{"bug_id": "118850", "issue_url": "https://github.com/llvm/llvm-project/issues/118850", "bug_type": "crash", "base_commit": "9ad22cf0ee9eb686ea9fd76fd9cc41f4c006ccbc", "knowledge_cutoff": "2024-12-05T18:11:28Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "4f7f71b7bccdc38f37b82981e8fa9ceb536a7016", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp": [[168, 174]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp": ["UnrollState::unrollWidenInductionByUF"]}}, "patch": "commit 4f7f71b7bccdc38f37b82981e8fa9ceb536a7016\nAuthor: Florian Hahn \nDate: Fri Dec 6 16:28:47 2024 +0000\n\n [VPlan] Compare APInt instead of getSExtValue to fix crash in unroll.\n \n getSExtValue assumes the result fits in 64 bits, but this may not be the\n case for indcutions with wider types. Instead, directly perform the\n compare on the APInt for the ConstantInt.\n \n Fixes https://github.com/llvm/llvm-project/issues/118850.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp b/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp\nindex f653269713b3..ff6c9295ee20 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp\n@@ -168,7 +168,7 @@ void UnrollState::unrollWidenInductionByUF(\n auto *ConstStep = ScalarStep->isLiveIn()\n ? dyn_cast(ScalarStep->getLiveInIRValue())\n : nullptr;\n- if (!ConstStep || ConstStep->getZExtValue() != 1) {\n+ if (!ConstStep || ConstStep->getValue() != 1) {\n if (TypeInfo.inferScalarType(ScalarStep) != IVTy) {\n ScalarStep =\n Builder.createWidenCast(Instruction::Trunc, ScalarStep, IVTy);\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/interleave-with-i65-induction.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=2 -S %s"], "tests": [{"test_name": "i65_induction_with_negative_step", "test_body": "define void @i65_induction_with_negative_step(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.i65 = phi i65 [ 0, %entry ], [ %iv.i65.next, %loop ]\n %for = phi i64 [ 0, %entry ], [ %trunc, %loop ]\n %trunc = trunc i65 %iv.i65 to i64\n %gep = getelementptr inbounds i64, ptr %dst, i64 %trunc\n store i64 %for, ptr %gep, align 8\n %iv.next = add i64 %iv, 1\n %icmp = icmp eq i64 %iv.next, 1000\n %iv.i65.next = add i65 %iv.i65, -1\n br i1 %icmp, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "Assertion `getActiveBits() <= 64 && \"Too many bits for uint64_t\"' failed.", "body": "To reproduce run the test below with -passes=loop-vectorize\n```\n; ModuleID = './reduced.ll'\nsource_filename = \"./reduced.ll\"\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-Fn32-ni:1-p2:32:8:8:32-ni:2\"\ntarget triple = \"aarch64-none-linux-gnu\"\n\ndefine double @wombat() gc \"statepoint-example\" {\nbb:\n br label %bb2\n\nbb1: ; preds = %bb2\n ret double 0.000000e+00\n\nbb2: ; preds = %bb2, %bb\n %phi = phi i64 [ 0, %bb ], [ %add, %bb2 ]\n %phi3 = phi i65 [ 0, %bb ], [ %add5, %bb2 ]\n %phi4 = phi i64 [ 0, %bb ], [ %trunc, %bb2 ]\n %trunc = trunc i65 %phi3 to i64\n store i64 0, ptr addrspace(1) null, align 8\n %add = add i64 %phi, 1\n %icmp = icmp ugt i64 %add, 0\n %add5 = add i65 %phi3, -1\n br i1 %icmp, label %bb2, label %bb1\n}\n```\nReproducer: https://godbolt.org/z/cjEcv8b4Y\n\nStack dump:\n```\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=loop-vectorize \n1.\tRunning pass \"function(loop-vectorize)\" on module \"\"\n2.\tRunning pass \"loop-vectorize\" on function \"wombat\"\n #0 0x00000000051b3748 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x51b3748)\n #1 0x00000000051b113c SignalHandler(int) Signals.cpp:0:0\n #2 0x000071a06ec42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #3 0x000071a06ec969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #4 0x000071a06ec42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #5 0x000071a06ec287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #6 0x000071a06ec2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #7 0x000071a06ec39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #8 0x0000000000b2d8a9 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xb2d8a9)\n #9 0x0000000004308ae6 (anonymous namespace)::UnrollState::unrollBlock(llvm::VPBlockBase*) VPlanUnroll.cpp:0:0\n#10 0x000000000430788c (anonymous namespace)::UnrollState::unrollBlock(llvm::VPBlockBase*) VPlanUnroll.cpp:0:0\n#11 0x0000000004309060 llvm::VPlanTransforms::unrollByUF(llvm::VPlan&, unsigned int, llvm::LLVMContext&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4309060)\n#12 0x0000000004176b76 llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4176b76)\n#13 0x000000000418cb6f llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x418cb6f)\n#14 0x000000000418ebd9 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x418ebd9)\n#15 0x000000000418f213 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x418f213)\n#16 0x0000000002fcba8e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2fcba8e)\n#17 0x0000000004fb2700 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb2700)\n#18 0x0000000000e33a9e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe33a9e)\n#19 0x0000000004fb101b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb101b)\n#20 0x0000000000e3312e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3312e)\n#21 0x0000000004fb0a20 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb0a20)\n#22 0x0000000000910a1a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x910a1a)\n#23 0x0000000000903099 optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x903099)\n#24 0x000071a06ec29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#25 0x000071a06ec29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#26 0x00000000008fa97e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fa97e)\nProgram terminated with signal: SIGSEGV\nCompiler returned: 139\n```\n", "author": "TatyanaDoubts", "labels": ["vectorizers", "crash"], "comments": [{"author": "huang-me", "body": "Hi @TatyanaDoubts, \nHow did you get the IR snippet? \nThe IR you provided is weird. I'll have to check where i65 constant 0 in `%phi3 = phi i65 [ 0, %bb ], [ %add5, %bb2 ]` come from since it seems it's the origin of the problem."}, {"author": "fhahn", "body": "I suspect this was found via a fuzzer, thanks for sharing the test case!"}, {"author": "TatyanaDoubts", "body": "Hi @huang-me, @fhahn, yes this crash was found by the fuzzer generated test. "}]}, "verified": true} {"bug_id": "115575", "issue_url": "https://github.com/llvm/llvm-project/issues/115575", "bug_type": "miscompilation", "base_commit": "6fb2a6044f11e251c3847d227049d9dae8b87796", "knowledge_cutoff": "2024-11-09T02:09:23Z", "lit_test_dir": ["llvm/test/Transforms/VectorCombine"], "hints": {"fix_commit": "958e37cd1feabf29fb1cc3fb5ac82051ad8d43eb", "components": ["VectorCombine"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": [[952, 957]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": ["VectorCombine::scalarizeBinopOrCmp"]}}, "patch": "commit 958e37cd1feabf29fb1cc3fb5ac82051ad8d43eb\nAuthor: Simon Pilgrim \nDate: Sat Nov 9 15:59:52 2024 +0000\n\n [VectorCombine] scalarizeBinopOrCmp - check for out of bounds element indices\n \n Fixes #115575\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\nindex 04ea12ef0f12..b8754b03c2eb 100644\n--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n@@ -952,6 +952,12 @@ bool VectorCombine::scalarizeBinopOrCmp(Instruction &I) {\n if (!IsConst0 && !IsConst1 && Index0 != Index1)\n return false;\n \n+ auto *VecTy0 = cast(Ins0->getType());\n+ auto *VecTy1 = cast(Ins1->getType());\n+ if (VecTy0->getElementCount().getKnownMinValue() <= Index0 ||\n+ VecTy1->getElementCount().getKnownMinValue() <= Index1)\n+ return false;\n+\n // Bail for single insertion if it is a load.\n // TODO: Handle this once getVectorInstrCost can cost for load/stores.\n auto *I0 = dyn_cast_or_null(V0);\n", "tests": [{"file": "llvm/test/Transforms/VectorCombine/X86/pr115575.ll", "commands": ["opt < %s -passes=vector-combine -S -mtriple=x86_64--"], "tests": [{"test_name": "PR115575", "test_body": "define <2 x i8> @PR115575(i8 %x) {\n %ins = insertelement <2 x i8> poison, i8 %x, i32 3\n %bo = sdiv <2 x i8> %ins, \n ret <2 x i8> %bo\n}\n"}]}], "issue": {"title": "[VectorCombine] UB triggered after optimization", "body": "https://github.com/llvm/llvm-project/blob/c93e001ca695e905cb965b36d63f7a348d1dd809/llvm/lib/Transforms/Vectorize/VectorCombine.cpp#L1013-L1031\r\n\r\nAlive2 report: https://alive2.llvm.org/ce/z/78CM4x\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine <2 x i8> @sdiv_constant_op1_not_undef_lane.2(i8 %x) {\r\n#0:\r\n %ins = insertelement <2 x i8> poison, i8 %x, i32 3\r\n %bo = sdiv <2 x i8> %ins, { 5, 2 }\r\n ret <2 x i8> %bo\r\n}\r\n=>\r\ndefine <2 x i8> @sdiv_constant_op1_not_undef_lane.2(i8 %x) {\r\n#0:\r\n %bo.scalar = sdiv i8 %x, poison\r\n %bo = insertelement <2 x i8> poison, i8 %bo.scalar, i64 3\r\n ret <2 x i8> %bo\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Source is more defined than target\r\n\r\nExample:\r\ni8 %x = poison\r\n\r\nSource:\r\n<2 x i8> %ins = < poison, poison >\r\n<2 x i8> %bo = < poison, poison >\r\n\r\nTarget:\r\ni8 %bo.scalar = UB triggered!\r\n\r\n\r\n\r\n----------------------------------------\r\ndefine <2 x i64> @urem_constant_op1_not_undef_lane.2(i64 %x) {\r\n#0:\r\n %ins = insertelement <2 x i64> poison, i64 %x, i32 4294967295\r\n %bo = urem <2 x i64> %ins, { 5, 2 }\r\n ret <2 x i64> %bo\r\n}\r\n=>\r\ndefine <2 x i64> @urem_constant_op1_not_undef_lane.2(i64 %x) {\r\n#0:\r\n %bo.scalar = urem i64 %x, poison\r\n %bo = insertelement <2 x i64> poison, i64 %bo.scalar, i64 4294967295\r\n ret <2 x i64> %bo\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Source is more defined than target\r\n\r\nExample:\r\ni64 %x = poison\r\n\r\nSource:\r\n<2 x i64> %ins = < poison, poison >\r\n<2 x i64> %bo = < poison, poison >\r\n\r\nTarget:\r\ni64 %bo.scalar = UB triggered!\r\n\r\n\r\nSummary:\r\n 0 correct transformations\r\n 2 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "vectorizers"], "comments": []}, "verified": true} {"bug_id": "114082", "issue_url": "https://github.com/llvm/llvm-project/issues/114082", "bug_type": "crash", "base_commit": "0f040433d325aa68ec6840aa179f3f314c26153a", "knowledge_cutoff": "2024-10-29T16:26:31Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "77bec78878762e34150fe23734fa43df796c873c", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[13398, 13404]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::getLastInstructionInBundle"]}}, "patch": "commit 77bec78878762e34150fe23734fa43df796c873c\nAuthor: Alexey Bataev \nDate: Fri Nov 8 04:16:01 2024 -0800\n\n [SLP]Do not look for last instruction in schedule block for buildvectors\n \n If looking for the insertion point for the node and the node is\n a buildvector node, the compiler should not use scheduling info for such\n nodes, they may contain only partial info, which is not fully correct\n and may cause compiler crash.\n \n Fixes #114082\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex b2f677fb84f9..c5b3537bc57b 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -13398,7 +13398,7 @@ Instruction &BoUpSLP::getLastInstructionInBundle(const TreeEntry *E) {\n // scheduled, and the last instruction is VL.back(). So we start with\n // VL.back() and iterate over schedule data until we reach the end of the\n // bundle. The end of the bundle is marked by null ScheduleData.\n- if (BlocksSchedules.count(BB)) {\n+ if (BlocksSchedules.count(BB) && !E->isGather()) {\n Value *V = E->isOneOf(E->Scalars.back());\n if (doesNotNeedToBeScheduled(V))\n V = *find_if_not(E->Scalars, doesNotNeedToBeScheduled);\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/buildvector-schedule-for-subvector.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -slp-threshold=-99999 < %s"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nbb:\n %icmp = icmp samesign ult i32 0, 0\n %select = select i1 %icmp, i32 0, i32 0\n %zext = zext i32 %select to i64\n %getelementptr = getelementptr ptr addrspace(1), ptr addrspace(1) null, i64 %zext\n store ptr addrspace(1) null, ptr addrspace(1) %getelementptr, align 8\n %icmp1 = icmp eq i32 0, 0\n %icmp2 = icmp eq i32 0, 0\n %icmp3 = icmp samesign ult i32 0, 0\n %icmp4 = icmp eq i32 0, 0\n %add = add i32 1, 0\n %icmp5 = icmp samesign ult i32 %add, 0\n store volatile i32 0, ptr addrspace(1) null, align 4\n %call = call i32 null(<2 x double> zeroinitializer)\n %icmp6 = icmp eq i32 %call, 0\n %icmp7 = icmp samesign ult i32 0, 0\n ret void\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Instruction does not dominate all uses! LLVM ERROR: Broken module found, compilation aborted!", "body": "To reproduce run opt with the test below with -passes=slp-vectorizer -slp-threshold=-99999\r\n```; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat() #0 gc \"statepoint-example\" {\r\nbb:\r\n %icmp = icmp samesign ult i32 0, 0\r\n %select = select i1 %icmp, i32 0, i32 0\r\n %zext = zext i32 %select to i64\r\n %getelementptr = getelementptr ptr addrspace(1), ptr addrspace(1) null, i64 %zext\r\n store ptr addrspace(1) null, ptr addrspace(1) %getelementptr, align 8\r\n %icmp1 = icmp eq i32 0, 0\r\n %icmp2 = icmp eq i32 0, 0\r\n %icmp3 = icmp samesign ult i32 0, 0\r\n %icmp4 = icmp eq i32 0, 0\r\n %add = add i32 1, 0\r\n %icmp5 = icmp samesign ult i32 %add, 0\r\n store volatile i32 0, ptr addrspace(1) null, align 4\r\n %call = call i32 @llvm.x86.sse2.cvttsd2si(<2 x double> zeroinitializer)\r\n %icmp6 = icmp eq i32 %call, 0\r\n %icmp7 = icmp samesign ult i32 0, 0\r\n ret void\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(none)\r\ndeclare i32 @llvm.x86.sse2.cvttsd2si(<2 x double>) #1\r\n\r\nattributes #0 = { \"target-features\"=\"+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-avx10.2-256,-gfni,-avxvnniint16,-amx-fp16,-zu,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,-nf,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-avx10.2-512,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f\" }\r\nattributes #1 = { nocallback nofree nosync nounwind willreturn memory(none) \"target-features\"=\"+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-avx10.2-256,-gfni,-avxvnniint16,-amx-fp16,-zu,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,-nf,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-avx10.2-512,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f\" }\r\n```\r\nReproducer: https://godbolt.org/z/nTE84xq7n\r\n\r\nStack dump\r\n```\r\nInstruction does not dominate all uses!\r\n %5 = icmp eq <4 x i32> %4, zeroinitializer\r\n %2 = call <8 x i1> @llvm.vector.insert.v8i1.v4i1(<8 x i1> poison, <4 x i1> %5, i64 0)\r\nLLVM ERROR: Broken module found, compilation aborted!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer -slp-threshold=-99999 \r\n1.\tRunning pass \"verify\" on module \"\"\r\n #0 0x00000000050e7a98 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50e7a98)\r\n #1 0x00000000050e548c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000078c8a0042520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000078c8a00969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000078c8a0042476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000078c8a00287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00000000007d7d7b llvm::json::operator==(llvm::json::Value const&, llvm::json::Value const&) (.cold) JSON.cpp:0:0\r\n #7 0x00000000050265b8 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50265b8)\r\n #8 0x0000000004f24790 llvm::VerifierPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f24790)\r\n #9 0x00000000009015ce llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x9015ce)\r\n#10 0x0000000004ee5440 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ee5440)\r\n#11 0x000000000090c1ea llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x90c1ea)\r\n#12 0x00000000008ff258 optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ff258)\r\n#13 0x000078c8a0029d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#14 0x000078c8a0029e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#15 0x00000000008f6d2e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f6d2e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "114738", "issue_url": "https://github.com/llvm/llvm-project/issues/114738", "bug_type": "miscompilation", "base_commit": "2588b8be5624f721c56f44fd2a69f7327de8c0c1", "knowledge_cutoff": "2024-11-04T05:50:07Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "899336735aeb2b41a48b6ac2c895da5e0f22dbf0", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[19490, 19498]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["tryToReduce"]}}, "patch": "commit 899336735aeb2b41a48b6ac2c895da5e0f22dbf0\nAuthor: Alexey Bataev \nDate: Mon Nov 4 06:02:06 2024 -0800\n\n [SLP]Be more pessimistic about poisonous reductions\n \n Consider all possible reductions ops as being non-poisoning boolean\n logical operations, which require freeze to be fully correct.\n \n https://alive2.llvm.org/ce/z/TKWDMP\n \n Fixes #114738\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 427b8bd0e75a..3c6daf7b9fbb 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -19490,9 +19490,7 @@ public:\n // To prevent poison from leaking across what used to be sequential,\n // safe, scalar boolean logic operations, the reduction operand must be\n // frozen.\n- if ((isBoolLogicOp(RdxRootInst) ||\n- (AnyBoolLogicOp && VL.size() != TrackedVals.size())) &&\n- !isGuaranteedNotToBePoison(VectorizedRoot))\n+ if (AnyBoolLogicOp && !isGuaranteedNotToBePoison(VectorizedRoot))\n VectorizedRoot = Builder.CreateFreeze(VectorizedRoot);\n \n // Emit code to correctly handle reused reduced values, if required.\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/reudction-or-non-poisoned.ll", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "test", "test_body": "define i1 @test(i32 %x, i32 %a, i32 %b, i32 %c, i32 %d) {\n %cmp = icmp sgt i32 %x, 1\n %cmp2 = icmp sgt i32 %b, 1\n %cmp3 = icmp sgt i32 %c, 1\n %cmp4 = icmp sgt i32 %d, 1\n %sel2 = select i1 %cmp4, i1 true, i1 %cmp2\n %sel3 = select i1 %sel2, i1 true, i1 %cmp3\n %sel4 = select i1 %cmp, i1 true, i1 %cmp4\n %ret = or i1 %sel3, %sel4\n ret i1 %ret\n}\n"}]}], "issue": {"title": "[SLPVectorizer] wrong vectorization to `reduce_or` due to poisonous value", "body": "Alive2 report: https://alive2.llvm.org/ce/z/f8PDZj\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine i1 @test1.2(i32 %x, i32 %a, i32 %b, i32 %c, i32 %d) {\r\n#0:\r\n %cmp = icmp sgt i32 %x, 1\r\n %cmp2 = icmp sgt i32 %b, 1\r\n %cmp3 = icmp sgt i32 %c, 1\r\n %cmp4 = icmp sgt i32 %d, 1\r\n %sel2 = select i1 %cmp4, i1 1, i1 %cmp2\r\n %sel3 = select i1 %sel2, i1 1, i1 %cmp3\r\n %sel4 = select i1 %cmp, i1 1, i1 %cmp4\r\n %ret = or i1 %sel3, %sel4\r\n ret i1 %ret\r\n}\r\n=>\r\ndefine i1 @test1.2(i32 %x, i32 %a, i32 %b, i32 %c, i32 %d) {\r\n#0:\r\n %#1 = insertelement <4 x i32> poison, i32 %d, i32 0\r\n %#2 = insertelement <4 x i32> %#1, i32 %b, i32 1\r\n %#3 = insertelement <4 x i32> %#2, i32 %c, i32 2\r\n %#4 = insertelement <4 x i32> %#3, i32 %x, i32 3\r\n %#5 = icmp sgt <4 x i32> %#4, { 1, 1, 1, 1 }\r\n %#6 = reduce_or <4 x i1> %#5\r\n ret i1 %#6\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\ni32 %x = #x00000000 (0)\r\ni32 %a = poison\r\ni32 %b = #x00000002 (2)\r\ni32 %c = poison\r\ni32 %d = #x00000000 (0)\r\n\r\nSource:\r\ni1 %cmp = #x0 (0)\r\ni1 %cmp2 = #x1 (1)\r\ni1 %cmp3 = poison\r\ni1 %cmp4 = #x0 (0)\r\ni1 %sel2 = #x1 (1)\r\ni1 %sel3 = #x1 (1)\r\ni1 %sel4 = #x0 (0)\r\ni1 %ret = #x1 (1)\r\n\r\nTarget:\r\n<4 x i32> %#1 = < #x00000000 (0), poison, poison, poison >\r\n<4 x i32> %#2 = < #x00000000 (0), #x00000002 (2), poison, poison >\r\n<4 x i32> %#3 = < #x00000000 (0), #x00000002 (2), poison, poison >\r\n<4 x i32> %#4 = < #x00000000 (0), #x00000002 (2), poison, #x00000000 (0) >\r\n<4 x i1> %#5 = < #x0 (0), #x1 (1), poison, #x0 (0) >\r\ni1 %#6 = poison\r\nSource value: #x1 (1)\r\nTarget value: poison\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": []}, "verified": true} {"bug_id": "107016", "issue_url": "https://github.com/llvm/llvm-project/issues/107016", "bug_type": "crash", "base_commit": "a586b5a49dbd3b6c658f9edbf0b4a9be0b108a14", "knowledge_cutoff": "2024-09-02T19:14:36Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "50a02e7c68f964c5d8338369746c849ed8d3bef4", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[965, 972]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPWidenCallRecipe::computeCost"]}}, "patch": "commit 50a02e7c68f964c5d8338369746c849ed8d3bef4\nAuthor: Florian Hahn \nDate: Mon Sep 2 20:47:36 2024 +0100\n\n [VPlan] Pass intrinsic inst to TTI in VPWidenCallRecipe::computeCost.\n \n Follow-up to 9ccf825, adjust computeCost to also pass IntrinsicInst to\n TTI if available, as there are multiple places in TTI which use the\n IntrinsicInst.\n \n Fixes https://github.com/llvm/llvm-project/issues/107016.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex e782ac444d02..c91ea2d1663c 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -965,8 +965,10 @@ InstructionCost VPWidenCallRecipe::computeCost(ElementCount VF,\n ParamTys.push_back(\n ToVectorTy(Ctx.Types.inferScalarType(getOperand(I)), VF));\n \n- IntrinsicCostAttributes CostAttrs(VectorIntrinsicID, RetTy, Arguments,\n- ParamTys, FMF);\n+ // TODO: Rework TTI interface to avoid reliance on underlying IntrinsicInst.\n+ IntrinsicCostAttributes CostAttrs(\n+ VectorIntrinsicID, RetTy, Arguments, ParamTys, FMF,\n+ dyn_cast_or_null(getUnderlyingValue()));\n return Ctx.TTI.getIntrinsicInstrCost(CostAttrs, CostKind);\n }\n \n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/call-costs.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "fshl_operand_first_order_recurrence", "test_body": "target triple = \"arm64-apple-macosx11.0.0\"\n\ndefine void @fshl_operand_first_order_recurrence(ptr %dst, ptr noalias %src) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %recur = phi i64 [ 0, %entry ], [ %l, %loop ]\n %gep.src = getelementptr inbounds i64, ptr %src, i64 %iv\n %l = load i64, ptr %gep.src, align 8\n %or = tail call i64 @llvm.fshl.i64(i64 1, i64 %recur, i64 1)\n %gep.dst = getelementptr inbounds i64, ptr %dst, i64 %iv\n store i64 %or, ptr %gep.dst, align 8\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv, 100\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i64 @llvm.fshl.i64(i64, i64, i64) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "powi_call", "test_body": "target triple = \"arm64-apple-macosx11.0.0\"\n\ndefine void @powi_call(ptr %P) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr inbounds double, ptr %P, i64 %iv\n %l = load double, ptr %gep, align 8\n %powi = tail call double @llvm.powi.f64.i32(double %l, i32 3)\n store double %powi, ptr %gep, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare double @llvm.powi.f64.i32(double, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[VPlan] clang crash", "body": "Failure since https://github.com/llvm/llvm-project/pull/106047\r\n\r\n[repro.zip](https://github.com/user-attachments/files/16840758/repro.zip)\r\n\r\n```\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /scratch3/packages/llvm/main-20240902/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -disable-free -clear-ast-before-backend -main-file-name PairCorrEstimator.cpp -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -menable-no-infs -menable-no-nans -fapprox-func -funsafe-math-optimizations -fno-signed-zeros -mreassociate -freciprocal-math -ffp-contract=fast -fno-rounding-math -ffast-math -ffinite-math-only -complex-range=basic -mconstructor-aliases -funwind-tables=2 -target-cpu skylake-avx512 -target-feature +prfchw -target-feature -cldemote -target-feature +avx -target-feature +aes -target-feature +sahf -target-feature +pclmul -target-feature -xop -target-feature +crc32 -target-feature +xsaves -target-feature -avx512fp16 -target-feature -usermsr -target-feature -sm4 -target-feature -egpr -target-feature +sse4.1 -target-feature -avx512ifma -target-feature +xsave -target-feature +sse4.2 -target-feature -tsxldtrk -target-feature -sm3 -target-feature -ptwrite -target-feature -widekl -target-feature +invpcid -target-feature +64bit -target-feature +xsavec -target-feature -avx10.1-512 -target-feature -avx512vpopcntdq -target-feature +cmov -target-feature -avx512vp2intersect -target-feature +avx512cd -target-feature +movbe -target-feature -avxvnniint8 -target-feature -ccmp -target-feature -amx-int8 -target-feature -kl -target-feature -avx10.1-256 -target-feature +evex512 -target-feature -avxvnni -target-feature +rtm -target-feature +adx -target-feature +avx2 -target-feature -hreset -target-feature -movdiri -target-feature -serialize -target-feature -sha512 -target-feature -vpclmulqdq -target-feature +avx512vl -target-feature -uintr -target-feature -cf -target-feature +clflushopt -target-feature -raoint -target-feature -cmpccxadd -target-feature +bmi -target-feature -amx-tile -target-feature +sse -target-feature -avx10.2-256 -target-feature -gfni -target-feature -avxvnniint16 -target-feature -amx-fp16 -target-feature -zu -target-feature -ndd -target-feature +xsaveopt -target-feature +rdrnd -target-feature +avx512f -target-feature -amx-bf16 -target-feature -avx512bf16 -target-feature -avx512vnni -target-feature -push2pop2 -target-feature +cx8 -target-feature +avx512bw -target-feature +sse3 -target-feature +pku -target-feature -nf -target-feature +fsgsbase -target-feature -clzero -target-feature -mwaitx -target-feature -lwp -target-feature +lzcnt -target-feature -sha -target-feature -movdir64b -target-feature -ppx -target-feature -wbnoinvd -target-feature -enqcmd -target-feature -avx10.2-512 -target-feature -avxneconvert -target-feature -tbm -target-feature -pconfig -target-feature -amx-complex -target-feature +ssse3 -target-feature +cx16 -target-feature +bmi2 -target-feature +fma -target-feature +popcnt -target-feature -avxifma -target-feature +f16c -target-feature -avx512bitalg -target-feature -rdpru -target-feature +clwb -target-feature +mmx -target-feature +sse2 -target-feature +rdseed -target-feature -avx512vbmi2 -target-feature -prefetchi -target-feature -rdpid -target-feature -fma4 -target-feature -avx512vbmi -target-feature -shstk -target-feature -vaes -target-feature -waitpkg -target-feature -sgx -target-feature +fxsr -target-feature +avx512dq -target-feature -sse4a -debugger-tuning=gdb -fdebug-compilation-dir=/scratch3/opt/qmcpack/build_clangPatched_real/src/QMCHamiltonians -fcoverage-compilation-dir=/scratch3/opt/qmcpack/build_clangPatched_real/src/QMCHamiltonians -sys-header-deps -D ADD_ -D H5_USE_110_API -D HAVE_CONFIG_H -D HAVE_EINSPLINE -D HAVE_LIBBOOST -D HAVE_LIBFFTW -D HAVE_MKL -D HAVE_MKL_VML -D MPICH_SKIP_MPICXX -D OMPI_SKIP_MPICXX -D OPENMP_NO_COMPLEX -D _MPICC_H -D restrict=__restrict__ -D NDEBUG -O3 -Wvla -Wall -Wno-unused-variable -Wno-overloaded-virtual -Wno-unused-private-field -Wno-unused-local-typedef -Wsuggest-override -Wno-unknown-pragmas -Wmisleading-indentation -std=c++17 -fdeprecated-macro -ferror-limit 19 -fopenmp -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcxx-exceptions -fexceptions -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -x c++ PairCorrEstimator-40412c.cpp\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"PairCorrEstimator-40412c.cpp\"\r\n4.\tRunning pass \"loop-vectorize\" on function \"_ZN11qmcplusplus17PairCorrEstimator15set_norm_factorEv\"\r\n #0 0x00005590c1746c00 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x2cb6c00)\r\n #1 0x00005590c174400f llvm::sys::RunSignalHandlers() (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x2cb400f)\r\n #2 0x00005590c1744165 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007fb046e53520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00005590bfa8725a llvm::BasicTTIImplBase::getIntrinsicInstrCost(llvm::IntrinsicCostAttributes const&, llvm::TargetTransformInfo::TargetCostKind) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0xff725a)\r\n #5 0x00005590c07fac92 llvm::TargetTransformInfo::getIntrinsicInstrCost(llvm::IntrinsicCostAttributes const&, llvm::TargetTransformInfo::TargetCostKind) const (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x1d6ac92)\r\n #6 0x00005590c31a87b9 llvm::VPWidenCallRecipe::computeCost(llvm::ElementCount, llvm::VPCostContext&) const (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x47187b9)\r\n #7 0x00005590c31ab682 llvm::VPRecipeBase::cost(llvm::ElementCount, llvm::VPCostContext&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x471b682)\r\n #8 0x00005590c31744af llvm::VPBasicBlock::cost(llvm::ElementCount, llvm::VPCostContext&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x46e44af)\r\n #9 0x00005590c317b4b5 llvm::VPRegionBlock::cost(llvm::ElementCount, llvm::VPCostContext&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x46eb4b5)\r\n#10 0x00005590c305cfb4 llvm::LoopVectorizationPlanner::cost(llvm::VPlan&, llvm::ElementCount) const (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x45ccfb4)\r\n#11 0x00005590c3064878 llvm::LoopVectorizationPlanner::computeBestVF() (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x45d4878)\r\n#12 0x00005590c307c6fb llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x45ec6fb)\r\n#13 0x00005590c307f711 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x45ef711)\r\n#14 0x00005590c307fd87 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x45efd87)\r\n#15 0x00005590c2bdf2e6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x414f2e6)\r\n#16 0x00005590c10e652f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x265652f)\r\n#17 0x00005590bfaa9996 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x1019996)\r\n#18 0x00005590c10e52ab llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x26552ab)\r\n#19 0x00005590bfaaa356 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x101a356)\r\n#20 0x00005590c10e32ed llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x26532ed)\r\n#21 0x00005590c19d36b6 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#22 0x00005590c19d70e9 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x2f470e9)\r\n#23 0x00005590c206a4b5 clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x35da4b5)\r\n#24 0x00005590c3cfd42c clang::ParseAST(clang::Sema&, bool, bool) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x526d42c)\r\n#25 0x00005590c2318d19 clang::FrontendAction::Execute() (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x3888d19)\r\n#26 0x00005590c22972b5 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x38072b5)\r\n#27 0x00005590c23ef9d3 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0x395f9d3)\r\n#28 0x00005590bfa2dc87 cc1_main(llvm::ArrayRef, char const*, void*) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0xf9dc87)\r\n#29 0x00005590bfa269ea ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#30 0x00005590bfa29be3 clang_main(int, char**, llvm::ToolContext const&) (/scratch3/packages/llvm/main-20240902/bin/clang-20+0xf99be3)\r\n#31 0x00005590bf955c4b main (/scratch3/packages/llvm/main-20240902/bin/clang-20+0xec5c4b)\r\n#32 0x00007fb046e3ad90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#33 0x00007fb046e3ae40 call_init ./csu/../csu/libc-start.c:128:20\r\n#34 0x00007fb046e3ae40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#35 0x00005590bfa26115 _start (/scratch3/packages/llvm/main-20240902/bin/clang-20+0xf96115)\r\nPairCorrEstimator-40412c.sh: line 4: 840679 Segmentation fault (core dumped) \"/scratch3/packages/llvm/main-20240902/bin/clang-20\" \"-cc1\" \"-triple\" \"x86_64-unknown-linux-gnu\" \"-emit-obj\" \"-disable-free\" \"-clear-ast-before-backend\" \"-main-file-name\" \"PairCorrEstimator.cpp\" \"-mrelocation-model\" \"pic\" \"-pic-level\" \"2\" \"-pic-is-pie\" \"-mframe-pointer=none\" \"-menable-no-infs\" \"-menable-no-nans\" \"-fapprox-func\" \"-funsafe-math-optimizations\" \"-fno-signed-zeros\" \"-mreassociate\" \"-freciprocal-math\" \"-ffp-contract=fast\" \"-fno-rounding-math\" \"-ffast-math\" \"-ffinite-math-only\" \"-complex-range=basic\" \"-mconstructor-aliases\" \"-funwind-tables=2\" \"-target-cpu\" \"skylake-avx512\" \"-target-feature\" \"+prfchw\" \"-target-feature\" \"-cldemote\" \"-target-feature\" \"+avx\" \"-target-feature\" \"+aes\" \"-target-feature\" \"+sahf\" \"-target-feature\" \"+pclmul\" \"-target-feature\" \"-xop\" \"-target-feature\" \"+crc32\" \"-target-feature\" \"+xsaves\" \"-target-feature\" \"-avx512fp16\" \"-target-feature\" \"-usermsr\" \"-target-feature\" \"-sm4\" \"-target-feature\" \"-egpr\" \"-target-feature\" \"+sse4.1\" \"-target-feature\" \"-avx512ifma\" \"-target-feature\" \"+xsave\" \"-target-feature\" \"+sse4.2\" \"-target-feature\" \"-tsxldtrk\" \"-target-feature\" \"-sm3\" \"-target-feature\" \"-ptwrite\" \"-target-feature\" \"-widekl\" \"-target-feature\" \"+invpcid\" \"-target-feature\" \"+64bit\" \"-target-feature\" \"+xsavec\" \"-target-feature\" \"-avx10.1-512\" \"-target-feature\" \"-avx512vpopcntdq\" \"-target-feature\" \"+cmov\" \"-target-feature\" \"-avx512vp2intersect\" \"-target-feature\" \"+avx512cd\" \"-target-feature\" \"+movbe\" \"-target-feature\" \"-avxvnniint8\" \"-target-feature\" \"-ccmp\" \"-target-feature\" \"-amx-int8\" \"-target-feature\" \"-kl\" \"-target-feature\" \"-avx10.1-256\" \"-target-feature\" \"+evex512\" \"-target-feature\" \"-avxvnni\" \"-target-feature\" \"+rtm\" \"-target-feature\" \"+adx\" \"-target-feature\" \"+avx2\" \"-target-feature\" \"-hreset\" \"-target-feature\" \"-movdiri\" \"-target-feature\" \"-serialize\" \"-target-feature\" \"-sha512\" \"-target-feature\" \"-vpclmulqdq\" \"-target-feature\" \"+avx512vl\" \"-target-feature\" \"-uintr\" \"-target-feature\" \"-cf\" \"-target-feature\" \"+clflushopt\" \"-target-feature\" \"-raoint\" \"-target-feature\" \"-cmpccxadd\" \"-target-feature\" \"+bmi\" \"-target-feature\" \"-amx-tile\" \"-target-feature\" \"+sse\" \"-target-feature\" \"-avx10.2-256\" \"-target-feature\" \"-gfni\" \"-target-feature\" \"-avxvnniint16\" \"-target-feature\" \"-amx-fp16\" \"-target-feature\" \"-zu\" \"-target-feature\" \"-ndd\" \"-target-feature\" \"+xsaveopt\" \"-target-feature\" \"+rdrnd\" \"-target-feature\" \"+avx512f\" \"-target-feature\" \"-amx-bf16\" \"-target-feature\" \"-avx512bf16\" \"-target-feature\" \"-avx512vnni\" \"-target-feature\" \"-push2pop2\" \"-target-feature\" \"+cx8\" \"-target-feature\" \"+avx512bw\" \"-target-feature\" \"+sse3\" \"-target-feature\" \"+pku\" \"-target-feature\" \"-nf\" \"-target-feature\" \"+fsgsbase\" \"-target-feature\" \"-clzero\" \"-target-feature\" \"-mwaitx\" \"-target-feature\" \"-lwp\" \"-target-feature\" \"+lzcnt\" \"-target-feature\" \"-sha\" \"-target-feature\" \"-movdir64b\" \"-target-feature\" \"-ppx\" \"-target-feature\" \"-wbnoinvd\" \"-target-feature\" \"-enqcmd\" \"-target-feature\" \"-avx10.2-512\" \"-target-feature\" \"-avxneconvert\" \"-target-feature\" \"-tbm\" \"-target-feature\" \"-pconfig\" \"-target-feature\" \"-amx-complex\" \"-target-feature\" \"+ssse3\" \"-target-feature\" \"+cx16\" \"-target-feature\" \"+bmi2\" \"-target-feature\" \"+fma\" \"-target-feature\" \"+popcnt\" \"-target-feature\" \"-avxifma\" \"-target-feature\" \"+f16c\" \"-target-feature\" \"-avx512bitalg\" \"-target-feature\" \"-rdpru\" \"-target-feature\" \"+clwb\" \"-target-feature\" \"+mmx\" \"-target-feature\" \"+sse2\" \"-target-feature\" \"+rdseed\" \"-target-feature\" \"-avx512vbmi2\" \"-target-feature\" \"-prefetchi\" \"-target-feature\" \"-rdpid\" \"-target-feature\" \"-fma4\" \"-target-feature\" \"-avx512vbmi\" \"-target-feature\" \"-shstk\" \"-target-feature\" \"-vaes\" \"-target-feature\" \"-waitpkg\" \"-target-feature\" \"-sgx\" \"-target-feature\" \"+fxsr\" \"-target-feature\" \"+avx512dq\" \"-target-feature\" \"-sse4a\" \"-debugger-tuning=gdb\" \"-fdebug-compilation-dir=/scratch3/opt/qmcpack/build_clangPatched_real/src/QMCHamiltonians\" \"-fcoverage-compilation-dir=/scratch3/opt/qmcpack/build_clangPatched_real/src/QMCHamiltonians\" \"-sys-header-deps\" \"-D\" \"ADD_\" \"-D\" \"H5_USE_110_API\" \"-D\" \"HAVE_CONFIG_H\" \"-D\" \"HAVE_EINSPLINE\" \"-D\" \"HAVE_LIBBOOST\" \"-D\" \"HAVE_LIBFFTW\" \"-D\" \"HAVE_MKL\" \"-D\" \"HAVE_MKL_VML\" \"-D\" \"MPICH_SKIP_MPICXX\" \"-D\" \"OMPI_SKIP_MPICXX\" \"-D\" \"OPENMP_NO_COMPLEX\" \"-D\" \"_MPICC_H\" \"-D\" \"restrict=__restrict__\" \"-D\" \"NDEBUG\" \"-O3\" \"-Wvla\" \"-Wall\" \"-Wno-unused-variable\" \"-Wno-overloaded-virtual\" \"-Wno-unused-private-field\" \"-Wno-unused-local-typedef\" \"-Wsuggest-override\" \"-Wno-unknown-pragmas\" \"-Wmisleading-indentation\" \"-std=c++17\" \"-fdeprecated-macro\" \"-ferror-limit\" \"19\" \"-fopenmp\" \"-fgnuc-version=4.2.1\" \"-fskip-odr-check-in-gmf\" \"-fcxx-exceptions\" \"-fexceptions\" \"-fcolor-diagnostics\" \"-vectorize-loops\" \"-vectorize-slp\" \"-faddrsig\" \"-D__GCC_HAVE_DWARF2_CFI_ASM=1\" \"-x\" \"c++\" \"PairCorrEstimator-40412c.cpp\"\r\n```", "author": "ye-luo", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "112356", "issue_url": "https://github.com/llvm/llvm-project/issues/112356", "bug_type": "miscompilation", "base_commit": "a3a9ba80337ec1b276a7571049eda6ea03276ea3", "knowledge_cutoff": "2024-10-15T13:21:27Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "9b7491e8669126180253480821d5addde34874d0", "components": ["IR"], "bug_location_lineno": {"llvm/lib/IR/Operator.cpp": [[50, 55]]}, "bug_location_funcname": {"llvm/lib/IR/Operator.cpp": ["Operator::hasPoisonGeneratingFlags"]}}, "patch": "commit 9b7491e8669126180253480821d5addde34874d0\nAuthor: Yingwei Zheng \nDate: Tue Oct 15 23:07:16 2024 +0800\n\n [IR] Add support for `samesign` in `Operator::hasPoisonGeneratingFlags` (#112358)\n \n Fix https://github.com/llvm/llvm-project/issues/112356.\n\ndiff --git a/llvm/lib/IR/Operator.cpp b/llvm/lib/IR/Operator.cpp\nindex f93ff8f6fc8a..199eb4d90f55 100644\n--- a/llvm/lib/IR/Operator.cpp\n+++ b/llvm/lib/IR/Operator.cpp\n@@ -50,6 +50,8 @@ bool Operator::hasPoisonGeneratingFlags() const {\n if (auto *NNI = dyn_cast(this))\n return NNI->hasNonNeg();\n return false;\n+ case Instruction::ICmp:\n+ return cast(this)->hasSameSign();\n default:\n if (const auto *FP = dyn_cast(this))\n return FP->hasNoNaNs() || FP->hasNoInfs();\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/icmp.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "icmp_samesign_logical_or", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @icmp_samesign_logical_or(i32 %In) {\n %c1 = icmp samesign slt i32 %In, 0\n %c2 = icmp samesign ne i32 %In, 1\n %V = select i1 %c1, i1 true, i1 %c2\n ret i1 %V\n}\n"}, {"test_name": "icmp_and_inv_pow2_or_zero_ne_0", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\ndefine i1 @icmp_and_inv_pow2_or_zero_ne_0(i32 %A, i32 %B) {\n %popcnt = tail call i32 @llvm.ctpop.i32(i32 %A)\n %cond = icmp ult i32 %popcnt, 2\n call void @llvm.assume(i1 %cond)\n %inv = xor i32 %B, -1\n %and = and i32 %A, %inv\n %cmp = icmp ne i32 %and, 0\n ret i1 %cmp\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #1\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\nattributes #1 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "icmp_samesign_logical_and", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @icmp_samesign_logical_and(i32 %In) {\n %c1 = icmp samesign sgt i32 %In, -1\n %c2 = icmp samesign eq i32 %In, 1\n %V = select i1 %c1, i1 %c2, i1 false\n ret i1 %V\n}\n"}]}], "issue": {"title": "[InstCombine] `samesign` flag should be dropped", "body": "Reproducer:\r\n```\r\ndefine i1 @src(i32 %In) {\r\n %c1 = icmp samesign sgt i32 %In, -1\r\n %c2 = icmp samesign eq i32 %In, 1\r\n %V = select i1 %c1, i1 %c2, i1 false\r\n ret i1 %V\r\n}\r\n\r\ndefine i1 @tgt(i32 %In) {\r\n %c2 = icmp samesign eq i32 %In, 1\r\n ret i1 %c2\r\n}\r\n```\r\n\r\nWe should check this flag in `Operator::hasPoisonGeneratingFlags`.\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:ir"], "comments": []}, "verified": true} {"bug_id": "94897", "issue_url": "https://github.com/llvm/llvm-project/issues/94897", "bug_type": "miscompilation", "base_commit": "add89088fb8de84fdbeb97c0386a831e51d090a1", "knowledge_cutoff": "2024-06-09T10:15:35Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "e4b0655b297e4f4b0eb26548c8be48d527ebc178", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": [[2441, 2449]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": ["InstCombinerImpl::foldICmpShlConstant"]}}, "patch": "commit e4b0655b297e4f4b0eb26548c8be48d527ebc178\nAuthor: Yingwei Zheng \nDate: Mon Jun 10 03:17:01 2024 +0800\n\n [InstCombine] Fix missing argument typo in `InstCombinerImpl::foldICmpShlConstant` (#94899)\n \n Closes #94897.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\nindex 4203147bc6a5..3fce4a3c20b3 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n@@ -2441,9 +2441,10 @@ Instruction *InstCombinerImpl::foldICmpShlConstant(ICmpInst &Cmp,\n Type *TruncTy = ShType->getWithNewBitWidth(TypeBits - Amt);\n Constant *NewC =\n ConstantInt::get(TruncTy, RHSC.ashr(*ShiftAmt).trunc(TypeBits - Amt));\n- return new ICmpInst(\n- CmpPred, Builder.CreateTrunc(X, TruncTy, \"\", Shl->hasNoSignedWrap()),\n- NewC);\n+ return new ICmpInst(CmpPred,\n+ Builder.CreateTrunc(X, TruncTy, \"\", /*IsNUW=*/false,\n+ Shl->hasNoSignedWrap()),\n+ NewC);\n }\n }\n \n", "tests": [{"file": "llvm/test/Transforms/InstCombine/icmp.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "test_icmp_shl_sgt", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @test_icmp_shl_sgt(i64 %x) {\n %shl = shl i64 %x, 32\n %cmp = icmp sgt i64 %shl, 8589934591\n ret i1 %cmp\n}\n"}, {"test_name": "test_icmp_shl_nuw_i31", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @test_icmp_shl_nuw_i31(i31 %x) {\n %shl = shl nuw i31 %x, 23\n %cmp = icmp ugt i31 %shl, -50331648\n ret i1 %cmp\n}\n"}, {"test_name": "test_icmp_shl_vec", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine <2 x i1> @test_icmp_shl_vec(<2 x i64> %x) {\n %shl = shl <2 x i64> %x, splat (i64 32)\n %cmp = icmp ult <2 x i64> %shl, splat (i64 8589934593)\n ret <2 x i1> %cmp\n}\n"}, {"test_name": "test_icmp_shl_nsw_i31", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @test_icmp_shl_nsw_i31(i31 %x) {\n %shl = shl nsw i31 %x, 23\n %cmp = icmp ugt i31 %shl, -50331648\n ret i1 %cmp\n}\n"}, {"test_name": "test_icmp_shl_nsw", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @test_icmp_shl_nsw(i64 %x) {\n %shl = shl nsw i64 %x, 32\n %cmp = icmp ult i64 %shl, 8589934593\n ret i1 %cmp\n}\n"}, {"test_name": "pr94897", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @pr94897(i32 range(i32 -2147483648, 0) %x) {\n %shl = shl nsw i32 %x, 24\n %cmp = icmp ugt i32 %shl, -50331648\n ret i1 %cmp\n}\n"}, {"test_name": "test_icmp_shl_nuw", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @test_icmp_shl_nuw(i64 %x) {\n %shl = shl nuw i64 %x, 32\n %cmp = icmp ult i64 %shl, 8589934593\n ret i1 %cmp\n}\n"}]}], "issue": {"title": "wrong code at -O1 and above on x86_64-linux-gnu", "body": "This appears to be a recent regression as it doesn't reproduce with 18.1.0 and earlier.\r\n\r\nCompiler Explorer: https://godbolt.org/z/T9b4a7z3v\r\n\r\n```console\r\n% clangtk -v\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git 338cbfef03e0ab58d7b52f3301928c58b194a1b4)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n% \r\n% clangtk -O0 small.c; ./a.out\r\n% clangtk -O1 small.c\r\n% ./a.out\r\nAborted\r\n% cat small.c\r\nint a;\r\nchar b(char c, char d) { return c - d; }\r\nint main() {\r\n int e;\r\n for (a = -10; a > -11; a--)\r\n e = b(a, -1);\r\n if (e > -2)\r\n __builtin_abort();\r\n return 0;\r\n}\r\n```", "author": "zhendongsu", "labels": ["miscompilation", "confirmed", "llvm:instcombine"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/_3cyuT\r\n```ll\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\n\r\ndefine i1 @src() {\r\nentry:\r\n br label %for.body\r\n\r\nfor.body:\r\n %storemerge4 = phi i32 [ -10, %entry ], [ %dec, %for.body ]\r\n %dec = add nsw i32 %storemerge4, -1\r\n %cmp = icmp ugt i32 %dec, -11\r\n br i1 %cmp, label %for.body, label %for.end\r\n\r\nfor.end:\r\n %sub.i.le = shl nsw i32 %storemerge4, 24\r\n %cmp2 = icmp ugt i32 %sub.i.le, -50331648\r\n ret i1 %cmp2\r\n}\r\n\r\ndefine i1 @tgt() {\r\nentry:\r\n br label %for.body\r\n\r\nfor.body:\r\n %storemerge4 = phi i32 [ -10, %entry ], [ %dec, %for.body ]\r\n %dec = add nsw i32 %storemerge4, -1\r\n %cmp = icmp ugt i32 %dec, -11\r\n br i1 %cmp, label %for.body, label %for.end\r\n\r\nfor.end:\r\n %0 = trunc nuw i32 %storemerge4 to i8\r\n %cmp2 = icmp ugt i8 %0, -3\r\n ret i1 %cmp2\r\n}\r\n```\r\n\r\n`nuw` flag is incorrectly added."}, {"author": "dtcxzyw", "body": "Caused by https://github.com/llvm/llvm-project/commit/5c7c1f6aba467b1c5a51d58fc591314ecd7147ce."}]}, "verified": true} {"bug_id": "77320", "issue_url": "https://github.com/llvm/llvm-project/issues/77320", "bug_type": "miscompilation", "base_commit": "ade7ae4760a0b0e74cddd8f852830ca946295930", "knowledge_cutoff": "2024-01-08T14:43:50Z", "lit_test_dir": ["llvm/test/Transforms/InstSimplify"], "hints": {"fix_commit": "97e3220d6312ae00bcbe08673f218bd0f705776b", "components": ["InstructionSimplify"], "bug_location_lineno": {"llvm/lib/Analysis/InstructionSimplify.cpp": [[4313, 4319]]}, "bug_location_funcname": {"llvm/lib/Analysis/InstructionSimplify.cpp": ["simplifyWithOpReplaced"]}}, "patch": "commit 97e3220d6312ae00bcbe08673f218bd0f705776b\nAuthor: Nikita Popov \nDate: Mon Jan 8 15:49:47 2024 +0100\n\n [InstSimplify] Consider bitcast as potential cross-lane operation\n \n The bitcast might change the number of vector lanes, in which case\n it will be a cross-lane operation.\n \n Fixes https://github.com/llvm/llvm-project/issues/77320.\n\ndiff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp\nindex 241bdd81b75a..d0c27cae0dff 100644\n--- a/llvm/lib/Analysis/InstructionSimplify.cpp\n+++ b/llvm/lib/Analysis/InstructionSimplify.cpp\n@@ -4313,7 +4313,7 @@ static Value *simplifyWithOpReplaced(Value *V, Value *Op, Value *RepOp,\n // For vector types, the simplification must hold per-lane, so forbid\n // potentially cross-lane operations like shufflevector.\n if (!I->getType()->isVectorTy() || isa(I) ||\n- isa(I))\n+ isa(I) || isa(I))\n return nullptr;\n }\n \n", "tests": [{"file": "llvm/test/Transforms/InstSimplify/select.ll", "commands": ["opt < %s -passes=instsimplify -S"], "tests": [{"test_name": "select_vector_cmp_with_bitcasts", "test_body": "define <4 x i32> @select_vector_cmp_with_bitcasts(<2 x i64> %x, <4 x i32> %y) {\n %x.bc = bitcast <2 x i64> %x to <4 x i32>\n %y.bc = bitcast <4 x i32> %y to <2 x i64>\n %sub = sub <2 x i64> %x, %y.bc\n %sub.bc = bitcast <2 x i64> %sub to <4 x i32>\n %cmp = icmp eq <4 x i32> %y, %x.bc\n %sel = select <4 x i1> %cmp, <4 x i32> %sub.bc, <4 x i32> zeroinitializer\n ret <4 x i32> %sel\n}\n"}, {"test_name": "select_or_disjoint_eq", "test_body": "define i8 @select_or_disjoint_eq(i8 %x, i8 %y) {\n %cmp = icmp eq i8 %x, %y\n %or = or disjoint i8 %x, %y\n %sel = select i1 %cmp, i8 %x, i8 %or\n ret i8 %sel\n}\n"}]}], "issue": {"title": "[InstSimplify] Miscompile with select of vectors and bitcast", "body": "https://alive2.llvm.org/ce/z/xhCFyk\r\n\r\n```llvm\r\ndefine <4 x i32> @src(<2 x i64> %x, <4 x i32> %y) {\r\n %x.bc = bitcast <2 x i64> %x to <4 x i32>\r\n %y.bc = bitcast <4 x i32> %y to <2 x i64>\r\n %sub = sub <2 x i64> %x, %y.bc\r\n %sub.bc = bitcast <2 x i64> %sub to <4 x i32>\r\n %cmp = icmp eq <4 x i32> %y, %x.bc\r\n %sel = select <4 x i1> %cmp, <4 x i32> %sub.bc, <4 x i32> zeroinitializer\r\n ret <4 x i32> %sel\r\n}\r\n\r\ndefine <4 x i32> @tgt(<2 x i64> %x, <4 x i32> %y) {\r\n ret <4 x i32> zeroinitializer\r\n}\r\n```", "author": "nikic", "labels": ["test-suite", "miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "118798", "issue_url": "https://github.com/llvm/llvm-project/issues/118798", "bug_type": "hang", "base_commit": "8a90b5b31776957cb9cfa4f9547d19544640c5cc", "knowledge_cutoff": "2024-12-05T13:19:25Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "27eaa8a40ef338bea611f693fa8e810b1fc8d78d", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp": [[427, 433]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp": ["InstCombinerImpl::commonShiftTransforms"]}}, "patch": "commit 27eaa8a40ef338bea611f693fa8e810b1fc8d78d\nAuthor: Maurice Heumann \nDate: Thu Dec 5 16:57:27 2024 +0100\n\n [InstCombine] Prevent infinite loop with two shifts (#118806)\n \n The following pattern: `(C2 << X) << C1` will usually be transformed\n into `(C2 << C1) << X`, essentially swapping `X` and `C1`.\n \n However, this should only be done when `C1` is an immediate constant,\n otherwise thiscan lead to both constants being swapped forever.\n \n This fixes #118798.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp\nindex 10c3ccdb2243..d511e79e3e48 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp\n@@ -427,7 +427,8 @@ Instruction *InstCombinerImpl::commonShiftTransforms(BinaryOperator &I) {\n if (Instruction *R = FoldOpIntoSelect(I, SI))\n return R;\n \n- if (Constant *CUI = dyn_cast(Op1))\n+ Constant *CUI;\n+ if (match(Op1, m_ImmConstant(CUI)))\n if (Instruction *Res = FoldShiftByConstant(Op0, CUI, I))\n return Res;\n \n", "tests": [{"file": "llvm/test/Transforms/InstCombine/shl-twice-constant.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "testfunc", "test_body": "@c = external constant i8\n@c2 = external constant i8\n\ndefine i64 @testfunc() {\n %shl1 = shl i64 1, ptrtoint (ptr @c2 to i64)\n %shl2 = shl i64 %shl1, ptrtoint (ptr @c to i64)\n ret i64 %shl2\n}\n"}]}], "issue": {"title": "[InstCombine] Infinite loop when combining two `shl` instructions", "body": "The following sample triggers an infinite loop in instcombine:\n\n```\n@c = external constant i8\n@c2 = external constant i8\n\ndefine i16 @testfunc() {\nentry:\n %0 = shl i64 1, ptrtoint (ptr @c2 to i64)\n %1 = shl i64 %0, ptrtoint (ptr @c to i64)\n %2 = inttoptr i64 %1 to ptr\n %3 = load i16, ptr %2, align 1\n ret i16 %3\n}\n```\n\nhttps://godbolt.org/z/WP6YvGjPq\n", "author": "momo5502", "labels": ["llvm:instcombine", "llvm:hang"], "comments": [{"author": "momo5502", "body": "I have a potential fix and will create a PR soon"}]}, "verified": true} {"bug_id": "117133", "issue_url": "https://github.com/llvm/llvm-project/issues/117133", "bug_type": "miscompilation", "base_commit": "94df95de6bab20930c1a77bc785221ff2f670dba", "knowledge_cutoff": "2024-11-21T09:18:16Z", "lit_test_dir": ["llvm/test/Analysis/ScalarEvolution", "llvm/test/Transforms/IndVarSimplify"], "hints": {"fix_commit": "f7ef0721d60f85e1f699f8d1b83d4402ae19b122", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/include/llvm/Analysis/ScalarEvolution.h": [[2187, 2192]], "llvm/lib/Analysis/ScalarEvolution.cpp": [[4307, 4321], [5933, 5950], [7324, 7329]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::createAddRecFromPHI", "ScalarEvolution::getSequentialMinMaxExpr", "ScalarEvolution::isGuaranteedNotToBePoison", "ScalarEvolution::isSCEVExprNeverPoison"]}}, "patch": "commit f7ef0721d60f85e1f699f8d1b83d4402ae19b122\nAuthor: Yingwei Zheng \nDate: Sun Dec 1 20:11:09 2024 +0800\n\n [SCEV] Do not allow refinement in the rewriting of BEValue (#117152)\n \n See the following case:\n ```\n ; bin/opt -passes=\"print\" test.ll --disable-output\n define i32 @widget() {\n b:\n br label %b1\n \n b1: ; preds = %b5, %b\n %phi = phi i32 [ 0, %b ], [ %udiv6, %b5 ]\n %phi2 = phi i32 [ 1, %b ], [ %add, %b5 ]\n %icmp = icmp eq i32 %phi, 0\n br i1 %icmp, label %b3, label %b8\n \n b3: ; preds = %b1\n %udiv = udiv i32 10, %phi2\n %urem = urem i32 %udiv, 10\n %icmp4 = icmp eq i32 %urem, 0\n br i1 %icmp4, label %b7, label %b5\n \n b5: ; preds = %b3\n %udiv6 = udiv i32 %phi2, 0\n %add = add i32 %phi2, 1\n br label %b1\n \n b7: ; preds = %b3\n ret i32 5\n \n b8: ; preds = %b1\n ret i32 7\n }\n ```\n ```\n %phi2 = phi i32 [ 1, %b ], [ %add, %b5 ] --> {1,+,1}<%b1>\n %udiv6 = udiv i32 %phi2, 0 --> ({1,+,1}<%b1> /u 0)\n %phi = phi i32 [ 0, %b ], [ %udiv6, %b5 ] --> ({0,+,1}<%b1> /u 0)\n ```\n `ScalarEvolution::createAddRecFromPHI` gives a wrong SCEV result for\n `%phi`:\n \n https://github.com/llvm/llvm-project/blob/d7d6fb1804415b0f3e7f1cc9290bfb3d711cb707/llvm/lib/Analysis/ScalarEvolution.cpp#L5926-L5950\n It converts `phi(0, ({1,+,1}<%b1> /u 0))` into `phi(0 / 0,\n ({1,+,1}<%b1> /u 0))`. Then it simplifies the expr into\n `{0,+,1}<%b1> /u 0`.\n \n As we did in\n https://github.com/llvm/llvm-project/commit/acd700a24b6f767413db3d525e06d03e4245aa40,\n this patch disallows udiv simplification if we cannot prove that the\n denominator is a well-defined non-zero value.\n \n Fixes https://github.com/llvm/llvm-project/issues/117133.\n\ndiff --git a/llvm/include/llvm/Analysis/ScalarEvolution.h b/llvm/include/llvm/Analysis/ScalarEvolution.h\nindex 885c5985f9d2..b20c6a13cb6b 100644\n--- a/llvm/include/llvm/Analysis/ScalarEvolution.h\n+++ b/llvm/include/llvm/Analysis/ScalarEvolution.h\n@@ -2187,6 +2187,9 @@ private:\n bool isGuaranteedToTransferExecutionTo(const Instruction *A,\n const Instruction *B);\n \n+ /// Returns true if \\p Op is guaranteed not to cause immediate UB.\n+ bool isGuaranteedNotToCauseUB(const SCEV *Op);\n+\n /// Returns true if \\p Op is guaranteed to not be poison.\n static bool isGuaranteedNotToBePoison(const SCEV *Op);\n \ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex c3f296b9ff33..882e938e69c0 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -4307,15 +4307,7 @@ ScalarEvolution::getSequentialMinMaxExpr(SCEVTypes Kind,\n }\n \n for (unsigned i = 1, e = Ops.size(); i != e; ++i) {\n- bool MayBeUB = SCEVExprContains(Ops[i], [this](const SCEV *S) {\n- auto *UDiv = dyn_cast(S);\n- // The UDiv may be UB if the divisor is poison or zero. Unless the divisor\n- // is a non-zero constant, we have to assume the UDiv may be UB.\n- return UDiv && (!isKnownNonZero(UDiv->getOperand(1)) ||\n- !isGuaranteedNotToBePoison(UDiv->getOperand(1)));\n- });\n-\n- if (MayBeUB)\n+ if (!isGuaranteedNotToCauseUB(Ops[i]))\n continue;\n // We can replace %x umin_seq %y with %x umin %y if either:\n // * %y being poison implies %x is also poison.\n@@ -5933,18 +5925,22 @@ const SCEV *ScalarEvolution::createAddRecFromPHI(PHINode *PN) {\n // We can generalize this saying that i is the shifted value of BEValue\n // by one iteration:\n // PHI(f(0), f({1,+,1})) --> f({0,+,1})\n- const SCEV *Shifted = SCEVShiftRewriter::rewrite(BEValue, L, *this);\n- const SCEV *Start = SCEVInitRewriter::rewrite(Shifted, L, *this, false);\n- if (Shifted != getCouldNotCompute() &&\n- Start != getCouldNotCompute()) {\n- const SCEV *StartVal = getSCEV(StartValueV);\n- if (Start == StartVal) {\n- // Okay, for the entire analysis of this edge we assumed the PHI\n- // to be symbolic. We now need to go back and purge all of the\n- // entries for the scalars that use the symbolic expression.\n- forgetMemoizedResults(SymbolicName);\n- insertValueToMap(PN, Shifted);\n- return Shifted;\n+\n+ // Do not allow refinement in rewriting of BEValue.\n+ if (isGuaranteedNotToCauseUB(BEValue)) {\n+ const SCEV *Shifted = SCEVShiftRewriter::rewrite(BEValue, L, *this);\n+ const SCEV *Start = SCEVInitRewriter::rewrite(Shifted, L, *this, false);\n+ if (Shifted != getCouldNotCompute() && Start != getCouldNotCompute() &&\n+ ::impliesPoison(BEValue, Start)) {\n+ const SCEV *StartVal = getSCEV(StartValueV);\n+ if (Start == StartVal) {\n+ // Okay, for the entire analysis of this edge we assumed the PHI\n+ // to be symbolic. We now need to go back and purge all of the\n+ // entries for the scalars that use the symbolic expression.\n+ forgetMemoizedResults(SymbolicName);\n+ insertValueToMap(PN, Shifted);\n+ return Shifted;\n+ }\n }\n }\n }\n@@ -7324,6 +7320,16 @@ bool ScalarEvolution::isGuaranteedNotToBePoison(const SCEV *Op) {\n return PC.MaybePoison.empty();\n }\n \n+bool ScalarEvolution::isGuaranteedNotToCauseUB(const SCEV *Op) {\n+ return !SCEVExprContains(Op, [this](const SCEV *S) {\n+ auto *UDiv = dyn_cast(S);\n+ // The UDiv may be UB if the divisor is poison or zero. Unless the divisor\n+ // is a non-zero constant, we have to assume the UDiv may be UB.\n+ return UDiv && (!isKnownNonZero(UDiv->getOperand(1)) ||\n+ !isGuaranteedNotToBePoison(UDiv->getOperand(1)));\n+ });\n+}\n+\n bool ScalarEvolution::isSCEVExprNeverPoison(const Instruction *I) {\n // Only proceed if we can prove that I does not yield poison.\n if (!programUndefinedIfPoison(I))\n", "tests": [{"file": "llvm/test/Analysis/ScalarEvolution/pr117133.ll", "commands": ["opt -S -disable-output \"-passes=print\" < %s 2>&1"], "tests": [{"test_name": "test_poisonous", "test_body": "define i64 @test_poisonous(i64 %a, i32 %n) {\nentry:\n br label %loop.body\n\nloop.body: ; preds = %loop.body, %entry\n %indvar1 = phi i32 [ 0, %entry ], [ %indvar1.next, %loop.body ]\n %indvar2 = phi i64 [ 0, %entry ], [ %mul, %loop.body ]\n %indvar1.next = add i32 %indvar1, 1\n %ext = zext i32 %indvar1.next to i64\n %mul = mul i64 %ext, %a\n %exitcond = icmp sgt i32 %indvar1.next, %n\n br i1 %exitcond, label %loop.exit, label %loop.body\n\nloop.exit: ; preds = %loop.body\n ret i64 %mul\n}\n"}, {"test_name": "widget", "test_body": "define i32 @widget() {\nb:\n br label %b1\n\nb1: ; preds = %b5, %b\n %phi = phi i32 [ 0, %b ], [ %udiv6, %b5 ]\n %phi2 = phi i32 [ 1, %b ], [ %add, %b5 ]\n %icmp = icmp eq i32 %phi, 0\n br i1 %icmp, label %b3, label %b8\n\nb3: ; preds = %b1\n %udiv = udiv i32 10, %phi2\n %urem = urem i32 %udiv, 10\n %icmp4 = icmp eq i32 %urem, 0\n br i1 %icmp4, label %b7, label %b5\n\nb5: ; preds = %b3\n %udiv6 = udiv i32 %phi2, 0\n %add = add i32 %phi2, 1\n br label %b1\n\nb7: ; preds = %b3\n ret i32 5\n\nb8: ; preds = %b1\n ret i32 7\n}\n"}]}, {"file": "llvm/test/Transforms/IndVarSimplify/pr117133.ll", "commands": ["opt -S -passes=indvars < %s"], "tests": [{"test_name": "widget", "test_body": "define i32 @widget() {\nb:\n br label %b1\n\nb1: ; preds = %b5, %b\n %phi = phi i32 [ 0, %b ], [ %udiv6, %b5 ]\n %phi2 = phi i32 [ 1, %b ], [ %add, %b5 ]\n %icmp = icmp eq i32 %phi, 0\n br i1 %icmp, label %b3, label %b8\n\nb3: ; preds = %b1\n %udiv = udiv i32 10, %phi2\n %urem = urem i32 %udiv, 10\n %icmp4 = icmp eq i32 %urem, 0\n br i1 %icmp4, label %b7, label %b5\n\nb5: ; preds = %b3\n %udiv6 = udiv i32 %phi2, 0\n %add = add i32 %phi2, 1\n br label %b1\n\nb7: ; preds = %b3\n ret i32 5\n\nb8: ; preds = %b1\n ret i32 7\n}\n"}]}], "issue": {"title": "[indvars] Miscompile when loop body has an operation with Undefined Behaviour", "body": "Here is the problematic IR:\r\n```\r\ndefine i32 @widget() {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb5, %bb\r\n %phi = phi i32 [ 0, %bb ], [ %udiv6, %bb5 ]\r\n %phi2 = phi i32 [ 1, %bb ], [ %add, %bb5 ]\r\n %icmp = icmp eq i32 %phi, 0\r\n br i1 %icmp, label %bb3, label %bb8\r\n\r\nbb3: ; preds = %bb1\r\n %udiv = udiv i32 10, %phi2\r\n %urem = urem i32 %udiv, 10\r\n %icmp4 = icmp eq i32 %urem, 0\r\n br i1 %icmp4, label %bb7, label %bb5\r\n\r\nbb5: ; preds = %bb3\r\n %udiv6 = udiv i32 %phi2, 0\r\n %add = add i32 %phi2, 1\r\n br label %bb1\r\n\r\nbb7: ; preds = %bb3\r\n ret i32 5\r\n\r\nbb8: ; preds = %bb1\r\n ret i32 7\r\n}\r\n```\r\nproduces incorrect IR when `indvars` pass is run. I suspect this is because of the `%udiv6 = udiv i32 %phi2, 0` divide by zero operation. \r\nLook at the indvars transformation here: https://godbolt.org/z/cz1r5178h\r\nThe original IR must return 5, while the transformed IR returns 7\r\n\r\nProof of wrong transformation: https://alive2.llvm.org/ce/z/vPFhzg", "author": "Nirhar", "labels": ["miscompilation", "llvm:SCEV"], "comments": [{"author": "nikic", "body": "Is this a fuzzer-generated test case?"}, {"author": "Nirhar", "body": "> Is this a fuzzer-generated test case?\n\nHi @nikic , yes this test was fuzzer-generated."}]}, "verified": true} {"bug_id": "113794", "issue_url": "https://github.com/llvm/llvm-project/issues/113794", "bug_type": "crash", "base_commit": "21895a84cf466f0a0632cfca0175f7e617a6dd5a", "knowledge_cutoff": "2024-10-27T05:55:32Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "17bad1a9da228ade0c0b9b020ee99769b188d9b3", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6389, 6397]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInstructionCost", "LoopVectorizationCostModel::shouldConsiderInvariant"]}}, "patch": "commit 17bad1a9da228ade0c0b9b020ee99769b188d9b3\nAuthor: Florian Hahn \nDate: Fri Nov 1 20:51:25 2024 +0000\n\n [LV] Bail out on header phis in shouldConsiderInvariant.\n \n This fixes an infinite recursion in rare cases.\n \n Fixes https://github.com/llvm/llvm-project/issues/113794.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 659b4c30a58a..1c64bd2982d7 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6389,9 +6389,11 @@ bool LoopVectorizationCostModel::shouldConsiderInvariant(Value *Op) {\n return false;\n // Consider Op invariant, if it or its operands aren't predicated\n // instruction in the loop. In that case, it is not trivially hoistable.\n- return !isa(Op) || !TheLoop->contains(cast(Op)) ||\n- (!isPredicatedInst(cast(Op)) &&\n- all_of(cast(Op)->operands(),\n+ auto *OpI = dyn_cast(Op);\n+ return !OpI || !TheLoop->contains(OpI) ||\n+ (!isPredicatedInst(OpI) &&\n+ (!isa(OpI) || OpI->getParent() != TheLoop->getHeader()) &&\n+ all_of(OpI->operands(),\n [this](Value *Op) { return shouldConsiderInvariant(Op); }));\n }\n \n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/phi-cost.ll", "commands": ["opt < %s -passes='function(loop-vectorize,instcombine)' -force-vector-width=2 -force-vector-interleave=1 -debug-only=loop-vectorize -disable-output -print-after=instcombine 2>&1"], "tests": [{"test_name": "red_phi_0", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\n\ndefine i32 @red_phi_0(i32 %start, ptr %src) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %red = phi i32 [ %start, %entry ], [ %red.next, %loop ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %red.next = add i32 0, %red\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 100\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n %res = phi i32 [ %red.next, %loop ]\n ret i32 %res\n}\n"}, {"test_name": "phi_three_incoming_values", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\n\ndefine void @phi_three_incoming_values(ptr noalias %a, ptr noalias %b, i64 %n) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %if.end, %entry\n %i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]\n %tmp0 = getelementptr inbounds i32, ptr %a, i64 %i\n %tmp1 = load i32, ptr %tmp0, align 4\n %tmp2 = getelementptr inbounds i32, ptr %b, i64 %i\n %tmp3 = load i32, ptr %tmp2, align 4\n %tmp4 = icmp sgt i32 %tmp1, %tmp3\n br i1 %tmp4, label %if.then, label %if.end\n\nif.then: ; preds = %for.body\n %tmp5 = icmp sgt i32 %tmp1, 19\n br i1 %tmp5, label %if.end, label %if.else\n\nif.else: ; preds = %if.then\n %tmp6 = icmp slt i32 %tmp3, 4\n %tmp7 = select i1 %tmp6, i32 4, i32 5\n br label %if.end\n\nif.end: ; preds = %if.else, %if.then, %for.body\n %tmp8 = phi i32 [ 9, %for.body ], [ 3, %if.then ], [ %tmp7, %if.else ]\n store i32 %tmp8, ptr %tmp0, align 4\n %i.next = add i64 %i, 1\n %cond = icmp eq i64 %i, %n\n br i1 %cond, label %for.end, label %for.body\n\nfor.end: ; preds = %if.end\n ret void\n}\n"}, {"test_name": "phi_two_incoming_values", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\n\ndefine void @phi_two_incoming_values(ptr noalias %a, ptr noalias %b, i64 %n) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %if.end, %entry\n %i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]\n %tmp0 = getelementptr inbounds i32, ptr %a, i64 %i\n %tmp1 = load i32, ptr %tmp0, align 4\n %tmp2 = getelementptr inbounds i32, ptr %b, i64 %i\n %tmp3 = icmp sgt i32 %tmp1, 0\n br i1 %tmp3, label %if.then, label %if.end\n\nif.then: ; preds = %for.body\n %tmp4 = add i32 %tmp1, 1\n br label %if.end\n\nif.end: ; preds = %if.then, %for.body\n %tmp5 = phi i32 [ %tmp1, %for.body ], [ %tmp4, %if.then ]\n store i32 %tmp5, ptr %tmp2, align 4\n %i.next = add i64 %i, 1\n %cond = icmp eq i64 %i, %n\n br i1 %cond, label %for.end, label %for.body\n\nfor.end: ; preds = %if.end\n ret void\n}\n"}]}], "issue": {"title": "[clang] Segmentation fault in loop vectorization cost model during optimization pass", "body": "When I compiled this code with -O2 flag, it crashed:\r\n\r\n```c\r\n#include \r\n#define UNDEFINED(x) ((x), 0)\r\nlong b;\r\nstatic int64_t j = 1;\r\nstatic int64_t *k = &j;\r\nstatic int64_t **l = &k;\r\nint64_t m;\r\nstatic uint64_t n;\r\nint16_t ***o;\r\nstatic int64_t ***p = &l;\r\nuint64_t **q;\r\n\r\nuint16_t\r\nsafe_mod_func_uint16_t_u_u(uint16_t ui1, uint16_t ui2)\r\n{\r\n return \r\n#ifndef UNSAFE\r\n (ui2 == 0) ? \r\n (UNDEFINED(ui1)) : \r\n#endif\r\n (ui1 % ui2);\r\n}\r\n\r\nint64_t\r\nsafe_mod_func_int64_t_s_s(int64_t si1, int64_t si2)\r\n{\r\n return \r\n#ifndef UNSAFE\r\n ((si2 == 0) || ((si1 == INT64_MIN) && (si2 == (-1)))) ? \r\n (UNDEFINED(si1)) : \r\n#endif\r\n (si1 % si2);\r\n}\r\n\r\nint a(int, int, int, int, int) {}\r\nlong c(int d, int *e) {\r\n switch (d) {\r\n case 9:\r\n return b;\r\n case 8:\r\n return *e;\r\n }\r\n}\r\nlong f(int g, long, int h) {\r\n long i = c(g, &h);\r\n return i;\r\n}\r\nint32_t r() { int32_t *s[] = {&k, &m}; }\r\nint32_t t(int32_t *aa) {\r\n uint64_t ab = 2599693332;\r\n uint8_t ac = 142;\r\n for (; n <= 4; n++)\r\n *aa = (safe_mod_func_uint16_t_u_u(\r\n safe_mod_func_int64_t_s_s(a(0, 0, 0, 0, **l + 20) + **l, **q),\r\n ***o |= f(**l, 0, ***p)),\r\n **l + ac + ab - -1695273821 + *aa);\r\n}\r\n```\r\n\r\nThe crash is: clang: error: unable to execute command: Segmentation fault (core dumped)\r\nclang: error: clang frontend command failed due to signal (use -v to see invocation)\r\nCompiler returned: 254\r\n\r\nDetails can be found here: https://godbolt.org/z/3h9j7W6dG", "author": "cardigan1008", "labels": ["regression", "vectorizers", "crash"], "comments": [{"author": "shafik", "body": "This looks like a clang trunk regression: https://godbolt.org/z/MoKeds3ev"}]}, "verified": true} {"bug_id": "99411", "issue_url": "https://github.com/llvm/llvm-project/issues/99411", "bug_type": "crash", "base_commit": "cb3de24b5c0a662ba4a03c6c06a2d765d558bf62", "knowledge_cutoff": "2024-07-18T00:03:03Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "b634e057ddecc41dce046887d0f0854fed305374", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[11852, 11859]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["castToScalarTyElem"]}}, "patch": "commit b634e057ddecc41dce046887d0f0854fed305374\nAuthor: Han-Kuan Chen \nDate: Thu Jul 18 19:54:46 2024 +0800\n\n [SLP][REVEC] Fix false assumption of the source for castToScalarTyElem. (#99424)\n \n The argument V may come from adjustExtracts, which is the vector operand\n of ExtractElementInst. In addition, it is not existed in getTreeEntry.\n \n The vector operand of ExtractElementInst may have a type of <1 x Ty>,\n ensuring that the number of elements in ScalarTy and VecTy are equal.\n \n reference: https://github.com/llvm/llvm-project/issues/99411\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex d88c6307b994..b994645cece6 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -11852,8 +11852,7 @@ class BoUpSLP::ShuffleInstructionBuilder final : public BaseShuffleAnalysis {\n Value *castToScalarTyElem(Value *V,\n std::optional IsSigned = std::nullopt) {\n auto *VecTy = cast(V->getType());\n- assert(getNumElements(ScalarTy) < getNumElements(VecTy) &&\n- (getNumElements(VecTy) % getNumElements(ScalarTy) == 0));\n+ assert(getNumElements(VecTy) % getNumElements(ScalarTy) == 0);\n if (VecTy->getElementType() == ScalarTy->getScalarType())\n return V;\n return Builder.CreateIntCast(\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/revec-fix-99411.ll", "commands": ["opt -mtriple x86_64-unknown-linux-gnu -passes=slp-vectorizer -S %s"], "tests": [{"test_name": "", "test_body": "\ndefine void @e() {\n;\nentry:\n %0 = extractelement <1 x i64> zeroinitializer, i64 0\n %bf.value = and i64 %0, 0\n %bf.set = or i64 0, %bf.value\n store i64 %bf.set, ptr getelementptr inbounds (i8, ptr null, i64 8), align 8\n %bf.value2 = and i64 0, 0\n %bf.set4 = or i64 0, %bf.value2\n store i64 %bf.set4, ptr null, align 8\n ret void\n}"}]}], "issue": {"title": "A/F: `getNumElements(ScalarTy) < getNumElements(VecTy) && (getNumElements(VecTy) % getNumElements(ScalarTy) == 0)' failed after 1813ffd", "body": "Several of our internal tests started hitting an assertion failure while building, one of which I bisected back to commit 1813ffd. Consider the following code:\r\n```c++\r\ntypedef long long __m64 __attribute__((__vector_size__(8), __aligned__(8)));\r\nstatic __inline__ long long foo(__m64 m) { return (long long) m; }\r\nstruct c {\r\n long a : 43;\r\n long b : 61;\r\n} d;\r\nvoid e() {\r\n volatile __m64 f;\r\n long g, i = d.b = g = foo(f);\r\n d.a = i;\r\n}\r\n```\r\nWhen compiled with optimizations with a compiler that includes 1813ffd, it hits an assertion failure:\r\n```\r\n$ ~/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang -c -O2 repro.cpp \r\nclang: /home/dyung/src/upstream/llvm_clean_git/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:11853: llvm::Value* llvm::slpvectorizer::BoUpSLP::ShuffleInstructionBuilder::castToScalarTyElem(llvm::Value*, std::optional): Assertion `getNumElements(ScalarTy) < getNumElements(VecTy) && (getNumElements(VecTy) % getNumElements(ScalarTy) == 0)' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0. Program arguments: /home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang -c -O2 repro.cpp\r\n1. parser at end of file\r\n2. Optimizer\r\n3. Running pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"repro.cpp\" \r\n4. Running pass \"slp-vectorizer\" on function \"_Z7test119v\"\r\n #0 0x00005650f775179f llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x41a279f) \r\n #1 0x00005650f774f2fc llvm::sys::CleanupOnSignal(unsigned long) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x41a02fc) \r\n #2 0x00005650f76947e8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0 \r\n #3 0x00007ff89a3f9420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420) \r\n #4 0x00007ff899ec600b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1 \r\n #5 0x00007ff899ea5859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7 \r\n #6 0x00007ff899ea5729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8 \r\n #7 0x00007ff899ea5729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34 \r\n #8 0x00007ff899eb6fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6) \r\n #9 0x00005650f903005c llvm::slpvectorizer::BoUpSLP::ShuffleInstructionBuilder::castToScalarTyElem(llvm::Value*, std::optional) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5a8105c)\r\n#10 0x00005650f9071f00 llvm::slpvectorizer::BoUpSLP::ShuffleInstructionBuilder::adjustExtracts(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::MutableArrayRef, llvm::ArrayRef>, unsigned int, bool&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5ac2f00) \r\n#11 0x00005650f909dfdd llvm::Value* llvm::slpvectorizer::BoUpSLP::processBuildVector, llvm::slpvectorizer::BoUpSLP>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::IRBuilder&, llvm::slpvectorizer::BoUpSLP&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5aeefdd)\r\n#12 0x00005650f90a941a llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5afa41a)\r\n#13 0x00005650f90a80e3 llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5af90e3)\r\n#14 0x00005650f90a9484 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5afa484)\r\n#15 0x00005650f90a7cdf llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5af8cdf)\r\n#16 0x00005650f90ac099 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5afd099)\r\n#17 0x00005650f90a7cdf llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5af8cdf)\r\n#18 0x00005650f90a9e6c llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5afae6c)\r\n#19 0x00005650f90c7263 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::MapVector, llvm::DenseMap, llvm::detail::DenseMapPair>, llvm::SmallVector>, 0u>> const&, llvm::SmallVectorImpl>&, llvm::Instruction*) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5b18263)\r\n#20 0x00005650f90cacbc llvm::slpvectorizer::BoUpSLP::vectorizeTree() (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5b1bcbc)\r\n#21 0x00005650f90e045b llvm::SLPVectorizerPass::vectorizeStoreChain(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, unsigned int, unsigned int, unsigned int&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5b3145b) \r\n#22 0x00005650f90e1a5e llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::'lambda'(std::set, llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::StoreDistCompare, std::allocator>> const&)::operator()(std::set, llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::StoreDistCompare, std::allocator>> const&) const SLPVectorizer.cpp:0:0\r\n#23 0x00005650f90e39a0 llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5b349a0)\r\n#24 0x00005650f90e42c3 llvm::SLPVectorizerPass::vectorizeStoreChains(llvm::slpvectorizer::BoUpSLP&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5b352c3)\r\n#25 0x00005650f90e57dd llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#26 0x00005650f90e6253 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x5b37253)\r\n#27 0x00005650f8baedb6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x55ffdb6)\r\n#28 0x00005650f70f6421 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x3b47421)\r\n#29 0x00005650f4cb9ca6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x170aca6)\r\n#30 0x00005650f70f4ded llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x3b45ded)\r\n#31 0x00005650f4cbac56 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x170bc56)\r\n#32 0x00005650f70f2f31 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x3b43f31)\r\n#33 0x00005650f7a062c8 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#34 0x00005650f7a09645 (anonymous namespace)::EmitAssemblyHelper::EmitAssembly(clang::BackendAction, std::unique_ptr>, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#35 0x00005650f7a09ce6 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x445ace6)\r\n#36 0x00005650f808091c clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x4ad191c)\r\n#37 0x00005650f9e7fe6c clang::ParseAST(clang::Sema&, bool, bool) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x68d0e6c)\r\n#38 0x00005650f8080d28 clang::CodeGenAction::ExecuteAction() (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x4ad1d28)\r\n#39 0x00005650f8318339 clang::FrontendAction::Execute() (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x4d69339)\r\n#40 0x00005650f829a98e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x4ceb98e)\r\n#41 0x00005650f8403dc6 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x4e54dc6)\r\n#42 0x00005650f489f105 cc1_main(llvm::ArrayRef, char const*, void*) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x12f0105)\r\n#43 0x00005650f489811a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#44 0x00005650f80c610d void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\r\n#45 0x00005650f7694cf0 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x40e5cf0)\r\n#46 0x00005650f80c672f clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\r\n#47 0x00005650f808aa6c clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x4adba6c)\r\n#48 0x00005650f808b4fd clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x4adc4fd)\r\n#49 0x00005650f80935d5 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x4ae45d5)\r\n#50 0x00005650f489c37b clang_main(int, char**, llvm::ToolContext const&) (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x12ed37b)\r\n#51 0x00005650f478ee1b main (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x11dfe1b)\r\n#52 0x00007ff899ea7083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#53 0x00005650f4897bae _start (/home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin/clang+0x12e8bae)\r\nclang: error: clang frontend command failed with exit code 134 (use -v to see invocation)\r\nclang version 19.0.0 (https://github.com/llvm/llvm-project.git 1813ffd6b2eb04ee2c296a4399a18748740a439d)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /home/dyung/src/upstream/1813ffd6b2eb04ee2c296a4399a18748740a439d-linux/bin\r\nBuild config: +assertions\r\n```\r\nIf built using the a compiler built from fa0e529 (the parent commit of 1813ffd), the compilation succeeds:\r\n```\r\n$ ~/src/upstream/fa0e52995929ab67dfb468d71fe793be5e1c7f03-linux/bin/clang -c -O2 repro.cpp\r\n$\r\n```\r\n", "author": "dyung", "labels": ["llvm:SLPVectorizer", "crash"], "comments": [{"author": "madhur13490", "body": "@dyung https://godbolt.org/z/ofneqo6zK I don't see here. Am I missing anything?"}, {"author": "dyung", "body": "> @dyung https://godbolt.org/z/ofneqo6zK I don't see here. Am I missing anything?\r\n\r\nI don't know if it reproduces on an ARM targeting compiler, I used an x86-64 targeting x86-64 compiler. Also the compiler currently used by godbolt was built from commit f56cdd4a45b7bbe84be5d4ba9442eb7071605efc which does not include 1813ffd6b2eb04ee2c296a4399a18748740a439d which introduced the issue."}, {"author": "HanKuanChen", "body": "Close the issue since https://github.com/llvm/llvm-project/commit/b634e057ddecc41dce046887d0f0854fed305374 is merged."}]}, "verified": true} {"bug_id": "106780", "issue_url": "https://github.com/llvm/llvm-project/issues/106780", "bug_type": "crash", "base_commit": "5aa83eb677d2f8cba0c2ad3dcc14f3f4e80a0bba", "knowledge_cutoff": "2024-08-30T19:02:15Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "654bb4e9f287f9ccb5142d1c62a9ab17c16e29d7", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6817, 6822]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::collectValuesToIgnore"]}}, "patch": "commit 654bb4e9f287f9ccb5142d1c62a9ab17c16e29d7\nAuthor: Florian Hahn \nDate: Sun Sep 1 20:35:28 2024 +0100\n\n [LV] Don't consider branches leaving loop in collectValuesToIgnore.\n \n Branches exiting the loop will remain regardless, so don't consider them\n in collectValuesToIgnore.\n \n This fixes another divergence between legacy and VPlan-based cost model.\n \n Fixes https://github.com/llvm/llvm-project/issues/106780.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex fa05b8dd2242..a8722db654f5 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6817,6 +6817,9 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n if (auto *Br = dyn_cast_or_null(Op)) {\n BasicBlock *ThenBB = Br->getSuccessor(0);\n BasicBlock *ElseBB = Br->getSuccessor(1);\n+ // Don't considers branches leaving the loop for simplification.