8 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rsi, %r11 movq 0x80(%rsp), %rcx imulq %rcx, %r11 addq 0x70(%rsp), %r11 imulq %rcx, %r10 movq %rsi, 0x40(%rsp) leaq (,%rsi,8), %rbx addq 0x28(%rbp), %r9 xorl %ecx, %ecx movq %r8, %rdx cmpl 0x68(%rsp), %ecx jg 0x2a147b movq (%rbp), %rdi xorl %r13d, %r13d cmpl 0x10(%rsp), %r13d jg 0x2a1474 movq -0x18(%rdi), %rsi movslq 0xe8(%rbp,%rsi), %r14 movslq %ecx, %r12 imulq %r14, %r12 imulq %r10, %r12 movl 0xe4(%rbp,%rsi), %ebp imull %r13d, %ebp shll $0x3, %ebp movslq %ebp, %r14 addq %r11, %r14 addq %r12, %r14 vpxor %xmm3, %xmm3, %xmm3 xorl %r12d, %r12d vpxor %xmm4, %xmm4, %xmm4 cmpq %r12, %r15 je 0x2a0bcc movslq (%rax,%r12,4), %rbp vmovq (%r14,%rbp,8), %xmm5 vpcmpgtb %xmm5, %xmm0, %xmm6 vpunpcklbw %xmm6, %xmm5, %xmm5 # xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7] vmovq (%r9,%r12,8), %xmm6 vpcmpgtb %xmm6, %xmm0, %xmm11 vpunpcklbw %xmm11, %xmm6, %xmm6 # xmm6 = xmm6[0],xmm11[0],xmm6[1],xmm11[1],xmm6[2],xmm11[2],xmm6[3],xmm11[3],xmm6[4],xmm11[4],xmm6[5],xmm11[5],xmm6[6],xmm11[6],xmm6[7],xmm11[7] vpmullw %xmm5, %xmm6, %xmm11 vpmulhw %xmm6, %xmm5, %xmm5 vpunpcklwd %xmm5, %xmm11, %xmm6 # xmm6 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3] vpunpckhwd %xmm5, %xmm11, %xmm5 # xmm5 = xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7] vpaddd %xmm6, %xmm4, %xmm4 vpaddd %xmm5, %xmm3, %xmm3 incq %r12 jmp 0x2a0b89 movq 0x8(%rsp), %rbp movq 0x1f8(%rbp,%rsi), %r14 movq 0x240(%rbp,%rsi), %r12 vmovups (%r14,%rbx,4), %xmm5 vmovups 0x10(%r14,%rbx,4), %xmm6 vmulps (%r12,%rbx,4), %xmm5, %xmm11 vmulps 0x10(%r12,%rbx,4), %xmm6, %xmm13 vrcpps %xmm11, %xmm11 vrcpps %xmm13, %xmm13 vcmpneqps %xmm0, %xmm5, %xmm5 vandps %xmm5, %xmm11, %xmm5 vcmpneqps %xmm0, %xmm6, %xmm6 vandps %xmm6, %xmm13, %xmm6 vcvtdq2ps %xmm4, %xmm4 vmulps %xmm4, %xmm5, %xmm5 vcvtdq2ps %xmm3, %xmm3 vmulps %xmm3, %xmm6, %xmm6 cmpl $0x0, 0x100(%rbp,%rsi) je 0x2a0c46 movq 0x1b0(%rbp,%rsi), %r14 vaddps (%r14,%rbx,4), %xmm5, %xmm5 vaddps 0x10(%r14,%rbx,4), %xmm6, %xmm6 movl 0x110(%rbp,%rsi), %r14d decl %r14d cmpl $0x5, %r14d ja 0x2a13e7 leaq 0x157a06(%rip), %r12 # 0x3f8668 movslq (%r12,%r14,4), %r14 addq %r12, %r14 vmaxps %xmm0, %xmm5, %xmm4 vmaxps %xmm0, %xmm6, %xmm3 jmpq *%r14 movq 0x118(%rbp,%rsi), %r14 vminps %xmm0, %xmm5, %xmm5 vbroadcastss (%r14), %xmm11 vfmadd231ps %xmm5, %xmm11, %xmm4 # xmm4 = (xmm11 * xmm5) + xmm4 vminps %xmm0, %xmm6, %xmm5 vfmadd231ps %xmm5, %xmm11, %xmm3 # xmm3 = (xmm11 * xmm5) + xmm3 jmp 0x2a13ef vbroadcastss 0x150517(%rip), %xmm1 # 0x3f11b8 vminps %xmm1, %xmm5, %xmm3 vbroadcastss 0x15050e(%rip), %xmm1 # 0x3f11bc vmaxps %xmm1, %xmm3, %xmm3 vbroadcastss 0x150505(%rip), %xmm15 # 0x3f11c0 vmovaps %xmm15, %xmm4 vbroadcastss 0x14d34c(%rip), %xmm8 # 0x3ee014 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm11 vcvtdq2ps %xmm11, %xmm11 vcmpltps %xmm11, %xmm4, %xmm4 vbroadcastss 0x14dfa3(%rip), %xmm1 # 0x3eec88 vandps %xmm1, %xmm4, %xmm4 vsubps %xmm4, %xmm11, %xmm4 vbroadcastss 0x1504ce(%rip), %xmm7 # 0x3f11c4 vfmsub231ps %xmm7, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm7) - xmm3 vmovaps %xmm7, %xmm9 vbroadcastss 0x1504c0(%rip), %xmm7 # 0x3f11c8 vfnmsub231ps %xmm7, %xmm4, %xmm3 # xmm3 = -(xmm4 * xmm7) - xmm3 vmovaps %xmm7, %xmm2 vmulps %xmm3, %xmm3, %xmm11 vbroadcastss 0x1504ae(%rip), %xmm13 # 0x3f11cc vbroadcastss 0x1504a9(%rip), %xmm7 # 0x3f11d0 vfmadd213ps %xmm7, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm7 vbroadcastss 0x15049f(%rip), %xmm12 # 0x3f11d4 vfmadd213ps %xmm12, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm12 vbroadcastss 0x150495(%rip), %xmm10 # 0x3f11d8 vfmadd213ps %xmm10, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm10 vbroadcastss 0x15048b(%rip), %xmm7 # 0x3f11dc vfmadd213ps %xmm7, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm7 vfmadd213ps %xmm8, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm8 vfmadd213ps %xmm3, %xmm11, %xmm13 # xmm13 = (xmm11 * xmm13) + xmm3 vaddps %xmm1, %xmm13, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm1, %xmm4, %xmm4 vfmadd213ps %xmm1, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm1 vcmpleps %xmm0, %xmm4, %xmm3 vbroadcastss 0x15045c(%rip), %xmm7 # 0x3f11e0 vmaxps %xmm7, %xmm4, %xmm4 vpsrld $0x17, %xmm4, %xmm11 vbroadcastss 0x15044e(%rip), %xmm7 # 0x3f11e4 vandps %xmm7, %xmm4, %xmm4 vorps %xmm4, %xmm8, %xmm4 vbroadcastss 0x150441(%rip), %xmm7 # 0x3f11e8 vpaddd %xmm7, %xmm11, %xmm11 vcvtdq2ps %xmm11, %xmm11 vbroadcastss 0x150433(%rip), %xmm7 # 0x3f11ec vcmpltps %xmm7, %xmm4, %xmm13 vandps %xmm4, %xmm13, %xmm7 vbroadcastss 0x150425(%rip), %xmm12 # 0x3f11f0 vaddps %xmm4, %xmm12, %xmm4 vaddps %xmm7, %xmm4, %xmm4 vandps %xmm1, %xmm13, %xmm7 vsubps %xmm7, %xmm11, %xmm7 vmulps %xmm4, %xmm4, %xmm11 vbroadcastss 0x15040c(%rip), %xmm13 # 0x3f11f4 vbroadcastss 0x150407(%rip), %xmm12 # 0x3f11f8 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503fd(%rip), %xmm12 # 0x3f11fc vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503f3(%rip), %xmm12 # 0x3f1200 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503e9(%rip), %xmm12 # 0x3f1204 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503df(%rip), %xmm12 # 0x3f1208 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503d5(%rip), %xmm12 # 0x3f120c vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503cb(%rip), %xmm12 # 0x3f1210 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vbroadcastss 0x1503c1(%rip), %xmm12 # 0x3f1214 vfmadd213ps %xmm12, %xmm4, %xmm13 # xmm13 = (xmm4 * xmm13) + xmm12 vmulps %xmm4, %xmm11, %xmm14 vmulps %xmm13, %xmm14, %xmm13 vmovaps %xmm2, %xmm14 vfmadd231ps %xmm2, %xmm7, %xmm13 # xmm13 = (xmm7 * xmm2) + xmm13 vfmsub231ps %xmm11, %xmm8, %xmm13 # xmm13 = (xmm8 * xmm11) - xmm13 vsubps %xmm4, %xmm13, %xmm4 vfmsub231ps %xmm7, %xmm9, %xmm4 # xmm4 = (xmm9 * xmm7) - xmm4 vbroadcastss 0x151897(%rip), %xmm2 # 0x3f2718 vmulps %xmm2, %xmm4, %xmm4 vbroadcastss 0x150a72(%rip), %xmm2 # 0x3f1900 vblendvps %xmm3, %xmm2, %xmm4, %xmm3 vbroadcastss 0x15031b(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm3, %xmm3 vbroadcastss 0x150312(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm3, %xmm3 vmovaps %xmm15, %xmm4 vfmadd213ps %xmm8, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm8 vcvttps2dq %xmm4, %xmm7 vcvtdq2ps %xmm7, %xmm7 vcmpltps %xmm7, %xmm4, %xmm4 vandps %xmm1, %xmm4, %xmm4 vsubps %xmm4, %xmm7, %xmm4 vfmsub231ps %xmm9, %xmm4, %xmm3 # xmm3 = (xmm4 * xmm9) - xmm3 vfnmsub231ps %xmm14, %xmm4, %xmm3 # xmm3 = -(xmm4 * xmm14) - xmm3 vmovaps %xmm14, %xmm13 vmulps %xmm3, %xmm3, %xmm7 vbroadcastss 0x1502e4(%rip), %xmm9 # 0x3f11cc vmovaps %xmm9, %xmm11 vbroadcastss 0x1502da(%rip), %xmm14 # 0x3f11d0 vfmadd213ps %xmm14, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm14 vbroadcastss 0x1502d0(%rip), %xmm10 # 0x3f11d4 vfmadd213ps %xmm10, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm10 vbroadcastss 0x1502c6(%rip), %xmm2 # 0x3f11d8 vfmadd213ps %xmm2, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm2 vbroadcastss 0x1502bc(%rip), %xmm12 # 0x3f11dc vfmadd213ps %xmm12, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm12 vfmadd213ps %xmm8, %xmm3, %xmm11 # xmm11 = (xmm3 * xmm11) + xmm8 vfmadd213ps %xmm3, %xmm7, %xmm11 # xmm11 = (xmm7 * xmm11) + xmm3 vaddps %xmm1, %xmm11, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm1, %xmm4, %xmm4 vfmadd213ps %xmm1, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm1 vrcpps %xmm4, %xmm3 vaddps %xmm3, %xmm3, %xmm7 vbroadcastss 0x1517b2(%rip), %xmm2 # 0x3f2708 vfmsub213ps %xmm2, %xmm7, %xmm4 # xmm4 = (xmm7 * xmm4) - xmm2 vfnmadd213ps %xmm7, %xmm3, %xmm4 # xmm4 = -(xmm3 * xmm4) + xmm7 vbroadcastss 0x15024f(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm6, %xmm3 vbroadcastss 0x150246(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm3, %xmm3 vmovaps %xmm15, %xmm7 vfmadd213ps %xmm8, %xmm3, %xmm7 # xmm7 = (xmm3 * xmm7) + xmm8 vcvttps2dq %xmm7, %xmm11 vcvtdq2ps %xmm11, %xmm11 vcmpltps %xmm11, %xmm7, %xmm7 vandps %xmm1, %xmm7, %xmm7 vsubps %xmm7, %xmm11, %xmm7 vbroadcastss 0x150221(%rip), %xmm2 # 0x3f11c4 vfmsub231ps %xmm2, %xmm7, %xmm3 # xmm3 = (xmm7 * xmm2) - xmm3 vfnmsub231ps %xmm13, %xmm7, %xmm3 # xmm3 = -(xmm7 * xmm13) - xmm3 vmulps %xmm3, %xmm3, %xmm11 vmovaps %xmm9, %xmm13 vfmadd213ps %xmm14, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm14 vfmadd213ps %xmm10, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm10 vbroadcastss 0x15020f(%rip), %xmm10 # 0x3f11d8 vfmadd213ps %xmm10, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm10 vfmadd213ps %xmm12, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm12 vfmadd213ps %xmm8, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm8 vfmadd213ps %xmm3, %xmm11, %xmm13 # xmm13 = (xmm11 * xmm13) + xmm3 vaddps %xmm1, %xmm13, %xmm11 vcvttps2dq %xmm7, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm1, %xmm3, %xmm3 vfmadd213ps %xmm1, %xmm11, %xmm3 # xmm3 = (xmm11 * xmm3) + xmm1 vbroadcastss 0x1501e4(%rip), %xmm2 # 0x3f11e0 vmaxps %xmm2, %xmm3, %xmm7 vpsrld $0x17, %xmm7, %xmm11 vbroadcastss 0x1501d6(%rip), %xmm2 # 0x3f11e4 vandps %xmm2, %xmm7, %xmm7 vorps %xmm7, %xmm8, %xmm7 vbroadcastss 0x1501c9(%rip), %xmm2 # 0x3f11e8 vpaddd %xmm2, %xmm11, %xmm11 vcvtdq2ps %xmm11, %xmm11 vbroadcastss 0x1501bb(%rip), %xmm2 # 0x3f11ec vcmpltps %xmm2, %xmm7, %xmm13 vandps %xmm7, %xmm13, %xmm14 vbroadcastss 0x1501ad(%rip), %xmm2 # 0x3f11f0 vaddps %xmm2, %xmm7, %xmm7 vaddps %xmm7, %xmm14, %xmm7 vandps %xmm1, %xmm13, %xmm13 vsubps %xmm13, %xmm11, %xmm11 vmulps %xmm7, %xmm7, %xmm13 vbroadcastss 0x150193(%rip), %xmm14 # 0x3f11f4 vbroadcastss 0x15018e(%rip), %xmm2 # 0x3f11f8 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150184(%rip), %xmm2 # 0x3f11fc vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x15017a(%rip), %xmm2 # 0x3f1200 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150170(%rip), %xmm2 # 0x3f1204 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150166(%rip), %xmm2 # 0x3f1208 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x15015c(%rip), %xmm2 # 0x3f120c vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150152(%rip), %xmm2 # 0x3f1210 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vbroadcastss 0x150148(%rip), %xmm2 # 0x3f1214 vfmadd213ps %xmm2, %xmm7, %xmm14 # xmm14 = (xmm7 * xmm14) + xmm2 vmovaps %xmm15, %xmm2 vmulps %xmm7, %xmm13, %xmm15 vmulps %xmm14, %xmm15, %xmm14 vmovaps %xmm2, %xmm15 vbroadcastss 0x1500dd(%rip), %xmm9 # 0x3f11c8 vfmadd231ps %xmm9, %xmm11, %xmm14 # xmm14 = (xmm11 * xmm9) + xmm14 vfmsub231ps %xmm13, %xmm8, %xmm14 # xmm14 = (xmm8 * xmm13) - xmm14 vcmpleps %xmm0, %xmm3, %xmm3 vsubps %xmm7, %xmm14, %xmm7 vbroadcastss 0x1500bd(%rip), %xmm12 # 0x3f11c4 vfmsub231ps %xmm11, %xmm12, %xmm7 # xmm7 = (xmm12 * xmm11) - xmm7 vbroadcastss 0x151603(%rip), %xmm2 # 0x3f2718 vmulps %xmm2, %xmm7, %xmm7 vbroadcastss 0x1507de(%rip), %xmm2 # 0x3f1900 vblendvps %xmm3, %xmm2, %xmm7, %xmm3 vbroadcastss 0x150087(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm3, %xmm3 vbroadcastss 0x15007e(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm3, %xmm3 vmovaps %xmm15, %xmm7 vfmadd213ps %xmm8, %xmm3, %xmm7 # xmm7 = (xmm3 * xmm7) + xmm8 vcvttps2dq %xmm7, %xmm11 vcvtdq2ps %xmm11, %xmm11 vcmpltps %xmm11, %xmm7, %xmm7 vandps %xmm1, %xmm7, %xmm7 vsubps %xmm7, %xmm11, %xmm7 vfmsub231ps %xmm12, %xmm7, %xmm3 # xmm3 = (xmm7 * xmm12) - xmm3 vfnmsub231ps %xmm9, %xmm7, %xmm3 # xmm3 = -(xmm7 * xmm9) - xmm3 vmulps %xmm3, %xmm3, %xmm11 vbroadcastss 0x150053(%rip), %xmm13 # 0x3f11cc vbroadcastss 0x15004e(%rip), %xmm2 # 0x3f11d0 vfmadd213ps %xmm2, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm2 vbroadcastss 0x150044(%rip), %xmm2 # 0x3f11d4 vfmadd213ps %xmm2, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm2 vfmadd213ps %xmm10, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm10 vbroadcastss 0x150039(%rip), %xmm2 # 0x3f11dc vfmadd213ps %xmm2, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm2 vfmadd213ps %xmm8, %xmm3, %xmm13 # xmm13 = (xmm3 * xmm13) + xmm8 vfmadd213ps %xmm3, %xmm11, %xmm13 # xmm13 = (xmm11 * xmm13) + xmm3 vaddps %xmm1, %xmm13, %xmm11 vcvttps2dq %xmm7, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm1, %xmm3, %xmm3 vfmadd213ps %xmm1, %xmm11, %xmm3 # xmm3 = (xmm11 * xmm3) + xmm1 vrcpps %xmm3, %xmm7 vaddps %xmm7, %xmm7, %xmm11 vbroadcastss 0x15152f(%rip), %xmm1 # 0x3f2708 vfmsub213ps %xmm1, %xmm11, %xmm3 # xmm3 = (xmm11 * xmm3) - xmm1 vfnmadd213ps %xmm11, %xmm7, %xmm3 # xmm3 = -(xmm7 * xmm3) + xmm11 vfmsub213ps %xmm5, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) - xmm5 vfmsub213ps %xmm6, %xmm6, %xmm3 # xmm3 = (xmm6 * xmm3) - xmm6 jmp 0x2a13ef movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm3 vbroadcastss 0x4(%r14), %xmm11 vmaxps %xmm3, %xmm5, %xmm4 vminps %xmm4, %xmm11, %xmm4 vmaxps %xmm3, %xmm6, %xmm3 vminps %xmm3, %xmm11, %xmm3 jmp 0x2a13ef vbroadcastss 0x14ff91(%rip), %xmm1 # 0x3f11b4 vxorps %xmm1, %xmm5, %xmm3 vbroadcastss 0x14ff88(%rip), %xmm9 # 0x3f11b8 vminps %xmm3, %xmm9, %xmm3 vbroadcastss 0x14ff7f(%rip), %xmm10 # 0x3f11bc vmaxps %xmm3, %xmm10, %xmm5 vbroadcastss 0x14ff76(%rip), %xmm8 # 0x3f11c0 vmovaps %xmm8, %xmm3 vbroadcastss 0x14cdbd(%rip), %xmm1 # 0x3ee014 vfmadd213ps %xmm1, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm3) + xmm1 vcvttps2dq %xmm3, %xmm4 vcvtdq2ps %xmm4, %xmm4 vcmpltps %xmm4, %xmm3, %xmm3 vbroadcastss 0x14da16(%rip), %xmm2 # 0x3eec88 vandps %xmm2, %xmm3, %xmm3 vsubps %xmm3, %xmm4, %xmm3 vbroadcastss 0x14ff41(%rip), %xmm7 # 0x3f11c4 vfmsub231ps %xmm7, %xmm3, %xmm5 # xmm5 = (xmm3 * xmm7) - xmm5 vbroadcastss 0x14ff37(%rip), %xmm13 # 0x3f11c8 vfnmsub231ps %xmm13, %xmm3, %xmm5 # xmm5 = -(xmm3 * xmm13) - xmm5 vmulps %xmm5, %xmm5, %xmm11 vbroadcastss 0x14ff29(%rip), %xmm12 # 0x3f11cc vmovaps %xmm12, %xmm4 vbroadcastss 0x14ff20(%rip), %xmm14 # 0x3f11d0 vfmadd213ps %xmm14, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm14 vbroadcastss 0x14ff16(%rip), %xmm15 # 0x3f11d4 vfmadd213ps %xmm15, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm15 vbroadcastss 0x14ff0c(%rip), %xmm7 # 0x3f11d8 vfmadd213ps %xmm7, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm7 vbroadcastss 0x14ff02(%rip), %xmm7 # 0x3f11dc vfmadd213ps %xmm7, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm7 vfmadd213ps %xmm1, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm1 vfmadd213ps %xmm5, %xmm11, %xmm4 # xmm4 = (xmm11 * xmm4) + xmm5 vbroadcastss 0x14fec2(%rip), %xmm5 # 0x3f11b4 vxorps %xmm5, %xmm6, %xmm5 vminps %xmm5, %xmm9, %xmm5 vmaxps %xmm5, %xmm10, %xmm5 vfmadd213ps %xmm1, %xmm5, %xmm8 # xmm8 = (xmm5 * xmm8) + xmm1 vcvttps2dq %xmm8, %xmm11 vcvtdq2ps %xmm11, %xmm11 vcmpltps %xmm11, %xmm8, %xmm6 vandps %xmm2, %xmm6, %xmm6 vsubps %xmm6, %xmm11, %xmm6 vbroadcastss 0x14fea0(%rip), %xmm8 # 0x3f11c4 vfmsub231ps %xmm8, %xmm6, %xmm5 # xmm5 = (xmm6 * xmm8) - xmm5 vfnmsub231ps %xmm13, %xmm6, %xmm5 # xmm5 = -(xmm6 * xmm13) - xmm5 vmulps %xmm5, %xmm5, %xmm11 vfmadd213ps %xmm14, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm14 vfmadd213ps %xmm15, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm15 vbroadcastss 0x14fe93(%rip), %xmm8 # 0x3f11d8 vfmadd213ps %xmm8, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm8 vfmadd213ps %xmm7, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm7 vfmadd213ps %xmm1, %xmm5, %xmm12 # xmm12 = (xmm5 * xmm12) + xmm1 vfmadd213ps %xmm5, %xmm11, %xmm12 # xmm12 = (xmm11 * xmm12) + xmm5 vaddps %xmm2, %xmm4, %xmm5 vcvttps2dq %xmm3, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm2, %xmm3, %xmm4 vfmadd213ps %xmm2, %xmm5, %xmm4 # xmm4 = (xmm5 * xmm4) + xmm2 vrcpps %xmm4, %xmm3 vfmsub213ps %xmm2, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) - xmm2 vfnmadd132ps %xmm3, %xmm3, %xmm4 # xmm4 = -(xmm4 * xmm3) + xmm3 vaddps %xmm2, %xmm12, %xmm5 vcvttps2dq %xmm6, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm2, %xmm3, %xmm3 vfmadd213ps %xmm2, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm3) + xmm2 vrcpps %xmm3, %xmm5 vfmsub213ps %xmm2, %xmm5, %xmm3 # xmm3 = (xmm5 * xmm3) - xmm2 vfnmadd132ps %xmm5, %xmm5, %xmm3 # xmm3 = -(xmm3 * xmm5) + xmm5 jmp 0x2a13ef movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm3 vbroadcastss 0x4(%r14), %xmm7 vmovaps %xmm5, %xmm4 vfmadd213ps %xmm7, %xmm3, %xmm4 # xmm4 = (xmm3 * xmm4) + xmm7 vmaxps %xmm0, %xmm4, %xmm4 vbroadcastss 0x14d8bc(%rip), %xmm1 # 0x3eec88 vminps %xmm1, %xmm4, %xmm4 vmulps %xmm5, %xmm4, %xmm4 vfmadd213ps %xmm7, %xmm6, %xmm3 # xmm3 = (xmm6 * xmm3) + xmm7 vmaxps %xmm0, %xmm3, %xmm3 vminps %xmm1, %xmm3, %xmm3 vmulps %xmm6, %xmm3, %xmm3 jmp 0x2a13ef vmovaps %xmm5, %xmm4 vmovaps %xmm6, %xmm3 cmpl $0x65, 0x38(%rsp) jl 0x2a145d movq 0x288(%rbp,%rsi), %rsi vmulps (%rsi,%rbx,4), %xmm4, %xmm4 vmulps 0x10(%rsi,%rbx,4), %xmm3, %xmm3 vbroadcastss 0x14fda2(%rip), %xmm1 # 0x3f11b4 vandps %xmm1, %xmm4, %xmm5 vandps %xmm1, %xmm3, %xmm6 vbroadcastss 0x14cbf1(%rip), %xmm1 # 0x3ee014 vorps %xmm1, %xmm5, %xmm5 vorps %xmm1, %xmm6, %xmm6 vaddps %xmm5, %xmm4, %xmm4 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vcvttps2dq %xmm3, %xmm3 vpackssdw %xmm3, %xmm4, %xmm3 vpminsw 0x157109(%rip), %xmm3, %xmm3 # 0x3f8550 vpmaxsw 0x157111(%rip), %xmm3, %xmm3 # 0x3f8560 vpacksswb %xmm3, %xmm3, %xmm3 vmovq %xmm3, (%rdx) addq $0x8, %rdx jmp 0x2a146c vmovups %xmm4, (%r8) vmovups %xmm3, 0x10(%r8) addq $0x20, %r8 incl %r13d jmp 0x2a0b45 incl %ecx jmp 0x2a0b34 movq 0x40(%rsp), %rsi incq %rsi movq 0x30(%rsp), %rdx addl 0x28(%rsp), %edx movq 0x18(%rsp), %r14 jmp 0x2a0acf pushq $0x3 popq %r13 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x2a15a1 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x2a15a1 cmpl $0x2, 0xe4(%rbp,%rdx) jne 0x2a15a1 cmpl $0x2, 0xe8(%rbp,%rdx) jne 0x2a15a1 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x2a15a1 cmpl $0x65, 0x38(%rsp) jl 0x2a20a9 vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %r13 vmovaps %xmm0, (%r13) andq $0x0, 0x10(%r13) xorl %ebx, %ebx movq %rbp, %rcx leaq 0xe0(%rsp), %rbp leaq 0x198(%rsp), %r12 movq -0x18(%rax), %r14 movslq 0x108(%rcx,%r14), %rax cmpq %rax, %rbx jge 0x2a2400 movq 0x1f8(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2a155c movq 0x240(%rcx,%r14), %rax vmulss (%rax,%rbx,4), %xmm1, %xmm0 vmovss 0x14d730(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm0 vmovss %xmm0, 0x198(%rsp) movq %r13, %rdi movq %rbp, %rsi callq 0x1ea12c movq %r13, %rdi movq %r12, %rsi callq 0x1ea12c incq %rbx movq 0x8(%rsp), %rcx movq (%rcx), %rax jmp 0x2a1517 imull %ecx, %r13d movslq %r13d, %rsi leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %rsi, 0xc8(%rsp) callq 0x73bbe movq (%rbp), %r9 movq -0x18(%r9), %rax movl 0x18(%rsp), %r10d imull 0xe0(%rbp,%rax), %r10d movq 0x130(%rsp), %rbx movl 0xdc(%rbp,%rax), %ecx imull 0xd4(%rbp,%rax), %ecx subl %ecx, %r10d xorl %ecx, %ecx xorl %edx, %edx xorl %esi, %esi cmpl 0xd8(%rbp,%rax), %esi jge 0x2a1634 movslq %ecx, %rcx leaq (%rbx,%rcx,4), %r8 xorl %edi, %edi cmpl 0xd4(%rbp,%rax), %edi jge 0x2a162a movl %edx, (%r8,%rdi,4) movq -0x18(%r9), %rax addl 0xdc(%rbp,%rax), %edx incq %rdi jmp 0x2a160d addl %r10d, %edx incl %esi addq %rdi, %rcx jmp 0x2a15fb xorl %edx, %edx testl %r13d, %r13d movl $0x0, %ecx movq %rcx, 0x50(%rsp) cmovlel %edx, %r13d vbroadcastss 0x14fb64(%rip), %xmm4 # 0x3f11b4 leaq 0x156ff9(%rip), %r8 # 0x3f8650 xorl %ebp, %ebp movq %r9, 0xc0(%rsp) movq 0x8(%rsp), %rcx movslq 0x108(%rcx,%rax), %rax cmpq %rax, %rbp jge 0x2a19d1 movq 0x10(%r14), %r9 imulq %rbp, %r9 imulq 0x40(%r14), %r9 addq (%r14), %r9 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rbp, %r11 movq 0x80(%rsp), %rax imulq %rax, %r11 addq 0x70(%rsp), %r11 imulq %rax, %r10 movq 0x28(%rcx), %r15 addq 0x50(%rsp), %r15 xorl %eax, %eax movq %r9, 0x18(%rsp) movq %r10, 0x28(%rsp) movq %r11, 0x48(%rsp) movq %rbp, 0xd0(%rsp) cmpl 0x68(%rsp), %eax jg 0x2a19a6 movl %eax, 0x30(%rsp) movq 0x8(%rsp), %rax movq (%rax), %rax movq %rax, 0x40(%rsp) xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d jg 0x2a199b movq 0x40(%rsp), %rax movq -0x18(%rax), %r12 movq 0x8(%rsp), %rcx movslq 0xe8(%rcx,%r12), %rax movslq 0x30(%rsp), %rdx imulq %rax, %rdx imulq %r10, %rdx movslq 0xe4(%rcx,%r12), %rax movslq %r14d, %rcx imulq %rax, %rcx addq %r11, %rcx addq %rdx, %rcx xorl %edx, %edx xorl %eax, %eax cmpq %rdx, %r13 je 0x2a1753 movslq (%rbx,%rdx,4), %rsi movsbl (%rcx,%rsi), %esi movsbl (%r15,%rdx), %edi imull %esi, %edi addl %edi, %eax incq %rdx jmp 0x2a1737 movq 0x8(%rsp), %rdx movq 0x1f8(%rdx,%r12), %rcx vmovss (%rcx,%rbp,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2a1788 movq 0x240(%rdx,%r12), %rcx vmulss (%rcx,%rbp,4), %xmm1, %xmm0 vmovss 0x14d504(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vcvtsi2ss %eax, %xmm6, %xmm1 vmulss %xmm1, %xmm0, %xmm5 cmpl $0x0, 0x100(%rdx,%r12) je 0x2a17a8 movq 0x1b0(%rdx,%r12), %rax vaddss (%rax,%rbp,4), %xmm5, %xmm5 movl 0x110(%rdx,%r12), %eax decl %eax cmpl $0x5, %eax ja 0x2a1920 movslq (%r8,%rax,4), %rax addq %r8, %rax jmpq *%rax vmaxss 0x14c844(%rip), %xmm5, %xmm0 # 0x3ee010 jmp 0x2a1924 vmovaps %xmm5, %xmm0 movq %r9, 0x58(%rsp) vmovss %xmm5, 0xd8(%rsp) callq 0x5f410 vaddss 0x14d498(%rip), %xmm0, %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x48(%rsp), %r11 movq 0x28(%rsp), %r10 movq 0x58(%rsp), %r9 leaq 0x156e40(%rip), %r8 # 0x3f8650 vbroadcastss 0x14f99b(%rip), %xmm4 # 0x3f11b4 vmulss 0xd8(%rsp), %xmm0, %xmm0 jmp 0x2a1924 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss 0x4(%rax), %xmm1 vmaxss (%rax), %xmm5, %xmm0 vucomiss %xmm1, %xmm0 jbe 0x2a1924 vmovaps %xmm1, %xmm0 jmp 0x2a1924 vminss 0x14f960(%rip), %xmm5, %xmm0 # 0x3f11b8 vxorps %xmm4, %xmm0, %xmm1 vcmpltss 0x14f957(%rip), %xmm0, %xmm0 # 0x3f11bc vbroadcastss 0x14f94a(%rip), %xmm2 # 0x3f11b8 vblendvps %xmm0, %xmm2, %xmm1, %xmm0 movq %r9, %rbp callq 0x5f410 movq 0x48(%rsp), %r11 movq 0x28(%rsp), %r10 movq %rbp, %r9 movq 0xd0(%rsp), %rbp leaq 0x156db8(%rip), %r8 # 0x3f8650 vbroadcastss 0x14f913(%rip), %xmm4 # 0x3f11b4 vmovss 0x14d3df(%rip), %xmm1 # 0x3eec88 vaddss %xmm1, %xmm0, %xmm0 vdivss %xmm0, %xmm1, %xmm0 jmp 0x2a1924 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vxorps %xmm0, %xmm0, %xmm0 vcmpltss %xmm5, %xmm0, %xmm0 vmovss (%rax), %xmm1 vbroadcastss 0x14d3b2(%rip), %xmm2 # 0x3eec88 vblendvps %xmm0, %xmm2, %xmm1, %xmm0 vmulss %xmm5, %xmm0, %xmm0 jmp 0x2a1924 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss (%rax), %xmm1 vmovss 0x4(%rax), %xmm2 vxorps %xmm4, %xmm2, %xmm0 vdivss %xmm1, %xmm0, %xmm3 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm3, %xmm5 jb 0x2a1924 vmovss 0x14d376(%rip), %xmm0 # 0x3eec88 vdivss %xmm1, %xmm0, %xmm0 vaddss %xmm0, %xmm3, %xmm0 vucomiss %xmm0, %xmm5 jbe 0x2a1990 vmovaps %xmm5, %xmm0 cmpl $0x65, 0x38(%rsp) jl 0x2a1976 movq 0x8(%rsp), %rax movq 0x288(%rax,%r12), %rax vmulss (%rax,%rbp,4), %xmm0, %xmm0 vandps %xmm4, %xmm0, %xmm1 vbroadcastss 0x156cba(%rip), %xmm2 # 0x3f8604 vorps %xmm2, %xmm1, %xmm1 vaddss %xmm1, %xmm0, %xmm0 vroundss $0xb, %xmm0, %xmm0, %xmm0 vcvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movb %al, (%r9) incq %r9 jmp 0x2a1988 movq 0x18(%rsp), %rax vmovss %xmm0, (%rax) addq $0x4, %rax movq %rax, 0x18(%rsp) incl %r14d jmp 0x2a16f0 vfmadd213ss %xmm2, %xmm5, %xmm1 # xmm1 = (xmm5 * xmm1) + xmm2 vmulss %xmm5, %xmm1, %xmm0 jmp 0x2a1924 movl 0x30(%rsp), %eax incl %eax jmp 0x2a16d2 incq %rbp movq 0xc0(%rsp), %r9 movq -0x18(%r9), %rax movq 0x50(%rsp), %rcx addq 0xc8(%rsp), %rcx movq %rcx, 0x50(%rsp) movq 0x20(%rsp), %r14 jmp 0x2a1661 leaq 0x130(%rsp), %rdi callq 0x624be xorl %r12d, %r12d jmp 0x2a1ab2 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0xe0(%rsp), %xmm0 vmovups %xmm0, (%rbx) movq 0xf0(%rsp), %rax movq %rax, 0x10(%rbx) movl 0xf8(%rsp), %eax movl %eax, 0x18(%rbx) movq 0x100(%rsp), %rax movq %rax, 0x20(%rbx) vmovups 0x108(%rsp), %xmm0 vmovups %xmm0, 0x28(%rbx) movl 0x118(%rsp), %eax movl %eax, 0x38(%rbx) movq 0x120(%rsp), %rax movq %rax, 0x40(%rbx) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2a1a78 lock decl (%rax) jne 0x2a1a78 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x2a1a70 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a1a78 movq %rsi, %rdi callq 0x5f3e0 movq 0x138(%rsp), %rax testq %rax, %rax movl %r14d, %r12d je 0x2a1ab2 lock decl (%rax) jne 0x2a1ab2 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi je 0x2a1aaa movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a1ab2 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a1ae3 lock decl (%rax) jne 0x2a1ae3 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a1adb movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a1ae3 movq %rsi, %rdi callq 0x5f3e0 movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x2a1b1a lock decl (%rax) jne 0x2a1b1a movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi je 0x2a1b12 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a1b1a movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x318, %rsp # imm = 0x318 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovaps %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2a1ba5 movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x14d104(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2a1b4f leaq 0x2a0(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdx movq 0x20(%rsp), %rcx movslq 0x2c(%rcx), %rsi movslq 0x30(%rcx), %rax movl 0x38(%rcx), %edi movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movl 0x34(%rcx), %r8d movl %r8d, 0x58(%rsp) movq (%rcx), %r8 movq 0x10(%rcx), %r11 movq 0x40(%rcx), %r15 imulq %r11, %r15 movq %rax, %rbx movq %rsi, 0x38(%rsp) imulq %rsi, %rbx movq %r11, 0xd0(%rsp) movq %rbx, 0x50(%rsp) imulq %rbx, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0xc8(%rsp) movq 0x80(%rsp), %rsi imulq 0xb0(%rsp), %rsi movq %rsi, 0xc0(%rsp) xorl %esi, %esi testl %eax, %eax movl $0x0, %r11d cmovgl %eax, %r11d movl %r11d, 0x10(%rsp) movl 0x28(%rcx), %eax movl %eax, 0x64(%rsp) testl %edi, %edi cmovlel %esi, %edi movq %rdi, 0x18(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x190(%rsp) movq %rdx, 0x28(%rsp) leaq (%rdx,%rdx), %rax movq %rax, 0x188(%rsp) movq 0x2a0(%rsp), %rax movq %rax, 0x180(%rsp) movq %r8, 0xd8(%rsp) xorl %esi, %esi movq %r10, 0x48(%rsp) movq %r15, 0x68(%rsp) cmpq 0x18(%rsp), %rsi je 0x2a1e5f movq 0xc8(%rsp), %rax xorl %edx, %edx divq 0xd0(%rsp) cmpl $0x4, 0x64(%rsp) cmoveq 0x50(%rsp), %rax testq %r10, %r10 je 0x2a1cd7 vmovss (%r10,%rsi,4), %xmm1 jmp 0x2a1cdb vxorps %xmm1, %xmm1, %xmm1 imulq %rsi, %r15 addq 0xd8(%rsp), %r15 movq 0x180(%rsp), %rcx vmovss (%rcx,%rsi,4), %xmm0 imull 0x58(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2a1d14 vmovss %xmm1, (%r8,%rcx,4) incq %rcx jmp 0x2a1d05 leaq (%rsi,%rsi,8), %rax movq 0xc0(%rsp), %r13 movq %rsi, 0x40(%rsp) imulq %rsi, %r13 addq 0x190(%rsp), %r13 movq 0x28(%rsp), %rcx leaq (%rcx,%r13), %r11 movq 0x188(%rsp), %rcx leaq (%rcx,%r13), %rbx xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d je 0x2a1e45 xorl %ecx, %ecx movq %r15, %rdx movq 0x38(%rsp), %rsi movl %esi, %r10d testl %r10d, %r10d jle 0x2a1e25 movsbl (%r13,%rcx), %r12d movsbl (%r9,%rax), %esi imull %r12d, %esi movsbl 0x1(%r13,%rcx), %r12d movsbl 0x1(%r9,%rax), %edi imull %r12d, %edi addl %esi, %edi movsbl 0x2(%r13,%rcx), %esi movsbl 0x2(%r9,%rax), %r12d imull %esi, %r12d movsbl (%r11,%rcx), %esi movsbl 0x3(%r9,%rax), %ebp imull %esi, %ebp addl %r12d, %ebp addl %edi, %ebp movsbl 0x1(%r11,%rcx), %esi movsbl 0x4(%r9,%rax), %edi imull %esi, %edi movsbl 0x2(%r11,%rcx), %esi movsbl 0x5(%r9,%rax), %r12d imull %esi, %r12d addl %edi, %r12d movsbl (%rbx,%rcx), %esi movsbl 0x6(%r9,%rax), %edi imull %esi, %edi addl %r12d, %edi addl %ebp, %edi movsbl 0x1(%rbx,%rcx), %esi movsbl 0x7(%r9,%rax), %ebp imull %esi, %ebp movsbl 0x2(%rbx,%rcx), %esi movsbl 0x8(%r9,%rax), %r12d imull %esi, %r12d addl %ebp, %r12d addl %edi, %r12d vcvtsi2ss %r12d, %xmm2, %xmm1 vfmadd213ss (%r15,%rcx,4), %xmm0, %xmm1 # xmm1 = (xmm0 * xmm1) + mem vmovss %xmm1, (%r15,%rcx,4) addq $0x4, %rdx decl %r10d incq %rcx jmp 0x2a1d61 addq %rcx, %r13 addq $0x2, %r13 addq %rcx, %r11 addq $0x2, %r11 addq %rcx, %rbx addq $0x2, %rbx incl %r14d movq %rdx, %r15 jmp 0x2a1d49 movq 0x40(%rsp), %rsi incq %rsi movq 0x68(%rsp), %r15 addq %r15, %r8 movq 0x48(%rsp), %r10 jmp 0x2a1ca2 leaq 0x2a0(%rsp), %rdi jmp 0x2a23e4 leaq 0x2b8(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rax movq 0x20(%rsp), %r15 movl 0x2c(%r15), %ecx movl %ecx, 0x38(%rsp) movl 0x30(%r15), %ecx movl 0x38(%r15), %edx movq 0x28(%rbp), %rdi movq 0x1b0(%rbp,%r13), %rsi xorl %r9d, %r9d testl %ecx, %ecx cmovlel %r9d, %ecx movl %ecx, 0x10(%rsp) movq %rax, 0x28(%rsp) addq %rax, %rax movq %rax, 0x58(%rsp) testl %edx, %edx cmovlel %r9d, %edx movq %rdx, 0x18(%rsp) vbroadcastss 0x14f2dc(%rip), %xmm0 # 0x3f11b4 vbroadcastss 0x156723(%rip), %xmm1 # 0x3f8604 pushq $-0x7f popq %r8 pushq $0x7f popq %rbx movq %rsi, 0x48(%rsp) cmpq 0x18(%rsp), %r9 je 0x2a209c testq %rsi, %rsi je 0x2a1f05 vmovss (%rsi,%r9,4), %xmm2 jmp 0x2a1f09 vxorps %xmm2, %xmm2, %xmm2 movq 0x40(%r15), %r14 imulq %r9, %r14 imulq 0x10(%r15), %r14 addq (%r15), %r14 movq 0x2b8(%rsp), %rax vmovss (%rax,%r9,8), %xmm3 vmovss 0x4(%rax,%r9,8), %xmm4 movq 0xb0(%rsp), %r15 imulq %r9, %r15 imulq 0x80(%rsp), %r15 addq 0x70(%rsp), %r15 movq %r9, 0x40(%rsp) leaq (%r9,%r9,8), %r12 movq 0x28(%rsp), %rax leaq (%r15,%rax), %r13 movq 0x58(%rsp), %rax leaq (%r15,%rax), %rbp xorl %eax, %eax cmpl 0x10(%rsp), %eax je 0x2a2085 xorl %r10d, %r10d movl 0x38(%rsp), %ecx testl %ecx, %ecx jle 0x2a2066 movsbl (%r15,%r10), %r11d movsbl (%rdi,%r12), %r9d imull %r11d, %r9d movsbl 0x1(%r15,%r10), %r11d movsbl 0x1(%rdi,%r12), %esi imull %r11d, %esi addl %r9d, %esi movsbl 0x2(%r15,%r10), %r9d movsbl 0x2(%rdi,%r12), %r11d imull %r9d, %r11d movsbl (%r13,%r10), %r9d movsbl 0x3(%rdi,%r12), %edx imull %r9d, %edx addl %r11d, %edx addl %esi, %edx movsbl 0x1(%r13,%r10), %esi movsbl 0x4(%rdi,%r12), %r9d imull %esi, %r9d movsbl 0x2(%r13,%r10), %esi movsbl 0x5(%rdi,%r12), %r11d imull %esi, %r11d addl %r9d, %r11d movsbl (%rbp,%r10), %esi movsbl 0x6(%rdi,%r12), %r9d imull %esi, %r9d addl %r11d, %r9d addl %edx, %r9d movsbl 0x1(%rbp,%r10), %edx movsbl 0x7(%rdi,%r12), %esi imull %edx, %esi movsbl 0x2(%rbp,%r10), %edx movsbl 0x8(%rdi,%r12), %r11d imull %edx, %r11d addl %esi, %r11d addl %r9d, %r11d vcvtsi2ss %r11d, %xmm7, %xmm5 vfmadd213ss %xmm2, %xmm3, %xmm5 # xmm5 = (xmm3 * xmm5) + xmm2 vmulss %xmm4, %xmm5, %xmm5 vandps %xmm0, %xmm5, %xmm6 vorps %xmm1, %xmm6, %xmm6 vaddss %xmm6, %xmm5, %xmm5 vroundss $0xb, %xmm5, %xmm5, %xmm5 vcvttss2si %xmm5, %r11d cmpl $-0x7e, %r11d jge 0x2a204f movl %r8d, %r11d cmpl $0x7f, %r11d jl 0x2a2058 movl %ebx, %r11d movb %r11b, (%r14,%r10) decl %ecx incq %r10 jmp 0x2a1f76 addq %r10, %r15 addq $0x2, %r15 addq %r10, %r13 addq $0x2, %r13 addq %r10, %rbp addq $0x2, %rbp incl %eax addq %r10, %r14 jmp 0x2a1f65 movq 0x40(%rsp), %r9 incq %r9 movq 0x20(%rsp), %r15 movq 0x48(%rsp), %rsi jmp 0x2a1eed leaq 0x2b8(%rsp), %rdi jmp 0x2a262f vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovaps %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2a211f movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x14cb8a(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2a20c9 leaq 0x270(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdi movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movslq 0x30(%rdx), %rax movl 0x38(%rdx), %r8d movl %edi, %ecx subl %esi, %ecx addl %ecx, %ecx movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movq 0x10(%rdx), %r11 movq 0x40(%rdx), %rbx imulq %r11, %rbx movq %rax, %r14 movq %rsi, 0x38(%rsp) imulq %rsi, %r14 movq %r11, 0x58(%rsp) movq %r14, 0xd0(%rsp) imulq %r14, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0x68(%rsp) movq 0x80(%rsp), %rsi imulq 0xb0(%rsp), %rsi movq %rsi, 0x50(%rsp) xorl %r11d, %r11d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi movl %esi, 0x10(%rsp) movl 0x34(%rdx), %eax movl %eax, 0xc8(%rsp) movq (%rdx), %r12 movslq %ecx, %rsi testl %r8d, %r8d cmovlel %r11d, %r8d movq %r8, 0x18(%rsp) movl 0x28(%rdx), %eax movl %eax, 0x64(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x190(%rsp) movq %rdi, 0x28(%rsp) leaq (%rdi,%rdi), %rax movq %rax, 0x188(%rsp) movq 0x270(%rsp), %rax movq %rax, 0x180(%rsp) movq %r12, 0xc0(%rsp) xorl %edi, %edi movq %r10, 0x48(%rsp) movq %rbx, 0xd8(%rsp) cmpq 0x18(%rsp), %rdi je 0x2a23dc movq 0x68(%rsp), %rax xorl %edx, %edx divq 0x58(%rsp) cmpl $0x4, 0x64(%rsp) cmoveq 0xd0(%rsp), %rax testq %r10, %r10 je 0x2a2256 vmovss (%r10,%rdi,4), %xmm1 jmp 0x2a225a vxorps %xmm1, %xmm1, %xmm1 movq %rbx, %rdx imulq %rdi, %rdx addq 0xc0(%rsp), %rdx movq 0x180(%rsp), %rcx vmovss (%rcx,%rdi,4), %xmm0 imull 0xc8(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2a2299 vmovss %xmm1, (%r12,%rcx,4) incq %rcx jmp 0x2a228a leaq (%rdi,%rdi,8), %rax movq 0x50(%rsp), %r8 movq %rdi, 0x40(%rsp) imulq %rdi, %r8 addq 0x190(%rsp), %r8 movq 0x28(%rsp), %rcx leaq (%r8,%rcx), %r13 movq 0x188(%rsp), %rcx leaq (%r8,%rcx), %r11 xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d je 0x2a23bf xorl %ecx, %ecx movq 0x38(%rsp), %rdi movl %edi, %r10d testl %r10d, %r10d jle 0x2a23a5 movsbl (%r8,%rcx), %ebx movsbl (%r9,%rax), %r15d imull %ebx, %r15d movsbl 0x1(%r8,%rcx), %ebx movsbl 0x1(%r9,%rax), %edi imull %ebx, %edi addl %r15d, %edi movsbl 0x2(%r8,%rcx), %ebx movsbl 0x2(%r9,%rax), %r15d imull %ebx, %r15d movsbl (%r13,%rcx), %ebx movsbl 0x3(%r9,%rax), %ebp imull %ebx, %ebp addl %r15d, %ebp addl %edi, %ebp movsbl 0x1(%r13,%rcx), %edi movsbl 0x4(%r9,%rax), %ebx imull %edi, %ebx movsbl 0x2(%r13,%rcx), %edi movsbl 0x5(%r9,%rax), %r15d imull %edi, %r15d addl %ebx, %r15d movsbl (%r11,%rcx), %edi movsbl 0x6(%r9,%rax), %ebx imull %edi, %ebx addl %r15d, %ebx addl %ebp, %ebx movsbl 0x1(%r11,%rcx), %edi movsbl 0x7(%r9,%rax), %ebp imull %edi, %ebp movsbl 0x2(%r11,%rcx), %edi movsbl 0x8(%r9,%rax), %r15d imull %edi, %r15d addl %ebp, %r15d addl %ebx, %r15d vcvtsi2ss %r15d, %xmm2, %xmm1 vfmadd213ss (%rdx), %xmm0, %xmm1 # xmm1 = (xmm0 * xmm1) + mem vmovss %xmm1, (%rdx) addq $0x4, %rdx decl %r10d addq $0x2, %rcx jmp 0x2a22e0 addq %rsi, %r8 addq %rcx, %r8 addq %rsi, %r13 addq %rcx, %r13 addq %rsi, %r11 addq %rcx, %r11 incl %r14d jmp 0x2a22cb movq 0x40(%rsp), %rdi incq %rdi movq 0xd8(%rsp), %rbx addq %rbx, %r12 movq 0x48(%rsp), %r10 jmp 0x2a2224 leaq 0x270(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x20(%rsp), %r15 jmp 0x2a2641 leaq 0x288(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rsi movq 0x20(%rsp), %r15 movl 0x2c(%r15), %eax movl 0x30(%r15), %edx movl 0x38(%r15), %r8d movl %esi, %ecx movl %eax, 0x10(%rsp) subl %eax, %ecx addl %ecx, %ecx movq 0x8(%rsp), %rax movq 0x28(%rax), %rdi movq 0x1b0(%rax,%r14), %rax movq %rsi, 0x18(%rsp) addq %rsi, %rsi movq %rsi, 0x48(%rsp) movslq %ecx, %r10 xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %r8d, %r8d cmovlel %esi, %r8d movq %r8, 0x40(%rsp) vbroadcastss 0x14ed3f(%rip), %xmm0 # 0x3f11b4 vbroadcastss 0x156186(%rip), %xmm1 # 0x3f8604 movq %rax, 0x28(%rsp) cmpq 0x40(%rsp), %rsi je 0x2a2627 testq %rax, %rax je 0x2a249a vmovss (%rax,%rsi,4), %xmm2 jmp 0x2a249e vxorps %xmm2, %xmm2, %xmm2 movq 0x40(%r15), %rbx imulq %rsi, %rbx imulq 0x10(%r15), %rbx addq (%r15), %rbx movq 0x288(%rsp), %rcx vmovss (%rcx,%rsi,8), %xmm3 vmovss 0x4(%rcx,%rsi,8), %xmm4 movq 0xb0(%rsp), %r14 imulq %rsi, %r14 imulq 0x80(%rsp), %r14 addq 0x70(%rsp), %r14 movq %rsi, 0x38(%rsp) leaq (%rsi,%rsi,8), %r15 movq 0x18(%rsp), %rax leaq (%r14,%rax), %r12 movq 0x48(%rsp), %rax leaq (%r14,%rax), %r13 xorl %ebp, %ebp cmpl %edx, %ebp je 0x2a2610 xorl %r9d, %r9d movl 0x10(%rsp), %ecx testl %ecx, %ecx jle 0x2a25f7 movsbl (%r14,%r9), %esi movsbl (%rdi,%r15), %r8d imull %esi, %r8d movsbl 0x1(%r14,%r9), %esi movsbl 0x1(%rdi,%r15), %r11d imull %esi, %r11d addl %r8d, %r11d movsbl 0x2(%r14,%r9), %esi movsbl 0x2(%rdi,%r15), %r8d imull %esi, %r8d movsbl (%r12,%r9), %esi movsbl 0x3(%rdi,%r15), %eax imull %esi, %eax addl %r8d, %eax addl %r11d, %eax movsbl 0x1(%r12,%r9), %esi movsbl 0x4(%rdi,%r15), %r8d imull %esi, %r8d movsbl 0x2(%r12,%r9), %esi movsbl 0x5(%rdi,%r15), %r11d imull %esi, %r11d addl %r8d, %r11d movsbl (%r13,%r9), %esi movsbl 0x6(%rdi,%r15), %r8d imull %esi, %r8d addl %r11d, %r8d addl %eax, %r8d movsbl 0x1(%r13,%r9), %eax movsbl 0x7(%rdi,%r15), %esi imull %eax, %esi movsbl 0x2(%r13,%r9), %eax movsbl 0x8(%rdi,%r15), %r11d imull %eax, %r11d addl %esi, %r11d addl %r8d, %r11d vcvtsi2ss %r11d, %xmm7, %xmm5 vfmadd213ss %xmm2, %xmm3, %xmm5 # xmm5 = (xmm3 * xmm5) + xmm2 vmulss %xmm4, %xmm5, %xmm5 vandps %xmm0, %xmm5, %xmm6 vorps %xmm1, %xmm6, %xmm6 vaddss %xmm6, %xmm5, %xmm5 vroundss $0xb, %xmm5, %xmm5, %xmm5 vcvttss2si %xmm5, %esi cmpl $-0x7e, %esi jge 0x2a25de pushq $-0x7f popq %rsi cmpl $0x7f, %esi jl 0x2a25e6 pushq $0x7f popq %rsi movb %sil, (%rbx) incq %rbx decl %ecx addq $0x2, %r9 jmp 0x2a2507 addq %r10, %r14 addq %r9, %r14 addq %r10, %r12 addq %r9, %r12 addq %r10, %r13 addq %r9, %r13 incl %ebp jmp 0x2a24f8 movq 0x38(%rsp), %rsi incq %rsi movq 0x20(%rsp), %r15 movq 0x28(%rsp), %rax jmp 0x2a2483 leaq 0x288(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x8(%rsp), %rax xorl %r12d, %r12d movq 0x8(%rax), %rdi testq %rdi, %rdi je 0x2a1ab2 movq (%rdi), %rax movq %r15, %rsi movq 0x30(%rsp), %rdx callq *0x48(%rax) jmp 0x2a1ab2 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 jmp 0x2a2677 movq %rax, %rbx leaq 0x130(%rsp), %rdi callq 0x621c2 jmp 0x2a27f9 jmp 0x2a2712 jmp 0x2a2712 jmp 0x2a2873 jmp 0x2a2873 movq %rax, %rbx jmp 0x2a27c2 jmp 0x2a2873 jmp 0x2a2712 movq %rax, %rbx jmp 0x2a278b jmp 0x2a2873 jmp 0x2a2873 jmp 0x2a2712 jmp 0x2a2873 jmp 0x2a2873 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a282a lock decl (%rax) jne 0x2a282a movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a281a movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a282a jmp 0x2a2873 movq %rax, %rbx jmp 0x2a282a movq %rax, %rbx jmp 0x2a27f9 movq %rax, %rbx movq 0x230(%rsp), %rax testq %rax, %rax je 0x2a2754 lock decl (%rax) jne 0x2a2754 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi jne 0x2a274e movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a2754 movq (%rdi), %rax callq *0x18(%rax) movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x2a278b lock decl (%rax) jne 0x2a278b movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi jne 0x2a2785 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a278b movq (%rdi), %rax callq *0x18(%rax) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2a27c2 lock decl (%rax) jne 0x2a27c2 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi jne 0x2a27bc movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a27c2 movq (%rdi), %rax callq *0x18(%rax) movq 0x138(%rsp), %rax testq %rax, %rax je 0x2a27f9 lock decl (%rax) jne 0x2a27f9 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi jne 0x2a27f3 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a27f9 movq (%rdi), %rax callq *0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a282a lock decl (%rax) jne 0x2a282a movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2a2824 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a282a movq (%rdi), %rax callq *0x18(%rax) movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x2a2861 lock decl (%rax) jne 0x2a2861 movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi jne 0x2a285b movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a2861 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2a2873 jmp 0x2a2873 jmp 0x2a2873 jmp 0x2a2873 jmp 0x2a2873 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_fma::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_fma::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float*)bias_data) + g * 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float*)bias_data) + g * 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float*)bias_data) + g * 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x29be02

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

ncnn::ConvolutionDepthWise_x86_fma::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_fma::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w *= weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w *= bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x168, %rsp # imm = 0x168 movq %rsi, %rbp movq %rdi, %r13 movq (%rsi), %r14 leaq 0x48(%r14), %rdi movq (%rdx), %rax movq %rax, 0xb8(%rsp) movl 0x60(%r14), %ebx movl 0x74(%r14), %eax movl %eax, 0x1c(%rsp) movl 0x78(%r14), %eax movl %eax, 0x18(%rsp) imull 0x80(%r14), %ebx leaq 0x70(%rsp), %rsi andq $0x0, 0x40(%rsi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rsi) vmovups %xmm0, 0xc(%rsi) vmovaps %xmm0, 0x20(%rsi) vmovups %xmm0, 0x2c(%rsi) movq %rcx, %r15 movq %rcx, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2a2ea9 movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x2a2ea9 movslq 0x88(%rsp), %rcx movl 0x9c(%rsp), %eax imull %ecx, %eax movl %eax, 0x9c(%rsp) movq 0x80(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x80(%rsp) movl $0x1, 0x88(%rsp) andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) vmovaps %xmm0, 0x40(%rsp) vmovups %xmm0, 0x4c(%rsp) movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r13,%rax) je 0x2a29f2 movl $0x90, %edi addq (%rbp), %rdi leaq 0x20(%rsp), %rsi movq %r15, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x20(%rsp) je 0x2a2e7b movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x2a2e7b movslq 0x38(%rsp), %rcx movl 0x4c(%rsp), %eax imull %ecx, %eax movl %eax, 0x4c(%rsp) movq 0x30(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x30(%rsp) movl $0x1, 0x38(%rsp) pushq $0x2a popq %rdi callq 0x782bf movq %rax, %r12 leaq 0x8(%rsp), %rdi callq 0x71548 leaq 0x8(%rsp), %rdi xorl %esi, %esi movl %ebx, %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi movl 0x1c(%rsp), %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0xb popq %rsi movl 0x18(%rsp), %edx callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xdc(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xd popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xec(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x4 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xf popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xe popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x10 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax vmovss 0xfc(%r13,%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x12 popq %rsi callq 0x71952 movq (%r13), %rax movq -0x18(%rax), %rax movl 0x100(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x5 popq %rsi callq 0x7193a movl 0x9c(%rsp), %edx leaq 0x8(%rsp), %rdi pushq $0x6 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x108(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x7 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x10c(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x110(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax leaq (%rax,%r13), %rdx addq $0x118, %rdx # imm = 0x118 leaq 0x8(%rsp), %rdi pushq $0xa popq %rsi callq 0x7196c movq (%r12), %rax leaq 0x8(%rsp), %rsi movq %r12, %rdi callq *0x10(%rax) andq $0x0, 0x110(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0xd0(%rsp) vmovups %xmm0, 0xdc(%rsp) vmovaps %xmm0, 0xf0(%rsp) vmovups %xmm0, 0xfc(%rsp) andq $0x0, 0x158(%rsp) vmovups %xmm0, 0x118(%rsp) vmovups %xmm0, 0x124(%rsp) vmovups %xmm0, 0x138(%rsp) vmovups %xmm0, 0x144(%rsp) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a2c60 lock incl (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2a2c97 lock decl (%rax) jne 0x2a2c97 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x2a2c8f movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a2c97 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0xd0(%rsp) movq 0x80(%rsp), %rax movq %rax, 0xe0(%rsp) movl 0x88(%rsp), %eax movl %eax, 0xe8(%rsp) movq 0x90(%rsp), %rax movq %rax, 0xf0(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0xf8(%rsp) movl 0xa8(%rsp), %eax movl %eax, 0x108(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x110(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a2d11 lock incl (%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x2a2d48 lock decl (%rax) jne 0x2a2d48 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x2a2d40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a2d48 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 leaq 0xd0(%rsp), %rsi vmovups %xmm0, 0x48(%rsi) movq 0x30(%rsp), %rax movq %rax, 0x58(%rsi) movl 0x38(%rsp), %eax movl %eax, 0x60(%rsi) movq 0x40(%rsp), %rax movq %rax, 0x68(%rsi) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsi) movl 0x58(%rsp), %eax movl %eax, 0x80(%rsi) movq 0x60(%rsp), %rax movq %rax, 0x88(%rsi) leaq 0xc0(%rsp), %rdi callq 0x6b00e movq (%r12), %rax leaq 0xc0(%rsp), %rsi movq %r12, %rdi callq *0x18(%rax) leaq 0xc0(%rsp), %rdi callq 0x6b03a movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq *0x20(%rax) movq (%r12), %rax movq %r12, %rdi movq %r14, %rsi movq 0xb8(%rsp), %rdx movq %r15, %rcx callq *0x38(%rax) movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq *0x28(%rax) movq (%r12), %rax movq %r12, %rdi callq *0x8(%rax) pushq $0x48 popq %rbx vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%rbx), %rax testq %rax, %rax je 0x2a2e43 lock decl (%rax) jne 0x2a2e43 movq 0xd0(%rsp,%rbx), %rsi movq 0xf0(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x2a2e37 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2a2e43 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%rbx), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x2a2e04 leaq 0x8(%rsp), %rdi callq 0x71614 xorl %r12d, %r12d movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a2ea9 lock decl (%rax) jne 0x2a2ea9 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2a2ea1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a2ea9 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a2eda lock decl (%rax) jne 0x2a2eda movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a2ed2 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a2eda movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x168, %rsp # imm = 0x168 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2a300f jmp 0x2a300f movq %rax, %rbx leaq 0xc0(%rsp), %rdi callq 0x6b03a jmp 0x2a2f1c jmp 0x2a2f19 jmp 0x2a2f11 jmp 0x2a2f11 movq %rax, %rbx jmp 0x2a2fa6 movq %rax, %rbx pushq $0x48 popq %r14 vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%r14), %rax testq %rax, %rax je 0x2a2f63 lock decl (%rax) jne 0x2a2f63 movq 0xd0(%rsp,%r14), %rsi movq 0xf0(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2a2f57 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2a2f63 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2a2f24 jmp 0x2a2f9c jmp 0x2a300f movq %rax, %rbx jmp 0x2a2fd4 jmp 0x2a300f movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a2fd4 lock decl (%rax) jne 0x2a2fd4 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2a2fce movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a2fd4 movq (%rdi), %rax callq *0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a3005 lock decl (%rax) jne 0x2a3005 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2a2fff movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a3005 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2a300f movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_fma::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_fma::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w *= weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w *= bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }

movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x2a2888

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_fma.cpp

ncnn::ConvolutionDepthWise_x86_avx::create_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x50, %rsp movq (%rdi), %rax movq -0x18(%rax), %r13 cmpl $0x0, 0x160(%rdi,%r13) je 0x2a30ed xorl %eax, %eax addq $0x50, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 retq movq %rsi, %r14 movq %rdi, %rbx movl 0x110(%rdi,%r13), %ecx decl %ecx cmpl $0x5, %ecx ja 0x2a3444 leaq 0x15559b(%rip), %rax # 0x3f86a8 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx jmpq *%rcx pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x2a3280 pushq $0x47 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x2a3280 pushq $0x36 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax vmovss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) jmp 0x2a3280 pushq $0x1e popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq (%r15), %rax movq %r15, %rdi movq %r12, %rsi callq *0x10(%rax) jmp 0x2a3280 pushq $0x1a popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) jmp 0x2a3280 pushq $0x43 popq %rdi callq 0x782bf movq %rax, %r15 leaq 0x8(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq 0x118(%rbx,%r13), %rax vmovss (%rax), %xmm0 movq %r12, %rdi xorl %esi, %esi callq 0x71952 movq 0x118(%rbx,%r13), %rax vmovss 0x4(%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi callq 0x71952 movq (%r15), %rax leaq 0x8(%rsp), %rsi movq %r15, %rdi callq *0x10(%rax) leaq 0x8(%rsp), %rdi callq 0x71614 movq (%r15), %rax movq %r15, %rdi movq %r14, %rsi callq *0x20(%rax) movq (%rbx), %rax movq %r15, 0x8(%rbx) movq -0x18(%rax), %r15 cmpb $0x1, 0x1e(%r14) jne 0x2a32c3 cmpq $0x1, 0x178(%rbx,%r15) jne 0x2a32c3 movq %rbx, %rdi movq %r14, %rsi callq 0x2a3642 jmp 0x2a30dd movl 0xd0(%rbx,%r15), %esi movl 0xd4(%rbx,%r15), %r10d movl 0xd8(%rbx,%r15), %r11d movl %r11d, %r8d imull %r10d, %r8d movl 0x104(%rbx,%r15), %eax movl 0x108(%rbx,%r15), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x2a33e2 imull %ecx, %eax cmpl %ecx, %eax jne 0x2a33e2 leaq (%rbx,%r15), %rsi cmpb $0x1, 0x27(%r14) jne 0x2a3339 testb $0x7, %cl je 0x2a344c testb $0x3, %cl je 0x2a34ad xorl $0x3, %r10d xorl $0x3, %r11d orl %r10d, %r11d jne 0x2a33e2 cmpl $0x1, 0xdc(%rsi) jne 0x2a336e cmpl $0x1, 0xe0(%rsi) jne 0x2a336e cmpl $0x1, 0xe4(%rsi) jne 0x2a336e cmpl $0x1, 0xe8(%rsi) je 0x2a3392 cmpl $0x1, 0xdc(%rsi) jne 0x2a33e2 cmpl $0x1, 0xe0(%rsi) jne 0x2a33e2 cmpl $0x2, 0xe4(%rsi) jne 0x2a33e2 cmpl $0x2, 0xe8(%rsi) jne 0x2a33e2 addq $0x168, %r15 # imm = 0x168 cmpq $0x28, %r15 je 0x2a33ed movq 0x8(%rbx,%r15), %rax testq %rax, %rax je 0x2a33ac lock incl (%rax) leaq 0x28(%rbx), %r12 movq 0x30(%rbx), %rax testq %rax, %rax je 0x2a3552 lock decl (%rax) jne 0x2a3552 movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x2a354a movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a3552 movq %rbx, %rdi movq %r14, %rsi callq 0x2a38ae cmpb $0x1, (%r14) jne 0x2a30dd movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x2a3512 lock decl (%rax) jne 0x2a3512 movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x2a350a movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a3512 xorl %r15d, %r15d jmp 0x2a3299 addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x8 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x2a33ed lock decl (%rax) jne 0x2a33ed movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x2a353d movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a33ed addq $0x168, %rsi # imm = 0x168 leaq 0x8(%rsp), %r15 movq %r15, %rdi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e leaq 0x28(%rbx), %rsi pushq $0x4 popq %rdx movq %r15, %rdi movq %r14, %rcx callq 0x64e3b movq 0x10(%rsp), %rax testq %rax, %rax je 0x2a33ed lock decl (%rax) jne 0x2a33ed movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x2a353d movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a33ed movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rbx) vmovups %xmm0, (%rbx) vmovups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x2a30dd movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a33ed movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%r12) vmovups %xmm0, (%r12) andl $0x0, 0x60(%rbx) vmovups %xmm0, 0x50(%rbx) vmovups (%rbx,%r15), %xmm0 vmovups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r15), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r15), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r15), %rax movq %rax, 0x48(%rbx) vmovups 0x28(%rbx,%r15), %xmm0 vmovups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r15), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r15), %rax movq %rax, 0x68(%rbx) jmp 0x2a33ed jmp 0x2a361a jmp 0x2a361a movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x2a3639 lock decl (%rax) jne 0x2a3639 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi je 0x2a3608 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a3639 jmp 0x2a361a movq %rax, %rbx movq 0x10(%rsp), %rax testq %rax, %rax je 0x2a3639 lock decl (%rax) jne 0x2a3639 movq 0x8(%rsp), %rsi movq 0x28(%rsp), %rdi testq %rdi, %rdi jne 0x2a3612 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a3639 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a3639 movq %rax, %rdi callq 0x61d68 jmp 0x2a362c jmp 0x2a362c jmp 0x2a362c jmp 0x2a362c jmp 0x2a362c movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq %rbx, %rdi callq 0x5f340 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

ncnn::ConvolutionDepthWise_x86_avx::create_pipeline_int8_x86(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx::create_pipeline_int8_x86(const Option& opt) { const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { elempack = channels % 8 == 0 ? 8 : 1; } #endif // __SSE2__ if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } if (elempack == 1) { weight_data_tm = weight_data; } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %r15 pushq %r14 pushq %r12 pushq %rbx subq $0x48, %rsp movq %rsi, %r15 movq %rdi, %rbx movq (%rdi), %rax movq -0x18(%rax), %r14 movl 0xd8(%rdi,%r14), %r8d imull 0xd4(%rdi,%r14), %r8d movl 0xd0(%rdi,%r14), %esi movl 0x104(%rdi,%r14), %eax movl 0x108(%rdi,%r14), %ecx cltd idivl %ecx cltd idivl %r8d movl %eax, %edi movl %esi, %eax cltd idivl %ecx movl %eax, %r9d movl %edi, %eax cltd idivl %r9d cmpl %esi, %ecx jne 0x2a3724 imull %ecx, %eax cmpl %ecx, %eax jne 0x2a3724 testb $0x7, %cl jne 0x2a3777 cmpb $0x0, 0x27(%r15) je 0x2a3777 addq %rbx, %r14 addq $0x168, %r14 # imm = 0x168 movq %rsp, %r12 movq %r12, %rdi movq %r14, %rsi movl %r8d, %edx xorl %r8d, %r8d callq 0x62e4e addq $0x28, %rbx pushq $0x8 popq %rdx movq %r12, %rdi movq %rbx, %rsi movq %r15, %rcx callq 0x64e3b movq 0x8(%rsp), %rax testq %rax, %rax je 0x2a385d lock decl (%rax) jne 0x2a385d movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2a37ec movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a385d movq %rbx, %rdi movq %r15, %rsi callq 0x2a38ae cmpb $0x1, (%r15) jne 0x2a385d movq (%rbx), %rax movq -0x18(%rax), %rax leaq (%rbx,%rax), %r14 addq %rax, %rbx addq $0x168, %rbx # imm = 0x168 movq 0x8(%rbx), %rax testq %rax, %rax je 0x2a37c4 lock decl (%rax) jne 0x2a37c4 movq 0x168(%r14), %rsi movq 0x188(%r14), %rdi testq %rdi, %rdi je 0x2a37bc movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a37c4 addq $0x168, %r14 # imm = 0x168 cmpq $0x28, %r14 je 0x2a385d movq 0x8(%rbx,%r14), %rax testq %rax, %rax je 0x2a3795 lock incl (%rax) leaq 0x28(%rbx), %r15 movq 0x30(%rbx), %rax testq %rax, %rax je 0x2a37fe lock decl (%rax) jne 0x2a37fe movq 0x28(%rbx), %rsi movq 0x48(%rbx), %rdi testq %rdi, %rdi je 0x2a37f6 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a37fe movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x1a8(%r14) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rbx) vmovups %xmm0, (%rbx) vmovups %xmm0, 0x190(%r14) andl $0x0, 0x1a0(%r14) jmp 0x2a385d movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a385d movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x68(%rbx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%r15) vmovups %xmm0, (%r15) andl $0x0, 0x60(%rbx) vmovups %xmm0, 0x50(%rbx) vmovups (%rbx,%r14), %xmm0 vmovups %xmm0, 0x28(%rbx) movq 0x10(%rbx,%r14), %rax movq %rax, 0x38(%rbx) movl 0x18(%rbx,%r14), %eax movl %eax, 0x40(%rbx) movq 0x20(%rbx,%r14), %rax movq %rax, 0x48(%rbx) vmovups 0x28(%rbx,%r14), %xmm0 vmovups %xmm0, 0x50(%rbx) movl 0x38(%rbx,%r14), %eax movl %eax, 0x60(%rbx) movq 0x40(%rbx,%r14), %rax movq %rax, 0x68(%rbx) xorl %eax, %eax addq $0x48, %rsp popq %rbx popq %r12 popq %r14 popq %r15 retq jmp 0x2a38a5 movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2a389d lock decl (%rax) jne 0x2a389d movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2a3897 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a389d movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

ncnn::ConvolutionDepthWise_x86_avx::create_group_ops(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx::create_group_ops(const Option& opt) { // create Convolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Convolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[5]; weights[0] = weight_data_g; weights[1] = bias_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[2] = weight_data_int8_scales_g; weights[3] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[4] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[4]; weights[0] = weight_data_g; #if NCNN_INT8 if (int8_scale_term) { Mat weight_data_int8_scales_g(num_output_g); weight_data_int8_scales_g.fill(weight_data_int8_scales[g]); weights[1] = weight_data_int8_scales_g; weights[2] = bottom_blob_int8_scales.range(g, 1); } if (int8_scale_term > 100) { weights[3] = top_blob_int8_scales.range(g, 1); } #endif op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x278, %rsp # imm = 0x278 movq %rsi, 0x250(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x104(%rdi,%rdx), %eax movl 0x108(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x10(%rdi), %rax movq %rax, 0x1e0(%rsp) xorl %r12d, %r12d movq 0x10(%r14), %rax movq 0x18(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x2a394b movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x2a3946 movq (%rdi), %rax callq *0x8(%rax) incq %r12 jmp 0x2a391d imull %r15d, %ebx cmpq %rax, %rcx je 0x2a3958 movq %rax, 0x18(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x108(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0x1e0(%rsp), %rdi callq 0x6fbc2 leaq 0x118(%r14), %rax movq %rax, 0x258(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0x6c(%rsp) movslq %ebx, %rax movq %rax, 0x260(%rsp) movslq %r15d, %rax movq %rax, 0x248(%rsp) pushq $0x1 popq %rbx xorl %edx, %edx movl %r15d, 0x14(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rcx cmpq %rcx, %rdx jge 0x2a46c2 movq %rdx, %rcx movq %rdx, 0x8(%rsp) movq 0x260(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdx, 0x80(%rsp) movl %esi, 0x88(%rsp) movq %rax, 0x90(%rsp) movl %ebx, 0x98(%rsp) movl %edi, 0x9c(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa0(%rsp) movl %ebx, 0xa8(%rsp) movq %rdi, 0xb0(%rsp) leaq 0x200(%rsp), %rdi leaq 0x70(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a3aa9 lock decl (%rax) jne 0x2a3aa9 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a3aa1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a3aa9 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x2a3b19 movq 0x8(%rsp), %r15 movq 0x248(%rsp), %rcx imulq %rcx, %r15 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0x1f8(%rsp) imulq %rsi, %r15 addq 0x1b0(%r14,%rax), %r15 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x1c(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0x1f0(%rsp) movl %ebx, %r13d movl 0x14(%rsp), %eax movl %eax, 0x18(%rsp) movq %rcx, 0x1e8(%rsp) jmp 0x2a3b53 xorl %r15d, %r15d movq $0x0, 0x1f8(%rsp) movl $0x0, 0x1c(%rsp) movq $0x0, 0x1f0(%rsp) xorl %r13d, %r13d movl $0x0, 0x18(%rsp) movq $0x0, 0x1e8(%rsp) pushq $0x6 popq %rdi callq 0x782bf movq %rax, %rbp leaq 0x268(%rsp), %r12 movq %r12, %rdi callq 0x71548 movq %r12, %rdi xorl %esi, %esi movl 0x14(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r12, %rdi movl %ebx, %esi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r12, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r12, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r12, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r12, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r12, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r12, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r12, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r12, %rdi pushq $0x6 popq %rsi movl 0x6c(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r12, %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x110(%r14,%rax), %edx movq %r12, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x258(%rsp), %rdx movq %r12, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r12, %rsi callq *0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r14,%rax) je 0x2a3d40 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x1a8, %rax # imm = 0x1A8 jne 0x2a3cce movq 0x208(%rsp), %rax testq %rax, %rax je 0x2a3d08 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a3dc1 lock decl (%rax) jne 0x2a3dc1 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a3db9 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a3dc1 pushq $0x40 popq %rax vxorps %xmm0, %xmm0, %xmm0 andq $0x0, 0x70(%rsp,%rax) vmovups %xmm0, 0x30(%rsp,%rax) vmovups %xmm0, 0x3c(%rsp,%rax) vmovups %xmm0, 0x50(%rsp,%rax) vmovups %xmm0, 0x5c(%rsp,%rax) addq $0x48, %rax cmpq $0x160, %rax # imm = 0x160 jne 0x2a3d47 movq 0x208(%rsp), %rax testq %rax, %rax je 0x2a3d81 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a4178 lock decl (%rax) jne 0x2a4178 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a4170 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a4178 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x2a3e65 lock decl (%rax) jne 0x2a3e65 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x2a3e5d movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a3e65 movq %rsi, %rdi callq 0x5f3e0 movq %r15, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq 0x1f8(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x1c(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x1f0(%rsp), %rax movq %rax, 0xd8(%rsp) movl %r13d, 0xe0(%rsp) movl 0x18(%rsp), %eax movl %eax, 0xe4(%rsp) movl %r13d, 0xe8(%rsp) movl %r13d, 0xec(%rsp) movl %r13d, 0xf0(%rsp) movq 0x1e8(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2a40fa andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x2a3f6a vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x2a3f5c movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a3f77 lock incl (%rax) movq 0x108(%rsp), %rax testq %rax, %rax je 0x2a3fae lock decl (%rax) jne 0x2a3fae movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x2a3fa6 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a3fae movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovaps %xmm0, 0x100(%rsp) movq 0x30(%rsp), %rax movq %rax, 0x110(%rsp) movl 0x38(%rsp), %eax movl %eax, 0x118(%rsp) movq 0x40(%rsp), %rax movq %rax, 0x120(%rsp) vmovups 0x48(%rsp), %xmm0 vmovups %xmm0, 0x128(%rsp) movl 0x58(%rsp), %eax movl %eax, 0x138(%rsp) movq 0x60(%rsp), %rax movq %rax, 0x140(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %r12d movq 0x260(%r14,%rax), %rbx movq 0x150(%rsp), %rax testq %rax, %rax je 0x2a4070 lock decl (%rax) jne 0x2a4070 movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x2a4068 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a4070 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %r12d, 0x160(%rsp) movq %rbx, 0x168(%rsp) vbroadcastss 0x15455e(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x2a40f3 lock decl (%rax) jne 0x2a40f3 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2a40eb movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a40f3 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x2a431a movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x198(%rsp), %rax testq %rax, %rax je 0x2a42c6 lock decl (%rax) jne 0x2a42c6 movq 0x190(%rsp), %rsi movq 0x1b0(%rsp), %rdi testq %rdi, %rdi je 0x2a42be movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a42c6 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x200(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsp) movq 0x210(%rsp), %rax movq %rax, 0x80(%rsp) movl 0x218(%rsp), %eax movl %eax, 0x88(%rsp) movq 0x220(%rsp), %rax movq %rax, 0x90(%rsp) vmovups 0x228(%rsp), %xmm0 vmovups %xmm0, 0x98(%rsp) movl 0x238(%rsp), %eax movl %eax, 0xa8(%rsp) movq 0x240(%rsp), %rax movq %rax, 0xb0(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2a44fd andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) leaq 0x40(%rsp), %rax vmovups %xmm0, 0xc(%rax) vmovaps %xmm0, (%rax) leaq 0x20(%rsp), %rdi movl 0x14(%rsp), %esi pushq $0x4 popq %rdx xorl %ecx, %ecx callq 0x635fa movq (%r14), %rax movq -0x18(%rax), %rax movq 0x1f8(%r14,%rax), %rax movq 0x8(%rsp), %rcx vmovss (%rax,%rcx,4), %xmm0 movl 0x58(%rsp), %eax imull 0x60(%rsp), %eax movq 0x20(%rsp), %rcx testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax xorl %edx, %edx cmpl %edx, %eax je 0x2a4273 vmovss %xmm0, (%rcx,%rdx,4) incq %rdx jmp 0x2a4265 movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a4280 lock incl (%rax) movq 0xc0(%rsp), %rax testq %rax, %rax je 0x2a43b2 lock decl (%rax) jne 0x2a43b2 movq 0xb8(%rsp), %rsi movq 0xd8(%rsp), %rdi testq %rdi, %rdi je 0x2a43aa movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a43b2 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x190(%rsp) andq $0x0, 0x198(%rsp) movq %r15, 0x1a0(%rsp) movl %ebx, 0x1a8(%rsp) movq %r12, 0x1b0(%rsp) vbroadcastss 0x154309(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x1b8(%rsp) movl $0x1, 0x1c8(%rsp) movq $0x1, 0x1d0(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0x120, %r15d # imm = 0x120 movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x2a4378 lock decl (%rax) jne 0x2a4378 movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x2a4370 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a4378 movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x2a4347 jmp 0x2a464b movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 vmovups %xmm0, 0xb8(%rsp) movq 0x30(%rsp), %rax movq %rax, 0xc8(%rsp) movl 0x38(%rsp), %eax movl %eax, 0xd0(%rsp) movq 0x40(%rsp), %rax movq %rax, 0xd8(%rsp) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movl 0x58(%rsp), %eax movl %eax, 0xf0(%rsp) movq 0x60(%rsp), %rax movq %rax, 0xf8(%rsp) movq (%r14), %rax movq -0x18(%rax), %rax movq 0x250(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x240(%r14,%rax), %r13 movl 0x258(%r14,%rax), %ebx movq 0x260(%r14,%rax), %r12 movq 0x108(%rsp), %rax testq %rax, %rax je 0x2a4474 lock decl (%rax) jne 0x2a4474 movq 0x100(%rsp), %rsi movq 0x120(%rsp), %rdi testq %rdi, %rdi je 0x2a446c movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a4474 movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x100(%rsp) andq $0x0, 0x108(%rsp) movq %r15, 0x110(%rsp) movl %ebx, 0x118(%rsp) movq %r12, 0x120(%rsp) vbroadcastss 0x15415b(%rip), %xmm0 # 0x3f8600 vmovups %xmm0, 0x128(%rsp) movl $0x1, 0x138(%rsp) movq $0x1, 0x140(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax pushq $0x1 popq %rbx je 0x2a44f6 lock decl (%rax) jne 0x2a44f6 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2a44ee movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a44f6 movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x65, 0x10c(%r14,%rax) jl 0x2a45c0 movq 0x298(%r14,%rax), %r15 movq %r15, %r13 imulq 0x8(%rsp), %r13 addq 0x288(%r14,%rax), %r13 movl 0x2a0(%r14,%rax), %ebx movq 0x2a8(%r14,%rax), %r12 movq 0x150(%rsp), %rax testq %rax, %rax je 0x2a456c lock decl (%rax) jne 0x2a456c movq 0x148(%rsp), %rsi movq 0x168(%rsp), %rdi testq %rdi, %rdi je 0x2a4564 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a456c movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x148(%rsp) andq $0x0, 0x150(%rsp) movq %r15, 0x158(%rsp) movl %ebx, 0x160(%rsp) movq %r12, 0x168(%rsp) vbroadcastss 0x154063(%rip), %xmm0 # 0x3f8600 vmovaps %xmm0, 0x170(%rsp) movl $0x1, 0x180(%rsp) movq $0x1, 0x188(%rsp) pushq $0x1 popq %rbx leaq 0x20(%rsp), %r15 movq %r15, %rdi leaq 0x70(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x18(%rax) movq %r15, %rdi callq 0x6b03a movl $0xd8, %r15d movq 0x78(%rsp,%r15), %rax testq %rax, %rax je 0x2a461e lock decl (%rax) jne 0x2a461e movq 0x70(%rsp,%r15), %rsi movq 0x90(%rsp,%r15), %rdi testq %rdi, %rdi je 0x2a4616 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a461e movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%r15), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r15 cmpq $-0x48, %r15 jne 0x2a45ed movq (%rbp), %rax movq %rbp, %rdi movq 0x250(%rsp), %rsi callq *0x20(%rax) leaq 0x268(%rsp), %rdi movq 0x1e0(%rsp), %rax movq (%rax), %rax movq 0x8(%rsp), %rcx movq %rbp, (%rax,%rcx,8) callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x2a46b5 lock decl (%rax) jne 0x2a46b5 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi je 0x2a46ad movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a46b5 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rdx incq %rdx jmp 0x2a39c3 xorl %eax, %eax addq $0x278, %rsp # imm = 0x278 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2a478a jmp 0x2a4921 jmp 0x2a46e2 movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a479c lock decl (%rax) jne 0x2a479c movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2a4718 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a479c movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a479c jmp 0x2a4921 jmp 0x2a4814 jmp 0x2a478a jmp 0x2a472e movq %rax, %rbx movq 0x28(%rsp), %rax testq %rax, %rax je 0x2a4826 lock decl (%rax) jne 0x2a4826 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2a4764 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a4826 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a4826 jmp 0x2a4921 jmp 0x2a4921 jmp 0x2a4921 jmp 0x2a4921 jmp 0x2a478a jmp 0x2a4814 movq %rax, %rbx jmp 0x2a479c movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0xd8, %r14d vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x2a47df lock decl (%rax) jne 0x2a47df movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2a47d3 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2a47df movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2a47a6 jmp 0x2a48d5 jmp 0x2a4921 jmp 0x2a4814 movq %rax, %rbx jmp 0x2a4826 movq %rax, %rbx leaq 0x20(%rsp), %rdi callq 0x6b03a movl $0x120, %r14d # imm = 0x120 vxorps %xmm0, %xmm0, %xmm0 movq 0x78(%rsp,%r14), %rax testq %rax, %rax je 0x2a4869 lock decl (%rax) jne 0x2a4869 movq 0x70(%rsp,%r14), %rsi movq 0x90(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2a485d movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2a4869 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x70, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2a4830 jmp 0x2a48d5 jmp 0x2a4921 jmp 0x2a48c9 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a4919 lock decl (%rax) jne 0x2a4919 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a4909 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a4919 jmp 0x2a4921 movq %rax, %rbx jmp 0x2a48e2 jmp 0x2a4921 jmp 0x2a4921 movq %rax, %rbx leaq 0x268(%rsp), %rdi callq 0x71614 movq 0x208(%rsp), %rax testq %rax, %rax je 0x2a4919 lock decl (%rax) jne 0x2a4919 movq 0x200(%rsp), %rsi movq 0x220(%rsp), %rdi testq %rdi, %rdi jne 0x2a4913 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a4919 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_avx::create_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx::create_pipeline(const Option& opt) { if (dynamic_weight) return 0; activation = create_activation_layer(activation_type, activation_params, opt); #if NCNN_INT8 if (opt.use_int8_inference && weight_data.elemsize == (size_t)1u) { return create_pipeline_int8_x86(opt); } #endif const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; // depth-wise if (channels == group && group == num_output) { int elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ elempack = channels % 16 == 0 ? 16 : channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #elif __AVX__ elempack = channels % 8 == 0 ? 8 : channels % 4 == 0 ? 4 : 1; #else elempack = channels % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ #if __SSE2__ #if __AVX__ // pack16 #if __AVX512F__ if (elempack == 16) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 16, opt); } #endif // __AVX512F__ // pack8 if (elempack == 8) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 8, opt); } #endif // __AVX__ // pack4 if (elempack == 4) { Mat weight_data_r2 = weight_data.reshape(maxk, group); convert_packing(weight_data_r2, weight_data_tm, 4, opt); } #endif // __SSE2__ if (elempack == 1) { // depth-wise specific if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { weight_data_tm = weight_data; } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { weight_data_tm = weight_data; } else { create_group_ops(opt); } } if (opt.lightmode) { weight_data.release(); } return 0; } // group convolution create_group_ops(opt); if (opt.lightmode) { weight_data.release(); } return 0; }

pushq %rax movq (%rdi), %rax addq -0x30(%rax), %rdi callq 0x2a30be xorl %eax, %eax popq %rcx retq nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

ncnn::ConvolutionDepthWise_x86_avx::destroy_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }

pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r14 movq %rdi, %rbx movq 0x8(%rdi), %rdi testq %rdi, %rdi je 0x2a496d movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x8(%rbx), %rdi testq %rdi, %rdi je 0x2a4968 movq (%rdi), %rax callq *0x8(%rax) andq $0x0, 0x8(%rbx) xorl %r15d, %r15d movq 0x10(%rbx), %rax movq 0x18(%rbx), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r15 jge 0x2a49af movq (%rax,%r15,8), %rdi movq (%rdi), %rax movq %r14, %rsi callq *0x28(%rax) movq 0x10(%rbx), %rax movq (%rax,%r15,8), %rdi testq %rdi, %rdi je 0x2a49aa movq (%rdi), %rax callq *0x8(%rax) incq %r15 jmp 0x2a4970 cmpq %rax, %rcx je 0x2a49b8 movq %rax, 0x18(%rbx) xorl %eax, %eax popq %rbx popq %r14 popq %r15 retq