\n+ if (!TheLoop->contains(ThenBB) || !TheLoop->contains(ElseBB))\n+ continue;\n bool ThenEmpty = IsEmptyBlock(ThenBB);\n bool ElseEmpty = IsEmptyBlock(ElseBB);\n if ((ThenEmpty && ElseEmpty) ||\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll", "commands": ["opt -p loop-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S %s"], "tests": [{"test_name": "dead_load", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @dead_load(ptr %p, i16 %start) {\nentry:\n %start.ext = sext i16 %start to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %start.ext, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr i16, ptr %p, i64 %iv\n store i16 0, ptr %gep, align 2\n %l = load i16, ptr %gep, align 2\n %iv.next = add i64 %iv, 3\n %cmp = icmp slt i64 %iv, 111\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "cost_of_exit_branch_and_cond_insts", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i16 @llvm.umax.i16(i16, i16) #0\n\ndefine i32 @cost_of_exit_branch_and_cond_insts(ptr %a, ptr %b, i1 %c, i16 %x) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %then, label %loop.exiting\n\nthen: ; preds = %loop.header\n %gep = getelementptr inbounds i32, ptr %b, i32 %iv\n store i1 false, ptr %a, align 1\n store i32 0, ptr %gep, align 4\n br label %loop.exiting\n\nloop.exiting: ; preds = %then, %loop.header\n %iv.next = add i32 %iv, 1\n %umax = tail call i16 @llvm.umax.i16(i16 %x, i16 111)\n %umax.ext = zext i16 %umax to i32\n %sub = sub i32 770, %umax.ext\n %ec = icmp slt i32 %iv, %sub\n br i1 %ec, label %loop.latch, label %exit\n\nloop.latch: ; preds = %loop.exiting\n br label %loop.header\n\nexit: ; preds = %loop.exiting\n br label %return\n\nreturn: ; preds = %exit\n ret i32 0\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "dead_live_out_due_to_scalar_epilogue_required", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine i8 @dead_live_out_due_to_scalar_epilogue_required(ptr %src, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %idxprom = sext i32 %iv to i64\n %gep.src = getelementptr i8, ptr %src, i64 %idxprom\n %l = load i8, ptr %gep.src, align 1\n %gep.dst = getelementptr i8, ptr %dst, i64 %idxprom\n store i8 0, ptr %gep.dst, align 1\n %iv.next = add i32 %iv, 4\n %cmp = icmp ult i32 %iv, 1001\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n %r = phi i8 [ %l, %loop ]\n ret i8 %r\n}\n"}]}], "issue": {"title": "[VPlan] Assertion \" VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Tested using ef7b18a53c0d186dcda1e322be6035407fdedb55\r\n\r\nTestcase:\r\n```c\r\nunsigned short a;\r\nextern int b;\r\nvoid c(_Bool d, int e, long long f[], unsigned short g[][16],\r\n unsigned short h[][16], unsigned short i[][13], short j[][13],\r\n int k[][21], unsigned char l[][21][21], signed char m[][24]) {\r\n for (short n = 0; n < 5; n += -5532)\r\n for (char o = 0; o < 3; o += d)\r\n for (int p = 0; p < -({ g[d][1] > 0115 ? g[d][1] : 0115; }) + 770; p++)\r\n for (_Bool q = 0; q < (_Bool)e; q = 5)\r\n for (_Bool r = 2048749519 ? 8 ? f[d] : 0 : 0; r < 1; r = h)\r\n b = a = 0;\r\n}\r\nvoid hash();\r\nunsigned short a = 6;\r\nlong long f[6];\r\nunsigned short g[6][16];\r\nunsigned short h[1][16];\r\nunsigned short i[1][13];\r\nshort j[1][13];\r\nint k[1][21];\r\nunsigned char l[1][21][21];\r\nsigned char m[1][24];\r\nint main() {\r\n for (long s = 0; s < 6; ++s)\r\n f[s] = 6;\r\n for (long s = 0; s < 16; ++s)\r\n for (long n = 0; n < 16; ++n)\r\n g[s][n] = 7;\r\n c(1, 407418382874, f, g, h, i, j, k, l, m);\r\n hash(a);\r\n}\r\n```\r\nc command:\r\n```\r\n/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/clang -fno-strict-aliasing -march=rv64gcv -flto -O3 -fuse-ld=lld red.c -o rv64gcv.out -w\r\n```\r\n\r\nLLVM IR testcase:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i16 @llvm.umax.i16(i16, i16) #0\r\n\r\ndefine i32 @main(ptr %a) #1 {\r\nentry:\r\n br label %for.cond22.preheader.us.us76.us.us.us.us.i.1\r\n\r\nfor.cond22.preheader.us.us76.us.us.us.us.i.1: ; preds = %for.cond22.preheader.us.us76.us.us.us.usthread-pre-split.i.1, %entry\r\n %p.072.us.us77.us.us.us.us.i.1 = phi i32 [ %inc.us.us80.us.us.us.us.i.1, %for.cond22.preheader.us.us76.us.us.us.usthread-pre-split.i.1 ], [ 0, %entry ]\r\n br i1 false, label %for.body41.lr.ph.us.us.us.us.us.us.us.i.1, label %for.cond22.for.cond.cleanup28_crit_edge.us.us.us.us.us.us.i.1\r\n\r\nfor.body41.lr.ph.us.us.us.us.us.us.us.i.1: ; preds = %for.cond22.preheader.us.us76.us.us.us.us.i.1\r\n store i1 false, ptr %a, align 1\r\n store i32 0, ptr null, align 4\r\n br label %for.cond22.for.cond.cleanup28_crit_edge.us.us.us.us.us.us.i.1\r\n\r\nfor.cond22.for.cond.cleanup28_crit_edge.us.us.us.us.us.us.i.1: ; preds = %for.body41.lr.ph.us.us.us.us.us.us.us.i.1, %for.cond22.preheader.us.us76.us.us.us.us.i.1\r\n %inc.us.us80.us.us.us.us.i.1 = add i32 %p.072.us.us77.us.us.us.us.i.1, 1\r\n %narrow.us.us81.us.us.us.us.i.1 = tail call i16 @llvm.umax.i16(i16 0, i16 0)\r\n %spec.select.us.us82.us.us.us.us.i.1 = zext i16 %narrow.us.us81.us.us.us.us.i.1 to i32\r\n %add.us.us83.us.us.us.us.i.1 = sub i32 770, %spec.select.us.us82.us.us.us.us.i.1\r\n %cmp18.us.us84.us.us.us.us.i.1 = icmp slt i32 %p.072.us.us77.us.us.us.us.i.1, %add.us.us83.us.us.us.us.i.1\r\n br i1 %cmp18.us.us84.us.us.us.us.i.1, label %for.cond22.preheader.us.us76.us.us.us.usthread-pre-split.i.1, label %for.cond22.for.cond.cleanup28_crit_edge.us.us.us.us.us.us.i.1.for.cond.cleanup20.us.us.us.us.i.1_crit_edge\r\n\r\nfor.cond22.for.cond.cleanup28_crit_edge.us.us.us.us.us.us.i.1.for.cond.cleanup20.us.us.us.us.i.1_crit_edge: ; preds = %for.cond22.for.cond.cleanup28_crit_edge.us.us.us.us.us.us.i.1\r\n br label %for.cond.cleanup20.us.us.us.us.i.1\r\n\r\nfor.cond.cleanup20.us.us.us.us.i.1: ; preds = %for.cond22.for.cond.cleanup28_crit_edge.us.us.us.us.us.us.i.1.for.cond.cleanup20.us.us.us.us.i.1_crit_edge\r\n ret i32 0\r\n\r\nfor.cond22.preheader.us.us76.us.us.us.usthread-pre-split.i.1: ; preds = %for.cond22.for.cond.cleanup28_crit_edge.us.us.us.us.us.us.i.1\r\n br label %for.cond22.preheader.us.us76.us.us.us.us.i.1\r\n}\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\nattributes #1 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7366: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"main\"\r\n #0 0x00005e8221f7d690 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x301e690)\r\n #1 0x00005e8221f7aa9f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x301ba9f)\r\n #2 0x00005e8221f7abf5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x0000772c88c42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x0000772c88c969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x0000772c88c969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x0000772c88c969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x0000772c88c42476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x0000772c88c287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x0000772c88c2871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x0000772c88c39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00005e8220f4a3e0 llvm::LoopVectorizationPlanner::computeBestVF() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1feb3e0)\r\n#12 0x00005e8220f60ffb llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2001ffb)\r\n#13 0x00005e8220f64011 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2005011)\r\n#14 0x00005e8220f64687 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2005687)\r\n#15 0x00005e821fdc0fb6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe61fb6)\r\n#16 0x00005e8221d8858f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e2958f)\r\n#17 0x00005e821fdc4146 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe65146)\r\n#18 0x00005e8221d8730b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e2830b)\r\n#19 0x00005e821fdc03e6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe613e6)\r\n#20 0x00005e8221d8534d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e2634d)\r\n#21 0x00005e821f5e6ee6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x687ee6)\r\n#22 0x00005e821f5d8c71 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x679c71)\r\n#23 0x0000772c88c29d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#24 0x0000772c88c29e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#25 0x0000772c88c29e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#26 0x00005e821f5cf025 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x670025)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\nhttps://godbolt.org/z/ej1Gax953\r\n\r\nFound via fuzzer\r\n\r\nRelated/prior issue: https://github.com/llvm/llvm-project/issues/106641\r\n\r\ncc @fhahn", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "76986", "issue_url": "https://github.com/llvm/llvm-project/issues/76986", "bug_type": "crash", "base_commit": "71c17424b5d5c22c0ce6b4c41acaa0401515baca", "knowledge_cutoff": "2024-01-04T18:14:09Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "2ab5c47c8752b444885d6bfaf6f570a482fb4cdf", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": [[829, 843]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": ["simplifyRecipe"]}}, "patch": "commit 2ab5c47c8752b444885d6bfaf6f570a482fb4cdf\nAuthor: Florian Hahn \nDate: Thu Jan 4 20:39:44 2024 +0000\n\n [VPlan] Don't replace scalarizing recipe with VPWidenCastRecipe.\n \n Don't replace a scalarizing recipe with a VPWidenCastRecipe. This would\n introduce wide (vectorizing) recipes when interleaving only.\n \n Fixes https://github.com/llvm/llvm-project/issues/76986\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\nindex 33132880d5a4..5c430620a2dc 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n@@ -829,15 +829,20 @@ static void simplifyRecipe(VPRecipeBase &R, VPTypeAnalysis &TypeInfo) {\n Type *ATy = TypeInfo.inferScalarType(A);\n if (TruncTy == ATy) {\n Trunc->replaceAllUsesWith(A);\n- } else if (ATy->getScalarSizeInBits() < TruncTy->getScalarSizeInBits()) {\n- auto *VPC =\n- new VPWidenCastRecipe(Instruction::CastOps(ExtOpcode), A, TruncTy);\n- VPC->insertBefore(&R);\n- Trunc->replaceAllUsesWith(VPC);\n- } else if (ATy->getScalarSizeInBits() > TruncTy->getScalarSizeInBits()) {\n- auto *VPC = new VPWidenCastRecipe(Instruction::Trunc, A, TruncTy);\n- VPC->insertBefore(&R);\n- Trunc->replaceAllUsesWith(VPC);\n+ } else {\n+ // Don't replace a scalarizing recipe with a widened cast.\n+ if (isa(&R))\n+ break;\n+ if (ATy->getScalarSizeInBits() < TruncTy->getScalarSizeInBits()) {\n+ auto *VPC =\n+ new VPWidenCastRecipe(Instruction::CastOps(ExtOpcode), A, TruncTy);\n+ VPC->insertBefore(&R);\n+ Trunc->replaceAllUsesWith(VPC);\n+ } else if (ATy->getScalarSizeInBits() > TruncTy->getScalarSizeInBits()) {\n+ auto *VPC = new VPWidenCastRecipe(Instruction::Trunc, A, TruncTy);\n+ VPC->insertBefore(&R);\n+ Trunc->replaceAllUsesWith(VPC);\n+ }\n }\n #ifndef NDEBUG\n // Verify that the cached type info is for both A and its users is still\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/interleave-and-scalarize-only.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=1 -force-vector-interleave=2 -debug -disable-output %s 2>&1", "opt -passes=loop-vectorize -force-vector-width=1 -force-vector-interleave=2 -S %s"], "tests": [{"test_name": "test_scalarize_call", "test_body": "define void @test_scalarize_call(i32 %start, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ %start, %entry ], [ %iv.next, %loop ]\n %min = tail call i32 @llvm.smin.i32(i32 %iv, i32 65535)\n %arrayidx = getelementptr inbounds i32, ptr %dst, i32 %iv\n store i32 %min, ptr %arrayidx, align 8\n %iv.next = add nsw i32 %iv, 1\n %tobool.not = icmp eq i32 %iv.next, 1000\n br i1 %tobool.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.smin.i32(i32, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "scalarize_ptrtoint", "test_body": "define void @scalarize_ptrtoint(ptr %src, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr ptr, ptr %src, i64 %iv\n %l = load ptr, ptr %gep, align 8\n %cast = ptrtoint ptr %l to i64\n %add = add i64 %cast, 10\n %cast.2 = inttoptr i64 %add to ptr\n store ptr %cast.2, ptr %dst, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 0\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "first_order_recurrence_using_induction", "test_body": "define void @first_order_recurrence_using_induction(i32 %n, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %for = phi i32 [ 0, %entry ], [ %iv.trunc, %loop ]\n %iv.trunc = trunc i64 %iv to i32\n store i32 %for, ptr %dst, align 4\n %iv.next = add nuw nsw i64 %iv, 1\n %iv.next.trunc = trunc i64 %iv.next to i32\n %ec = icmp slt i32 %iv.next.trunc, %n\n br i1 %ec, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "reduction_with_casts", "test_body": "define i16 @reduction_with_casts() {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %count.0.in1 = phi i32 [ 0, %entry ], [ %add, %loop ]\n %iv = phi i16 [ 1, %entry ], [ %iv.next, %loop ]\n %conv1 = and i32 %count.0.in1, 65535\n %add = add nuw nsw i32 %conv1, 1\n %iv.next = add i16 %iv, 1\n %cmp = icmp eq i16 %iv.next, 10000\n br i1 %cmp, label %exit, label %loop\n\nexit: ; preds = %loop\n %add.lcssa = phi i32 [ %add, %loop ]\n %count.0 = trunc i32 %add.lcssa to i16\n ret i16 %count.0\n}\n"}, {"test_name": "pr76986_trunc_sext_interleaving_only", "test_body": "define void @pr76986_trunc_sext_interleaving_only(i16 %arg, ptr noalias %src, ptr noalias %dst) {\nbb:\n br label %loop\n\nloop: ; preds = %loop, %bb\n %iv = phi i64 [ 0, %bb ], [ %iv.next, %loop ]\n %gep.src = getelementptr inbounds i8, ptr %src, i64 %iv\n %l = load i8, ptr %gep.src, align 1\n %sext = sext i8 %l to i32\n %trunc = trunc i32 %sext to i16\n %sdiv = sdiv i16 %trunc, %arg\n %gep.dst = getelementptr inbounds i16, ptr %dst, i64 %iv\n store i16 %sdiv, ptr %gep.dst, align 2\n %iv.next = add i64 %iv, 1\n %icmp = icmp ult i64 %iv, 14933\n br i1 %icmp, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "test_scalarize_with_branch_cond", "test_body": "define void @test_scalarize_with_branch_cond(ptr %src, ptr %dst) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %d = phi i1 [ false, %entry ], [ %d.next, %loop.latch ]\n %d.next = xor i1 %d, true\n br i1 %d, label %cond.false, label %loop.latch\n\ncond.false: ; preds = %loop.header\n %gep.src = getelementptr inbounds i32, ptr %src, i64 %iv\n %gep.dst = getelementptr inbounds i32, ptr %dst, i64 %iv\n %l = load i32, ptr %gep.src, align 4\n store i32 %l, ptr %gep.dst, align 4\n br label %loop.latch\n\nloop.latch: ; preds = %cond.false, %loop.header\n %iv.next = add nsw i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}], "issue": {"title": "Assertion `State.VF.isVector() && \"Not vectorizing?\"' failed.", "body": "Run opt with -passes=loop-vectorize\r\n\r\nhttps://godbolt.org/z/xdrTfEddx\r\n\r\nTest.ll\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"aarch64-none-linux-gnu\"\r\n\r\ndefine void @wombat(i16 %arg) gc \"statepoint-example\" {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb1, %bb\r\n %phi = phi i64 [ 0, %bb ], [ %add, %bb1 ]\r\n %sext = sext i8 0 to i32\r\n %add = add i64 %phi, 1\r\n %trunc = trunc i32 %sext to i16\r\n %sdiv = sdiv i16 0, %arg\r\n %sext2 = sext i16 %trunc to i32\r\n %sext3 = sext i32 %sext2 to i64\r\n %icmp = icmp ult i64 %phi, 14933\r\n br i1 %icmp, label %bb1, label %bb4\r\n\r\nbb4: ; preds = %bb1\r\n %phi5 = phi i64 [ %sext3, %bb1 ]\r\n ret void\r\n}\r\n```", "author": "TatyanaDoubts", "labels": ["vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "Taking a look now!"}]}, "verified": true} {"bug_id": "91434", "issue_url": "https://github.com/llvm/llvm-project/issues/91434", "bug_type": "crash", "base_commit": "932f0de43a9e334e161a69a50bd6b01cd51e238e", "knowledge_cutoff": "2024-05-08T06:10:09Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "83d9aa27680b6a7f3556fcf13ada70b4be95bab2", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp": [[171, 176]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp": ["VPTypeAnalysis::inferScalarTypeForRecipe"]}}, "patch": "commit 83d9aa27680b6a7f3556fcf13ada70b4be95bab2\nAuthor: Pietro Ghiglio \nDate: Wed May 15 15:03:21 2024 +0200\n\n [VPlan] Add scalar inferencing support for addrspace cast (#92107)\n \n Fixes https://github.com/llvm/llvm-project/issues/91434\n \n PR: https://github.com/llvm/llvm-project/pull/92107\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp\nindex 5f93339083f0..efe8c21874a3 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp\n@@ -171,6 +171,7 @@ Type *VPTypeAnalysis::inferScalarTypeForRecipe(const VPReplicateRecipe *R) {\n case Instruction::ICmp:\n case Instruction::FCmp:\n return IntegerType::get(Ctx, 1);\n+ case Instruction::AddrSpaceCast:\n case Instruction::Alloca:\n case Instruction::BitCast:\n case Instruction::Trunc:\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/as_cast.ll", "commands": ["opt -passes=loop-vectorize %s -force-vector-width=1 -force-vector-interleave=2 -S -o -"], "tests": [{"test_name": "", "test_body": "\ndefine void @foo(ptr addrspace(1) %in) {\nentry:\n br label %loop\n\nloop:\n %iter = phi i64 [ %next, %loop ], [ 0, %entry ]\n %ascast = addrspacecast ptr addrspace(1) %in to ptr\n %next = add i64 %iter, 1\n %arrayidx = getelementptr inbounds i64, ptr %ascast, i64 %next\n store i64 %next, ptr %arrayidx, align 4\n\n; check that we find the two interleaved blocks with ascast, gep and store:\n\n\n %cmp = icmp eq i64 %next, 7\n br i1 %cmp, label %exit, label %loop\n\n; check that we branch to the exit block\n\nexit:\n ret void\n}\n"}]}], "issue": {"title": "[Loop Vectorizer] Unhandled opcode UNREACHABLE executed at llvm/llvm-project/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp:208!", "body": "```\r\nopt -passes=loop-vectorize before-loop-vectorize.ll\r\n\r\nUnhandled opcode\r\nUNREACHABLE executed at /llvm/llvm-project/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp:208!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /llvm/llvm-project/build/bin/opt -passes=loop-vectorize before-loop-vectorize.ll\r\n #0 0x0000000003167161 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/llvm/llvm-project/build/bin/opt+0x3167161)\r\n #1 0x00000000031646c4 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007fcb77b7bdb0 __restore_rt (/lib64/libc.so.6+0x59db0)\r\n #3 0x00007fcb77bc842c __pthread_kill_implementation (/lib64/libc.so.6+0xa642c)\r\n #4 0x00007fcb77b7bd06 gsignal (/lib64/libc.so.6+0x59d06)\r\n #5 0x00007fcb77b4e7d3 abort (/lib64/libc.so.6+0x2c7d3)\r\n #6 0x00000000030ac16a (/llvm/llvm-project/build/bin/opt+0x30ac16a)\r\n #7 0x00000000022ffc8d llvm::VPTypeAnalysis::inferScalarTypeForRecipe(llvm::VPReplicateRecipe const*) (/llvm/llvm-project/build/bin/opt+0x22ffc8d)\r\n #8 0x00000000022ff04c llvm::VPTypeAnalysis::inferScalarType(llvm::VPValue const*) (/llvm/llvm-project/build/bin/opt+0x22ff04c)\r\n #9 0x00000000021cb7f4 llvm::InnerLoopVectorizer::scalarizeInstruction(llvm::Instruction const*, llvm::VPReplicateRecipe*, llvm::VPIteration const&, llvm::VPTransformState&) (/llvm/llvm-project/build/bin/opt+0x21cb7f4)\r\n#10 0x00000000021cecb5 llvm::VPReplicateRecipe::execute(llvm::VPTransformState&) (/llvm/llvm-project/build/bin/opt+0x21cecb5)\r\n#11 0x00000000022f834f llvm::VPBasicBlock::execute(llvm::VPTransformState*) (/llvm/llvm-project/build/bin/opt+0x22f834f)\r\n#12 0x00000000022f0130 llvm::VPRegionBlock::execute(llvm::VPTransformState*) (/llvm/llvm-project/build/bin/opt+0x22f0130)\r\n#13 0x00000000022f03d8 llvm::VPRegionBlock::execute(llvm::VPTransformState*) (/llvm/llvm-project/build/bin/opt+0x22f03d8)\r\n#14 0x00000000022fb806 llvm::VPlan::execute(llvm::VPTransformState*) (/llvm/llvm-project/build/bin/opt+0x22fb806)\r\n#15 0x00000000021e983b llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/llvm/llvm-project/build/bin/opt+0x21e983b)\r\n#16 0x00000000021fbbe0 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/llvm/llvm-project/build/bin/opt+0x21fbbe0)\r\n#17 0x00000000021fde5e llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/llvm/llvm-project/build/bin/opt+0x21fde5e)\r\n#18 0x00000000021fef78 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/llvm/llvm-project/build/bin/opt+0x21fef78)\r\n#19 0x0000000001125c8e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/llvm/llvm-project/build/bin/opt+0x1125c8e)\r\n#20 0x0000000002f8671a llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/llvm/llvm-project/build/bin/opt+0x2f8671a)\r\n#21 0x000000000082c95e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/llvm/llvm-project/build/bin/opt+0x82c95e)\r\n#22 0x0000000002f853cd llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/llvm/llvm-project/build/bin/opt+0x2f853cd)\r\n#23 0x000000000082da6e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/llvm/llvm-project/build/bin/opt+0x82da6e)\r\n#24 0x0000000002f832cc llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/llvm/llvm-project/build/bin/opt+0x2f832cc)\r\n#25 0x00000000007a0d5d llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/llvm/llvm-project/build/bin/opt+0x7a0d5d)\r\n#26 0x0000000000793a7c optMain (/llvm/llvm-project/build/bin/opt+0x793a7c)\r\n#27 0x00007fcb77b66e50 __libc_start_call_main (/lib64/libc.so.6+0x44e50)\r\n#28 0x00007fcb77b66efc __libc_start_main@GLIBC_2.2.5 (/lib64/libc.so.6+0x44efc)\r\n#29 0x0000000000789e15 _start (/llvm/llvm-project/build/bin/opt+0x789e15)\r\nAborted (core dumped)\r\n```\r\nInput IR file before-loop-vectorize.ll is:\r\n```\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\n%\"class.sycl::_V1::range\" = type { %\"class.sycl::_V1::detail::array\" }\r\n%\"class.sycl::_V1::detail::array\" = type { [1 x i64] }\r\n\r\n$_ZTSN4sycl3_V16detail18RoundedRangeKernelINS0_4itemILi1ELb1EEELi1EZZ12foo_parallelRNS0_5queueERNS0_6bufferIfLi1ENS1_17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESD_EUlNS0_2idILi1EEEE_EE = comdat any\r\n\r\n$_ZTSZZ12foo_parallelRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_EUlNS0_2idILi1EEEE_ = comdat any\r\n\r\n$_ZTSZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_EUlvE_ = comdat any\r\n\r\n; Function Attrs: convergent norecurse nounwind\r\ndefine weak_odr dso_local spir_kernel void @_ZTSN4sycl3_V16detail18RoundedRangeKernelINS0_4itemILi1ELb1EEELi1EZZ12foo_parallelRNS0_5queueERNS0_6bufferIfLi1ENS1_17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESD_EUlNS0_2idILi1EEEE_EE(ptr noundef byval(%\"class.sycl::_V1::range\") align 8 %_arg_UserRange, ptr addrspace(1) noundef align 4 %_arg_acc, ptr noundef byval(%\"class.sycl::_V1::range\") align 8 %_arg_acc3) local_unnamed_addr #0 comdat !srcloc !7 !kernel_arg_buffer_location !8 !kernel_arg_runtime_aligned !9 !kernel_arg_exclusive_ptr !9 !sycl_fixed_targets !10 !sycl_kernel_omit_args !11 {\r\nentry:\r\n %0 = load i64, ptr %_arg_UserRange, align 8\r\n %agg.tmp7.sroa.0.0.copyload = load i64, ptr %_arg_acc3, align 8\r\n %add.ptr.i = getelementptr inbounds float, ptr addrspace(1) %_arg_acc, i64 %agg.tmp7.sroa.0.0.copyload\r\n %call.i = tail call i64 @_Z21__spirv_WorkgroupId_xv() #5, !noalias !12\r\n %call1.i = tail call i64 @_Z23__spirv_WorkgroupSize_xv() #5, !noalias !12\r\n %mul.i = mul i64 %call1.i, %call.i\r\n %call2.i = tail call i64 @_Z27__spirv_LocalInvocationId_xv() #5, !noalias !12\r\n %add.i = add i64 %mul.i, %call2.i\r\n %call3.i = tail call i64 @_Z22__spirv_GlobalOffset_xv() #5, !noalias !12\r\n %add4.i = add i64 %add.i, %call3.i\r\n %call.i1 = tail call i64 @_Z23__spirv_NumWorkgroups_xv() #5, !noalias !12\r\n %call1.i2 = tail call i64 @_Z23__spirv_WorkgroupSize_xv() #5, !noalias !12\r\n %mul.i3 = mul i64 %call1.i2, %call.i1\r\n %call.i.i.i9.i.i = tail call noundef i64 @_Z22__spirv_GlobalOffset_xv() #6, !noalias !12\r\n %cmp6.not.i.not.i = icmp ult i64 %add4.i, %0\r\n br i1 %cmp6.not.i.not.i, label %for.body.i.preheader, label %_ZNK4sycl3_V16detail18RoundedRangeKernelINS0_4itemILi1ELb1EEELi1EZZ12foo_parallelRNS0_5queueERNS0_6bufferIfLi1ENS1_17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESD_EUlNS0_2idILi1EEEE_EclES4_.exit\r\n\r\nfor.body.i.preheader: ; preds = %entry\r\n br label %for.body.i\r\n\r\nfor.body.i: ; preds = %for.body.i.preheader, %for.body.i\r\n %Gen.sroa.0.019.i = phi i64 [ %add.i.i, %for.body.i ], [ %add4.i, %for.body.i.preheader ]\r\n %conv.i.i = uitofp i64 %Gen.sroa.0.019.i to float\r\n %arrayidx.i.i = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %Gen.sroa.0.019.i\r\n %arrayidx.ascast.i.i = addrspacecast ptr addrspace(1) %arrayidx.i.i to ptr\r\n store float %conv.i.i, ptr %arrayidx.ascast.i.i, align 4, !tbaa !17\r\n %add.i.i = add i64 %Gen.sroa.0.019.i, %mul.i3\r\n %cmp6.i.not.i = icmp ult i64 %add.i.i, %0\r\n br i1 %cmp6.i.not.i, label %for.body.i, label %_ZNK4sycl3_V16detail18RoundedRangeKernelINS0_4itemILi1ELb1EEELi1EZZ12foo_parallelRNS0_5queueERNS0_6bufferIfLi1ENS1_17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESD_EUlNS0_2idILi1EEEE_EclES4_.exit.loopexit, !llvm.loop !21\r\n\r\n_ZNK4sycl3_V16detail18RoundedRangeKernelINS0_4itemILi1ELb1EEELi1EZZ12foo_parallelRNS0_5queueERNS0_6bufferIfLi1ENS1_17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESD_EUlNS0_2idILi1EEEE_EclES4_.exit.loopexit: ; preds = %for.body.i\r\n br label %_ZNK4sycl3_V16detail18RoundedRangeKernelINS0_4itemILi1ELb1EEELi1EZZ12foo_parallelRNS0_5queueERNS0_6bufferIfLi1ENS1_17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESD_EUlNS0_2idILi1EEEE_EclES4_.exit\r\n\r\n_ZNK4sycl3_V16detail18RoundedRangeKernelINS0_4itemILi1ELb1EEELi1EZZ12foo_parallelRNS0_5queueERNS0_6bufferIfLi1ENS1_17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESD_EUlNS0_2idILi1EEEE_EclES4_.exit: ; preds = %_ZNK4sycl3_V16detail18RoundedRangeKernelINS0_4itemILi1ELb1EEELi1EZZ12foo_parallelRNS0_5queueERNS0_6bufferIfLi1ENS1_17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESD_EUlNS0_2idILi1EEEE_EclES4_.exit.loopexit, %entry\r\n ret void\r\n}\r\n\r\n; Function Attrs: convergent nounwind\r\ndeclare dso_local noundef i64 @_Z22__spirv_GlobalOffset_xv() local_unnamed_addr #1\r\n\r\n; Function Attrs: convergent norecurse nounwind\r\ndefine weak_odr dso_local spir_kernel void @_ZTSZZ12foo_parallelRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_EUlNS0_2idILi1EEEE_(ptr addrspace(1) noundef align 4 %_arg_acc, ptr noundef byval(%\"class.sycl::_V1::range\") align 8 %_arg_acc3) local_unnamed_addr #0 comdat !srcloc !23 !kernel_arg_buffer_location !24 !kernel_arg_runtime_aligned !25 !kernel_arg_exclusive_ptr !25 !sycl_fixed_targets !10 !sycl_kernel_omit_args !26 {\r\nentry:\r\n %agg.tmp6.sroa.0.0.copyload = load i64, ptr %_arg_acc3, align 8\r\n %add.ptr.i = getelementptr inbounds float, ptr addrspace(1) %_arg_acc, i64 %agg.tmp6.sroa.0.0.copyload\r\n %call.i = tail call i64 @_Z21__spirv_WorkgroupId_xv() #5, !noalias !27\r\n %call1.i = tail call i64 @_Z23__spirv_WorkgroupSize_xv() #5, !noalias !27\r\n %mul.i = mul i64 %call1.i, %call.i\r\n %call2.i = tail call i64 @_Z27__spirv_LocalInvocationId_xv() #5, !noalias !27\r\n %add.i = add i64 %mul.i, %call2.i\r\n %call3.i = tail call i64 @_Z22__spirv_GlobalOffset_xv() #5, !noalias !27\r\n %add4.i = add i64 %add.i, %call3.i\r\n %call.i1 = tail call i64 @_Z23__spirv_NumWorkgroups_xv() #5, !noalias !27\r\n %call1.i2 = tail call i64 @_Z23__spirv_WorkgroupSize_xv() #5, !noalias !27\r\n %call.i.i.i9.i.i = tail call noundef i64 @_Z22__spirv_GlobalOffset_xv() #6, !noalias !27\r\n %cmp.i.i = icmp ult i64 %add4.i, 2147483648\r\n tail call void @llvm.assume(i1 %cmp.i.i)\r\n %conv.i = uitofp i64 %add4.i to float\r\n %arrayidx.i = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %add4.i\r\n %arrayidx.ascast.i = addrspacecast ptr addrspace(1) %arrayidx.i to ptr\r\n store float %conv.i, ptr %arrayidx.ascast.i, align 4, !tbaa !17\r\n ret void\r\n}\r\n\r\n; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\r\ndeclare void @llvm.assume(i1 noundef) #2\r\n\r\n; Function Attrs: mustprogress norecurse nounwind\r\ndefine weak_odr dso_local spir_kernel void @_ZTSZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_EUlvE_(i64 noundef %_arg_n, ptr addrspace(1) noundef align 4 %_arg_acc, ptr noundef byval(%\"class.sycl::_V1::range\") align 8 %_arg_acc3) local_unnamed_addr #3 comdat !srcloc !32 !kernel_arg_buffer_location !8 !kernel_arg_runtime_aligned !9 !kernel_arg_exclusive_ptr !9 !sycl_fixed_targets !10 !sycl_kernel_omit_args !11 {\r\nentry:\r\n %agg.tmp6.sroa.0.0.copyload = load i64, ptr %_arg_acc3, align 8\r\n %add.ptr.i = getelementptr inbounds float, ptr addrspace(1) %_arg_acc, i64 %agg.tmp6.sroa.0.0.copyload\r\n %cmp.i12.not = icmp eq i64 %_arg_n, 0\r\n br i1 %cmp.i12.not, label %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit, label %for.body.i.preheader\r\n\r\nfor.body.i.preheader: ; preds = %entry\r\n %xtraiter = and i64 %_arg_n, 7\r\n %0 = icmp ult i64 %_arg_n, 8\r\n br i1 %0, label %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit.loopexit.unr-lcssa, label %for.body.i.preheader.new\r\n\r\nfor.body.i.preheader.new: ; preds = %for.body.i.preheader\r\n %unroll_iter = and i64 %_arg_n, -8\r\n br label %for.body.i\r\n\r\nfor.body.i: ; preds = %for.body.i, %for.body.i.preheader.new\r\n %i.0.i13 = phi i64 [ 0, %for.body.i.preheader.new ], [ %inc.i.7, %for.body.i ]\r\n %conv.i = uitofp i64 %i.0.i13 to float\r\n %arrayidx.i = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %i.0.i13\r\n %arrayidx.ascast.i = addrspacecast ptr addrspace(1) %arrayidx.i to ptr\r\n store float %conv.i, ptr %arrayidx.ascast.i, align 4, !tbaa !17\r\n %inc.i = or disjoint i64 %i.0.i13, 1\r\n %conv.i.1 = uitofp i64 %inc.i to float\r\n %arrayidx.i.1 = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %inc.i\r\n %arrayidx.ascast.i.1 = addrspacecast ptr addrspace(1) %arrayidx.i.1 to ptr\r\n store float %conv.i.1, ptr %arrayidx.ascast.i.1, align 4, !tbaa !17\r\n %inc.i.1 = or disjoint i64 %i.0.i13, 2\r\n %conv.i.2 = uitofp i64 %inc.i.1 to float\r\n %arrayidx.i.2 = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %inc.i.1\r\n %arrayidx.ascast.i.2 = addrspacecast ptr addrspace(1) %arrayidx.i.2 to ptr\r\n store float %conv.i.2, ptr %arrayidx.ascast.i.2, align 4, !tbaa !17\r\n %inc.i.2 = or disjoint i64 %i.0.i13, 3\r\n %conv.i.3 = uitofp i64 %inc.i.2 to float\r\n %arrayidx.i.3 = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %inc.i.2\r\n %arrayidx.ascast.i.3 = addrspacecast ptr addrspace(1) %arrayidx.i.3 to ptr\r\n store float %conv.i.3, ptr %arrayidx.ascast.i.3, align 4, !tbaa !17\r\n %inc.i.3 = or disjoint i64 %i.0.i13, 4\r\n %conv.i.4 = uitofp i64 %inc.i.3 to float\r\n %arrayidx.i.4 = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %inc.i.3\r\n %arrayidx.ascast.i.4 = addrspacecast ptr addrspace(1) %arrayidx.i.4 to ptr\r\n store float %conv.i.4, ptr %arrayidx.ascast.i.4, align 4, !tbaa !17\r\n %inc.i.4 = or disjoint i64 %i.0.i13, 5\r\n %conv.i.5 = uitofp i64 %inc.i.4 to float\r\n %arrayidx.i.5 = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %inc.i.4\r\n %arrayidx.ascast.i.5 = addrspacecast ptr addrspace(1) %arrayidx.i.5 to ptr\r\n store float %conv.i.5, ptr %arrayidx.ascast.i.5, align 4, !tbaa !17\r\n %inc.i.5 = or disjoint i64 %i.0.i13, 6\r\n %conv.i.6 = uitofp i64 %inc.i.5 to float\r\n %arrayidx.i.6 = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %inc.i.5\r\n %arrayidx.ascast.i.6 = addrspacecast ptr addrspace(1) %arrayidx.i.6 to ptr\r\n store float %conv.i.6, ptr %arrayidx.ascast.i.6, align 4, !tbaa !17\r\n %inc.i.6 = or disjoint i64 %i.0.i13, 7\r\n %conv.i.7 = uitofp i64 %inc.i.6 to float\r\n %arrayidx.i.7 = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %inc.i.6\r\n %arrayidx.ascast.i.7 = addrspacecast ptr addrspace(1) %arrayidx.i.7 to ptr\r\n store float %conv.i.7, ptr %arrayidx.ascast.i.7, align 4, !tbaa !17\r\n %inc.i.7 = add nuw i64 %i.0.i13, 8\r\n %niter.ncmp.7 = icmp eq i64 %inc.i.7, %unroll_iter\r\n br i1 %niter.ncmp.7, label %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit.loopexit.unr-lcssa, label %for.body.i, !llvm.loop !33\r\n\r\n_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit.loopexit.unr-lcssa: ; preds = %for.body.i, %for.body.i.preheader\r\n %i.0.i13.unr = phi i64 [ 0, %for.body.i.preheader ], [ %unroll_iter, %for.body.i ]\r\n %lcmp.mod.not = icmp eq i64 %xtraiter, 0\r\n br i1 %lcmp.mod.not, label %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit, label %for.body.i.epil\r\n\r\nfor.body.i.epil: ; preds = %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit.loopexit.unr-lcssa, %for.body.i.epil\r\n %i.0.i13.epil = phi i64 [ %inc.i.epil, %for.body.i.epil ], [ %i.0.i13.unr, %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit.loopexit.unr-lcssa ]\r\n %epil.iter = phi i64 [ %epil.iter.next, %for.body.i.epil ], [ 0, %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit.loopexit.unr-lcssa ]\r\n %conv.i.epil = uitofp i64 %i.0.i13.epil to float\r\n %arrayidx.i.epil = getelementptr inbounds float, ptr addrspace(1) %add.ptr.i, i64 %i.0.i13.epil\r\n %arrayidx.ascast.i.epil = addrspacecast ptr addrspace(1) %arrayidx.i.epil to ptr\r\n store float %conv.i.epil, ptr %arrayidx.ascast.i.epil, align 4, !tbaa !17\r\n %inc.i.epil = add nuw nsw i64 %i.0.i13.epil, 1\r\n %epil.iter.next = add nuw nsw i64 %epil.iter, 1\r\n %epil.iter.cmp.not = icmp eq i64 %epil.iter.next, %xtraiter\r\n br i1 %epil.iter.cmp.not, label %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit, label %for.body.i.epil, !llvm.loop !34\r\n\r\n_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit: ; preds = %for.body.i.epil, %_ZZZ15foo_single_taskRN4sycl3_V15queueERNS0_6bufferIfLi1ENS0_6detail17aligned_allocatorIfEEvEEmENKUlRNS0_7handlerEE_clESA_ENKUlvE_clEv.exit.loopexit.unr-lcssa, %entry\r\n ret void\r\n}\r\n\r\n; Function Attrs: convergent nounwind\r\ndeclare dso_local i64 @_Z21__spirv_WorkgroupId_xv() local_unnamed_addr #4\r\n\r\n; Function Attrs: convergent nounwind\r\ndeclare dso_local i64 @_Z23__spirv_WorkgroupSize_xv() local_unnamed_addr #4\r\n\r\n; Function Attrs: convergent nounwind\r\ndeclare dso_local i64 @_Z27__spirv_LocalInvocationId_xv() local_unnamed_addr #4\r\n\r\n; Function Attrs: convergent nounwind\r\ndeclare dso_local i64 @_Z23__spirv_NumWorkgroups_xv() local_unnamed_addr #4\r\n\r\nattributes #0 = { convergent norecurse nounwind \"min-legal-vector-width\"=\"0\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"sycl-module-id\"=\"native_cpu.cpp\" \"sycl-optlevel\"=\"2\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" \"uniform-work-group-size\"=\"true\" }\r\nattributes #1 = { convergent nounwind \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\r\nattributes #2 = { mustprogress nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\r\nattributes #3 = { mustprogress norecurse nounwind \"min-legal-vector-width\"=\"0\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"sycl-module-id\"=\"native_cpu.cpp\" \"sycl-optlevel\"=\"2\" \"sycl-single-task\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" \"uniform-work-group-size\"=\"true\" }\r\nattributes #4 = { convergent nounwind \"frame-pointer\"=\"all\" \"no-builtins\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+avx,+avx2,+avx512f,+cmov,+crc32,+cx8,+evex512,+f16c,+fma,+fxsr,+mmx,+popcnt,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87,+xsave\" \"tune-cpu\"=\"generic\" }\r\nattributes #5 = { convergent nobuiltin nounwind \"no-builtins\" }\r\nattributes #6 = { convergent nounwind }\r\n\r\n!opencl.spir.version = !{!0}\r\n!spirv.Source = !{!1}\r\n!llvm.ident = !{!2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2, !2}\r\n!llvm.module.flags = !{!3, !4, !5, !6}\r\n\r\n!0 = !{i32 1, i32 2}\r\n!1 = !{i32 4, i32 100000}\r\n!2 = !{!\"clang version 19.0.0git (https://github.com/intel/llvm.git f56d7d7ffd798e78146510bef7ba02575d53bbb2)\"}\r\n!3 = !{i32 1, !\"wchar_size\", i32 4}\r\n!4 = !{i32 1, !\"is-native-cpu\", i32 1}\r\n!5 = !{i32 7, !\"uwtable\", i32 2}\r\n!6 = !{i32 7, !\"frame-pointer\", i32 2}\r\n!7 = !{i32 8377463}\r\n!8 = !{i32 -1, i32 -1, i32 -1}\r\n!9 = !{i1 false, i1 true, i1 false}\r\n!10 = !{}\r\n!11 = !{i1 false, i1 false, i1 true, i1 true, i1 false}\r\n!12 = !{!13, !15}\r\n!13 = distinct !{!13, !14, !\"_ZN4sycl3_V16detail7Builder7getItemILi1ELb1EEENSt9enable_ifIXT0_EKNS0_4itemIXT_EXT0_EEEE4typeEv: %agg.result\"}\r\n!14 = distinct !{!14, !\"_ZN4sycl3_V16detail7Builder7getItemILi1ELb1EEENSt9enable_ifIXT0_EKNS0_4itemIXT_EXT0_EEEE4typeEv\"}\r\n!15 = distinct !{!15, !16, !\"_ZN4sycl3_V16detail7Builder10getElementILi1ELb1EEEDTcl7getItemIXT_EXT0_EEEEPNS0_4itemIXT_EXT0_EEE: %agg.result\"}\r\n!16 = distinct !{!16, !\"_ZN4sycl3_V16detail7Builder10getElementILi1ELb1EEEDTcl7getItemIXT_EXT0_EEEEPNS0_4itemIXT_EXT0_EEE\"}\r\n!17 = !{!18, !18, i64 0}\r\n!18 = !{!\"float\", !19, i64 0}\r\n!19 = !{!\"omnipotent char\", !20, i64 0}\r\n!20 = !{!\"Simple C++ TBAA\"}\r\n!21 = distinct !{!21, !22}\r\n!22 = !{!\"llvm.loop.mustprogress\"}\r\n!23 = !{i32 738}\r\n!24 = !{i32 -1, i32 -1}\r\n!25 = !{i1 true, i1 false}\r\n!26 = !{i1 false, i1 true, i1 true, i1 false}\r\n!27 = !{!28, !30}\r\n!28 = distinct !{!28, !29, !\"_ZN4sycl3_V16detail7Builder7getItemILi1ELb1EEENSt9enable_ifIXT0_EKNS0_4itemIXT_EXT0_EEEE4typeEv: %agg.result\"}\r\n!29 = distinct !{!29, !\"_ZN4sycl3_V16detail7Builder7getItemILi1ELb1EEENSt9enable_ifIXT0_EKNS0_4itemIXT_EXT0_EEEE4typeEv\"}\r\n!30 = distinct !{!30, !31, !\"_ZN4sycl3_V16detail7Builder10getElementILi1ELb1EEEDTcl7getItemIXT_EXT0_EEEEPNS0_4itemIXT_EXT0_EEE: %agg.result\"}\r\n!31 = distinct !{!31, !\"_ZN4sycl3_V16detail7Builder10getElementILi1ELb1EEEDTcl7getItemIXT_EXT0_EEEEPNS0_4itemIXT_EXT0_EEE\"}\r\n!32 = !{i32 1035}\r\n!33 = distinct !{!33, !22}\r\n!34 = distinct !{!34, !35}\r\n!35 = !{!\"llvm.loop.unroll.disable\"}\r\n\r\n```", "author": "wenju-he", "labels": ["vectorizers", "crash"], "comments": [{"author": "PietroGhg", "body": "Hi, this should be addressed by https://github.com/llvm/llvm-project/pull/92107"}]}, "verified": true} {"bug_id": "106641", "issue_url": "https://github.com/llvm/llvm-project/issues/106641", "bug_type": "crash", "base_commit": "c8568f09577e9332d15edf98beb5376dc8d0672e", "knowledge_cutoff": "2024-08-29T23:08:58Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "f0e34f381866b82a26241f7e9aa5964f0dd11ebd", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[7147, 7153], [7250, 7259], [7284, 7299], [7364, 7371]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationPlanner::computeBestVF", "LoopVectorizationPlanner::precomputeCosts", "planContainsAdditionalSimplifications"]}}, "patch": "commit f0e34f381866b82a26241f7e9aa5964f0dd11ebd\nAuthor: Florian Hahn \nDate: Fri Aug 30 11:29:29 2024 +0100\n\n [VPlan] Don't skip optimizable truncs in planContainsAdditionalSimps.\n \n A optimizable cast can also be removed by VPlan simplifications. Remove\n the restriction from planContainsAdditionalSimplifications, as this\n causes it to miss relevant simplifications, triggering false positives\n for the cost decision verification.\n \n Also adds debug output for printing additional cost-precomputations.\n \n Fixes https://github.com/llvm/llvm-project/issues/106641.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 6babfd1eee91..fa05b8dd2242 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -7147,7 +7147,12 @@ LoopVectorizationPlanner::precomputeCosts(VPlan &Plan, ElementCount VF,\n if (!OrigLoop->contains(CondI) ||\n !CostCtx.SkipCostComputation.insert(CondI).second)\n continue;\n- Cost += CostCtx.getLegacyCost(CondI, VF);\n+ InstructionCost CondICost = CostCtx.getLegacyCost(CondI, VF);\n+ LLVM_DEBUG({\n+ dbgs() << \"Cost of \" << CondICost << \" for VF \" << VF\n+ << \": exit condition instruction \" << *CondI << \"\\n\";\n+ });\n+ Cost += CondICost;\n for (Value *Op : CondI->operands()) {\n auto *OpI = dyn_cast(Op);\n if (!OpI || any_of(OpI->users(), [&ExitInstrs, this](User *U) {\n@@ -7250,10 +7255,9 @@ InstructionCost LoopVectorizationPlanner::cost(VPlan &Plan,\n /// not have corresponding recipes in \\p Plan and are not marked to be ignored\n /// in \\p CostCtx. This means the VPlan contains simplification that the legacy\n /// cost-model did not account for.\n-static bool\n-planContainsAdditionalSimplifications(VPlan &Plan, ElementCount VF,\n- VPCostContext &CostCtx, Loop *TheLoop,\n- LoopVectorizationCostModel &CM) {\n+static bool planContainsAdditionalSimplifications(VPlan &Plan,\n+ VPCostContext &CostCtx,\n+ Loop *TheLoop) {\n // First collect all instructions for the recipes in Plan.\n auto GetInstructionForCost = [](const VPRecipeBase *R) -> Instruction * {\n if (auto *S = dyn_cast(R))\n@@ -7284,16 +7288,13 @@ planContainsAdditionalSimplifications(VPlan &Plan, ElementCount VF,\n // Return true if the loop contains any instructions that are not also part of\n // the VPlan or are skipped for VPlan-based cost computations. This indicates\n // that the VPlan contains extra simplifications.\n- return any_of(\n- TheLoop->blocks(), [&SeenInstrs, VF, &CostCtx, &CM](BasicBlock *BB) {\n- return any_of(*BB, [&SeenInstrs, VF, &CostCtx, &CM](Instruction &I) {\n- if (isa(&I))\n- return false;\n- return !SeenInstrs.contains(&I) &&\n- !CostCtx.skipCostComputation(&I, true) &&\n- !CM.canTruncateToMinimalBitwidth(&I, VF);\n- });\n- });\n+ return any_of(TheLoop->blocks(), [&SeenInstrs, &CostCtx](BasicBlock *BB) {\n+ return any_of(*BB, [&SeenInstrs, &CostCtx](Instruction &I) {\n+ if (isa(&I))\n+ return false;\n+ return !SeenInstrs.contains(&I) && !CostCtx.skipCostComputation(&I, true);\n+ });\n+ });\n }\n #endif\n \n@@ -7364,8 +7365,7 @@ VectorizationFactor LoopVectorizationPlanner::computeBestVF() {\n precomputeCosts(BestPlan, BestFactor.Width, CostCtx);\n assert((BestFactor.Width == LegacyVF.Width ||\n planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width),\n- BestFactor.Width, CostCtx,\n- OrigLoop, CM)) &&\n+ CostCtx, OrigLoop)) &&\n \" VPlan cost model and legacy cost model disagreed\");\n assert((BestFactor.Width.isScalar() || BestFactor.ScalarCost > 0) &&\n \"when vectorizing, the scalar cost must be computed.\");\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-cost.ll", "commands": ["opt -p loop-vectorize -mattr=+v -S %s"], "tests": [{"test_name": "truncate_to_i1_used_by_branch", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @truncate_to_i1_used_by_branch(i8 %x, ptr %dst) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %f.039 = phi i8 [ 0, %entry ], [ %add, %loop.latch ]\n %0 = or i8 23, %x\n %extract.t = trunc i8 %0 to i1\n br i1 %extract.t, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n store i8 0, ptr %dst, align 1\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %add = add i8 %f.039, 1\n %conv = sext i8 %f.039 to i32\n %cmp = icmp slt i32 %conv, 1\n br i1 %cmp, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v,+zvl256b\" }\n"}, {"test_name": "test_pr98413_sext_removed", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @test_pr98413_sext_removed(ptr %src, ptr noalias %dst, i64 %x) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src = getelementptr inbounds i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep.src, align 8\n %ext.l = sext i16 %l to i64\n %and = and i64 %x, %ext.l\n %trunc.and = trunc i64 %and to i8\n %gep.dst = getelementptr inbounds i8, ptr %dst, i64 %iv\n store i8 %trunc.and, ptr %gep.dst, align 1\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv, 96\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "test_pr98413_zext_removed", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @test_pr98413_zext_removed(ptr %src, ptr noalias %dst, i64 %x) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src = getelementptr inbounds i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep.src, align 8\n %ext.l = zext i16 %l to i64\n %and = and i64 %x, %ext.l\n %trunc.and = trunc i64 %and to i8\n %gep.dst = getelementptr inbounds i8, ptr %dst, i64 %iv\n store i8 %trunc.and, ptr %gep.dst, align 1\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv, 96\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "[VPlan] Assertion \" VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Tested using c4906588ce47de33d59bcd95f3e82ce2c3e61c23\r\n\r\nTestcase:\r\n```c\r\n_Bool a;\r\nchar b;\r\nunsigned char c;\r\nvoid d(_Bool e[][6]) {\r\n for (signed char f = 0; f < c; f += 3)\r\n for (_Bool g = ({\r\n __typeof__(0) h = e[1][2];\r\n e[f][f] > h ?: h;\r\n });\r\n g; g = b)\r\n for (_Bool i = 0; i < (0 >= 0); i = 1)\r\n a = 0;\r\n}\r\n```\r\nhttps://godbolt.org/z/Eab8s5e1K\r\n\r\nReduced LLVM IR Testcase:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @d() #0 {\r\nentry:\r\n br label %for.body\r\n\r\nfor.cond.cleanup.loopexit: ; preds = %for.cond.cleanup14\r\n ret void\r\n\r\nfor.body: ; preds = %for.cond.cleanup14, %entry\r\n %f.039 = phi i8 [ %add, %for.cond.cleanup14 ], [ 0, %entry ]\r\n %0 = or i8 0, 0\r\n %extract.t = trunc i8 %0 to i1\r\n br i1 %extract.t, label %for.body22.lr.ph, label %for.cond.cleanup14\r\n\r\nfor.body22.lr.ph: ; preds = %for.body\r\n store i8 0, ptr null, align 1\r\n br label %for.cond.cleanup14\r\n\r\nfor.cond.cleanup14: ; preds = %for.body22.lr.ph, %for.body\r\n %add = add i8 %f.039, 1\r\n %conv = sext i8 %f.039 to i32\r\n %cmp = icmp slt i32 %conv, 1\r\n br i1 %cmp, label %for.body, label %for.cond.cleanup.loopexit\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl256b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7365: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), BestFactor.Width, CostCtx, OrigLoop, CM)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"d\"\r\n #0 0x000064cd5b412560 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x301e560)\r\n #1 0x000064cd5b40f96f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x301b96f)\r\n #2 0x000064cd5b40fac5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x000072ec5a042520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x000072ec5a0969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x000072ec5a0969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x000072ec5a0969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x000072ec5a042476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x000072ec5a0287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x000072ec5a02871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x000072ec5a039e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x000064cd5a3dd109 llvm::LoopVectorizationPlanner::computeBestVF() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fe9109)\r\n#12 0x000064cd5a3f3ccb llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fffccb)\r\n#13 0x000064cd5a3f6ce1 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2002ce1)\r\n#14 0x000064cd5a3f7357 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2003357)\r\n#15 0x000064cd592541b6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe601b6)\r\n#16 0x000064cd5b21d41f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e2941f)\r\n#17 0x000064cd59257346 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe63346)\r\n#18 0x000064cd5b21c19b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e2819b)\r\n#19 0x000064cd592535e6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5f5e6)\r\n#20 0x000064cd5b21a1dd llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e261dd)\r\n#21 0x000064cd58a7c106 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x688106)\r\n#22 0x000064cd58a6de91 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x679e91)\r\n#23 0x000072ec5a029d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#24 0x000072ec5a029e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#25 0x000072ec5a029e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#26 0x000064cd58a64245 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x670245)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\nhttps://godbolt.org/z/EzdEYKW7c\r\n\r\nFound via fuzzer\r\n\r\nRelated/prior issue: #106417\r\n\r\ncc @fhahn", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "113520", "issue_url": "https://github.com/llvm/llvm-project/issues/113520", "bug_type": "miscompilation", "base_commit": "907c13638819b36ecc5154775217068b1a1f4cfc", "knowledge_cutoff": "2024-10-24T04:03:39Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "e914421d7fbf22059ea2180233add0c914ec80f0", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[12068, 12075], [15784, 15791]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::getTreeCost", "BoUpSLP::vectorizeTree"]}}, "patch": "commit e914421d7fbf22059ea2180233add0c914ec80f0\nAuthor: Alexey Bataev \nDate: Thu Oct 24 08:57:07 2024 -0700\n\n [SLP]Do correct signedness analysis for externally used scalars\n \n If the scalars is used externally is in the root node, it may have\n incorrect signedness info because of the conflict with the demanded bits\n analysis. Need to perform exact signedness analysis and compute it\n rather than rely on the precomputed value, which might be incorrect for\n alternate zext/sext nodes.\n \n Fixes #113520\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 012d85353f08..2afd02dae3a8 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -12068,8 +12068,9 @@ InstructionCost BoUpSLP::getTreeCost(ArrayRef VectorizedVals) {\n auto It = MinBWs.find(Entry);\n if (It != MinBWs.end()) {\n auto *MinTy = IntegerType::get(F->getContext(), It->second.first);\n- unsigned Extend =\n- It->second.second ? Instruction::SExt : Instruction::ZExt;\n+ unsigned Extend = isKnownNonNegative(EU.Scalar, SimplifyQuery(*DL))\n+ ? Instruction::ZExt\n+ : Instruction::SExt;\n VecTy = getWidenedType(MinTy, BundleWidth);\n ExtraCost = TTI->getExtractWithExtendCost(Extend, EU.Scalar->getType(),\n VecTy, EU.Lane);\n@@ -15784,8 +15785,9 @@ BoUpSLP::vectorizeTree(const ExtraValueToDebugLocsMap &ExternallyUsedValues,\n // to the larger type.\n ExV = Ex;\n if (Scalar->getType() != Ex->getType())\n- ExV = Builder.CreateIntCast(Ex, Scalar->getType(),\n- MinBWs.find(E)->second.second);\n+ ExV = Builder.CreateIntCast(\n+ Ex, Scalar->getType(),\n+ !isKnownNonNegative(Scalar, SimplifyQuery(*DL)));\n auto *I = dyn_cast(Ex);\n ScalarToEEs[Scalar].try_emplace(I ? I->getParent()\n : &F->getEntryBlock(),\n", "tests": [{"file": "", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "add3_i96", "test_body": "@0 = external global i1\n\ndefine <3 x i32> @add3_i96.2(<3 x i32> %0, <3 x i32> %1) {\n %4 = extractelement <3 x i32> %0, i64 0\n %5 = sext i32 %4 to i64\n %6 = extractelement <3 x i32> %1, i64 0\n %7 = zext i32 %6 to i64\n %8 = add nuw nsw i64 %7, %5\n %9 = extractelement <3 x i32> %0, i64 1\n %10 = zext i32 %9 to i64\n %11 = extractelement <3 x i32> %1, i64 1\n %12 = zext i32 %11 to i64\n %13 = add nuw nsw i64 %12, %10\n %14 = lshr i64 %8, 32\n %15 = add nuw nsw i64 %13, %14\n %16 = extractelement <3 x i32> %0, i64 2\n %17 = extractelement <3 x i32> %1, i64 2\n %18 = add i32 %17, %16\n %19 = or i64 %15, 32\n %20 = trunc i64 %19 to i32\n %21 = add i32 %18, %17\n %22 = trunc i64 %8 to i32\n %23 = insertelement <3 x i32> %1, i32 %22, i32 0\n ret <3 x i32> %23\n}\n"}]}], "issue": {"title": "`slp-vectorizer` miscompilation", "body": "Alive2 report: https://alive2.llvm.org/ce/z/FmxNvP\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine <3 x i32> @add3_i96.2(<3 x i32> %#0, <3 x i32> %#1) {\r\n#2:\r\n %#3 = extractelement <3 x i32> %#0, i64 0\r\n %#4 = sext i32 %#3 to i64\r\n %#5 = extractelement <3 x i32> %#1, i64 0\r\n %#6 = zext i32 %#5 to i64\r\n %#7 = add nsw nuw i64 %#6, %#4\r\n %#21 = trunc i64 %#7 to i32\r\n %#22 = insertelement <3 x i32> %#1, i32 %#21, i32 0\r\n ret <3 x i32> %#22\r\n}\r\n=>\r\ndefine <3 x i32> @add3_i96.2(<3 x i32> %#0, <3 x i32> %#1) {\r\n#2:\r\n %#3 = shufflevector <3 x i32> %#1, <3 x i32> %#0, 0, 3\r\n %#4 = extractelement <2 x i32> %#3, i32 0\r\n %#5 = sext i32 %#4 to i64\r\n %#6 = extractelement <2 x i32> %#3, i32 1\r\n %#7 = sext i32 %#6 to i64\r\n %#8 = add nsw nuw i64 %#5, %#7\r\n %#22 = trunc i64 %#8 to i32\r\n %#23 = insertelement <3 x i32> %#1, i32 %#22, i32 0\r\n ret <3 x i32> %#23\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\n<3 x i32> %#0 = < #x00000008 (8), #x00000000 (0), #x00000000 (0) >\r\n<3 x i32> %#1 = < #xfffffff9 (4294967289, -7), #x00000000 (0), #x00000000 (0) >\r\n\r\nSource:\r\ni32 %#3 = #x00000008 (8)\r\ni64 %#4 = #x0000000000000008 (8)\r\ni32 %#5 = #xfffffff9 (4294967289, -7)\r\ni64 %#6 = #x00000000fffffff9 (4294967289)\r\ni64 %#7 = #x0000000100000001 (4294967297)\r\ni32 %#21 = #x00000001 (1)\r\n<3 x i32> %#22 = < #x00000001 (1), #x00000000 (0), #x00000000 (0) >\r\n\r\nTarget:\r\n<2 x i32> %#3 = < #xfffffff9 (4294967289, -7), #x00000008 (8) >\r\ni32 %#4 = #xfffffff9 (4294967289, -7)\r\ni64 %#5 = #xfffffffffffffff9 (18446744073709551609, -7)\r\ni32 %#6 = #x00000008 (8)\r\ni64 %#7 = #x0000000000000008 (8)\r\ni64 %#8 = poison\r\ni32 %#22 = poison\r\n<3 x i32> %#23 = < poison, #x00000000 (0), #x00000000 (0) >\r\nSource value: < #x00000001 (1), #x00000000 (0), #x00000000 (0) >\r\nTarget value: < poison, #x00000000 (0), #x00000000 (0) >\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": []}, "verified": true} {"bug_id": "125357", "issue_url": "https://github.com/llvm/llvm-project/issues/125357", "bug_type": "crash", "base_commit": "cdeeb390a9ea21540fc44ba10dede66fbc0b2fc8", "knowledge_cutoff": "2025-02-01T20:11:30Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "0c70a26f46e4efd5a29eb281ff99d2cf7f04c6f6", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[20147, 20153]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["tryToReduce"]}}, "patch": "commit 0c70a26f46e4efd5a29eb281ff99d2cf7f04c6f6\nAuthor: Alexey Bataev \nDate: Mon Feb 3 04:37:49 2025 -0800\n\n [SLP]Clear root node reordering only if the root node is not re-used in graph\n \n The reordering of the root node can be safely cleared only if the root\n node is not reused, otherwise the graph might be broken\n \n Fixes #125357\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 4f83b9cfa18d..539c9227af7e 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -20147,7 +20147,7 @@ public:\n }\n V.reorderTopToBottom();\n // No need to reorder the root node at all.\n- V.reorderBottomToTop(/*IgnoreReorder=*/true);\n+ V.reorderBottomToTop(!V.doesRootHaveInTreeUses());\n // Keep extracted other reduction values, if they are used in the\n // vectorization trees.\n BoUpSLP::ExtraValueToDebugLocsMap LocalExternallyUsedValues(\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/root-node-reordered-reused.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "", "test_body": "; ModuleID = 'test.ll'\nsource_filename = \"test.c\"\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i32 @main(i32 %0) {\nentry:\n %tobool.not = icmp eq i32 %0, 0\n br i1 %tobool.not, label %if.end6, label %if.then\n\nif.then: ; preds = %entry\n br i1 %tobool.not, label %if.end, label %i\n\nif.end: ; preds = %if.then\n br i1 %tobool.not, label %if.end6, label %i\n\nif.end6: ; preds = %if.end, %entry\n %h.0 = phi i32 [ %0, %if.end ], [ 0, %entry ]\n br i1 %tobool.not, label %i, label %if.end13\n\ni: ; preds = %if.end13, %if.end6, %if.end, %if.then\n %h.1 = phi i32 [ %0, %if.then ], [ %h.2, %if.end13 ], [ %0, %if.end ], [ %h.0, %if.end6 ]\n %j.0 = phi i32 [ undef, %if.then ], [ %j.1, %if.end13 ], [ undef, %if.end ], [ %0, %if.end6 ]\n %k.0 = phi i32 [ undef, %if.then ], [ %k.1, %if.end13 ], [ undef, %if.end ], [ %0, %if.end6 ]\n %l.0 = phi i32 [ undef, %if.then ], [ %l.1, %if.end13 ], [ undef, %if.end ], [ %0, %if.end6 ]\n %or = or i32 %j.0, %h.1\n %or9 = or i32 %or, %k.0\n %or10 = or i32 %or9, %l.0\n %tobool11.not = icmp eq i32 %or10, 0\n br i1 %tobool11.not, label %if.end13, label %land.rhs\n\nland.rhs: ; preds = %i\n br label %if.end13\n\nif.end13: ; preds = %i, %land.rhs, %if.end6\n %h.2 = phi i32 [ %h.0, %if.end6 ], [ %h.1, %land.rhs ], [ %h.1, %i ]\n %j.1 = phi i32 [ %0, %if.end6 ], [ %j.0, %land.rhs ], [ %j.0, %i ]\n %k.1 = phi i32 [ %0, %if.end6 ], [ %k.0, %land.rhs ], [ %k.0, %i ]\n %l.1 = phi i32 [ %0, %if.end6 ], [ %l.0, %land.rhs ], [ %l.0, %i ]\n br i1 true, label %if.end15, label %i\n\nif.end15: ; preds = %if.end13\n ret i32 0\n\n; uselistorder directives\n uselistorder i32 %k.0, { 1, 2, 0 }\n uselistorder i32 %l.0, { 1, 2, 0 }\n}\n"}]}], "issue": {"title": "clang crashes on valid code at -O{s,2,3} on x86_64-linux-gnu: Assertion `TE->isSame(VL) && \"Expected same scalars.\"' failed", "body": "It appears to be a recent regression as it doesn't reproduce with 19.1.0 and earlier.\n\nCompiler Explorer: https://godbolt.org/z/6ca9841xE\n\n```\n[512] % clangtk -v\nclang version 21.0.0git (https://github.com/llvm/llvm-project.git 7612dcc6e8d8e7f19b364084effbb01946294720)\nTarget: x86_64-unknown-linux-gnu\nThread model: posix\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\nBuild config: +assertions\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\nCandidate multilib: .;@m64\nSelected multilib: .;@m64\n[513] % \n[513] % clangtk -O3 small.c\nclang-21: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:3011: llvm::slpvectorizer::BoUpSLP::TreeEntry* llvm::slpvectorizer::BoUpSLP::getVectorizedOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int): Assertion `TE->isSame(VL) && \"Expected same scalars.\"' failed.\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\nStack dump:\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-21 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20250201-clangtk-m64-O3-build-174536/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20250201-clangtk-m64-O3-build-174536/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/21 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/21/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O3 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-b035e7.o -x c small.c\n1.\t parser at end of file\n2.\tOptimizer\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\n4.\tRunning pass \"slp-vectorizer\" on function \"main\"\n #0 0x00005596296c84af llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x479d4af)\n #1 0x00005596296c5c84 SignalHandler(int) Signals.cpp:0:0\n #2 0x00007fe817057420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\n #3 0x00007fe816a8e00b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\n #4 0x00007fe816a6d859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\n #5 0x00007fe816a6d729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\n #6 0x00007fe816a6d729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\n #7 0x00007fe816a7efd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\n #8 0x000055962b1df5c7 llvm::slpvectorizer::BoUpSLP::getSpillCost() (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x62b45c7)\n #9 0x000055962b243030 llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x6318030)\n#10 0x000055962b247f34 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&, llvm::AssumptionCache*) SLPVectorizer.cpp:0:0\n#11 0x000055962b24a205 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::SmallVectorImpl&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x631f205)\n#12 0x000055962b24ee25 bool llvm::SLPVectorizerPass::vectorizeCmpInsts>(llvm::iterator_range>, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x6323e25)\n#13 0x000055962b24fedb llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&)::'lambda3'(bool)::operator()(bool) const SLPVectorizer.cpp:0:0\n#14 0x000055962b253720 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x6328720)\n#15 0x000055962b259f46 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\n#16 0x000055962b25aaa3 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x632faa3)\n#17 0x000055962ac7c7e6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x5d517e6)\n#18 0x000055962903ed49 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x4113d49)\n#19 0x000055962693a376 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x1a0f376)\n#20 0x000055962903d5b2 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x41125b2)\n#21 0x0000559626938aa6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x1a0daa6)\n#22 0x000055962903cfd1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x4111fd1)\n#23 0x0000559629991596 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\n#24 0x0000559629995253 clang::emitBackendOutput(clang::CompilerInstance&, clang::CodeGenOptions&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x4a6a253)\n#25 0x000055962a09f2ce clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x51742ce)\n#26 0x000055962ba0f29c clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x6ae429c)\n#27 0x000055962a09f708 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x5174708)\n#28 0x000055962a3719d9 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x54469d9)\n#29 0x000055962a2f25be clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x53c75be)\n#30 0x000055962a466746 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x553b746)\n#31 0x00005596264d4a87 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x15a9a87)\n#32 0x00005596264cc33a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\n#33 0x00005596264d0471 clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x15a5471)\n#34 0x00005596263bb29b main (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x149029b)\n#35 0x00007fe816a6f083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\n#36 0x00005596264cbdce _start (/local/suz-local/software/local/clang-trunk/bin/clang-21+0x15a0dce)\nclangtk: error: unable to execute command: Aborted\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\nclang version 21.0.0git (https://github.com/llvm/llvm-project.git 7612dcc6e8d8e7f19b364084effbb01946294720)\nTarget: x86_64-unknown-linux-gnu\nThread model: posix\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\nBuild config: +assertions\nclangtk: note: diagnostic msg: \n********************\n\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\nPreprocessed source(s) and associated run script(s) are located at:\nclangtk: note: diagnostic msg: /tmp/small-401f68.c\nclangtk: note: diagnostic msg: /tmp/small-401f68.sh\nclangtk: note: diagnostic msg: \n\n********************\n[514] % \n[514] % cat small.c\nint a, b, c, d, e;\nvolatile int f;\nint main() {\n int h = 0, j, k, l;\n if (d) {\n h = c;\n if (f)\n goto i;\n if (f)\n goto i;\n }\n l = j = k = d;\n if (!b)\n i:\n (l | (h | j) | k) && f;\n if (a != e)\n goto i;\n return 0;\n}\n```\n\n\n", "author": "zhendongsu", "labels": ["llvm:SLPVectorizer", "crash", "generated by fuzzer"], "comments": [{"author": "dtcxzyw", "body": "Reduced reproducer: https://godbolt.org/z/TWadzP8Gc\n"}]}, "verified": true} {"bug_id": "112467", "issue_url": "https://github.com/llvm/llvm-project/issues/112467", "bug_type": "miscompilation", "base_commit": "9df8d8d05c2650b51bd4233e1759206d163f3133", "knowledge_cutoff": "2024-10-16T03:31:20Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "3bf2295ee0ebd1eafe66ca15dff44bdb31e6198a", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": [[3369, 3376]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": ["InstCombinerImpl::foldAndOrOfICmps"]}}, "patch": "commit 3bf2295ee0ebd1eafe66ca15dff44bdb31e6198a\nAuthor: Yingwei Zheng \nDate: Wed Oct 16 16:24:44 2024 +0800\n\n [InstCombine] Drop `samesign` flag in `foldAndOrOfICmpsWithConstEq` (#112489)\n \n In\n https://github.com/llvm/llvm-project/commit/5dbfca30c1a672cd0c5089df2b4fdd171436643a\n we assume that RHS is poison implies LHS is also poison. It doesn't hold\n after introducing samesign flag.\n \n This patch drops the `samesign` flag on RHS if the original expression\n is a logical and/or.\n \n Closes #112467.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\nindex 64bee4ab974e..c8407e8ba5ab 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n@@ -3369,8 +3369,14 @@ Value *InstCombinerImpl::foldAndOrOfICmps(ICmpInst *LHS, ICmpInst *RHS,\n // We can convert this case to bitwise and, because both operands are used\n // on the LHS, and as such poison from both will propagate.\n if (Value *V = foldAndOrOfICmpsWithConstEq(RHS, LHS, IsAnd,\n- /*IsLogical*/ false, Builder, Q))\n+ /*IsLogical=*/false, Builder, Q)) {\n+ // If RHS is still used, we should drop samesign flag.\n+ if (IsLogical && RHS->hasSameSign() && !RHS->use_empty()) {\n+ RHS->setSameSign(false);\n+ addToWorklist(RHS);\n+ }\n return V;\n+ }\n \n if (Value *V = foldIsPowerOf2OrZero(LHS, RHS, IsAnd, Builder, *this))\n return V;\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/and-or-icmp-min-max.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "sge_and_max_commute", "test_body": "define i1 @sge_and_max_commute(i8 %x, i8 %y) {\n %cmp = icmp sge i8 %x, %y\n %cmpeq = icmp eq i8 %x, 127\n %r = and i1 %cmpeq, %cmp\n ret i1 %r\n}\n"}, {"test_name": "sge_and_max_logical", "test_body": "define i1 @sge_and_max_logical(i8 %x, i8 %y) {\n %cmp = icmp sge i8 %x, %y\n %cmpeq = icmp eq i8 %x, 127\n %r = select i1 %cmp, i1 %cmpeq, i1 false\n ret i1 %r\n}\n"}, {"test_name": "sge_and_max_logical_samesign", "test_body": "define i1 @sge_and_max_logical_samesign(i8 %x, i8 %y) {\n %cmp = icmp sge i8 %x, %y\n %cmpeq = icmp samesign eq i8 %x, 127\n %r = select i1 %cmp, i1 %cmpeq, i1 false\n ret i1 %r\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/and-or-icmp-nullptr.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "sgt_and_min_logical_samesign", "test_body": "define i1 @sgt_and_min_logical_samesign(ptr %x, ptr %y) {\n %cmp = icmp sgt ptr %x, %y\n %cmpeq = icmp samesign eq ptr %x, null\n %r = select i1 %cmp, i1 %cmpeq, i1 false\n ret i1 %r\n}\n"}, {"test_name": "sgt_and_min_logical", "test_body": "define i1 @sgt_and_min_logical(ptr %x, ptr %y) {\n %cmp = icmp sgt ptr %x, %y\n %cmpeq = icmp eq ptr %x, null\n %r = select i1 %cmp, i1 %cmpeq, i1 false\n ret i1 %r\n}\n"}, {"test_name": "sle_or_not_min", "test_body": "define i1 @sle_or_not_min(ptr %x, ptr %y) {\n %cmp = icmp sle ptr %x, %y\n %cmpeq = icmp ne ptr %x, null\n %r = or i1 %cmp, %cmpeq\n ret i1 %r\n}\n"}]}], "issue": {"title": "[InstCombine] `samesign` flag should be dropped in `foldAndOrOfICmpsWithConstEq`", "body": "Reproducer:\r\n```\r\n; bin/opt -passes=instcombine test.ll -S\r\ndefine i1 @src(i8 %x, i8 %y) {\r\n %cmp = icmp sgt i8 %x, %y\r\n %cmpeq = icmp samesign ne i8 %x, 0\r\n %r = select i1 %cmp, i1 true, i1 %cmpeq\r\n ret i1 %r\r\n}\r\n```\r\n```\r\ndefine i1 @tgt(i8 %x, i8 %y) {\r\n %cmpeq = icmp samesign ne i8 %x, 0\r\n %1 = icmp slt i8 %y, 0\r\n %r = or i1 %cmpeq, %1\r\n ret i1 %r\r\n}\r\n```\r\n```\r\n----------------------------------------\r\ndefine i1 @src(i8 %x, i8 %y) {\r\n#0:\r\n %cmp = icmp sgt i8 %x, %y\r\n %cmpeq = icmp samesign ne i8 %x, 0\r\n %r = select i1 %cmp, i1 1, i1 %cmpeq\r\n ret i1 %r\r\n}\r\n=>\r\ndefine i1 @tgt(i8 %x, i8 %y) {\r\n#0:\r\n %cmpeq = icmp samesign ne i8 %x, 0\r\n %#1 = icmp slt i8 %y, 0\r\n %r = or i1 %cmpeq, %#1\r\n ret i1 %r\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\ni8 %x = #x81 (129, -127)\r\ni8 %y = #x80 (128, -128)\r\n\r\nSource:\r\ni1 %cmp = #x1 (1)\r\ni1 %cmpeq = poison\r\ni1 %r = #x1 (1)\r\n\r\nTarget:\r\ni1 %cmpeq = poison\r\ni1 %#1 = #x1 (1)\r\ni1 %r = poison\r\nSource value: #x1 (1)\r\nTarget value: poison\r\n```\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "106523", "issue_url": "https://github.com/llvm/llvm-project/issues/106523", "bug_type": "crash", "base_commit": "2002533802dbe74c82476e30d093baf6d4cdee50", "knowledge_cutoff": "2024-08-29T10:12:29Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "2dfb1c664c0a0afc512b900c45989e247406e523", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": [[772, 777], [795, 801]], "llvm/lib/Transforms/Vectorize/VPlanTransforms.h": [[36, 46]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": ["VPlanTransforms::adjustFixedOrderRecurrences", "sinkRecurrenceUsersAfterPrevious"]}}, "patch": "commit 2dfb1c664c0a0afc512b900c45989e247406e523\nAuthor: Florian Hahn \nDate: Wed Oct 23 13:12:03 2024 -0700\n\n [VPlan] Try to hoist Previous (and operands), if sinking fails for FORs. (#108945)\n \n In some cases, Previous (and its operands) can be hoisted. This allows\n supporting additional cases where sinking of all users of to FOR fails,\n e.g. due having to sink recipes with side-effects.\n \n This fixes a crash where we fail to create a scalar VPlan for a\n first-order recurrence, but can create a vector VPlan, because the trunc\n instruction of an IV which generates the previous value of the\n recurrence has been optimized to a truncated induction recipe, thus\n hoisting it to the beginning.\n \n Fixes https://github.com/llvm/llvm-project/issues/106523.\n \n PR: https://github.com/llvm/llvm-project/pull/108945\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\nindex d50f3c0c3f3e..c6e09c4f2e6e 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n@@ -772,6 +772,105 @@ sinkRecurrenceUsersAfterPrevious(VPFirstOrderRecurrencePHIRecipe *FOR,\n return true;\n }\n \n+/// Try to hoist \\p Previous and its operands before all users of \\p FOR.\n+static bool hoistPreviousBeforeFORUsers(VPFirstOrderRecurrencePHIRecipe *FOR,\n+ VPRecipeBase *Previous,\n+ VPDominatorTree &VPDT) {\n+ if (Previous->mayHaveSideEffects() || Previous->mayReadFromMemory())\n+ return false;\n+\n+ // Collect recipes that need hoisting.\n+ SmallVector HoistCandidates;\n+ SmallPtrSet Visited;\n+ VPRecipeBase *HoistPoint = nullptr;\n+ // Find the closest hoist point by looking at all users of FOR and selecting\n+ // the recipe dominating all other users.\n+ for (VPUser *U : FOR->users()) {\n+ auto *R = dyn_cast(U);\n+ if (!R)\n+ continue;\n+ if (!HoistPoint || VPDT.properlyDominates(R, HoistPoint))\n+ HoistPoint = R;\n+ }\n+ assert(all_of(FOR->users(),\n+ [&VPDT, HoistPoint](VPUser *U) {\n+ auto *R = dyn_cast(U);\n+ return !R || HoistPoint == R ||\n+ VPDT.properlyDominates(HoistPoint, R);\n+ }) &&\n+ \"HoistPoint must dominate all users of FOR\");\n+\n+ auto NeedsHoisting = [HoistPoint, &VPDT,\n+ &Visited](VPValue *HoistCandidateV) -> VPRecipeBase * {\n+ VPRecipeBase *HoistCandidate = HoistCandidateV->getDefiningRecipe();\n+ if (!HoistCandidate)\n+ return nullptr;\n+ VPRegionBlock *EnclosingLoopRegion =\n+ HoistCandidate->getParent()->getEnclosingLoopRegion();\n+ assert((!HoistCandidate->getParent()->getParent() ||\n+ HoistCandidate->getParent()->getParent() == EnclosingLoopRegion) &&\n+ \"CFG in VPlan should still be flat, without replicate regions\");\n+ // Hoist candidate was already visited, no need to hoist.\n+ if (!Visited.insert(HoistCandidate).second)\n+ return nullptr;\n+\n+ // Candidate is outside loop region or a header phi, dominates FOR users w/o\n+ // hoisting.\n+ if (!EnclosingLoopRegion || isa(HoistCandidate))\n+ return nullptr;\n+\n+ // If we reached a recipe that dominates HoistPoint, we don't need to\n+ // hoist the recipe.\n+ if (VPDT.properlyDominates(HoistCandidate, HoistPoint))\n+ return nullptr;\n+ return HoistCandidate;\n+ };\n+ auto CanHoist = [&](VPRecipeBase *HoistCandidate) {\n+ // Avoid hoisting candidates with side-effects, as we do not yet analyze\n+ // associated dependencies.\n+ return !HoistCandidate->mayHaveSideEffects();\n+ };\n+\n+ if (!NeedsHoisting(Previous->getVPSingleValue()))\n+ return true;\n+\n+ // Recursively try to hoist Previous and its operands before all users of FOR.\n+ HoistCandidates.push_back(Previous);\n+\n+ for (unsigned I = 0; I != HoistCandidates.size(); ++I) {\n+ VPRecipeBase *Current = HoistCandidates[I];\n+ assert(Current->getNumDefinedValues() == 1 &&\n+ \"only recipes with a single defined value expected\");\n+ if (!CanHoist(Current))\n+ return false;\n+\n+ for (VPValue *Op : Current->operands()) {\n+ // If we reach FOR, it means the original Previous depends on some other\n+ // recurrence that in turn depends on FOR. If that is the case, we would\n+ // also need to hoist recipes involving the other FOR, which may break\n+ // dependencies.\n+ if (Op == FOR)\n+ return false;\n+\n+ if (auto *R = NeedsHoisting(Op))\n+ HoistCandidates.push_back(R);\n+ }\n+ }\n+\n+ // Order recipes to hoist by dominance so earlier instructions are processed\n+ // first.\n+ sort(HoistCandidates, [&VPDT](const VPRecipeBase *A, const VPRecipeBase *B) {\n+ return VPDT.properlyDominates(A, B);\n+ });\n+\n+ for (VPRecipeBase *HoistCandidate : HoistCandidates) {\n+ HoistCandidate->moveBefore(*HoistPoint->getParent(),\n+ HoistPoint->getIterator());\n+ }\n+\n+ return true;\n+}\n+\n bool VPlanTransforms::adjustFixedOrderRecurrences(VPlan &Plan,\n VPBuilder &LoopBuilder) {\n VPDominatorTree VPDT;\n@@ -795,7 +894,8 @@ bool VPlanTransforms::adjustFixedOrderRecurrences(VPlan &Plan,\n Previous = PrevPhi->getBackedgeValue()->getDefiningRecipe();\n }\n \n- if (!sinkRecurrenceUsersAfterPrevious(FOR, Previous, VPDT))\n+ if (!sinkRecurrenceUsersAfterPrevious(FOR, Previous, VPDT) &&\n+ !hoistPreviousBeforeFORUsers(FOR, Previous, VPDT))\n return false;\n \n // Introduce a recipe to combine the incoming and previous values of a\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.h b/llvm/lib/Transforms/Vectorize/VPlanTransforms.h\nindex 60a44bfb0dca..11e094db6294 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.h\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.h\n@@ -36,11 +36,11 @@ struct VPlanTransforms {\n GetIntOrFpInductionDescriptor,\n ScalarEvolution &SE, const TargetLibraryInfo &TLI);\n \n- /// Sink users of fixed-order recurrences after the recipe defining their\n- /// previous value. Then introduce FirstOrderRecurrenceSplice VPInstructions\n- /// to combine the value from the recurrence phis and previous values. The\n- /// current implementation assumes all users can be sunk after the previous\n- /// value, which is enforced by earlier legality checks.\n+ /// Try to have all users of fixed-order recurrences appear after the recipe\n+ /// defining their previous value, by either sinking users or hoisting recipes\n+ /// defining their previous value (and its operands). Then introduce\n+ /// FirstOrderRecurrenceSplice VPInstructions to combine the value from the\n+ /// recurrence phis and previous values.\n /// \\returns true if all users of fixed-order recurrences could be re-arranged\n /// as needed or false if it is not possible. In the latter case, \\p Plan is\n /// not valid.\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/fixed-order-recurrence.ll", "commands": ["opt -passes=loop-vectorize -S -o - %s"], "tests": [{"test_name": "test_pr62954_scalar_epilogue_required", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc_linux\"\n\ndefine i64 @test_pr62954_scalar_epilogue_required(ptr %A, ptr noalias %B, ptr %C) {\nentry:\n %gep = getelementptr i8, ptr %A, i64 872\n %rec.start = load i64, ptr %gep, align 8\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 1, %entry ], [ %iv.next, %loop ]\n %for = phi i64 [ %rec.start, %entry ], [ %neg.iv, %loop ]\n %gep.B = getelementptr double, ptr %B, i64 %iv\n %l.B = load double, ptr %gep.B, align 8\n %neg.iv = sub nsw i64 0, %iv\n store i64 %neg.iv, ptr %gep, align 8\n %iv.next = add nuw nsw i64 %iv, 2\n %ec = icmp ugt i64 %iv, 74\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n %.in.lcssa = phi i64 [ %for, %loop ]\n %.lcssa = phi double [ %l.B, %loop ]\n store double %.lcssa, ptr %C, align 8\n ret i64 %.in.lcssa\n}\n"}, {"test_name": "thirdorderrec", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc_linux\"\n\ndefine void @thirdorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapture noundef writeonly %y, i32 noundef %n) {\nentry:\n %cmp38 = icmp sgt i32 %n, 3\n br i1 %cmp38, label %for.body.preheader, label %for.cond.cleanup\n\nfor.body.preheader: ; preds = %entry\n %wide.trip.count = zext i32 %n to i64\n %.pre = load i8, ptr %x, align 1\n %arrayidx5.phi.trans.insert = getelementptr inbounds i8, ptr %x, i64 1\n %.pre44 = load i8, ptr %arrayidx5.phi.trans.insert, align 1\n %arrayidx12.phi.trans.insert = getelementptr inbounds i8, ptr %x, i64 2\n %.pre45 = load i8, ptr %arrayidx12.phi.trans.insert, align 1\n br label %for.body\n\nfor.cond.cleanup: ; preds = %for.body, %entry\n ret void\n\nfor.body: ; preds = %for.body, %for.body.preheader\n %0 = phi i8 [ %.pre45, %for.body.preheader ], [ %3, %for.body ]\n %1 = phi i8 [ %.pre44, %for.body.preheader ], [ %0, %for.body ]\n %2 = phi i8 [ %.pre, %for.body.preheader ], [ %1, %for.body ]\n %indvars.iv = phi i64 [ 3, %for.body.preheader ], [ %indvars.iv.next, %for.body ]\n %add8 = add i8 %1, %2\n %add15 = add i8 %add8, %0\n %arrayidx18 = getelementptr inbounds i8, ptr %x, i64 %indvars.iv\n %3 = load i8, ptr %arrayidx18, align 1\n %add21 = add i8 %add15, %3\n %arrayidx24 = getelementptr inbounds i8, ptr %y, i64 %indvars.iv\n store i8 %add21, ptr %arrayidx24, align 1\n %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\n %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count\n br i1 %exitcond.not, label %for.cond.cleanup, label %for.body\n}\n"}, {"test_name": "for_iv_trunc_optimized", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc_linux\"\n\ndefine void @for_iv_trunc_optimized(ptr %dst) {\nbb:\n br label %loop\n\nloop: ; preds = %loop, %bb\n %iv = phi i64 [ 1, %bb ], [ %add, %loop ]\n %for.1 = phi i32 [ 1, %bb ], [ %trunc, %loop ]\n %for.2 = phi i32 [ 0, %bb ], [ %or, %loop ]\n %or = or i32 %for.1, 3\n %add = add i64 %iv, 1\n store i32 %for.2, ptr %dst, align 4\n %icmp = icmp ult i64 %iv, 337\n %trunc = trunc i64 %iv to i32\n br i1 %icmp, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "firstorderrec", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc_linux\"\n\ndefine void @firstorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapture noundef writeonly %y, i32 noundef %n) {\nentry:\n %cmp18 = icmp sgt i32 %n, 1\n br i1 %cmp18, label %for.body.preheader, label %for.cond.cleanup\n\nfor.body.preheader: ; preds = %entry\n %wide.trip.count = zext i32 %n to i64\n %.pre = load i8, ptr %x, align 1\n br label %for.body\n\nfor.cond.cleanup: ; preds = %for.body, %entry\n ret void\n\nfor.body: ; preds = %for.body, %for.body.preheader\n %0 = phi i8 [ %.pre, %for.body.preheader ], [ %1, %for.body ]\n %indvars.iv = phi i64 [ 1, %for.body.preheader ], [ %indvars.iv.next, %for.body ]\n %arrayidx4 = getelementptr inbounds i8, ptr %x, i64 %indvars.iv\n %1 = load i8, ptr %arrayidx4, align 1\n %add7 = add i8 %1, %0\n %arrayidx10 = getelementptr inbounds i8, ptr %y, i64 %indvars.iv\n store i8 %add7, ptr %arrayidx10, align 1\n %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\n %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count\n br i1 %exitcond.not, label %for.cond.cleanup, label %for.body\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/first-order-recurrence-chains-vplan.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -debug-only=loop-vectorize -disable-output -S %s 2>&1"], "tests": [{"test_name": "test_chained_first_order_recurrences_4", "test_body": "define i32 @test_chained_first_order_recurrences_4(ptr %base, i64 %x) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]\n %for.x = phi i64 [ %for.x.next, %loop ], [ 0, %entry ]\n %for.y = phi i32 [ %for.x.prev, %loop ], [ 0, %entry ]\n %iv.next = add i64 %iv, 1\n %gep = getelementptr i64, ptr %base, i64 %iv\n %for.x.prev = trunc i64 %for.x to i32\n %for.y.i64 = sext i32 %for.y to i64\n store i64 %for.y.i64, ptr %gep, align 4\n %for.x.next = mul i64 %x, 2\n %icmp = icmp ugt i64 %iv, 4096\n br i1 %icmp, label %ret, label %loop\n\nret: ; preds = %loop\n ret i32 0\n}\n"}, {"test_name": "test_chained_first_order_recurrences_3", "test_body": "define void @test_chained_first_order_recurrences_3(ptr %ptr) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %for.1 = phi i16 [ 22, %entry ], [ %for.1.next, %loop ]\n %for.2 = phi i16 [ 33, %entry ], [ %for.1, %loop ]\n %for.3 = phi i16 [ 33, %entry ], [ %for.2, %loop ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.next = add nuw nsw i64 %iv, 1\n %gep.ptr = getelementptr inbounds i16, ptr %ptr, i64 %iv\n %for.1.next = load i16, ptr %gep.ptr, align 2\n %add.1 = add i16 %for.1, %for.2\n %add.2 = add i16 %add.1, %for.3\n store i16 %add.2, ptr %gep.ptr, align 2\n %exitcond.not = icmp eq i64 %iv.next, 1000\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "test_chained_first_order_recurrences_1", "test_body": "define void @test_chained_first_order_recurrences_1(ptr %ptr) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %for.1 = phi i16 [ 22, %entry ], [ %for.1.next, %loop ]\n %for.2 = phi i16 [ 33, %entry ], [ %for.1, %loop ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.next = add nuw nsw i64 %iv, 1\n %gep.ptr = getelementptr inbounds i16, ptr %ptr, i64 %iv\n %for.1.next = load i16, ptr %gep.ptr, align 2\n %add = add i16 %for.1, %for.2\n store i16 %add, ptr %gep.ptr, align 2\n %exitcond.not = icmp eq i64 %iv.next, 1000\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "test_chained_first_order_recurrences_5_hoist_to_load", "test_body": "define i32 @test_chained_first_order_recurrences_5_hoist_to_load(ptr %base) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]\n %for.x = phi i64 [ %for.x.next, %loop ], [ 0, %entry ]\n %for.y = phi i32 [ %for.x.prev, %loop ], [ 0, %entry ]\n %iv.next = add i64 %iv, 1\n %gep = getelementptr i64, ptr %base, i64 %iv\n %l = load i64, ptr %gep, align 4\n %for.x.prev = trunc i64 %for.x to i32\n %for.y.i64 = sext i32 %for.y to i64\n store i64 %for.y.i64, ptr %gep, align 4\n %for.x.next = mul i64 %l, 2\n %icmp = icmp ugt i64 %iv, 4096\n br i1 %icmp, label %ret, label %loop\n\nret: ; preds = %loop\n ret i32 0\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/first-order-recurrence-multiply-recurrences.ll", "commands": [], "tests": [{"test_name": "test_pr54223_sink_after_insertion_order", "test_body": "define void @test_pr54223_sink_after_insertion_order(ptr noalias %a, ptr noalias %b, ptr noalias %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %for.1 = phi float [ 0.000000e+00, %entry ], [ %for.1.next, %loop ]\n %for.2 = phi float [ 0.000000e+00, %entry ], [ %for.2.next, %loop ]\n %neg = fneg float %for.2\n %muladd = call float @llvm.fmuladd.f32(float %for.1, float %neg, float 0.000000e+00)\n %dst.gep = getelementptr inbounds float, ptr %dst, i64 %iv\n %iv.next = add nuw nsw i64 %iv, 1\n %for.1.next = load float, ptr %a, align 4\n %for.2.next = load float, ptr %b, align 4\n store float %muladd, ptr %dst.gep, align 4\n %exitcond.not = icmp eq i64 %iv.next, 10000\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.fmuladd.f32(float, float, float) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "test_pr54233_for_depend_on_each_other", "test_body": "define void @test_pr54233_for_depend_on_each_other(ptr noalias %a, ptr noalias %b) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %for.1 = phi i32 [ 0, %entry ], [ %for.1.next, %loop ]\n %for.2 = phi i32 [ 0, %entry ], [ %for.2.next, %loop ]\n %or = or i32 %for.2, 10\n %shl = shl i32 %for.2, %for.1\n %xor = xor i32 %shl, 255\n %and = and i32 %xor, %or\n %for.1.next = xor i32 12, %for.2\n %for.2.next = load i32, ptr %b, align 4\n %a.gep = getelementptr inbounds i32, ptr %a, i64 %iv\n store i32 %and, ptr %a.gep, align 4\n %iv.next = add nuw i64 %iv, 1\n %exitcond = icmp eq i64 %iv, 1000\n br i1 %exitcond, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "Assertion `hasPlanWithVF(ScalarVF) && \"More than a single plan/VF w/o any plan having scalar VF\"' failed.", "body": "To reproduce run the following test with -passes=loop-vectorize\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat() gc \"statepoint-example\" {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb1, %bb\r\n %phi = phi i64 [ %add, %bb1 ], [ 1, %bb ]\r\n %phi2 = phi i32 [ %trunc, %bb1 ], [ 1, %bb ]\r\n %phi3 = phi i32 [ %or, %bb1 ], [ 0, %bb ]\r\n %or = or i32 %phi2, 0\r\n %add = add i64 %phi, 1\r\n store i32 %phi3, ptr addrspace(1) null, align 4\r\n %icmp = icmp ult i64 %phi, 337\r\n %trunc = trunc i64 %phi to i32\r\n br i1 %icmp, label %bb1, label %bb4\r\n\r\nbb4: ; preds = %bb1\r\n ret void\r\n}\r\n```\r\n\r\nReproducer: https://godbolt.org/z/e14eWvxMb\r\n\r\nStack dump:\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7304: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `hasPlanWithVF(ScalarVF) && \"More than a single plan/VF w/o any plan having scalar VF\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=loop-vectorize \r\n1.\tRunning pass \"function(loop-vectorize)\" on module \"\"\r\n2.\tRunning pass \"loop-vectorize\" on function \"wombat\"\r\n #0 0x0000000004fa4138 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fa4138)\r\n #1 0x0000000004fa18ac SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000073e073a42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000073e073a969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000073e073a42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000073e073a287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000073e073a2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000073e073a39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000003fe5276 llvm::LoopVectorizationPlanner::computeBestVF() (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3fe5276)\r\n #9 0x0000000004001214 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4001214)\r\n#10 0x0000000004003eb9 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4003eb9)\r\n#11 0x0000000004004543 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4004543)\r\n#12 0x0000000002ec85ae llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2ec85ae)\r\n#13 0x0000000004da08f8 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4da08f8)\r\n#14 0x0000000000df11be llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdf11be)\r\n#15 0x0000000004d9f2fe llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d9f2fe)\r\n#16 0x0000000000df0aae llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdf0aae)\r\n#17 0x0000000004d9d490 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d9d490)\r\n#18 0x00000000008e8002 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e8002)\r\n#19 0x00000000008daedc optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8daedc)\r\n#20 0x000073e073a29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#21 0x000073e073a29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#22 0x00000000008d296e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d296e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139", "author": "TatyanaDoubts", "labels": ["vectorizers", "crash-on-valid"], "comments": [{"author": "fhahn", "body": "Very interesting issue thanks! Proposed fix https://github.com/llvm/llvm-project/issues/106523"}]}, "verified": true} {"bug_id": "87394", "issue_url": "https://github.com/llvm/llvm-project/issues/87394", "bug_type": "crash", "base_commit": "d5022d9ad4aec250f77d21c819a9810a97b8b6a8", "knowledge_cutoff": "2024-04-02T18:48:53Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "adb0126ef11b0083d5a78be1534ccefa53def1cf", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp": [[35, 40], [44, 49]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp": ["VPTypeAnalysis::inferScalarTypeForRecipe"]}}, "patch": "commit adb0126ef11b0083d5a78be1534ccefa53def1cf\nAuthor: Patrick O'Neill \nDate: Tue Apr 23 07:48:43 2024 -0700\n\n [VPlan] Add scalar inferencing support for Not and Or insns (#89160)\n \n Fixes #87394.\n \n PR: https://github.com/llvm/llvm-project/pull/89160\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp\nindex ad4ea648cd61..5f93339083f0 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp\n@@ -35,6 +35,7 @@ Type *VPTypeAnalysis::inferScalarTypeForRecipe(const VPInstruction *R) {\n CachedTypes[OtherV] = ResTy;\n return ResTy;\n }\n+ case Instruction::Or:\n case Instruction::ICmp:\n case VPInstruction::FirstOrderRecurrenceSplice: {\n Type *ResTy = inferScalarType(R->getOperand(0));\n@@ -44,6 +45,12 @@ Type *VPTypeAnalysis::inferScalarTypeForRecipe(const VPInstruction *R) {\n CachedTypes[OtherV] = ResTy;\n return ResTy;\n }\n+ case VPInstruction::Not: {\n+ Type *ResTy = inferScalarType(R->getOperand(0));\n+ assert(IntegerType::get(Ctx, 1) == ResTy &&\n+ \"unexpected scalar type inferred for operand\");\n+ return ResTy;\n+ }\n case VPInstruction::PtrAdd:\n // Return the type based on the pointer argument (i.e. first operand).\n return inferScalarType(R->getOperand(0));\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/vplan-infer-not-or-type.ll", "commands": ["opt < %s -passes=loop-vectorize -S"], "tests": [{"test_name": "foo", "test_body": "define void @foo(i8 %arg.0, i8 %arg.1) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %latch, %entry\n %incrementor = phi i8 [ %add, %latch ], [ 0, %entry ]\n %and = and i8 %arg.0, %arg.1\n %extract.t = trunc i8 %and to i1\n br i1 %extract.t, label %latch, label %indirect.latch\n\nindirect.latch: ; preds = %loop.header\n br label %latch\n\nlatch: ; preds = %indirect.latch, %loop.header\n %add = add i8 %incrementor, 1\n %conv = zext i8 %incrementor to i32\n %cmp = icmp ult i32 %conv, 1\n br i1 %cmp, label %loop.header, label %loop.exit\n\nloop.exit: ; preds = %latch\n ret void\n}\n"}]}], "issue": {"title": "[LoopVectorize][VPlan] Unreachable executed \"Unhandled opcode!\"", "body": "Reduced LLVM IR:\r\n```llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main() #0 {\r\nentry:\r\n br label %for.body\r\n\r\nfor.cond.cleanup.loopexit: ; preds = %cond.end\r\n ret i32 0\r\n\r\nfor.body: ; preds = %cond.end, %entry\r\n %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %cond.end ]\r\n %conv3 = zext i1 false to i64\r\n %conv4 = trunc i64 %conv3 to i16\r\n br i1 false, label %cond.false, label %cond.end\r\n\r\ncond.false: ; preds = %for.body\r\n br label %cond.end\r\n\r\ncond.end: ; preds = %cond.false, %for.body\r\n %indvars.iv.next = add i64 %indvars.iv, 1\r\n %cmp = icmp ult i64 %indvars.iv, 1\r\n br i1 %cmp, label %for.body, label %for.cond.cleanup.loopexit\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\r\n```\r\n\r\nBacktrace:\r\n```\r\nUnhandled opcode!\r\nUNREACHABLE executed at /scratch/tc-testing/tc-apr-2/llvm/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp:58!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt --passes=loop-vectorize reduced.ll\r\n #0 0x00005f05c836eb60 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2d84b60)\r\n #1 0x00005f05c836bf6f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2d81f6f)\r\n #2 0x00005f05c836c0c5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x000071a702442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x000071a7024969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x000071a7024969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x000071a7024969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x000071a702442476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x000071a7024287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00005f05c82a2afe (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2cb8afe)\r\n#10 0x00005f05c74dfe67 llvm::VPTypeAnalysis::inferScalarTypeForRecipe(llvm::VPInstruction const*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1ef5e67)\r\n#11 0x00005f05c74dc88c llvm::VPTypeAnalysis::inferScalarType(llvm::VPValue const*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1ef288c)\r\n#12 0x00005f05c750790e simplifyRecipes(llvm::VPlan&, llvm::LLVMContext&) VPlanTransforms.cpp:0:0\r\n#13 0x00005f05c750e606 llvm::VPlanTransforms::optimize(llvm::VPlan&, llvm::ScalarEvolution&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1f24606)\r\n#14 0x00005f05c73ded66 llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1df4d66)\r\n#15 0x00005f05c73e5319 llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1dfb319)\r\n#16 0x00005f05c73e88a9 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1dfe8a9)\r\n#17 0x00005f05c73eb73e llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1e0173e)\r\n#18 0x00005f05c73ec88d llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1e0288d)\r\n#19 0x00005f05c62f3276 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd09276)\r\n#20 0x00005f05c8195141 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2bab141)\r\n#21 0x00005f05c62eabd6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd00bd6)\r\n#22 0x00005f05c8193e3b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2ba9e3b)\r\n#23 0x00005f05c62f2dc6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd08dc6)\r\n#24 0x00005f05c8191cb1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2ba7cb1)\r\n#25 0x00005f05c5b6cce5 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x582ce5)\r\n#26 0x00005f05c5b5f316 optMain (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x575316)\r\n#27 0x000071a702429d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#28 0x000071a702429e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#29 0x000071a702429e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#30 0x00005f05c5b54f95 _start (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x56af95)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt\r\n```\r\nGodbolt: https://godbolt.org/z/bvY138WfG\r\n\r\nTriggered assertion: https://github.com/llvm/llvm-project/blob/f5991161529511ca6ebc058da2a0507c2fc5283e/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp#L58\r\n\r\nFound via fuzzer", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "110494", "issue_url": "https://github.com/llvm/llvm-project/issues/110494", "bug_type": "crash", "base_commit": "1ba19d6c49663a35baf5d6cdc4a149e2aca9ae6c", "knowledge_cutoff": "2024-09-30T11:42:07Z", "lit_test_dir": ["llvm/test/Transforms/LoopStrengthReduce"], "hints": {"fix_commit": "7506872afc6c1cc7556eed5128a2030b6eb8ca29", "components": ["LoopStrengthReduce"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp": [[6870, 6875]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp": ["SalvageDVI"]}}, "patch": "commit 7506872afc6c1cc7556eed5128a2030b6eb8ca29\nAuthor: Orlando Cazalet-Hyams \nDate: Thu Oct 3 11:47:08 2024 +0100\n\n [DebugInfo][LSR] Fix assertion failure salvaging IV with offset > 64 bits wide (#110979)\n \n Fixes #110494\n\ndiff --git a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp\nindex f69db57c25ed..575395eda1c5 100644\n--- a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp\n+++ b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp\n@@ -6870,6 +6870,8 @@ static bool SalvageDVI(llvm::Loop *L, ScalarEvolution &SE,\n SE.computeConstantDifference(DVIRec.SCEVs[i], SCEVInductionVar)) {\n if (Offset->getSignificantBits() <= 64)\n SalvageExpr->createOffsetExpr(Offset->getSExtValue(), LSRInductionVar);\n+ else\n+ return false;\n } else if (!SalvageExpr->createIterCountExpr(DVIRec.SCEVs[i], IterCountExpr,\n SE))\n return false;\n", "tests": [{"file": "llvm/test/Transforms/LoopStrengthReduce/debuginfo-scev-salvage-large-width.ll", "commands": ["opt -S -loop-reduce %s"], "tests": [{"test_name": "main", "test_body": "define i16 @main() {\nentry:\n br label %for.cond29\n\nfor.cond29: ; preds = %for.body32, %entry\n %il_1000.0 = phi i128 [ 0, %entry ], [ %inc72, %for.body32 ]\n %l_995.0 = phi i128 [ 4704496199548239085565, %entry ], [ %inc70, %for.body32 ]\n #dbg_value(i128 %l_995.0, !4, !DIExpression(), !9)\n %cmp30 = icmp slt i128 %il_1000.0, 0\n br i1 %cmp30, label %for.body32, label %for.cond.cleanup31\n\nfor.cond.cleanup31: ; preds = %for.cond29\n ret i16 0\n\nfor.body32: ; preds = %for.cond29\n %inc70 = add i128 %l_995.0, 1\n %inc72 = add i128 %il_1000.0, 1\n br label %for.cond29\n}\n\n!llvm.dbg.cu = !{!0}\n!llvm.module.flags = !{!3}\n\n!0 = distinct !DICompileUnit(language: DW_LANG_C11, file: !1, producer: \"clang\", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, retainedTypes: !2, globals: !2, splitDebugInlining: false, nameTableKind: None)\n!1 = !DIFile(filename: \"foo.c\", directory: \"/tmp\")\n!2 = !{}\n!3 = !{i32 2, !\"Debug Info Version\", i32 3}\n!4 = !DILocalVariable(name: \"l_995\", scope: !5, file: !1, line: 414, type: !7)\n!5 = distinct !DISubprogram(name: \"main\", scope: !1, file: !1, line: 397, type: !6, scopeLine: 398, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)\n!6 = !DISubroutineType(types: !2)\n!7 = !DIDerivedType(tag: DW_TAG_typedef, name: \"__uint128_t\", file: !1, baseType: !8)\n!8 = !DIBasicType(name: \"unsigned __int128\", size: 128, encoding: DW_ATE_unsigned)\n!9 = !DILocation(line: 0, scope: !5)\n"}]}], "issue": {"title": "llc: ../lib/Transforms/Scalar/LoopStrengthReduce.cpp:6630: [...] Assertion `!LocationOps.empty() && \"Expected the location ops to contain the IV.\"' failed.", "body": "llvm commit: 82e594a022d\r\nReproduce with: ```llc bbi-99605.ll -o /dev/null```\r\nResult:\r\n```\r\nllc: ../lib/Transforms/Scalar/LoopStrengthReduce.cpp:6630: void (anonymous namespace)::SCEVDbgValueBuilder::appendToVectors(SmallVectorImpl &, SmallVectorImpl &): Assertion `!LocationOps.empty() && \"Expected the location ops to contain the IV.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: build-all/bin/llc bbi-99605.ll -o /dev/null\r\n1.\tRunning pass 'Function Pass Manager' on module 'bbi-99605.ll'.\r\n2.\tRunning pass 'Loop Pass Manager' on function '@main'\r\n3.\tRunning pass 'Loop Strength Reduction' on basic block '%for.cond29'\r\n #0 0x000055e150599557 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (build-all/bin/llc+0x6dee557)\r\n #1 0x000055e1505970be llvm::sys::RunSignalHandlers() (build-all/bin/llc+0x6dec0be)\r\n #2 0x000055e150599bff SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007f66b410bcf0 __restore_rt (/lib64/libpthread.so.0+0x12cf0)\r\n #4 0x00007f66b1cc4acf raise (/lib64/libc.so.6+0x4eacf)\r\n #5 0x00007f66b1c97ea5 abort (/lib64/libc.so.6+0x21ea5)\r\n #6 0x00007f66b1c97d79 _nl_load_domain.cold.0 (/lib64/libc.so.6+0x21d79)\r\n #7 0x00007f66b1cbd426 (/lib64/libc.so.6+0x47426)\r\n #8 0x000055e15005db21 (build-all/bin/llc+0x68b2b21)\r\n #9 0x000055e15005c0c3 SalvageDVI(llvm::Loop*, llvm::ScalarEvolution&, llvm::PHINode*, (anonymous namespace)::DVIRecoveryRec&, llvm::SCEV const*, (anonymous namespace)::SCEVDbgValueBuilder) LoopStrengthReduce.cpp:0:0\r\n#10 0x000055e15002dca8 ReduceLoopStrength(llvm::Loop*, llvm::IVUsers&, llvm::ScalarEvolution&, llvm::DominatorTree&, llvm::LoopInfo&, llvm::TargetTransformInfo const&, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::MemorySSA*) LoopStrengthReduce.cpp:0:0\r\n#11 0x000055e15005e7e8 (anonymous namespace)::LoopStrengthReduce::runOnLoop(llvm::Loop*, llvm::LPPassManager&) LoopStrengthReduce.cpp:0:0\r\n#12 0x000055e14f26f0f9 llvm::LPPassManager::runOnFunction(llvm::Function&) (build-all/bin/llc+0x5ac40f9)\r\n#13 0x000055e14fba9e38 llvm::FPPassManager::runOnFunction(llvm::Function&) (build-all/bin/llc+0x63fee38)\r\n#14 0x000055e14fbb2171 llvm::FPPassManager::runOnModule(llvm::Module&) (build-all/bin/llc+0x6407171)\r\n#15 0x000055e14fbaa8b5 llvm::legacy::PassManagerImpl::run(llvm::Module&) (build-all/bin/llc+0x63ff8b5)\r\n#16 0x000055e14d988be8 compileModule(char**, llvm::LLVMContext&) llc.cpp:0:0\r\n#17 0x000055e14d98626d main (build-all/bin/llc+0x41db26d)\r\n#18 0x00007f66b1cb0d85 __libc_start_main (/lib64/libc.so.6+0x3ad85)\r\n#19 0x000055e14d9822ee _start (build-all/bin/llc+0x41d72ee)\r\nAbort (core dumped)\r\n```\r\nThe input contains the new debug info:\r\n```#dbg_value(i128 %l_995.0, !4, !DIExpression(), !9)```\r\n\r\n[bbi-99605.ll.gz](https://github.com/user-attachments/files/17189125/bbi-99605.ll.gz)\r\n", "author": "mikaelholmen", "labels": ["debuginfo", "crash", "llvm:transforms"], "comments": [{"author": "OCHyams", "body": "Thanks for the small reproducer. That is fixed in #110979 - this PR just avoids us crash/asserting when a salvage attempt is made with an offset over 64 bits wide and poisons the dbg record instead. I didn't look into whether we could theoretically or practically salvage it.\r\n\r\nn.b., this case was broken with debug intrinsics too, not just the new debug record format."}]}, "verified": true} {"bug_id": "122691", "issue_url": "https://github.com/llvm/llvm-project/issues/122691", "bug_type": "crash", "base_commit": "26e13091ea5ac3a53d11b50265a506f88129d6ff", "knowledge_cutoff": "2025-01-13T11:28:13Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "092d6283838dea79670750b9415955c5f0cb5178", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[8091, 8096], [17818, 17824]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::buildTree_rec", "BoUpSLP::collectValuesToDemote"]}}, "patch": "commit 092d6283838dea79670750b9415955c5f0cb5178\nAuthor: Alexey Bataev \nDate: Mon Jan 13 09:02:56 2025 -0800\n\n [SLP]Check for div/rem instructions before extending with poisons\n \n Need to check if the instructions can be safely extended with poison\n before actually doing this to avoid incorrect transformations.\n \n Fixes #122691\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex df46c69ff3ab..4b0ed5b30179 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -8091,6 +8091,13 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n NonUniqueValueVL.append(\n PWSz - UniqueValues.size(),\n PoisonValue::get(UniqueValues.front()->getType()));\n+ // Check that extended with poisons operations are still valid for\n+ // vectorization (div/rem are not allowed).\n+ if (!getSameOpcode(NonUniqueValueVL, *TLI).valid()) {\n+ LLVM_DEBUG(dbgs() << \"SLP: Scalar used twice in bundle.\\n\");\n+ newTreeEntry(VL, std::nullopt /*not vectorized*/, S, UserTreeIdx);\n+ return false;\n+ }\n VL = NonUniqueValueVL;\n }\n return true;\n@@ -17818,7 +17825,7 @@ bool BoUpSLP::collectValuesToDemote(\n };\n if (E.isGather() || !Visited.insert(&E).second ||\n any_of(E.Scalars, [&](Value *V) {\n- return all_of(V->users(), [&](User *U) {\n+ return !isa(V) && all_of(V->users(), [&](User *U) {\n return isa(U) && !getTreeEntry(U);\n });\n }))\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/div-possibly-extended-with-poisons.ll", "commands": ["opt -S -O2 < %s"], "tests": [{"test_name": "", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@g_3 = internal unnamed_addr global i64 0\n@g_127 = external local_unnamed_addr global [3 x i32]\n@g_211 = external local_unnamed_addr global i32\n\ndefine i8 @func_125() {\nentry:\n %call42 = call ptr @func_178()\n ret i8 0\n}\n\ndefine ptr @func_178() {\nentry:\n %0 = load i32, ptr @g_211, align 4\n %tobool158.not = icmp eq i32 %0, 0\n br i1 %tobool158.not, label %for.inc434, label %for.cond166.preheader\n\nfor.cond166.preheader: ; preds = %entry\n br label %for.cond166\n\nfor.cond166: ; preds = %for.cond166.preheader, %for.cond166\n %1 = phi i64 [ %2, %for.cond166 ], [ poison, %for.cond166.preheader ]\n %2 = phi i64 [ %inc255, %for.cond166 ], [ 0, %for.cond166.preheader ]\n %cmp167 = icmp samesign ult i64 %2, 61\n %inc255 = add nuw nsw i64 %2, 1\n br i1 %cmp167, label %for.cond166, label %for.inc434.loopexit\n\nfor.inc434.loopexit: ; preds = %for.cond166\n %.lcssa = phi i64 [ %1, %for.cond166 ]\n store i64 %.lcssa, ptr @g_3, align 8\n br label %for.inc434\n\nfor.inc434: ; preds = %for.inc434.loopexit, %entry\n store i32 1, ptr @g_211, align 4\n %3 = load i64, ptr @g_3, align 8\n %conv.i = trunc nuw nsw i64 %3 to i32\n %4 = load i32, ptr @g_127, align 4\n %and.i = and i32 %4, %conv.i\n %cmp.i = icmp eq i32 %and.i, 0\n %conv8.i = sext i32 %and.i to i64\n %sext = shl i64 %3, 32\n %conv9.i = ashr exact i64 %sext, 32\n %g_211.promoted5 = load i32, ptr @g_211, align 4\n br label %for.body447\n\nfor.body447: ; preds = %for.inc434, %cleanup552\n %p_181.addr.08 = phi i64 [ -13, %for.inc434 ], [ %conv564, %cleanup552 ]\n %and14.i1.lcssa67 = phi i32 [ %g_211.promoted5, %for.inc434 ], [ %and14.i.lcssa, %cleanup552 ]\n br label %if.then489\n\nif.then489: ; preds = %for.body447, %safe_div_func_int64_t_s_s.exit\n %l_317.03 = phi i32 [ 2, %for.body447 ], [ %sub, %safe_div_func_int64_t_s_s.exit ]\n %and14.i12 = phi i32 [ %and14.i1.lcssa67, %for.body447 ], [ %and14.i, %safe_div_func_int64_t_s_s.exit ]\n br i1 %cmp.i, label %safe_div_func_int64_t_s_s.exit, label %lor.lhs.false.i\n\nlor.lhs.false.i: ; preds = %if.then489\n %div.i = sdiv i64 %conv8.i, %conv9.i\n %5 = trunc i64 %div.i to i16\n br label %safe_div_func_int64_t_s_s.exit\n\nsafe_div_func_int64_t_s_s.exit: ; preds = %lor.lhs.false.i, %if.then489\n %cond.i = phi i16 [ %5, %lor.lhs.false.i ], [ 0, %if.then489 ]\n %call12.i = tail call i16 @llvm.bswap.i16(i16 %cond.i)\n %conv13.i = zext i16 %call12.i to i32\n %and14.i = and i32 %and14.i12, %conv13.i\n %call26.i = load volatile ptr, ptr null, align 8\n %sub = add nsw i32 %l_317.03, -1\n %cmp482 = icmp sgt i32 %l_317.03, 0\n br i1 %cmp482, label %if.then489, label %cleanup552\n\ncleanup552: ; preds = %safe_div_func_int64_t_s_s.exit\n %and14.i.lcssa = phi i32 [ %and14.i, %safe_div_func_int64_t_s_s.exit ]\n %add.i = add nsw i64 %p_181.addr.08, 7\n %conv564 = and i64 %add.i, 255\n %6 = and i64 %p_181.addr.08, 255\n %cmp445.not = icmp eq i64 %6, 22\n br i1 %cmp445.not, label %for.end565, label %for.body447\n\nfor.end565: ; preds = %cleanup552\n %and14.i.lcssa.lcssa = phi i32 [ %and14.i.lcssa, %cleanup552 ]\n store i32 %conv.i, ptr getelementptr inbounds nuw (i8, ptr @g_127, i64 8), align 4\n store i32 %and14.i.lcssa.lcssa, ptr @g_211, align 4\n ret ptr null\n}\n\ndefine ptr @func_183(i64 %p_185) {\nentry:\n %call = call i64 @builtin_uaddl_overflow(i64 %p_185, i64 0)\n unreachable\n}\n\ndefine i64 @builtin_uaddl_overflow(i64 %x, i64 %y) {\nentry:\n %0 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %x, i64 %y)\n %1 = extractvalue { i64, i1 } %0, 1\n %conv = zext i1 %1 to i64\n ret i64 %conv\n}"}]}], "issue": {"title": "[SLPVectorizer] Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.", "body": "Reproducer: https://godbolt.org/z/68MKbnE4W\n```\n; bin/opt -O2 reduced.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@g_3 = internal unnamed_addr global i64 0\n@g_127 = external local_unnamed_addr global [3 x i32]\n@g_211 = external local_unnamed_addr global i32\n\ndefine i8 @func_125() {\nentry:\n %call42 = call ptr @func_178()\n ret i8 0\n}\n\ndefine ptr @func_178() {\nentry:\n %0 = load i32, ptr @g_211, align 4\n %tobool158.not = icmp eq i32 %0, 0\n br i1 %tobool158.not, label %for.inc434, label %for.cond166.preheader\n\nfor.cond166.preheader: ; preds = %entry\n br label %for.cond166\n\nfor.cond166: ; preds = %for.cond166.preheader, %for.cond166\n %1 = phi i64 [ %2, %for.cond166 ], [ poison, %for.cond166.preheader ]\n %2 = phi i64 [ %inc255, %for.cond166 ], [ 0, %for.cond166.preheader ]\n %cmp167 = icmp samesign ult i64 %2, 61\n %inc255 = add nuw nsw i64 %2, 1\n br i1 %cmp167, label %for.cond166, label %for.inc434.loopexit\n\nfor.inc434.loopexit: ; preds = %for.cond166\n %.lcssa = phi i64 [ %1, %for.cond166 ]\n store i64 %.lcssa, ptr @g_3, align 8\n br label %for.inc434\n\nfor.inc434: ; preds = %for.inc434.loopexit, %entry\n store i32 1, ptr @g_211, align 4\n %3 = load i64, ptr @g_3, align 8\n %conv.i = trunc nuw nsw i64 %3 to i32\n %4 = load i32, ptr @g_127, align 4\n %and.i = and i32 %4, %conv.i\n %cmp.i = icmp eq i32 %and.i, 0\n %conv8.i = sext i32 %and.i to i64\n %sext = shl i64 %3, 32\n %conv9.i = ashr exact i64 %sext, 32\n %g_211.promoted5 = load i32, ptr @g_211, align 4\n br label %for.body447\n\nfor.body447: ; preds = %for.inc434, %cleanup552\n %p_181.addr.08 = phi i64 [ -13, %for.inc434 ], [ %conv564, %cleanup552 ]\n %and14.i1.lcssa67 = phi i32 [ %g_211.promoted5, %for.inc434 ], [ %and14.i.lcssa, %cleanup552 ]\n br label %if.then489\n\nif.then489: ; preds = %for.body447, %safe_div_func_int64_t_s_s.exit\n %l_317.03 = phi i32 [ 2, %for.body447 ], [ %sub, %safe_div_func_int64_t_s_s.exit ]\n %and14.i12 = phi i32 [ %and14.i1.lcssa67, %for.body447 ], [ %and14.i, %safe_div_func_int64_t_s_s.exit ]\n br i1 %cmp.i, label %safe_div_func_int64_t_s_s.exit, label %lor.lhs.false.i\n\nlor.lhs.false.i: ; preds = %if.then489\n %div.i = sdiv i64 %conv8.i, %conv9.i\n %5 = trunc i64 %div.i to i16\n br label %safe_div_func_int64_t_s_s.exit\n\nsafe_div_func_int64_t_s_s.exit: ; preds = %lor.lhs.false.i, %if.then489\n %cond.i = phi i16 [ %5, %lor.lhs.false.i ], [ 0, %if.then489 ]\n %call12.i = tail call i16 @llvm.bswap.i16(i16 %cond.i)\n %conv13.i = zext i16 %call12.i to i32\n %and14.i = and i32 %and14.i12, %conv13.i\n %call26.i = load volatile ptr, ptr null, align 8\n %sub = add nsw i32 %l_317.03, -1\n %cmp482 = icmp sgt i32 %l_317.03, 0\n br i1 %cmp482, label %if.then489, label %cleanup552\n\ncleanup552: ; preds = %safe_div_func_int64_t_s_s.exit\n %and14.i.lcssa = phi i32 [ %and14.i, %safe_div_func_int64_t_s_s.exit ]\n %add.i = add nsw i64 %p_181.addr.08, 7\n %conv564 = and i64 %add.i, 255\n %6 = and i64 %p_181.addr.08, 255\n %cmp445.not = icmp eq i64 %6, 22\n br i1 %cmp445.not, label %for.end565, label %for.body447\n\nfor.end565: ; preds = %cleanup552\n %and14.i.lcssa.lcssa = phi i32 [ %and14.i.lcssa, %cleanup552 ]\n store i32 %conv.i, ptr getelementptr inbounds nuw (i8, ptr @g_127, i64 8), align 4\n store i32 %and14.i.lcssa.lcssa, ptr @g_211, align 4\n ret ptr null\n}\n\ndefine ptr @func_183(i64 %p_185) {\nentry:\n %call = call i64 @builtin_uaddl_overflow(i64 %p_185, i64 0)\n unreachable\n}\n\ndefine i64 @builtin_uaddl_overflow(i64 %x, i64 %y) {\nentry:\n %0 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %x, i64 %y)\n %1 = extractvalue { i64, i1 } %0, 1\n %conv = zext i1 %1 to i64\n ret i64 %conv\n}\n```\n```\nopt: /root/llvm-project/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = llvm::Instruction; From = llvm::Value]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\nStack dump:\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -O2 \n1.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"\"\n2.\tRunning pass \"slp-vectorizer\" on function \"func_125\"\n #0 0x00000000052803a8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x52803a8)\n #1 0x000000000527ddac SignalHandler(int) Signals.cpp:0:0\n #2 0x0000776446c42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #3 0x0000776446c969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #4 0x0000776446c42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #5 0x0000776446c287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #6 0x0000776446c2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #7 0x0000776446c39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #8 0x0000000000b6a41a (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xb6a41a)\n #9 0x0000000004253c98 llvm::slpvectorizer::BoUpSLP::collectValuesToDemote(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, unsigned int&, llvm::SmallVectorImpl&, llvm::DenseSet>&, llvm::SmallDenseSet> const&, unsigned int&, bool&, bool) const::'lambda10'(unsigned int, unsigned int)::operator()(unsigned int, unsigned int) const::'lambda'(llvm::Value*)::operator()(llvm::Value*) const SLPVectorizer.cpp:0:0\n#10 0x0000000004253f67 bool llvm::function_ref::callback_fn&, llvm::DenseSet>&, llvm::SmallDenseSet> const&, unsigned int&, bool&, bool) const::'lambda10'(unsigned int, unsigned int)>(long, unsigned int, unsigned int) SLPVectorizer.cpp:0:0\n#11 0x00000000042e2312 llvm::slpvectorizer::BoUpSLP::collectValuesToDemote(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, unsigned int&, llvm::SmallVectorImpl&, llvm::DenseSet>&, llvm::SmallDenseSet> const&, unsigned int&, bool&, bool) const::'lambda6'(unsigned int&, llvm::ArrayRef, llvm::function_ref)::operator()(unsigned int&, llvm::ArrayRef, llvm::function_ref) const SLPVectorizer.cpp:0:0\n#12 0x00000000042deacb llvm::slpvectorizer::BoUpSLP::collectValuesToDemote(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, unsigned int&, llvm::SmallVectorImpl&, llvm::DenseSet>&, llvm::SmallDenseSet> const&, unsigned int&, bool&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42deacb)\n#13 0x00000000042df8a0 llvm::slpvectorizer::BoUpSLP::computeMinimumValueSizes()::'lambda0'(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, bool, unsigned int, unsigned int, bool, bool)::operator()(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, bool, unsigned int, unsigned int, bool, bool) const SLPVectorizer.cpp:0:0\n#14 0x00000000042e03d0 llvm::slpvectorizer::BoUpSLP::computeMinimumValueSizes() (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42e03d0)\n#15 0x000000000431e02d (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&, llvm::AssumptionCache*) SLPVectorizer.cpp:0:0\n#16 0x000000000432015d llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x432015d)\n#17 0x000000000432494b llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\n#18 0x000000000432904f llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x432904f)\n#19 0x000000000432fa86 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\n#20 0x00000000043305bb llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x43305bb)\n#21 0x0000000003036f1e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3036f1e)\n#22 0x000000000507d7f0 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x507d7f0)\n#23 0x0000000000e7023e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe7023e)\n#24 0x000000000507c10b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x507c10b)\n#25 0x0000000000e6f8ce llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe6f8ce)\n#26 0x000000000507bb10 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x507bb10)\n#27 0x000000000094448a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x94448a)\n#28 0x0000000000936a29 optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x936a29)\n#29 0x0000776446c29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#30 0x0000776446c29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#31 0x000000000092e305 _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x92e305)\nProgram terminated with signal: SIGSEGV\nCompiler returned: 139\n```\n\nllvm version: 7e01a322f850e86be9eefde8ae5a30e532d22cfa\nI am working to provide a single-pass reproducer.\n", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": [{"author": "dtcxzyw", "body": "Missing poison check: https://github.com/llvm/llvm-project/blob/6c5941b09fca487efc5000c82bbce6054bf36a7c/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp#L18000\n"}, {"author": "dtcxzyw", "body": "> I am working to provide a single-pass reproducer.\n\nIt is weird that the single-pass reproducer doesn't work even with `--preserve-ll-uselistorder`.\n"}]}, "verified": true} {"bug_id": "120076", "issue_url": "https://github.com/llvm/llvm-project/issues/120076", "bug_type": "miscompilation", "base_commit": "2402bccc805614069efb0a47e1ef43349ea7ba1e", "knowledge_cutoff": "2024-12-16T13:09:17Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "d1a7225076218ce224cd29c74259b715b393dc9d", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[2931, 2943], [17515, 17522], [17528, 17533], [17623, 17630], [17926, 17931], [17949, 17954], [18071, 18078], [18206, 18212], [18214, 18219]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::collectValuesToDemote", "BoUpSLP::computeMinimumValueSizes"]}}, "patch": "commit d1a7225076218ce224cd29c74259b715b393dc9d\nAuthor: Alexey Bataev \nDate: Mon Dec 16 07:52:32 2024 -0800\n\n [SLP]Check if the node must keep its original bitwidth\n \n Need to check if during previous analysis the node has requested to keep\n its original bitwidth to avoid incorrect codegen.\n \n Fixes #120076\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 3bd983ee6e12..d967813075bb 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -2931,13 +2931,11 @@ private:\n /// truncation. We collect the entries that will be demoted in ToDemote.\n /// \\param E Node for analysis\n /// \\param ToDemote indices of the nodes to be demoted.\n- bool collectValuesToDemote(const TreeEntry &E, bool IsProfitableToDemoteRoot,\n- unsigned &BitWidth,\n- SmallVectorImpl &ToDemote,\n- DenseSet &Visited,\n- unsigned &MaxDepthLevel,\n- bool &IsProfitableToDemote,\n- bool IsTruncRoot) const;\n+ bool collectValuesToDemote(\n+ const TreeEntry &E, bool IsProfitableToDemoteRoot, unsigned &BitWidth,\n+ SmallVectorImpl &ToDemote, DenseSet &Visited,\n+ const SmallDenseSet &NodesToKeepBWs, unsigned &MaxDepthLevel,\n+ bool &IsProfitableToDemote, bool IsTruncRoot) const;\n \n /// Check if the operands on the edges \\p Edges of the \\p UserTE allows\n /// reordering (i.e. the operands can be reordered because they have only one\n@@ -17515,8 +17513,8 @@ unsigned BoUpSLP::getVectorElementSize(Value *V) {\n bool BoUpSLP::collectValuesToDemote(\n const TreeEntry &E, bool IsProfitableToDemoteRoot, unsigned &BitWidth,\n SmallVectorImpl &ToDemote, DenseSet &Visited,\n- unsigned &MaxDepthLevel, bool &IsProfitableToDemote,\n- bool IsTruncRoot) const {\n+ const SmallDenseSet &NodesToKeepBWs, unsigned &MaxDepthLevel,\n+ bool &IsProfitableToDemote, bool IsTruncRoot) const {\n // We can always demote constants.\n if (all_of(E.Scalars, IsaPred))\n return true;\n@@ -17528,6 +17526,10 @@ bool BoUpSLP::collectValuesToDemote(\n return true;\n }\n \n+ // Check if the node was analyzed already and must keep its original bitwidth.\n+ if (NodesToKeepBWs.contains(E.Idx))\n+ return false;\n+\n // If the value is not a vectorized instruction in the expression and not used\n // by the insertelement instruction and not used in multiple vector nodes, it\n // cannot be demoted.\n@@ -17623,8 +17625,8 @@ bool BoUpSLP::collectValuesToDemote(\n for (const TreeEntry *Op : Operands) {\n unsigned Level = InitLevel;\n if (!collectValuesToDemote(*Op, IsProfitableToDemoteRoot, BitWidth,\n- ToDemote, Visited, Level, IsProfitableToDemote,\n- IsTruncRoot)) {\n+ ToDemote, Visited, NodesToKeepBWs, Level,\n+ IsProfitableToDemote, IsTruncRoot)) {\n if (!IsProfitableToDemote)\n return false;\n NeedToExit = true;\n@@ -17926,6 +17928,7 @@ void BoUpSLP::computeMinimumValueSizes() {\n bool IsTruncRoot = false;\n bool IsProfitableToDemoteRoot = !IsStoreOrInsertElt;\n SmallVector RootDemotes;\n+ SmallDenseSet NodesToKeepBWs;\n if (NodeIdx != 0 &&\n VectorizableTree[NodeIdx]->State == TreeEntry::Vectorize &&\n VectorizableTree[NodeIdx]->getOpcode() == Instruction::Trunc) {\n@@ -17949,6 +17952,7 @@ void BoUpSLP::computeMinimumValueSizes() {\n // Check if the root is trunc and the next node is gather/buildvector, then\n // keep trunc in scalars, which is free in most cases.\n if (E.isGather() && IsTruncRoot && E.UserTreeIndices.size() == 1 &&\n+ !NodesToKeepBWs.contains(E.Idx) &&\n E.Idx > (IsStoreOrInsertElt ? 2u : 1u) &&\n all_of(E.Scalars, [&](Value *V) {\n return V->hasOneUse() || isa(V) ||\n@@ -18071,8 +18075,8 @@ void BoUpSLP::computeMinimumValueSizes() {\n bool NeedToDemote = IsProfitableToDemote;\n \n if (!collectValuesToDemote(E, IsProfitableToDemoteRoot, MaxBitWidth,\n- ToDemote, Visited, MaxDepthLevel, NeedToDemote,\n- IsTruncRoot) ||\n+ ToDemote, Visited, NodesToKeepBWs, MaxDepthLevel,\n+ NeedToDemote, IsTruncRoot) ||\n (MaxDepthLevel <= Limit &&\n !(((Opcode == Instruction::SExt || Opcode == Instruction::ZExt) &&\n (!IsTopRoot || !(IsStoreOrInsertElt || UserIgnoreList) ||\n@@ -18206,7 +18210,7 @@ void BoUpSLP::computeMinimumValueSizes() {\n });\n }\n \n- // If the maximum bit width we compute is less than the with of the roots'\n+ // If the maximum bit width we compute is less than the width of the roots'\n // type, we can proceed with the narrowing. Otherwise, do nothing.\n if (MaxBitWidth == 0 ||\n MaxBitWidth >=\n@@ -18214,6 +18218,7 @@ void BoUpSLP::computeMinimumValueSizes() {\n ->getBitWidth()) {\n if (UserIgnoreList)\n AnalyzedMinBWVals.insert(TreeRoot.begin(), TreeRoot.end());\n+ NodesToKeepBWs.insert(ToDemote.begin(), ToDemote.end());\n continue;\n }\n \n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/minbw-node-used-twice.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux < %s"], "tests": [{"test_name": "test", "test_body": "define i8 @test() {\nentry:\n %conv4.i.i = zext i16 0 to i32\n %conv7.i.i = sext i16 0 to i32\n %cmp8.i.i = icmp slt i32 %conv7.i.i, %conv4.i.i\n %conv9.i.i = zext i1 %cmp8.i.i to i32\n %or10.i.i = or i32 %conv9.i.i, %conv4.i.i\n %cmp11.i.i = icmp eq i32 %or10.i.i, %conv4.i.i\n %sub.i.i79.peel.i = sub i16 0, 1\n %xor5.i81.peel.i = zext i16 %sub.i.i79.peel.i to i32\n %conv7.i84.peel.i = sext i16 0 to i32\n %cmp8.i85.peel.i = icmp slt i32 %conv7.i84.peel.i, %xor5.i81.peel.i\n %conv9.i86.peel.i = zext i1 %cmp8.i85.peel.i to i32\n %or10.i87.peel.i = or i32 %conv9.i86.peel.i, %xor5.i81.peel.i\n %cmp11.i88.peel.i = icmp eq i32 %or10.i87.peel.i, %xor5.i81.peel.i\n %conv13.i89.peel.i = zext i1 %cmp8.i85.peel.i to i8\n ret i8 %conv13.i89.peel.i\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Miscompile at O3", "body": "Reproducer: https://alive2.llvm.org/ce/z/_88AAB\n```\n; bin/opt -passes=slp-vectorizer reduced.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i8 @func_129() {\nentry:\n %conv4.i.i = zext i16 0 to i32\n %conv7.i.i = sext i16 0 to i32\n %cmp8.i.i = icmp slt i32 %conv7.i.i, %conv4.i.i\n %conv9.i.i = zext i1 %cmp8.i.i to i32\n %or10.i.i = or i32 %conv9.i.i, %conv4.i.i\n %cmp11.i.i = icmp eq i32 %or10.i.i, %conv4.i.i\n %sub.i.i79.peel.i = sub i16 0, 1\n %xor5.i81.peel.i = zext i16 %sub.i.i79.peel.i to i32\n %conv7.i84.peel.i = sext i16 0 to i32\n %cmp8.i85.peel.i = icmp slt i32 %conv7.i84.peel.i, %xor5.i81.peel.i\n %conv9.i86.peel.i = zext i1 %cmp8.i85.peel.i to i32\n %or10.i87.peel.i = or i32 %conv9.i86.peel.i, %xor5.i81.peel.i\n %cmp11.i88.peel.i = icmp eq i32 %or10.i87.peel.i, %xor5.i81.peel.i\n %conv13.i89.peel.i = zext i1 %cmp8.i85.peel.i to i8\n ret i8 %conv13.i89.peel.i\n}\n```\n```\ndefine i8 @func_129() {\nentry:\n %sub.i.i79.peel.i = sub i16 0, 1\n %0 = insertelement <2 x i16> , i16 %sub.i.i79.peel.i, i32 0\n %1 = icmp slt <2 x i16> zeroinitializer, %0\n %2 = zext <2 x i1> %1 to <2 x i16>\n %3 = or <2 x i16> %2, %0\n %4 = icmp eq <2 x i16> %3, %0\n %5 = extractelement <2 x i1> %1, i32 0\n %conv13.i89.peel.i = zext i1 %5 to i8\n ret i8 %conv13.i89.peel.i\n}\n```\n```\n----------------------------------------\ndefine i8 @src() {\nentry:\n %sub.i.i79.peel.i = sub i16 0, 1\n %xor5.i81.peel.i = zext i16 %sub.i.i79.peel.i to i32\n %conv7.i84.peel.i = sext i16 0 to i32\n %cmp8.i85.peel.i = icmp slt i32 %conv7.i84.peel.i, %xor5.i81.peel.i\n %conv13.i89.peel.i = zext i1 %cmp8.i85.peel.i to i8\n ret i8 %conv13.i89.peel.i\n}\n=>\ndefine i8 @tgt() {\nentry:\n %sub.i.i79.peel.i = sub i16 0, 1\n %#0 = insertelement <2 x i16> { poison, 0 }, i16 %sub.i.i79.peel.i, i32 0\n %#1 = icmp slt <2 x i16> { 0, 0 }, %#0\n %#5 = extractelement <2 x i1> %#1, i32 0\n %conv13.i89.peel.i = zext i1 %#5 to i8\n ret i8 %conv13.i89.peel.i\n}\nTransformation doesn't verify!\n\nERROR: Value mismatch\n\nExample:\n\nSource:\ni16 %sub.i.i79.peel.i = #xffff (65535, -1)\ni32 %xor5.i81.peel.i = #x0000ffff (65535)\ni32 %conv7.i84.peel.i = #x00000000 (0)\ni1 %cmp8.i85.peel.i = #x1 (1)\ni8 %conv13.i89.peel.i = #x01 (1)\n\nTarget:\ni16 %sub.i.i79.peel.i = #xffff (65535, -1)\n<2 x i16> %#0 = < #xffff (65535, -1), #x0000 (0) >\n<2 x i1> %#1 = < #x0 (0), #x0 (0) >\ni1 %#5 = #x0 (0)\ni8 %conv13.i89.peel.i = #x00 (0)\nSource value: #x01 (1)\nTarget value: #x00 (0)\n```\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": [{"author": "fhahn", "body": "cc @alexey-bataev \n"}]}, "verified": true} {"bug_id": "122558", "issue_url": "https://github.com/llvm/llvm-project/issues/122558", "bug_type": "crash", "base_commit": "25f28ddd69ed2453726c0934ba6feea8ae6f10f8", "knowledge_cutoff": "2025-01-11T00:36:48Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "1de3dc7d23dd6b856efad3a3a04f2396328726d7", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[4052, 4058], [4064, 4069]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::computeMaxVF"]}}, "patch": "commit 1de3dc7d23dd6b856efad3a3a04f2396328726d7\nAuthor: Florian Hahn \nDate: Tue Jan 14 22:07:38 2025 +0000\n\n [LV] Bail out early if BTC+1 wraps.\n \n Currently we fail to detect the case where BTC + 1 wraps, i.e. the\n vector trip count is 0, In those cases, the minimum iteration count\n check will fail, and the vector code will never be executed.\n \n Explicitly check for this condition in computeMaxVF and avoid trying to\n vectorize alltogether.\n \n Note that a number of tests needed to be updated, because the vector\n loop would never be executed given the input IR.\n \n Fixes https://github.com/llvm/llvm-project/issues/122558.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 744faef19243..fe2fb5e9faae 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -4052,7 +4052,8 @@ LoopVectorizationCostModel::computeMaxVF(ElementCount UserVF, unsigned UserIC) {\n return FixedScalableVFPair::getNone();\n }\n \n- unsigned TC = PSE.getSE()->getSmallConstantTripCount(TheLoop);\n+ ScalarEvolution *SE = PSE.getSE();\n+ unsigned TC = SE->getSmallConstantTripCount(TheLoop);\n unsigned MaxTC = PSE.getSmallConstantMaxTripCount();\n LLVM_DEBUG(dbgs() << \"LV: Found trip count: \" << TC << '\\n');\n if (TC != MaxTC)\n@@ -4064,6 +4065,22 @@ LoopVectorizationCostModel::computeMaxVF(ElementCount UserVF, unsigned UserIC) {\n return FixedScalableVFPair::getNone();\n }\n \n+ // If BTC matches the widest induction type and is -1 then the trip count\n+ // computation will wrap to 0 and the vector trip count will be 0. Do not try\n+ // to vectorize.\n+ const SCEV *BTC = SE->getBackedgeTakenCount(TheLoop);\n+ if (!isa(BTC) &&\n+ BTC->getType()->getScalarSizeInBits() >=\n+ Legal->getWidestInductionType()->getScalarSizeInBits() &&\n+ SE->isKnownPredicate(CmpInst::ICMP_EQ, BTC,\n+ SE->getMinusOne(BTC->getType()))) {\n+ reportVectorizationFailure(\n+ \"Trip count computation wrapped\",\n+ \"backedge-taken count is -1, loop trip count wrapped to 0\",\n+ \"TripCountWrapped\", ORE, TheLoop);\n+ return FixedScalableVFPair::getNone();\n+ }\n+\n switch (ScalarEpilogueStatus) {\n case CM_ScalarEpilogueAllowed:\n return computeFeasibleMaxVF(MaxTC, UserVF, false);\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-epilogue.ll", "commands": ["opt -mattr=+dotprod -passes=loop-vectorize -force-vector-interleave=1 -S < %s"], "tests": [{"test_name": "dotp", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @dotp(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-mixed.ll", "commands": ["opt -passes=loop-vectorize -enable-epilogue-vectorization=false -mattr=+i8mm,+dotprod -S < %s", "opt -passes=loop-vectorize -enable-epilogue-vectorization=false -mattr=+dotprod -S < %s"], "tests": [{"test_name": "dotp_z_s", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @dotp_z_s(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = sext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "dotp_s_z", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @dotp_s_z(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = sext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-neon.ll", "commands": ["opt -passes=loop-vectorize -enable-epilogue-vectorization=false -mattr=+neon,+dotprod -force-vector-interleave=1 -S < %s", "opt -passes=loop-vectorize -enable-epilogue-vectorization=false -mattr=+neon,+dotprod -S < %s", "opt -passes=loop-vectorize -enable-epilogue-vectorization=false -mattr=+neon,+dotprod -force-vector-interleave=1 -vectorizer-maximize-bandwidth -S < %s"], "tests": [{"test_name": "dotp_unrolled", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum3 = phi i32 [ 0, %entry ], [ %add.a3, %for.body ]\n %accum2 = phi i32 [ 0, %entry ], [ %add.a2, %for.body ]\n %accum1 = phi i32 [ 0, %entry ], [ %add.a1, %for.body ]\n %accum0 = phi i32 [ 0, %entry ], [ %add.a0, %for.body ]\n %gep.a0 = getelementptr inbounds i8, ptr %a, i64 %iv\n %gep.b0 = getelementptr inbounds i8, ptr %b, i64 %iv\n %offset.1 = or disjoint i64 %iv, 1\n %gep.a1 = getelementptr inbounds i8, ptr %a, i64 %offset.1\n %gep.b1 = getelementptr inbounds i8, ptr %b, i64 %offset.1\n %offset.2 = or disjoint i64 %iv, 2\n %gep.a2 = getelementptr inbounds i8, ptr %a, i64 %offset.2\n %gep.b2 = getelementptr inbounds i8, ptr %b, i64 %offset.2\n %offset.3 = or disjoint i64 %iv, 3\n %gep.a3 = getelementptr inbounds i8, ptr %a, i64 %offset.3\n %gep.b3 = getelementptr inbounds i8, ptr %b, i64 %offset.3\n %load.a0 = load i8, ptr %gep.a0, align 1\n %ext.a0 = sext i8 %load.a0 to i32\n %load.b0 = load i8, ptr %gep.b0, align 1\n %ext.b0 = sext i8 %load.b0 to i32\n %mul.a0 = mul nsw i32 %ext.b0, %ext.a0\n %add.a0 = add nsw i32 %mul.a0, %accum0\n %load.a1 = load i8, ptr %gep.a1, align 1\n %ext.a1 = sext i8 %load.a1 to i32\n %load.b1 = load i8, ptr %gep.b1, align 1\n %ext.b1 = sext i8 %load.b1 to i32\n %mul.a1 = mul nsw i32 %ext.a1, %ext.b1\n %add.a1 = add nsw i32 %mul.a1, %accum1\n %load.a2 = load i8, ptr %gep.a2, align 1\n %ext.a2 = sext i8 %load.a2 to i32\n %load.b2 = load i8, ptr %gep.b2, align 1\n %ext.b2 = sext i8 %load.b2 to i32\n %mul.a2 = mul nsw i32 %ext.a2, %ext.b2\n %add.a2 = add nsw i32 %mul.a2, %accum2\n %load.a3 = load i8, ptr %gep.a3, align 1\n %ext.a3 = sext i8 %load.a3 to i32\n %load.b3 = load i8, ptr %gep.b3, align 1\n %ext.b3 = sext i8 %load.b3 to i32\n %mul.a3 = mul nsw i32 %ext.a3, %ext.b3\n %add.a3 = add nsw i32 %mul.a3, %accum3\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, %num_in\n br i1 %exitcond.not, label %exit, label %for.body\n\nexit: ; preds = %for.body\n %result0 = add nsw i32 %add.a0, %add.a1\n %result1 = add nsw i32 %add.a2, %add.a3\n %result = add nsw i32 %result0, %result1\n ret i32 %result\n}\n"}, {"test_name": "dotp", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @dotp(ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n"}, {"test_name": "not_dotp_predicated_pragma", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @not_dotp_predicated_pragma(i64 %N, ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr inbounds i8, ptr %b, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = sext i8 %load.a to i32\n %gep.a2 = getelementptr inbounds i8, ptr %a, i64 %iv\n %load.b = load i8, ptr %gep.a2, align 1\n %ext.b = sext i8 %load.b to i32\n %mul = mul nsw i32 %ext.b, %ext.a\n %add = add nsw i32 %mul, %accum\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, %N\n br i1 %exitcond.not, label %exit, label %for.body, !llvm.loop !0\n\nexit: ; preds = %for.body\n ret i32 %add\n}\n\n!0 = distinct !{!0, !1, !2, !3}\n!1 = !{!\"llvm.loop.mustprogress\"}\n!2 = !{!\"llvm.loop.vectorize.predicate.enable\", i1 true}\n!3 = !{!\"llvm.loop.vectorize.enable\", i1 true}\n"}, {"test_name": "not_dotp_not_loop_carried", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %mul, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n"}, {"test_name": "not_dotp_different_types", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @not_dotp_different_types(ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i16, ptr %gep.b, align 2\n %ext.b = zext i16 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n"}, {"test_name": "not_dotp_not_phi", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @not_dotp_not_phi(ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %ext.b\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n"}, {"test_name": "not_dotp_extend_user", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @not_dotp_extend_user(ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n %result = add i32 %add, %ext.b\n ret i32 %result\n}\n"}, {"test_name": "not_dotp_predicated", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @not_dotp_predicated(i64 %N, ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr inbounds i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = sext i8 %load.a to i32\n %gep.b = getelementptr inbounds i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = sext i8 %load.b to i32\n %mul = mul nsw i32 %ext.b, %ext.a\n %add = add nsw i32 %mul, %accum\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, %N\n br i1 %exitcond.not, label %exit, label %for.body\n\nexit: ; preds = %for.body\n ret i32 %add\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product.ll", "commands": ["opt -passes=loop-vectorize -force-vector-interleave=1 -enable-epilogue-vectorization=false -S < %s", "opt -passes=loop-vectorize -enable-epilogue-vectorization=false -S < %s", "opt -passes=loop-vectorize -force-vector-interleave=1 -vectorizer-maximize-bandwidth -enable-epilogue-vectorization=false -S < %s"], "tests": [{"test_name": "not_dotp_not_phi2", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine void @not_dotp_not_phi2(ptr %matrix, i32 %n) #0 {\nentry:\n %cmp = icmp sgt i32 %n, 0\n br i1 %cmp, label %for.preheader, label %exit\n\nfor.preheader: ; preds = %entry\n %load.a = load i8, ptr null, align 1\n %load.a1 = load i8, ptr inttoptr (i64 1 to ptr), align 1\n %a.ext = sext i8 %load.a to i32\n %a.ext1 = sext i8 %load.a1 to i32\n br label %for.body\n\nfor.body: ; preds = %for.body, %for.preheader\n %iv = phi i32 [ %iv.next, %for.body ], [ 0, %for.preheader ]\n %ptr = phi ptr [ %scevgep, %for.body ], [ %matrix, %for.preheader ]\n %accum = phi i32 [ %add.1, %for.body ], [ 0, %for.preheader ]\n %gep.b = getelementptr i8, ptr %ptr, i64 1\n %gep.b1 = getelementptr i8, ptr %ptr, i64 2\n %load.b = load i8, ptr %gep.b, align 1\n %b.ext = sext i8 %load.b to i32\n %mul = mul nsw i32 %a.ext, %b.ext\n %add = add i32 %mul, %accum\n %load.b1 = load i8, ptr %gep.b1, align 1\n %b.ext1 = sext i8 %load.b1 to i32\n %mul.1 = mul nsw i32 %a.ext1, %b.ext1\n %add.1 = add i32 %mul.1, %add\n %scevgep = getelementptr i8, ptr %ptr, i64 16\n %iv.next = add nuw nsw i32 %iv, 1\n %exitcond.not = icmp eq i32 %iv.next, %n\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n %add.1.lcssa = phi i32 [ %add.1, %for.body ]\n %add.float = sitofp i32 %add.1.lcssa to float\n br label %exit\n\nexit: ; preds = %for.exit, %entry\n %result = phi float [ 0.000000e+00, %entry ], [ %add.float, %for.exit ]\n store float %result, ptr %matrix, align 4\n ret void\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "not_dotp_ext_outside_plan2", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i64 @not_dotp_ext_outside_plan2(ptr %a, i16 %b, i64 %n) #0 {\nentry:\n %cmp = icmp eq i64 %n, 0\n br i1 %cmp, label %exit, label %for.ph\n\nfor.ph: ; preds = %entry\n %ext.b = zext i16 %b to i64\n br label %for.body\n\nfor.body: ; preds = %for.body, %for.ph\n %iv = phi i64 [ 0, %for.ph ], [ %iv.next, %for.body ]\n %accum = phi i64 [ 0, %for.ph ], [ %add, %for.body ]\n %gep.a = getelementptr inbounds nuw i16, ptr %a, i64 %iv\n %load.a = load i16, ptr %gep.a, align 2\n %ext.a = zext i16 %load.a to i64\n %mul = mul nuw nsw i64 %ext.b, %ext.a\n %add = add i64 %mul, %accum\n %iv.next = add nuw nsw i64 %iv, 1\n %cmp.1 = icmp eq i64 %iv.next, %n\n br i1 %cmp.1, label %exit, label %for.body\n\nexit: ; preds = %for.body, %entry\n %result = phi i64 [ 0, %entry ], [ %add, %for.body ]\n ret i64 %result\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "dotp_unrolled", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum3 = phi i32 [ 0, %entry ], [ %add.a3, %for.body ]\n %accum2 = phi i32 [ 0, %entry ], [ %add.a2, %for.body ]\n %accum1 = phi i32 [ 0, %entry ], [ %add.a1, %for.body ]\n %accum0 = phi i32 [ 0, %entry ], [ %add.a0, %for.body ]\n %gep.a0 = getelementptr inbounds i8, ptr %a, i64 %iv\n %gep.b0 = getelementptr inbounds i8, ptr %b, i64 %iv\n %offset.1 = or disjoint i64 %iv, 1\n %gep.a1 = getelementptr inbounds i8, ptr %a, i64 %offset.1\n %gep.b1 = getelementptr inbounds i8, ptr %b, i64 %offset.1\n %offset.2 = or disjoint i64 %iv, 2\n %gep.a2 = getelementptr inbounds i8, ptr %a, i64 %offset.2\n %gep.b2 = getelementptr inbounds i8, ptr %b, i64 %offset.2\n %offset.3 = or disjoint i64 %iv, 3\n %gep.a3 = getelementptr inbounds i8, ptr %a, i64 %offset.3\n %gep.b3 = getelementptr inbounds i8, ptr %b, i64 %offset.3\n %load.a0 = load i8, ptr %gep.a0, align 1\n %ext.a0 = sext i8 %load.a0 to i32\n %load.b0 = load i8, ptr %gep.b0, align 1\n %ext.b0 = sext i8 %load.b0 to i32\n %mul.a0 = mul nsw i32 %ext.b0, %ext.a0\n %add.a0 = add nsw i32 %mul.a0, %accum0\n %load.a1 = load i8, ptr %gep.a1, align 1\n %ext.a1 = sext i8 %load.a1 to i32\n %load.b1 = load i8, ptr %gep.b1, align 1\n %ext.b1 = sext i8 %load.b1 to i32\n %mul.a1 = mul nsw i32 %ext.a1, %ext.b1\n %add.a1 = add nsw i32 %mul.a1, %accum1\n %load.a2 = load i8, ptr %gep.a2, align 1\n %ext.a2 = sext i8 %load.a2 to i32\n %load.b2 = load i8, ptr %gep.b2, align 1\n %ext.b2 = sext i8 %load.b2 to i32\n %mul.a2 = mul nsw i32 %ext.a2, %ext.b2\n %add.a2 = add nsw i32 %mul.a2, %accum2\n %load.a3 = load i8, ptr %gep.a3, align 1\n %ext.a3 = sext i8 %load.a3 to i32\n %load.b3 = load i8, ptr %gep.b3, align 1\n %ext.b3 = sext i8 %load.b3 to i32\n %mul.a3 = mul nsw i32 %ext.a3, %ext.b3\n %add.a3 = add nsw i32 %mul.a3, %accum3\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, %num_in\n br i1 %exitcond.not, label %exit, label %for.body\n\nexit: ; preds = %for.body\n %result0 = add nsw i32 %add.a0, %add.a1\n %result1 = add nsw i32 %add.a2, %add.a3\n %result = add nsw i32 %result0, %result1\n ret i32 %result\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "dotp", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @dotp(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "not_dotp_predicated_pragma", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @not_dotp_predicated_pragma(i64 %N, ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr inbounds i8, ptr %b, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = sext i8 %load.a to i32\n %gep.a2 = getelementptr inbounds i8, ptr %a, i64 %iv\n %load.b = load i8, ptr %gep.a2, align 1\n %ext.b = sext i8 %load.b to i32\n %mul = mul nsw i32 %ext.b, %ext.a\n %add = add nsw i32 %mul, %accum\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, %N\n br i1 %exitcond.not, label %exit, label %for.body, !llvm.loop !0\n\nexit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n\n!0 = distinct !{!0, !1, !2, !3}\n!1 = !{!\"llvm.loop.mustprogress\"}\n!2 = !{!\"llvm.loop.vectorize.predicate.enable\", i1 true}\n!3 = !{!\"llvm.loop.vectorize.enable\", i1 true}\n"}, {"test_name": "dotp_cost_disagreement", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i.iv = phi i64 [ 0, %entry ], [ %i.iv.next, %for.body ]\n %sum = phi i64 [ 0, %entry ], [ %add, %for.body ]\n %arrayidx = getelementptr inbounds nuw i8, ptr %a, i64 %i.iv\n %0 = load i8, ptr %arrayidx, align 1\n %conv = zext i8 %0 to i64\n %i.iv.next = add nuw nsw i64 %i.iv, 1\n %arrayidx2 = getelementptr inbounds nuw i8, ptr %b, i64 %i.iv.next\n %1 = load i8, ptr %arrayidx2, align 1\n %conv3 = zext i8 %1 to i64\n %mul = mul nuw nsw i64 %conv3, %conv\n %add = add i64 %sum, %mul\n %exitcond.not = icmp eq i64 %i.iv.next, 16\n br i1 %exitcond.not, label %exit, label %for.body\n\nexit: ; preds = %for.body\n ret i64 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "not_dotp_not_loop_carried", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %mul, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "not_dotp_different_types", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @not_dotp_different_types(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i16, ptr %gep.b, align 2\n %ext.b = zext i16 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "not_dotp_not_phi", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @not_dotp_not_phi(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %ext.b\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "not_dotp_extend_user", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @not_dotp_extend_user(ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n %result = add i32 %add, %ext.b\n ret i32 %result\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "not_dotp_ext_outside_plan", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i64 @not_dotp_ext_outside_plan(ptr %a, i16 %b, i64 %n) #0 {\nentry:\n %cmp = icmp eq i64 %n, 0\n br i1 %cmp, label %exit, label %for.ph\n\nfor.ph: ; preds = %entry\n %ext.b = zext i16 %b to i64\n br label %for.body\n\nfor.body: ; preds = %for.body, %for.ph\n %iv = phi i64 [ 0, %for.ph ], [ %iv.next, %for.body ]\n %accum = phi i64 [ 0, %for.ph ], [ %add, %for.body ]\n %gep.a = getelementptr inbounds nuw i16, ptr %a, i64 %iv\n %load.a = load i16, ptr %gep.a, align 2\n %ext.a = zext i16 %load.a to i64\n %mul = mul nuw nsw i64 %ext.a, %ext.b\n %add = add i64 %mul, %accum\n %iv.next = add nuw nsw i64 %iv, 1\n %cmp.1 = icmp eq i64 %iv.next, %n\n br i1 %cmp.1, label %exit, label %for.body\n\nexit: ; preds = %for.body, %entry\n %result = phi i64 [ 0, %entry ], [ %add, %for.body ]\n ret i64 %result\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}, {"test_name": "not_dotp_predicated", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\n; Function Attrs: vscale_range(1,16)\ndefine i32 @not_dotp_predicated(i64 %N, ptr %a, ptr %b) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr inbounds i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = sext i8 %load.a to i32\n %gep.b = getelementptr inbounds i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = sext i8 %load.b to i32\n %mul = mul nsw i32 %ext.b, %ext.a\n %add = add nsw i32 %mul, %accum\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, %N\n br i1 %exitcond.not, label %exit, label %for.body\n\nexit: ; preds = %for.body\n ret i32 %add\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-no-dotprod.ll", "commands": ["opt -passes=loop-vectorize -enable-epilogue-vectorization=false -mattr=+neon -S < %s"], "tests": [{"test_name": "not_dotp", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @not_dotp(ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1000\n br i1 %exitcond.not, label %for.exit, label %for.body\n\nfor.exit: ; preds = %for.body\n ret i32 %add\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll", "commands": ["opt -p loop-vectorize -S %s", "opt -p loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -S %s"], "tests": [{"test_name": "trunc_store", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"arm64-apple-macosx14.0.0\"\n\n; Function Attrs: vscale_range(1,16)\ndefine void @trunc_store(ptr %dst, ptr %src, i16 %x) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %x.ext = zext i16 %x to i64\n %l = load i64, ptr %src, align 8\n %and = and i64 %l, %x.ext\n %trunc = trunc i64 %and to i8\n %gep = getelementptr i8, ptr %dst, i64 %iv\n store i8 %trunc, ptr %gep, align 1\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { vscale_range(1,16) \"target-features\"=\"+sve\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll", "commands": ["opt -mattr=+neon,+dotprod -passes=loop-vectorize -debug-only=loop-vectorize -force-vector-interleave=1 -disable-output %s 2>&1"], "tests": [{"test_name": "print_partial_reduction", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-none-unknown-elf\"\n\ndefine i32 @print_partial_reduction(ptr %a, ptr %b) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %accum = phi i32 [ 0, %entry ], [ %add, %for.body ]\n %gep.a = getelementptr i8, ptr %a, i64 %iv\n %load.a = load i8, ptr %gep.a, align 1\n %ext.a = zext i8 %load.a to i32\n %gep.b = getelementptr i8, ptr %b, i64 %iv\n %load.b = load i8, ptr %gep.b, align 1\n %ext.b = zext i8 %load.b to i32\n %mul = mul i32 %ext.b, %ext.a\n %add = add i32 %mul, %accum\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1024\n br i1 %exitcond.not, label %exit, label %for.body\n\nexit: ; preds = %for.body\n ret i32 %add\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/Hexagon/maximum-vf-crash.ll", "commands": ["opt -mtriple=hexagon -hexagon-autohvx -passes=loop-vectorize -S < %s 2>&1"], "tests": [{"test_name": "f", "test_body": "target datalayout = \"e-m:e-p:32:32:32-a:0-n16:32-i64:64:64-i32:32:32-i16:16:16-i1:8:8-f32:32:32-f64:64:64-v32:32:32-v64:64:64-v512:512:512-v1024:1024:1024-v2048:2048:2048\"\n\n; Function Attrs: optsize\ndefine i32 @f(ptr %src) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %g.016 = phi i32 [ 0, %entry ], [ %g.1.lcssa, %loop ]\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %0 = load i8, ptr %src, align 1\n %g.1.lcssa = add i32 %g.016, 1\n %iv.next = add nsw i32 %iv, 1\n %exitcond = icmp eq i32 %iv.next, 1000\n br i1 %exitcond, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i32 %g.1.lcssa\n}\n\nattributes #0 = { optsize \"target-features\"=\"+hvx-length128b\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/cost-constant-known-via-scev.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "second_lshr_operand_zero_via_scev", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i64 @second_lshr_operand_zero_via_scev() {\nentry:\n %ext.0 = sext i8 0 to i32\n br label %loops\n\nloops: ; preds = %loops, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loops ]\n %red = phi i64 [ 0, %entry ], [ %red.next, %loops ]\n %c = icmp eq i64 %iv, 0\n %and = and i64 %iv, 0\n %0 = trunc i64 %iv to i32\n %shr = lshr i32 %0, %ext.0\n %conv.1 = zext i32 %shr to i64\n %red.next.v = select i1 %c, i64 %and, i64 %conv.1\n %red.next = or i64 %red.next.v, %red\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loops\n\nexit: ; preds = %loops\n %res = phi i64 [ %red.next, %loops ]\n ret i64 %res\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/cost-model.ll", "commands": ["opt < %s -passes=loop-vectorize -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -S"], "tests": [{"test_name": "narrowed_reduction", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i32 @narrowed_reduction(ptr %a, i1 %cmp) #0 {\nentry:\n %conv = zext i1 %cmp to i32\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 1, %entry ], [ %inc, %loop ]\n %or13 = phi i32 [ 0, %entry ], [ %or, %loop ]\n %and = and i32 %or13, 1\n %or = or i32 %and, %conv\n %inc = add i32 %iv, 1\n %ec = icmp eq i32 %iv, 16\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i32 %or\n}\n\nattributes #0 = { \"target-cpu\"=\"penryn\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/drop-poison-generating-flags.ll", "commands": ["opt %s -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S"], "tests": [{"test_name": "drop_zext_nneg", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind uwtable\ndefine void @drop_zext_nneg(ptr noalias %p, ptr noalias %p1) #0 {\nentry:\n br label %body\n\nbody: ; preds = %else, %entry\n %iv = phi i64 [ %next, %else ], [ 0, %entry ]\n %0 = trunc i64 %iv to i32\n %c = icmp eq i32 %0, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %body\n %zext = zext nneg i32 %0 to i64\n %idx1 = getelementptr double, ptr %p, i64 %zext\n %idx2 = getelementptr double, ptr %p, i64 %zext\n %1 = load double, ptr %idx2, align 8\n br label %else\n\nelse: ; preds = %then, %body\n %phi = phi double [ %1, %then ], [ 0.000000e+00, %body ]\n store double %phi, ptr %p1, align 8\n %next = add i64 %iv, 1\n %cmp = icmp eq i64 %next, 1024\n br i1 %cmp, label %exit, label %body\n\nexit: ; preds = %else\n ret void\n}\n\nattributes #0 = { noinline nounwind uwtable \"target-features\"=\"+avx512bw,+avx512cd,+avx512dq,+avx512f,+avx512vl\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/interleave-cost.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "test_free_instructions_feeding_geps_for_interleave_groups", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.15.0\"\n\ndefine void @test_free_instructions_feeding_geps_for_interleave_groups(ptr noalias %p.invar, ptr noalias %dst.1, ptr noalias %dst.2) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %l.0 = load float, ptr %p.invar, align 4\n %iv.mul = shl i64 %iv, 2\n %gep.dst.19 = getelementptr float, ptr %dst.1, i64 %iv.mul\n store float %l.0, ptr %gep.dst.19, align 4\n %l.1 = load float, ptr %p.invar, align 4\n %add.1 = or disjoint i64 %iv.mul, 1\n %gep.dst.119 = getelementptr float, ptr %dst.1, i64 %add.1\n store float %l.1, ptr %gep.dst.119, align 4\n %add.2 = or disjoint i64 %iv.mul, 2\n %gep.dst.129 = getelementptr float, ptr %dst.1, i64 %add.2\n store float 0.000000e+00, ptr %gep.dst.129, align 4\n %add.3 = or disjoint i64 %iv.mul, 3\n %gep.dst.140 = getelementptr float, ptr %dst.1, i64 %add.3\n store float 0.000000e+00, ptr %gep.dst.140, align 4\n %l.2 = load float, ptr %p.invar, align 4\n %gep.dst.247 = getelementptr float, ptr %dst.2, i64 %iv.mul\n store float %l.2, ptr %gep.dst.247, align 4\n %gep.dst.255 = getelementptr float, ptr %dst.2, i64 %add.1\n store float 0.000000e+00, ptr %gep.dst.255, align 4\n %gep.dst.265 = getelementptr float, ptr %dst.2, i64 %add.2\n store float 0.000000e+00, ptr %gep.dst.265, align 4\n %gep.dst.276 = getelementptr float, ptr %dst.2, i64 %add.3\n store float 0.000000e+00, ptr %gep.dst.276, align 4\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 1024\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/replicate-uniform-call.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "smax_call_uniform", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @smax_call_uniform(ptr %dst, i64 %x) {\nentry:\n %c = icmp ult i8 -68, -69\n %mul = mul nuw nsw i64 %x, 0\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %loop.latch, label %else\n\nelse: ; preds = %loop.header\n %rem = urem i64 %mul, %x\n %smax = tail call i64 @llvm.smax.i64(i64 %rem, i64 0)\n br label %loop.latch\n\nloop.latch: ; preds = %else, %loop.header\n %p = phi i64 [ 1, %loop.header ], [ %smax, %else ]\n %add = add i64 %p, 1\n %gep = getelementptr i64, ptr %dst, i64 %add\n store i64 0, ptr %gep, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 1024\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i64 @llvm.smax.i64(i64, i64) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/epilog-vectorization-vector-trip-count-zero.ll", "commands": ["opt -p loop-vectorize -force-vector-width=2 -epilogue-vectorization-force-VF=2 -S %s"], "tests": [{"test_name": "vector_trip_count_0_as_btc_is_all_1", "test_body": "target datalayout = \"E-m:e-i1:8:16-i8:8:16-i64:64-f128:64-v128:64-a:8:16-n32:64\"\n\ndefine void @vector_trip_count_0_as_btc_is_all_1(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.next = add i32 %iv, -1\n %gep = getelementptr inbounds i32, ptr %dst, i32 %iv\n store i32 %iv.next, ptr %gep, align 4\n %ec = icmp eq i32 %iv.next, 0\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/first-order-recurrence-chains.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "test_resinking_required", "test_body": "define double @test_resinking_required(ptr %p, ptr noalias %a, ptr noalias %b) {\nEntry:\n br label %Loop\n\nLoop: ; preds = %Loop, %Entry\n %for.1 = phi double [ %l1, %Loop ], [ 0.000000e+00, %Entry ]\n %for.2 = phi double [ %l2, %Loop ], [ 0.000000e+00, %Entry ]\n %for.3 = phi double [ %for.2, %Loop ], [ 0.000000e+00, %Entry ]\n %iv = phi i64 [ %iv.next, %Loop ], [ 0, %Entry ]\n %USE_2_FORS = fdiv double %for.3, %for.1\n %div = fdiv double 0.000000e+00, %for.1\n %l1 = load double, ptr %a, align 8\n %iv.next = add nuw nsw i64 %iv, 1\n %l2 = load double, ptr %b, align 8\n store double %div, ptr %p, align 8\n %cond = icmp eq i64 %iv.next, 1000\n br i1 %cond, label %End, label %Loop\n\nEnd: ; preds = %Loop\n %res.1 = fadd double %for.1, %for.2\n %res.2 = fadd double %res.1, %for.3\n ret double %res.2\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/if-pred-stores.ll", "commands": ["opt -S -vectorize-num-stores-pred=1 -force-vector-width=1 -force-vector-interleave=2 -passes=loop-vectorize,simplifycfg -verify-loop-info -simplifycfg-require-and-preserve-domtree=1 < %s", "opt -S -vectorize-num-stores-pred=1 -force-vector-width=1 -force-vector-interleave=2 -passes=loop-vectorize -verify-loop-info < %s", "opt -S -vectorize-num-stores-pred=1 -force-vector-width=2 -force-vector-interleave=1 -passes=loop-vectorize,simplifycfg -verify-loop-info -simplifycfg-require-and-preserve-domtree=1 < %s"], "tests": [{"test_name": "minimal_bit_widths_with_aliasing_store", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @minimal_bit_widths_with_aliasing_store(i1 %c, ptr %ptr) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.inc, %entry\n %tmp0 = phi i64 [ %tmp6, %for.inc ], [ 0, %entry ]\n %tmp1 = phi i64 [ %tmp7, %for.inc ], [ 1000, %entry ]\n %tmp2 = getelementptr i8, ptr %ptr, i64 %tmp0\n %tmp3 = load i8, ptr %tmp2, align 1\n store i8 0, ptr %tmp2, align 1\n br i1 %c, label %if.then, label %for.inc\n\nif.then: ; preds = %for.body\n %tmp4 = zext i8 %tmp3 to i32\n %tmp5 = trunc i32 %tmp4 to i8\n store i8 %tmp5, ptr %tmp2, align 1\n br label %for.inc\n\nfor.inc: ; preds = %if.then, %for.body\n %tmp6 = add nuw nsw i64 %tmp0, 1\n %tmp7 = add i64 %tmp1, -1\n %tmp8 = icmp eq i64 %tmp7, 0\n br i1 %tmp8, label %for.end, label %for.body\n\nfor.end: ; preds = %for.inc\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/induction.ll", "commands": ["opt < %s -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 -force-widen-divrem-via-safe-divisor=0 -S", "opt < %s -passes=loop-vectorize,instcombine -force-vector-interleave=1 -force-vector-width=2 -force-widen-divrem-via-safe-divisor=0 -S", "opt < %s -passes=loop-vectorize,instcombine -force-vector-interleave=2 -force-vector-width=2 -force-widen-divrem-via-safe-divisor=0 -S", "opt < %s -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=2 -force-widen-divrem-via-safe-divisor=0 -S", "opt < %s -passes=loop-vectorize,instcombine -force-vector-interleave=2 -force-vector-width=4 -force-widen-divrem-via-safe-divisor=0 -enable-interleaved-mem-accesses -S"], "tests": [{"test_name": "max_i32_backedgetaken", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\n\ndefine i32 @max_i32_backedgetaken() {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %a.0 = phi i32 [ 1, %entry ], [ %a.0.and, %loop ]\n %b.0 = phi i32 [ 0, %entry ], [ %b.next, %loop ]\n %a.0.and = and i32 %a.0, 4\n %b.next = add i32 %b.0, -1\n %ec = icmp eq i32 %b.next, 0\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i32 %a.0.and\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/interleave-and-scalarize-only.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=1 -force-vector-interleave=2 -debug -disable-output %s 2>&1", "opt -passes=loop-vectorize -force-vector-width=1 -force-vector-interleave=2 -S %s"], "tests": [{"test_name": "scalarize_ptrtoint", "test_body": "define void @scalarize_ptrtoint(ptr %src, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr ptr, ptr %src, i64 %iv\n %l = load ptr, ptr %gep, align 8\n %cast = ptrtoint ptr %l to i64\n %add = add i64 %cast, 10\n %cast.2 = inttoptr i64 %add to ptr\n store ptr %cast.2, ptr %dst, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 1024\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/is_fpclass.ll", "commands": ["opt < %s -passes=loop-vectorize -force-vector-width=2 -S"], "tests": [{"test_name": "d", "test_body": "define void @d() {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %i = phi i64 [ 0, %entry ], [ %i7, %loop ]\n %i3 = load float, ptr null, align 4\n %i4 = getelementptr float, ptr @d, i64 %i\n %i5 = tail call i1 @llvm.is.fpclass.f32(float %i3, i32 0)\n %i6 = select i1 %i5, float 0.000000e+00, float 0.000000e+00\n store float %i6, ptr %i4, align 4\n %i7 = add i64 %i, 1\n %i8 = icmp eq i64 %i7, 128\n br i1 %i8, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i1 @llvm.is.fpclass.f32(float, i32 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization.ll", "commands": ["opt -passes='loop-vectorize' -force-vector-width=4 -enable-epilogue-vectorization -epilogue-vectorization-force-VF=4 -S %s", "opt -passes='loop-vectorize' -enable-epilogue-vectorization -epilogue-vectorization-force-VF=2 -S %s"], "tests": [{"test_name": "multiple_ivs_wide", "test_body": "target datalayout = \"e-m:e-i64:64-n32:64-v128:128:128\"\n\ndefine void @multiple_ivs_wide(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.2 = phi i32 [ -64, %entry ], [ %iv.2.next, %loop ]\n %iv.next = add i32 %iv, 2\n %iv.2.next = add i32 %iv.2, 2\n %gep = getelementptr inbounds i32, ptr %dst, i32 %iv\n store i32 %iv.2.next, ptr %gep, align 4\n %ec = icmp eq i32 %iv.next, 128\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/runtime-checks-hoist.ll", "commands": ["opt < %s -passes='loop-vectorize' -force-vector-interleave=1 -S -force-vector-width=4 -debug-only=loop-accesses,loop-vectorize,loop-utils"], "tests": [{"test_name": "stride_check_known_via_loop_guard", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\n\ndefine void @stride_check_known_via_loop_guard(ptr %C, ptr %A, i32 %Acols) {\nentry:\n %pre.c = icmp ugt i32 %Acols, 0\n br i1 %pre.c, label %exit, label %outer.header\n\nouter.header: ; preds = %outer.latch, %entry\n %outer.iv = phi i32 [ 0, %entry ], [ %outer.iv.next, %outer.latch ]\n %mul.us = mul i32 %outer.iv, %Acols\n %arrayidx.us = getelementptr double, ptr %A, i32 %mul.us\n br label %inner\n\ninner: ; preds = %inner, %outer.header\n %inner.iv = phi i32 [ 0, %outer.header ], [ %inner.iv.next, %inner ]\n %gep.C = getelementptr inbounds double, ptr %C, i32 %inner.iv\n %l = load double, ptr %arrayidx.us, align 8\n store double %l, ptr %gep.C, align 8\n %inner.iv.next = add i32 %inner.iv, 1\n %inner.c = icmp eq i32 %inner.iv.next, 1000\n br i1 %inner.c, label %outer.latch, label %inner\n\nouter.latch: ; preds = %inner\n %outer.iv.next = add i32 %outer.iv, 1\n %outer.c = icmp ult i32 %outer.iv, 128\n br i1 %outer.c, label %exit, label %outer.header\n\nexit: ; preds = %outer.latch, %entry\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -lv-strided-pointer-ivs=true -S %s"], "tests": [{"test_name": "step_direction_unknown", "test_body": "define void @step_direction_unknown(i32 %arg, ptr %dst) {\nentry:\n %add = add i32 %arg, 1\n br label %loop\n\nloop: ; preds = %loop, %entry\n %phi = phi i64 [ 0, %entry ], [ %add2, %loop ]\n %trunc = trunc i64 %phi to i32\n %mul = mul i32 %add, %trunc\n %zext = zext i32 %mul to i64\n %getelementptr = getelementptr double, ptr %dst, i64 %zext\n store double 0.000000e+00, ptr %getelementptr, align 8\n %add2 = add i64 %phi, 1\n %icmp = icmp eq i64 %add2, 1024\n br i1 %icmp, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "integer_induction_wraps_scev_predicate_known", "test_body": "define void @integer_induction_wraps_scev_predicate_known(i32 %x, ptr %call, ptr %start) {\nentry:\n %mul = shl i32 %x, 1\n br label %for.cond\n\nfor.cond: ; preds = %for.cond, %entry\n %iv = phi i32 [ 30, %entry ], [ %inc, %for.cond ]\n %p.0 = phi ptr [ %start, %entry ], [ %add.ptr, %for.cond ]\n %add.ptr = getelementptr i8, ptr %p.0, i32 %mul\n %arrayidx = getelementptr ptr, ptr %call, i32 %iv\n store ptr %p.0, ptr %arrayidx, align 4\n %inc = add i32 %iv, 1\n %tobool.not = icmp eq i32 %iv, 1024\n br i1 %tobool.not, label %for.end, label %for.cond\n\nfor.end: ; preds = %for.cond\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge-vf1.ll", "commands": ["opt -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=1 -debug -disable-output %s 2>&1"], "tests": [{"test_name": "sink_with_sideeffects", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\n\ndefine void @sink_with_sideeffects(i1 %c, ptr %ptr) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.inc, %entry\n %tmp0 = phi i64 [ %tmp6, %for.inc ], [ 0, %entry ]\n %tmp1 = phi i64 [ %tmp7, %for.inc ], [ 1024, %entry ]\n %tmp2 = getelementptr i8, ptr %ptr, i64 %tmp0\n %tmp3 = load i8, ptr %tmp2, align 1\n store i8 0, ptr %tmp2, align 1\n %tmp4 = zext i8 %tmp3 to i32\n %tmp5 = trunc i32 %tmp4 to i8\n br i1 %c, label %if.then, label %for.inc\n\nif.then: ; preds = %for.body\n store i8 %tmp5, ptr %tmp2, align 1\n br label %for.inc\n\nfor.inc: ; preds = %if.then, %for.body\n %tmp6 = add nuw nsw i64 %tmp0, 1\n %tmp7 = add i64 %tmp1, -1\n %tmp8 = icmp eq i64 %tmp7, 0\n br i1 %tmp8, label %for.end, label %for.body\n\nfor.end: ; preds = %for.inc\n ret void\n}\n"}]}], "issue": {"title": "[LoopVectorizer] Assertion `EPResumeVal && \"must have a resume value for the canonical IV\"' failed.", "body": "clang -Wno-incompatible-pointer-types -O3 -march=z13 -S -c crash19.i -o a.out -w -mllvm -disable-licm-promotion -mllvm -epilogue-vectorization-force-VF=2 \n\n[crash19.tar.gz](https://github.com/user-attachments/files/18383534/crash19.tar.gz)\n\n#9 0x000002aa3ff5e440 preparePlanForEpilogueVectorLoop\n#12 0x000002aa3ffa098a llvm::LoopVectorizePass::run\n\n@bmahjour \n@fhahn ", "author": "JonPsson1", "labels": ["vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "@JonPsson1 Was the test case fuzzer generated? Just curious, because it had a loop with unsigned max as iteration count."