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_avx::destroy_pipeline(ncnn::Option const&)

int ConvolutionDepthWise_x86_avx::destroy_pipeline(const Option& opt) { if (activation) { activation->destroy_pipeline(opt); delete activation; activation = 0; } for (int i = 0; i < (int)group_ops.size(); i++) { group_ops[i]->destroy_pipeline(opt); delete group_ops[i]; } group_ops.clear(); return 0; }

pushq %rax movq (%rdi), %rax addq -0x38(%rax), %rdi callq 0x2a493c xorl %eax, %eax popq %rcx retq nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

ncnn::ConvolutionDepthWise_x86_avx::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float*)bias_data) + g * 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float*)bias_data) + g * 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float*)bias_data) + g * 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x3b8, %rsp # imm = 0x3B8 movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdi movq %rcx, 0x48(%rsp) cmpb $0x1, 0x1e(%rcx) jne 0x2a4a21 cmpl $0x0, 0x10c(%r14,%rdi) je 0x2a4a21 movq %r14, %rdi movq 0x48(%rsp), %rcx addq $0x3b8, %rsp # imm = 0x3B8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x2a98c2 movq %rdx, 0x18(%rsp) movl 0x38(%rsi), %ecx movq %rcx, 0xf0(%rsp) movq 0x10(%rsi), %r13 movslq 0x18(%rsi), %r15 movl 0xd4(%r14,%rdi), %ebx decl %ebx imull 0xdc(%r14,%rdi), %ebx movl 0xd8(%r14,%rdi), %ebp decl %ebp imull 0xe0(%r14,%rdi), %ebp leaq 0x90(%rsp), %rdx andq $0x0, 0x40(%rdx) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rdx) vmovups %xmm0, 0xc(%rdx) vmovaps %xmm0, 0x20(%rdx) vmovups %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %r14, %rdi movq 0x48(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0x90(%rsp) je 0x2a96d0 movslq 0xc8(%rsp), %rax imulq 0xd0(%rsp), %rax testq %rax, %rax je 0x2a96d0 notl %ebx notl %ebp movl 0xbc(%rsp), %edi addl %edi, %ebx movq (%r14), %rax movq -0x18(%rax), %rcx movl %ebx, %eax cltd idivl 0xe4(%r14,%rcx) movl %eax, %ebx leal 0x1(%rbx), %esi addl 0xc0(%rsp), %ebp movl %ebp, %eax cltd idivl 0xe8(%r14,%rcx) movq %rax, 0x50(%rsp) leal 0x1(%rax), %r8d movl 0xd0(%r14,%rcx), %ecx movq 0x48(%rsp), %rax cmpb $0x1, 0x27(%rax) pushq $0x8 popq %rbp movl %edi, 0x20(%rsp) jne 0x2a4b30 xorl %eax, %eax testb $0x3, %cl sete %al testb $0x7, %cl leal 0x1(%rax,%rax,2), %r9d cmovel %ebp, %r9d jmp 0x2a4b34 pushq $0x1 popq %r9 movq %r13, %rax xorl %edx, %edx divq %r15 movq %rax, %r10 movl %r9d, %eax movq %rax, 0x70(%rsp) imulq %rax, %r10 movl %ecx, %eax cltd idivl %r9d movq 0x48(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq 0x18(%rsp), %r13 movq %r13, %rdi movl %esi, 0x110(%rsp) movl %r8d, 0x130(%rsp) movl %r8d, %edx movl %eax, %ecx movq %r10, 0x170(%rsp) movq %r10, %r8 callq 0x628f2 movq (%r13), %r9 testq %r9, %r9 je 0x2a96d0 movq 0x40(%r13), %r11 movslq 0x38(%r13), %rax imulq %r11, %rax testq %rax, %rax je 0x2a96d0 movq %r15, %r10 movl %r10d, %eax imull 0xf0(%rsp), %eax movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xd0(%r14,%rdi), %ecx movl 0x108(%r14,%rdi), %esi cmpl %esi, %eax jne 0x2a75fb cmpl %ecx, %eax jne 0x2a75fb addq %r14, %rdi cmpl $0x1, %r10d je 0x2a6634 movslq %ebx, %rbx cmpl $0x4, %r10d movq %r14, 0x10(%rsp) je 0x2a59a4 movl %eax, %ecx cmpl $0x8, %r10d jne 0x2a75fb movl 0xd4(%rdi), %eax movl 0xd8(%rdi), %r15d cmpl $0x5, %eax je 0x2a70c6 cmpl $0x3, %eax jne 0x2a93b2 cmpl $0x3, %r15d jne 0x2a93b2 cmpl $0x1, 0xdc(%rdi) jne 0x2a8abf cmpl $0x1, 0xe0(%rdi) jne 0x2a8abf cmpl $0x1, 0xe4(%rdi) jne 0x2a8abf cmpl $0x1, 0xe8(%rdi) jne 0x2a8abf movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xbc(%rsp), %edx movl 0xc8(%rsp), %esi movq 0x1b0(%rdi), %r8 leal 0x10(,%rdx,8), %edx movslq %edx, %rdi leal (,%rax,8), %edx movslq %edx, %rdx xorl %r9d, %r9d testl %esi, %esi cmovlel %r9d, %esi movq %rsi, 0x160(%rsp) shlq $0x2, %rdx movq %rdx, 0x190(%rsp) shlq $0x2, %rdi movq %r8, 0x150(%rsp) cmpq 0x160(%rsp), %r9 je 0x2a9725 testq %r8, %r8 je 0x2a4ce5 movq %r9, %rdx shlq $0x5, %rdx vmovups (%r8,%rdx), %ymm15 jmp 0x2a4cea vxorps %xmm15, %xmm15, %xmm15 movq 0x40(%r13), %r10 imulq %r9, %r10 movq 0x10(%r13), %rsi imulq %rsi, %r10 addq (%r13), %r10 movslq 0x2c(%r13), %rdx movslq 0x54(%r14), %r11 imulq %r9, %r11 imulq 0x38(%r14), %r11 movq 0x28(%r14), %rbx imulq %rsi, %rdx addq %r10, %rdx movslq 0xbc(%rsp), %rsi movq 0xd0(%rsp), %r15 movq %r9, 0xe0(%rsp) imulq %r9, %r15 movq 0xa0(%rsp), %r8 imulq %r8, %r15 addq 0x90(%rsp), %r15 imulq %r8, %rsi leaq (%r15,%rsi), %r12 leaq (%r15,%rsi,2), %r13 leaq (%rsi,%rsi,2), %rsi addq %r15, %rsi xorl %ebp, %ebp vmovups %ymm15, 0x20(%rsp) movl %ebp, %r8d orl $0x1, %r8d cmpl %ecx, %r8d jge 0x2a597f xorl %r9d, %r9d xorl %r8d, %r8d leal 0x3(%r8), %r14d cmpl %eax, %r14d jge 0x2a53c6 vmovaps 0x20(%r15,%r9), %ymm0 vmovaps 0x40(%r15,%r9), %ymm2 vmovaps 0x60(%r15,%r9), %ymm6 vmovaps 0x80(%r15,%r9), %ymm7 vmovaps (%rbx,%r11), %ymm3 vmovaps 0x20(%rbx,%r11), %ymm5 vmovaps 0x40(%rbx,%r11), %ymm4 vmovaps 0x60(%rbx,%r11), %ymm1 vmulps (%r15,%r9), %ymm3, %ymm8 vaddps %ymm15, %ymm8, %ymm8 vmulps %ymm3, %ymm0, %ymm9 vaddps %ymm15, %ymm9, %ymm9 vmulps %ymm3, %ymm2, %ymm10 vaddps %ymm15, %ymm10, %ymm10 vmulps %ymm3, %ymm6, %ymm11 vaddps %ymm15, %ymm11, %ymm11 vmulps %ymm5, %ymm0, %ymm0 vmulps %ymm5, %ymm2, %ymm12 vmulps %ymm5, %ymm6, %ymm13 vmulps %ymm5, %ymm7, %ymm14 vmulps %ymm4, %ymm2, %ymm2 vaddps %ymm2, %ymm0, %ymm0 vaddps %ymm0, %ymm8, %ymm0 vmovups %ymm0, 0xf0(%rsp) vmulps %ymm4, %ymm6, %ymm0 vaddps %ymm0, %ymm12, %ymm0 vaddps %ymm0, %ymm9, %ymm0 vmovups %ymm0, 0x110(%rsp) vmulps %ymm4, %ymm7, %ymm0 vaddps %ymm0, %ymm13, %ymm0 vmulps 0xa0(%r15,%r9), %ymm4, %ymm2 vaddps %ymm0, %ymm10, %ymm0 vmovups %ymm0, 0x1f0(%rsp) vaddps %ymm2, %ymm14, %ymm0 vaddps %ymm0, %ymm11, %ymm0 vmovups %ymm0, 0x130(%rsp) vmovaps 0x80(%rbx,%r11), %ymm2 vmovaps (%r12,%r9), %ymm11 vmovaps 0x20(%r12,%r9), %ymm7 vmovaps 0x40(%r12,%r9), %ymm15 vmovaps 0x60(%r12,%r9), %ymm14 vmulps %ymm1, %ymm11, %ymm0 vmulps %ymm1, %ymm7, %ymm6 vmulps %ymm2, %ymm7, %ymm8 vaddps %ymm0, %ymm8, %ymm9 vmulps %ymm2, %ymm15, %ymm0 vaddps %ymm0, %ymm6, %ymm8 vmulps %ymm1, %ymm15, %ymm0 vmulps %ymm2, %ymm14, %ymm6 vaddps %ymm6, %ymm0, %ymm10 vmovaps 0x80(%r12,%r9), %ymm0 vmulps %ymm1, %ymm14, %ymm6 vmulps %ymm2, %ymm0, %ymm13 vaddps %ymm6, %ymm13, %ymm6 vmulps %ymm5, %ymm7, %ymm13 vmulps %ymm4, %ymm15, %ymm12 vaddps %ymm12, %ymm13, %ymm12 vmulps %ymm3, %ymm11, %ymm11 vaddps 0x20(%rsp), %ymm11, %ymm11 vaddps %ymm12, %ymm11, %ymm11 vmovups %ymm11, 0x50(%rsp) vmulps %ymm5, %ymm15, %ymm11 vmulps %ymm4, %ymm14, %ymm12 vaddps %ymm12, %ymm11, %ymm11 vmulps %ymm3, %ymm7, %ymm7 vaddps 0x20(%rsp), %ymm7, %ymm7 vaddps %ymm7, %ymm11, %ymm7 vmovups %ymm7, 0x70(%rsp) vmulps %ymm5, %ymm14, %ymm11 vmulps %ymm4, %ymm0, %ymm12 vaddps %ymm12, %ymm11, %ymm11 vmulps %ymm3, %ymm15, %ymm12 vaddps 0x20(%rsp), %ymm12, %ymm12 vaddps %ymm11, %ymm12, %ymm7 vmovups %ymm7, 0x170(%rsp) vmulps %ymm5, %ymm0, %ymm5 vmovaps 0xa0(%r12,%r9), %ymm12 vmulps %ymm4, %ymm12, %ymm4 vaddps %ymm4, %ymm5, %ymm4 vmulps %ymm3, %ymm14, %ymm3 vaddps 0x20(%rsp), %ymm3, %ymm3 vaddps %ymm4, %ymm3, %ymm3 vmovups %ymm3, 0x270(%rsp) vmovaps 0xa0(%rbx,%r11), %ymm13 vmulps %ymm13, %ymm15, %ymm4 vaddps %ymm4, %ymm9, %ymm4 vaddps 0xf0(%rsp), %ymm4, %ymm3 vmovups %ymm3, 0xf0(%rsp) vmulps %ymm13, %ymm14, %ymm5 vaddps %ymm5, %ymm8, %ymm5 vaddps 0x110(%rsp), %ymm5, %ymm3 vmovups %ymm3, 0x110(%rsp) vmulps %ymm0, %ymm13, %ymm0 vaddps %ymm0, %ymm10, %ymm0 vaddps 0x1f0(%rsp), %ymm0, %ymm0 vmovups %ymm0, 0x1f0(%rsp) vmulps %ymm13, %ymm12, %ymm0 vaddps %ymm0, %ymm6, %ymm0 vaddps 0x130(%rsp), %ymm0, %ymm0 vmovups %ymm0, 0x130(%rsp) vmovaps (%r13,%r9), %ymm6 vmovaps 0x20(%r13,%r9), %ymm9 vmovaps 0x40(%r13,%r9), %ymm10 vmulps %ymm1, %ymm6, %ymm12 vmulps %ymm2, %ymm9, %ymm14 vaddps %ymm14, %ymm12, %ymm12 vmulps %ymm1, %ymm9, %ymm14 vmulps %ymm2, %ymm10, %ymm15 vaddps %ymm15, %ymm14, %ymm8 vmovaps 0x60(%r13,%r9), %ymm14 vmulps %ymm1, %ymm10, %ymm15 vmulps %ymm2, %ymm14, %ymm7 vaddps %ymm7, %ymm15, %ymm7 vmulps %ymm1, %ymm14, %ymm1 vmovaps 0x80(%r13,%r9), %ymm15 vmulps %ymm2, %ymm15, %ymm2 vaddps %ymm2, %ymm1, %ymm5 vmovaps 0xc0(%rbx,%r11), %ymm1 vmulps %ymm1, %ymm6, %ymm6 vmovaps 0xe0(%rbx,%r11), %ymm2 vmulps %ymm2, %ymm9, %ymm11 vaddps %ymm6, %ymm11, %ymm11 vmulps %ymm1, %ymm9, %ymm6 vmulps %ymm2, %ymm10, %ymm9 vaddps %ymm6, %ymm9, %ymm0 vmulps %ymm1, %ymm10, %ymm6 vmulps %ymm2, %ymm14, %ymm9 vaddps %ymm6, %ymm9, %ymm3 vmulps %ymm1, %ymm14, %ymm6 vmulps %ymm2, %ymm15, %ymm9 vaddps %ymm6, %ymm9, %ymm4 vmulps %ymm13, %ymm10, %ymm6 vaddps %ymm6, %ymm12, %ymm6 vaddps 0x50(%rsp), %ymm6, %ymm6 vmovups %ymm6, 0x50(%rsp) vmulps %ymm13, %ymm14, %ymm9 vaddps %ymm9, %ymm8, %ymm8 vaddps 0x70(%rsp), %ymm8, %ymm9 vmulps %ymm13, %ymm15, %ymm8 vaddps %ymm7, %ymm8, %ymm7 vaddps 0x170(%rsp), %ymm7, %ymm7 vmovaps 0xa0(%r13,%r9), %ymm8 vmulps %ymm13, %ymm8, %ymm12 vaddps %ymm5, %ymm12, %ymm5 vaddps 0x270(%rsp), %ymm5, %ymm13 vmovaps 0x100(%rbx,%r11), %ymm12 vmulps %ymm12, %ymm10, %ymm5 vaddps %ymm5, %ymm11, %ymm5 vaddps 0xf0(%rsp), %ymm5, %ymm10 vmulps %ymm12, %ymm14, %ymm5 vaddps %ymm5, %ymm0, %ymm0 vaddps 0x110(%rsp), %ymm0, %ymm5 vmulps %ymm12, %ymm15, %ymm0 vaddps %ymm0, %ymm3, %ymm0 vaddps 0x1f0(%rsp), %ymm0, %ymm3 vmulps %ymm12, %ymm8, %ymm0 vaddps %ymm0, %ymm4, %ymm0 vaddps 0x130(%rsp), %ymm0, %ymm0 vmovaps 0x20(%rsi,%r9), %ymm4 vmovaps 0x40(%rsi,%r9), %ymm8 vmulps (%rsi,%r9), %ymm1, %ymm11 vmovaps 0x60(%rsi,%r9), %ymm14 vmulps %ymm1, %ymm4, %ymm15 vmulps %ymm2, %ymm4, %ymm4 vaddps %ymm4, %ymm11, %ymm4 vmulps %ymm2, %ymm8, %ymm11 vaddps %ymm11, %ymm15, %ymm11 vmulps %ymm1, %ymm8, %ymm15 vmulps %ymm2, %ymm14, %ymm6 vaddps %ymm6, %ymm15, %ymm6 vmulps %ymm1, %ymm14, %ymm1 vmovaps 0x80(%rsi,%r9), %ymm15 vmulps %ymm2, %ymm15, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps %ymm12, %ymm8, %ymm2 vaddps %ymm2, %ymm4, %ymm2 vaddps 0x50(%rsp), %ymm2, %ymm2 vmulps %ymm12, %ymm14, %ymm4 vaddps %ymm4, %ymm11, %ymm4 vaddps %ymm4, %ymm9, %ymm4 vmulps %ymm12, %ymm15, %ymm8 vmovups 0x20(%rsp), %ymm15 vaddps %ymm6, %ymm8, %ymm6 vaddps %ymm6, %ymm7, %ymm6 vmulps 0xa0(%rsi,%r9), %ymm12, %ymm7 vaddps %ymm7, %ymm1, %ymm1 vaddps %ymm1, %ymm13, %ymm1 vmovaps %ymm10, (%r10,%r9) vmovaps %ymm5, 0x20(%r10,%r9) vmovaps %ymm3, 0x40(%r10,%r9) vmovaps %ymm0, 0x60(%r10,%r9) vmovaps %ymm2, (%rdx,%r9) vmovaps %ymm4, 0x20(%rdx,%r9) vmovaps %ymm6, 0x40(%rdx,%r9) vmovaps %ymm1, 0x60(%rdx,%r9) addl $0x4, %r8d subq $-0x80, %r9 jmp 0x2a4d7b vmovaps 0x20(%r15,%r9), %ymm1 vmovaps 0x40(%r15,%r9), %ymm3 vmovaps (%rbx,%r11), %ymm5 vmovaps 0x20(%rbx,%r11), %ymm0 vmovaps 0x40(%rbx,%r11), %ymm4 vmulps (%r15,%r9), %ymm5, %ymm6 vmovaps 0x60(%rbx,%r11), %ymm2 vaddps %ymm6, %ymm15, %ymm6 vmulps %ymm5, %ymm1, %ymm7 vaddps %ymm7, %ymm15, %ymm7 vmulps %ymm0, %ymm1, %ymm1 vmulps %ymm0, %ymm3, %ymm8 vmulps %ymm4, %ymm3, %ymm3 vaddps %ymm3, %ymm1, %ymm1 vaddps %ymm1, %ymm6, %ymm6 vmulps 0x60(%r15,%r9), %ymm4, %ymm1 vaddps %ymm1, %ymm8, %ymm1 vaddps %ymm1, %ymm7, %ymm7 vmovaps 0x80(%rbx,%r11), %ymm3 vmovaps 0xa0(%rbx,%r11), %ymm1 vmovaps (%r12,%r9), %ymm8 vmovaps 0x20(%r12,%r9), %ymm9 vmovaps 0x40(%r12,%r9), %ymm10 vmovaps 0x60(%r12,%r9), %ymm11 vmulps %ymm2, %ymm8, %ymm12 vmulps %ymm2, %ymm9, %ymm13 vmulps %ymm5, %ymm8, %ymm8 vaddps %ymm15, %ymm8, %ymm8 vmulps %ymm5, %ymm9, %ymm5 vaddps %ymm5, %ymm15, %ymm5 vmulps %ymm3, %ymm9, %ymm14 vaddps %ymm14, %ymm12, %ymm12 vmulps %ymm3, %ymm10, %ymm14 vaddps %ymm14, %ymm13, %ymm13 vmulps %ymm0, %ymm9, %ymm9 vmulps %ymm0, %ymm10, %ymm14 vmulps %ymm1, %ymm10, %ymm0 vaddps %ymm0, %ymm12, %ymm0 vaddps %ymm0, %ymm6, %ymm0 vmulps %ymm1, %ymm11, %ymm6 vaddps %ymm6, %ymm13, %ymm6 vaddps %ymm6, %ymm7, %ymm6 vmulps %ymm4, %ymm10, %ymm7 vaddps %ymm7, %ymm9, %ymm7 vaddps %ymm7, %ymm8, %ymm7 vmulps %ymm4, %ymm11, %ymm4 vaddps %ymm4, %ymm14, %ymm4 vaddps %ymm4, %ymm5, %ymm8 vmovaps 0xc0(%rbx,%r11), %ymm5 vmovaps 0xe0(%rbx,%r11), %ymm4 vmovaps (%r13,%r9), %ymm9 vmovaps 0x20(%r13,%r9), %ymm10 vmovaps 0x40(%r13,%r9), %ymm11 vmulps %ymm5, %ymm9, %ymm12 vmulps %ymm5, %ymm10, %ymm13 vmulps %ymm4, %ymm10, %ymm14 vaddps %ymm14, %ymm12, %ymm12 vmulps %ymm4, %ymm11, %ymm14 vaddps %ymm14, %ymm13, %ymm13 vmulps %ymm2, %ymm9, %ymm9 vmulps %ymm2, %ymm10, %ymm2 vmulps %ymm3, %ymm10, %ymm10 vaddps %ymm10, %ymm9, %ymm9 vmovaps 0x100(%rbx,%r11), %ymm10 vmulps %ymm3, %ymm11, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps %ymm10, %ymm11, %ymm3 vaddps %ymm3, %ymm12, %ymm3 vmovaps 0x60(%r13,%r9), %ymm12 vaddps %ymm3, %ymm0, %ymm0 vmulps %ymm10, %ymm12, %ymm3 vaddps %ymm3, %ymm13, %ymm3 vaddps %ymm3, %ymm6, %ymm3 vmulps %ymm1, %ymm11, %ymm6 vaddps %ymm6, %ymm9, %ymm6 vaddps %ymm6, %ymm7, %ymm6 vmulps %ymm1, %ymm12, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vaddps %ymm1, %ymm8, %ymm1 vmulps (%rsi,%r9), %ymm5, %ymm2 vmovaps 0x20(%rsi,%r9), %ymm7 vmulps %ymm5, %ymm7, %ymm5 vmulps %ymm4, %ymm7, %ymm7 vaddps %ymm7, %ymm2, %ymm2 vmovaps 0x40(%rsi,%r9), %ymm7 vmulps %ymm4, %ymm7, %ymm4 vaddps %ymm4, %ymm5, %ymm4 vmulps %ymm7, %ymm10, %ymm5 vaddps %ymm5, %ymm2, %ymm2 vaddps %ymm2, %ymm6, %ymm2 vmulps 0x60(%rsi,%r9), %ymm10, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vaddps %ymm4, %ymm1, %ymm1 vmovaps %ymm0, (%r10,%r9) vmovaps %ymm3, 0x20(%r10,%r9) vmovaps %ymm2, (%rdx,%r9) vmovaps %ymm1, 0x20(%rdx,%r9) addl $0x2, %r8d addq $0x40, %r9 leal 0x1(%r8), %r14d cmpl %eax, %r14d jl 0x2a51d2 movq 0x10(%rsp), %r14 cmpl %eax, %r8d jge 0x2a5510 vmovaps (%rbx,%r11), %ymm0 vmovaps 0x20(%rbx,%r11), %ymm1 vmovaps 0x40(%rbx,%r11), %ymm2 vmulps (%r15,%r9), %ymm0, %ymm3 vmovaps 0x60(%rbx,%r11), %ymm4 vaddps %ymm3, %ymm15, %ymm3 vmulps 0x20(%r15,%r9), %ymm1, %ymm5 vmulps 0x40(%r15,%r9), %ymm2, %ymm6 vaddps %ymm6, %ymm5, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmovaps 0x80(%rbx,%r11), %ymm5 vmovaps 0xa0(%rbx,%r11), %ymm6 vmovaps (%r12,%r9), %ymm7 vmovaps 0x20(%r12,%r9), %ymm8 vmovaps 0x40(%r12,%r9), %ymm9 vmulps %ymm4, %ymm7, %ymm10 vmulps %ymm0, %ymm7, %ymm0 vaddps %ymm0, %ymm15, %ymm0 vmulps %ymm5, %ymm8, %ymm7 vaddps %ymm7, %ymm10, %ymm7 vmulps %ymm1, %ymm8, %ymm1 vmulps %ymm6, %ymm9, %ymm8 vaddps %ymm7, %ymm8, %ymm7 vaddps %ymm7, %ymm3, %ymm3 vmulps %ymm2, %ymm9, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps 0xc0(%rbx,%r11), %ymm1 vmovaps 0xe0(%rbx,%r11), %ymm2 vmovaps 0x100(%rbx,%r11), %ymm7 vmovaps (%r13,%r9), %ymm8 vmovaps 0x20(%r13,%r9), %ymm9 vmovaps 0x40(%r13,%r9), %ymm10 vmulps %ymm1, %ymm8, %ymm11 vmulps %ymm4, %ymm8, %ymm4 vmulps %ymm2, %ymm9, %ymm8 vaddps %ymm8, %ymm11, %ymm8 vmulps %ymm5, %ymm9, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps %ymm7, %ymm10, %ymm5 vaddps %ymm5, %ymm8, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmulps %ymm6, %ymm10, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps (%rsi,%r9), %ymm1, %ymm1 vaddps %ymm4, %ymm0, %ymm0 vmulps 0x20(%rsi,%r9), %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x40(%rsi,%r9), %ymm7, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps %ymm3, (%r10,%r9) vmovaps %ymm0, (%rdx,%r9) incl %r8d addq $0x20, %r9 jmp 0x2a53d8 addq %rdi, %r15 addq %r9, %r15 addq %rdi, %r12 addq %r9, %r12 addq %rdi, %r13 addq %r9, %r13 addq %rdi, %rsi addq %r9, %rsi movq 0x190(%rsp), %r8 addq %r8, %r10 addq %r9, %r10 addq %r8, %rdx addq %r9, %rdx addl $0x2, %ebp jmp 0x2a4d65 xorl %edx, %edx xorl %esi, %esi leal 0x3(%rsi), %r8d cmpl %eax, %r8d jge 0x2a5891 vmovaps 0x20(%r15,%rdx), %ymm1 vmovaps 0x40(%r15,%rdx), %ymm2 vmovaps 0x60(%r15,%rdx), %ymm3 vmovaps 0x80(%r15,%rdx), %ymm4 vmovaps (%rbx,%r11), %ymm5 vmovaps 0x20(%rbx,%r11), %ymm6 vmovaps 0x40(%rbx,%r11), %ymm7 vmulps (%r15,%rdx), %ymm5, %ymm8 vmovaps 0x60(%rbx,%r11), %ymm0 vaddps %ymm15, %ymm8, %ymm8 vmulps %ymm5, %ymm1, %ymm9 vaddps %ymm15, %ymm9, %ymm9 vmulps %ymm5, %ymm2, %ymm10 vaddps %ymm15, %ymm10, %ymm10 vmulps %ymm5, %ymm3, %ymm5 vaddps %ymm5, %ymm15, %ymm5 vmulps %ymm6, %ymm1, %ymm1 vmulps %ymm6, %ymm2, %ymm11 vmulps %ymm6, %ymm3, %ymm12 vmulps %ymm6, %ymm4, %ymm6 vmulps %ymm7, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm8, %ymm1 vmulps %ymm7, %ymm3, %ymm2 vaddps %ymm2, %ymm11, %ymm2 vaddps %ymm2, %ymm9, %ymm2 vmulps %ymm7, %ymm4, %ymm3 vaddps %ymm3, %ymm12, %ymm3 vaddps %ymm3, %ymm10, %ymm3 vmulps 0xa0(%r15,%rdx), %ymm7, %ymm4 vaddps %ymm4, %ymm6, %ymm4 vaddps %ymm4, %ymm5, %ymm4 vmovaps 0x80(%rbx,%r11), %ymm5 vmovaps 0xa0(%rbx,%r11), %ymm6 vmovaps 0x20(%r12,%rdx), %ymm7 vmovaps 0x40(%r12,%rdx), %ymm8 vmovaps 0x60(%r12,%rdx), %ymm9 vmulps (%r12,%rdx), %ymm0, %ymm10 vmovaps 0x80(%r12,%rdx), %ymm11 vmulps %ymm0, %ymm7, %ymm12 vmulps %ymm0, %ymm8, %ymm13 vmulps %ymm0, %ymm9, %ymm0 vmulps %ymm5, %ymm7, %ymm7 vaddps %ymm7, %ymm10, %ymm7 vmulps %ymm5, %ymm8, %ymm10 vaddps %ymm10, %ymm12, %ymm10 vmulps %ymm5, %ymm9, %ymm12 vaddps %ymm12, %ymm13, %ymm12 vmulps %ymm5, %ymm11, %ymm5 vaddps %ymm5, %ymm0, %ymm0 vmulps %ymm6, %ymm8, %ymm5 vaddps %ymm5, %ymm7, %ymm5 vaddps %ymm5, %ymm1, %ymm1 vmulps %ymm6, %ymm9, %ymm5 vaddps %ymm5, %ymm10, %ymm5 vaddps %ymm5, %ymm2, %ymm2 vmulps %ymm6, %ymm11, %ymm5 vaddps %ymm5, %ymm12, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmulps 0xa0(%r12,%rdx), %ymm6, %ymm5 vaddps %ymm5, %ymm0, %ymm0 vaddps %ymm0, %ymm4, %ymm0 vmovaps 0xc0(%rbx,%r11), %ymm4 vmovaps 0xe0(%rbx,%r11), %ymm5 vmovaps 0x100(%rbx,%r11), %ymm6 vmovaps 0x20(%r13,%rdx), %ymm7 vmovaps 0x40(%r13,%rdx), %ymm8 vmovaps 0x60(%r13,%rdx), %ymm9 vmovaps 0x80(%r13,%rdx), %ymm10 vmulps (%r13,%rdx), %ymm4, %ymm11 vmulps %ymm4, %ymm7, %ymm12 vmulps %ymm4, %ymm8, %ymm13 vmulps %ymm4, %ymm9, %ymm4 vmulps %ymm5, %ymm7, %ymm7 vaddps %ymm7, %ymm11, %ymm7 vmulps %ymm5, %ymm8, %ymm11 vaddps %ymm11, %ymm12, %ymm11 vmulps %ymm5, %ymm9, %ymm12 vaddps %ymm12, %ymm13, %ymm12 vmulps %ymm5, %ymm10, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps %ymm6, %ymm8, %ymm5 vaddps %ymm5, %ymm7, %ymm5 vaddps %ymm5, %ymm1, %ymm1 vmulps %ymm6, %ymm9, %ymm5 vaddps %ymm5, %ymm11, %ymm5 vaddps %ymm5, %ymm2, %ymm2 vmulps %ymm6, %ymm10, %ymm5 vaddps %ymm5, %ymm12, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmulps 0xa0(%r13,%rdx), %ymm6, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmovaps %ymm1, (%r10,%rdx) vmovaps %ymm2, 0x20(%r10,%rdx) vmovaps %ymm3, 0x40(%r10,%rdx) vmovaps %ymm0, 0x60(%r10,%rdx) addl $0x4, %esi subq $-0x80, %rdx jmp 0x2a5548 vmovaps 0x20(%r15,%rdx), %ymm0 vmovaps 0x40(%r15,%rdx), %ymm1 vmovaps (%rbx,%r11), %ymm2 vmovaps 0x20(%rbx,%r11), %ymm3 vmovaps 0x40(%rbx,%r11), %ymm4 vmovaps 0x60(%rbx,%r11), %ymm5 vmulps (%r15,%rdx), %ymm2, %ymm6 vaddps %ymm6, %ymm15, %ymm6 vmulps %ymm2, %ymm0, %ymm2 vaddps %ymm2, %ymm15, %ymm2 vmulps %ymm3, %ymm0, %ymm0 vmulps %ymm3, %ymm1, %ymm3 vmulps %ymm4, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmulps 0x60(%r15,%rdx), %ymm4, %ymm1 vaddps %ymm0, %ymm6, %ymm0 vaddps %ymm1, %ymm3, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vmovaps 0x80(%rbx,%r11), %ymm2 vmovaps 0xa0(%rbx,%r11), %ymm3 vmovaps 0x20(%r12,%rdx), %ymm4 vmulps (%r12,%rdx), %ymm5, %ymm6 vmovaps 0x40(%r12,%rdx), %ymm7 vmulps %ymm5, %ymm4, %ymm5 vmulps %ymm2, %ymm4, %ymm4 vaddps %ymm4, %ymm6, %ymm4 vmulps %ymm2, %ymm7, %ymm2 vaddps %ymm2, %ymm5, %ymm2 vmulps %ymm3, %ymm7, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmulps 0x60(%r12,%rdx), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0xc0(%rbx,%r11), %ymm2 vmovaps 0xe0(%rbx,%r11), %ymm3 vmovaps 0x100(%rbx,%r11), %ymm4 vmovaps 0x20(%r13,%rdx), %ymm5 vmovaps 0x40(%r13,%rdx), %ymm6 vmulps (%r13,%rdx), %ymm2, %ymm7 vmulps %ymm2, %ymm5, %ymm2 vmulps %ymm3, %ymm5, %ymm5 vaddps %ymm5, %ymm7, %ymm5 vmulps %ymm3, %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps %ymm4, %ymm6, %ymm3 vaddps %ymm3, %ymm5, %ymm3 vaddps %ymm3, %ymm0, %ymm0 vmulps 0x60(%r13,%rdx), %ymm4, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps %ymm0, (%r10,%rdx) vmovaps %ymm1, 0x20(%r10,%rdx) addl $0x2, %esi addq $0x40, %rdx leal 0x1(%rsi), %r8d cmpl %eax, %r8d jl 0x2a5766 jmp 0x2a595d vmovaps (%r15,%rdx), %ymm0 vmovaps 0x20(%r15,%rdx), %ymm1 vmovaps 0x40(%r15,%rdx), %ymm2 vmulps (%rbx,%r11), %ymm0, %ymm0 vmulps 0x20(%rbx,%r11), %ymm1, %ymm1 vaddps %ymm0, %ymm15, %ymm0 vmulps 0x40(%rbx,%r11), %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps (%r12,%rdx), %ymm1 vmovaps 0x20(%r12,%rdx), %ymm2 vmovaps 0x40(%r12,%rdx), %ymm3 vmulps 0x60(%rbx,%r11), %ymm1, %ymm1 vmulps 0x80(%rbx,%r11), %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0xa0(%rbx,%r11), %ymm3, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps (%r13,%rdx), %ymm1 vmovaps 0x20(%r13,%rdx), %ymm2 vmovaps 0x40(%r13,%rdx), %ymm3 vmulps 0xc0(%rbx,%r11), %ymm1, %ymm1 vmulps 0xe0(%rbx,%r11), %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x100(%rbx,%r11), %ymm3, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps %ymm0, (%r10,%rdx) incl %esi addq $0x20, %rdx cmpl %eax, %esi jl 0x2a58a3 addq %rdx, %r15 addq $0x40, %r15 addq %rdx, %r12 addq $0x40, %r12 addq %rdx, %r13 addq $0x40, %r13 incl %ebp addq %rdx, %r10 cmpl %ecx, %ebp jl 0x2a5544 movq 0xe0(%rsp), %r9 incq %r9 movq 0x18(%rsp), %r13 movq 0x150(%rsp), %r8 jmp 0x2a4cc3 movl 0xd4(%rdi), %eax movl 0xd8(%rdi), %r15d cmpl $0x5, %eax je 0x2a6998 cmpl $0x3, %eax jne 0x2a841c cmpl $0x3, %r15d jne 0x2a841c cmpl $0x1, 0xdc(%rdi) jne 0x2a7b18 cmpl $0x1, 0xe0(%rdi) jne 0x2a7b18 cmpl $0x1, 0xe4(%rdi) jne 0x2a7b18 cmpl $0x1, 0xe8(%rdi) jne 0x2a7b18 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xc8(%rsp), %edx movq 0x1b0(%rdi), %rsi xorl %edi, %edi testl %ecx, %ecx cmovlel %edi, %ecx testl %edx, %edx cmovlel %edi, %edx movq %rdx, 0x2b8(%rsp) movl $0x80, %r8d movq %rsi, 0x2b0(%rsp) cmpq 0x2b8(%rsp), %rdi je 0x2a9725 testq %rsi, %rsi je 0x2a5a5a movq %rdi, %rdx shlq $0x4, %rdx vmovups (%rsi,%rdx), %xmm0 jmp 0x2a5a5e vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) movq 0x40(%r13), %r9 imulq %rdi, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %rdx movslq 0x54(%r14), %rsi imulq %rdi, %rsi imulq 0x38(%r14), %rsi movslq 0xbc(%rsp), %r10 movq 0xd0(%rsp), %r13 imulq %rdi, %r13 movq 0xa0(%rsp), %r11 imulq %r11, %r13 addq 0x90(%rsp), %r13 imulq %r11, %r10 leaq (%r10,%r13), %rbp leaq (,%r10,2), %r11 addq %r13, %r11 vmovaps (%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x50(%rsp) vmovaps 0x10(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x70(%rsp) vmovaps 0x20(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x110(%rsp) vmovaps 0x30(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0xf0(%rsp) vmovaps 0x40(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x130(%rsp) vmovaps 0x50(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x170(%rsp) vmovaps 0x60(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x190(%rsp) vmovaps 0x70(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x270(%rsp) vmovaps 0x80(%rdx,%rsi), %xmm0 vmovaps %xmm0, 0x1f0(%rsp) xorl %r10d, %r10d cmpl %ecx, %r10d je 0x2a661f leaq 0x20(%r13), %r15 leaq 0x20(%rbp), %rbx leaq 0x20(%r11), %rdx xorl %esi, %esi xorl %r12d, %r12d leal 0x7(%r12), %r14d cmpl %eax, %r14d jge 0x2a63ad vmovaps 0x50(%rsp), %xmm0 vmulps (%r13,%rsi), %xmm0, %xmm0 vaddps 0x20(%rsp), %xmm0, %xmm1 vmovaps 0x10(%r13,%rsi), %xmm3 vmovaps 0x20(%r13,%rsi), %xmm0 vmovaps %xmm0, 0xe0(%rsp) vmovaps 0x30(%r13,%rsi), %xmm10 vmulps 0x70(%rsp), %xmm3, %xmm2 vmulps 0x110(%rsp), %xmm0, %xmm4 vaddps %xmm4, %xmm2, %xmm2 vmovaps 0xf0(%rsp), %xmm0 vmulps (%rbp,%rsi), %xmm0, %xmm4 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0x10(%rbp,%rsi), %xmm13 vmovaps 0x20(%rbp,%rsi), %xmm0 vmovaps %xmm0, 0x260(%rsp) vmovaps 0x30(%rbp,%rsi), %xmm2 vmovaps %xmm2, 0x160(%rsp) vmovaps 0x130(%rsp), %xmm5 vmulps %xmm5, %xmm13, %xmm12 vmovaps %xmm5, %xmm9 vaddps %xmm4, %xmm12, %xmm4 vmulps 0x170(%rsp), %xmm0, %xmm12 vaddps %xmm4, %xmm12, %xmm4 vaddps %xmm4, %xmm1, %xmm1 vmovaps 0x190(%rsp), %xmm11 vmulps (%r11,%rsi), %xmm11, %xmm14 vmovaps 0x10(%r11,%rsi), %xmm15 vmovaps 0x20(%r11,%rsi), %xmm0 vmovaps 0x30(%r11,%rsi), %xmm12 vmovaps 0x270(%rsp), %xmm7 vmulps %xmm7, %xmm15, %xmm5 vaddps %xmm5, %xmm14, %xmm5 vmovaps 0x1f0(%rsp), %xmm8 vmulps %xmm0, %xmm8, %xmm14 vaddps %xmm5, %xmm14, %xmm5 vaddps %xmm5, %xmm1, %xmm1 vmovaps %xmm1, (%r9,%rsi) vmulps 0x50(%rsp), %xmm3, %xmm1 vaddps 0x20(%rsp), %xmm1, %xmm1 vmovaps 0xe0(%rsp), %xmm2 vmulps 0x70(%rsp), %xmm2, %xmm3 vmovaps 0x110(%rsp), %xmm2 vmovaps %xmm10, %xmm14 vmovaps %xmm10, 0x2a0(%rsp) vmulps %xmm2, %xmm10, %xmm5 vmovaps 0xf0(%rsp), %xmm4 vmulps %xmm4, %xmm13, %xmm13 vaddps %xmm3, %xmm13, %xmm3 vaddps %xmm3, %xmm1, %xmm1 vmovaps %xmm1, 0x150(%rsp) vmovaps %xmm9, %xmm10 vmovaps 0x260(%rsp), %xmm9 vmulps %xmm10, %xmm9, %xmm3 vmovaps 0x160(%rsp), %xmm1 vmovaps 0x170(%rsp), %xmm6 vmulps %xmm6, %xmm1, %xmm13 vaddps %xmm5, %xmm13, %xmm5 vmulps %xmm11, %xmm15, %xmm13 vaddps %xmm3, %xmm13, %xmm3 vmulps %xmm7, %xmm0, %xmm13 vmovaps %xmm0, %xmm11 vaddps %xmm3, %xmm13, %xmm3 vaddps 0x150(%rsp), %xmm3, %xmm7 vmulps %xmm8, %xmm12, %xmm3 vmovaps %xmm12, 0x290(%rsp) vaddps %xmm3, %xmm5, %xmm3 vaddps %xmm3, %xmm7, %xmm5 vmovaps 0x40(%r13,%rsi), %xmm3 vmovaps 0x40(%rbp,%rsi), %xmm7 vmovaps 0x40(%r11,%rsi), %xmm8 vmovaps %xmm5, 0x10(%r9,%rsi) vmovaps 0xe0(%rsp), %xmm0 vmulps 0x50(%rsp), %xmm0, %xmm0 vaddps 0x20(%rsp), %xmm0, %xmm0 vmulps 0x70(%rsp), %xmm14, %xmm5 vmulps %xmm2, %xmm3, %xmm14 vmulps %xmm4, %xmm9, %xmm2 vmovaps %xmm10, %xmm13 vmulps %xmm1, %xmm10, %xmm15 vaddps %xmm5, %xmm15, %xmm5 vmulps %xmm6, %xmm7, %xmm15 vmovaps %xmm7, 0x150(%rsp) vaddps %xmm15, %xmm14, %xmm14 vmovaps 0x190(%rsp), %xmm1 vmulps %xmm1, %xmm11, %xmm4 vmovaps %xmm1, %xmm9 vaddps %xmm4, %xmm2, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x270(%rsp), %xmm10 vmulps %xmm10, %xmm12, %xmm2 vaddps %xmm2, %xmm5, %xmm2 vaddps %xmm2, %xmm0, %xmm4 vmovaps %xmm8, %xmm0 vmovaps %xmm8, 0x350(%rsp) vmovaps 0x1f0(%rsp), %xmm6 vmulps %xmm6, %xmm8, %xmm2 vaddps %xmm2, %xmm14, %xmm2 vaddps %xmm2, %xmm4, %xmm5 vmovaps 0x50(%r13,%rsi), %xmm1 vmovaps %xmm1, 0x260(%rsp) vmovaps 0x50(%rbp,%rsi), %xmm4 vmovaps %xmm4, 0x360(%rsp) vmovaps 0x50(%r11,%rsi), %xmm2 vmovaps %xmm2, 0xe0(%rsp) vmovaps %xmm5, 0x20(%r9,%rsi) vmovaps 0x2a0(%rsp), %xmm2 vmulps 0x50(%rsp), %xmm2, %xmm5 vaddps 0x20(%rsp), %xmm5, %xmm5 vmulps 0x70(%rsp), %xmm3, %xmm12 vmovaps 0x110(%rsp), %xmm2 vmulps %xmm2, %xmm1, %xmm14 vmovaps 0x160(%rsp), %xmm1 vmulps 0xf0(%rsp), %xmm1, %xmm11 vmulps %xmm7, %xmm13, %xmm15 vaddps %xmm15, %xmm12, %xmm8 vmovaps 0x170(%rsp), %xmm13 vmulps %xmm4, %xmm13, %xmm15 vaddps %xmm15, %xmm14, %xmm14 vmulps 0x290(%rsp), %xmm9, %xmm12 vmovaps %xmm9, %xmm15 vaddps %xmm12, %xmm11, %xmm11 vaddps %xmm5, %xmm11, %xmm5 vmulps %xmm0, %xmm10, %xmm11 vaddps %xmm11, %xmm8, %xmm10 vaddps %xmm5, %xmm10, %xmm5 vmovaps 0xe0(%rsp), %xmm0 vmulps %xmm6, %xmm0, %xmm10 vaddps %xmm10, %xmm14, %xmm10 vaddps %xmm5, %xmm10, %xmm5 vmovaps 0x60(%r13,%rsi), %xmm8 vmovaps 0x60(%rbp,%rsi), %xmm1 vmovaps %xmm1, 0x290(%rsp) vmovaps 0x60(%r11,%rsi), %xmm14 vmovaps %xmm5, 0x30(%r9,%rsi) vmulps 0x50(%rsp), %xmm3, %xmm3 vaddps 0x20(%rsp), %xmm3, %xmm3 vmovaps 0x260(%rsp), %xmm7 vmulps 0x70(%rsp), %xmm7, %xmm5 vmulps %xmm2, %xmm8, %xmm10 vmovaps %xmm8, 0x160(%rsp) vmovaps %xmm2, %xmm11 vmovaps 0xf0(%rsp), %xmm4 vmulps 0x150(%rsp), %xmm4, %xmm2 vmovaps 0x130(%rsp), %xmm6 vmovaps 0x360(%rsp), %xmm9 vmulps %xmm6, %xmm9, %xmm12 vaddps %xmm5, %xmm12, %xmm5 vmulps %xmm1, %xmm13, %xmm12 vaddps %xmm12, %xmm10, %xmm10 vmulps 0x350(%rsp), %xmm15, %xmm12 vaddps %xmm2, %xmm12, %xmm1 vaddps %xmm1, %xmm3, %xmm1 vmovaps 0x270(%rsp), %xmm12 vmulps %xmm0, %xmm12, %xmm3 vaddps %xmm3, %xmm5, %xmm3 vaddps %xmm3, %xmm1, %xmm1 vmovaps 0x1f0(%rsp), %xmm2 vmulps %xmm2, %xmm14, %xmm3 vaddps %xmm3, %xmm10, %xmm3 vaddps %xmm3, %xmm1, %xmm1 vmovaps 0x70(%r13,%rsi), %xmm5 vmovaps %xmm5, 0x2a0(%rsp) vmovaps 0x70(%rbp,%rsi), %xmm0 vmovaps 0x70(%r11,%rsi), %xmm10 vmovaps %xmm1, 0x40(%r9,%rsi) vmulps 0x50(%rsp), %xmm7, %xmm1 vaddps 0x20(%rsp), %xmm1, %xmm1 vmulps 0x70(%rsp), %xmm8, %xmm3 vmulps %xmm5, %xmm11, %xmm7 vmulps %xmm4, %xmm9, %xmm13 vmovaps %xmm4, %xmm11 vmovaps 0x290(%rsp), %xmm9 vmulps %xmm6, %xmm9, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmovaps 0x170(%rsp), %xmm8 vmulps %xmm0, %xmm8, %xmm5 vmovaps %xmm0, %xmm4 vmovaps %xmm0, 0x150(%rsp) vaddps %xmm5, %xmm7, %xmm5 vmulps 0xe0(%rsp), %xmm15, %xmm0 vaddps %xmm0, %xmm13, %xmm0 vaddps %xmm0, %xmm1, %xmm0 vmulps %xmm12, %xmm14, %xmm1 vaddps %xmm1, %xmm3, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmulps %xmm2, %xmm10, %xmm1 vmovaps %xmm10, %xmm2 vaddps %xmm1, %xmm5, %xmm1 vaddps %xmm1, %xmm0, %xmm3 vmovaps 0x80(%r13,%rsi), %xmm1 vmovaps 0x80(%rbp,%rsi), %xmm10 vmovaps 0x80(%r11,%rsi), %xmm0 vmovaps %xmm0, 0xe0(%rsp) vmovaps %xmm3, 0x50(%r9,%rsi) vmovaps 0x2a0(%rsp), %xmm7 vmulps 0x70(%rsp), %xmm7, %xmm3 vmovaps 0x110(%rsp), %xmm13 vmulps %xmm1, %xmm13, %xmm0 vmulps %xmm6, %xmm4, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmulps %xmm8, %xmm10, %xmm5 vaddps %xmm5, %xmm0, %xmm4 vmovaps %xmm11, %xmm0 vmulps %xmm11, %xmm9, %xmm5 vmulps %xmm15, %xmm14, %xmm14 vmovaps %xmm15, %xmm6 vaddps %xmm5, %xmm14, %xmm5 vmovaps 0x50(%rsp), %xmm14 vmulps 0x160(%rsp), %xmm14, %xmm11 vaddps 0x20(%rsp), %xmm11, %xmm11 vaddps %xmm5, %xmm11, %xmm5 vmulps %xmm2, %xmm12, %xmm11 vmovaps %xmm2, %xmm15 vaddps %xmm3, %xmm11, %xmm3 vaddps %xmm3, %xmm5, %xmm3 vmovaps 0x1f0(%rsp), %xmm2 vmovaps 0xe0(%rsp), %xmm9 vmulps %xmm2, %xmm9, %xmm5 vaddps %xmm5, %xmm4, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps 0x90(%r13,%rsi), %xmm13, %xmm4 vmulps 0x90(%rbp,%rsi), %xmm8, %xmm5 vaddps %xmm5, %xmm4, %xmm4 vmulps 0x90(%r11,%rsi), %xmm2, %xmm5 vaddps %xmm5, %xmm4, %xmm4 vmovaps %xmm3, 0x60(%r9,%rsi) vmulps 0x70(%rsp), %xmm1, %xmm2 vmulps 0x130(%rsp), %xmm10, %xmm1 vaddps %xmm1, %xmm2, %xmm1 vmulps 0x150(%rsp), %xmm0, %xmm2 vmulps %xmm6, %xmm15, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps %xmm7, %xmm14, %xmm3 vaddps 0x20(%rsp), %xmm3, %xmm3 vaddps %xmm2, %xmm3, %xmm2 vmulps %xmm12, %xmm9, %xmm0 vaddps %xmm0, %xmm1, %xmm0 vaddps %xmm0, %xmm2, %xmm0 vaddps %xmm4, %xmm0, %xmm0 vmovaps %xmm0, 0x70(%r9,%rsi) addl $0x8, %r12d addq %r8, %rsi addq %r8, %r15 addq %r8, %rbx addq %r8, %rdx jmp 0x2a5b61 vmovaps 0x50(%rsp), %xmm0 vmulps (%r13,%rsi), %xmm0, %xmm0 vaddps 0x20(%rsp), %xmm0, %xmm5 vmovaps 0x10(%r13,%rsi), %xmm1 vmovaps 0x20(%r13,%rsi), %xmm0 vmovaps 0x30(%r13,%rsi), %xmm2 vmovaps %xmm2, 0xe0(%rsp) vmulps 0x70(%rsp), %xmm1, %xmm2 vmovaps 0x110(%rsp), %xmm12 vmulps %xmm0, %xmm12, %xmm3 vmovaps %xmm0, %xmm10 vmovaps %xmm0, 0x150(%rsp) vaddps %xmm3, %xmm2, %xmm2 vmovaps 0xf0(%rsp), %xmm0 vmulps (%rbp,%rsi), %xmm0, %xmm3 vaddps %xmm2, %xmm5, %xmm2 vmovaps 0x10(%rbp,%rsi), %xmm4 vmovaps 0x20(%rbp,%rsi), %xmm13 vmovaps 0x30(%rbp,%rsi), %xmm5 vmovaps %xmm5, 0x160(%rsp) vmulps 0x130(%rsp), %xmm4, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmovaps 0x170(%rsp), %xmm9 vmulps %xmm9, %xmm13, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm3 vmovaps 0x190(%rsp), %xmm11 vmulps (%r11,%rsi), %xmm11, %xmm5 vmovaps 0x10(%r11,%rsi), %xmm14 vmovaps 0x20(%r11,%rsi), %xmm8 vmovaps 0x30(%r11,%rsi), %xmm2 vmovaps 0x270(%rsp), %xmm6 vmulps %xmm6, %xmm14, %xmm15 vaddps %xmm5, %xmm15, %xmm5 vmovaps 0x1f0(%rsp), %xmm7 vmulps %xmm7, %xmm8, %xmm15 vaddps %xmm5, %xmm15, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmovaps %xmm3, (%r9,%rsi) vmulps 0x50(%rsp), %xmm1, %xmm1 vaddps 0x20(%rsp), %xmm1, %xmm1 vmulps 0x70(%rsp), %xmm10, %xmm3 vmovaps 0xe0(%rsp), %xmm15 vmulps %xmm12, %xmm15, %xmm5 vmulps %xmm0, %xmm4, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm1, %xmm10 vmovaps 0x130(%rsp), %xmm1 vmulps %xmm1, %xmm13, %xmm3 vmovaps 0x160(%rsp), %xmm0 vmulps %xmm0, %xmm9, %xmm4 vaddps %xmm4, %xmm5, %xmm4 vmulps %xmm11, %xmm14, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vmulps %xmm6, %xmm8, %xmm5 vaddps %xmm5, %xmm3, %xmm3 vaddps %xmm3, %xmm10, %xmm5 vmulps %xmm7, %xmm2, %xmm3 vmovaps %xmm2, %xmm10 vaddps %xmm3, %xmm4, %xmm3 vaddps %xmm3, %xmm5, %xmm5 vmovaps 0x40(%r13,%rsi), %xmm4 vmovaps 0x40(%rbp,%rsi), %xmm3 vmovaps 0x40(%r11,%rsi), %xmm2 vmovaps %xmm2, 0x260(%rsp) vmovaps %xmm5, 0x10(%r9,%rsi) vmulps 0x70(%rsp), %xmm15, %xmm5 vmulps %xmm4, %xmm12, %xmm14 vmulps %xmm1, %xmm0, %xmm15 vaddps %xmm5, %xmm15, %xmm5 vmulps %xmm3, %xmm9, %xmm15 vaddps %xmm15, %xmm14, %xmm14 vmovaps 0xf0(%rsp), %xmm15 vmulps %xmm15, %xmm13, %xmm0 vmulps %xmm11, %xmm8, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x150(%rsp), %xmm2 vmulps 0x50(%rsp), %xmm2, %xmm2 vmovaps 0x20(%rsp), %xmm13 vaddps %xmm2, %xmm13, %xmm2 vaddps %xmm0, %xmm2, %xmm0 vmulps %xmm6, %xmm10, %xmm2 vaddps %xmm2, %xmm5, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x260(%rsp), %xmm8 vmulps %xmm7, %xmm8, %xmm2 vaddps %xmm2, %xmm14, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmulps 0x50(%r13,%rsi), %xmm12, %xmm2 vmulps 0x50(%rbp,%rsi), %xmm9, %xmm5 vaddps %xmm5, %xmm2, %xmm2 vmulps 0x50(%r11,%rsi), %xmm7, %xmm5 vaddps %xmm5, %xmm2, %xmm2 vmovaps %xmm0, 0x20(%r9,%rsi) vmulps 0x70(%rsp), %xmm4, %xmm0 vmulps %xmm1, %xmm3, %xmm3 vaddps %xmm3, %xmm0, %xmm0 vmulps 0x160(%rsp), %xmm15, %xmm3 vmulps %xmm11, %xmm10, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmovaps 0xe0(%rsp), %xmm1 vmulps 0x50(%rsp), %xmm1, %xmm4 vmovaps %xmm13, 0x20(%rsp) vaddps %xmm4, %xmm13, %xmm4 vaddps %xmm3, %xmm4, %xmm3 vmulps %xmm6, %xmm8, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm0, %xmm3, %xmm0 vaddps %xmm2, %xmm0, %xmm0 vmovaps %xmm0, 0x30(%r9,%rsi) addl $0x4, %r12d addq $0x40, %rsi addq $0x40, %r15 addq $0x40, %rbx addq $0x40, %rdx leal 0x3(%r12), %r14d cmpl %eax, %r14d jl 0x2a6129 jmp 0x2a6534 vmovaps 0x10(%r13,%rsi), %xmm0 vmovaps %xmm0, 0xe0(%rsp) vmovaps 0x20(%r13,%rsi), %xmm4 vmovaps 0x10(%rbp,%rsi), %xmm5 vmovaps 0x20(%rbp,%rsi), %xmm3 vmovaps %xmm3, 0x160(%rsp) vmovaps 0x10(%r11,%rsi), %xmm2 vmovaps 0x50(%rsp), %xmm1 vmulps (%r13,%rsi), %xmm1, %xmm10 vmovaps 0x20(%r11,%rsi), %xmm1 vmovaps %xmm1, 0x150(%rsp) vmovaps 0x20(%rsp), %xmm14 vaddps %xmm14, %xmm10, %xmm10 vmovaps 0x70(%rsp), %xmm13 vmulps %xmm0, %xmm13, %xmm11 vmovaps 0x110(%rsp), %xmm6 vmulps %xmm6, %xmm4, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vaddps %xmm11, %xmm10, %xmm10 vmovaps 0xf0(%rsp), %xmm7 vmulps (%rbp,%rsi), %xmm7, %xmm11 vmovaps 0x130(%rsp), %xmm8 vmulps %xmm5, %xmm8, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vmovaps 0x170(%rsp), %xmm9 vmulps %xmm3, %xmm9, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vaddps %xmm11, %xmm10, %xmm10 vmovaps 0x190(%rsp), %xmm15 vmulps (%r11,%rsi), %xmm15, %xmm11 vmovaps 0x270(%rsp), %xmm3 vmulps %xmm3, %xmm2, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vmovaps 0x1f0(%rsp), %xmm0 vmulps %xmm0, %xmm1, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vaddps %xmm11, %xmm10, %xmm10 vmulps 0x30(%r13,%rsi), %xmm6, %xmm11 vmulps 0x30(%rbp,%rsi), %xmm9, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vmulps 0x30(%r11,%rsi), %xmm0, %xmm12 vaddps %xmm12, %xmm11, %xmm11 vmovaps %xmm10, (%r9,%rsi) vmulps %xmm4, %xmm13, %xmm4 vmulps %xmm7, %xmm5, %xmm5 vaddps %xmm5, %xmm4, %xmm4 vmovaps 0xe0(%rsp), %xmm0 vmulps 0x50(%rsp), %xmm0, %xmm0 vaddps %xmm0, %xmm14, %xmm0 vaddps %xmm4, %xmm0, %xmm0 vmulps 0x160(%rsp), %xmm8, %xmm1 vmulps %xmm2, %xmm15, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x150(%rsp), %xmm3, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm0, %xmm11, %xmm0 vmovaps %xmm0, 0x10(%r9,%rsi) addl $0x2, %r12d addq $0x20, %rsi addq $0x20, %r15 addq $0x20, %rbx addq $0x20, %rdx leal 0x1(%r12), %r14d cmpl %eax, %r14d jl 0x2a63c0 addq %rsi, %r9 movq 0x10(%rsp), %r14 cmpl %eax, %r12d jge 0x2a660e vmovaps 0x50(%rsp), %xmm0 vmulps -0x20(%r15), %xmm0, %xmm0 vaddps 0x20(%rsp), %xmm0, %xmm0 vmovaps 0x70(%rsp), %xmm1 vmulps -0x10(%r15), %xmm1, %xmm1 vmovaps 0x110(%rsp), %xmm2 vmulps (%r15), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmovaps 0xf0(%rsp), %xmm1 vmulps -0x20(%rbx), %xmm1, %xmm1 vmovaps 0x130(%rsp), %xmm2 vmulps -0x10(%rbx), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0x170(%rsp), %xmm2 vmulps (%rbx), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0x190(%rsp), %xmm2 vmulps -0x20(%rdx), %xmm2, %xmm2 vmovaps 0x270(%rsp), %xmm3 vmulps -0x10(%rdx), %xmm3, %xmm3 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm3, %xmm2, %xmm1 vmovaps 0x1f0(%rsp), %xmm2 vmulps (%rdx), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmovaps %xmm0, (%r9) addq $0x10, %r9 incl %r12d addq $0x10, %r15 addq $0x10, %rbx addq $0x10, %rdx jmp 0x2a654a incl %r10d movq %r15, %r13 movq %rbx, %rbp movq %rdx, %r11 jmp 0x2a5b47 incq %rdi movq 0x18(%rsp), %r13 movq 0x2b0(%rsp), %rsi jmp 0x2a5a39 cmpl $0x3, 0xd4(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x3, 0xd8(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x1, 0xdc(%rdi) movq %r9, 0x50(%rsp) jne 0x2a742a cmpl $0x1, 0xe0(%rdi) jne 0x2a742a cmpl $0x1, 0xe4(%rdi) jne 0x2a742a cmpl $0x1, 0xe8(%rdi) jne 0x2a742a movslq 0xbc(%rsp), %rax movl 0x2c(%r13), %edx movl 0x30(%r13), %r8d movl 0xc8(%rsp), %ecx movq 0x28(%r14), %r10 imulq 0x10(%r13), %r11 movq 0x1b0(%rdi), %rsi movslq %edx, %rdi movq 0x90(%rsp), %rbx movq %rbx, 0x1f0(%rsp) movq 0xa0(%rsp), %rbx imulq 0xd0(%rsp), %rbx movq %rbx, 0x270(%rsp) leaq (%rax,%rax), %rbx movq %rbx, 0x190(%rsp) leaq (%rax,%rax,2), %rbx movq %rbx, 0xe0(%rsp) movq %rax, 0x130(%rsp) addl $0x2, %eax movslq %eax, %rbp xorl %eax, %eax testl %ecx, %ecx cmovlel %eax, %ecx movq %rcx, 0xf0(%rsp) shlq $0x2, %rbp movq %rdi, 0x10(%rsp) leaq (,%rdi,4), %rcx movq %rcx, 0x20(%rsp) movq %rsi, 0x170(%rsp) movq %r11, 0x110(%rsp) cmpq 0xf0(%rsp), %rax je 0x2a9725 testq %rsi, %rsi je 0x2a6753 vmovss (%rsi,%rax,4), %xmm0 jmp 0x2a6757 vxorps %xmm0, %xmm0, %xmm0 movq %r11, %rsi imulq %rax, %rsi addq %r9, %rsi imulq $0x24, %rax, %r13 movq 0x270(%rsp), %r15 movq %rax, 0x70(%rsp) imulq %rax, %r15 addq 0x1f0(%rsp), %r15 movq 0x130(%rsp), %rax leaq (%r15,%rax,4), %rax movq 0x190(%rsp), %rcx leaq (%r15,%rcx,4), %r9 movq 0xe0(%rsp), %rcx leaq (%r15,%rcx,4), %r11 xorl %edi, %edi movq %rsi, %rbx movl %edi, %ecx orl $0x1, %ecx cmpl %r8d, %ecx jge 0x2a68d4 movq 0x10(%rsp), %rcx leaq (%rsi,%rcx,4), %rsi xorl %r12d, %r12d movl %edx, %ecx testl %ecx, %ecx jle 0x2a68a9 vmovsd 0x4(%r15,%r12), %xmm1 vmovss (%r15,%r12), %xmm2 vmovlhps %xmm1, %xmm2, %xmm2 # xmm2 = xmm2[0],xmm1[0] vshufps $0xd8, %xmm1, %xmm2, %xmm2 # xmm2 = xmm2[0,2],xmm1[1,3] vmovss (%rax,%r12), %xmm1 vinsertps $0x30, %xmm1, %xmm2, %xmm2 # xmm2 = xmm2[0,1,2],xmm1[0] vmovsd 0x4(%rax,%r12), %xmm3 vmovss (%r9,%r12), %xmm4 vinsertps $0x20, %xmm4, %xmm3, %xmm5 # xmm5 = xmm3[0,1],xmm4[0],xmm3[3] vinsertps $0x30, 0x4(%r9,%r12), %xmm5, %xmm5 # xmm5 = xmm5[0,1,2],mem[0] vmovups 0x10(%r10,%r13), %xmm6 vmovss 0x20(%r10,%r13), %xmm7 vmulss 0x8(%r9,%r12), %xmm7, %xmm8 vmulps %xmm5, %xmm6, %xmm5 vhaddps %xmm5, %xmm5, %xmm5 vmovups (%r10,%r13), %xmm9 vhaddps %xmm5, %xmm5, %xmm5 vmulps %xmm2, %xmm9, %xmm2 vhaddps %xmm2, %xmm2, %xmm2 vhaddps %xmm2, %xmm2, %xmm2 vaddss %xmm2, %xmm8, %xmm2 vmovlhps %xmm3, %xmm9, %xmm8 # xmm8 = xmm9[0],xmm3[0] vshufps $0xd8, %xmm3, %xmm8, %xmm3 # xmm3 = xmm8[0,2],xmm3[1,3] vinsertps $0x30, %xmm4, %xmm3, %xmm3 # xmm3 = xmm3[0,1,2],xmm4[0] vmovsd 0x4(%r9,%r12), %xmm4 vmovhps (%r11,%r12), %xmm4, %xmm4 # xmm4 = xmm4[0,1],mem[0,1] vmulss 0x8(%r11,%r12), %xmm7, %xmm7 vmulps %xmm6, %xmm4, %xmm4 vhaddps %xmm4, %xmm4, %xmm4 vmovss %xmm1, %xmm9, %xmm1 # xmm1 = xmm1[0],xmm9[1,2,3] vhaddps %xmm4, %xmm4, %xmm4 vmulps %xmm1, %xmm3, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm0, %xmm2, %xmm2 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm2, %xmm5, %xmm2 vaddss %xmm7, %xmm4, %xmm3 vaddss %xmm0, %xmm1, %xmm1 vaddss %xmm1, %xmm3, %xmm1 vmovss %xmm2, (%rbx,%r12) vmovss %xmm1, (%rsi,%r12) decl %ecx addq $0x4, %r12 jmp 0x2a67c3 addq %rbp, %r15 addq %r12, %r15 addq %rbp, %rax addq %r12, %rax addq %rbp, %r9 addq %r12, %r9 addq %rbp, %r11 addq %r12, %r11 addq 0x20(%rsp), %rbx addq %r12, %rbx addl $0x2, %edi addq %r12, %rsi jmp 0x2a67a7 movq 0x110(%rsp), %r11 cmpl %r8d, %edi jge 0x2a6979 xorl %ecx, %ecx movl %edx, %esi testl %esi, %esi jle 0x2a695a vmovsd (%r15,%rcx), %xmm1 vinsertps $0x20, 0x8(%r15,%rcx), %xmm1, %xmm1 # xmm1 = xmm1[0,1],mem[0],xmm1[3] vinsertps $0x30, (%rax,%rcx), %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],mem[0] vmovsd 0x4(%rax,%rcx), %xmm2 vinsertf128 $0x1, %xmm2, %ymm1, %ymm2 vmovsd (%r9,%rcx), %xmm3 vinsertf128 $0x1, %xmm3, %ymm1, %ymm1 vshufpd $0x2, %ymm1, %ymm2, %ymm1 # ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[2] vmulps (%r10,%r13), %ymm1, %ymm1 vmovss 0x20(%r10,%r13), %xmm2 vmulss 0x8(%r9,%rcx), %xmm2, %xmm2 vextractf128 $0x1, %ymm1, %xmm3 vhaddps %xmm1, %xmm3, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm0, %xmm2, %xmm2 vaddss %xmm1, %xmm2, %xmm1 vmovss %xmm1, (%rbx,%rcx) decl %esi addq $0x4, %rcx jmp 0x2a68e9 addq %rcx, %r15 addq $0x8, %r15 addq %rcx, %rax addq $0x8, %rax addq %rcx, %r9 addq $0x8, %r9 incl %edi addq %rcx, %rbx jmp 0x2a68dc movq 0x70(%rsp), %rax incq %rax movq 0x18(%rsp), %r13 movq 0x50(%rsp), %r9 movq 0x170(%rsp), %rsi jmp 0x2a6739 cmpl $0x5, %r15d jne 0x2a841c cmpl $0x1, 0xdc(%rdi) jne 0x2a80c5 cmpl $0x1, 0xe0(%rdi) jne 0x2a80c5 cmpl $0x1, 0xe4(%rdi) jne 0x2a80c5 cmpl $0x1, 0xe8(%rdi) jne 0x2a80c5 movl 0x2c(%r13), %r8d movl 0x30(%r13), %eax movl %eax, 0x10(%rsp) movl 0xbc(%rsp), %eax movl 0xc8(%rsp), %ecx movq 0x1b0(%rdi), %rdx leal 0x10(,%rax,4), %eax movslq %eax, %rdi leal (,%r8,4), %eax cltq xorl %esi, %esi testl %r8d, %r8d cmovlel %esi, %r8d testl %ecx, %ecx cmovlel %esi, %ecx movq %rcx, 0xf0(%rsp) shlq $0x2, %rax movq %rax, 0x50(%rsp) shlq $0x2, %rdi movq %rdx, 0x110(%rsp) movq %r8, 0x20(%rsp) cmpq 0xf0(%rsp), %rsi je 0x2a9725 testq %rdx, %rdx je 0x2a6a5c movq %rsi, %rax shlq $0x4, %rax vmovups (%rdx,%rax), %xmm0 jmp 0x2a6a60 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r10 imulq %rsi, %r10 movq 0x10(%r13), %rax imulq %rax, %r10 addq (%r13), %r10 movslq 0x2c(%r13), %r9 movq 0x28(%r14), %r11 movslq 0x54(%r14), %rbx imulq %rsi, %rbx imulq 0x38(%r14), %rbx imulq %rax, %r9 addq %r10, %r9 movslq 0xbc(%rsp), %rax movq 0xd0(%rsp), %r15 movq %rsi, 0x70(%rsp) imulq %rsi, %r15 movq 0xa0(%rsp), %rcx imulq %rcx, %r15 addq 0x90(%rsp), %r15 imulq %rcx, %rax leaq (%r15,%rax), %r12 leaq (%r15,%rax,2), %r13 leaq (%rax,%rax,2), %rbp addq %r15, %rbp leaq (%r15,%rax,4), %rdx leaq (%rax,%rax,4), %rax addq %r15, %rax xorl %esi, %esi movl %esi, %ecx orl $0x1, %ecx cmpl 0x10(%rsp), %ecx jge 0x2a70a2 movl %r8d, %ecx xorl %r8d, %r8d subl $0x1, %ecx jb 0x2a6e36 vmovaps (%r11,%rbx), %xmm1 vmovaps 0x10(%r11,%rbx), %xmm2 vmovaps 0x20(%r11,%rbx), %xmm3 vmovaps 0x30(%r11,%rbx), %xmm4 vmulps (%r15,%r8), %xmm1, %xmm5 vmovaps 0x40(%r11,%rbx), %xmm6 vaddps %xmm0, %xmm5, %xmm5 vmulps 0x10(%r15,%r8), %xmm2, %xmm7 vmulps 0x20(%r15,%r8), %xmm3, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps 0x30(%r15,%r8), %xmm4, %xmm7 vmulps 0x40(%r15,%r8), %xmm6, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vmovaps (%r12,%r8), %xmm8 vmovaps 0x10(%r12,%r8), %xmm9 vmovaps 0x20(%r12,%r8), %xmm10 vmovaps 0x30(%r12,%r8), %xmm11 vmovaps 0x40(%r12,%r8), %xmm12 vmulps %xmm1, %xmm8, %xmm1 vaddps %xmm0, %xmm1, %xmm1 vmulps %xmm2, %xmm9, %xmm2 vmulps %xmm3, %xmm10, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps %xmm4, %xmm11, %xmm2 vmulps %xmm6, %xmm12, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0x50(%r11,%rbx), %xmm3 vmovaps 0x60(%r11,%rbx), %xmm4 vmovaps 0x70(%r11,%rbx), %xmm6 vmovaps 0x80(%r11,%rbx), %xmm13 vmovaps 0x90(%r11,%rbx), %xmm14 vmulps %xmm3, %xmm8, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm4, %xmm9, %xmm7 vmulps %xmm6, %xmm10, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vmulps %xmm11, %xmm13, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vmulps %xmm12, %xmm14, %xmm8 vaddps %xmm7, %xmm8, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmovaps (%r13,%r8), %xmm7 vmovaps 0x10(%r13,%r8), %xmm8 vmovaps 0x20(%r13,%r8), %xmm9 vmovaps 0x30(%r13,%r8), %xmm10 vmovaps 0x40(%r13,%r8), %xmm11 vmulps %xmm3, %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps %xmm4, %xmm8, %xmm2 vmulps %xmm6, %xmm9, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps %xmm13, %xmm10, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps %xmm14, %xmm11, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm2 vmovaps 0xa0(%r11,%rbx), %xmm3 vmovaps 0xb0(%r11,%rbx), %xmm4 vmovaps 0xc0(%r11,%rbx), %xmm6 vmovaps 0xd0(%r11,%rbx), %xmm12 vmovaps 0xe0(%r11,%rbx), %xmm13 vmulps %xmm7, %xmm3, %xmm1 vmulps %xmm4, %xmm8, %xmm7 vaddps %xmm7, %xmm1, %xmm1 vmulps %xmm6, %xmm9, %xmm7 vaddps %xmm7, %xmm1, %xmm1 vmulps %xmm10, %xmm12, %xmm7 vaddps %xmm7, %xmm1, %xmm1 vmulps %xmm11, %xmm13, %xmm7 vaddps %xmm7, %xmm1, %xmm1 vaddps %xmm1, %xmm5, %xmm1 vmovaps (%rbp,%r8), %xmm5 vmovaps 0x10(%rbp,%r8), %xmm7 vmovaps 0x20(%rbp,%r8), %xmm8 vmovaps 0x30(%rbp,%r8), %xmm9 vmovaps 0x40(%rbp,%r8), %xmm10 vmulps %xmm3, %xmm5, %xmm3 vmulps %xmm4, %xmm7, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm6, %xmm8, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm12, %xmm9, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm13, %xmm10, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0xf0(%r11,%rbx), %xmm3 vmovaps 0x100(%r11,%rbx), %xmm4 vmovaps 0x110(%r11,%rbx), %xmm6 vmovaps 0x120(%r11,%rbx), %xmm11 vmovaps 0x130(%r11,%rbx), %xmm12 vmulps %xmm5, %xmm3, %xmm5 vmulps %xmm7, %xmm4, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm6, %xmm8, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm9, %xmm11, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm10, %xmm12, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmovaps (%rdx,%r8), %xmm7 vmovaps 0x10(%rdx,%r8), %xmm8 vmovaps 0x20(%rdx,%r8), %xmm9 vmovaps 0x30(%rdx,%r8), %xmm10 vmovaps 0x40(%rdx,%r8), %xmm13 vmulps %xmm3, %xmm7, %xmm3 vmulps %xmm4, %xmm8, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm6, %xmm9, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm11, %xmm10, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps %xmm12, %xmm13, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmovaps 0x140(%r11,%rbx), %xmm4 vmovaps 0x150(%r11,%rbx), %xmm6 vmovaps 0x160(%r11,%rbx), %xmm11 vmovaps 0x170(%r11,%rbx), %xmm12 vmovaps 0x180(%r11,%rbx), %xmm14 vmulps %xmm7, %xmm4, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vaddps %xmm5, %xmm1, %xmm1 vmulps %xmm6, %xmm8, %xmm5 vmulps %xmm9, %xmm11, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm10, %xmm12, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vmulps %xmm13, %xmm14, %xmm7 vaddps %xmm7, %xmm5, %xmm5 vaddps %xmm5, %xmm1, %xmm1 vmulps (%rax,%r8), %xmm4, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps 0x10(%rax,%r8), %xmm6, %xmm4 vmulps 0x20(%rax,%r8), %xmm11, %xmm5 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm5, %xmm4, %xmm3 vmulps 0x30(%rax,%r8), %xmm12, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vmulps 0x40(%rax,%r8), %xmm14, %xmm4 vaddps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps %xmm1, (%r10,%r8) vmovaps %xmm2, (%r9,%r8) addq $0x10, %r8 jmp 0x2a6af2 addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp addq %rdi, %rdx addq %r8, %rdx addq %rdi, %rax addq %r8, %rax movq 0x50(%rsp), %rcx addq %rcx, %r10 addq %r8, %r10 addq %rcx, %r9 addq %r8, %r9 addl $0x2, %esi movq 0x20(%rsp), %r8 jmp 0x2a6add movl %r8d, %ecx xorl %eax, %eax subl $0x1, %ecx jb 0x2a707c vmovaps (%r11,%rbx), %xmm1 vmovaps 0x10(%r11,%rbx), %xmm2 vmovaps 0x20(%r11,%rbx), %xmm3 vmovaps 0x30(%r11,%rbx), %xmm4 vmulps (%r15,%rax), %xmm1, %xmm1 vmovaps 0x40(%r11,%rbx), %xmm5 vaddps %xmm0, %xmm1, %xmm1 vmulps 0x10(%r15,%rax), %xmm2, %xmm2 vmulps 0x20(%r15,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x30(%r15,%rax), %xmm4, %xmm2 vmulps 0x40(%r15,%rax), %xmm5, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0x50(%r11,%rbx), %xmm3 vmovaps 0x60(%r11,%rbx), %xmm4 vmovaps 0x70(%r11,%rbx), %xmm5 vmovaps 0x80(%r11,%rbx), %xmm6 vmovaps 0x90(%r11,%rbx), %xmm7 vmulps (%r12,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x10(%r12,%rax), %xmm4, %xmm2 vmulps 0x20(%r12,%rax), %xmm5, %xmm3 vmulps 0x30(%r12,%rax), %xmm6, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x40(%r12,%rax), %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0xa0(%r11,%rbx), %xmm2 vmovaps 0xb0(%r11,%rbx), %xmm3 vmovaps 0xc0(%r11,%rbx), %xmm4 vmovaps 0xd0(%r11,%rbx), %xmm5 vmulps (%r13,%rax), %xmm2, %xmm2 vmovaps 0xe0(%r11,%rbx), %xmm6 vmulps 0x10(%r13,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x20(%r13,%rax), %xmm4, %xmm3 