}, {"author": "JonPsson1", "body": "> [@JonPsson1](https://github.com/JonPsson1) Was the test case fuzzer generated? Just curious, because it had a loop with unsigned max as iteration count.\n\nNo, it was a CSmith test case."}, {"author": "fhahn", "body": "@JonPsson1 Could you check if it fixes the full reproducer? If not, it would be great if you could share the full reproducer?"}]}, "verified": true} {"bug_id": "108700", "issue_url": "https://github.com/llvm/llvm-project/issues/108700", "bug_type": "crash", "base_commit": "4c6f313cb340c435f47ac032857030768c81c341", "knowledge_cutoff": "2024-09-14T13:45:54Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "f564a48f0ea4d2100c0cadfa6e6f20f97244025e", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[7481, 7487], [7492, 7506]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::buildTree_rec"]}}, "patch": "commit f564a48f0ea4d2100c0cadfa6e6f20f97244025e\nAuthor: Alexey Bataev \nDate: Mon Sep 16 09:19:42 2024 -0700\n\n [SLP]Fix PR108700: correctly identify id of the operand node\n \n If the operand node for truncs is not created during construction, but\n one of the previous ones is reused instead, need to correctly identify\n its index, to correctly emit the code.\n \n Fixes https://github.com/llvm/llvm-project/issues/108700\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 5f2bf082fb87..282bb8eac7e2 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -7481,7 +7481,16 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n PrevMaxBW),\n std::min(DL->getTypeSizeInBits(VL0->getType()),\n PrevMinBW));\n- ExtraBitWidthNodes.insert(VectorizableTree.size() + 1);\n+ }\n+ TreeEntry *TE = newTreeEntry(VL, Bundle /*vectorized*/, S, UserTreeIdx,\n+ ReuseShuffleIndices);\n+ LLVM_DEBUG(dbgs() << \"SLP: added a vector of casts.\\n\");\n+\n+ TE->setOperandsInOrder();\n+ for (unsigned I : seq(0, VL0->getNumOperands()))\n+ buildTree_rec(TE->getOperand(I), Depth + 1, {TE, I});\n+ if (ShuffleOrOp == Instruction::Trunc) {\n+ ExtraBitWidthNodes.insert(getOperandEntry(TE, 0)->Idx);\n } else if (ShuffleOrOp == Instruction::SIToFP ||\n ShuffleOrOp == Instruction::UIToFP) {\n unsigned NumSignBits =\n@@ -7492,15 +7501,8 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n }\n if (NumSignBits * 2 >=\n DL->getTypeSizeInBits(VL0->getOperand(0)->getType()))\n- ExtraBitWidthNodes.insert(VectorizableTree.size() + 1);\n+ ExtraBitWidthNodes.insert(getOperandEntry(TE, 0)->Idx);\n }\n- TreeEntry *TE = newTreeEntry(VL, Bundle /*vectorized*/, S, UserTreeIdx,\n- ReuseShuffleIndices);\n- LLVM_DEBUG(dbgs() << \"SLP: added a vector of casts.\\n\");\n-\n- TE->setOperandsInOrder();\n- for (unsigned I : seq(0, VL0->getNumOperands()))\n- buildTree_rec(TE->getOperand(I), Depth + 1, {TE, I});\n return;\n }\n case Instruction::ICmp:\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/minbitwidth-user-not-min.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define void @test(ptr %block, ptr noalias %pixels, i1 %b) {\nentry:\n %0 = load i16, ptr %block, align 2\n %tobool.not.i78 = icmp ult i16 %0, 0\n %conv.i80 = sext i1 true to i8\n %conv1.i81 = trunc i16 %0 to i8\n %retval.0.i82 = select i1 %tobool.not.i78, i8 %conv1.i81, i8 %conv.i80\n store i8 %retval.0.i82, ptr %pixels, align 1\n %arrayidx2 = getelementptr i8, ptr %block, i64 2\n %1 = load i16, ptr %arrayidx2, align 2\n %tobool.not.i73 = icmp ult i16 %1, 0\n %conv.i75 = sext i1 %b to i8\n %conv1.i76 = trunc i16 %1 to i8\n %retval.0.i77 = select i1 %tobool.not.i73, i8 %conv1.i76, i8 %conv.i75\n %arrayidx5 = getelementptr i8, ptr %pixels, i64 1\n store i8 %retval.0.i77, ptr %arrayidx5, align 1\n %arrayidx6 = getelementptr i8, ptr %block, i64 4\n %2 = load i16, ptr %arrayidx6, align 2\n %tobool.not.i68 = icmp ult i16 %2, 0\n %conv.i70 = sext i1 false to i8\n %conv1.i71 = trunc i16 %2 to i8\n %retval.0.i72 = select i1 %tobool.not.i68, i8 %conv1.i71, i8 %conv.i70\n %arrayidx9 = getelementptr i8, ptr %pixels, i64 2\n store i8 %retval.0.i72, ptr %arrayidx9, align 1\n %arrayidx10 = getelementptr i8, ptr %block, i64 6\n %3 = load i16, ptr %arrayidx10, align 2\n %tobool.not.i63 = icmp ult i16 %3, 0\n %conv.i65 = sext i1 false to i8\n %conv1.i66 = trunc i16 %3 to i8\n %retval.0.i67 = select i1 %tobool.not.i63, i8 %conv1.i66, i8 %conv.i65\n %arrayidx13 = getelementptr i8, ptr %pixels, i64 3\n store i8 %retval.0.i67, ptr %arrayidx13, align 1\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/SLPVectorizer/X86/trunc-node-reused.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define i16 @test() {\nentry:\n %conv73 = xor i64 0, 0\n %and.i = and i64 0, 0\n %xor2.i = or i64 %and.i, 0\n %sub.i = or i64 %xor2.i, 0\n %xor3.i = xor i64 %sub.i, %conv73\n %and4.i = and i64 %xor3.i, 0\n %cmp.i = icmp slt i64 %and4.i, 0\n %0 = trunc i64 %conv73 to i16\n %1 = or i16 0, %0\n %conv73i = xor i64 0, 0\n %andi.i = and i64 0, 0\n %xor2i.i = or i64 %andi.i, 0\n %subi.i = or i64 %xor2i.i, 0\n %xor3i.i = xor i64 %subi.i, %conv73i\n %and4i.i = and i64 %xor3i.i, 0\n %cmpi.i = icmp slt i64 %and4i.i, 0\n %2 = trunc i64 %conv73i to i16\n %3 = or i16 0, %2\n %4 = select i1 %cmpi.i, i16 0, i16 %3\n %5 = select i1 %cmp.i, i16 0, i16 %1\n %6 = zext i32 0 to i64\n %add.ip = or i64 %6, 0\n %orp = or i64 %add.ip, 0\n %conv72p = shl i64 %orp, 0\n %sextp = ashr i64 %conv72p, 0\n %conv73p = xor i64 %sextp, 0\n %and.ip = and i64 0, 0\n %xor2.ip = or i64 %and.ip, 0\n %sub.ip = or i64 %xor2.ip, 0\n %xor3.ip = xor i64 %sub.ip, %conv73p\n %and4.ip = and i64 %xor3.ip, 0\n %cmp.ip = icmp slt i64 %and4.ip, 0\n %7 = trunc i64 %conv73p to i16\n %8 = or i16 0, %7\n %9 = select i1 %cmp.ip, i16 0, i16 %8\n %conv76i = and i16 %4, %5\n %conv76p = and i16 %conv76i, %9\n %10 = zext i32 0 to i64\n %add.ip1 = or i64 %10, 0\n %orp1 = or i64 %add.ip1, 0\n %conv72p1 = shl i64 %orp1, 0\n %sextp1 = ashr i64 %conv72p1, 0\n %conv73p1 = xor i64 %sextp1, 0\n %and.ip1 = and i64 0, 0\n %xor2.ip1 = or i64 %and.ip1, 0\n %sub.ip1 = or i64 %xor2.ip1, 0\n %xor3.ip1 = xor i64 %sub.ip1, %conv73p1\n %and4.ip1 = and i64 %xor3.ip1, 0\n %cmp.ip1 = icmp slt i64 %and4.ip1, 0\n %11 = trunc i64 %conv73p1 to i16\n %12 = or i16 0, %11\n %13 = select i1 %cmp.ip1, i16 0, i16 %12\n %conv76p2 = and i16 %conv76p, %13\n ret i16 %conv76p2\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Assertion `OpIdx < Operands.size() && \"Off bounds\"' failed.", "body": "Reproducer: https://godbolt.org/z/8Es17cqcW\r\n```\r\n; bin/opt -passes=slp-vectorizer reduced.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i16 @main() {\r\nentry:\r\n %conv73.i.peel.i.i.i.i.i = xor i64 0, 0\r\n %and.i.i.peel.i.i.i.i.i = and i64 0, 0\r\n %xor2.i.i.peel.i.i.i.i.i = or i64 %and.i.i.peel.i.i.i.i.i, 0\r\n %sub.i.i.peel.i.i.i.i.i = or i64 %xor2.i.i.peel.i.i.i.i.i, 0\r\n %xor3.i.i.peel.i.i.i.i.i = xor i64 %sub.i.i.peel.i.i.i.i.i, %conv73.i.peel.i.i.i.i.i\r\n %and4.i.i.peel.i.i.i.i.i = and i64 %xor3.i.i.peel.i.i.i.i.i, 0\r\n %cmp.i.i.peel.i.i.i.i.i = icmp slt i64 %and4.i.i.peel.i.i.i.i.i, 0\r\n %0 = trunc i64 %conv73.i.peel.i.i.i.i.i to i16\r\n %1 = or i16 0, %0\r\n %conv73.i.i.i.i.i.i = xor i64 0, 0\r\n %and.i.i.i.i.i.i.i = and i64 0, 0\r\n %xor2.i.i.i.i.i.i.i = or i64 %and.i.i.i.i.i.i.i, 0\r\n %sub.i.i.i.i.i.i.i = or i64 %xor2.i.i.i.i.i.i.i, 0\r\n %xor3.i.i.i.i.i.i.i = xor i64 %sub.i.i.i.i.i.i.i, %conv73.i.i.i.i.i.i\r\n %and4.i.i.i.i.i.i.i = and i64 %xor3.i.i.i.i.i.i.i, 0\r\n %cmp.i.i.i.i.i.i.i = icmp slt i64 %and4.i.i.i.i.i.i.i, 0\r\n %2 = trunc i64 %conv73.i.i.i.i.i.i to i16\r\n %3 = or i16 0, %2\r\n %4 = select i1 %cmp.i.i.i.i.i.i.i, i16 0, i16 %3\r\n %5 = select i1 %cmp.i.i.peel.i.i.i.i.i, i16 0, i16 %1\r\n %6 = zext i32 0 to i64\r\n %add.i.i.peel.i369.i.i.i.i = or i64 %6, 0\r\n %or.i.peel.i371.i.i.i.i = or i64 %add.i.i.peel.i369.i.i.i.i, 0\r\n %conv72.i.peel.i372.i.i.i.i = shl i64 %or.i.peel.i371.i.i.i.i, 0\r\n %sext.i.peel.i373.i.i.i.i = ashr i64 %conv72.i.peel.i372.i.i.i.i, 0\r\n %conv73.i.peel.i374.i.i.i.i = xor i64 %sext.i.peel.i373.i.i.i.i, 0\r\n %and.i.i.peel.i376.i.i.i.i = and i64 0, 0\r\n %xor2.i.i.peel.i377.i.i.i.i = or i64 %and.i.i.peel.i376.i.i.i.i, 0\r\n %sub.i.i.peel.i378.i.i.i.i = or i64 %xor2.i.i.peel.i377.i.i.i.i, 0\r\n %xor3.i.i.peel.i379.i.i.i.i = xor i64 %sub.i.i.peel.i378.i.i.i.i, %conv73.i.peel.i374.i.i.i.i\r\n %and4.i.i.peel.i380.i.i.i.i = and i64 %xor3.i.i.peel.i379.i.i.i.i, 0\r\n %cmp.i.i.peel.i381.i.i.i.i = icmp slt i64 %and4.i.i.peel.i380.i.i.i.i, 0\r\n %7 = trunc i64 %conv73.i.peel.i374.i.i.i.i to i16\r\n %8 = or i16 0, %7\r\n %9 = select i1 %cmp.i.i.peel.i381.i.i.i.i, i16 0, i16 %8\r\n %conv76.i.i.i.i.i.i = and i16 %4, %5\r\n %conv76.i.peel.i384.i.i.i.i = and i16 %conv76.i.i.i.i.i.i, %9\r\n %10 = zext i32 0 to i64\r\n %add.i.i.peel.i417.i.i.i.i = or i64 %10, 0\r\n %or.i.peel.i419.i.i.i.i = or i64 %add.i.i.peel.i417.i.i.i.i, 0\r\n %conv72.i.peel.i420.i.i.i.i = shl i64 %or.i.peel.i419.i.i.i.i, 0\r\n %sext.i.peel.i421.i.i.i.i = ashr i64 %conv72.i.peel.i420.i.i.i.i, 0\r\n %conv73.i.peel.i422.i.i.i.i = xor i64 %sext.i.peel.i421.i.i.i.i, 0\r\n %and.i.i.peel.i424.i.i.i.i = and i64 0, 0\r\n %xor2.i.i.peel.i425.i.i.i.i = or i64 %and.i.i.peel.i424.i.i.i.i, 0\r\n %sub.i.i.peel.i426.i.i.i.i = or i64 %xor2.i.i.peel.i425.i.i.i.i, 0\r\n %xor3.i.i.peel.i427.i.i.i.i = xor i64 %sub.i.i.peel.i426.i.i.i.i, %conv73.i.peel.i422.i.i.i.i\r\n %and4.i.i.peel.i428.i.i.i.i = and i64 %xor3.i.i.peel.i427.i.i.i.i, 0\r\n %cmp.i.i.peel.i429.i.i.i.i = icmp slt i64 %and4.i.i.peel.i428.i.i.i.i, 0\r\n %11 = trunc i64 %conv73.i.peel.i422.i.i.i.i to i16\r\n %12 = or i16 0, %11\r\n %13 = select i1 %cmp.i.i.peel.i429.i.i.i.i, i16 0, i16 %12\r\n %conv76.i.peel.i432.i.i.i.i = and i16 %conv76.i.peel.i384.i.i.i.i, %13\r\n ret i16 %conv76.i.peel.i432.i.i.i.i\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:3180: llvm::slpvectorizer::BoUpSLP::ValueList& llvm::slpvectorizer::BoUpSLP::TreeEntry::getOperand(unsigned int): Assertion `OpIdx < Operands.size() && \"Off bounds\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S --passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"main\"\r\n #0 0x0000000005016bb8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x5016bb8)\r\n #1 0x00000000050142fc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007955e2242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007955e22969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007955e2242476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007955e22287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007955e222871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007955e2239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x000000000406f2b8 llvm::slpvectorizer::BoUpSLP::TreeEntry::getOperand(unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x406f2b8)\r\n #9 0x00000000040f675e llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&)::'lambda0'(llvm::function_ref, llvm::function_ref)::operator()(llvm::function_ref, llvm::function_ref) const SLPVectorizer.cpp:0:0\r\n#10 0x00000000040f74aa llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40f74aa)\r\n#11 0x0000000004114641 llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4114641)\r\n#12 0x00000000041198c8 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#13 0x000000000411bfe1 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x411bfe1)\r\n#14 0x0000000004120876 llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*) (.constprop.0) SLPVectorizer.cpp:0:0\r\n#15 0x0000000004124e16 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4124e16)\r\n#16 0x000000000412b2e1 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#17 0x000000000412bf6b llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x412bf6b)\r\n#18 0x0000000002f0034e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f0034e)\r\n#19 0x0000000004e18ff8 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e18ff8)\r\n#20 0x0000000000e06a0e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe06a0e)\r\n#21 0x0000000004e17a1e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e17a1e)\r\n#22 0x0000000000e0621e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe0621e)\r\n#23 0x0000000004e17450 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e17450)\r\n#24 0x00000000008fce52 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fce52)\r\n#25 0x00000000008efcdc optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8efcdc)\r\n#26 0x00007955e2229d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#27 0x00007955e2229e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#28 0x00000000008e77ae _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e77ae)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "121432", "issue_url": "https://github.com/llvm/llvm-project/issues/121432", "bug_type": "miscompilation", "base_commit": "b51a082e1afd707f8cf1473e4c49374e7e951d2d", "knowledge_cutoff": "2025-01-01T11:07:27Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "c05599966cd594c15834378f57669719d221c4d4", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp": [[2673, 2680]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp": ["InstCombinerImpl::visitCallInst"]}}, "patch": "commit c05599966cd594c15834378f57669719d221c4d4\nAuthor: Yingwei Zheng \nDate: Mon Jan 6 16:23:46 2025 +0800\n\n [InstCombine] Fix FMF propagation in `copysign Mag, (copysign ?, X) -> copysign Mag, X` (#121574)\n \n Closes https://github.com/llvm/llvm-project/issues/121432.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\nindex cdb2c11ef046..5494c70b34b1 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\n@@ -2673,8 +2673,12 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {\n // Propagate sign argument through nested calls:\n // copysign Mag, (copysign ?, X) --> copysign Mag, X\n Value *X;\n- if (match(Sign, m_Intrinsic(m_Value(), m_Value(X))))\n- return replaceOperand(*II, 1, X);\n+ if (match(Sign, m_Intrinsic(m_Value(), m_Value(X)))) {\n+ Value *CopySign = Builder.CreateCopySign(\n+ Mag, X,\n+ II->getFastMathFlags() & cast(Sign)->getFastMathFlags());\n+ return replaceInstUsesWith(*II, CopySign);\n+ }\n \n // Clear sign-bit of constant magnitude:\n // copysign -MagC, X --> copysign MagC, X\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/copysign.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "copysign_sign_arg_mixed", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.copysign.f32(float, float) #0\n\ndefine float @copysign_sign_arg_mixed(float %x, float %y, float %z) {\n %s = call ninf nsz float @llvm.copysign.f32(float %y, float %z)\n %r = call nnan nsz float @llvm.copysign.f32(float %x, float %s)\n ret float %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "copysign_sign_arg", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.copysign.f32(float, float) #0\n\ndefine float @copysign_sign_arg(float %x, float %y, float %z) {\n %s = call float @llvm.copysign.f32(float %y, float %z)\n %r = call ninf float @llvm.copysign.f32(float %x, float %s)\n ret float %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "fneg_mag", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.copysign.f32(float, float) #0\n\ndefine float @fneg_mag(float %x, float %y) {\n %n = fneg float %x\n %r = call float @llvm.copysign.f32(float %n, float %y)\n ret float %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "copysign_sign_arg_nnan", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.copysign.f32(float, float) #0\n\ndefine float @copysign_sign_arg_nnan(float %x, float %y, float %z) {\n %s = call nnan float @llvm.copysign.f32(float %y, float %z)\n %r = call nnan float @llvm.copysign.f32(float %x, float %s)\n ret float %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "not_known_positive_sign_arg", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.copysign.f32(float, float) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.maxnum.f32(float, float) #0\n\ndefine float @not_known_positive_sign_arg(float %x, float %y) {\n %max = call float @llvm.maxnum.f32(float %x, float 0.000000e+00)\n %r = call ninf float @llvm.copysign.f32(float %y, float %max)\n ret float %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[InstCombine] FMF flags should not be propagated", "body": "Alive2: https://alive2.llvm.org/ce/z/ajCxTh\n```\ndefine float @src1(float noundef %x, float noundef %y, float noundef %z) {\n %s = call float @llvm.copysign.f32(float %y, float %z)\n %r = call nnan float @llvm.copysign.f32(float %x, float %s)\n ret float %r\n}\n\ndefine float @tgt1(float noundef %x, float noundef %y, float noundef %z) {\n %r = call nnan float @llvm.copysign.f32(float %x, float %z)\n ret float %r\n}\n\ndefine float @src2(float noundef %x, float noundef %y, float noundef %z) {\n %s = call float @llvm.copysign.f32(float %y, float %z)\n %r = call ninf float @llvm.copysign.f32(float %x, float %s)\n ret float %r\n}\n\ndefine float @tgt2(float noundef %x, float noundef %y, float noundef %z) {\n %r = call ninf float @llvm.copysign.f32(float %x, float %z)\n ret float %r\n}\n\ndefine float @src3(float noundef %x, float noundef %y, float noundef %z) {\n %s = call float @llvm.copysign.f32(float %y, float %z)\n %r = call nsz float @llvm.copysign.f32(float %x, float %s)\n ret float %r\n}\n\ndefine float @tgt3(float noundef %x, float noundef %y, float noundef %z) {\n %r = call nsz float @llvm.copysign.f32(float %x, float %z)\n ret float %r\n}\n```\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine", "floating-point"], "comments": []}, "verified": true} {"bug_id": "113986", "issue_url": "https://github.com/llvm/llvm-project/issues/113986", "bug_type": "crash", "base_commit": "0c1c37bfbed08c9d4e414a10f46cbed9a3e4c870", "knowledge_cutoff": "2024-10-29T00:38:49Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "18311093abe6481388a0d963d58438d743b47569", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp": [[2902, 2909]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp": ["InstCombinerImpl::visitShuffleVectorInst"]}}, "patch": "commit 18311093abe6481388a0d963d58438d743b47569\nAuthor: Yingwei Zheng \nDate: Tue Oct 29 10:39:07 2024 +0800\n\n [InstCombine] Do not fold `shufflevector(select)` if the select condition is a vector (#113993)\n \n Since `shufflevector` is not element-wise, we cannot do fold it into\n select when the select condition is a vector.\n For shufflevector that doesn't change the length, it doesn't crash, but\n it is still a miscompilation: https://alive2.llvm.org/ce/z/s8saCx\n \n Fixes https://github.com/llvm/llvm-project/issues/113986.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\nindex 75e7c1c97018..454fe5a91d37 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\n@@ -2902,8 +2902,12 @@ Instruction *InstCombinerImpl::visitShuffleVectorInst(ShuffleVectorInst &SVI) {\n \n if (match(RHS, m_Constant())) {\n if (auto *SI = dyn_cast(LHS)) {\n- if (Instruction *I = FoldOpIntoSelect(SVI, SI))\n- return I;\n+ // We cannot do this fold for elementwise select since ShuffleVector is\n+ // not elementwise.\n+ if (SI->getCondition()->getType()->isIntegerTy()) {\n+ if (Instruction *I = FoldOpIntoSelect(SVI, SI))\n+ return I;\n+ }\n }\n if (auto *PN = dyn_cast(LHS)) {\n if (Instruction *I = foldOpIntoPhi(SVI, PN))\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/vec_shuffle.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "foldselect0", "test_body": "define <2 x i32> @foldselect0(i1 %c) {\n %sel = select i1 %c, <2 x i32> , <2 x i32> \n %shuf = shufflevector <2 x i32> %sel, <2 x i32> poison, <2 x i32> \n ret <2 x i32> %shuf\n}\n"}, {"test_name": "shuf_larger_length_vec_select", "test_body": "define <4 x float> @shuf_larger_length_vec_select(<2 x i1> %cond) {\n %sel = select <2 x i1> %cond, <2 x float> zeroinitializer, <2 x float> splat (float 1.000000e+00)\n %shuf = shufflevector <2 x float> %sel, <2 x float> zeroinitializer, <4 x i32> \n ret <4 x float> %shuf\n}\n"}, {"test_name": "shuf_same_length_vec_select", "test_body": "define <4 x i32> @shuf_same_length_vec_select(<4 x i1> %cond) {\n %sel = select <4 x i1> %cond, <4 x i32> , <4 x i32> \n %shuf = shufflevector <4 x i32> %sel, <4 x i32> , <4 x i32> \n ret <4 x i32> %shuf\n}\n"}]}], "issue": {"title": "[InstCombine] Assertion `!areInvalidOperands(C, S1, S2) && \"Invalid operands for select\"' failed.", "body": "Reproducer: \r\n```\r\n; bin/opt -passes=instcombine reduced.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-pc-linux-gnu\"\r\n\r\ndefine <32 x float> @process(<8 x i1> %0) {\r\n %3 = select <8 x i1> %0, <8 x float> zeroinitializer, <8 x float> \r\n %4 = shufflevector <8 x float> %3, <8 x float> zeroinitializer, <16 x i32> \r\n %5 = shufflevector <16 x float> %4, <16 x float> zeroinitializer, <32 x i32> \r\n ret <32 x float> %5\r\n}\r\n```\r\n```\r\nopt: /home/dtcxzyw/WorkSpace/Projects/compilers/llvm-project/llvm/include/llvm/IR/Instructions.h:1614: void llvm::SelectInst::init(llvm::Value*, llvm::Value*, llvm::Value*): Assertion `!areInvalidOperands(C, S1, S2) && \"Invalid operands for select\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: bin/opt -passes=instcombine reduced.ll\r\n1. Running pass \"function(instcombine)\" on module \"reduced.ll\"\r\n2. Running pass \"instcombine\" on function \"process\"\r\n #0 0x00007ffff7e15b02 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.20.0git+0x215b02)\r\n #1 0x00007ffff7e129af llvm::sys::RunSignalHandlers() (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.20.0git+0x2129af)\r\n #2 0x00007ffff7e12af5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007ffff7842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007ffff78969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007ffff78969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007ffff78969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007ffff7842476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007ffff78287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007ffff782871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007ffff7839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00007ffff1272963 llvm::InstCombinerImpl::FoldOpIntoSelect(llvm::Instruction&, llvm::SelectInst*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMInstCombine.so.20.0git+0x60963)\r\n#12 0x00007ffff13a09d1 llvm::InstCombinerImpl::visitShuffleVectorInst(llvm::ShuffleVectorInst&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMInstCombine.so.20.0git+0x18e9d1)\r\n#13 0x00007ffff1275da8 llvm::InstCombinerImpl::run() (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMInstCombine.so.20.0git+0x63da8)\r\n#14 0x00007ffff1276f09 combineInstructionsOverFunction(llvm::Function&, llvm::InstructionWorklist&, llvm::AAResults*, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::BranchProbabilityInfo*, llvm::ProfileSummaryInfo*, llvm::InstCombineOptions const&) InstructionCombining.cpp:0:0\r\n#15 0x00007ffff1277dfb llvm::InstCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMInstCombine.so.20.0git+0x65dfb)\r\n#16 0x00007ffff30d9955 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMPasses.so.20.0git+0xd9955)\r\n#17 0x00007ffff01678bd llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x3678bd)\r\n#18 0x00007ffff68d9e65 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.20.0git+0xd9e65)\r\n#19 0x00007ffff01659f6 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x3659f6)\r\n#20 0x00007ffff68da825 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.20.0git+0xda825)\r\n#21 0x00007ffff016657a llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x36657a)\r\n#22 0x00007ffff7f95051 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.20.0git+0x2d051)\r\n#23 0x00007ffff7fa073f optMain (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.20.0git+0x3873f)\r\n#24 0x00007ffff7829d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#25 0x00007ffff7829e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#26 0x00007ffff7829e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#27 0x0000555555555095 _start (bin/opt+0x1095)\r\nAborted (core dumped)\r\n```\r\n\r\nIt is caused by https://github.com/llvm/llvm-project/pull/113746. We cannot do this fold if the output and input operands don't have the exact same type. It was reported by llvm-opt-benchmark pre-commit test https://github.com/dtcxzyw/llvm-opt-benchmark/pull/1543. I am sorry I didn't notice about this. Fortunately this issue was captured by a post-commit test https://github.com/dtcxzyw/llvm-opt-benchmark/issues/1560.\r\n\r\ncc @MatzeB @RKSimon @goldsteinn \r\n\r\nI will post a fix later.\r\n\r\n\r\n", "author": "dtcxzyw", "labels": ["llvm:instcombine", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "93152", "issue_url": "https://github.com/llvm/llvm-project/issues/93152", "bug_type": "crash", "base_commit": "d3ce1078186389ce39505f06c2a0100dce9187a5", "knowledge_cutoff": "2024-05-23T08:42:34Z", "lit_test_dir": ["llvm/test/Transforms/LowerSwitch"], "hints": {"fix_commit": "67aec0cd62d607b4e5b7198769be061454ce67b3", "components": ["LowerSwitch"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/LowerSwitch.cpp": [[208, 214]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/LowerSwitch.cpp": ["NewLeafBlock"]}}, "patch": "commit 67aec0cd62d607b4e5b7198769be061454ce67b3\nAuthor: Peter Rong \nDate: Mon May 27 10:58:22 2024 -0700\n\n [LowerSwitch] Use unsigned integer for range comparison (#93237)\n \n Commit 1db51d8eb2 switched from int64_t to `APInt` to prevent high precision integer overflow.\n However, when comparing the \"range\" of switch cases, we should switch to unsigned integer to prevent overflow.\n This patch fixes https://github.com/llvm/llvm-project/issues/93152.\n Some test cases are added.\n \n Signed-off-by: Peter Rong \n\ndiff --git a/llvm/lib/Transforms/Utils/LowerSwitch.cpp b/llvm/lib/Transforms/Utils/LowerSwitch.cpp\nindex f5921e5ccb09..f4ef6d02abf0 100644\n--- a/llvm/lib/Transforms/Utils/LowerSwitch.cpp\n+++ b/llvm/lib/Transforms/Utils/LowerSwitch.cpp\n@@ -208,7 +208,7 @@ BasicBlock *NewLeafBlock(CaseRange &Leaf, Value *Val, ConstantInt *LowerBound,\n PHINode *PN = cast(I);\n // Remove all but one incoming entries from the cluster\n APInt Range = Leaf.High->getValue() - Leaf.Low->getValue();\n- for (APInt j(Range.getBitWidth(), 0, true); j.slt(Range); ++j) {\n+ for (APInt j(Range.getBitWidth(), 0, false); j.ult(Range); ++j) {\n PN->removeIncomingValue(OrigBlock);\n }\n \n", "tests": [{"file": "llvm/test/Transforms/LowerSwitch/93152.ll", "commands": ["opt < %s -passes=lower-switch -S"], "tests": [{"test_name": "", "test_body": "define void @i3_range_4(i3 %0) {\n;\nbb.0:\n switch i3 %0, label %bb.2 [\n i3 -1, label %bb.1\n i3 -2, label %bb.1\n i3 2, label %bb.1\n i3 1, label %bb.1\n i3 0, label %bb.1\n ]\n\nbb.1: ; preds = %bb.0, %bb.0, %bb.0, %bb.0, %bb.0\n %tmp = phi i3 [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ]\n br label %bb.2\n\nbb.2: ; preds = %bb.1, %bb.0\n ret void\n}\n\ndefine void @i3_range_6(i3 %0) {\n;\nbb.0:\n switch i3 %0, label %bb.2 [\n i3 -1, label %bb.1\n i3 -2, label %bb.1\n i3 -3, label %bb.1\n i3 3, label %bb.1\n i3 2, label %bb.1\n i3 1, label %bb.1\n i3 0, label %bb.1\n ]\n\nbb.1: ; preds = %bb.0, %bb.0, %bb.0, %bb.0, %bb.0\n %tmp = phi i3 [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ]\n br label %bb.2\n\nbb.2: ; preds = %bb.1, %bb.0\n ret void\n}\n\n\ndefine void @i3_range_7(i3 %0) {\n;\nbb.0:\n switch i3 %0, label %bb.2 [\n i3 -1, label %bb.1\n i3 -2, label %bb.1\n i3 -3, label %bb.1\n i3 -4, label %bb.1\n i3 3, label %bb.1\n i3 2, label %bb.1\n i3 1, label %bb.1\n i3 0, label %bb.1\n ]\n\nbb.1: ; preds = %bb.0, %bb.0, %bb.0, %bb.0, %bb.0\n %tmp = phi i3 [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ]\n br label %bb.2\n\nbb.2: ; preds = %bb.1, %bb.0\n ret void\n}"}]}], "issue": {"title": "LLVM ERROR: Broken function found after pass \"LowerSwitchPass\", compilation aborted!", "body": "llvm commit: 4cc6d0f4dfb26d\r\nReproduce with:\r\n```opt -passes=\"lower-switch\" -verify-each bbi-95731.ll -o /dev/null```\r\nResult:\r\n```\r\nPHINode should have one entry for each predecessor of its parent basic block!\r\n %tmp = phi i3 [ 0, %LeafBlock ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ]\r\nLLVM ERROR: Broken function found after pass \"LowerSwitchPass\", compilation aborted!\r\n```\r\nIf we look at what lower-switch actually produces (-print-after-all) we see\r\n```\r\n; *** IR Dump After LowerSwitchPass on v_257_0 ***\r\ndefine void @v_257_0(i3 %0) {\r\nbb.0:\r\n br label %LeafBlock\r\n\r\nLeafBlock: ; preds = %bb.0\r\n %.off = add i3 %0, 2\r\n %SwitchLeaf = icmp ule i3 %.off, -4\r\n br i1 %SwitchLeaf, label %bb.1, label %bb.2\r\n\r\nbb.1: ; preds = %LeafBlock\r\n %tmp = phi i3 [ 0, %LeafBlock ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ]\r\n br label %bb.2\r\n\r\nbb.2: ; preds = %LeafBlock, %bb.1\r\n ret void\r\n}\r\nPHINode should have one entry for each predecessor of its parent basic block!\r\n %tmp = phi i3 [ 0, %LeafBlock ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ]\r\n```\r\nSo it has removed the switch instruction but it only updated one of the five phi operands for %bb.0.\r\nThe input looked like\r\n```\r\ndefine void @v_257_0(i3 %0) {\r\nbb.0:\r\n switch i3 %0, label %bb.2 [\r\n i3 -1, label %bb.1\r\n i3 -2, label %bb.1\r\n i3 2, label %bb.1\r\n i3 1, label %bb.1\r\n i3 0, label %bb.1\r\n ]\r\n\r\nbb.1: ; preds = %bb.0, %bb.0, %bb.0, %bb.0, %bb.0\r\n %tmp = phi i3 [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ], [ 0, %bb.0 ]\r\n br label %bb.2\r\n\r\nbb.2: ; preds = %bb.1, %bb.0\r\n ret void\r\n}\r\n```\r\nSo there is one phi operand for each of the values that makes it jump from %bb.0 to %bb.1 and lower-switch only updated one of them when it changed things.\r\n[bbi-95731.ll.gz](https://github.com/llvm/llvm-project/files/15414163/bbi-95731.ll.gz)\r\n", "author": "mikaelholmen", "labels": ["crash-on-valid", "llvm:transforms"], "comments": [{"author": "mikaelholmen", "body": "This starts happening with 1db51d8eb2d\r\n```\r\n[Transform] Rewrite LowerSwitch using APInt\r\n\r\nThis rewrite fixes https://github.com/llvm/llvm-project/issues/59316.\r\n\r\nPreviously LowerSwitch uses int64_t, which will crash on case branches using integers with more than 64 bits.\r\nUsing APInt fixes this problem. This patch also includes a test\r\n\r\nReviewed By: RKSimon\r\n\r\nDifferential Revision: https://reviews.llvm.org/D140747\r\n```"}, {"author": "DataCorrupted", "body": "I am taking a look. Some initial thoughts:\r\n\r\n- https://github.com/llvm/llvm-project/commit/1db51d8eb2d220a4f0000555ada310990098cf5b didn't attempt to change how it optimizes the switch. Maybe we forgot to change the API in that optimization.\r\n- `i3` is playing a role here: if you switch it to `i32`, [it'll pass](https://godbolt.org/z/c5Yveb8Kf). "}, {"author": "DataCorrupted", "body": "I found the problem. Fixing it now. Will push a patch shortly"}]}, "verified": true} {"bug_id": "97053", "issue_url": "https://github.com/llvm/llvm-project/issues/97053", "bug_type": "miscompilation", "base_commit": "a3571376ad9555de07e4d8f74f92de1eaa4c486e", "knowledge_cutoff": "2024-06-28T12:58:31Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "4d2ae88d1617a910ec3a1436ce53579523ac2f97", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp": [[419, 424], [426, 431], [471, 478]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp": ["InstCombinerImpl::visitExtractElementInst"]}}, "patch": "commit 4d2ae88d1617a910ec3a1436ce53579523ac2f97\nAuthor: Nikita Popov \nDate: Wed Jul 3 10:58:38 2024 +0200\n\n [InstCombine] Fix invalid scalarization of div\n \n If the binop is not speculatable, and the extract index is out of\n range, then scalarizing will perform the operation on a poison\n operand, resulting in immediate UB, instead of the previous\n poison result.\n \n Fixes https://github.com/llvm/llvm-project/issues/97053.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\nindex 3de56a403803..753ed55523c8 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\n@@ -419,6 +419,7 @@ Instruction *InstCombinerImpl::visitExtractElementInst(ExtractElementInst &EI) {\n // If extracting a specified index from the vector, see if we can recursively\n // find a previously computed scalar that was inserted into the vector.\n auto *IndexC = dyn_cast(Index);\n+ bool HasKnownValidIndex = false;\n if (IndexC) {\n // Canonicalize type of constant indices to i64 to simplify CSE\n if (auto *NewIdx = getPreferredVectorIndex(IndexC))\n@@ -426,6 +427,7 @@ Instruction *InstCombinerImpl::visitExtractElementInst(ExtractElementInst &EI) {\n \n ElementCount EC = EI.getVectorOperandType()->getElementCount();\n unsigned NumElts = EC.getKnownMinValue();\n+ HasKnownValidIndex = IndexC->getValue().ult(NumElts);\n \n if (IntrinsicInst *II = dyn_cast(SrcVec)) {\n Intrinsic::ID IID = II->getIntrinsicID();\n@@ -471,8 +473,11 @@ Instruction *InstCombinerImpl::visitExtractElementInst(ExtractElementInst &EI) {\n return UnaryOperator::CreateWithCopiedFlags(UO->getOpcode(), E, UO);\n }\n \n+ // If the binop is not speculatable, we cannot hoist the extractelement if\n+ // it may make the operand poison.\n BinaryOperator *BO;\n- if (match(SrcVec, m_BinOp(BO)) && cheapToScalarize(SrcVec, Index)) {\n+ if (match(SrcVec, m_BinOp(BO)) && cheapToScalarize(SrcVec, Index) &&\n+ (HasKnownValidIndex || isSafeToSpeculativelyExecute(BO))) {\n // extelt (binop X, Y), Index --> binop (extelt X, Index), (extelt Y, Index)\n Value *X = BO->getOperand(0), *Y = BO->getOperand(1);\n Value *E0 = Builder.CreateExtractElement(X, Index);\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/scalarization.ll", "commands": ["opt -passes=instcombine -S < %s"], "tests": [{"test_name": "extract_element_binop_splat_variable_index_may_trap", "test_body": "define i8 @extract_element_binop_splat_variable_index_may_trap(<4 x i8> %x, <4 x i8> %y, i32 %z) {\n %b = sdiv <4 x i8> splat (i8 42), %y\n %r = extractelement <4 x i8> %b, i32 %z\n ret i8 %r\n}\n"}, {"test_name": "extract_element_binop_splat_variable_index", "test_body": "define i8 @extract_element_binop_splat_variable_index(<4 x i8> %x, i32 %y) {\n %b = sdiv <4 x i8> %x, splat (i8 42)\n %r = extractelement <4 x i8> %b, i32 %y\n ret i8 %r\n}\n"}]}], "issue": {"title": "Wrong swap between `sdiv` and `extractelement` by InstCombine", "body": "https://alive2.llvm.org/ce/z/mGNeLx\r\n\r\n\r\n```ll\r\n----------------------------------------\r\ndefine i8 @fun0(<4 x i8> %val0, i32 %val1) {\r\n#0:\r\n %val2 = sdiv <4 x i8> { 42, 42, 42, 42 }, %val0\r\n %val3 = extractelement <4 x i8> %val2, i32 %val1\r\n ret i8 %val3\r\n}\r\n=>\r\ndefine i8 @fun0(<4 x i8> %val0, i32 %val1) {\r\n#0:\r\n %#1 = extractelement <4 x i8> %val0, i32 %val1\r\n %val3 = sdiv i8 42, %#1\r\n ret i8 %val3\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Source is more defined than target\r\n\r\nExample:\r\n<4 x i8> %val0 = < #x01 (1), #x08 (8), #x02 (2), #x04 (4) >\r\ni32 %val1 = #x20000000 (536870912)\r\n\r\nSource:\r\n<4 x i8> %val2 = < #x2a (42), #x05 (5), #x15 (21), #x0a (10) >\r\ni8 %val3 = poison\r\n\r\nTarget:\r\ni8 %#1 = poison\r\ni8 %val3 = UB triggered!\r\n\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```\r\n\r\n\r\n`sdiv` and `extractelement` cannot be swapped for vector operands as it can trigger UB, diving an integer by `poison`.", "author": "bongjunj", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "121890", "issue_url": "https://github.com/llvm/llvm-project/issues/121890", "bug_type": "miscompilation", "base_commit": "7612dcc6e8d8e7f19b364084effbb01946294720", "knowledge_cutoff": "2025-01-07T06:21:54Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "9725595f3acc0c1aaa354e15ac4ee2b1f8ff4cc9", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": [[834, 841]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": ["InstCombinerImpl::foldGEPICmp"]}}, "patch": "commit 9725595f3acc0c1aaa354e15ac4ee2b1f8ff4cc9\nAuthor: Yingwei Zheng \nDate: Sat Feb 1 20:41:15 2025 +0800\n\n [InstCombine] Check nowrap flags when folding comparison of GEPs with the same base pointer (#121892)\n \n Alive2: https://alive2.llvm.org/ce/z/P5XbMx\n Closes https://github.com/llvm/llvm-project/issues/121890\n \n TODO: It is still safe to perform this transform without nowrap flags if\n the corresponding scale factor is 1 byte:\n https://alive2.llvm.org/ce/z/J-JCJd\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\nindex 9e9739ac2440..a8a5cb2b20d7 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n@@ -834,8 +834,9 @@ Instruction *InstCombinerImpl::foldGEPICmp(GEPOperator *GEPLHS, Value *RHS,\n return replaceInstUsesWith(\n I, // No comparison is needed here.\n ConstantInt::get(I.getType(), ICmpInst::isTrueWhenEqual(Cond)));\n-\n- else if (NumDifferences == 1 && CanFold(NW)) {\n+ // If two GEPs only differ by an index, compare them.\n+ // Note that nowrap flags are always needed when comparing two indices.\n+ else if (NumDifferences == 1 && NW != GEPNoWrapFlags::none()) {\n Value *LHSV = GEPLHS->getOperand(DiffOperand);\n Value *RHSV = GEPRHS->getOperand(DiffOperand);\n return NewICmp(NW, LHSV, RHSV);\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/opaque-ptr.ll", "commands": ["opt -S -passes='instcombine' < %s"], "tests": [{"test_name": "cmp_gep_same_base_same_type_nusw", "test_body": "target datalayout = \"p:8:8:8\"\ndefine i1 @cmp_gep_same_base_same_type_nusw(ptr %ptr, i8%idx1, i8%idx2) {\n %gep1 = getelementptr nusw i32, ptr %ptr, i8%idx1\n %gep2 = getelementptr nusw i32, ptr %ptr, i8%idx2\n %cmp = icmp eq ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}, {"test_name": "cmp_gep_same_base_same_type_maywrap", "test_body": "target datalayout = \"p:8:8:8\"\ndefine i1 @cmp_gep_same_base_same_type_maywrap(ptr %ptr, i8%idx1, i8%idx2) {\n %gep1 = getelementptr i32, ptr %ptr, i8%idx1\n %gep2 = getelementptr i32, ptr %ptr, i8%idx2\n %cmp = icmp eq ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}, {"test_name": "cmp_gep_same_base_same_type_nuw", "test_body": "target datalayout = \"p:8:8:8\"\ndefine i1 @cmp_gep_same_base_same_type_nuw(ptr %ptr, i8%idx1, i8%idx2) {\n %gep1 = getelementptr nuw i32, ptr %ptr, i8%idx1\n %gep2 = getelementptr nuw i32, ptr %ptr, i8%idx2\n %cmp = icmp eq ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}, {"test_name": "cmp_gep_same_base_different_type", "test_body": "target datalayout = \"p:8:8:8\"\ndefine i1 @cmp_gep_same_base_different_type(ptr %ptr, i8%idx1, i8%idx2) {\n %gep1 = getelementptr inbounds i32, ptr %ptr, i8%idx1\n %gep2 = getelementptr inbounds i64, ptr %ptr, i8%idx2\n %cmp = icmp ult ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}, {"test_name": "cmp_gep_same_base_same_type", "test_body": "target datalayout = \"p:8:8:8\"\ndefine i1 @cmp_gep_same_base_same_type(ptr %ptr, i8%idx1, i8%idx2) {\n %gep1 = getelementptr inbounds i32, ptr %ptr, i8%idx1\n %gep2 = getelementptr inbounds i32, ptr %ptr, i8%idx2\n %cmp = icmp ult ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}]}], "issue": {"title": "[InstCombine] Missing flag check when folding comparison of geps with the same base pointer", "body": "Alive2: https://alive2.llvm.org/ce/z/P5XbMx\n```\n\n----------------------------------------\ndefine i1 @src(ptr %ptr, i8 %idx1, i8 %idx2) {\n#0:\n %gep1 = gep ptr %ptr, 4 x i8 %idx1\n %gep2 = gep ptr %ptr, 4 x i8 %idx2\n %cmp = icmp eq ptr %gep1, %gep2, offsetonly\n ret i1 %cmp\n}\n=>\ndefine i1 @tgt(ptr %ptr, i8 %idx1, i8 %idx2) {\n#0:\n %cmp = icmp eq i8 %idx1, %idx2\n ret i1 %cmp\n}\nTransformation doesn't verify!\n\nERROR: Value mismatch\n\nExample:\nptr %ptr = null\ni8 %idx1 = #x00 (0)\ni8 %idx2 = #x40 (64)\n\nSource:\nptr %gep1 = null\nptr %gep2 = null\ni1 %cmp = #x1 (1)\n\nSOURCE MEMORY STATE\n===================\nNON-LOCAL BLOCKS:\nBlock 0 >\tsize: 0\talign: 1\talloc type: 0\talive: false\taddress: 0\nBlock 1 >\tsize: 0\talign: 1\talloc type: 0\talive: true\taddress: 1\n\nTarget:\ni1 %cmp = #x0 (0)\nSource value: #x1 (1)\nTarget value: #x0 (0)\n\n\n----------------------------------------\ndefine i1 @src1(ptr %ptr, i8 %idx1, i8 %idx2) {\n#0:\n %gep1 = gep nuw ptr %ptr, 4 x i8 %idx1\n %gep2 = gep nuw ptr %ptr, 4 x i8 %idx2\n %cmp = icmp eq ptr %gep1, %gep2, offsetonly\n ret i1 %cmp\n}\n=>\ndefine i1 @tgt1(ptr %ptr, i8 %idx1, i8 %idx2) {\n#0:\n %cmp = icmp eq i8 %idx1, %idx2\n ret i1 %cmp\n}\nTransformation seems to be correct!\n\n\n----------------------------------------\ndefine i1 @src2(ptr %ptr, i8 %idx1, i8 %idx2) {\n#0:\n %gep1 = gep nusw ptr %ptr, 4 x i8 %idx1\n %gep2 = gep nusw ptr %ptr, 4 x i8 %idx2\n %cmp = icmp eq ptr %gep1, %gep2, offsetonly\n ret i1 %cmp\n}\n=>\ndefine i1 @tgt2(ptr %ptr, i8 %idx1, i8 %idx2) {\n#0:\n %cmp = icmp eq i8 %idx1, %idx2\n ret i1 %cmp\n}\nTransformation seems to be correct!\n\nSummary:\n 2 correct transformations\n 1 incorrect transformations\n 0 failed-to-prove transformations\n 0 Alive2 errors\n```", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine", "generated by fuzzer"], "comments": []}, "verified": true} {"bug_id": "124275", "issue_url": "https://github.com/llvm/llvm-project/issues/124275", "bug_type": "miscompilation", "base_commit": "35df525fd00c2037ef144189ee818b7d612241ff", "knowledge_cutoff": "2025-01-24T14:33:26Z", "lit_test_dir": ["llvm/test/Analysis/ValueTracking"], "hints": {"fix_commit": "c2fba023475fddb893eac29dc9f34dfbdb221cd5", "components": ["ValueTracking"], "bug_location_lineno": {"llvm/lib/Analysis/ValueTracking.cpp": [[593, 603], [610, 620], [1673, 1680], [1705, 1713]]}, "bug_location_funcname": {"llvm/lib/Analysis/ValueTracking.cpp": ["breakSelfRecursivePHI", "computeKnownBitsFromOperator"]}}, "patch": "commit c2fba023475fddb893eac29dc9f34dfbdb221cd5\nAuthor: goldsteinn <35538541+goldsteinn@users.noreply.github.com>\nDate: Tue Jan 28 15:54:00 2025 -0600\n\n [ValueTracking] Fix bug of using wrong condition for deducing KnownBits (#124481)\n \n - **[ValueTracking] Add test for issue 124275**\n - **[ValueTracking] Fix bug of using wrong condition for deducing\n KnownBits**\n \n Fixes https://github.com/llvm/llvm-project/issues/124275\n \n Bug was introduced by https://github.com/llvm/llvm-project/pull/114689\n \n Now that computeKnownBits supports breaking out of recursive Phi\n nodes, `IncValue` can be an operand of a different Phi than `P`. This\n breaks the previous assumptions we had when using the possibly\n condition at `CxtI` to constrain `IncValue`.\n\ndiff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp\nindex eba728c7c8c3..b63a0a07f7de 100644\n--- a/llvm/lib/Analysis/ValueTracking.cpp\n+++ b/llvm/lib/Analysis/ValueTracking.cpp\n@@ -593,11 +593,14 @@ static bool cmpExcludesZero(CmpInst::Predicate Pred, const Value *RHS) {\n }\n \n static void breakSelfRecursivePHI(const Use *U, const PHINode *PHI,\n- Value *&ValOut, Instruction *&CtxIOut) {\n+ Value *&ValOut, Instruction *&CtxIOut,\n+ const PHINode **PhiOut = nullptr) {\n ValOut = U->get();\n if (ValOut == PHI)\n return;\n CtxIOut = PHI->getIncomingBlock(*U)->getTerminator();\n+ if (PhiOut)\n+ *PhiOut = PHI;\n Value *V;\n // If the Use is a select of this phi, compute analysis on other arm to break\n // recursion.\n@@ -610,11 +613,13 @@ static void breakSelfRecursivePHI(const Use *U, const PHINode *PHI,\n // incoming value to break recursion.\n // TODO: We could handle any number of incoming edges as long as we only have\n // two unique values.\n- else if (auto *IncPhi = dyn_cast(ValOut);\n- IncPhi && IncPhi->getNumIncomingValues() == 2) {\n+ if (auto *IncPhi = dyn_cast(ValOut);\n+ IncPhi && IncPhi->getNumIncomingValues() == 2) {\n for (int Idx = 0; Idx < 2; ++Idx) {\n if (IncPhi->getIncomingValue(Idx) == PHI) {\n ValOut = IncPhi->getIncomingValue(1 - Idx);\n+ if (PhiOut)\n+ *PhiOut = IncPhi;\n CtxIOut = IncPhi->getIncomingBlock(1 - Idx)->getTerminator();\n break;\n }\n@@ -1673,8 +1678,9 @@ static void computeKnownBitsFromOperator(const Operator *I,\n Known.One.setAllBits();\n for (const Use &U : P->operands()) {\n Value *IncValue;\n+ const PHINode *CxtPhi;\n Instruction *CxtI;\n- breakSelfRecursivePHI(&U, P, IncValue, CxtI);\n+ breakSelfRecursivePHI(&U, P, IncValue, CxtI, &CxtPhi);\n // Skip direct self references.\n if (IncValue == P)\n continue;\n@@ -1705,9 +1711,10 @@ static void computeKnownBitsFromOperator(const Operator *I,\n m_Br(m_c_ICmp(Pred, m_Specific(IncValue), m_APInt(RHSC)),\n m_BasicBlock(TrueSucc), m_BasicBlock(FalseSucc)))) {\n // Check for cases of duplicate successors.\n- if ((TrueSucc == P->getParent()) != (FalseSucc == P->getParent())) {\n+ if ((TrueSucc == CxtPhi->getParent()) !=\n+ (FalseSucc == CxtPhi->getParent())) {\n // If we're using the false successor, invert the predicate.\n- if (FalseSucc == P->getParent())\n+ if (FalseSucc == CxtPhi->getParent())\n Pred = CmpInst::getInversePredicate(Pred);\n // Get the knownbits implied by the incoming phi condition.\n auto CR = ConstantRange::makeExactICmpRegion(Pred, *RHSC);\n", "tests": [{"file": "llvm/test/Analysis/ValueTracking/phi-known-bits.ll", "commands": ["opt -O1 < %s -S"], "tests": [{"test_name": "issue_124275_wrong_br_direction", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@c = dso_local global i32 1, align 4\n@a = dso_local global i32 0, align 4\n@b = dso_local global i32 0, align 4\n\ndefine dso_local i32 @main() #0 {\nentry:\n %retval = alloca i32, align 4\n %d = alloca i32, align 4\n store i32 0, ptr %retval, align 4\n store i32 -1, ptr @a, align 4\n %0 = load i32, ptr @a, align 4\n %cmp = icmp ult i32 %0, 1\n br i1 %cmp, label %if.then, label %if.end\n\nif.then:\n br label %e\n\nif.end:\n %1 = load i32, ptr @c, align 4\n store i32 %1, ptr %d, align 4\n %2 = load i32, ptr %d, align 4\n %tobool = icmp ne i32 %2, 0\n br i1 %tobool, label %if.then1, label %if.end2\n\nif.then1:\n br label %e\n\ne:\n br label %if.end2\n\nif.end2:\n %3 = load i32, ptr %d, align 4\n %tobool3 = icmp ne i32 %3, 0\n br i1 %tobool3, label %if.end5, label %if.then4\n\nif.then4:\n unreachable\n\nif.end5:\n %4 = load i32, ptr @b, align 4\n %tobool6 = icmp ne i32 %4, 0\n br i1 %tobool6, label %if.then7, label %if.end8\n\nif.then7:\n br label %e\n\nif.end8:\n ret i32 0\n}\n\nattributes #0 = { noinline nounwind uwtable \"frame-pointer\"=\"all\" \"min-legal-vector-width\"=\"0\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\n\n!llvm.module.flags = !{!0, !1, !2, !3, !4, !5}\n!llvm.ident = !{!6}\n\n!0 = !{i32 7, !\"Dwarf Version\", i32 4}\n!1 = !{i32 1, !\"wchar_size\", i32 4}\n!2 = !{i32 8, !\"PIC Level\", i32 2}\n!3 = !{i32 7, !\"PIE Level\", i32 2}\n!4 = !{i32 7, !\"uwtable\", i32 2}\n!5 = !{i32 7, !\"frame-pointer\", i32 2}\n!6 = !{!\"clang version 21.0.0git (https://github.com/llvm/llvm-project.git 5c3c0a8cec9205efba78583bd0b8a646270968a1)\"}\n", "additional_args": "-src-unroll=4 -tgt-unroll=4"}]}], "issue": {"title": "wrong code at -O1 and above on x86_64-linux-gnu", "body": "It appears to be a recent regression from 19.1.0. \n\nCompiler Explorer: https://godbolt.org/z/haMfGoo6h\n\n```\n[548] % clangtk -v\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git ddd2f57b29661f21308eec0400fa92a6d075b0c6)\nTarget: x86_64-unknown-linux-gnu\nThread model: posix\nInstalledDir: /local/home/suz/suz-local/software/local/clang-trunk/bin\nBuild config: +assertions\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\nCandidate multilib: .;@m64\nSelected multilib: .;@m64\nFound CUDA installation: /usr/local/cuda, version 12.1\n[549] % \n[549] % clangtk -O1 small.c\n[550] % ./a.out\nAborted\n[551] % \n[551] % cat small.c\nint printf(const char *, ...);\nunsigned a;\nint b, c = 1;\nint main() {\n int d;\n a = -1;\n if (a < 1)\n goto e;\n d = c;\n if (d) {\n e:;\n }\n if (!d)\n __builtin_abort();\n if (b)\n goto e;\n return 0;\n}\n```", "author": "zhendongsu", "labels": ["miscompilation", "llvm:analysis"], "comments": [{"author": "zhendongsu", "body": "Another likely related test:\n```\n[552] % clangtk -v\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git ddd2f57b29661f21308eec0400fa92a6d075b0c6)\nTarget: x86_64-unknown-linux-gnu\nThread model: posix\nInstalledDir: /local/home/suz/suz-local/software/local/clang-trunk/bin\nBuild config: +assertions\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\nCandidate multilib: .;@m64\nSelected multilib: .;@m64\nFound CUDA installation: /usr/local/cuda, version 12.1\n[553] % \n[553] % clangtk -O0 small.c\n[554] % ./a.out\n[555] % \n[555] % clangtk -O1 small.c\n[556] % timeout -s 9 5 ./a.out\nKilled\n[557] % cat small.c\nint a, b;\nint main() {\n {\n int c;\n if (a && 0)\n goto d;\n }\n int e = -~b;\n if (e) {\n d:;\n }\n while (!e)\n goto d;\n return 0;\n}\n```"}, {"author": "antoniofrighetto", "body": "Reduced to:\n\n```llvm\n; opt -p instcombine small.ll\ndefine i32 @small(i32 noundef %0) {\n %2 = xor i32 %0, 1\n %3 = sub i32 0, %2\n %4 = icmp ne i32 %3, 0\n br i1 %4, label %5, label %6\n\n5: ; preds = %6, %1\n %.0 = phi i32 [ %.1, %6 ], [ %3, %1 ]\n br label %6\n\n6: ; preds = %5, %1\n %.1 = phi i32 [ %.0, %5 ], [ 0, %1 ]\n %7 = icmp ne i32 %.1, 0\n %8 = xor i1 %7, true\n br i1 %8, label %5, label %9\n\n9: ; preds = %6\n ret i32 0\n}\n```"}, {"author": "goldsteinn", "body": "Fix at: https://github.com/llvm/llvm-project/pull/124481"}]}, "verified": true} {"bug_id": "106083", "issue_url": "https://github.com/llvm/llvm-project/issues/106083", "bug_type": "crash", "base_commit": "d075debc508898d5f365f8e909c54d6f4edada85", "knowledge_cutoff": "2024-08-26T14:36:57Z", "lit_test_dir": ["llvm/test/Transforms/DFAJumpThreading"], "hints": {"fix_commit": "d4a38c8ff5c993e14c42895b51a47272fb03a857", "components": ["DFAJumpThreading"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp": [[194, 200], [202, 207], [223, 254]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp": ["unfold"]}}, "patch": "commit d4a38c8ff5c993e14c42895b51a47272fb03a857\nAuthor: Usman Nadeem \nDate: Tue Sep 24 08:54:36 2024 -0700\n\n [DFAJumpThreading] Handle select unfolding when user phi is not a dir… (#109511)\n \n …ect successor\n \n Previously the code assumed that the select instruction is defined in a\n block that is a direct predecessor of the block where the PHINode uses\n it. So, we were hitting an assertion when we tried to access the def\n block as an incoming block for the user phi node.\n \n This patch handles that case by using the correct end block and creating\n a new phi node that aggregates both the values of the select in that end\n block, and then using that new unfolded phi to overwrite the original\n user phi node.\n \n Fixes #106083\n \n Change-Id: Ie471994cca232318f74a6e6438efa21e561c2dc0\n\ndiff --git a/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp b/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp\nindex ef9c264482a6..0e2b5c925a6a 100644\n--- a/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp\n+++ b/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp\n@@ -194,7 +194,6 @@ void unfold(DomTreeUpdater *DTU, LoopInfo *LI, SelectInstToUnfold SIToUnfold,\n SelectInst *SI = SIToUnfold.getInst();\n PHINode *SIUse = SIToUnfold.getUse();\n BasicBlock *StartBlock = SI->getParent();\n- BasicBlock *EndBlock = SIUse->getParent();\n BranchInst *StartBlockTerm =\n dyn_cast(StartBlock->getTerminator());\n \n@@ -202,6 +201,7 @@ void unfold(DomTreeUpdater *DTU, LoopInfo *LI, SelectInstToUnfold SIToUnfold,\n assert(SI->hasOneUse());\n \n if (StartBlockTerm->isUnconditional()) {\n+ BasicBlock *EndBlock = StartBlock->getUniqueSuccessor();\n // Arbitrarily choose the 'false' side for a new input value to the PHI.\n BasicBlock *NewBlock = BasicBlock::Create(\n SI->getContext(), Twine(SI->getName(), \".si.unfold.false\"),\n@@ -223,32 +223,44 @@ void unfold(DomTreeUpdater *DTU, LoopInfo *LI, SelectInstToUnfold SIToUnfold,\n NewBlock->getFirstInsertionPt());\n NewPhi->addIncoming(SIOp2, StartBlock);\n \n- if (auto *OpSi = dyn_cast(SIOp1))\n- NewSIsToUnfold->push_back(SelectInstToUnfold(OpSi, SIUse));\n- if (auto *OpSi = dyn_cast(SIOp2))\n- NewSIsToUnfold->push_back(SelectInstToUnfold(OpSi, NewPhi));\n-\n- // Update the phi node of SI.\n- for (unsigned Idx = 0; Idx < SIUse->getNumIncomingValues(); ++Idx) {\n- if (SIUse->getIncomingBlock(Idx) == StartBlock)\n- SIUse->setIncomingValue(Idx, SIOp1);\n+ // Update any other PHI nodes in EndBlock.\n+ for (PHINode &Phi : EndBlock->phis()) {\n+ if (SIUse == &Phi)\n+ continue;\n+ Phi.addIncoming(Phi.getIncomingValueForBlock(StartBlock), NewBlock);\n }\n- SIUse->addIncoming(NewPhi, NewBlock);\n \n- // Update any other PHI nodes in EndBlock.\n- for (auto II = EndBlock->begin(); PHINode *Phi = dyn_cast(II);\n- ++II) {\n- if (Phi != SIUse)\n- Phi->addIncoming(Phi->getIncomingValueForBlock(StartBlock), NewBlock);\n+ // Update the phi node of SI, which is its only use.\n+ if (EndBlock == SIUse->getParent()) {\n+ SIUse->addIncoming(NewPhi, NewBlock);\n+ SIUse->replaceUsesOfWith(SI, SIOp1);\n+ } else {\n+ PHINode *EndPhi = PHINode::Create(SIUse->getType(), pred_size(EndBlock),\n+ Twine(SI->getName(), \".si.unfold.phi\"),\n+ EndBlock->getFirstInsertionPt());\n+ for (BasicBlock *Pred : predecessors(EndBlock)) {\n+ if (Pred != StartBlock && Pred != NewBlock)\n+ EndPhi->addIncoming(EndPhi, Pred);\n+ }\n+\n+ EndPhi->addIncoming(SIOp1, StartBlock);\n+ EndPhi->addIncoming(NewPhi, NewBlock);\n+ SIUse->replaceUsesOfWith(SI, EndPhi);\n+ SIUse = EndPhi;\n }\n \n- StartBlockTerm->eraseFromParent();\n+ if (auto *OpSi = dyn_cast(SIOp1))\n+ NewSIsToUnfold->push_back(SelectInstToUnfold(OpSi, SIUse));\n+ if (auto *OpSi = dyn_cast(SIOp2))\n+ NewSIsToUnfold->push_back(SelectInstToUnfold(OpSi, NewPhi));\n \n // Insert the real conditional branch based on the original condition.\n+ StartBlockTerm->eraseFromParent();\n BranchInst::Create(EndBlock, NewBlock, SI->getCondition(), StartBlock);\n DTU->applyUpdates({{DominatorTree::Insert, StartBlock, EndBlock},\n {DominatorTree::Insert, StartBlock, NewBlock}});\n } else {\n+ BasicBlock *EndBlock = SIUse->getParent();\n BasicBlock *NewBlockT = BasicBlock::Create(\n SI->getContext(), Twine(SI->getName(), \".si.unfold.true\"),\n EndBlock->getParent(), EndBlock);\n", "tests": [{"file": "llvm/test/Transforms/DFAJumpThreading/dfa-jump-threading-transform.ll", "commands": ["opt -S -passes=dfa-jump-threading %s"], "tests": [{"test_name": "test2", "test_body": "define i32 @test2(i32 %init) {\nentry:\n %cmp = icmp eq i32 %init, 0\n %sel = select i1 %cmp, i32 0, i32 2\n br label %loop.1\n\nloop.1: ; preds = %loop.1.backedge, %entry\n %state.1 = phi i32 [ %sel, %entry ], [ %state.1.be2, %loop.1.backedge ]\n br label %loop.2\n\nloop.2: ; preds = %loop.2.backedge, %loop.1\n %state.2 = phi i32 [ %state.1, %loop.1 ], [ %state.2.be, %loop.2.backedge ]\n br label %loop.3\n\nloop.3: ; preds = %case2, %loop.2\n %state = phi i32 [ %state.2, %loop.2 ], [ 3, %case2 ]\n switch i32 %state, label %infloop.i [\n i32 2, label %case2\n i32 3, label %case3\n i32 4, label %case4\n i32 0, label %case0\n i32 1, label %case1\n ]\n\ncase2: ; preds = %loop.3\n br i1 %cmp, label %loop.3, label %loop.1.backedge\n\ncase3: ; preds = %loop.3\n br i1 %cmp, label %loop.2.backedge, label %case4\n\ncase4: ; preds = %case3, %loop.3\n br i1 %cmp, label %loop.2.backedge, label %loop.1.backedge\n\nloop.1.backedge: ; preds = %case4, %case2\n %state.1.be = phi i32 [ 2, %case4 ], [ 4, %case2 ]\n %state.1.be2 = select i1 %cmp, i32 1, i32 %state.1.be\n br label %loop.1\n\nloop.2.backedge: ; preds = %case4, %case3\n %state.2.be = phi i32 [ 3, %case4 ], [ 0, %case3 ]\n br label %loop.2\n\ncase0: ; preds = %loop.3\n br label %exit\n\ncase1: ; preds = %loop.3\n br label %exit\n\ninfloop.i: ; preds = %infloop.i, %loop.3\n br label %infloop.i\n\nexit: ; preds = %case1, %case0\n ret i32 0\n}\n"}, {"test_name": "pr106083_select_dead_uses", "test_body": "define void @pr106083_select_dead_uses(i1 %cmp1, i1 %not, ptr %p) {\nbb:\n %spec.select = select i1 %cmp1, i32 0, i32 1\n br label %.loopexit6\n\n.loopexit6: ; preds = %select.unfold, %bb\n br i1 %not, label %select.unfold, label %bb1\n\nbb1: ; preds = %.loopexit6\n %i = load i32, ptr %p, align 4\n %not2 = icmp eq i32 0, 0\n %spec.select7 = select i1 %not2, i32 %spec.select, i32 0\n br label %select.unfold\n\nselect.unfold: ; preds = %bb1, %.loopexit6\n %_2 = phi i32 [ 0, %.loopexit6 ], [ %spec.select7, %bb1 ]\n switch i32 %_2, label %bb2 [\n i32 0, label %.preheader.preheader\n i32 1, label %.loopexit6\n ]\n\n.preheader.preheader: ; preds = %select.unfold\n ret void\n\nbb2: ; preds = %select.unfold\n unreachable\n}\n"}, {"test_name": "test1", "test_body": "define i32 @test1(i32 %num) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.inc, %entry\n %count = phi i32 [ 0, %entry ], [ %inc, %for.inc ]\n %state = phi i32 [ 1, %entry ], [ %state.next, %for.inc ]\n switch i32 %state, label %for.inc [\n i32 1, label %case1\n i32 2, label %case2\n ]\n\ncase1: ; preds = %for.body\n br label %for.inc\n\ncase2: ; preds = %for.body\n %cmp = icmp eq i32 %count, 50\n %sel = select i1 %cmp, i32 1, i32 2\n br label %for.inc\n\nfor.inc: ; preds = %case2, %case1, %for.body\n %state.next = phi i32 [ %sel, %case2 ], [ 1, %for.body ], [ 2, %case1 ]\n %inc = add nsw i32 %count, 1\n %cmp.exit = icmp slt i32 %inc, %num\n br i1 %cmp.exit, label %for.body, label %for.end\n\nfor.end: ; preds = %for.inc\n ret i32 0\n}\n"}, {"test_name": "pr106083_invalidBBarg_fold", "test_body": "define void @pr106083_invalidBBarg_fold(i1 %cmp1, i1 %cmp2, i1 %not, ptr %d) {\nbb:\n %sel = select i1 %cmp1, i32 0, i32 1\n br label %BB1\n\nBB1: ; preds = %BB1.backedge, %BB7, %bb\n %i = phi i16 [ 0, %BB1.backedge ], [ 0, %bb ], [ 1, %BB7 ]\n br i1 %not, label %BB7, label %BB2\n\nBB2: ; preds = %BB1\n store i16 0, ptr %d, align 2\n %spec.select = select i1 %cmp2, i32 %sel, i32 0\n br label %BB7\n\nBB7: ; preds = %BB2, %BB1\n %d.promoted4 = phi i16 [ 0, %BB1 ], [ 1, %BB2 ]\n %_3 = phi i32 [ 0, %BB1 ], [ %spec.select, %BB2 ]\n switch i32 %_3, label %BB1.backedge [\n i32 0, label %BB1\n i32 1, label %BB8\n ]\n\nBB1.backedge: ; preds = %BB7\n br label %BB1\n\nBB8: ; preds = %BB7\n ret void\n}\n"}, {"test_name": "pr78059_bitwidth", "test_body": "define void @pr78059_bitwidth() {\n.split.preheader:\n br label %.split\n\n.split: ; preds = %.split, %.split.preheader\n %0 = phi i128 [ 0, %.split.preheader ], [ -1, %.split ]\n switch i128 %0, label %end [\n i128 -1, label %end\n i128 0, label %.split\n ]\n\nend: ; preds = %.split, %.split\n ret void\n}\n"}]}], "issue": {"title": "[DFAJumpThreading] crash since b167ada", "body": "Seeing many assertions that go away if I revert b167ada \"[DFAJumpThreading] Rewrite the way paths are enumerated (#96127)\".\r\n\r\n```\r\nclang-20 -cc1 -triple s390x-unknown-linux-gnu -target-cpu z15 -O3 -w -mllvm -enable-dfa-jump-thread -o /dev/null -emit-llvm -x ir tc_dfa-jt_invalidbbarg.ll\r\n...\r\n10 0x000002aa0a73e544 (anonymous namespace)::DFAJumpThreading::run\r\n...\r\n```\r\n\r\n```\r\nclang-20 -cc1 -triple s390x-unknown-linux-gnu -target-cpu z16 -O3 -w -mllvm -enable-dfa-jump-thread -o /dev/null -emit-llvm -x ir tc_select_mustbedeadnow.ll\r\n...\r\n#10 0x000002aa1b0c0968 llvm::DFAJumpThreadingPass::run\r\n...\r\n\r\n```\r\n[tcs_dfa-jt.tar.gz](https://github.com/user-attachments/files/16750050/tcs_dfa-jt.tar.gz)\r\n", "author": "JonPsson1", "labels": ["crash", "llvm:transforms"], "comments": [{"author": "mikaelholmen", "body": "Perhaps the same assertion I reported here:\r\n https://github.com/llvm/llvm-project/pull/96127#issuecomment-2303862037\r\n?"}, {"author": "JonPsson1", "body": "> Perhaps the same assertion I reported here: [#96127 (comment)](https://github.com/llvm/llvm-project/pull/96127#issuecomment-2303862037) ?\r\n\r\nYes, same assertion, and in addition my second test case above triggered another assertion."}]}, "verified": true} {"bug_id": "113810", "issue_url": "https://github.com/llvm/llvm-project/issues/113810", "bug_type": "crash", "base_commit": "ce0368eb8417f2d369499bb98b1f0ccbe2219598", "knowledge_cutoff": "2024-10-27T14:16:44Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "7152bf3bc805b8d9b1873058ab0a084d7b6079d6", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[7947, 7954]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::buildTree_rec"]}}, "patch": "commit 7152bf3bc805b8d9b1873058ab0a084d7b6079d6\nAuthor: Alexey Bataev \nDate: Mon Oct 28 06:37:24 2024 -0700\n\n [SLP]Do not create new vector node if scalars fully overlap with the existing one\n \n If the list of scalars vectorized as the part of the same vector node,\n no need to generate vector node again, it will be handled as part of\n overlapping matching.\n \n Fixes #113810\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 2afd02dae3a8..268546fe99e1 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -7947,8 +7947,13 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n Nodes.insert(E);\n SmallPtrSet Values(VL.begin(), VL.end());\n if (any_of(Nodes, [&](const TreeEntry *E) {\n- return all_of(E->Scalars,\n- [&](Value *V) { return Values.contains(V); });\n+ if (all_of(E->Scalars,\n+ [&](Value *V) { return Values.contains(V); }))\n+ return true;\n+ SmallPtrSet EValues(E->Scalars.begin(),\n+ E->Scalars.end());\n+ return (\n+ all_of(VL, [&](Value *V) { return EValues.contains(V); }));\n })) {\n LLVM_DEBUG(dbgs() << \"SLP: Gathering due to full overlap.\\n\");\n if (TryToFindDuplicates(S))\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/full-overlap-non-schedulable.ll", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "test", "test_body": "define void @test(ptr %p1, ptr %0, i32 %1, i1 %c1, ptr %p2) {\ntop:\n %2 = getelementptr i8, ptr %0, i64 8\n %3 = getelementptr i8, ptr %0, i64 12\n %4 = getelementptr i8, ptr %0, i64 16\n %5 = getelementptr i8, ptr %0, i64 20\n br i1 %c1, label %L42, label %L41\n\nL41: ; preds = %top\n %.not276 = icmp eq ptr %2, null\n %6 = load i32, ptr %2, align 4\n %7 = select i1 %.not276, i32 0, i32 %6\n %.not277 = icmp eq ptr %3, null\n %8 = load i32, ptr %3, align 4\n %9 = select i1 %.not277, i32 0, i32 %8\n %.not278 = icmp eq ptr %4, null\n %10 = load i32, ptr %4, align 4\n %11 = select i1 %.not278, i32 0, i32 %10\n %.not279 = icmp eq ptr %5, null\n %12 = load i32, ptr %5, align 4\n %13 = select i1 %.not279, i32 0, i32 %12\n br label %L112\n\nL42: ; preds = %top\n %14 = load i32, ptr %2, align 4\n %.not280 = icmp eq i32 %14, 0\n br i1 %.not280, label %L112, label %L47\n\nL47: ; preds = %L42\n %15 = load i32, ptr %3, align 4\n %.not282 = icmp eq ptr %4, null\n %16 = load i32, ptr %4, align 4\n %17 = select i1 %.not282, i32 0, i32 %16\n %.not283 = icmp eq ptr %5, null\n %18 = load i32, ptr %5, align 4\n %19 = select i1 %.not283, i32 0, i32 %18\n br label %L112\n\nL112: ; preds = %L47, %L42, %L41\n %value_phi13336 = phi i32 [ %19, %L47 ], [ %13, %L41 ], [ 0, %L42 ]\n %value_phi12335 = phi i32 [ %17, %L47 ], [ %11, %L41 ], [ %1, %L42 ]\n %value_phi11334 = phi i32 [ %15, %L47 ], [ %9, %L41 ], [ 0, %L42 ]\n %value_phi10333 = phi i32 [ 0, %L47 ], [ %7, %L41 ], [ 0, %L42 ]\n store i32 %value_phi10333, ptr %p2, align 4\n store i32 %value_phi11334, ptr %p1, align 4\n store i32 %value_phi12335, ptr %p2, align 4\n store i32 %value_phi13336, ptr %p1, align 4\n ret void\n}\n"}]}], "issue": {"title": "SLPVectorizer miscompilation", "body": "This IR\r\n```llvm\r\n; Function Attrs: alignstack(16)\r\ndefine swiftcc { ptr addrspace(10), i8 } @julia__zip_iterate_all_17381(ptr noalias nocapture noundef nonnull align 4 dereferenceable(96) %union_bytes_return, ptr nonnull swiftself %pgcstack_arg, ptr addrspace(11) nocapture noundef nonnull readonly align 4 dereferenceable(48) %0, ptr addrspace(11) nocapture noundef nonnull readonly align 4 dereferenceable(48) %1) #0 {\r\ntop:\r\n %pgcstack = call ptr @julia.get_pgcstack()\r\n %ptls_field = getelementptr inbounds i8, ptr %pgcstack, i32 8\r\n %ptls_load = load ptr, ptr %ptls_field, align 4\r\n %2 = getelementptr inbounds i8, ptr %ptls_load, i32 16\r\n %safepoint = load ptr, ptr %2, align 4\r\n fence syncscope(\"singlethread\") seq_cst\r\n call void @julia.safepoint(ptr %safepoint)\r\n fence syncscope(\"singlethread\") seq_cst\r\n %3 = load i32, ptr addrspace(11) %1, align 4\r\n %4 = load i32, ptr addrspace(11) %0, align 4\r\n %.not = icmp eq i32 %3, %4\r\n br i1 %.not, label %L26, label %L20\r\n\r\nL20: ; preds = %top\r\n %5 = add i32 %3, 1\r\n %6 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 4\r\n %7 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 8\r\n %8 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 12\r\n %9 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 16\r\n %10 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 20\r\n %.not265 = icmp eq ptr addrspace(11) %6, null\r\n %11 = load i32, ptr addrspace(11) %6, align 4\r\n %12 = select i1 %.not265, i32 undef, i32 %11\r\n %.not266 = icmp eq ptr addrspace(11) %7, null\r\n %13 = load i32, ptr addrspace(11) %7, align 4\r\n %14 = select i1 %.not266, i32 undef, i32 %13\r\n %.not267 = icmp eq ptr addrspace(11) %8, null\r\n %15 = load i32, ptr addrspace(11) %8, align 4\r\n %16 = select i1 %.not267, i32 undef, i32 %15\r\n %.not268 = icmp eq ptr addrspace(11) %9, null\r\n %17 = load i32, ptr addrspace(11) %9, align 4\r\n %18 = select i1 %.not268, i32 undef, i32 %17\r\n %.not269 = icmp eq ptr addrspace(11) %10, null\r\n %19 = load i32, ptr addrspace(11) %10, align 4\r\n %20 = select i1 %.not269, i32 undef, i32 %19\r\n br label %L112\r\n\r\nL26: ; preds = %top\r\n %21 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 4\r\n %22 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 8\r\n %23 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 12\r\n %24 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 16\r\n %25 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 20\r\n %26 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 4\r\n %27 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 12\r\n %28 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 16\r\n %29 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 20\r\n %30 = load i32, ptr addrspace(11) %21, align 4\r\n %31 = load i32, ptr addrspace(11) %26, align 4\r\n %.not275 = icmp eq i32 %30, %31\r\n br i1 %.not275, label %L42, label %L41\r\n\r\nL41: ; preds = %L26\r\n %32 = add i32 %30, 1\r\n %.not276 = icmp eq ptr addrspace(11) %22, null\r\n %33 = load i32, ptr addrspace(11) %22, align 4\r\n %34 = select i1 %.not276, i32 undef, i32 %33\r\n %.not277 = icmp eq ptr addrspace(11) %23, null\r\n %35 = load i32, ptr addrspace(11) %23, align 4\r\n %36 = select i1 %.not277, i32 undef, i32 %35\r\n %.not278 = icmp eq ptr addrspace(11) %24, null\r\n %37 = load i32, ptr addrspace(11) %24, align 4\r\n %38 = select i1 %.not278, i32 undef, i32 %37\r\n %.not279 = icmp eq ptr addrspace(11) %25, null\r\n %39 = load i32, ptr addrspace(11) %25, align 4\r\n %40 = select i1 %.not279, i32 undef, i32 %39\r\n br label %L112\r\n\r\nL42: ; preds = %L26\r\n %41 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 8\r\n %42 = load i32, ptr addrspace(11) %22, align 4\r\n %43 = load i32, ptr addrspace(11) %41, align 4\r\n %.not280 = icmp eq i32 %42, %43\r\n br i1 %.not280, label %L48, label %L47\r\n\r\nL47: ; preds = %L42\r\n %44 = add i32 %42, 1\r\n %.not281 = icmp eq ptr addrspace(11) %23, null\r\n %45 = load i32, ptr addrspace(11) %23, align 4\r\n %46 = select i1 %.not281, i32 undef, i32 %45\r\n %.not282 = icmp eq ptr addrspace(11) %24, null\r\n %47 = load i32, ptr addrspace(11) %24, align 4\r\n %48 = select i1 %.not282, i32 undef, i32 %47\r\n %.not283 = icmp eq ptr addrspace(11) %25, null\r\n %49 = load i32, ptr addrspace(11) %25, align 4\r\n %50 = select i1 %.not283, i32 undef, i32 %49\r\n br label %L112\r\n\r\nL48: ; preds = %L42\r\n %51 = load i32, ptr addrspace(11) %23, align 4\r\n %52 = load i32, ptr addrspace(11) %27, align 4\r\n %.not284 = icmp eq i32 %51, %52\r\n br i1 %.not284, label %L54, label %L53\r\n\r\nL53: ; preds = %L48\r\n %53 = add i32 %51, 1\r\n %.not285 = icmp eq ptr addrspace(11) %24, null\r\n %54 = load i32, ptr addrspace(11) %24, align 4\r\n %55 = select i1 %.not285, i32 undef, i32 %54\r\n %.not286 = icmp eq ptr addrspace(11) %25, null\r\n %56 = load i32, ptr addrspace(11) %25, align 4\r\n %57 = select i1 %.not286, i32 undef, i32 %56\r\n br label %L112\r\n\r\nL54: ; preds = %L48\r\n %58 = load i32, ptr addrspace(11) %24, align 4\r\n %59 = load i32, ptr addrspace(11) %28, align 4\r\n %.not287 = icmp eq i32 %58, %59\r\n br i1 %.not287, label %L87, label %L59\r\n\r\nL59: ; preds = %L54\r\n %60 = add i32 %58, 1\r\n %.not288 = icmp eq ptr addrspace(11) %25, null\r\n %61 = load i32, ptr addrspace(11) %25, align 4\r\n %62 = select i1 %.not288, i32 undef, i32 %61\r\n br label %L112\r\n\r\nL87: ; preds = %L54\r\n %63 = load i32, ptr addrspace(11) %25, align 4\r\n %64 = add i32 %63, 1\r\n %65 = load i32, ptr addrspace(11) %29, align 4\r\n %66 = icmp eq i32 %63, %65\r\n br i1 %66, label %common.ret, label %L112\r\n\r\nL112: ; preds = %L87, %L20, %L41, %L47, %L53, %L59\r\n %value_phi13336 = phi i32 [ %64, %L87 ], [ %62, %L59 ], [ %57, %L53 ], [ %50, %L47 ], [ %40, %L41 ], [ %20, %L20 ]\r\n %value_phi12335 = phi i32 [ 1, %L87 ], [ %60, %L59 ], [ %55, %L53 ], [ %48, %L47 ], [ %38, %L41 ], [ %18, %L20 ]\r\n %value_phi11334 = phi i32 [ 1, %L87 ], [ 1, %L59 ], [ %53, %L53 ], [ %46, %L47 ], [ %36, %L41 ], [ %16, %L20 ]\r\n %value_phi10333 = phi i32 [ 1, %L87 ], [ 1, %L59 ], [ 1, %L53 ], [ %44, %L47 ], [ %34, %L41 ], [ %14, %L20 ]\r\n %value_phi9332 = phi i32 [ 1, %L87 ], [ 1, %L59 ], [ 1, %L53 ], [ 1, %L47 ], [ %32, %L41 ], [ %12, %L20 ]\r\n %value_phi2326 = phi i32 [ 1, %L87 ], [ 1, %L59 ], [ 1, %L53 ], [ 1, %L47 ], [ 1, %L41 ], [ %5, %L20 ]\r\n %67 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 24\r\n %68 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 24\r\n %69 = load i32, ptr addrspace(11) %68, align 4\r\n %70 = load i32, ptr addrspace(11) %67, align 4\r\n %.not289 = icmp eq i32 %69, %70\r\n br i1 %.not289, label %L134, label %L128\r\n\r\nL128: ; preds = %L112\r\n %71 = add i32 %69, 1\r\n %72 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 28\r\n %73 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 32\r\n %74 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 36\r\n %75 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 40\r\n %76 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 44\r\n %.not290 = icmp eq ptr addrspace(11) %72, null\r\n %77 = load i32, ptr addrspace(11) %72, align 4\r\n %78 = select i1 %.not290, i32 undef, i32 %77\r\n %.not291 = icmp eq ptr addrspace(11) %73, null\r\n %79 = load i32, ptr addrspace(11) %73, align 4\r\n %80 = select i1 %.not291, i32 undef, i32 %79\r\n %.not292 = icmp eq ptr addrspace(11) %74, null\r\n %81 = load i32, ptr addrspace(11) %74, align 4\r\n %82 = select i1 %.not292, i32 undef, i32 %81\r\n %.not293 = icmp eq ptr addrspace(11) %75, null\r\n %83 = load i32, ptr addrspace(11) %75, align 4\r\n %84 = select i1 %.not293, i32 undef, i32 %83\r\n %.not294 = icmp eq ptr addrspace(11) %76, null\r\n %85 = load i32, ptr addrspace(11) %76, align 4\r\n %86 = select i1 %.not294, i32 undef, i32 %85\r\n br label %L230\r\n\r\nL134: ; preds = %L112\r\n %87 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 28\r\n %88 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 32\r\n %89 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 36\r\n %90 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 40\r\n %91 = getelementptr inbounds i8, ptr addrspace(11) %1, i32 44\r\n %92 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 28\r\n %93 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 36\r\n %94 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 40\r\n %95 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 44\r\n %96 = load i32, ptr addrspace(11) %87, align 4\r\n %97 = load i32, ptr addrspace(11) %92, align 4\r\n %.not300 = icmp eq i32 %96, %97\r\n br i1 %.not300, label %L150, label %L149\r\n\r\nL149: ; preds = %L134\r\n %98 = add i32 %96, 1\r\n %.not301 = icmp eq ptr addrspace(11) %88, null\r\n %99 = load i32, ptr addrspace(11) %88, align 4\r\n %100 = select i1 %.not301, i32 undef, i32 %99\r\n %.not302 = icmp eq ptr addrspace(11) %89, null\r\n %101 = load i32, ptr addrspace(11) %89, align 4\r\n %102 = select i1 %.not302, i32 undef, i32 %101\r\n %.not303 = icmp eq ptr addrspace(11) %90, null\r\n %103 = load i32, ptr addrspace(11) %90, align 4\r\n %104 = select i1 %.not303, i32 undef, i32 %103\r\n %.not304 = icmp eq ptr addrspace(11) %91, null\r\n %105 = load i32, ptr addrspace(11) %91, align 4\r\n %106 = select i1 %.not304, i32 undef, i32 %105\r\n br label %L230\r\n\r\nL150: ; preds = %L134\r\n %107 = getelementptr inbounds i8, ptr addrspace(11) %0, i32 32\r\n %108 = load i32, ptr addrspace(11) %88, align 4\r\n %109 = load i32, ptr addrspace(11) %107, align 4\r\n %.not305 = icmp eq i32 %108, %109\r\n br i1 %.not305, label %L156, label %L155\r\n\r\nL155: ; preds = %L150\r\n %110 = add i32 %108, 1\r\n %.not306 = icmp eq ptr addrspace(11) %89, null\r\n %111 = load i32, ptr addrspace(11) %89, align 4\r\n %112 = select i1 %.not306, i32 undef, i32 %111\r\n %.not307 = icmp eq ptr addrspace(11) %90, null\r\n %113 = load i32, ptr addrspace(11) %90, align 4\r\n %114 = select i1 %.not307, i32 undef, i32 %113\r\n %.not308 = icmp eq ptr addrspace(11) %91, null\r\n %115 = load i32, ptr addrspace(11) %91, align 4\r\n %116 = select i1 %.not308, i32 undef, i32 %115\r\n br label %L230\r\n\r\nL156: ; preds = %L150\r\n %117 = load i32, ptr addrspace(11) %89, align 4\r\n %118 = load i32, ptr addrspace(11) %93, align 4\r\n %.not309 = icmp eq i32 %117, %118\r\n br i1 %.not309, label %L162, label %L161\r\n\r\nL161: ; preds = %L156\r\n %119 = add i32 %117, 1\r\n %.not310 = icmp eq ptr addrspace(11) %90, null\r\n %120 = load i32, ptr addrspace(11) %90, align 4\r\n %121 = select i1 %.not310, i32 undef, i32 %120\r\n %.not311 = icmp eq ptr addrspace(11) %91, null\r\n %122 = load i32, ptr addrspace(11) %91, align 4\r\n %123 = select i1 %.not311, i32 undef, i32 %122\r\n br label %L230\r\n\r\nL162: ; preds = %L156\r\n %124 = load i32, ptr addrspace(11) %90, align 4\r\n %125 = load i32, ptr addrspace(11) %94, align 4\r\n %.not312 = icmp eq i32 %124, %125\r\n br i1 %.not312, label %L195, label %L167\r\n\r\nL167: ; preds = %L162\r\n %126 = add i32 %124, 1\r\n %.not313 = icmp eq ptr addrspace(11) %91, null\r\n %127 = load i32, ptr addrspace(11) %91, align 4\r\n %128 = select i1 %.not313, i32 undef, i32 %127\r\n br label %L230\r\n\r\nL195: ; preds = %L162\r\n %129 = load i32, ptr addrspace(11) %91, align 4\r\n %130 = add i32 %129, 1\r\n %131 = load i32, ptr addrspace(11) %95, align 4\r\n %132 = icmp eq i32 %129, %131\r\n br i1 %132, label %common.ret, label %L230\r\n\r\ncommon.ret: ; preds = %L87, %L195, %L230\r\n %common.ret.op = phi { ptr addrspace(10), i8 } [ { ptr addrspace(10) null, i8 2 }, %L230 ], [ { ptr addrspace(10) null, i8 1 }, %L195 ], [ { ptr addrspace(10) null, i8 1 }, %L87 ]\r\n ret { ptr addrspace(10), i8 } %common.ret.op\r\n\r\nL230: ; preds = %L195, %L128, %L149, %L155, %L161, %L167\r\n %value_phi29359 = phi i32 [ %130, %L195 ], [ %128, %L167 ], [ %123, %L161 ], [ %116, %L155 ], [ %106, %L149 ], [ %86, %L128 ]\r\n %value_phi28358 = phi i32 [ 1, %L195 ], [ %126, %L167 ], [ %121, %L161 ], [ %114, %L155 ], [ %104, %L149 ], [ %84, %L128 ]\r\n %value_phi27357 = phi i32 [ 1, %L195 ], [ 1, %L167 ], [ %119, %L161 ], [ %112, %L155 ], [ %102, %L149 ], [ %82, %L128 ]\r\n %value_phi26356 = phi i32 [ 1, %L195 ], [ 1, %L167 ], [ 1, %L161 ], [ %110, %L155 ], [ %100, %L149 ], [ %80, %L128 ]\r\n %value_phi25355 = phi i32 [ 1, %L195 ], [ 1, %L167 ], [ 1, %L161 ], [ 1, %L155 ], [ %98, %L149 ], [ %78, %L128 ]\r\n %value_phi18349 = phi i32 [ 1, %L195 ], [ 1, %L167 ], [ 1, %L161 ], [ 1, %L155 ], [ 1, %L149 ], [ %71, %L128 ]\r\n store i32 %value_phi2326, ptr %union_bytes_return, align 4\r\n %.sroa.0429.sroa.0.sroa.2.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 4\r\n store i32 %value_phi9332, ptr %.sroa.0429.sroa.0.sroa.2.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.0.sroa.3.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 8\r\n store i32 %value_phi10333, ptr %.sroa.0429.sroa.0.sroa.3.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.0.sroa.4.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 12\r\n store i32 %value_phi11334, ptr %.sroa.0429.sroa.0.sroa.4.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.0.sroa.5.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 16\r\n store i32 %value_phi12335, ptr %.sroa.0429.sroa.0.sroa.5.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.0.sroa.6.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 20\r\n store i32 %value_phi13336, ptr %.sroa.0429.sroa.0.sroa.6.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.2.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 24\r\n store i32 %value_phi18349, ptr %.sroa.0429.sroa.2.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.3.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 28\r\n store i32 %value_phi25355, ptr %.sroa.0429.sroa.3.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.4.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 32\r\n store i32 %value_phi26356, ptr %.sroa.0429.sroa.4.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.5.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 36\r\n store i32 %value_phi27357, ptr %.sroa.0429.sroa.5.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.6.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 40\r\n store i32 %value_phi28358, ptr %.sroa.0429.sroa.6.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.7.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 44\r\n store i32 %value_phi29359, ptr %.sroa.0429.sroa.7.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.8.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 48\r\n store i32 %value_phi2326, ptr %.sroa.0429.sroa.8.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.0429.sroa.8.sroa.2.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 52\r\n store i32 %value_phi9332, ptr %.sroa.0429.sroa.8.sroa.2.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx, align 4\r\n %.sroa.0429.sroa.8.sroa.3.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 56\r\n store i32 %value_phi10333, ptr %.sroa.0429.sroa.8.sroa.3.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx, align 4\r\n %.sroa.0429.sroa.8.sroa.4.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 60\r\n store i32 %value_phi11334, ptr %.sroa.0429.sroa.8.sroa.4.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx, align 4\r\n %.sroa.0429.sroa.8.sroa.5.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 64\r\n store i32 %value_phi12335, ptr %.sroa.0429.sroa.8.sroa.5.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx, align 4\r\n %.sroa.0429.sroa.8.sroa.6.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 68\r\n store i32 %value_phi13336, ptr %.sroa.0429.sroa.8.sroa.6.0..sroa.0429.sroa.8.0.union_bytes_return.sroa_idx.sroa_idx, align 4\r\n %.sroa.4430.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 72\r\n store i32 %value_phi18349, ptr %.sroa.4430.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.5431.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 76\r\n store i32 %value_phi25355, ptr %.sroa.5431.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.6432.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 80\r\n store i32 %value_phi26356, ptr %.sroa.6432.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.7433.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 84\r\n store i32 %value_phi27357, ptr %.sroa.7433.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.8434.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 88\r\n store i32 %value_phi28358, ptr %.sroa.8434.0.union_bytes_return.sroa_idx, align 4\r\n %.sroa.9435.0.union_bytes_return.sroa_idx = getelementptr inbounds i8, ptr %union_bytes_return, i32 92\r\n store i32 %value_phi29359, ptr %.sroa.9435.0.union_bytes_return.sroa_idx, align 4\r\n br label %common.ret\r\n}\r\n\r\ndeclare ptr @julia.get_pgcstack()\r\n\r\n; Function Attrs: memory(argmem: readwrite, inaccessiblemem: readwrite)\r\ndeclare void @julia.safepoint(ptr) #2\r\n\r\n; Function Attrs: willreturn memory(argmem: read, inaccessiblemem: readwrite)\r\ndeclare nonnull align 16 dereferenceable(16) ptr addrspace(10) @jl_alloc_genericmemory(ptr addrspace(10), i32) #3\r\n\r\n; Function Attrs: nounwind willreturn allockind(\"alloc\") allocsize(1) memory(argmem: read, inaccessiblemem: readwrite)\r\ndeclare noalias nonnull ptr addrspace(10) @julia.gc_alloc_obj(ptr, i32, ptr addrspace(10)) #4\r\n\r\n; Function Attrs: norecurse nosync nounwind speculatable willreturn memory(none)\r\ndeclare noundef nonnull ptr addrspace(13) @julia.gc_loaded(ptr addrspace(10) nocapture noundef nonnull readnone, ptr noundef nonnull readnone) #5\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)\r\ndeclare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #6\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)\r\ndeclare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #6\r\n```\r\ngives error \r\n```\r\nInstruction does not dominate all uses!\r\n %20 = shufflevector <4 x ptr addrspace(11)> %19, <4 x ptr addrspace(11)> poison, <4 x i32> zeroinitializer\r\n %17 = shufflevector <4 x ptr addrspace(11)> %20, <4 x ptr addrspace(11)> poison, <2 x i32> zeroinitializer\r\nInstruction does not dominate all uses!\r\n %90 = shufflevector <4 x ptr addrspace(11)> %89, <4 x ptr addrspace(11)> poison, <4 x i32> zeroinitializer\r\n %87 = shufflevector <4 x ptr addrspace(11)> %90, <4 x ptr addrspace(11)> poison, <2 x i32> zeroinitializer\r\nLLVM ERROR: Broken module found, compilation aborted!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"verify\" on module \"\"\r\n #0 0x00000000050e6958 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50e6958)\r\n #1 0x00000000050e434c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007508e6a42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007508e6a969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007508e6a42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007508e6a287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00000000007d7105 llvm::json::operator==(llvm::json::Value const&, llvm::json::Value const&) (.cold) JSON.cpp:0:0\r\n #7 0x00000000050254d8 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50254d8)\r\n #8 0x0000000004f23440 llvm::VerifierPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f23440)\r\n #9 0x00000000009000ae llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x9000ae)\r\n#10 0x0000000004ee6850 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ee6850)\r\n#11 0x000000000090adda llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x90adda)\r\n#12 0x00000000008fdd2a optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fdd2a)\r\n#13 0x00007508e6a29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#14 0x00007508e6a29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#15 0x00000000008f57ae _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f57ae)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\nhttps://godbolt.org/z/sah34KE77", "author": "Zentrik", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": [{"author": "Zentrik", "body": "Reduced IR:\r\n```llvm\r\ndefine swiftcc { ptr addrspace(10), i8 } @julia__zip_iterate_all_17381(ptr addrspace(11) %0, i32 %1) {\r\ntop:\r\n %2 = getelementptr i8, ptr addrspace(11) %0, i32 8\r\n %3 = getelementptr i8, ptr addrspace(11) %0, i32 12\r\n %4 = getelementptr i8, ptr addrspace(11) %0, i32 16\r\n %5 = getelementptr i8, ptr addrspace(11) %0, i32 20\r\n br i1 false, label %L42, label %L41\r\n\r\nL41: ; preds = %top\r\n %.not276 = icmp eq ptr addrspace(11) %2, null\r\n %6 = select i1 %.not276, i32 0, i32 0\r\n %.not277 = icmp eq ptr addrspace(11) %3, null\r\n %7 = select i1 %.not277, i32 0, i32 0\r\n %.not278 = icmp eq ptr addrspace(11) %4, null\r\n %8 = select i1 %.not278, i32 0, i32 0\r\n %.not279 = icmp eq ptr addrspace(11) %5, null\r\n %9 = select i1 %.not279, i32 0, i32 0\r\n br label %common.ret\r\n\r\nL42: ; preds = %top\r\n %.not282 = icmp eq ptr addrspace(11) %4, null\r\n %10 = select i1 %.not282, i32 0, i32 0\r\n %.not283 = icmp eq ptr addrspace(11) %5, null\r\n %11 = select i1 %.not283, i32 0, i32 0\r\n br label %common.ret\r\n\r\nL59: ; No predecessors!\r\n %12 = select i1 false, i32 0, i32 0\r\n br label %common.ret\r\n\r\ncommon.ret: ; preds = %L59, %L42, %L41\r\n %value_phi13336 = phi i32 [ %12, %L59 ], [ %11, %L42 ], [ %9, %L41 ]\r\n %value_phi12335 = phi i32 [ 0, %L59 ], [ %10, %L42 ], [ %8, %L41 ]\r\n %value_phi11334 = phi i32 [ 0, %L59 ], [ %1, %L42 ], [ %7, %L41 ]\r\n %value_phi10333 = phi i32 [ 0, %L59 ], [ 0, %L42 ], [ %6, %L41 ]\r\n ret { ptr addrspace(10), i8 } zeroinitializer\r\n}\r\n```\r\nhttps://godbolt.org/z/a8xzqGYo9"}, {"author": "Zentrik", "body": "Bisected to 69332bb8995aef60d830406de12cb79a5039026 and f3d2609af3031ddb54030548e86335f295cf49ca, @alexey-bataev "}]}, "verified": true} {"bug_id": "101337", "issue_url": "https://github.com/llvm/llvm-project/issues/101337", "bug_type": "crash", "base_commit": "1b936e4812ab395328203f65cacd4ffe02bb318e", "knowledge_cutoff": "2024-07-31T14:11:24Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "cd08fadd03904806fa26a1f117879ddae34fbf67", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[7080, 7086]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationPlanner::cost"]}}, "patch": "commit cd08fadd03904806fa26a1f117879ddae34fbf67\nAuthor: Florian Hahn \nDate: Mon Aug 12 14:45:43 2024 +0100\n\n [LV] Include chains feeding inductions in cost precomputation.\n \n Include chain of ops feeding inductions in cost precomputation for\n inductions, not just the induction increment. In VPlan, those\n instructions will be cleaned up, as both phi and increment are generated\n by VPWidenIntOrFpInductionRecipe independently.\n \n Fixes https://github.com/llvm/llvm-project/issues/101337.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 542d74ef0e1e..41e4b0fcd9b0 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -7080,7 +7080,16 @@ InstructionCost LoopVectorizationPlanner::cost(VPlan &Plan,\n for (const auto &[IV, IndDesc] : Legal->getInductionVars()) {\n Instruction *IVInc = cast(\n IV->getIncomingValueForBlock(OrigLoop->getLoopLatch()));\n- SmallVector IVInsts = {IV, IVInc};\n+ SmallVector IVInsts = {IVInc};\n+ for (unsigned I = 0; I != IVInsts.size(); I++) {\n+ for (Value *Op : IVInsts[I]->operands()) {\n+ auto *OpI = dyn_cast(Op);\n+ if (Op == IV || !OpI || !OrigLoop->contains(OpI) || !Op->hasOneUse())\n+ continue;\n+ IVInsts.push_back(OpI);\n+ }\n+ }\n+ IVInsts.push_back(IV);\n for (User *U : IV->users()) {\n auto *CI = cast(U);\n if (!CostCtx.CM.isOptimizableIVTruncate(CI, VF))\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/induction-costs.ll", "commands": ["opt -p loop-vectorize -mtriple=x86_64-apple-macosx -S %s"], "tests": [{"test_name": "iv_and_step_trunc", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine i16 @iv_and_step_trunc() {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %rec = phi i16 [ 0, %entry ], [ %rec.next, %loop ]\n %iv.next = add i64 %iv, 1\n %0 = trunc i64 %iv to i16\n %1 = trunc i64 %iv.next to i16\n %rec.next = mul i16 %0, %1\n %ec = icmp eq i64 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i16 %rec\n}\n"}, {"test_name": "wombat2", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine void @wombat2(i32 %arg, ptr %dst) #0 {\nentry:\n %mul = mul i32 %arg, 3\n %zext = zext i32 %arg to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %phi = phi i64 [ 4, %entry ], [ %add, %loop ]\n %phi2 = phi i32 [ %mul, %entry ], [ %trunc.1, %loop ]\n %getelementptr = getelementptr i32, ptr %dst, i64 %phi\n %and = and i32 %phi2, 12\n store i32 %and, ptr %getelementptr, align 4\n %mul3 = mul i64 %phi, %zext\n %add = add i64 %phi, 1\n %icmp = icmp ugt i64 %phi, 65\n %trunc.0 = trunc i64 %mul3 to i60\n %trunc.1 = trunc i60 %trunc.0 to i32\n br i1 %icmp, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"skylake-avx512\" \"target-features\"=\"-avx512f\" }\n"}, {"test_name": "with_dead_use", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine void @with_dead_use(i32 %arg, ptr %dst) #0 {\nentry:\n %mul = mul i32 %arg, 3\n %zext = zext i32 %arg to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %phi = phi i64 [ 4, %entry ], [ %add, %loop ]\n %phi2 = phi i32 [ %mul, %entry ], [ %trunc, %loop ]\n %getelementptr = getelementptr i32, ptr %dst, i64 %phi\n %and = and i32 %phi2, 12\n store i32 %and, ptr %getelementptr, align 4\n %mul3 = mul i64 %phi, %zext\n %add = add i64 %phi, 1\n %icmp = icmp ugt i64 %phi, 65\n %trunc = trunc i64 %mul3 to i32\n %dead.and = and i32 %trunc, 123\n br i1 %icmp, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"skylake-avx512\" \"target-features\"=\"-avx512f\" }\n"}, {"test_name": "test_scalar_predicated_cost", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine i32 @test_scalar_predicated_cost(i64 %x, i64 %y, ptr %A) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %cmp9.not = icmp ule i64 %iv, %y\n br i1 %cmp9.not, label %loop.latch, label %if.then\n\nif.then: ; preds = %loop.header\n %or = or i64 %x, %iv\n %gep = getelementptr i32, ptr %A, i64 %iv\n %t = trunc i64 %or to i32\n store i32 %t, ptr %gep, align 4\n br label %loop.latch\n\nloop.latch: ; preds = %if.then, %loop.header\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 100\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret i32 0\n}\n\nattributes #0 = { \"min-legal-vector-width\"=\"0\" \"target-cpu\"=\"skylake-avx512\" }\n"}, {"test_name": "truncated_ivs_with_wide_and_scalar_uses", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine void @truncated_ivs_with_wide_and_scalar_uses(i1 %c, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.16 = trunc i64 %iv to i16\n %iv.32 = trunc i64 %iv to i32\n %gep = getelementptr i16, ptr %dst, i32 %iv.32\n %sel = select i1 %c, i16 %iv.16, i16 10\n store i16 %sel, ptr %gep, align 2\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 64\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "multiple_truncated_ivs_with_wide_uses", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine void @multiple_truncated_ivs_with_wide_uses(i1 %c, ptr %A, ptr %B) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.16 = trunc i64 %iv to i16\n %iv.32 = trunc i64 %iv to i32\n %sel = select i1 %c, i16 %iv.16, i16 10\n %gep.A = getelementptr i16, ptr %A, i64 %iv\n store i16 %sel, ptr %gep.A, align 2\n %gep.B = getelementptr i32, ptr %B, i64 %iv\n store i32 %iv.32, ptr %gep.B, align 4\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 64\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "wide_iv_trunc_reuse", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine void @wide_iv_trunc_reuse(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 1, %entry ], [ %iv.next, %loop ]\n %iv.2 = phi i32 [ 0, %entry ], [ %iv.trunc, %loop ]\n store i32 %iv.2, ptr %dst, align 4\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 0\n %iv.trunc = trunc i64 %iv to i32\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "iv_used_widened_and_truncated", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine i32 @iv_used_widened_and_truncated(ptr %dst, i64 %N) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr { i32, [8 x i32] }, ptr %dst, i64 %iv\n %t = trunc i64 %iv to i32\n store i32 %t, ptr %gep, align 8\n %iv.next = add i64 %iv, 1\n %c = icmp eq i64 %iv, %N\n br i1 %c, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i32 0\n}\n\nattributes #0 = { \"min-legal-vector-width\"=\"0\" \"target-cpu\"=\"skylake-avx512\" }\n"}, {"test_name": "wombat", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine void @wombat(i32 %arg, ptr %dst) #0 {\nentry:\n %mul = mul i32 %arg, 3\n %zext = zext i32 %arg to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %phi = phi i64 [ 4, %entry ], [ %add, %loop ]\n %phi2 = phi i32 [ %mul, %entry ], [ %trunc, %loop ]\n %getelementptr = getelementptr i32, ptr %dst, i64 %phi\n %and = and i32 %phi2, 12\n store i32 %and, ptr %getelementptr, align 4\n %mul3 = mul i64 %phi, %zext\n %add = add i64 %phi, 1\n %icmp = icmp ugt i64 %phi, 65\n %trunc = trunc i64 %mul3 to i32\n br i1 %icmp, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"skylake-avx512\" \"target-features\"=\"-avx512f\" }\n"}, {"test_name": "wide_iv_trunc", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine void @wide_iv_trunc(ptr %dst, i64 %N) {\nentry:\n %cmp = icmp ult i64 %N, 8\n br i1 %cmp, label %loop, label %exit\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.trunc = trunc i64 %iv to i32\n store i32 %iv.trunc, ptr %dst, align 4\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, %N\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop, %entry\n ret void\n}\n"}, {"test_name": "multiple_pointer_ivs_with_scalar_uses_only", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\ndefine void @multiple_pointer_ivs_with_scalar_uses_only(ptr %A, ptr %B) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv.1 = phi i32 [ 100, %entry ], [ %dec, %loop ]\n %iv.2 = phi i32 [ 2048, %entry ], [ %add38, %loop ]\n %ptr.iv.1 = phi ptr [ %A, %entry ], [ %outptr.0, %loop ]\n %ptr.iv.2 = phi ptr [ %B, %entry ], [ %incdec.ptr36, %loop ]\n %ptr.iv.3 = phi ptr [ %B, %entry ], [ %incdec.ptr33, %loop ]\n %incdec.ptr33 = getelementptr i8, ptr %ptr.iv.3, i64 1\n %0 = load i8, ptr %ptr.iv.3, align 1\n %conv34 = zext i8 %0 to i32\n %incdec.ptr36 = getelementptr i8, ptr %ptr.iv.2, i64 1\n %1 = load i8, ptr %ptr.iv.2, align 1\n %conv37 = zext i8 %1 to i32\n %add38 = add i32 %conv34, %conv37\n %shr42 = lshr i32 %iv.2, 1\n %conv43 = trunc i32 %shr42 to i8\n store i8 %conv43, ptr %ptr.iv.1, align 1\n %dec = add i32 %iv.1, 1\n %outptr.0 = getelementptr i8, ptr %ptr.iv.1, i64 2\n %cmp30.not = icmp eq i32 %dec, 0\n br i1 %cmp30.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"min-legal-vector-width\"=\"0\" \"target-cpu\"=\"skylake-avx512\" }\n"}]}], "issue": {"title": "Assertion `VF.Width == BestVF && \"VPlan cost model and legacy cost model disagreed\"' failed.", "body": "To reproduce run opt with \r\n```\r\n-passes=loop-vectorize --mcpu=skylake-avx512 --mattr=+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-gfni,-avxvnniint16,-amx-fp16,-zu,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,-nf,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f\r\n```\r\nwith the following test:\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat(i32 %arg) #0 gc \"statepoint-example\" {\r\nbb:\r\n %mul = mul i32 %arg, 3\r\n %zext = zext i32 %arg to i64\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb1, %bb\r\n %phi = phi i64 [ %add, %bb1 ], [ 4, %bb ]\r\n %phi2 = phi i32 [ %trunc, %bb1 ], [ %mul, %bb ]\r\n %getelementptr = getelementptr i32, ptr addrspace(1) null, i64 %phi\r\n %and = and i32 %phi2, 0\r\n store i32 0, ptr addrspace(1) %getelementptr, align 4\r\n %mul3 = mul i64 %phi, %zext\r\n %add = add i64 %phi, 1\r\n %icmp = icmp ugt i64 %phi, 65\r\n %trunc = trunc i64 %mul3 to i32\r\n br i1 %icmp, label %bb4, label %bb1\r\n\r\nbb4: ; preds = %bb1\r\n ret void\r\n}\r\n```\r\nReproducer: https://godbolt.org/z/nPsP7h51n\r\n\r\nStack dump:\r\n```\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=loop-vectorize --mcpu=skylake-avx512 --mattr=+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-gfni,-avxvnniint16,-amx-fp16,-zu,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,-nf,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f \r\n1.\tRunning pass \"function(loop-vectorize)\" on module \"\"\r\n2.\tRunning pass \"loop-vectorize\" on function \"wombat\"\r\n #0 0x0000000004f053c8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f053c8)\r\n #1 0x0000000004f02b3c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007458cc042520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007458cc0969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007458cc042476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007458cc0287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007458cc02871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007458cc039e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000003f83f37 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f83f37)\r\n #9 0x0000000003f86659 llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f86659)\r\n#10 0x0000000003f86cd2 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f86cd2)\r\n#11 0x0000000002e596ce llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2e596ce)\r\n#12 0x0000000004d02758 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d02758)\r\n#13 0x0000000000de089e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xde089e)\r\n#14 0x0000000004d0114e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d0114e)\r\n#15 0x0000000000de0eee llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xde0eee)\r\n#16 0x0000000004cff2b0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4cff2b0)\r\n#17 0x00000000008e3752 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e3752)\r\n#18 0x00000000008d667c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d667c)\r\n#19 0x00007458cc029d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#20 0x00007458cc029e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#21 0x00000000008cdade _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8cdade)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["vectorizers", "crash"], "comments": [{"author": "patrick-rivos", "body": "Potentially related: https://github.com/llvm/llvm-project/issues/100591"}]}, "verified": true} {"bug_id": "113869", "issue_url": "https://github.com/llvm/llvm-project/issues/113869", "bug_type": "miscompilation", "base_commit": "c752efbdcc1ebd81f879633d6b798248064e2ae2", "knowledge_cutoff": "2024-10-28T06:34:27Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "e577f14b670ee2ae6bb717133310b215be4331b3", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[1056, 1062]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["canonicalizeSaturatedAdd"]}}, "patch": "commit e577f14b670ee2ae6bb717133310b215be4331b3\nAuthor: Yingwei Zheng \nDate: Fri Nov 1 22:18:44 2024 +0800\n\n [InstCombine] Use `m_NotForbidPoison` when folding `(X u< Y) ? -1 : (~X + Y) --> uadd.sat(~X, Y)` (#114345)\n \n Alive2: https://alive2.llvm.org/ce/z/mTGCo-\n We cannot reuse `~X` if `m_AllOnes` matches a vector constant with some\n poison elts. An alternative solution is to create a new not instead of\n reusing `~X`. But it doesn't worth the effort because we need to add a\n one-use check.\n \n Fixes https://github.com/llvm/llvm-project/issues/113869.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex 162ebebf1b8f..999ad1adff20 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -1056,7 +1056,7 @@ static Value *canonicalizeSaturatedAdd(ICmpInst *Cmp, Value *TVal, Value *FVal,\n // Strictness of the comparison is irrelevant.\n X = Cmp0;\n Y = Cmp1;\n- if (match(FVal, m_c_Add(m_Not(m_Specific(X)), m_Specific(Y)))) {\n+ if (match(FVal, m_c_Add(m_NotForbidPoison(m_Specific(X)), m_Specific(Y)))) {\n // (X u< Y) ? -1 : (~X + Y) --> uadd.sat(~X, Y)\n // (X u< Y) ? -1 : (Y + ~X) --> uadd.sat(Y, ~X)\n BinaryOperator *BO = cast(FVal);\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/saturating-add-sub.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "uadd_sat_not_commute_select", "test_body": "define i32 @uadd_sat_not_commute_select(i32 %x, i32 %y) {\n %notx = xor i32 %x, -1\n %a = add i32 %notx, %y\n %c = icmp ult i32 %y, %x\n %r = select i1 %c, i32 %a, i32 -1\n ret i32 %r\n}\n"}, {"test_name": "uadd_sat_not_ugt_commute_add", "test_body": "define <2 x i32> @uadd_sat_not_ugt_commute_add(<2 x i32> %x, <2 x i32> %yp) {\n %y = sdiv <2 x i32> %yp, \n %notx = xor <2 x i32> %x, splat (i32 -1)\n %a = add <2 x i32> %y, %notx\n %c = icmp ugt <2 x i32> %y, %x\n %r = select <2 x i1> %c, <2 x i32> splat (i32 -1), <2 x i32> %a\n ret <2 x i32> %r\n}\n"}, {"test_name": "uadd_sat_not_ugt_commute_add_partial_poison", "test_body": "define <2 x i32> @uadd_sat_not_ugt_commute_add_partial_poison(<2 x i32> %x, <2 x i32> %yp) {\n %notx = xor <2 x i32> %x, \n %a = add nuw <2 x i32> %yp, %notx\n %c = icmp ugt <2 x i32> %yp, %x\n %r = select <2 x i1> %c, <2 x i32> splat (i32 -1), <2 x i32> %a\n ret <2 x i32> %r\n}\n"}]}], "issue": {"title": "Poison value wrongly propagates through vector operations on an optimized code", "body": "https://github.com/llvm/llvm-project/blob/5d4a0d54b5269bad1410e6db957836fe98634069/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp#L1050-L1060\r\n\r\nAlive2 report: https://alive2.llvm.org/ce/z/6tSdhb\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine <2 x i32> @uadd_sat_not_ugt_commute_add.2(<2 x i32> %x, <2 x i32> %yp) {\r\n#0:\r\n %#1 = sdiv <2 x i32> { 2442, 4242 }, %yp\r\n %notx = xor <2 x i32> %x, { 4294967295, poison }\r\n %a = add nuw <2 x i32> %#1, %notx\r\n %c = icmp ugt <2 x i32> %#1, %x\r\n %r = select <2 x i1> %c, <2 x i32> { 4294967295, 4294967295 }, <2 x i32> %a\r\n ret <2 x i32> %r\r\n}\r\n=>\r\ndefine <2 x i32> @uadd_sat_not_ugt_commute_add.2(<2 x i32> %x, <2 x i32> %yp) {\r\n#0:\r\n %#1 = sdiv <2 x i32> { 2442, 4242 }, %yp\r\n %notx = xor <2 x i32> %x, { 4294967295, poison }\r\n %r = uadd_sat <2 x i32> %#1, %notx\r\n ret <2 x i32> %r\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\n<2 x i32> %x = < poison, #x00000000 (0) >\r\n<2 x i32> %yp = < #x00000008 (8), #x0000000e (14) >\r\n\r\nSource:\r\n<2 x i32> %#1 = < #x00000131 (305), #x0000012f (303) >\r\n<2 x i32> %notx = < poison, poison >\r\n<2 x i32> %a = < poison, poison >\r\n<2 x i1> %c = < poison, #x1 (1) >\r\n<2 x i32> %r = < poison, #xffffffff (4294967295, -1) >\r\n\r\nTarget:\r\n<2 x i32> %#1 = < #x00000131 (305), #x0000012f (303) >\r\n<2 x i32> %notx = < poison, poison >\r\n<2 x i32> %r = < poison, poison >\r\nSource value: < poison, #xffffffff (4294967295, -1) >\r\nTarget value: < poison, poison >\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "81415", "issue_url": "https://github.com/llvm/llvm-project/issues/81415", "bug_type": "crash", "base_commit": "0ea64ad88abcbb939547a92000d98efb2b00d95f", "knowledge_cutoff": "2024-02-11T14:16:02Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "0dacba3ad1b5214ca35eda14987c0b76568a070e", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": [[977, 985], [990, 1000]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": ["VPlanTransforms::truncateToMinimalBitwidths"]}}, "patch": "commit 0dacba3ad1b5214ca35eda14987c0b76568a070e\nAuthor: Florian Hahn \nDate: Fri Feb 16 12:58:39 2024 +0000\n\n [VPlan] Handle truncating ICMPs in truncateToMinimalBWs.\n \n Update truncateToMinimalBitwidths to handle truncating ICMPs. For ICMPs,\n the new target type will be the same as the original type. In that case,\n only truncate the operands, but skip the extend. This is in line with\n what the original truncateToMinimalBitwidths did for compares.\n \n Fixes https://github.com/llvm/llvm-project/issues/81415.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\nindex 16855a9227a1..99198cfa9955 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n@@ -977,9 +977,6 @@ void VPlanTransforms::truncateToMinimalBitwidths(\n Type *OldResTy = TypeInfo.inferScalarType(ResultVPV);\n unsigned OldResSizeInBits = OldResTy->getScalarSizeInBits();\n assert(OldResTy->isIntegerTy() && \"only integer types supported\");\n- if (OldResSizeInBits == NewResSizeInBits)\n- continue;\n- assert(OldResSizeInBits > NewResSizeInBits && \"Nothing to shrink?\");\n (void)OldResSizeInBits;\n \n auto *NewResTy = IntegerType::get(Ctx, NewResSizeInBits);\n@@ -990,11 +987,17 @@ void VPlanTransforms::truncateToMinimalBitwidths(\n if (auto *VPW = dyn_cast(&R))\n VPW->dropPoisonGeneratingFlags();\n \n- // Extend result to original width.\n- auto *Ext = new VPWidenCastRecipe(Instruction::ZExt, ResultVPV, OldResTy);\n- Ext->insertAfter(&R);\n- ResultVPV->replaceAllUsesWith(Ext);\n- Ext->setOperand(0, ResultVPV);\n+ if (OldResSizeInBits != NewResSizeInBits) {\n+ // Extend result to original width.\n+ auto *Ext =\n+ new VPWidenCastRecipe(Instruction::ZExt, ResultVPV, OldResTy);\n+ Ext->insertAfter(&R);\n+ ResultVPV->replaceAllUsesWith(Ext);\n+ Ext->setOperand(0, ResultVPV);\n+ assert(OldResSizeInBits > NewResSizeInBits && \"Nothing to shrink?\");\n+ } else\n+ assert(cast(&R)->getOpcode() == Instruction::ICmp &&\n+ \"Only ICmps should not need extending the result.\");\n \n if (isa(&R)) {\n assert(!cast(&R)->isStore() && \"stores cannot be narrowed\");\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/trunc-reductions.ll", "commands": ["opt -passes=loop-vectorize,dce,instcombine -force-vector-interleave=1 -force-vector-width=8 -S < %s"], "tests": [{"test_name": "reduction_and_or", "test_body": "define i32 @reduction_and_or(i16 %a, i32 %b, ptr %src) {\nentry:\n %ext1 = zext i16 %a to i32\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %or67 = phi i32 [ 10, %entry ], [ %or, %loop ]\n %t = trunc i32 %b to i16\n %ext = sext i16 %t to i32\n %cmp = icmp sgt i32 %ext, %ext1\n %ext2 = zext i1 %cmp to i32\n %cmp3 = icmp sge i32 %iv, %ext2\n %ext4 = zext i1 %cmp3 to i32\n %div = sdiv i32 %ext4, %b\n %and = and i32 %div, 0\n %gep = getelementptr inbounds i32, ptr %src, i32 %iv\n %l = load i32, ptr %gep, align 4\n %add = add i32 %and, %l\n %or = or i32 %or67, %add\n %iv.next = add nsw i32 %iv, 1\n %tobool.not = icmp eq i32 %iv.next, 999\n br i1 %tobool.not, label %exit, label %loop\n\nexit: ; preds = %loop\n %or.lcssa = phi i32 [ %or, %loop ]\n ret i32 %or.lcssa\n}\n"}]}], "issue": {"title": "clang crashes at -O{s,2,3} on x86_64-linux-gnu: Assertion `MinBWs.size() == NumProcessedRecipes && \"some entries in MinBWs haven't been processed\"' failed", "body": "It appears to be a recent regression as it does not reproduce with 17.0.1 and earlier versions.\r\n\r\nCompiler Explorer: https://godbolt.org/z/vGGoa684a\r\n\r\n```\r\n[511] % clangtk -v\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git 5932fcc47855fdd209784f38820422d2369b84b2)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/opfuzz/bin\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[512] % \r\n[512] % clangtk -Os small.c\r\nclang-19: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp:1038: static void llvm::VPlanTransforms::truncateToMinimalBitwidths(llvm::VPlan&, const llvm::MapVector&, llvm::LLVMContext&): Assertion `MinBWs.size() == NumProcessedRecipes && \"some entries in MinBWs haven't been processed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-19 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20240211-clangtk-m64-Os-Weverything-pipe-fPIC-build-024130/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20240211-clangtk-m64-Os-Weverything-pipe-fPIC-build-024130/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/19 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/19/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -Os -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-d74fff.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n #0 0x000055fb61397fcf llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x3f04fcf)\r\n #1 0x000055fb613954f4 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007fdd88b17420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007fdd8851e00b raise /build/glibc-wuryBv/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007fdd884fd859 abort /build/glibc-wuryBv/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007fdd884fd729 get_sysdep_segment_value /build/glibc-wuryBv/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007fdd884fd729 _nl_load_domain /build/glibc-wuryBv/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007fdd8850efd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x000055fb62c44895 llvm::VPlanTransforms::truncateToMinimalBitwidths(llvm::VPlan&, llvm::MapVector, llvm::detail::DenseMapPair>, llvm::SmallVector, 0u>> const&, llvm::LLVMContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x57b1895)\r\n #9 0x000055fb62b2644a llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x569344a)\r\n#10 0x000055fb62b2d861 llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x569a861)\r\n#11 0x000055fb62b30a1f llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x569da1f)\r\n#12 0x000055fb62b33acd llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x56a0acd)\r\n#13 0x000055fb62b34d20 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x56a1d20)\r\n#14 0x000055fb62717536 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x5284536)\r\n#15 0x000055fb5ea6d61d llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x15da61d)\r\n#16 0x000055fb60d5cb3d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x38c9b3d)\r\n#17 0x000055fb5ea509a6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x15bd9a6)\r\n#18 0x000055fb60d5aa6d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x38c7a6d)\r\n#19 0x000055fb6161dc90 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#20 0x000055fb61620f60 (anonymous namespace)::EmitAssemblyHelper::EmitAssembly(clang::BackendAction, std::unique_ptr>, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#21 0x000055fb616215cb clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x418e5cb)\r\n#22 0x000055fb61c5730a clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x47c430a)\r\n#23 0x000055fb63944ee9 clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x64b1ee9)\r\n#24 0x000055fb61c566a8 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x47c36a8)\r\n#25 0x000055fb61ed4429 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x4a41429)\r\n#26 0x000055fb61e508de clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x49bd8de)\r\n#27 0x000055fb61fb8d46 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x4b25d46)\r\n#28 0x000055fb5e655986 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x11c2986)\r\n#29 0x000055fb5e64cfba ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#30 0x000055fb5e652434 clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x11bf434)\r\n#31 0x000055fb5e564aab main (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x10d1aab)\r\n#32 0x00007fdd884ff083 __libc_start_main /build/glibc-wuryBv/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#33 0x000055fb5e64ca7e _start (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x11b9a7e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git 5932fcc47855fdd209784f38820422d2369b84b2)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/opfuzz/bin\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-ca7a0b.c\r\nclangtk: note: diagnostic msg: /tmp/small-ca7a0b.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[513] % \r\n[513] % cat small.c\r\nstatic int *a, **b = &a, f = 1;\r\nint c, d, h;\r\nshort e, g;\r\nint main() {\r\n int **i[9] = {&a, &a, &a, &a, &a, &a, &a, &a, &a};\r\n for (; d; d++)\r\n if ((**b < e) | f) {\r\n int *j = &f, k;\r\n c = h;\r\n k = g > ((long)j == 1);\r\n }\r\n return 0;\r\n}\r\n```", "author": "zhendongsu", "labels": ["vectorizers", "crash"], "comments": [{"author": "zhendongsu", "body": "Another related test that reproduces the crash only at -Os.\r\n\r\n```\r\n[513] % clangtk -v\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git e769fb8699e3fa8e40623764f7713bfc783b0330)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/opfuzz/bin\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[514] % \r\n[514] % clangtk -Os small.c\r\nclang-19: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp:1038: static void llvm::VPlanTransforms::truncateToMinimalBitwidths(llvm::VPlan&, const llvm::MapVector&, llvm::LLVMContext&): Assertion `MinBWs.size() == NumProcessedRecipes && \"some entries in MinBWs haven't been processed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-19 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/DUP/20240215-clangtk-m64-Os-Weverything-pipe-fPIC-build-144944/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/DUP/20240215-clangtk-m64-Os-Weverything-pipe-fPIC-build-144944/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/19 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/19/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -Os -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-a3d3db.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n #0 0x000055bf09b51e2f llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x3f38e2f)\r\n #1 0x000055bf09b4f354 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007fdb52bbc420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007fdb525c300b raise /build/glibc-wuryBv/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007fdb525a2859 abort /build/glibc-wuryBv/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007fdb525a2729 get_sysdep_segment_value /build/glibc-wuryBv/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007fdb525a2729 _nl_load_domain /build/glibc-wuryBv/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007fdb525b3fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x000055bf0b405565 llvm::VPlanTransforms::truncateToMinimalBitwidths(llvm::VPlan&, llvm::MapVector, llvm::detail::DenseMapPair>, llvm::SmallVector, 0u>> const&, llvm::LLVMContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x57ec565)\r\n #9 0x000055bf0b2e62ba llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x56cd2ba)\r\n#10 0x000055bf0b2ed661 llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x56d4661)\r\n#11 0x000055bf0b2f081f llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x56d781f)\r\n#12 0x000055bf0b2f38cd llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x56da8cd)\r\n#13 0x000055bf0b2f4b20 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x56dbb20)\r\n#14 0x000055bf0aed6086 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x52bd086)\r\n#15 0x000055bf0721ff8d llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x1606f8d)\r\n#16 0x000055bf0951611d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x38fd11d)\r\n#17 0x000055bf072032e6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x15ea2e6)\r\n#18 0x000055bf0951404d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x38fb04d)\r\n#19 0x000055bf09dd88c0 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#20 0x000055bf09ddbb80 (anonymous namespace)::EmitAssemblyHelper::EmitAssembly(clang::BackendAction, std::unique_ptr>, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#21 0x000055bf09ddc1eb clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x41c31eb)\r\n#22 0x000055bf0a41298a clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x47f998a)\r\n#23 0x000055bf0c10e439 clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x64f5439)\r\n#24 0x000055bf0a411d18 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x47f8d18)\r\n#25 0x000055bf0a6903e9 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x4a773e9)\r\n#26 0x000055bf0a60d21e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x49f421e)\r\n#27 0x000055bf0a776736 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x4b5d736)\r\n#28 0x000055bf06e06b76 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x11edb76)\r\n#29 0x000055bf06dfe1aa ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#30 0x000055bf06e03624 clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x11ea624)\r\n#31 0x000055bf06d15abb main (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x10fcabb)\r\n#32 0x00007fdb525a4083 __libc_start_main /build/glibc-wuryBv/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#33 0x000055bf06dfdc6e _start (/local/suz-local/software/local/clang-trunk/bin/clang-19+0x11e4c6e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git e769fb8699e3fa8e40623764f7713bfc783b0330)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/opfuzz/bin\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-585ede.c\r\nclangtk: note: diagnostic msg: /tmp/small-585ede.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[515] % \r\n[515] % cat small.c\r\nint a = 1, b, c, d;\r\nshort e() { return a == 1 ? a : a % b; }\r\nstatic int f;\r\nshort g;\r\nunsigned short h;\r\nint main() {\r\n int *i[10] = {&f};\r\n for (c = -2; c; c++) {\r\n g = e();\r\n d |= f & ((b >= (g > h)) / a);\r\n }\r\n return 0;\r\n}\r\n```"}, {"author": "fhahn", "body": "Thanks for the test case, should be fixed! \r\n\r\nIt looks like this only gets triggered when the resulting truncated compare is dead. Please let me know if you see any issues with similar cases where the resulting loop/compares in the loop aren't dead."}]}, "verified": true} {"bug_id": "92887", "issue_url": "https://github.com/llvm/llvm-project/issues/92887", "bug_type": "miscompilation", "base_commit": "f78b1a40864470f7eb30dfda44a6ea8e845346a5", "knowledge_cutoff": "2024-05-21T09:38:14Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "263224e4481b0850539a93a272184aac4abe86d4", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp": [[2901, 2907]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp": ["InstCombinerImpl::visitShuffleVectorInst"]}}, "patch": "commit 263224e4481b0850539a93a272184aac4abe86d4\nAuthor: Nikita Popov \nDate: Tue May 21 14:57:19 2024 +0200\n\n [InstCombine] Require poison operand in canEvaluateShuffled transform\n \n This transform works on single-source shuffles, which require that\n the second operand is poison, not undef. Otherwise we may convert\n undef to poison.\n \n Fixes https://github.com/llvm/llvm-project/issues/92887.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\nindex 99f1f8eb34bb..745ccbfe9dc7 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\n@@ -2901,7 +2901,7 @@ Instruction *InstCombinerImpl::visitShuffleVectorInst(ShuffleVectorInst &SVI) {\n if (Instruction *I = foldIdentityPaddedShuffles(SVI))\n return I;\n \n- if (match(RHS, m_Undef()) && canEvaluateShuffled(LHS, Mask)) {\n+ if (match(RHS, m_Poison()) && canEvaluateShuffled(LHS, Mask)) {\n Value *V = evaluateInDifferentElementOrder(LHS, Mask, Builder);\n return replaceInstUsesWith(SVI, V);\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/vec_shuffle.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "pr92887", "test_body": "define i16 @pr92887(<2 x i16> %v) {\n %v0 = extractelement <2 x i16> %v, i64 0\n %v0lo = and i16 %v0, 1\n %v1 = extractelement <2 x i16> %v, i64 1\n %v1lo = and i16 %v1, 1\n %ins1 = insertelement <4 x i16> poison, i16 %v0lo, i64 0\n %ins2 = insertelement <4 x i16> %ins1, i16 %v1lo, i64 1\n %shuf = shufflevector <4 x i16> %ins2, <4 x i16> , <4 x i32> \n %extract = extractelement <4 x i16> %shuf, i32 2\n ret i16 %extract\n}\n"}, {"test_name": "blend_elements_from_load", "test_body": "define <4 x i16> @blend_elements_from_load(ptr align 8 %_0) {\n %load = load <3 x i16>, ptr %_0, align 8\n %rv = shufflevector <3 x i16> , <3 x i16> %load, <4 x i32> \n ret <4 x i16> %rv\n}\n"}]}], "issue": {"title": "Miscompile with opt -passes=\"instcombine\"", "body": "llvm commit: 502bea25bdc07d1\r\n\r\nReproduce with:\r\n```opt -passes=\"instcombine\" bbi-95633_2.ll -S -o -```\r\n\r\nResult:\r\n```\r\ndefine i16 @foo() {\r\nentry:\r\n store i16 0, ptr @bit, align 2\r\n store i16 0, ptr getelementptr inbounds (i8, ptr @bit, i64 2), align 2\r\n store i16 0, ptr getelementptr inbounds (i8, ptr @bit, i64 4), align 2\r\n store i16 0, ptr getelementptr inbounds (i8, ptr @bit, i64 6), align 2\r\n store i16 2, ptr getelementptr inbounds (i8, ptr @bit, i64 8), align 2\r\n store i16 0, ptr getelementptr inbounds (i8, ptr @bit, i64 10), align 2\r\n ret i16 poison\r\n}\r\n```\r\nNote the poison return value which I think is wrong.\r\n\r\nBefore commit e93d324adb4ff8\r\n```\r\n[InstCombine] Preserve poison in evaluateInDifferentElementOrder()\r\n\r\nDon't unnecessarily replace poison with undef.\r\n```\r\nwe instead got\r\n```\r\ndefine i16 @foo() {\r\nentry:\r\n store i16 0, ptr @bit, align 2\r\n store i16 0, ptr getelementptr inbounds (%struct.bit, ptr @bit, i64 0, i32 1), align 2\r\n store i16 0, ptr getelementptr inbounds (%struct.bit, ptr @bit, i64 0, i32 2), align 2\r\n store i16 0, ptr getelementptr inbounds (%struct.bit, ptr @bit, i64 0, i32 3), align 2\r\n store i16 2, ptr getelementptr inbounds (%struct.bit, ptr @bit, i64 0, i32 4), align 2\r\n store i16 0, ptr getelementptr inbounds (%struct.bit, ptr @bit, i64 0, i32 5), align 2\r\n ret i16 1\r\n}\r\n```\r\nwhere we return 1 instead of poison.\r\n\r\nI tried to use Alive2 on the input/output in\r\n https://alive2.llvm.org/ce/z/1GVkRh\r\nand it says\r\n```\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n```\r\n\r\n[bbi-95633_2.ll.gz](https://github.com/llvm/llvm-project/files/15387836/bbi-95633_2.ll.gz)\r\n", "author": "mikaelholmen", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "mikaelholmen", "body": "@nikic : Any idea about this?"}, {"author": "artagnon", "body": "Here's a reduced test case:\r\n\r\n```llvm\r\n; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5\r\n; RUN: opt -passes=instcombine -S %s | FileCheck %s\r\n\r\n%struct.widget = type { i16, i16, i16, i16, i16, i16 }\r\n\r\n; Miscompile. Correct output: 1, incorrect output: poison.\r\ndefine i16 @pr92887(ptr %p, i16 %x, i16 %y) {\r\n; CHECK-LABEL: define i16 @pr92887(\r\n; CHECK-SAME: ptr [[P:%.*]], i16 [[X:%.*]], i16 [[Y:%.*]]) {\r\n; CHECK-NEXT: store i16 0, ptr [[P]], align 2\r\n; CHECK-NEXT: [[P_REPACK1:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 2\r\n; CHECK-NEXT: store i16 0, ptr [[P_REPACK1]], align 2\r\n; CHECK-NEXT: [[P_REPACK2:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 4\r\n; CHECK-NEXT: store i16 0, ptr [[P_REPACK2]], align 2\r\n; CHECK-NEXT: [[P_REPACK3:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 6\r\n; CHECK-NEXT: store i16 0, ptr [[P_REPACK3]], align 2\r\n; CHECK-NEXT: [[P_REPACK4:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 8\r\n; CHECK-NEXT: store i16 2, ptr [[P_REPACK4]], align 2\r\n; CHECK-NEXT: [[P_REPACK5:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 10\r\n; CHECK-NEXT: store i16 0, ptr [[P_REPACK5]], align 2\r\n; CHECK-NEXT: ret i16 poison\r\n;\r\n store %struct.widget { i16 0, i16 0, i16 0, i16 0, i16 2, i16 0 }, ptr %p\r\n %insert.y = insertelement <2 x i16> zeroinitializer, i16 %y, i64 0\r\n %insert.x = insertelement <2 x i16> %insert.y, i16 %x, i64 1\r\n %shufflevector = shufflevector <2 x i16> %insert.x, <2 x i16> zeroinitializer, <4 x i32> \r\n %extract = extractelement <4 x i16> %shufflevector, i32 2 ; undef\r\n %and = and i16 %extract, 0 ; 0\r\n %gep = getelementptr %struct.widget, ptr %p, i32 0, i32 4\r\n %load = load i16, ptr %gep ; 2\r\n %or = or i16 %and, %load ; 2\r\n %ashr = ashr i16 %or, 1 ; 1\r\n ret i16 %ashr ; 1\r\n}\r\n```\r\n\r\nEDIT: Sorry, Alive2 verifies the transformation as correct."}, {"author": "nikic", "body": "The problematic transform is:\r\n```llvm\r\ndefine i16 @src(<2 x i16> %v) {\r\n %v0 = extractelement <2 x i16> %v, i64 0\r\n %v0lo = and i16 %v0, 1\r\n %v1 = extractelement <2 x i16> %v, i64 1\r\n %v1lo = and i16 %v1, 1\r\n %ins1 = insertelement <4 x i16> poison, i16 %v0lo, i64 0\r\n %ins2 = insertelement <4 x i16> %ins1, i16 %v1lo, i64 1\r\n %shuf = shufflevector <4 x i16> %ins2, <4 x i16> , <4 x i32> \r\n %extract = extractelement <4 x i16> %shuf, i32 2\r\n ret i16 %extract\r\n}\r\n```\r\nTo:\r\n```\r\ndefine i16 @src(<2 x i16> %v) {\r\n %v0 = extractelement <2 x i16> %v, i64 0\r\n %v0lo = and i16 %v0, 1\r\n %v1 = extractelement <2 x i16> %v, i64 1\r\n %v1lo = and i16 %v1, 1\r\n %1 = insertelement <4 x i16> poison, i16 %v0lo, i64 0\r\n %shuf = insertelement <4 x i16> %1, i16 %v1lo, i64 1\r\n %extract = extractelement <4 x i16> %shuf, i32 2\r\n ret i16 %extract\r\n}\r\n```\r\n\r\nAt a glance, the problem is likely that https://github.com/llvm/llvm-project/blob/e93d324adb4ff80dcbf0e5b678ed3342350eb2de/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp#L2881 should be matching m_Poison, not m_Undef, otherwise it's not a real single-source shuffle."}, {"author": "mikaelholmen", "body": "Thanks for the quick fix!\r\nI've verified that it solves the problem we saw."}, {"author": "nikic", "body": "As a bonus, fixed a similar miscompile in https://github.com/llvm/llvm-project/commit/ecd269e8305330c185bbedbd5a59e887122333ba."}, {"author": "nikic", "body": "Another one in https://github.com/llvm/llvm-project/commit/8f1c984325bd679b2634a6173db69548da87ac71."}, {"author": "nikic", "body": "And... one more in https://github.com/llvm/llvm-project/commit/d0e0205bfc5a147f8744a176a10f185af7520c26."}, {"author": "mikaelholmen", "body": "Nice, four fixes from just one issue! :)"}]}, "verified": true} {"bug_id": "110382", "issue_url": "https://github.com/llvm/llvm-project/issues/110382", "bug_type": "crash", "base_commit": "72a957ba4c8ef059f1572f6d4ee0cba8dc615268", "knowledge_cutoff": "2024-09-28T19:00:52Z", "lit_test_dir": ["llvm/test/Transforms/VectorCombine"], "hints": {"fix_commit": "c136d3237a3c6230cfe1ab3f0f6790f903c54a27", "components": ["VectorCombine"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": [[1984, 2016], [2038, 2044]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": ["VectorCombine::foldShuffleToIdentity"]}}, "patch": "commit c136d3237a3c6230cfe1ab3f0f6790f903c54a27\nAuthor: David Green \nDate: Wed Oct 9 16:20:03 2024 +0100\n\n [VectorCombine] Do not try to operate on OperandBundles. (#111635)\n \n This bails out if we see an intrinsic with an operand bundle on it, to\n make sure we don't process the bundles incorrectly.\n \n Fixes #110382.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\nindex a2ab5d966640..627edb680dfa 100644\n--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n@@ -1984,33 +1984,35 @@ bool VectorCombine::foldShuffleToIdentity(Instruction &I) {\n \n // We need each element to be the same type of value, and check that each\n // element has a single use.\n- if (all_of(drop_begin(Item), [Item](InstLane IL) {\n- Value *FrontV = Item.front().first->get();\n- if (!IL.first)\n- return true;\n- Value *V = IL.first->get();\n- if (auto *I = dyn_cast(V); I && !I->hasOneUse())\n- return false;\n- if (V->getValueID() != FrontV->getValueID())\n- return false;\n- if (auto *CI = dyn_cast(V))\n- if (CI->getPredicate() != cast(FrontV)->getPredicate())\n- return false;\n- if (auto *CI = dyn_cast(V))\n- if (CI->getSrcTy() != cast(FrontV)->getSrcTy())\n- return false;\n- if (auto *SI = dyn_cast(V))\n- if (!isa(SI->getOperand(0)->getType()) ||\n- SI->getOperand(0)->getType() !=\n- cast(FrontV)->getOperand(0)->getType())\n- return false;\n- if (isa(V) && !isa(V))\n- return false;\n- auto *II = dyn_cast(V);\n- return !II || (isa(FrontV) &&\n- II->getIntrinsicID() ==\n- cast(FrontV)->getIntrinsicID());\n- })) {\n+ auto CheckLaneIsEquivalentToFirst = [Item](InstLane IL) {\n+ Value *FrontV = Item.front().first->get();\n+ if (!IL.first)\n+ return true;\n+ Value *V = IL.first->get();\n+ if (auto *I = dyn_cast(V); I && !I->hasOneUse())\n+ return false;\n+ if (V->getValueID() != FrontV->getValueID())\n+ return false;\n+ if (auto *CI = dyn_cast(V))\n+ if (CI->getPredicate() != cast(FrontV)->getPredicate())\n+ return false;\n+ if (auto *CI = dyn_cast(V))\n+ if (CI->getSrcTy() != cast(FrontV)->getSrcTy())\n+ return false;\n+ if (auto *SI = dyn_cast(V))\n+ if (!isa(SI->getOperand(0)->getType()) ||\n+ SI->getOperand(0)->getType() !=\n+ cast(FrontV)->getOperand(0)->getType())\n+ return false;\n+ if (isa(V) && !isa(V))\n+ return false;\n+ auto *II = dyn_cast(V);\n+ return !II || (isa(FrontV) &&\n+ II->getIntrinsicID() ==\n+ cast(FrontV)->getIntrinsicID() &&\n+ !II->hasOperandBundles());\n+ };\n+ if (all_of(drop_begin(Item), CheckLaneIsEquivalentToFirst)) {\n // Check the operator is one that we support.\n if (isa(FrontU)) {\n // We exclude div/rem in case they hit UB from poison lanes.\n@@ -2038,7 +2040,8 @@ bool VectorCombine::foldShuffleToIdentity(Instruction &I) {\n Worklist.push_back(generateInstLaneVectorFromOperand(Item, 2));\n continue;\n } else if (auto *II = dyn_cast(FrontU);\n- II && isTriviallyVectorizable(II->getIntrinsicID())) {\n+ II && isTriviallyVectorizable(II->getIntrinsicID()) &&\n+ !II->hasOperandBundles()) {\n for (unsigned Op = 0, E = II->getNumOperands() - 1; Op < E; Op++) {\n if (isVectorIntrinsicWithScalarOpAtArg(II->getIntrinsicID(), Op)) {\n if (!all_of(drop_begin(Item), [Item, Op](InstLane &IL) {\n", "tests": [{"file": "llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll", "commands": ["opt -passes=vector-combine -S %s"], "tests": [{"test_name": "add_same_operands", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i32> @add_same_operands(<4 x i32> %x) {\n %shuf = shufflevector <4 x i32> %x, <4 x i32> poison, <4 x i32> \n %add = add <4 x i32> %shuf, %shuf\n %revshuf = shufflevector <4 x i32> %add, <4 x i32> poison, <4 x i32> \n ret <4 x i32> %revshuf\n}\n"}, {"test_name": "extrause_add_same_operands", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i32> @extrause_add_same_operands(<4 x i32> %x) {\n %shuf = shufflevector <4 x i32> %x, <4 x i32> poison, <4 x i32> \n %add = add <4 x i32> %shuf, %shuf\n %revshuf = shufflevector <4 x i32> %add, <4 x i32> poison, <4 x i32> \n %add2 = add <4 x i32> %shuf, %revshuf\n ret <4 x i32> %add2\n}\n"}, {"test_name": "callinst2", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @callinst2(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = call <4 x half> @llvm.fabs.v4f16(<4 x half> %at)\n %abb = call <4 x half> @othercall(<4 x half> %ab)\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n\ndeclare <4 x half> @othercall(<4 x half>)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x half> @llvm.fabs.v4f16(<4 x half>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "poison_intrinsic", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i32> @poison_intrinsic(<2 x i16> %l256) {\n %l266 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> %l256, i1 false)\n %l267 = zext <2 x i16> %l266 to <2 x i32>\n %l271 = shufflevector <2 x i32> %l267, <2 x i32> poison, <4 x i32> \n ret <4 x i32> %l271\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i16> @llvm.abs.v2i16(<2 x i16>, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "splattwice", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @splattwice(<8 x half> %a, <8 x half> %b) {\n %as = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> zeroinitializer\n %bs = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> zeroinitializer\n %ab1 = fadd <4 x half> %as, %bs\n %ab2 = fadd <4 x half> %as, %bs\n %r = shufflevector <4 x half> %ab1, <4 x half> %ab2, <8 x i32> \n ret <8 x half> %r\n}\n"}, {"test_name": "fadd_mismatched_types", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x float> @fadd_mismatched_types(<4 x float> %x, <4 x float> %y) {\n %shuf.x = shufflevector <4 x float> %x, <4 x float> poison, <2 x i32> \n %shuf.y = shufflevector <4 x float> %y, <4 x float> poison, <2 x i32> \n %fadd = fadd fast <2 x float> %shuf.x, %shuf.y\n %extshuf = shufflevector <2 x float> %fadd, <2 x float> poison, <4 x i32> \n ret <4 x float> %extshuf\n}\n"}, {"test_name": "intrinsics_minmax", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @intrinsics_minmax(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.smin.v4i8(<4 x i8> %at, <4 x i8> %bt)\n %abb = call <4 x i8> @llvm.smin.v4i8(<4 x i8> %ab, <4 x i8> %bb)\n %abt1 = call <4 x i8> @llvm.smax.v4i8(<4 x i8> %abt, <4 x i8> %bt)\n %abb1 = call <4 x i8> @llvm.smax.v4i8(<4 x i8> %abb, <4 x i8> %bb)\n %abt2 = call <4 x i8> @llvm.umin.v4i8(<4 x i8> %abt1, <4 x i8> %bt)\n %abb2 = call <4 x i8> @llvm.umin.v4i8(<4 x i8> %abb1, <4 x i8> %bb)\n %abt3 = call <4 x i8> @llvm.umax.v4i8(<4 x i8> %abt2, <4 x i8> %bt)\n %abb3 = call <4 x i8> @llvm.umax.v4i8(<4 x i8> %abb2, <4 x i8> %bb)\n %r = shufflevector <4 x i8> %abt3, <4 x i8> %abb3, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.smax.v4i8(<4 x i8>, <4 x i8>) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.smin.v4i8(<4 x i8>, <4 x i8>) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.umax.v4i8(<4 x i8>, <4 x i8>) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.umin.v4i8(<4 x i8>, <4 x i8>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "constantdiff2", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @constantdiff2(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = add <4 x i8> %at, \n %abb = add <4 x i8> %ab, \n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "fcmpsel", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @fcmpsel(<8 x half> %a, <8 x half> %b, <8 x i8> %c, <8 x i8> %d) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %bb = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> \n %bt = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> \n %cb = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> \n %ct = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> \n %db = shufflevector <8 x i8> %d, <8 x i8> poison, <4 x i32> \n %dt = shufflevector <8 x i8> %d, <8 x i8> poison, <4 x i32> \n %abt1 = fcmp olt <4 x half> %at, %bt\n %abb1 = fcmp olt <4 x half> %ab, %bb\n %abt = select <4 x i1> %abt1, <4 x i8> %ct, <4 x i8> %dt\n %abb = select <4 x i1> %abb1, <4 x i8> %cb, <4 x i8> %db\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "inner_shuffle", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @inner_shuffle(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %cs = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> zeroinitializer\n %abt = mul <4 x i8> %at, %bt\n %abb = mul <4 x i8> %ab, %bb\n %abt2 = shufflevector <4 x i8> %abt, <4 x i8> poison, <4 x i32> \n %abb2 = shufflevector <4 x i8> %abb, <4 x i8> poison, <4 x i32> \n %abt3 = add <4 x i8> %abt2, %cs\n %abb3 = add <4 x i8> %abb2, %cs\n %r = shufflevector <4 x i8> %abt3, <4 x i8> %abb3, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "zext_types", "test_body": "target triple = \"aarch64\"\n\ndefine void @zext_types(<8 x i16> %a, <8 x i32> %b, ptr %p) {\n %ab = shufflevector <8 x i16> %a, <8 x i16> poison, <4 x i32> \n %at = shufflevector <8 x i32> %b, <8 x i32> poison, <4 x i32> \n %ab1 = zext <4 x i16> %ab to <4 x i64>\n %at1 = zext <4 x i32> %at to <4 x i64>\n %r = shufflevector <4 x i64> %ab1, <4 x i64> %at1, <8 x i32> \n store <8 x i64> %r, ptr %p, align 64\n ret void\n}\n"}, {"test_name": "add_different_operands", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @add_different_operands(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = add <4 x i8> %at, %bt\n %abb = add <4 x i8> %ab, %bb\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "constantdiff", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @constantdiff(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = add <4 x i8> %at, \n %abb = add <4 x i8> %ab, \n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "extrause_add_different_operands", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @extrause_add_different_operands(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = add <4 x i8> %at, %bt\n %abb = add <4 x i8> %ab, %bb\n call void @use(<4 x i8> %abb)\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\ndeclare void @use(<4 x i8>)\n"}, {"test_name": "icmpsel_diffentcond", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @icmpsel_diffentcond(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c, <8 x i8> %d) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %cb = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> \n %ct = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> \n %db = shufflevector <8 x i8> %d, <8 x i8> poison, <4 x i32> \n %dt = shufflevector <8 x i8> %d, <8 x i8> poison, <4 x i32> \n %abt1 = icmp slt <4 x i8> %at, %bt\n %abb1 = icmp ult <4 x i8> %ab, %bb\n %abt = select <4 x i1> %abt1, <4 x i8> %ct, <4 x i8> %dt\n %abb = select <4 x i1> %abb1, <4 x i8> %cb, <4 x i8> %db\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "exttrunc", "test_body": "target triple = \"aarch64\"\n\ndefine void @exttrunc(<8 x i32> %a, <8 x i32> %b, ptr %p) {\n %ab = shufflevector <8 x i32> %a, <8 x i32> poison, <4 x i32> \n %at = shufflevector <8 x i32> %a, <8 x i32> poison, <4 x i32> \n %bb = shufflevector <8 x i32> %b, <8 x i32> poison, <4 x i32> \n %bt = shufflevector <8 x i32> %b, <8 x i32> poison, <4 x i32> \n %ab1 = zext <4 x i32> %ab to <4 x i64>\n %at1 = zext <4 x i32> %at to <4 x i64>\n %bb1 = sext <4 x i32> %bb to <4 x i64>\n %bt1 = sext <4 x i32> %bt to <4 x i64>\n %abb = add <4 x i64> %ab1, %bb1\n %abt = add <4 x i64> %at1, %bt1\n %abb1 = trunc <4 x i64> %abb to <4 x i32>\n %abt1 = trunc <4 x i64> %abt to <4 x i32>\n %r = shufflevector <4 x i32> %abb1, <4 x i32> %abt1, <8 x i32> \n store <8 x i32> %r, ptr %p, align 32\n ret void\n}\n"}, {"test_name": "lrint", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i32> @lrint(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = call <4 x i32> @llvm.lrint.v4i32.v4f16(<4 x half> %at)\n %abb = call <4 x i32> @llvm.lrint.v4i32.v4f16(<4 x half> %ab)\n %r = shufflevector <4 x i32> %abt, <4 x i32> %abb, <8 x i32> \n ret <8 x i32> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i32> @llvm.lrint.v4i32.v4f16(<4 x half>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "fadd", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @fadd(<8 x half> %a, <8 x half> %b) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %bb = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> \n %bt = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> \n %abt = fadd <4 x half> %at, %bt\n %abb = fadd <4 x half> %ab, %bb\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n"}, {"test_name": "abs_different", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @abs_different(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %at, i1 true)\n %abb = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %ab, i1 false)\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.abs.v4i8(<4 x i8>, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "wrong_addsub", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @wrong_addsub(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = add <4 x i8> %at, %bt\n %abb = sub <4 x i8> %ab, %bb\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "single_zext", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i64> @single_zext(<4 x i32> %x) {\n %shuf = shufflevector <4 x i32> %x, <4 x i32> poison, <4 x i32> \n %zext = zext <4 x i32> %shuf to <4 x i64>\n %revshuf = shufflevector <4 x i64> %zext, <4 x i64> poison, <4 x i32> \n ret <4 x i64> %revshuf\n}\n"}, {"test_name": "v8f64interleave", "test_body": "target triple = \"aarch64\"\n\ndefine void @v8f64interleave(i64 %0, ptr %1, ptr %x, double %z) {\nentry:\n %broadcast.splatinsert = insertelement <2 x double> poison, double %z, i64 0\n %broadcast.splat = shufflevector <2 x double> %broadcast.splatinsert, <2 x double> poison, <2 x i32> zeroinitializer\n %wide.vec = load <16 x double>, ptr %1, align 8\n %strided.vec = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> \n %strided.vec27 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> \n %strided.vec28 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> \n %strided.vec29 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> \n %strided.vec30 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> \n %strided.vec31 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> \n %strided.vec32 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> \n %strided.vec33 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> \n %2 = fmul fast <2 x double> %strided.vec, %broadcast.splat\n %3 = getelementptr inbounds double, ptr %x, i64 %0\n %wide.vec34 = load <16 x double>, ptr %3, align 8\n %strided.vec35 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> \n %strided.vec36 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> \n %strided.vec37 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> \n %strided.vec38 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> \n %strided.vec39 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> \n %strided.vec40 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> \n %strided.vec41 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> \n %strided.vec42 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> \n %4 = fadd fast <2 x double> %strided.vec35, %2\n %5 = fmul fast <2 x double> %strided.vec27, %broadcast.splat\n %6 = fadd fast <2 x double> %strided.vec36, %5\n %7 = fmul fast <2 x double> %strided.vec28, %broadcast.splat\n %8 = fadd fast <2 x double> %strided.vec37, %7\n %9 = fmul fast <2 x double> %strided.vec29, %broadcast.splat\n %10 = fadd fast <2 x double> %strided.vec38, %9\n %11 = fmul fast <2 x double> %strided.vec30, %broadcast.splat\n %12 = fadd fast <2 x double> %strided.vec39, %11\n %13 = fmul fast <2 x double> %strided.vec31, %broadcast.splat\n %14 = fadd fast <2 x double> %strided.vec40, %13\n %15 = fmul fast <2 x double> %strided.vec32, %broadcast.splat\n %16 = fadd fast <2 x double> %strided.vec41, %15\n %17 = or disjoint i64 %0, 7\n %18 = fmul fast <2 x double> %strided.vec33, %broadcast.splat\n %19 = getelementptr inbounds double, ptr %x, i64 %17\n %20 = fadd fast <2 x double> %strided.vec42, %18\n %21 = getelementptr inbounds i8, ptr %19, i64 -56\n %22 = shufflevector <2 x double> %4, <2 x double> %6, <4 x i32> \n %23 = shufflevector <2 x double> %8, <2 x double> %10, <4 x i32> \n %24 = shufflevector <2 x double> %12, <2 x double> %14, <4 x i32> \n %25 = shufflevector <2 x double> %16, <2 x double> %20, <4 x i32> \n %26 = shufflevector <4 x double> %22, <4 x double> %23, <8 x i32> \n %27 = shufflevector <4 x double> %24, <4 x double> %25, <8 x i32> \n %interleaved.vec = shufflevector <8 x double> %26, <8 x double> %27, <16 x i32> \n store <16 x double> %interleaved.vec, ptr %21, align 8\n ret void\n}\n"}, {"test_name": "singleop", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i8> @singleop(<4 x i8> %a, <4 x i8> %b) {\n %a1 = shufflevector <4 x i8> %a, <4 x i8> poison, <4 x i32> \n %b1 = shufflevector <4 x i8> %b, <4 x i8> poison, <4 x i32> zeroinitializer\n %a2 = zext <4 x i8> %a1 to <4 x i16>\n %b2 = zext <4 x i8> %b1 to <4 x i16>\n %ab = add <4 x i16> %a2, %b2\n %t = trunc <4 x i16> %ab to <4 x i8>\n %r = shufflevector <4 x i8> %t, <4 x i8> poison, <4 x i32> \n ret <4 x i8> %r\n}\n"}, {"test_name": "not_bitcast", "test_body": "target triple = \"aarch64\"\n\ndefine i32 @not_bitcast(<4 x i8> %x) {\n %shuf = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> \n %bitcast = bitcast <4 x i8> %shuf to i32\n ret i32 %bitcast\n}\n"}, {"test_name": "zext_chain", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i64> @zext_chain(<4 x i16> %x) {\n %shuf = shufflevector <4 x i16> %x, <4 x i16> poison, <4 x i32> \n %zext = zext <4 x i16> %shuf to <4 x i32>\n %sext = sext <4 x i32> %zext to <4 x i64>\n %revshuf = shufflevector <4 x i64> %sext, <4 x i64> poison, <4 x i32> \n ret <4 x i64> %revshuf\n}\n"}, {"test_name": "intrinsics_different", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @intrinsics_different(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.smin.v4i8(<4 x i8> %at, <4 x i8> %bt)\n %abb = call <4 x i8> @llvm.umin.v4i8(<4 x i8> %ab, <4 x i8> %bb)\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.smin.v4i8(<4 x i8>, <4 x i8>) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.umin.v4i8(<4 x i8>, <4 x i8>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "operandbundles_second", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @operandbundles_second(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %at, i1 false)\n %abb = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %ab, i1 false) [ \"jl_roots\"(ptr addrspace(10) null, ptr addrspace(10) null) ]\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.abs.v4i8(<4 x i8>, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "cast_mismatched_types", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i64> @cast_mismatched_types(<4 x i32> %x) {\n %shuf = shufflevector <4 x i32> %x, <4 x i32> poison, <2 x i32> \n %zext = zext <2 x i32> %shuf to <2 x i64>\n %extshuf = shufflevector <2 x i64> %zext, <2 x i64> poison, <4 x i32> \n ret <4 x i64> %extshuf\n}\n"}, {"test_name": "operandbundles_first", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @operandbundles_first(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %at, i1 false) [ \"jl_roots\"(ptr addrspace(10) null, ptr addrspace(10) null) ]\n %abb = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %ab, i1 false)\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.abs.v4i8(<4 x i8>, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "wrong_lanes", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @wrong_lanes(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = add <4 x i8> %at, %bt\n %abb = add <4 x i8> %ab, %bb\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "first_scalar_select", "test_body": "target triple = \"aarch64\"\n\ndefine <2 x float> @first_scalar_select(<2 x float> %0, <2 x float> %1, float %x) {\nentry:\n %cmp.i903 = fcmp ogt float %x, 0.000000e+00\n %sel1639 = select i1 %cmp.i903, <2 x float> %0, <2 x float> %1\n %2 = fcmp ogt <2 x float> %0, zeroinitializer\n %sel48.i913 = select <2 x i1> %2, <2 x float> %0, <2 x float> %1\n %3 = shufflevector <2 x float> %sel1639, <2 x float> %sel48.i913, <2 x i32> \n ret <2 x float> %3\n}\n"}, {"test_name": "splat0", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @splat0(<8 x half> %a, <8 x half> %b) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %bs = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> zeroinitializer\n %abt = fadd <4 x half> %at, %bs\n %abb = fadd <4 x half> %ab, %bs\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n"}, {"test_name": "splatandidentity", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @splatandidentity(<8 x half> %a, <8 x half> %b) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %bs = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> zeroinitializer\n %abt = fadd <4 x half> %at, %bs\n %abb = fadd <4 x half> %ab, %bs\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n"}, {"test_name": "trivial", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @trivial(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %r = shufflevector <4 x i8> %at, <4 x i8> %ab, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "constantsplat", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @constantsplat(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = add <4 x i8> %at, splat (i8 10)\n %abb = add <4 x i8> %ab, splat (i8 10)\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "bitcast_srcty_mismatch", "test_body": "target triple = \"aarch64\"\n\ndefine void @bitcast_srcty_mismatch() {\nentry:\n %shuffle.i.i = shufflevector <2 x i64> zeroinitializer, <2 x i64> zeroinitializer, <2 x i32> \n %0 = bitcast <4 x i32> zeroinitializer to <4 x float>\n %1 = bitcast <2 x i64> %shuffle.i.i to <4 x float>\n %shufp.i196 = shufflevector <4 x float> %0, <4 x float> %1, <4 x i32> \n store <4 x float> %shufp.i196, ptr null, align 16\n ret void\n}\n"}, {"test_name": "zext_add_chain", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i64> @zext_add_chain(<4 x i32> %x) {\n %shuf = shufflevector <4 x i32> %x, <4 x i32> poison, <4 x i32> \n %zext = zext <4 x i32> %shuf to <4 x i64>\n %add = add <4 x i64> %zext, %zext\n %revshuf = shufflevector <4 x i64> %add, <4 x i64> poison, <4 x i32> \n ret <4 x i64> %revshuf\n}\n"}, {"test_name": "add_chain", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i32> @add_chain(<4 x i32> %x) {\n %shuf = shufflevector <4 x i32> %x, <4 x i32> poison, <4 x i32> \n %add = add <4 x i32> %shuf, %shuf\n %add2 = add <4 x i32> %add, %add\n %revshuf = shufflevector <4 x i32> %add2, <4 x i32> poison, <4 x i32> \n ret <4 x i32> %revshuf\n}\n"}, {"test_name": "splat2", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @splat2(<8 x half> %a, <8 x half> %b) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %bs = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> \n %abt = fadd <4 x half> %at, %bs\n %abb = fadd <4 x half> %ab, %bs\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n"}, {"test_name": "trunc", "test_body": "target triple = \"aarch64\"\n\ndefine void @trunc(<8 x i64> %a, <8 x i64> %b, ptr %p) {\n %ab = shufflevector <8 x i64> %a, <8 x i64> poison, <4 x i32> \n %at = shufflevector <8 x i64> %a, <8 x i64> poison, <4 x i32> \n %abb1 = trunc <4 x i64> %ab to <4 x i32>\n %abt1 = trunc <4 x i64> %at to <4 x i32>\n %r = shufflevector <4 x i32> %abb1, <4 x i32> %abt1, <8 x i32> \n store <8 x i32> %r, ptr %p, align 32\n ret void\n}\n"}, {"test_name": "sext", "test_body": "target triple = \"aarch64\"\n\ndefine void @sext(<8 x i16> %a, <8 x i16> %b, ptr %p) {\n %ab = shufflevector <8 x i16> %a, <8 x i16> poison, <4 x i32> \n %at = shufflevector <8 x i16> %a, <8 x i16> poison, <4 x i32> \n %bb = shufflevector <8 x i16> %b, <8 x i16> poison, <4 x i32> \n %bt = shufflevector <8 x i16> %b, <8 x i16> poison, <4 x i32> \n %ab1 = sext <4 x i16> %ab to <4 x i32>\n %at1 = sext <4 x i16> %at to <4 x i32>\n %bb1 = sext <4 x i16> %bb to <4 x i32>\n %bt1 = sext <4 x i16> %bt to <4 x i32>\n %abb = add <4 x i32> %ab1, %bb1\n %abt = add <4 x i32> %at1, %bt1\n %r = shufflevector <4 x i32> %abb, <4 x i32> %abt, <8 x i32> \n store <8 x i32> %r, ptr %p, align 32\n ret void\n}\n"}, {"test_name": "powi", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @powi(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = call <4 x half> @llvm.powi.v4f16.i32(<4 x half> %at, i32 10)\n %abb = call <4 x half> @llvm.powi.v4f16.i32(<4 x half> %ab, i32 10)\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x half> @llvm.powi.v4f16.i32(<4 x half>, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "callinst", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @callinst(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = call <4 x half> @othercall(<4 x half> %at)\n %abb = call <4 x half> @llvm.fabs.v4f16(<4 x half> %ab)\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n\ndeclare <4 x half> @othercall(<4 x half>)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x half> @llvm.fabs.v4f16(<4 x half>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "fma", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @fma(<8 x half> %a, <8 x half> %b, <8 x half> %c) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %bb = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> \n %bt = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> \n %cb = shufflevector <8 x half> %c, <8 x half> poison, <4 x i32> \n %ct = shufflevector <8 x half> %c, <8 x half> poison, <4 x i32> \n %abb = call <4 x half> @llvm.fma.v4f16(<4 x half> %ab, <4 x half> %bb, <4 x half> %cb)\n %abt = call <4 x half> @llvm.fma.v4f16(<4 x half> %at, <4 x half> %bt, <4 x half> %ct)\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x half> @llvm.fma.v4f16(<4 x half>, <4 x half>, <4 x half>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "extrause_shuffle", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @extrause_shuffle(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n call void @use(<4 x i8> %bt)\n %abt = add <4 x i8> %at, %bt\n %abb = add <4 x i8> %ab, %bb\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\ndeclare void @use(<4 x i8>)\n"}, {"test_name": "intrinsics_addsat", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @intrinsics_addsat(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.sadd.sat.v4i8(<4 x i8> %at, <4 x i8> %bt)\n %abb = call <4 x i8> @llvm.sadd.sat.v4i8(<4 x i8> %ab, <4 x i8> %bb)\n %abt1 = call <4 x i8> @llvm.ssub.sat.v4i8(<4 x i8> %abt, <4 x i8> %bt)\n %abb1 = call <4 x i8> @llvm.ssub.sat.v4i8(<4 x i8> %abb, <4 x i8> %bb)\n %abt2 = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> %abt1, <4 x i8> %bt)\n %abb2 = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> %abb1, <4 x i8> %bb)\n %abt3 = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> %abt2, <4 x i8> %bt)\n %abb3 = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> %abb2, <4 x i8> %bb)\n %r = shufflevector <4 x i8> %abt3, <4 x i8> %abb3, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.sadd.sat.v4i8(<4 x i8>, <4 x i8>) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.ssub.sat.v4i8(<4 x i8>, <4 x i8>) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8>, <4 x i8>) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.usub.sat.v4i8(<4 x i8>, <4 x i8>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "zext", "test_body": "target triple = \"aarch64\"\n\ndefine void @zext(<8 x i16> %a, <8 x i16> %b, ptr %p) {\n %ab = shufflevector <8 x i16> %a, <8 x i16> poison, <4 x i32> \n %at = shufflevector <8 x i16> %a, <8 x i16> poison, <4 x i32> \n %bb = shufflevector <8 x i16> %b, <8 x i16> poison, <4 x i32> \n %bt = shufflevector <8 x i16> %b, <8 x i16> poison, <4 x i32> \n %ab1 = zext <4 x i16> %ab to <4 x i32>\n %at1 = zext <4 x i16> %at to <4 x i32>\n %bb1 = zext <4 x i16> %bb to <4 x i32>\n %bt1 = zext <4 x i16> %bt to <4 x i32>\n %abb = add <4 x i32> %ab1, %bb1\n %abt = add <4 x i32> %at1, %bt1\n %r = shufflevector <4 x i32> %abb, <4 x i32> %abt, <8 x i32> \n store <8 x i32> %r, ptr %p, align 32\n ret void\n}\n"}, {"test_name": "div", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @div(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = udiv <4 x i8> %at, %bt\n %abb = udiv <4 x i8> %ab, %bb\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "bitcast_smax_v8i32_v4i32", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i64> @bitcast_smax_v8i32_v4i32(<4 x i64> %a, <4 x i64> %b) {\n %a.bc0 = bitcast <4 x i64> %a to <8 x i32>\n %b.bc0 = bitcast <4 x i64> %b to <8 x i32>\n %cmp = icmp slt <8 x i32> %a.bc0, %b.bc0\n %cmp.lo = shufflevector <8 x i1> %cmp, <8 x i1> poison, <4 x i32> \n %cmp.hi = shufflevector <8 x i1> %cmp, <8 x i1> poison, <4 x i32> \n %a.bc1 = bitcast <4 x i64> %a to <8 x i32>\n %b.bc1 = bitcast <4 x i64> %b to <8 x i32>\n %a.lo = shufflevector <8 x i32> %a.bc1, <8 x i32> poison, <4 x i32> \n %b.lo = shufflevector <8 x i32> %b.bc1, <8 x i32> poison, <4 x i32> \n %lo = select <4 x i1> %cmp.lo, <4 x i32> %b.lo, <4 x i32> %a.lo\n %a.bc2 = bitcast <4 x i64> %a to <8 x i32>\n %b.bc2 = bitcast <4 x i64> %b to <8 x i32>\n %a.hi = shufflevector <8 x i32> %a.bc2, <8 x i32> poison, <4 x i32> \n %b.hi = shufflevector <8 x i32> %b.bc2, <8 x i32> poison, <4 x i32> \n %hi = select <4 x i1> %cmp.hi, <4 x i32> %b.hi, <4 x i32> %a.hi\n %concat = shufflevector <4 x i32> %lo, <4 x i32> %hi, <8 x i32> \n %res = bitcast <8 x i32> %concat to <4 x i64>\n ret <4 x i64> %res\n}\n"}, {"test_name": "not_bitcast2", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i16> @not_bitcast2(<4 x i32> %x, <8 x i16> %y) {\n %cast = bitcast <4 x i32> %x to <8 x i16>\n %out = shufflevector <8 x i16> %y, <8 x i16> %cast, <8 x i32> \n ret <8 x i16> %out\n}\n"}, {"test_name": "abs", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @abs(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %at, i1 false)\n %abb = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %ab, i1 false)\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.abs.v4i8(<4 x i8>, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "maximal_legal_fpmath", "test_body": "target triple = \"aarch64\"\n\ndefine void @maximal_legal_fpmath(ptr %addr1, ptr %addr2, ptr %result, float %val) {\n %splatinsert = insertelement <4 x float> poison, float %val, i64 0\n %incoming.vec = shufflevector <4 x float> %splatinsert, <4 x float> poison, <4 x i32> zeroinitializer\n %vec1 = load <16 x float>, ptr %addr1, align 4\n %strided.vec1 = shufflevector <16 x float> %vec1, <16 x float> poison, <4 x i32> \n %strided.vec2 = shufflevector <16 x float> %vec1, <16 x float> poison, <4 x i32> \n %strided.vec3 = shufflevector <16 x float> %vec1, <16 x float> poison, <4 x i32> \n %strided.vec4 = shufflevector <16 x float> %vec1, <16 x float> poison, <4 x i32> \n %vec2 = load <16 x float>, ptr %addr2, align 4\n %strided.vec6 = shufflevector <16 x float> %vec2, <16 x float> poison, <4 x i32> \n %strided.vec7 = shufflevector <16 x float> %vec2, <16 x float> poison, <4 x i32> \n %strided.vec8 = shufflevector <16 x float> %vec2, <16 x float> poison, <4 x i32> \n %strided.vec9 = shufflevector <16 x float> %vec2, <16 x float> poison, <4 x i32> \n %1 = fmul fast <4 x float> %incoming.vec, %strided.vec6\n %2 = fadd fast <4 x float> %strided.vec1, %1\n %3 = fmul contract <4 x float> %incoming.vec, %strided.vec7\n %4 = fadd fast <4 x float> %strided.vec2, %3\n %5 = fmul reassoc contract <4 x float> %incoming.vec, %strided.vec8\n %6 = fadd fast <4 x float> %strided.vec3, %5\n %7 = fmul reassoc contract <4 x float> %incoming.vec, %strided.vec9\n %8 = fadd reassoc contract <4 x float> %strided.vec4, %7\n %9 = shufflevector <4 x float> %2, <4 x float> %4, <8 x i32> \n %10 = shufflevector <4 x float> %6, <4 x float> %8, <8 x i32> \n %interleaved.vec = shufflevector <8 x float> %9, <8 x float> %10, <16 x i32> \n store <16 x float> %interleaved.vec, ptr %result, align 4\n ret void\n}\n"}, {"test_name": "not_zext", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i64> @not_zext(<4 x i32> %x) {\n %zext = zext <4 x i32> %x to <4 x i64>\n %revshuf = shufflevector <4 x i64> %zext, <4 x i64> poison, <4 x i32> \n ret <4 x i64> %revshuf\n}\n"}, {"test_name": "undeflane", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @undeflane(<8 x i8> %a, <8 x i8> %b) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %abt = add <4 x i8> %at, %bt\n %abb = add <4 x i8> %ab, %bb\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "operandbundles", "test_body": "target triple = \"aarch64\"\n\ndefine <16 x i64> @operandbundles(<4 x i64> %a, <4 x i64> %b, <4 x i64> %c) {\n %call = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a, <4 x i64> %b, <4 x i64> %c) [ \"jl_roots\"(ptr addrspace(10) null, ptr addrspace(10) null) ]\n %shufflevector = shufflevector <4 x i64> %call, <4 x i64> poison, <16 x i32> \n %shufflevector1 = shufflevector <16 x i64> %shufflevector, <16 x i64> undef, <16 x i32> \n ret <16 x i64> %shufflevector1\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i64> @llvm.fshl.v4i64(<4 x i64>, <4 x i64>, <4 x i64>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "fneg", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @fneg(<8 x half> %a, <8 x half> %b) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = fneg <4 x half> %at\n %abb = fneg <4 x half> %ab\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n"}, {"test_name": "icmpsel", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @icmpsel(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c, <8 x i8> %d) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> \n %cb = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> \n %ct = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> \n %db = shufflevector <8 x i8> %d, <8 x i8> poison, <4 x i32> \n %dt = shufflevector <8 x i8> %d, <8 x i8> poison, <4 x i32> \n %abt1 = icmp slt <4 x i8> %at, %bt\n %abb1 = icmp slt <4 x i8> %ab, %bb\n %abt = select <4 x i1> %abt1, <4 x i8> %ct, <4 x i8> %dt\n %abb = select <4 x i1> %abb1, <4 x i8> %cb, <4 x i8> %db\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n"}, {"test_name": "szext", "test_body": "target triple = \"aarch64\"\n\ndefine void @szext(<8 x i32> %a, <8 x i32> %b, ptr %p) {\n %ab = shufflevector <8 x i32> %a, <8 x i32> poison, <4 x i32> \n %at = shufflevector <8 x i32> %a, <8 x i32> poison, <4 x i32> \n %ab1 = sext <4 x i32> %ab to <4 x i64>\n %at1 = zext <4 x i32> %at to <4 x i64>\n %r = shufflevector <4 x i64> %ab1, <4 x i64> %at1, <8 x i32> \n store <8 x i64> %r, ptr %p, align 64\n ret void\n}\n"}, {"test_name": "constantsplatf", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @constantsplatf(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = fadd <4 x half> %at, splat (half 0xH4900)\n %abb = fadd <4 x half> %ab, splat (half 0xH4900)\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n"}, {"test_name": "not_zext2", "test_body": "target triple = \"aarch64\"\n\ndefine <4 x i64> @not_zext2(<4 x i32> %x) {\n %shuf = shufflevector <4 x i32> %x, <4 x i32> poison, <4 x i32> \n %zext = zext <4 x i32> %shuf to <4 x i64>\n ret <4 x i64> %zext\n}\n"}]}], "issue": {"title": "Assertion Failure in Vector Combine", "body": "Runnning vector-combine pass on \r\n```llvm\r\nsource_filename = \"text\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:10:11:12:13\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine nonnull ptr addrspace(10) @wombat() {\r\n %call = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> zeroinitializer, <4 x i64> zeroinitializer, <4 x i64> ) [ \"jl_roots\"(ptr addrspace(10) null, ptr addrspace(10) null) ]\r\n %shufflevector = shufflevector <4 x i64> %call, <4 x i64> poison, <16 x i32> \r\n %shufflevector1 = shufflevector <16 x i64> %shufflevector, <16 x i64> undef, <16 x i32> \r\n ret ptr addrspace(10) null\r\n}\r\n\r\ndeclare <4 x i64> @llvm.fshl.v4i64(<4 x i64>, <4 x i64>, <4 x i64>)\r\n```\r\nproduces\r\n```\r\nopt: /root/llvm-project/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = llvm::FixedVectorType; From = llvm::Type]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=vector-combine \r\n1.