vmulps 0x30(%r13,%rax), %xmm5, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x40(%r13,%rax), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0xf0(%r11,%rbx), %xmm2 vmovaps 0x100(%r11,%rbx), %xmm3 vmovaps 0x110(%r11,%rbx), %xmm4 vmovaps 0x120(%r11,%rbx), %xmm5 vmulps (%rbp,%rax), %xmm2, %xmm2 vmovaps 0x130(%r11,%rbx), %xmm6 vmulps 0x10(%rbp,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x20(%rbp,%rax), %xmm4, %xmm3 vmulps 0x30(%rbp,%rax), %xmm5, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x40(%rbp,%rax), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0x140(%r11,%rbx), %xmm3 vmovaps 0x150(%r11,%rbx), %xmm4 vmovaps 0x160(%r11,%rbx), %xmm5 vmovaps 0x170(%r11,%rbx), %xmm6 vmovaps 0x180(%r11,%rbx), %xmm7 vmulps (%rdx,%rax), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x10(%rdx,%rax), %xmm4, %xmm2 vmulps 0x20(%rdx,%rax), %xmm5, %xmm3 vmulps 0x30(%rdx,%rax), %xmm6, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x40(%rdx,%rax), %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps %xmm1, (%r10,%rax) addq $0x10, %rax jmp 0x2a6e7d addq %rax, %r15 addq $0x40, %r15 leaq 0x40(%r12,%rax), %r12 addq %rax, %r13 addq $0x40, %r13 addq %rax, %rbp addq $0x40, %rbp addq %rax, %rdx addq $0x40, %rdx incl %esi addq %rax, %r10 cmpl 0x10(%rsp), %esi jl 0x2a6e78 movq 0x70(%rsp), %rsi incq %rsi movq 0x18(%rsp), %r13 movq 0x110(%rsp), %rdx jmp 0x2a6a3b cmpl $0x5, %r15d jne 0x2a93b2 cmpl $0x1, 0xdc(%rdi) jne 0x2a9040 cmpl $0x1, 0xe0(%rdi) jne 0x2a9040 cmpl $0x1, 0xe4(%rdi) jne 0x2a9040 cmpl $0x1, 0xe8(%rdi) jne 0x2a9040 movl 0x2c(%r13), %esi movl 0x30(%r13), %ecx movl 0xc8(%rsp), %eax xorl %edx, %edx testl %esi, %esi cmovlel %edx, %esi movq 0x1b0(%rdi), %rdi testl %ecx, %ecx cmovlel %edx, %ecx testl %eax, %eax cmovlel %edx, %eax movq %rax, 0x20(%rsp) movl $0x80, %r8d movq %rdi, 0x50(%rsp) cmpq 0x20(%rsp), %rdx je 0x2a9725 testq %rdi, %rdi je 0x2a7159 movq %rdx, %rax shlq $0x5, %rax vmovups (%rdi,%rax), %ymm0 jmp 0x2a715d vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %rdx, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %rdx, %r11 imulq 0x38(%r14), %r11 movq 0xd0(%rsp), %rbx movq %rdx, 0x10(%rsp) imulq %rdx, %rbx movq 0xa0(%rsp), %rax imulq %rax, %rbx addq 0x90(%rsp), %rbx movslq 0xbc(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x2a7413 movl %esi, %eax xorl %edi, %edi subl $0x1, %eax jb 0x2a73eb vmovaps (%r10,%r11), %ymm1 vmovaps 0x20(%r10,%r11), %ymm2 vmovaps 0x40(%r10,%r11), %ymm3 vmovaps 0x60(%r10,%r11), %ymm4 vmulps (%rbx,%rdi), %ymm1, %ymm1 vmovaps 0x80(%r10,%r11), %ymm5 vaddps %ymm0, %ymm1, %ymm1 vmulps 0x20(%rbx,%rdi), %ymm2, %ymm2 vmulps 0x40(%rbx,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x60(%rbx,%rdi), %ymm4, %ymm2 vmulps 0x80(%rbx,%rdi), %ymm5, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmovaps 0xa0(%r10,%r11), %ymm3 vmovaps 0xc0(%r10,%r11), %ymm4 vmovaps 0xe0(%r10,%r11), %ymm5 vmovaps 0x100(%r10,%r11), %ymm6 vmovaps 0x120(%r10,%r11), %ymm7 vmulps (%r15,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x20(%r15,%rdi), %ymm4, %ymm2 vmulps 0x40(%r15,%rdi), %ymm5, %ymm3 vmulps 0x60(%r15,%rdi), %ymm6, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x80(%r15,%rdi), %ymm7, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0x140(%r10,%r11), %ymm2 vmovaps 0x160(%r10,%r11), %ymm3 vmovaps 0x180(%r10,%r11), %ymm4 vmovaps 0x1a0(%r10,%r11), %ymm5 vmulps (%r12,%rdi), %ymm2, %ymm2 vmovaps 0x1c0(%r10,%r11), %ymm6 vmulps 0x20(%r12,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x40(%r12,%rdi), %ymm4, %ymm3 vmulps 0x60(%r12,%rdi), %ymm5, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x80(%r12,%rdi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0x1e0(%r10,%r11), %ymm2 vmovaps 0x200(%r10,%r11), %ymm3 vmovaps 0x220(%r10,%r11), %ymm4 vmovaps 0x240(%r10,%r11), %ymm5 vmulps (%r13,%rdi), %ymm2, %ymm2 vmovaps 0x260(%r10,%r11), %ymm6 vmulps 0x20(%r13,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x40(%r13,%rdi), %ymm4, %ymm3 vmulps 0x60(%r13,%rdi), %ymm5, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x80(%r13,%rdi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmovaps 0x280(%r10,%r11), %ymm3 vmovaps 0x2a0(%r10,%r11), %ymm4 vmovaps 0x2c0(%r10,%r11), %ymm5 vmovaps 0x2e0(%r10,%r11), %ymm6 vmovaps 0x300(%r10,%r11), %ymm7 vmulps (%rbp,%rdi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x20(%rbp,%rdi), %ymm4, %ymm2 vmulps 0x40(%rbp,%rdi), %ymm5, %ymm3 vmulps 0x60(%rbp,%rdi), %ymm6, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x80(%rbp,%rdi), %ymm7, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps %ymm1, (%r9,%rdi) addq $0x20, %rdi jmp 0x2a71d1 addq %rdi, %rbx addq %r8, %rbx addq %rdi, %r15 addq %r8, %r15 addq %rdi, %r12 addq %r8, %r12 addq %rdi, %r13 addq %r8, %r13 addq %rdi, %rbp addq %r8, %rbp incl %edx addq %rdi, %r9 jmp 0x2a71c5 movq 0x10(%rsp), %rdx incq %rdx movq 0x18(%rsp), %r13 movq 0x50(%rsp), %rdi jmp 0x2a713b cmpl $0x1, 0xdc(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x1, 0xe0(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x2, 0xe4(%rdi) movl %eax, %ecx jne 0x2a75fb cmpl $0x2, 0xe8(%rdi) movl %eax, %ecx jne 0x2a75fb movslq 0xbc(%rsp), %rdx movl 0x2c(%r13), %eax movl 0x30(%r13), %r8d movl 0xc8(%rsp), %esi movl %edx, %ecx subl %eax, %ecx addl %ecx, %ecx movq 0x28(%r14), %r10 movq 0x1b0(%rdi), %rdi imulq 0x10(%r13), %r11 movq 0x90(%rsp), %rbx movq %rbx, 0xf0(%rsp) movq 0xa0(%rsp), %rbx imulq 0xd0(%rsp), %rbx movq %rbx, 0x110(%rsp) movq %rdx, 0x20(%rsp) addq %rdx, %rdx movq %rdx, 0x130(%rsp) movslq %ecx, %r15 xorl %r12d, %r12d testl %r8d, %r8d cmovlel %r12d, %r8d testl %esi, %esi cmovlel %r12d, %esi movq %rsi, 0x10(%rsp) shlq $0x2, %r15 movq %rdi, 0x70(%rsp) cmpq 0x10(%rsp), %r12 je 0x2a9725 testq %rdi, %rdi je 0x2a7505 vmovss (%rdi,%r12,4), %xmm0 jmp 0x2a7509 vxorps %xmm0, %xmm0, %xmm0 movq %r11, %rdi movq %r11, %r13 imulq %r12, %r13 addq %r9, %r13 imulq $0x24, %r12, %rbp movq 0x110(%rsp), %r11 imulq %r12, %r11 addq 0xf0(%rsp), %r11 movq 0x20(%rsp), %rcx leaq (%r11,%rcx,4), %rsi movq 0x130(%rsp), %rcx leaq (%r11,%rcx,4), %rbx xorl %r9d, %r9d cmpl %r8d, %r9d je 0x2a75e1 xorl %edx, %edx movl %eax, %ecx testl %ecx, %ecx jle 0x2a75c7 vmovsd (%r11,%rdx), %xmm1 vinsertps $0x20, 0x8(%r11,%rdx), %xmm1, %xmm1 # xmm1 = xmm1[0,1],mem[0],xmm1[3] vinsertps $0x30, (%rsi,%rdx), %xmm1, %xmm1 # xmm1 = xmm1[0,1,2],mem[0] vmovsd 0x4(%rsi,%rdx), %xmm2 vinsertf128 $0x1, %xmm2, %ymm1, %ymm2 vmovsd (%rbx,%rdx), %xmm3 vinsertf128 $0x1, %xmm3, %ymm1, %ymm1 vshufpd $0x2, %ymm1, %ymm2, %ymm1 # ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[2] vmulps (%r10,%rbp), %ymm1, %ymm1 vmovss 0x20(%r10,%rbp), %xmm2 vmulss 0x8(%rbx,%rdx), %xmm2, %xmm2 vextractf128 $0x1, %ymm1, %xmm3 vhaddps %xmm1, %xmm3, %xmm1 vaddss %xmm0, %xmm2, %xmm2 vhaddps %xmm1, %xmm1, %xmm1 vhaddps %xmm1, %xmm1, %xmm1 vaddss %xmm1, %xmm2, %xmm1 vmovss %xmm1, (%r13) addq $0x4, %r13 decl %ecx addq $0x8, %rdx jmp 0x2a7553 addq %r15, %r11 addq %rdx, %r11 addq %r15, %rsi addq %rdx, %rsi addq %r15, %rbx addq %rdx, %rbx incl %r9d jmp 0x2a7546 incq %r12 movq 0x18(%rsp), %r13 movq 0x50(%rsp), %r9 movq %rdi, %r11 movq 0x70(%rsp), %rdi jmp 0x2a74ed cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %r8d movq 0x48(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x2a7641 xorl %eax, %eax testb $0x3, %dil sete %al testb $0x7, %dil leal 0x1(%rax,%rax,2), %eax cmovel %ebp, %eax movl %eax, 0x20(%rsp) xorl %eax, %eax testb $0x3, %r8b sete %al testb $0x7, %r8b leal 0x1(%rax,%rax,2), %eax cmovel %ebp, %eax jmp 0x2a7648 pushq $0x1 popq %rax movl %eax, 0x20(%rsp) movl %eax, 0x10(%rsp) movq 0x70(%rsp), %r15 movq 0x98(%rsp), %rax vmovaps 0x90(%rsp), %xmm0 vmovaps %xmm0, 0x210(%rsp) movq 0xa0(%rsp), %rcx movq %rcx, 0x220(%rsp) movl 0xa8(%rsp), %ecx movl %ecx, 0x228(%rsp) movq 0xb0(%rsp), %rcx movq %rcx, 0x230(%rsp) vmovups 0xb8(%rsp), %xmm0 vmovups %xmm0, 0x238(%rsp) movl 0xc8(%rsp), %ecx movl %ecx, 0x248(%rsp) movq 0xd0(%rsp), %rcx movq %rcx, 0x250(%rsp) testq %rax, %rax je 0x2a76d1 lock incl (%rax) cmpl 0x20(%rsp), %r10d jle 0x2a7725 movl %r8d, %ebx movl %edi, %ebp movq 0x48(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0x1a0(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x90(%rsp), %rdi leaq 0x210(%rsp), %rsi movl 0x20(%rsp), %edx vzeroupper callq 0x64e3b movl %ebp, %edi movl %ebx, %r8d movq 0x18(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm0 vmovaps %xmm0, 0x1a0(%rsp) movq 0x10(%rdx), %rcx movq %rcx, 0x1b0(%rsp) movl 0x18(%rdx), %ecx movl %ecx, 0x1b8(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x1c0(%rsp) vmovups 0x28(%rdx), %xmm0 vmovups %xmm0, 0x1c8(%rsp) movl 0x38(%rdx), %ecx movl %ecx, 0x1d8(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x1e0(%rsp) testq %rax, %rax je 0x2a7789 lock incl (%rax) cmpl %r15d, 0x10(%rsp) jae 0x2a7821 movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd0(%r14,%rax), %eax cltd movl 0x10(%rsp), %r9d idivl %r9d movl %eax, %ecx movq 0x170(%rsp), %rax xorl %edx, %edx divq %r15 movl %edi, %ebp movl %r8d, %ebx movl %r9d, %r8d imulq %rax, %r8 movq 0x48(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0x1a0(%rsp), %rdi movl 0x110(%rsp), %esi movl 0x130(%rsp), %edx callq 0x628f2 pushq $-0x64 popq %r12 cmpq $0x0, 0x1a0(%rsp) je 0x2a9662 movl %ebx, %r8d movl %ebp, %edi movslq 0x1d8(%rsp), %rax imulq 0x1e0(%rsp), %rax testq %rax, %rax je 0x2a9662 xorl %r15d, %r15d xorl %r12d, %r12d xorl %ebp, %ebp movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x108(%r14,%rax), %rax cmpq %rax, %rbp jge 0x2a7a8f movl %r15d, %eax cltd movl 0x20(%rsp), %esi idivl %esi movl %eax, %ecx movl %edi, %r13d movl %edi, %eax cltd idivl %esi movslq %ecx, %rdx imulq 0x250(%rsp), %rdx movq 0x220(%rsp), %rcx imulq %rcx, %rdx addq 0x210(%rsp), %rdx movl 0x228(%rsp), %esi movq 0x230(%rsp), %rdi movq %rdx, 0x308(%rsp) andq $0x0, 0x310(%rsp) movq %rcx, 0x318(%rsp) movl %esi, 0x320(%rsp) movq %rdi, 0x328(%rsp) movl %eax, 0x340(%rsp) vmovups 0x238(%rsp), %xmm0 movslq 0x244(%rsp), %rax movslq 0x23c(%rsp), %rdx movslq 0x240(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x348(%rsp) vmovups %xmm0, 0x330(%rsp) movl %r12d, %eax cltd movl 0x10(%rsp), %esi idivl %esi movl %eax, %ecx movl %r8d, %ebx movl %r8d, %eax cltd idivl %esi movslq %ecx, %rdx imulq 0x1e0(%rsp), %rdx movq 0x1b0(%rsp), %rsi imulq %rsi, %rdx addq 0x1a0(%rsp), %rdx movl 0x1b8(%rsp), %edi movq 0x1c0(%rsp), %rcx movq %rdx, 0x2c0(%rsp) andq $0x0, 0x2c8(%rsp) movq %rsi, 0x2d0(%rsp) movl %edi, 0x2d8(%rsp) movq %rcx, 0x2e0(%rsp) movl %eax, 0x2f8(%rsp) vmovups 0x1c8(%rsp), %xmm0 movslq 0x1d4(%rsp), %rax movslq 0x1cc(%rsp), %rdx movslq 0x1d0(%rsp), %rdi imulq %rdx, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x300(%rsp) vmovups %xmm0, 0x2e8(%rsp) movq 0x10(%r14), %rax movq (%rax,%rbp,8), %rdi movq 0x48(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 vmovups %ymm0, 0x370(%rsp) vmovups %ymm1, 0x390(%rsp) movq %rcx, 0x378(%rsp) movq (%rdi), %rax leaq 0x308(%rsp), %rsi leaq 0x2c0(%rsp), %rdx leaq 0x370(%rsp), %rcx vzeroupper callq *0x38(%rax) movq 0x2c8(%rsp), %rax testq %rax, %rax je 0x2a7a44 lock decl (%rax) jne 0x2a7a44 movq 0x2c0(%rsp), %rsi movq 0x2e0(%rsp), %rdi testq %rdi, %rdi je 0x2a7a3c movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a7a44 movq %rsi, %rdi callq 0x5f3e0 movq 0x310(%rsp), %rax testq %rax, %rax je 0x2a7a7b lock decl (%rax) jne 0x2a7a7b movq 0x308(%rsp), %rsi movq 0x328(%rsp), %rdi testq %rdi, %rdi je 0x2a7a73 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a7a7b movq %rsi, %rdi callq 0x5f3e0 incq %rbp movl %ebx, %r8d addl %ebx, %r12d movl %r13d, %edi addl %r13d, %r15d jmp 0x2a7829 movq 0x70(%rsp), %rdx cmpl %edx, 0x10(%rsp) jae 0x2a7ab9 xorl %r12d, %r12d leaq 0x1a0(%rsp), %rdi movq 0x18(%rsp), %rsi movq 0x48(%rsp), %rcx callq 0x64e3b jmp 0x2a9662 xorl %r12d, %r12d leaq 0x1a0(%rsp), %rax movq 0x18(%rsp), %rcx cmpq %rcx, %rax je 0x2a9662 movq 0x1a8(%rsp), %rax testq %rax, %rax je 0x2a7ae2 lock incl (%rax) movq 0x8(%rcx), %rax testq %rax, %rax je 0x2a960a lock decl (%rax) jne 0x2a960a movq 0x18(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x2a9602 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a960a pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x2a841c cmpl $0x1, 0xe0(%rdi) jne 0x2a841c cmpl $0x2, 0xe4(%rdi) jne 0x2a841c cmpl $0x2, 0xe8(%rdi) jne 0x2a841c movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xbc(%rsp), %r8d movl 0xc8(%rsp), %edx subl %eax, %r8d shll $0x3, %r8d movq 0x1b0(%rdi), %rsi movslq %r8d, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x2a9725 testq %rsi, %rsi je 0x2a7ba8 movq %r8, %r9 shlq $0x4, %r9 vmovups (%rsi,%r9), %xmm15 jmp 0x2a7bad vxorps %xmm15, %xmm15, %xmm15 movq 0x40(%r13), %r9 imulq %r8, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq %r14, %r10 movq 0x28(%r14), %r14 movslq 0x54(%r10), %r15 imulq %r8, %r15 imulq 0x38(%r10), %r15 movslq 0xbc(%rsp), %rbx movq 0xd0(%rsp), %r10 imulq %r8, %r10 movq 0xa0(%rsp), %r11 imulq %r11, %r10 addq 0x90(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx vmovaps (%r14,%r15), %xmm3 vmovaps 0x10(%r14,%r15), %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovaps 0x20(%r14,%r15), %xmm0 vmovaps %xmm0, 0x170(%rsp) vmovaps 0x30(%r14,%r15), %xmm6 vmovaps 0x40(%r14,%r15), %xmm7 vmovaps 0x50(%r14,%r15), %xmm14 vmovaps 0x60(%r14,%r15), %xmm0 vmovaps %xmm0, 0x50(%rsp) vmovaps 0x70(%r14,%r15), %xmm0 vmovaps %xmm0, 0xf0(%rsp) vmovaps 0x80(%r14,%r15), %xmm0 vmovaps %xmm0, 0x130(%rsp) xorl %ebp, %ebp vmovaps %xmm14, 0x1f0(%rsp) vmovaps %xmm3, 0x70(%rsp) vmovaps %xmm15, 0x110(%rsp) cmpl %ecx, %ebp je 0x2a80b8 xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r14d cmpl %eax, %r14d jge 0x2a7ed4 vmulps (%r10,%r15), %xmm3, %xmm10 vaddps %xmm15, %xmm10, %xmm11 vmovaps 0x20(%rsp), %xmm0 vmulps 0x10(%r10,%r15), %xmm0, %xmm12 vmovaps 0x20(%r10,%r15), %xmm14 vmovaps 0x40(%r10,%r15), %xmm10 vmovaps 0x170(%rsp), %xmm8 vmulps %xmm8, %xmm14, %xmm13 vaddps %xmm13, %xmm12, %xmm12 vaddps %xmm12, %xmm11, %xmm12 vmulps (%r11,%r15), %xmm6, %xmm11 vmulps 0x10(%r11,%r15), %xmm7, %xmm13 vaddps %xmm13, %xmm11, %xmm13 vmovaps 0x20(%r11,%r15), %xmm15 vmovaps 0x40(%r11,%r15), %xmm11 vmovaps 0x1f0(%rsp), %xmm9 vmulps %xmm9, %xmm15, %xmm0 vaddps %xmm0, %xmm13, %xmm0 vaddps %xmm0, %xmm12, %xmm0 vmovaps 0x50(%rsp), %xmm5 vmulps (%rbx,%r15), %xmm5, %xmm12 vmovaps 0xf0(%rsp), %xmm4 vmulps 0x10(%rbx,%r15), %xmm4, %xmm13 vaddps %xmm13, %xmm12, %xmm12 vmovaps 0x20(%rbx,%r15), %xmm1 vmovaps 0x40(%rbx,%r15), %xmm13 vmovaps 0x130(%rsp), %xmm3 vmulps %xmm3, %xmm1, %xmm2 vaddps %xmm2, %xmm12, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x20(%rsp), %xmm2 vmulps 0x30(%r10,%r15), %xmm2, %xmm2 vmulps 0x30(%r11,%r15), %xmm7, %xmm12 vaddps %xmm2, %xmm12, %xmm2 vmulps 0x30(%rbx,%r15), %xmm4, %xmm12 vaddps %xmm2, %xmm12, %xmm2 vmovaps %xmm0, (%r9) vmulps 0x70(%rsp), %xmm14, %xmm0 vaddps 0x110(%rsp), %xmm0, %xmm0 vmulps %xmm8, %xmm10, %xmm12 vmulps %xmm6, %xmm15, %xmm14 vmulps %xmm9, %xmm11, %xmm15 vaddps %xmm15, %xmm12, %xmm12 vmulps %xmm5, %xmm1, %xmm1 vaddps %xmm1, %xmm14, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm2, %xmm0, %xmm0 vmulps %xmm3, %xmm13, %xmm1 vaddps %xmm1, %xmm12, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmovaps 0x60(%r10,%r15), %xmm12 vmovaps 0x60(%r11,%r15), %xmm14 vmovaps 0x20(%rsp), %xmm1 vmulps 0x50(%r10,%r15), %xmm1, %xmm1 vmulps 0x50(%r11,%r15), %xmm7, %xmm2 vmovaps 0x60(%rbx,%r15), %xmm15 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x50(%rbx,%r15), %xmm4, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps %xmm0, 0x10(%r9) vmulps %xmm8, %xmm12, %xmm0 vmulps %xmm9, %xmm14, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmulps 0x70(%rsp), %xmm10, %xmm2 vaddps 0x110(%rsp), %xmm2, %xmm2 vmulps %xmm6, %xmm11, %xmm10 vmulps %xmm5, %xmm13, %xmm11 vaddps %xmm11, %xmm10, %xmm10 vaddps %xmm2, %xmm10, %xmm2 vaddps %xmm1, %xmm2, %xmm1 vmulps %xmm3, %xmm15, %xmm2 vaddps %xmm2, %xmm0, %xmm0 vmovaps 0x20(%rsp), %xmm2 vmulps 0x70(%r10,%r15), %xmm2, %xmm2 vmulps 0x80(%r10,%r15), %xmm8, %xmm10 vmulps 0x70(%r11,%r15), %xmm7, %xmm11 vaddps %xmm0, %xmm1, %xmm0 vaddps %xmm2, %xmm11, %xmm1 vmulps 0x80(%r11,%r15), %xmm9, %xmm2 vaddps %xmm2, %xmm10, %xmm2 vmulps 0x70(%rbx,%r15), %xmm4, %xmm10 vaddps %xmm1, %xmm10, %xmm1 vmulps 0x80(%rbx,%r15), %xmm3, %xmm10 vaddps %xmm2, %xmm10, %xmm2 vmovaps %xmm0, 0x20(%r9) vmulps %xmm6, %xmm14, %xmm0 vmulps %xmm5, %xmm15, %xmm10 vmovaps 0x110(%rsp), %xmm15 vmovaps 0x70(%rsp), %xmm3 vaddps %xmm0, %xmm10, %xmm0 vmulps %xmm3, %xmm12, %xmm10 vaddps %xmm15, %xmm10, %xmm10 vaddps %xmm0, %xmm10, %xmm0 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm2, %xmm0, %xmm0 vmovaps %xmm0, 0x30(%r9) addq $0x40, %r9 addl $0x4, %r12d subq $-0x80, %r15 jmp 0x2a7c96 vmovaps 0x1f0(%rsp), %xmm14 leal 0x1(%r12), %r14d cmpl %eax, %r14d jge 0x2a8096 vmovaps 0x20(%r10,%r15), %xmm10 vmovaps 0x20(%r11,%r15), %xmm11 vmulps (%r10,%r15), %xmm3, %xmm0 vmovaps 0x20(%rbx,%r15), %xmm12 vaddps %xmm0, %xmm15, %xmm0 vmovaps 0x20(%rsp), %xmm13 vmulps 0x10(%r10,%r15), %xmm13, %xmm1 vmovaps 0x170(%rsp), %xmm5 vmulps %xmm5, %xmm10, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps (%r11,%r15), %xmm6, %xmm2 vaddps %xmm1, %xmm0, %xmm0 vmulps 0x10(%r11,%r15), %xmm7, %xmm1 vaddps %xmm1, %xmm2, %xmm1 vmulps %xmm14, %xmm11, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0x50(%rsp), %xmm9 vmulps (%rbx,%r15), %xmm9, %xmm2 vaddps %xmm1, %xmm0, %xmm0 vmovaps 0xf0(%rsp), %xmm8 vmulps 0x10(%rbx,%r15), %xmm8, %xmm1 vaddps %xmm1, %xmm2, %xmm1 vmovaps 0x130(%rsp), %xmm4 vmulps %xmm4, %xmm12, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x30(%r10,%r15), %xmm13, %xmm2 vmulps 0x40(%r10,%r15), %xmm5, %xmm13 vmovaps %xmm14, %xmm5 vmulps 0x30(%r11,%r15), %xmm7, %xmm14 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm2, %xmm14, %xmm1 vmovaps %xmm5, %xmm14 vmulps 0x40(%r11,%r15), %xmm5, %xmm2 vaddps %xmm2, %xmm13, %xmm2 vmulps 0x30(%rbx,%r15), %xmm8, %xmm13 vaddps %xmm1, %xmm13, %xmm1 vmulps 0x40(%rbx,%r15), %xmm4, %xmm13 vaddps %xmm2, %xmm13, %xmm2 vmovaps %xmm0, (%r9) vmulps %xmm6, %xmm11, %xmm0 vmulps %xmm9, %xmm12, %xmm11 vaddps %xmm0, %xmm11, %xmm0 vmulps %xmm3, %xmm10, %xmm10 vaddps %xmm15, %xmm10, %xmm10 vaddps %xmm0, %xmm10, %xmm0 vaddps %xmm1, %xmm0, %xmm0 vaddps %xmm2, %xmm0, %xmm0 vmovaps %xmm0, 0x10(%r9) addq $0x20, %r9 addl $0x2, %r12d addq $0x40, %r15 jmp 0x2a7edd vmulps (%r10,%r15), %xmm3, %xmm0 vaddps %xmm0, %xmm15, %xmm0 vmovaps 0x20(%rsp), %xmm1 vmulps 0x10(%r10,%r15), %xmm1, %xmm1 vmovaps 0x170(%rsp), %xmm2 vmulps 0x20(%r10,%r15), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmulps (%r11,%r15), %xmm6, %xmm1 vmulps 0x10(%r11,%r15), %xmm7, %xmm2 vmulps 0x20(%r11,%r15), %xmm14, %xmm10 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm10, %xmm1 vmovaps 0x50(%rsp), %xmm2 vmulps (%rbx,%r15), %xmm2, %xmm2 vaddps %xmm1, %xmm0, %xmm0 vmovaps 0xf0(%rsp), %xmm1 vmulps 0x10(%rbx,%r15), %xmm1, %xmm1 vaddps %xmm1, %xmm2, %xmm1 vmovaps 0x130(%rsp), %xmm2 vmulps 0x20(%rbx,%r15), %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm1, %xmm0, %xmm0 vmovaps %xmm0, (%r9) addq $0x10, %r9 incl %r12d addq $0x20, %r15 cmpl %eax, %r12d jl 0x2a7fff addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x2a7c88 incq %r8 movq 0x10(%rsp), %r14 jmp 0x2a7b8b pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x2a841c cmpl $0x1, 0xe0(%rdi) jne 0x2a841c cmpl $0x2, 0xe4(%rdi) jne 0x2a841c cmpl $0x2, 0xe8(%rdi) jne 0x2a841c movl 0x2c(%r13), %edx movl 0x30(%r13), %ecx movl 0xbc(%rsp), %eax movl 0xc8(%rsp), %esi subl %edx, %eax shll $0x3, %eax movq 0x1b0(%rdi), %r9 xorl %r15d, %r15d testl %edx, %edx cmovlel %r15d, %edx movl %edx, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %esi, %esi cmovlel %r15d, %esi movq %rsi, 0x20(%rsp) shlq $0x2, %r8 movq %r9, 0x50(%rsp) cmpq 0x20(%rsp), %r15 je 0x2a9725 testq %r9, %r9 je 0x2a8166 movq %r15, %rax shlq $0x4, %rax vmovups (%r9,%rax), %xmm0 jmp 0x2a816a vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r15, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %r15, %r11 imulq 0x38(%r14), %r11 movq 0xd0(%rsp), %rbx movq %r15, 0x10(%rsp) imulq %r15, %rbx movq 0xa0(%rsp), %rax imulq %rax, %rbx addq 0x90(%rsp), %rbx movslq 0xbc(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x2a8405 movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x2a83e0 vmovaps (%r10,%r11), %xmm1 vmovaps 0x10(%r10,%r11), %xmm2 vmovaps 0x20(%r10,%r11), %xmm3 vmovaps 0x30(%r10,%r11), %xmm4 vmovaps 0x40(%r10,%r11), %xmm5 vmulps (%rbx,%rsi), %xmm1, %xmm1 vmulps 0x10(%rbx,%rsi), %xmm2, %xmm2 vmulps 0x20(%rbx,%rsi), %xmm3, %xmm3 vaddps %xmm0, %xmm1, %xmm1 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x30(%rbx,%rsi), %xmm4, %xmm3 vaddps %xmm2, %xmm1, %xmm1 vmulps 0x40(%rbx,%rsi), %xmm5, %xmm2 vaddps %xmm2, %xmm3, %xmm2 vmovaps 0x50(%r10,%r11), %xmm3 vmovaps 0x60(%r10,%r11), %xmm4 vmovaps 0x70(%r10,%r11), %xmm5 vmovaps 0x80(%r10,%r11), %xmm6 vmovaps 0x90(%r10,%r11), %xmm7 vmulps (%r15,%rsi), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x10(%r15,%rsi), %xmm4, %xmm3 vmulps 0x20(%r15,%rsi), %xmm5, %xmm4 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm4, %xmm3, %xmm2 vmulps 0x30(%r15,%rsi), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x40(%r15,%rsi), %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0xa0(%r10,%r11), %xmm2 vmovaps 0xb0(%r10,%r11), %xmm3 vmovaps 0xc0(%r10,%r11), %xmm4 vmovaps 0xd0(%r10,%r11), %xmm5 vmovaps 0xe0(%r10,%r11), %xmm6 vmulps (%r12,%rsi), %xmm2, %xmm2 vmulps 0x10(%r12,%rsi), %xmm3, %xmm3 vmulps 0x20(%r12,%rsi), %xmm4, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x30(%r12,%rsi), %xmm5, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x40(%r12,%rsi), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps 0xf0(%r10,%r11), %xmm2 vmovaps 0x100(%r10,%r11), %xmm3 vmovaps 0x110(%r10,%r11), %xmm4 vmovaps 0x120(%r10,%r11), %xmm5 vmovaps 0x130(%r10,%r11), %xmm6 vmulps (%r13,%rsi), %xmm2, %xmm2 vmulps 0x10(%r13,%rsi), %xmm3, %xmm3 vmulps 0x20(%r13,%rsi), %xmm4, %xmm4 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm4, %xmm2, %xmm2 vmulps 0x30(%r13,%rsi), %xmm5, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x40(%r13,%rsi), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmovaps 0x140(%r10,%r11), %xmm3 vmovaps 0x150(%r10,%r11), %xmm4 vmovaps 0x160(%r10,%r11), %xmm5 vmovaps 0x170(%r10,%r11), %xmm6 vmovaps 0x180(%r10,%r11), %xmm7 vmulps (%rbp,%rsi), %xmm3, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x10(%rbp,%rsi), %xmm4, %xmm3 vmulps 0x20(%rbp,%rsi), %xmm5, %xmm4 vaddps %xmm2, %xmm1, %xmm1 vaddps %xmm4, %xmm3, %xmm2 vmulps 0x30(%rbp,%rsi), %xmm6, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vmulps 0x40(%rbp,%rsi), %xmm7, %xmm3 vaddps %xmm3, %xmm2, %xmm2 vaddps %xmm2, %xmm1, %xmm1 vmovaps %xmm1, (%r9) addq $0x10, %r9 addq $0x20, %rsi jmp 0x2a81de addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x2a81d2 movq 0x10(%rsp), %r15 incq %r15 movq 0x18(%rsp), %r13 movq 0x50(%rsp), %r9 jmp 0x2a8147 imull %eax, %r15d movslq %r15d, %rsi leaq 0x210(%rsp), %rdi leaq 0x1a0(%rsp), %rdx callq 0x73bbe movq (%r14), %rcx movq -0x18(%rcx), %rdx movl 0x20(%rsp), %r11d imull 0xe0(%r14,%rdx), %r11d movq 0x210(%rsp), %rax movl 0xdc(%r14,%rdx), %esi imull 0xd4(%r14,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%r14,%rdx), %r8d jge 0x2a84ae movslq %esi, %rsi leaq (%rax,%rsi,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%r14,%rdx), %r9d jge 0x2a84a3 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%r14,%rdx), %edi incq %r9 jmp 0x2a8484 addl %r11d, %edi incl %r8d addq %r9, %rsi jmp 0x2a8470 leal (,%r15,4), %ecx movl %ecx, 0x170(%rsp) movl 0x110(%rsp), %ecx shll $0x2, %ecx xorl %edx, %edx testl %r15d, %r15d cmovlel %edx, %r15d movq 0xf0(%rsp), %r12 testl %r12d, %r12d cmovlel %edx, %r12d movq %r12, 0xf0(%rsp) movslq %ecx, %rcx movq %rcx, 0x70(%rsp) shlq $0x2, %r15 leaq 0x1501c6(%rip), %rdi # 0x3f86c0 vxorps %xmm0, %xmm0, %xmm0 vbroadcastss 0x145b0d(%rip), %xmm5 # 0x3ee014 vbroadcastss 0x146778(%rip), %xmm6 # 0x3eec88 xorl %ecx, %ecx vbroadcastss 0x148ca5(%rip), %xmm9 # 0x3f11c0 vbroadcastss 0x148ca8(%rip), %xmm10 # 0x3f11cc vbroadcastss 0x148ca3(%rip), %xmm11 # 0x3f11d0 vbroadcastss 0x148c9e(%rip), %xmm12 # 0x3f11d4 vbroadcastss 0x148c99(%rip), %xmm13 # 0x3f11d8 cmpq 0xf0(%rsp), %rcx je 0x2a95ea movq %rdx, 0x110(%rsp) movslq %edx, %r9 movq 0x40(%r13), %r10 imulq %rcx, %r10 imulq 0x10(%r13), %r10 shlq $0x2, %r9 addq (%r13), %r10 movslq 0xbc(%rsp), %r11 movq 0xd0(%rsp), %r13 imulq %rcx, %r13 movq %rcx, %rdx movq 0xa0(%rsp), %rcx imulq %rcx, %r13 addq 0x90(%rsp), %r13 imulq %rcx, %r11 movq %rdx, 0x130(%rsp) shlq $0x4, %rdx movq %rdx, 0x20(%rsp) addq 0x28(%r14), %r9 xorl %ecx, %ecx cmpl 0x50(%rsp), %ecx jg 0x2a8a9b movq (%r14), %rdx xorl %r8d, %r8d cmpq %rbx, %r8 jg 0x2a8a8b movq -0x18(%rdx), %r12 cmpl $0x0, 0x100(%r14,%r12) je 0x2a85ef movq 0x1b0(%r14,%r12), %rsi movq 0x20(%rsp), %rbp vmovups (%rsi,%rbp), %xmm1 jmp 0x2a85f3 vxorps %xmm1, %xmm1, %xmm1 movslq 0xe8(%r14,%r12), %rsi movq %r14, %rbp movslq %ecx, %r14 imulq %rsi, %r14 imulq %r11, %r14 addq %r13, %r14 movl 0xe4(%rbp,%r12), %esi imull %r8d, %esi shll $0x2, %esi movslq %esi, %rsi leaq (%r14,%rsi,4), %rsi xorl %r14d, %r14d cmpq %r14, %r15 je 0x2a8647 movslq (%rax,%r14), %rbp shlq $0x4, %rbp vmovups (%r9,%r14,4), %xmm2 vmulps (%rsi,%rbp), %xmm2, %xmm2 vaddps %xmm1, %xmm2, %xmm1 addq $0x4, %r14 jmp 0x2a8625 movq 0x10(%rsp), %r14 movl 0x110(%r14,%r12), %esi decl %esi cmpl $0x5, %esi ja 0x2a8a72 movslq (%rdi,%rsi,4), %rsi addq %rdi, %rsi vmaxps %xmm0, %xmm1, %xmm7 jmpq *%rsi movq 0x118(%r14,%r12), %rsi vminps %xmm0, %xmm1, %xmm1 vbroadcastss (%rsi), %xmm15 vmulps %xmm1, %xmm15, %xmm1 vaddps %xmm7, %xmm1, %xmm7 jmp 0x2a8a76 vbroadcastss 0x148b25(%rip), %xmm4 # 0x3f11b8 vminps %xmm4, %xmm1, %xmm7 vbroadcastss 0x148b1c(%rip), %xmm8 # 0x3f11bc vmaxps %xmm7, %xmm8, %xmm7 vmulps %xmm7, %xmm9, %xmm14 vaddps %xmm5, %xmm14, %xmm14 vcvttps2dq %xmm14, %xmm15 vcvtdq2ps %xmm15, %xmm15 vcmpltps %xmm15, %xmm14, %xmm14 vandps %xmm6, %xmm14, %xmm14 vsubps %xmm14, %xmm15, %xmm14 vbroadcastss 0x149252(%rip), %xmm4 # 0x3f1920 vmulps %xmm4, %xmm14, %xmm15 vsubps %xmm15, %xmm7, %xmm7 vmulps %xmm7, %xmm7, %xmm15 vmulps %xmm7, %xmm10, %xmm2 vaddps %xmm2, %xmm11, %xmm2 vmulps %xmm7, %xmm2, %xmm2 vaddps %xmm2, %xmm12, %xmm2 vmulps %xmm7, %xmm2, %xmm2 vaddps %xmm2, %xmm13, %xmm2 vmulps %xmm7, %xmm2, %xmm2 vmovaps %xmm5, %xmm0 vbroadcastss 0x148ad8(%rip), %xmm5 # 0x3f11dc vaddps %xmm5, %xmm2, %xmm2 vmulps %xmm7, %xmm2, %xmm2 vaddps %xmm0, %xmm2, %xmm2 vmulps %xmm2, %xmm15, %xmm2 vaddps %xmm6, %xmm7, %xmm7 vaddps %xmm2, %xmm7, %xmm2 vcvttps2dq %xmm14, %xmm7 vpslld $0x17, %xmm7, %xmm7 vpaddd %xmm6, %xmm7, %xmm7 vmulps %xmm7, %xmm2, %xmm2 vaddps %xmm6, %xmm2, %xmm7 vbroadcastss 0x148aa5(%rip), %xmm2 # 0x3f11e0 vmaxps %xmm2, %xmm7, %xmm2 vpsrld $0x17, %xmm2, %xmm14 vbroadcastss 0x148a9b(%rip), %xmm3 # 0x3f11e8 vpaddd %xmm3, %xmm14, %xmm14 vbroadcastss 0x148a8a(%rip), %xmm3 # 0x3f11e4 vandps %xmm3, %xmm2, %xmm2 vorps %xmm0, %xmm2, %xmm2 vcvtdq2ps %xmm14, %xmm14 vbroadcastss 0x148a7c(%rip), %xmm3 # 0x3f11ec vcmpltps %xmm3, %xmm2, %xmm15 vandps %xmm2, %xmm15, %xmm3 vbroadcastss 0x148a6e(%rip), %xmm8 # 0x3f11f0 vaddps %xmm2, %xmm8, %xmm2 vaddps %xmm3, %xmm2, %xmm2 vandps %xmm6, %xmm15, %xmm3 vsubps %xmm3, %xmm14, %xmm3 vmulps %xmm2, %xmm2, %xmm14 vbroadcastss 0x148a55(%rip), %xmm15 # 0x3f11f4 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a4c(%rip), %xmm5 # 0x3f11f8 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a3f(%rip), %xmm5 # 0x3f11fc vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a32(%rip), %xmm5 # 0x3f1200 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a25(%rip), %xmm5 # 0x3f1204 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a18(%rip), %xmm5 # 0x3f1208 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x148a0b(%rip), %xmm5 # 0x3f120c vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x1489fe(%rip), %xmm5 # 0x3f1210 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x1489f1(%rip), %xmm5 # 0x3f1214 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm2, %xmm15, %xmm15 vbroadcastss 0x1490f0(%rip), %xmm5 # 0x3f1924 vaddps %xmm5, %xmm15, %xmm15 vmulps %xmm15, %xmm14, %xmm14 vcmpleps 0x14583a(%rip), %xmm7, %xmm7 # 0x3ee080 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm2, %xmm3, %xmm2 vaddps %xmm2, %xmm14, %xmm2 vbroadcastss 0x149ebd(%rip), %xmm3 # 0x3f2718 vmulps %xmm3, %xmm2, %xmm2 vbroadcastss 0x149098(%rip), %xmm3 # 0x3f1900 vblendvps %xmm7, %xmm3, %xmm2, %xmm2 vbroadcastss 0x148941(%rip), %xmm3 # 0x3f11b8 vminps %xmm3, %xmm2, %xmm2 vbroadcastss 0x148938(%rip), %xmm3 # 0x3f11bc vmaxps %xmm3, %xmm2, %xmm2 vmulps %xmm2, %xmm9, %xmm3 vaddps %xmm0, %xmm3, %xmm3 vcvttps2dq %xmm3, %xmm7 vcvtdq2ps %xmm7, %xmm7 vcmpltps %xmm7, %xmm3, %xmm3 vandps %xmm6, %xmm3, %xmm3 vsubps %xmm3, %xmm7, %xmm3 vmulps %xmm4, %xmm3, %xmm7 vsubps %xmm7, %xmm2, %xmm2 vmulps %xmm2, %xmm2, %xmm7 vmulps %xmm2, %xmm10, %xmm14 vaddps %xmm11, %xmm14, %xmm14 vmulps %xmm2, %xmm14, %xmm14 vaddps %xmm12, %xmm14, %xmm14 vmulps %xmm2, %xmm14, %xmm14 vaddps %xmm13, %xmm14, %xmm14 vmulps %xmm2, %xmm14, %xmm14 vbroadcastss 0x148903(%rip), %xmm4 # 0x3f11dc vaddps %xmm4, %xmm14, %xmm14 vmovaps %xmm0, %xmm5 vxorps %xmm0, %xmm0, %xmm0 vmulps %xmm2, %xmm14, %xmm14 vaddps %xmm5, %xmm14, %xmm14 vmulps %xmm7, %xmm14, %xmm7 vaddps %xmm6, %xmm2, %xmm2 vaddps %xmm7, %xmm2, %xmm2 vcvttps2dq %xmm3, %xmm3 vpslld $0x17, %xmm3, %xmm3 vpaddd %xmm6, %xmm3, %xmm3 vmulps %xmm3, %xmm2, %xmm2 vaddps %xmm6, %xmm2, %xmm2 vrcpps %xmm2, %xmm3 vaddps %xmm3, %xmm3, %xmm7 vmulps %xmm7, %xmm2, %xmm2 vbroadcastss 0x149de5(%rip), %xmm4 # 0x3f2708 vsubps %xmm2, %xmm4, %xmm2 vmulps %xmm2, %xmm3, %xmm2 vaddps %xmm7, %xmm8, %xmm3 vaddps %xmm2, %xmm3, %xmm2 jmp 0x2a8a6c movq 0x118(%r14,%r12), %rsi vbroadcastss (%rsi), %xmm7 vbroadcastss 0x4(%rsi), %xmm15 vmaxps %xmm7, %xmm1, %xmm1 vminps %xmm1, %xmm15, %xmm7 jmp 0x2a8a76 vbroadcastss 0x148853(%rip), %xmm2 # 0x3f11b4 vxorps %xmm2, %xmm1, %xmm1 vbroadcastss 0x14884a(%rip), %xmm2 # 0x3f11b8 vminps %xmm2, %xmm1, %xmm1 vbroadcastss 0x148841(%rip), %xmm2 # 0x3f11bc vmaxps %xmm2, %xmm1, %xmm1 vbroadcastss 0x148838(%rip), %xmm2 # 0x3f11c0 vmulps %xmm2, %xmm1, %xmm7 vaddps %xmm5, %xmm7, %xmm7 vcvttps2dq %xmm7, %xmm15 vcvtdq2ps %xmm15, %xmm15 vcmpltps %xmm15, %xmm7, %xmm7 vandps %xmm6, %xmm7, %xmm7 vsubps %xmm7, %xmm15, %xmm7 vbroadcastss 0x14a030(%rip), %xmm2 # 0x3f29e0 vmulps %xmm2, %xmm7, %xmm15 vaddps %xmm1, %xmm15, %xmm1 vmulps %xmm1, %xmm1, %xmm15 vbroadcastss 0x148807(%rip), %xmm2 # 0x3f11cc vmulps %xmm2, %xmm1, %xmm14 vbroadcastss 0x1487fe(%rip), %xmm2 # 0x3f11d0 vaddps %xmm2, %xmm14, %xmm14 vmulps %xmm1, %xmm14, %xmm14 vbroadcastss 0x1487f1(%rip), %xmm2 # 0x3f11d4 vaddps %xmm2, %xmm14, %xmm14 vmulps %xmm1, %xmm14, %xmm14 vbroadcastss 0x1487e4(%rip), %xmm2 # 0x3f11d8 vaddps %xmm2, %xmm14, %xmm14 vmulps %xmm1, %xmm14, %xmm14 vbroadcastss 0x1487d7(%rip), %xmm2 # 0x3f11dc vaddps %xmm2, %xmm14, %xmm14 vmulps %xmm1, %xmm14, %xmm14 vaddps %xmm5, %xmm14, %xmm14 vmulps %xmm14, %xmm15, %xmm14 vaddps %xmm6, %xmm1, %xmm1 vaddps %xmm1, %xmm14, %xmm1 vcvttps2dq %xmm7, %xmm7 vpslld $0x17, %xmm7, %xmm7 vpaddd %xmm6, %xmm7, %xmm7 vmulps %xmm7, %xmm1, %xmm1 vaddps %xmm6, %xmm1, %xmm1 vrcpps %xmm1, %xmm7 vmulps %xmm7, %xmm1, %xmm1 vsubps %xmm1, %xmm6, %xmm1 vmulps %xmm1, %xmm7, %xmm1 vaddps %xmm1, %xmm7, %xmm7 jmp 0x2a8a76 movq 0x118(%r14,%r12), %rsi vbroadcastss (%rsi), %xmm2 vbroadcastss 0x4(%rsi), %xmm3 vmulps %xmm1, %xmm2, %xmm2 vaddps %xmm3, %xmm2, %xmm2 vmaxps %xmm0, %xmm2, %xmm2 vminps %xmm6, %xmm2, %xmm2 vmulps %xmm1, %xmm2, %xmm7 jmp 0x2a8a76 vmovaps %xmm1, %xmm7 movq %r8, %rsi shlq $0x4, %rsi vmovups %xmm7, (%r10,%rsi) incq %r8 jmp 0x2a85c3 movq 0x70(%rsp), %rdx leaq (%r10,%rdx,4), %r10 incl %ecx jmp 0x2a85b3 movq 0x130(%rsp), %rcx incq %rcx movq 0x110(%rsp), %rdx addl 0x170(%rsp), %edx movq 0x18(%rsp), %r13 jmp 0x2a853f pushq $0x3 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x2a93b2 cmpl $0x1, 0xe0(%rdi) jne 0x2a93b2 cmpl $0x2, 0xe4(%rdi) jne 0x2a93b2 cmpl $0x2, 0xe8(%rdi) jne 0x2a93b2 movl 0x2c(%r13), %eax movl 0x30(%r13), %ecx movl 0xbc(%rsp), %r8d movl 0xc8(%rsp), %edx subl %eax, %r8d shll $0x4, %r8d movq 0x1b0(%rdi), %rsi movslq %r8d, %rdi xorl %r8d, %r8d testl %ecx, %ecx cmovlel %r8d, %ecx testl %edx, %edx cmovlel %r8d, %edx shlq $0x2, %rdi cmpq %rdx, %r8 je 0x2a9725 testq %rsi, %rsi je 0x2a8b4f movq %r8, %r9 shlq $0x5, %r9 vmovups (%rsi,%r9), %ymm10 jmp 0x2a8b54 vxorps %xmm10, %xmm10, %xmm10 movq 0x40(%r13), %r9 imulq %r8, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq %r14, %r10 movq 0x28(%r14), %r14 movslq 0x54(%r10), %r15 imulq %r8, %r15 imulq 0x38(%r10), %r15 movslq 0xbc(%rsp), %rbx movq 0xd0(%rsp), %r10 imulq %r8, %r10 movq 0xa0(%rsp), %r11 imulq %r11, %r10 addq 0x90(%rsp), %r10 imulq %r11, %rbx leaq (%r10,%rbx), %r11 leaq (%r10,%rbx,2), %rbx vmovaps (%r14,%r15), %ymm11 vmovaps 0x20(%r14,%r15), %ymm12 vmovaps 0x40(%r14,%r15), %ymm13 vmovaps 0x60(%r14,%r15), %ymm14 vmovaps 0x80(%r14,%r15), %ymm15 vmovaps 0xa0(%r14,%r15), %ymm6 vmovaps 0xc0(%r14,%r15), %ymm7 vmovaps 0xe0(%r14,%r15), %ymm8 vmovaps 0x100(%r14,%r15), %ymm9 xorl %ebp, %ebp vmovups %ymm10, 0x20(%rsp) vmovups %ymm11, 0x50(%rsp) vmovups %ymm12, 0x70(%rsp) vmovups %ymm13, 0xf0(%rsp) vmovups %ymm14, 0x110(%rsp) vmovups %ymm15, 0x130(%rsp) vmovups %ymm6, 0x170(%rsp) cmpl %ecx, %ebp je 0x2a9033 xorl %r15d, %r15d xorl %r12d, %r12d leal 0x3(%r12), %r14d cmpl %eax, %r14d jge 0x2a8e76 vmovaps 0x40(%r10,%r15), %ymm10 vmovaps 0x80(%r10,%r15), %ymm11 vmovaps 0xc0(%r10,%r15), %ymm12 vmovups 0x50(%rsp), %ymm0 vmulps (%r10,%r15), %ymm0, %ymm13 vmovups 0x20(%rsp), %ymm1 vaddps %ymm1, %ymm13, %ymm13 vmulps %ymm0, %ymm10, %ymm14 vaddps %ymm1, %ymm14, %ymm14 vmulps %ymm0, %ymm11, %ymm15 vaddps %ymm1, %ymm15, %ymm15 vmulps %ymm0, %ymm12, %ymm0 vaddps %ymm1, %ymm0, %ymm0 vmovups 0x70(%rsp), %ymm4 vmulps 0x20(%r10,%r15), %ymm4, %ymm1 vmulps 0x60(%r10,%r15), %ymm4, %ymm2 vmulps 0xa0(%r10,%r15), %ymm4, %ymm3 vmulps 0xe0(%r10,%r15), %ymm4, %ymm4 vmovups 0xf0(%rsp), %ymm5 vmulps %ymm5, %ymm10, %ymm10 vaddps %ymm1, %ymm10, %ymm1 vaddps %ymm1, %ymm13, %ymm1 vmulps %ymm5, %ymm11, %ymm10 vaddps %ymm2, %ymm10, %ymm2 vaddps %ymm2, %ymm14, %ymm2 vmulps %ymm5, %ymm12, %ymm10 vaddps %ymm3, %ymm10, %ymm3 vmulps 0x100(%r10,%r15), %ymm5, %ymm10 vaddps %ymm3, %ymm15, %ymm3 vaddps %ymm4, %ymm10, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmovaps 0x40(%r11,%r15), %ymm4 vmovaps 0x80(%r11,%r15), %ymm10 vmovaps 0xc0(%r11,%r15), %ymm11 vmovups 0x110(%rsp), %ymm5 vmulps (%r11,%r15), %ymm5, %ymm12 vmulps %ymm5, %ymm4, %ymm13 vmulps %ymm5, %ymm10, %ymm14 vmovups 0x130(%rsp), %ymm6 vmulps 0x20(%r11,%r15), %ymm6, %ymm15 vmulps %ymm5, %ymm11, %ymm5 vaddps %ymm15, %ymm12, %ymm12 vmulps 0x60(%r11,%r15), %ymm6, %ymm15 vaddps %ymm15, %ymm13, %ymm13 vmulps 0xa0(%r11,%r15), %ymm6, %ymm15 vaddps %ymm15, %ymm14, %ymm14 vmulps 0xe0(%r11,%r15), %ymm6, %ymm15 vaddps %ymm5, %ymm15, %ymm5 vmovups 0x170(%rsp), %ymm6 vmulps %ymm6, %ymm4, %ymm4 vaddps %ymm4, %ymm12, %ymm4 vaddps %ymm4, %ymm1, %ymm1 vmulps %ymm6, %ymm10, %ymm4 vaddps %ymm4, %ymm13, %ymm4 vaddps %ymm4, %ymm2, %ymm2 vmulps %ymm6, %ymm11, %ymm4 vaddps %ymm4, %ymm14, %ymm4 vaddps %ymm4, %ymm3, %ymm3 vmulps 0x100(%r11,%r15), %ymm6, %ymm4 vaddps %ymm4, %ymm5, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmovaps 0x40(%rbx,%r15), %ymm4 vmovaps 0x80(%rbx,%r15), %ymm5 vmulps (%rbx,%r15), %ymm7, %ymm10 vmovaps 0xc0(%rbx,%r15), %ymm11 vmulps %ymm7, %ymm4, %ymm12 vmulps 0x20(%rbx,%r15), %ymm8, %ymm13 vmulps %ymm7, %ymm5, %ymm14 vaddps %ymm13, %ymm10, %ymm10 vmulps 0x60(%rbx,%r15), %ymm8, %ymm13 vmulps %ymm7, %ymm11, %ymm15 vaddps %ymm13, %ymm12, %ymm12 vmulps 0xa0(%rbx,%r15), %ymm8, %ymm13 vaddps %ymm13, %ymm14, %ymm13 vmulps 0xe0(%rbx,%r15), %ymm8, %ymm14 vaddps %ymm14, %ymm15, %ymm14 vmulps %ymm4, %ymm9, %ymm4 vaddps %ymm4, %ymm10, %ymm4 vaddps %ymm4, %ymm1, %ymm1 vmulps %ymm5, %ymm9, %ymm4 vaddps %ymm4, %ymm12, %ymm4 vaddps %ymm4, %ymm2, %ymm2 vmulps %ymm9, %ymm11, %ymm4 vaddps %ymm4, %ymm13, %ymm4 vaddps %ymm4, %ymm3, %ymm3 vmulps 0x100(%rbx,%r15), %ymm9, %ymm4 vaddps %ymm4, %ymm14, %ymm4 vaddps %ymm4, %ymm0, %ymm0 vmovaps %ymm1, (%r9) vmovaps %ymm2, 0x20(%r9) vmovaps %ymm3, 0x40(%r9) vmovaps %ymm0, 0x60(%r9) subq $-0x80, %r9 addl $0x4, %r12d addq $0x100, %r15 # imm = 0x100 jmp 0x2a8c40 vmovups 0x20(%rsp), %ymm10 vmovups 0x50(%rsp), %ymm11 vmovups 0x70(%rsp), %ymm12 vmovups 0xf0(%rsp), %ymm13 vmovups 0x110(%rsp), %ymm14 vmovups 0x130(%rsp), %ymm15 vmovups 0x170(%rsp), %ymm6 leal 0x1(%r12), %r14d cmpl %eax, %r14d jge 0x2a9015 vmovaps 0x40(%r10,%r15), %ymm0 vmulps (%r10,%r15), %ymm11, %ymm1 vaddps %ymm1, %ymm10, %ymm1 vmulps %ymm0, %ymm11, %ymm2 vaddps %ymm2, %ymm10, %ymm2 vmulps 0x20(%r10,%r15), %ymm12, %ymm3 vmulps 0x60(%r10,%r15), %ymm12, %ymm4 vmulps %ymm0, %ymm13, %ymm0 vaddps %ymm0, %ymm3, %ymm0 vaddps %ymm0, %ymm1, %ymm0 vmulps 0x80(%r10,%r15), %ymm13, %ymm1 vaddps %ymm1, %ymm4, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vmovaps 0x40(%r11,%r15), %ymm2 vmulps (%r11,%r15), %ymm14, %ymm3 vmulps %ymm2, %ymm14, %ymm4 vmulps 0x20(%r11,%r15), %ymm15, %ymm5 vaddps %ymm5, %ymm3, %ymm3 vmulps 0x60(%r11,%r15), %ymm15, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps %ymm6, %ymm2, %ymm2 vaddps %ymm2, %ymm3, %ymm2 vaddps %ymm2, %ymm0, %ymm0 vmulps 0x80(%r11,%r15), %ymm6, %ymm2 vaddps %ymm2, %ymm4, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmulps (%rbx,%r15), %ymm7, %ymm2 vmovaps 0x40(%rbx,%r15), %ymm3 vmulps %ymm7, %ymm3, %ymm4 vmulps 0x20(%rbx,%r15), %ymm8, %ymm5 vaddps %ymm5, %ymm2, %ymm2 vmulps 0x60(%rbx,%r15), %ymm8, %ymm5 vaddps %ymm5, %ymm4, %ymm4 vmulps %ymm3, %ymm9, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%rbx,%r15), %ymm9, %ymm3 vaddps %ymm2, %ymm0, %ymm0 vaddps %ymm3, %ymm4, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps %ymm0, (%r9) vmovaps %ymm1, 0x20(%r9) addq $0x40, %r9 addl $0x2, %r12d subq $-0x80, %r15 jmp 0x2a8eac vmulps (%r10,%r15), %ymm11, %ymm0 vaddps %ymm0, %ymm10, %ymm0 vmulps 0x20(%r10,%r15), %ymm12, %ymm1 vmulps 0x40(%r10,%r15), %ymm13, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmulps (%r11,%r15), %ymm14, %ymm1 vmulps 0x20(%r11,%r15), %ymm15, %ymm2 vmulps 0x40(%r11,%r15), %ymm6, %ymm3 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm3, %ymm1, %ymm1 vmulps (%rbx,%r15), %ymm7, %ymm2 vaddps %ymm1, %ymm0, %ymm0 vmulps 0x20(%rbx,%r15), %ymm8, %ymm1 vaddps %ymm1, %ymm2, %ymm1 vmulps 0x40(%rbx,%r15), %ymm9, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm1, %ymm0, %ymm0 vmovaps %ymm0, (%r9) addq $0x20, %r9 incl %r12d addq $0x40, %r15 cmpl %eax, %r12d jl 0x2a8fa5 addq %rdi, %r10 addq %r15, %r10 addq %rdi, %r11 addq %r15, %r11 addq %rdi, %rbx addq %r15, %rbx incl %ebp jmp 0x2a8c32 incq %r8 movq 0x10(%rsp), %r14 jmp 0x2a8b32 pushq $0x5 popq %r15 cmpl $0x1, 0xdc(%rdi) jne 0x2a93b2 cmpl $0x1, 0xe0(%rdi) jne 0x2a93b2 cmpl $0x2, 0xe4(%rdi) jne 0x2a93b2 cmpl $0x2, 0xe8(%rdi) jne 0x2a93b2 movl 0x2c(%r13), %edx movl 0x30(%r13), %ecx movl 0xbc(%rsp), %eax movl 0xc8(%rsp), %esi subl %edx, %eax shll $0x4, %eax movq 0x1b0(%rdi), %r9 xorl %r15d, %r15d testl %edx, %edx cmovlel %r15d, %edx movl %edx, %edi movslq %eax, %r8 testl %ecx, %ecx cmovlel %r15d, %ecx testl %esi, %esi cmovlel %r15d, %esi movq %rsi, 0x20(%rsp) shlq $0x2, %r8 movq %r9, 0x50(%rsp) cmpq 0x20(%rsp), %r15 je 0x2a9725 testq %r9, %r9 je 0x2a90e1 movq %r15, %rax shlq $0x5, %rax vmovups (%r9,%rax), %ymm0 jmp 0x2a90e5 vxorps %xmm0, %xmm0, %xmm0 movq 0x40(%r13), %r9 imulq %r15, %r9 imulq 0x10(%r13), %r9 addq (%r13), %r9 movq 0x28(%r14), %r10 movslq 0x54(%r14), %r11 imulq %r15, %r11 imulq 0x38(%r14), %r11 movq 0xd0(%rsp), %rbx movq %r15, 0x10(%rsp) imulq %r15, %rbx movq 0xa0(%rsp), %rax imulq %rax, %rbx addq 0x90(%rsp), %rbx movslq 0xbc(%rsp), %rdx imulq %rax, %rdx leaq (%rbx,%rdx), %r15 leaq (%rbx,%rdx,2), %r12 leaq (%rdx,%rdx,2), %r13 addq %rbx, %r13 leaq (%rbx,%rdx,4), %rbp xorl %edx, %edx cmpl %ecx, %edx je 0x2a939b movl %edi, %eax xorl %esi, %esi subl $0x1, %eax jb 0x2a9376 vmovaps (%r10,%r11), %ymm1 vmovaps 0x20(%r10,%r11), %ymm2 vmovaps 0x40(%r10,%r11), %ymm3 vmovaps 0x60(%r10,%r11), %ymm4 vmovaps 0x80(%r10,%r11), %ymm5 vmulps (%rbx,%rsi), %ymm1, %ymm1 vmulps 0x20(%rbx,%rsi), %ymm2, %ymm2 vmulps 0x40(%rbx,%rsi), %ymm3, %ymm3 vaddps %ymm0, %ymm1, %ymm1 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x60(%rbx,%rsi), %ymm4, %ymm3 vaddps %ymm2, %ymm1, %ymm1 vmulps 0x80(%rbx,%rsi), %ymm5, %ymm2 vaddps %ymm2, %ymm3, %ymm2 vmovaps 0xa0(%r10,%r11), %ymm3 vmovaps 0xc0(%r10,%r11), %ymm4 vmovaps 0xe0(%r10,%r11), %ymm5 vmovaps 0x100(%r10,%r11), %ymm6 vmovaps 0x120(%r10,%r11), %ymm7 vmulps (%r15,%rsi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x20(%r15,%rsi), %ymm4, %ymm3 vmulps 0x40(%r15,%rsi), %ymm5, %ymm4 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm4, %ymm3, %ymm2 vmulps 0x60(%r15,%rsi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%r15,%rsi), %ymm7, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0x140(%r10,%r11), %ymm2 vmovaps 0x160(%r10,%r11), %ymm3 vmovaps 0x180(%r10,%r11), %ymm4 vmovaps 0x1a0(%r10,%r11), %ymm5 vmovaps 0x1c0(%r10,%r11), %ymm6 vmulps (%r12,%rsi), %ymm2, %ymm2 vmulps 0x20(%r12,%rsi), %ymm3, %ymm3 vmulps 0x40(%r12,%rsi), %ymm4, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x60(%r12,%rsi), %ymm5, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%r12,%rsi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps 0x1e0(%r10,%r11), %ymm2 vmovaps 0x200(%r10,%r11), %ymm3 vmovaps 0x220(%r10,%r11), %ymm4 vmovaps 0x240(%r10,%r11), %ymm5 vmovaps 0x260(%r10,%r11), %ymm6 vmulps (%r13,%rsi), %ymm2, %ymm2 vmulps 0x20(%r13,%rsi), %ymm3, %ymm3 vmulps 0x40(%r13,%rsi), %ymm4, %ymm4 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm4, %ymm2, %ymm2 vmulps 0x60(%r13,%rsi), %ymm5, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%r13,%rsi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmovaps 0x280(%r10,%r11), %ymm3 vmovaps 0x2a0(%r10,%r11), %ymm4 vmovaps 0x2c0(%r10,%r11), %ymm5 vmovaps 0x2e0(%r10,%r11), %ymm6 vmovaps 0x300(%r10,%r11), %ymm7 vmulps (%rbp,%rsi), %ymm3, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x20(%rbp,%rsi), %ymm4, %ymm3 vmulps 0x40(%rbp,%rsi), %ymm5, %ymm4 vaddps %ymm2, %ymm1, %ymm1 vaddps %ymm4, %ymm3, %ymm2 vmulps 0x60(%rbp,%rsi), %ymm6, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vmulps 0x80(%rbp,%rsi), %ymm7, %ymm3 vaddps %ymm3, %ymm2, %ymm2 vaddps %ymm2, %ymm1, %ymm1 vmovaps %ymm1, (%r9) addq $0x20, %r9 addq $0x40, %rsi jmp 0x2a9159 addq %r8, %rbx addq %rsi, %rbx addq %r8, %r15 addq %rsi, %r15 addq %r8, %r12 addq %rsi, %r12 addq %r8, %r13 addq %rsi, %r13 addq %r8, %rbp addq %rsi, %rbp incl %edx jmp 0x2a914d movq 0x10(%rsp), %r15 incq %r15 movq 0x18(%rsp), %r13 movq 0x50(%rsp), %r9 jmp 0x2a90c2 imull %eax, %r15d movslq %r15d, %rsi leaq 0x210(%rsp), %rdi leaq 0x1a0(%rsp), %rdx callq 0x73bbe movq (%r14), %rcx movq -0x18(%rcx), %rdx movl 0x20(%rsp), %r11d imull 0xe0(%r14,%rdx), %r11d movq 0x210(%rsp), %rax movl 0xdc(%r14,%rdx), %esi imull 0xd4(%r14,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%r14,%rdx), %esi jge 0x2a9443 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%r14,%rdx), %r9d jge 0x2a9439 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%r14,%rdx), %edi incq %r9 jmp 0x2a941a addl %r11d, %edi incl %esi addq %r9, %r8 jmp 0x2a9406 leal (,%r15,8), %ecx movl %ecx, 0x130(%rsp) movl 0x110(%rsp), %ecx shll $0x3, %ecx movslq %ecx, %rcx movq %rcx, 0x20(%rsp) xorl %ecx, %ecx testl %r15d, %r15d cmovlel %ecx, %r15d movq 0xf0(%rsp), %r12 testl %r12d, %r12d cmovlel %ecx, %r12d movq %r12, 0xf0(%rsp) shlq $0x2, %r15 xorl %edx, %edx cmpq 0xf0(%rsp), %rdx je 0x2a95d0 movq %rcx, 0x70(%rsp) movslq %ecx, %r8 movq 0x40(%r13), %r9 imulq %rdx, %r9 imulq 0x10(%r13), %r9 shlq $0x2, %r8 addq (%r13), %r9 movslq 0xbc(%rsp), %r10 movq 0xd0(%rsp), %r11 imulq %rdx, %r11 movq 0xa0(%rsp), %rcx imulq %rcx, %r11 addq 0x90(%rsp), %r11 imulq %rcx, %r10 movq %rdx, 0x110(%rsp) movq %rdx, %r13 shlq $0x5, %r13 addq 0x28(%r14), %r8 xorl %ebp, %ebp cmpl 0x50(%rsp), %ebp jg 0x2a95af movq (%r14), %rcx xorl %r12d, %r12d cmpq %rbx, %r12 jg 0x2a959f movq -0x18(%rcx), %rsi cmpl $0x0, 0x100(%r14,%rsi) je 0x2a952e movq 0x1b0(%r14,%rsi), %rdi vmovups (%rdi,%r13), %ymm0 jmp 0x2a9532 vxorps %xmm0, %xmm0, %xmm0 movslq 0xe8(%r14,%rsi), %rdi movq %r14, %rdx movslq %ebp, %r14 imulq %rdi, %r14 imulq %r10, %r14 addq %r11, %r14 movl 0xe4(%rdx,%rsi), %esi imull %r12d, %esi shll $0x3, %esi movslq %esi, %rsi leaq (%r14,%rsi,4), %rsi xorl %edi, %edi cmpq %rdi, %r15 je 0x2a9585 movslq (%rax,%rdi), %r14 shlq $0x5, %r14 vmovups (%r8,%rdi,8), %ymm1 vmulps (%rsi,%r14), %ymm1, %ymm1 vaddps %ymm0, %ymm1, %ymm0 addq $0x4, %rdi jmp 0x2a9562 movq %r12, %rsi shlq $0x5, %rsi vmovups %ymm0, (%r9,%rsi) incq %r12 movq 0x10(%rsp), %r14 jmp 0x2a9506 movq 0x20(%rsp), %rcx leaq (%r9,%rcx,4), %r9 incl %ebp jmp 0x2a94f6 movq 0x110(%rsp), %rdx incq %rdx movq 0x70(%rsp), %rcx addl 0x130(%rsp), %ecx movq 0x18(%rsp), %r13 jmp 0x2a948a movq 0x8(%r14), %rdi testq %rdi, %rdi je 0x2a95ea movq (%rdi), %rax movq %r13, %rsi movq 0x48(%rsp), %rdx vzeroupper callq *0x48(%rax) leaq 0x210(%rsp), %rdi vzeroupper callq 0x624be xorl %r12d, %r12d jmp 0x2a96d0 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x1a0(%rsp), %xmm0 movq 0x18(%rsp), %rcx vmovups %xmm0, (%rcx) movq 0x1b0(%rsp), %rax movq %rax, 0x10(%rcx) movl 0x1b8(%rsp), %eax movl %eax, 0x18(%rcx) movq 0x1c0(%rsp), %rax movq %rax, 0x20(%rcx) vmovups 0x1c8(%rsp), %xmm0 vmovups %xmm0, 0x28(%rcx) movl 0x1d8(%rsp), %eax movl %eax, 0x38(%rcx) movq 0x1e0(%rsp), %rax movq %rax, 0x40(%rcx) movq 0x1a8(%rsp), %rax testq %rax, %rax je 0x2a9699 lock decl (%rax) jne 0x2a9699 movq 0x1a0(%rsp), %rsi movq 0x1c0(%rsp), %rdi testq %rdi, %rdi je 0x2a9691 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a9699 movq %rsi, %rdi callq 0x5f3e0 movq 0x218(%rsp), %rax testq %rax, %rax je 0x2a96d0 lock decl (%rax) jne 0x2a96d0 movq 0x210(%rsp), %rsi movq 0x230(%rsp), %rdi testq %rdi, %rdi je 0x2a96c8 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a96d0 movq %rsi, %rdi callq 0x5f3e0 movq 0x98(%rsp), %rax testq %rax, %rax je 0x2a970d lock decl (%rax) jne 0x2a970d movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi je 0x2a9702 movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x2a970d movq %rsi, %rdi vzeroupper callq 0x5f3e0 movl %r12d, %eax addq $0x3b8, %rsp # imm = 0x3B8 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movq 0x8(%r14), %rdi testq %rdi, %rdi je 0x2a95fa movq (%rdi), %rax movq %r13, %rsi movq 0x48(%rsp), %rdx vzeroupper callq *0x48(%rax) jmp 0x2a95fa movq %rax, %rbx leaq 0x210(%rsp), %rdi callq 0x624be jmp 0x2a9873 jmp 0x2a978c jmp 0x2a978c jmp 0x2a98ba jmp 0x2a98ba movq %rax, %rbx jmp 0x2a983c movq %rax, %rbx jmp 0x2a9805 jmp 0x2a98ba jmp 0x2a978c jmp 0x2a98ba jmp 0x2a98ba movq %rax, %rbx jmp 0x2a9873 movq %rax, %rbx movq 0x2c8(%rsp), %rax testq %rax, %rax je 0x2a97ce lock decl (%rax) jne 0x2a97ce movq 0x2c0(%rsp), %rsi movq 0x2e0(%rsp), %rdi testq %rdi, %rdi jne 0x2a97c8 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a97ce movq (%rdi), %rax callq *0x18(%rax) movq 0x310(%rsp), %rax testq %rax, %rax je 0x2a9805 lock decl (%rax) jne 0x2a9805 movq 0x308(%rsp), %rsi movq 0x328(%rsp), %rdi testq %rdi, %rdi jne 0x2a97ff movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a9805 movq (%rdi), %rax callq *0x18(%rax) movq 0x1a8(%rsp), %rax testq %rax, %rax je 0x2a983c lock decl (%rax) jne 0x2a983c movq 0x1a0(%rsp), %rsi movq 0x1c0(%rsp), %rdi testq %rdi, %rdi jne 0x2a9836 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a983c movq (%rdi), %rax callq *0x18(%rax) movq 0x218(%rsp), %rax testq %rax, %rax je 0x2a9873 lock decl (%rax) jne 0x2a9873 movq 0x210(%rsp), %rsi movq 0x230(%rsp), %rdi testq %rdi, %rdi jne 0x2a986d movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a9873 movq (%rdi), %rax callq *0x18(%rax) movq 0x98(%rsp), %rax testq %rax, %rax je 0x2a98aa lock decl (%rax) jne 0x2a98aa movq 0x90(%rsp), %rsi movq 0xb0(%rsp), %rdi testq %rdi, %rdi jne 0x2a98a4 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2a98aa movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2a98ba jmp 0x2a98ba jmp 0x2a98ba jmp 0x2a98ba movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