\tRunning pass \"function(vector-combine)\" on module \"\"\r\n2.\tRunning pass \"vector-combine\" on function \"wombat\"\r\n #0 0x0000000005079b38 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x5079b38)\r\n #1 0x00000000050774ec SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000074d0dce42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000074d0dce969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000074d0dce42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000074d0dce287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000074d0dce2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000074d0dce39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000004186458 (anonymous namespace)::VectorCombine::foldShuffleToIdentity(llvm::Instruction&) VectorCombine.cpp:0:0\r\n #9 0x0000000004190f01 (anonymous namespace)::VectorCombine::run()::'lambda'(llvm::Instruction&)::operator()(llvm::Instruction&) const (.isra.0) VectorCombine.cpp:0:0\r\n#10 0x0000000004191c25 llvm::VectorCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4191c25)\r\n#11 0x0000000002f2defe llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f2defe)\r\n#12 0x0000000004e7c978 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7c978)\r\n#13 0x0000000000e157ce llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe157ce)\r\n#14 0x0000000004e7b39e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7b39e)\r\n#15 0x0000000000e14f7e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe14f7e)\r\n#16 0x0000000004e7add0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7add0)\r\n#17 0x0000000000904cc2 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x904cc2)\r\n#18 0x00000000008f7b4c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f7b4c)\r\n#19 0x000074d0dce29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#20 0x000074d0dce29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#21 0x00000000008ef61e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ef61e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\nGodbolt: https://godbolt.org/z/MGefYs3M5", "author": "Zentrik", "labels": ["release:backport", "vectorizers", "crash"], "comments": [{"author": "Heresh83", "body": "> Runnning vector-combine pass on \n> ```llvm\n> source_filename = \"text\"\n> target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:10:11:12:13\"\n> target triple = \"x86_64-unknown-linux-gnu\"\n> \n> define nonnull ptr addrspace(10) @wombat() {\n> %call = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> zeroinitializer, <4 x i64> zeroinitializer, <4 x i64> ) [ \"jl_roots\"(ptr addrspace(10) null, ptr addrspace(10) null) ]\n> %shufflevector = shufflevector <4 x i64> %call, <4 x i64> poison, <16 x i32> \n> %shufflevector1 = shufflevector <16 x i64> %shufflevector, <16 x i64> undef, <16 x i32> \n> ret ptr addrspace(10) null\n> }\n> \n> declare <4 x i64> @llvm.fshl.v4i64(<4 x i64>, <4 x i64>, <4 x i64>)\n> ```\n> produces\n> ```\n> opt: /root/llvm-project/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = llvm::FixedVectorType; From = llvm::Type]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\n> PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\n> Stack dump:\n> 0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=vector-combine \n> 1.\tRunning pass \"function(vector-combine)\" on module \"\"\n> 2.\tRunning pass \"vector-combine\" on function \"wombat\"\n> #0 0x0000000005079b38 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x5079b38)\n> #1 0x00000000050774ec SignalHandler(int) Signals.cpp:0:0\n> #2 0x000074d0dce42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n> #3 0x000074d0dce969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n> #4 0x000074d0dce42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n> #5 0x000074d0dce287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n> #6 0x000074d0dce2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n> #7 0x000074d0dce39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n> #8 0x0000000004186458 (anonymous namespace)::VectorCombine::foldShuffleToIdentity(llvm::Instruction&) VectorCombine.cpp:0:0\n> #9 0x0000000004190f01 (anonymous namespace)::VectorCombine::run()::'lambda'(llvm::Instruction&)::operator()(llvm::Instruction&) const (.isra.0) VectorCombine.cpp:0:0\n> #10 0x0000000004191c25 llvm::VectorCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4191c25)\n> #11 0x0000000002f2defe llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f2defe)\n> #12 0x0000000004e7c978 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7c978)\n> #13 0x0000000000e157ce llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe157ce)\n> #14 0x0000000004e7b39e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7b39e)\n> #15 0x0000000000e14f7e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe14f7e)\n> #16 0x0000000004e7add0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7add0)\n> #17 0x0000000000904cc2 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x904cc2)\n> #18 0x00000000008f7b4c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f7b4c)\n> #19 0x000074d0dce29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n> #20 0x000074d0dce29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n> #21 0x00000000008ef61e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ef61e)\n> Program terminated with signal: SIGSEGV\n> Compiler returned: 139\n> ```\n> Godbolt: https://godbolt.org/z/MGefYs3M5\n\n"}, {"author": "RKSimon", "body": "Looks like foldShuffleToIdentity is attempting to fold the operand bundle on the fshl call"}]}, "verified": true} {"bug_id": "120433", "issue_url": "https://github.com/llvm/llvm-project/issues/120433", "bug_type": "crash", "base_commit": "ac8a9f8fffb605514056f53f12f1fba94ee933e9", "knowledge_cutoff": "2024-12-18T15:02:42Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "e7a4d78ad328d02bf515b2fa4af8b2c188a6a636", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[21148, 21155]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["SLPVectorizerPass::vectorizeCmpInsts"]}}, "patch": "commit e7a4d78ad328d02bf515b2fa4af8b2c188a6a636\nAuthor: DianQK \nDate: Thu Dec 19 06:21:57 2024 +0800\n\n [SLP] Check if instructions exist after vectorization (#120434)\n \n Fixes #120433.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex d967813075bb..88049897b29a 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -21148,8 +21148,11 @@ bool SLPVectorizerPass::vectorizeCmpInsts(iterator_range CmpInsts,\n if (R.isDeleted(I))\n continue;\n for (Value *Op : I->operands())\n- if (auto *RootOp = dyn_cast(Op))\n+ if (auto *RootOp = dyn_cast(Op)) {\n Changed |= vectorizeRootInstruction(nullptr, RootOp, BB, R);\n+ if (R.isDeleted(I))\n+ break;\n+ }\n }\n // Try to vectorize operands as vector bundles.\n for (CmpInst *I : CmpInsts) {\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/slp-deleted-inst.ll", "commands": ["opt -S -passes=slp-vectorizer < %s"], "tests": [{"test_name": "foo", "test_body": "define void @foo() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb3, %bb\n %i = phi i32 [ %i26, %bb3 ], [ 0, %bb ]\n %i2 = phi i32 [ %i24, %bb3 ], [ 0, %bb ]\n br label %bb3\n\nbb3: ; preds = %bb1\n %i4 = zext i32 %i2 to i64\n %i5 = mul i64 %i4, 0\n %i10 = or i64 0, %i5\n %i11 = trunc i64 %i10 to i32\n %i12 = and i32 %i11, 0\n %i13 = zext i32 %i to i64\n %i14 = mul i64 %i13, 0\n %i19 = or i64 0, %i14\n %i20 = trunc i64 %i19 to i32\n %i21 = and i32 %i20, 0\n %i22 = or i32 %i12, %i21\n %i23 = icmp ult i32 %i22, 0\n %i24 = select i1 %i23, i32 0, i32 0\n %i25 = icmp ult i32 0, 0\n %i26 = select i1 %i25, i32 0, i32 0\n br label %bb1\n}\n"}]}], "issue": {"title": "[SLP] Assertion `detail::isPresent(Val) && \"dyn_cast on a non-existent value\"' failed.", "body": "This IR crashes on SLP: https://llvm.godbolt.org/z/8oar4o4d1\n\n```llvm\ndefine void @foo() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb3, %bb\n %i = phi i32 [ %i26, %bb3 ], [ 0, %bb ]\n %i2 = phi i32 [ %i24, %bb3 ], [ 0, %bb ]\n br label %bb3\n\nbb3: ; preds = %bb1\n %i4 = zext i32 %i2 to i64\n %i5 = mul i64 %i4, 0\n %i10 = or i64 0, %i5\n %i11 = trunc i64 %i10 to i32\n %i12 = and i32 %i11, 0\n %i13 = zext i32 %i to i64\n %i14 = mul i64 %i13, 0\n %i19 = or i64 0, %i14\n %i20 = trunc i64 %i19 to i32\n %i21 = and i32 %i20, 0\n %i22 = or i32 %i12, %i21\n %i23 = icmp ult i32 %i22, 0\n %i24 = select i1 %i23, i32 0, i32 0\n %i25 = icmp ult i32 0, 0\n %i26 = select i1 %i25, i32 0, i32 0\n br label %bb1\n}\n```", "author": "DianQK", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "110295", "issue_url": "https://github.com/llvm/llvm-project/issues/110295", "bug_type": "crash", "base_commit": "51e0a997ca607f64beafb838ba1325f0465eecd4", "knowledge_cutoff": "2024-09-27T16:39:37Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "2c8836c899015cce49a399a6bc47b260a24a22e7", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6539, 6546]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInstructionCost"]}}, "patch": "commit 2c8836c899015cce49a399a6bc47b260a24a22e7\nAuthor: Florian Hahn \nDate: Sun Sep 29 20:31:23 2024 +0100\n\n [LV] Don't consider predicated insts as invariant unconditionally in CM.\n \n Predicated instructions cannot hoisted trivially, so don't treat them as\n uniform value in the cost model.\n \n This fixes a difference between legacy and VPlan-based cost model.\n \n Fixes https://github.com/llvm/llvm-project/issues/110295.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex cb346be8ffe5..792e0e17dd87 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6539,8 +6539,16 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I,\n Op2 = cast(PSE.getSCEV(Op2))->getValue();\n }\n auto Op2Info = TTI.getOperandInfo(Op2);\n- if (Op2Info.Kind == TargetTransformInfo::OK_AnyValue &&\n- Legal->isInvariant(Op2))\n+ auto IsInvariant = [this](Value *Op) {\n+ if (!Legal->isInvariant(Op))\n+ return false;\n+ // Consider Op2 invariant, if it is not a predicated instruction in the\n+ // loop. In that case, it is not trivially hoistable.\n+ return !isa(Op) ||\n+ !TheLoop->contains(cast(Op)) ||\n+ !isPredicatedInst(cast(Op));\n+ };\n+ if (Op2Info.Kind == TargetTransformInfo::OK_AnyValue && IsInvariant(Op2))\n Op2Info.Kind = TargetTransformInfo::OK_UniformValue;\n \n SmallVector Operands(I->operand_values());\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/predicated-instruction-cost.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "predicated_urem_shl_cost", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @predicated_urem_shl_cost(ptr %A, i32 %x, i1 %c) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 1, %entry ], [ %iv.next, %loop.latch ]\n %gep = getelementptr inbounds i32, ptr %A, i32 %iv\n %l = load i32, ptr %gep, align 4\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n %rem = urem i32 2, %x\n %shl = shl i32 %l, %rem\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %p = phi i32 [ 0, %loop.header ], [ %shl, %then ]\n store i32 %p, ptr %gep, align 4\n %iv.next = add i32 %iv, 1\n %ec = icmp eq i32 %iv, 0\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}], "issue": {"title": "clang crashes at -O{2,3} on x86_64-linux-gnu: Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed ", "body": "Compiler Explorer: https://godbolt.org/z/YhKsza3dx\r\n\r\nRelated to https://github.com/llvm/llvm-project/issues/108697, but it doesn't seem to trigger anymore with the latest build.\r\n\r\n```\r\n[519] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 631bcbe9de13e160d427ad7452a7ef2ca67911ab)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[520] % \r\n[520] % clangtk -O2 small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7380: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20240927-clangtk-m64-O3-build-064435/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20240927-clangtk-m64-O3-build-064435/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-c6819e.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4.\tRunning pass \"loop-vectorize\" on function \"main\"\r\n #0 0x0000560d31f97bef llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x43b3bef)\r\n #1 0x0000560d31f953c4 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f3f4386e420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007f3f432a500b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007f3f43284859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007f3f43284729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007f3f43284729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007f3f43295fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x0000560d3393ac39 llvm::LoopVectorizationPlanner::computeBestVF() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5d56c39)\r\n #9 0x0000560d3395443b llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5d7043b)\r\n#10 0x0000560d33957301 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5d73301)\r\n#11 0x0000560d33957973 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5d73973)\r\n#12 0x0000560d334c3406 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x58df406)\r\n#13 0x0000560d31924a81 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3d40a81)\r\n#14 0x0000560d2f43df76 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1859f76)\r\n#15 0x0000560d319233ed llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3d3f3ed)\r\n#16 0x0000560d2f43cb56 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1858b56)\r\n#17 0x0000560d31922e21 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3d3ee21)\r\n#18 0x0000560d322504b3 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#19 0x0000560d32253a88 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x466fa88)\r\n#20 0x0000560d3293154c clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4d4d54c)\r\n#21 0x0000560d347d56cc clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x6bf16cc)\r\n#22 0x0000560d32931968 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4d4d968)\r\n#23 0x0000560d32bf9249 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5015249)\r\n#24 0x0000560d32b752fe clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4f912fe)\r\n#25 0x0000560d32ce5926 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5101926)\r\n#26 0x0000560d2f004695 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1420695)\r\n#27 0x0000560d2effc1aa ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#28 0x0000560d2f000bde clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x141cbde)\r\n#29 0x0000560d2eeed47b main (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x130947b)\r\n#30 0x00007f3f43286083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#31 0x0000560d2effbc3e _start (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1417c3e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 631bcbe9de13e160d427ad7452a7ef2ca67911ab)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-49ab95.c\r\nclangtk: note: diagnostic msg: /tmp/small-49ab95.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[521] % \r\n[521] % cat small.c\r\nint a, b, c, e, f, i, g, h, k, l, m;\r\nunsigned d;\r\nvolatile int j;\r\nint n(short o) { return o ? a % o : 0; }\r\nint main() {\r\n short q;\r\n if (h)\r\n while (1)\r\n ;\r\nr:\r\n e && j;\r\n if (a && j)\r\n goto r;\r\ns:\r\n while (g)\r\n if (l)\r\n goto s;\r\n for (; i; i++) {\r\n f = !d ? : 2 % d;\r\n q = f > 1 ? 0 : c << f;\r\n m = q;\r\n b = n(k ^ q);\r\n }\r\n return 0;\r\n}\r\n```\r\n", "author": "zhendongsu", "labels": ["regression", "vectorizers", "confirmed", "crash"], "comments": [{"author": "shafik", "body": "This looks like a regression in trunk: https://godbolt.org/z/c5Wxf5Teh\r\n\r\nI know godbolt does not have clang assertions build beyond trunk but you can use clang++ and just use `-x c` to check out if the assert exists in previous versions.\r\n\r\nCC @fhahn "}, {"author": "zhendongsu", "body": "> I know godbolt does not have clang assertions build beyond trunk but you can use clang++ and just use `-x c` to check out if the assert exists in previous versions.\r\n\r\n@shafik: Thanks for the tip!\r\n\r\n"}]}, "verified": true} {"bug_id": "97600", "issue_url": "https://github.com/llvm/llvm-project/issues/97600", "bug_type": "miscompilation", "base_commit": "97a2bd8415dc6792b99ec0f091ad7570673c3f37", "knowledge_cutoff": "2024-07-03T16:13:45Z", "lit_test_dir": ["llvm/test/Analysis/BasicAA"], "hints": {"fix_commit": "ebab105670a409e426ddcb0278578711a622b1b2", "components": ["IR", "ValueTracking"], "bug_location_lineno": {"llvm/lib/Analysis/ValueTracking.cpp": [[6403, 6411]], "llvm/lib/IR/Value.cpp": [[652, 660]]}, "bug_location_funcname": {"llvm/lib/Analysis/ValueTracking.cpp": ["llvm::getUnderlyingObject"], "llvm/lib/IR/Value.cpp": ["stripPointerCastsAndOffsets"]}}, "patch": "commit ebab105670a409e426ddcb0278578711a622b1b2\nAuthor: Nikita Popov \nDate: Thu Jul 4 09:43:16 2024 +0200\n\n [IR] Don't strip through pointer to vector of pointer bitcasts\n \n When using stripPointerCasts() and getUnderlyingObject(), don't\n strip through a bitcast from ptr to <1 x ptr>, which is not a\n no-op pointer cast. Calling code is generally not prepared to\n handle that situation, resulting in incorrect alias analysis\n results for example.\n \n Fixes https://github.com/llvm/llvm-project/issues/97600.\n\ndiff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp\nindex 5476dc5d8518..258576f0cdff 100644\n--- a/llvm/lib/Analysis/ValueTracking.cpp\n+++ b/llvm/lib/Analysis/ValueTracking.cpp\n@@ -6403,9 +6403,10 @@ const Value *llvm::getUnderlyingObject(const Value *V, unsigned MaxLookup) {\n V = GEP->getPointerOperand();\n } else if (Operator::getOpcode(V) == Instruction::BitCast ||\n Operator::getOpcode(V) == Instruction::AddrSpaceCast) {\n- V = cast(V)->getOperand(0);\n- if (!V->getType()->isPointerTy())\n+ Value *NewV = cast(V)->getOperand(0);\n+ if (!NewV->getType()->isPointerTy())\n return V;\n+ V = NewV;\n } else if (auto *GA = dyn_cast(V)) {\n if (GA->isInterposable())\n return V;\ndiff --git a/llvm/lib/IR/Value.cpp b/llvm/lib/IR/Value.cpp\nindex 8522747ccf12..b2ee75811fbb 100644\n--- a/llvm/lib/IR/Value.cpp\n+++ b/llvm/lib/IR/Value.cpp\n@@ -652,9 +652,10 @@ static const Value *stripPointerCastsAndOffsets(\n }\n V = GEP->getPointerOperand();\n } else if (Operator::getOpcode(V) == Instruction::BitCast) {\n- V = cast(V)->getOperand(0);\n- if (!V->getType()->isPointerTy())\n+ Value *NewV = cast(V)->getOperand(0);\n+ if (!NewV->getType()->isPointerTy())\n return V;\n+ V = NewV;\n } else if (StripKind != PSK_ZeroIndicesSameRepresentation &&\n Operator::getOpcode(V) == Instruction::AddrSpaceCast) {\n // TODO: If we know an address space cast will not change the\n", "tests": [{"file": "llvm/test/Analysis/BasicAA/ptr-vector.ll", "commands": ["opt -S -passes=licm < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "@c = dso_local global i32 0, align 4\n@d = dso_local local_unnamed_addr global ptr @c, align 8\n\ndefine i32 @main() {\nentry:\n br label %for.cond\n\nfor.cond: ; preds = %f.exit.split, %entry\n %b.0 = phi i32 [ 0, %entry ], [ %inc6, %f.exit.split ]\n %cmp = icmp ult i32 %b.0, 2\n br i1 %cmp, label %for.cond1.preheader, label %for.cond.cleanup\n\nfor.cond1.preheader: ; preds = %for.cond\n br label %for.cond1\n\nfor.cond.cleanup: ; preds = %for.cond\n %0 = load i32, ptr @c, align 4, !tbaa !13\n %retval = xor i32 %0, 7\n ret i32 %retval\n\nfor.cond1: ; preds = %for.cond1, %for.cond1.preheader\n %i.0 = phi i32 [ %inc, %for.cond1 ], [ 0, %for.cond1.preheader ]\n %cmp2 = icmp ult i32 %i.0, 2\n %inc = add nuw nsw i32 %i.0, 1\n br i1 %cmp2, label %for.cond1, label %for.body.i.preheader, !llvm.loop !17\n\nfor.body.i.preheader: ; preds = %for.cond1\n %cmp2.lcssa = phi i1 [ %cmp2, %for.cond1 ]\n %1 = xor i1 %cmp2.lcssa, true\n %2 = bitcast i1 %1 to <1 x i1>\n %3 = call <1 x ptr> @llvm.masked.load.v1p0.p0(ptr @d, i32 8, <1 x i1> %2, <1 x ptr> poison), !tbaa !9\n %4 = bitcast <1 x ptr> %3 to ptr\n store i32 0, ptr @c, align 4, !tbaa !13\n %5 = load i32, ptr %4, align 4, !tbaa !13\n %tobool1.not.i = icmp ne i32 %5, 0\n %tobool1.not.i.fr = freeze i1 %tobool1.not.i\n br i1 %tobool1.not.i.fr, label %f.exit.split, label %for.body.i.preheader.split\n\nfor.body.i.preheader.split: ; preds = %for.body.i.preheader\n br label %for.body.i\n\nfor.body.i: ; preds = %for.body.i.preheader.split, %for.body.i\n %n.04.i = phi i8 [ %add.i, %for.body.i ], [ -66, %for.body.i.preheader.split ]\n %add.i = add nsw i8 %n.04.i, 1\n %tobool.not.i = icmp eq i8 %add.i, 0\n br i1 %tobool.not.i, label %f.exit, label %for.body.i, !llvm.loop !15\n\nf.exit: ; preds = %for.body.i\n br label %f.exit.split\n\nf.exit.split: ; preds = %for.body.i.preheader, %f.exit\n store i32 7, ptr %4, align 4, !tbaa !13\n %inc6 = add nuw nsw i32 %b.0, 1\n br label %for.cond, !llvm.loop !18\n}\n\n!9 = !{!10, !10, i64 0}\n!10 = !{!\"any pointer\", !11, i64 0}\n!11 = !{!\"omnipotent char\", !12, i64 0}\n!12 = !{!\"Simple C/C++ TBAA\"}\n!13 = !{!14, !14, i64 0}\n!14 = !{!\"int\", !11, i64 0}\n!15 = distinct !{!15, !16}\n!16 = !{!\"llvm.loop.mustprogress\"}\n!17 = distinct !{!17, !16}\n!18 = distinct !{!18, !16}", "lli_expected_out": ""}]}], "issue": {"title": "[LICM] Miscompilation caused by wrong AA result", "body": "Reproducer: https://godbolt.org/z/Y1vvbndf8\r\n```\r\n@c = dso_local global i32 0, align 4\r\n@d = dso_local local_unnamed_addr global ptr @c, align 8\r\n\r\ndefine i32 @main() {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %f.exit.split, %entry\r\n %b.0 = phi i32 [ 0, %entry ], [ %inc6, %f.exit.split ]\r\n %cmp = icmp ult i32 %b.0, 2\r\n br i1 %cmp, label %for.cond1.preheader, label %for.cond.cleanup\r\n\r\nfor.cond1.preheader: ; preds = %for.cond\r\n br label %for.cond1\r\n\r\nfor.cond.cleanup: ; preds = %for.cond\r\n %0 = load i32, ptr @c, align 4, !tbaa !13\r\n ret i32 %0\r\n\r\nfor.cond1: ; preds = %for.cond1, %for.cond1.preheader\r\n %i.0 = phi i32 [ %inc, %for.cond1 ], [ 0, %for.cond1.preheader ]\r\n %cmp2 = icmp ult i32 %i.0, 2\r\n %inc = add nuw nsw i32 %i.0, 1\r\n br i1 %cmp2, label %for.cond1, label %for.body.i.preheader, !llvm.loop !17\r\n\r\nfor.body.i.preheader: ; preds = %for.cond1\r\n %cmp2.lcssa = phi i1 [ %cmp2, %for.cond1 ]\r\n %1 = xor i1 %cmp2.lcssa, true\r\n %2 = bitcast i1 %1 to <1 x i1>\r\n %3 = call <1 x ptr> @llvm.masked.load.v1p0.p0(ptr @d, i32 8, <1 x i1> %2, <1 x ptr> poison), !tbaa !9\r\n %4 = bitcast <1 x ptr> %3 to ptr\r\n store i32 0, ptr @c, align 4, !tbaa !13\r\n %5 = load i32, ptr %4, align 4, !tbaa !13\r\n %tobool1.not.i = icmp ne i32 %5, 0\r\n %tobool1.not.i.fr = freeze i1 %tobool1.not.i\r\n br i1 %tobool1.not.i.fr, label %f.exit.split, label %for.body.i.preheader.split\r\n\r\nfor.body.i.preheader.split: ; preds = %for.body.i.preheader\r\n br label %for.body.i\r\n\r\nfor.body.i: ; preds = %for.body.i.preheader.split, %for.body.i\r\n %n.04.i = phi i8 [ %add.i, %for.body.i ], [ -66, %for.body.i.preheader.split ]\r\n %add.i = add nsw i8 %n.04.i, 1\r\n %tobool.not.i = icmp eq i8 %add.i, 0\r\n br i1 %tobool.not.i, label %f.exit, label %for.body.i, !llvm.loop !15\r\n\r\nf.exit: ; preds = %for.body.i\r\n br label %f.exit.split\r\n\r\nf.exit.split: ; preds = %for.body.i.preheader, %f.exit\r\n store i32 7, ptr %4, align 4, !tbaa !13\r\n %inc6 = add nuw nsw i32 %b.0, 1\r\n br label %for.cond, !llvm.loop !18\r\n}\r\n\r\n!9 = !{!10, !10, i64 0}\r\n!10 = !{!\"any pointer\", !11, i64 0}\r\n!11 = !{!\"omnipotent char\", !12, i64 0}\r\n!12 = !{!\"Simple C/C++ TBAA\"}\r\n!13 = !{!14, !14, i64 0}\r\n!14 = !{!\"int\", !11, i64 0}\r\n!15 = distinct !{!15, !16}\r\n!16 = !{!\"llvm.loop.mustprogress\"}\r\n!17 = distinct !{!17, !16}\r\n!18 = distinct !{!18, !16}\r\n```\r\nAfter LICM:\r\n```\r\nsource_filename = \"/app/example.ll\"\r\n\r\ndefine i32 @main() {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %f.exit.split, %entry\r\n %0 = phi i32 [ poison, %entry ], [ 0, %f.exit.split ]\r\n %b.0 = phi i32 [ 0, %entry ], [ %inc6, %f.exit.split ]\r\n %cmp = icmp ult i32 %b.0, 2\r\n br i1 %cmp, label %for.cond1.preheader, label %for.cond.cleanup\r\n\r\nfor.cond1.preheader: ; preds = %for.cond\r\n br label %for.cond1\r\n\r\nfor.cond.cleanup: ; preds = %for.cond\r\n %.lcssa = phi i32 [ %0, %for.cond ]\r\n store i32 %.lcssa, ptr @c, align 4, !tbaa !0\r\n %1 = load i32, ptr @c, align 4, !tbaa !0\r\n ret i32 %1\r\n\r\nfor.cond1: ; preds = %for.cond1, %for.cond1.preheader\r\n %i.0 = phi i32 [ %inc, %for.cond1 ], [ 0, %for.cond1.preheader ]\r\n %cmp2 = icmp ult i32 %i.0, 2\r\n %inc = add nuw nsw i32 %i.0, 1\r\n br i1 %cmp2, label %for.cond1, label %for.body.i.preheader, !llvm.loop !4\r\n\r\nfor.body.i.preheader: ; preds = %for.cond1\r\n %cmp2.lcssa = phi i1 [ %cmp2, %for.cond1 ]\r\n %2 = xor i1 %cmp2.lcssa, true\r\n %3 = bitcast i1 %2 to <1 x i1>\r\n %4 = call <1 x ptr> @llvm.masked.load.v1p0.p0(ptr @d, i32 8, <1 x i1> %3, <1 x ptr> poison), !tbaa !6\r\n %5 = bitcast <1 x ptr> %4 to ptr\r\n %6 = load i32, ptr %5, align 4, !tbaa !0\r\n %tobool1.not.i = icmp ne i32 %6, 0\r\n %tobool1.not.i.fr = freeze i1 %tobool1.not.i\r\n br i1 %tobool1.not.i.fr, label %f.exit.split, label %for.body.i.preheader.split\r\n\r\nfor.body.i.preheader.split: ; preds = %for.body.i.preheader\r\n br label %for.body.i\r\n\r\nfor.body.i: ; preds = %for.body.i, %for.body.i.preheader.split\r\n %n.04.i = phi i8 [ %add.i, %for.body.i ], [ -66, %for.body.i.preheader.split ]\r\n %add.i = add nsw i8 %n.04.i, 1\r\n %tobool.not.i = icmp eq i8 %add.i, 0\r\n br i1 %tobool.not.i, label %f.exit, label %for.body.i, !llvm.loop !8\r\n\r\nf.exit: ; preds = %for.body.i\r\n br label %f.exit.split\r\n\r\nf.exit.split: ; preds = %f.exit, %for.body.i.preheader\r\n store i32 7, ptr %5, align 4, !tbaa !0\r\n %inc6 = add nuw nsw i32 %b.0, 1\r\n br label %for.cond, !llvm.loop !9\r\n}\r\n\r\ndeclare <1 x ptr> @llvm.masked.load.v1p0.p0(ptr nocapture, i32 immarg, <1 x i1>, <1 x ptr>) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(argmem: read) }\r\n\r\n!0 = !{!1, !1, i64 0}\r\n!1 = !{!\"int\", !2, i64 0}\r\n!2 = !{!\"omnipotent char\", !3, i64 0}\r\n!3 = !{!\"Simple C/C++ TBAA\"}\r\n!4 = distinct !{!4, !5}\r\n!5 = !{!\"llvm.loop.mustprogress\"}\r\n!6 = !{!7, !7, i64 0}\r\n!7 = !{!\"any pointer\", !2, i64 0}\r\n!8 = distinct !{!8, !5}\r\n!9 = distinct !{!9, !5}\r\n```\r\n```\r\n> lli test.ll\r\n> echo $?\r\n7\r\n> bin/opt -passes=licm test.ll -S -o out.ll\r\n> lli out.ll\r\n> echo $?\r\n0\r\n```\r\n`store i32 7, ptr %5, align 4, !tbaa !0` may write something to `@c`. So this transform is incorrect.\r\n\r\nRelated patch: https://github.com/llvm/llvm-project/pull/96878\r\nOriginal C code (needs x86 +cf or riscv +zicldst):\r\n```\r\n#include \"csmith.h\"\r\nuint32_t c = 0;\r\nint32_t *d = &c;\r\nvoid f(int32_t *o) {\r\n uint8_t n;\r\n for (n = 190; n; n += 1) {\r\n *o = 0;\r\n if (*d)\r\n break;\r\n }\r\n}\r\nint main() {\r\n for (int b = 0; b <= 1; b++) {\r\n uint64_t j[2];\r\n int i;\r\n for (i = 0; i < 2; i++)\r\n j[i] = 1;\r\n f(&c);\r\n safe_add_func_int32_t_s_s(0, j[1]);\r\n *d = 7;\r\n }\r\n printf(\"%d\\n\", c);\r\n return 0;\r\n}\r\n```\r\n\r\ncc @nikic @fhahn @KanRobert \r\n", "author": "dtcxzyw", "labels": ["miscompilation", "TBAA", "llvm:ir", "llvm:analysis"], "comments": [{"author": "KanRobert", "body": "Which patch supports `zicldst` for riscv? @dtcxzyw "}, {"author": "nikic", "body": "Hm, we probably have a lot of places that don't handle things like `bitcast <1 x ptr> %3 to ptr` correctly, and assume that bitcast of pointer is always a pointer to pointer cast."}, {"author": "dtcxzyw", "body": "> Which patch supports `zicldst` for riscv? @dtcxzyw\r\n\r\nWe have a prototype which is still working in progress:\r\n+ llvm support: https://github.com/llvm/llvm-project/compare/main...dtcxzyw:llvm-project:zicldst-support\r\n+ qemu support: https://github.com/qemu/qemu/compare/master...cyyself:qemu:exp_rv_zicldst\r\n"}]}, "verified": true} {"bug_id": "82230", "issue_url": "https://github.com/llvm/llvm-project/issues/82230", "bug_type": "crash", "base_commit": "769c22f25b81b74e4da7871d4e552584605caef1", "knowledge_cutoff": "2024-02-19T09:37:19Z", "lit_test_dir": ["llvm/test/DebugInfo"], "hints": {"fix_commit": "5b342e6041f8567a39354ae3c5346cf74b3bb84f", "components": ["IR"], "bug_location_lineno": {"llvm/lib/IR/AsmWriter.cpp": [[293, 299], [4886, 4893]]}, "bug_location_funcname": {"llvm/lib/IR/AsmWriter.cpp": ["DPValue::print", "PrintCallingConv", "getModuleFromDPI"]}}, "patch": "commit 5b342e6041f8567a39354ae3c5346cf74b3bb84f\nAuthor: Stephen Tozer \nDate: Mon Feb 19 11:38:04 2024 +0000\n\n [RemoveDIs][DebugInfo] Check for null marker when printing DPValues (#82238)\n \n The function to print DPValues currently tries to incorporate the\n function it is part of, which is found through its marker; this means\n when we try to print a DPValue with no marker, we dereference a nullptr.\n We can print instructions without parents, and so the same should be\n true for DPValues; this patch changes DPValue::print to check for a null\n marker and avoid dereferencing it.\n \n Fixes issue: https://github.com/llvm/llvm-project/issues/82230\n\ndiff --git a/llvm/lib/IR/AsmWriter.cpp b/llvm/lib/IR/AsmWriter.cpp\nindex 0ae720e8b7ce..d3c64a57f7fd 100644\n--- a/llvm/lib/IR/AsmWriter.cpp\n+++ b/llvm/lib/IR/AsmWriter.cpp\n@@ -293,7 +293,7 @@ static const Module *getModuleFromDPI(const DPMarker *Marker) {\n }\n \n static const Module *getModuleFromDPI(const DPValue *DPV) {\n- return getModuleFromDPI(DPV->getMarker());\n+ return DPV->getMarker() ? getModuleFromDPI(DPV->getMarker()) : nullptr;\n }\n \n static void PrintCallingConv(unsigned cc, raw_ostream &Out) {\n@@ -4886,8 +4886,9 @@ void DPValue::print(raw_ostream &ROS, ModuleSlotTracker &MST,\n if (F)\n MST.incorporateFunction(*F);\n };\n- incorporateFunction(Marker->getParent() ? Marker->getParent()->getParent()\n- : nullptr);\n+ incorporateFunction(Marker && Marker->getParent()\n+ ? Marker->getParent()->getParent()\n+ : nullptr);\n AssemblyWriter W(OS, SlotTable, getModuleFromDPI(this), nullptr, IsForDebug);\n W.printDPValue(*this);\n }\n", "tests": [{"file": "llvm/test/DebugInfo/dpvalue-print-nocrash.ll", "commands": ["opt -passes=\"instcombine\" -debug %s -o /dev/null 2>&1"], "tests": [{"test_name": "", "test_body": ";; Tests that we can debug-print DPValues that have no markers attached.\n; REQUIRES: asserts\n\n\ndefine ptr @func_10(i32 %p_11) {\nentry:\n %conv108 = zext i32 %p_11 to i64\n tail call void @llvm.dbg.value(metadata i64 %conv108, metadata !4, metadata !DIExpression()), !dbg !12\n br label %func_29.exit\n\nfunc_29.exit: ; preds = %entry\n store i64 %conv108, ptr null, align 1\n ret ptr null\n}\n\ndeclare void @llvm.dbg.value(metadata, metadata, metadata)\n\n!llvm.dbg.cu = !{!0}\n!llvm.module.flags = !{!3}\n\n!0 = distinct !DICompileUnit(language: DW_LANG_C11, file: !1, producer: \"clang\", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, retainedTypes: !2, globals: !2, splitDebugInlining: false, nameTableKind: None)\n!1 = !DIFile(filename: \"csmith5961503756960.c\", directory: \"/llvm\")\n!2 = !{}\n!3 = !{i32 2, !\"Debug Info Version\", i32 3}\n!4 = !DILocalVariable(name: \"p_31\", arg: 2, scope: !5, file: !1, line: 148, type: !7)\n!5 = distinct !DISubprogram(name: \"func_29\", scope: !1, file: !1, line: 148, type: !6, scopeLine: 149, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)\n!6 = !DISubroutineType(types: !2)\n!7 = !DIDerivedType(tag: DW_TAG_typedef, name: \"uint64_t\", file: !8, line: 60, baseType: !9)\n!8 = !DIFile(filename: \"/foo/_stdint.h\", directory: \"\")\n!9 = !DIDerivedType(tag: DW_TAG_typedef, name: \"__uint64_t\", file: !10, line: 108, baseType: !11)\n!10 = !DIFile(filename: \"/foo/_default_types.h\", directory: \"\")\n!11 = !DIBasicType(name: \"unsigned long long\", size: 64, encoding: DW_ATE_unsigned)\n!12 = !DILocation(line: 0, scope: !5)"}]}], "issue": {"title": "opt -passes=\"instcombine\" -debug crashes with non-instrinsic debug-info", "body": "llvm commit: d61864f813e3\r\nReproduce with:\r\n```\r\nopt -passes=\"instcombine\" -debug bbi-92509.ll -o /dev/null\r\n```\r\n[bbi-92509.ll.gz](https://github.com/llvm/llvm-project/files/14329283/bbi-92509.ll.gz)\r\n\r\nResult:\r\n```\r\nINSTCOMBINE ITERATION #1 on func_10\r\nADD: ret ptr null\r\nADD: store i64 %conv108, ptr null, align 1\r\nADD: br label %func_29.exit\r\nADD: %conv108 = zext i32 %p_11 to i64\r\nCLONE: PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: ../../main-github/llvm/build-all/bin/opt -passes=instcombine -debug bbi-92509.ll -o /dev/null\r\n #0 0x000055e12968f4a7 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (../../main-github/llvm/build-all/bin/opt+0x3db14a7)\r\n #1 0x000055e12968cf7e llvm::sys::RunSignalHandlers() (../../main-github/llvm/build-all/bin/opt+0x3daef7e)\r\n #2 0x000055e12968fe6f SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007fdf33999630 __restore_rt sigaction.c:0:0\r\n #4 0x000055e12971f8e0 llvm::DPMarker::getParent() const (../../main-github/llvm/build-all/bin/opt+0x3e418e0)\r\n #5 0x000055e129761abf llvm::DPValue::print(llvm::raw_ostream&, bool) const (../../main-github/llvm/build-all/bin/opt+0x3e83abf)\r\n #6 0x000055e12a244957 llvm::InstCombinerImpl::tryToSinkInstructionDPValues(llvm::Instruction*, llvm::ilist_iterator_w_bits, false, false>, llvm::BasicBlock*, llvm::BasicBlock*, llvm::SmallVectorImpl&) InstructionCombining.cpp:0:0\r\n #7 0x000055e12a242b0c llvm::InstCombinerImpl::tryToSinkInstruction(llvm::Instruction*, llvm::BasicBlock*) InstructionCombining.cpp:0:0\r\n #8 0x000055e12a2453fc llvm::InstCombinerImpl::run() InstructionCombining.cpp:0:0\r\n #9 0x000055e12a248c79 combineInstructionsOverFunction(llvm::Function&, llvm::InstructionWorklist&, llvm::AAResults*, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::ProfileSummaryInfo*, llvm::LoopInfo*, llvm::InstCombineOptions const&) InstructionCombining.cpp:0:0\r\n#10 0x000055e12a2482ea llvm::InstCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x496a2ea)\r\n#11 0x000055e12a95e5bd llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) PassBuilderPipelines.cpp:0:0\r\n#12 0x000055e12989a8e4 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x3fbc8e4)\r\n#13 0x000055e12a9644dd llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) PassBuilderPipelines.cpp:0:0\r\n#14 0x000055e12989f9ae llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x3fc19ae)\r\n#15 0x000055e12a95dd9d llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) PassBuilderPipelines.cpp:0:0\r\n#16 0x000055e129899456 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x3fbb456)\r\n#17 0x000055e12a90a42b llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (../../main-github/llvm/build-all/bin/opt+0x502c42b)\r\n#18 0x000055e129657a28 optMain (../../main-github/llvm/build-all/bin/opt+0x3d79a28)\r\n#19 0x00007fdf310cc555 __libc_start_main (/lib64/libc.so.6+0x22555)\r\n#20 0x000055e129651369 _start (../../main-github/llvm/build-all/bin/opt+0x3d73369)\r\nSegmentation fault (core dumped)\r\n```\r\n\r\nThis starts happening with a93a4ec7dd205\r\n```\r\nReapply \"[DebugInfo][RemoveDIs] Turn on non-instrinsic debug-info by default\"\r\n```\r\n", "author": "mikaelholmen", "labels": ["debuginfo", "llvm:crash", "llvm:ir"], "comments": [{"author": "SLTozer", "body": "Should be fixed by: [5b342e6](https://github.com/llvm/llvm-project/commit/5b342e6041f8567a39354ae3c5346cf74b3bb84f)"}, {"author": "mikaelholmen", "body": "> Should be fixed by: [5b342e6](https://github.com/llvm/llvm-project/commit/5b342e6041f8567a39354ae3c5346cf74b3bb84f)\r\n\r\nYep, thanks!"}]}, "verified": true} {"bug_id": "113997", "issue_url": "https://github.com/llvm/llvm-project/issues/113997", "bug_type": "miscompilation", "base_commit": "bef3b54ea10a564a2de72f658f2efd64f537c079", "knowledge_cutoff": "2024-10-29T04:11:49Z", "lit_test_dir": ["llvm/test/Transforms/GVN", "llvm/test/Transforms/NewGVN"], "hints": {"fix_commit": "f16bff1261a92169992c6edf6bc6b38d1c815c8d", "components": ["GVN", "NewGVN", "Local"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/GVN.cpp": [[2189, 2194], [2734, 2740], [3000, 3006]], "llvm/lib/Transforms/Scalar/NewGVN.cpp": [[3854, 3859], [3963, 3968], [4069, 4076], [4112, 4128]], "llvm/lib/Transforms/Utils/Local.cpp": [[3508, 3513]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/GVN.cpp": ["GVNPass::performScalarPRE", "GVNPass::processAssumeIntrinsic", "GVNPass::processInstruction", "patchAndReplaceAllUsesWith"], "llvm/lib/Transforms/Scalar/NewGVN.cpp": ["NewGVN::eliminateInstructions", "NewGVN::findPHIOfOpsLeader"], "llvm/lib/Transforms/Utils/Local.cpp": ["llvm::patchReplacementInstruction"]}}, "patch": "commit f16bff1261a92169992c6edf6bc6b38d1c815c8d\nAuthor: Yingwei Zheng \nDate: Fri Nov 1 12:44:33 2024 +0800\n\n [GVN][NewGVN][Local] Handle attributes for function calls after CSE (#114011)\n \n This patch intersects attributes of two calls to avoid introducing UB.\n It also skips incompatible call pairs in GVN/NewGVN. However, I cannot\n provide negative tests for these changes.\n \n Fixes https://github.com/llvm/llvm-project/issues/113997.\n\ndiff --git a/llvm/lib/Transforms/Scalar/GVN.cpp b/llvm/lib/Transforms/Scalar/GVN.cpp\nindex 2ba600497e00..ad9b1217089d 100644\n--- a/llvm/lib/Transforms/Scalar/GVN.cpp\n+++ b/llvm/lib/Transforms/Scalar/GVN.cpp\n@@ -2189,6 +2189,16 @@ bool GVNPass::processAssumeIntrinsic(AssumeInst *IntrinsicI) {\n return Changed;\n }\n \n+// Return true iff V1 can be replaced with V2.\n+static bool canBeReplacedBy(Value *V1, Value *V2) {\n+ if (auto *CB1 = dyn_cast(V1))\n+ if (auto *CB2 = dyn_cast(V2))\n+ return CB1->getAttributes()\n+ .intersectWith(CB2->getContext(), CB2->getAttributes())\n+ .has_value();\n+ return true;\n+}\n+\n static void patchAndReplaceAllUsesWith(Instruction *I, Value *Repl) {\n patchReplacementInstruction(I, Repl);\n I->replaceAllUsesWith(Repl);\n@@ -2734,7 +2744,7 @@ bool GVNPass::processInstruction(Instruction *I) {\n // Perform fast-path value-number based elimination of values inherited from\n // dominators.\n Value *Repl = findLeader(I->getParent(), Num);\n- if (!Repl) {\n+ if (!Repl || !canBeReplacedBy(I, Repl)) {\n // Failure, just remember this instance for future use.\n LeaderTable.insert(Num, I, I->getParent());\n return false;\n@@ -3000,7 +3010,7 @@ bool GVNPass::performScalarPRE(Instruction *CurInst) {\n \n uint32_t TValNo = VN.phiTranslate(P, CurrentBlock, ValNo, *this);\n Value *predV = findLeader(P, TValNo);\n- if (!predV) {\n+ if (!predV || !canBeReplacedBy(CurInst, predV)) {\n predMap.push_back(std::make_pair(static_cast(nullptr), P));\n PREPred = P;\n ++NumWithout;\ndiff --git a/llvm/lib/Transforms/Scalar/NewGVN.cpp b/llvm/lib/Transforms/Scalar/NewGVN.cpp\nindex 13d9e8f186b4..6800ad51cc0a 100644\n--- a/llvm/lib/Transforms/Scalar/NewGVN.cpp\n+++ b/llvm/lib/Transforms/Scalar/NewGVN.cpp\n@@ -3854,6 +3854,16 @@ Value *NewGVN::findPHIOfOpsLeader(const Expression *E,\n return nullptr;\n }\n \n+// Return true iff V1 can be replaced with V2.\n+static bool canBeReplacedBy(Value *V1, Value *V2) {\n+ if (auto *CB1 = dyn_cast(V1))\n+ if (auto *CB2 = dyn_cast(V2))\n+ return CB1->getAttributes()\n+ .intersectWith(CB2->getContext(), CB2->getAttributes())\n+ .has_value();\n+ return true;\n+}\n+\n bool NewGVN::eliminateInstructions(Function &F) {\n // This is a non-standard eliminator. The normal way to eliminate is\n // to walk the dominator tree in order, keeping track of available\n@@ -3963,6 +3973,9 @@ bool NewGVN::eliminateInstructions(Function &F) {\n MembersLeft.insert(Member);\n continue;\n }\n+ if (!canBeReplacedBy(Member, Leader))\n+ continue;\n+\n LLVM_DEBUG(dbgs() << \"Found replacement \" << *(Leader) << \" for \"\n << *Member << \"\\n\");\n auto *I = cast(Member);\n@@ -4069,8 +4082,11 @@ bool NewGVN::eliminateInstructions(Function &F) {\n if (DominatingLeader != Def) {\n // Even if the instruction is removed, we still need to update\n // flags/metadata due to downstreams users of the leader.\n- if (!match(DefI, m_Intrinsic()))\n+ if (!match(DefI, m_Intrinsic())) {\n+ if (!canBeReplacedBy(DefI, DominatingLeader))\n+ continue;\n patchReplacementInstruction(DefI, DominatingLeader);\n+ }\n \n markInstructionForDeletion(DefI);\n }\n@@ -4112,17 +4128,21 @@ bool NewGVN::eliminateInstructions(Function &F) {\n // Don't replace our existing users with ourselves.\n if (U->get() == DominatingLeader)\n continue;\n- LLVM_DEBUG(dbgs()\n- << \"Found replacement \" << *DominatingLeader << \" for \"\n- << *U->get() << \" in \" << *(U->getUser()) << \"\\n\");\n \n // If we replaced something in an instruction, handle the patching of\n // metadata. Skip this if we are replacing predicateinfo with its\n // original operand, as we already know we can just drop it.\n auto *ReplacedInst = cast(U->get());\n auto *PI = PredInfo->getPredicateInfoFor(ReplacedInst);\n- if (!PI || DominatingLeader != PI->OriginalOp)\n+ if (!PI || DominatingLeader != PI->OriginalOp) {\n+ if (!canBeReplacedBy(ReplacedInst, DominatingLeader))\n+ continue;\n patchReplacementInstruction(ReplacedInst, DominatingLeader);\n+ }\n+\n+ LLVM_DEBUG(dbgs()\n+ << \"Found replacement \" << *DominatingLeader << \" for \"\n+ << *U->get() << \" in \" << *(U->getUser()) << \"\\n\");\n U->set(DominatingLeader);\n // This is now a use of the dominating leader, which means if the\n // dominating leader was dead, it's now live!\ndiff --git a/llvm/lib/Transforms/Utils/Local.cpp b/llvm/lib/Transforms/Utils/Local.cpp\nindex 65c1669f92b4..47a704925596 100644\n--- a/llvm/lib/Transforms/Utils/Local.cpp\n+++ b/llvm/lib/Transforms/Utils/Local.cpp\n@@ -3508,6 +3508,17 @@ void llvm::patchReplacementInstruction(Instruction *I, Value *Repl) {\n else if (!isa(I))\n ReplInst->andIRFlags(I);\n \n+ // Handle attributes.\n+ if (auto *CB1 = dyn_cast(ReplInst)) {\n+ if (auto *CB2 = dyn_cast(I)) {\n+ bool Success = CB1->tryIntersectAttributes(CB2);\n+ assert(Success && \"We should not be trying to sink callbases \"\n+ \"with non-intersectable attributes\");\n+ // For NDEBUG Compile.\n+ (void)Success;\n+ }\n+ }\n+\n // FIXME: If both the original and replacement value are part of the\n // same control-flow region (meaning that the execution of one\n // guarantees the execution of the other), then we can combine the\n", "tests": [{"file": "llvm/test/Transforms/GVN/pr113997.ll", "commands": ["opt -S -passes=gvn < %s"], "tests": [{"test_name": "bucket", "test_body": "define i1 @bucket(i32 noundef %x) {\n %cmp1 = icmp sgt i32 %x, 0\n %ctpop1 = tail call range(i32 1, 32) i32 @llvm.ctpop.i32(i32 %x)\n %cmp2 = icmp samesign ult i32 %ctpop1, 2\n %cond = select i1 %cmp1, i1 %cmp2, i1 false\n br i1 %cond, label %if.then, label %if.else\n\nif.else: ; preds = %0\n %ctpop2 = tail call range(i32 0, 33) i32 @llvm.ctpop.i32(i32 %x)\n %res = icmp eq i32 %ctpop2, 1\n ret i1 %res\n\nif.then: ; preds = %0\n ret i1 false\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}, {"file": "llvm/test/Transforms/NewGVN/pr113997.ll", "commands": ["opt -S -passes=newgvn < %s"], "tests": [{"test_name": "bucket", "test_body": "define i1 @bucket(i32 noundef %x) {\n %cmp1 = icmp sgt i32 %x, 0\n %ctpop1 = tail call range(i32 1, 32) i32 @llvm.ctpop.i32(i32 %x)\n %cmp2 = icmp samesign ult i32 %ctpop1, 2\n %cond = select i1 %cmp1, i1 %cmp2, i1 false\n br i1 %cond, label %if.then, label %if.else\n\nif.else: ; preds = %0\n %ctpop2 = tail call range(i32 0, 33) i32 @llvm.ctpop.i32(i32 %x)\n %res = icmp eq i32 %ctpop2, 1\n ret i1 %res\n\nif.then: ; preds = %0\n ret i1 false\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[GVNPass] Range attribute should be handled after CSE", "body": "Reproducer: https://alive2.llvm.org/ce/z/HAnoKn\r\n```\r\ndefine i1 @src(i32 noundef %x) {\r\n %cmp1 = icmp sgt i32 %x, 0\r\n %ctpop1 = tail call range(i32 1, 32) i32 @llvm.ctpop.i32(i32 %x)\r\n %cmp2 = icmp samesign ult i32 %ctpop1, 2\r\n %cond = select i1 %cmp1, i1 %cmp2, i1 false\r\n br i1 %cond, label %if.then, label %if.else\r\n\r\nif.else:\r\n %ctpop2 = tail call range(i32 0, 33) i32 @llvm.ctpop.i32(i32 %x)\r\n %res = icmp eq i32 %ctpop2, 1\r\n ret i1 %res\r\n\r\nif.then:\r\n ret i1 false\r\n}\r\n\r\ndefine i1 @tgt(i32 noundef %x) {\r\n %cmp1 = icmp sgt i32 %x, 0\r\n %ctpop1 = tail call range(i32 1, 32) i32 @llvm.ctpop.i32(i32 %x)\r\n %cmp2 = icmp samesign ult i32 %ctpop1, 2\r\n %cond = select i1 %cmp1, i1 %cmp2, i1 false\r\n br i1 %cond, label %if.then, label %if.else\r\n\r\nif.else:\r\n %res = icmp eq i32 %ctpop1, 1\r\n ret i1 %res\r\n\r\nif.then:\r\n ret i1 false\r\n}\r\n```\r\n```\r\n\r\n----------------------------------------\r\ndefine i1 @src(i32 noundef %x) {\r\n#0:\r\n %cmp1 = icmp sgt i32 noundef %x, 0\r\n %ctpop1 = ctpop i32 noundef %x\r\n %#range_0_%ctpop1 = !range i32 %ctpop1, i32 1, i32 32\r\n %cmp2 = icmp samesign ult i32 %#range_0_%ctpop1, 2\r\n %cond = select i1 %cmp1, i1 %cmp2, i1 0\r\n br i1 %cond, label %if.then, label %if.else\r\n\r\nif.else:\r\n %ctpop2 = ctpop i32 noundef %x\r\n %#range_1_%ctpop2 = !range i32 %ctpop2, i32 0, i32 33\r\n %res = icmp eq i32 %#range_1_%ctpop2, 1\r\n ret i1 %res\r\n\r\nif.then:\r\n ret i1 0\r\n}\r\n=>\r\ndefine i1 @tgt(i32 noundef %x) {\r\n#0:\r\n %cmp1 = icmp sgt i32 noundef %x, 0\r\n %ctpop1 = ctpop i32 noundef %x\r\n %#range_0_%ctpop1 = !range i32 %ctpop1, i32 1, i32 32\r\n %cmp2 = icmp samesign ult i32 %#range_0_%ctpop1, 2\r\n %cond = select i1 %cmp1, i1 %cmp2, i1 0\r\n br i1 %cond, label %if.then, label %if.else\r\n\r\nif.else:\r\n %res = icmp eq i32 %#range_0_%ctpop1, 1\r\n ret i1 %res\r\n\r\nif.then:\r\n ret i1 0\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\ni32 noundef %x = #x00000000 (0)\r\n\r\nSource:\r\ni1 %cmp1 = #x0 (0)\r\ni32 %ctpop1 = #x00000000 (0)\r\ni32 %#range_0_%ctpop1 = poison\r\ni1 %cmp2 = poison\r\ni1 %cond = #x0 (0)\r\n >> Jump to %if.else\r\ni32 %ctpop2 = #x00000000 (0)\r\ni32 %#range_1_%ctpop2 = #x00000000 (0)\r\ni1 %res = #x0 (0)\r\n\r\nTarget:\r\ni1 %cmp1 = #x0 (0)\r\ni32 %ctpop1 = #x00000000 (0)\r\ni32 %#range_0_%ctpop1 = poison\r\ni1 %cmp2 = poison\r\ni1 %cond = #x0 (0)\r\n >> Jump to %if.else\r\ni1 %res = poison\r\nSource value: #x0 (0)\r\nTarget value: poison\r\n```\r\n\r\nReported by @pranavk. See https://github.com/llvm/llvm-project/pull/100899#issuecomment-2442780517.\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:GVN"], "comments": []}, "verified": true} {"bug_id": "120823", "issue_url": "https://github.com/llvm/llvm-project/issues/120823", "bug_type": "miscompilation", "base_commit": "9d0a5d4620a2aa5dc01b150e5ebe7613238cae1c", "knowledge_cutoff": "2024-12-21T05:25:23Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "030829a7e53fad0eab9b87b5dd49427e9fb13303", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[15483, 15488]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree"]}}, "patch": "commit 030829a7e53fad0eab9b87b5dd49427e9fb13303\nAuthor: Alexey Bataev \nDate: Mon Dec 23 16:53:37 2024 -0800\n\n [SLP]Drop samesign flag if the vector node has reduced bitwidth\n \n If the operands of the icmp instructions has reduced bitwidth after\n MinBitwidth analysis, need to drop samesign flag to preserve correctness\n of the transformation.\n \n Fixes #120823\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex d22379429d00..b5d68c075b98 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -15483,6 +15483,8 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n CmpInst::Predicate P0 = cast(VL0)->getPredicate();\n Value *V = Builder.CreateCmp(P0, L, R);\n propagateIRFlags(V, E->Scalars, VL0);\n+ if (auto *ICmp = dyn_cast(V); ICmp && It == MinBWs.end())\n+ ICmp->setSameSign(/*B=*/false);\n // Do not cast for cmps.\n VecTy = cast(V->getType());\n V = FinalShuffle(V, E);\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/buildvector-schedule-for-subvector.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -slp-threshold=-99999 < %s"], "tests": [{"test_name": "test", "test_body": "define i1 @test() {\nentry:\n %and.i1698.1.i = zext i16 0 to i32\n %and19.i1699.2.i = and i32 %and.i1698.1.i, 0\n %and.i1698.2.i = zext i16 0 to i32\n %cmp25.i1700.2.i2 = icmp samesign uge i32 %and19.i1699.2.i, %and.i1698.1.i\n %and19.i1699.11841.i = and i32 %and.i1698.2.i, 0\n %cmp25.i1700.11842.i3 = icmp samesign uge i32 %and19.i1699.11841.i, %and.i1698.2.i\n %and.i1698.1.1.i = zext i16 0 to i32\n %and19.i1699.2.1.i = and i32 %and.i1698.1.1.i, 0\n %0 = add i16 1, 0\n %and.i1698.2.1.i = zext i16 %0 to i32\n %cmp25.i1700.2.1.i4 = icmp samesign uge i32 %and19.i1699.2.1.i, %and.i1698.1.1.i\n %and19.i1699.21846.i = and i32 %and.i1698.2.1.i, 0\n %cmp25.i1700.21847.i = icmp samesign uge i32 %and19.i1699.21846.i, %and.i1698.2.1.i\n ret i1 %cmp25.i1700.21847.i\n}\n"}]}], "issue": {"title": "[SLPVectorizer] `samesign` flags should be dropped after narrowing down the width of operands", "body": "Reproducer: https://alive2.llvm.org/ce/z/AvBhy9\n```\n; bin/opt -passes=slp-vectorizer test.ll -S\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i1 @test() {\nentry:\n %and.i1698.1.i = zext i16 0 to i32\n %and19.i1699.2.i = and i32 %and.i1698.1.i, 0\n %and.i1698.2.i = zext i16 0 to i32\n %cmp25.i1700.2.i2 = icmp samesign uge i32 %and19.i1699.2.i, %and.i1698.1.i\n %and19.i1699.11841.i = and i32 %and.i1698.2.i, 0\n %cmp25.i1700.11842.i3 = icmp samesign uge i32 %and19.i1699.11841.i, %and.i1698.2.i\n %and.i1698.1.1.i = zext i16 0 to i32\n %and19.i1699.2.1.i = and i32 %and.i1698.1.1.i, 0\n %0 = add i16 1, 0\n %and.i1698.2.1.i = zext i16 %0 to i32\n %cmp25.i1700.2.1.i4 = icmp samesign uge i32 %and19.i1699.2.1.i, %and.i1698.1.1.i\n %and19.i1699.21846.i = and i32 %and.i1698.2.1.i, 0\n %cmp25.i1700.21847.i = icmp samesign uge i32 %and19.i1699.21846.i, %and.i1698.2.1.i\n ret i1 %cmp25.i1700.21847.i\n}\n```\n```\ndefine i1 @test() {\nentry:\n %0 = add i16 1, 0\n %1 = insertelement <4 x i16> , i16 %0, i32 0\n %2 = trunc <4 x i16> %1 to <4 x i1>\n %3 = and <4 x i1> %2, zeroinitializer\n %4 = icmp samesign uge <4 x i1> %3, %2\n %5 = extractelement <4 x i1> %4, i32 0\n ret i1 %5\n}\n```\n```\n\n----------------------------------------\ndefine i1 @src() {\nentry:\n %#0 = add i16 1, 0\n %and.i1698.2.1.i = zext i16 %#0 to i32\n %and19.i1699.21846.i = and i32 %and.i1698.2.1.i, 0\n %cmp25.i1700.21847.i = icmp samesign uge i32 %and19.i1699.21846.i, %and.i1698.2.1.i\n ret i1 %cmp25.i1700.21847.i\n}\n=>\ndefine i1 @src() {\nentry:\n %#0 = add i16 1, 0\n %#1 = insertelement <4 x i16> { poison, 0, 0, 0 }, i16 %#0, i32 0\n %#2 = trunc <4 x i16> %#1 to <4 x i1>\n %#3 = and <4 x i1> %#2, { 0, 0, 0, 0 }\n %#4 = icmp samesign uge <4 x i1> %#3, %#2\n %#5 = extractelement <4 x i1> %#4, i32 0\n ret i1 %#5\n}\nTransformation doesn't verify!\n\nERROR: Target is more poisonous than source\n\nExample:\n\nSource:\ni16 %#0 = #x0001 (1)\ni32 %and.i1698.2.1.i = #x00000001 (1)\ni32 %and19.i1699.21846.i = #x00000000 (0)\ni1 %cmp25.i1700.21847.i = #x0 (0)\n\nTarget:\ni16 %#0 = #x0001 (1)\n<4 x i16> %#1 = < #x0001 (1), #x0000 (0), #x0000 (0), #x0000 (0) >\n<4 x i1> %#2 = < #x1 (1), #x0 (0), #x0 (0), #x0 (0) >\n<4 x i1> %#3 = < #x0 (0), #x0 (0), #x0 (0), #x0 (0) >\n<4 x i1> %#4 = < poison, #x1 (1), #x1 (1), #x1 (1) >\ni1 %#5 = poison\nSource value: #x0 (0)\nTarget value: poison\n\nSummary:\n 0 correct transformations\n 1 incorrect transformations\n 0 failed-to-prove transformations\n 0 Alive2 errors\n```\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": []}, "verified": true} {"bug_id": "91417", "issue_url": "https://github.com/llvm/llvm-project/issues/91417", "bug_type": "miscompilation", "base_commit": "d9507a3e10d1750d88dd518c14b9a9a62b9eefcd", "knowledge_cutoff": "2024-05-08T01:40:37Z", "lit_test_dir": ["llvm/test/Transforms/Reassociate"], "hints": {"fix_commit": "645fb04a3389e69801d401e669eae9ee42d70217", "components": ["Reassociate"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/Reassociate.cpp": [[302, 399], [475, 481], [490, 497], [509, 515], [518, 525], [528, 534], [562, 568], [625, 634], [642, 648], [1188, 1195], [2368, 2374]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/Reassociate.cpp": ["CarmichaelShift", "IncorporateWeight", "LinearizeExprTree", "LowerNegateToMultiply", "ReassociatePass::ReassociateExpression", "ReassociatePass::RemoveFactorFromExpression"]}}, "patch": "commit 645fb04a3389e69801d401e669eae9ee42d70217\nAuthor: Yingwei Zheng \nDate: Sat Jun 8 22:28:56 2024 +0800\n\n [Reassociate] Use uint64_t for repeat count (#94232)\n \n This patch relands #91469 and uses `uint64_t` for repeat count to avoid\n a miscompilation caused by overflow\n https://github.com/llvm/llvm-project/pull/91469#discussion_r1623925158.\n\ndiff --git a/llvm/lib/Transforms/Scalar/Reassociate.cpp b/llvm/lib/Transforms/Scalar/Reassociate.cpp\nindex c73d7c8d83be..f36e21b296bd 100644\n--- a/llvm/lib/Transforms/Scalar/Reassociate.cpp\n+++ b/llvm/lib/Transforms/Scalar/Reassociate.cpp\n@@ -302,98 +302,7 @@ static BinaryOperator *LowerNegateToMultiply(Instruction *Neg) {\n return Res;\n }\n \n-/// Returns k such that lambda(2^Bitwidth) = 2^k, where lambda is the Carmichael\n-/// function. This means that x^(2^k) === 1 mod 2^Bitwidth for\n-/// every odd x, i.e. x^(2^k) = 1 for every odd x in Bitwidth-bit arithmetic.\n-/// Note that 0 <= k < Bitwidth, and if Bitwidth > 3 then x^(2^k) = 0 for every\n-/// even x in Bitwidth-bit arithmetic.\n-static unsigned CarmichaelShift(unsigned Bitwidth) {\n- if (Bitwidth < 3)\n- return Bitwidth - 1;\n- return Bitwidth - 2;\n-}\n-\n-/// Add the extra weight 'RHS' to the existing weight 'LHS',\n-/// reducing the combined weight using any special properties of the operation.\n-/// The existing weight LHS represents the computation X op X op ... op X where\n-/// X occurs LHS times. The combined weight represents X op X op ... op X with\n-/// X occurring LHS + RHS times. If op is \"Xor\" for example then the combined\n-/// operation is equivalent to X if LHS + RHS is odd, or 0 if LHS + RHS is even;\n-/// the routine returns 1 in LHS in the first case, and 0 in LHS in the second.\n-static void IncorporateWeight(APInt &LHS, const APInt &RHS, unsigned Opcode) {\n- // If we were working with infinite precision arithmetic then the combined\n- // weight would be LHS + RHS. But we are using finite precision arithmetic,\n- // and the APInt sum LHS + RHS may not be correct if it wraps (it is correct\n- // for nilpotent operations and addition, but not for idempotent operations\n- // and multiplication), so it is important to correctly reduce the combined\n- // weight back into range if wrapping would be wrong.\n-\n- // If RHS is zero then the weight didn't change.\n- if (RHS.isMinValue())\n- return;\n- // If LHS is zero then the combined weight is RHS.\n- if (LHS.isMinValue()) {\n- LHS = RHS;\n- return;\n- }\n- // From this point on we know that neither LHS nor RHS is zero.\n-\n- if (Instruction::isIdempotent(Opcode)) {\n- // Idempotent means X op X === X, so any non-zero weight is equivalent to a\n- // weight of 1. Keeping weights at zero or one also means that wrapping is\n- // not a problem.\n- assert(LHS == 1 && RHS == 1 && \"Weights not reduced!\");\n- return; // Return a weight of 1.\n- }\n- if (Instruction::isNilpotent(Opcode)) {\n- // Nilpotent means X op X === 0, so reduce weights modulo 2.\n- assert(LHS == 1 && RHS == 1 && \"Weights not reduced!\");\n- LHS = 0; // 1 + 1 === 0 modulo 2.\n- return;\n- }\n- if (Opcode == Instruction::Add || Opcode == Instruction::FAdd) {\n- // TODO: Reduce the weight by exploiting nsw/nuw?\n- LHS += RHS;\n- return;\n- }\n-\n- assert((Opcode == Instruction::Mul || Opcode == Instruction::FMul) &&\n- \"Unknown associative operation!\");\n- unsigned Bitwidth = LHS.getBitWidth();\n- // If CM is the Carmichael number then a weight W satisfying W >= CM+Bitwidth\n- // can be replaced with W-CM. That's because x^W=x^(W-CM) for every Bitwidth\n- // bit number x, since either x is odd in which case x^CM = 1, or x is even in\n- // which case both x^W and x^(W - CM) are zero. By subtracting off multiples\n- // of CM like this weights can always be reduced to the range [0, CM+Bitwidth)\n- // which by a happy accident means that they can always be represented using\n- // Bitwidth bits.\n- // TODO: Reduce the weight by exploiting nsw/nuw? (Could do much better than\n- // the Carmichael number).\n- if (Bitwidth > 3) {\n- /// CM - The value of Carmichael's lambda function.\n- APInt CM = APInt::getOneBitSet(Bitwidth, CarmichaelShift(Bitwidth));\n- // Any weight W >= Threshold can be replaced with W - CM.\n- APInt Threshold = CM + Bitwidth;\n- assert(LHS.ult(Threshold) && RHS.ult(Threshold) && \"Weights not reduced!\");\n- // For Bitwidth 4 or more the following sum does not overflow.\n- LHS += RHS;\n- while (LHS.uge(Threshold))\n- LHS -= CM;\n- } else {\n- // To avoid problems with overflow do everything the same as above but using\n- // a larger type.\n- unsigned CM = 1U << CarmichaelShift(Bitwidth);\n- unsigned Threshold = CM + Bitwidth;\n- assert(LHS.getZExtValue() < Threshold && RHS.getZExtValue() < Threshold &&\n- \"Weights not reduced!\");\n- unsigned Total = LHS.getZExtValue() + RHS.getZExtValue();\n- while (Total >= Threshold)\n- Total -= CM;\n- LHS = Total;\n- }\n-}\n-\n-using RepeatedValue = std::pair;\n+using RepeatedValue = std::pair;\n \n /// Given an associative binary expression, return the leaf\n /// nodes in Ops along with their weights (how many times the leaf occurs). The\n@@ -475,7 +384,6 @@ static bool LinearizeExprTree(Instruction *I,\n assert((isa(I) || isa(I)) &&\n \"Expected a UnaryOperator or BinaryOperator!\");\n LLVM_DEBUG(dbgs() << \"LINEARIZE: \" << *I << '\\n');\n- unsigned Bitwidth = I->getType()->getScalarType()->getPrimitiveSizeInBits();\n unsigned Opcode = I->getOpcode();\n assert(I->isAssociative() && I->isCommutative() &&\n \"Expected an associative and commutative operation!\");\n@@ -490,8 +398,8 @@ static bool LinearizeExprTree(Instruction *I,\n // with their weights, representing a certain number of paths to the operator.\n // If an operator occurs in the worklist multiple times then we found multiple\n // ways to get to it.\n- SmallVector, 8> Worklist; // (Op, Weight)\n- Worklist.push_back(std::make_pair(I, APInt(Bitwidth, 1)));\n+ SmallVector, 8> Worklist; // (Op, Weight)\n+ Worklist.push_back(std::make_pair(I, 1));\n bool Changed = false;\n \n // Leaves of the expression are values that either aren't the right kind of\n@@ -509,7 +417,7 @@ static bool LinearizeExprTree(Instruction *I,\n \n // Leaves - Keeps track of the set of putative leaves as well as the number of\n // paths to each leaf seen so far.\n- using LeafMap = DenseMap;\n+ using LeafMap = DenseMap;\n LeafMap Leaves; // Leaf -> Total weight so far.\n SmallVector LeafOrder; // Ensure deterministic leaf output order.\n const DataLayout DL = I->getModule()->getDataLayout();\n@@ -518,8 +426,8 @@ static bool LinearizeExprTree(Instruction *I,\n SmallPtrSet Visited; // For checking the iteration scheme.\n #endif\n while (!Worklist.empty()) {\n- std::pair P = Worklist.pop_back_val();\n- I = P.first; // We examine the operands of this binary operator.\n+ // We examine the operands of this binary operator.\n+ auto [I, Weight] = Worklist.pop_back_val();\n \n if (isa(I)) {\n Flags.HasNUW &= I->hasNoUnsignedWrap();\n@@ -528,7 +436,6 @@ static bool LinearizeExprTree(Instruction *I,\n \n for (unsigned OpIdx = 0; OpIdx < I->getNumOperands(); ++OpIdx) { // Visit operands.\n Value *Op = I->getOperand(OpIdx);\n- APInt Weight = P.second; // Number of paths to this operand.\n LLVM_DEBUG(dbgs() << \"OPERAND: \" << *Op << \" (\" << Weight << \")\\n\");\n assert(!Op->use_empty() && \"No uses, so how did we get to it?!\");\n \n@@ -562,7 +469,8 @@ static bool LinearizeExprTree(Instruction *I,\n \"In leaf map but not visited!\");\n \n // Update the number of paths to the leaf.\n- IncorporateWeight(It->second, Weight, Opcode);\n+ It->second += Weight;\n+ assert(It->second >= Weight && \"Weight overflows\");\n \n // If we still have uses that are not accounted for by the expression\n // then it is not safe to modify the value.\n@@ -625,10 +533,7 @@ static bool LinearizeExprTree(Instruction *I,\n // Node initially thought to be a leaf wasn't.\n continue;\n assert(!isReassociableOp(V, Opcode) && \"Shouldn't be a leaf!\");\n- APInt Weight = It->second;\n- if (Weight.isMinValue())\n- // Leaf already output or weight reduction eliminated it.\n- continue;\n+ uint64_t Weight = It->second;\n // Ensure the leaf is only output once.\n It->second = 0;\n Ops.push_back(std::make_pair(V, Weight));\n@@ -642,7 +547,7 @@ static bool LinearizeExprTree(Instruction *I,\n if (Ops.empty()) {\n Constant *Identity = ConstantExpr::getBinOpIdentity(Opcode, I->getType());\n assert(Identity && \"Associative operation without identity!\");\n- Ops.emplace_back(Identity, APInt(Bitwidth, 1));\n+ Ops.emplace_back(Identity, 1);\n }\n \n return Changed;\n@@ -1188,8 +1093,7 @@ Value *ReassociatePass::RemoveFactorFromExpression(Value *V, Value *Factor) {\n Factors.reserve(Tree.size());\n for (unsigned i = 0, e = Tree.size(); i != e; ++i) {\n RepeatedValue E = Tree[i];\n- Factors.append(E.second.getZExtValue(),\n- ValueEntry(getRank(E.first), E.first));\n+ Factors.append(E.second, ValueEntry(getRank(E.first), E.first));\n }\n \n bool FoundFactor = false;\n@@ -2368,7 +2272,7 @@ void ReassociatePass::ReassociateExpression(BinaryOperator *I) {\n SmallVector Ops;\n Ops.reserve(Tree.size());\n for (const RepeatedValue &E : Tree)\n- Ops.append(E.second.getZExtValue(), ValueEntry(getRank(E.first), E.first));\n+ Ops.append(E.second, ValueEntry(getRank(E.first), E.first));\n \n LLVM_DEBUG(dbgs() << \"RAIn:\\t\"; PrintOps(I, Ops); dbgs() << '\\n');\n \n", "tests": [{"file": "llvm/test/Transforms/Reassociate/repeats.ll", "commands": ["opt < %s -passes=reassociate -S"], "tests": [{"test_name": "foo4x11", "test_body": "define i4 @foo4x11(i4 %x) {\n %tmp1 = mul i4 %x, %x\n %tmp2 = mul i4 %tmp1, %x\n %tmp3 = mul i4 %tmp2, %x\n %tmp4 = mul i4 %tmp3, %x\n %tmp5 = mul i4 %tmp4, %x\n %tmp6 = mul i4 %tmp5, %x\n %tmp7 = mul i4 %tmp6, %x\n %tmp8 = mul i4 %tmp7, %x\n %tmp9 = mul i4 %tmp8, %x\n %tmp10 = mul i4 %tmp9, %x\n ret i4 %tmp10\n}\n"}, {"test_name": "foo3x5_nsw", "test_body": "define i3 @foo3x5_nsw(i3 %x) {\n %tmp1 = mul i3 %x, %x\n %tmp2 = mul i3 %tmp1, %x\n %tmp3 = mul i3 %tmp2, %x\n %tmp4 = mul nsw i3 %tmp3, %x\n ret i3 %tmp4\n}\n"}, {"test_name": "foo3x5", "test_body": "define i3 @foo3x5(i3 %x) {\n %tmp1 = mul i3 %x, %x\n %tmp2 = mul i3 %tmp1, %x\n %tmp3 = mul i3 %tmp2, %x\n %tmp4 = mul i3 %tmp3, %x\n ret i3 %tmp4\n}\n"}, {"test_name": "foo4x10", "test_body": "define i4 @foo4x10(i4 %x) {\n %tmp1 = mul i4 %x, %x\n %tmp2 = mul i4 %tmp1, %x\n %tmp3 = mul i4 %tmp2, %x\n %tmp4 = mul i4 %tmp3, %x\n %tmp5 = mul i4 %tmp4, %x\n %tmp6 = mul i4 %tmp5, %x\n %tmp7 = mul i4 %tmp6, %x\n %tmp8 = mul i4 %tmp7, %x\n %tmp9 = mul i4 %tmp8, %x\n ret i4 %tmp9\n}\n"}, {"test_name": "foo4x13", "test_body": "define i4 @foo4x13(i4 %x) {\n %tmp1 = mul i4 %x, %x\n %tmp2 = mul i4 %tmp1, %x\n %tmp3 = mul i4 %tmp2, %x\n %tmp4 = mul i4 %tmp3, %x\n %tmp5 = mul i4 %tmp4, %x\n %tmp6 = mul i4 %tmp5, %x\n %tmp7 = mul i4 %tmp6, %x\n %tmp8 = mul i4 %tmp7, %x\n %tmp9 = mul i4 %tmp8, %x\n %tmp10 = mul i4 %tmp9, %x\n %tmp11 = mul i4 %tmp10, %x\n %tmp12 = mul i4 %tmp11, %x\n ret i4 %tmp12\n}\n"}, {"test_name": "foo4x8", "test_body": "define i4 @foo4x8(i4 %x) {\n %tmp1 = mul i4 %x, %x\n %tmp2 = mul i4 %tmp1, %x\n %tmp3 = mul i4 %tmp2, %x\n %tmp4 = mul i4 %tmp3, %x\n %tmp5 = mul i4 %tmp4, %x\n %tmp6 = mul i4 %tmp5, %x\n %tmp7 = mul i4 %tmp6, %x\n ret i4 %tmp7\n}\n"}, {"test_name": "foo4x12", "test_body": "define i4 @foo4x12(i4 %x) {\n %tmp1 = mul i4 %x, %x\n %tmp2 = mul i4 %tmp1, %x\n %tmp3 = mul i4 %tmp2, %x\n %tmp4 = mul i4 %tmp3, %x\n %tmp5 = mul i4 %tmp4, %x\n %tmp6 = mul i4 %tmp5, %x\n %tmp7 = mul i4 %tmp6, %x\n %tmp8 = mul i4 %tmp7, %x\n %tmp9 = mul i4 %tmp8, %x\n %tmp10 = mul i4 %tmp9, %x\n %tmp11 = mul i4 %tmp10, %x\n ret i4 %tmp11\n}\n"}, {"test_name": "foo4x15", "test_body": "define i4 @foo4x15(i4 %x) {\n %tmp1 = mul i4 %x, %x\n %tmp2 = mul i4 %tmp1, %x\n %tmp3 = mul i4 %tmp2, %x\n %tmp4 = mul i4 %tmp3, %x\n %tmp5 = mul i4 %tmp4, %x\n %tmp6 = mul i4 %tmp5, %x\n %tmp7 = mul i4 %tmp6, %x\n %tmp8 = mul i4 %tmp7, %x\n %tmp9 = mul i4 %tmp8, %x\n %tmp10 = mul i4 %tmp9, %x\n %tmp11 = mul i4 %tmp10, %x\n %tmp12 = mul i4 %tmp11, %x\n %tmp13 = mul i4 %tmp12, %x\n %tmp14 = mul i4 %tmp13, %x\n ret i4 %tmp14\n}\n"}, {"test_name": "foo4x9", "test_body": "define i4 @foo4x9(i4 %x) {\n %tmp1 = mul i4 %x, %x\n %tmp2 = mul i4 %tmp1, %x\n %tmp3 = mul i4 %tmp2, %x\n %tmp4 = mul i4 %tmp3, %x\n %tmp5 = mul i4 %tmp4, %x\n %tmp6 = mul i4 %tmp5, %x\n %tmp7 = mul i4 %tmp6, %x\n %tmp8 = mul i4 %tmp7, %x\n ret i4 %tmp8\n}\n"}, {"test_name": "foo4x14", "test_body": "define i4 @foo4x14(i4 %x) {\n %tmp1 = mul i4 %x, %x\n %tmp2 = mul i4 %tmp1, %x\n %tmp3 = mul i4 %tmp2, %x\n %tmp4 = mul i4 %tmp3, %x\n %tmp5 = mul i4 %tmp4, %x\n %tmp6 = mul i4 %tmp5, %x\n %tmp7 = mul i4 %tmp6, %x\n %tmp8 = mul i4 %tmp7, %x\n %tmp9 = mul i4 %tmp8, %x\n %tmp10 = mul i4 %tmp9, %x\n %tmp11 = mul i4 %tmp10, %x\n %tmp12 = mul i4 %tmp11, %x\n %tmp13 = mul i4 %tmp12, %x\n ret i4 %tmp13\n}\n"}, {"test_name": "foo3x6", "test_body": "define i3 @foo3x6(i3 %x) {\n %tmp1 = mul i3 %x, %x\n %tmp2 = mul i3 %tmp1, %x\n %tmp3 = mul i3 %tmp2, %x\n %tmp4 = mul i3 %tmp3, %x\n %tmp5 = mul i3 %tmp4, %x\n ret i3 %tmp5\n}\n"}, {"test_name": "foo3x7", "test_body": "define i3 @foo3x7(i3 %x) {\n %tmp1 = mul i3 %x, %x\n %tmp2 = mul i3 %tmp1, %x\n %tmp3 = mul i3 %tmp2, %x\n %tmp4 = mul i3 %tmp3, %x\n %tmp5 = mul i3 %tmp4, %x\n %tmp6 = mul i3 %tmp5, %x\n ret i3 %tmp6\n}\n"}]}], "issue": {"title": "incorrect reassociation at low bitwidth", "body": "here's a function:\r\n```llvm\r\ndefine i3 @f(i3 %0) {\r\n %2 = mul i3 %0, %0\r\n %3 = mul i3 %2, %0\r\n %4 = mul i3 %3, %0\r\n %5 = mul nsw i3 %4, %0\r\n ret i3 %5\r\n}\r\n```\r\n\r\nreassociate seems to be getting it wrong:\r\n```\r\nJohns-MacBook-Pro:reduce regehr$ ~/alive2-regehr/build/alive-tv reduced.ll -passes=reassociate\r\n\r\n----------------------------------------\r\ndefine i3 @f(i3 %#0) {\r\n#1:\r\n %#2 = mul i3 %#0, %#0\r\n %#3 = mul i3 %#2, %#0\r\n %#4 = mul i3 %#3, %#0\r\n %#5 = mul nsw i3 %#4, %#0\r\n ret i3 %#5\r\n}\r\n=>\r\ndefine i3 @f(i3 %#0) {\r\n#1:\r\n %#2 = mul i3 %#0, %#0\r\n %#3 = mul nsw i3 %#2, %#0\r\n ret i3 %#3\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\ni3 %#0 = #x6 (6, -2)\r\n\r\nSource:\r\ni3 %#2 = #x4 (4, -4)\r\ni3 %#3 = #x0 (0)\r\ni3 %#4 = #x0 (0)\r\ni3 %#5 = #x0 (0)\r\n\r\nTarget:\r\ni3 %#2 = #x4 (4, -4)\r\ni3 %#3 = poison\r\nSource value: #x0 (0)\r\nTarget value: poison\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\nJohns-MacBook-Pro:reduce regehr$ \r\n```\r\n\r\ncc @nunoplopes ", "author": "regehr", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "dtcxzyw", "body": "Introduced by https://github.com/llvm/llvm-project/commit/d7aeefebd6b049f017711cd7c6ef5f217a17b673.\r\n"}]}, "verified": true} {"bug_id": "100464", "issue_url": "https://github.com/llvm/llvm-project/issues/100464", "bug_type": "crash", "base_commit": "ba461f8c6278a7b2fd7695454c1f184c38897ecd", "knowledge_cutoff": "2024-07-24T20:37:49Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "b72689a5cbd8645c183476cd87e32948308c5b64", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6678, 6683], [6693, 6700], [6727, 6742]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::collectValuesToIgnore"]}}, "patch": "commit b72689a5cbd8645c183476cd87e32948308c5b64\nAuthor: Florian Hahn \nDate: Thu Jul 25 11:16:09 2024 +0100\n\n [LV] Ignore live-out users in cost model if scalar epilogue is required.\n \n Follow-up to ba8126b6fef79.\n \n If a scalar epilogue is required, users outside the loop won't use\n live-outs from the vector loop but from the scalar epilogue. Ignore them if\n that is the case.\n \n This fixes another case where the VPlan-based cost-model more accurately\n computes cost.\n \n Fixes https://github.com/llvm/llvm-project/issues/100464.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex b709ccc3c35a..224d98d5c4f9 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6678,6 +6678,15 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n \n SmallVector DeadInterleavePointerOps;\n SmallVector DeadOps;\n+\n+ // If a scalar epilogue is required, users outside the loop won't use\n+ // live-outs from the vector loop but from the scalar epilogue. Ignore them if\n+ // that is the case.\n+ bool RequiresScalarEpilogue = requiresScalarEpilogue(true);\n+ auto IsLiveOutDead = [this, RequiresScalarEpilogue](User *U) {\n+ return RequiresScalarEpilogue &&\n+ !TheLoop->contains(cast(U)->getParent());\n+ };\n for (BasicBlock *BB : TheLoop->blocks())\n for (Instruction &I : *BB) {\n // Find all stores to invariant variables. Since they are going to sink\n@@ -6693,8 +6702,9 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n // Add instructions that would be trivially dead and are only used by\n // values already ignored to DeadOps to seed worklist.\n if (wouldInstructionBeTriviallyDead(&I, TLI) &&\n- all_of(I.users(), [this](User *U) {\n- return VecValuesToIgnore.contains(U) || ValuesToIgnore.contains(U);\n+ all_of(I.users(), [this, IsLiveOutDead](User *U) {\n+ return VecValuesToIgnore.contains(U) ||\n+ ValuesToIgnore.contains(U) || IsLiveOutDead(U);\n }))\n DeadOps.push_back(&I);\n \n@@ -6727,16 +6737,22 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n \n // Mark ops that would be trivially dead and are only used by ignored\n // instructions as free.\n+ BasicBlock *Header = TheLoop->getHeader();\n for (unsigned I = 0; I != DeadOps.size(); ++I) {\n auto *Op = dyn_cast(DeadOps[I]);\n // Skip any op that shouldn't be considered dead.\n if (!Op || !TheLoop->contains(Op) ||\n+ (isa(Op) && Op->getParent() == Header) ||\n !wouldInstructionBeTriviallyDead(Op, TLI) ||\n- any_of(Op->users(), [this](User *U) {\n- return !VecValuesToIgnore.contains(U) && !ValuesToIgnore.contains(U);\n+ any_of(Op->users(), [this, IsLiveOutDead](User *U) {\n+ return !VecValuesToIgnore.contains(U) && ValuesToIgnore.contains(U) &&\n+ !IsLiveOutDead(U);\n }))\n continue;\n \n+ if (!TheLoop->contains(Op->getParent()))\n+ continue;\n+\n // If all of Op's users are in ValuesToIgnore, add it to ValuesToIgnore\n // which applies for both scalar and vector versions. Otherwise it is only\n // dead in vector versions, so only add it to VecValuesToIgnore.\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll", "commands": ["opt -p loop-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S %s"], "tests": [{"test_name": "dead_live_out_due_to_scalar_epilogue_required", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine i8 @dead_live_out_due_to_scalar_epilogue_required(ptr %src, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %idxprom = sext i32 %iv to i64\n %gep.src = getelementptr i8, ptr %src, i64 %idxprom\n %l = load i8, ptr %gep.src, align 1\n %gep.dst = getelementptr i8, ptr %dst, i64 %idxprom\n store i8 0, ptr %gep.dst, align 1\n %iv.next = add i32 %iv, 4\n %cmp = icmp ult i32 %iv, 1001\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n %r = phi i8 [ %l, %loop ]\n ret i8 %r\n}\n"}, {"test_name": "dead_load", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @dead_load(ptr %p, i16 %start) {\nentry:\n %start.ext = sext i16 %start to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %start.ext, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr i16, ptr %p, i64 %iv\n store i16 0, ptr %gep, align 2\n %l = load i16, ptr %gep, align 2\n %iv.next = add i64 %iv, 3\n %cmp = icmp slt i64 %iv, 111\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/RISCV/riscv-vector-reverse.ll", "commands": ["opt -passes=loop-vectorize,dce,instcombine -mtriple riscv64-linux-gnu -mattr=+v -debug-only=loop-vectorize -scalable-vectorization=on -riscv-v-vector-bits-min=128 -disable-output < %s 2>&1"], "tests": [{"test_name": "vector_reverse_i64", "test_body": "define void @vector_reverse_i64(ptr nocapture noundef writeonly %A, ptr nocapture noundef readonly %B, i32 noundef signext %n) {\nentry:\n %cmp7 = icmp sgt i32 %n, 0\n br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup\n\nfor.body.preheader: ; preds = %entry\n %0 = zext i32 %n to i64\n br label %for.body\n\nfor.cond.cleanup: ; preds = %for.body, %entry\n ret void\n\nfor.body: ; preds = %for.body, %for.body.preheader\n %indvars.iv = phi i64 [ %0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]\n %i.0.in8 = phi i32 [ %n, %for.body.preheader ], [ %i.0, %for.body ]\n %i.0 = add nsw i32 %i.0.in8, -1\n %idxprom = zext i32 %i.0 to i64\n %arrayidx = getelementptr inbounds i32, ptr %B, i64 %idxprom\n %1 = load i32, ptr %arrayidx, align 4\n %add9 = add i32 %1, 1\n %arrayidx3 = getelementptr inbounds i32, ptr %A, i64 %idxprom\n store i32 %add9, ptr %arrayidx3, align 4\n %cmp = icmp ugt i64 %indvars.iv, 1\n %indvars.iv.next = add nsw i64 %indvars.iv, -1\n br i1 %cmp, label %for.body, label %for.cond.cleanup, !llvm.loop !0\n}\n\n!0 = distinct !{!0, !1, !2, !3, !4}\n!1 = !{!\"llvm.loop.mustprogress\"}\n!2 = !{!\"llvm.loop.vectorize.width\", i32 4}\n!3 = !{!\"llvm.loop.vectorize.scalable.enable\", i1 true}\n!4 = !{!\"llvm.loop.vectorize.enable\", i1 true}\n"}, {"test_name": "vector_reverse_f32", "test_body": "define void @vector_reverse_f32(ptr nocapture noundef writeonly %A, ptr nocapture noundef readonly %B, i32 noundef signext %n) {\nentry:\n %cmp7 = icmp sgt i32 %n, 0\n br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup\n\nfor.body.preheader: ; preds = %entry\n %0 = zext i32 %n to i64\n br label %for.body\n\nfor.cond.cleanup: ; preds = %for.body, %entry\n ret void\n\nfor.body: ; preds = %for.body, %for.body.preheader\n %indvars.iv = phi i64 [ %0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]\n %i.0.in8 = phi i32 [ %n, %for.body.preheader ], [ %i.0, %for.body ]\n %i.0 = add nsw i32 %i.0.in8, -1\n %idxprom = zext i32 %i.0 to i64\n %arrayidx = getelementptr inbounds float, ptr %B, i64 %idxprom\n %1 = load float, ptr %arrayidx, align 4\n %conv1 = fadd float %1, 1.000000e+00\n %arrayidx3 = getelementptr inbounds float, ptr %A, i64 %idxprom\n store float %conv1, ptr %arrayidx3, align 4\n %cmp = icmp ugt i64 %indvars.iv, 1\n %indvars.iv.next = add nsw i64 %indvars.iv, -1\n br i1 %cmp, label %for.body, label %for.cond.cleanup, !llvm.loop !0\n}\n\n!0 = distinct !