ncnn::ConvolutionDepthWise_x86_avx::forward_int8_x86(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx::forward_int8_x86(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; int elempack = bottom_blob.elempack; int elembits = bottom_blob.elembits(); const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_int8 = bottom_blob; if (elembits != 8) { const int channels_g = channels * elempack / group; Mat scales(channels * elempack); { float* ps = scales; for (int g = 0; g < group; g++) { float scale = bottom_blob_int8_scales[g]; for (int q = 0; q < channels_g; q++) { *ps++ = scale; } } } Option opt_q = opt; opt_q.blob_allocator = opt.workspace_allocator; quantize_to_int8(bottom_blob, bottom_blob_int8, scales, opt_q); } Mat bottom_blob_bordered; make_padding(bottom_blob_int8, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; channels = bottom_blob_bordered.c; elempack = bottom_blob_bordered.elempack; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; // depth-wise if (channels * elempack == group && group == num_output) { int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { out_elempack = num_output % 8 == 0 ? 8 : 1; } #endif // __SSE2__ bool use_int8_requantize = int8_scale_term > 100; size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #if __SSE2__ if (elempack == 8) { { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128i _sum0 = _mm_setzero_si128(); __m128i _sum1 = _mm_setzero_si128(); const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { // TODO use _mm_cvtepi8_epi16 on sse4.1 __m128i _val = _mm_loadl_epi64((const __m128i*)(sptr + space_ofs[k] * 8)); _val = _mm_unpacklo_epi8(_val, _mm_cmpgt_epi8(_mm_setzero_si128(), _val)); __m128i _w = _mm_loadl_epi64((const __m128i*)(kptr + k * 8)); _w = _mm_unpacklo_epi8(_w, _mm_cmpgt_epi8(_mm_setzero_si128(), _w)); __m128i _sl = _mm_mullo_epi16(_val, _w); __m128i _sh = _mm_mulhi_epi16(_val, _w); __m128i _s0 = _mm_unpacklo_epi16(_sl, _sh); __m128i _s1 = _mm_unpackhi_epi16(_sl, _sh); _sum0 = _mm_add_epi32(_sum0, _s0); _sum1 = _mm_add_epi32(_sum1, _s1); } __m128 _scale_in0; __m128 _scale_in1; { __m128 _bottom_blob_int8_scales0 = _mm_loadu_ps((const float*)bottom_blob_int8_scales + g * 8); __m128 _bottom_blob_int8_scales1 = _mm_loadu_ps((const float*)bottom_blob_int8_scales + g * 8 + 4); __m128 _weight_data_int8_scales0 = _mm_loadu_ps((const float*)weight_data_int8_scales + g * 8); __m128 _weight_data_int8_scales1 = _mm_loadu_ps((const float*)weight_data_int8_scales + g * 8 + 4); _scale_in0 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales0, _weight_data_int8_scales0)); _scale_in1 = _mm_rcp_ps(_mm_mul_ps(_bottom_blob_int8_scales1, _weight_data_int8_scales1)); __m128 _m0 = _mm_cmpneq_ps(_weight_data_int8_scales0, _mm_setzero_ps()); __m128 _m1 = _mm_cmpneq_ps(_weight_data_int8_scales1, _mm_setzero_ps()); _scale_in0 = _mm_and_ps(_scale_in0, _m0); _scale_in1 = _mm_and_ps(_scale_in1, _m1); } __m128 _sumfp32_0 = _mm_mul_ps(_mm_cvtepi32_ps(_sum0), _scale_in0); __m128 _sumfp32_1 = _mm_mul_ps(_mm_cvtepi32_ps(_sum1), _scale_in1); if (bias_term) { __m128 _bias0 = _mm_loadu_ps((const float*)bias_data + g * 8); __m128 _bias1 = _mm_loadu_ps((const float*)bias_data + g * 8 + 4); _sumfp32_0 = _mm_add_ps(_sumfp32_0, _bias0); _sumfp32_1 = _mm_add_ps(_sumfp32_1, _bias1); } _sumfp32_0 = activation_sse(_sumfp32_0, activation_type, activation_params); _sumfp32_1 = activation_sse(_sumfp32_1, activation_type, activation_params); if (use_int8_requantize) { // requantize and relu __m128 _scale_out0 = _mm_loadu_ps((const float*)top_blob_int8_scales + g * 8); __m128 _scale_out1 = _mm_loadu_ps((const float*)top_blob_int8_scales + g * 8 + 4); _sumfp32_0 = _mm_mul_ps(_sumfp32_0, _scale_out0); _sumfp32_1 = _mm_mul_ps(_sumfp32_1, _scale_out1); int64_t _sum8 = float2int8_sse(_sumfp32_0, _sumfp32_1); *(int64_t*)outptr_s8 = _sum8; outptr_s8 += 8; } else { // dequantize and relu _mm_storeu_ps(outptr_f32, _sumfp32_0); _mm_storeu_ps(outptr_f32 + 4, _sumfp32_1); outptr_f32 += 8; } } } } } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && stride_w == 1 && stride_h == 1 && dilation_w == 1 && dilation_h == 1 && (activation_type == 0 || activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s1_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s1_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2 && (activation_type == 0 || activation_type == 1)) { if (use_int8_requantize) { std::vector<float> requantize_scales; for (int g = 0; g < group; g++) { float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float scale_out = top_blob_int8_scales[g]; requantize_scales.push_back(scale_in); requantize_scales.push_back(scale_out); } convdw3x3s2_int8_requant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, requantize_scales, opt); } else { std::vector<float> dequantize_scales; for (int g = 0; g < group; g++) { float top_rescale = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); dequantize_scales.push_back(top_rescale); } convdw3x3s2_int8_dequant_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, dequantize_scales, opt); } if (activation) { activation->forward_inplace(top_blob, opt); } } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { signed char* outptr_s8 = top_blob.channel(g); float* outptr_f32 = top_blob.channel(g); const signed char* kptr = (const signed char*)weight_data_tm + maxk * g; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { int sum = 0; const signed char* sptr = m.row<const signed char>(i * stride_h) + j * stride_w; for (int k = 0; k < maxk; k++) { signed char val = sptr[space_ofs[k]]; signed char w = kptr[k]; sum += val * w; } float scale_in; if (weight_data_int8_scales[g] == 0) scale_in = 0; else scale_in = 1.f / (bottom_blob_int8_scales[g] * weight_data_int8_scales[g]); float sumfp32 = sum * scale_in; if (bias_term) sumfp32 += bias_data[g]; sumfp32 = activation_ss(sumfp32, activation_type, activation_params); if (use_int8_requantize) { // requantize float scale_out = top_blob_int8_scales[g]; signed char sums8 = float2int8(sumfp32 * scale_out); outptr_s8[0] = sums8; outptr_s8 += 1; } else { // dequantize outptr_f32[0] = sumfp32; outptr_f32 += 1; } } } } } } return 0; } bool use_int8_requantize = int8_scale_term > 100; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { if (use_int8_requantize) out_elempack = num_output % 8 == 0 ? 8 : 1; else out_elempack = num_output % 4 == 0 ? 4 : 1; } #endif // __SSE2__ size_t out_elemsize = use_int8_requantize ? 1u * out_elempack : 4u * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { g_elempack = channels_g % 8 == 0 ? 8 : 1; if (use_int8_requantize) out_g_elempack = num_output_g % 8 == 0 ? 8 : 1; else out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x318, %rsp # imm = 0x318 movq %rcx, 0x30(%rsp) movq %rdx, 0x20(%rsp) movq %rsi, %r15 movl 0x38(%rsi), %ecx movl 0x18(%rsi), %r12d movq 0x10(%rsi), %rsi testl %r12d, %r12d je 0x2a9903 leal (,%rsi,8), %eax cltd idivl %r12d cmpl $0x8, %eax sete %dl jmp 0x2a9905 xorl %edx, %edx movq (%rdi), %rax movq -0x18(%rax), %r9 movq %rdi, %r8 movl 0xd4(%rdi,%r9), %ebp movl 0xd8(%rdi,%r9), %ebx decl %ebp imull 0xdc(%rdi,%r9), %ebp decl %ebx imull 0xe0(%rdi,%r9), %ebx movq 0x8(%r15), %rdi vmovups (%r15), %xmm0 vmovaps %xmm0, 0x1e0(%rsp) movq %rsi, 0x1f0(%rsp) movl %r12d, 0x1f8(%rsp) movq 0x20(%r15), %rsi movq %rsi, 0x200(%rsp) vmovdqu 0x28(%r15), %xmm0 vmovdqu %xmm0, 0x208(%rsp) movl %ecx, 0x218(%rsp) movq 0x40(%r15), %rsi movq %rsi, 0x220(%rsp) testq %rdi, %rdi je 0x2a9990 lock incl (%rdi) movq (%r8), %rax movq %r8, 0x8(%rsp) testb %dl, %dl je 0x2a99a3 movq 0x8(%rsp), %rcx jmp 0x2a9ab6 imull %ecx, %r12d movq -0x18(%rax), %rax movq 0x8(%rsp), %r13 movl 0x108(%r13,%rax), %eax movl %eax, 0x10(%rsp) leaq 0x70(%rsp), %rdi andq $0x0, 0x40(%rdi) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdi) vmovdqu %xmm0, 0xc(%rdi) vmovdqa %xmm0, 0x20(%rdi) vmovdqu %xmm0, 0x2c(%rdi) xorl %r14d, %r14d pushq $0x4 popq %rdx movl %r12d, %esi xorl %ecx, %ecx callq 0x635fa movl %r12d, %eax cltd idivl 0x10(%rsp) movq 0x70(%rsp), %rcx movq (%r13), %rdx testl %eax, %eax cmovlel %r14d, %eax movq 0x8(%rsp), %r8 movq -0x18(%rdx), %rsi movslq 0x108(%r8,%rsi), %rdi cmpq %rdi, %r14 jge 0x2a9a3e movq 0x240(%r8,%rsi), %rsi vmovd (%rsi,%r14,4), %xmm0 movl %eax, %esi subl $0x1, %esi jb 0x2a9a39 vmovd %xmm0, (%rcx) addq $0x4, %rcx jmp 0x2a9a2a incq %r14 jmp 0x2a9a09 movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0x130(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x1e0(%rsp), %rsi leaq 0x70(%rsp), %rdx movq %r15, %rdi vzeroupper callq 0x652e3 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2a9aae lock decl (%rax) jne 0x2a9aae movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2a9aa6 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2a9aae movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx movq (%rcx), %rax leaq 0x70(%rsp), %rdx andq $0x0, 0x40(%rdx) vpxor %xmm0, %xmm0, %xmm0 vmovdqa %xmm0, (%rdx) vmovdqu %xmm0, 0xc(%rdx) vmovdqa %xmm0, 0x20(%rdx) vmovdqu %xmm0, 0x2c(%rdx) movq -0x18(%rax), %rdi addq %rcx, %rdi leaq 0x1e0(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x287daa pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2ab463 movslq 0xa8(%rsp), %r14 movq 0xb0(%rsp), %rax imulq %r14, %rax testq %rax, %rax je 0x2ab463 notl %ebp movl 0x88(%rsp), %r13d movl 0x9c(%rsp), %edi addl %edi, %ebp movq 0x8(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx movl %ebp, %eax movq %rdx, %rbp cltd idivl 0xe4(%rbp,%rcx) movl %eax, %esi notl %ebx addl 0xa0(%rsp), %ebx movl %ebx, %eax cltd idivl 0xe8(%rbp,%rcx) movl %eax, %edx movq %rsi, 0x10(%rsp) incl %esi movq %rdx, 0x68(%rsp) leal 0x1(%rdx), %r10d movl %r13d, %ebx imull %r14d, %ebx pushq $0x8 popq %r15 cmpl 0x108(%rbp,%rcx), %ebx jne 0x2a9cca cmpl 0xd0(%rbp,%rcx), %ebx jne 0x2a9cca movl %edi, 0x18(%rsp) movq 0x30(%rsp), %rdi cmpb $0x0, 0x27(%rdi) pushq $0x1 popq %rax cmovel %eax, %r15d testb $0x7, %bl cmovnel %eax, %r15d movl 0x10c(%rbp,%rcx), %eax leal (,%r15,4), %r8d movl %eax, 0x38(%rsp) cmpl $0x65, %eax cmovgel %r15d, %r8d movl %ebx, %eax cltd idivl %r15d movq 0x8(%rdi), %rcx movq %rcx, (%rsp) movq 0x20(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %eax, %ecx movl %r15d, %r9d callq 0x628f2 cmpq $0x0, (%rbx) je 0x2ab463 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ab463 xorl %r12d, %r12d cmpl $0x1, %r13d je 0x2aa2a7 cmpl $0x8, %r13d jne 0x2ab463 movq (%rbp), %rax movq -0x18(%rax), %rax movslq 0xd4(%rbp,%rax), %rcx movslq 0xd8(%rbp,%rax), %r15 imulq %rcx, %r15 leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %r15, %rsi callq 0x73bbe movq (%rbp), %rcx movq -0x18(%rcx), %rdx movl 0x18(%rsp), %r11d imull 0xe0(%rbp,%rdx), %r11d movq 0x130(%rsp), %rax movl 0xdc(%rbp,%rdx), %esi imull 0xd4(%rbp,%rdx), %esi subl %esi, %r11d xorl %esi, %esi xorl %edi, %edi xorl %r8d, %r8d cmpl 0xd8(%rbp,%rdx), %esi jge 0x2aa3d8 movslq %r8d, %r8 leaq (%rax,%r8,4), %r10 xorl %r9d, %r9d cmpl 0xd4(%rbp,%rdx), %r9d jge 0x2a9cc0 movl %edi, (%r10,%r9,4) movq -0x18(%rcx), %rdx addl 0xdc(%rbp,%rdx), %edi incq %r9 jmp 0x2a9ca2 addl %r11d, %edi incl %esi addq %r9, %r8 jmp 0x2a9c8b movl 0xd0(%rbp,%rcx), %eax movl 0x10c(%rbp,%rcx), %r14d movq 0x30(%rsp), %rcx cmpb $0x1, 0x27(%rcx) movl %r13d, 0x40(%rsp) movl %r12d, %r13d jne 0x2a9cfe cmpl $0x65, %r14d jl 0x2a9d04 testb $0x7, %al pushq $0x1 popq %r9 cmovel %r15d, %r9d jmp 0x2a9d12 pushq $0x1 popq %r9 jmp 0x2a9d12 xorl %ecx, %ecx testb $0x3, %al sete %cl leal (%rcx,%rcx,2), %r9d incl %r9d leal (,%r9,4), %r8d cmpl $0x65, %r14d cmovgel %r9d, %r8d cltd idivl %r9d movq 0x30(%rsp), %rcx movq 0x8(%rcx), %rcx movq %rcx, (%rsp) movq 0x20(%rsp), %r12 movq %r12, %rdi movl %esi, 0x18(%rsp) movl %r10d, 0x28(%rsp) movl %r10d, %edx movl %eax, %ecx movq %r8, 0x48(%rsp) movq %r9, 0x38(%rsp) callq 0x628f2 cmpq $0x0, (%r12) je 0x2a9dd1 movslq 0x38(%r12), %rax imulq 0x40(%r12), %rax testq %rax, %rax movl %r13d, %r12d movl 0x40(%rsp), %r8d je 0x2ab463 movq (%rbp), %rax movq -0x18(%rax), %rax movl 0xd0(%rbp,%rax), %ecx movl 0x108(%rbp,%rax), %esi movl %ebx, %eax cltd idivl %esi movl %eax, %edi movl %ecx, %eax cltd idivl %esi movl %eax, %ebp movq 0x30(%rsp), %rax cmpb $0x1, 0x27(%rax) jne 0x2a9dd9 testb $0x7, %dil pushq $0x1 popq %rbx movl %ebx, %r12d cmovel %r15d, %r12d cmpl $0x65, %r14d movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d jl 0x2a9deb testb $0x7, %bpl cmovel %r15d, %ebx jmp 0x2a9df9 movl %r13d, %r12d jmp 0x2ab463 pushq $0x1 popq %r12 movl %r12d, %ebx movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d jmp 0x2a9df9 xorl %eax, %eax testb $0x3, %bpl sete %al leal (%rax,%rax,2), %ebx incl %ebx movq 0x78(%rsp), %rax vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0x130(%rsp) movq 0x80(%rsp), %rcx movq %rcx, 0x140(%rsp) movl 0x88(%rsp), %ecx movl %ecx, 0x148(%rsp) movq 0x90(%rsp), %rcx movq %rcx, 0x150(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0x158(%rsp) movl 0xa8(%rsp), %ecx movl %ecx, 0x168(%rsp) movq 0xb0(%rsp), %rcx movq %rcx, 0x170(%rsp) testq %rax, %rax movq 0x20(%rsp), %r14 je 0x2a9e78 lock incl (%rax) cmpl %r12d, %r8d movl %edi, 0x10(%rsp) jle 0x2a9ecd movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 leaq 0xe0(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovups %ymm0, (%rcx) movq 0x10(%rax), %rax movq %rax, 0x8(%rcx) leaq 0x70(%rsp), %rdi leaq 0x130(%rsp), %rsi movl %r12d, %edx vzeroupper callq 0x64e3b movl 0x10(%rsp), %edi movl 0x18(%rsp), %esi movl 0x28(%rsp), %r9d movq 0x8(%r14), %rax vmovups (%r14), %xmm0 vmovaps %xmm0, 0xe0(%rsp) movq 0x10(%r14), %rcx movq %rcx, 0xf0(%rsp) movl 0x18(%r14), %ecx movl %ecx, 0xf8(%rsp) movq 0x20(%r14), %rcx movq %rcx, 0x100(%rsp) vmovups 0x28(%r14), %xmm0 vmovups %xmm0, 0x108(%rsp) movl 0x38(%r14), %ecx movl %ecx, 0x118(%rsp) movq 0x40(%r14), %rcx movq %rcx, 0x120(%rsp) testq %rax, %rax je 0x2a9f30 lock incl (%rax) cmpl 0x38(%rsp), %ebx jae 0x2a9fb0 movq 0x8(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xd0(%rcx,%rax), %eax cltd idivl %ebx movl %eax, %ecx movzbl 0x48(%rsp), %eax divb 0x38(%rsp) movzbl %al, %r8d imull %ebx, %r8d movq 0x30(%rsp), %rax movq 0x10(%rax), %rax movq %rax, (%rsp) leaq 0xe0(%rsp), %rdi movl %r9d, %edx movl %ebx, %r9d callq 0x628f2 movl 0x10(%rsp), %edi pushq $-0x64 popq %r14 cmpq $0x0, 0xe0(%rsp) je 0x2ab3f2 movslq 0x118(%rsp), %rax imulq 0x120(%rsp), %rax testq %rax, %rax je 0x2ab3f2 xorl %r15d, %r15d xorl %r13d, %r13d xorl %r14d, %r14d movq 0x8(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rax movslq 0x108(%r8,%rax), %rax cmpq %rax, %r14 jge 0x2aa225 movl %r15d, %eax cltd idivl %r12d movl %eax, %ecx movl %edi, %eax cltd idivl %r12d movslq %ecx, %rdx imulq 0x170(%rsp), %rdx movq 0x140(%rsp), %rcx imulq %rcx, %rdx addq 0x130(%rsp), %rdx movl 0x148(%rsp), %esi movq 0x150(%rsp), %rdi movq %rdx, 0x198(%rsp) andq $0x0, 0x1a0(%rsp) movq %rcx, 0x1a8(%rsp) movl %esi, 0x1b0(%rsp) movq %rdi, 0x1b8(%rsp) movl %eax, 0x1d0(%rsp) vmovups 0x158(%rsp), %xmm0 movslq 0x164(%rsp), %rax movslq 0x15c(%rsp), %rdx movslq 0x160(%rsp), %rsi imulq %rdx, %rsi imulq %rcx, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rcx movq %rax, 0x1d8(%rsp) vmovups %xmm0, 0x1c0(%rsp) movl %r13d, %eax cltd idivl %ebx movl %eax, %ecx movl %ebp, %eax cltd idivl %ebx movslq %ecx, %rdx imulq 0x120(%rsp), %rdx movq 0xf0(%rsp), %rsi imulq %rsi, %rdx addq 0xe0(%rsp), %rdx movl 0xf8(%rsp), %edi movq 0x100(%rsp), %rcx movq %rdx, 0x228(%rsp) andq $0x0, 0x230(%rsp) movq %rsi, 0x238(%rsp) movl %edi, 0x240(%rsp) movq %rcx, 0x248(%rsp) movl %eax, 0x260(%rsp) vmovups 0x108(%rsp), %xmm0 movslq 0x114(%rsp), %rax movslq 0x10c(%rsp), %rdx movslq 0x110(%rsp), %rdi imulq %rdx, %rdi imulq %rsi, %rax imulq %rdi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rsi movq %rax, 0x268(%rsp) vmovups %xmm0, 0x250(%rsp) movq 0x10(%r8), %rax movq (%rax,%r14,8), %rdi movq 0x30(%rsp), %rax vmovups (%rax), %ymm0 vmovups 0x20(%rax), %ymm1 vmovups %ymm0, 0x2d0(%rsp) vmovups %ymm1, 0x2f0(%rsp) movq %rcx, 0x2d8(%rsp) movq (%rdi), %rax leaq 0x198(%rsp), %rsi leaq 0x228(%rsp), %rdx leaq 0x2d0(%rsp), %rcx vzeroupper callq *0x38(%rax) movq 0x230(%rsp), %rax testq %rax, %rax movl 0x10(%rsp), %edi je 0x2aa1d8 lock decl (%rax) jne 0x2aa1d8 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi je 0x2aa1cc movq (%rdi), %rax callq *0x18(%rax) movl 0x10(%rsp), %edi jmp 0x2aa1d8 movq %rsi, %rdi callq 0x5f3e0 movl 0x10(%rsp), %edi movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x2aa217 lock decl (%rax) jne 0x2aa217 movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi je 0x2aa20b movq (%rdi), %rax callq *0x18(%rax) movl 0x10(%rsp), %edi jmp 0x2aa217 movq %rsi, %rdi callq 0x5f3e0 movl 0x10(%rsp), %edi incq %r14 addl %ebp, %r13d addl %edi, %r15d jmp 0x2a9fb9 movq 0x38(%rsp), %rdx cmpl %edx, %ebx jae 0x2aa24d xorl %r14d, %r14d leaq 0xe0(%rsp), %rdi movq 0x20(%rsp), %rsi movq 0x30(%rsp), %rcx callq 0x64e3b jmp 0x2ab3f2 leaq 0xe0(%rsp), %rax movq 0x20(%rsp), %rbx xorl %r14d, %r14d cmpq %rbx, %rax je 0x2ab3f2 movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2aa276 lock incl (%rax) movq 0x8(%rbx), %rax testq %rax, %rax je 0x2ab39f lock decl (%rax) jne 0x2ab39f movq (%rbx), %rsi movq 0x20(%rbx), %rdi testq %rdi, %rdi je 0x2ab397 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ab39f movq (%rbp), %rax movq -0x18(%rax), %rdx movl 0xd4(%rbp,%rdx), %ecx movl 0xd8(%rbp,%rdx), %r13d movl %ecx, %esi xorl $0x3, %esi movl %r13d, %edi xorl $0x3, %edi orl %esi, %edi movq 0x20(%rsp), %r14 jne 0x2aaf50 cmpl $0x1, 0xe4(%rbp,%rdx) jne 0x2aae45 cmpl $0x1, 0xe8(%rbp,%rdx) jne 0x2aae45 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x2aae45 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x2aae45 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x2aae45 cmpl $0x65, 0x38(%rsp) jl 0x2ab4e0 vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx vmovaps %xmm0, (%rbx) andq $0x0, 0x10(%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 leaq 0x198(%rsp), %r12 movq -0x18(%rax), %r13 movslq 0x108(%rbp,%r13), %rax cmpq %rax, %r14 jge 0x2ab821 movq 0x1f8(%rbp,%r13), %rax vmovss (%rax,%r14,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2aa396 movq 0x240(%rbp,%r13), %rax vmulss (%rax,%r14,4), %xmm1, %xmm0 vmovss 0x1448f6(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rbp,%r13), %rax vmovss (%rax,%r14,4), %xmm0 vmovss %xmm0, 0x198(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c movq %rbx, %rdi movq %r12, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2aa34f leal (,%r15,8), %ecx movl %ecx, 0x28(%rsp) xorl %edx, %edx testl %r15d, %r15d cmovlel %edx, %r15d testl %r14d, %r14d cmovlel %edx, %r14d vpxor %xmm0, %xmm0, %xmm0 vbroadcastss 0x143c13(%rip), %xmm2 # 0x3ee014 vbroadcastss 0x14487e(%rip), %xmm8 # 0x3eec88 xorl %esi, %esi movq %r14, 0x18(%rsp) cmpq %r14, %rsi je 0x2ab382 movq 0x20(%rsp), %rcx movq 0x10(%rcx), %r8 imulq %rsi, %r8 imulq 0x40(%rcx), %r8 movq %rdx, 0x30(%rsp) movslq %edx, %r9 addq (%rcx), %r8 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rsi, %r11 movq 0x80(%rsp), %rcx imulq %rcx, %r11 addq 0x70(%rsp), %r11 imulq %rcx, %r10 movq %rsi, 0x40(%rsp) leaq (,%rsi,8), %rbx addq 0x28(%rbp), %r9 xorl %ecx, %ecx movq %r8, %rdx cmpl 0x68(%rsp), %ecx jg 0x2aae2a movq (%rbp), %rdi xorl %r13d, %r13d cmpl 0x10(%rsp), %r13d jg 0x2aae23 movq -0x18(%rdi), %rsi movslq 0xe8(%rbp,%rsi), %r14 movslq %ecx, %r12 imulq %r14, %r12 imulq %r10, %r12 movl 0xe4(%rbp,%rsi), %ebp imull %r13d, %ebp shll $0x3, %ebp movslq %ebp, %r14 addq %r11, %r14 addq %r12, %r14 vpxor %xmm3, %xmm3, %xmm3 xorl %r12d, %r12d vpxor %xmm4, %xmm4, %xmm4 cmpq %r12, %r15 je 0x2aa50d movslq (%rax,%r12,4), %rbp vmovq (%r14,%rbp,8), %xmm5 vpcmpgtb %xmm5, %xmm0, %xmm6 vpunpcklbw %xmm6, %xmm5, %xmm5 # xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7] vmovq (%r9,%r12,8), %xmm6 vpcmpgtb %xmm6, %xmm0, %xmm7 vpunpcklbw %xmm7, %xmm6, %xmm6 # xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3],xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7] vpmullw %xmm5, %xmm6, %xmm7 vpmulhw %xmm6, %xmm5, %xmm5 vpunpcklwd %xmm5, %xmm7, %xmm6 # xmm6 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3] vpunpckhwd %xmm5, %xmm7, %xmm5 # xmm5 = xmm7[4],xmm5[4],xmm7[5],xmm5[5],xmm7[6],xmm5[6],xmm7[7],xmm5[7] vpaddd %xmm6, %xmm4, %xmm4 vpaddd %xmm5, %xmm3, %xmm3 incq %r12 jmp 0x2aa4cb movq 0x8(%rsp), %rbp movq 0x1f8(%rbp,%rsi), %r14 movq 0x240(%rbp,%rsi), %r12 vmovups (%r14,%rbx,4), %xmm5 vmovups 0x10(%r14,%rbx,4), %xmm15 vmulps (%r12,%rbx,4), %xmm5, %xmm6 vmulps 0x10(%r12,%rbx,4), %xmm15, %xmm7 vrcpps %xmm6, %xmm6 vrcpps %xmm7, %xmm7 vcmpneqps %xmm0, %xmm5, %xmm5 vandps %xmm6, %xmm5, %xmm5 vcmpneqps %xmm0, %xmm15, %xmm6 vandps %xmm7, %xmm6, %xmm6 vcvtdq2ps %xmm4, %xmm4 vmulps %xmm4, %xmm5, %xmm4 vcvtdq2ps %xmm3, %xmm3 vmulps %xmm3, %xmm6, %xmm5 cmpl $0x0, 0x100(%rbp,%rsi) je 0x2aa585 movq 0x1b0(%rbp,%rsi), %r14 vaddps (%r14,%rbx,4), %xmm4, %xmm4 vaddps 0x10(%r14,%rbx,4), %xmm5, %xmm5 movl 0x110(%rbp,%rsi), %r14d decl %r14d cmpl $0x5, %r14d ja 0x2aad9f leaq 0x14e14f(%rip), %r12 # 0x3f86f0 movslq (%r12,%r14,4), %r14 addq %r12, %r14 vmaxps %xmm0, %xmm4, %xmm15 vmaxps %xmm0, %xmm5, %xmm3 jmpq *%r14 movq 0x118(%rbp,%rsi), %r14 vminps %xmm0, %xmm4, %xmm4 vbroadcastss (%r14), %xmm6 vmulps %xmm4, %xmm6, %xmm4 vaddps %xmm4, %xmm15, %xmm15 vminps %xmm0, %xmm5, %xmm4 vmulps %xmm4, %xmm6, %xmm4 jmp 0x2aad5c vbroadcastss 0x146bd6(%rip), %xmm1 # 0x3f11b8 vminps %xmm1, %xmm4, %xmm3 vbroadcastss 0x146bcd(%rip), %xmm1 # 0x3f11bc vmaxps %xmm1, %xmm3, %xmm3 vbroadcastss 0x146bc4(%rip), %xmm1 # 0x3f11c0 vmulps %xmm1, %xmm3, %xmm6 vaddps %xmm2, %xmm6, %xmm6 vcvttps2dq %xmm6, %xmm7 vcvtdq2ps %xmm7, %xmm7 vcmpltps %xmm7, %xmm6, %xmm6 vandps %xmm6, %xmm8, %xmm6 vsubps %xmm6, %xmm7, %xmm6 vbroadcastss 0x1472fe(%rip), %xmm9 # 0x3f1920 vmulps %xmm6, %xmm9, %xmm7 vsubps %xmm7, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm7 vbroadcastss 0x146b95(%rip), %xmm10 # 0x3f11cc vmulps %xmm3, %xmm10, %xmm15 vmovaps %xmm10, %xmm12 vbroadcastss 0x146b87(%rip), %xmm10 # 0x3f11d0 vaddps %xmm10, %xmm15, %xmm15 vmovaps %xmm10, %xmm13 vmulps %xmm3, %xmm15, %xmm15 vbroadcastss 0x146b74(%rip), %xmm11 # 0x3f11d4 vaddps %xmm11, %xmm15, %xmm15 vmovaps %xmm11, %xmm14 vmulps %xmm3, %xmm15, %xmm15 vbroadcastss 0x146b61(%rip), %xmm1 # 0x3f11d8 vaddps %xmm1, %xmm15, %xmm15 vmulps %xmm3, %xmm15, %xmm15 vbroadcastss 0x146b54(%rip), %xmm10 # 0x3f11dc vaddps %xmm10, %xmm15, %xmm15 vmulps %xmm3, %xmm15, %xmm15 vaddps %xmm2, %xmm15, %xmm15 vmulps %xmm7, %xmm15, %xmm7 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm7, %xmm3, %xmm3 vcvttps2dq %xmm6, %xmm6 vpslld $0x17, %xmm6, %xmm6 vpaddd %xmm6, %xmm8, %xmm6 vmulps %xmm6, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm6 vcmpleps %xmm0, %xmm6, %xmm3 vbroadcastss 0x146b1c(%rip), %xmm7 # 0x3f11e0 vmaxps %xmm7, %xmm6, %xmm6 vpsrld $0x17, %xmm6, %xmm7 vbroadcastss 0x146b0e(%rip), %xmm10 # 0x3f11e4 vandps %xmm6, %xmm10, %xmm6 vorps %xmm2, %xmm6, %xmm6 vbroadcastss 0x146b01(%rip), %xmm10 # 0x3f11e8 vpaddd %xmm7, %xmm10, %xmm7 vcvtdq2ps %xmm7, %xmm7 vbroadcastss 0x146af4(%rip), %xmm10 # 0x3f11ec vcmpltps %xmm10, %xmm6, %xmm15 vandps %xmm6, %xmm15, %xmm10 vbroadcastss 0x146ae5(%rip), %xmm1 # 0x3f11f0 vaddps %xmm1, %xmm6, %xmm6 vaddps %xmm6, %xmm10, %xmm6 vandps %xmm8, %xmm15, %xmm10 vsubps %xmm10, %xmm7, %xmm7 vmulps %xmm6, %xmm6, %xmm10 vbroadcastss 0x146aca(%rip), %xmm15 # 0x3f11f4 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146ac1(%rip), %xmm11 # 0x3f11f8 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146ab3(%rip), %xmm11 # 0x3f11fc vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146aa5(%rip), %xmm11 # 0x3f1200 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a97(%rip), %xmm11 # 0x3f1204 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a89(%rip), %xmm11 # 0x3f1208 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a7b(%rip), %xmm11 # 0x3f120c vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a6d(%rip), %xmm11 # 0x3f1210 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vbroadcastss 0x146a5f(%rip), %xmm11 # 0x3f1214 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm6, %xmm15, %xmm15 vmulps %xmm7, %xmm9, %xmm7 vbroadcastss 0x147159(%rip), %xmm11 # 0x3f1924 vaddps %xmm11, %xmm15, %xmm15 vmulps %xmm15, %xmm10, %xmm10 vaddps %xmm6, %xmm7, %xmm6 vaddps %xmm6, %xmm10, %xmm6 vbroadcastss 0x147f32(%rip), %xmm11 # 0x3f2718 vmulps %xmm6, %xmm11, %xmm6 vbroadcastss 0x14710d(%rip), %xmm11 # 0x3f1900 vblendvps %xmm3, %xmm11, %xmm6, %xmm3 vbroadcastss 0x1469b6(%rip), %xmm1 # 0x3f11b8 vminps %xmm1, %xmm3, %xmm3 vbroadcastss 0x1469ad(%rip), %xmm1 # 0x3f11bc vmaxps %xmm1, %xmm3, %xmm3 vbroadcastss 0x1469a4(%rip), %xmm1 # 0x3f11c0 vmulps %xmm1, %xmm3, %xmm6 vaddps %xmm2, %xmm6, %xmm6 vcvttps2dq %xmm6, %xmm7 vcvtdq2ps %xmm7, %xmm7 vcmpltps %xmm7, %xmm6, %xmm6 vandps %xmm6, %xmm8, %xmm6 vsubps %xmm6, %xmm7, %xmm6 vmulps %xmm6, %xmm9, %xmm7 vsubps %xmm7, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm7 vmovaps %xmm12, %xmm1 vmulps %xmm3, %xmm12, %xmm10 vmovaps %xmm13, %xmm12 vaddps %xmm13, %xmm10, %xmm10 vmulps %xmm3, %xmm10, %xmm10 vmovaps %xmm14, %xmm13 vaddps %xmm14, %xmm10, %xmm10 vmulps %xmm3, %xmm10, %xmm10 vbroadcastss 0x146966(%rip), %xmm14 # 0x3f11d8 vaddps %xmm14, %xmm10, %xmm10 vmulps %xmm3, %xmm10, %xmm10 vbroadcastss 0x146958(%rip), %xmm11 # 0x3f11dc vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm3, %xmm10, %xmm10 vaddps %xmm2, %xmm10, %xmm10 vmulps %xmm7, %xmm10, %xmm7 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm7, %xmm3, %xmm3 vcvttps2dq %xmm6, %xmm6 vpslld $0x17, %xmm6, %xmm6 vpaddd %xmm6, %xmm8, %xmm6 vmulps %xmm6, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vrcpps %xmm3, %xmm6 vaddps %xmm6, %xmm6, %xmm7 vmulps %xmm7, %xmm3, %xmm3 vbroadcastss 0x147e41(%rip), %xmm11 # 0x3f2708 vsubps %xmm3, %xmm11, %xmm3 vmulps %xmm3, %xmm6, %xmm3 vbroadcastss 0x146918(%rip), %xmm11 # 0x3f11f0 vaddps %xmm7, %xmm11, %xmm6 vaddps %xmm3, %xmm6, %xmm3 vmulps %xmm4, %xmm3, %xmm15 vbroadcastss 0x1468cb(%rip), %xmm3 # 0x3f11b8 vminps %xmm3, %xmm5, %xmm3 vbroadcastss 0x1468c2(%rip), %xmm4 # 0x3f11bc vmaxps %xmm4, %xmm3, %xmm3 vbroadcastss 0x1468b9(%rip), %xmm4 # 0x3f11c0 vmulps %xmm4, %xmm3, %xmm4 vaddps %xmm2, %xmm4, %xmm4 vcvttps2dq %xmm4, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm4, %xmm4 vandps %xmm4, %xmm8, %xmm4 vsubps %xmm4, %xmm6, %xmm4 vmulps %xmm4, %xmm9, %xmm6 vsubps %xmm6, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm6 vmulps %xmm1, %xmm3, %xmm7 vaddps %xmm7, %xmm12, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm13, %xmm7 vmovaps %xmm13, %xmm12 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm14, %xmm7 vmovaps %xmm14, %xmm13 vmulps %xmm3, %xmm7, %xmm7 vbroadcastss 0x14687d(%rip), %xmm14 # 0x3f11dc vaddps %xmm7, %xmm14, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm2, %xmm7, %xmm7 vmulps %xmm7, %xmm6, %xmm6 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm4, %xmm8, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vbroadcastss 0x14684b(%rip), %xmm1 # 0x3f11e0 vmaxps %xmm1, %xmm3, %xmm4 vpsrld $0x17, %xmm4, %xmm6 vbroadcastss 0x14683d(%rip), %xmm1 # 0x3f11e4 vandps %xmm1, %xmm4, %xmm4 vorps %xmm2, %xmm4, %xmm4 vbroadcastss 0x146830(%rip), %xmm1 # 0x3f11e8 vpaddd %xmm1, %xmm6, %xmm6 vcvtdq2ps %xmm6, %xmm6 vbroadcastss 0x146823(%rip), %xmm1 # 0x3f11ec vcmpltps %xmm1, %xmm4, %xmm7 vandps %xmm4, %xmm7, %xmm10 vaddps %xmm4, %xmm11, %xmm4 vmovaps %xmm11, %xmm1 vaddps %xmm4, %xmm10, %xmm4 vandps %xmm7, %xmm8, %xmm7 vsubps %xmm7, %xmm6, %xmm6 vmulps %xmm4, %xmm4, %xmm7 vbroadcastss 0x146801(%rip), %xmm10 # 0x3f11f4 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467f8(%rip), %xmm11 # 0x3f11f8 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467ea(%rip), %xmm11 # 0x3f11fc vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467dc(%rip), %xmm11 # 0x3f1200 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467ce(%rip), %xmm11 # 0x3f1204 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467c0(%rip), %xmm11 # 0x3f1208 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467b2(%rip), %xmm11 # 0x3f120c vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x1467a4(%rip), %xmm11 # 0x3f1210 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x146796(%rip), %xmm11 # 0x3f1214 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm4, %xmm10, %xmm10 vbroadcastss 0x146e94(%rip), %xmm11 # 0x3f1924 vaddps %xmm11, %xmm10, %xmm10 vmulps %xmm7, %xmm10, %xmm7 vcmpleps %xmm0, %xmm3, %xmm3 vmulps %xmm6, %xmm9, %xmm6 vaddps %xmm4, %xmm6, %xmm4 vaddps %xmm7, %xmm4, %xmm4 vbroadcastss 0x147c65(%rip), %xmm6 # 0x3f2718 vmulps %xmm6, %xmm4, %xmm4 vbroadcastss 0x146e40(%rip), %xmm6 # 0x3f1900 vblendvps %xmm3, %xmm6, %xmm4, %xmm3 vbroadcastss 0x1466e9(%rip), %xmm4 # 0x3f11b8 vminps %xmm4, %xmm3, %xmm3 vbroadcastss 0x1466e0(%rip), %xmm4 # 0x3f11bc vmaxps %xmm4, %xmm3, %xmm3 vbroadcastss 0x1466d7(%rip), %xmm4 # 0x3f11c0 vmulps %xmm4, %xmm3, %xmm4 vaddps %xmm2, %xmm4, %xmm4 vcvttps2dq %xmm4, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm4, %xmm4 vandps %xmm4, %xmm8, %xmm4 vsubps %xmm4, %xmm6, %xmm4 vmulps %xmm4, %xmm9, %xmm6 vsubps %xmm6, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm6 vbroadcastss 0x1466b1(%rip), %xmm7 # 0x3f11cc vmulps %xmm7, %xmm3, %xmm7 vbroadcastss 0x1466a8(%rip), %xmm9 # 0x3f11d0 vaddps %xmm7, %xmm9, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm12, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm13, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm7, %xmm14, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm2, %xmm7, %xmm7 vmulps %xmm7, %xmm6, %xmm6 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm4, %xmm8, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vrcpps %xmm3, %xmm4 vaddps %xmm4, %xmm4, %xmm6 vmulps %xmm6, %xmm3, %xmm3 vbroadcastss 0x147b86(%rip), %xmm7 # 0x3f2708 vsubps %xmm3, %xmm7, %xmm3 vmulps %xmm3, %xmm4, %xmm3 vaddps %xmm1, %xmm6, %xmm4 vaddps %xmm3, %xmm4, %xmm3 jmp 0x2aad99 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm3 vbroadcastss 0x4(%r14), %xmm6 vmaxps %xmm3, %xmm4, %xmm4 vminps %xmm6, %xmm4, %xmm15 vmaxps %xmm3, %xmm5, %xmm3 vminps %xmm6, %xmm3, %xmm3 jmp 0x2aada7 vbroadcastss 0x1465ec(%rip), %xmm12 # 0x3f11b4 vxorps %xmm4, %xmm12, %xmm3 vbroadcastss 0x1465e3(%rip), %xmm13 # 0x3f11b8 vminps %xmm3, %xmm13, %xmm3 vbroadcastss 0x1465da(%rip), %xmm14 # 0x3f11bc vmaxps %xmm3, %xmm14, %xmm3 vbroadcastss 0x1465d1(%rip), %xmm9 # 0x3f11c0 vmulps %xmm3, %xmm9, %xmm4 vaddps %xmm2, %xmm4, %xmm4 vcvttps2dq %xmm4, %xmm6 vcvtdq2ps %xmm6, %xmm6 vcmpltps %xmm6, %xmm4, %xmm4 vandps %xmm4, %xmm8, %xmm4 vsubps %xmm4, %xmm6, %xmm4 vbroadcastss 0x146d0b(%rip), %xmm1 # 0x3f1920 vmulps %xmm1, %xmm4, %xmm6 vsubps %xmm6, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm6 vbroadcastss 0x1465a2(%rip), %xmm10 # 0x3f11cc vmulps %xmm3, %xmm10, %xmm7 vbroadcastss 0x146599(%rip), %xmm11 # 0x3f11d0 vaddps %xmm7, %xmm11, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vmovaps %xmm2, %xmm1 vbroadcastss 0x146588(%rip), %xmm2 # 0x3f11d4 vaddps %xmm2, %xmm7, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vbroadcastss 0x14657b(%rip), %xmm10 # 0x3f11d8 vaddps %xmm7, %xmm10, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vbroadcastss 0x14656e(%rip), %xmm11 # 0x3f11dc vaddps %xmm7, %xmm11, %xmm7 vmulps %xmm3, %xmm7, %xmm7 vaddps %xmm1, %xmm7, %xmm7 vmulps %xmm7, %xmm6, %xmm6 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm4, %xmm8, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vrcpps %xmm3, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vsubps %xmm3, %xmm8, %xmm3 vmulps %xmm3, %xmm4, %xmm3 vaddps %xmm3, %xmm4, %xmm15 vxorps %xmm5, %xmm12, %xmm3 vminps %xmm3, %xmm13, %xmm3 vmaxps %xmm3, %xmm14, %xmm3 vmulps %xmm3, %xmm9, %xmm4 vaddps %xmm1, %xmm4, %xmm4 vcvttps2dq %xmm4, %xmm5 vcvtdq2ps %xmm5, %xmm5 vcmpltps %xmm5, %xmm4, %xmm4 vandps %xmm4, %xmm8, %xmm4 vsubps %xmm4, %xmm5, %xmm4 vbroadcastss 0x146c3f(%rip), %xmm5 # 0x3f1920 vmulps %xmm5, %xmm4, %xmm5 vsubps %xmm5, %xmm3, %xmm3 vmulps %xmm3, %xmm3, %xmm5 vbroadcastss 0x1464d6(%rip), %xmm6 # 0x3f11cc vmulps %xmm6, %xmm3, %xmm6 vbroadcastss 0x1464cd(%rip), %xmm7 # 0x3f11d0 vaddps %xmm7, %xmm6, %xmm6 vmulps %xmm3, %xmm6, %xmm6 vaddps %xmm2, %xmm6, %xmm6 vmovaps %xmm1, %xmm2 vmulps %xmm3, %xmm6, %xmm6 vaddps %xmm6, %xmm10, %xmm6 vmulps %xmm3, %xmm6, %xmm6 vaddps %xmm6, %xmm11, %xmm6 vmulps %xmm3, %xmm6, %xmm6 vaddps %xmm1, %xmm6, %xmm6 vmulps %xmm6, %xmm5, %xmm5 vaddps %xmm3, %xmm8, %xmm3 vaddps %xmm5, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vpslld $0x17, %xmm4, %xmm4 vpaddd %xmm4, %xmm8, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vaddps %xmm3, %xmm8, %xmm3 vrcpps %xmm3, %xmm4 vmulps %xmm4, %xmm3, %xmm3 vsubps %xmm3, %xmm8, %xmm3 vmulps %xmm3, %xmm4, %xmm3 vaddps %xmm3, %xmm4, %xmm3 jmp 0x2aada7 movq 0x118(%rbp,%rsi), %r14 vbroadcastss (%r14), %xmm3 vbroadcastss 0x4(%r14), %xmm6 vmulps %xmm4, %xmm3, %xmm7 vaddps %xmm6, %xmm7, %xmm7 vmaxps %xmm0, %xmm7, %xmm7 vminps %xmm7, %xmm8, %xmm7 vmulps %xmm4, %xmm7, %xmm15 vmulps %xmm5, %xmm3, %xmm3 vaddps %xmm6, %xmm3, %xmm3 vmaxps %xmm0, %xmm3, %xmm3 vminps %xmm3, %xmm8, %xmm3 vmulps %xmm5, %xmm3, %xmm3 jmp 0x2aada7 vmovaps %xmm4, %xmm15 vmovaps %xmm5, %xmm3 cmpl $0x65, 0x38(%rsp) jl 0x2aae0c movq 0x288(%rbp,%rsi), %rsi vmulps (%rsi,%rbx,4), %xmm15, %xmm4 vmulps 0x10(%rsi,%rbx,4), %xmm3, %xmm3 vbroadcastss 0x1463ea(%rip), %xmm1 # 0x3f11b4 vandps %xmm1, %xmm4, %xmm5 vandps %xmm1, %xmm3, %xmm6 vorps %xmm2, %xmm5, %xmm5 vorps %xmm2, %xmm6, %xmm6 vaddps %xmm5, %xmm4, %xmm4 vaddps %xmm6, %xmm3, %xmm3 vcvttps2dq %xmm4, %xmm4 vcvttps2dq %xmm3, %xmm3 vpackssdw %xmm3, %xmm4, %xmm3 vpminsw 0x14d75a(%rip), %xmm3, %xmm3 # 0x3f8550 vpmaxsw 0x14d762(%rip), %xmm3, %xmm3 # 0x3f8560 vpacksswb %xmm3, %xmm3, %xmm3 vmovq %xmm3, (%rdx) addq $0x8, %rdx jmp 0x2aae1b vmovups %xmm15, (%r8) vmovups %xmm3, 0x10(%r8) addq $0x20, %r8 incl %r13d jmp 0x2aa487 incl %ecx jmp 0x2aa476 movq 0x40(%rsp), %rsi incq %rsi movq 0x30(%rsp), %rdx addl 0x28(%rsp), %edx movq 0x18(%rsp), %r14 jmp 0x2aa411 pushq $0x3 popq %r13 cmpl $0x1, 0xdc(%rbp,%rdx) jne 0x2aaf50 cmpl $0x1, 0xe0(%rbp,%rdx) jne 0x2aaf50 cmpl $0x2, 0xe4(%rbp,%rdx) jne 0x2aaf50 cmpl $0x2, 0xe8(%rbp,%rdx) jne 0x2aaf50 cmpl $0x1, 0x110(%rbp,%rdx) ja 0x2aaf50 cmpl $0x65, 0x38(%rsp) jl 0x2aba61 vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %r13 vmovaps %xmm0, (%r13) andq $0x0, 0x10(%r13) xorl %ebx, %ebx movq %rbp, %rcx leaq 0xe0(%rsp), %rbp leaq 0x198(%rsp), %r12 movq -0x18(%rax), %r14 movslq 0x108(%rcx,%r14), %rax cmpq %rax, %rbx jge 0x2abdbb movq 0x1f8(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2aaf0b movq 0x240(%rcx,%r14), %rax vmulss (%rax,%rbx,4), %xmm1, %xmm0 vmovss 0x143d81(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq 0x288(%rcx,%r14), %rax vmovss (%rax,%rbx,4), %xmm0 vmovss %xmm0, 0x198(%rsp) movq %r13, %rdi movq %rbp, %rsi callq 0x1ea12c movq %r13, %rdi movq %r12, %rsi callq 0x1ea12c incq %rbx movq 0x8(%rsp), %rcx movq (%rcx), %rax jmp 0x2aaec6 imull %ecx, %r13d movslq %r13d, %rsi leaq 0x130(%rsp), %rdi leaq 0xe0(%rsp), %rdx movq %rsi, 0xc8(%rsp) callq 0x73bbe movq (%rbp), %r9 movq -0x18(%r9), %rax movl 0x18(%rsp), %r10d imull 0xe0(%rbp,%rax), %r10d movq 0x130(%rsp), %rbx movl 0xdc(%rbp,%rax), %ecx imull 0xd4(%rbp,%rax), %ecx subl %ecx, %r10d xorl %ecx, %ecx xorl %edx, %edx xorl %esi, %esi cmpl 0xd8(%rbp,%rax), %esi jge 0x2aafe3 movslq %ecx, %rcx leaq (%rbx,%rcx,4), %r8 xorl %edi, %edi cmpl 0xd4(%rbp,%rax), %edi jge 0x2aafd9 movl %edx, (%r8,%rdi,4) movq -0x18(%r9), %rax addl 0xdc(%rbp,%rax), %edx incq %rdi jmp 0x2aafbc addl %r10d, %edx incl %esi addq %rdi, %rcx jmp 0x2aafaa xorl %edx, %edx testl %r13d, %r13d movl $0x0, %ecx movq %rcx, 0x50(%rsp) cmovlel %edx, %r13d vbroadcastss 0x1461b5(%rip), %xmm4 # 0x3f11b4 leaq 0x14d6d2(%rip), %r8 # 0x3f86d8 xorl %ebp, %ebp movq %r9, 0xc0(%rsp) movq 0x8(%rsp), %rcx movslq 0x108(%rcx,%rax), %rax cmpq %rax, %rbp jge 0x2ab382 movq 0x10(%r14), %r9 imulq %rbp, %r9 imulq 0x40(%r14), %r9 addq (%r14), %r9 movslq 0x9c(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %rbp, %r11 movq 0x80(%rsp), %rax imulq %rax, %r11 addq 0x70(%rsp), %r11 imulq %rax, %r10 movq 0x28(%rcx), %r15 addq 0x50(%rsp), %r15 xorl %eax, %eax movq %r9, 0x18(%rsp) movq %r10, 0x28(%rsp) movq %r11, 0x48(%rsp) movq %rbp, 0xd0(%rsp) cmpl 0x68(%rsp), %eax jg 0x2ab357 movl %eax, 0x30(%rsp) movq 0x8(%rsp), %rax movq (%rax), %rax movq %rax, 0x40(%rsp) xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d jg 0x2ab34c movq 0x40(%rsp), %rax movq -0x18(%rax), %r12 movq 0x8(%rsp), %rcx movslq 0xe8(%rcx,%r12), %rax movslq 0x30(%rsp), %rdx imulq %rax, %rdx imulq %r10, %rdx movslq 0xe4(%rcx,%r12), %rax movslq %r14d, %rcx imulq %rax, %rcx addq %r11, %rcx addq %rdx, %rcx xorl %edx, %edx xorl %eax, %eax cmpq %rdx, %r13 je 0x2ab102 movslq (%rbx,%rdx,4), %rsi movsbl (%rcx,%rsi), %esi movsbl (%r15,%rdx), %edi imull %esi, %edi addl %edi, %eax incq %rdx jmp 0x2ab0e6 movq 0x8(%rsp), %rdx movq 0x1f8(%rdx,%r12), %rcx vmovss (%rcx,%rbp,4), %xmm1 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm1, %xmm0 je 0x2ab137 movq 0x240(%rdx,%r12), %rcx vmulss (%rcx,%rbp,4), %xmm1, %xmm0 vmovss 0x143b55(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vcvtsi2ss %eax, %xmm6, %xmm1 vmulss %xmm1, %xmm0, %xmm5 cmpl $0x0, 0x100(%rdx,%r12) je 0x2ab157 movq 0x1b0(%rdx,%r12), %rax vaddss (%rax,%rbp,4), %xmm5, %xmm5 movl 0x110(%rdx,%r12), %eax decl %eax cmpl $0x5, %eax ja 0x2ab2cf movslq (%r8,%rax,4), %rax addq %r8, %rax jmpq *%rax vmaxss 0x142e95(%rip), %xmm5, %xmm0 # 0x3ee010 jmp 0x2ab2d3 vmovaps %xmm5, %xmm0 movq %r9, 0x58(%rsp) vmovss %xmm5, 0xd8(%rsp) callq 0x5f410 vaddss 0x143ae9(%rip), %xmm0, %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x48(%rsp), %r11 movq 0x28(%rsp), %r10 movq 0x58(%rsp), %r9 leaq 0x14d519(%rip), %r8 # 0x3f86d8 vbroadcastss 0x145fec(%rip), %xmm4 # 0x3f11b4 vmulss 0xd8(%rsp), %xmm0, %xmm0 jmp 0x2ab2d3 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss 0x4(%rax), %xmm1 vmaxss (%rax), %xmm5, %xmm0 vucomiss %xmm1, %xmm0 jbe 0x2ab2d3 vmovaps %xmm1, %xmm0 jmp 0x2ab2d3 vminss 0x145fb1(%rip), %xmm5, %xmm0 # 0x3f11b8 vxorps %xmm4, %xmm0, %xmm1 vcmpltss 0x145fa8(%rip), %xmm0, %xmm0 # 0x3f11bc vbroadcastss 0x145f9b(%rip), %xmm2 # 0x3f11b8 vblendvps %xmm0, %xmm2, %xmm1, %xmm0 movq %r9, %rbp callq 0x5f410 movq 0x48(%rsp), %r11 movq 0x28(%rsp), %r10 movq %rbp, %r9 movq 0xd0(%rsp), %rbp leaq 0x14d491(%rip), %r8 # 0x3f86d8 vbroadcastss 0x145f64(%rip), %xmm4 # 0x3f11b4 vmovss 0x143a30(%rip), %xmm1 # 0x3eec88 vaddss %xmm1, %xmm0, %xmm0 vdivss %xmm0, %xmm1, %xmm0 jmp 0x2ab2d3 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vxorps %xmm0, %xmm0, %xmm0 vcmpltss %xmm5, %xmm0, %xmm0 vmovss (%rax), %xmm1 vbroadcastss 0x143a03(%rip), %xmm2 # 0x3eec88 vblendvps %xmm0, %xmm2, %xmm1, %xmm0 vmulss %xmm5, %xmm0, %xmm0 jmp 0x2ab2d3 movq 0x8(%rsp), %rax movq 0x118(%rax,%r12), %rax vmovss (%rax), %xmm1 vmovss 0x4(%rax), %xmm2 vxorps %xmm4, %xmm2, %xmm0 vdivss %xmm1, %xmm0, %xmm3 vxorps %xmm0, %xmm0, %xmm0 vucomiss %xmm3, %xmm5 jb 0x2ab2d3 vmovss 0x1439c7(%rip), %xmm0 # 0x3eec88 vdivss %xmm1, %xmm0, %xmm0 vaddss %xmm0, %xmm3, %xmm0 vucomiss %xmm0, %xmm5 jbe 0x2ab33f vmovaps %xmm5, %xmm0 cmpl $0x65, 0x38(%rsp) jl 0x2ab325 movq 0x8(%rsp), %rax movq 0x288(%rax,%r12), %rax vmulss (%rax,%rbp,4), %xmm0, %xmm0 vandps %xmm4, %xmm0, %xmm1 vbroadcastss 0x14d30b(%rip), %xmm2 # 0x3f8604 vorps %xmm2, %xmm1, %xmm1 vaddss %xmm1, %xmm0, %xmm0 vroundss $0xb, %xmm0, %xmm0, %xmm0 vcvttss2si %xmm0, %eax cmpl $-0x7e, %eax pushq $-0x7f popq %rcx cmovll %ecx, %eax cmpl $0x7f, %eax pushq $0x7f popq %rcx cmovgel %ecx, %eax movb %al, (%r9) incq %r9 jmp 0x2ab337 movq 0x18(%rsp), %rax vmovss %xmm0, (%rax) addq $0x4, %rax movq %rax, 0x18(%rsp) incl %r14d jmp 0x2ab09f vmulss %xmm5, %xmm1, %xmm0 vaddss %xmm2, %xmm0, %xmm0 jmp 0x2ab28b movl 0x30(%rsp), %eax incl %eax jmp 0x2ab081 incq %rbp movq 0xc0(%rsp), %r9 movq -0x18(%r9), %rax movq 0x50(%rsp), %rcx addq 0xc8(%rsp), %rcx movq %rcx, 0x50(%rsp) movq 0x20(%rsp), %r14 jmp 0x2ab010 leaq 0x130(%rsp), %rdi callq 0x624be xorl %r12d, %r12d jmp 0x2ab463 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0xe0(%rsp), %xmm0 vmovups %xmm0, (%rbx) movq 0xf0(%rsp), %rax movq %rax, 0x10(%rbx) movl 0xf8(%rsp), %eax movl %eax, 0x18(%rbx) movq 0x100(%rsp), %rax movq %rax, 0x20(%rbx) vmovups 0x108(%rsp), %xmm0 vmovups %xmm0, 0x28(%rbx) movl 0x118(%rsp), %eax movl %eax, 0x38(%rbx) movq 0x120(%rsp), %rax movq %rax, 0x40(%rbx) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2ab429 lock decl (%rax) jne 0x2ab429 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi je 0x2ab421 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ab429 movq %rsi, %rdi callq 0x5f3e0 movq 0x138(%rsp), %rax testq %rax, %rax movl %r14d, %r12d je 0x2ab463 lock decl (%rax) jne 0x2ab463 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi je 0x2ab45b movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ab463 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ab494 lock decl (%rax) jne 0x2ab494 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2ab48c movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ab494 movq %rsi, %rdi callq 0x5f3e0 movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x2ab4cb lock decl (%rax) jne 0x2ab4cb movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi je 0x2ab4c3 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ab4cb movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x318, %rsp # imm = 0x318 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovaps %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2ab556 movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x143753(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2ab500 leaq 0x2a0(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdx movq 0x20(%rsp), %rcx movslq 0x2c(%rcx), %rsi movslq 0x30(%rcx), %rax movl 0x38(%rcx), %edi movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movl 0x34(%rcx), %r8d movl %r8d, 0x58(%rsp) movq (%rcx), %r8 movq 0x10(%rcx), %r11 movq 0x40(%rcx), %r15 imulq %r11, %r15 movq %rax, %rbx movq %rsi, 0x38(%rsp) imulq %rsi, %rbx movq %r11, 0xd0(%rsp) movq %rbx, 0x50(%rsp) imulq %rbx, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0xc8(%rsp) movq 0x80(%rsp), %rsi imulq 0xb0(%rsp), %rsi movq %rsi, 0xc0(%rsp) xorl %esi, %esi testl %eax, %eax movl $0x0, %r11d cmovgl %eax, %r11d movl %r11d, 0x10(%rsp) movl 0x28(%rcx), %eax movl %eax, 0x64(%rsp) testl %edi, %edi cmovlel %esi, %edi movq %rdi, 0x18(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x190(%rsp) movq %rdx, 0x28(%rsp) leaq (%rdx,%rdx), %rax movq %rax, 0x188(%rsp) movq 0x2a0(%rsp), %rax movq %rax, 0x180(%rsp) movq %r8, 0xd8(%rsp) xorl %esi, %esi movq %r10, 0x48(%rsp) movq %r15, 0x68(%rsp) cmpq 0x18(%rsp), %rsi je 0x2ab814 movq 0xc8(%rsp), %rax xorl %edx, %edx divq 0xd0(%rsp) cmpl $0x4, 0x64(%rsp) cmoveq 0x50(%rsp), %rax testq %r10, %r10 je 0x2ab688 vmovss (%r10,%rsi,4), %xmm1 jmp 0x2ab68c vxorps %xmm1, %xmm1, %xmm1 imulq %rsi, %r15 addq 0xd8(%rsp), %r15 movq 0x180(%rsp), %rcx vmovss (%rcx,%rsi,4), %xmm0 imull 0x58(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2ab6c5 vmovss %xmm1, (%r8,%rcx,4) incq %rcx jmp 0x2ab6b6 leaq (%rsi,%rsi,8), %rax movq 0xc0(%rsp), %r13 movq %rsi, 0x40(%rsp) imulq %rsi, %r13 addq 0x190(%rsp), %r13 movq 0x28(%rsp), %rcx leaq (%rcx,%r13), %r11 movq 0x188(%rsp), %rcx leaq (%rcx,%r13), %rbx xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d je 0x2ab7fa xorl %ecx, %ecx movq %r15, %rdx movq 0x38(%rsp), %rsi movl %esi, %r10d testl %r10d, %r10d jle 0x2ab7da movsbl (%r13,%rcx), %r12d movsbl (%r9,%rax), %esi imull %r12d, %esi movsbl 0x1(%r13,%rcx), %r12d movsbl 0x1(%r9,%rax), %edi imull %r12d, %edi addl %esi, %edi movsbl 0x2(%r13,%rcx), %esi movsbl 0x2(%r9,%rax), %r12d imull %esi, %r12d movsbl (%r11,%rcx), %esi movsbl 0x3(%r9,%rax), %ebp imull %esi, %ebp addl %r12d, %ebp addl %edi, %ebp movsbl 0x1(%r11,%rcx), %esi movsbl 0x4(%r9,%rax), %edi imull %esi, %edi movsbl 0x2(%r11,%rcx), %esi movsbl 0x5(%r9,%rax), %r12d imull %esi, %r12d addl %edi, %r12d movsbl (%rbx,%rcx), %esi movsbl 0x6(%r9,%rax), %edi imull %esi, %edi addl %r12d, %edi addl %ebp, %edi movsbl 0x1(%rbx,%rcx), %esi movsbl 0x7(%r9,%rax), %ebp imull %esi, %ebp movsbl 0x2(%rbx,%rcx), %esi movsbl 0x8(%r9,%rax), %r12d imull %esi, %r12d addl %ebp, %r12d addl %edi, %r12d vcvtsi2ss %r12d, %xmm2, %xmm1 vmulss %xmm1, %xmm0, %xmm1 vaddss (%r15,%rcx,4), %xmm1, %xmm1 vmovss %xmm1, (%r15,%rcx,4) addq $0x4, %rdx decl %r10d incq %rcx jmp 0x2ab712 addq %rcx, %r13 addq $0x2, %r13 addq %rcx, %r11 addq $0x2, %r11 addq %rcx, %rbx addq $0x2, %rbx incl %r14d movq %rdx, %r15 jmp 0x2ab6fa movq 0x40(%rsp), %rsi incq %rsi movq 0x68(%rsp), %r15 addq %r15, %r8 movq 0x48(%rsp), %r10 jmp 0x2ab653 leaq 0x2a0(%rsp), %rdi jmp 0x2abd9f leaq 0x2b8(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rax movq 0x20(%rsp), %r15 movl 0x2c(%r15), %ecx movl %ecx, 0x38(%rsp) movl 0x30(%r15), %ecx movl 0x38(%r15), %edx movq 0x28(%rbp), %rdi movq 0x1b0(%rbp,%r13), %rsi xorl %r9d, %r9d testl %ecx, %ecx cmovlel %r9d, %ecx movl %ecx, 0x10(%rsp) movq %rax, 0x28(%rsp) addq %rax, %rax movq %rax, 0x58(%rsp) testl %edx, %edx cmovlel %r9d, %edx movq %rdx, 0x18(%rsp) vbroadcastss 0x145927(%rip), %xmm0 # 0x3f11b4 vbroadcastss 0x14cd6e(%rip), %xmm1 # 0x3f8604 pushq $-0x7f popq %r8 pushq $0x7f popq %rbx movq %rsi, 0x48(%rsp) cmpq 0x18(%rsp), %r9 je 0x2aba54 testq %rsi, %rsi je 0x2ab8ba vmovss (%rsi,%r9,4), %xmm2 jmp 0x2ab8be vxorps %xmm2, %xmm2, %xmm2 movq 0x40(%r15), %r14 imulq %r9, %r14 imulq 0x10(%r15), %r14 addq (%r15), %r14 movq 0x2b8(%rsp), %rax vmovss (%rax,%r9,8), %xmm3 vmovss 0x4(%rax,%r9,8), %xmm4 movq 0xb0(%rsp), %r15 imulq %r9, %r15 imulq 0x80(%rsp), %r15 addq 0x70(%rsp), %r15 movq %r9, 0x40(%rsp) leaq (%r9,%r9,8), %r12 movq 0x28(%rsp), %rax leaq (%r15,%rax), %r13 movq 0x58(%rsp), %rax leaq (%r15,%rax), %rbp xorl %eax, %eax cmpl 0x10(%rsp), %eax je 0x2aba3d xorl %r10d, %r10d movl 0x38(%rsp), %ecx testl %ecx, %ecx jle 0x2aba1e movsbl (%r15,%r10), %r11d movsbl (%rdi,%r12), %r9d imull %r11d, %r9d movsbl 0x1(%r15,%r10), %r11d movsbl 0x1(%rdi,%r12), %esi imull %r11d, %esi addl %r9d, %esi movsbl 0x2(%r15,%r10), %r9d movsbl 0x2(%rdi,%r12), %r11d imull %r9d, %r11d movsbl (%r13,%r10), %r9d movsbl 0x3(%rdi,%r12), %edx imull %r9d, %edx addl %r11d, %edx addl %esi, %edx movsbl 0x1(%r13,%r10), %esi movsbl 0x4(%rdi,%r12), %r9d imull %esi, %r9d movsbl 0x2(%r13,%r10), %esi movsbl 0x5(%rdi,%r12), %r11d imull %esi, %r11d addl %r9d, %r11d movsbl (%rbp,%r10), %esi movsbl 0x6(%rdi,%r12), %r9d imull %esi, %r9d addl %r11d, %r9d addl %edx, %r9d movsbl 0x1(%rbp,%r10), %edx movsbl 0x7(%rdi,%r12), %esi imull %edx, %esi movsbl 0x2(%rbp,%r10), %edx movsbl 0x8(%rdi,%r12), %r11d imull %edx, %r11d addl %esi, %r11d addl %r9d, %r11d vcvtsi2ss %r11d, %xmm7, %xmm5 vmulss %xmm5, %xmm3, %xmm5 vaddss %xmm2, %xmm5, %xmm5 vmulss %xmm4, %xmm5, %xmm5 vandps %xmm0, %xmm5, %xmm6 vorps %xmm1, %xmm6, %xmm6 vaddss %xmm6, %xmm5, %xmm5 vroundss $0xb, %xmm5, %xmm5, %xmm5 vcvttss2si %xmm5, %r11d cmpl $-0x7e, %r11d jge 0x2aba07 movl %r8d, %r11d cmpl $0x7f, %r11d jl 0x2aba10 movl %ebx, %r11d movb %r11b, (%r14,%r10) decl %ecx incq %r10 jmp 0x2ab92b addq %r10, %r15 addq $0x2, %r15 addq %r10, %r13 addq $0x2, %r13 addq %r10, %rbp addq $0x2, %rbp incl %eax addq %r10, %r14 jmp 0x2ab91a movq 0x40(%rsp), %r9 incq %r9 movq 0x20(%rsp), %r15 movq 0x48(%rsp), %rsi jmp 0x2ab8a2 leaq 0x2b8(%rsp), %rdi jmp 0x2abfed vxorps %xmm0, %xmm0, %xmm0 leaq 0x130(%rsp), %rbx andq $0x0, 0x10(%rbx) vmovaps %xmm0, (%rbx) xorl %r14d, %r14d leaq 0xe0(%rsp), %r15 movq -0x18(%rax), %r12 movslq 0x108(%rbp,%r12), %rax cmpq %rax, %r14 jge 0x2abad7 movq 0x1f8(%rbp,%r12), %rax movq 0x240(%rbp,%r12), %rcx vmovss (%rax,%r14,4), %xmm0 vmulss (%rcx,%r14,4), %xmm0, %xmm0 vmovss 0x1431d2(%rip), %xmm1 # 0x3eec88 vdivss %xmm0, %xmm1, %xmm0 vmovss %xmm0, 0xe0(%rsp) movq %rbx, %rdi movq %r15, %rsi callq 0x1ea12c incq %r14 movq (%rbp), %rax jmp 0x2aba81 leaq 0x270(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rdi movq 0x20(%rsp), %rdx movslq 0x2c(%rdx), %rsi movslq 0x30(%rdx), %rax movl 0x38(%rdx), %r8d movl %edi, %ecx subl %esi, %ecx addl %ecx, %ecx movq 0x28(%rbp), %r9 movq 0x1b0(%rbp,%r12), %r10 movq 0x10(%rdx), %r11 movq 0x40(%rdx), %rbx imulq %r11, %rbx movq %rax, %r14 movq %rsi, 0x38(%rsp) imulq %rsi, %r14 movq %r11, 0x58(%rsp) movq %r14, 0xd0(%rsp) imulq %r14, %r11 addq $0xf, %r11 andq $-0x10, %r11 movq %r11, 0x68(%rsp) movq 0x80(%rsp), %rsi imulq 0xb0(%rsp), %rsi movq %rsi, 0x50(%rsp) xorl %r11d, %r11d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi movl %esi, 0x10(%rsp) movl 0x34(%rdx), %eax movl %eax, 0xc8(%rsp) movq (%rdx), %r12 movslq %ecx, %rsi testl %r8d, %r8d cmovlel %r11d, %r8d movq %r8, 0x18(%rsp) movl 0x28(%rdx), %eax movl %eax, 0x64(%rsp) movq 0x70(%rsp), %rax movq %rax, 0x190(%rsp) movq %rdi, 0x28(%rsp) leaq (%rdi,%rdi), %rax movq %rax, 0x188(%rsp) movq 0x270(%rsp), %rax movq %rax, 0x180(%rsp) movq %r12, 0xc0(%rsp) xorl %edi, %edi movq %r10, 0x48(%rsp) movq %rbx, 0xd8(%rsp) cmpq 0x18(%rsp), %rdi je 0x2abd97 movq 0x68(%rsp), %rax xorl %edx, %edx divq 0x58(%rsp) cmpl $0x4, 0x64(%rsp) cmoveq 0xd0(%rsp), %rax testq %r10, %r10 je 0x2abc0e vmovss (%r10,%rdi,4), %xmm1 jmp 0x2abc12 vxorps %xmm1, %xmm1, %xmm1 movq %rbx, %rdx imulq %rdi, %rdx addq 0xc0(%rsp), %rdx movq 0x180(%rsp), %rcx vmovss (%rcx,%rdi,4), %xmm0 imull 0xc8(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2abc51 vmovss %xmm1, (%r12,%rcx,4) incq %rcx jmp 0x2abc42 leaq (%rdi,%rdi,8), %rax movq 0x50(%rsp), %r8 movq %rdi, 0x40(%rsp) imulq %rdi, %r8 addq 0x190(%rsp), %r8 movq 0x28(%rsp), %rcx leaq (%r8,%rcx), %r13 movq 0x188(%rsp), %rcx leaq (%r8,%rcx), %r11 xorl %r14d, %r14d cmpl 0x10(%rsp), %r14d je 0x2abd7a xorl %ecx, %ecx movq 0x38(%rsp), %rdi movl %edi, %r10d testl %r10d, %r10d jle 0x2abd60 movsbl (%r8,%rcx), %ebx movsbl (%r9,%rax), %r15d imull %ebx, %r15d movsbl 0x1(%r8,%rcx), %ebx movsbl 0x1(%r9,%rax), %edi imull %ebx, %edi addl %r15d, %edi movsbl 0x2(%r8,%rcx), %ebx movsbl 0x2(%r9,%rax), %r15d imull %ebx, %r15d movsbl (%r13,%rcx), %ebx movsbl 0x3(%r9,%rax), %ebp imull %ebx, %ebp addl %r15d, %ebp addl %edi, %ebp movsbl 0x1(%r13,%rcx), %edi movsbl 0x4(%r9,%rax), %ebx imull %edi, %ebx movsbl 0x2(%r13,%rcx), %edi movsbl 0x5(%r9,%rax), %r15d imull %edi, %r15d addl %ebx, %r15d movsbl (%r11,%rcx), %edi movsbl 0x6(%r9,%rax), %ebx imull %edi, %ebx addl %r15d, %ebx addl %ebp, %ebx movsbl 0x1(%r11,%rcx), %edi movsbl 0x7(%r9,%rax), %ebp imull %edi, %ebp movsbl 0x2(%r11,%rcx), %edi movsbl 0x8(%r9,%rax), %r15d imull %edi, %r15d addl %ebp, %r15d addl %ebx, %r15d vcvtsi2ss %r15d, %xmm2, %xmm1 vmulss %xmm1, %xmm0, %xmm1 vaddss (%rdx), %xmm1, %xmm1 vmovss %xmm1, (%rdx) addq $0x4, %rdx decl %r10d addq $0x2, %rcx jmp 0x2abc98 addq %rsi, %r8 addq %rcx, %r8 addq %rsi, %r13 addq %rcx, %r13 addq %rsi, %r11 addq %rcx, %r11 incl %r14d jmp 0x2abc83 movq 0x40(%rsp), %rdi incq %rdi movq 0xd8(%rsp), %rbx addq %rbx, %r12 movq 0x48(%rsp), %r10 jmp 0x2abbdc leaq 0x270(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x20(%rsp), %r15 jmp 0x2abfff leaq 0x288(%rsp), %rdi leaq 0x130(%rsp), %rsi callq 0x290632 movslq 0x9c(%rsp), %rsi movq 0x20(%rsp), %r15 movl 0x2c(%r15), %eax movl 0x30(%r15), %edx movl 0x38(%r15), %r8d movl %esi, %ecx movl %eax, 0x10(%rsp) subl %eax, %ecx addl %ecx, %ecx movq 0x8(%rsp), %rax movq 0x28(%rax), %rdi movq 0x1b0(%rax,%r14), %rax movq %rsi, 0x18(%rsp) addq %rsi, %rsi movq %rsi, 0x48(%rsp) movslq %ecx, %r10 xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %r8d, %r8d cmovlel %esi, %r8d movq %r8, 0x40(%rsp) vbroadcastss 0x145384(%rip), %xmm0 # 0x3f11b4 vbroadcastss 0x14c7cb(%rip), %xmm1 # 0x3f8604 movq %rax, 0x28(%rsp) cmpq 0x40(%rsp), %rsi je 0x2abfe5 testq %rax, %rax je 0x2abe55 vmovss (%rax,%rsi,4), %xmm2 jmp 0x2abe59 vxorps %xmm2, %xmm2, %xmm2 movq 0x40(%r15), %rbx imulq %rsi, %rbx imulq 0x10(%r15), %rbx addq (%r15), %rbx movq 0x288(%rsp), %rcx vmovss (%rcx,%rsi,8), %xmm3 vmovss 0x4(%rcx,%rsi,8), %xmm4 movq 0xb0(%rsp), %r14 imulq %rsi, %r14 imulq 0x80(%rsp), %r14 addq 0x70(%rsp), %r14 movq %rsi, 0x38(%rsp) leaq (%rsi,%rsi,8), %r15 movq 0x18(%rsp), %rax leaq (%r14,%rax), %r12 movq 0x48(%rsp), %rax leaq (%r14,%rax), %r13 xorl %ebp, %ebp cmpl %edx, %ebp je 0x2abfce xorl %r9d, %r9d movl 0x10(%rsp), %ecx testl %ecx, %ecx jle 0x2abfb5 movsbl (%r14,%r9), %esi movsbl (%rdi,%r15), %r8d imull %esi, %r8d movsbl 0x1(%r14,%r9), %esi movsbl 0x1(%rdi,%r15), %r11d imull %esi, %r11d addl %r8d, %r11d movsbl 0x2(%r14,%r9), %esi movsbl 0x2(%rdi,%r15), %r8d imull %esi, %r8d movsbl (%r12,%r9), %esi movsbl 0x3(%rdi,%r15), %eax imull %esi, %eax addl %r8d, %eax addl %r11d, %eax movsbl 0x1(%r12,%r9), %esi movsbl 0x4(%rdi,%r15), %r8d imull %esi, %r8d movsbl 0x2(%r12,%r9), %esi movsbl 0x5(%rdi,%r15), %r11d imull %esi, %r11d addl %r8d, %r11d movsbl (%r13,%r9), %esi movsbl 0x6(%rdi,%r15), %r8d imull %esi, %r8d addl %r11d, %r8d addl %eax, %r8d movsbl 0x1(%r13,%r9), %eax movsbl 0x7(%rdi,%r15), %esi imull %eax, %esi movsbl 0x2(%r13,%r9), %eax movsbl 0x8(%rdi,%r15), %r11d imull %eax, %r11d addl %esi, %r11d addl %r8d, %r11d vcvtsi2ss %r11d, %xmm7, %xmm5 vmulss %xmm5, %xmm3, %xmm5 vaddss %xmm2, %xmm5, %xmm5 vmulss %xmm4, %xmm5, %xmm5 vandps %xmm0, %xmm5, %xmm6 vorps %xmm1, %xmm6, %xmm6 vaddss %xmm6, %xmm5, %xmm5 vroundss $0xb, %xmm5, %xmm5, %xmm5 vcvttss2si %xmm5, %esi cmpl $-0x7e, %esi jge 0x2abf9c pushq $-0x7f popq %rsi cmpl $0x7f, %esi jl 0x2abfa4 pushq $0x7f popq %rsi movb %sil, (%rbx) incq %rbx decl %ecx addq $0x2, %r9 jmp 0x2abec2 addq %r10, %r14 addq %r9, %r14 addq %r10, %r12 addq %r9, %r12 addq %r10, %r13 addq %r9, %r13 incl %ebp jmp 0x2abeb3 movq 0x38(%rsp), %rsi incq %rsi movq 0x20(%rsp), %r15 movq 0x28(%rsp), %rax jmp 0x2abe3e leaq 0x288(%rsp), %rdi callq 0x621c2 leaq 0x130(%rsp), %rdi callq 0x621c2 movq 0x8(%rsp), %rax xorl %r12d, %r12d movq 0x8(%rax), %rdi testq %rdi, %rdi je 0x2ab463 movq (%rdi), %rax movq %r15, %rsi movq 0x30(%rsp), %rdx callq *0x48(%rax) jmp 0x2ab463 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 jmp 0x2ac035 movq %rax, %rbx leaq 0x130(%rsp), %rdi callq 0x621c2 jmp 0x2ac1b7 jmp 0x2ac0d0 jmp 0x2ac0d0 jmp 0x2ac231 jmp 0x2ac231 movq %rax, %rbx jmp 0x2ac180 jmp 0x2ac231 jmp 0x2ac0d0 movq %rax, %rbx jmp 0x2ac149 jmp 0x2ac231 jmp 0x2ac231 jmp 0x2ac0d0 jmp 0x2ac231 jmp 0x2ac231 movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac1e8 lock decl (%rax) jne 0x2ac1e8 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2ac1d8 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ac1e8 jmp 0x2ac231 movq %rax, %rbx jmp 0x2ac1e8 movq %rax, %rbx jmp 0x2ac1b7 movq %rax, %rbx movq 0x230(%rsp), %rax testq %rax, %rax je 0x2ac112 lock decl (%rax) jne 0x2ac112 movq 0x228(%rsp), %rsi movq 0x248(%rsp), %rdi testq %rdi, %rdi jne 0x2ac10c movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac112 movq (%rdi), %rax callq *0x18(%rax) movq 0x1a0(%rsp), %rax testq %rax, %rax je 0x2ac149 lock decl (%rax) jne 0x2ac149 movq 0x198(%rsp), %rsi movq 0x1b8(%rsp), %rdi testq %rdi, %rdi jne 0x2ac143 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac149 movq (%rdi), %rax callq *0x18(%rax) movq 0xe8(%rsp), %rax testq %rax, %rax je 0x2ac180 lock decl (%rax) jne 0x2ac180 movq 0xe0(%rsp), %rsi movq 0x100(%rsp), %rdi testq %rdi, %rdi jne 0x2ac17a movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac180 movq (%rdi), %rax callq *0x18(%rax) movq 0x138(%rsp), %rax testq %rax, %rax je 0x2ac1b7 lock decl (%rax) jne 0x2ac1b7 movq 0x130(%rsp), %rsi movq 0x150(%rsp), %rdi testq %rdi, %rdi jne 0x2ac1b1 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac1b7 movq (%rdi), %rax callq *0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac1e8 lock decl (%rax) jne 0x2ac1e8 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2ac1e2 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac1e8 movq (%rdi), %rax callq *0x18(%rax) movq 0x1e8(%rsp), %rax testq %rax, %rax je 0x2ac21f lock decl (%rax) jne 0x2ac21f movq 0x1e0(%rsp), %rsi movq 0x200(%rsp), %rdi testq %rdi, %rdi jne 0x2ac219 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac21f movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2ac231 jmp 0x2ac231 jmp 0x2ac231 jmp 0x2ac231 jmp 0x2ac231 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_avx::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { #if NCNN_INT8 if (opt.use_int8_inference && int8_scale_term) { return forward_int8_x86(bottom_blob, top_blob, opt); } #endif int w = bottom_blob.w; int h = bottom_blob.h; int channels = bottom_blob.c; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; Mat bottom_blob_bordered; make_padding(bottom_blob, bottom_blob_bordered, opt); if (bottom_blob_bordered.empty()) return -100; w = bottom_blob_bordered.w; h = bottom_blob_bordered.h; int outw = (w - kernel_extent_w) / stride_w + 1; int outh = (h - kernel_extent_h) / stride_h + 1; int out_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ out_elempack = num_output % 16 == 0 ? 16 : num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #elif __AVX__ out_elempack = num_output % 8 == 0 ? 8 : num_output % 4 == 0 ? 4 : 1; #else out_elempack = num_output % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ size_t out_elemsize = elemsize / elempack * out_elempack; top_blob.create(outw, outh, num_output / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; // depth-wise if (channels * elempack == group && group == num_output) { #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack16_avx512(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 16; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m512 _sum = _mm512_set1_ps(0.f); if (bias_term) { _sum = _mm512_loadu_ps(((const float*)bias_data) + g * 16); } const float* sptr = m.row(i * stride_h) + j * stride_w * 16; for (int k = 0; k < maxk; k++) { __m512 _val = _mm512_loadu_ps(sptr + space_ofs[k] * 16); __m512 _w = _mm512_loadu_ps(kptr + k * 16); _sum = _mm512_fmadd_ps(_val, _w, _sum); } _mm512_storeu_ps(outptr, _sum); outptr += 16; } } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX512F__ if (elempack == 8) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack8_avx(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } else { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 8; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m256 _sum = _mm256_set1_ps(0.f); if (bias_term) { _sum = _mm256_loadu_ps(((const float*)bias_data) + g * 8); } const float* sptr = m.row(i * stride_h) + j * stride_w * 8; for (int k = 0; k < maxk; k++) { __m256 _val = _mm256_loadu_ps(sptr + space_ofs[k] * 8); __m256 _w = _mm256_loadu_ps(kptr + k * 8); _sum = _mm256_comp_fmadd_ps(_val, _w, _sum); } _mm256_storeu_ps(outptr + j * 8, _sum); } outptr += outw * 8; } } if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } #endif // __AVX__ if (elempack == 4) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw5x5s1_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 5 && kernel_h == 5 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw5x5s2_pack4_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } { const int maxk = kernel_w * kernel_h; // kernel offsets std::vector<int> _space_ofs(maxk); int* space_ofs = &_space_ofs[0]; { int p1 = 0; int p2 = 0; int gap = w * dilation_h - kernel_w * dilation_w; for (int i = 0; i < kernel_h; i++) { for (int j = 0; j < kernel_w; j++) { space_ofs[p1] = p2; p1++; p2 += dilation_w; } p2 += gap; } } #pragma omp parallel for num_threads(opt.num_threads) for (int g = 0; g < channels; g++) { float* outptr = top_blob.channel(g); const float* kptr = (const float*)weight_data_tm + maxk * g * 4; const Mat m = bottom_blob_bordered.channel(g); for (int i = 0; i < outh; i++) { for (int j = 0; j < outw; j++) { __m128 _sum = _mm_set1_ps(0.f); if (bias_term) { _sum = _mm_loadu_ps(((const float*)bias_data) + g * 4); } const float* sptr = m.row(i * stride_h) + j * stride_w * 4; for (int k = 0; k < maxk; k++) { __m128 _val = _mm_loadu_ps(sptr + space_ofs[k] * 4); __m128 _w = _mm_loadu_ps(kptr + k * 4); _sum = _mm_add_ps(_mm_mul_ps(_val, _w), _sum); } _sum = activation_sse(_sum, activation_type, activation_params); _mm_storeu_ps(outptr + j * 4, _sum); } outptr += outw * 4; } } return 0; } } #endif // __SSE2__ if (elempack == 1) { if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && stride_h == 1) { convdw3x3s1_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && stride_h == 2) { convdw3x3s2_sse(bottom_blob_bordered, top_blob, weight_data_tm, bias_data, opt); if (activation) { activation->forward_inplace(top_blob, opt); } return 0; } } } // group convolution const int channels_g = channels * elempack / group; const int num_output_g = num_output / group; int g_elempack = 1; int out_g_elempack = 1; #if __SSE2__ if (opt.use_packing_layout) { #if __AVX512F__ g_elempack = channels_g % 16 == 0 ? 16 : channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 16 == 0 ? 16 : num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #elif __AVX__ g_elempack = channels_g % 8 == 0 ? 8 : channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 8 == 0 ? 8 : num_output_g % 4 == 0 ? 4 : 1; #else g_elempack = channels_g % 4 == 0 ? 4 : 1; out_g_elempack = num_output_g % 4 == 0 ? 4 : 1; #endif } #endif // __SSE2__ // unpacking Mat bottom_blob_bordered_unpacked = bottom_blob_bordered; if (elempack > g_elempack) { Option opt_p = opt; opt_p.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob_bordered, bottom_blob_bordered_unpacked, g_elempack, opt_p); } Mat top_blob_unpacked = top_blob; if (out_g_elempack < out_elempack) { top_blob_unpacked.create(outw, outh, num_output / out_g_elempack, out_elemsize / out_elempack * out_g_elempack, out_g_elempack, opt.workspace_allocator); if (top_blob_unpacked.empty()) return -100; } for (int g = 0; g < group; g++) { const Mat bottom_blob_bordered_g = bottom_blob_bordered_unpacked.channel_range(channels_g * g / g_elempack, channels_g / g_elempack); Mat top_blob_g = top_blob_unpacked.channel_range(num_output_g * g / out_g_elempack, num_output_g / out_g_elempack); const ncnn::Layer* op = group_ops[g]; Option opt_g = opt; opt_g.blob_allocator = top_blob_unpacked.allocator; // forward op->forward(bottom_blob_bordered_g, top_blob_g, opt_g); } // packing if (out_g_elempack < out_elempack) { convert_packing(top_blob_unpacked, top_blob, out_elempack, opt); } else { top_blob = top_blob_unpacked; } return 0; }

movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2a49d2

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

ncnn::ConvolutionDepthWise_x86_avx::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w *= weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w *= bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x168, %rsp # imm = 0x168 movq %rsi, %rbp movq %rdi, %r13 movq (%rsi), %r14 leaq 0x48(%r14), %rdi movq (%rdx), %rax movq %rax, 0xb8(%rsp) movl 0x60(%r14), %ebx movl 0x74(%r14), %eax movl %eax, 0x1c(%rsp) movl 0x78(%r14), %eax movl %eax, 0x18(%rsp) imull 0x80(%r14), %ebx leaq 0x70(%rsp), %rsi andq $0x0, 0x40(%rsi) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, (%rsi) vmovups %xmm0, 0xc(%rsi) vmovaps %xmm0, 0x20(%rsi) vmovups %xmm0, 0x2c(%rsi) movq %rcx, %r15 movq %rcx, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x70(%rsp) je 0x2ac867 movslq 0xa8(%rsp), %rax imulq 0xb0(%rsp), %rax testq %rax, %rax je 0x2ac867 movslq 0x88(%rsp), %rcx movl 0x9c(%rsp), %eax imull %ecx, %eax movl %eax, 0x9c(%rsp) movq 0x80(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x80(%rsp) movl $0x1, 0x88(%rsp) andq $0x0, 0x60(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0x20(%rsp) vmovups %xmm0, 0x2c(%rsp) vmovaps %xmm0, 0x40(%rsp) vmovups %xmm0, 0x4c(%rsp) movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x100(%r13,%rax) je 0x2ac3b0 movl $0x90, %edi addq (%rbp), %rdi leaq 0x20(%rsp), %rsi movq %r15, %rdx callq 0x64ee7 pushq $-0x64 popq %r12 cmpq $0x0, 0x20(%rsp) je 0x2ac839 movslq 0x58(%rsp), %rax imulq 0x60(%rsp), %rax testq %rax, %rax je 0x2ac839 movslq 0x38(%rsp), %rcx movl 0x4c(%rsp), %eax imull %ecx, %eax movl %eax, 0x4c(%rsp) movq 0x30(%rsp), %rax xorl %edx, %edx divq %rcx movq %rax, 0x30(%rsp) movl $0x1, 0x38(%rsp) pushq $0x2a popq %rdi callq 0x782bf movq %rax, %r12 leaq 0x8(%rsp), %rdi callq 0x71548 leaq 0x8(%rsp), %rdi xorl %esi, %esi movl %ebx, %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0x1 popq %rsi movl 0x1c(%rsp), %edx callq 0x7193a leaq 0x8(%rsp), %rdi pushq $0xb popq %rsi movl 0x18(%rsp), %edx callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xdc(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xe8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xd popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xec(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x4 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf0(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xf popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf4(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0xe popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0xf8(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x10 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax vmovss 0xfc(%r13,%rax), %xmm0 leaq 0x8(%rsp), %rdi pushq $0x12 popq %rsi callq 0x71952 movq (%r13), %rax movq -0x18(%rax), %rax movl 0x100(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x5 popq %rsi callq 0x7193a movl 0x9c(%rsp), %edx leaq 0x8(%rsp), %rdi pushq $0x6 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x108(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x7 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x10c(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x8 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax movl 0x110(%r13,%rax), %edx leaq 0x8(%rsp), %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r13), %rax movq -0x18(%rax), %rax leaq (%rax,%r13), %rdx addq $0x118, %rdx # imm = 0x118 leaq 0x8(%rsp), %rdi pushq $0xa popq %rsi callq 0x7196c movq (%r12), %rax leaq 0x8(%rsp), %rsi movq %r12, %rdi callq *0x10(%rax) andq $0x0, 0x110(%rsp) vxorps %xmm0, %xmm0, %xmm0 vmovaps %xmm0, 0xd0(%rsp) vmovups %xmm0, 0xdc(%rsp) vmovaps %xmm0, 0xf0(%rsp) vmovups %xmm0, 0xfc(%rsp) andq $0x0, 0x158(%rsp) vmovups %xmm0, 0x118(%rsp) vmovups %xmm0, 0x124(%rsp) vmovups %xmm0, 0x138(%rsp) vmovups %xmm0, 0x144(%rsp) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac61e lock incl (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2ac655 lock decl (%rax) jne 0x2ac655 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x2ac64d movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ac655 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x70(%rsp), %xmm0 vmovaps %xmm0, 0xd0(%rsp) movq 0x80(%rsp), %rax movq %rax, 0xe0(%rsp) movl 0x88(%rsp), %eax movl %eax, 0xe8(%rsp) movq 0x90(%rsp), %rax movq %rax, 0xf0(%rsp) vmovups 0x98(%rsp), %xmm0 vmovups %xmm0, 0xf8(%rsp) movl 0xa8(%rsp), %eax movl %eax, 0x108(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x110(%rsp) movq 0x28(%rsp), %rax testq %rax, %rax je 0x2ac6cf lock incl (%rax) movq 0x120(%rsp), %rax testq %rax, %rax je 0x2ac706 lock decl (%rax) jne 0x2ac706 movq 0x118(%rsp), %rsi movq 0x138(%rsp), %rdi testq %rdi, %rdi je 0x2ac6fe movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ac706 movq %rsi, %rdi callq 0x5f3e0 vmovaps 0x20(%rsp), %xmm0 leaq 0xd0(%rsp), %rsi vmovups %xmm0, 0x48(%rsi) movq 0x30(%rsp), %rax movq %rax, 0x58(%rsi) movl 0x38(%rsp), %eax movl %eax, 0x60(%rsi) movq 0x40(%rsp), %rax movq %rax, 0x68(%rsi) vmovups 0x48(%rsp), %xmm0 vmovaps %xmm0, 0x70(%rsi) movl 0x58(%rsp), %eax movl %eax, 0x80(%rsi) movq 0x60(%rsp), %rax movq %rax, 0x88(%rsi) leaq 0xc0(%rsp), %rdi callq 0x6b00e movq (%r12), %rax leaq 0xc0(%rsp), %rsi movq %r12, %rdi callq *0x18(%rax) leaq 0xc0(%rsp), %rdi callq 0x6b03a movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq *0x20(%rax) movq (%r12), %rax movq %r12, %rdi movq %r14, %rsi movq 0xb8(%rsp), %rdx movq %r15, %rcx callq *0x38(%rax) movq (%r12), %rax movq %r12, %rdi movq %r15, %rsi callq *0x28(%rax) movq (%r12), %rax movq %r12, %rdi callq *0x8(%rax) pushq $0x48 popq %rbx vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%rbx), %rax testq %rax, %rax je 0x2ac801 lock decl (%rax) jne 0x2ac801 movq 0xd0(%rsp,%rbx), %rsi movq 0xf0(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x2ac7f5 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2ac801 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%rbx), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x2ac7c2 leaq 0x8(%rsp), %rdi callq 0x71614 xorl %r12d, %r12d movq 0x28(%rsp), %rax testq %rax, %rax je 0x2ac867 lock decl (%rax) jne 0x2ac867 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi je 0x2ac85f movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ac867 movq %rsi, %rdi callq 0x5f3e0 movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac898 lock decl (%rax) jne 0x2ac898 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2ac890 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ac898 movq %rsi, %rdi callq 0x5f3e0 movl %r12d, %eax addq $0x168, %rsp # imm = 0x168 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2ac9cd jmp 0x2ac9cd movq %rax, %rbx leaq 0xc0(%rsp), %rdi callq 0x6b03a jmp 0x2ac8da jmp 0x2ac8d7 jmp 0x2ac8cf jmp 0x2ac8cf movq %rax, %rbx jmp 0x2ac964 movq %rax, %rbx pushq $0x48 popq %r14 vxorps %xmm0, %xmm0, %xmm0 movq 0xd8(%rsp,%r14), %rax testq %rax, %rax je 0x2ac921 lock decl (%rax) jne 0x2ac921 movq 0xd0(%rsp,%r14), %rsi movq 0xf0(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2ac915 movq (%rdi), %rax callq *0x18(%rax) vxorps %xmm0, %xmm0, %xmm0 jmp 0x2ac921 movq %rsi, %rdi callq 0x5f3e0 vxorps %xmm0, %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0xd0, %rax andq $0x0, 0x40(%rax) vmovups %xmm0, 0xc(%rax) vmovups %xmm0, (%rax) andl $0x0, 0x38(%rax) vmovups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2ac8e2 jmp 0x2ac95a jmp 0x2ac9cd movq %rax, %rbx jmp 0x2ac992 jmp 0x2ac9cd movq %rax, %rbx leaq 0x8(%rsp), %rdi callq 0x71614 movq 0x28(%rsp), %rax testq %rax, %rax je 0x2ac992 lock decl (%rax) jne 0x2ac992 movq 0x20(%rsp), %rsi movq 0x40(%rsp), %rdi testq %rdi, %rdi jne 0x2ac98c movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac992 movq (%rdi), %rax callq *0x18(%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2ac9c3 lock decl (%rax) jne 0x2ac9c3 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2ac9bd movq %rsi, %rdi callq 0x5f3e0 jmp 0x2ac9c3 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2ac9cd movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

virtual thunk to ncnn::ConvolutionDepthWise_x86_avx::forward(std::vector<ncnn::Mat, std::allocator<ncnn::Mat>> const&, std::vector<ncnn::Mat, std::allocator<ncnn::Mat>>&, ncnn::Option const&) const

int ConvolutionDepthWise_x86_avx::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const { const Mat& bottom_blob = bottom_blobs[0]; const Mat& _weight_data = bottom_blobs[1]; Mat& top_blob = top_blobs[0]; const int _kernel_w = _weight_data.w; const int _kernel_h = _weight_data.h; const int _num_output = _weight_data.c * _weight_data.elempack; Mat weight_data_flattened; flatten(_weight_data, weight_data_flattened, opt); if (weight_data_flattened.empty()) return -100; // weight_data_flattened as pack1 weight_data_flattened.w *= weight_data_flattened.elempack; weight_data_flattened.elemsize /= weight_data_flattened.elempack; weight_data_flattened.elempack = 1; Mat bias_data_flattened; if (bias_term) { const Mat& _bias_data = bottom_blobs[2]; flatten(_bias_data, bias_data_flattened, opt); if (bias_data_flattened.empty()) return -100; // bias_data_flattened as pack1 bias_data_flattened.w *= bias_data_flattened.elempack; bias_data_flattened.elemsize /= bias_data_flattened.elempack; bias_data_flattened.elempack = 1; } ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::ConvolutionDepthWise); ncnn::ParamDict pd; pd.set(0, _num_output); pd.set(1, _kernel_w); pd.set(11, _kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, pad_left); pd.set(15, pad_right); pd.set(14, pad_top); pd.set(16, pad_bottom); pd.set(18, pad_value); pd.set(5, bias_term); pd.set(6, weight_data_flattened.w); pd.set(7, group); pd.set(8, int8_scale_term); pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); ncnn::Mat weights[2]; weights[0] = weight_data_flattened; weights[1] = bias_data_flattened; op->load_model(ncnn::ModelBinFromMatArray(weights)); op->create_pipeline(opt); op->forward(bottom_blob, top_blob, opt); op->destroy_pipeline(opt); delete op; return 0; }

movq (%rdi), %rax addq -0x40(%rax), %rdi jmp 0x2ac246 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/convolutiondepthwise_x86_avx.cpp

ncnn::Padding::load_param(ncnn::ParamDict const&)

int Padding::load_param(const ParamDict& pd) { top = pd.get(0, 0); bottom = pd.get(1, 0); left = pd.get(2, 0); right = pd.get(3, 0); type = pd.get(4, 0); value = pd.get(5, 0.f); per_channel_pad_data_size = pd.get(6, 0); front = pd.get(7, 0); behind = pd.get(8, 0); return 0; }

pushq %r14 pushq %rbx pushq %rax movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd4(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x3 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xdc(%rbx) pushq $0x4 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe0(%rbx) pushq $0x5 popq %rsi xorps %xmm0, %xmm0 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xe4(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xf0(%rbx) pushq $0x7 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe8(%rbx) pushq $0x8 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xec(%rbx) xorl %eax, %eax addq $0x8, %rsp popq %rbx popq %r14 retq nop

/csukuangfj[P]ncnn/src/layer/padding.cpp

ncnn::Padding::load_model(ncnn::ModelBin const&)

int Padding::load_model(const ModelBin& mb) { if (per_channel_pad_data_size) { per_channel_pad_data = mb.load(per_channel_pad_data_size, 1); } return 0; }

pushq %r14 pushq %rbx subq $0x48, %rsp movl 0xf0(%rdi), %edx testl %edx, %edx je 0x2acbe2 movq %rdi, %rbx movq (%rsi), %rax movq %rsp, %r14 pushq $0x1 popq %rcx movq %r14, %rdi callq *0x10(%rax) leaq 0xf8(%rbx), %rcx movq 0x8(%r14), %rax cmpq %r14, %rcx je 0x2acbba testq %rax, %rax je 0x2acb2b lock incl (%rax) movq 0x100(%rbx), %rax testq %rax, %rax je 0x2acb5f lock decl (%rax) jne 0x2acb5f movq 0xf8(%rbx), %rsi movq 0x118(%rbx), %rdi testq %rdi, %rdi je 0x2acb57 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2acb5f movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0xf8(%rbx) movq 0x8(%rsp), %rax movq %rax, 0x100(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0x108(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0x110(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x118(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x120(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x130(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x138(%rbx) testq %rax, %rax je 0x2acbe2 lock decl (%rax) jne 0x2acbe2 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2acbda movq (%rdi), %rax callq *0x18(%rax) jmp 0x2acbe2 movq %rsi, %rdi callq 0x5f3e0 xorl %eax, %eax addq $0x48, %rsp popq %rbx popq %r14 retq movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2acc1c lock decl (%rax) jne 0x2acc1c movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2acc16 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2acc1c movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 jmp 0x2acc26 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/padding.cpp

ncnn::Padding::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int outw = w + left + right; if (dims == 1) { top_blob.create(outw, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; if (elemsize == 1) copy_make_border_image<signed char>(bottom_blob, top_blob, 0, left, type, static_cast<signed char>(value)); if (elemsize == 2) copy_make_border_image<unsigned short>(bottom_blob, top_blob, 0, left, type, support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(value) : float32_to_bfloat16(value)); if (elemsize == 4) copy_make_border_image<float>(bottom_blob, top_blob, 0, left, type, value); return 0; } int outh = h + top + bottom; if (dims == 2) { top_blob.create(outw, outh, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; if (elemsize == 1) copy_make_border_image<signed char>(bottom_blob, top_blob, top, left, type, static_cast<signed char>(value)); if (elemsize == 2) copy_make_border_image<unsigned short>(bottom_blob, top_blob, top, left, type, support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(value) : float32_to_bfloat16(value)); if (elemsize == 4) copy_make_border_image<float>(bottom_blob, top_blob, top, left, type, value); return 0; } if (dims == 3) { int outc = channels + front + behind; top_blob.create(outw, outh, outc, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc; q++) { Mat borderm = top_blob.channel(q); float pad_value = per_channel_pad_data_size ? per_channel_pad_data[q] : value; //Channel padding if (((q < front) || (q >= (channels + front))) && type == 0) { if (elemsize == 1) { borderm.fill(static_cast<signed char>(pad_value)); } if (elemsize == 2) { borderm.fill(support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(pad_value) : float32_to_bfloat16(pad_value)); } if (elemsize == 4) { borderm.fill(pad_value); } } else { int q_ = q - front; if (type == 1) { q_ = q_ <= 0 ? 0 : q_; q_ = q_ >= channels - 1 ? channels - 1 : q_; } if (type == 2) { q_ = abs(q_); q_ = (channels - 1) - abs(q_ - (channels - 1)); } const Mat m = bottom_blob.channel(q_); if (elemsize == 1) copy_make_border_image<signed char>(m, borderm, top, left, type, static_cast<signed char>(pad_value)); if (elemsize == 2) copy_make_border_image<unsigned short>(m, borderm, top, left, type, support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(pad_value) : float32_to_bfloat16(pad_value)); if (elemsize == 4) copy_make_border_image<float>(m, borderm, top, left, type, pad_value); } } return 0; } if (dims == 4) { int outd = d + front + behind; top_blob.create(outw, outh, outd, channels, elemsize, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { float pad_value = per_channel_pad_data_size ? per_channel_pad_data[q] : value; for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if (((z < front) || (z >= (d + front))) && type == 0) { if (elemsize == 1) { borderm.fill(static_cast<signed char>(pad_value)); } if (elemsize == 2) { borderm.fill(support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(pad_value) : float32_to_bfloat16(pad_value)); } if (elemsize == 4) { borderm.fill(pad_value); } } else { int z_ = z - front; if (type == 1) { z_ = z_ <= 0 ? 0 : z_; z_ = z_ >= d - 1 ? d - 1 : z_; } if (type == 2) { z_ = abs(z_); z_ = (d - 1) - abs(z_ - (d - 1)); } const Mat m = bottom_blob.channel(q).depth(z_); if (elemsize == 1) copy_make_border_image<signed char>(m, borderm, top, left, type, static_cast<signed char>(pad_value)); if (elemsize == 2) copy_make_border_image<unsigned short>(m, borderm, top, left, type, support_fp16_storage && opt.use_fp16_storage ? float32_to_float16(pad_value) : float32_to_bfloat16(pad_value)); if (elemsize == 4) copy_make_border_image<float>(m, borderm, top, left, type, pad_value); } } } return 0; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x118, %rsp # imm = 0x118 movq %rcx, 0x20(%rsp) movq %rdx, %rbx movq %rsi, %r14 movl 0xd0(%rdi), %eax movl 0xd4(%rdi), %edx testl %eax, %eax sete %cl testl %edx, %edx sete %sil movq 0xd8(%rdi), %xmm1 movq %rdi, 0x8(%rsp) movq 0xe8(%rdi), %xmm0 movdqa %xmm1, %xmm2 punpcklqdq %xmm0, %xmm2 # xmm2 = xmm2[0],xmm0[0] pxor %xmm3, %xmm3 pcmpeqd %xmm2, %xmm3 movmskps %xmm3, %edi xorl $0xf, %edi sete %dil andb %cl, %sil andb %dil, %sil cmpb $0x1, %sil jne 0x2acceb movq $0x0, 0x10(%rsp) cmpq %r14, %rbx je 0x2ad8eb movq 0x8(%r14), %rax testq %rax, %rax je 0x2accba lock incl (%rax) movq 0x8(%rbx), %rax testq %rax, %rax je 0x2ad760 lock decl (%rax) jne 0x2ad760 movq (%rbx), %rsi movq 0x20(%rbx), %rdi testq %rdi, %rdi je 0x2ad758 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2ad760 movl 0x28(%r14), %ecx movq 0x10(%r14), %r8 pshufd $0x55, %xmm1, %xmm2 # xmm2 = xmm1[1,1,1,1] paddd %xmm1, %xmm2 movd %xmm2, %esi addl 0x2c(%r14), %esi pushq $-0x64 popq %rdi movq %rdi, 0x10(%rsp) cmpl $0x1, %ecx movq %r8, 0x28(%rsp) jne 0x2acdaf movq 0x20(%rsp), %rax movq 0x8(%rax), %rcx movq %rbx, %rdi movq %r8, %rdx callq 0x635fa cmpq $0x0, (%rbx) je 0x2ad8eb movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ad8eb movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad7ef cmpq $0x2, %rax je 0x2ad7b8 movq $0x0, 0x10(%rsp) cmpq $0x1, %rax jne 0x2ad8eb movq 0x8(%rsp), %rax movl 0xd8(%rax), %ecx movl 0xe0(%rax), %r8d cvttss2si 0xe4(%rax), %eax movsbl %al, %r9d movq $0x0, 0x10(%rsp) movq %r14, %rdi movq %rbx, %rsi xorl %edx, %edx callq 0x2ad910 jmp 0x2ad8eb movl 0x38(%r14), %edi movq %rdi, 0x40(%rsp) addl %eax, %edx addl 0x30(%r14), %edx pshufd $0x55, %xmm0, %xmm1 # xmm1 = xmm0[1,1,1,1] cmpl $0x4, %ecx movq %r14, 0x60(%rsp) je 0x2ad2bc cmpl $0x3, %ecx je 0x2ace6c cmpl $0x2, %ecx jne 0x2ad8e2 movq 0x20(%rsp), %rax movq 0x8(%rax), %r8 movq %rbx, %rdi movq 0x28(%rsp), %rcx callq 0x636fa cmpq $0x0, (%rbx) je 0x2ad8eb movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ad8eb movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad892 cmpq $0x2, %rax je 0x2ad854 cmpq $0x1, %rax jne 0x2ad8e2 movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx movl 0xe0(%rax), %r8d cvttss2si 0xe4(%rax), %eax movsbl %al, %r9d movq %r14, %rdi movq %rbx, %rsi callq 0x2ad910 jmp 0x2ad8e2 paddd %xmm0, %xmm1 movd %xmm1, %ecx addl 0x40(%rsp), %ecx movq 0x20(%rsp), %rax movq 0x8(%rax), %r9 movq %rbx, %rdi movq %rcx, %r14 movq 0x28(%rsp), %r8 callq 0x63810 cmpq $0x0, (%rbx) je 0x2ad8eb movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ad8eb movq 0x8(%rsp), %rax addq $0xe4, %rax movq %rax, 0x50(%rsp) xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x10(%rsp) cmovlel %ecx, %r14d movq %r14, 0x48(%rsp) movq 0x40(%rsp), %rax decl %eax movl %eax, 0x68(%rsp) xorl %esi, %esi cmpq 0x48(%rsp), %rsi je 0x2ad8eb movslq 0x2c(%rbx), %rax movslq 0x30(%rbx), %rcx movl 0x34(%rbx), %r14d movq (%rbx), %r12 movq 0x10(%rbx), %rbp movq 0x40(%rbx), %r13 movq %r13, %rdx movq %rsi, 0x38(%rsp) imulq %rsi, %rdx imulq %rbp, %rdx addq %r12, %rdx movl 0x18(%rbx), %esi movq 0x20(%rbx), %rdi movq %rdx, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq %rbp, 0xc8(%rsp) movl %esi, 0xd0(%rsp) movq %rdi, 0xd8(%rsp) movl %eax, 0xe4(%rsp) movl %ecx, 0xe8(%rsp) movl $0x1, 0xec(%rsp) movl %r14d, 0xf0(%rsp) imulq %rax, %rcx movq %rbp, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rbp movq %rax, %r15 movq %rax, 0xf8(%rsp) movl 0x28(%rbx), %eax leal -0x1(%rax), %edx movl %edx, 0xe0(%rsp) cmpl $0x4, %eax jne 0x2acfa7 movq %rcx, 0xf8(%rsp) movq %rcx, %r15 movq 0x38(%rsp), %rdi leaq (,%rdi,4), %rax movq 0x8(%rsp), %rcx addq 0xf8(%rcx), %rax cmpl $0x0, 0xf0(%rcx) cmoveq 0x50(%rsp), %rax movd (%rax), %xmm0 movl 0xe0(%rcx), %r8d movl 0xe8(%rcx), %edx movslq %edx, %rax cmpq %rax, %rdi setl %cl addl 0x40(%rsp), %eax cltq cmpq %rax, %rdi setge %al orb %cl, %al testl %r8d, %r8d sete %sil andb %al, %sil movd %xmm0, %ecx cmpb $0x1, %sil jne 0x2ad061 movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad201 cmpq $0x2, %rax je 0x2ad19f cmpq $0x1, %rax jne 0x2ad2af cvttss2si %xmm0, %eax imull %r15d, %r14d testl %r14d, %r14d movl $0x0, %ecx cmovlel %ecx, %r14d imulq %rbp, %r13 imulq 0x38(%rsp), %r13 addq %r13, %r12 xorl %ecx, %ecx cmpq %rcx, %r14 je 0x2ad2af movb %al, (%r12,%rcx) incq %rcx jmp 0x2ad04f movl %edi, %eax subl %edx, %eax cmpl $0x2, %r8d movl %r8d, %ebp je 0x2ad08d cmpl $0x1, %ebp movq 0x60(%rsp), %r11 jne 0x2ad0a8 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax movl 0x68(%rsp), %edx cmpl %edx, %eax cmovgel %edx, %eax jmp 0x2ad0a8 movl %eax, %edx negl %edx cmovsl %eax, %edx movl 0x68(%rsp), %esi subl %esi, %edx movl %edx, %eax negl %eax cmovnsl %edx, %eax addl %esi, %eax movq 0x60(%rsp), %r11 movslq 0x2c(%r11), %rdx movslq 0x30(%r11), %rsi cltq imulq 0x40(%r11), %rax movq 0x10(%r11), %rdi imulq %rdi, %rax addq (%r11), %rax movl 0x34(%r11), %r8d movl 0x18(%r11), %r9d movq 0x20(%r11), %r10 movq %rax, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdi, 0x80(%rsp) movl %r9d, 0x88(%rsp) movq %r10, 0x90(%rsp) movl %edx, 0x9c(%rsp) movl %esi, 0xa0(%rsp) movl $0x1, 0xa4(%rsp) movl %r8d, 0xa8(%rsp) imulq %rdx, %rsi movq %rdi, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xb0(%rsp) movl 0x28(%r11), %eax leal -0x1(%rax), %edx movl %edx, 0x98(%rsp) cmpl $0x4, %eax jne 0x2ad14d movq %rsi, 0xb0(%rsp) movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad234 cmpq $0x2, %rax je 0x2ad1c9 cmpq $0x1, %rax jne 0x2ad2af movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx cvttss2si %xmm0, %eax movsbl %al, %r9d leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %ebp, %r8d callq 0x2ad910 jmp 0x2ad2af movq 0x8(%rsp), %rax cmpb $0x1, 0xd(%rax) jne 0x2ad25c movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) jne 0x2ad25c callq 0x6460e movl %eax, %ecx jmp 0x2ad25f movq 0x8(%rsp), %rax movl 0xd0(%rax), %r15d movl 0xd8(%rax), %r14d cmpb $0x1, 0xd(%rax) jne 0x2ad28d movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) jne 0x2ad28d callq 0x6460e movl %eax, %ecx jmp 0x2ad290 imull %r15d, %r14d testl %r14d, %r14d movl $0x0, %eax cmovlel %eax, %r14d imulq %rbp, %r13 imulq 0x38(%rsp), %r13 addq %r13, %r12 xorl %eax, %eax cmpl %eax, %r14d je 0x2ad2af movss %xmm0, (%r12,%rax,4) incq %rax jmp 0x2ad220 movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %ebp, %r8d callq 0x2ae697 jmp 0x2ad2af shrl $0x10, %ecx imull %r15d, %r14d testl %r14d, %r14d movl $0x0, %eax cmovlel %eax, %r14d imulq %rbp, %r13 imulq 0x38(%rsp), %r13 addq %r13, %r12 xorl %eax, %eax cmpq %rax, %r14 je 0x2ad2af movw %cx, (%r12,%rax,2) incq %rax jmp 0x2ad27e shrl $0x10, %ecx movzwl %cx, %r9d leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %r15d, %edx movl %r14d, %ecx movl %ebp, %r8d callq 0x2adf5b movq 0x38(%rsp), %rsi incq %rsi jmp 0x2acee2 movl 0x34(%r14), %eax paddd %xmm0, %xmm1 movd %xmm1, %ecx movq %rax, 0x100(%rsp) addl %eax, %ecx movq 0x20(%rsp), %rax movq 0x8(%rax), %rax movq %rax, (%rsp) movq %rbx, %rdi movq %rcx, %r14 movq 0x40(%rsp), %r8 movq 0x28(%rsp), %r9 callq 0x6393c cmpq $0x0, (%rbx) je 0x2ad8eb movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2ad8eb movq 0x8(%rsp), %rax addq $0xe4, %rax movq %rax, 0x108(%rsp) xorl %ecx, %ecx testl %r14d, %r14d cmovlel %ecx, %r14d movq %r14, 0x68(%rsp) movq 0x100(%rsp), %rax decl %eax movl %eax, 0x5c(%rsp) movq 0x40(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x10(%rsp) cmovlel %ecx, %edx movq %rdx, 0x40(%rsp) xorl %edx, %edx movq %rbx, 0x110(%rsp) cmpq 0x40(%rsp), %rdx je 0x2ad8eb leaq (,%rdx,4), %rax movq 0x8(%rsp), %rcx addq 0xf8(%rcx), %rax cmpl $0x0, 0xf0(%rcx) cmoveq 0x108(%rsp), %rax movd (%rax), %xmm0 movd %xmm0, %eax shrl $0x10, %eax movl %eax, 0x34(%rsp) cvttss2si %xmm0, %eax movl %eax, 0x38(%rsp) xorl %edi, %edi movss %xmm0, 0x1c(%rsp) movq %rdx, 0x50(%rsp) cmpq 0x68(%rsp), %rdi je 0x2ad750 movslq 0x2c(%rbx), %r15 movslq 0x30(%rbx), %r14 movq (%rbx), %r13 movq %rbx, %rsi movq 0x10(%rbx), %rbx movq 0x40(%rsi), %r12 imulq %rdx, %r12 movq %r12, %rax imulq %rbx, %rax addq %r13, %rax movl 0x18(%rsi), %ecx movq %r14, %rbp imulq %r15, %rbp movq %rdi, %rdx imulq %rbx, %rdx imulq %rbp, %rdx addq %rax, %rdx movq 0x20(%rsi), %rax movq %rdx, 0xb8(%rsp) andq $0x0, 0xc0(%rsp) movq %rbx, 0xc8(%rsp) movl %ecx, 0xd0(%rsp) movq %rax, 0xd8(%rsp) movl $0x2, 0xe0(%rsp) movl %r15d, 0xe4(%rsp) movl %r14d, 0xe8(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xec(%rsp) movq %rbp, 0xf8(%rsp) movq 0x8(%rsp), %rsi movl 0xe8(%rsi), %ecx movslq %ecx, %rax cmpq %rax, %rdi setl %dl addl 0x100(%rsp), %eax cltq cmpq %rax, %rdi movl 0xe0(%rsi), %esi setge %al orb %dl, %al testl %esi, %esi sete %dl andb %al, %dl cmpb $0x1, %dl movq %rdi, 0x48(%rsp) jne 0x2ad4f3 movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad6d4 cmpq $0x2, %rax je 0x2ad627 cmpq $0x1, %rax jne 0x2ad736 testl %ebp, %ebp movl $0x0, %eax cmovlel %eax, %ebp imulq 0x48(%rsp), %r14 imulq %r15, %r14 addq %r12, %r14 imulq %r14, %rbx addq %rbx, %r13 xorl %eax, %eax cmpq %rax, %rbp je 0x2ad736 movl 0x38(%rsp), %ecx movb %cl, (%r13,%rax) incq %rax jmp 0x2ad4dc movl %edi, %eax subl %ecx, %eax cmpl $0x2, %esi je 0x2ad523 cmpl $0x1, %esi movq 0x60(%rsp), %r9 movq 0x50(%rsp), %rdi movl %esi, %r14d jne 0x2ad546 testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax movl 0x5c(%rsp), %ecx cmpl %ecx, %eax cmovgel %ecx, %eax jmp 0x2ad546 movl %esi, %r14d movl %eax, %ecx negl %ecx cmovsl %eax, %ecx movl 0x5c(%rsp), %edx subl %edx, %ecx movl %ecx, %eax negl %eax cmovnsl %ecx, %eax addl %edx, %eax movq 0x60(%rsp), %r9 movq 0x50(%rsp), %rdi movslq 0x2c(%r9), %rcx movslq 0x30(%r9), %rdx movq 0x40(%r9), %rsi imulq %rdi, %rsi movq 0x10(%r9), %rdi imulq %rdi, %rsi addq (%r9), %rsi movl 0x18(%r9), %r8d movq 0x20(%r9), %r9 movq %rdx, %r10 imulq %rcx, %r10 cltq imulq %rdi, %rax imulq %r10, %rax addq %rsi, %rax movq %rax, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %rdi, 0x80(%rsp) movl %r8d, 0x88(%rsp) movq %r9, 0x90(%rsp) movl $0x2, 0x98(%rsp) movl %ecx, 0x9c(%rsp) movl %edx, 0xa0(%rsp) movabsq $0x100000001, %rax # imm = 0x100000001 movq %rax, 0xa4(%rsp) movq %r10, 0xb0(%rsp) movq 0x28(%rsp), %rax cmpq $0x4, %rax je 0x2ad704 cmpq $0x2, %rax je 0x2ad684 cmpq $0x1, %rax jne 0x2ad736 movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx movsbl 0x38(%rsp), %r9d leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %r14d, %r8d callq 0x2ad910 jmp 0x2ad730 movq 0x8(%rsp), %rax cmpb $0x1, 0xd(%rax) movl 0x34(%rsp), %eax jne 0x2ad650 movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) movl 0x34(%rsp), %eax jne 0x2ad650 callq 0x6460e movss 0x1c(%rsp), %xmm0 testl %ebp, %ebp movl $0x0, %ecx cmovlel %ecx, %ebp imulq 0x48(%rsp), %r14 imulq %r15, %r14 addq %r12, %r14 imulq %r14, %rbx addq %rbx, %r13 xorl %ecx, %ecx cmpq %rcx, %rbp je 0x2ad736 movw %ax, (%r13,%rcx,2) incq %rcx jmp 0x2ad670 movq 0x8(%rsp), %rax movl 0xd0(%rax), %r13d movl 0xd8(%rax), %ebp cmpb $0x1, 0xd(%rax) movl 0x34(%rsp), %eax jne 0x2ad6b4 movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) movl 0x34(%rsp), %eax jne 0x2ad6b4 callq 0x6460e movzwl %ax, %r9d leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %r13d, %edx movl %ebp, %ecx movl %r14d, %r8d callq 0x2adf5b jmp 0x2ad730 testl %ebp, %ebp movl $0x0, %eax cmovlel %eax, %ebp imulq 0x48(%rsp), %r14 imulq %r15, %r14 addq %r12, %r14 imulq %r14, %rbx addq %rbx, %r13 xorl %eax, %eax cmpl %eax, %ebp je 0x2ad736 movss %xmm0, (%r13,%rax,4) incq %rax jmp 0x2ad6f4 movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx leaq 0x70(%rsp), %rdi leaq 0xb8(%rsp), %rsi movl %r14d, %r8d movss 0x1c(%rsp), %xmm0 callq 0x2ae697 movss 0x1c(%rsp), %xmm0 movq 0x48(%rsp), %rdi incq %rdi movq 0x110(%rsp), %rbx movq 0x50(%rsp), %rdx jmp 0x2ad3b5 incq %rdx jmp 0x2ad362 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rbx) pxor %xmm0, %xmm0 movdqu %xmm0, (%rbx) movdqu %xmm0, 0xc(%rbx) andl $0x0, 0x38(%rbx) movdqu %xmm0, 0x28(%rbx) movups (%r14), %xmm0 movups %xmm0, (%rbx) movq 0x10(%r14), %rax movq %rax, 0x10(%rbx) movl 0x18(%r14), %eax movl %eax, 0x18(%rbx) movq 0x20(%r14), %rax movq %rax, 0x20(%rbx) movdqu 0x28(%r14), %xmm0 movdqu %xmm0, 0x28(%rbx) movl 0x38(%r14), %eax movl %eax, 0x38(%rbx) movq 0x40(%r14), %rax movq %rax, 0x40(%rbx) jmp 0x2ad8eb movq 0x8(%rsp), %rax movl 0xd8(%rax), %ebp movl 0xe0(%rax), %r15d cmpb $0x1, 0xd(%rax) jne 0x2ad824 movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) jne 0x2ad824 movq 0x8(%rsp), %rax movd 0xe4(%rax), %xmm0 callq 0x6460e jmp 0x2ad830 movq 0x8(%rsp), %rax movl 0xd8(%rax), %ecx movl 0xe0(%rax), %r8d movd 0xe4(%rax), %xmm0 movq $0x0, 0x10(%rsp) movq %r14, %rdi movq %rbx, %rsi xorl %edx, %edx callq 0x2ae697 jmp 0x2ad8eb movq 0x8(%rsp), %rax movzwl 0xe6(%rax), %eax movzwl %ax, %r9d movq $0x0, 0x10(%rsp) movq %r14, %rdi movq %rbx, %rsi xorl %edx, %edx movl %ebp, %ecx movl %r15d, %r8d callq 0x2adf5b jmp 0x2ad8eb movq 0x8(%rsp), %rax movl 0xd0(%rax), %ebp movl 0xd8(%rax), %r15d movl 0xe0(%rax), %r13d cmpb $0x1, 0xd(%rax) jne 0x2ad8bf movq 0x20(%rsp), %rax cmpb $0x1, 0x22(%rax) jne 0x2ad8bf movq 0x8(%rsp), %rax movd 0xe4(%rax), %xmm0 callq 0x6460e jmp 0x2ad8cb movq 0x8(%rsp), %rax movl 0xd0(%rax), %edx movl 0xd8(%rax), %ecx movl 0xe0(%rax), %r8d movd 0xe4(%rax), %xmm0 movq %r14, %rdi movq %rbx, %rsi callq 0x2ae697 jmp 0x2ad8e2 movq 0x8(%rsp), %rax movzwl 0xe6(%rax), %eax movzwl %ax, %r9d movq %r14, %rdi movq %rbx, %rsi movl %ebp, %edx movl %r15d, %ecx movl %r13d, %r8d callq 0x2adf5b movq $0x0, 0x10(%rsp) movq 0x10(%rsp), %rax addq $0x118, %rsp # imm = 0x118 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2ad908 jmp 0x2ad908 jmp 0x2ad908 movq %rax, %rdi callq 0x5f340

/csukuangfj[P]ncnn/src/layer/padding.cpp

void ncnn::copy_make_border_image<signed char>(ncnn::Mat const&, ncnn::Mat&, int, int, int, signed char)

static void copy_make_border_image(const Mat& src, Mat& dst, int top, int left, int type, T v) { int w = dst.w; int h = dst.h; const T* ptr = src; T* outptr = dst; if (type == 0) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = v; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = v; } ptr += src.w; outptr += w; } // fill bottom for (; y < h; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } } if (type == 1) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } ptr += src.w; outptr += w; } // fill bottom ptr -= src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } } if (type == 2) { int y = 0; // fill top ptr += top * src.w; for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } ptr += src.w; outptr += w; } // fill bottom ptr -= 2 * src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } } }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x38, %rsp movl %r9d, %r13d movl %ecx, %r9d movq %rdi, %r14 movslq 0x2c(%rsi), %r12 movl 0x30(%rsi), %eax movl %eax, 0x2c(%rsp) movq (%rdi), %r15 movq (%rsi), %rbp movslq %ecx, %rax movq %rax, (%rsp) testl %r8d, %r8d movl %ecx, 0xc(%rsp) movl %edx, 0x8(%rsp) jne 0x2ada5f movl %r8d, 0x30(%rsp) xorl %eax, %eax testl %r12d, %r12d movl $0x0, %ebx cmovgl %r12d, %ebx testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi cmpl %esi, %eax je 0x2ad989 xorl %ecx, %ecx cmpq %rcx, %rbx je 0x2ad982 movb %r13b, (%rbp,%rcx) incq %rcx jmp 0x2ad973 addq %r12, %rbp incl %eax jmp 0x2ad96d xorl %edi, %edi movq (%rsp), %rax testl %eax, %eax cmovgl %r9d, %edi negq %rax movq %rax, 0x18(%rsp) movq %rdi, 0x20(%rsp) movl 0x30(%r14), %eax addl 0x8(%rsp), %eax cmpl %eax, %esi jge 0x2ada54 xorl %eax, %eax cmpq %rax, %rdi je 0x2ad9c3 movb %r13b, (%rbp,%rax) incq %rax jmp 0x2ad9b4 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2ad9f4 movq 0x18(%rsp), %rax leaq (%r15,%rax), %rcx movq %rdi, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2ada1c movb (%rcx,%rax), %dl movb %dl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2ad9d9 movq (%rsp), %rax leaq (%rax,%rbp), %rdi movl %esi, 0x10(%rsp) movq %r15, %rsi callq 0x5f3c0 movq 0x20(%rsp), %rdi movl 0x10(%rsp), %esi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %edi, %eax cltq cmpq %r12, %rax jge 0x2ada2d movb %r13b, (%rbp,%rax) incq %rax jmp 0x2ada1e movslq 0x2c(%r14), %rax addq %rax, %r15 addq %r12, %rbp incl %esi jmp 0x2ad9a2 xorl %eax, %eax cmpq %rax, %rbx je 0x2ada4f movb %r13b, (%rbp,%rax) incq %rax jmp 0x2ada40 addq %r12, %rbp incl %esi cmpl 0x2c(%rsp), %esi jl 0x2ada3e movl 0x30(%rsp), %r8d cmpl $0x1, %r8d jne 0x2adb3c movl %r8d, 0x30(%rsp) xorl %r8d, %r8d movq (%rsp), %rax testl %eax, %eax movl $0x0, %ebx cmovgl %r9d, %ebx movl 0x8(%rsp), %ecx testl %ecx, %ecx movl $0x0, %esi cmovgl %ecx, %esi leaq (%rbx,%rbp), %rdi movq %r15, %r13 subq %rax, %r13 movl %esi, 0x20(%rsp) cmpl %esi, %r8d je 0x2adb45 xorl %eax, %eax cmpq %rax, %rbx je 0x2adab8 movb (%r15), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adaa7 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2adae2 movq %rbx, %rax addl %r9d, %edx movslq %edx, %rcx cmpq %rcx, %rax jge 0x2adb15 movb (%r13,%rax), %cl movb %cl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2adac5 movq (%rsp), %rax movq %rdi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movq %r15, %rsi movq %r8, 0x18(%rsp) callq 0x5f3c0 movq 0x18(%rsp), %r8 movq 0x10(%rsp), %rdi movl 0x20(%rsp), %esi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %ebx, %eax cltq cmpq %r12, %rax jge 0x2adb2e movslq 0x2c(%r14), %rcx movb -0x1(%rcx,%r15), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adb17 addq %r12, %rbp incl %r8d addq %r12, %rdi jmp 0x2ada9c movl 0x8(%rsp), %edx jmp 0x2adcb4 movq (%rsp), %rax negq %rax movq %rax, 0x18(%rsp) movl 0x30(%r14), %eax addl 0x8(%rsp), %eax cmpl %eax, %esi jge 0x2adbf9 xorl %eax, %eax cmpq %rax, %rbx je 0x2adb74 movb (%r15), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adb63 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2adba5 movq 0x18(%rsp), %rax leaq (%r15,%rax), %rcx movq %rbx, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2adbd0 movb (%rcx,%rax), %dl movb %dl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2adb8a movq (%rsp), %rax movq %rdi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movl %esi, %r13d movq %r15, %rsi callq 0x5f3c0 movq 0x10(%rsp), %rdi movl %r13d, %esi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %ebx, %eax cltq movslq 0x2c(%r14), %rcx cmpq %r12, %rax jge 0x2adbe9 movb -0x1(%rcx,%r15), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adbd2 addq %rcx, %r15 addq %r12, %rbp incl %esi addq %r12, %rdi jmp 0x2adb51 movslq 0x2c(%r14), %rax movq %r15, %r13 subq %rax, %r13 addq (%rsp), %rax subq %rax, %r15 cmpl 0x2c(%rsp), %esi jge 0x2adca8 xorl %eax, %eax cmpq %rax, %rbx je 0x2adc28 movb (%r13), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adc16 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2adc55 movq %rdi, %rcx movq %rbx, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2adc82 movb (%r15,%rax), %dl movb %dl, (%rcx) incq %rax movl 0x2c(%r14), %edx incq %rcx jmp 0x2adc38 movq (%rsp), %rax movq %rdi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movl %esi, 0x20(%rsp) movq %r13, %rsi callq 0x5f3c0 movq 0x10(%rsp), %rdi movl 0x20(%rsp), %esi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %ebx, %eax cltq cmpq %r12, %rax jge 0x2adc9b movslq 0x2c(%r14), %rcx movb -0x1(%r13,%rcx), %cl movb %cl, (%rbp,%rax) incq %rax jmp 0x2adc84 addq %r12, %rbp incl %esi addq %r12, %rdi jmp 0x2adc0a movq %r13, %r15 movl 0x8(%rsp), %edx movl 0x30(%rsp), %r8d cmpl $0x2, %r8d jne 0x2adf4c movl 0x2c(%r14), %eax movl %eax, %ecx imull %edx, %ecx movslq %ecx, %rcx addq %rcx, %r15 xorl %esi, %esi movq (%rsp), %rcx testl %ecx, %ecx movl $0x0, %r13d cmovgl %ecx, %r13d testl %edx, %edx movl $0x0, %edi cmovgl %edx, %edi movq %rbp, %rbx addq %r13, %rbx movq %rcx, %rdx negq %rdx movq %rdx, 0x20(%rsp) leal -0x2(%rcx), %r10d movl %r10d, 0x18(%rsp) movl %edi, 0x10(%rsp) cmpl %edi, %esi je 0x2addc4 movq (%rsp), %rax addq %r15, %rax xorl %ecx, %ecx cmpq %rcx, %r13 je 0x2add2b movb (%rax), %dl movb %dl, (%rbp,%rcx) incq %rcx decq %rax jmp 0x2add18 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2add5c movq 0x20(%rsp), %rax leaq (%r15,%rax), %rcx movq %r13, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2add8b movb (%rcx,%rax), %dl movb %dl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2add41 movq (%rsp), %rax leaq (%rax,%rbp), %rdi movq %rsi, 0x30(%rsp) movq %r15, %rsi callq 0x5f3c0 movq 0x30(%rsp), %rsi movl 0x18(%rsp), %r10d movl 0x10(%rsp), %edi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %r13d, %eax cltq movl %r10d, %ecx subl %eax, %ecx cmpq %r12, %rax jge 0x2addb0 movl 0x2c(%r14), %edx leal (%rcx,%rdx,2), %edx movslq %edx, %rdx movb (%r15,%rdx), %dl movb %dl, (%rbp,%rax) incq %rax decl %ecx jmp 0x2add92 addq %r12, %rbp movslq 0x2c(%r14), %rax subq %rax, %r15 incl %esi addq %r12, %rbx jmp 0x2add07 movl 0x8(%rsp), %esi movl 0x30(%r14), %ecx addl %esi, %ecx cmpl %ecx, %edi jge 0x2ade88 movq (%rsp), %rax addq %r15, %rax xorl %ecx, %ecx cmpq %rcx, %r13 je 0x2addf2 movb (%rax), %dl movb %dl, (%rbp,%rcx) incq %rcx decq %rax jmp 0x2adddf movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2ade23 movq 0x20(%rsp), %rax leaq (%r15,%rax), %rcx movq %r13, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2ade50 movb (%rcx,%rax), %dl movb %dl, (%rbp,%rax) incq %rax movl 0x2c(%r14), %edx jmp 0x2ade08 movq (%rsp), %rax movl %edi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movq %r15, %rsi callq 0x5f3c0 movl 0x8(%rsp), %esi movl 0x18(%rsp), %r10d movl 0x10(%rsp), %edi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %r13d, %eax movslq %eax, %rcx movl %r10d, %edx subl %ecx, %edx movl 0x2c(%r14), %eax cmpq %r12, %rcx jge 0x2ade75 leal (%rdx,%rax,2), %eax cltq movb (%r15,%rax), %al movb %al, (%rbp,%rcx) incq %rcx decl %edx jmp 0x2ade58 movslq %eax, %rcx addq %rcx, %r15 addq %r12, %rbp incl %edi addq %r12, %rbx jmp 0x2addc8 addl %eax, %eax cltq subq %rax, %r15 cmpl 0x2c(%rsp), %edi jge 0x2adf4c movq (%rsp), %rax addq %r15, %rax xorl %ecx, %ecx cmpq %rcx, %r13 je 0x2adeb5 movb (%rax), %dl movb %dl, (%rbp,%rcx) incq %rcx decq %rax jmp 0x2adea2 movslq 0x2c(%r14), %rdx cmpq $0xb, %rdx jg 0x2adeea movq 0x20(%rsp), %rax leaq (%r15,%rax), %rcx movq %rbx, %rsi movq %r13, %rax addl %r9d, %edx movslq %edx, %rdx cmpq %rdx, %rax jge 0x2adf13 movb (%rcx,%rax), %dl movb %dl, (%rsi) incq %rax movl 0x2c(%r14), %edx incq %rsi jmp 0x2adece movq (%rsp), %rax movl %edi, 0x10(%rsp) leaq (%rax,%rbp), %rdi movq %r15, %rsi callq 0x5f3c0 movl 0x18(%rsp), %r10d movl 0x10(%rsp), %edi movl 0xc(%rsp), %r9d movl 0x2c(%r14), %eax addl %r13d, %eax cltq movl %r10d, %ecx subl %eax, %ecx cmpq %r12, %rax jge 0x2adf38 movl 0x2c(%r14), %edx leal (%rcx,%rdx,2), %edx movslq %edx, %rdx movb (%r15,%rdx), %dl movb %dl, (%rbp,%rax) incq %rax decl %ecx jmp 0x2adf1a addq %r12, %rbp movslq 0x2c(%r14), %rax subq %rax, %r15 incl %edi addq %r12, %rbx jmp 0x2ade8f addq $0x38, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq

/csukuangfj[P]ncnn/src/layer/padding.cpp

void ncnn::copy_make_border_image<unsigned short>(ncnn::Mat const&, ncnn::Mat&, int, int, int, unsigned short)

static void copy_make_border_image(const Mat& src, Mat& dst, int top, int left, int type, T v) { int w = dst.w; int h = dst.h; const T* ptr = src; T* outptr = dst; if (type == 0) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = v; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = v; } ptr += src.w; outptr += w; } // fill bottom for (; y < h; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } } if (type == 1) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } ptr += src.w; outptr += w; } // fill bottom ptr -= src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } } if (type == 2) { int y = 0; // fill top ptr += top * src.w; for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } ptr += src.w; outptr += w; } // fill bottom ptr -= 2 * src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } } }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x58, %rsp movslq 0x2c(%rsi), %rbx movl 0x30(%rsi), %eax movl %eax, 0x4c(%rsp) movq (%rdi), %r12 movq (%rsi), %rbp movq %rcx, 0x28(%rsp) movslq %ecx, %rax movq %rax, 0x18(%rsp) testl %r8d, %r8d movq %rdi, 0x40(%rsp) movl %edx, 0xc(%rsp) jne 0x2ae0d4 movl %r9d, %r13d movl %r8d, 0x30(%rsp) xorl %eax, %eax testl %ebx, %ebx movl $0x0, %r14d cmovgl %ebx, %r14d testl %edx, %edx movl $0x0, %r9d cmovgl %edx, %r9d leaq (%rbx,%rbx), %r10 cmpl %r9d, %eax je 0x2adfdd xorl %ecx, %ecx cmpq %rcx, %r14 je 0x2adfd6 movw %r13w, (%rbp,%rcx,2) incq %rcx jmp 0x2adfc6 incl %eax addq %r10, %rbp jmp 0x2adfbf xorl %r15d, %r15d movq 0x18(%rsp), %rax testl %eax, %eax cmovgl 0x28(%rsp), %r15d addq %rax, %rax negq %rax movq %rax, 0x38(%rsp) movq %r10, 0x20(%rsp) movl 0x30(%rdi), %eax addl 0xc(%rsp), %eax cmpl %eax, %r9d jge 0x2ae0ac xorl %eax, %eax cmpq %rax, %r15 je 0x2ae01f movw %r13w, (%rbp,%rax,2) incq %rax jmp 0x2ae00f movl 0x2c(%rdi), %edx cmpl $0xb, %edx jg 0x2ae052 movq 0x28(%rsp), %rax addl %edx, %eax movslq %eax, %rcx movq 0x38(%rsp), %rax leaq (%r12,%rax), %rsi movq %r15, %rax cmpq %rcx, %rax jge 0x2ae088 movzwl (%rsi,%rax,2), %r8d movw %r8w, (%rbp,%rax,2) incq %rax jmp 0x2ae03d movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addq %rdx, %rdx movq %r12, %rsi movl %r9d, 0x10(%rsp) callq 0x5f3c0 movq 0x20(%rsp), %r10 movl 0x10(%rsp), %r9d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r15), %eax cltq cmpq %rbx, %rax jge 0x2ae09a movw %r13w, (%rbp,%rax,2) incq %rax jmp 0x2ae08a movslq %edx, %rax leaq (%r12,%rax,2), %r12 addq %r10, %rbp incl %r9d jmp 0x2adffd movl 0x4c(%rsp), %ecx cmpl %ecx, %r9d jge 0x2ae0cf xorl %eax, %eax cmpq %rax, %r14 je 0x2ae0c7 movw %r13w, (%rbp,%rax,2) incq %rax jmp 0x2ae0b7 addq %r10, %rbp incl %r9d jmp 0x2ae0b0 movl 0x30(%rsp), %r8d cmpl $0x1, %r8d jne 0x2ae1dc movl %r8d, 0x30(%rsp) xorl %esi, %esi movq 0x18(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl 0x28(%rsp), %r14d movl 0xc(%rsp), %ecx testl %ecx, %ecx movl $0x0, %r8d cmovgl %ecx, %r8d leaq (%rbx,%rbx), %r9 leaq (%rbp,%r14,2), %r10 addq %rax, %rax movq %r12, %r15 movq %rax, 0x50(%rsp) subq %rax, %r15 movq %r9, 0x20(%rsp) cmpl %r8d, %esi je 0x2ae1e5 xorl %eax, %eax cmpq %rax, %r14 je 0x2ae143 movzwl (%r12), %ecx movw %cx, (%rbp,%rax,2) incq %rax jmp 0x2ae12f movl 0x2c(%rdi), %edx cmpl $0xb, %edx jg 0x2ae169 addl 0x28(%rsp), %edx movslq %edx, %rcx movq %r14, %rax cmpq %rcx, %rax jge 0x2ae1b1 movzwl (%r15,%rax,2), %edx movw %dx, (%rbp,%rax,2) incq %rax jmp 0x2ae155 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addq %rdx, %rdx movq %rsi, 0x38(%rsp) movq %r12, %rsi movl %r8d, %r13d movq %r10, 0x10(%rsp) callq 0x5f3c0 movq 0x38(%rsp), %rsi movq 0x10(%rsp), %r10 movq 0x20(%rsp), %r9 movl %r13d, %r8d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %ecx leal (%rcx,%r14), %eax jmp 0x2ae1b4 movl 0x2c(%rdi), %ecx decl %ecx movslq %ecx, %rcx cltq cmpq %rbx, %rax jge 0x2ae1cf movzwl (%r12,%rcx,2), %edx movw %dx, (%rbp,%rax,2) incq %rax jmp 0x2ae1bb addq %r9, %rbp incl %esi addq %r9, %r10 jmp 0x2ae124 movl 0xc(%rsp), %ecx jmp 0x2ae38a movq 0x50(%rsp), %rax negq %rax movq %rax, 0x10(%rsp) movl 0x30(%rdi), %eax addl 0xc(%rsp), %eax cmpl %eax, %r8d jge 0x2ae2b4 xorl %eax, %eax cmpq %rax, %r14 je 0x2ae218 movzwl (%r12), %ecx movw %cx, (%rbp,%rax,2) incq %rax jmp 0x2ae204 movl 0x2c(%rdi), %edx cmpl $0xb, %edx jg 0x2ae246 addl 0x28(%rsp), %edx movslq %edx, %rax movq 0x10(%rsp), %rcx leaq (%r12,%rcx), %rdx movq %r14, %rcx cmpq %rax, %rcx jge 0x2ae280 movzwl (%rdx,%rcx,2), %esi movw %si, (%rbp,%rcx,2) incq %rcx jmp 0x2ae233 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addq %rdx, %rdx movq %r12, %rsi movl %r8d, %r13d movq %r10, %r15 callq 0x5f3c0 movq %r15, %r10 movq 0x20(%rsp), %r9 movl %r13d, %r8d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r14), %ecx jmp 0x2ae283 movl 0x2c(%rdi), %eax leal -0x1(%rax), %edx movslq %edx, %rdx movslq %ecx, %rcx cmpq %rbx, %rcx jge 0x2ae2a0 movzwl (%r12,%rdx,2), %esi movw %si, (%rbp,%rcx,2) incq %rcx jmp 0x2ae28c cltq leaq (%r12,%rax,2), %r12 addq %r9, %rbp incl %r8d addq %r9, %r10 jmp 0x2ae1f2 movslq 0x2c(%rdi), %rdx movq %r12, %r13 subq %rdx, %r13 subq %rdx, %r13 movq 0x50(%rsp), %rax leaq (%rax,%rdx,2), %rax subq %rax, %r12 cmpl 0x4c(%rsp), %r8d jge 0x2ae37e xorl %eax, %eax cmpq %rax, %r14 je 0x2ae2ee movzwl (%r13), %ecx movw %cx, (%rbp,%rax,2) incq %rax jmp 0x2ae2da cmpl $0xb, %edx jg 0x2ae316 addl 0x28(%rsp), %edx movslq %edx, %rcx movq %r10, %rdx movq %r14, %rax cmpq %rcx, %rax jge 0x2ae353 movzwl (%r12,%rax,2), %esi movw %si, (%rdx) incq %rax addq $0x2, %rdx jmp 0x2ae300 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addl %edx, %edx movq %r13, %rsi movl %r8d, %r15d movq %r10, 0x10(%rsp) callq 0x5f3c0 movq 0x10(%rsp), %r10 movq 0x20(%rsp), %r9 movl %r15d, %r8d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r14), %eax jmp 0x2ae356 movl 0x2c(%rdi), %edx movslq %edx, %rcx cltq cmpq %rbx, %rax jge 0x2ae370 movzwl -0x2(%r13,%rcx,2), %esi movw %si, (%rbp,%rax,2) incq %rax jmp 0x2ae35b addq %r9, %rbp incl %r8d addq %r9, %r10 jmp 0x2ae2cd movq %r13, %r12 movl 0xc(%rsp), %ecx movl 0x30(%rsp), %r8d cmpl $0x2, %r8d jne 0x2ae688 movl 0x2c(%rdi), %edx movl %edx, %eax imull %ecx, %eax cltq leaq (%r12,%rax,2), %r15 xorl %r8d, %r8d movq 0x18(%rsp), %rax testl %eax, %eax movl $0x0, %r12d cmovgl %eax, %r12d testl %ecx, %ecx movl $0x0, %r13d cmovgl %ecx, %r13d leaq (%rbx,%rbx), %r9 leaq (,%r12,2), %r10 addq %rbp, %r10 leaq (%rax,%rax), %r11 movq %r11, %rcx negq %rcx movq %rcx, 0x38(%rsp) leal -0x2(%rax), %r14d movq %r9, 0x10(%rsp) movq %r11, 0x20(%rsp) movq %r14, 0x30(%rsp) movl %r13d, 0x50(%rsp) cmpl %r13d, %r8d je 0x2ae4cb leaq (%r15,%r11), %rax xorl %ecx, %ecx cmpq %rcx, %r12 je 0x2ae41d movzwl (%rax), %esi movw %si, (%rbp,%rcx,2) incq %rcx addq $-0x2, %rax jmp 0x2ae407 cmpl $0xb, %edx jg 0x2ae448 addl 0x28(%rsp), %edx movslq %edx, %rcx movq 0x38(%rsp), %rax leaq (%r15,%rax), %rdx movq %r12, %rax cmpq %rcx, %rax jge 0x2ae490 movzwl (%rdx,%rax,2), %esi movw %si, (%rbp,%rax,2) incq %rax jmp 0x2ae435 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addl %edx, %edx movq %r15, %rsi movq %r10, %r13 movq %r8, %r14 callq 0x5f3c0 movq %r14, %r8 movq 0x20(%rsp), %r11 movq %r13, %r10 movl 0x50(%rsp), %r13d movq 0x30(%rsp), %r14 movq 0x10(%rsp), %r9 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r12), %eax jmp 0x2ae493 movl 0x2c(%rdi), %edx cltq leal (%r14,%rdx,2), %ecx subl %eax, %ecx cmpq %rbx, %rax jge 0x2ae4b4 movslq %ecx, %rcx movzwl (%r15,%rcx,2), %esi movw %si, (%rbp,%rax,2) incq %rax decl %ecx jmp 0x2ae49b addq %r9, %rbp movslq %edx, %rax addq %rax, %rax subq %rax, %r15 incl %r8d addq %r9, %r10 jmp 0x2ae3f8 movl 0xc(%rsp), %r8d movl 0x30(%rdi), %eax addl %r8d, %eax cmpl %eax, %r13d jge 0x2ae5ac leaq (%r15,%r11), %rax xorl %ecx, %ecx cmpq %rcx, %r12 je 0x2ae4fb movzwl (%rax), %esi movw %si, (%rbp,%rcx,2) incq %rcx addq $-0x2, %rax jmp 0x2ae4e5 cmpl $0xb, %edx jg 0x2ae526 addl 0x28(%rsp), %edx movslq %edx, %rcx movq 0x38(%rsp), %rax leaq (%r15,%rax), %rdx movq %r12, %rax cmpq %rcx, %rax jge 0x2ae573 movzwl (%rdx,%rax,2), %esi movw %si, (%rbp,%rax,2) incq %rax jmp 0x2ae513 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addl %edx, %edx movq %r15, %rsi movl %r13d, 0x50(%rsp) movq %r10, %r13 movl %r8d, %r14d callq 0x5f3c0 movl %r14d, %r8d movq 0x20(%rsp), %r11 movq %r13, %r10 movl 0x50(%rsp), %r13d movq 0x30(%rsp), %r14 movq 0x10(%rsp), %r9 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r12), %eax jmp 0x2ae576 movl 0x2c(%rdi), %edx cltq leal (%r14,%rdx,2), %ecx subl %eax, %ecx cmpq %rbx, %rax jge 0x2ae597 movslq %ecx, %rcx movzwl (%r15,%rcx,2), %esi movw %si, (%rbp,%rax,2) incq %rax decl %ecx jmp 0x2ae57e movslq %edx, %rax leaq (%r15,%rax,2), %r15 addq %r9, %rbp incl %r13d addq %r9, %r10 jmp 0x2ae4d0 leal (%rdx,%rdx), %eax cltq addq %rax, %rax subq %rax, %r15 cmpl 0x4c(%rsp), %r13d jge 0x2ae688 leaq (%r15,%r11), %rax xorl %ecx, %ecx cmpq %rcx, %r12 je 0x2ae5de movzwl (%rax), %esi movw %si, (%rbp,%rcx,2) incq %rcx addq $-0x2, %rax jmp 0x2ae5c8 cmpl $0xb, %edx jg 0x2ae610 addl 0x28(%rsp), %edx movslq %edx, %rcx movq 0x38(%rsp), %rax leaq (%r15,%rax), %rdx movq %r10, %rsi movq %r12, %rax cmpq %rcx, %rax jge 0x2ae64d movzwl (%rdx,%rax,2), %r8d movw %r8w, (%rsi) incq %rax addq $0x2, %rsi jmp 0x2ae5f9 movq 0x18(%rsp), %rax leaq (,%rax,2), %rdi addq %rbp, %rdi addl %edx, %edx movq %r15, %rsi movq %r10, %r14 callq 0x5f3c0 movq 0x20(%rsp), %r11 movq %r14, %r10 movq 0x30(%rsp), %r14 movq 0x10(%rsp), %r9 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r12), %eax jmp 0x2ae650 movl 0x2c(%rdi), %edx cltq leal (%r14,%rdx,2), %ecx subl %eax, %ecx cmpq %rbx, %rax jge 0x2ae671 movslq %ecx, %rcx movzwl (%r15,%rcx,2), %esi movw %si, (%rbp,%rax,2) incq %rax decl %ecx jmp 0x2ae658 addq %r9, %rbp movslq %edx, %rax addq %rax, %rax subq %rax, %r15 incl %r13d addq %r9, %r10 jmp 0x2ae5b7 addq $0x58, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq

/csukuangfj[P]ncnn/src/layer/padding.cpp

void ncnn::copy_make_border_image<float>(ncnn::Mat const&, ncnn::Mat&, int, int, int, float)

static void copy_make_border_image(const Mat& src, Mat& dst, int top, int left, int type, T v) { int w = dst.w; int h = dst.h; const T* ptr = src; T* outptr = dst; if (type == 0) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = v; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = v; } ptr += src.w; outptr += w; } // fill bottom for (; y < h; y++) { int x = 0; for (; x < w; x++) { outptr[x] = v; } outptr += w; } } if (type == 1) { int y = 0; // fill top for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } ptr += src.w; outptr += w; } // fill bottom ptr -= src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[0]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - 1]; } outptr += w; } } if (type == 2) { int y = 0; // fill top ptr += top * src.w; for (; y < top; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } // fill center for (; y < (top + src.h); y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } ptr += src.w; outptr += w; } // fill bottom ptr -= 2 * src.w; for (; y < h; y++) { int x = 0; for (; x < left; x++) { outptr[x] = ptr[left - x]; } if (src.w < 12) { for (; x < (left + src.w); x++) { outptr[x] = ptr[x - left]; } } else { memcpy(outptr + left, ptr, src.w * sizeof(T)); x += src.w; } for (; x < w; x++) { outptr[x] = ptr[src.w - (x - left - src.w) - 2]; } outptr += w; ptr -= src.w; } } }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x58, %rsp movslq 0x2c(%rsi), %rbx movl 0x30(%rsi), %eax movl %eax, 0x4c(%rsp) movq (%rdi), %r12 movq (%rsi), %rbp movq %rcx, 0x30(%rsp) movslq %ecx, %rax movq %rax, 0x18(%rsp) testl %r8d, %r8d movq %rdi, 0x40(%rsp) movl %edx, 0xc(%rsp) jne 0x2ae81b movl %r8d, 0x38(%rsp) xorl %eax, %eax testl %ebx, %ebx movl $0x0, %r14d cmovgl %ebx, %r14d testl %edx, %edx movl $0x0, %r9d cmovgl %edx, %r9d leaq (,%rbx,4), %r15 cmpl %r9d, %eax je 0x2ae71a xorl %ecx, %ecx cmpq %rcx, %r14 je 0x2ae713 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2ae703 incl %eax addq %r15, %rbp jmp 0x2ae6fc xorl %r13d, %r13d movq 0x18(%rsp), %rax testl %eax, %eax cmovgl 0x30(%rsp), %r13d leaq (,%rax,4), %rax negq %rax movq %rax, 0x20(%rsp) movss %xmm0, 0x28(%rsp) movl 0x30(%rdi), %eax addl 0xc(%rsp), %eax cmpl %eax, %r9d jge 0x2ae80b xorl %eax, %eax cmpq %rax, %r13 je 0x2ae762 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae752 movslq 0x2c(%rdi), %rdx cmpq $0xb, %rdx jg 0x2ae797 movq 0x30(%rsp), %rax addl %edx, %eax movslq %eax, %rcx movq 0x20(%rsp), %rax leaq (%r12,%rax), %rsi movq %r13, %rax cmpq %rcx, %rax jge 0x2ae7cf movss (%rsi,%rax,4), %xmm1 movss %xmm1, (%rbp,%rax,4) incq %rax jmp 0x2ae782 movq 0x18(%rsp), %rax leaq (,%rax,4), %rdi addq %rbp, %rdi shlq $0x2, %rdx movq %r12, %rsi movl %r9d, 0x10(%rsp) callq 0x5f3c0 movl 0x10(%rsp), %r9d movss 0x28(%rsp), %xmm0 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %edx leal (%rdx,%r13), %eax cltq cmpq %rbx, %rax jge 0x2ae7e1 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae7d1 movslq %edx, %rax leaq (%r12,%rax,4), %r12 addq %r15, %rbp incl %r9d jmp 0x2ae740 xorl %eax, %eax cmpq %rax, %r14 je 0x2ae805 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae7f5 addq %r15, %rbp incl %r9d cmpl 0x4c(%rsp), %r9d jl 0x2ae7f3 movl 0xc(%rsp), %edx movl 0x38(%rsp), %r8d cmpl $0x1, %r8d jne 0x2aeaef movl %r8d, 0x38(%rsp) xorl %r10d, %r10d movq 0x18(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl 0x30(%rsp), %r14d testl %edx, %edx movl $0x0, %esi cmovgl %edx, %esi leaq (,%rbx,4), %r8 leaq (%rbp,%r14,4), %r9 leaq (,%rax,4), %rax movq %r12, %r15 movq %rax, 0x50(%rsp) subq %rax, %r15 movq %r8, 0x10(%rsp) movl %esi, 0x28(%rsp) cmpl %esi, %r10d je 0x2ae92f xorl %eax, %eax cmpq %rax, %r14 je 0x2ae894 movss (%r12), %xmm0 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae87e movslq 0x2c(%rdi), %rdx cmpq $0xb, %rdx jg 0x2ae8be addl 0x30(%rsp), %edx movslq %edx, %rcx movq %r14, %rax cmpq %rcx, %rax jge 0x2ae901 movss (%r15,%rax,4), %xmm0 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae8a8 movq 0x18(%rsp), %rax leaq (,%rax,4), %rdi addq %rbp, %rdi shlq $0x2, %rdx movq %r12, %rsi movq %r9, %r13 movq %r10, 0x20(%rsp) callq 0x5f3c0 movq 0x20(%rsp), %r10 movq %r13, %r9 movq 0x10(%rsp), %r8 movl 0x28(%rsp), %esi movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %ecx leal (%rcx,%r14), %eax jmp 0x2ae904 movl 0x2c(%rdi), %ecx decl %ecx movslq %ecx, %rcx cltq cmpq %rbx, %rax jge 0x2ae921 movss (%r12,%rcx,4), %xmm0 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae90b addq %r8, %rbp incl %r10d addq %r8, %r9 jmp 0x2ae873 movq 0x50(%rsp), %rax negq %rax movq %rax, 0x20(%rsp) movl 0xc(%rsp), %r10d movl 0x30(%rdi), %eax addl %r10d, %eax cmpl %eax, %esi jge 0x2aea0e xorl %eax, %eax cmpq %rax, %r14 je 0x2ae967 movss (%r12), %xmm0 movss %xmm0, (%rbp,%rax,4) incq %rax jmp 0x2ae951 movslq 0x2c(%rdi), %rdx cmpq $0xb, %rdx jg 0x2ae999 addl 0x30(%rsp), %edx movslq %edx, %rax movq 0x20(%rsp), %rcx leaq (%r12,%rcx), %rdx movq %r14, %rcx cmpq %rax, %rcx jge 0x2ae9d9 movss (%rdx,%rcx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2ae984 movq 0x18(%rsp), %rax leaq (,%rax,4), %rdi addq %rbp, %rdi shlq $0x2, %rdx movl %esi, %r13d movq %r12, %rsi movq %r9, %r15 callq 0x5f3c0 movq %r15, %r9 movq 0x10(%rsp), %r8 movl %r13d, %esi movl 0xc(%rsp), %r10d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r14), %ecx jmp 0x2ae9dc movl 0x2c(%rdi), %eax leal -0x1(%rax), %edx movslq %edx, %rdx movslq %ecx, %rcx cmpq %rbx, %rcx jge 0x2ae9fb movss (%r12,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2ae9e5 cltq leaq (%r12,%rax,4), %r12 addq %r8, %rbp incl %esi addq %r8, %r9 jmp 0x2ae941 movslq 0x2c(%rdi), %rax leaq (,%rax,4), %rcx movq %r12, %r13 subq %rcx, %r13 movq 0x50(%rsp), %rcx leaq (%rcx,%rax,4), %rcx subq %rcx, %r12 cmpl 0x4c(%rsp), %esi jge 0x2aeae3 xorl %ecx, %ecx cmpq %rcx, %r14 je 0x2aea4e movss (%r13), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2aea38 cmpl $0xb, %eax jg 0x2aea77 addl 0x30(%rsp), %eax cltq movq %r9, %rdx movq %r14, %rcx cmpq %rax, %rcx jge 0x2aeab6 movss (%r12,%rcx,4), %xmm0 movss %xmm0, (%rdx) incq %rcx addq $0x4, %rdx jmp 0x2aea5f movq 0x18(%rsp), %rcx leaq (,%rcx,4), %rdi addq %rbp, %rdi movl %eax, %edx shlq $0x2, %rdx movl %esi, 0x28(%rsp) movq %r13, %rsi movq %r9, %r15 callq 0x5f3c0 movq %r15, %r9 movq 0x10(%rsp), %r8 movl 0x28(%rsp), %esi movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r14), %ecx jmp 0x2aeab9 movl 0x2c(%rdi), %eax movslq %eax, %rdx movslq %ecx, %rcx cmpq %rbx, %rcx jge 0x2aead6 movss -0x4(%r13,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2aeabf addq %r8, %rbp incl %esi addq %r8, %r9 jmp 0x2aea2c movq %r13, %r12 movl 0x38(%rsp), %r8d movl 0xc(%rsp), %edx cmpl $0x2, %r8d jne 0x2aedff movl 0x2c(%rdi), %eax movl %eax, %ecx imull %edx, %ecx movslq %ecx, %rcx leaq (%r12,%rcx,4), %r15 xorl %esi, %esi movq 0x18(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r12d cmovgl %ecx, %r12d testl %edx, %edx movl $0x0, %r9d cmovgl %edx, %r9d leaq (,%rbx,4), %rdx movq %rdx, 0x10(%rsp) leaq (,%r12,4), %r13 addq %rbp, %r13 leaq (,%rcx,4), %r10 movq %r10, %rdx negq %rdx movq %rdx, 0x20(%rsp) leal -0x2(%rcx), %r11d movq %r10, 0x28(%rsp) movq %r11, 0x38(%rsp) cmpl %r9d, %esi je 0x2aec3c leaq (%r15,%r10), %rcx xorl %edx, %edx cmpq %rdx, %r12 je 0x2aeb87 movss (%rcx), %xmm0 movss %xmm0, (%rbp,%rdx,4) incq %rdx addq $-0x4, %rcx jmp 0x2aeb6f cmpl $0xb, %eax jg 0x2aebb3 addl 0x30(%rsp), %eax cltq movq 0x20(%rsp), %rcx leaq (%r15,%rcx), %rdx movq %r12, %rcx cmpq %rax, %rcx jge 0x2aebf9 movss (%rdx,%rcx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2aeb9e movq 0x18(%rsp), %rcx leaq (,%rcx,4), %rdi addq %rbp, %rdi movl %eax, %edx shlq $0x2, %rdx movq %rsi, 0x50(%rsp) movq %r15, %rsi movl %r9d, %r14d callq 0x5f3c0 movq 0x50(%rsp), %rsi movq 0x38(%rsp), %r11 movq 0x28(%rsp), %r10 movl %r14d, %r9d movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r12), %ecx jmp 0x2aebfc movl 0x2c(%rdi), %eax movslq %ecx, %rcx leal (%r11,%rax,2), %edx subl %ecx, %edx cmpq %rbx, %rcx jge 0x2aec20 movslq %edx, %rdx movss (%r15,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx decl %edx jmp 0x2aec05 movq 0x10(%rsp), %rdx addq %rdx, %rbp movslq %eax, %rcx shlq $0x2, %rcx subq %rcx, %r15 incl %esi addq %rdx, %r13 jmp 0x2aeb60 movq 0x10(%rsp), %r14 movl 0x30(%rdi), %ecx addl 0xc(%rsp), %ecx cmpl %ecx, %r9d jge 0x2aed18 leaq (%r15,%r10), %rcx xorl %edx, %edx cmpq %rdx, %r12 je 0x2aec6f movss (%rcx), %xmm0 movss %xmm0, (%rbp,%rdx,4) incq %rdx addq $-0x4, %rcx jmp 0x2aec57 cmpl $0xb, %eax jg 0x2aec9b addl 0x30(%rsp), %eax cltq movq 0x20(%rsp), %rcx leaq (%r15,%rcx), %rdx movq %r12, %rcx cmpq %rax, %rcx jge 0x2aecdc movss (%rdx,%rcx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx jmp 0x2aec86 movq 0x18(%rsp), %rcx leaq (,%rcx,4), %rdi addq %rbp, %rdi movl %eax, %edx shlq $0x2, %rdx movq %r15, %rsi movl %r9d, %r14d callq 0x5f3c0 movq 0x38(%rsp), %r11 movq 0x28(%rsp), %r10 movl %r14d, %r9d movq 0x10(%rsp), %r14 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r12), %ecx jmp 0x2aecdf movl 0x2c(%rdi), %eax movslq %ecx, %rcx leal (%r11,%rax,2), %edx subl %ecx, %edx cmpq %rbx, %rcx jge 0x2aed03 movslq %edx, %rdx movss (%r15,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx decl %edx jmp 0x2aece8 movslq %eax, %rcx leaq (%r15,%rcx,4), %r15 addq %r14, %rbp incl %r9d addq %r14, %r13 jmp 0x2aec41 leal (%rax,%rax), %ecx movslq %ecx, %rcx shlq $0x2, %rcx subq %rcx, %r15 cmpl 0x4c(%rsp), %r9d jge 0x2aedff leaq (%r15,%r10), %rcx xorl %edx, %edx cmpq %rdx, %r12 je 0x2aed4e movss (%rcx), %xmm0 movss %xmm0, (%rbp,%rdx,4) incq %rdx addq $-0x4, %rcx jmp 0x2aed36 cmpl $0xb, %eax jg 0x2aed7f addl 0x30(%rsp), %eax cltq movq 0x20(%rsp), %rcx leaq (%r15,%rcx), %rdx movq %r13, %rsi movq %r12, %rcx cmpq %rax, %rcx jge 0x2aedc0 movss (%rdx,%rcx,4), %xmm0 movss %xmm0, (%rsi) incq %rcx addq $0x4, %rsi jmp 0x2aed68 movq 0x18(%rsp), %rcx leaq (,%rcx,4), %rdi addq %rbp, %rdi movl %eax, %edx shlq $0x2, %rdx movq %r15, %rsi movl %r9d, %r14d callq 0x5f3c0 movq 0x38(%rsp), %r11 movq 0x28(%rsp), %r10 movl %r14d, %r9d movq 0x10(%rsp), %r14 movq 0x40(%rsp), %rdi movl 0x2c(%rdi), %eax leal (%rax,%r12), %ecx jmp 0x2aedc3 movl 0x2c(%rdi), %eax movslq %ecx, %rcx leal (%r11,%rax,2), %edx subl %ecx, %edx cmpq %rbx, %rcx jge 0x2aede7 movslq %edx, %rdx movss (%r15,%rdx,4), %xmm0 movss %xmm0, (%rbp,%rcx,4) incq %rcx decl %edx jmp 0x2aedcc addq %r14, %rbp movslq %eax, %rcx shlq $0x2, %rcx subq %rcx, %r15 incl %r9d addq %r14, %r13 jmp 0x2aed25 addq $0x58, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq

/csukuangfj[P]ncnn/src/layer/padding.cpp

ncnn::Padding_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0xe8, %rsp movq %rcx, %rbx movq %rdx, %r15 movq %rsi, %rcx movq (%rdi), %rax movq -0x18(%rax), %rbp movq %rdi, (%rsp) movl 0xd0(%rdi,%rbp), %r10d testl %r10d, %r10d movq %rsi, 0x8(%rsp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xd4(%rax,%rbp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xd8(%rax,%rbp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xdc(%rax,%rbp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xe8(%rax,%rbp) jne 0x2aeef1 movq (%rsp), %rax cmpl $0x0, 0xec(%rax,%rbp) je 0x2afc57 movl 0x18(%rcx), %r11d movq 0x10(%rcx), %r9 testl %r11d, %r11d je 0x2aef32 leal (,%r9,8), %eax cltd idivl %r11d cmpl $0x8, %eax jne 0x2aef32 movq (%rsp), %rdi movq %rcx, %rsi movq %r15, %rdx movq %rbx, %rcx addq $0xe8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x2afd44 movl 0x28(%rcx), %r13d movl 0x2c(%rcx), %esi movl 0x30(%rcx), %r14d movl 0x34(%rcx), %eax movl %eax, 0x18(%rsp) movl 0x38(%rcx), %eax movq %rax, 0x10(%rsp) cmpl $0x4, %r11d jne 0x2afab9 leal -0x1(%r13), %eax cmpl $0x3, %eax ja 0x2afab9 leaq 0x1497d6(%rip), %rcx # 0x3f8740 movslq (%rcx,%rax,4), %rax addq %rcx, %rax pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) jmpq *%rax movq (%rsp), %rcx movl 0xd8(%rcx,%rbp), %r8d leal (%r8,%rsi,4), %eax addl 0xdc(%rcx,%rbp), %eax testb $0x3, %al sete %r10b movq %r9, %rdx shrq $0x2, %rdx leal (%r10,%r10), %ecx shlq %cl, %rdx testb $0x3, %r8b sete %cl andb %r10b, %cl cmpb $0x1, %cl movq 0x8(%rsp), %rcx jne 0x2afab9 movq (%rsp), %rdi cmpl $0x0, 0xe0(%rdi,%rbp) jne 0x2afab9 sarl $0x2, %eax movq 0x8(%rbx), %r8 pushq $0x4 popq %rbp movq %r15, %rdi movl %eax, %esi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%r15) je 0x2afba1 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2afba1 movq (%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx movss 0xe4(%rdx,%rcx), %xmm0 shufps $0x0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq $0x0, 0x20(%rsp) movq 0x8(%rsp), %rdi movq %r15, %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b0c1b jmp 0x2afba1 movq (%rsp), %rdx movl 0xe8(%rdx,%rbp), %eax movq 0x10(%rsp), %rcx leal (%rax,%rcx,4), %r12d addl 0xec(%rdx,%rbp), %r12d testb $0x3, %r12b sete %dl movq %r9, %r8 shrq $0x2, %r8 leal (%rdx,%rdx), %ecx shlq %cl, %r8 testb $0x3, %al sete %al andb %dl, %al cmpb $0x1, %al movq 0x8(%rsp), %rcx jne 0x2afab9 movq (%rsp), %rcx movl 0xd8(%rcx,%rbp), %eax addl %esi, %eax addl 0xdc(%rcx,%rbp), %eax addl %r14d, %r10d addl 0xd4(%rcx,%rbp), %r10d movq 0x10(%rsp), %rcx leal (,%rcx,4), %ecx cmpl %ecx, %r12d movq 0x8(%rsp), %rcx je 0x2af0e0 movq (%rsp), %rdx cmpl $0x0, 0xe0(%rdx,%rbp) jne 0x2afab9 sarl $0x2, %r12d subq $0x8, %rsp pushq $0x4 popq %r9 movq %r15, %rdi movl %eax, %esi movl %r10d, %edx movl %r12d, %ecx pushq 0x8(%rbx) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%r15) je 0x2afba1 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2afba1 movq (%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax movl 0xe8(%rcx,%rax), %eax pushq $-0x4 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r12d, %r12d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r12d movl %edx, 0xbc(%rsp) movl %edx, %r14d xorl %r9d, %r9d movq %r12, 0xe0(%rsp) movq (%rsp), %rbx cmpq %r12, %r9 je 0x2afba1 movslq 0x2c(%r15), %rax movslq 0x30(%r15), %rcx movq 0x40(%r15), %r13 imulq %r9, %r13 movq 0x10(%r15), %rdi imulq %rdi, %r13 addq (%r15), %r13 movl 0x34(%r15), %esi movl 0x18(%r15), %edx movq 0x20(%r15), %r8 movq %r13, 0x28(%rsp) andq $0x0, 0x30(%rsp) movq %rdi, 0x38(%rsp) movl %edx, 0x40(%rsp) movq %r8, 0x48(%rsp) movl %eax, 0x54(%rsp) movl %ecx, 0x58(%rsp) movl $0x1, 0x5c(%rsp) movl %esi, 0x60(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0x68(%rsp) movl 0x28(%r15), %edx leal -0x1(%rdx), %edi movl %edi, 0x50(%rsp) cmpl $0x4, %edx jne 0x2af1f7 movq %rcx, 0x68(%rsp) movq %rcx, %rax movq (%rsp), %rdx movq (%rdx), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%rdx,%rdi) je 0x2af226 movq (%rsp), %rdx movq 0xf8(%rdx,%rdi), %rdx movq %r9, %r8 shlq $0x4, %r8 movups (%rdx,%r8), %xmm0 jmp 0x2af237 movq (%rsp), %rdx movss 0xe4(%rdx,%rdi), %xmm0 shufps $0x0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] movl %r14d, %r14d movl 0xbc(%rsp), %edx movq %r9, 0x18(%rsp) addl %r9d, %edx setns %r8b cmpl 0x10(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2af417 movq %r14, 0xc0(%rsp) movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r14 movq 0x40(%r10), %r11 movl %edx, %ebp imulq %r11, %rbp imulq %r14, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %r14, 0x80(%rsp) movl %edx, 0x88(%rsp) movq %r9, 0x90(%rsp) movl %eax, 0x9c(%rsp) movl %esi, 0xa0(%rsp) movl $0x1, 0xa4(%rsp) movl %r8d, 0xa8(%rsp) imulq %rax, %rsi movq %r14, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r14 movq %rax, 0xb0(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x98(%rsp) cmpl $0x4, %eax jne 0x2af315 movq %rsi, 0xb0(%rsp) movq (%rsp), %rax movl 0xe0(%rax,%rdi), %eax testl %eax, %eax jne 0x2af36f movl 0xd0(%rbx,%rdi), %edx movl 0xd4(%rbx,%rdi), %ecx movl 0xd8(%rbx,%rdi), %r8d movl 0xdc(%rbx,%rdi), %r9d leaq 0x70(%rsp), %rdi leaq 0x28(%rsp), %rsi movq %r11, 0xd8(%rsp) callq 0x2b0c1b movq 0xd8(%rsp), %r11 movq (%rbx), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%rbx,%rdi), %eax cmpl $0x1, %eax jne 0x2af521 movq (%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r14, %r11 imulq 0xc0(%rsp), %r11 addq %r11, %r12 cmpl %r9d, %r8d je 0x2af43d movaps (%rbp), %xmm0 movl %eax, %edi subl $0x1, %edi jb 0x2af3da movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af3ca xorl %edi, %edi xorl %r10d, %r10d cmpl 0x9c(%rsp), %r10d jge 0x2af3fd movaps (%r12,%rdi), %xmm0 movaps %xmm0, (%r13,%rdi) incl %r10d addq $0x10, %rdi jmp 0x2af3df addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2af412 movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af402 incl %r8d jmp 0x2af3bf imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0x18(%rsp), %r9 subl $0x1, %esi jb 0x2af7be movups %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af429 xorl %edi, %edi cmpl 0xa0(%rsp), %edi jge 0x2af49b movaps (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2af460 movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af44f xorl %r8d, %r8d cmpl 0x9c(%rsp), %r8d jge 0x2af483 movaps (%rbp), %xmm0 movaps %xmm0, (%r13) addq $0x10, %rbp addq $0x10, %r13 incl %r8d jmp 0x2af463 movl %edx, %r8d subl $0x1, %r8d jb 0x2af497 movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af486 incl %edi jmp 0x2af43f movl 0x9c(%rsp), %edi shll $0x2, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2af512 movaps (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2af4d2 movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af4c1 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x9c(%rsp), %r9d jge 0x2af4f7 movaps (%rbp,%r8), %xmm0 movaps %xmm0, (%r13,%r8) incl %r9d addq $0x10, %r8 jmp 0x2af4d8 addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2af50e movaps %xmm0, (%r13) addq $0x10, %r13 jmp 0x2af4fd incl %edi jmp 0x2af4b6 movq -0x18(%rcx), %rdi movq (%rsp), %rax movl 0xe0(%rax,%rdi), %eax cmpl $0x2, %eax pushq $-0x20 popq %r13 jne 0x2af748 movq (%rsp), %r8 movl 0xd0(%r8,%rdi), %r9d movslq 0xd8(%r8,%rdi), %rax movl 0x9c(%rsp), %ecx imull %r9d, %ecx shll $0x2, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x70(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r8,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r8,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x28(%rsp), %r8 shlq $0x4, %rax shlq $0x4, %rdx shlq $0x4, %rsi cmpl %r9d, %r10d je 0x2af626 leaq (%rcx,%rax), %r11 xorl %r12d, %r12d xorl %r14d, %r14d movq %rdx, %r13 addq %r14, %r13 je 0x2af5c6 movaps (%r11,%r14), %xmm0 movaps %xmm0, (%r8,%r12) addq $-0x10, %r14 addq $0x10, %r12 jmp 0x2af5aa subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 pushq $-0x20 popq %r13 cmpl 0x9c(%rsp), %ebp jge 0x2af5ef movaps (%r11), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r11 addq $0x10, %r8 incl %ebp jmp 0x2af5d2 movq %r13, %r14 leaq (%rsi,%r14), %r12 cmpq $-0x20, %r12 je 0x2af60f movaps (%r11,%r14), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r14 jmp 0x2af5f2 movslq 0x9c(%rsp), %r11 shlq $0x4, %r11 subq %r11, %rcx incl %r10d jmp 0x2af597 xorl %r9d, %r9d cmpl 0xa0(%rsp), %r9d jge 0x2af6a2 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r12 addq %r11, %r12 je 0x2af659 movaps (%r10,%r11), %xmm0 movaps %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2af63d subq %r11, %r8 xorl %r10d, %r10d cmpl 0x9c(%rsp), %r10d jge 0x2af67d movaps (%rcx), %xmm0 movaps %xmm0, (%r8) addq $0x10, %rcx addq $0x10, %r8 incl %r10d jmp 0x2af65f movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x20, %r11 je 0x2af69d movaps (%rcx,%r10), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r10 jmp 0x2af680 incl %r9d jmp 0x2af629 movslq 0x9c(%rsp), %r9 shlq $0x5, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2af748 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r12 addq %r11, %r12 je 0x2af6e9 movaps (%r10,%r11), %xmm0 movaps %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2af6cd subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x9c(%rsp), %r11d jge 0x2af711 movaps (%r10), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r10 addq $0x10, %r8 incl %r11d jmp 0x2af6f2 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x20, %r14 je 0x2af731 movaps (%r10,%r11), %xmm0 movaps %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r11 jmp 0x2af714 movslq 0x9c(%rsp), %r10 shlq $0x4, %r10 subq %r10, %rcx incl %r9d jmp 0x2af6ba movq 0x78(%rsp), %rax testq %rax, %rax movq 0xc0(%rsp), %r14 je 0x2af781 lock decl (%rax) jne 0x2af781 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2af779 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2af781 movq %rsi, %rdi callq 0x5f3e0 movq 0x30(%rsp), %rax testq %rax, %rax movq 0x18(%rsp), %r9 je 0x2af7be lock decl (%rax) jne 0x2af7be movq 0x28(%rsp), %rsi movq 0x48(%rsp), %rdi testq %rdi, %rdi je 0x2af7b1 movq (%rdi), %rax callq *0x18(%rax) movq 0x18(%rsp), %r9 jmp 0x2af7be movq %rsi, %rdi callq 0x5f3e0 movq 0x18(%rsp), %r9 incq %r9 incl %r14d movq 0xe0(%rsp), %r12 jmp 0x2af165 movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%rbp) movq 0x8(%rsp), %rcx jne 0x2afab9 movq (%rsp), %rax movl 0xe8(%rax,%rbp), %r12d addl 0x18(%rsp), %r12d addl 0xec(%rax,%rbp), %r12d addl %r14d, %r10d addl 0xd4(%rax,%rbp), %r10d addl 0xd8(%rax,%rbp), %esi addl 0xdc(%rax,%rbp), %esi movq %r15, %rdi movl %r10d, %edx movl %r12d, %ecx movq 0x10(%rsp), %r8 pushq 0x8(%rbx) pushq $0x4 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%r15) je 0x2afba1 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2afba1 xorl %ecx, %ecx testl %r12d, %r12d cmovlel %ecx, %r12d movq 0x10(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x10(%rsp) movabsq $0x100000001, %r13 # imm = 0x100000001 xorl %ebx, %ebx cmpq 0x10(%rsp), %rbx je 0x2afba1 movq (%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rcx,%rax) je 0x2af8b9 movq (%rsp), %rcx movq 0xf8(%rcx,%rax), %rax movq %rbx, %rcx shlq $0x4, %rcx movups (%rax,%rcx), %xmm0 jmp 0x2af8ca movq (%rsp), %rcx movss 0xe4(%rcx,%rax), %xmm0 shufps $0x0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] xorl %ebp, %ebp movaps %xmm0, 0xc0(%rsp) cmpq %r12, %rbp je 0x2afa67 movslq 0x2c(%r15), %rdx movslq 0x30(%r15), %rsi movq 0x40(%r15), %rdi imulq %rbx, %rdi movq 0x10(%r15), %r8 imulq %r8, %rdi addq (%r15), %rdi movl 0x18(%r15), %r9d movq 0x20(%r15), %r10 movq %rsi, %rax imulq %rdx, %rax movq %rbp, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x28(%rsp) andq $0x0, 0x30(%rsp) movq %r8, 0x38(%rsp) movl %r9d, 0x40(%rsp) movq %r10, 0x48(%rsp) movl $0x2, 0x50(%rsp) movl %edx, 0x54(%rsp) movl %esi, 0x58(%rsp) movq %r13, 0x5c(%rsp) movq %rax, 0x68(%rsp) movq (%rsp), %rdi movq (%rdi), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rdi,%rsi), %edi movl %ebp, %edx subl %edi, %edx setns %dil cmpl 0x18(%rsp), %edx setl %r8b testb %r8b, %dil je 0x2afa4f movq 0x8(%rsp), %rcx movslq 0x2c(%rcx), %rax movq %r15, %r14 movslq 0x30(%rcx), %r15 movq 0x40(%rcx), %rdi imulq %rbx, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %r15, %rcx imulq %rax, %rcx movl %edx, %edx movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x70(%rsp) andq $0x0, 0x78(%rsp) movq %r8, 0x80(%rsp) movl %r9d, 0x88(%rsp) movq %r10, 0x90(%rsp) movl $0x2, 0x98(%rsp) movl %eax, 0x9c(%rsp) movl %r15d, 0xa0(%rsp) movq (%rsp), %rax movq %r14, %r15 movq %r13, 0xa4(%rsp) movq %rcx, 0xb0(%rsp) movl 0xd0(%rax,%rsi), %edx movl 0xd4(%rax,%rsi), %ecx movl 0xd8(%rax,%rsi), %r8d movl 0xdc(%rax,%rsi), %r9d leaq 0x70(%rsp), %rdi leaq 0x28(%rsp), %rsi movaps 0xc0(%rsp), %xmm0 callq 0x2b0c1b movaps 0xc0(%rsp), %xmm0 incq %rbp jmp 0x2af8d4 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2afa47 movups %xmm0, (%rcx) addq $0x10, %rcx jmp 0x2afa59 incq %rbx jmp 0x2af880 leal (%r10,%r14,4), %edx movq (%rsp), %rax addl 0xd4(%rax,%rbp), %edx testb $0x3, %dl sete %r8b movq %r9, %rax shrq $0x2, %rax leal (%r8,%r8), %ecx shlq %cl, %rax testb $0x3, %r10b sete %cl andb %r8b, %cl cmpb $0x1, %cl movq 0x8(%rsp), %rcx jne 0x2afab9 movq (%rsp), %rdi cmpl $0x0, 0xe0(%rdi,%rbp) je 0x2afbb8 movq (%rcx), %rax movq %rax, 0x28(%rsp) movq 0x8(%rcx), %rax movq %rax, 0x30(%rsp) movq %r9, 0x38(%rsp) movl %r11d, 0x40(%rsp) movq %rcx, %rdx movq 0x20(%rcx), %rcx movq %rcx, 0x48(%rsp) movl %r13d, 0x50(%rsp) movl %esi, 0x54(%rsp) movl %r14d, 0x58(%rsp) movl 0x18(%rsp), %ecx movl %ecx, 0x5c(%rsp) movq 0x10(%rsp), %rcx movl %ecx, 0x60(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x68(%rsp) testq %rax, %rax je 0x2afb10 lock incl (%rax) cmpl $0x1, %r11d je 0x2afb53 movups (%rbx), %xmm0 movups 0x10(%rbx), %xmm1 movups 0x20(%rbx), %xmm2 movups 0x30(%rbx), %xmm3 leaq 0x70(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x10(%rbx), %rax movq %rax, 0x8(%rcx) leaq 0x28(%rsp), %rsi pushq $0x1 popq %rdx movq 0x8(%rsp), %rdi callq 0x64e3b movq (%rsp), %rdi movq (%rdi), %rax addq -0x18(%rax), %rdi leaq 0x28(%rsp), %rsi movq %r15, %rdx movq %rbx, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x30(%rsp), %rax testq %rax, %rax je 0x2afba1 lock decl (%rax) jne 0x2afba1 movq 0x28(%rsp), %rsi movq 0x48(%rsp), %rdi testq %rdi, %rdi je 0x2afb99 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2afba1 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0xe8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq movq (%rsp), %rcx addl 0xd8(%rcx,%rbp), %esi addl 0xdc(%rcx,%rbp), %esi sarl $0x2, %edx movq 0x8(%rbx), %r9 pushq $0x4 popq %rbp movq %r15, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%r15) je 0x2afba1 movslq 0x38(%r15), %rax imulq 0x40(%r15), %rax testq %rax, %rax je 0x2afba1 movq (%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi movss 0xe4(%rdi,%rsi), %xmm0 shufps $0x0, %xmm0, %xmm0 # xmm0 = xmm0[0,0,0,0] movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq 0x8(%rsp), %rdi movq %r15, %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b0c1b movq $0x0, 0x20(%rsp) jmp 0x2afba1 movq $0x0, 0x20(%rsp) cmpq %rcx, %r15 je 0x2afba1 movq 0x8(%rcx), %rax testq %rax, %rax je 0x2afc75 lock incl (%rax) movq 0x8(%r15), %rax testq %rax, %rax je 0x2afca4 lock decl (%rax) jne 0x2afca4 movq (%r15), %rsi movq 0x20(%r15), %rdi testq %rdi, %rdi je 0x2afc97 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2afc9f movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx andq $0x0, 0x40(%r15) xorps %xmm0, %xmm0 movups %xmm0, (%r15) movups %xmm0, 0xc(%r15) andl $0x0, 0x38(%r15) movups %xmm0, 0x28(%r15) movups (%rcx), %xmm0 movups %xmm0, (%r15) movq 0x10(%rcx), %rax movq %rax, 0x10(%r15) movl 0x18(%rcx), %eax movl %eax, 0x18(%r15) movq 0x20(%rcx), %rax movq %rax, 0x20(%r15) movups 0x28(%rcx), %xmm0 movups %xmm0, 0x28(%r15) movl 0x38(%rcx), %eax movl %eax, 0x38(%r15) movq 0x40(%rcx), %rax movq %rax, 0x40(%r15) jmp 0x2afba1 jmp 0x2afd3b jmp 0x2afd3b jmp 0x2afd3b jmp 0x2afd02 movq %rax, %rbx movq 0x30(%rsp), %rax testq %rax, %rax je 0x2afd33 lock decl (%rax) jne 0x2afd33 movq 0x28(%rsp), %rsi movq 0x48(%rsp), %rdi testq %rdi, %rdi jne 0x2afd2d movq %rsi, %rdi callq 0x5f3e0 jmp 0x2afd33 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/x86/padding_x86.cpp

ncnn::Padding_x86::forward_int8(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_int8_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_int8_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x118, %rsp # imm = 0x118 movq %rcx, %r12 movq %rdx, %rbx movq %rsi, %rax movq %rdi, %r13 movl 0x2c(%rsi), %esi movl 0x30(%rax), %edx movl 0x34(%rax), %ecx movl %ecx, 0x18(%rsp) movl 0x38(%rax), %ebp movq 0x10(%rax), %r9 movl 0x18(%rax), %edi movq %rax, 0x28(%rsp) movl 0x28(%rax), %r11d cmpl $0x8, %edi jne 0x2b09db leal -0x1(%r11), %eax cmpl $0x3, %eax ja 0x2b09db leaq 0x1489b2(%rip), %rcx # 0x3f8750 movslq (%rcx,%rax,4), %rax addq %rcx, %rax pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) movq %r13, 0x10(%rsp) jmpq *%rax movq (%r13), %rax movq -0x18(%rax), %r8 movl 0xd8(%r13,%r8), %r14d leal (%r14,%rsi,8), %eax addl 0xdc(%r13,%r8), %eax xorl %r15d, %r15d testb $0x7, %al sete %r15b movq %r9, %r10 shrq $0x3, %r10 leal (%r15,%r15,2), %ecx shlq %cl, %r10 testb $0x7, %r14b sete %cl andb %cl, %r15b cmpb $0x1, %r15b jne 0x2b09db cmpl $0x0, 0xe0(%r13,%r8) jne 0x2b09db sarl $0x3, %eax movq 0x8(%r12), %r8 pushq $0x8 popq %rbp movq %rbx, %rdi movl %eax, %esi movq %r10, %rdx movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b0ad5 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b0ad5 movq 0x10(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rcx cvttss2si 0xe4(%r8,%rcx), %rax movq %rax, %rdx shlq $0x8, %rdx movq %rax, %rsi shlq $0x10, %rsi orq %rdx, %rsi movq %rax, %rdx shlq $0x18, %rdx movq %rax, %rdi shlq $0x20, %rdi orq %rdx, %rdi orq %rsi, %rdi movq %rax, %rdx shlq $0x28, %rdx movq %rax, %rsi shlq $0x30, %rsi orq %rdx, %rsi movq %rax, %r9 shlq $0x38, %r9 orq %rsi, %r9 orq %rdi, %r9 orq %rax, %r9 movl 0xd8(%r8,%rcx), %eax movl 0xdc(%r8,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq %r9, (%rsp) movq $0x0, 0x20(%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b0cbe jmp 0x2b0ad5 movq (%r13), %rax movq -0x18(%rax), %r10 movl 0xe8(%r13,%r10), %eax leal (%rax,%rbp,8), %r15d addl 0xec(%r13,%r10), %r15d xorl %r14d, %r14d testb $0x7, %r15b sete %r14b movq %r9, %r8 shrq $0x3, %r8 leal (%r14,%r14,2), %ecx shlq %cl, %r8 testb $0x7, %al sete %al andb %al, %r14b cmpb $0x1, %r14b jne 0x2b09db movl 0xd8(%r13,%r10), %eax addl %esi, %eax addl 0xdc(%r13,%r10), %eax movl 0xd0(%r13,%r10), %ecx addl %edx, %ecx addl 0xd4(%r13,%r10), %ecx movq %rbp, 0x30(%rsp) movq 0x30(%rsp), %r14 leal (,%r14,8), %ebp cmpl %ebp, %r15d movq 0x30(%rsp), %rbp je 0x2aff73 cmpl $0x0, 0xe0(%r13,%r10) jne 0x2b09db sarl $0x3, %r15d movq 0x8(%r12), %rdx movq %rdx, (%rsp) pushq $0x8 popq %r9 movq %rbx, %rdi movl %eax, %esi movl %ecx, %edx movl %r15d, %ecx callq 0x628f2 cmpq $0x0, (%rbx) je 0x2b0ad5 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b0ad5 movq %r15, %r13 movq 0x10(%rsp), %r12 movq (%r12), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r13d, %r13d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r13d movl %edx, 0xa4(%rsp) movl %edx, %r10d xorl %ecx, %ecx movq %r13, 0x110(%rsp) cmpq %r13, %rcx je 0x2b0ad5 movslq 0x2c(%rbx), %rax movslq 0x30(%rbx), %r9 movl 0x34(%rbx), %esi movq (%rbx), %rdi movq 0x10(%rbx), %r11 movq 0x40(%rbx), %r8 movq %r8, %rbp movq %rcx, 0x48(%rsp) imulq %rcx, %rbp imulq %r11, %rbp addq %rdi, %rbp movl 0x18(%rbx), %ecx movq 0x20(%rbx), %rdx movq %rbp, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %r11, 0x60(%rsp) movl %ecx, 0x68(%rsp) movq %rdx, 0x70(%rsp) movl %eax, 0x7c(%rsp) movl %r9d, 0x80(%rsp) movl $0x1, 0x84(%rsp) movl %esi, 0x88(%rsp) imulq %rax, %r9 movq %r11, %rax imulq %r9, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r11 movq %rax, 0x90(%rsp) movl 0x28(%rbx), %ecx leal -0x1(%rcx), %edx movl %edx, 0x78(%rsp) cmpl $0x4, %ecx jne 0x2b00a8 movq %r9, 0x90(%rsp) movq %r9, %rax movq 0x98(%rsp), %rcx movq -0x18(%rcx), %r9 cvttss2si 0xe4(%r12,%r9), %rcx movl %r10d, %r10d movq %r10, 0x18(%rsp) movq %rcx, %rdx shlq $0x8, %rdx movq %rcx, %r10 shlq $0x10, %r10 orq %rdx, %r10 movq %rcx, %rdx shlq $0x18, %rdx movq %rcx, %r14 shlq $0x20, %r14 orq %rdx, %r14 orq %r10, %r14 movq %rcx, %rdx shlq $0x28, %rdx movq %rcx, %r15 shlq $0x30, %r15 orq %rdx, %r15 movq %rcx, %r10 shlq $0x38, %r10 orq %r15, %r10 orq %r14, %r10 orq %rcx, %r10 movl 0xa4(%rsp), %ecx movq 0x48(%rsp), %r15 addl %r15d, %ecx setns %dl cmpl 0x30(%rsp), %ecx setl %r14b testb %r14b, %dl je 0x2b031e movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rdi movslq 0x30(%rsi), %r8 movl 0x34(%rsi), %eax movq (%rsi), %r13 movq 0x10(%rsi), %r11 movq 0x40(%rsi), %rdx movl %ecx, %r12d movq %rdx, 0x38(%rsp) imulq %rdx, %r12 imulq %r11, %r12 addq %r13, %r12 movl 0x18(%rsi), %ecx movq 0x20(%rsi), %rdx movq %r12, 0xc0(%rsp) andq $0x0, 0xc8(%rsp) movq %r11, 0xd0(%rsp) movl %ecx, 0xd8(%rsp) movq %rdx, 0xe0(%rsp) movl %edi, 0xec(%rsp) movl %r8d, 0xf0(%rsp) movl $0x1, 0xf4(%rsp) movl %eax, 0xf8(%rsp) movq %r8, 0xb0(%rsp) movq %r8, %rcx movq %rdi, 0x40(%rsp) imulq %rdi, %rcx movq %r11, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx movq %r11, 0xa8(%rsp) divq %r11 movq %rax, 0x100(%rsp) movl 0x28(%rsi), %eax leal -0x1(%rax), %edx movl %edx, 0xe8(%rsp) cmpl $0x4, %eax jne 0x2b01ff movq %rcx, 0x100(%rsp) movq 0x10(%rsp), %r14 movl 0xe0(%r14,%r9), %eax testl %eax, %eax jne 0x2b025d movl 0xd0(%r14,%r9), %edx movl 0xd4(%r14,%r9), %ecx movl 0xd8(%r14,%r9), %r8d movl 0xdc(%r14,%r9), %r9d movq %r10, (%rsp) leaq 0xc0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b0cbe movq (%r14), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %r9 movl 0xe0(%r14,%r9), %eax cmpl $0x1, %eax jne 0x2b0463 movq 0x10(%rsp), %rax movl 0xd0(%rax,%r9), %r8d movl 0xd4(%rax,%r9), %ecx movl %ecx, 0xa0(%rsp) movl 0xd8(%rax,%r9), %edx xorl %r10d, %r10d testl %edx, %edx cmovlel %r10d, %edx movq 0x40(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r11d cmovgl %ecx, %r11d movl 0xdc(%rax,%r9), %esi testl %esi, %esi cmovlel %r10d, %esi testl %r8d, %r8d cmovlel %r10d, %r8d movq 0x38(%rsp), %r9 imulq 0xa8(%rsp), %r9 imulq 0x18(%rsp), %r9 addq %r13, %r9 movq %rbp, %rdi cmpl %r8d, %r10d je 0x2b034d movl %edx, %ecx subl $0x1, %ecx jb 0x2b02ec movq (%r12), %r14 movq %r14, (%rdi) addq $0x8, %rdi jmp 0x2b02da xorl %r14d, %r14d cmpl %r14d, %r11d je 0x2b0304 movq (%r9,%r14,8), %rcx movq %rcx, (%rdi) addq $0x8, %rdi incq %r14 jmp 0x2b02ef movl %esi, %ecx subl $0x1, %ecx jb 0x2b0319 movq -0x8(%r9,%r14,8), %r15 movq %r15, (%rdi) addq $0x8, %rdi jmp 0x2b0306 incl %r10d jmp 0x2b02d3 imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi imulq %r11, %r8 imulq %r15, %r8 addq %r8, %rdi xorl %eax, %eax movq %r15, %rcx cmpq %rax, %rsi je 0x2b067e movq %r10, (%rdi,%rax,8) incq %rax jmp 0x2b033b movslq 0xec(%rsp), %rax xorl %r11d, %r11d testl %eax, %eax movl $0x0, %r8d movq %rax, 0x108(%rsp) cmovgl %eax, %r8d movq 0xb0(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl %eax, %r14d movq %r12, %r9 cmpl %r14d, %r11d je 0x2b03df movl %edx, %ecx subl $0x1, %ecx jb 0x2b039b movq (%r9), %r10 movq %r10, (%rdi) addq $0x8, %rdi jmp 0x2b038a xorl %r15d, %r15d xorl %r10d, %r10d cmpl %r15d, %r8d je 0x2b03b7 movq (%r9,%r15,8), %rcx movq %rcx, (%rdi,%r15,8) addq $-0x8, %r10 incq %r15 jmp 0x2b03a1 movq %r9, %rcx subq %r10, %rcx subq %r10, %rdi movl %esi, %r10d subl $0x1, %r10d jb 0x2b03d7 movq -0x8(%r9,%r15,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b03c3 incl %r11d movq %rcx, %r9 jmp 0x2b0383 movq 0x108(%rsp), %rax shlq $0x3, %rax subq %rax, %r9 xorl %r10d, %r10d movl 0xa0(%rsp), %r14d testl %r14d, %r14d cmovlel %r10d, %r14d cmpl %r14d, %r10d je 0x2b044a movl %edx, %ecx subl $0x1, %ecx jb 0x2b0418 movq (%r9), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b0407 xorl %r11d, %r11d cmpl %r11d, %r8d je 0x2b0430 movq (%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi incq %r11 jmp 0x2b041b movl %esi, %ecx subl $0x1, %ecx jb 0x2b0445 movq -0x8(%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b0432 incl %r10d jmp 0x2b0400 movq 0x98(%rsp), %rax movq -0x18(%rax), %r9 movq 0x10(%rsp), %rax movl 0xe0(%rax,%r9), %eax cmpl $0x2, %eax jne 0x2b066c movq 0x10(%rsp), %rcx movl 0xd0(%rcx,%r9), %r14d movslq 0xd8(%rcx,%r9), %rax movq %rcx, %r8 movl %r14d, %r11d movq 0x40(%rsp), %rcx imull %ecx, %r11d movq %rcx, %rdx negq %rdx xorl %r15d, %r15d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi testl %ecx, %ecx movl $0x0, %edi cmovgl %ecx, %edi movl 0xdc(%r8,%r9), %r8d testl %r8d, %r8d cmovlel %r15d, %r8d testl %r14d, %r14d cmovlel %r15d, %r14d movq 0x38(%rsp), %r10 imulq 0xa8(%rsp), %r10 movslq %r11d, %rcx leaq (%r12,%rcx,8), %r11 imulq 0x18(%rsp), %r10 leaq (%r10,%rax,8), %r10 leaq (%r10,%rcx,8), %r10 movq 0x10(%rsp), %r12 movl 0xd4(%r12,%r9), %r9d movl %r9d, 0x38(%rsp) shlq $0x3, %rax addq %r10, %r13 movq 0x40(%rsp), %rcx leaq (,%rcx,8), %r10 negq %r10 negq %rsi negq %r8 cmpl %r14d, %r15d je 0x2b0575 xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b0533 movq (%r13,%rcx,8), %r12 movq %r12, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b051c xorl %ecx, %ecx movq %r11, %r12 cmpl %edi, %ecx je 0x2b0550 movq (%r12), %r9 addq $0x8, %r12 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b0538 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b0569 movq -0x10(%r12,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b0552 leaq (%r11,%rdx,8), %r11 incl %r15d addq %r10, %r13 jmp 0x2b0515 xorl %r14d, %r14d movq 0xb0(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r15d cmovgl %ecx, %r15d movl 0x38(%rsp), %r13d cmpl %r15d, %r14d je 0x2b05ea leaq (%r11,%rax), %rcx xorl %r12d, %r12d cmpq %r12, %rsi je 0x2b05b3 movq (%rcx,%r12,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %r12 jmp 0x2b059d xorl %ecx, %ecx cmpl %edi, %ecx je 0x2b05cc movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b05b5 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b05e5 movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b05ce incl %r14d jmp 0x2b0591 movq 0x40(%rsp), %rcx addl %ecx, %ecx movslq %ecx, %rcx shlq $0x3, %rcx movq %r11, %r14 subq %rcx, %r14 xorl %r15d, %r15d testl %r13d, %r13d cmovlel %r15d, %r13d subq %rcx, %rax addq %r11, %rax cmpl %r13d, %r15d je 0x2b066c xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b062b movq (%rax,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b0615 xorl %ecx, %ecx movq %r14, %r11 cmpl %edi, %ecx je 0x2b0647 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b0630 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b0660 movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b0649 leaq (%r14,%rdx,8), %r14 incl %r15d addq %r10, %rax jmp 0x2b060e movq 0x10(%rsp), %r12 movq 0x110(%rsp), %r13 movq 0x48(%rsp), %rcx incq %rcx movq 0x18(%rsp), %r10 incl %r10d jmp 0x2afffe movq (%r13), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xe0(%r13,%rax) jne 0x2b09db movl 0xe8(%r13,%rax), %ecx addl 0x18(%rsp), %ecx addl 0xec(%r13,%rax), %ecx addl 0xd0(%r13,%rax), %edx addl 0xd4(%r13,%rax), %edx addl 0xd8(%r13,%rax), %esi addl 0xdc(%r13,%rax), %esi movq 0x8(%r12), %rax movq %rax, 0x8(%rsp) movl $0x8, (%rsp) movq %rbx, %rdi movq %rcx, %r14 movl %ebp, %r8d callq 0x62a26 cmpq $0x0, (%rbx) je 0x2b0ad5 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b0ad5 xorl %ecx, %ecx testl %r14d, %r14d cmovlel %ecx, %r14d movq %r14, 0x48(%rsp) testl %ebp, %ebp movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %ebp movq %rbp, 0x30(%rsp) movabsq $0x100000001, %r12 # imm = 0x100000001 xorl %ebp, %ebp movq %rbx, 0xb8(%rsp) movq 0x10(%rsp), %r13 cmpq 0x30(%rsp), %rbp je 0x2b0ad5 movq 0x10(%rsp), %rcx movq (%rcx), %rax movq -0x18(%rax), %rax cvttss2si 0xe4(%rcx,%rax), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rsi shlq $0x20, %rsi orq %rcx, %rsi orq %rdx, %rsi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r14 shlq $0x38, %r14 orq %rdx, %r14 orq %rsi, %r14 orq %rax, %r14 xorl %r15d, %r15d cmpq 0x48(%rsp), %r15 je 0x2b0981 movslq 0x2c(%rbx), %rdi movslq 0x30(%rbx), %rdx movq (%rbx), %rax movq 0x10(%rbx), %rsi movq 0x40(%rbx), %r8 imulq %rbp, %r8 movq %r8, %r9 imulq %rsi, %r9 addq %rax, %r9 movl 0x18(%rbx), %r10d movq %rdx, %rcx imulq %rdi, %rcx movq %r15, %r11 imulq %rsi, %r11 imulq %rcx, %r11 addq %r9, %r11 movq 0x20(%rbx), %r9 movq %r11, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %rsi, 0x60(%rsp) movl %r10d, 0x68(%rsp) movq %r9, 0x70(%rsp) movl $0x2, 0x78(%rsp) movl %edi, 0x7c(%rsp) movl %edx, 0x80(%rsp) movq %r12, 0x84(%rsp) movq %rcx, 0x90(%rsp) movq (%r13), %r9 movq -0x18(%r9), %r9 movl 0xe8(%r13,%r9), %r11d movl %r15d, %r10d subl %r11d, %r10d setns %r11b cmpl 0x18(%rsp), %r10d movq %r12, %rbx setl %r12b testb %r12b, %r11b je 0x2b0948 movq 0x28(%rsp), %r8 movslq 0x2c(%r8), %rax movslq 0x30(%r8), %rcx movq 0x40(%r8), %rdx imulq %rbp, %rdx movq 0x10(%r8), %rsi imulq %rsi, %rdx addq (%r8), %rdx movl 0x18(%r8), %edi movq 0x20(%r8), %r8 movq %rcx, %r11 imulq %rax, %r11 movl %r10d, %r10d movq %r11, %r12 imulq %rsi, %r12 imulq %r10, %r12 addq %rdx, %r12 movq %r12, 0xc0(%rsp) andq $0x0, 0xc8(%rsp) movq %rsi, 0xd0(%rsp) movl %edi, 0xd8(%rsp) movq %r8, 0xe0(%rsp) movl $0x2, 0xe8(%rsp) movl %eax, 0xec(%rsp) movl %ecx, 0xf0(%rsp) movq %rbx, 0xf4(%rsp) movq %r11, 0x100(%rsp) movl 0xd0(%r13,%r9), %edx movl 0xd4(%r13,%r9), %ecx movl 0xd8(%r13,%r9), %r8d movl 0xdc(%r13,%r9), %r9d movq %r14, (%rsp) leaq 0xc0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b0cbe movq %rbx, %r12 movq 0xb8(%rsp), %rbx incq %r15 jmp 0x2b07b9 testl %ecx, %ecx movl $0x0, %r9d cmovlel %r9d, %ecx imulq %r15, %rdx imulq %rdi, %rdx addq %r8, %rdx imulq %rdx, %rsi addq %rsi, %rax xorl %edx, %edx movq %rbx, %r12 movq 0xb8(%rsp), %rbx cmpq %rdx, %rcx je 0x2b0940 movq %r14, (%rax,%rdx,8) incq %rdx jmp 0x2b0973 incq %rbp jmp 0x2b074f movq (%r13), %rax movq -0x18(%rax), %r10 movl 0xd0(%r13,%r10), %r14d leal (%r14,%rdx,8), %eax addl 0xd4(%r13,%r10), %eax xorl %r15d, %r15d testb $0x7, %al sete %r15b movq %r9, %r8 shrq $0x3, %r8 leal (%r15,%r15,2), %ecx shlq %cl, %r8 testb $0x7, %r14b sete %cl andb %cl, %r15b cmpb $0x1, %r15b jne 0x2b09db cmpl $0x0, 0xe0(%r13,%r10) je 0x2b0aec movq 0x28(%rsp), %r8 movq (%r8), %rax movq %rax, 0x50(%rsp) movq 0x8(%r8), %rax movq %rax, 0x58(%rsp) movq %r9, 0x60(%rsp) movl %edi, 0x68(%rsp) movq 0x20(%r8), %rcx movq %rcx, 0x70(%rsp) movl %r11d, 0x78(%rsp) movl %esi, 0x7c(%rsp) movl %edx, 0x80(%rsp) movl 0x18(%rsp), %ecx movl %ecx, 0x84(%rsp) movl %ebp, 0x88(%rsp) movq 0x40(%r8), %rcx movq %rcx, 0x90(%rsp) testq %rax, %rax je 0x2b0a39 lock incl (%rax) cmpl $0x1, %edi je 0x2b0a87 movups (%r12), %xmm0 movups 0x10(%r12), %xmm1 movups 0x20(%r12), %xmm2 movups 0x30(%r12), %xmm3 leaq 0xc0(%rsp), %rcx movaps %xmm3, 0x30(%rcx) movaps %xmm2, 0x20(%rcx) movaps %xmm1, 0x10(%rcx) movaps %xmm0, (%rcx) movq 0x10(%r12), %rax movq %rax, 0x8(%rcx) leaq 0x50(%rsp), %rsi pushq $0x1 popq %rdx movq 0x28(%rsp), %rdi callq 0x64e3b movq (%r13), %rax addq -0x18(%rax), %r13 leaq 0x50(%rsp), %rsi movq %r13, %rdi movq %rbx, %rdx movq %r12, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b0ad5 lock decl (%rax) jne 0x2b0ad5 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2b0acd movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b0ad5 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0x118, %rsp # imm = 0x118 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq addl 0xd8(%r13,%r10), %esi addl 0xdc(%r13,%r10), %esi sarl $0x3, %eax movq 0x8(%r12), %r9 pushq $0x8 popq %rbp movq %rbx, %rdi movl %eax, %edx movq %r8, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b0ad5 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b0ad5 movq 0x10(%rsp), %r9 movq (%r9), %rax movq -0x18(%rax), %rsi cvttss2si 0xe4(%r9,%rsi), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rdi shlq $0x20, %rdi orq %rcx, %rdi orq %rdx, %rdi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r11 shlq $0x38, %r11 orq %rdx, %r11 orq %rdi, %r11 orq %rax, %r11 movl 0xd0(%r9,%rsi), %eax movl 0xd4(%r9,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%r9,%rsi), %r8d movl 0xdc(%r9,%rsi), %r9d movq %r11, (%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b0cbe movq $0x0, 0x20(%rsp) jmp 0x2b0ad5 jmp 0x2b0c13 jmp 0x2b0bda movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b0c0b lock decl (%rax) jne 0x2b0c0b movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2b0c05 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b0c0b movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/x86/padding_x86.cpp

virtual thunk to ncnn::Padding_x86::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2aee70

/csukuangfj[P]ncnn/src/layer/x86/padding_x86.cpp

ncnn::padding_constant_pack8_int8_sse(ncnn::Mat const&, ncnn::Mat&, int, int, int, int, long)

static void padding_constant_pack8_int8_sse(const Mat& src, Mat& dst, int top, int bottom, int left, int right, int64_t _v) { const int64_t* ptr = src; int64_t* outptr = dst; // fill top for (int y = 0; y < top; y++) { for (int x = 0; x < dst.w; x++) { *outptr++ = _v; } } // fill center for (int y = 0; y < src.h; y++) { for (int x = 0; x < left; x++) { *outptr++ = _v; } for (int x = 0; x < src.w; x++) { *outptr++ = *ptr++; } for (int x = 0; x < right; x++) { *outptr++ = _v; } } // fill bottom for (int y = 0; y < bottom; y++) { for (int x = 0; x < dst.w; x++) { *outptr++ = _v; } } }

pushq %rbp pushq %r15 pushq %r14 pushq %r12 pushq %rbx movq 0x30(%rsp), %rax movq (%rdi), %r11 movl 0x2c(%rsi), %ebx xorl %ebp, %ebp testl %ebx, %ebx cmovlel %ebp, %ebx movq (%rsi), %r10 testl %edx, %edx cmovlel %ebp, %edx cmpl %edx, %ebp je 0x2b0cfa movl %ebx, %r14d subl $0x1, %r14d jb 0x2b0cf6 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b0ce7 incl %ebp jmp 0x2b0ce0 xorl %edx, %edx testl %r8d, %r8d cmovlel %edx, %r8d movl 0x2c(%rdi), %ebx testl %ebx, %ebx cmovlel %edx, %ebx testl %r9d, %r9d cmovlel %edx, %r9d movl 0x30(%rdi), %edi testl %edi, %edi cmovlel %edx, %edi cmpl %edi, %edx je 0x2b0d66 movl %r8d, %ebp subl $0x1, %ebp jb 0x2b0d2f movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b0d21 xorl %r15d, %r15d xorl %r14d, %r14d cmpl %r15d, %ebx je 0x2b0d4b movq (%r11,%r15,8), %r12 movq %r12, (%r10,%r15,8) addq $-0x8, %r14 incq %r15 jmp 0x2b0d35 subq %r14, %r11 subq %r14, %r10 movl %r9d, %ebp subl $0x1, %ebp jb 0x2b0d62 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b0d54 incl %edx jmp 0x2b0d1a movl 0x2c(%rsi), %edx xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %ecx, %ecx cmovlel %esi, %ecx movl %edx, %edi cmpl %ecx, %esi je 0x2b0d8d subl $0x1, %edi jb 0x2b0d89 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b0d7b incl %esi jmp 0x2b0d75 popq %rbx popq %r12 popq %r14 popq %r15 popq %rbp retq nop

/csukuangfj[P]ncnn/src/layer/x86/padding_pack8_int8.h

ncnn::Padding_x86_avx512::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86_avx512::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x118, %rsp # imm = 0x118 movq %rcx, %rbp movq %rdx, 0x8(%rsp) movq %rsi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rax leaq (%rdi,%rax), %r11 movq %rdi, 0x10(%rsp) movl 0xd0(%rdi,%rax), %r10d testl %r10d, %r10d jne 0x2b0e4f cmpl $0x0, 0xd4(%r11) jne 0x2b0e4f cmpl $0x0, 0xd8(%r11) jne 0x2b0e4f cmpl $0x0, 0xdc(%r11) jne 0x2b0e4f cmpl $0x0, 0xe8(%r11) jne 0x2b0e4f cmpl $0x0, 0xec(%r11) je 0x2b351e movl 0x18(%r14), %r15d movq 0x10(%r14), %r9 testl %r15d, %r15d je 0x2b0e93 leal (,%r9,8), %eax cltd idivl %r15d cmpl $0x8, %eax jne 0x2b0ea5 movq 0x10(%rsp), %rdi movq %r14, %rsi movq 0x8(%rsp), %rdx movq %rbp, %rcx addq $0x118, %rsp # imm = 0x118 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x2b362c vmovdqu 0x2c(%r14), %xmm0 movq %r14, %rdi movl 0x28(%r14), %r13d jmp 0x2b1157 vmovdqu 0x2c(%r14), %xmm0 movq %r14, (%rsp) movl 0x28(%r14), %ecx vpextrd $0x2, %xmm0, %r12d pushq $0x4 popq %r13 vpextrd $0x1, %xmm0, %eax vmovd %xmm0, %esi vpextrd $0x3, %xmm0, %edx movq %rdx, 0x18(%rsp) pushq $-0x64 popq %rdx movq %rdx, 0x20(%rsp) cmpl $0x4, %r15d je 0x2b104a cmpl $0x8, %r15d je 0x2b0f4a cmpl $0x10, %r15d jne 0x2b1150 leal -0x1(%rcx), %edx cmpl $0x3, %edx ja 0x2b1150 leaq 0x147891(%rip), %rcx # 0x3f8798 movslq (%rcx,%rdx,4), %rdx addq %rcx, %rdx jmpq *%rdx shll $0x4, %esi movl 0xd8(%r11), %eax addl %eax, %esi addl 0xdc(%r11), %esi testb $0xf, %sil je 0x2b2bce testb $0x7, %sil je 0x2b3371 xorl %ecx, %ecx testb $0x3, %sil sete %cl leal (%rcx,%rcx,2), %ecx incl %ecx jmp 0x2b3374 leal -0x1(%rcx), %edx cmpl $0x3, %edx ja 0x2b1150 leaq 0x14782b(%rip), %rcx # 0x3f8788 movslq (%rcx,%rdx,4), %rdx addq %rcx, %rdx jmpq *%rdx movl 0xd8(%r11), %edx leal (%rdx,%rsi,8), %eax addl 0xdc(%r11), %eax testb $0x3, %al sete %cl movq %r9, %r10 shrq $0x3, %r10 addb %cl, %cl testb $0x7, %al movzbl %cl, %esi pushq $0x3 popq %rcx cmovnel %esi, %ecx shlq %cl, %r10 orl %eax, %edx pushq $0x1 popq %r13 testb $0x7, %dl jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 pushq $0x8 popq %r12 cltd idivl %r12d movq 0x8(%rbp), %r8 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %esi movq %r10, %rdx movl %r12d, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vpbroadcastd 0xe4(%rdx,%rcx), %ymm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %r12d movl %eax, %r8d movl %ecx, %eax cltd idivl %r12d movq $0x0, 0x20(%rsp) movq (%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b4597 jmp 0x2b1239 leal -0x1(%rcx), %edx cmpl $0x3, %edx ja 0x2b1150 leaq 0x14771b(%rip), %rcx # 0x3f8778 movslq (%rcx,%rdx,4), %rdx addq %rcx, %rdx jmpq *%rdx movl 0xd8(%r11), %r8d leal (%r8,%rsi,4), %esi addl 0xdc(%r11), %esi testb $0x3, %sil sete %cl movq %r9, %rdx shrq $0x2, %rdx addb %cl, %cl movl %esi, %eax andl $0x7, %eax movzbl %cl, %r10d pushq $0x3 popq %rcx cmovnel %r10d, %ecx shlq %cl, %rdx pushq $0x1 popq %r13 testb $0x3, %r8b jne 0x2b1153 cmpl $0x4, %eax jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 sarl $0x2, %esi movq 0x8(%rbp), %r8 pushq $0x4 popq %rbp movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vpbroadcastd 0xe4(%rdx,%rcx), %xmm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq $0x0, 0x20(%rsp) movq (%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b4637 jmp 0x2b1239 movl %ecx, %r13d movq (%rsp), %rdi movq 0x8(%rdi), %rax vmovups (%rdi), %xmm1 vmovaps %xmm1, 0x80(%rsp) movq %r9, 0x90(%rsp) movl %r15d, 0x98(%rsp) movq 0x20(%rdi), %rcx movq %rcx, 0xa0(%rsp) movl %r13d, 0xa8(%rsp) vmovdqu %xmm0, 0xac(%rsp) movq 0x40(%rdi), %rcx movq %rcx, 0xc0(%rsp) testq %rax, %rax je 0x2b11a9 lock incl (%rax) cmpl $0x1, %r15d je 0x2b11dc vmovdqu64 (%rbp), %zmm0 leaq 0x30(%rsp), %rcx vmovdqu64 %zmm0, (%rcx) movq 0x10(%rbp), %rax movq %rax, 0x8(%rcx) leaq 0x80(%rsp), %rsi pushq $0x1 popq %rdx vzeroupper callq 0x64e3b movq 0x10(%rsp), %rdi movq (%rdi), %rax addq -0x18(%rax), %rdi leaq 0x80(%rsp), %rsi movq 0x8(%rsp), %rdx movq %rbp, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x88(%rsp), %rax testq %rax, %rax je 0x2b1239 lock decl (%rax) jne 0x2b1239 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b1231 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b1239 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0x118, %rsp # imm = 0x118 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq movl 0xe8(%r11), %r14d movq 0x18(%rsp), %rdi leal (%r14,%rdi,8), %edx addl 0xec(%r11), %edx testb $0x3, %dl sete %cl movq %r9, %r8 shrq $0x3, %r8 addb %cl, %cl testb $0x7, %dl movzbl %cl, %ecx pushq $0x3 popq %r13 cmovel %r13d, %ecx shlq %cl, %r8 orl %edx, %r14d testb $0x7, %r14b jne 0x2b1153 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi addl %eax, %r10d addl 0xd4(%r11), %r10d leal (,%rdi,8), %eax cmpl %eax, %edx movq (%rsp), %rdi je 0x2b12cc cmpl $0x0, 0xe0(%r11) jne 0x2b1157 pushq $0x8 popq %r9 movl %edx, %eax cltd idivl %r9d subq $0x8, %rsp movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movq %rax, %r14 movl %eax, %ecx pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xc8(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xc8(%rsp), %r9 je 0x2b1239 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b1419 movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b1443 movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x5, %r8 vmovdqu (%rdx,%r8), %ymm0 jmp 0x2b144d vpbroadcastd 0xe4(%r12,%rdi), %ymm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b160a movq (%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b1504 movq %rsi, 0x70(%rsp) movq 0x10(%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b1560 movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b4597 movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b1634 movq 0x10(%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b163e vmovdqa (%rbp), %ymm0 movl %eax, %edi subl $0x1, %edi jb 0x2b15cd vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b15bc xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b15ef vmovdqa (%r12,%rdi), %ymm0 vmovdqa %ymm0, (%r13,%rdi) incl %r10d addq $0x20, %rdi jmp 0x2b15d2 addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b1605 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b15f4 incl %r8d jmp 0x2b15ac imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b19db vmovdqu %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b161f movq 0x10(%rsp), %r12 jmp 0x2b1722 xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b169b vmovdqa (%rbp), %ymm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b1660 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b164e xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b1682 vmovdqa (%rbp), %ymm0 vmovdqa %ymm0, (%r13) addq $0x20, %rbp addq $0x20, %r13 incl %r8d jmp 0x2b1663 movl %edx, %r8d subl $0x1, %r8d jb 0x2b1697 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b1685 incl %edi jmp 0x2b1640 movl 0x5c(%rsp), %edi shll $0x3, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b1711 vmovdqa (%rbp), %ymm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b16d1 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b16bf xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b16f5 vmovdqa (%rbp,%r8), %ymm0 vmovdqa %ymm0, (%r13,%r8) incl %r9d addq $0x20, %r8 jmp 0x2b16d7 addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b170d vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b16fb incl %edi jmp 0x2b16b3 movq -0x18(%rcx), %rdi movq 0x10(%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x40 popq %r13 jne 0x2b1943 movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x3, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x5, %rax shlq $0x5, %rdx shlq $0x5, %rsi cmpl %r9d, %r10d je 0x2b1824 leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b17c5 vmovdqa (%r11,%r14), %ymm0 vmovdqa %ymm0, (%r8,%r15) addq $-0x20, %r14 addq $0x20, %r15 jmp 0x2b17a7 subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b17e9 vmovdqa (%r11), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r11 addq $0x20, %r8 incl %ebp jmp 0x2b17cd movq %r13, %r14 movq 0x10(%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x40, %r15 je 0x2b1810 vmovdqa (%r11,%r14), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r14 jmp 0x2b17f1 movslq 0x5c(%rsp), %r11 shlq $0x5, %r11 subq %r11, %rcx incl %r10d jmp 0x2b1794 xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b18a0 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b1856 vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8,%r14) addq $-0x20, %r11 addq $0x20, %r14 jmp 0x2b1838 subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b1879 vmovdqa (%rcx), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %rcx addq $0x20, %r8 incl %r10d jmp 0x2b185c movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x40, %r11 je 0x2b189b vmovdqa (%rcx,%r10), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r10 jmp 0x2b187c incl %r9d jmp 0x2b1827 movslq 0x5c(%rsp), %r9 shlq $0x6, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b1943 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b18e6 vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8,%r14) addq $-0x20, %r11 addq $0x20, %r14 jmp 0x2b18c8 subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b190d vmovdqa (%r10), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r10 addq $0x20, %r8 incl %r11d jmp 0x2b18ef movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x40, %r14 je 0x2b192f vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r11 jmp 0x2b1910 movslq 0x5c(%rsp), %r10 shlq $0x5, %r10 subq %r10, %rcx incl %r9d jmp 0x2b18b5 movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b1977 lock decl (%rax) jne 0x2b1977 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b196c movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x2b1977 movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b19db lock decl (%rax) jne 0x2b19db movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b19c3 movq (%rdi), %rax vzeroupper callq *0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b19db movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b1359 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 movl 0xe8(%r11), %r13d addl %r12d, %r13d addl 0xec(%r11), %r13d addl %eax, %r10d addl 0xd4(%r11), %r10d addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x8 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %r11 # imm = 0x100000001 xorl %r14d, %r14d movq 0x10(%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b1239 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b1aca movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x5, %rcx vmovdqu (%rax,%rcx), %ymm0 jmp 0x2b1ad4 vpbroadcastd 0xe4(%rbp,%rax), %ymm0 xorl %r15d, %r15d vmovdqu %ymm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b1c86 movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %r11, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b1c6d movq (%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %r13, %rbx movq %r11, %r13 movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq 0x10(%rsp), %rbp movq %r13, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovups 0xd0(%rsp), %ymm0 callq 0x2b4597 vmovdqu 0xd0(%rsp), %ymm0 movq %r13, %r11 movq %rbx, %r13 incq %r15 jmp 0x2b1ae0 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b1c65 vmovdqu %ymm0, (%rcx) addq $0x20, %rcx jmp 0x2b1c77 incq %r14 jmp 0x2b1a97 movl 0xe8(%r11), %r14d movq 0x18(%rsp), %rdi leal (%r14,%rdi,4), %edx addl 0xec(%r11), %edx testb $0x3, %dl sete %cl movq %r9, %r8 shrq $0x2, %r8 addb %cl, %cl movq %rdx, %r12 andl $0x7, %edx movzbl %cl, %ecx pushq $0x3 popq %r13 cmovel %r13d, %ecx shlq %cl, %r8 testb $0x3, %r14b jne 0x2b1153 cmpl $0x4, %edx jne 0x2b1153 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi addl %eax, %r10d addl 0xd4(%r11), %r10d leal (,%rdi,4), %eax cmpl %eax, %r12d movq (%rsp), %rdi je 0x2b1d11 cmpl $0x0, 0xe0(%r11) jne 0x2b1157 movq %r12, %rcx sarl $0x2, %ecx subq $0x8, %rsp pushq $0x4 popq %r9 movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movq %rcx, %r14 pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x4 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xc8(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xc8(%rsp), %r9 je 0x2b1239 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b1e5c movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b1e86 movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x4, %r8 vmovdqu (%rdx,%r8), %xmm0 jmp 0x2b1e90 vpbroadcastd 0xe4(%r12,%rdi), %xmm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b204d movq (%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b1f47 movq %rsi, 0x70(%rsp) movq 0x10(%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b1fa3 movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b4637 movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b2077 movq 0x10(%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b2081 vmovdqa (%rbp), %xmm0 movl %eax, %edi subl $0x1, %edi jb 0x2b2010 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b1fff xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b2032 vmovdqa (%r12,%rdi), %xmm0 vmovdqa %xmm0, (%r13,%rdi) incl %r10d addq $0x10, %rdi jmp 0x2b2015 addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b2048 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b2037 incl %r8d jmp 0x2b1fef imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b2412 vmovdqu %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b2062 movq 0x10(%rsp), %r12 jmp 0x2b2165 xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b20de vmovdqa (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b20a3 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b2091 xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b20c5 vmovdqa (%rbp), %xmm0 vmovdqa %xmm0, (%r13) addq $0x10, %rbp addq $0x10, %r13 incl %r8d jmp 0x2b20a6 movl %edx, %r8d subl $0x1, %r8d jb 0x2b20da vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b20c8 incl %edi jmp 0x2b2083 movl 0x5c(%rsp), %edi shll $0x2, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b2154 vmovdqa (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b2114 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b2102 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b2138 vmovdqa (%rbp,%r8), %xmm0 vmovdqa %xmm0, (%r13,%r8) incl %r9d addq $0x10, %r8 jmp 0x2b211a addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b2150 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b213e incl %edi jmp 0x2b20f6 movq -0x18(%rcx), %rdi movq 0x10(%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x20 popq %r13 jne 0x2b2386 movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x2, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x4, %rax shlq $0x4, %rdx shlq $0x4, %rsi cmpl %r9d, %r10d je 0x2b2267 leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b2208 vmovdqa (%r11,%r14), %xmm0 vmovdqa %xmm0, (%r8,%r15) addq $-0x10, %r14 addq $0x10, %r15 jmp 0x2b21ea subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b222c vmovdqa (%r11), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r11 addq $0x10, %r8 incl %ebp jmp 0x2b2210 movq %r13, %r14 movq 0x10(%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x20, %r15 je 0x2b2253 vmovdqa (%r11,%r14), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r14 jmp 0x2b2234 movslq 0x5c(%rsp), %r11 shlq $0x4, %r11 subq %r11, %rcx incl %r10d jmp 0x2b21d7 xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b22e3 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b2299 vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2b227b subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b22bc vmovdqa (%rcx), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %rcx addq $0x10, %r8 incl %r10d jmp 0x2b229f movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x20, %r11 je 0x2b22de vmovdqa (%rcx,%r10), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r10 jmp 0x2b22bf incl %r9d jmp 0x2b226a movslq 0x5c(%rsp), %r9 shlq $0x5, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b2386 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b2329 vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2b230b subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b2350 vmovdqa (%r10), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r10 addq $0x10, %r8 incl %r11d jmp 0x2b2332 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x20, %r14 je 0x2b2372 vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r11 jmp 0x2b2353 movslq 0x5c(%rsp), %r10 shlq $0x4, %r10 subq %r10, %rcx incl %r9d jmp 0x2b22f8 movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b23b4 lock decl (%rax) jne 0x2b23b4 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b23ac movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b23b4 movq %rsi, %rdi callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b2412 lock decl (%rax) jne 0x2b2412 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b23fd movq (%rdi), %rax callq *0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b2412 movq %rsi, %rdi callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b1d9c movq 0x18(%rsp), %rdx shll $0x4, %edx movl 0xe8(%r11), %r14d leal (%r14,%rdx), %ecx addl 0xec(%r11), %ecx testb $0xf, %cl je 0x2b2bca testb $0x7, %cl je 0x2b2bdc xorl %r8d, %r8d testb $0x3, %cl sete %r8b leal (%r8,%r8,2), %r8d incl %r8d jmp 0x2b2be0 leal (%r10,%rax,8), %eax addl 0xd4(%r11), %eax testb $0x3, %al sete %cl movq %r9, %r8 shrq $0x3, %r8 addb %cl, %cl testb $0x7, %al movzbl %cl, %edx pushq $0x3 popq %rcx cmovnel %edx, %ecx shlq %cl, %r8 orl %eax, %r10d pushq $0x2 popq %r13 testb $0x7, %r10b jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi pushq $0x8 popq %r12 cltd idivl %r12d movq 0x8(%rbp), %r9 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %edx movq %r8, %rcx movl %r12d, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vpbroadcastd 0xe4(%rdi,%rsi), %ymm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %r12d movl %eax, %r10d movl %ecx, %eax cltd idivl %r12d movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq (%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b4597 jmp 0x2b3510 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 movl 0xe8(%r11), %r13d addl %r12d, %r13d addl 0xec(%r11), %r13d addl %eax, %r10d addl 0xd4(%r11), %r10d addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x4 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %r11 # imm = 0x100000001 xorl %r14d, %r14d movq 0x10(%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b1239 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b2636 movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x4, %rcx vmovdqu (%rax,%rcx), %xmm0 jmp 0x2b2640 vpbroadcastd 0xe4(%rbp,%rax), %xmm0 xorl %r15d, %r15d vmovdqa %xmm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b27f8 movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %r11, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b27df movq (%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %r13, %rbx movl %r12d, %r13d movq %r11, %r12 movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq 0x10(%rsp), %rbp movq %r12, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovaps 0xd0(%rsp), %xmm0 callq 0x2b4637 vmovdqa 0xd0(%rsp), %xmm0 movq %r12, %r11 movl %r13d, %r12d movq %rbx, %r13 incq %r15 jmp 0x2b264c testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b27d7 vmovdqu %xmm0, (%rcx) addq $0x10, %rcx jmp 0x2b27e9 incq %r14 jmp 0x2b2603 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 movl 0xe8(%r11), %r13d addl %r12d, %r13d addl 0xec(%r11), %r13d addl %eax, %r10d addl 0xd4(%r11), %r10d addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x10 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %rbx # imm = 0x100000001 xorl %r14d, %r14d movq 0x10(%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b1239 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b28e6 movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x6, %rcx vmovdqu64 (%rax,%rcx), %zmm0 jmp 0x2b28ee vpbroadcastd 0xe4(%rbp,%rax), %zmm0 xorl %r15d, %r15d vmovdqu64 %zmm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b2a9c movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %rbx, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b2a81 movq (%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq 0x10(%rsp), %rbp movq %rbx, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovups 0xd0(%rsp), %zmm0 callq 0x2b44ec vmovdqu64 0xd0(%rsp), %zmm0 incq %r15 jmp 0x2b28fc testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b2a79 vmovdqu64 %zmm0, (%rcx) addq $0x40, %rcx jmp 0x2b2a8b incq %r14 jmp 0x2b28b1 leal (%r10,%rax,4), %edx addl 0xd4(%r11), %edx testb $0x3, %dl sete %cl movq %r9, %rax shrq $0x2, %rax addb %cl, %cl movl %edx, %r8d andl $0x7, %r8d movzbl %cl, %r14d pushq $0x3 popq %rcx cmovnel %r14d, %ecx shlq %cl, %rax pushq $0x2 popq %r13 testb $0x3, %r10b jne 0x2b1153 cmpl $0x4, %r8d jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi sarl $0x2, %edx movq 0x8(%rbp), %r9 pushq $0x4 popq %rbp movq 0x8(%rsp), %rbx movq %rbx, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vpbroadcastd 0xe4(%rdi,%rsi), %xmm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq (%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b4637 jmp 0x2b3510 shll $0x4, %eax addl %r10d, %eax addl 0xd4(%r11), %eax testb $0xf, %al je 0x2b2bd5 testb $0x7, %al je 0x2b343b xorl %ecx, %ecx testb $0x3, %al sete %cl leal (%rcx,%rcx,2), %ecx incl %ecx jmp 0x2b343e pushq $0x10 jmp 0x2b2bde pushq $0x10 jmp 0x2b3373 pushq $0x10 jmp 0x2b343d pushq $0x8 popq %r8 pushq $0x3 popq %r13 testb $0xf, %r14b jne 0x2b1153 cmpl $0x10, %r8d jne 0x2b1153 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi addl %eax, %r10d addl 0xd4(%r11), %r10d cmpl %edx, %ecx movq (%rsp), %rdi je 0x2b2c26 cmpl $0x0, 0xe0(%r11) jne 0x2b1157 shrq $0x4, %r9 movl %r8d, %r8d imulq %r9, %r8 pushq $0x10 popq %r9 movl %ecx, %eax cltd idivl %r9d subq $0x8, %rsp movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movq %rax, %r14 movl %eax, %ecx pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x10 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xc8(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xc8(%rsp), %r9 je 0x2b1239 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b2d7e movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b2da9 movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x6, %r8 vmovdqu64 (%rdx,%r8), %zmm0 jmp 0x2b2db1 vpbroadcastd 0xe4(%r12,%rdi), %zmm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b2f74 movq (%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b2e68 movq %rsi, 0x70(%rsp) movq 0x10(%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b2ec4 movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b44ec movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b2f9f movq 0x10(%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b2fa9 vmovdqa64 (%rbp), %zmm0 movl %eax, %edi subl $0x1, %edi jb 0x2b2f34 vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2f22 xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b2f58 vmovdqa64 (%r12,%rdi), %zmm0 vmovdqa64 %zmm0, (%r13,%rdi) incl %r10d addq $0x40, %rdi jmp 0x2b2f39 addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b2f6f vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2f5d incl %r8d jmp 0x2b2f10 imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b3366 vmovdqu64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2f89 movq 0x10(%rsp), %r12 jmp 0x2b309a xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b300d vmovdqa64 (%rbp), %zmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b2fce vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2fbb xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b2ff3 vmovdqa64 (%rbp), %zmm0 vmovdqa64 %zmm0, (%r13) addq $0x40, %rbp addq $0x40, %r13 incl %r8d jmp 0x2b2fd1 movl %edx, %r8d subl $0x1, %r8d jb 0x2b3009 vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b2ff6 incl %edi jmp 0x2b2fab movl 0x5c(%rsp), %edi shll $0x4, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b3089 vmovdqa64 (%rbp), %zmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b3046 vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b3033 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b306c vmovdqa64 (%rbp,%r8), %zmm0 vmovdqa64 %zmm0, (%r13,%r8) incl %r9d addq $0x40, %r8 jmp 0x2b304c addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b3085 vmovdqa64 %zmm0, (%r13) addq $0x40, %r13 jmp 0x2b3072 incl %edi jmp 0x2b3025 movq -0x18(%rcx), %rdi movq 0x10(%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x80 popq %r13 jne 0x2b32ce movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x4, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x6, %rax shlq $0x6, %rdx shlq $0x6, %rsi cmpl %r9d, %r10d je 0x2b31a2 leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b313f vmovdqa64 (%r11,%r14), %zmm0 vmovdqa64 %zmm0, (%r8,%r15) addq $-0x40, %r14 addq $0x40, %r15 jmp 0x2b311f subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b3165 vmovdqa64 (%r11), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r11 addq $0x40, %r8 incl %ebp jmp 0x2b3147 movq %r13, %r14 movq 0x10(%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x80, %r15 je 0x2b318e vmovdqa64 (%r11,%r14), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r8 addq $-0x40, %r14 jmp 0x2b316d movslq 0x5c(%rsp), %r11 shlq $0x6, %r11 subq %r11, %rcx incl %r10d jmp 0x2b310c xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b3225 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b31d6 vmovdqa64 (%r10,%r11), %zmm0 vmovdqa64 %zmm0, (%r8,%r14) addq $-0x40, %r11 addq $0x40, %r14 jmp 0x2b31b6 subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b31fc vmovdqa64 (%rcx), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %rcx addq $0x40, %r8 incl %r10d jmp 0x2b31dc movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x80, %r11 je 0x2b3220 vmovdqa64 (%rcx,%r10), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r8 addq $-0x40, %r10 jmp 0x2b31ff incl %r9d jmp 0x2b31a5 movslq 0x5c(%rsp), %r9 shlq $0x7, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b32ce leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b326d vmovdqa64 (%r10,%r11), %zmm0 vmovdqa64 %zmm0, (%r8,%r14) addq $-0x40, %r11 addq $0x40, %r14 jmp 0x2b324d subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b3296 vmovdqa64 (%r10), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r10 addq $0x40, %r8 incl %r11d jmp 0x2b3276 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x80, %r14 je 0x2b32ba vmovdqa64 (%r10,%r11), %zmm0 vmovdqa64 %zmm0, (%r8) addq $0x40, %r8 addq $-0x40, %r11 jmp 0x2b3299 movslq 0x5c(%rsp), %r10 shlq $0x6, %r10 subq %r10, %rcx incl %r9d jmp 0x2b323a movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b3302 lock decl (%rax) jne 0x2b3302 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b32f7 movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x2b3302 movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b3366 lock decl (%rax) jne 0x2b3366 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b334e movq (%rdi), %rax vzeroupper callq *0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b3366 movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b2cbe pushq $0x8 popq %rcx pushq $0x1 popq %r13 testb $0xf, %al jne 0x2b1153 cmpl $0x10, %ecx jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 shrq $0x4, %r9 movl %ecx, %ecx imulq %r9, %rcx pushq $0x10 popq %r12 movl %esi, %eax cltd idivl %r12d movq 0x8(%rbp), %r8 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %esi movq %rcx, %rdx movl %r12d, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vpbroadcastd 0xe4(%rdx,%rcx), %zmm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %r12d movl %eax, %r8d movl %ecx, %eax cltd idivl %r12d movq $0x0, 0x20(%rsp) movq (%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b44ec jmp 0x2b1239 pushq $0x8 popq %rcx pushq $0x2 popq %r13 testb $0xf, %r10b jne 0x2b1153 cmpl $0x10, %ecx jne 0x2b1153 cmpl $0x0, 0xe0(%r11) movq (%rsp), %rdi jne 0x2b1157 addl 0xd8(%r11), %esi addl 0xdc(%r11), %esi shrq $0x4, %r9 movl %ecx, %ecx imulq %r9, %rcx pushq $0x10 popq %r12 cltd idivl %r12d movq 0x8(%rbp), %r9 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %edx movl %r12d, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b1239 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b1239 movq 0x10(%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vpbroadcastd 0xe4(%rdi,%rsi), %zmm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %r12d movl %eax, %r10d movl %ecx, %eax cltd idivl %r12d movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq (%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b44ec movq $0x0, 0x20(%rsp) jmp 0x2b1239 movq $0x0, 0x20(%rsp) cmpq %r14, 0x8(%rsp) je 0x2b1239 movq 0x8(%r14), %rax testq %rax, %rax je 0x2b353e lock incl (%rax) movq 0x8(%rsp), %rax movq 0x8(%rax), %rax testq %rax, %rax je 0x2b3572 lock decl (%rax) jne 0x2b3572 movq 0x8(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x2b356a movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b3572 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx andq $0x0, 0x40(%rcx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, (%rcx) vmovups %xmm0, 0xc(%rcx) andl $0x0, 0x38(%rcx) vmovups %xmm0, 0x28(%rcx) vmovups (%r14), %xmm0 vmovups %xmm0, (%rcx) movq 0x10(%r14), %rax movq %rax, 0x10(%rcx) movl 0x18(%r14), %eax movl %eax, 0x18(%rcx) movq 0x20(%r14), %rax movq %rax, 0x20(%rcx) vmovdqu 0x28(%r14), %xmm0 vmovdqu %xmm0, 0x28(%rcx) movl 0x38(%r14), %eax movl %eax, 0x38(%rcx) movq 0x40(%r14), %rax movq %rax, 0x40(%rcx) jmp 0x2b1239 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b3623 jmp 0x2b35e1 movq %rax, %rbx movq 0x88(%rsp), %rax testq %rax, %rax je 0x2b361b lock decl (%rax) jne 0x2b361b movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi jne 0x2b3615 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b361b movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_avx512.cpp