{!0, !1, !2, !3, !4}\n!1 = !{!\"llvm.loop.mustprogress\"}\n!2 = !{!\"llvm.loop.vectorize.width\", i32 4}\n!3 = !{!\"llvm.loop.vectorize.scalable.enable\", i1 true}\n!4 = !{!\"llvm.loop.vectorize.enable\", i1 true}\n"}]}], "issue": {"title": "[VPlan] VPlan cost model and legacy cost model disagreed", "body": "Related: #99701, #92555. @fhahn Please let me know if you'd prefer these assert triggering testcases to be all in the same issue or prefer a new issue when a new testcase is found.\r\n\r\nTested using llvm f719a339a89b\r\n\r\nTestcase:\r\n```llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main(ptr %0) #0 {\r\nentry:\r\n br label %for.body\r\n\r\nfor.cond.cleanup: ; preds = %for.body\r\n %.lcssa = phi i8 [ %1, %for.body ]\r\n ret i32 0\r\n\r\nfor.body: ; preds = %for.body, %entry\r\n %l.011 = phi i32 [ 0, %entry ], [ %add, %for.body ]\r\n %idxprom = sext i32 %l.011 to i64\r\n %arrayidx1 = getelementptr [15 x i8], ptr %0, i64 0, i64 %idxprom\r\n %1 = load i8, ptr %arrayidx1, align 1\r\n %arrayidx3 = getelementptr [0 x i8], ptr null, i64 0, i64 %idxprom\r\n store i8 0, ptr %arrayidx3, align 1\r\n %add = add i32 %l.011, 4\r\n %cmp = icmp ult i32 %l.011, -9\r\n br i1 %cmp, label %for.body, label %for.cond.cleanup\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10098: bool llvm::LoopVectorizePass::processLoop(llvm::Loop*): Assertion `VF.Width == Width && \"VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"main\"\r\n #0 0x0000647de5790dd0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f1ddd0)\r\n #1 0x0000647de578e1ef llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f1b1ef)\r\n #2 0x0000647de578e345 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007b2a02842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007b2a028969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007b2a028969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007b2a028969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007b2a02842476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007b2a028287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007b2a0282871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007b2a02839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x0000647de47acc98 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f39c98)\r\n#12 0x0000647de47af851 llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f3c851)\r\n#13 0x0000647de47affd0 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f3cfd0)\r\n#14 0x0000647de3621ff6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdaeff6)\r\n#15 0x0000647de55a8e7e llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d35e7e)\r\n#16 0x0000647de3625186 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdb2186)\r\n#17 0x0000647de55a7c0b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d34c0b)\r\n#18 0x0000647de36214d6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdae4d6)\r\n#19 0x0000647de55a5d5d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d32d5d)\r\n#20 0x0000647de2e6b5d6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5f85d6)\r\n#21 0x0000647de2e5d211 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5ea211)\r\n#22 0x00007b2a02829d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#23 0x00007b2a02829e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#24 0x00007b2a02829e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#25 0x0000647de2e530a5 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5e00a5)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\n\r\nGodbolt: https://godbolt.org/z/z7ee8hWWP", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "@patrick-rivos thanks for the report. I think it would be best to keep filing separate issues, as the underlying issues usually are different. Although the fuzzers may generate many instances of the same issue as was the case for https://github.com/llvm/llvm-project/issues/99701. Not sure if there's an easy way to catch that other than waiting for a day or so after filing the first issue."}]}, "verified": true} {"bug_id": "77064", "issue_url": "https://github.com/llvm/llvm-project/issues/77064", "bug_type": "crash", "base_commit": "255f95a40377677dd762df5a1aa65bcbb4f75c79", "knowledge_cutoff": "2024-01-05T09:28:42Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "a001e9718fd974859f2797a9f9ed7bf87d364e4f", "components": ["SimplifyLibCalls"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp": [[3735, 3760]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp": ["LibCallSimplifier::optimizeCall"]}}, "patch": "commit a001e9718fd974859f2797a9f9ed7bf87d364e4f\nAuthor: Nikita Popov \nDate: Fri Jan 5 10:41:44 2024 +0100\n\n [SimplifyLibCalls] Don't try to manually reprocess calls\n \n The current code for reprocessing the result of fortified libcall\n simplifications is not correct, because we might simplify to an\n argument of the original call, and if that is again a libcall,\n mistakenly think that this is actually the simplification result.\n \n Instead of trying to fix this, simply remove the code entirely,\n because InstCombine nowadays correctly handles reprocessing of\n SimplifyLibCall results.\n \n Fixes https://github.com/llvm/llvm-project/issues/77064.\n\ndiff --git a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp\nindex 760a626c8b6f..a7cd68e860e4 100644\n--- a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp\n+++ b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp\n@@ -3735,26 +3735,8 @@ Value *LibCallSimplifier::optimizeCall(CallInst *CI, IRBuilderBase &Builder) {\n \n // Also try to simplify calls to fortified library functions.\n if (Value *SimplifiedFortifiedCI =\n- FortifiedSimplifier.optimizeCall(CI, Builder)) {\n- // Try to further simplify the result.\n- CallInst *SimplifiedCI = dyn_cast(SimplifiedFortifiedCI);\n- if (SimplifiedCI && SimplifiedCI->getCalledFunction()) {\n- // Ensure that SimplifiedCI's uses are complete, since some calls have\n- // their uses analyzed.\n- replaceAllUsesWith(CI, SimplifiedCI);\n-\n- // Set insertion point to SimplifiedCI to guarantee we reach all uses\n- // we might replace later on.\n- IRBuilderBase::InsertPointGuard Guard(Builder);\n- Builder.SetInsertPoint(SimplifiedCI);\n- if (Value *V = optimizeStringMemoryLibCall(SimplifiedCI, Builder)) {\n- // If we were able to further simplify, remove the now redundant call.\n- substituteInParent(SimplifiedCI, V);\n- return V;\n- }\n- }\n+ FortifiedSimplifier.optimizeCall(CI, Builder))\n return SimplifiedFortifiedCI;\n- }\n \n // Then check for known library functions.\n if (TLI->getLibFunc(*Callee, Func) && isLibFuncEmittable(M, TLI, Func)) {\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/pr77064.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "main", "test_body": "define void @main(ptr %ptr) {\n %opendir = call fastcc ptr @opendir(ptr %ptr)\n %memset = call ptr @__memset_chk(ptr %opendir, i32 0, i64 596, i64 -1)\n ret void\n}\n\ndeclare ptr @__memset_chk(ptr, i32, i64, i64)\n\ndeclare fastcc ptr @opendir(ptr)\n"}]}], "issue": {"title": "[SimplifyLibCalls] Calling convention assertion fails", "body": "```llvm\r\n; RUN: opt -S -passes=instcombine\r\ndefine void @main(ptr %ptr) {\r\n %opendir = call fastcc ptr @opendir(ptr %ptr)\r\n %memset = call ptr @__memset_chk(ptr %opendir, i32 0, i64 596, i64 -1)\r\n ret void\r\n}\r\n\r\ndeclare ptr @__memset_chk(ptr, i32, i64, i64)\r\n\r\ndeclare fastcc ptr @opendir(ptr)\r\n```\r\n\r\n> opt: /home/npopov/repos/llvm-project/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp:3488: Value *llvm::LibCallSimplifier::optimizeStringMemoryLibCall(CallInst *, IRBuilderBase &): Assertion `(ignoreCallingConv(Func) || TargetLibraryInfoImpl::isCallingConvCCompatible(CI)) && \"Optimizing string/memory libcall would change the calling convention\"' failed.\r\n\r\nEncountered when building openjpeg2.", "author": "nikic", "labels": ["llvm:crash", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "115732", "issue_url": "https://github.com/llvm/llvm-project/issues/115732", "bug_type": "crash", "base_commit": "da78ac5d331953d3386fd56cd7979022be7400cf", "knowledge_cutoff": "2024-11-11T16:03:33Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "058ac837bc35419bbbb34f3206f5aa229c669811", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[3045, 3051], [9167, 9174], [13454, 13460], [13560, 13566], [14022, 14028]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::gather", "BoUpSLP::setInsertPointAfterBundle", "createFreeze", "createShuffle"]}}, "patch": "commit 058ac837bc35419bbbb34f3206f5aa229c669811\nAuthor: Alexey Bataev \nDate: Mon Nov 11 10:48:15 2024 -0800\n\n [SLP]Use generic createShuffle for buildvector\n \n Use generic createShuffle function, which know how to adjust the vectors\n correctly, to avoid compiler crash when trying to build a buildvector as\n a shuffle\n \n Fixes #115732\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex da8e0d8cc09a..1bf082d57b8b 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -3045,7 +3045,9 @@ private:\n \n /// \\returns a vector from a collection of scalars in \\p VL. if \\p Root is not\n /// specified, the starting vector value is poison.\n- Value *gather(ArrayRef VL, Value *Root, Type *ScalarTy);\n+ Value *\n+ gather(ArrayRef VL, Value *Root, Type *ScalarTy,\n+ function_ref)> CreateShuffle);\n \n /// \\returns whether the VectorizableTree is fully vectorizable and will\n /// be beneficial even the tree height is tiny.\n@@ -9167,8 +9169,9 @@ protected:\n int VF = Mask.size();\n if (auto *FTy = dyn_cast(V1->getType()))\n VF = FTy->getNumElements();\n- if (V2 &&\n- !isUndefVector(V2, buildUseMask(VF, Mask, UseMask::SecondArg)).all()) {\n+ if (V2 && !isUndefVector(\n+ V2, buildUseMask(VF, Mask, UseMask::SecondArg))\n+ .all()) {\n // Peek through shuffles.\n Value *Op1 = V1;\n Value *Op2 = V2;\n@@ -13454,7 +13457,9 @@ void BoUpSLP::setInsertPointAfterBundle(const TreeEntry *E) {\n Builder.SetCurrentDebugLocation(Front->getDebugLoc());\n }\n \n-Value *BoUpSLP::gather(ArrayRef VL, Value *Root, Type *ScalarTy) {\n+Value *BoUpSLP::gather(\n+ ArrayRef VL, Value *Root, Type *ScalarTy,\n+ function_ref)> CreateShuffle) {\n // List of instructions/lanes from current block and/or the blocks which are\n // part of the current loop. These instructions will be inserted at the end to\n // make it possible to optimize loops and hoist invariant instructions out of\n@@ -13560,7 +13565,7 @@ Value *BoUpSLP::gather(ArrayRef VL, Value *Root, Type *ScalarTy) {\n if (isa(Vec)) {\n Vec = OriginalRoot;\n } else {\n- Vec = Builder.CreateShuffleVector(Root, Vec, Mask);\n+ Vec = CreateShuffle(Root, Vec, Mask);\n if (auto *OI = dyn_cast(OriginalRoot);\n OI && OI->hasNUses(0))\n eraseInstruction(OI);\n@@ -14022,7 +14027,10 @@ public:\n }\n Value *gather(ArrayRef VL, unsigned MaskVF = 0,\n Value *Root = nullptr) {\n- return R.gather(VL, Root, ScalarTy);\n+ return R.gather(VL, Root, ScalarTy,\n+ [&](Value *V1, Value *V2, ArrayRef Mask) {\n+ return createShuffle(V1, V2, Mask);\n+ });\n }\n Value *createFreeze(Value *V) { return Builder.CreateFreeze(V); }\n /// Finalize emission of the shuffles.\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/buildvector-shuffle-with-root.ll", "commands": ["opt -S --passes=slp-vectorizer -slp-threshold=-99999 -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "", "test_body": "\ndefine void @test(i16 %arg) {\n;\nbb:\n %sitofp = sitofp i16 %arg to float\n %fadd = fadd float 0.000000e+00, 0.000000e+00\n %fsub = fsub float 0.000000e+00, %sitofp\n %fsub1 = fsub float 0.000000e+00, %sitofp\n %fsub2 = fsub float 0.000000e+00, %sitofp\n %sitofp3 = sitofp i16 0 to float\n %fsub4 = fsub float %fadd, %sitofp3\n store float %fsub4, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 20), align 4\n %fsub5 = fsub float %fsub, %sitofp\n store float %fsub5, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 24), align 4\n %fsub6 = fsub float %fsub1, %sitofp\n store float %fsub6, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 28), align 4\n %fsub7 = fsub float %fsub2, %sitofp\n store float %fsub7, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 32), align 4\n ret void\n}"}]}], "issue": {"title": "Assertion `isValidOperands(V1, V2, Mask) && \"Invalid shuffle vector instruction operands!\"' failed.", "body": "To reproduce run the test below with -passes=slp-vectorizer.\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat(i16 %arg) gc \"statepoint-example\" {\r\nbb:\r\n %sitofp = sitofp i16 %arg to float\r\n %fadd = fadd float 0.000000e+00, 0.000000e+00\r\n %fsub = fsub float 0.000000e+00, %sitofp\r\n %fsub1 = fsub float 0.000000e+00, %sitofp\r\n %fsub2 = fsub float 0.000000e+00, %sitofp\r\n %sitofp3 = sitofp i16 0 to float\r\n %fsub4 = fsub float %fadd, %sitofp3\r\n store float %fsub4, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 20), align 4\r\n %fsub5 = fsub float %fsub, %sitofp\r\n store float %fsub5, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 24), align 4\r\n %fsub6 = fsub float %fsub1, %sitofp\r\n store float %fsub6, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 28), align 4\r\n %fsub7 = fsub float %fsub2, %sitofp\r\n store float %fsub7, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 32), align 4\r\n ret void\r\n}\r\n```\r\nReproducer: https://godbolt.org/z/6h4xzhGa1\r\nStack dump:\r\n```\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"wombat\"\r\n #0 0x00000000051494c8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x51494c8)\r\n #1 0x0000000005146ebc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000072f583642520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000072f5836969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000072f583642476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000072f5836287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000072f58362871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000072f583639e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000004eb74ce llvm::ShuffleVectorInst::ShuffleVectorInst(llvm::Value*, llvm::Value*, llvm::ArrayRef, llvm::Twine const&, llvm::InsertPosition) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4eb74ce)\r\n #9 0x0000000000d09e93 llvm::IRBuilderBase::CreateShuffleVector(llvm::Value*, llvm::Value*, llvm::ArrayRef, llvm::Twine const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xd09e93)\r\n#10 0x00000000041a0bdd llvm::slpvectorizer::BoUpSLP::gather(llvm::ArrayRef, llvm::Value*, llvm::Type*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41a0bdd)\r\n#11 0x00000000041a13fc void llvm::function_ref&)>::callback_fn, llvm::slpvectorizer::BoUpSLP>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::IRBuilder&, llvm::slpvectorizer::BoUpSLP&)::'lambda15'(llvm::Value*&, llvm::SmallVectorImpl&)>(long, llvm::Value*&, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41a13fc)\r\n#12 0x00000000041a26b9 llvm::slpvectorizer::BoUpSLP::ShuffleInstructionBuilder::finalize(llvm::ArrayRef, llvm::ArrayRef>, llvm::ArrayRef, unsigned int, llvm::function_ref&)>) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41a26b9)\r\n#13 0x00000000041d5335 llvm::Value* llvm::slpvectorizer::BoUpSLP::processBuildVector, llvm::slpvectorizer::BoUpSLP>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::IRBuilder&, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41d5335)\r\n#14 0x00000000041e168e llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41e168e)\r\n#15 0x000000000420234a llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::SmallDenseSet> const&, llvm::Instruction*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x420234a)\r\n#16 0x0000000004205c7d llvm::slpvectorizer::BoUpSLP::vectorizeTree() (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4205c7d)\r\n#17 0x0000000004224402 llvm::SLPVectorizerPass::vectorizeStoreChain(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, unsigned int, unsigned int, unsigned int&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4224402)\r\n#18 0x0000000004225e6b llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::'lambda'(std::set, llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::StoreDistCompare, std::allocator>> const&)::operator()(std::set, llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::StoreDistCompare, std::allocator>> const&) const SLPVectorizer.cpp:0:0\r\n#19 0x0000000004227d98 llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4227d98)\r\n#20 0x00000000042285e9 llvm::SLPVectorizerPass::vectorizeStoreChains(llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42285e9)\r\n#21 0x0000000004229b35 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#22 0x000000000422a63b llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x422a63b)\r\n#23 0x0000000002f9948e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f9948e)\r\n#24 0x0000000004f47488 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f47488)\r\n#25 0x0000000000e3bf3e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3bf3e)\r\n#26 0x0000000004f45eae llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f45eae)\r\n#27 0x0000000000e3b64e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3b64e)\r\n#28 0x0000000004f458e0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f458e0)\r\n#29 0x000000000091103a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x91103a)\r\n#30 0x000000000090399a optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x90399a)\r\n#31 0x000072f583629d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#32 0x000072f583629e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#33 0x00000000008fb41e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fb41e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "crash"], "comments": [{"author": "RKSimon", "body": "CC @alexey-bataev "}]}, "verified": true} {"bug_id": "116691", "issue_url": "https://github.com/llvm/llvm-project/issues/116691", "bug_type": "crash", "base_commit": "a4e1a3dc8bc9bb971d8a38130254b4570f8b7a03", "knowledge_cutoff": "2024-11-18T20:36:44Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "ad9c0b369e86e75d56e229f294782a4eaf527226", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[6815, 6830], [6861, 6866], [6903, 6909]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::tryToVectorizeGatheredLoads"]}}, "patch": "commit ad9c0b369e86e75d56e229f294782a4eaf527226\nAuthor: Alexey Bataev \nDate: Mon Nov 18 13:49:04 2024 -0800\n\n [SLP]Check if the gathered loads form full vector before attempting build it\n \n Need to check that the number of gathered loads in the slice forms the\n build vector to avoid compiler crash.\n \n Fixes #116691\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 918d7663548f..dc0dffd9fcbf 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -6815,16 +6815,7 @@ void BoUpSLP::tryToVectorizeGatheredLoads(\n // Check if it is profitable to try vectorizing gathered loads. It is\n // profitable if we have more than 3 consecutive loads or if we have\n // less but all users are vectorized or deleted.\n- bool AllowToVectorize =\n- NumElts >= 3 ||\n- any_of(ValueToGatherNodes.at(Slice.front()),\n- [=](const TreeEntry *TE) {\n- return TE->Scalars.size() == 2 &&\n- ((TE->Scalars.front() == Slice.front() &&\n- TE->Scalars.back() == Slice.back()) ||\n- (TE->Scalars.front() == Slice.back() &&\n- TE->Scalars.back() == Slice.front()));\n- });\n+ bool AllowToVectorize = false;\n // Check if it is profitable to vectorize 2-elements loads.\n if (NumElts == 2) {\n bool IsLegalBroadcastLoad = TTI->isLegalBroadcastLoad(\n@@ -6861,6 +6852,19 @@ void BoUpSLP::tryToVectorizeGatheredLoads(\n return true;\n };\n AllowToVectorize = CheckIfAllowed(Slice);\n+ } else {\n+ AllowToVectorize =\n+ (NumElts >= 3 ||\n+ any_of(ValueToGatherNodes.at(Slice.front()),\n+ [=](const TreeEntry *TE) {\n+ return TE->Scalars.size() == 2 &&\n+ ((TE->Scalars.front() == Slice.front() &&\n+ TE->Scalars.back() == Slice.back()) ||\n+ (TE->Scalars.front() == Slice.back() &&\n+ TE->Scalars.back() == Slice.front()));\n+ })) &&\n+ hasFullVectorsOrPowerOf2(*TTI, Slice.front()->getType(),\n+ Slice.size());\n }\n if (AllowToVectorize) {\n SmallVector PointerOps;\n@@ -6903,7 +6907,8 @@ void BoUpSLP::tryToVectorizeGatheredLoads(\n }\n // Mark masked gathers candidates as vectorized, if any.\n for (unsigned Cnt : MaskedGatherVectorized) {\n- ArrayRef Slice = ArrayRef(Loads).slice(Cnt, NumElts);\n+ ArrayRef Slice = ArrayRef(Loads).slice(\n+ Cnt, std::min(NumElts, Loads.size() - Cnt));\n ArrayRef Values(\n reinterpret_cast(Slice.begin()), Slice.size());\n Results.emplace_back(Values, LoadsState::ScatterVectorize);\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/gathered-loads-non-full-reg.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux -mcpu=cascadelake < %s"], "tests": [{"test_name": "", "test_body": "\n@solid_ = external global [608 x i8]\n\ndefine void @test(ptr noalias %0) {\n;\n.lr.ph1019:\n %1 = getelementptr i8, ptr %0, i64 8\n %2 = getelementptr i8, ptr %0, i64 32\n %3 = getelementptr i8, ptr %0, i64 128\n %4 = getelementptr i8, ptr %0, i64 200\n %5 = getelementptr i8, ptr %0, i64 208\n %6 = getelementptr i8, ptr %0, i64 232\n %7 = getelementptr i8, ptr %0, i64 288\n %8 = getelementptr i8, ptr %0, i64 320\n %9 = getelementptr i8, ptr %0, i64 304\n %10 = getelementptr i8, ptr %0, i64 424\n %11 = getelementptr i8, ptr %0, i64 480\n %12 = getelementptr i8, ptr %0, i64 504\n %13 = getelementptr i8, ptr %0, i64 632\n %14 = getelementptr i8, ptr %0, i64 720\n %15 = load double, ptr %1, align 8\n %16 = load double, ptr %2, align 8\n %17 = fadd double %16, %15\n %18 = load double, ptr %3, align 8\n %19 = load double, ptr %4, align 8\n %20 = load double, ptr %5, align 8\n %21 = load double, ptr %6, align 8\n %22 = fadd double %21, %20\n %23 = load double, ptr %7, align 8\n %24 = load double, ptr %8, align 8\n %25 = load double, ptr %9, align 8\n %26 = load double, ptr %10, align 8\n %27 = load double, ptr %11, align 8\n %28 = load double, ptr %12, align 8\n %29 = fadd double %28, %27\n %30 = fmul double %22, %18\n %31 = fmul double %30, 0.000000e+00\n %32 = fsub double 0.000000e+00, %25\n %33 = fmul double %32, 0.000000e+00\n %34 = fadd double %33, 0.000000e+00\n %35 = fmul double %34, 0.000000e+00\n %36 = fmul double %29, %26\n %37 = fmul double %36, 0.000000e+00\n %38 = fadd double %37, 0.000000e+00\n %39 = fsub double %17, %19\n %40 = fmul double %39, %23\n %41 = fmul double %40, 0.000000e+00\n %42 = load double, ptr %0, align 8\n %43 = load double, ptr %13, align 8\n %44 = fmul double %43, %31\n %45 = load double, ptr %14, align 8\n %46 = fmul double %35, 0.000000e+00\n %47 = fadd double %44, 0.000000e+00\n %48 = fmul double %45, %38\n %49 = fmul double %45, %41\n store double %46, ptr getelementptr inbounds (i8, ptr @solid_, i64 384), align 8\n store double %47, ptr getelementptr inbounds (i8, ptr @solid_, i64 408), align 8\n store double %48, ptr getelementptr inbounds (i8, ptr @solid_, i64 392), align 8\n store double %49, ptr getelementptr inbounds (i8, ptr @solid_, i64 400), align 8\n %.neg965 = fmul double %48, %24\n %reass.add993 = fadd double %.neg965, 0.000000e+00\n %50 = fadd double %42, %reass.add993\n %51 = fsub double 0.000000e+00, %50\n store double %51, ptr getelementptr inbounds (i8, ptr @solid_, i64 296), align 8\n %.neg969 = fmul double %49, 0.000000e+00\n %reass.add996 = fadd double %.neg969, 0.000000e+00\n %52 = fadd double %45, %reass.add996\n %53 = fsub double 0.000000e+00, %52\n store double %53, ptr getelementptr inbounds (i8, ptr @solid_, i64 304), align 8\n ret void\n}"}]}], "issue": {"title": "Crash in SLPVectorizer.cpp", "body": "I'm seeing a problem that shows up in the flang compiler when compiling a large, proprietary benchmark. Since the benchmark is proprietary, I can't include the source code in this bug report.\r\n\r\nThe problem happens in in the file `.../llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp` in the function `BoUpSLP::tryToVectorizeGatheredLoads`.\r\n\r\nIn the code around line 6906, is the statement:\r\n\r\n```\r\nArrayRef Slice = ArrayRef(Loads).slice(Cnt, NumElts);\r\n```\r\n\r\nThe code for `slice()` is in `llvm-project/llvm/include/llvm/ADT/ArrayRef.h`. It contains the line:\r\n\r\n```\r\nassert(N+M <= size() && \"Invalid specifier\");\r\n```\r\n\r\nThis assertion fails, and the compiler crashes.\r\n\r\nI added debug code to `SLPVectorizer.cpp` to trace how the variables `Cnt` and `NumElts` got their values leading up to the point of failure. I've included the output of running the compiler on the test case in the file `log` along with the source code of the modified version of `SLPVectorizer.cpp` in the attached zip file.\r\n\r\nAt the point of failure, the value of `N` (which corresponds to `Cnt`) is 8. The value of `M` (passed from `NumElts`) is also 8. The call to `size()` returns 14. Since 8 + 8 is not less than or equal to 14, the assert fails and the compiler crashes.\r\n\r\nI captured the LLVM IR of the original test case and included it below in the file `x13.f90.ll` in the attached zip file.\r\n\r\nIf you run this command after downloading `x13.f90.ll`, you'll see the error:\r\n\r\n```\r\n#!/bin/bash\r\n../bin/clang -c -O3 -march=native -mllvm -locally-hot-callsite-threshold=4500 x13.f90.ll\r\n```\r\n\r\nI also ran the `x13.f90.ll` file through `llvm-reduce` using the following command:\r\n\r\n```\r\nllvm-reduce --test pcrash.sh x13.f90.ll\r\n```\r\n\r\nThis produced a file called `reduced.ll` that I've included in the attached zip file. I've also included the log file `reduced.log` that shows what happend when you run the command:\r\n\r\n```\r\n#!/bin/bash\r\n../bin/clang -c -O3 -march=native -mllvm -locally-hot-callsite-threshold=4500 reduced.ll\r\n```\r\n\r\n@alexey-bataev, please let me know if someone else should look at this or if I can help out by adding additional debug information.\r\n\r\n[Archive.zip](https://github.com/user-attachments/files/17805978/Archive.zip)\r\n", "author": "psteinfeld", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "118030", "issue_url": "https://github.com/llvm/llvm-project/issues/118030", "bug_type": "crash", "base_commit": "6f4b4f41ca41d7d0b44a32d9968aef7667c68184", "knowledge_cutoff": "2024-11-28T19:14:48Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "f4974e09312afc9549fcfc0304b738d235da9eca", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[17799, 17804]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::collectValuesToDemote"]}}, "patch": "commit f4974e09312afc9549fcfc0304b738d235da9eca\nAuthor: Alexey Bataev \nDate: Fri Nov 29 06:20:19 2024 -0800\n\n [SLP] Add a check for poison value in AShrChecker\n \n Need to check if the value in AShrChecker is a poison before casting it\n to instruction to avoid compiler crash\n \n Fixes #118030\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 48a8520a966f..7723442bc0fb 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -17799,6 +17799,8 @@ bool BoUpSLP::collectValuesToDemote(\n // original type and the sign bit of the truncate type are similar.\n auto AShrChecker = [&](unsigned BitWidth, unsigned OrigBitWidth) {\n return all_of(E.Scalars, [&](Value *V) {\n+ if (isa(V))\n+ return true;\n auto *I = cast(V);\n KnownBits AmtKnownBits = computeKnownBits(I->getOperand(1), *DL);\n unsigned ShiftedBits = OrigBitWidth - BitWidth;\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/ashr-node-with-poison.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux < %s"], "tests": [{"test_name": "", "test_body": "\ndefine i32 @test(ptr %n, i32 %conv57, i1 %tobool5.not, i64 %bf.load14) {\n;\nentry:\n br i1 false, label %entry.if.end54_crit_edge, label %while.cond\n\nentry.if.end54_crit_edge:\n br label %if.end54\n\ntthread-pre-split:\n br label %t\n\nt:\n %w.2 = phi i32 [ 1, %tthread-pre-split ], [ 0, %if.end83 ]\n br label %if.end7\n\nt.u_crit_edge:\n br label %if.end7\n\nif.end7:\n %w.4 = phi i32 [ %w.2, %t ], [ %w.2, %t.u_crit_edge ], [ %bf.cast25, %while.body ]\n %a.4 = phi i32 [ 0, %t ], [ %conv57, %t.u_crit_edge ], [ %bf.cast2910, %while.body ]\n %b.4 = phi i32 [ 0, %t ], [ 0, %t.u_crit_edge ], [ %bf.cast2910, %while.body ]\n %c.4 = phi i32 [ 0, %t ], [ 1, %t.u_crit_edge ], [ poison, %while.body ]\n br label %v\n\nwhile.cond:\n %bf.load66.pre.pre1135 = phi i64 [ 0, %entry ], [ %spec.select, %if.end42 ]\n br i1 %tobool5.not, label %if.end54, label %while.body\n\nwhile.body:\n %bf.ashr24 = ashr i64 %bf.load14, 33\n %bf.cast25 = trunc nsw i64 %bf.ashr24 to i32\n %bf.ashr28 = lshr i64 %bf.load14, 1\n %bf.cast2910 = trunc i64 %bf.ashr28 to i32\n br label %if.end7\n\nif.end36:\n br label %v\n\nv:\n %c.7 = phi i32 [ %c.4, %if.end7 ], [ 0, %if.end36 ]\n br i1 true, label %if.end42, label %v.if.end83_crit_edge\n\nv.if.end83_crit_edge:\n br label %if.end83\n\nif.end42:\n %tobool43.not = icmp eq i32 %b.4, 0\n %narrow = select i1 %tobool43.not, i32 0, i32 %w.4\n %spec.select = zext i32 %narrow to i64\n %bf.value48 = zext i32 %a.4 to i64\n store i64 %bf.value48, ptr %n, align 8\n store i32 %c.7, ptr %n, align 4\n br label %while.cond\n\nif.end54:\n %bf.load66.pre.pre113125 = phi i64 [ %bf.load66.pre.pre1135, %while.cond ], [ poison, %entry.if.end54_crit_edge ]\n %0 = icmp eq i64 %bf.load66.pre.pre113125, 0\n br i1 %0, label %if.end83, label %ai.if.end76_crit_edge\n\nai.if.end76_crit_edge:\n br label %if.end83\n\nif.end83:\n br label %t\n}"}]}], "issue": {"title": "clang crashes on valid code at -O{s,2,3} on x86_64-linux-gnu: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed", "body": "It appears to be a recent regression as it doesn't reproduce with 19.1.0 and earlier.\n\nCompiler Explorer: https://godbolt.org/z/cb67rEodY\n\n```\n[516] % clangtk -v\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 5208bc3694473f9378fc0d98eeee63e5ad5c5f04)\nTarget: x86_64-unknown-linux-gnu\nThread model: posix\nInstalledDir: /local/home/suz/suz-local/software/local/clang-trunk/bin\nBuild config: +assertions\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\nCandidate multilib: .;@m64\nSelected multilib: .;@m64\nFound CUDA installation: /usr/local/cuda, version 11.8\n[517] % \n[517] % clangtk -O2 -w small.c\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = llvm::Instruction; From = llvm::Value]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\nStack dump:\n0.\tProgram arguments: /local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20241125-clangtk-m64-O3-fpointer-tbaa-build-014338/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20241125-clangtk-m64-O3-fpointer-tbaa-build-014338/delta -resource-dir /local/home/suz/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/home/suz/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/12/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -w -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-99895f.o -x c small.c\n1.\t parser at end of file\n2.\tOptimizer\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\n4.\tRunning pass \"slp-vectorizer\" on function \"main\"\n #0 0x0000562e34749b50 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x455db50)\n #1 0x0000562e34746f5f llvm::sys::RunSignalHandlers() (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x455af5f)\n #2 0x0000562e347470b5 SignalHandler(int) Signals.cpp:0:0\n #3 0x00007f99ac5bf520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #4 0x00007f99ac613a7c pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x96a7c)\n #5 0x00007f99ac5bf476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #6 0x00007f99ac5a57f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #7 0x00007f99ac5a571b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #8 0x00007f99ac5b6e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #9 0x0000562e31712868 (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x1526868)\n#10 0x0000562e3616cf4e llvm::slpvectorizer::BoUpSLP::collectValuesToDemote(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, unsigned int&, llvm::SmallVectorImpl&, llvm::DenseSet>&, unsigned int&, bool&, bool) const::'lambda9'(unsigned int, unsigned int)::operator()(unsigned int, unsigned int) const::'lambda'(llvm::Value*)::operator()(llvm::Value*) const SLPVectorizer.cpp:0:0\n#11 0x0000562e3616d208 bool llvm::function_ref::callback_fn&, llvm::DenseSet>&, unsigned int&, bool&, bool) const::'lambda9'(unsigned int, unsigned int)>(long, unsigned int, unsigned int) SLPVectorizer.cpp:0:0\n#12 0x0000562e361c72e9 llvm::slpvectorizer::BoUpSLP::collectValuesToDemote(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, unsigned int&, llvm::SmallVectorImpl&, llvm::DenseSet>&, unsigned int&, bool&, bool) const::'lambda6'(unsigned int&, llvm::ArrayRef, llvm::function_ref)::operator()(unsigned int&, llvm::ArrayRef, llvm::function_ref) const SLPVectorizer.cpp:0:0\n#13 0x0000562e361c6065 llvm::slpvectorizer::BoUpSLP::collectValuesToDemote(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, unsigned int&, llvm::SmallVectorImpl&, llvm::DenseSet>&, unsigned int&, bool&, bool) const (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5fda065)\n#14 0x0000562e361c6d09 llvm::slpvectorizer::BoUpSLP::computeMinimumValueSizes()::'lambda0'(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, bool, unsigned int, unsigned int, bool, bool)::operator()(llvm::slpvectorizer::BoUpSLP::TreeEntry const&, bool, bool, unsigned int, unsigned int, bool, bool) const SLPVectorizer.cpp:0:0\n#15 0x0000562e361db4e3 llvm::slpvectorizer::BoUpSLP::computeMinimumValueSizes() (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5fef4e3)\n#16 0x0000562e3623dc1e llvm::SLPVectorizerPass::tryToVectorizeList(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, bool) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x6051c1e)\n#17 0x0000562e362427c0 bool tryToVectorizeSequence(llvm::SmallVectorImpl&, llvm::function_ref, llvm::function_ref, llvm::function_ref, bool)>, bool, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\n#18 0x0000562e36244162 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x6058162)\n#19 0x0000562e3624bc96 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\n#20 0x0000562e3624c802 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x6060802)\n#21 0x0000562e35c92cf6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5aa6cf6)\n#22 0x0000562e340cd78f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3ee178f)\n#23 0x0000562e31add026 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x18f1026)\n#24 0x0000562e340cbb1b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3edfb1b)\n#25 0x0000562e31adb986 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x18ef986)\n#26 0x0000562e340cc67d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3ee067d)\n#27 0x0000562e349f1d16 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\n#28 0x0000562e349f566c clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x480966c)\n#29 0x0000562e350be425 clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x4ed2425)\n#30 0x0000562e36cf2c7c clang::ParseAST(clang::Sema&, bool, bool) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x6b06c7c)\n#31 0x0000562e3539f2c9 clang::FrontendAction::Execute() (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x51b32c9)\n#32 0x0000562e353190f5 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x512d0f5)\n#33 0x0000562e354836b3 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x52976b3)\n#34 0x0000562e316862d1 cc1_main(llvm::ArrayRef, char const*, void*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x149a2d1)\n#35 0x0000562e3167cc8a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\n#36 0x0000562e31681067 clang_main(int, char**, llvm::ToolContext const&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x1495067)\n#37 0x0000562e31579a7b main (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x138da7b)\n#38 0x00007f99ac5a6d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#39 0x00007f99ac5a6e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#40 0x0000562e3167c3b5 _start (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x14903b5)\nclangtk: error: unable to execute command: Aborted\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 5208bc3694473f9378fc0d98eeee63e5ad5c5f04)\nTarget: x86_64-unknown-linux-gnu\nThread model: posix\nInstalledDir: /local/home/suz/suz-local/software/local/clang-trunk/bin\nBuild config: +assertions\nclangtk: note: diagnostic msg: \n********************\n\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\nPreprocessed source(s) and associated run script(s) are located at:\nclangtk: note: diagnostic msg: /tmp/small-522eca.c\nclangtk: note: diagnostic msg: /tmp/small-522eca.sh\nclangtk: note: diagnostic msg: \n\n********************\n[518] % \n[518] % cat small.c\nstruct {\n int b : 31;\n int c : 6;\n} f, g, j;\nint d, e, h, i, k, l, m, n, o;\nint main() {\n if (m) {\n short q = d;\n if (d)\n r:\n d = q;\n if (0) {\n s:\n t:\n if (h)\n u:\n o = g.c;\n if (n)\n goto v;\n }\n if (n)\n while (g.c) {\n if (e)\n k;\n int w = j.b, a, b = g.c, c = ~b;\n a = f.b;\n if (l) {\n c = 0 % 0;\n goto u;\n }\n if (a)\n goto s;\n l;\n v:\n if (i)\n goto ae;\n if (b)\n j.b = w;\n f.b = a;\n l = c;\n }\n short p = n;\n if (0) {\n ag:\n h = p;\n if (f.b)\n goto t;\n }\n if (j.b) {\n ae:\n e = 0;\n ai:\n if (q)\n goto r;\n }\n }\n if (j.b)\n goto ai;\n if (d)\n goto ag;\n return 0;\n}\n```", "author": "zhendongsu", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "101213", "issue_url": "https://github.com/llvm/llvm-project/issues/101213", "bug_type": "crash", "base_commit": "8364a6e108faea1a6f05f7bb53720a6fc101fed2", "knowledge_cutoff": "2024-07-30T18:00:09Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "f70f1228035c9610de38e0e376afdacb647c4ad9", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[13909, 13919]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree"]}}, "patch": "commit f70f1228035c9610de38e0e376afdacb647c4ad9\nAuthor: Alexey Bataev \nDate: Tue Jul 30 14:02:51 2024 -0700\n\n [SLP]Fix PR101213: Reuse extractelement, only if its vector operand comes before new vector value.\n \n When trying to reuse extractelement instruction, need to check that it\n is inserted into proper position. Its original vector operand should\n come before new vector value, otherwise new extractelement instruction\n must be generated.\n \n Fixes https://github.com/llvm/llvm-project/issues/101213\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 6501a14d8778..cf87b8621027 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -13909,11 +13909,16 @@ Value *BoUpSLP::vectorizeTree(\n }\n if (!Ex) {\n // \"Reuse\" the existing extract to improve final codegen.\n- if (auto *ES = dyn_cast(Scalar)) {\n+ if (auto *ES = dyn_cast(Scalar);\n+ ES && isa(Vec)) {\n Value *V = ES->getVectorOperand();\n if (const TreeEntry *ETE = getTreeEntry(V))\n V = ETE->VectorizedValue;\n- Ex = Builder.CreateExtractElement(V, ES->getIndexOperand());\n+ if (auto *IV = dyn_cast(V);\n+ !IV || IV == Vec || IV->comesBefore(cast(Vec)))\n+ Ex = Builder.CreateExtractElement(V, ES->getIndexOperand());\n+ else\n+ Ex = Builder.CreateExtractElement(Vec, Lane);\n } else if (ReplaceGEP) {\n // Leave the GEPs as is, they are free in most cases and better to\n // keep them as GEPs.\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/extract-vectorized-operand.ll", "commands": ["opt -S --passes=slp-vectorizer -slp-threshold=-99999 < %s -mtriple=x86_64-unknown-linux-gnu"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nbb:\n %0 = shufflevector <2 x ptr addrspace(1)> zeroinitializer, <2 x ptr addrspace(1)> zeroinitializer, <2 x i32> \n %1 = extractelement <2 x ptr addrspace(1)> %0, i32 0\n %2 = extractelement <2 x ptr addrspace(1)> %0, i32 1\n br label %bb43\n\nbb20: ; No predecessors!\n br label %bb105\n\nbb43: ; preds = %bb51, %bb\n %phi441 = phi ptr addrspace(1) [ %4, %bb51 ], [ %2, %bb ]\n %phi452 = phi ptr addrspace(1) [ %5, %bb51 ], [ %1, %bb ]\n br i1 false, label %bb105, label %bb51\n\nbb51: ; preds = %bb54, %bb43\n %3 = phi <2 x ptr addrspace(1)> [ poison, %bb54 ], [ zeroinitializer, %bb43 ]\n %4 = extractelement <2 x ptr addrspace(1)> %3, i32 0\n %5 = extractelement <2 x ptr addrspace(1)> %3, i32 1\n br label %bb43\n\nbb54: ; No predecessors!\n br label %bb51\n\nbb105: ; preds = %bb43, %bb20\n %phi106 = phi ptr addrspace(1) [ %1, %bb20 ], [ null, %bb43 ]\n ret void\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Instruction does not dominate all uses! LLVM ERROR: Broken module found, compilation aborted!", "body": "To reproduce run the following test with -passes slp-vectorizer -slp-threshold=-99999\r\n[Test.ll.tgz](https://github.com/user-attachments/files/16431764/Test.ll.tgz)\r\n\r\nReproducer : https://godbolt.org/z/GW41a7WP6\r\n\r\nStack dump:\r\n```\r\nInstruction does not dominate all uses!\r\n %2 = shufflevector <2 x ptr addrspace(1)> %0, <2 x ptr addrspace(1)> poison, <2 x i32> \r\n %1 = extractelement <2 x ptr addrspace(1)> %2, i32 0\r\nLLVM ERROR: Broken module found, compilation aborted!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes slp-vectorizer -slp-threshold=-99999 \r\n1.\tRunning pass \"verify\" on module \"\"\r\n #0 0x0000000004eff778 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4eff778)\r\n #1 0x0000000004efceec SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007c0269e42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007c0269e969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007c0269e42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007c0269e287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00000000007b3ff7 llvm::json::operator==(llvm::json::Value const&, llvm::json::Value const&) (.cold) JSON.cpp:0:0\r\n #7 0x0000000004e396a8 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e396a8)\r\n #8 0x0000000004d360f0 llvm::VerifierPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d360f0)\r\n #9 0x00000000008d865e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d865e)\r\n#10 0x0000000004cf9c90 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4cf9c90)\r\n#11 0x00000000008e33d2 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e33d2)\r\n#12 0x00000000008d62fc optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d62fc)\r\n#13 0x00007c0269e29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#14 0x00007c0269e29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#15 0x00000000008cd75e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8cd75e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\n", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "llvm:crash"], "comments": [{"author": "Zentrik", "body": "@alexey-bataev Hi, is there a way to backport this fix for llvm 19, I tried just cherry picking f70f1228035c9610de38e0e376afdacb647c4ad9 but I ended up getting assertion failures of the form `/workspace/srcdir/llvm-project/llvm/lib/IR/Instruction.cpp:321: bool llvm::Instruction::comesBefore(const llvm::Instruction*) const: Assertion 'getParent() == Other->getParent() && \"cross-BB instruction order comparison\"' failed.'`. \r\nI can create an IR reproducer for that failure if need be.\r\nThank you.\r\n\r\nEDIT: Would just adding a check that they're in the same basic block work?"}, {"author": "alexey-bataev", "body": "> @alexey-bataev Hi, is there a way to backport this fix for llvm 19, I tried just cherry picking [f70f122](https://github.com/llvm/llvm-project/commit/f70f1228035c9610de38e0e376afdacb647c4ad9) but I ended up getting assertion failures of the form `/workspace/srcdir/llvm-project/llvm/lib/IR/Instruction.cpp:321: bool llvm::Instruction::comesBefore(const llvm::Instruction*) const: Assertion 'getParent() == Other->getParent() && \"cross-BB instruction order comparison\"' failed.'`. I can create an IR reproducer for that failure if need be. Thank you.\r\n> \r\n> EDIT: Would just adding a check that they're in the same basic block work?\r\n\r\nYes, check for the same parent should fix it"}]}, "verified": true} {"bug_id": "114879", "issue_url": "https://github.com/llvm/llvm-project/issues/114879", "bug_type": "miscompilation", "base_commit": "3850801ca57575640a6ad3a5a421a416dc5c6f12", "knowledge_cutoff": "2024-11-04T22:12:00Z", "lit_test_dir": ["llvm/test/Transforms/IndVarSimplify", "llvm/test/Transforms/LoopStrengthReduce", "llvm/test/Transforms/LoopUnroll", "llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "0b9f1cc024ca6c7e8d60524be07c0ddfcd08b23c", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[6023, 6029]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::createNodeForPHI"]}}, "patch": "commit 0b9f1cc024ca6c7e8d60524be07c0ddfcd08b23c\nAuthor: Yingwei Zheng \nDate: Thu Nov 7 15:53:51 2024 +0800\n\n [SCEV] Disallow simplifying phi(undef, X) to X (#115109)\n \n See the following case:\n ```\n @GlobIntONE = global i32 0, align 4\n \n define ptr @src() {\n entry:\n br label %for.body.peel.begin\n \n for.body.peel.begin: ; preds = %entry\n br label %for.body.peel\n \n for.body.peel: ; preds = %for.body.peel.begin\n br i1 true, label %cleanup.peel, label %cleanup.loopexit.peel\n \n cleanup.loopexit.peel: ; preds = %for.body.peel\n br label %cleanup.peel\n \n cleanup.peel: ; preds = %cleanup.loopexit.peel, %for.body.peel\n %retval.2.peel = phi ptr [ undef, %for.body.peel ], [ @GlobIntONE, %cleanup.loopexit.peel ]\n br i1 true, label %for.body.peel.next, label %cleanup7\n \n for.body.peel.next: ; preds = %cleanup.peel\n br label %for.body.peel.next1\n \n for.body.peel.next1: ; preds = %for.body.peel.next\n br label %entry.peel.newph\n \n entry.peel.newph: ; preds = %for.body.peel.next1\n br label %for.body\n \n for.body: ; preds = %cleanup, %entry.peel.newph\n %retval.0 = phi ptr [ %retval.2.peel, %entry.peel.newph ], [ %retval.2, %cleanup ]\n br i1 false, label %cleanup, label %cleanup.loopexit\n \n cleanup.loopexit: ; preds = %for.body\n br label %cleanup\n \n cleanup: ; preds = %cleanup.loopexit, %for.body\n %retval.2 = phi ptr [ %retval.0, %for.body ], [ @GlobIntONE, %cleanup.loopexit ]\n br i1 false, label %for.body, label %cleanup7.loopexit\n \n cleanup7.loopexit: ; preds = %cleanup\n %retval.2.lcssa.ph = phi ptr [ %retval.2, %cleanup ]\n br label %cleanup7\n \n cleanup7: ; preds = %cleanup7.loopexit, %cleanup.peel\n %retval.2.lcssa = phi ptr [ %retval.2.peel, %cleanup.peel ], [ %retval.2.lcssa.ph, %cleanup7.loopexit ]\n ret ptr %retval.2.lcssa\n }\n \n define ptr @tgt() {\n entry:\n br label %for.body.peel.begin\n \n for.body.peel.begin: ; preds = %entry\n br label %for.body.peel\n \n for.body.peel: ; preds = %for.body.peel.begin\n br i1 true, label %cleanup.peel, label %cleanup.loopexit.peel\n \n cleanup.loopexit.peel: ; preds = %for.body.peel\n br label %cleanup.peel\n \n cleanup.peel: ; preds = %cleanup.loopexit.peel, %for.body.peel\n %retval.2.peel = phi ptr [ undef, %for.body.peel ], [ @GlobIntONE, %cleanup.loopexit.peel ]\n br i1 true, label %for.body.peel.next, label %cleanup7\n \n for.body.peel.next: ; preds = %cleanup.peel\n br label %for.body.peel.next1\n \n for.body.peel.next1: ; preds = %for.body.peel.next\n br label %entry.peel.newph\n \n entry.peel.newph: ; preds = %for.body.peel.next1\n br label %for.body\n \n for.body: ; preds = %cleanup, %entry.peel.newph\n br i1 false, label %cleanup, label %cleanup.loopexit\n \n cleanup.loopexit: ; preds = %for.body\n br label %cleanup\n \n cleanup: ; preds = %cleanup.loopexit, %for.body\n br i1 false, label %for.body, label %cleanup7.loopexit\n \n cleanup7.loopexit: ; preds = %cleanup\n %retval.2.lcssa.ph = phi ptr [ %retval.2.peel, %cleanup ]\n br label %cleanup7\n \n cleanup7: ; preds = %cleanup7.loopexit, %cleanup.peel\n %retval.2.lcssa = phi ptr [ %retval.2.peel, %cleanup.peel ], [ %retval.2.lcssa.ph, %cleanup7.loopexit ]\n ret ptr %retval.2.lcssa\n }\n ```\n 1. `simplifyInstruction(%retval.2.peel)` returns `@GlobIntONE`. Thus,\n `ScalarEvolution::createNodeForPHI` returns SCEV expr `@GlobIntONE` for\n `%retval.2.peel`.\n 2. `SimplifyIndvar::replaceIVUserWithLoopInvariant` tries to replace the\n use of `%retval.2.peel` in `%retval.2.lcssa.ph` with `@GlobIntONE`.\n 3. `simplifyLoopAfterUnroll -> simplifyLoopIVs -> SCEVExpander::expand`\n reuses `%retval.2.peel = phi ptr [ undef, %for.body.peel ], [\n @GlobIntONE, %cleanup.loopexit.peel ]` to generate code for\n `@GlobIntONE`. It is incorrect.\n \n This patch disallows simplifying `phi(undef, X)` to `X` by setting\n `CanUseUndef` to false.\n Closes https://github.com/llvm/llvm-project/issues/114879.\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex 58e23e9556f1..b10811133770 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -6023,7 +6023,11 @@ const SCEV *ScalarEvolution::createNodeForPHI(PHINode *PN) {\n if (const SCEV *S = createAddRecFromPHI(PN))\n return S;\n \n- if (Value *V = simplifyInstruction(PN, {getDataLayout(), &TLI, &DT, &AC}))\n+ // We do not allow simplifying phi (undef, X) to X here, to avoid reusing the\n+ // phi node for X.\n+ if (Value *V = simplifyInstruction(\n+ PN, {getDataLayout(), &TLI, &DT, &AC, /*CtxI=*/nullptr,\n+ /*UseInstrInfo=*/true, /*CanUseUndef=*/false}))\n return getSCEV(V);\n \n if (const SCEV *S = createNodeFromSelectLikePHI(PN))\n", "tests": [{"file": "llvm/test/Transforms/IndVarSimplify/invalidate-modified-lcssa-phi.ll", "commands": ["opt -passes='loop(indvars,loop-deletion)' -verify-scev -S %s"], "tests": [{"test_name": "test_pr52023", "test_body": "define i8 @test_pr52023(i1 %c.1, i1 %c.2) {\nentry:\n br label %loop.1\n\nloop.1: ; preds = %loop.1.latch, %entry\n %inc79 = phi i8 [ %inc.lcssa, %loop.1.latch ], [ 0, %entry ]\n br label %loop.2\n\nloop.2: ; preds = %loop.2.latch, %loop.1\n %inc6 = phi i8 [ %inc79, %loop.1 ], [ %inc, %loop.2.latch ]\n %inc = add i8 %inc6, 1\n br i1 %c.1, label %loop.2.latch, label %loop.1.latch\n\nloop.2.latch: ; preds = %loop.2\n br i1 false, label %loop.2, label %loop.1.latch\n\nloop.1.latch: ; preds = %loop.2.latch, %loop.2\n %inc.lcssa = phi i8 [ %inc, %loop.2.latch ], [ undef, %loop.2 ]\n br i1 %c.2, label %exit, label %loop.1\n\nexit: ; preds = %loop.1.latch\n %inc.lcssa.lcssa = phi i8 [ %inc.lcssa, %loop.1.latch ]\n ret i8 %inc.lcssa.lcssa\n}\n"}]}, {"file": "llvm/test/Transforms/IndVarSimplify/no-iv-rewrite.ll", "commands": ["opt < %s -passes=indvars -S -indvars-predicate-loops=0"], "tests": [{"test_name": "identityphi", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\n; Function Attrs: nounwind\ndefine void @identityphi(i32 %limit) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %control, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %control ]\n br i1 undef, label %if.then, label %control\n\nif.then: ; preds = %loop\n br label %control\n\ncontrol: ; preds = %if.then, %loop\n %iv.next = phi i32 [ %iv, %loop ], [ undef, %if.then ]\n %cmp = icmp slt i32 %iv.next, %limit\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %control\n ret void\n}\n\nattributes #0 = { nounwind }\n"}]}, {"file": "llvm/test/Transforms/LoopStrengthReduce/X86/2012-01-13-phielim.ll", "commands": ["opt < %s -S -loop-reduce -mtriple=x86_64-- -mcpu=core2"], "tests": [{"test_name": "test3", "test_body": "; Function Attrs: nounwind ssp uwtable\ndefine fastcc void @test3(ptr nocapture %u) #0 {\nentry:\n br i1 undef, label %meshBB1, label %meshBB5\n\nfor.inc8.us.i: ; preds = %for.body3.us.i\n br i1 undef, label %meshBB1, label %meshBB\n\nfor.body3.us.i: ; preds = %meshBB, %for.body3.lr.ph.us.i\n %indvars.iv.i.SV.phi = phi i64 [ %indvars.iv.next.i, %meshBB ], [ 0, %for.body3.lr.ph.us.i ]\n %storemerge13.us.i.SV.phi = phi i32 [ 0, %meshBB ], [ 0, %for.body3.lr.ph.us.i ]\n %Opq.sa.calc12 = sub i32 undef, 227\n %0 = add nsw i64 %indvars.iv.i.SV.phi, %indvars.iv8.i.SV.phi26\n %1 = trunc i64 %0 to i32\n %mul.i.us.i = mul nsw i32 0, %1\n %arrayidx5.us.i = getelementptr inbounds double, ptr %u, i64 %indvars.iv.i.SV.phi\n %2 = load double, ptr %arrayidx5.us.i, align 8\n %indvars.iv.next.i = add i64 %indvars.iv.i.SV.phi, 1\n br i1 undef, label %for.inc8.us.i, label %meshBB\n\nfor.body3.lr.ph.us.i: ; preds = %meshBB1, %meshBB\n %indvars.iv8.i.SV.phi26 = phi i64 [ undef, %meshBB1 ], [ %indvars.iv8.i.SV.phi24, %meshBB ]\n %arrayidx.us.i = getelementptr inbounds double, ptr undef, i64 %indvars.iv8.i.SV.phi26\n %3 = add i64 %indvars.iv8.i.SV.phi26, 1\n br label %for.body3.us.i\n\nfor.inc8.us.i2: ; preds = %meshBB5\n unreachable\n\neval_At_times_u.exit: ; preds = %meshBB5\n ret void\n\nmeshBB: ; preds = %for.body3.us.i, %for.inc8.us.i\n %indvars.iv8.i.SV.phi24 = phi i64 [ undef, %for.body3.us.i ], [ %3, %for.inc8.us.i ]\n %meshStackVariable.phi = phi i32 [ %Opq.sa.calc12, %for.body3.us.i ], [ undef, %for.inc8.us.i ]\n br i1 undef, label %for.body3.lr.ph.us.i, label %for.body3.us.i\n\nmeshBB1: ; preds = %for.inc8.us.i, %entry\n br label %for.body3.lr.ph.us.i\n\nmeshBB5: ; preds = %entry\n br i1 undef, label %eval_At_times_u.exit, label %for.inc8.us.i2\n}\n\nattributes #0 = { nounwind ssp uwtable }\n"}]}, {"file": "llvm/test/Transforms/LoopUnroll/pr114879.ll", "commands": ["opt -S -passes=loop-unroll-full -unroll-full-max-count=1 %s"], "tests": [{"test_name": "test", "test_body": "@GlobIntONE = external global i32, align 4\n\ndefine ptr @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %cleanup, %entry\n %retval.0 = phi ptr [ undef, %entry ], [ %retval.2, %cleanup ]\n %cmp1.not = phi i1 [ true, %entry ], [ false, %cleanup ]\n br i1 %cmp1.not, label %cleanup, label %cleanup.loopexit\n\ncleanup.loopexit: ; preds = %for.body\n br label %cleanup\n\ncleanup: ; preds = %cleanup.loopexit, %for.body\n %retval.2 = phi ptr [ %retval.0, %for.body ], [ @GlobIntONE, %cleanup.loopexit ]\n br i1 %cmp1.not, label %for.body, label %cleanup2\n\ncleanup2: ; preds = %cleanup\n %retval.2.lcssa = phi ptr [ %retval.2, %cleanup ]\n ret ptr %retval.2.lcssa\n}\n", "additional_args": "--src-unroll=10 -tgt-unroll=10"}]}, {"file": "llvm/test/Transforms/LoopVectorize/uniform-blend.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=4 -S %s"], "tests": [{"test_name": "blend_uniform_iv", "test_body": "@dst = external global [32 x i16], align 1\n\ndefine void @blend_uniform_iv(i1 %c) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %loop.next, label %loop.latch\n\nloop.next: ; preds = %loop.header\n br label %loop.latch\n\nloop.latch: ; preds = %loop.next, %loop.header\n %blend = phi i64 [ poison, %loop.header ], [ %iv, %loop.next ]\n %dst.ptr = getelementptr inbounds [32 x i16], ptr @dst, i16 0, i64 %blend\n store i16 0, ptr %dst.ptr, align 2\n %iv.next = add nuw nsw i64 %iv, 1\n %cmp439 = icmp ult i64 %iv, 31\n br i1 %cmp439, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}, {"test_name": "blend_uniform_iv_trunc", "test_body": "@dst = external global [32 x i16], align 1\n\ndefine void @blend_uniform_iv_trunc(i1 %c) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %iv.trunc.2 = trunc i64 %iv to i16\n br i1 %c, label %loop.next, label %loop.latch\n\nloop.next: ; preds = %loop.header\n br label %loop.latch\n\nloop.latch: ; preds = %loop.next, %loop.header\n %blend = phi i16 [ poison, %loop.header ], [ %iv.trunc.2, %loop.next ]\n %dst.ptr = getelementptr inbounds [32 x i16], ptr @dst, i16 0, i16 %blend\n store i16 0, ptr %dst.ptr, align 2\n %iv.next = add nuw nsw i64 %iv, 1\n %cmp439 = icmp ult i64 %iv, 31\n br i1 %cmp439, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}], "issue": {"title": "WRONG code: GVN? Loop opts?", "body": "cat wrong0.i\r\n```\r\nint printf(const char *, ...);\r\nunsigned int IntArr[6];\r\nunsigned int GlobIntONE = 1, GlobIntZERO = 0;\r\nconst unsigned int *GlobIntPtr = 0;\r\nunsigned long Res;\r\nunsigned long *ResPtr = &Res;\r\nconst unsigned int **GlobIntPtrPTr = &GlobIntPtr;\r\nunsigned int *func() {\r\n int *GlobIntONEPtr = &GlobIntONE;\r\n for (int Idx = 0; Idx <= 7; Idx += 1) {\r\n int Idx2 = 1;\r\n if (Idx > 0) {\r\n for (; Idx2 <= 7; Idx2 += 1)\r\n ;\r\n return GlobIntONEPtr;\r\n }\r\n }\r\n 0 != &GlobIntONEPtr;\r\n return &GlobIntZERO;\r\n}\r\n\r\nint main() {\r\n IntArr[GlobIntZERO] = GlobIntZERO;\r\n *GlobIntPtrPTr = func();\r\n unsigned char Byte = *GlobIntPtr;\r\n *ResPtr = Byte;\r\n printf(\"checksum = %X\\n\", Res);\r\n}\r\n```\r\n\r\n\r\nIn the above file, func() should obviously should return GlobIntONEPtr, but it does not if compiled with '-fno-inline -mllvm -unroll-full-max-count=1'.\r\n\r\n```\r\nclang -O0 -march=z16 wrong0.i -o a.out -w -fno-inline ; ./a.out\r\nchecksum = 1\r\n\r\nclang -O3 -march=z16 wrong0.i -o a.out -w -fno-inline ; ./a.out\r\nchecksum = 1\r\n\r\nclang -O3 -march=z16 wrong0.i -o a.out -w -fno-inline -mllvm -unroll-full-max-count=1 ; ./a.out\r\nchecksum = 0\r\n\r\n```\r\nIt may be that GVN does something wrong, as after it @func() is always returning ptr undef. Or it could very well be some of the loop optimizations that is run just before it.\r\n\r\nA bisect shows that\r\n\r\n```\r\nd77067d08a3f56dc2d0e6c95bd2852c943df743a is the first bad commit\r\ncommit d77067d08a3f56dc2d0e6c95bd2852c943df743a\r\nAuthor: Nikita Popov \r\nDate: Wed Dec 6 14:17:18 2023 +0100\r\n\r\n [ValueTracking] Add dominating condition support in computeKnownBits() (#73662)\r\n\r\n```\r\n@nikic ", "author": "JonPsson1", "labels": ["miscompilation", "llvm:SCEV"], "comments": [{"author": "dtcxzyw", "body": "Bisected to LoopUnrollFullPass:\r\n```\r\n; bin/opt -passes=loop-unroll-full -unroll-full-max-count=1 test.ll -S\r\n\r\n@GlobIntONE = dso_local global i32 1, align 4\r\n@GlobIntZERO = dso_local local_unnamed_addr global i32 0, align 4\r\n\r\ndefine i32 @main() local_unnamed_addr {\r\nentry:\r\n br label %for.body\r\n\r\nfor.body: ; preds = %entry, %cleanup\r\n %retval.0 = phi ptr [ undef, %entry ], [ %retval.2, %cleanup ]\r\n %cmp1.not = phi i1 [ true, %entry ], [ false, %cleanup ]\r\n br i1 %cmp1.not, label %cleanup, label %for.cond2.preheader\r\n\r\nfor.cond2.preheader: ; preds = %for.body\r\n br label %cleanup.loopexit\r\n\r\ncleanup.loopexit: ; preds = %for.cond2.preheader\r\n br label %cleanup\r\n\r\ncleanup: ; preds = %cleanup.loopexit, %for.body\r\n %retval.2 = phi ptr [ %retval.0, %for.body ], [ @GlobIntONE, %cleanup.loopexit ]\r\n br i1 %cmp1.not, label %for.body, label %cleanup7\r\n\r\ncleanup7: ; preds = %cleanup\r\n %retval.2.lcssa = phi ptr [ %retval.2, %cleanup ]\r\n %res = load i32, ptr %retval.2.lcssa, align 4\r\n ret i32 %res\r\n}\r\n```\r\n```\r\nsource_filename = \"test.ll\"\r\n\r\n@GlobIntONE = dso_local global i32 1, align 4\r\n@GlobIntZERO = dso_local local_unnamed_addr global i32 0, align 4\r\n\r\ndefine i32 @main() local_unnamed_addr {\r\nentry:\r\n br label %for.body.peel.begin\r\n\r\nfor.body.peel.begin: ; preds = %entry\r\n br label %for.body.peel\r\n\r\nfor.body.peel: ; preds = %for.body.peel.begin\r\n br i1 true, label %cleanup.peel, label %for.cond2.preheader.peel\r\n\r\nfor.cond2.preheader.peel: ; preds = %for.body.peel\r\n br label %cleanup.loopexit.peel\r\n\r\ncleanup.loopexit.peel: ; preds = %for.cond2.preheader.peel\r\n br label %cleanup.peel\r\n\r\ncleanup.peel: ; preds = %cleanup.loopexit.peel, %for.body.peel\r\n %retval.2.peel = phi ptr [ undef, %for.body.peel ], [ @GlobIntONE, %cleanup.loopexit.peel ]\r\n br i1 true, label %for.body.peel.next, label %cleanup7\r\n\r\nfor.body.peel.next: ; preds = %cleanup.peel\r\n br label %for.body.peel.next1\r\n\r\nfor.body.peel.next1: ; preds = %for.body.peel.next\r\n br label %entry.peel.newph\r\n\r\nentry.peel.newph: ; preds = %for.body.peel.next1\r\n br label %for.body\r\n\r\nfor.body: ; preds = %cleanup, %entry.peel.newph\r\n br i1 false, label %cleanup, label %for.cond2.preheader\r\n\r\nfor.cond2.preheader: ; preds = %for.body\r\n br label %cleanup.loopexit\r\n\r\ncleanup.loopexit: ; preds = %for.cond2.preheader\r\n br label %cleanup\r\n\r\ncleanup: ; preds = %cleanup.loopexit, %for.body\r\n br i1 false, label %for.body, label %cleanup7.loopexit, !llvm.loop !0\r\n\r\ncleanup7.loopexit: ; preds = %cleanup\r\n %retval.2.lcssa.ph = phi ptr [ %retval.2.peel, %cleanup ]\r\n br label %cleanup7\r\n\r\ncleanup7: ; preds = %cleanup7.loopexit, %cleanup.peel\r\n %retval.2.lcssa = phi ptr [ %retval.2.peel, %cleanup.peel ], [ %retval.2.lcssa.ph, %cleanup7.loopexit ]\r\n %res = load i32, ptr %retval.2.lcssa, align 4\r\n ret i32 %res\r\n}\r\n\r\n!0 = distinct !{!0, !1}\r\n!1 = !{!\"llvm.loop.peeled.count\", i32 1}\r\n```"}, {"author": "dtcxzyw", "body": "`simplifyLoopAfterUnroll -> simplifyLoopIVs` causes the miscompilation: https://alive2.llvm.org/ce/z/yiBsZt\r\nI will post a fix later.\r\n"}, {"author": "dtcxzyw", "body": "The root cause is:\r\n1. `simplifyInstruction(%retval.2.peel)` returns `@GlobIntONE`. Thus, `ScalarEvolution::createNodeForPHI` returns SCEV expr `@GlobIntONE` for `%retval.2.peel`.\r\n2. `SimplifyIndvar::replaceIVUserWithLoopInvariant` tries to replace the use of `%retval.2.peel` in `%retval.2.lcssa.ph` with `@GlobIntONE`.\r\n3. `SCEVExpander::expand` reuses `%retval.2.peel = phi ptr [ undef, %for.body.peel ], [ @GlobIntONE, %cleanup.loopexit.peel ]` to generate code for `@GlobIntONE`. It is incorrect.\r\n\r\nTwo possible solutions:\r\n1. Set `CanUseUndef` to false in `simplifyInstruction`.\r\n2. Block this case in `ScalarEvolution::canReuseInstruction`.\r\n\r\n@nikic @preames Any thoughts?\r\n"}, {"author": "nikic", "body": "I think setting `CanUseUndef=false` would be the correct thing to do. The reverse replacement here is particularly fishy, but more generally, if we make an assumption about the value of undef at a particular use, we have to be consistent about that assumption, and SCEV cannot really guarantee that this is the case."}]}, "verified": true} {"bug_id": "87798", "issue_url": "https://github.com/llvm/llvm-project/issues/87798", "bug_type": "crash", "base_commit": "49ef12a08c4c7d7ae4765929e72fe2320a12b08c", "knowledge_cutoff": "2024-04-05T16:02:46Z", "lit_test_dir": ["llvm/test/Analysis/ScalarEvolution"], "hints": {"fix_commit": "54a9f0007cb4f19d2e9df30405c5027229f5def0", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[928, 938]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["BinomialCoefficient"]}}, "patch": "commit 54a9f0007cb4f19d2e9df30405c5027229f5def0\nAuthor: annamthomas \nDate: Wed Apr 10 09:02:23 2024 -0400\n\n [SCEV] Fix BinomialCoefficient Iteration to fit in W bits (#88010)\n \n BinomialCoefficient computes the value of W-bit IV at iteration It of a loop. When W is 1, we can call multiplicative inverse on 0 which triggers an assert since 1b76120.\n \n Since the arithmetic is supposed to wrap if It or K does not fit in W bits, do the truncation into W bits after we do the shift.\n \n Fixes #87798\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex e030b9fc7dac..9fcce797f559 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -928,11 +928,9 @@ static const SCEV *BinomialCoefficient(const SCEV *It, unsigned K,\n APInt OddFactorial(W, 1);\n unsigned T = 1;\n for (unsigned i = 3; i <= K; ++i) {\n- APInt Mult(W, i);\n- unsigned TwoFactors = Mult.countr_zero();\n+ unsigned TwoFactors = countr_zero(i);\n T += TwoFactors;\n- Mult.lshrInPlace(TwoFactors);\n- OddFactorial *= Mult;\n+ OddFactorial *= (i >> TwoFactors);\n }\n \n // We need at least W + T bits for the multiplication step\n", "tests": [{"file": "llvm/test/Analysis/ScalarEvolution/pr87798.ll", "commands": ["opt -disable-output -passes='print' -verify-scev < %s 2>&1"], "tests": [{"test_name": "pr87798", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i32 @pr87798() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb1, %bb\n %phi = phi i32 [ 0, %bb ], [ %add4, %bb1 ]\n %phi2 = phi i32 [ 0, %bb ], [ %add, %bb1 ]\n %phi3 = phi i32 [ 0, %bb ], [ %add5, %bb1 ]\n %add = add i32 %phi2, %phi3\n %mul = mul i32 %phi2, %phi3\n %add4 = add i32 %mul, %phi\n %and = and i32 %phi, 1\n %add5 = add i32 %phi3, 1\n br i1 true, label %preheader, label %bb1\n\npreheader: ; preds = %bb1\n %phi9 = phi i32 [ %and, %bb1 ]\n br label %loop\n\nloop: ; preds = %loop, %preheader\n br label %loop\n\nbb7: ; No predecessors!\n %zext = zext i32 %phi9 to i64\n ret i32 0\n}\n"}]}], "issue": {"title": "Assertion failure: \"multiplicative inverse is only defined for odd numbers\" exposed through usage in SCEV", "body": "With this change [1b76120](https://github.com/llvm/llvm-project/commit/1b761205f2686516cebadbcbc37f798197d9c482) landed in https://github.com/llvm/llvm-project/pull/87610, we started seeing an assertion failure on the introduced API.\r\n\r\nThis change caused an assertion failure in usage of multiplicativeInverse in SCEV's BinomialCoefficient algorithm (which was also updated by the change to use this overloaded API). The `oddFactorial` in this algorithm becomes 0 and the assertion in newly introduced API gets triggered. Note that the old API for multiplicativeInverse (which passed in the explicit modulo) returned 0 when the input was also 0, and probably hid this bug in SCEV. \r\n\r\nHere's a minimal reproducer for trunk:\r\n```\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i32 @\"static jint Test.iMeth(jlong, jint, jint)2272037699586\"() {\r\nbci_0:\r\n br label %bci_71.preheader\r\n\r\nbci_71.preheader: ; preds = %bci_71.preheader, %bci_0\r\n %local_9_242 = phi i32 [ 0, %bci_0 ], [ %1, %bci_71.preheader ]\r\n %local_7_241 = phi i32 [ 0, %bci_0 ], [ %0, %bci_71.preheader ]\r\n %local_4_239 = phi i32 [ 0, %bci_0 ], [ %3, %bci_71.preheader ]\r\n %0 = add i32 %local_7_241, %local_4_239\r\n %.neg14 = mul i32 %local_7_241, %local_4_239\r\n %1 = add i32 %.neg14, %local_9_242\r\n %2 = and i32 %local_9_242, 1\r\n %3 = add i32 %local_4_239, 1\r\n br i1 true, label %bci_10.i38.preheader, label %bci_71.preheader\r\n\r\nbci_10.i38: ; preds = %bci_10.i38.preheader, %bci_10.i38\r\n br label %bci_10.i38\r\n\r\n\"static jlong FuzzerUtils.checkSum(jobject)2306397437964.exit\": ; No predecessors!\r\n %4 = zext i32 %.lcssa596 to i64\r\n ret i32 0\r\n\r\nbci_10.i38.preheader: ; preds = %bci_71.preheader\r\n %.lcssa596 = phi i32 [ %2, %bci_71.preheader ]\r\n br label %bci_10.i38\r\n}\r\n```\r\n\r\nopt -passes=indvars reduced.ll produces:\r\n```\r\nopt: /root/llvm-project/llvm/lib/Support/APInt.cpp:1245: llvm::APInt llvm::APInt::multiplicativeInverse() const: Assertion `(*this)[0] && \"multiplicative inverse is only defined for odd numbers!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=indvars \r\n #0 0x0000000004cfc568 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4cfc568)\r\n #1 0x0000000004cf9cbc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f7209e42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007f7209e969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007f7209e42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007f7209e287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007f7209e2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007f7209e39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000004bf9eff llvm::APInt::multiplicativeInverse() const (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4bf9eff)\r\n #9 0x000000000443a4b5 llvm::SCEVAddRecExpr::evaluateAtIteration(llvm::ArrayRef, llvm::SCEV const*, llvm::ScalarEvolution&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x443a4b5)\r\n#10 0x000000000443cb4f llvm::ScalarEvolution::computeSCEVAtScope(llvm::SCEV const*, llvm::Loop const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x443cb4f)\r\n#11 0x000000000443cfee llvm::ScalarEvolution::getSCEVAtScope(llvm::SCEV const*, llvm::Loop const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x443cfee)\r\n#12 0x000000000443c525 llvm::ScalarEvolution::computeSCEVAtScope(llvm::SCEV const*, llvm::Loop const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x443c525)\r\n#13 0x000000000443cfee llvm::ScalarEvolution::getSCEVAtScope(llvm::SCEV const*, llvm::Loop const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x443cfee)\r\n#14 0x000000000404eca0 llvm::rewriteLoopExitValues(llvm::Loop*, llvm::LoopInfo*, llvm::TargetLibraryInfo*, llvm::ScalarEvolution*, llvm::TargetTransformInfo const*, llvm::SCEVExpander&, llvm::DominatorTree*, llvm::ReplaceExitVal, llvm::SmallVector&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x404eca0)\r\n#15 0x0000000003a10fd8 (anonymous namespace)::IndVarSimplify::run(llvm::Loop*) IndVarSimplify.cpp:0:0\r\n#16 0x0000000003a11c1a llvm::IndVarSimplifyPass::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3a11c1a)\r\n#17 0x00000000029ffd4e llvm::detail::PassModel, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x29ffd4e)\r\n#18 0x0000000003aaef56 std::optional llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::runSinglePass, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>, std::default_delete, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>>>>(llvm::Loop&, std::unique_ptr, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>, std::default_delete, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>>>&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&, llvm::PassInstrumentation&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3aaef56)\r\n#19 0x0000000003aaf1ee llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::runWithoutLoopNestPasses(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3aaf1ee)\r\n#20 0x0000000003ab05d9 llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3ab05d9)\r\n...\r\n```", "author": "annamthomas", "labels": ["llvm:SCEV", "crash"], "comments": [{"author": "topperc", "body": "CC @jayfoad "}, {"author": "annamthomas", "body": "I think the old API (multiplicativeInverse when modulo provided) was hiding this bug since multiplicative inverse shouldn't be asked for 0.\r\nAlso, the reason the `oddFactorial` becomes 0 here is because of this:\r\n```\r\n// Calculate K! / 2^T and T; we divide out the factors of two before\r\n // multiplying for calculating K! / 2^T to avoid overflow.\r\n // Other overflow doesn't matter because we only care about the bottom\r\n // W bits of the result.\r\n APInt OddFactorial(W, 1);\r\n unsigned T = 1;\r\n for (unsigned i = 3; i <= K; ++i) {\r\n APInt Mult(W, i);\r\n unsigned TwoFactors = Mult.countr_zero();\r\n T += TwoFactors;\r\n Mult.lshrInPlace(TwoFactors);\r\n OddFactorial *= Mult; \r\n }\r\n ```\r\n\r\nHere, W is one (bit width = 1 and I believe this is the bug) and we are trying to represent Mult=4 in one bit. That makes `Mult` 0 and hence `OddFactorial` 0.\r\n\r\nOne thing we can obviously do is in the newly introduced API, explicitly check if the APInt is 0 and return 0 (this makes behaviour same as the old API).\r\nBut I think this is hiding the bug. "}, {"author": "jayfoad", "body": "Suggestion:\r\n```diff\r\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\r\nindex e030b9fc7dac..3f8c390614ff 100644\r\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\r\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\r\n@@ -928,11 +928,9 @@ static const SCEV *BinomialCoefficient(const SCEV *It, unsigned K,\r\n APInt OddFactorial(W, 1);\r\n unsigned T = 1;\r\n for (unsigned i = 3; i <= K; ++i) {\r\n- APInt Mult(W, i);\r\n- unsigned TwoFactors = Mult.countr_zero();\r\n+ unsigned TwoFactors = countr_zero(i);\r\n T += TwoFactors;\r\n- Mult.lshrInPlace(TwoFactors);\r\n- OddFactorial *= Mult;\r\n+ OddFactorial *= i >> TwoFactors;\r\n }\r\n \r\n // We need at least W + T bits for the multiplication step\r\n```"}, {"author": "annamthomas", "body": "yes, that would probably work as well (since we avoid representing Mult in W bit width), but how can the bitWidth of the SCEV result`W` be 1 in the first place? "}, {"author": "jayfoad", "body": "> yes, that would probably work as well (since we avoid representing Mult in W bit width), but how can the bitWidth of the SCEV result`W` be 1 in the first place?\r\n\r\nNo idea! I have only looked at `BinomialCoefficient` itself, not at its callers. I assume it is supposed to work for any W > 0 and K > 0. At least I don't see it mentioned that K has to be representable in W bits, or anything like that."}, {"author": "efriedma-quic", "body": "The function is supposed to compute `BC(It, K) mod 2^W`, i.e. the binomial coefficient truncated to fit into a W-bit integer. This is useful for computing the value of a W-bit induction variable at iteration \"It\" of a loop. There isn't any expectation that \"It\" or \"K\" fit losslessly into \"W\": the arithmetic is supposed to wrap.\r\n\r\nGiven the way the function works, truncating \"i\" after the shift looks like the right solution."}]}, "verified": true} {"bug_id": "126181", "issue_url": "https://github.com/llvm/llvm-project/issues/126181", "bug_type": "miscompilation", "base_commit": "317a644ae6d501f1a1ec54d17ea8559bcdea35c0", "knowledge_cutoff": "2025-02-07T05:51:32Z", "lit_test_dir": ["llvm/test/Transforms/DeadStoreElimination"], "hints": {"fix_commit": "2d31a12dbe2339d20844ede70cbb54dbaf4ceea9", "components": ["DeadStoreElimination"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp": [[2283, 2289]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp": ["DSEState::getInitializesArgMemLoc"]}}, "patch": "commit 2d31a12dbe2339d20844ede70cbb54dbaf4ceea9\nAuthor: Nikita Popov \nDate: Mon Feb 10 10:34:03 2025 +0100\n\n [DSE] Don't use initializes on byval argument (#126259)\n \n There are two ways we can fix this problem, depending on how the\n semantics of byval and initializes should interact:\n \n * Don't infer initializes on byval arguments. initializes on byval\n refers to the original caller memory (or having both attributes is made\n a verifier error).\n * Infer initializes on byval, but don't use it in DSE. initializes on\n byval refers to the callee copy. This matches the semantics of readonly\n on byval. This is slightly more powerful, for example, we could do a\n backend optimization where byval + initializes will allocate the full\n size of byval on the stack but not copy over the parts covered by\n initializes.\n \n I went with the second variant here, skipping byval + initializes in DSE\n (FunctionAttrs already doesn't propagate initializes past byval). I'm\n open to going in the other direction though.\n \n Fixes https://github.com/llvm/llvm-project/issues/126181.\n\ndiff --git a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\nindex 05b4f176bfc3..38454053b039 100644\n--- a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\n+++ b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\n@@ -2283,7 +2283,9 @@ DSEState::getInitializesArgMemLoc(const Instruction *I) {\n for (unsigned Idx = 0, Count = CB->arg_size(); Idx < Count; ++Idx) {\n ConstantRangeList Inits;\n Attribute InitializesAttr = CB->getParamAttr(Idx, Attribute::Initializes);\n- if (InitializesAttr.isValid())\n+ // initializes on byval arguments refers to the callee copy, not the\n+ // original memory the caller passed in.\n+ if (InitializesAttr.isValid() && !CB->isByValArgument(Idx))\n Inits = InitializesAttr.getValueAsConstantRangeList();\n \n Value *CurArg = CB->getArgOperand(Idx);\n", "tests": [{"file": "llvm/test/Transforms/DeadStoreElimination/inter-procedural.ll", "commands": ["opt < %s -passes=dse -enable-dse-initializes-attr-improvement -S"], "tests": [{"test_name": "test_byval", "test_body": "%struct.a = type { i32, i64, i32, i32 }\n\n@e = dso_local global { i32, [4 x i8], i64, i32, i32 } { i32 5, [4 x i8] zeroinitializer, i64 0, i32 1, i32 90986701 }, align 8\n\ndefine i8 @ad(ptr byval(%struct.a) %ah) {\nentry:\n store i32 0, ptr %ah, align 8\n %call = call i64 @af(ptr %ah)\n %0 = load i8, ptr %ah, align 1\n ret i8 %0\n}\n\ndefine i64 @af(ptr byval(%struct.a) initializes((0, 24)) %am) {\nentry:\n call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(24) %am, ptr noundef nonnull align 8 dereferenceable(24) @e, i64 24, i1 false)\n ret i64 0\n}\n\ndefine i32 @main() {\n %res = call i8 @ad(ptr @e)\n %val = sext i8 %res to i32\n ret i32 %val\n}\n", "lli_expected_out": ""}]}], "issue": {"title": "[FuncAttrs] `initializes` is incorrectly set on parameters with `byval`", "body": "Reproducer: https://godbolt.org/z/Mcx1nY4E7\n```\n; bin/opt -passes=function-attrs reduced.ll -S\n%struct.a = type { i32, i64, i32, i32 }\n\n@e = dso_local global { i32, [4 x i8], i64, i32, i32 } { i32 5, [4 x i8] zeroinitializer, i64 0, i32 1, i32 90986701 }, align 8\n\ndefine i8 @ad(ptr byval(%struct.a) %ah) {\nentry:\n store i32 0, ptr %ah, align 8\n %call = call i64 @af(ptr %ah)\n %0 = load i8, ptr %ah, align 1\n ret i8 %0\n}\n\ndefine i64 @af(ptr byval(%struct.a) %am) {\nentry:\n call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(24) %am, ptr noundef nonnull align 8 dereferenceable(24) @e, i64 24, i1 false)\n ret i64 0\n}\n\n```\n```\ndefine i8 @ad(ptr byval(%struct.a) captures(none) initializes((0, 4)) %ah) #0 {\n store i32 0, ptr %ah, align 8\n %call = call i64 @af(ptr %ah)\n %0 = load i8, ptr %ah, align 1\n ret i8 %0\n}\n\ndefine noundef i64 @af(ptr writeonly byval(%struct.a) captures(none) initializes((0, 24)) %am) #0 {\n call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(24) %am, ptr noundef nonnull align 8 dereferenceable(24) @e, i64 24, i1 false)\n ret i64 0\n}\n```\nAfter DSE:\n```\ndefine i8 @ad(ptr byval(%struct.a) captures(none) initializes((0, 4)) %ah) #0 {\n %call = call i64 @af(ptr %ah)\n %0 = load i8, ptr %ah, align 1\n ret i8 %0\n}\n\ndefine noundef i64 @af(ptr writeonly byval(%struct.a) captures(none) initializes((0, 24)) %am) #0 {\n ret i64 0\n}\n```\n\nllvm version: d21fc58aeeaa7f0369a24dbe70a0360e0edbf76f\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "nikic", "body": "> We should also emit the padding for %struct.a.\n\nWithout commenting on whether we should or shouldn't, why is not having the padding a miscompile? As this is not a packed struct, for a i64:64 target, LLVM will implicitly insert the padding and it should have the correct size."}, {"author": "dtcxzyw", "body": "> As this is not a packed struct, for a i64:64 target, LLVM will implicitly insert the padding and it should have the correct size.\n\nSorry. It is a DSE bug.\n"}, {"author": "dtcxzyw", "body": "cc @haopliu "}, {"author": "nikic", "body": "@dtcxzyw I think your new IR is over-reduced. You have `initializes((0, 24)` on `@af`, so I think eliminating the store is correct."}, {"author": "dtcxzyw", "body": "> [@dtcxzyw](https://github.com/dtcxzyw) I think your new IR is over-reduced. You have `initializes((0, 24)` on `@af`, so I think eliminating the store is correct.\n\nFixed."}, {"author": "nikic", "body": "There are two ways we can fix this, depending on what we want the semantics to do:\n\n 1. Don't infer initializes on byval arguments. initializes on byval refers to the original caller memory (or having both attributes is made a verifier error).\n 2. Infer initializes on byval, but don't use it in DSE. initializes on byval refers to the callee copy. This matches the semantics of readonly on byval.