ncnn::Padding_x86_avx512::forward_int8(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86_avx512::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_int8_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_int8_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x128, %rsp # imm = 0x128 movq %rcx, %r12 movq %rdx, %rbx movq %rdi, %r14 movl 0x38(%rsi), %r15d movq 0x10(%rsi), %r9 movl 0x18(%rsi), %edi movq %rsi, 0x28(%rsp) vmovdqu 0x28(%rsi), %xmm0 cmpl $0x8, %edi jne 0x2b42dc vmovd %xmm0, %eax decl %eax cmpl $0x3, %eax ja 0x2b42dc leaq 0x14512e(%rip), %rcx # 0x3f87a8 movslq (%rcx,%rax,4), %rax vpextrd $0x1, %xmm0, %esi addq %rcx, %rax vpextrd $0x2, %xmm0, %edx pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) movq %r14, 0x18(%rsp) jmpq *%rax movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %r8d leal (%r8,%rsi,8), %esi addl 0xdc(%r14,%rax), %esi xorl %r10d, %r10d testb $0x7, %sil sete %r10b movq %r9, %rdx shrq $0x3, %rdx leal (%r10,%r10,2), %ecx shlq %cl, %rdx testb $0x7, %r8b sete %cl andb %cl, %r10b cmpb $0x1, %r10b jne 0x2b42dc cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b42dc sarl $0x3, %esi movq 0x8(%r12), %r8 pushq $0x8 popq %rbp movq %rbx, %rdi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b43a8 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b43a8 movq 0x18(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rcx vcvttss2si 0xe4(%r8,%rcx), %rax movq %rax, %rdx shlq $0x8, %rdx movq %rax, %rsi shlq $0x10, %rsi orq %rdx, %rsi movq %rax, %rdx shlq $0x18, %rdx movq %rax, %rdi shlq $0x20, %rdi orq %rdx, %rdi orq %rsi, %rdi movq %rax, %rdx shlq $0x28, %rdx movq %rax, %rsi shlq $0x30, %rsi orq %rdx, %rsi movq %rax, %r9 shlq $0x38, %r9 orq %rsi, %r9 orq %rdi, %r9 orq %rax, %r9 movl 0xd8(%r8,%rcx), %eax movl 0xdc(%r8,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq %r9, (%rsp) movq $0x0, 0x20(%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b46e0 jmp 0x2b43a8 movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %r10d leal (%r10,%r15,8), %r13d addl 0xec(%r14,%rax), %r13d xorl %r11d, %r11d testb $0x7, %r13b sete %r11b movq %r9, %r8 shrq $0x3, %r8 leal (%r11,%r11,2), %ecx shlq %cl, %r8 testb $0x7, %r10b sete %cl andb %cl, %r11b cmpb $0x1, %r11b jne 0x2b42dc addl 0xd8(%r14,%rax), %esi addl 0xdc(%r14,%rax), %esi addl 0xd0(%r14,%rax), %edx addl 0xd4(%r14,%rax), %edx leal (,%r15,8), %ecx cmpl %ecx, %r13d je 0x2b3845 cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b42dc sarl $0x3, %r13d movq 0x8(%r12), %rax movq %rax, (%rsp) pushq $0x8 popq %r9 movq %rbx, %rdi movl %r13d, %ecx callq 0x628f2 cmpq $0x0, (%rbx) je 0x2b43a8 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b43a8 movq 0x18(%rsp), %rcx movq (%rcx), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %rax movl 0xe8(%rcx,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %r12d xorl %ecx, %ecx testl %r13d, %r13d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r13d movl %r12d, %r10d xorl %ecx, %ecx movq %r15, 0xa0(%rsp) movq %r13, 0xc0(%rsp) movl %r12d, 0xac(%rsp) cmpq %r13, %rcx je 0x2b43a8 movslq 0x2c(%rbx), %rax movslq 0x30(%rbx), %r9 movl 0x34(%rbx), %esi movq (%rbx), %rdi movq 0x10(%rbx), %r11 movq 0x40(%rbx), %r8 movq %r8, %rbp movq %rcx, 0x48(%rsp) imulq %rcx, %rbp imulq %r11, %rbp addq %rdi, %rbp movl 0x18(%rbx), %ecx movq 0x20(%rbx), %rdx movq %rbp, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %r11, 0x60(%rsp) movl %ecx, 0x68(%rsp) movq %rdx, 0x70(%rsp) movl %eax, 0x7c(%rsp) movl %r9d, 0x80(%rsp) movl $0x1, 0x84(%rsp) movl %esi, 0x88(%rsp) imulq %rax, %r9 movq %r11, %rax imulq %r9, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r11 movq %rax, 0x90(%rsp) movl 0x28(%rbx), %ecx leal -0x1(%rcx), %edx movl %edx, 0x78(%rsp) cmpl $0x4, %ecx jne 0x2b397b movq %r9, 0x90(%rsp) movq %r9, %rax movq 0x98(%rsp), %rcx movq -0x18(%rcx), %r9 movq 0x18(%rsp), %rcx vcvttss2si 0xe4(%rcx,%r9), %rcx movl %r10d, %r10d movq %r10, 0x30(%rsp) movq %rcx, %rdx shlq $0x8, %rdx movq %rcx, %r10 shlq $0x10, %r10 orq %rdx, %r10 movq %rcx, %rdx shlq $0x18, %rdx movq %rcx, %r14 shlq $0x20, %r14 orq %rdx, %r14 orq %r10, %r14 movq %rcx, %rdx shlq $0x28, %rdx movq %rcx, %r15 shlq $0x30, %r15 orq %rdx, %r15 movq %rcx, %r10 shlq $0x38, %r10 orq %r15, %r10 orq %r14, %r10 orq %rcx, %r10 movl %r12d, %ecx movl %r12d, %r13d movq 0x48(%rsp), %r12 addl %r12d, %ecx setns %dl cmpl 0xa0(%rsp), %ecx setl %r14b testb %r14b, %dl je 0x2b3bfc movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rdi movslq 0x30(%rsi), %r8 movl 0x34(%rsi), %eax movq (%rsi), %r13 movq 0x10(%rsi), %r11 movq 0x40(%rsi), %rdx movl %ecx, %r12d movq %rdx, 0x38(%rsp) imulq %rdx, %r12 imulq %r11, %r12 addq %r13, %r12 movl 0x18(%rsi), %ecx movq 0x20(%rsi), %rdx movq %r12, 0xd0(%rsp) andq $0x0, 0xd8(%rsp) movq %r11, 0xe0(%rsp) movl %ecx, 0xe8(%rsp) movq %rdx, 0xf0(%rsp) movl %edi, 0xfc(%rsp) movl %r8d, 0x100(%rsp) movl $0x1, 0x104(%rsp) movl %eax, 0x108(%rsp) movq %r8, 0xb8(%rsp) movq %r8, %rcx movq %rdi, 0x40(%rsp) imulq %rdi, %rcx movq %r11, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx movq %r11, 0xb0(%rsp) divq %r11 movq %rax, 0x110(%rsp) movl 0x28(%rsi), %eax leal -0x1(%rax), %edx movl %edx, 0xf8(%rsp) cmpl $0x4, %eax jne 0x2b3ad9 movq %rcx, 0x110(%rsp) movq 0x18(%rsp), %r14 movl 0xe0(%r14,%r9), %eax testl %eax, %eax jne 0x2b3b37 movl 0xd0(%r14,%r9), %edx movl 0xd4(%r14,%r9), %ecx movl 0xd8(%r14,%r9), %r8d movl 0xdc(%r14,%r9), %r9d movq %r10, (%rsp) leaq 0xd0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b46e0 movq (%r14), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %r9 movl 0xe0(%r14,%r9), %eax cmpl $0x1, %eax jne 0x2b3d4c movq 0x18(%rsp), %rax movl 0xd0(%rax,%r9), %r8d movl 0xd4(%rax,%r9), %ecx movl %ecx, 0xa8(%rsp) movl 0xd8(%rax,%r9), %edx xorl %r10d, %r10d testl %edx, %edx cmovlel %r10d, %edx movq 0x40(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r11d cmovgl %ecx, %r11d movl 0xdc(%rax,%r9), %esi testl %esi, %esi cmovlel %r10d, %esi testl %r8d, %r8d cmovlel %r10d, %r8d movq 0x38(%rsp), %r9 imulq 0xb0(%rsp), %r9 imulq 0x30(%rsp), %r9 addq %r13, %r9 movq %rbp, %rdi cmpl %r8d, %r10d je 0x2b3c36 movl %edx, %ecx subl $0x1, %ecx jb 0x2b3bca movq (%r12), %r14 movq %r14, (%rdi) addq $0x8, %rdi jmp 0x2b3bb8 xorl %r14d, %r14d cmpl %r14d, %r11d je 0x2b3be2 movq (%r9,%r14,8), %rcx movq %rcx, (%rdi) addq $0x8, %rdi incq %r14 jmp 0x2b3bcd movl %esi, %ecx subl $0x1, %ecx jb 0x2b3bf7 movq -0x8(%r9,%r14,8), %r15 movq %r15, (%rdi) addq $0x8, %rdi jmp 0x2b3be4 incl %r10d jmp 0x2b3bad imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi imulq %r11, %r8 imulq %r12, %r8 addq %r8, %rdi xorl %eax, %eax movq %r12, %rcx movl %r13d, %r12d movq 0xc0(%rsp), %r13 cmpq %rax, %rsi je 0x2b3f6a movq %r10, (%rdi,%rax,8) incq %rax jmp 0x2b3c24 movslq 0xfc(%rsp), %rax xorl %r11d, %r11d testl %eax, %eax movl $0x0, %r8d movq %rax, 0x120(%rsp) cmovgl %eax, %r8d movq 0xb8(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl %eax, %r14d movq %r12, %r9 cmpl %r14d, %r11d je 0x2b3cc8 movl %edx, %ecx subl $0x1, %ecx jb 0x2b3c84 movq (%r9), %r10 movq %r10, (%rdi) addq $0x8, %rdi jmp 0x2b3c73 xorl %r15d, %r15d xorl %r10d, %r10d cmpl %r15d, %r8d je 0x2b3ca0 movq (%r9,%r15,8), %rcx movq %rcx, (%rdi,%r15,8) addq $-0x8, %r10 incq %r15 jmp 0x2b3c8a movq %r9, %rcx subq %r10, %rcx subq %r10, %rdi movl %esi, %r10d subl $0x1, %r10d jb 0x2b3cc0 movq -0x8(%r9,%r15,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b3cac incl %r11d movq %rcx, %r9 jmp 0x2b3c6c movq 0x120(%rsp), %rax shlq $0x3, %rax subq %rax, %r9 xorl %r10d, %r10d movl 0xa8(%rsp), %r14d testl %r14d, %r14d cmovlel %r10d, %r14d cmpl %r14d, %r10d je 0x2b3d33 movl %edx, %ecx subl $0x1, %ecx jb 0x2b3d01 movq (%r9), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b3cf0 xorl %r11d, %r11d cmpl %r11d, %r8d je 0x2b3d19 movq (%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi incq %r11 jmp 0x2b3d04 movl %esi, %ecx subl $0x1, %ecx jb 0x2b3d2e movq -0x8(%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b3d1b incl %r10d jmp 0x2b3ce9 movq 0x98(%rsp), %rax movq -0x18(%rax), %r9 movq 0x18(%rsp), %rax movl 0xe0(%rax,%r9), %eax cmpl $0x2, %eax jne 0x2b3f55 movq 0x18(%rsp), %rcx movl 0xd0(%rcx,%r9), %r14d movslq 0xd8(%rcx,%r9), %rax movq %rcx, %r8 movl %r14d, %r11d movq 0x40(%rsp), %rcx imull %ecx, %r11d movq %rcx, %rdx negq %rdx xorl %r15d, %r15d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi testl %ecx, %ecx movl $0x0, %edi cmovgl %ecx, %edi movl 0xdc(%r8,%r9), %r8d testl %r8d, %r8d cmovlel %r15d, %r8d testl %r14d, %r14d cmovlel %r15d, %r14d movq 0x38(%rsp), %r10 imulq 0xb0(%rsp), %r10 movslq %r11d, %rcx leaq (%r12,%rcx,8), %r11 imulq 0x30(%rsp), %r10 leaq (%r10,%rax,8), %r10 leaq (%r10,%rcx,8), %r10 movq 0x18(%rsp), %r12 movl 0xd4(%r12,%r9), %r9d movl %r9d, 0x38(%rsp) shlq $0x3, %rax addq %r10, %r13 movq 0x40(%rsp), %rcx leaq (,%rcx,8), %r10 negq %r10 negq %rsi negq %r8 cmpl %r14d, %r15d je 0x2b3e5e xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b3e1c movq (%r13,%rcx,8), %r12 movq %r12, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3e05 xorl %ecx, %ecx movq %r11, %r12 cmpl %edi, %ecx je 0x2b3e39 movq (%r12), %r9 addq $0x8, %r12 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b3e21 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b3e52 movq -0x10(%r12,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3e3b leaq (%r11,%rdx,8), %r11 incl %r15d addq %r10, %r13 jmp 0x2b3dfe xorl %r14d, %r14d movq 0xb8(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r15d cmovgl %ecx, %r15d movl 0x38(%rsp), %r13d cmpl %r15d, %r14d je 0x2b3ed3 leaq (%r11,%rax), %rcx xorl %r12d, %r12d cmpq %r12, %rsi je 0x2b3e9c movq (%rcx,%r12,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %r12 jmp 0x2b3e86 xorl %ecx, %ecx cmpl %edi, %ecx je 0x2b3eb5 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b3e9e xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b3ece movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3eb7 incl %r14d jmp 0x2b3e7a movq 0x40(%rsp), %rcx addl %ecx, %ecx movslq %ecx, %rcx shlq $0x3, %rcx movq %r11, %r14 subq %rcx, %r14 xorl %r15d, %r15d testl %r13d, %r13d cmovlel %r15d, %r13d subq %rcx, %rax addq %r11, %rax cmpl %r13d, %r15d je 0x2b3f55 xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b3f14 movq (%rax,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3efe xorl %ecx, %ecx movq %r14, %r11 cmpl %edi, %ecx je 0x2b3f30 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b3f19 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b3f49 movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b3f32 leaq (%r14,%rdx,8), %r14 incl %r15d addq %r10, %rax jmp 0x2b3ef7 movq 0xc0(%rsp), %r13 movl 0xac(%rsp), %r12d movq 0x48(%rsp), %rcx incq %rcx movq 0x30(%rsp), %r10 incl %r10d jmp 0x2b38d1 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b42dc vpextrd $0x3, %xmm0, %edi movl 0xe8(%r14,%rax), %ecx movl %edi, 0x48(%rsp) addl %edi, %ecx addl 0xec(%r14,%rax), %ecx addl 0xd0(%r14,%rax), %edx addl 0xd4(%r14,%rax), %edx addl 0xd8(%r14,%rax), %esi addl 0xdc(%r14,%rax), %esi movq 0x8(%r12), %rax movq %rax, 0x8(%rsp) movl $0x8, (%rsp) movq %rbx, %rdi movq %rcx, %r14 movl %r15d, %r8d callq 0x62a26 cmpq $0x0, (%rbx) je 0x2b43a8 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b43a8 xorl %ecx, %ecx testl %r14d, %r14d cmovlel %ecx, %r14d movq %r14, 0x30(%rsp) testl %r15d, %r15d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r15d movq %r15, 0xa0(%rsp) movabsq $0x100000001, %r13 # imm = 0x100000001 xorl %r14d, %r14d movq 0x18(%rsp), %rbp movq %rbx, 0xc8(%rsp) cmpq 0xa0(%rsp), %r14 je 0x2b43a8 movq (%rbp), %rax movq -0x18(%rax), %rax vcvttss2si 0xe4(%rbp,%rax), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rsi shlq $0x20, %rsi orq %rcx, %rsi orq %rdx, %rsi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r15 shlq $0x38, %r15 orq %rdx, %r15 orq %rsi, %r15 orq %rax, %r15 xorl %r12d, %r12d cmpq 0x30(%rsp), %r12 je 0x2b4282 movslq 0x2c(%rbx), %rdi movslq 0x30(%rbx), %rdx movq (%rbx), %rax movq 0x10(%rbx), %rsi movq 0x40(%rbx), %r8 imulq %r14, %r8 movq %r8, %r9 imulq %rsi, %r9 addq %rax, %r9 movl 0x18(%rbx), %r10d movq %rdx, %rcx imulq %rdi, %rcx movq %r12, %r11 imulq %rsi, %r11 imulq %rcx, %r11 addq %r9, %r11 movq 0x20(%rbx), %r9 movq %r11, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %rsi, 0x60(%rsp) movl %r10d, 0x68(%rsp) movq %r9, 0x70(%rsp) movl $0x2, 0x78(%rsp) movl %edi, 0x7c(%rsp) movl %edx, 0x80(%rsp) movq %r13, 0x84(%rsp) movq %rcx, 0x90(%rsp) movq (%rbp), %r9 movq -0x18(%r9), %r9 movl 0xe8(%rbp,%r9), %r11d movl %r12d, %r10d subl %r11d, %r10d setns %r11b cmpl 0x48(%rsp), %r10d movq %r13, %rbx movq %rbp, %r13 setl %bpl testb %bpl, %r11b je 0x2b4246 movq 0x28(%rsp), %r8 movslq 0x2c(%r8), %rax movslq 0x30(%r8), %rcx movq 0x40(%r8), %rdx imulq %r14, %rdx movq 0x10(%r8), %rsi imulq %rsi, %rdx addq (%r8), %rdx movl 0x18(%r8), %edi movq 0x20(%r8), %r8 movq %rcx, %r11 imulq %rax, %r11 movl %r10d, %r10d movq %r11, %rbp imulq %rsi, %rbp imulq %r10, %rbp addq %rdx, %rbp movq %rbp, 0xd0(%rsp) andq $0x0, 0xd8(%rsp) movq %rsi, 0xe0(%rsp) movl %edi, 0xe8(%rsp) movq %r8, 0xf0(%rsp) movl $0x2, 0xf8(%rsp) movl %eax, 0xfc(%rsp) movl %ecx, 0x100(%rsp) movq %rbx, 0x104(%rsp) movq %r11, 0x110(%rsp) movl 0xd0(%r13,%r9), %edx movl 0xd4(%r13,%r9), %ecx movl 0xd8(%r13,%r9), %r8d movl 0xdc(%r13,%r9), %r9d movq %r15, (%rsp) leaq 0xd0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b46e0 movq %r13, %rbp movq %rbx, %r13 movq 0xc8(%rsp), %rbx incq %r12 jmp 0x2b40b1 testl %ecx, %ecx movl $0x0, %r9d cmovlel %r9d, %ecx imulq %r12, %rdx imulq %rdi, %rdx addq %r8, %rdx imulq %rdx, %rsi addq %rsi, %rax xorl %edx, %edx movq %r13, %rbp movq %rbx, %r13 movq 0xc8(%rsp), %rbx cmpq %rdx, %rcx je 0x2b423e movq %r15, (%rax,%rdx,8) incq %rdx jmp 0x2b4274 incq %r14 jmp 0x2b4048 movq (%r14), %rax movq -0x18(%rax), %r8 movl 0xd0(%r14,%r8), %r10d leal (%r10,%rdx,8), %edx addl 0xd4(%r14,%r8), %edx xorl %r11d, %r11d testb $0x7, %dl sete %r11b movq %r9, %rax shrq $0x3, %rax leal (%r11,%r11,2), %ecx shlq %cl, %rax testb $0x7, %r10b sete %cl andb %cl, %r11b cmpb $0x1, %r11b jne 0x2b42dc cmpl $0x0, 0xe0(%r14,%r8) je 0x2b43bf movq 0x28(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm1 vmovaps %xmm1, 0x50(%rsp) movq %r9, 0x60(%rsp) movl %edi, 0x68(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x70(%rsp) vmovdqu %xmm0, 0x78(%rsp) movl %r15d, 0x88(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x90(%rsp) testq %rax, %rax je 0x2b4323 lock incl (%rax) cmpl $0x1, %edi je 0x2b435b vmovdqu64 (%r12), %zmm0 leaq 0xd0(%rsp), %rcx vmovdqu64 %zmm0, (%rcx) movq 0x10(%r12), %rax movq %rax, 0x8(%rcx) leaq 0x50(%rsp), %rsi pushq $0x1 popq %rdx movq 0x28(%rsp), %rdi vzeroupper callq 0x64e3b movq (%r14), %rax addq -0x18(%rax), %r14 leaq 0x50(%rsp), %rsi movq %r14, %rdi movq %rbx, %rdx movq %r12, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b43a8 lock decl (%rax) jne 0x2b43a8 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2b43a0 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b43a8 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0x128, %rsp # imm = 0x128 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq addl 0xd8(%r14,%r8), %esi addl 0xdc(%r14,%r8), %esi sarl $0x3, %edx movq 0x8(%r12), %r9 pushq $0x8 popq %rbp movq %rbx, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b43a8 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b43a8 movq 0x18(%rsp), %r9 movq (%r9), %rax movq -0x18(%rax), %rsi vcvttss2si 0xe4(%r9,%rsi), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rdi shlq $0x20, %rdi orq %rcx, %rdi orq %rdx, %rdi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r11 shlq $0x38, %r11 orq %rdx, %r11 orq %rdi, %r11 orq %rax, %r11 movl 0xd0(%r9,%rsi), %eax movl 0xd4(%r9,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%r9,%rsi), %r8d movl 0xdc(%r9,%rsi), %r9d movq %r11, (%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b46e0 movq $0x0, 0x20(%rsp) jmp 0x2b43a8 jmp 0x2b44e4 jmp 0x2b44ab movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b44dc lock decl (%rax) jne 0x2b44dc movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2b44d6 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b44dc movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_avx512.cpp

virtual thunk to ncnn::Padding_x86_avx512::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86_avx512::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2b0de0

/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_avx512.cpp

ncnn::padding_constant_pack8_int8_sse(ncnn::Mat const&, ncnn::Mat&, int, int, int, int, long)

static void padding_constant_pack8_int8_sse(const Mat& src, Mat& dst, int top, int bottom, int left, int right, int64_t _v) { const int64_t* ptr = src; int64_t* outptr = dst; // fill top for (int y = 0; y < top; y++) { for (int x = 0; x < dst.w; x++) { *outptr++ = _v; } } // fill center for (int y = 0; y < src.h; y++) { for (int x = 0; x < left; x++) { *outptr++ = _v; } for (int x = 0; x < src.w; x++) { *outptr++ = *ptr++; } for (int x = 0; x < right; x++) { *outptr++ = _v; } } // fill bottom for (int y = 0; y < bottom; y++) { for (int x = 0; x < dst.w; x++) { *outptr++ = _v; } } }

pushq %rbp pushq %r15 pushq %r14 pushq %r12 pushq %rbx movq 0x30(%rsp), %rax movq (%rdi), %r11 movl 0x2c(%rsi), %ebx xorl %ebp, %ebp testl %ebx, %ebx cmovlel %ebp, %ebx movq (%rsi), %r10 testl %edx, %edx cmovlel %ebp, %edx cmpl %edx, %ebp je 0x2b471c movl %ebx, %r14d subl $0x1, %r14d jb 0x2b4718 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b4709 incl %ebp jmp 0x2b4702 xorl %edx, %edx testl %r8d, %r8d cmovlel %edx, %r8d movl 0x2c(%rdi), %ebx testl %ebx, %ebx cmovlel %edx, %ebx testl %r9d, %r9d cmovlel %edx, %r9d movl 0x30(%rdi), %edi testl %edi, %edi cmovlel %edx, %edi cmpl %edi, %edx je 0x2b4788 movl %r8d, %ebp subl $0x1, %ebp jb 0x2b4751 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b4743 xorl %r15d, %r15d xorl %r14d, %r14d cmpl %r15d, %ebx je 0x2b476d movq (%r11,%r15,8), %r12 movq %r12, (%r10,%r15,8) addq $-0x8, %r14 incq %r15 jmp 0x2b4757 subq %r14, %r11 subq %r14, %r10 movl %r9d, %ebp subl $0x1, %ebp jb 0x2b4784 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b4776 incl %edx jmp 0x2b473c movl 0x2c(%rsi), %edx xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %ecx, %ecx cmovlel %esi, %ecx movl %edx, %edi cmpl %ecx, %esi je 0x2b47af subl $0x1, %edi jb 0x2b47ab movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b479d incl %esi jmp 0x2b4797 popq %rbx popq %r12 popq %r14 popq %r15 popq %rbp retq

/csukuangfj[P]ncnn/src/layer/x86/padding_pack8_int8.h

ncnn::Padding_x86_fma::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86_fma::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0xf8, %rsp movq %rcx, %rbp movq %rdx, 0x8(%rsp) movq %rsi, %r14 movq (%rdi), %rax movq -0x18(%rax), %r11 movl 0xd0(%rdi,%r11), %r10d testl %r10d, %r10d jne 0x2b486b cmpl $0x0, 0xd4(%rdi,%r11) jne 0x2b486b cmpl $0x0, 0xd8(%rdi,%r11) jne 0x2b486b cmpl $0x0, 0xdc(%rdi,%r11) jne 0x2b486b cmpl $0x0, 0xe8(%rdi,%r11) jne 0x2b486b cmpl $0x0, 0xec(%rdi,%r11) je 0x2b62f3 movq %rdi, (%rsp) movl 0x18(%r14), %r15d movq 0x10(%r14), %r9 testl %r15d, %r15d je 0x2b48b2 leal (,%r9,8), %eax cltd idivl %r15d cmpl $0x8, %eax jne 0x2b48c4 movq (%rsp), %rdi movq %r14, %rsi movq 0x8(%rsp), %rdx movq %rbp, %rcx addq $0xf8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp jmp 0x2b63fc vmovdqu 0x2c(%r14), %xmm0 movq %r14, %rdi movl 0x28(%r14), %r13d jmp 0x2b4b30 vmovdqu 0x2c(%r14), %xmm0 movq %r14, 0x10(%rsp) movl 0x28(%r14), %eax vpextrd $0x2, %xmm0, %r12d pushq $0x4 popq %r13 vpextrd $0x1, %xmm0, %edx vmovd %xmm0, %esi vpextrd $0x3, %xmm0, %ecx movq %rcx, 0x18(%rsp) pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) cmpl $0x4, %r15d je 0x2b4a16 cmpl $0x8, %r15d jne 0x2b4b28 leal -0x1(%rax), %ecx cmpl $0x3, %ecx ja 0x2b4b28 leaq 0x143ec3(%rip), %rax # 0x3f87e4 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx pushq $0x8 popq %r14 jmpq *%rcx movq (%rsp), %rdi movl 0xd8(%rdi,%r11), %edx leal (%rdx,%rsi,8), %eax addl 0xdc(%rdi,%r11), %eax testb $0x3, %al sete %cl movq %r9, %r10 shrq $0x3, %r10 addb %cl, %cl testb $0x7, %al movzbl %cl, %esi pushq $0x3 popq %rcx cmovnel %esi, %ecx shlq %cl, %r10 orl %eax, %edx pushq $0x1 popq %r13 testb $0x7, %dl jne 0x2b4b2b cmpl $0x0, 0xe0(%rdi,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 cltd idivl %r14d movq 0x8(%rbp), %r8 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %esi movq %r10, %rdx movl %r14d, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vbroadcastss 0xe4(%rdx,%rcx), %ymm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %r14d movl %eax, %r8d movl %ecx, %eax cltd idivl %r14d movq $0x0, 0x20(%rsp) movq 0x10(%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b72c5 jmp 0x2b4c17 leal -0x1(%rax), %ecx cmpl $0x3, %ecx ja 0x2b4b28 leaq 0x143dab(%rip), %rax # 0x3f87d4 movslq (%rax,%rcx,4), %rcx addq %rax, %rcx jmpq *%rcx movq (%rsp), %rax movl 0xd8(%rax,%r11), %r8d leal (%r8,%rsi,4), %esi addl 0xdc(%rax,%r11), %esi testb $0x3, %sil sete %cl movq %r9, %rdx shrq $0x2, %rdx addb %cl, %cl movl %esi, %eax andl $0x7, %eax movzbl %cl, %r10d pushq $0x3 popq %rcx cmovnel %r10d, %ecx shlq %cl, %rdx pushq $0x1 popq %r13 testb $0x3, %r8b jne 0x2b4b2b cmpl $0x4, %eax jne 0x2b4b2b movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 sarl $0x2, %esi movq 0x8(%rbp), %r8 pushq $0x4 popq %rbp movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %rdx movq (%rdx), %rax movq -0x18(%rax), %rcx vbroadcastss 0xe4(%rdx,%rcx), %xmm0 movl 0xd8(%rdx,%rcx), %eax movl 0xdc(%rdx,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq $0x0, 0x20(%rsp) movq 0x10(%rsp), %rdi movq 0x8(%rsp), %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b7365 jmp 0x2b4c17 movl %eax, %r13d movq 0x10(%rsp), %rdi movq 0x8(%rdi), %rax vmovups (%rdi), %xmm1 vmovaps %xmm1, 0x80(%rsp) movq %r9, 0x90(%rsp) movl %r15d, 0x98(%rsp) movq 0x20(%rdi), %rcx movq %rcx, 0xa0(%rsp) movl %r13d, 0xa8(%rsp) vmovdqu %xmm0, 0xac(%rsp) movq 0x40(%rdi), %rcx movq %rcx, 0xc0(%rsp) testq %rax, %rax je 0x2b4b82 lock incl (%rax) cmpl $0x1, %r15d je 0x2b4bbb vmovdqu (%rbp), %ymm0 vmovups 0x20(%rbp), %ymm1 leaq 0x30(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovdqu %ymm0, (%rcx) movq 0x10(%rbp), %rax movq %rax, 0x8(%rcx) leaq 0x80(%rsp), %rsi pushq $0x1 popq %rdx vzeroupper callq 0x64e3b movq (%rsp), %rdi movq (%rdi), %rax addq -0x18(%rax), %rdi leaq 0x80(%rsp), %rsi movq 0x8(%rsp), %rdx movq %rbp, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x88(%rsp), %rax testq %rax, %rax je 0x2b4c17 lock decl (%rax) jne 0x2b4c17 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b4c0f movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b4c17 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0xf8, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp vzeroupper retq pushq $0x3 popq %r13 movq (%rsp), %rdi movl 0xe8(%rdi,%r11), %r14d movq 0x18(%rsp), %r12 leal (%r14,%r12,8), %eax addl 0xec(%rdi,%r11), %eax testb $0x3, %al sete %cl movq %r9, %r8 shrq $0x3, %r8 addb %cl, %cl testb $0x7, %al movzbl %cl, %ecx cmovel %r13d, %ecx shlq %cl, %r8 orl %eax, %r14d testb $0x7, %r14b jne 0x2b4b2b addl 0xd8(%rdi,%r11), %esi addl 0xdc(%rdi,%r11), %esi addl %edx, %r10d addl 0xd4(%rdi,%r11), %r10d leal (,%r12,8), %ecx cmpl %ecx, %eax movq 0x10(%rsp), %rdi pushq $0x8 popq %r14 je 0x2b4cbc movq (%rsp), %rcx cmpl $0x0, 0xe0(%rcx,%r11) jne 0x2b4b30 cltd idivl %r14d subq $0x8, %rsp movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r14d, %r9d movq %rax, %r14 movl %eax, %ecx pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xf0(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xf0(%rsp), %r9 je 0x2b4c17 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b4e05 movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b4e2f movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x5, %r8 vmovups (%rdx,%r8), %ymm0 jmp 0x2b4e39 vbroadcastss 0xe4(%r12,%rdi), %ymm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b4ff5 movq 0x10(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b4ef1 movq %rsi, 0x70(%rsp) movq (%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b4f4c movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b72c5 movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b501f movq (%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b5028 vmovdqa (%rbp), %ymm0 movl %eax, %edi subl $0x1, %edi jb 0x2b4fb8 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b4fa7 xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b4fda vmovdqa (%r12,%rdi), %ymm0 vmovdqa %ymm0, (%r13,%rdi) incl %r10d addq $0x20, %rdi jmp 0x2b4fbd addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b4ff0 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b4fdf incl %r8d jmp 0x2b4f97 imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b53c3 vmovups %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b500a movq (%rsp), %r12 jmp 0x2b510b xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b5085 vmovdqa (%rbp), %ymm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b504a vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b5038 xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b506c vmovdqa (%rbp), %ymm0 vmovdqa %ymm0, (%r13) addq $0x20, %rbp addq $0x20, %r13 incl %r8d jmp 0x2b504d movl %edx, %r8d subl $0x1, %r8d jb 0x2b5081 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b506f incl %edi jmp 0x2b502a movl 0x5c(%rsp), %edi shll $0x3, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b50fb vmovdqa (%rbp), %ymm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b50bb vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b50a9 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b50df vmovdqa (%rbp,%r8), %ymm0 vmovdqa %ymm0, (%r13,%r8) incl %r9d addq $0x20, %r8 jmp 0x2b50c1 addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b50f7 vmovdqa %ymm0, (%r13) addq $0x20, %r13 jmp 0x2b50e5 incl %edi jmp 0x2b509d movq -0x18(%rcx), %rdi movq (%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x40 popq %r13 jne 0x2b532b movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x3, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x5, %rax shlq $0x5, %rdx shlq $0x5, %rsi cmpl %r9d, %r10d je 0x2b520c leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b51ae vmovdqa (%r11,%r14), %ymm0 vmovdqa %ymm0, (%r8,%r15) addq $-0x20, %r14 addq $0x20, %r15 jmp 0x2b5190 subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b51d2 vmovdqa (%r11), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r11 addq $0x20, %r8 incl %ebp jmp 0x2b51b6 movq %r13, %r14 movq (%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x40, %r15 je 0x2b51f8 vmovdqa (%r11,%r14), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r14 jmp 0x2b51d9 movslq 0x5c(%rsp), %r11 shlq $0x5, %r11 subq %r11, %rcx incl %r10d jmp 0x2b517d xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b5288 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b523e vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8,%r14) addq $-0x20, %r11 addq $0x20, %r14 jmp 0x2b5220 subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b5261 vmovdqa (%rcx), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %rcx addq $0x20, %r8 incl %r10d jmp 0x2b5244 movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x40, %r11 je 0x2b5283 vmovdqa (%rcx,%r10), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r10 jmp 0x2b5264 incl %r9d jmp 0x2b520f movslq 0x5c(%rsp), %r9 shlq $0x6, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b532b leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b52ce vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8,%r14) addq $-0x20, %r11 addq $0x20, %r14 jmp 0x2b52b0 subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b52f5 vmovdqa (%r10), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r10 addq $0x20, %r8 incl %r11d jmp 0x2b52d7 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x40, %r14 je 0x2b5317 vmovdqa (%r10,%r11), %ymm0 vmovdqa %ymm0, (%r8) addq $0x20, %r8 addq $-0x20, %r11 jmp 0x2b52f8 movslq 0x5c(%rsp), %r10 shlq $0x5, %r10 subq %r10, %rcx incl %r9d jmp 0x2b529d movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b535f lock decl (%rax) jne 0x2b535f movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b5354 movq (%rdi), %rax vzeroupper callq *0x18(%rax) jmp 0x2b535f movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b53c3 lock decl (%rax) jne 0x2b53c3 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b53ab movq (%rdi), %rax vzeroupper callq *0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b53c3 movq %rsi, %rdi vzeroupper callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b4d45 movq (%rsp), %rax movl 0xe8(%rax,%r11), %r14d movq 0x18(%rsp), %rdi leal (%r14,%rdi,4), %r13d addl 0xec(%rax,%r11), %r13d testb $0x3, %r13b sete %cl movq %r9, %r8 shrq $0x2, %r8 addb %cl, %cl movq %r13, %r12 movl %r13d, %eax andl $0x7, %eax movzbl %cl, %ecx pushq $0x3 popq %r13 cmovel %r13d, %ecx shlq %cl, %r8 testb $0x3, %r14b jne 0x2b4b2b cmpl $0x4, %eax jne 0x2b4b2b movq (%rsp), %rax addl 0xd8(%rax,%r11), %esi addl 0xdc(%rax,%r11), %esi addl %edx, %r10d addl 0xd4(%rax,%r11), %r10d leal (,%rdi,4), %eax cmpl %eax, %r12d movq 0x10(%rsp), %rdi je 0x2b5468 movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%r11) jne 0x2b4b30 movq %r12, %rcx sarl $0x2, %ecx subq $0x8, %rsp pushq $0x4 popq %r9 movq 0x10(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movq %rcx, %r14 pushq 0x8(%rbp) callq 0x628f2 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %r12 movq (%r12), %rax movq -0x18(%rax), %rax movl 0xe8(%r12,%rax), %eax pushq $-0x4 popq %rcx cltd idivl %ecx movl %eax, %edx xorl %ecx, %ecx testl %r14d, %r14d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r14d movq %r14, 0xf0(%rsp) movl %edx, 0x7c(%rsp) movl %edx, %r11d xorl %r9d, %r9d cmpq 0xf0(%rsp), %r9 je 0x2b4c17 movq 0x8(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rcx movq 0x40(%r10), %r13 imulq %r9, %r13 movq 0x10(%r10), %rdi imulq %rdi, %r13 addq (%r10), %r13 movl 0x34(%r10), %esi movl 0x18(%r10), %edx movq 0x20(%r10), %r8 movq %r13, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %rdi, 0x90(%rsp) movl %edx, 0x98(%rsp) movq %r8, 0xa0(%rsp) movl %eax, 0xac(%rsp) movl %ecx, 0xb0(%rsp) movl $0x1, 0xb4(%rsp) movl %esi, 0xb8(%rsp) imulq %rax, %rcx movq %rdi, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %rdi movq %rax, 0xc0(%rsp) movl 0x28(%r10), %edx leal -0x1(%rdx), %edi movl %edi, 0xa8(%rsp) cmpl $0x4, %edx jne 0x2b55b2 movq %rcx, 0xc0(%rsp) movq %rcx, %rax movq (%r12), %rcx movq -0x18(%rcx), %rdi cmpl $0x0, 0xf0(%r12,%rdi) je 0x2b55dc movq 0xf8(%r12,%rdi), %rdx movq %r9, %r8 shlq $0x4, %r8 vmovups (%rdx,%r8), %xmm0 jmp 0x2b55e6 vbroadcastss 0xe4(%r12,%rdi), %xmm0 movl %r11d, %r11d movl 0x7c(%rsp), %edx movq %r9, 0xd0(%rsp) addl %r9d, %edx setns %r8b cmpl 0x18(%rsp), %edx setl %r9b testb %r9b, %r8b je 0x2b57a2 movq 0x10(%rsp), %r10 movslq 0x2c(%r10), %rax movslq 0x30(%r10), %rsi movl 0x34(%r10), %r8d movq (%r10), %r12 movq 0x10(%r10), %r15 movq 0x40(%r10), %rbx movl %edx, %ebp imulq %rbx, %rbp imulq %r15, %rbp addq %r12, %rbp movl 0x18(%r10), %edx movq 0x20(%r10), %r9 movq %rbp, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r15, 0x40(%rsp) movl %edx, 0x48(%rsp) movq %r9, 0x50(%rsp) movl %eax, 0x5c(%rsp) movl %esi, 0x60(%rsp) movl $0x1, 0x64(%rsp) movl %r8d, 0x68(%rsp) imulq %rax, %rsi movq %r15, %rax imulq %rsi, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r15 movq %rax, 0x70(%rsp) movl 0x28(%r10), %eax leal -0x1(%rax), %edx movl %edx, 0x58(%rsp) cmpl $0x4, %eax jne 0x2b569e movq %rsi, 0x70(%rsp) movq (%rsp), %r14 movl 0xe0(%r14,%rdi), %eax testl %eax, %eax movq %r11, 0x28(%rsp) jne 0x2b56f9 movl 0xd0(%r14,%rdi), %edx movl 0xd4(%r14,%rdi), %ecx movl 0xd8(%r14,%rdi), %r8d movl 0xdc(%r14,%rdi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi callq 0x2b7365 movq 0x28(%rsp), %r11 movq (%r14), %rcx movq -0x18(%rcx), %rdi movl 0xe0(%r14,%rdi), %eax cmpl $0x1, %eax jne 0x2b57cc movq (%rsp), %rsi movl 0xd8(%rsi,%rdi), %eax xorl %r8d, %r8d testl %eax, %eax cmovlel %r8d, %eax movl 0xdc(%rsi,%rdi), %edx testl %edx, %edx cmovlel %r8d, %edx movl 0xd0(%rsi,%rdi), %r9d testl %r9d, %r9d cmovlel %r8d, %r9d movl 0xd4(%rsi,%rdi), %esi imulq %r15, %rbx imulq %r11, %rbx addq %rbx, %r12 cmpl %r9d, %r8d je 0x2b57d5 vmovdqa (%rbp), %xmm0 movl %eax, %edi subl $0x1, %edi jb 0x2b5765 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b5754 xorl %edi, %edi xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b5787 vmovdqa (%r12,%rdi), %xmm0 vmovdqa %xmm0, (%r13,%rdi) incl %r10d addq $0x10, %rdi jmp 0x2b576a addq %rdi, %r13 movl %edx, %edi subl $0x1, %edi jb 0x2b579d vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b578c incl %r8d jmp 0x2b5744 imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi movq 0xd0(%rsp), %r9 subl $0x1, %esi jb 0x2b5b64 vmovups %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b57b7 movq (%rsp), %r12 jmp 0x2b58b8 xorl %edi, %edi cmpl 0x60(%rsp), %edi jge 0x2b5832 vmovdqa (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b57f7 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b57e5 xorl %r8d, %r8d cmpl 0x5c(%rsp), %r8d jge 0x2b5819 vmovdqa (%rbp), %xmm0 vmovdqa %xmm0, (%r13) addq $0x10, %rbp addq $0x10, %r13 incl %r8d jmp 0x2b57fa movl %edx, %r8d subl $0x1, %r8d jb 0x2b582e vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b581c incl %edi jmp 0x2b57d7 movl 0x5c(%rsp), %edi shll $0x2, %edi movslq %edi, %rdi shlq $0x2, %rdi subq %rdi, %rbp xorl %edi, %edi testl %esi, %esi cmovlel %edi, %esi cmpl %esi, %edi je 0x2b58a8 vmovdqa (%rbp), %xmm0 movl %eax, %r8d subl $0x1, %r8d jb 0x2b5868 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b5856 xorl %r8d, %r8d xorl %r9d, %r9d cmpl 0x5c(%rsp), %r9d jge 0x2b588c vmovdqa (%rbp,%r8), %xmm0 vmovdqa %xmm0, (%r13,%r8) incl %r9d addq $0x10, %r8 jmp 0x2b586e addq %r8, %r13 movl %edx, %r8d subl $0x1, %r8d jb 0x2b58a4 vmovdqa %xmm0, (%r13) addq $0x10, %r13 jmp 0x2b5892 incl %edi jmp 0x2b584a movq -0x18(%rcx), %rdi movq (%rsp), %r12 movl 0xe0(%r12,%rdi), %eax cmpl $0x2, %eax pushq $-0x20 popq %r13 jne 0x2b5ad8 movl 0xd0(%r12,%rdi), %r9d movslq 0xd8(%r12,%rdi), %rax movl 0x5c(%rsp), %ecx imull %r9d, %ecx shll $0x2, %ecx movslq %ecx, %rcx shlq $0x2, %rcx addq 0x30(%rsp), %rcx xorl %r10d, %r10d testl %eax, %eax movl $0x0, %edx cmovgl %eax, %edx movl 0xdc(%r12,%rdi), %esi testl %esi, %esi cmovlel %r10d, %esi movl 0xd4(%r12,%rdi), %edi testl %r9d, %r9d cmovlel %r10d, %r9d movq 0x80(%rsp), %r8 shlq $0x4, %rax shlq $0x4, %rdx shlq $0x4, %rsi cmpl %r9d, %r10d je 0x2b59b9 leaq (%rcx,%rax), %r11 xorl %r15d, %r15d xorl %r14d, %r14d movq %rdx, %r12 addq %r14, %r12 je 0x2b595b vmovdqa (%r11,%r14), %xmm0 vmovdqa %xmm0, (%r8,%r15) addq $-0x10, %r14 addq $0x10, %r15 jmp 0x2b593d subq %r14, %r8 xorl %ebp, %ebp movq %rcx, %r11 cmpl 0x5c(%rsp), %ebp jge 0x2b597f vmovdqa (%r11), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r11 addq $0x10, %r8 incl %ebp jmp 0x2b5963 movq %r13, %r14 movq (%rsp), %r12 leaq (%rsi,%r14), %r15 cmpq $-0x20, %r15 je 0x2b59a5 vmovdqa (%r11,%r14), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r14 jmp 0x2b5986 movslq 0x5c(%rsp), %r11 shlq $0x4, %r11 subq %r11, %rcx incl %r10d jmp 0x2b592a xorl %r9d, %r9d cmpl 0x60(%rsp), %r9d jge 0x2b5a35 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b59eb vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2b59cd subq %r11, %r8 xorl %r10d, %r10d cmpl 0x5c(%rsp), %r10d jge 0x2b5a0e vmovdqa (%rcx), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %rcx addq $0x10, %r8 incl %r10d jmp 0x2b59f1 movq %r13, %r10 leaq (%rsi,%r10), %r11 cmpq $-0x20, %r11 je 0x2b5a30 vmovdqa (%rcx,%r10), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r10 jmp 0x2b5a11 incl %r9d jmp 0x2b59bc movslq 0x5c(%rsp), %r9 shlq $0x5, %r9 subq %r9, %rcx xorl %r9d, %r9d testl %edi, %edi cmovlel %r9d, %edi cmpl %edi, %r9d je 0x2b5ad8 leaq (%rcx,%rax), %r10 xorl %r14d, %r14d xorl %r11d, %r11d movq %rdx, %r15 addq %r11, %r15 je 0x2b5a7b vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8,%r14) addq $-0x10, %r11 addq $0x10, %r14 jmp 0x2b5a5d subq %r11, %r8 xorl %r11d, %r11d movq %rcx, %r10 cmpl 0x5c(%rsp), %r11d jge 0x2b5aa2 vmovdqa (%r10), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r10 addq $0x10, %r8 incl %r11d jmp 0x2b5a84 movq %r13, %r11 leaq (%rsi,%r11), %r14 cmpq $-0x20, %r14 je 0x2b5ac4 vmovdqa (%r10,%r11), %xmm0 vmovdqa %xmm0, (%r8) addq $0x10, %r8 addq $-0x10, %r11 jmp 0x2b5aa5 movslq 0x5c(%rsp), %r10 shlq $0x4, %r10 subq %r10, %rcx incl %r9d jmp 0x2b5a4a movq 0x38(%rsp), %rax testq %rax, %rax je 0x2b5b06 lock decl (%rax) jne 0x2b5b06 movq 0x30(%rsp), %rsi movq 0x50(%rsp), %rdi testq %rdi, %rdi je 0x2b5afe movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b5b06 movq %rsi, %rdi callq 0x5f3e0 movq 0x88(%rsp), %rax testq %rax, %rax movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 je 0x2b5b64 lock decl (%rax) jne 0x2b5b64 movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi je 0x2b5b4f movq (%rdi), %rax callq *0x18(%rax) movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 jmp 0x2b5b64 movq %rsi, %rdi callq 0x5f3e0 movq 0x28(%rsp), %r11 movq 0xd0(%rsp), %r9 incq %r9 incl %r11d jmp 0x2b54f2 movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 movq (%rsp), %rax movl 0xe8(%rax,%r11), %r13d addl %r12d, %r13d addl 0xec(%rax,%r11), %r13d addl %edx, %r10d addl 0xd4(%rax,%r11), %r10d addl 0xd8(%rax,%r11), %esi addl 0xdc(%rax,%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x8 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %r11 # imm = 0x100000001 xorl %r14d, %r14d movq (%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b4c17 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b5c61 movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x5, %rcx vmovups (%rax,%rcx), %ymm0 jmp 0x2b5c6b vbroadcastss 0xe4(%rbp,%rax), %ymm0 xorl %r15d, %r15d vmovups %ymm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b5e1d movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %r11, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b5e04 movq 0x10(%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %r13, %rbx movq %r11, %r13 movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq (%rsp), %rbp movq %r13, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovups 0xd0(%rsp), %ymm0 callq 0x2b72c5 vmovups 0xd0(%rsp), %ymm0 movq %r13, %r11 movq %rbx, %r13 incq %r15 jmp 0x2b5c77 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b5dfc vmovups %ymm0, (%rcx) addq $0x20, %rcx jmp 0x2b5e0e incq %r14 jmp 0x2b5c2e leal (%r10,%rdx,8), %eax movq (%rsp), %rdi addl 0xd4(%rdi,%r11), %eax testb $0x3, %al sete %cl movq %r9, %r8 shrq $0x3, %r8 addb %cl, %cl testb $0x7, %al movzbl %cl, %edx pushq $0x3 popq %rcx cmovnel %edx, %ecx shlq %cl, %r8 orl %eax, %r10d pushq $0x2 popq %r13 testb $0x7, %r10b jne 0x2b4b2b cmpl $0x0, 0xe0(%rdi,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 movq (%rsp), %rcx addl 0xd8(%rcx,%r11), %esi addl 0xdc(%rcx,%r11), %esi cltd idivl %r14d movq 0x8(%rbp), %r9 movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %eax, %edx movq %r8, %rcx movl %r14d, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vbroadcastss 0xe4(%rdi,%rsi), %ymm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %r14d movl %eax, %r10d movl %ecx, %eax cltd idivl %r14d movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq 0x10(%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b72c5 jmp 0x2b62e5 movq (%rsp), %rax cmpl $0x0, 0xe0(%rax,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 movq (%rsp), %rax movl 0xe8(%rax,%r11), %r13d addl %r12d, %r13d addl 0xec(%rax,%r11), %r13d addl %edx, %r10d addl 0xd4(%rax,%r11), %r10d addl 0xd8(%rax,%r11), %esi addl 0xdc(%rax,%r11), %esi movq 0x8(%rsp), %rbx movq %rbx, %rdi movl %r10d, %edx movl %r13d, %ecx movq 0x18(%rsp), %r8 pushq 0x8(%rbp) pushq $0x4 callq 0x62a26 addq $0x10, %rsp cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 xorl %ecx, %ecx testl %r13d, %r13d cmovlel %ecx, %r13d movq 0x18(%rsp), %rdx testl %edx, %edx movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %edx movq %rdx, 0x18(%rsp) movabsq $0x100000001, %r11 # imm = 0x100000001 xorl %r14d, %r14d movq (%rsp), %rbp cmpq 0x18(%rsp), %r14 je 0x2b4c17 movq (%rbp), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xf0(%rbp,%rax) je 0x2b6013 movq 0xf8(%rbp,%rax), %rax movq %r14, %rcx shlq $0x4, %rcx vmovups (%rax,%rcx), %xmm0 jmp 0x2b601d vbroadcastss 0xe4(%rbp,%rax), %xmm0 xorl %r15d, %r15d vmovaps %xmm0, 0xd0(%rsp) cmpq %r13, %r15 je 0x2b61d5 movq 0x8(%rsp), %rax movslq 0x2c(%rax), %rdx movslq 0x30(%rax), %rsi movq 0x40(%rax), %rdi imulq %r14, %rdi movq 0x10(%rax), %r8 imulq %r8, %rdi addq (%rax), %rdi movl 0x18(%rax), %r9d movq 0x20(%rax), %r10 movq %rsi, %rax imulq %rdx, %rax movq %r15, %rcx imulq %r8, %rcx imulq %rax, %rcx addq %rdi, %rcx movq %rcx, 0x80(%rsp) andq $0x0, 0x88(%rsp) movq %r8, 0x90(%rsp) movl %r9d, 0x98(%rsp) movq %r10, 0xa0(%rsp) movl $0x2, 0xa8(%rsp) movl %edx, 0xac(%rsp) movl %esi, 0xb0(%rsp) movq %r11, 0xb4(%rsp) movq %rax, 0xc0(%rsp) movq (%rbp), %rdx movq -0x18(%rdx), %rsi movl 0xe8(%rbp,%rsi), %edi movl %r15d, %edx subl %edi, %edx setns %dil cmpl %r12d, %edx setl %r8b testb %r8b, %dil je 0x2b61bc movq 0x10(%rsp), %rcx movslq 0x2c(%rcx), %rax movslq 0x30(%rcx), %rbp movq 0x40(%rcx), %rdi imulq %r14, %rdi movq 0x10(%rcx), %r8 imulq %r8, %rdi addq (%rcx), %rdi movl 0x18(%rcx), %r9d movq 0x20(%rcx), %r10 movq %rbp, %rcx imulq %rax, %rcx movl %edx, %edx movq %r13, %rbx movl %r12d, %r13d movq %r11, %r12 movq %rcx, %r11 imulq %r8, %r11 imulq %rdx, %r11 addq %rdi, %r11 movq %r11, 0x30(%rsp) andq $0x0, 0x38(%rsp) movq %r8, 0x40(%rsp) movl %r9d, 0x48(%rsp) movq %r10, 0x50(%rsp) movl $0x2, 0x58(%rsp) movl %eax, 0x5c(%rsp) movl %ebp, 0x60(%rsp) movq (%rsp), %rbp movq %r12, 0x64(%rsp) movq %rcx, 0x70(%rsp) movl 0xd0(%rbp,%rsi), %edx movl 0xd4(%rbp,%rsi), %ecx movl 0xd8(%rbp,%rsi), %r8d movl 0xdc(%rbp,%rsi), %r9d leaq 0x30(%rsp), %rdi leaq 0x80(%rsp), %rsi vmovaps 0xd0(%rsp), %xmm0 callq 0x2b7365 vmovaps 0xd0(%rsp), %xmm0 movq %r12, %r11 movl %r13d, %r12d movq %rbx, %r13 incq %r15 jmp 0x2b6029 testl %eax, %eax movl $0x0, %edx cmovlel %edx, %eax subl $0x1, %eax jb 0x2b61b4 vmovups %xmm0, (%rcx) addq $0x10, %rcx jmp 0x2b61c6 incq %r14 jmp 0x2b5fe0 leal (%r10,%rdx,4), %edx movq (%rsp), %rax addl 0xd4(%rax,%r11), %edx testb $0x3, %dl sete %cl movq %r9, %rax shrq $0x2, %rax addb %cl, %cl movl %edx, %r8d andl $0x7, %r8d movzbl %cl, %r14d pushq $0x3 popq %rcx cmovnel %r14d, %ecx shlq %cl, %rax pushq $0x2 popq %r13 testb $0x3, %r10b jne 0x2b4b2b cmpl $0x4, %r8d jne 0x2b4b2b movq (%rsp), %rcx cmpl $0x0, 0xe0(%rcx,%r11) movq 0x10(%rsp), %rdi jne 0x2b4b30 movq (%rsp), %rcx addl 0xd8(%rcx,%r11), %esi addl 0xdc(%rcx,%r11), %esi sarl $0x2, %edx movq 0x8(%rbp), %r9 pushq $0x4 popq %rbp movq 0x8(%rsp), %rbx movq %rbx, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b4c17 movq 0x8(%rsp), %rcx movslq 0x38(%rcx), %rax imulq 0x40(%rcx), %rax testq %rax, %rax je 0x2b4c17 movq (%rsp), %rdi movq (%rdi), %rax movq -0x18(%rax), %rsi vbroadcastss 0xe4(%rdi,%rsi), %xmm0 movl 0xd0(%rdi,%rsi), %eax movl 0xd4(%rdi,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%rdi,%rsi), %r8d movl 0xdc(%rdi,%rsi), %r9d movq 0x10(%rsp), %rdi movq 0x8(%rsp), %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b7365 movq $0x0, 0x20(%rsp) jmp 0x2b4c17 movq $0x0, 0x20(%rsp) cmpq %r14, 0x8(%rsp) je 0x2b4c17 movq 0x8(%r14), %rax testq %rax, %rax je 0x2b6313 lock incl (%rax) movq 0x8(%rsp), %rax movq 0x8(%rax), %rax testq %rax, %rax je 0x2b6347 lock decl (%rax) jne 0x2b6347 movq 0x8(%rsp), %rax movq (%rax), %rsi movq 0x20(%rax), %rdi testq %rdi, %rdi je 0x2b633f movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b6347 movq %rsi, %rdi callq 0x5f3e0 movq 0x8(%rsp), %rcx andq $0x0, 0x40(%rcx) vxorps %xmm0, %xmm0, %xmm0 vmovups %xmm0, (%rcx) vmovups %xmm0, 0xc(%rcx) andl $0x0, 0x38(%rcx) vmovups %xmm0, 0x28(%rcx) vmovups (%r14), %xmm0 vmovups %xmm0, (%rcx) movq 0x10(%r14), %rax movq %rax, 0x10(%rcx) movl 0x18(%r14), %eax movl %eax, 0x18(%rcx) movq 0x20(%r14), %rax movq %rax, 0x20(%rcx) vmovdqu 0x28(%r14), %xmm0 vmovdqu %xmm0, 0x28(%rcx) movl 0x38(%r14), %eax movl %eax, 0x38(%rcx) movq 0x40(%r14), %rax movq %rax, 0x40(%rcx) jmp 0x2b4c17 jmp 0x2b63f4 jmp 0x2b63f4 jmp 0x2b63f4 jmp 0x2b63f4 jmp 0x2b63f4 jmp 0x2b63b2 movq %rax, %rbx movq 0x88(%rsp), %rax testq %rax, %rax je 0x2b63ec lock decl (%rax) jne 0x2b63ec movq 0x80(%rsp), %rsi movq 0xa0(%rsp), %rdi testq %rdi, %rdi jne 0x2b63e6 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b63ec movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_fma.cpp

ncnn::Padding_x86_fma::forward_int8(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86_fma::forward_int8(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); padding_constant_pack8_int8_sse(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_int8_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_int8_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { // TODO perchannel // int64_t pad_value = per_channel_pad_data_size ? vld1_s8(per_channel_pad_data + q * 8) : vdup_n_s8((signed char)value); int64_t v8 = (int64_t)value; int64_t pad_value = v8 | (v8 << 8) | (v8 << 16) | (v8 << 24) | (v8 << 32) | (v8 << 40) | (v8 << 48) | (v8 << 56); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill<int64_t>(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_int8_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x128, %rsp # imm = 0x128 movq %rcx, %r12 movq %rdx, %rbx movq %rdi, %r14 movl 0x38(%rsi), %r15d movq 0x10(%rsi), %r9 movl 0x18(%rsi), %edi movq %rsi, 0x28(%rsp) vmovdqu 0x28(%rsi), %xmm0 cmpl $0x8, %edi jne 0x2b70ac vmovd %xmm0, %eax decl %eax cmpl $0x3, %eax ja 0x2b70ac leaq 0x1423aa(%rip), %rcx # 0x3f87f4 movslq (%rcx,%rax,4), %rax vpextrd $0x1, %xmm0, %esi addq %rcx, %rax vpextrd $0x2, %xmm0, %edx pushq $-0x64 popq %rcx movq %rcx, 0x20(%rsp) movq %r14, 0x18(%rsp) jmpq *%rax movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %r8d leal (%r8,%rsi,8), %esi addl 0xdc(%r14,%rax), %esi xorl %r10d, %r10d testb $0x7, %sil sete %r10b movq %r9, %rdx shrq $0x3, %rdx leal (%r10,%r10,2), %ecx shlq %cl, %rdx testb $0x7, %r8b sete %cl andb %cl, %r10b cmpb $0x1, %r10b jne 0x2b70ac cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b70ac sarl $0x3, %esi movq 0x8(%r12), %r8 pushq $0x8 popq %rbp movq %rbx, %rdi movl %ebp, %ecx callq 0x626da cmpq $0x0, (%rbx) je 0x2b7181 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b7181 movq 0x18(%rsp), %r8 movq (%r8), %rax movq -0x18(%rax), %rcx vcvttss2si 0xe4(%r8,%rcx), %rax movq %rax, %rdx shlq $0x8, %rdx movq %rax, %rsi shlq $0x10, %rsi orq %rdx, %rsi movq %rax, %rdx shlq $0x18, %rdx movq %rax, %rdi shlq $0x20, %rdi orq %rdx, %rdi orq %rsi, %rdi movq %rax, %rdx shlq $0x28, %rdx movq %rax, %rsi shlq $0x30, %rsi orq %rdx, %rsi movq %rax, %r9 shlq $0x38, %r9 orq %rsi, %r9 orq %rdi, %r9 orq %rax, %r9 movl 0xd8(%r8,%rcx), %eax movl 0xdc(%r8,%rcx), %ecx cltd idivl %ebp movl %eax, %r8d movl %ecx, %eax cltd idivl %ebp movq %r9, (%rsp) movq $0x0, 0x20(%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi xorl %edx, %edx xorl %ecx, %ecx movl %eax, %r9d callq 0x2b740e jmp 0x2b7181 movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %r10d leal (%r10,%r15,8), %r13d addl 0xec(%r14,%rax), %r13d xorl %r11d, %r11d testb $0x7, %r13b sete %r11b movq %r9, %r8 shrq $0x3, %r8 leal (%r11,%r11,2), %ecx shlq %cl, %r8 testb $0x7, %r10b sete %cl andb %cl, %r11b cmpb $0x1, %r11b jne 0x2b70ac addl 0xd8(%r14,%rax), %esi addl 0xdc(%r14,%rax), %esi addl 0xd0(%r14,%rax), %edx addl 0xd4(%r14,%rax), %edx leal (,%r15,8), %ecx cmpl %ecx, %r13d je 0x2b6615 cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b70ac sarl $0x3, %r13d movq 0x8(%r12), %rax movq %rax, (%rsp) pushq $0x8 popq %r9 movq %rbx, %rdi movl %r13d, %ecx callq 0x628f2 cmpq $0x0, (%rbx) je 0x2b7181 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b7181 movq 0x18(%rsp), %rcx movq (%rcx), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %rax movl 0xe8(%rcx,%rax), %eax pushq $-0x8 popq %rcx cltd idivl %ecx movl %eax, %r12d xorl %ecx, %ecx testl %r13d, %r13d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r13d movl %r12d, %r10d xorl %ecx, %ecx movq %r15, 0xa0(%rsp) movq %r13, 0xc0(%rsp) movl %r12d, 0xac(%rsp) cmpq %r13, %rcx je 0x2b7181 movslq 0x2c(%rbx), %rax movslq 0x30(%rbx), %r9 movl 0x34(%rbx), %esi movq (%rbx), %rdi movq 0x10(%rbx), %r11 movq 0x40(%rbx), %r8 movq %r8, %rbp movq %rcx, 0x48(%rsp) imulq %rcx, %rbp imulq %r11, %rbp addq %rdi, %rbp movl 0x18(%rbx), %ecx movq 0x20(%rbx), %rdx movq %rbp, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %r11, 0x60(%rsp) movl %ecx, 0x68(%rsp) movq %rdx, 0x70(%rsp) movl %eax, 0x7c(%rsp) movl %r9d, 0x80(%rsp) movl $0x1, 0x84(%rsp) movl %esi, 0x88(%rsp) imulq %rax, %r9 movq %r11, %rax imulq %r9, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx divq %r11 movq %rax, 0x90(%rsp) movl 0x28(%rbx), %ecx leal -0x1(%rcx), %edx movl %edx, 0x78(%rsp) cmpl $0x4, %ecx jne 0x2b674b movq %r9, 0x90(%rsp) movq %r9, %rax movq 0x98(%rsp), %rcx movq -0x18(%rcx), %r9 movq 0x18(%rsp), %rcx vcvttss2si 0xe4(%rcx,%r9), %rcx movl %r10d, %r10d movq %r10, 0x30(%rsp) movq %rcx, %rdx shlq $0x8, %rdx movq %rcx, %r10 shlq $0x10, %r10 orq %rdx, %r10 movq %rcx, %rdx shlq $0x18, %rdx movq %rcx, %r14 shlq $0x20, %r14 orq %rdx, %r14 orq %r10, %r14 movq %rcx, %rdx shlq $0x28, %rdx movq %rcx, %r15 shlq $0x30, %r15 orq %rdx, %r15 movq %rcx, %r10 shlq $0x38, %r10 orq %r15, %r10 orq %r14, %r10 orq %rcx, %r10 movl %r12d, %ecx movl %r12d, %r13d movq 0x48(%rsp), %r12 addl %r12d, %ecx setns %dl cmpl 0xa0(%rsp), %ecx setl %r14b testb %r14b, %dl je 0x2b69cc movq 0x28(%rsp), %rsi movslq 0x2c(%rsi), %rdi movslq 0x30(%rsi), %r8 movl 0x34(%rsi), %eax movq (%rsi), %r13 movq 0x10(%rsi), %r11 movq 0x40(%rsi), %rdx movl %ecx, %r12d movq %rdx, 0x38(%rsp) imulq %rdx, %r12 imulq %r11, %r12 addq %r13, %r12 movl 0x18(%rsi), %ecx movq 0x20(%rsi), %rdx movq %r12, 0xd0(%rsp) andq $0x0, 0xd8(%rsp) movq %r11, 0xe0(%rsp) movl %ecx, 0xe8(%rsp) movq %rdx, 0xf0(%rsp) movl %edi, 0xfc(%rsp) movl %r8d, 0x100(%rsp) movl $0x1, 0x104(%rsp) movl %eax, 0x108(%rsp) movq %r8, 0xb8(%rsp) movq %r8, %rcx movq %rdi, 0x40(%rsp) imulq %rdi, %rcx movq %r11, %rax imulq %rcx, %rax addq $0xf, %rax andq $-0x10, %rax xorl %edx, %edx movq %r11, 0xb0(%rsp) divq %r11 movq %rax, 0x110(%rsp) movl 0x28(%rsi), %eax leal -0x1(%rax), %edx movl %edx, 0xf8(%rsp) cmpl $0x4, %eax jne 0x2b68a9 movq %rcx, 0x110(%rsp) movq 0x18(%rsp), %r14 movl 0xe0(%r14,%r9), %eax testl %eax, %eax jne 0x2b6907 movl 0xd0(%r14,%r9), %edx movl 0xd4(%r14,%r9), %ecx movl 0xd8(%r14,%r9), %r8d movl 0xdc(%r14,%r9), %r9d movq %r10, (%rsp) leaq 0xd0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b740e movq (%r14), %rax movq %rax, 0x98(%rsp) movq -0x18(%rax), %r9 movl 0xe0(%r14,%r9), %eax cmpl $0x1, %eax jne 0x2b6b1c movq 0x18(%rsp), %rax movl 0xd0(%rax,%r9), %r8d movl 0xd4(%rax,%r9), %ecx movl %ecx, 0xa8(%rsp) movl 0xd8(%rax,%r9), %edx xorl %r10d, %r10d testl %edx, %edx cmovlel %r10d, %edx movq 0x40(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r11d cmovgl %ecx, %r11d movl 0xdc(%rax,%r9), %esi testl %esi, %esi cmovlel %r10d, %esi testl %r8d, %r8d cmovlel %r10d, %r8d movq 0x38(%rsp), %r9 imulq 0xb0(%rsp), %r9 imulq 0x30(%rsp), %r9 addq %r13, %r9 movq %rbp, %rdi cmpl %r8d, %r10d je 0x2b6a06 movl %edx, %ecx subl $0x1, %ecx jb 0x2b699a movq (%r12), %r14 movq %r14, (%rdi) addq $0x8, %rdi jmp 0x2b6988 xorl %r14d, %r14d cmpl %r14d, %r11d je 0x2b69b2 movq (%r9,%r14,8), %rcx movq %rcx, (%rdi) addq $0x8, %rdi incq %r14 jmp 0x2b699d movl %esi, %ecx subl $0x1, %ecx jb 0x2b69c7 movq -0x8(%r9,%r14,8), %r15 movq %r15, (%rdi) addq $0x8, %rdi jmp 0x2b69b4 incl %r10d jmp 0x2b697d imull %eax, %esi testl %esi, %esi movl $0x0, %eax cmovlel %eax, %esi imulq %r11, %r8 imulq %r12, %r8 addq %r8, %rdi xorl %eax, %eax movq %r12, %rcx movl %r13d, %r12d movq 0xc0(%rsp), %r13 cmpq %rax, %rsi je 0x2b6d3a movq %r10, (%rdi,%rax,8) incq %rax jmp 0x2b69f4 movslq 0xfc(%rsp), %rax xorl %r11d, %r11d testl %eax, %eax movl $0x0, %r8d movq %rax, 0x120(%rsp) cmovgl %eax, %r8d movq 0xb8(%rsp), %rax testl %eax, %eax movl $0x0, %r14d cmovgl %eax, %r14d movq %r12, %r9 cmpl %r14d, %r11d je 0x2b6a98 movl %edx, %ecx subl $0x1, %ecx jb 0x2b6a54 movq (%r9), %r10 movq %r10, (%rdi) addq $0x8, %rdi jmp 0x2b6a43 xorl %r15d, %r15d xorl %r10d, %r10d cmpl %r15d, %r8d je 0x2b6a70 movq (%r9,%r15,8), %rcx movq %rcx, (%rdi,%r15,8) addq $-0x8, %r10 incq %r15 jmp 0x2b6a5a movq %r9, %rcx subq %r10, %rcx subq %r10, %rdi movl %esi, %r10d subl $0x1, %r10d jb 0x2b6a90 movq -0x8(%r9,%r15,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b6a7c incl %r11d movq %rcx, %r9 jmp 0x2b6a3c movq 0x120(%rsp), %rax shlq $0x3, %rax subq %rax, %r9 xorl %r10d, %r10d movl 0xa8(%rsp), %r14d testl %r14d, %r14d cmovlel %r10d, %r14d cmpl %r14d, %r10d je 0x2b6b03 movl %edx, %ecx subl $0x1, %ecx jb 0x2b6ad1 movq (%r9), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b6ac0 xorl %r11d, %r11d cmpl %r11d, %r8d je 0x2b6ae9 movq (%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi incq %r11 jmp 0x2b6ad4 movl %esi, %ecx subl $0x1, %ecx jb 0x2b6afe movq -0x8(%r9,%r11,8), %rax movq %rax, (%rdi) addq $0x8, %rdi jmp 0x2b6aeb incl %r10d jmp 0x2b6ab9 movq 0x98(%rsp), %rax movq -0x18(%rax), %r9 movq 0x18(%rsp), %rax movl 0xe0(%rax,%r9), %eax cmpl $0x2, %eax jne 0x2b6d25 movq 0x18(%rsp), %rcx movl 0xd0(%rcx,%r9), %r14d movslq 0xd8(%rcx,%r9), %rax movq %rcx, %r8 movl %r14d, %r11d movq 0x40(%rsp), %rcx imull %ecx, %r11d movq %rcx, %rdx negq %rdx xorl %r15d, %r15d testl %eax, %eax movl $0x0, %esi cmovgl %eax, %esi testl %ecx, %ecx movl $0x0, %edi cmovgl %ecx, %edi movl 0xdc(%r8,%r9), %r8d testl %r8d, %r8d cmovlel %r15d, %r8d testl %r14d, %r14d cmovlel %r15d, %r14d movq 0x38(%rsp), %r10 imulq 0xb0(%rsp), %r10 movslq %r11d, %rcx leaq (%r12,%rcx,8), %r11 imulq 0x30(%rsp), %r10 leaq (%r10,%rax,8), %r10 leaq (%r10,%rcx,8), %r10 movq 0x18(%rsp), %r12 movl 0xd4(%r12,%r9), %r9d movl %r9d, 0x38(%rsp) shlq $0x3, %rax addq %r10, %r13 movq 0x40(%rsp), %rcx leaq (,%rcx,8), %r10 negq %r10 negq %rsi negq %r8 cmpl %r14d, %r15d je 0x2b6c2e xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b6bec movq (%r13,%rcx,8), %r12 movq %r12, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6bd5 xorl %ecx, %ecx movq %r11, %r12 cmpl %edi, %ecx je 0x2b6c09 movq (%r12), %r9 addq $0x8, %r12 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b6bf1 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b6c22 movq -0x10(%r12,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6c0b leaq (%r11,%rdx,8), %r11 incl %r15d addq %r10, %r13 jmp 0x2b6bce xorl %r14d, %r14d movq 0xb8(%rsp), %rcx testl %ecx, %ecx movl $0x0, %r15d cmovgl %ecx, %r15d movl 0x38(%rsp), %r13d cmpl %r15d, %r14d je 0x2b6ca3 leaq (%r11,%rax), %rcx xorl %r12d, %r12d cmpq %r12, %rsi je 0x2b6c6c movq (%rcx,%r12,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %r12 jmp 0x2b6c56 xorl %ecx, %ecx cmpl %edi, %ecx je 0x2b6c85 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b6c6e xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b6c9e movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6c87 incl %r14d jmp 0x2b6c4a movq 0x40(%rsp), %rcx addl %ecx, %ecx movslq %ecx, %rcx shlq $0x3, %rcx movq %r11, %r14 subq %rcx, %r14 xorl %r15d, %r15d testl %r13d, %r13d cmovlel %r15d, %r13d subq %rcx, %rax addq %r11, %rax cmpl %r13d, %r15d je 0x2b6d25 xorl %ecx, %ecx cmpq %rcx, %rsi je 0x2b6ce4 movq (%rax,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6cce xorl %ecx, %ecx movq %r14, %r11 cmpl %edi, %ecx je 0x2b6d00 movq (%r11), %r9 addq $0x8, %r11 movq %r9, (%rbp) addq $0x8, %rbp incl %ecx jmp 0x2b6ce9 xorl %ecx, %ecx cmpq %rcx, %r8 je 0x2b6d19 movq -0x10(%r11,%rcx,8), %r9 movq %r9, (%rbp) addq $0x8, %rbp decq %rcx jmp 0x2b6d02 leaq (%r14,%rdx,8), %r14 incl %r15d addq %r10, %rax jmp 0x2b6cc7 movq 0xc0(%rsp), %r13 movl 0xac(%rsp), %r12d movq 0x48(%rsp), %rcx incq %rcx movq 0x30(%rsp), %r10 incl %r10d jmp 0x2b66a1 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0xe0(%r14,%rax) jne 0x2b70ac vpextrd $0x3, %xmm0, %edi movl 0xe8(%r14,%rax), %ecx movl %edi, 0x48(%rsp) addl %edi, %ecx addl 0xec(%r14,%rax), %ecx addl 0xd0(%r14,%rax), %edx addl 0xd4(%r14,%rax), %edx addl 0xd8(%r14,%rax), %esi addl 0xdc(%r14,%rax), %esi movq 0x8(%r12), %rax movq %rax, 0x8(%rsp) movl $0x8, (%rsp) movq %rbx, %rdi movq %rcx, %r14 movl %r15d, %r8d callq 0x62a26 cmpq $0x0, (%rbx) je 0x2b7181 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b7181 xorl %ecx, %ecx testl %r14d, %r14d cmovlel %ecx, %r14d movq %r14, 0x30(%rsp) testl %r15d, %r15d movl $0x0, %eax movq %rax, 0x20(%rsp) cmovlel %ecx, %r15d movq %r15, 0xa0(%rsp) movabsq $0x100000001, %r13 # imm = 0x100000001 xorl %r14d, %r14d movq 0x18(%rsp), %rbp movq %rbx, 0xc8(%rsp) cmpq 0xa0(%rsp), %r14 je 0x2b7181 movq (%rbp), %rax movq -0x18(%rax), %rax vcvttss2si 0xe4(%rbp,%rax), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rsi shlq $0x20, %rsi orq %rcx, %rsi orq %rdx, %rsi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r15 shlq $0x38, %r15 orq %rdx, %r15 orq %rsi, %r15 orq %rax, %r15 xorl %r12d, %r12d cmpq 0x30(%rsp), %r12 je 0x2b7052 movslq 0x2c(%rbx), %rdi movslq 0x30(%rbx), %rdx movq (%rbx), %rax movq 0x10(%rbx), %rsi movq 0x40(%rbx), %r8 imulq %r14, %r8 movq %r8, %r9 imulq %rsi, %r9 addq %rax, %r9 movl 0x18(%rbx), %r10d movq %rdx, %rcx imulq %rdi, %rcx movq %r12, %r11 imulq %rsi, %r11 imulq %rcx, %r11 addq %r9, %r11 movq 0x20(%rbx), %r9 movq %r11, 0x50(%rsp) andq $0x0, 0x58(%rsp) movq %rsi, 0x60(%rsp) movl %r10d, 0x68(%rsp) movq %r9, 0x70(%rsp) movl $0x2, 0x78(%rsp) movl %edi, 0x7c(%rsp) movl %edx, 0x80(%rsp) movq %r13, 0x84(%rsp) movq %rcx, 0x90(%rsp) movq (%rbp), %r9 movq -0x18(%r9), %r9 movl 0xe8(%rbp,%r9), %r11d movl %r12d, %r10d subl %r11d, %r10d setns %r11b cmpl 0x48(%rsp), %r10d movq %r13, %rbx movq %rbp, %r13 setl %bpl testb %bpl, %r11b je 0x2b7016 movq 0x28(%rsp), %r8 movslq 0x2c(%r8), %rax movslq 0x30(%r8), %rcx movq 0x40(%r8), %rdx imulq %r14, %rdx movq 0x10(%r8), %rsi imulq %rsi, %rdx addq (%r8), %rdx movl 0x18(%r8), %edi movq 0x20(%r8), %r8 movq %rcx, %r11 imulq %rax, %r11 movl %r10d, %r10d movq %r11, %rbp imulq %rsi, %rbp imulq %r10, %rbp addq %rdx, %rbp movq %rbp, 0xd0(%rsp) andq $0x0, 0xd8(%rsp) movq %rsi, 0xe0(%rsp) movl %edi, 0xe8(%rsp) movq %r8, 0xf0(%rsp) movl $0x2, 0xf8(%rsp) movl %eax, 0xfc(%rsp) movl %ecx, 0x100(%rsp) movq %rbx, 0x104(%rsp) movq %r11, 0x110(%rsp) movl 0xd0(%r13,%r9), %edx movl 0xd4(%r13,%r9), %ecx movl 0xd8(%r13,%r9), %r8d movl 0xdc(%r13,%r9), %r9d movq %r15, (%rsp) leaq 0xd0(%rsp), %rdi leaq 0x50(%rsp), %rsi callq 0x2b740e movq %r13, %rbp movq %rbx, %r13 movq 0xc8(%rsp), %rbx incq %r12 jmp 0x2b6e81 testl %ecx, %ecx movl $0x0, %r9d cmovlel %r9d, %ecx imulq %r12, %rdx imulq %rdi, %rdx addq %r8, %rdx imulq %rdx, %rsi addq %rsi, %rax xorl %edx, %edx movq %r13, %rbp movq %rbx, %r13 movq 0xc8(%rsp), %rbx cmpq %rdx, %rcx je 0x2b700e movq %r15, (%rax,%rdx,8) incq %rdx jmp 0x2b7044 incq %r14 jmp 0x2b6e18 movq (%r14), %rax movq -0x18(%rax), %r8 movl 0xd0(%r14,%r8), %r10d leal (%r10,%rdx,8), %edx addl 0xd4(%r14,%r8), %edx xorl %r11d, %r11d testb $0x7, %dl sete %r11b movq %r9, %rax shrq $0x3, %rax leal (%r11,%r11,2), %ecx shlq %cl, %rax testb $0x7, %r10b sete %cl andb %cl, %r11b cmpb $0x1, %r11b jne 0x2b70ac cmpl $0x0, 0xe0(%r14,%r8) je 0x2b7198 movq 0x28(%rsp), %rdx movq 0x8(%rdx), %rax vmovups (%rdx), %xmm1 vmovaps %xmm1, 0x50(%rsp) movq %r9, 0x60(%rsp) movl %edi, 0x68(%rsp) movq 0x20(%rdx), %rcx movq %rcx, 0x70(%rsp) vmovdqu %xmm0, 0x78(%rsp) movl %r15d, 0x88(%rsp) movq 0x40(%rdx), %rcx movq %rcx, 0x90(%rsp) testq %rax, %rax je 0x2b70f3 lock incl (%rax) cmpl $0x1, %edi je 0x2b7134 vmovdqu (%r12), %ymm0 vmovups 0x20(%r12), %ymm1 leaq 0xd0(%rsp), %rcx vmovups %ymm1, 0x20(%rcx) vmovdqu %ymm0, (%rcx) movq 0x10(%r12), %rax movq %rax, 0x8(%rcx) leaq 0x50(%rsp), %rsi pushq $0x1 popq %rdx movq 0x28(%rsp), %rdi vzeroupper callq 0x64e3b movq (%r14), %rax addq -0x18(%rax), %r14 leaq 0x50(%rsp), %rsi movq %r14, %rdi movq %rbx, %rdx movq %r12, %rcx callq 0x2acc2e movq %rax, 0x20(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b7181 lock decl (%rax) jne 0x2b7181 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2b7179 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2b7181 movq %rsi, %rdi callq 0x5f3e0 movq 0x20(%rsp), %rax addq $0x128, %rsp # imm = 0x128 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq addl 0xd8(%r14,%r8), %esi addl 0xdc(%r14,%r8), %esi sarl $0x3, %edx movq 0x8(%r12), %r9 pushq $0x8 popq %rbp movq %rbx, %rdi movq %rax, %rcx movl %ebp, %r8d callq 0x627de cmpq $0x0, (%rbx) je 0x2b7181 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax testq %rax, %rax je 0x2b7181 movq 0x18(%rsp), %r9 movq (%r9), %rax movq -0x18(%rax), %rsi vcvttss2si 0xe4(%r9,%rsi), %rax movq %rax, %rcx shlq $0x8, %rcx movq %rax, %rdx shlq $0x10, %rdx orq %rcx, %rdx movq %rax, %rcx shlq $0x18, %rcx movq %rax, %rdi shlq $0x20, %rdi orq %rcx, %rdi orq %rdx, %rdi movq %rax, %rcx shlq $0x28, %rcx movq %rax, %rdx shlq $0x30, %rdx orq %rcx, %rdx movq %rax, %r11 shlq $0x38, %r11 orq %rdx, %r11 orq %rdi, %r11 orq %rax, %r11 movl 0xd0(%r9,%rsi), %eax movl 0xd4(%r9,%rsi), %ecx cltd idivl %ebp movl %eax, %r10d movl %ecx, %eax cltd idivl %ebp movl 0xd8(%r9,%rsi), %r8d movl 0xdc(%r9,%rsi), %r9d movq %r11, (%rsp) movq 0x28(%rsp), %rdi movq %rbx, %rsi movl %r10d, %edx movl %eax, %ecx callq 0x2b740e movq $0x0, 0x20(%rsp) jmp 0x2b7181 jmp 0x2b72bd jmp 0x2b7284 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2b72b5 lock decl (%rax) jne 0x2b72b5 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2b72af movq %rsi, %rdi callq 0x5f3e0 jmp 0x2b72b5 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_fma.cpp

virtual thunk to ncnn::Padding_x86_fma::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Padding_x86_fma::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { if (top == 0 && bottom == 0 && left == 0 && right == 0 && front == 0 && behind == 0) { top_blob = bottom_blob; return 0; } int elembits = bottom_blob.elembits(); if (elembits == 8) return forward_int8(bottom_blob, top_blob, opt); int w = bottom_blob.w; int h = bottom_blob.h; int d = bottom_blob.d; int channels = bottom_blob.c; int dims = bottom_blob.dims; size_t elemsize = bottom_blob.elemsize; int elempack = bottom_blob.elempack; #if __SSE2__ #if __AVX__ #if __AVX512F__ if (elempack == 16) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 16 == 0 ? 16 : outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, 0, 0, left / 16, right / 16, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 16 == 0 ? 16 : outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 16 == 0 && out_elempack == 16 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m512 pad_value = _mm512_set1_ps(value); padding_constant_pack16_avx512(bottom_blob, top_blob, top / 16, bottom / 16, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 16 == 0 ? 16 : outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 16 == 0 && out_elempack == 16 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack16_avx512(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack16_avx512(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m512 pad_value = per_channel_pad_data_size ? _mm512_loadu_ps((const float*)per_channel_pad_data + q * 16) : _mm512_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack16_avx512(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX512F__ if (elempack == 8) { if (dims == 1) { int outw = w * elempack + left + right; int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, 0, 0, left / 8, right / 8, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 8 == 0 && out_elempack == 8 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m256 pad_value = _mm256_set1_ps(value); padding_constant_pack8_avx(bottom_blob, top_blob, top / 8, bottom / 8, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 8 == 0 && out_elempack == 8 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack8_avx(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack8_avx(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m256 pad_value = per_channel_pad_data_size ? _mm256_loadu_ps((const float*)per_channel_pad_data + q * 8) : _mm256_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack8_avx(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __AVX__ if (elempack == 4) { if (dims == 1) { int outw = w * elempack + left + right; #if __AVX__ int out_elempack = outw % 8 == 0 ? 8 : outw % 4 == 0 ? 4 : 1; #else int out_elempack = outw % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (left % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, 0, 0, left / 4, right / 4, pad_value); return 0; } } if (dims == 2) { int outw = w + left + right; int outh = h * elempack + top + bottom; #if __AVX__ int out_elempack = outh % 8 == 0 ? 8 : outh % 4 == 0 ? 4 : 1; #else int out_elempack = outh % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (top % 4 == 0 && out_elempack == 4 && type == 0) { top_blob.create(outw, outh / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; __m128 pad_value = _mm_set1_ps(value); padding_constant_pack4_sse(bottom_blob, top_blob, top / 4, bottom / 4, left, right, pad_value); return 0; } } if (dims == 3) { int outw = w + left + right; int outh = h + top + bottom; int outc = channels * elempack + front + behind; #if __AVX__ int out_elempack = outc % 8 == 0 ? 8 : outc % 4 == 0 ? 4 : 1; #else int out_elempack = outc % 4 == 0 ? 4 : 1; #endif size_t out_elemsize = elemsize / elempack * out_elempack; if (front % 4 == 0 && out_elempack == 4 && !(outc != channels * elempack && type != 0)) { top_blob.create(outw, outh, outc / out_elempack, out_elemsize, out_elempack, opt.blob_allocator); if (top_blob.empty()) return -100; int front_ = front / elempack; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < outc / out_elempack; q++) { Mat borderm = top_blob.channel(q); __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); //Channel padding if ((q - front_) < 0 || (q - front_) >= channels) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q - front_); if (type == 0) padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); if (type == 1) padding_replicate_pack4_sse(m, borderm, top, bottom, left, right); if (type == 2) padding_reflect_pack4_sse(m, borderm, top, bottom, left, right); } } return 0; } } if (dims == 4) { int outw = w + left + right; int outh = h + top + bottom; int outd = d + front + behind; if (type == 0) { top_blob.create(outw, outh, outd, channels, elemsize, elempack, opt.blob_allocator); if (top_blob.empty()) return -100; #pragma omp parallel for num_threads(opt.num_threads) for (int q = 0; q < channels; q++) { __m128 pad_value = per_channel_pad_data_size ? _mm_loadu_ps((const float*)per_channel_pad_data + q * 4) : _mm_set1_ps(value); for (int z = 0; z < outd; z++) { Mat borderm = top_blob.channel(q).depth(z); // depth padding if ((z - front) < 0 || (z - front) >= d) { borderm.fill(pad_value); } else { const Mat m = bottom_blob.channel(q).depth(z - front); padding_constant_pack4_sse(m, borderm, top, bottom, left, right, pad_value); } } } return 0; } } } #endif // __SSE2__ Mat bottom_blob_unpacked = bottom_blob; if (elempack != 1) { Option opt_pack1 = opt; opt_pack1.blob_allocator = opt.workspace_allocator; convert_packing(bottom_blob, bottom_blob_unpacked, 1, opt_pack1); } return Padding::forward(bottom_blob_unpacked, top_blob, opt); }

movq (%rdi), %rax addq -0x48(%rax), %rdi jmp 0x2b4800

/csukuangfj[P]ncnn/build_O2/src/layer/x86/padding_x86_fma.cpp

ncnn::padding_constant_pack8_int8_sse(ncnn::Mat const&, ncnn::Mat&, int, int, int, int, long)

static void padding_constant_pack8_int8_sse(const Mat& src, Mat& dst, int top, int bottom, int left, int right, int64_t _v) { const int64_t* ptr = src; int64_t* outptr = dst; // fill top for (int y = 0; y < top; y++) { for (int x = 0; x < dst.w; x++) { *outptr++ = _v; } } // fill center for (int y = 0; y < src.h; y++) { for (int x = 0; x < left; x++) { *outptr++ = _v; } for (int x = 0; x < src.w; x++) { *outptr++ = *ptr++; } for (int x = 0; x < right; x++) { *outptr++ = _v; } } // fill bottom for (int y = 0; y < bottom; y++) { for (int x = 0; x < dst.w; x++) { *outptr++ = _v; } } }

pushq %rbp pushq %r15 pushq %r14 pushq %r12 pushq %rbx movq 0x30(%rsp), %rax movq (%rdi), %r11 movl 0x2c(%rsi), %ebx xorl %ebp, %ebp testl %ebx, %ebx cmovlel %ebp, %ebx movq (%rsi), %r10 testl %edx, %edx cmovlel %ebp, %edx cmpl %edx, %ebp je 0x2b744a movl %ebx, %r14d subl $0x1, %r14d jb 0x2b7446 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b7437 incl %ebp jmp 0x2b7430 xorl %edx, %edx testl %r8d, %r8d cmovlel %edx, %r8d movl 0x2c(%rdi), %ebx testl %ebx, %ebx cmovlel %edx, %ebx testl %r9d, %r9d cmovlel %edx, %r9d movl 0x30(%rdi), %edi testl %edi, %edi cmovlel %edx, %edi cmpl %edi, %edx je 0x2b74b6 movl %r8d, %ebp subl $0x1, %ebp jb 0x2b747f movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b7471 xorl %r15d, %r15d xorl %r14d, %r14d cmpl %r15d, %ebx je 0x2b749b movq (%r11,%r15,8), %r12 movq %r12, (%r10,%r15,8) addq $-0x8, %r14 incq %r15 jmp 0x2b7485 subq %r14, %r11 subq %r14, %r10 movl %r9d, %ebp subl $0x1, %ebp jb 0x2b74b2 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b74a4 incl %edx jmp 0x2b746a movl 0x2c(%rsi), %edx xorl %esi, %esi testl %edx, %edx cmovlel %esi, %edx testl %ecx, %ecx cmovlel %esi, %ecx movl %edx, %edi cmpl %ecx, %esi je 0x2b74dd subl $0x1, %edi jb 0x2b74d9 movq %rax, (%r10) addq $0x8, %r10 jmp 0x2b74cb incl %esi jmp 0x2b74c5 popq %rbx popq %r12 popq %r14 popq %r15 popq %rbp retq nop

/csukuangfj[P]ncnn/src/layer/x86/padding_pack8_int8.h

ncnn::ExpandDims::load_param(ncnn::ParamDict const&)

int ExpandDims::load_param(const ParamDict& pd) { expand_w = pd.get(0, 0); expand_h = pd.get(1, 0); expand_d = pd.get(11, 0); expand_c = pd.get(2, 0); axes = pd.get(3, Mat()); return 0; }

pushq %r15 pushq %r14 pushq %rbx subq $0xa0, %rsp movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd4(%rbx) pushq $0xb popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xdc(%rbx) leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) xorps %xmm0, %xmm0 movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 pushq $0x3 popq %rdx movq %r15, %rdi movq %r14, %rsi callq 0x718da leaq 0xe0(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2bc8e4 testq %rax, %rax je 0x2bc855 lock incl (%rax) movq 0xe8(%rbx), %rax testq %rax, %rax je 0x2bc889 lock decl (%rax) jne 0x2bc889 movq 0xe0(%rbx), %rsi movq 0x100(%rbx), %rdi testq %rdi, %rdi je 0x2bc881 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2bc889 movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0xe0(%rbx) movq 0x8(%rsp), %rax movq %rax, 0xe8(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0xf0(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0xf8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x100(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x108(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x118(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x120(%rbx) testq %rax, %rax je 0x2bc90c lock decl (%rax) jne 0x2bc90c movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2bc904 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2bc90c movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2bc956 lock decl (%rax) jne 0x2bc956 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2bc94e movq (%rdi), %rax callq *0x18(%rax) jmp 0x2bc956 movq %rsi, %rdi callq 0x5f3e0 xorl %eax, %eax addq $0xa0, %rsp popq %rbx popq %r14 popq %r15 retq movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2bc995 lock decl (%rax) jne 0x2bc995 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2bc98f movq %rsi, %rdi callq 0x5f3e0 jmp 0x2bc995 movq (%rdi), %rax callq *0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2bc9bc jmp 0x2bc9f2 jmp 0x2bc9f2 jmp 0x2bc9f2 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2bc9ea lock decl (%rax) jne 0x2bc9ea movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2bc9e4 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2bc9ea movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68

/csukuangfj[P]ncnn/src/layer/expanddims.cpp

ncnn::PriorBox::load_param(ncnn::ParamDict const&)

int PriorBox::load_param(const ParamDict& pd) { min_sizes = pd.get(0, Mat()); max_sizes = pd.get(1, Mat()); aspect_ratios = pd.get(2, Mat()); variances[0] = pd.get(3, 0.1f); variances[1] = pd.get(4, 0.1f); variances[2] = pd.get(5, 0.2f); variances[3] = pd.get(6, 0.2f); flip = pd.get(7, 1); clip = pd.get(8, 0); image_width = pd.get(9, 0); image_height = pd.get(10, 0); step_width = pd.get(11, -233.f); step_height = pd.get(12, -233.f); offset = pd.get(13, 0.f); step_mmdetection = pd.get(14, 0); center_mmdetection = pd.get(15, 0); return 0; }

pushq %r15 pushq %r14 pushq %rbx subq $0xa0, %rsp movq %rsi, %r14 movq %rdi, %rbx leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) xorps %xmm0, %xmm0 movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 movq %r15, %rdi xorl %edx, %edx callq 0x718da leaq 0xd0(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2c0907 testq %rax, %rax je 0x2c0878 lock incl (%rax) movq 0xd8(%rbx), %rax testq %rax, %rax je 0x2c08ac lock decl (%rax) jne 0x2c08ac movq 0xd0(%rbx), %rsi movq 0xf0(%rbx), %rdi testq %rdi, %rdi je 0x2c08a4 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c08ac movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0xd0(%rbx) movq 0x8(%rsp), %rax movq %rax, 0xd8(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0xe0(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0xe8(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0xf0(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0xf8(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x108(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x110(%rbx) testq %rax, %rax je 0x2c092f lock decl (%rax) jne 0x2c092f movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2c0927 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c092f movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c097f lock decl (%rax) jne 0x2c097f movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0974 movq (%rdi), %rax callq *0x18(%rax) xorps %xmm0, %xmm0 jmp 0x2c097f movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 pushq $0x1 popq %rdx movq %r15, %rdi movq %r14, %rsi callq 0x718da leaq 0x118(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2c0a55 testq %rax, %rax je 0x2c09c6 lock incl (%rax) movq 0x120(%rbx), %rax testq %rax, %rax je 0x2c09fa lock decl (%rax) jne 0x2c09fa movq 0x118(%rbx), %rsi movq 0x138(%rbx), %rdi testq %rdi, %rdi je 0x2c09f2 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c09fa movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0x118(%rbx) movq 0x8(%rsp), %rax movq %rax, 0x120(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0x128(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0x130(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x138(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x140(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x150(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x158(%rbx) testq %rax, %rax je 0x2c0a7d lock decl (%rax) jne 0x2c0a7d movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2c0a75 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c0a7d movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0acd lock decl (%rax) jne 0x2c0acd movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0ac2 movq (%rdi), %rax callq *0x18(%rax) xorps %xmm0, %xmm0 jmp 0x2c0acd movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 pushq $0x2 popq %rdx movq %r15, %rdi movq %r14, %rsi callq 0x718da leaq 0x160(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2c0ba3 testq %rax, %rax je 0x2c0b14 lock incl (%rax) movq 0x168(%rbx), %rax testq %rax, %rax je 0x2c0b48 lock decl (%rax) jne 0x2c0b48 movq 0x160(%rbx), %rsi movq 0x180(%rbx), %rdi testq %rdi, %rdi je 0x2c0b40 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c0b48 movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0x160(%rbx) movq 0x8(%rsp), %rax movq %rax, 0x168(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0x170(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0x178(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x180(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x188(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x198(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x1a0(%rbx) testq %rax, %rax je 0x2c0bcb lock decl (%rax) jne 0x2c0bcb movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2c0bc3 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c0bcb movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0c15 lock decl (%rax) jne 0x2c0c15 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0c0d movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c0c15 movq %rsi, %rdi callq 0x5f3e0 pushq $0x3 popq %rsi movss 0x132fe8(%rip), %xmm0 # 0x3f3c08 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0x1a8(%rbx) pushq $0x4 popq %rsi movq %r14, %rdi movss 0x132fca(%rip), %xmm0 # 0x3f3c08 callq 0x718c0 movss %xmm0, 0x1ac(%rbx) pushq $0x5 popq %rsi movss 0x137c7e(%rip), %xmm0 # 0x3f88d4 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0x1b0(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi movss 0x137c60(%rip), %xmm0 # 0x3f88d4 callq 0x718c0 movss %xmm0, 0x1b4(%rbx) pushq $0x7 popq %rsi pushq $0x1 popq %rdx movq %r14, %rdi callq 0x718a6 movl %eax, 0x1b8(%rbx) pushq $0x8 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x1bc(%rbx) pushq $0x9 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x1c0(%rbx) pushq $0xa popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x1c4(%rbx) pushq $0xb popq %rsi movss 0x137bff(%rip), %xmm0 # 0x3f88d8 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0x1c8(%rbx) pushq $0xc popq %rsi movq %r14, %rdi movss 0x137be1(%rip), %xmm0 # 0x3f88d8 callq 0x718c0 movss %xmm0, 0x1cc(%rbx) pushq $0xd popq %rsi xorps %xmm0, %xmm0 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0x1d0(%rbx) pushq $0xe popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 testl %eax, %eax setne 0x1d4(%rbx) pushq $0xf popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 testl %eax, %eax setne 0x1d5(%rbx) xorl %eax, %eax addq $0xa0, %rsp popq %rbx popq %r14 popq %r15 retq movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2c0d85 lock decl (%rax) jne 0x2c0d85 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2c0d7f movq %rsi, %rdi callq 0x5f3e0 jmp 0x2c0d85 movq (%rdi), %rax callq *0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2c0e75 jmp 0x2c0f01 movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2c0ddb lock decl (%rax) jne 0x2c0ddb movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2c0dd5 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2c0ddb movq (%rdi), %rax callq *0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2c0ea0 jmp 0x2c0f01 movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2c0e31 lock decl (%rax) jne 0x2c0e31 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2c0e2b movq %rsi, %rdi callq 0x5f3e0 jmp 0x2c0e31 movq (%rdi), %rax callq *0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2c0ecb jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 jmp 0x2c0f01 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0ef1 lock decl (%rax) jne 0x2c0ef1 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0ee9 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c0ef1 jmp 0x2c0f01 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0ef1 lock decl (%rax) jne 0x2c0ef1 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2c0ee9 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c0ef1 jmp 0x2c0f01 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2c0ef1 lock decl (%rax) jne 0x2c0ef1 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2c0ef9 movq %rsi, %rdi callq 0x5f3e0 movq %rbx, %rdi callq 0x5f340 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c0ef1 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/priorbox.cpp

ncnn::DetectionOutput::load_param(ncnn::ParamDict const&)

int DetectionOutput::load_param(const ParamDict& pd) { num_class = pd.get(0, 0); nms_threshold = pd.get(1, 0.05f); nms_top_k = pd.get(2, 300); keep_top_k = pd.get(3, 100); confidence_threshold = pd.get(4, 0.5f); variances[0] = pd.get(5, 0.1f); variances[1] = pd.get(6, 0.1f); variances[2] = pd.get(7, 0.2f); variances[3] = pd.get(8, 0.2f); return 0; }

pushq %r14 pushq %rbx pushq %rax movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rsi movss 0x1371b7(%rip), %xmm0 # 0x3f88f0 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xd4(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi movl $0x12c, %edx # imm = 0x12C callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x3 popq %rsi pushq $0x64 popq %rdx movq %r14, %rdi callq 0x718a6 movl %eax, 0xdc(%rbx) pushq $0x4 popq %rsi movss 0x12c896(%rip), %xmm0 # 0x3ee014 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xe0(%rbx) pushq $0x5 popq %rsi movss 0x13246f(%rip), %xmm0 # 0x3f3c08 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xe4(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi movss 0x132451(%rip), %xmm0 # 0x3f3c08 callq 0x718c0 movss %xmm0, 0xe8(%rbx) pushq $0x7 popq %rsi movss 0x137105(%rip), %xmm0 # 0x3f88d4 movq %r14, %rdi callq 0x718c0 movss %xmm0, 0xec(%rbx) pushq $0x8 popq %rsi movq %r14, %rdi movss 0x1370e7(%rip), %xmm0 # 0x3f88d4 callq 0x718c0 movss %xmm0, 0xf0(%rbx) xorl %eax, %eax addq $0x8, %rsp popq %rbx popq %r14 retq

/csukuangfj[P]ncnn/src/layer/detectionoutput.cpp

ncnn::Interp::load_param(ncnn::ParamDict const&)

int Interp::load_param(const ParamDict& pd) { resize_type = pd.get(0, 0); height_scale = pd.get(1, 1.f); width_scale = pd.get(2, 1.f); output_height = pd.get(3, 0); output_width = pd.get(4, 0); dynamic_target_size = pd.get(5, 0); align_corner = pd.get(6, 0); if (resize_type < 0 || resize_type > 3) { NCNN_LOGE("unsupported resize type %d", resize_type); return -1; } if (dynamic_target_size == 1) { one_blob_only = false; } return 0; }

pushq %r15 pushq %r14 pushq %rbx movq %rsi, %r15 movq %rdi, %r14 xorl %ebx, %ebx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%r14) pushq $0x1 popq %rsi movss 0x12bc91(%rip), %xmm0 # 0x3eec88 movq %r15, %rdi callq 0x718c0 movss %xmm0, 0xd8(%r14) pushq $0x2 popq %rsi movq %r15, %rdi movss 0x12bc72(%rip), %xmm0 # 0x3eec88 callq 0x718c0 movss %xmm0, 0xd4(%r14) pushq $0x3 popq %rsi movq %r15, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe0(%r14) pushq $0x4 popq %rsi movq %r15, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xdc(%r14) pushq $0x5 popq %rsi movq %r15, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe4(%r14) pushq $0x6 popq %rsi movq %r15, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xe8(%r14) movl 0xd0(%r14), %edx cmpl $0x4, %edx jae 0x2c3097 cmpl $0x1, 0xe4(%r14) jne 0x2c308f movb $0x0, 0x8(%r14) movl %ebx, %eax popq %rbx popq %r14 popq %r15 retq movq 0x1ccf3a(%rip), %rbx # 0x48ffd8 movq (%rbx), %rdi leaq 0x1358a8(%rip), %rsi # 0x3f8950 xorl %eax, %eax callq 0x5f150 movq (%rbx), %rsi pushq $0xa popq %rdi callq 0x5f1c0 pushq $-0x1 popq %rbx jmp 0x2c308f nop

/csukuangfj[P]ncnn/src/layer/interp.cpp

ncnn::Interp::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int Interp::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; int outw = output_width; int outh = output_height; if (bottom_blob.dims == 1) { w = 1; h = 1; } if (outw == 0 || outh == 0) { outw = static_cast<int>(w * width_scale); outh = static_cast<int>(h * height_scale); } Mat reference_blob; reference_blob.w = outw; reference_blob.h = outh; std::vector<Mat> bottom_blobs(2); bottom_blobs[0] = bottom_blob; bottom_blobs[1] = reference_blob; std::vector<Mat> top_blobs(1); int ret = forward(bottom_blobs, top_blobs, opt); top_blob = top_blobs[0]; return ret; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x58, %rsp movq %rcx, %r14 movq %rdx, %rbx movq %rsi, %r12 movq %rdi, %r15 movq 0xdc(%rdi), %xmm0 movdqa %xmm0, 0x40(%rsp) pshufd $0x50, %xmm0, %xmm0 # xmm0 = xmm0[0,0,1,1] pxor %xmm1, %xmm1 pcmpeqd %xmm0, %xmm1 movmskpd %xmm1, %eax testl %eax, %eax je 0x2c314f xorl %eax, %eax cmpl $0x1, 0x28(%r12) sete %al movsd 0x2c(%r12), %xmm0 movd %eax, %xmm1 pshufd $0x50, %xmm1, %xmm1 # xmm1 = xmm1[0,0,1,1] pslld $0x1f, %xmm1 psrad $0x1f, %xmm1 cvtdq2ps %xmm0, %xmm0 movdqa %xmm1, %xmm2 pandn %xmm0, %xmm2 pand 0x12af7b(%rip), %xmm1 # 0x3ee0b0 por %xmm2, %xmm1 movsd 0xd4(%r15), %xmm0 mulps %xmm1, %xmm0 cvttps2dq %xmm0, %xmm0 movdqa %xmm0, 0x40(%rsp) leaq 0x28(%rsp), %r13 pushq $0x2 popq %rsi leaq 0x10(%rsp), %rdx movq %r13, %rdi callq 0x6fa72 movq (%r13), %r13 cmpq %r12, %r13 je 0x2c3212 movq 0x8(%r12), %rax testq %rax, %rax je 0x2c317e lock incl (%rax) movq 0x8(%r13), %rax testq %rax, %rax je 0x2c31a9 lock decl (%rax) jne 0x2c31a9 movq (%r13), %rsi movq 0x20(%r13), %rdi testq %rdi, %rdi je 0x2c31a1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c31a9 movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%r13) pxor %xmm0, %xmm0 movdqu %xmm0, (%r13) movdqu %xmm0, 0xc(%r13) andl $0x0, 0x38(%r13) movdqu %xmm0, 0x28(%r13) movups (%r12), %xmm0 movups %xmm0, (%r13) movq 0x10(%r12), %rax movq %rax, 0x10(%r13) movl 0x18(%r12), %eax movl %eax, 0x18(%r13) movq 0x20(%r12), %rax movq %rax, 0x20(%r13) movdqu 0x28(%r12), %xmm0 movdqu %xmm0, 0x28(%r13) movl 0x38(%r12), %eax movl %eax, 0x38(%r13) movq 0x40(%r12), %rax movq %rax, 0x40(%r13) movq 0x28(%rsp), %r13 movq 0x50(%r13), %rax testq %rax, %rax je 0x2c323d lock decl (%rax) jne 0x2c323d movq 0x48(%r13), %rsi movq 0x68(%r13), %rdi testq %rdi, %rdi je 0x2c3235 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c323d movq %rsi, %rdi callq 0x5f3e0 leaq 0x48(%r13), %rax pxor %xmm0, %xmm0 movdqu %xmm0, 0xc(%rax) movdqu %xmm0, (%rax) movdqu %xmm0, 0x68(%r13) movdqa 0x40(%rsp), %xmm0 movq %xmm0, 0x74(%r13) andq $0x0, 0x7c(%r13) andq $0x0, 0x88(%r13) leaq 0x10(%rsp), %rdi pushq $0x1 popq %rsi leaq 0xf(%rsp), %rdx callq 0x6fa72 movq (%r15), %rax leaq 0x28(%rsp), %rsi leaq 0x10(%rsp), %rdx movq %r15, %rdi movq %r14, %rcx callq *0x30(%rax) movl %eax, %ebp movq 0x10(%rsp), %r14 cmpq %rbx, %r14 je 0x2c332e movq 0x8(%r14), %rax testq %rax, %rax je 0x2c32b1 lock incl (%rax) movq 0x8(%rbx), %rax testq %rax, %rax je 0x2c32db lock decl (%rax) jne 0x2c32db movq (%rbx), %rsi movq 0x20(%rbx), %rdi testq %rdi, %rdi je 0x2c32d3 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2c32db movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rbx) pxor %xmm0, %xmm0 movdqu %xmm0, (%rbx) movdqu %xmm0, 0xc(%rbx) andl $0x0, 0x38(%rbx) movdqu %xmm0, 0x28(%rbx) movups (%r14), %xmm0 movups %xmm0, (%rbx) movq 0x10(%r14), %rax movq %rax, 0x10(%rbx) movl 0x18(%r14), %eax movl %eax, 0x18(%rbx) movq 0x20(%r14), %rax movq %rax, 0x20(%rbx) movdqu 0x28(%r14), %xmm0 movdqu %xmm0, 0x28(%rbx) movl 0x38(%r14), %eax movl %eax, 0x38(%rbx) movq 0x40(%r14), %rax movq %rax, 0x40(%rbx) leaq 0x10(%rsp), %rdi callq 0x6fac4 leaq 0x28(%rsp), %rdi callq 0x6fac4 movl %ebp, %eax addq $0x58, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2c3355 movq %rax, %rbx jmp 0x2c3367 movq %rax, %rbx leaq 0x10(%rsp), %rdi callq 0x6fac4 leaq 0x28(%rsp), %rdi callq 0x6fac4 movq %rbx, %rdi callq 0x5f340 nop

/csukuangfj[P]ncnn/src/layer/interp.cpp

ncnn::linear_coeffs(int, int, int*, float*, int)

static void linear_coeffs(int w, int outw, int* xofs, float* alpha, int align_corner) { double scale = (double)w / outw; if (align_corner) { scale = (double)(w - 1) / (outw - 1); } for (int dx = 0; dx < outw; dx++) { float fx = (float)((dx + 0.5) * scale - 0.5); if (align_corner) { fx = (float)(dx * scale); } int sx = (int)floorf(fx); fx -= sx; if (sx < 0) { sx = 0; fx = 0.f; } if (sx >= w - 1) { sx = w - 2; fx = 1.f; } xofs[dx] = sx; alpha[dx * 2] = 1.f - fx; alpha[dx * 2 + 1] = fx; } }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x18, %rsp movl %r8d, %ebx movq %rcx, %r14 movq %rdx, 0x10(%rsp) movl %esi, %r12d movl %edi, %r13d testl %r8d, %r8d je 0x2c7078 leal -0x1(%r13), %ebp cvtsi2sd %ebp, %xmm1 leal -0x1(%r12), %eax cvtsi2sd %eax, %xmm0 divsd %xmm0, %xmm1 movsd %xmm1, (%rsp) jmp 0x2c708f cvtsi2sd %r13d, %xmm1 cvtsi2sd %r12d, %xmm0 divsd %xmm0, %xmm1 movsd %xmm1, (%rsp) leal -0x1(%r13), %ebp addl $-0x2, %r13d xorl %eax, %eax testl %r12d, %r12d cmovlel %eax, %r12d xorl %r15d, %r15d cmpq %r15, %r12 je 0x2c7140 cvtsi2sd %r15d, %xmm0 movapd %xmm0, %xmm1 addsd 0x12700f(%rip), %xmm1 # 0x3ee0c8 mulsd (%rsp), %xmm1 addsd 0x127342(%rip), %xmm1 # 0x3ee408 testl %ebx, %ebx je 0x2c70d3 mulsd (%rsp), %xmm0 movapd %xmm0, %xmm1 cvtsd2ss %xmm1, %xmm0 movss %xmm0, 0xc(%rsp) callq 0x5f360 cvttss2si %xmm0, %eax xorps %xmm1, %xmm1 testl %eax, %eax js 0x2c70fe cvttps2dq %xmm0, %xmm0 cvtdq2ps %xmm0, %xmm0 movss 0xc(%rsp), %xmm1 subss %xmm0, %xmm1 movl $0x0, %ecx cmovlel %ecx, %eax cmpl %ebp, %eax jl 0x2c7112 movss 0x127b76(%rip), %xmm1 # 0x3eec88 cmovgel %r13d, %eax movq 0x10(%rsp), %rcx movl %eax, (%rcx,%r15,4) movss 0x127b61(%rip), %xmm0 # 0x3eec88 subss %xmm1, %xmm0 movss %xmm0, (%r14,%r15,8) movss %xmm1, 0x4(%r14,%r15,8) incq %r15 jmp 0x2c709f addq $0x18, %rsp popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq

/csukuangfj[P]ncnn/src/layer/x86/interp_bilinear.h

ncnn::cubic_coeffs(int, int, int*, float*, int)

static void cubic_coeffs(int w, int outw, int* xofs, float* alpha, int align_corner) { double scale = (double)w / outw; if (align_corner) { scale = (double)(w - 1) / (outw - 1); } for (int dx = 0; dx < outw; dx++) { float fx = (float)((dx + 0.5) * scale - 0.5); if (align_corner) { fx = (float)(dx * scale); } int sx = static_cast<int>(floor(fx)); fx -= sx; interpolate_cubic(fx, alpha + dx * 4); if (sx <= -1) { sx = 1; alpha[dx * 4 + 0] = 1.f - alpha[dx * 4 + 3]; alpha[dx * 4 + 1] = alpha[dx * 4 + 3]; alpha[dx * 4 + 2] = 0.f; alpha[dx * 4 + 3] = 0.f; } if (sx == 0) { sx = 1; alpha[dx * 4 + 0] = alpha[dx * 4 + 0] + alpha[dx * 4 + 1]; alpha[dx * 4 + 1] = alpha[dx * 4 + 2]; alpha[dx * 4 + 2] = alpha[dx * 4 + 3]; alpha[dx * 4 + 3] = 0.f; } if (sx == w - 2) { sx = w - 3; alpha[dx * 4 + 3] = alpha[dx * 4 + 2] + alpha[dx * 4 + 3]; alpha[dx * 4 + 2] = alpha[dx * 4 + 1]; alpha[dx * 4 + 1] = alpha[dx * 4 + 0]; alpha[dx * 4 + 0] = 0.f; } if (sx >= w - 1) { sx = w - 3; alpha[dx * 4 + 3] = 1.f - alpha[dx * 4 + 0]; alpha[dx * 4 + 2] = alpha[dx * 4 + 0]; alpha[dx * 4 + 1] = 0.f; alpha[dx * 4 + 0] = 0.f; } xofs[dx] = sx; } }

testl %r8d, %r8d je 0x2d3720 leal -0x1(%rdi), %eax vcvtsi2sd %eax, %xmm0, %xmm0 leal -0x1(%rsi), %r9d vcvtsi2sd %r9d, %xmm1, %xmm1 vdivsd %xmm1, %xmm0, %xmm0 jmp 0x2d372f vcvtsi2sd %edi, %xmm0, %xmm0 vcvtsi2sd %esi, %xmm1, %xmm1 vdivsd %xmm1, %xmm0, %xmm0 leal -0x1(%rdi), %eax pushq %rbp pushq %r14 pushq %rbx leal -0x2(%rdi), %r9d addl $-0x3, %edi xorl %r10d, %r10d testl %esi, %esi cmovlel %r10d, %esi leaq 0xc(%rcx), %r11 vmovss 0x11b539(%rip), %xmm3 # 0x3eec88 vmovss 0x1251b9(%rip), %xmm4 # 0x3f8910 vmovss 0x124129(%rip), %xmm5 # 0x3f7888 vmovsd 0x1251b9(%rip), %xmm6 # 0x3f8920 vmovss 0x1251a5(%rip), %xmm7 # 0x3f8914 vbroadcastss 0x1251a0(%rip), %xmm8 # 0x3f8918 vbroadcastss 0x11b507(%rip), %xmm9 # 0x3eec88 vmovss 0x11ef9b(%rip), %xmm10 # 0x3f2724 vmovss 0x125187(%rip), %xmm11 # 0x3f8918 vbroadcastss 0x11da1a(%rip), %xmm12 # 0x3f11b4 movl $0x1, %ebx cmpq %r10, %rsi je 0x2d3901 vcvtsi2sd %r10d, %xmm15, %xmm13 vaddsd 0x11a913(%rip), %xmm13, %xmm14 # 0x3ee0c8 vmulsd %xmm0, %xmm14, %xmm14 vaddsd 0x11ac47(%rip), %xmm14, %xmm14 # 0x3ee408 testl %r8d, %r8d je 0x2d37ca vmulsd %xmm0, %xmm13, %xmm14 vcvtsd2ss %xmm14, %xmm14, %xmm13 vroundss $0x9, %xmm13, %xmm13, %xmm14 vcvttss2si %xmm14, %ebp vsubss %xmm14, %xmm13, %xmm13 vaddss %xmm3, %xmm13, %xmm14 vsubss %xmm13, %xmm3, %xmm15 vmulss %xmm4, %xmm14, %xmm1 vaddss %xmm5, %xmm1, %xmm1 vblendps $0xe, %xmm9, %xmm1, %xmm1 # xmm1 = xmm1[0],xmm9[1,2,3] vinsertps $0x10, %xmm13, %xmm14, %xmm13 # xmm13 = xmm14[0],xmm13[0],xmm14[2,3] vmulps %xmm13, %xmm13, %xmm14 vmulps %xmm6, %xmm13, %xmm13 vsubps %xmm13, %xmm7, %xmm2 vaddps %xmm8, %xmm13, %xmm13 vblendps $0x2, %xmm13, %xmm2, %xmm2 # xmm2 = xmm2[0],xmm13[1],xmm2[2,3] vmulps %xmm2, %xmm14, %xmm2 vaddps %xmm1, %xmm2, %xmm13 vmovlps %xmm13, -0xc(%r11) vmulss %xmm15, %xmm15, %xmm1 vmulss %xmm10, %xmm15, %xmm2 vaddss %xmm2, %xmm11, %xmm2 vmulss %xmm2, %xmm1, %xmm1 vaddss %xmm3, %xmm1, %xmm2 vmovss %xmm2, -0x4(%r11) vmovshdup %xmm13, %xmm2 # xmm2 = xmm13[1,1,3,3] vaddss %xmm1, %xmm13, %xmm1 vaddss %xmm2, %xmm1, %xmm1 vxorps %xmm1, %xmm12, %xmm1 vmovss %xmm1, (%r11) testl %ebp, %ebp js 0x2d3878 jne 0x2d38a8 vaddss -0x8(%r11), %xmm13, %xmm13 vmovss %xmm13, -0xc(%r11) vmovsd -0x4(%r11), %xmm1 vmovsd %xmm1, -0x8(%r11) movq %r11, %r14 jmp 0x2d38a2 leaq (,%r10,4), %r14 leaq (%rcx,%r14,4), %r14 addq $0xc, %r14 vmovss (%r11), %xmm1 vsubss %xmm1, %xmm3, %xmm13 vmovss %xmm13, -0xc(%r11) vmovss %xmm1, -0x8(%r11) andl $0x0, -0x4(%r11) andl $0x0, (%r14) movl %ebx, %ebp cmpl %r9d, %ebp jne 0x2d38d1 vmovss (%r11), %xmm1 vaddss -0x4(%r11), %xmm1, %xmm1 vmovss -0x8(%r11), %xmm2 vmovss %xmm1, (%r11) vmovss %xmm2, -0x4(%r11) vmovss %xmm13, -0x8(%r11) jmp 0x2d38ea cmpl %eax, %ebp jl 0x2d38f1 vsubss %xmm13, %xmm3, %xmm1 vmovss %xmm1, (%r11) vmovss %xmm13, -0x4(%r11) andl $0x0, -0x8(%r11) andl $0x0, -0xc(%r11) movl %edi, %ebp movl %ebp, (%rdx,%r10,4) incq %r10 addq $0x10, %r11 jmp 0x2d379f popq %rbx popq %r14 popq %rbp retq

/csukuangfj[P]ncnn/src/layer/x86/interp_bicubic.h

ncnn::DeconvolutionDepthWise::load_param(ncnn::ParamDict const&)

int DeconvolutionDepthWise::load_param(const ParamDict& pd) { num_output = pd.get(0, 0); kernel_w = pd.get(1, 0); kernel_h = pd.get(11, kernel_w); dilation_w = pd.get(2, 1); dilation_h = pd.get(12, dilation_w); stride_w = pd.get(3, 1); stride_h = pd.get(13, stride_w); pad_left = pd.get(4, 0); pad_right = pd.get(15, pad_left); pad_top = pd.get(14, pad_left); pad_bottom = pd.get(16, pad_top); output_pad_right = pd.get(18, 0); output_pad_bottom = pd.get(19, output_pad_right); output_w = pd.get(20, 0); output_h = pd.get(21, output_w); bias_term = pd.get(5, 0); weight_data_size = pd.get(6, 0); group = pd.get(7, 1); activation_type = pd.get(9, 0); activation_params = pd.get(10, Mat()); return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %rbx subq $0x98, %rsp movq %rsi, %r14 movq %rdi, %rbx movq %rsi, %rdi xorl %esi, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd0(%rbx) pushq $0x1 popq %rbp movq %r14, %rdi movl %ebp, %esi xorl %edx, %edx callq 0x718a6 movl %eax, 0xd4(%rbx) pushq $0xb popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xd8(%rbx) pushq $0x2 popq %rsi movq %r14, %rdi movl %ebp, %edx callq 0x718a6 movl %eax, 0xdc(%rbx) pushq $0xc popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xe0(%rbx) pushq $0x3 popq %rsi movq %r14, %rdi movl %ebp, %edx callq 0x718a6 movl %eax, 0xe4(%rbx) pushq $0xd popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xe8(%rbx) pushq $0x4 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xec(%rbx) pushq $0xf popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xf0(%rbx) movl 0xec(%rbx), %edx pushq $0xe popq %rsi movq %r14, %rdi callq 0x718a6 movl %eax, 0xf4(%rbx) pushq $0x10 popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0xf8(%rbx) pushq $0x12 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0xfc(%rbx) pushq $0x13 popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0x100(%rbx) pushq $0x14 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x104(%rbx) pushq $0x15 popq %rsi movq %r14, %rdi movl %eax, %edx callq 0x718a6 movl %eax, 0x108(%rbx) pushq $0x5 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x10c(%rbx) pushq $0x6 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x110(%rbx) pushq $0x7 popq %rsi movq %r14, %rdi movl %ebp, %edx callq 0x718a6 movl %eax, 0x114(%rbx) pushq $0x9 popq %rsi movq %r14, %rdi xorl %edx, %edx callq 0x718a6 movl %eax, 0x118(%rbx) leaq 0x50(%rsp), %rcx andq $0x0, 0x40(%rcx) xorps %xmm0, %xmm0 movaps %xmm0, (%rcx) movups %xmm0, 0xc(%rcx) movaps %xmm0, 0x20(%rcx) movups %xmm0, 0x2c(%rcx) movq %rsp, %r15 pushq $0xa popq %rdx movq %r15, %rdi movq %r14, %rsi callq 0x718da leaq 0x120(%rbx), %rcx movq 0x8(%rsp), %rax cmpq %r15, %rcx je 0x2d3bf6 testq %rax, %rax je 0x2d3b67 lock incl (%rax) movq 0x128(%rbx), %rax testq %rax, %rax je 0x2d3b9b lock decl (%rax) jne 0x2d3b9b movq 0x120(%rbx), %rsi movq 0x140(%rbx), %rdi testq %rdi, %rdi je 0x2d3b93 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d3b9b movq %rsi, %rdi callq 0x5f3e0 movq (%rsp), %rax movq %rax, 0x120(%rbx) movq 0x8(%rsp), %rax movq %rax, 0x128(%rbx) movq 0x10(%rsp), %rcx movq %rcx, 0x130(%rbx) movl 0x18(%rsp), %ecx movl %ecx, 0x138(%rbx) movq 0x20(%rsp), %rcx movq %rcx, 0x140(%rbx) movups 0x28(%rsp), %xmm0 movups %xmm0, 0x148(%rbx) movl 0x38(%rsp), %ecx movl %ecx, 0x158(%rbx) movq 0x40(%rsp), %rcx movq %rcx, 0x160(%rbx) testq %rax, %rax je 0x2d3c1e lock decl (%rax) jne 0x2d3c1e movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi je 0x2d3c16 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d3c1e movq %rsi, %rdi callq 0x5f3e0 andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) andl $0x0, 0x38(%rsp) movups %xmm0, 0x28(%rsp) movq 0x58(%rsp), %rax testq %rax, %rax je 0x2d3c68 lock decl (%rax) jne 0x2d3c68 movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi je 0x2d3c60 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d3c68 movq %rsi, %rdi callq 0x5f3e0 xorl %eax, %eax addq $0x98, %rsp popq %rbx popq %r14 popq %r15 popq %rbp retq movq %rax, %rbx movq 0x8(%rsp), %rax testq %rax, %rax je 0x2d3ca8 lock decl (%rax) jne 0x2d3ca8 movq (%rsp), %rsi movq 0x20(%rsp), %rdi testq %rdi, %rdi jne 0x2d3ca2 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d3ca8 movq (%rdi), %rax callq *0x18(%rax) andq $0x0, 0x40(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, (%rsp) movups %xmm0, 0xc(%rsp) movups %xmm0, 0x28(%rsp) andl $0x0, 0x38(%rsp) jmp 0x2d3ccf jmp 0x2d3d05 jmp 0x2d3d05 jmp 0x2d3d05 movq %rax, %rbx movq 0x58(%rsp), %rax testq %rax, %rax je 0x2d3cfd lock decl (%rax) jne 0x2d3cfd movq 0x50(%rsp), %rsi movq 0x70(%rsp), %rdi testq %rdi, %rdi jne 0x2d3cf7 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d3cfd movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/deconvolutiondepthwise.cpp

ncnn::DeconvolutionDepthWise::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const

int DeconvolutionDepthWise::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const { int w = bottom_blob.w; int h = bottom_blob.h; size_t elemsize = bottom_blob.elemsize; const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; const int kernel_extent_h = dilation_h * (kernel_h - 1) + 1; int outw = (w - 1) * stride_w + kernel_extent_w + output_pad_right; int outh = (h - 1) * stride_h + kernel_extent_h + output_pad_bottom; Mat top_blob_bordered; if (pad_left > 0 || pad_right > 0 || pad_top > 0 || pad_bottom > 0 || (output_w > 0 && output_h > 0)) { top_blob_bordered.create(outw, outh, num_output, elemsize, opt.workspace_allocator); } else { top_blob_bordered = top_blob; top_blob_bordered.create(outw, outh, num_output, elemsize, opt.blob_allocator); } if (top_blob_bordered.empty()) return -100; int ret = deconvolutiondepthwise(bottom_blob, top_blob_bordered, weight_data, bias_data, kernel_w, kernel_h, stride_w, stride_h, dilation_w, dilation_h, group, activation_type, activation_params, opt); if (ret != 0) return ret; cut_padding(top_blob_bordered, top_blob, opt); if (top_blob.empty()) return -100; return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x198, %rsp # imm = 0x198 movl 0x2c(%rsi), %r14d movl 0x30(%rsi), %ebp movl 0xd4(%rdi), %ebx decl %ebx imull 0xdc(%rdi), %ebx movq %rsi, 0x20(%rsp) movq 0x10(%rsi), %r8 decl %r14d imull 0xe4(%rdi), %r14d movl 0xd8(%rdi), %r12d decl %r12d imull 0xe0(%rdi), %r12d decl %ebp imull 0xe8(%rdi), %ebp movl 0xfc(%rdi), %esi movl 0x100(%rdi), %r10d andq $0x0, 0xb0(%rsp) movq %rcx, %r9 movq %rdx, %r13 pxor %xmm0, %xmm0 movdqa %xmm0, 0x70(%rsp) movdqu %xmm0, 0x7c(%rsp) movdqa %xmm0, 0x90(%rsp) movdqu %xmm0, 0x9c(%rsp) movdqu 0xec(%rdi), %xmm1 pcmpgtd %xmm0, %xmm1 movmskps %xmm1, %eax pushq $0x10 popq %r15 testl %eax, %eax movq %rdi, 0x18(%rsp) movq %rcx, 0x118(%rsp) jne 0x2d4172 cmpl $0x0, 0x104(%rdi) jle 0x2d409b movq 0x18(%rsp), %rax cmpl $0x0, 0x108(%rax) jg 0x2d4172 pushq $0x8 popq %r15 leaq 0x70(%rsp), %rax cmpq %r13, %rax je 0x2d4172 movq 0x8(%r13), %rax testq %rax, %rax je 0x2d4117 lock incl (%rax) movq 0x78(%rsp), %rax testq %rax, %rax je 0x2d4117 lock decl (%rax) jne 0x2d4117 movq %rsi, 0x8(%rsp) movq %r10, 0x58(%rsp) movq %r8, 0x10(%rsp) movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2d4100 movq (%rdi), %rax callq *0x18(%rax) movq 0x10(%rsp), %r8 movq 0x58(%rsp), %r10 movq 0x8(%rsp), %rsi jmp 0x2d4117 movq %rsi, %rdi callq 0x5f3e0 movq 0x10(%rsp), %r8 movq 0x58(%rsp), %r10 movq 0x8(%rsp), %rsi movups (%r13), %xmm0 movaps %xmm0, 0x70(%rsp) movq 0x10(%r13), %rax movq %rax, 0x80(%rsp) movl 0x18(%r13), %eax movl %eax, 0x88(%rsp) movq 0x20(%r13), %rax movq %rax, 0x90(%rsp) movdqu 0x28(%r13), %xmm0 movdqu %xmm0, 0x98(%rsp) movl 0x38(%r13), %eax movl %eax, 0xa8(%rsp) movq 0x40(%r13), %rax movq %rax, 0xb0(%rsp) movq 0x118(%rsp), %r9 addl %r14d, %ebx addl %ebp, %r12d addl %ebx, %esi incl %esi leal 0x1(%r10,%r12), %edx movq 0x18(%rsp), %rax movl 0xd0(%rax), %ecx movq (%r9,%r15), %r9 leaq 0x70(%rsp), %rdi callq 0x63810 pushq $-0x64 popq %rbx cmpq $0x0, 0x70(%rsp) je 0x2d4c36 movslq 0xa8(%rsp), %rcx movq 0xb0(%rsp), %rax imulq %rcx, %rax testq %rax, %rax je 0x2d4c36 movq %r13, 0x150(%rsp) movq %rcx, 0x40(%rsp) movq 0x18(%rsp), %rax movl 0xd4(%rax), %r14d movl 0xd8(%rax), %r13d movslq 0xe4(%rax), %rcx movq %rcx, 0x58(%rsp) movslq 0xe8(%rax), %rcx movq %rcx, 0x60(%rsp) movl 0xdc(%rax), %ebp movl 0xe0(%rax), %ebx movl 0x114(%rax), %ecx movq %rcx, 0xc0(%rsp) movl 0x118(%rax), %eax movq %rax, 0x128(%rsp) movq 0x20(%rsp), %rax movl 0x38(%rax), %eax movl %eax, 0x10(%rsp) movl 0x9c(%rsp), %r15d movl %r13d, %eax imull %r14d, %eax movslq %eax, %r12 leaq 0x180(%rsp), %rdi leaq 0x37(%rsp), %rdx movq %r12, %rsi callq 0x73bbe imull %ebx, %r15d movl %ebp, %eax imull %r14d, %eax subl %eax, %r15d xorl %eax, %eax testl %r14d, %r14d cmovlel %eax, %r14d testl %r13d, %r13d cmovlel %eax, %r13d movq 0x180(%rsp), %rbx movl %r14d, %ecx xorl %edx, %edx xorl %esi, %esi cmpl %r13d, %eax je 0x2d42ae movslq %ecx, %rdi movslq %esi, %r8 leal (%r8,%r14), %esi cmpq %r8, %rdi je 0x2d42a4 movl %edx, (%rbx,%r8,4) incq %r8 addl %ebp, %edx jmp 0x2d4294 addl %r15d, %edx incl %eax addl %r14d, %ecx jmp 0x2d4285 movl 0x10(%rsp), %eax movl %eax, %edx movq 0xc0(%rsp), %rsi xorl %esi, %edx movq 0x40(%rsp), %r8 movl %r8d, %ecx xorl %esi, %ecx orl %edx, %ecx jne 0x2d46d2 movq 0x20(%rsp), %rdi movq 0x10(%rdi), %rax imulq 0x40(%rdi), %rax movq %rax, 0x38(%rsp) movq 0x18(%rsp), %rsi movq 0x1b0(%rsi), %rcx movslq 0x9c(%rsp), %rax movslq 0xa0(%rsp), %r9 movq 0x80(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %r10, %r11 movq %r9, %rdx imulq %rax, %rdx movq %r10, %r14 movq %rdx, 0x100(%rsp) imulq %rdx, %r14 addq $0xf, %r14 andq $-0x10, %r14 movq %r14, 0xf8(%rsp) movq %rcx, 0xc8(%rsp) testq %rcx, %rcx sete %cl movslq 0x1e8(%rsi), %rdx imulq 0x1f0(%rsi), %rdx testq %rdx, %rdx sete %bpl movslq 0x2c(%rdi), %rdx imull %eax, %r9d xorl %r13d, %r13d testl %r12d, %r12d movl $0x0, %r14d cmovgl %r12d, %r14d testl %edx, %edx movl $0x0, %r15d movq %rdx, 0x50(%rsp) cmovgl %edx, %r15d movl 0x30(%rdi), %edx testl %edx, %edx cmovlel %r13d, %edx movq %rdx, 0x8(%rsp) movq (%rdi), %rdx movq %rdx, 0xe8(%rsp) testl %r9d, %r9d cmovlel %r13d, %r9d testl %r8d, %r8d cmovlel %r13d, %r8d movq 0x60(%rsp), %rdx imulq %rax, %rdx movq 0x168(%rsi), %r13 orb %cl, %bpl movb %bpl, 0xf0(%rsp) movq %r10, 0x130(%rsp) imulq %r10, %rdx movq %rdx, 0x60(%rsp) shlq $0x2, %r12 movl 0xa4(%rsp), %eax movl %eax, 0xe0(%rsp) movq 0x70(%rsp), %rax movl 0x98(%rsp), %edx movl %edx, 0xd0(%rsp) movq %rax, 0xd8(%rsp) movq %rax, %rdi xorl %esi, %esi movq %r8, 0x40(%rsp) movq %r9, 0x48(%rsp) movq %r11, 0x68(%rsp) cmpq %r8, %rsi je 0x2d4bee movq 0xf8(%rsp), %rax xorl %edx, %edx divq 0x130(%rsp) cmpl $0x4, 0xd0(%rsp) cmoveq 0x100(%rsp), %rax pxor %xmm0, %xmm0 cmpb $0x0, 0xf0(%rsp) jne 0x2d4461 movq 0xc8(%rsp), %rcx movd (%rcx,%rsi,4), %xmm0 movq 0x38(%rsp), %rcx imulq %rsi, %rcx addq 0xe8(%rsp), %rcx movq %r11, %rdx movq %rsi, 0x28(%rsp) imulq %rsi, %rdx addq 0xd8(%rsp), %rdx imull 0xe0(%rsp), %eax testl %eax, %eax movl $0x0, %esi cmovlel %esi, %eax xorl %esi, %esi movq 0x18(%rsp), %rbp cmpl %esi, %eax je 0x2d44ad movd %xmm0, (%rdi,%rsi,4) incq %rsi jmp 0x2d449f xorl %eax, %eax movq %rdi, 0x10(%rsp) cmpq 0x8(%rsp), %rax je 0x2d4526 movq 0x60(%rsp), %rsi imulq %rax, %rsi addq %rdx, %rsi movq %rax, %rdi imulq 0x50(%rsp), %rdi leaq (%rcx,%rdi,4), %rdi xorl %r8d, %r8d cmpq %r15, %r8 je 0x2d451c movq %r8, %r9 imulq 0x58(%rsp), %r9 leaq (%rsi,%r9,4), %r9 movss (%rdi,%r8,4), %xmm0 xorl %r10d, %r10d cmpq %r10, %r14 je 0x2d4517 movss (%r13,%r10,4), %xmm1 mulss %xmm0, %xmm1 movslq (%rbx,%r10,4), %r11 addss (%r9,%r11,4), %xmm1 movss %xmm1, (%r9,%r11,4) incq %r10 jmp 0x2d44f2 incq %r8 jmp 0x2d44d7 incq %rax movq 0x10(%rsp), %rdi jmp 0x2d44b4 movq 0x120(%rbp), %rax movq %rax, 0x20(%rsp) xorl %ebp, %ebp movq 0x48(%rsp), %rcx leaq 0x1244b4(%rip), %rdx # 0x3f89f4 cmpq %rbp, %rcx je 0x2d46b5 movss (%rdi,%rbp,4), %xmm4 movq 0x128(%rsp), %rax decl %eax cmpl $0x5, %eax ja 0x2d456e movslq (%rdx,%rax,4), %rax addq %rdx, %rax jmpq *%rax maxss 0x119aa2(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2d46a8 movaps %xmm4, %xmm0 movaps %xmm4, 0x140(%rsp) callq 0x5f410 addss 0x11a6fa(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x10(%rsp), %rdi leaq 0x124450(%rip), %rdx # 0x3f89f4 movq 0x48(%rsp), %rcx mulss 0x140(%rsp), %xmm0 jmp 0x2d46a8 movq 0x20(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2d46a5 jmp 0x2d46a8 movss 0x11cbda(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x119aa4(%rip), %xmm0 # 0x3ee090 cmpltss 0x11cbc7(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x10(%rsp), %rdi leaq 0x1243df(%rip), %rdx # 0x3f89f4 movq 0x48(%rsp), %rcx movaps %xmm0, %xmm1 movss 0x11a663(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2d46a8 pxor %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x11a645(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x20(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2d46a8 movq 0x20(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x119a1d(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 pxor %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2d46a8 movss 0x11a600(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x2d456e mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 movd %xmm0, (%rdi,%rbp,4) incq %rbp jmp 0x2d4540 movq 0x28(%rsp), %rsi incq %rsi movq 0x68(%rsp), %r11 addq %r11, %rdi addq %r12, %r13 movq 0x40(%rsp), %r8 jmp 0x2d441a cltd idivl %esi movl %eax, %ecx movl %r8d, %eax cltd idivl %esi movl %ecx, %edx imull %r12d, %edx xorl %edi, %edi testl %r12d, %r12d movl $0x0, %ebp cmovgl %r12d, %ebp testl %ecx, %ecx movl $0x0, %r15d cmovgl %ecx, %r15d testl %eax, %eax movl $0x0, %r8d cmovgl %eax, %r8d movq %r8, 0x100(%rsp) movslq %edx, %rdx movslq %ecx, %rcx movq %rcx, 0x160(%rsp) cltq testl %esi, %esi cmovlel %edi, %esi movl %eax, %ecx imull %edx, %ecx movl %ecx, 0x10c(%rsp) shlq $0x2, %rdx movq %rdx, 0x178(%rsp) shlq $0x2, %r12 xorl %ecx, %ecx movq $0x0, 0x120(%rsp) xorl %r14d, %r14d movq %rsi, 0xc0(%rsp) movq %rax, 0x158(%rsp) cmpq %rsi, %r14 je 0x2d4bee movl %ecx, 0x110(%rsp) movslq %ecx, %rdi shlq $0x2, %rdi movq %r14, %rcx imulq %rax, %rcx movq %rcx, 0xf8(%rsp) movslq 0x9c(%rsp), %r8 movslq 0xa0(%rsp), %r9 movq 0x80(%rsp), %r10 movq 0xb0(%rsp), %r11 imulq %r10, %r11 movq %r9, %rax imulq %r8, %rax movq %r10, %rcx movq %rax, 0xe8(%rsp) imulq %rax, %rcx addq $0xf, %rcx andq $-0x10, %rcx movq %rcx, 0xe0(%rsp) movq 0x18(%rsp), %rdx movq 0x1b0(%rdx), %rax movq %rax, 0x170(%rsp) testq %rax, %rax sete 0x8(%rsp) movslq 0x1e8(%rdx), %rcx imulq 0x1f0(%rdx), %rcx testq %rcx, %rcx sete %r13b movq 0x20(%rsp), %rcx movl 0x2c(%rcx), %eax movq %r14, 0x168(%rsp) imulq 0x160(%rsp), %r14 imull %r8d, %r9d testl %eax, %eax movl $0x0, %esi cmovlel %esi, %eax movq %rax, 0x10(%rsp) movl 0x30(%rcx), %eax testl %eax, %eax cmovlel %esi, %eax movq %rax, 0x140(%rsp) testl %r9d, %r9d cmovlel %esi, %r9d movq %r9, 0x48(%rsp) movq 0x70(%rsp), %rax orb 0x8(%rsp), %r13b movb %r13b, 0xd8(%rsp) imulq 0x60(%rsp), %r8 movq %r10, 0xf0(%rsp) imulq %r10, %r8 movq %r8, 0x40(%rsp) movq %r11, 0x68(%rsp) imulq 0x120(%rsp), %r11 addq 0x168(%rdx), %rdi movq %rdi, 0x8(%rsp) movq %rax, 0xd0(%rsp) addq %rax, %r11 movl 0xa4(%rsp), %eax movl %eax, 0xc8(%rsp) movl 0x98(%rsp), %eax movl %eax, 0x114(%rsp) xorl %edx, %edx cmpq 0x100(%rsp), %rdx je 0x2d4bb8 movq 0xf8(%rsp), %rax movq %rdx, 0x38(%rsp) leaq (%rdx,%rax), %rsi movq 0xe0(%rsp), %rax xorl %edx, %edx divq 0xf0(%rsp) cmpl $0x4, 0x114(%rsp) cmoveq 0xe8(%rsp), %rax pxor %xmm0, %xmm0 cmpb $0x0, 0xd8(%rsp) jne 0x2d4916 movq 0x170(%rsp), %rcx movd (%rcx,%rsi,4), %xmm0 imulq 0x68(%rsp), %rsi addq 0xd0(%rsp), %rsi imull 0xc8(%rsp), %eax testl %eax, %eax movl $0x0, %ecx cmovlel %ecx, %eax xorl %ecx, %ecx cmpl %ecx, %eax je 0x2d4947 movd %xmm0, (%r11,%rcx,4) incq %rcx jmp 0x2d4938 xorl %eax, %eax movq %r11, 0x50(%rsp) movq %rsi, 0x28(%rsp) cmpq 0x140(%rsp), %rax je 0x2d4a0e movq 0x40(%rsp), %rdx imulq %rax, %rdx addq %rsi, %rdx xorl %esi, %esi cmpq 0x10(%rsp), %rsi je 0x2d49fc movq %rsi, %rcx imulq 0x58(%rsp), %rcx leaq (%rdx,%rcx,4), %rdi movq 0x20(%rsp), %r10 movslq 0x2c(%r10), %rcx movq 0x10(%r10), %r9 movq 0x40(%r10), %r8 imulq %rax, %rcx imulq %r9, %rcx addq (%r10), %rcx imulq %r9, %r8 leaq (%rcx,%rsi,4), %r9 movq 0x8(%rsp), %r10 xorl %r11d, %r11d cmpq %r15, %r11 je 0x2d49f4 leaq (%r11,%r14), %rcx imulq %r8, %rcx movss (%r9,%rcx), %xmm0 xorl %ecx, %ecx cmpq %rcx, %rbp je 0x2d49ec movss (%r10,%rcx,4), %xmm1 mulss %xmm0, %xmm1 movslq (%rbx,%rcx,4), %r13 addss (%rdi,%r13,4), %xmm1 movss %xmm1, (%rdi,%r13,4) incq %rcx jmp 0x2d49c8 incq %r11 addq %r12, %r10 jmp 0x2d49b3 incq %rsi jmp 0x2d496f incq %rax movq 0x50(%rsp), %r11 movq 0x28(%rsp), %rsi jmp 0x2d4953 movq 0x18(%rsp), %rax movq 0x120(%rax), %rax movq %rax, 0x28(%rsp) xorl %r13d, %r13d movq 0x38(%rsp), %rdx cmpq %r13, 0x48(%rsp) je 0x2d4b99 movss (%r11,%r13,4), %xmm4 movq 0x128(%rsp), %rax decl %eax cmpl $0x5, %eax ja 0x2d4a5f leaq 0x123f8e(%rip), %rcx # 0x3f89dc movslq (%rcx,%rax,4), %rax addq %rcx, %rax jmpq *%rax maxss 0x1195b1(%rip), %xmm4 # 0x3ee010 movaps %xmm4, %xmm0 jmp 0x2d4b8b movaps %xmm4, %xmm0 movaps %xmm4, 0x130(%rsp) callq 0x5f410 addss 0x11a209(%rip), %xmm0 # 0x3eec88 callq 0x5f200 callq 0x5f160 movq 0x38(%rsp), %rdx movq 0x50(%rsp), %r11 mulss 0x130(%rsp), %xmm0 jmp 0x2d4b8b movq 0x28(%rsp), %rax maxss (%rax), %xmm4 movss 0x4(%rax), %xmm1 ucomiss %xmm1, %xmm4 movaps %xmm4, %xmm0 ja 0x2d4b88 jmp 0x2d4b8b movss 0x11c6f0(%rip), %xmm2 # 0x3f11b8 minss %xmm2, %xmm4 movaps %xmm4, %xmm0 xorps 0x1195ba(%rip), %xmm0 # 0x3ee090 cmpltss 0x11c6dd(%rip), %xmm4 # 0x3f11bc movaps %xmm4, %xmm1 andnps %xmm0, %xmm1 andps %xmm2, %xmm4 orps %xmm1, %xmm4 movaps %xmm4, %xmm0 callq 0x5f410 movq 0x38(%rsp), %rdx movq 0x50(%rsp), %r11 movaps %xmm0, %xmm1 movss 0x11a180(%rip), %xmm0 # 0x3eec88 addss %xmm0, %xmm1 divss %xmm1, %xmm0 jmp 0x2d4b8b pxor %xmm0, %xmm0 cmpltss %xmm4, %xmm0 movaps %xmm0, %xmm1 movss 0x11a162(%rip), %xmm2 # 0x3eec88 andps %xmm2, %xmm1 movq 0x28(%rsp), %rax movss (%rax), %xmm2 andnps %xmm2, %xmm0 orps %xmm1, %xmm0 mulss %xmm4, %xmm0 jmp 0x2d4b8b movq 0x28(%rsp), %rax movss (%rax), %xmm1 movss 0x4(%rax), %xmm2 movaps %xmm2, %xmm3 xorps 0x11953a(%rip), %xmm3 # 0x3ee090 divss %xmm1, %xmm3 pxor %xmm0, %xmm0 ucomiss %xmm3, %xmm4 jb 0x2d4b8b movss 0x11a11d(%rip), %xmm0 # 0x3eec88 divss %xmm1, %xmm0 addss %xmm0, %xmm3 ucomiss %xmm3, %xmm4 ja 0x2d4a5f mulss %xmm4, %xmm1 addss %xmm2, %xmm1 mulss %xmm4, %xmm1 movaps %xmm1, %xmm0 movd %xmm0, (%r11,%r13,4) incq %r13 jmp 0x2d4a27 incq %rdx addq 0x68(%rsp), %r11 movq 0x8(%rsp), %rax addq 0x178(%rsp), %rax movq %rax, 0x8(%rsp) jmp 0x2d48b9 movq 0x168(%rsp), %r14 incq %r14 movq 0x158(%rsp), %rax addq %rax, 0x120(%rsp) movl 0x110(%rsp), %ecx addl 0x10c(%rsp), %ecx movq 0xc0(%rsp), %rsi jmp 0x2d4763 leaq 0x180(%rsp), %rdi callq 0x624be leaq 0x70(%rsp), %rsi movq 0x18(%rsp), %rdi movq 0x150(%rsp), %rbx movq %rbx, %rdx movq 0x118(%rsp), %rcx callq 0x2d4cc4 cmpq $0x0, (%rbx) je 0x2d4c33 movslq 0x38(%rbx), %rax imulq 0x40(%rbx), %rax xorl %ebx, %ebx testq %rax, %rax jne 0x2d4c36 pushq $-0x64 popq %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2d4c67 lock decl (%rax) jne 0x2d4c67 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi je 0x2d4c5f movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d4c67 movq %rsi, %rdi callq 0x5f3e0 movl %ebx, %eax addq $0x198, %rsp # imm = 0x198 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2d4cbb jmp 0x2d4c7f movq %rax, %rbx movq 0x78(%rsp), %rax testq %rax, %rax je 0x2d4cb3 lock decl (%rax) jne 0x2d4cb3 movq 0x70(%rsp), %rsi movq 0x90(%rsp), %rdi testq %rdi, %rdi jne 0x2d4cad movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d4cb3 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/deconvolutiondepthwise.cpp

ncnn::DeconvolutionDepthWise_x86::create_group_ops(ncnn::Option const&)

int DeconvolutionDepthWise_x86::create_group_ops(const Option& opt) { // create Deconvolution op for each group const int maxk = kernel_w * kernel_h; int channels = (weight_data_size / group) / maxk / (num_output / group) * group; for (int i = 0; i < (int)group_ops.size(); i++) delete group_ops[i]; group_ops.clear(); const int channels_g = channels / group; const int num_output_g = num_output / group; group_ops.resize(group); for (int g = 0; g < group; g++) { Mat weight_data_g = weight_data.range(maxk * channels_g * num_output_g * g, maxk * channels_g * num_output_g).clone(); Mat bias_data_g; if (bias_term) bias_data_g = bias_data.range(num_output_g * g, num_output_g); ncnn::Layer* op = ncnn::create_layer(ncnn::LayerType::Deconvolution); // set param ncnn::ParamDict pd; pd.set(0, num_output_g); // num_output pd.set(1, kernel_w); pd.set(11, kernel_h); pd.set(2, dilation_w); pd.set(12, dilation_h); pd.set(3, stride_w); pd.set(13, stride_h); pd.set(4, 0); // pad_w pd.set(14, 0); // pad_h pd.set(18, output_pad_right); pd.set(19, output_pad_bottom); pd.set(5, bias_term); pd.set(6, maxk * channels_g * num_output_g); // weight_data_size pd.set(9, activation_type); pd.set(10, activation_params); op->load_param(pd); // set weights if (bias_term) { ncnn::Mat weights[2]; weights[0] = weight_data_g; weights[1] = bias_data_g; op->load_model(ModelBinFromMatArray(weights)); } else { ncnn::Mat weights[1]; weights[0] = weight_data_g; op->load_model(ModelBinFromMatArray(weights)); } op->create_pipeline(opt); group_ops[g] = op; } return 0; }

pushq %rbp pushq %r15 pushq %r14 pushq %r13 pushq %r12 pushq %rbx subq $0x158, %rsp # imm = 0x158 movq %rsi, 0x120(%rsp) movq %rdi, %r14 movq (%rdi), %rax movq -0x18(%rax), %rdx movl 0xd0(%rdi,%rdx), %ecx movl 0xd8(%rdi,%rdx), %ebp imull 0xd4(%rdi,%rdx), %ebp movl 0x110(%rdi,%rdx), %eax movl 0x114(%rdi,%rdx), %r15d cltd idivl %r15d cltd idivl %ebp movl %eax, %esi movl %ecx, %eax cltd idivl %r15d movl %eax, %ecx movl %esi, %eax cltd idivl %ecx movl %eax, %ebx leaq 0x8(%rdi), %rax movq %rax, 0xa8(%rsp) xorl %r12d, %r12d movq 0x8(%r14), %rax movq 0x10(%r14), %rcx movq %rcx, %rdx subq %rax, %rdx shrq $0x3, %rdx movslq %edx, %rdx cmpq %rdx, %r12 jge 0x2d5365 movq (%rax,%r12,8), %rdi testq %rdi, %rdi je 0x2d5360 movq (%rdi), %rax callq *0x8(%rax) incq %r12 jmp 0x2d5337 imull %r15d, %ebx cmpq %rax, %rcx je 0x2d5372 movq %rax, 0x10(%r14) movq (%r14), %rax movq -0x18(%rax), %rcx movslq 0x114(%r14,%rcx), %rsi movl %ebx, %eax cltd idivl %esi movl %eax, %ebx movl 0xd0(%r14,%rcx), %eax cltd idivl %esi movl %eax, %r15d movq 0xa8(%rsp), %rdi callq 0x6fbc2 leaq 0x120(%r14), %rax movq %rax, 0x128(%rsp) imull %ebp, %ebx imull %r15d, %ebx movl %ebx, 0xa4(%rsp) movslq %ebx, %rax movq %rax, 0x130(%rsp) movl %r15d, 0xc(%rsp) movslq %r15d, %rax movq %rax, 0x118(%rsp) xorl %r12d, %r12d leaq 0x148(%rsp), %r15 movq (%r14), %rax movq -0x18(%rax), %rax movslq 0x114(%r14,%rax), %rcx cmpq %rcx, %r12 jge 0x2d5a6e movq %r12, %rcx movq 0x130(%rsp), %rdi imulq %rdi, %rcx movq 0x178(%r14,%rax), %rdx imulq %rdx, %rcx addq 0x168(%r14,%rax), %rcx movl 0x180(%r14,%rax), %esi movq 0x188(%r14,%rax), %rax movq %rcx, 0x10(%rsp) andq $0x0, 0x18(%rsp) movq %rdx, 0x20(%rsp) movl %esi, 0x28(%rsp) movq %rax, 0x30(%rsp) pushq $0x1 popq %rax movl %eax, 0x38(%rsp) movl %edi, 0x3c(%rsp) movabsq $0x100000001, %rcx # imm = 0x100000001 movq %rcx, 0x40(%rsp) movl %eax, 0x48(%rsp) movq %rdi, 0x50(%rsp) leaq 0xd0(%rsp), %rdi leaq 0x10(%rsp), %rsi xorl %edx, %edx callq 0x624f0 movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d54af lock decl (%rax) jne 0x2d54af movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d54a7 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d54af movq %rsi, %rdi callq 0x5f3e0 movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2d5521 movq %r12, %r13 movq 0x118(%rsp), %rcx imulq %rcx, %r13 movq 0x1c0(%r14,%rax), %rsi movq %rsi, 0xc0(%rsp) imulq %rsi, %r13 addq 0x1b0(%r14,%rax), %r13 movl 0x1c8(%r14,%rax), %edx movl %edx, 0x8(%rsp) movq 0x1d0(%r14,%rax), %rax movq %rax, 0xb8(%rsp) movl 0xc(%rsp), %eax movq %rax, 0xc8(%rsp) pushq $0x1 popq %rbx movq %rcx, 0xb0(%rsp) jmp 0x2d555e movq $0x0, 0xc8(%rsp) xorl %ebx, %ebx movq $0x0, 0xb8(%rsp) movl $0x0, 0x8(%rsp) movq $0x0, 0xc0(%rsp) xorl %r13d, %r13d movq $0x0, 0xb0(%rsp) pushq $0x8 popq %rdi callq 0x782bf movq %rax, %rbp movq %r15, %rdi callq 0x71548 movq %r15, %rdi xorl %esi, %esi movl 0xc(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd4(%r14,%rax), %edx movq %r15, %rdi pushq $0x1 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xd8(%r14,%rax), %edx movq %r15, %rdi pushq $0xb popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xdc(%r14,%rax), %edx movq %r15, %rdi pushq $0x2 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe0(%r14,%rax), %edx movq %r15, %rdi pushq $0xc popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe4(%r14,%rax), %edx movq %r15, %rdi pushq $0x3 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xe8(%r14,%rax), %edx movq %r15, %rdi pushq $0xd popq %rsi callq 0x7193a movq %r15, %rdi pushq $0x4 popq %rsi xorl %edx, %edx callq 0x7193a movq %r15, %rdi pushq $0xe popq %rsi xorl %edx, %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0xfc(%r14,%rax), %edx movq %r15, %rdi pushq $0x12 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x100(%r14,%rax), %edx movq %r15, %rdi pushq $0x13 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x10c(%r14,%rax), %edx movq %r15, %rdi pushq $0x5 popq %rsi callq 0x7193a movq %r15, %rdi pushq $0x6 popq %rsi movl 0xa4(%rsp), %edx callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rax movl 0x118(%r14,%rax), %edx movq %r15, %rdi pushq $0x9 popq %rsi callq 0x7193a movq (%r14), %rax movq -0x18(%rax), %rdx addq 0x128(%rsp), %rdx movq %r15, %rdi pushq $0xa popq %rsi callq 0x7196c movq (%rbp), %rax movq %rbp, %rdi movq %r15, %rsi callq *0x10(%rax) movq (%r14), %rax movq -0x18(%rax), %rax cmpl $0x0, 0x10c(%r14,%rax) je 0x2d575e andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) leaq 0x30(%rsp), %rax movups %xmm0, 0xc(%rax) movaps %xmm0, (%rax) andq $0x0, 0x98(%rsp) movups %xmm0, 0x34(%rax) movups %xmm0, 0x28(%rax) movups %xmm0, 0x54(%rax) movups %xmm0, 0x48(%rax) movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2d5730 lock incl (%rax) movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d57d6 lock decl (%rax) jne 0x2d57d6 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d57ce movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d57d6 andq $0x0, 0x50(%rsp) xorps %xmm0, %xmm0 movaps %xmm0, 0x10(%rsp) movups %xmm0, 0x1c(%rsp) leaq 0x30(%rsp), %rax movups %xmm0, 0xc(%rax) movaps %xmm0, (%rax) movq 0xd8(%rsp), %rax testq %rax, %rax leaq 0x138(%rsp), %rbx je 0x2d59b7 lock incl (%rax) movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d59b7 lock decl (%rax) jne 0x2d59b7 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d59af movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d59b7 movq %rsi, %rdi callq 0x5f3e0 movaps 0xd0(%rsp), %xmm0 movaps %xmm0, 0x10(%rsp) movq 0xe0(%rsp), %rax movq %rax, 0x20(%rsp) movl 0xe8(%rsp), %eax movl %eax, 0x28(%rsp) movq 0xf0(%rsp), %rax movq %rax, 0x30(%rsp) movups 0xf8(%rsp), %xmm0 movups %xmm0, 0x38(%rsp) movl 0x108(%rsp), %eax movl %eax, 0x48(%rsp) movq 0x110(%rsp), %rax movq %rax, 0x50(%rsp) movq 0x60(%rsp), %rax testq %rax, %rax je 0x2d585b lock decl (%rax) jne 0x2d585b movq 0x58(%rsp), %rsi movq 0x78(%rsp), %rdi testq %rdi, %rdi je 0x2d5853 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d585b movq %rsi, %rdi callq 0x5f3e0 movq %r13, 0x58(%rsp) andq $0x0, 0x60(%rsp) movq 0xc0(%rsp), %rax movq %rax, 0x68(%rsp) movl 0x8(%rsp), %eax movl %eax, 0x70(%rsp) movq 0xb8(%rsp), %rax movq %rax, 0x78(%rsp) movl %ebx, 0x80(%rsp) movq 0xc8(%rsp), %rax movl %eax, 0x84(%rsp) movl %ebx, 0x88(%rsp) movl %ebx, 0x8c(%rsp) movl %ebx, 0x90(%rsp) movq 0xb0(%rsp), %rax movq %rax, 0x98(%rsp) leaq 0x138(%rsp), %rbx movq %rbx, %rdi leaq 0x10(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %rbx, %rsi callq *0x18(%rax) movq %rbx, %rdi callq 0x6b03a pushq $0x48 popq %rbx movq 0x18(%rsp,%rbx), %rax testq %rax, %rax je 0x2d591e lock decl (%rax) jne 0x2d591e movq 0x10(%rsp,%rbx), %rsi movq 0x30(%rsp,%rbx), %rdi testq %rdi, %rdi je 0x2d5916 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d591e movq %rsi, %rdi callq 0x5f3e0 leaq (%rsp,%rbx), %rax addq $0x10, %rax andq $0x0, 0x40(%rax) xorps %xmm0, %xmm0 movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %rbx cmpq $-0x48, %rbx jne 0x2d58f0 movq (%rbp), %rax movq %rbp, %rdi movq 0x120(%rsp), %rsi callq *0x20(%rax) movq 0xa8(%rsp), %rax movq (%rax), %rax movq %rbp, (%rax,%r12,8) movq %r15, %rdi callq 0x71614 movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2d59a7 lock decl (%rax) jne 0x2d59a7 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi je 0x2d599f movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d59a7 movq %rsi, %rdi callq 0x5f3e0 incq %r12 jmp 0x2d53e6 movq %rsi, %rdi callq 0x5f3e0 movaps 0xd0(%rsp), %xmm0 movaps %xmm0, 0x10(%rsp) movq 0xe0(%rsp), %rax movq %rax, 0x20(%rsp) movl 0xe8(%rsp), %eax movl %eax, 0x28(%rsp) movq 0xf0(%rsp), %rax movq %rax, 0x30(%rsp) movups 0xf8(%rsp), %xmm0 movups %xmm0, 0x38(%rsp) movl 0x108(%rsp), %eax movl %eax, 0x48(%rsp) movq 0x110(%rsp), %rax movq %rax, 0x50(%rsp) movq %rbx, %rdi leaq 0x10(%rsp), %rsi callq 0x6b00e movq (%rbp), %rax movq %rbp, %rdi movq %rbx, %rsi callq *0x18(%rax) movq %rbx, %rdi callq 0x6b03a movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d5947 lock decl (%rax) jne 0x2d5947 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d5a61 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d5947 movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d5947 xorl %eax, %eax addq $0x158, %rsp # imm = 0x158 popq %rbx popq %r12 popq %r13 popq %r14 popq %r15 popq %rbp retq jmp 0x2d5a95 jmp 0x2d5bed jmp 0x2d5bed jmp 0x2d5bed jmp 0x2d5aff movq %rax, %rbx jmp 0x2d5aaa movq %rax, %rbx leaq 0x138(%rsp), %rdi callq 0x6b03a movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d5ba1 lock decl (%rax) jne 0x2d5ba1 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi jne 0x2d5add movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d5ba1 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d5ba1 jmp 0x2d5bed movq %rax, %rbx leaq 0x138(%rsp), %rdi callq 0x6b03a jmp 0x2d5b02 movq %rax, %rbx pushq $0x48 popq %r14 xorps %xmm0, %xmm0 movq 0x18(%rsp,%r14), %rax testq %rax, %rax je 0x2d5b3d lock decl (%rax) jne 0x2d5b3d movq 0x10(%rsp,%r14), %rsi movq 0x30(%rsp,%r14), %rdi testq %rdi, %rdi je 0x2d5b32 movq (%rdi), %rax callq *0x18(%rax) xorps %xmm0, %xmm0 jmp 0x2d5b3d movq %rsi, %rdi callq 0x5f3e0 xorps %xmm0, %xmm0 leaq (%rsp,%r14), %rax addq $0x10, %rax andq $0x0, 0x40(%rax) movups %xmm0, 0xc(%rax) movups %xmm0, (%rax) andl $0x0, 0x38(%rax) movups %xmm0, 0x28(%rax) addq $-0x48, %r14 cmpq $-0x48, %r14 jne 0x2d5b09 jmp 0x2d5ba1 jmp 0x2d5bed jmp 0x2d5b97 movq %rax, %rbx movq 0x18(%rsp), %rax testq %rax, %rax je 0x2d5be5 lock decl (%rax) jne 0x2d5be5 movq 0x10(%rsp), %rsi movq 0x30(%rsp), %rdi testq %rdi, %rdi je 0x2d5bd5 movq (%rdi), %rax callq *0x18(%rax) jmp 0x2d5be5 jmp 0x2d5bed movq %rax, %rbx jmp 0x2d5bae jmp 0x2d5bed movq %rax, %rbx leaq 0x148(%rsp), %rdi callq 0x71614 movq 0xd8(%rsp), %rax testq %rax, %rax je 0x2d5be5 lock decl (%rax) jne 0x2d5be5 movq 0xd0(%rsp), %rsi movq 0xf0(%rsp), %rdi testq %rdi, %rdi jne 0x2d5bdf movq %rsi, %rdi callq 0x5f3e0 jmp 0x2d5be5 movq (%rdi), %rax callq *0x18(%rax) movq %rbx, %rdi callq 0x5f340 movq %rax, %rdi callq 0x61d68 nop

/csukuangfj[P]ncnn/src/layer/x86/deconvolutiondepthwise_x86.cpp