\n\nI think variant 2 is marginally more powerful. For example, we could do a backend optimization where byval + initializes will allocate the full size of byval on the stack but not copy over the parts covered by initializes."}]}, "verified": true} {"bug_id": "80113", "issue_url": "https://github.com/llvm/llvm-project/issues/80113", "bug_type": "miscompilation", "base_commit": "24a804101b67676aa9fa7f1097043ddd9e2ac1b6", "knowledge_cutoff": "2024-01-31T08:09:28Z", "lit_test_dir": ["llvm/test/Transforms/BDCE"], "hints": {"fix_commit": "b210cbbd0eb8ef7cd2735e99570474e6e53ee00b", "components": ["BDCE"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/BDCE.cpp": [[45, 59], [79, 90]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/BDCE.cpp": ["clearAssumptionsOfUsers"]}}, "patch": "commit b210cbbd0eb8ef7cd2735e99570474e6e53ee00b\nAuthor: Nikita Popov \nDate: Wed Jan 31 11:01:03 2024 +0100\n\n [BDCE] Fix clearing of poison-generating flags\n \n If the demanded bits of an instruction are full, we don't have to\n recurse to its users, but we may still have to clear flags on the\n instruction itself.\n \n Fixes https://github.com/llvm/llvm-project/issues/80113.\n\ndiff --git a/llvm/lib/Transforms/Scalar/BDCE.cpp b/llvm/lib/Transforms/Scalar/BDCE.cpp\nindex e99210ce2f22..6e6fa6c8ffc4 100644\n--- a/llvm/lib/Transforms/Scalar/BDCE.cpp\n+++ b/llvm/lib/Transforms/Scalar/BDCE.cpp\n@@ -45,15 +45,17 @@ static void clearAssumptionsOfUsers(Instruction *I, DemandedBits &DB) {\n assert(I->getType()->isIntOrIntVectorTy() &&\n \"Trivializing a non-integer value?\");\n \n+ // If all bits of a user are demanded, then we know that nothing below that\n+ // in the def-use chain needs to be changed.\n+ if (DB.getDemandedBits(I).isAllOnes())\n+ return;\n+\n // Initialize the worklist with eligible direct users.\n SmallPtrSet Visited;\n SmallVector WorkList;\n for (User *JU : I->users()) {\n- // If all bits of a user are demanded, then we know that nothing below that\n- // in the def-use chain needs to be changed.\n- auto *J = dyn_cast(JU);\n- if (J && J->getType()->isIntOrIntVectorTy() &&\n- !DB.getDemandedBits(J).isAllOnes()) {\n+ auto *J = cast(JU);\n+ if (J->getType()->isIntOrIntVectorTy()) {\n Visited.insert(J);\n WorkList.push_back(J);\n }\n@@ -79,12 +81,14 @@ static void clearAssumptionsOfUsers(Instruction *I, DemandedBits &DB) {\n // 1. llvm.assume demands its operand, so trivializing can't change it.\n // 2. range metadata only applies to memory accesses which demand all bits.\n \n+ // If all bits of a user are demanded, then we know that nothing below\n+ // that in the def-use chain needs to be changed.\n+ if (DB.getDemandedBits(J).isAllOnes())\n+ continue;\n+\n for (User *KU : J->users()) {\n- // If all bits of a user are demanded, then we know that nothing below\n- // that in the def-use chain needs to be changed.\n- auto *K = dyn_cast(KU);\n- if (K && Visited.insert(K).second && K->getType()->isIntOrIntVectorTy() &&\n- !DB.getDemandedBits(K).isAllOnes())\n+ auto *K = cast(KU);\n+ if (Visited.insert(K).second && K->getType()->isIntOrIntVectorTy())\n WorkList.push_back(K);\n }\n }\n", "tests": [{"file": "llvm/test/Transforms/BDCE/invalidate-assumptions.ll", "commands": ["opt -passes=bdce %s -S"], "tests": [{"test_name": "disjoint_indirect", "test_body": "define i32 @disjoint_indirect(i64 %x) {\n %and = and i64 %x, 1\n %trunc = trunc i64 %and to i32\n %or = or disjoint i32 %trunc, -2\n ret i32 %or\n}\n"}, {"test_name": "disjoint", "test_body": "define i64 @disjoint(i64 %x) {\n %and = and i64 %x, 1\n %or = or disjoint i64 %and, -2\n ret i64 %or\n}\n"}]}], "issue": {"title": "Wrong code at -O2/3 on x86-64_linux-gnu (recent regression)", "body": "clang at -O2/3 produced the wrong code. This is a recent regression.\r\n\r\nBisected to 20737825c9122b6e0a8912731cfa7e0558fe025d, which was committed by @antoniofrighetto \r\n\r\nCompiler explorer: https://godbolt.org/z/55z1WT3MM\r\n\r\n```console\r\n% cat reduced.c\r\nint printf(const char *, ...);\r\nshort a, f = 6;\r\nint b, e, g;\r\nint c[7][1];\r\nstatic int *d = &b;\r\nlong h;\r\nint *const *i;\r\nstatic unsigned char j(long k, short *n) {\r\n int *l = &c[5][0];\r\n int *const **m = &i;\r\n i = &l;\r\n *d = 0;\r\n for (; e < 60; e++)\r\n ***m = ((h = *d == (1 ^ k)) & *n + f) - 2;\r\n return 0;\r\n}\r\nint main() {\r\n j(1, &a);\r\n printf(\"%d\\n\", c[5][0]);\r\n}\r\n%\r\n% clang -O0 reduced.c &&./a.out\r\n-2\r\n% clang -O2 reduced.c &&./a.out\r\n4\r\n%\r\n```", "author": "shao-hua-li", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "nikic", "body": "Looks like poison flags don't get cleared as expected."}, {"author": "nikic", "body": "Minimized:\r\n```llvm\r\ndefine i32 @test(i64 %x) {\r\n %and = and i64 %x, 1\r\n %trunc = trunc i64 %and to i32\r\n %or = or disjoint i32 %trunc, -2\r\n ret i32 %or\r\n}\r\n```\r\nThe `%and` gets dropped but `disjoint` is kept."}, {"author": "antoniofrighetto", "body": "@nikic, could you kindly elaborate this? I'm having a hard time getting this, the transformation looks correct with simple `or` (https://alive2.llvm.org/ce/z/ui3yFf). Semantics for `disjoint`:\r\n> If the disjoint flag is set, and the or has a one and the same bit in both arguments, the result is poison.\r\n\r\nHowever, here:\r\n```llvm\r\n %add = add nuw nsw i64 %conv5, %conv4\r\n %and = and i64 %add, 1\r\n %2 = trunc i64 %and to i32\r\n %conv7 = or disjoint i32 %2, -2\r\n```\r\nThe two or operands do not seem to have any same bit in both arguments."}, {"author": "antoniofrighetto", "body": "OK, I got it now, the `disjoint` should not be there at all after the transformation. Makes sense, thank you."}, {"author": "nikic", "body": "Fixed another issue I noticed in this code in https://github.com/llvm/llvm-project/commit/cb6240d247b3419dea29eb99261171ea239b1c5c."}]}, "verified": true} {"bug_id": "111040", "issue_url": "https://github.com/llvm/llvm-project/issues/111040", "bug_type": "crash", "base_commit": "b3e0bd3d284dec705386b1efcae40dd51b763010", "knowledge_cutoff": "2024-10-03T18:42:52Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "45b526afa26e76e0c351e947ac8f0e4b55aa760b", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6196, 6206]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::setVectorizedCallDecision"]}}, "patch": "commit 45b526afa26e76e0c351e947ac8f0e4b55aa760b\nAuthor: Florian Hahn \nDate: Sun Oct 6 10:34:37 2024 +0100\n\n [LV] Honor uniform-after-vectorization in setVectorizedCallDecision.\n \n The legacy cost model always computes the cost for uniforms as cost of\n VF = 1, but VPWidenCallRecipes would be created, as\n setVectorizedCallDecisions would not consider uniform calls.\n \n Fix setVectorizedCallDecision to set to Scalarize, if the call is\n uniform-after-vectorization.\n \n This fixes a bug in VPlan construction uncovered by the VPlan-based\n cost model.\n \n Fixes https://github.com/llvm/llvm-project/issues/111040.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 6e082b1c134d..35c042b3ab7f 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6196,11 +6196,12 @@ void LoopVectorizationCostModel::setVectorizedCallDecision(ElementCount VF) {\n getScalarizationOverhead(CI, VF, CostKind);\n \n ScalarCost = ScalarCallCost * VF.getKnownMinValue() + ScalarizationCost;\n- // Honor ForcedScalars decision.\n+ // Honor ForcedScalars and UniformAfterVectorization decisions.\n // TODO: For calls, it might still be more profitable to widen. Use\n // VPlan-based cost model to compare different options.\n- if (VF.isVector() && ForcedScalar != ForcedScalars.end() &&\n- ForcedScalar->second.contains(CI)) {\n+ if (VF.isVector() && ((ForcedScalar != ForcedScalars.end() &&\n+ ForcedScalar->second.contains(CI)) ||\n+ isUniformAfterVectorization(CI, VF))) {\n setCallWideningDecision(CI, VF, CM_Scalarize, nullptr,\n Intrinsic::not_intrinsic, std::nullopt,\n ScalarCost);\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/replicate-uniform-call.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "smax_call_uniform", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @smax_call_uniform(ptr %dst, i64 %x) {\nentry:\n %c = icmp ult i8 -68, -69\n %mul = mul nuw nsw i64 %x, 0\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %loop.latch, label %else\n\nelse: ; preds = %loop.header\n %rem = urem i64 %mul, %x\n %smax = tail call i64 @llvm.smax.i64(i64 %rem, i64 0)\n br label %loop.latch\n\nloop.latch: ; preds = %else, %loop.header\n %p = phi i64 [ 1, %loop.header ], [ %smax, %else ]\n %add = add i64 %p, 1\n %gep = getelementptr i64, ptr %dst, i64 %add\n store i64 0, ptr %gep, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 0\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i64 @llvm.smax.i64(i64, i64) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[clang] Assertion failed in LoopVectorize.cpp:7401", "body": "When I compiled the code with -O2 flag, it crashed:\r\n\r\n```c\r\n#include \r\nstatic uint8_t b = 188;\r\nvoid c();\r\nlong d(int e, int p2, short j) {\r\n long f = 0, a = 1;\r\n unsigned long g;\r\n int h;\r\n if (e < 0 || j < 0)\r\n return 1;\r\n h = 0;\r\n for (; h < p2; h++) {\r\n a *= j + 1;\r\n g = e * a;\r\n f += g % (j + 1);\r\n }\r\n if (f < 0)\r\n return 0;\r\n return f;\r\n}\r\nvoid i();\r\nint64_t k() {\r\n uint32_t l[][8][10] = {0, 4, 7, 1, 4, 3446557846};\r\n int64_t m[10];\r\n int h;\r\n for (h = 0; c + h < 10; h++)\r\n m[(int)d(5, l[0][0][5] - 3446557842, b - 187) + h] = 1;\r\n i(m);\r\n uint8_t *n = &b;\r\n uint8_t **o;\r\n o[0] = &n;\r\n}\r\n```\r\n\r\nThe crash is:\r\n```\r\nclang: /root/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7401: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\n```\r\n\r\nDetails can be found here: https://godbolt.org/z/hqP8vMG7b\r\n", "author": "cardigan1008", "labels": ["regression", "vectorizers", "crash"], "comments": [{"author": "shafik", "body": "This is a regression in trunk: https://godbolt.org/z/a53x7oxhq"}]}, "verified": true} {"bug_id": "106239", "issue_url": "https://github.com/llvm/llvm-project/issues/106239", "bug_type": "crash", "base_commit": "15405b32b1cdbefab9ce1b1f301a51ae25404037", "knowledge_cutoff": "2024-08-27T15:51:29Z", "lit_test_dir": ["llvm/test/Transforms/IndVarSimplify"], "hints": {"fix_commit": "c9a5e1b665dbba898e9981fd7d48881947e6560e", "components": ["SimplifyIndVar"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/SimplifyIndVar.cpp": [[1928, 1945]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/SimplifyIndVar.cpp": ["WidenIV::widenIVUse"]}}, "patch": "commit c9a5e1b665dbba898e9981fd7d48881947e6560e\nAuthor: Nikita Popov \nDate: Wed Aug 28 12:54:14 2024 +0200\n\n [IndVars] Check if WideInc available before trying to use it\n \n WideInc/WideIncExpr can be null. Previously this worked out\n because the comparison with WideIncExpr would fail. Now we have\n accesses to WideInc prior to that. Avoid the issue with an\n explicit check.\n \n Fixes https://github.com/llvm/llvm-project/issues/106239.\n\ndiff --git a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp\nindex 8e3a14bb4f6d..a950a4f57ef4 100644\n--- a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp\n+++ b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp\n@@ -1928,18 +1928,24 @@ Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU,\n if (!WideAddRec.first)\n return nullptr;\n \n- // Reuse the IV increment that SCEVExpander created. Recompute flags, unless\n- // the flags for both increments agree and it is safe to use the ones from\n- // the original inc. In that case, the new use of the wide increment won't\n- // be more poisonous.\n- bool NeedToRecomputeFlags =\n- !SCEVExpander::canReuseFlagsFromOriginalIVInc(OrigPhi, WidePhi,\n- DU.NarrowUse, WideInc) ||\n- DU.NarrowUse->hasNoUnsignedWrap() != WideInc->hasNoUnsignedWrap() ||\n- DU.NarrowUse->hasNoSignedWrap() != WideInc->hasNoSignedWrap();\n+ auto CanUseWideInc = [&]() {\n+ if (!WideInc)\n+ return false;\n+ // Reuse the IV increment that SCEVExpander created. Recompute flags,\n+ // unless the flags for both increments agree and it is safe to use the\n+ // ones from the original inc. In that case, the new use of the wide\n+ // increment won't be more poisonous.\n+ bool NeedToRecomputeFlags =\n+ !SCEVExpander::canReuseFlagsFromOriginalIVInc(\n+ OrigPhi, WidePhi, DU.NarrowUse, WideInc) ||\n+ DU.NarrowUse->hasNoUnsignedWrap() != WideInc->hasNoUnsignedWrap() ||\n+ DU.NarrowUse->hasNoSignedWrap() != WideInc->hasNoSignedWrap();\n+ return WideAddRec.first == WideIncExpr &&\n+ Rewriter.hoistIVInc(WideInc, DU.NarrowUse, NeedToRecomputeFlags);\n+ };\n+\n Instruction *WideUse = nullptr;\n- if (WideAddRec.first == WideIncExpr &&\n- Rewriter.hoistIVInc(WideInc, DU.NarrowUse, NeedToRecomputeFlags))\n+ if (CanUseWideInc())\n WideUse = WideInc;\n else {\n WideUse = cloneIVUser(DU, WideAddRec.first);\n", "tests": [{"file": "llvm/test/Transforms/IndVarSimplify/pr106239.ll", "commands": ["opt -S -passes=indvars < %s"], "tests": [{"test_name": "m", "test_body": "target datalayout = \"n8:16:32:64\"\n\ndefine i32 @m() {\nentry:\n %div.i4 = sdiv i32 1, 0\n br label %for.body.i6\n\nfor.body.i6: ; preds = %if.end.i, %entry\n %add57.i = phi i32 [ %add.i7, %if.end.i ], [ 0, %entry ]\n br i1 true, label %i.exit, label %if.end.i\n\nif.end.i: ; preds = %for.body.i6\n %add.i7 = add i32 %add57.i, %div.i4\n %conv.i = zext i32 %add57.i to i64\n store i64 %conv.i, ptr null, align 8\n br label %for.body.i6\n\ni.exit: ; preds = %for.body.i6\n ret i32 0\n}\n"}]}], "issue": {"title": "[clang] Crash with O2 optimization flag", "body": "I compiled this code with -O2 flag and it crashed. The code is:\r\n\r\n```c\r\n#include \r\nlong a, c;\r\nunsigned d;\r\nint32_t g;\r\nuint8_t h[];\r\nlong b(long p1, long k) {\r\n long e = k;\r\n int b = 0;\r\n while (e) {\r\n e /= 10;\r\n b++;\r\n }\r\n for (int f = 0; f < b; f++)\r\n p1 *= 10;\r\n a = p1 + k;\r\n return a;\r\n}\r\nlong i(int p1, int, int l) {\r\n int j = l / p1;\r\n for (int f = 0; f < p1; f++) {\r\n if (d > l)\r\n break;\r\n c = d += j;\r\n }\r\n}\r\nuint32_t m() {\r\n for (g = 8; b(35, 46) - 3546 + g >= 0; g--)\r\n ;\r\n h[i(g + 1, 0, -4294967221)];\r\n}\r\n```\r\n\r\nThe crash is: clang: /root/llvm-project/llvm/include/llvm/Support/Casting.h:662: decltype(auto) llvm::dyn_cast(From*) [with To = llvm::BinaryOperator; From = llvm::Instruction]: Assertion `detail::isPresent(Val) && \"dyn_cast on a non-existent value\"' failed.\r\n\r\n\r\nThe details can be found here: https://godbolt.org/z/37WsGxMcv", "author": "cardigan1008", "labels": ["regression", "release:backport", "confirmed", "crash", "llvm:transforms"], "comments": [{"author": "shafik", "body": "Looks like a regression, does not crash in clang-18: https://godbolt.org/z/1zYqs3sbz"}, {"author": "dtcxzyw", "body": "Reproducer: https://godbolt.org/z/bbdhPxx5E\r\n\r\n```\r\n; bin/opt -passes=indvars reduced.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\n\r\ndefine i32 @m() {\r\nentry:\r\n %div.i4 = sdiv i32 1, 0\r\n br label %for.body.i6\r\n\r\nfor.body.i6: ; preds = %if.end.i, %entry\r\n %add57.i = phi i32 [ %add.i7, %if.end.i ], [ 0, %entry ]\r\n br i1 true, label %i.exit, label %if.end.i\r\n\r\nif.end.i: ; preds = %for.body.i6\r\n %add.i7 = add i32 %add57.i, %div.i4\r\n %conv.i = zext i32 %add57.i to i64\r\n store i64 %conv.i, ptr null, align 8\r\n br label %for.body.i6\r\n\r\ni.exit: ; preds = %for.body.i6\r\n ret i32 0\r\n}\r\n```\r\ncc @nikic"}, {"author": "DimitryAndric", "body": "This regressed with `llvmorg-19-init-02822-g`4db93e5d564 (\"[IndVars] Recompute flags if needed in widenIVUse of IV increment. (#82352)\") by @fhahn."}]}, "verified": true} {"bug_id": "82243", "issue_url": "https://github.com/llvm/llvm-project/issues/82243", "bug_type": "miscompilation", "base_commit": "fcd6549e5801de938935b93fd2d13020b42eebdb", "knowledge_cutoff": "2024-02-19T12:48:24Z", "lit_test_dir": ["llvm/test/Transforms/IndVarSimplify"], "hints": {"fix_commit": "4db93e5d56440be44d8ad2036e2579208c82ad37", "components": ["ScalarEvolution", "SimplifyIndVar"], "bug_location_lineno": {"llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h": [[258, 263]], "llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp": [[771, 776]], "llvm/lib/Transforms/Utils/SimplifyIndVar.cpp": [[1131, 1137], [1731, 1737], [1825, 1835], [1996, 2006], [2024, 2030]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp": ["SCEVExpander::hoistIVInc"], "llvm/lib/Transforms/Utils/SimplifyIndVar.cpp": ["WidenIV::createWideIV", "WidenIV::widenIVUse"]}}, "patch": "commit 4db93e5d56440be44d8ad2036e2579208c82ad37\nAuthor: Florian Hahn \nDate: Tue Feb 20 14:03:25 2024 +0000\n\n [IndVars] Recompute flags if needed in widenIVUse of IV increment. (#82352)\n \n widenIVUse may hoist a wide induction increment and introduce new uses,\n but does not recompute the wrap flags. In some cases this can make the\n new uses of the wide IV inc more poisonous.\n \n Update the code to recompute flags if needed when hoisting an IV. If\n both the narrow and wide IV increment's flags match and we can re-use\n the flags from the increments, there's no need to recompute the flags,\n as the replacement won't make the new uses of the wide IV's increment\n more poisonous.\n \n Note that this also updates a stale comment which claimed that the widen\n increment is only used if it dominates the new use.\n \n The helper should also be used to guard the code added in da437330be,\n which I am planning on doing separately once the helper lands.\n \n Fixes https://github.com/llvm/llvm-project/issues/82243.\n\ndiff --git a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h\nindex 035705b7f4b7..fa10443f14bb 100644\n--- a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h\n+++ b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h\n@@ -258,6 +258,14 @@ public:\n bool hoistIVInc(Instruction *IncV, Instruction *InsertPos,\n bool RecomputePoisonFlags = false);\n \n+ /// Return true if both increments directly increment the corresponding IV PHI\n+ /// nodes and have the same opcode. It is not safe to re-use the flags from\n+ /// the original increment, if it is more complex and SCEV expansion may have\n+ /// yielded a more simplified wider increment.\n+ static bool canReuseFlagsFromOriginalIVInc(PHINode *OrigPhi, PHINode *WidePhi,\n+ Instruction *OrigInc,\n+ Instruction *WideInc);\n+\n /// replace congruent phis with their most canonical representative. Return\n /// the number of phis eliminated.\n unsigned replaceCongruentIVs(Loop *L, const DominatorTree *DT,\ndiff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\nindex 3a28909473d9..fbe1dba5b8d4 100644\n--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\n+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\n@@ -771,6 +771,15 @@ bool SCEVExpander::hoistIVInc(Instruction *IncV, Instruction *InsertPos,\n return true;\n }\n \n+bool SCEVExpander::canReuseFlagsFromOriginalIVInc(PHINode *OrigPhi,\n+ PHINode *WidePhi,\n+ Instruction *OrigInc,\n+ Instruction *WideInc) {\n+ return match(OrigInc, m_c_BinOp(m_Specific(OrigPhi), m_Value())) &&\n+ match(WideInc, m_c_BinOp(m_Specific(WidePhi), m_Value())) &&\n+ OrigInc->getOpcode() == WideInc->getOpcode();\n+}\n+\n /// Determine if this cyclic phi is in a form that would have been generated by\n /// LSR. We don't care if the phi was actually expanded in this pass, as long\n /// as it is in a low-cost form, for example, no implied multiplication. This\ndiff --git a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp\nindex 5aa6df49e7f6..66bba1ca2f1d 100644\n--- a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp\n+++ b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp\n@@ -1131,7 +1131,8 @@ protected:\n const SCEV *getSCEVByOpCode(const SCEV *LHS, const SCEV *RHS,\n unsigned OpCode) const;\n \n- Instruction *widenIVUse(NarrowIVDefUse DU, SCEVExpander &Rewriter);\n+ Instruction *widenIVUse(NarrowIVDefUse DU, SCEVExpander &Rewriter,\n+ PHINode *OrigPhi, PHINode *WidePhi);\n \n bool widenLoopCompare(NarrowIVDefUse DU);\n bool widenWithVariantUse(NarrowIVDefUse DU);\n@@ -1731,7 +1732,9 @@ bool WidenIV::widenWithVariantUse(WidenIV::NarrowIVDefUse DU) {\n \n /// Determine whether an individual user of the narrow IV can be widened. If so,\n /// return the wide clone of the user.\n-Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU, SCEVExpander &Rewriter) {\n+Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU,\n+ SCEVExpander &Rewriter, PHINode *OrigPhi,\n+ PHINode *WidePhi) {\n assert(ExtendKindMap.count(DU.NarrowDef) &&\n \"Should already know the kind of extension used to widen NarrowDef\");\n \n@@ -1825,11 +1828,18 @@ Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU, SCEVExpander &Rewri\n if (!WideAddRec.first)\n return nullptr;\n \n- // Reuse the IV increment that SCEVExpander created as long as it dominates\n- // NarrowUse.\n+ // Reuse the IV increment that SCEVExpander created. Recompute flags, unless\n+ // the flags for both increments agree and it is safe to use the ones from\n+ // the original inc. In that case, the new use of the wide increment won't\n+ // be more poisonous.\n+ bool NeedToRecomputeFlags =\n+ !SCEVExpander::canReuseFlagsFromOriginalIVInc(OrigPhi, WidePhi,\n+ DU.NarrowUse, WideInc) ||\n+ DU.NarrowUse->hasNoUnsignedWrap() != WideInc->hasNoUnsignedWrap() ||\n+ DU.NarrowUse->hasNoSignedWrap() != WideInc->hasNoSignedWrap();\n Instruction *WideUse = nullptr;\n if (WideAddRec.first == WideIncExpr &&\n- Rewriter.hoistIVInc(WideInc, DU.NarrowUse))\n+ Rewriter.hoistIVInc(WideInc, DU.NarrowUse, NeedToRecomputeFlags))\n WideUse = WideInc;\n else {\n WideUse = cloneIVUser(DU, WideAddRec.first);\n@@ -1996,11 +2006,9 @@ PHINode *WidenIV::createWideIV(SCEVExpander &Rewriter) {\n // the same opcode. It is not safe to re-use the flags from the original\n // increment, if it is more complex and SCEV expansion may have yielded a\n // more simplified wider increment.\n- bool MatchingOps =\n- match(OrigInc, m_c_BinOp(m_Specific(OrigPhi), m_Value())) &&\n- match(WideInc, m_c_BinOp(m_Specific(WidePhi), m_Value())) &&\n- OrigInc->getOpcode() == WideInc->getOpcode();\n- if (MatchingOps && isa(OrigInc) &&\n+ if (SCEVExpander::canReuseFlagsFromOriginalIVInc(OrigPhi, WidePhi,\n+ OrigInc, WideInc) &&\n+ isa(OrigInc) &&\n isa(WideInc)) {\n WideInc->setHasNoUnsignedWrap(WideInc->hasNoUnsignedWrap() ||\n OrigInc->hasNoUnsignedWrap());\n@@ -2024,7 +2032,7 @@ PHINode *WidenIV::createWideIV(SCEVExpander &Rewriter) {\n \n // Process a def-use edge. This may replace the use, so don't hold a\n // use_iterator across it.\n- Instruction *WideUse = widenIVUse(DU, Rewriter);\n+ Instruction *WideUse = widenIVUse(DU, Rewriter, OrigPhi, WidePhi);\n \n // Follow all def-use edges from the previous narrow use.\n if (WideUse)\n", "tests": [{"file": "llvm/test/Transforms/IndVarSimplify/hoist-wide-inc-for-narrow-use-recompute-flags.ll", "commands": ["opt -p indvars -S %s"], "tests": [{"test_name": "test_pr82243", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\n\ndefine void @test_pr82243(ptr %f) {\nentry:\n br label %outer.header\n\nouter.header: ; preds = %outer.latch, %entry\n %iv = phi i32 [ 1, %entry ], [ %iv.next, %outer.latch ]\n %iv.sext = sext i32 %iv to i64\n %gep.iv.ext = getelementptr i32, ptr %f, i64 %iv.sext\n store i32 1, ptr %gep.iv.ext, align 4\n %sub = add i32 %iv, -1\n %shl = shl i32 123, %sub\n %gep.shl = getelementptr i32, ptr %f, i32 %shl\n br label %inner.header\n\ninner.header: ; preds = %inner.latch, %outer.header\n store i32 0, ptr %gep.shl, align 4\n br i1 false, label %inner.latch, label %exit\n\ninner.latch: ; preds = %inner.header\n br i1 false, label %inner.header, label %outer.latch\n\nouter.latch: ; preds = %inner.latch\n %iv.next = add i32 %iv, -1\n br label %outer.header\n\nexit: ; preds = %inner.header\n ret void\n}\n"}]}], "issue": {"title": "Wrong code at -Os on x86-64_gnu-linux (recent regression)", "body": "Clang at -Os produced the wrong code.\r\n\r\nBisected to dce77a357948709e335910ddc07f9c3f2eb2ac4b, which was committed by @fhahn \r\n\r\nCompiler explorer: https://godbolt.org/z/fTrenPK4h\r\n\r\n```console\r\n% cat reduced.c\r\nint printf(const char *, ...);\r\nint a, b, c, d, e;\r\nint f[9];\r\nint g(int i) {\r\n c = 1 << i;\r\n if (b & c)\r\n return 4;\r\n return 0;\r\n}\r\nint j(int i) {\r\n int h = g(i);\r\n return h;\r\n}\r\nint k() {\r\n d = 6;\r\n for (; d; d--) {\r\n e = 0;\r\n for (; e <= 6; e++) {\r\n f[j(d - 1)] = f[d];\r\n for (; e + d;)\r\n return 0;\r\n }\r\n }\r\n return 0;\r\n}\r\nint main() {\r\n k();\r\n printf(\"%d\\n\", a);\r\n}\r\n%\r\n% clang -O3 reduced.c && ./a.out\r\n0\r\n% clang -Os reduced.c && ./a.out\r\n\r\n%\r\n```", "author": "shao-hua-li", "labels": ["miscompilation", "llvm:SCEV"], "comments": [{"author": "fhahn", "body": "The issue is that IndVars doesn't drop poison-generating flags when hoisting the IV to introduce a new use of the widened IV. working on a fix"}, {"author": "fhahn", "body": "Should be fixed, thanks again as always @shao-hua-li "}, {"author": "shao-hua-li", "body": "You're welcome @fhahn "}]}, "verified": true} {"bug_id": "111874", "issue_url": "https://github.com/llvm/llvm-project/issues/111874", "bug_type": "crash", "base_commit": "6a65e98fa7901dc1de91172d065fafb16ce89d77", "knowledge_cutoff": "2024-10-10T17:06:05Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "65da32c634a8345fcbe021f69fc6a609d074c08c", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6480, 6491]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInstructionCost"]}}, "patch": "commit 65da32c634a8345fcbe021f69fc6a609d074c08c\nAuthor: Florian Hahn \nDate: Fri Oct 11 11:26:57 2024 +0100\n\n [LV] Account for any-of reduction when computing costs of blend phis.\n \n Any-of reductions are narrowed to i1. Update the legacy cost model to\n use the correct type when computing the cost of a phi that gets lowered\n to selects (BLEND).\n \n This fixes a divergence between legacy and VPlan-based cost models after\n 36fc291b6ec6d.\n \n Fixes https://github.com/llvm/llvm-project/issues/111874.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 05dc58a42249..54f57fb0b6b5 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6480,12 +6480,32 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I,\n // Phi nodes in non-header blocks (not inductions, reductions, etc.) are\n // converted into select instructions. We require N - 1 selects per phi\n // node, where N is the number of incoming values.\n- if (VF.isVector() && Phi->getParent() != TheLoop->getHeader())\n+ if (VF.isVector() && Phi->getParent() != TheLoop->getHeader()) {\n+ Type *ResultTy = Phi->getType();\n+\n+ // All instructions in an Any-of reduction chain are narrowed to bool.\n+ // Check if that is the case for this phi node.\n+ auto *HeaderUser = cast_if_present(\n+ find_singleton(Phi->users(), [this](User *U, bool) -> User * {\n+ auto *Phi = dyn_cast(U);\n+ if (Phi && Phi->getParent() == TheLoop->getHeader())\n+ return Phi;\n+ return nullptr;\n+ }));\n+ if (HeaderUser) {\n+ auto &ReductionVars = Legal->getReductionVars();\n+ auto Iter = ReductionVars.find(HeaderUser);\n+ if (Iter != ReductionVars.end() &&\n+ RecurrenceDescriptor::isAnyOfRecurrenceKind(\n+ Iter->second.getRecurrenceKind()))\n+ ResultTy = Type::getInt1Ty(Phi->getContext());\n+ }\n return (Phi->getNumIncomingValues() - 1) *\n TTI.getCmpSelInstrCost(\n- Instruction::Select, ToVectorTy(Phi->getType(), VF),\n+ Instruction::Select, ToVectorTy(ResultTy, VF),\n ToVectorTy(Type::getInt1Ty(Phi->getContext()), VF),\n CmpInst::BAD_ICMP_PREDICATE, CostKind);\n+ }\n \n return TTI.getCFInstrCost(Instruction::PHI, CostKind);\n }\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/blend-any-of-reduction-cost.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "any_of_reduction_used_in_blend", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine i32 @any_of_reduction_used_in_blend(ptr %src, i64 %N, i1 %c.0, i1 %c.1) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %any.of.red = phi i32 [ 0, %entry ], [ %any.of.red.next, %loop.latch ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c.0, label %loop.latch, label %else.1\n\nelse.1: ; preds = %loop.header\n br i1 %c.1, label %loop.latch, label %else.2\n\nelse.2: ; preds = %else.1\n %l = load ptr, ptr %src, align 8\n %c.2 = icmp eq ptr %l, null\n %sel = select i1 %c.2, i32 0, i32 %any.of.red\n br label %loop.latch\n\nloop.latch: ; preds = %else.2, %else.1, %loop.header\n %any.of.red.next = phi i32 [ %any.of.red, %loop.header ], [ %any.of.red, %else.1 ], [ %sel, %else.2 ]\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n %res = phi i32 [ %any.of.red.next, %loop.latch ]\n ret i32 %res\n}\n\nattributes #0 = { \"target-cpu\"=\"sifive-p670\" }\n"}, {"test_name": "any_of_reduction_used_in_blend_with_mutliple_phis", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine i32 @any_of_reduction_used_in_blend_with_mutliple_phis(ptr %src, i64 %N, i1 %c.0, i1 %c.1) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %any.of.red = phi i32 [ 0, %entry ], [ %any.of.red.next, %loop.latch ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c.0, label %x.1, label %else.1\n\nelse.1: ; preds = %loop.header\n br i1 %c.1, label %x.1, label %else.2\n\nelse.2: ; preds = %else.1\n %l = load ptr, ptr %src, align 8\n %c.2 = icmp eq ptr %l, null\n %sel = select i1 %c.2, i32 0, i32 %any.of.red\n br label %loop.latch\n\nx.1: ; preds = %else.1, %loop.header\n %p = phi i32 [ %any.of.red, %loop.header ], [ %any.of.red, %else.1 ]\n br label %loop.latch\n\nloop.latch: ; preds = %x.1, %else.2\n %any.of.red.next = phi i32 [ %p, %x.1 ], [ %sel, %else.2 ]\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n %res = phi i32 [ %any.of.red.next, %loop.latch ]\n ret i32 %res\n}\n\nattributes #0 = { \"target-cpu\"=\"sifive-p670\" }\n"}]}], "issue": {"title": "Crash in LV: \"VPlan cost model and legacy cost model disagreed\"", "body": "Loop vectorizer crash for the following reproducer:\r\n```\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @init_dependence_analysis(ptr %iiv_used.i.i.i.sroa.0, i64 %indvars.iv71.i.i.i, i1 %cmp104.not.i.i.i, i1 %tobool108.not.i.i.i) #0 {\r\nentry:\r\n br label %for.cond101.preheader.i.i.i\r\n\r\nfor.cond101.preheader.i.i.i: ; preds = %for.inc121.i.i.i, %entry\r\n %separability.0.i.i = phi i32 [ 0, %entry ], [ %separability.2.i.i, %for.inc121.i.i.i ]\r\n br label %for.body103.i.i.i\r\n\r\nfor.body103.i.i.i: ; preds = %for.inc121.i.i.i, %for.cond101.preheader.i.i.i\r\n %separability.1.i.i = phi i32 [ %separability.0.i.i, %for.cond101.preheader.i.i.i ], [ %separability.2.i.i, %for.inc121.i.i.i ]\r\n %indvars.iv71.i.i.i5 = phi i64 [ 0, %for.cond101.preheader.i.i.i ], [ %indvars.iv.next72.i.i.i, %for.inc121.i.i.i ]\r\n br i1 %cmp104.not.i.i.i, label %for.inc121.i.i.i, label %land.lhs.true105.i.i.i\r\n\r\nland.lhs.true105.i.i.i: ; preds = %for.body103.i.i.i\r\n br i1 %tobool108.not.i.i.i, label %for.inc121.i.i.i, label %land.lhs.true109.i.i.i\r\n\r\nland.lhs.true109.i.i.i: ; preds = %land.lhs.true105.i.i.i\r\n %0 = load ptr, ptr %iiv_used.i.i.i.sroa.0, align 8\r\n %cmp118.i.i.i = icmp eq ptr %0, null\r\n %spec.select209.i.i = select i1 %cmp118.i.i.i, i32 0, i32 %separability.1.i.i\r\n br label %for.inc121.i.i.i\r\n\r\nfor.inc121.i.i.i: ; preds = %land.lhs.true109.i.i.i, %land.lhs.true105.i.i.i, %for.body103.i.i.i\r\n %separability.2.i.i = phi i32 [ %separability.1.i.i, %for.body103.i.i.i ], [ %separability.1.i.i, %land.lhs.true105.i.i.i ], [ %spec.select209.i.i, %land.lhs.true109.i.i.i ]\r\n %indvars.iv.next72.i.i.i = add i64 %indvars.iv71.i.i.i5, 1\r\n %exitcond74.not.i.i.i = icmp eq i64 %indvars.iv71.i.i.i5, %indvars.iv71.i.i.i\r\n br i1 %exitcond74.not.i.i.i, label %for.cond101.preheader.i.i.i, label %for.body103.i.i.i\r\n\r\n; uselistorder directives\r\n uselistorder i32 %separability.2.i.i, { 1, 0 }\r\n}\r\n\r\nattributes #0 = { \"target-cpu\"=\"sifive-p670\" }\r\n```\r\n\r\nopt -S --passes=loop-vectorize ./reduced.ll\r\nLoopVectorize.cpp:7430: VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\n", "author": "alexey-bataev", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "114191", "issue_url": "https://github.com/llvm/llvm-project/issues/114191", "bug_type": "miscompilation", "base_commit": "e577f14b670ee2ae6bb717133310b215be4331b3", "knowledge_cutoff": "2024-10-30T07:41:28Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine", "llvm/test/Transforms/InstSimplify"], "hints": {"fix_commit": "a77dedcacb4c5eb221395b69877981dd6ad98989", "components": ["IR", "InstCombine", "InstructionSimplify"], "bug_location_lineno": {"llvm/lib/Analysis/InstructionSimplify.cpp": [[1095, 1113]], "llvm/lib/IR/ConstantFold.cpp": [[902, 912]], "llvm/lib/Transforms/InstCombine/InstCombineInternal.h": [[102, 107]], "llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp": [[1158, 1186], [1189, 1194], [2138, 2166]]}, "bug_location_funcname": {"llvm/lib/Analysis/InstructionSimplify.cpp": ["simplifyDivRem"], "llvm/lib/IR/ConstantFold.cpp": ["llvm::ConstantFoldBinaryInstruction"], "llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp": ["InstCombinerImpl::commonIDivTransforms", "InstCombinerImpl::commonIRemTransforms", "foldIDivShl"]}}, "patch": "commit a77dedcacb4c5eb221395b69877981dd6ad98989\nAuthor: Yingwei Zheng \nDate: Fri Nov 1 22:56:22 2024 +0800\n\n [InstSimplify][InstCombine][ConstantFold] Move vector div/rem by zero fold to InstCombine (#114280)\n \n Previously we fold `div/rem X, C` into `poison` if any element of the\n constant divisor `C` is zero or undef. However, it is incorrect when\n threading udiv over an vector select:\n https://alive2.llvm.org/ce/z/3Ninx5\n ```\n define <2 x i32> @vec_select_udiv_poison(<2 x i1> %x) {\n %sel = select <2 x i1> %x, <2 x i32> , <2 x i32> \n %div = udiv <2 x i32> , %sel\n ret <2 x i32> %div\n }\n ```\n In this case, `threadBinOpOverSelect` folds `udiv , ` and `udiv , ` into\n `zeroinitializer` and `poison`, respectively. One solution is to\n introduce a new flag indicating that we are threading over a vector\n select. But it requires to modify both `InstSimplify` and\n `ConstantFold`.\n \n However, this optimization doesn't provide benefits to real-world\n programs:\n \n https://dtcxzyw.github.io/llvm-opt-benchmark/coverage/data/zyw/opt-ci/actions-runner/_work/llvm-opt-benchmark/llvm-opt-benchmark/llvm/llvm-project/llvm/lib/IR/ConstantFold.cpp.html#L908\n \n https://dtcxzyw.github.io/llvm-opt-benchmark/coverage/data/zyw/opt-ci/actions-runner/_work/llvm-opt-benchmark/llvm-opt-benchmark/llvm/llvm-project/llvm/lib/Analysis/InstructionSimplify.cpp.html#L1107\n \n This patch moves the fold into InstCombine to avoid breaking numerous\n existing tests.\n \n Fixes #114191 and #113866 (only poison-safety issue).\n\ndiff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp\nindex d08be1e55c85..2cb2612bf611 100644\n--- a/llvm/lib/Analysis/InstructionSimplify.cpp\n+++ b/llvm/lib/Analysis/InstructionSimplify.cpp\n@@ -1095,19 +1095,6 @@ static Value *simplifyDivRem(Instruction::BinaryOps Opcode, Value *Op0,\n if (match(Op1, m_Zero()))\n return PoisonValue::get(Ty);\n \n- // If any element of a constant divisor fixed width vector is zero or undef\n- // the behavior is undefined and we can fold the whole op to poison.\n- auto *Op1C = dyn_cast(Op1);\n- auto *VTy = dyn_cast(Ty);\n- if (Op1C && VTy) {\n- unsigned NumElts = VTy->getNumElements();\n- for (unsigned i = 0; i != NumElts; ++i) {\n- Constant *Elt = Op1C->getAggregateElement(i);\n- if (Elt && (Elt->isNullValue() || Q.isUndefValue(Elt)))\n- return PoisonValue::get(Ty);\n- }\n- }\n-\n // poison / X -> poison\n // poison % X -> poison\n if (isa(Op0))\ndiff --git a/llvm/lib/IR/ConstantFold.cpp b/llvm/lib/IR/ConstantFold.cpp\nindex 07dfbc41e79b..c2780faee403 100644\n--- a/llvm/lib/IR/ConstantFold.cpp\n+++ b/llvm/lib/IR/ConstantFold.cpp\n@@ -902,11 +902,6 @@ Constant *llvm::ConstantFoldBinaryInstruction(unsigned Opcode, Constant *C1,\n Constant *ExtractIdx = ConstantInt::get(Ty, i);\n Constant *LHS = ConstantExpr::getExtractElement(C1, ExtractIdx);\n Constant *RHS = ConstantExpr::getExtractElement(C2, ExtractIdx);\n-\n- // If any element of a divisor vector is zero, the whole op is poison.\n- if (Instruction::isIntDivRem(Opcode) && RHS->isNullValue())\n- return PoisonValue::get(VTy);\n-\n Constant *Res = ConstantExpr::isDesirableBinOp(Opcode)\n ? ConstantExpr::get(Opcode, LHS, RHS)\n : ConstantFoldBinaryInstruction(Opcode, LHS, RHS);\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\nindex 7a060cdab2d3..adbd9186c59c 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\n@@ -102,6 +102,7 @@ public:\n Instruction *visitSRem(BinaryOperator &I);\n Instruction *visitFRem(BinaryOperator &I);\n bool simplifyDivRemOfSelectWithZeroOp(BinaryOperator &I);\n+ Instruction *commonIDivRemTransforms(BinaryOperator &I);\n Instruction *commonIRemTransforms(BinaryOperator &I);\n Instruction *commonIDivTransforms(BinaryOperator &I);\n Instruction *visitUDiv(BinaryOperator &I);\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp\nindex b9c165da906d..f85a3c936513 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp\n@@ -1158,29 +1158,39 @@ static Value *foldIDivShl(BinaryOperator &I, InstCombiner::BuilderTy &Builder) {\n return nullptr;\n }\n \n-/// This function implements the transforms common to both integer division\n-/// instructions (udiv and sdiv). It is called by the visitors to those integer\n-/// division instructions.\n-/// Common integer divide transforms\n-Instruction *InstCombinerImpl::commonIDivTransforms(BinaryOperator &I) {\n- if (Instruction *Phi = foldBinopWithPhiOperands(I))\n- return Phi;\n-\n+/// Common integer divide/remainder transforms\n+Instruction *InstCombinerImpl::commonIDivRemTransforms(BinaryOperator &I) {\n+ assert(I.isIntDivRem() && \"Unexpected instruction\");\n Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);\n- bool IsSigned = I.getOpcode() == Instruction::SDiv;\n+\n+ // If any element of a constant divisor fixed width vector is zero or undef\n+ // the behavior is undefined and we can fold the whole op to poison.\n+ auto *Op1C = dyn_cast(Op1);\n Type *Ty = I.getType();\n+ auto *VTy = dyn_cast(Ty);\n+ if (Op1C && VTy) {\n+ unsigned NumElts = VTy->getNumElements();\n+ for (unsigned i = 0; i != NumElts; ++i) {\n+ Constant *Elt = Op1C->getAggregateElement(i);\n+ if (Elt && (Elt->isNullValue() || isa(Elt)))\n+ return replaceInstUsesWith(I, PoisonValue::get(Ty));\n+ }\n+ }\n+\n+ if (Instruction *Phi = foldBinopWithPhiOperands(I))\n+ return Phi;\n \n // The RHS is known non-zero.\n if (Value *V = simplifyValueKnownNonZero(I.getOperand(1), *this, I))\n return replaceOperand(I, 1, V);\n \n- // Handle cases involving: [su]div X, (select Cond, Y, Z)\n- // This does not apply for fdiv.\n+ // Handle cases involving: div/rem X, (select Cond, Y, Z)\n if (simplifyDivRemOfSelectWithZeroOp(I))\n return &I;\n \n // If the divisor is a select-of-constants, try to constant fold all div ops:\n- // C / (select Cond, TrueC, FalseC) --> select Cond, (C / TrueC), (C / FalseC)\n+ // C div/rem (select Cond, TrueC, FalseC) --> select Cond, (C div/rem TrueC),\n+ // (C div/rem FalseC)\n // TODO: Adapt simplifyDivRemOfSelectWithZeroOp to allow this and other folds.\n if (match(Op0, m_ImmConstant()) &&\n match(Op1, m_Select(m_Value(), m_ImmConstant(), m_ImmConstant()))) {\n@@ -1189,6 +1199,21 @@ Instruction *InstCombinerImpl::commonIDivTransforms(BinaryOperator &I) {\n return R;\n }\n \n+ return nullptr;\n+}\n+\n+/// This function implements the transforms common to both integer division\n+/// instructions (udiv and sdiv). It is called by the visitors to those integer\n+/// division instructions.\n+/// Common integer divide transforms\n+Instruction *InstCombinerImpl::commonIDivTransforms(BinaryOperator &I) {\n+ if (Instruction *Res = commonIDivRemTransforms(I))\n+ return Res;\n+\n+ Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);\n+ bool IsSigned = I.getOpcode() == Instruction::SDiv;\n+ Type *Ty = I.getType();\n+\n const APInt *C2;\n if (match(Op1, m_APInt(C2))) {\n Value *X;\n@@ -2138,29 +2163,11 @@ static Instruction *simplifyIRemMulShl(BinaryOperator &I,\n /// remainder instructions.\n /// Common integer remainder transforms\n Instruction *InstCombinerImpl::commonIRemTransforms(BinaryOperator &I) {\n- if (Instruction *Phi = foldBinopWithPhiOperands(I))\n- return Phi;\n+ if (Instruction *Res = commonIDivRemTransforms(I))\n+ return Res;\n \n Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);\n \n- // The RHS is known non-zero.\n- if (Value *V = simplifyValueKnownNonZero(I.getOperand(1), *this, I))\n- return replaceOperand(I, 1, V);\n-\n- // Handle cases involving: rem X, (select Cond, Y, Z)\n- if (simplifyDivRemOfSelectWithZeroOp(I))\n- return &I;\n-\n- // If the divisor is a select-of-constants, try to constant fold all rem ops:\n- // C % (select Cond, TrueC, FalseC) --> select Cond, (C % TrueC), (C % FalseC)\n- // TODO: Adapt simplifyDivRemOfSelectWithZeroOp to allow this and other folds.\n- if (match(Op0, m_ImmConstant()) &&\n- match(Op1, m_Select(m_Value(), m_ImmConstant(), m_ImmConstant()))) {\n- if (Instruction *R = FoldOpIntoSelect(I, cast(Op1),\n- /*FoldWithMultiUse*/ true))\n- return R;\n- }\n-\n if (isa(Op1)) {\n if (Instruction *Op0I = dyn_cast(Op0)) {\n if (SelectInst *SI = dyn_cast(Op0I)) {\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/div.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "sdiv_constant_dividend_select_of_constants_divisor_vec", "test_body": "define <2 x i8> @sdiv_constant_dividend_select_of_constants_divisor_vec(i1 %b) {\n %s = select i1 %b, <2 x i8> , <2 x i8> \n %r = sdiv <2 x i8> , %s\n ret <2 x i8> %r\n}\n"}, {"test_name": "udiv_constant_dividend_select_of_constants_divisor_vec", "test_body": "define <2 x i8> @udiv_constant_dividend_select_of_constants_divisor_vec(i1 %b) {\n %s = select i1 %b, <2 x i8> , <2 x i8> \n %r = udiv <2 x i8> , %s\n ret <2 x i8> %r\n}\n"}, {"test_name": "sdiv_constant_dividend_select_of_constants_divisor_vec_ub1", "test_body": "define <2 x i8> @sdiv_constant_dividend_select_of_constants_divisor_vec_ub1(i1 %b) {\n %s = select i1 %b, <2 x i8> , <2 x i8> \n %r = sdiv <2 x i8> , %s\n ret <2 x i8> %r\n}\n"}, {"test_name": "udiv_constant_dividend_select_of_constants_divisor_vec_ub1", "test_body": "define <2 x i8> @udiv_constant_dividend_select_of_constants_divisor_vec_ub1(i1 %b) {\n %s = select i1 %b, <2 x i8> , <2 x i8> \n %r = udiv <2 x i8> , %s\n ret <2 x i8> %r\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/rem.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "urem_constant_dividend_select_of_constants_divisor_vec_ub1", "test_body": "define <2 x i8> @urem_constant_dividend_select_of_constants_divisor_vec_ub1(i1 %b) {\n %s = select i1 %b, <2 x i8> , <2 x i8> \n %r = urem <2 x i8> , %s\n ret <2 x i8> %r\n}\n"}, {"test_name": "urem_constant_dividend_select_of_constants_divisor_vec", "test_body": "define <2 x i8> @urem_constant_dividend_select_of_constants_divisor_vec(i1 %b) {\n %s = select i1 %b, <2 x i8> , <2 x i8> \n %r = urem <2 x i8> , %s\n ret <2 x i8> %r\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/vector-udiv.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "vec_select_udiv_poison", "test_body": "define <2 x i32> @vec_select_udiv_poison(<2 x i1> %x) {\n %sel = select <2 x i1> %x, <2 x i32> splat (i32 -1), <2 x i32> \n %div = udiv <2 x i32> , %sel\n ret <2 x i32> %div\n}\n"}, {"test_name": "test_v4i32_zext_shl_const_pow2", "test_body": "define <4 x i32> @test_v4i32_zext_shl_const_pow2(<4 x i32> %a0, <4 x i16> %a1) {\n %1 = shl <4 x i16> , %a1\n %2 = zext <4 x i16> %1 to <4 x i32>\n %3 = udiv <4 x i32> %a0, %2\n ret <4 x i32> %3\n}\n"}]}, {"file": "llvm/test/Transforms/InstSimplify/div.ll", "commands": ["opt < %s -passes=instsimplify -S"], "tests": [{"test_name": "udiv_zero_elt_vec", "test_body": "define <2 x i8> @udiv_zero_elt_vec(<2 x i8> %x) {\n %div = udiv <2 x i8> %x, \n ret <2 x i8> %div\n}\n"}, {"test_name": "sdiv_zero_elt_vec", "test_body": "define <2 x i8> @sdiv_zero_elt_vec(<2 x i8> %x) {\n %div = sdiv <2 x i8> %x, \n ret <2 x i8> %div\n}\n"}, {"test_name": "udiv_poison_elt_vec", "test_body": "define <2 x i8> @udiv_poison_elt_vec(<2 x i8> %x) {\n %div = udiv <2 x i8> %x, \n ret <2 x i8> %div\n}\n"}, {"test_name": "sdiv_zero_elt_vec_constfold", "test_body": "define <2 x i8> @sdiv_zero_elt_vec_constfold(<2 x i8> %x) {\n %div = sdiv <2 x i8> , \n ret <2 x i8> %div\n}\n"}, {"test_name": "sdiv_poison_elt_vec", "test_body": "define <2 x i8> @sdiv_poison_elt_vec(<2 x i8> %x) {\n %div = sdiv <2 x i8> %x, \n ret <2 x i8> %div\n}\n"}, {"test_name": "udiv_zero_elt_vec_constfold", "test_body": "define <2 x i8> @udiv_zero_elt_vec_constfold(<2 x i8> %x) {\n %div = udiv <2 x i8> , \n ret <2 x i8> %div\n}\n"}, {"test_name": "zero_dividend_vector_poison_elt", "test_body": "define <2 x i32> @zero_dividend_vector_poison_elt(<2 x i32> %A) {\n %B = sdiv <2 x i32> , %A\n ret <2 x i32> %B\n}\n"}]}, {"file": "llvm/test/Transforms/InstSimplify/rem.ll", "commands": ["opt < %s -passes=instsimplify -S"], "tests": [{"test_name": "srem_zero_elt_vec_constfold", "test_body": "define <2 x i8> @srem_zero_elt_vec_constfold(<2 x i8> %x) {\n %rem = srem <2 x i8> , \n ret <2 x i8> %rem\n}\n"}, {"test_name": "urem_zero_elt_vec", "test_body": "define <2 x i8> @urem_zero_elt_vec(<2 x i8> %x) {\n %rem = urem <2 x i8> %x, \n ret <2 x i8> %rem\n}\n"}, {"test_name": "urem_zero_elt_vec_constfold", "test_body": "define <2 x i8> @urem_zero_elt_vec_constfold(<2 x i8> %x) {\n %rem = urem <2 x i8> , \n ret <2 x i8> %rem\n}\n"}, {"test_name": "srem_undef_elt_vec", "test_body": "define <2 x i8> @srem_undef_elt_vec(<2 x i8> %x) {\n %rem = srem <2 x i8> %x, \n ret <2 x i8> %rem\n}\n"}, {"test_name": "urem_undef_elt_vec", "test_body": "define <2 x i8> @urem_undef_elt_vec(<2 x i8> %x) {\n %rem = urem <2 x i8> %x, \n ret <2 x i8> %rem\n}\n"}, {"test_name": "zero_dividend_vector_poison_elt", "test_body": "define <2 x i32> @zero_dividend_vector_poison_elt(<2 x i32> %A) {\n %B = urem <2 x i32> , %A\n ret <2 x i32> %B\n}\n"}, {"test_name": "srem_zero_elt_vec", "test_body": "define <2 x i8> @srem_zero_elt_vec(<2 x i8> %x) {\n %rem = srem <2 x i8> %x, \n ret <2 x i8> %rem\n}\n"}]}], "issue": {"title": "[InstCombine] miscompilation of `or` of vectors", "body": "Alive2 report: https://alive2.llvm.org/ce/z/LyM8sY\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine <2 x i32> @or_sext_to_sel_constant_vec.3(<2 x i1> %y) {\r\n#0:\r\n %sext = sext <2 x i1> %y to <2 x i32>\r\n %#1 = or <2 x i32> %sext, { 0, 1 }\r\n %#2 = udiv <2 x i32> { 42, 4294967289 }, %#1\r\n ret <2 x i32> %#2\r\n}\r\n=>\r\ndefine <2 x i32> @or_sext_to_sel_constant_vec.3(<2 x i1> %y) {\r\n#0:\r\n ret <2 x i32> { 0, 0 }\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Value mismatch\r\n\r\nExample:\r\n<2 x i1> %y = < #x1 (1), #x0 (0) >\r\n\r\nSource:\r\n<2 x i32> %sext = < #xffffffff (4294967295, -1), #x00000000 (0) >\r\n<2 x i32> %#1 = < #xffffffff (4294967295, -1), #x00000001 (1) >\r\n<2 x i32> %#2 = < #x00000000 (0), #xfffffff9 (4294967289, -7) >\r\n\r\nTarget:\r\nSource value: < #x00000000 (0), #xfffffff9 (4294967289, -7) >\r\nTarget value: < #x00000000 (0), #x00000000 (0) >\r\n\r\n\r\n----------------------------------------\r\ndefine <2 x i32> @or_sext_to_sel_constant_vec.2(<2 x i1> %y) {\r\n#0:\r\n %sext = sext <2 x i1> %y to <2 x i32>\r\n %#1 = or <2 x i32> %sext, { 0, 1 }\r\n %#2 = sdiv <2 x i32> { 42, 4294967289 }, %#1\r\n %#3 = udiv exact <2 x i32> %#2, %#2\r\n %#4 = or <2 x i32> %#2, %#2\r\n ret <2 x i32> %#4\r\n}\r\n=>\r\ndefine <2 x i32> @or_sext_to_sel_constant_vec.2(<2 x i1> %y) {\r\n#0:\r\n ret <2 x i32> { 4294967254, 7 }\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Value mismatch\r\n\r\nExample:\r\n<2 x i1> %y = < #x1 (1), #x0 (0) >\r\n\r\nSource:\r\n<2 x i32> %sext = < #xffffffff (4294967295, -1), #x00000000 (0) >\r\n<2 x i32> %#1 = < #xffffffff (4294967295, -1), #x00000001 (1) >\r\n<2 x i32> %#2 = < #xffffffd6 (4294967254, -42), #xfffffff9 (4294967289, -7) >\r\n<2 x i32> %#3 = < #x00000001 (1), #x00000001 (1) >\r\n<2 x i32> %#4 = < #xffffffd6 (4294967254, -42), #xfffffff9 (4294967289, -7) >\r\n\r\nTarget:\r\nSource value: < #xffffffd6 (4294967254, -42), #xfffffff9 (4294967289, -7) >\r\nTarget value: < #xffffffd6 (4294967254, -42), #x00000007 (7) >\r\n\r\nSummary:\r\n 0 correct transformations\r\n 2 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:instcombine", "llvm:analysis"], "comments": [{"author": "dtcxzyw", "body": "@bongjunj Feel free to ping me if you find new InstCombine miscompilation bugs without involving undef values/NaN payloads."}, {"author": "dtcxzyw", "body": "https://github.com/llvm/llvm-project/blob/eac2c182c6f852fc187af9952250a43d6fb17b28/llvm/lib/Analysis/InstructionSimplify.cpp#L1098-L1109\r\n\r\nWe cannot do this simplification when threading udiv over selects.\r\n "}]}, "verified": true} {"bug_id": "115149", "issue_url": "https://github.com/llvm/llvm-project/issues/115149", "bug_type": "miscompilation", "base_commit": "0f44d72e0ee74970cf696ff4c791f63e0c3fa9b4", "knowledge_cutoff": "2024-11-06T11:22:14Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "929cbe7f596733f85cd274485acc19442dd34a80", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp": [[545, 551]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp": ["InstCombinerImpl::foldPHIArgGEPIntoPHI"]}}, "patch": "commit 929cbe7f596733f85cd274485acc19442dd34a80\nAuthor: Antonio Frighetto \nDate: Tue Nov 12 19:02:04 2024 +0100\n\n [InstCombine] Intersect nowrap flags between geps while folding into phi\n \n A miscompilation issue has been addressed with refined checking.\n \n Fixes: https://github.com/llvm/llvm-project/issues/115149.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp b/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp\nindex cb5c44730512..1fcf1c570add 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp\n@@ -545,7 +545,8 @@ Instruction *InstCombinerImpl::foldPHIArgGEPIntoPHI(PHINode &PN) {\n // especially bad when the PHIs are in the header of a loop.\n bool NeededPhi = false;\n \n- GEPNoWrapFlags NW = GEPNoWrapFlags::all();\n+ // Remember flags of the first phi-operand getelementptr.\n+ GEPNoWrapFlags NW = FirstInst->getNoWrapFlags();\n \n // Scan to see if all operands are the same opcode, and all have one user.\n for (Value *V : drop_begin(PN.incoming_values())) {\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/opaque-ptr.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "phi_of_gep_flags_1", "test_body": "define ptr @phi_of_gep_flags_1(i1 %c, ptr %p) {\n br i1 %c, label %if, label %else\n\nif: ; preds = %0\n %gep1 = getelementptr inbounds i32, ptr %p, i64 1\n br label %join\n\nelse: ; preds = %0\n %gep2 = getelementptr nusw nuw i32, ptr %p, i64 1\n br label %join\n\njoin: ; preds = %else, %if\n %phi = phi ptr [ %gep1, %if ], [ %gep2, %else ]\n ret ptr %phi\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/phi.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "wrong_gep_arg_into_phi", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i64 @wrong_gep_arg_into_phi(ptr noundef %ptr) {\nentry:\n %add.ptr = getelementptr i8, ptr %ptr, i64 1\n br label %for.cond\n\nfor.cond: ; preds = %for.cond, %entry\n %.pn = phi ptr [ %add.ptr, %entry ], [ %incdec.ptr, %for.cond ]\n %val = load i8, ptr %.pn, align 1\n %cond = icmp ne i8 %val, 0\n %incdec.ptr = getelementptr inbounds nuw i8, ptr %.pn, i64 1\n br i1 %cond, label %for.cond, label %exit\n\nexit: ; preds = %for.cond\n ret i64 0\n}\n"}]}], "issue": {"title": "[clang] Miscompile with -O3 and -O0/1/2 since 18.1.0", "body": "When I compiled this code with -O0/1/2, its output is 0. However, with -O3, it returned 143:\r\n\r\n```c\r\nint printf(const char *, ...);\r\nchar a, b;\r\nint c;\r\nchar *e = &b;\r\nint f(char *g, int *k) {\r\n char *d = g + *k;\r\n for (; *d && *d <= ' '; d++)\r\n ;\r\n if (*d)\r\n return 0;\r\n return 1;\r\n}\r\nint l(int g) {\r\n char h[] = {a, a, a};\r\n int i[] = {g};\r\n int j = f(h, i);\r\n return j;\r\n}\r\nlong m() {\r\n *e = 255;\r\n for (; l(b + 1);)\r\n return 0;\r\n for (;;)\r\n ;\r\n}\r\nint main() {\r\n m();\r\n printf(\"%d\\n\", c);\r\n}\r\n```\r\n\r\nDetails can be found here: https://godbolt.org/z/dnvvTPror", "author": "cardigan1008", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "antoniofrighetto", "body": "This looks like a nasty one. Getting the optimized IR via -O3 and setting noinline/optnone to main, we get: https://llvm.godbolt.org/z/vaPvs6d16. \r\n\r\nThis is the initial assembly output with -O1 (which leads to the infinite loop):\r\n\r\n```asm\r\nmain: # @main\r\n push rax\r\n mov rax, qword ptr [e]\r\n mov byte ptr [rax], -1\r\n movsx rax, byte ptr [b]\r\n mov cl, byte ptr [a]\r\n mov byte ptr [rsp + 6], cl\r\n mov byte ptr [rsp + 7], cl\r\n lea rdx, [rsp + 5]\r\n add rdx, rax\r\n mov rax, rdx\r\n.LBB0_1: # %for.cond.i.i.i\r\n add rax, 1\r\n mov cl, byte ptr [rdx + 1]\r\n ...\r\n```\r\nWhereas this is with -O0 (returns 0):\r\n```asm\r\nmain: # @main\r\n sub rsp, 24\r\n mov rax, qword ptr [e]\r\n mov byte ptr [rax], -1\r\n movsx rcx, byte ptr [b]\r\n mov al, byte ptr [a]\r\n mov byte ptr [rsp + 22], al\r\n mov byte ptr [rsp + 23], al\r\n lea rax, [rsp + 21]\r\n add rax, rcx\r\n mov qword ptr [rsp + 8], rax # 8-byte Spill\r\n.LBB0_1: # %for.cond.i.i.i\r\n mov rcx, qword ptr [rsp + 8] # 8-byte Reload\r\n mov rax, rcx\r\n add rax, 1\r\n mov dl, byte ptr [rcx + 1]\r\n```\r\nIn the optimized one, $rdx is set to point to $rsp+0x5, part of the stack slot containing $rax (undef), later dereferenced to set the condition leading to the loop. Initially I thought some bug in regalloc, but to me it seems more something in prologepiloginserter. Function prologue / epilogue seems set up correctly with triple arm64-apple-macosx."}, {"author": "antoniofrighetto", "body": "Looking at the IR better, it seems like InstCombine performs a wrong optimization immediately after inlining `l` into `m`, probably related with default argument conversion done by operator + (which should include integer promotion on types smaller than integer). Alive2 seems to confirm it: https://alive2.llvm.org/ce/z/XAzrip."}, {"author": "antoniofrighetto", "body": "Reduced to:\r\n```llvm\r\n@a = global i8 0, align 1\r\n\r\ndefine i64 @src(i8 noundef %arg) {\r\nentry:\r\n %h.i = alloca [3 x i8], align 1\r\n %conv = sext i8 %arg to i32\r\n %add = add nsw i32 %conv, 1\r\n %2 = load i8, ptr @a, align 1\r\n store i8 %2, ptr %h.i, align 1\r\n %idx.ext.i.i = sext i32 %add to i64\r\n %add.ptr.i.i = getelementptr inbounds i8, ptr %h.i, i64 %idx.ext.i.i\r\n br label %for.cond.i.i\r\n\r\nfor.cond.i.i: ; preds = %for.cond.i.i, %entry\r\n %d.0.i.i = phi ptr [ %add.ptr.i.i, %entry ], [ %incdec.ptr.i.i, %for.cond.i.i ]\r\n %3 = load i8, ptr %d.0.i.i, align 1\r\n %tobool.not.i.i = icmp ne i8 %3, 0\r\n %incdec.ptr.i.i = getelementptr inbounds nuw i8, ptr %d.0.i.i, i64 1\r\n br i1 %tobool.not.i.i, label %for.cond.i.i, label %l.exit\r\n\r\nl.exit: ; preds = %for.cond.i.i\r\n ret i64 0\r\n}\r\n```\r\nWrong sext of add of sext simplified to sext leads to wrong gep offset: https://alive2.llvm.org/ce/z/JaR4j9."}, {"author": "AZero13", "body": "Can we backport this to 19.x?"}, {"author": "antoniofrighetto", "body": "Should be fixed in 929cbe7f596733f85cd274485acc19442dd34a80, thanks for reporting!"}, {"author": "AZero13", "body": "@antoniofrighetto can we please backport this "}]}, "verified": true} {"bug_id": "120237", "issue_url": "https://github.com/llvm/llvm-project/issues/120237", "bug_type": "hang", "base_commit": "9e333872199b1e3bf488d71e222ff4b6f0370347", "knowledge_cutoff": "2024-12-17T14:27:28Z", "lit_test_dir": ["llvm/test/Analysis/ScalarEvolution"], "hints": {"fix_commit": "acfd26a93be3fb70013560f3fb894eb9086e7e32", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[15753, 15758], [15764, 15773]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::LoopGuards::collectFromBlock"]}}, "patch": "commit acfd26a93be3fb70013560f3fb894eb9086e7e32\nAuthor: Julian Nagele \nDate: Fri Dec 20 15:44:15 2024 +0100\n\n [SCEV] Fix exit condition for recursive loop guard collection (#120442)\n \n When assumptions are present `Terms.size()` does not actually count the\n number of conditions collected from dominating branches; introduce a\n separate counter.\n \n Fixes https://github.com/llvm/llvm-project/issues/120237\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex d55d09020fc1..1e4bb1d606cd 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -15753,6 +15753,7 @@ void ScalarEvolution::LoopGuards::collectFromBlock(\n // predecessors that can be found that have unique successors leading to the\n // original header.\n // TODO: share this logic with isLoopEntryGuardedByCond.\n+ unsigned NumCollectedConditions = 0;\n std::pair Pair(Pred, Block);\n for (; Pair.first;\n Pair = SE.getPredecessorWithUniqueSuccessorForBB(Pair.first)) {\n@@ -15764,10 +15765,11 @@ void ScalarEvolution::LoopGuards::collectFromBlock(\n \n Terms.emplace_back(LoopEntryPredicate->getCondition(),\n LoopEntryPredicate->getSuccessor(0) == Pair.second);\n+ NumCollectedConditions++;\n \n // If we are recursively collecting guards stop after 2\n- // predecessors to limit compile-time impact for now.\n- if (Depth > 0 && Terms.size() == 2)\n+ // conditions to limit compile-time impact for now.\n+ if (Depth > 0 && NumCollectedConditions == 2)\n break;\n }\n // Finally, if we stopped climbing the predecessor chain because\n", "tests": [{"file": "llvm/test/Analysis/ScalarEvolution/backedge-taken-count-guard-info-with-multiple-predecessors.ll", "commands": ["opt < %s -disable-output \"-passes=print\" -scalar-evolution-max-iterations=0 -scalar-evolution-classify-expressions=0 2>&1"], "tests": [{"test_name": "pr120442", "test_body": "; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\ndefine void @pr120442(i1 %c.1, i1 %c.2) {\nentry:\n call void @llvm.assume(i1 %c.1)\n call void @llvm.assume(i1 %c.2)\n br label %outer.header\n\nouter.header: ; preds = %bb, %entry\n %phi7 = phi i32 [ 0, %bb ], [ 0, %entry ]\n br label %inner.header\n\nbb: ; preds = %bb\n br i1 false, label %outer.header, label %bb\n\ninner.header: ; preds = %inner.header, %outer.header\n %phi = phi i32 [ %add, %inner.header ], [ 0, %outer.header ]\n %add = add i32 %phi, 1\n %icmp = icmp ugt i32 %add, 0\n br i1 %icmp, label %exit, label %inner.header\n\nexit: ; preds = %inner.header\n ret void\n}\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}]}], "issue": {"title": "Program terminated with signal: SIGKILL Compiler returned: 137", "body": "To reproduce run the following test with -passes=loop-vectorize on opt:\n```\n; ModuleID = './reduced.ll'\nsource_filename = \"./reduced.ll\"\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @wombat() #0 gc \"statepoint-example\" {\nbb:\n call void @llvm.assume(i1 false)\n call void @llvm.assume(i1 false)\n br label %bb6\n\nbb1: ; preds = %bb2, %bb6\n br label %bb2\n\nbb2: ; preds = %bb2, %bb1\n %phi = phi i32 [ %add4, %bb2 ], [ 0, %bb1 ]\n %phi3 = phi i32 [ 0, %bb2 ], [ 0, %bb1 ]\n %add = add i32 0, 0\n %add4 = add i32 %phi, 1\n %icmp = icmp ugt i32 %phi, 0\n br i1 %icmp, label %bb1, label %bb2\n\nbb5: ; preds = %bb5\n br i1 false, label %bb6, label %bb5\n\nbb6: ; preds = %bb5, %bb\n %phi7 = phi i32 [ 0, %bb5 ], [ 0, %bb ]\n br label %bb1\n}\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #1\n\nattributes #0 = { \"target-features\"=\"+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,-amx-fp8,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,-movrs,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-avx10.2-256,-gfni,-avxvnniint16,-amx-fp16,-zu,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,-nf,-amx-tf32,-amx-avx512,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-amx-transpose,-avx10.2-512,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-amx-movrs,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f\" }\nattributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) \"target-features\"=\"+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,-amx-fp8,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,-movrs,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-avx10.2-256,-gfni,-avxvnniint16,-amx-fp16,-zu,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,-nf,-amx-tf32,-amx-avx512,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-amx-transpose,-avx10.2-512,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-amx-movrs,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f\" }\n```\n\nReproducer: https://godbolt.org/z/x7WrWvoxa\n\n", "author": "TatyanaDoubts", "labels": ["llvm:SCEV", "crash"], "comments": [{"author": "TatyanaDoubts", "body": "The godbolt is not informative. But locally I see the following stack dump:\n```\nbin/opt -S -passes=loop-vectorize reduced-stripped.ll\nLLVM ERROR: SmallVector unable to grow. Requested capacity (4294967296) is larger than maximum value for size type (4294967295)\n0. Program arguments: bin/opt -S -passes=loop-vectorize reduced-stripped.ll\n1. Running pass \"function(loop-vectorize)\" on module \"reduced-stripped.ll\"\n2. Running pass \"loop-vectorize\" on function \"wombat\"\n #0 0x00007f945d6be47b llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) llvm/lib/Support/Unix/Signals.inc:727:3\n #1 0x00007f945d6bbfbb llvm::sys::RunSignalHandlers() llvm/lib/Support/Signals.cpp:105:20\n #2 0x00007f945d6bc2fe SignalHandler(int) llvm/lib/Support/Unix/Signals.inc:413:1\n #3 0x00007f9462a89980 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x12980)\n #4 0x00007f945c0fbe87 raise (/lib/x86_64-linux-gnu/libc.so.6+0x3ee87)\n #5 0x00007f945c0fd7f1 abort (/lib/x86_64-linux-gnu/libc.so.6+0x407f1)\n #6 0x00007f945d435674 std::__cxx11::basic_string, std::allocator>::_M_replace(unsigned long, unsigned long, char const*, unsigned long) gcc-13.3.0-rhel6/include/c++/13.3.0/bits/basic_string.tcc:545:21\n #7 0x00007f945d435674 std::__cxx11::basic_string, std::allocator>::assign(char const*) (.isra.0.cold) gcc-13.3.0-rhel6/include/c++/13.3.0/bits\n/basic_string.h:1684:19\n #8 0x00007f945d63c540 std::__cxx11::basic_string, std::allocator>::replace(unsigned long, unsigned long, char const*, unsigned long) gcc-13.3.0-rhel6/include/c++/13.3.0/bits/basic_string.h:2208:19\n #9 0x00007f945d63c540 std::__cxx11::basic_string, std::allocator>::insert(unsigned long, std::__cxx11::basic_string, std::allocator> const\n&) gcc-13.3.0-rhel6/include/c++/13.3.0/bits/basic_string.h:1895:29\n#10 0x00007f945d63c540 std::__cxx11::basic_string, std::allocator> std::operator+, std::allocator>(std::__cxx11::basic_string, std::allocator>&&, std::__cxx11::basic_string, std::allocator>&&) gcc-13.3.0-rhel6/include/c++/13.3.0/bits/basic_string.\nh:3666:35\n#11 0x00007f945d63c540 report_size_overflow(unsigned long, unsigned long) (.constprop.0) llvm/lib/Support/SmallVector.cpp:65:73\n#12 0x00007f945d63c79e ......\n#13 0x00007f945f6b3c9c llvm::PointerIntPairInfo>::updateInt(long, long)\nllvm/include/llvm/ADT/PointerIntPair.h:213:55\n#14 0x00007f945f6b3c9c llvm::PointerIntPair, llvm::PointerIntPairInfo>>::setPointer\nAndInt(llvm::Value*, bool) & llvm/include/llvm/ADT/PointerIntPair.h:111:11\n#15 0x00007f945f6b3c9c llvm::PointerIntPair, llvm::PointerIntPairInfo>>::PointerInt\nPair(llvm::Value*, bool) llvm/include/llvm/ADT/PointerIntPair.h:89:21\n#16 0x00007f945f6b3c9c llvm::PointerIntPair, llvm::PointerIntPairInfo>>& llvm::Smal\nlVectorTemplateBase, llvm::PointerIntPairInfo>>, true>::growAn\ndEmplaceBack(llvm::Value*&&, bool&&) llvm/include/llvm/ADT/SmallVector.h:559:14\n#17 0x00007f945f6b3c9c llvm::PointerIntPair, llvm::PointerIntPairInfo>>& llvm::Smal\nlVectorImpl, llvm::PointerIntPairInfo>>>::emplace_back(llvm::Value*&&, bool&&) llvm/include/llvm/ADT/SmallVector.h:946:38\n#18 0x00007f945f6b3c9c llvm::ScalarEvolution::LoopGuards::collectFromBlock(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::BasicBlock const*, llvm::BasicBlock const*, llvm::SmallPtrSetImpl&, unsigned int) llvm/lib/Analysis/ScalarEvolution.cpp:15708:23\n#19 0x00007f945f6b51fa llvm::DenseMapIterator, llvm::detail::DenseMapPair, false>::operator->() const llvm/include/llvm/ADT/DenseMap.h:1238:5\n#20 0x00007f945f6b51fa llvm::ScalarEvolution::LoopGuards::collectFromPHI(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::PHINode const&, llvm::SmallPtrSetImpl&, llvm::SmallDenseMap, llvm::detail::DenseMapPair>&, unsigned int)::'lambda'(unsigned int)::operator()(unsigned int) const llvm/lib/Analysis/ScalarEvolution.cpp:15285:25\n#21 0x00007f945f6b200c llvm::ScalarEvolution::LoopGuards::collectFromPHI(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::PHINode const&, llvm::SmallPtrSetImpl&, llvm::SmallDenseMap, llvm::detail::DenseMapPair>&, unsigned int) llvm/lib/Analysis/ScalarEvolution.cpp:15317:26\n#22 0x00007f945f6b40af llvm::BasicBlock::phi_iterator_impl, false, true>>::operator++() llvm/include/llvm/IR/BasicBlock.h:505:7\n#23 0x00007f945f6b40af llvm::ScalarEvolution::LoopGuards::collectFromBlock(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::BasicBlock const*, llvm::BasicBlock const*, llvm::SmallPtrSetImpl&, unsigned int) llvm/lib/Analysis/ScalarEvolution.cpp:15724:40\n#24 0x00007f945f6b534b llvm::SmallPtrSetImplBase::~SmallPtrSetImplBase() llvm/include/llvm/ADT/SmallPtrSet.h:87:5\n#25 0x00007f945f6b534b llvm::SmallPtrSetImpl::~SmallPtrSetImpl() llvm/include/llvm/ADT/SmallPtrSet.h:368:7\n#26 0x00007f945f6b534b llvm::SmallPtrSet::~SmallPtrSet() llvm/include/llvm/ADT/SmallPtrSet.h:524:7\n#27 0x00007f945f6b534b llvm::ScalarEvolution::LoopGuards::collect(llvm::Loop const*, llvm::ScalarEvolution&) llvm/lib/Analysis/ScalarEvolution.cpp:15266:1\n#28 0x00007f945f6b539a llvm::ScalarEvolution::LoopGuards::rewrite(llvm::SCEV const*) const llvm/lib/Analysis/ScalarEvolution.cpp:15892:3\n#29 0x00007f945f6b539a llvm::ScalarEvolution::applyLoopGuards(llvm::SCEV const*, llvm::ScalarEvolution::LoopGuards const&) llvm/lib/Analysis/ScalarEvolution.cpp:15905:24\n#30 0x00007f945f6b539a llvm::ScalarEvolution::applyLoopGuards(llvm::SCEV const*, llvm::Loop const*) llvm/lib/Analysis/ScalarEvolution.cpp:15900:25\n#31 0x00007f945f1e1c35 llvm::LoopVectorizationCostModel::computeMaxVF(llvm::ElementCount, unsigned int) llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:4119:38\n#32 0x00007f945f1e5bfb llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7150:51\n#33 0x00007f945f1e7b71 llvm::LoopVectorizePass::processLoop(llvm::Loop*) llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10095:45\n#34 0x00007f945f1eb181 llvm::LoopVectorizePass::runImpl(llvm::Function&) llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10441:27\n#35 0x00007f945f1eb777 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10478:39\n#36 0x00007f94606acedd llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) llvm/include/llvm/IR/PassManagerInternal.h:91:3\n#37 0x00007f945d8d889f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) llvm/include/llvm/IR/PassManagerImpl.h:85:18\n#38 0x00007f94601642ed llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) llvm/include/llvm/IR/PassManagerInternal.h:91:3\n#39 0x00007f945d8d8dea llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) llvm/lib/IR/PassManager.cpp:129:41\n#40 0x00007f94601646dd llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) llvm/include/llvm/IR/PassManagerInternal.h:91:3\n#41 0x00007f945d8da12f llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) llvm/include/llvm/IR/PassManagerImpl.h:85:18\n#42 0x000000000042ccba llvm::SmallPtrSetImplBase::~SmallPtrSetImplBase() llvm/include/llvm/ADT/SmallPtrSet.h:87:5\n#43 0x000000000042ccba llvm::SmallPtrSetImpl::~SmallPtrSetImpl() llvm/include/llvm/ADT/SmallPtrSet.h:368:7\n#44 0x000000000042ccba llvm::SmallPtrSet::~SmallPtrSet() llvm/include/llvm/ADT/SmallPtrSet.h:524:7\n#45 0x000000000042ccba llvm::PreservedAnalyses::~PreservedAnalyses() llvm/include/llvm/IR/Analysis.h:111:7\n#46 0x000000000042ccba llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) llvm/tools/opt/NewPMDriver.cpp:553:10\n#47 0x00000000004207e3 std::__cxx11::basic_string, std::allocator>::~basic_string() gcc-13.3.0-rhel6/include/c++/13.3.0/bits/basic_string.h:804:19\n#48 0x00000000004207e3 optMain llvm/tools/opt/optdriver.cpp:765:3\n#49 0x00007f945c0dec87 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x21c87)\n......\n```"}, {"author": "fhahn", "body": "@juliannagele could you take a look? Seems like a crash in `collectFromPHI` you recently added"}, {"author": "juliannagele", "body": "> [@juliannagele](https://github.com/juliannagele) could you take a look? Seems like a crash in `collectFromPHI` you recently added\n\nSure, checking!"}, {"author": "juliannagele", "body": "Thanks for the report, should be fixed here: https://github.com/llvm/llvm-project/pull/120442"}]}, "verified": true} {"bug_id": "112922", "issue_url": "https://github.com/llvm/llvm-project/issues/112922", "bug_type": "crash", "base_commit": "b5fa4fee46c1d0046cc395e3338ae13fe6e2cb84", "knowledge_cutoff": "2024-10-18T15:25:44Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "2a6b09e0d3d3c1a05d3d5165202a6e68900974b1", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[5738, 5751], [5760, 5767]], "llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[2958, 2968], [2976, 2983]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInterleaveGroupCost"], "llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPInterleaveRecipe::computeCost"]}}, "patch": "commit 2a6b09e0d3d3c1a05d3d5165202a6e68900974b1\nAuthor: Florian Hahn \nDate: Fri Oct 18 19:12:39 2024 -0700\n\n [LV] Use type from InsertPos for cost computation of interleave groups.\n \n Previously the legacy cost model would pick the type for the cost\n computation depending on the order of the members in the input IR.\n This is incompatible with the VPlan-based cost model (independent of\n original IR order) and also doesn't match code-gen, which uses the type\n of the insert position.\n \n Update the legacy cost model to use the type (and address space) from\n the Group's insert position.\n \n This brings the legacy cost model in line with the legacy cost model and\n fixes a divergence between both models.\n \n Note that the X86 cost model seems to assign different costs to groups\n with i64 and double types. Added a TODO to check.\n \n Fixes https://github.com/llvm/llvm-project/issues/112922.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 857efbdf687c..ce0903b838aa 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -5738,14 +5738,15 @@ LoopVectorizationCostModel::getGatherScatterCost(Instruction *I,\n InstructionCost\n LoopVectorizationCostModel::getInterleaveGroupCost(Instruction *I,\n ElementCount VF) {\n- Type *ValTy = getLoadStoreType(I);\n- auto *VectorTy = cast(ToVectorTy(ValTy, VF));\n- unsigned AS = getLoadStoreAddressSpace(I);\n- enum TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n-\n const auto *Group = getInterleavedAccessGroup(I);\n assert(Group && \"Fail to get an interleaved access group.\");\n \n+ Instruction *InsertPos = Group->getInsertPos();\n+ Type *ValTy = getLoadStoreType(InsertPos);\n+ auto *VectorTy = cast(ToVectorTy(ValTy, VF));\n+ unsigned AS = getLoadStoreAddressSpace(InsertPos);\n+ enum TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n+\n unsigned InterleaveFactor = Group->getFactor();\n auto *WideVecTy = VectorType::get(ValTy, VF * InterleaveFactor);\n \n@@ -5760,8 +5761,9 @@ LoopVectorizationCostModel::getInterleaveGroupCost(Instruction *I,\n (Group->requiresScalarEpilogue() && !isScalarEpilogueAllowed()) ||\n (isa(I) && (Group->getNumMembers() < Group->getFactor()));\n InstructionCost Cost = TTI.getInterleavedMemoryOpCost(\n- I->getOpcode(), WideVecTy, Group->getFactor(), Indices, Group->getAlign(),\n- AS, CostKind, Legal->isMaskRequired(I), UseMaskForGaps);\n+ InsertPos->getOpcode(), WideVecTy, Group->getFactor(), Indices,\n+ Group->getAlign(), AS, CostKind, Legal->isMaskRequired(I),\n+ UseMaskForGaps);\n \n if (Group->isReverse()) {\n // TODO: Add support for reversed masked interleaved access.\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex f4a1f58debba..41f13cc2d9a9 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -2958,11 +2958,20 @@ void VPInterleaveRecipe::print(raw_ostream &O, const Twine &Indent,\n \n InstructionCost VPInterleaveRecipe::computeCost(ElementCount VF,\n VPCostContext &Ctx) const {\n- Instruction *I = getInsertPos();\n+ Instruction *InsertPos = getInsertPos();\n+ // Find the VPValue index of the interleave group. We need to skip gaps.\n+ unsigned InsertPosIdx = 0;\n+ for (unsigned Idx = 0; IG->getFactor(); ++Idx)\n+ if (auto *Member = IG->getMember(Idx)) {\n+ if (Member == InsertPos)\n+ break;\n+ InsertPosIdx++;\n+ }\n Type *ValTy = Ctx.Types.inferScalarType(\n- getNumDefinedValues() > 0 ? getVPValue(0) : getStoredValues()[0]);\n+ getNumDefinedValues() > 0 ? getVPValue(InsertPosIdx)\n+ : getStoredValues()[InsertPosIdx]);\n auto *VectorTy = cast(ToVectorTy(ValTy, VF));\n- unsigned AS = getLoadStoreAddressSpace(I);\n+ unsigned AS = getLoadStoreAddressSpace(InsertPos);\n enum TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n \n unsigned InterleaveFactor = IG->getFactor();\n@@ -2976,8 +2985,8 @@ InstructionCost VPInterleaveRecipe::computeCost(ElementCount VF,\n \n // Calculate the cost of the whole interleaved group.\n InstructionCost Cost = Ctx.TTI.getInterleavedMemoryOpCost(\n- I->getOpcode(), WideVecTy, IG->getFactor(), Indices, IG->getAlign(), AS,\n- CostKind, getMask(), NeedsMaskForGaps);\n+ InsertPos->getOpcode(), WideVecTy, IG->getFactor(), Indices,\n+ IG->getAlign(), AS, CostKind, getMask(), NeedsMaskForGaps);\n \n if (!IG->isReverse())\n return Cost;\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/interleave-cost.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "geps_feeding_interleave_groups_with_reuse", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.15.0\"\n\ndefine void @geps_feeding_interleave_groups_with_reuse(ptr %arg, i64 %arg1, ptr %arg2) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %shl.iv.5 = shl i64 %iv, 5\n %gep.1 = getelementptr i8, ptr %arg, i64 %shl.iv.5\n %add.5 = or disjoint i64 %shl.iv.5, 16\n %gep.2 = getelementptr i8, ptr %arg, i64 %add.5\n %shl.iv.4 = shl i64 %iv, 4\n %gep.3 = getelementptr i8, ptr %arg2, i64 %shl.iv.4\n %l.1 = load float, ptr %gep.1, align 4\n %l.2 = load float, ptr %gep.2, align 4\n %add.1 = fadd float %l.1, %l.2\n %mul.1 = fmul float %add.1, 0.000000e+00\n store float %mul.1, ptr %gep.3, align 4\n %gep.4 = getelementptr i8, ptr %gep.1, i64 4\n %l.3 = load float, ptr %gep.4, align 4\n %gep.5 = getelementptr i8, ptr %gep.2, i64 4\n %l.4 = load float, ptr %gep.5, align 4\n %add.2 = fadd float %l.3, %l.4\n %mul.2 = fmul float %add.2, 0.000000e+00\n %gep.6 = getelementptr i8, ptr %gep.3, i64 4\n store float %mul.2, ptr %gep.6, align 4\n %gep.7 = getelementptr i8, ptr %gep.1, i64 8\n %l.5 = load float, ptr %gep.7, align 4\n %gep.8 = getelementptr i8, ptr %gep.2, i64 8\n %l.6 = load float, ptr %gep.8, align 4\n %add.3 = fadd float %l.5, %l.6\n %mul.3 = fmul float %add.3, 0.000000e+00\n %gep.9 = getelementptr i8, ptr %gep.3, i64 8\n store float %mul.3, ptr %gep.9, align 4\n %i27 = getelementptr i8, ptr %gep.1, i64 12\n %l.7 = load float, ptr %i27, align 4\n %gep.10 = getelementptr i8, ptr %gep.2, i64 12\n %l.8 = load float, ptr %gep.10, align 4\n %add.4 = fadd float %l.7, %l.8\n %mul.4 = fmul float %add.4, 0.000000e+00\n %gep.11 = getelementptr i8, ptr %gep.3, i64 12\n store float %mul.4, ptr %gep.11, align 4\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, %arg1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+sse4.2\" }\n"}, {"test_name": "geps_feeding_interleave_groups_with_reuse2", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.15.0\"\n\ndefine void @geps_feeding_interleave_groups_with_reuse2(ptr %A, ptr %B, i64 %N) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next.7, %loop ]\n %shr.1 = lshr exact i64 %iv, 1\n %gep.B = getelementptr nusw i32, ptr %B, i64 %shr.1\n %l = load i32, ptr %gep.B, align 4\n %gep.A = getelementptr i32, ptr %A, i64 %iv\n store i32 %l, ptr %gep.A, align 4\n %iv.next = or disjoint i64 %iv, 1\n %gep.A.1 = getelementptr i32, ptr %A, i64 %iv.next\n store i32 0, ptr %gep.A.1, align 4\n %iv.next.1 = or disjoint i64 %iv, 2\n %shr.2 = lshr exact i64 %iv.next.1, 1\n %gep.B.2 = getelementptr i32, ptr %B, i64 %shr.2\n %0 = load i32, ptr %gep.B.2, align 4\n %gep.A.2 = getelementptr i32, ptr %A, i64 %iv.next.1\n store i32 %0, ptr %gep.A.2, align 4\n %iv.next.2 = or disjoint i64 %iv, 3\n %gep.A.3 = getelementptr i32, ptr %A, i64 %iv.next.2\n store i32 0, ptr %gep.A.3, align 4\n %iv.next.3 = or disjoint i64 %iv, 4\n %gep.B.4 = getelementptr i32, ptr %B, i64 %iv\n %1 = load i32, ptr %gep.B.4, align 4\n %gep.A.4 = getelementptr i32, ptr %A, i64 %iv.next.3\n store i32 %1, ptr %gep.A.4, align 4\n %iv.next.4 = or disjoint i64 %iv, 5\n %gep.A.5 = getelementptr i32, ptr %A, i64 %iv.next.4\n store i32 0, ptr %gep.A.5, align 4\n %iv.next.5 = or disjoint i64 %iv, 6\n %gep.A.6 = getelementptr i32, ptr %A, i64 %iv.next.5\n store i32 0, ptr %gep.A.6, align 4\n %iv.next.6 = or disjoint i64 %iv, 7\n %gep.A.7 = getelementptr i32, ptr %A, i64 %iv.next.6\n store i32 0, ptr %gep.A.7, align 4\n %iv.next.7 = add nuw nsw i64 %iv, 8\n %ec = icmp eq i64 %iv, %N\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"min-legal-vector-width\"=\"0\" \"target-cpu\"=\"cascadelake\" }\n"}, {"test_name": "interleave_store_double_i64_2", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.15.0\"\n\ndefine void @interleave_store_double_i64_2(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.1 = getelementptr { i64, double }, ptr %dst, i64 %iv, i32 1\n store double 0.000000e+00, ptr %gep.1, align 8\n %gep.0 = getelementptr { i64, double }, ptr %dst, i64 %iv\n store i64 %iv, ptr %gep.0, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "interleave_store_i64_double", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.15.0\"\n\ndefine void @interleave_store_i64_double(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.0 = getelementptr { double, i64 }, ptr %dst, i64 %iv\n store double 0.000000e+00, ptr %gep.0, align 8\n %gep.1 = getelementptr { double, i64 }, ptr %dst, i64 %iv, i32 1\n store i64 %iv, ptr %gep.1, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "test_free_instructions_feeding_geps_for_interleave_groups", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.15.0\"\n\ndefine void @test_free_instructions_feeding_geps_for_interleave_groups(ptr noalias %p.invar, ptr noalias %dst.1, ptr noalias %dst.2) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %l.0 = load float, ptr %p.invar, align 4\n %iv.mul = shl i64 %iv, 2\n %gep.dst.19 = getelementptr float, ptr %dst.1, i64 %iv.mul\n store float %l.0, ptr %gep.dst.19, align 4\n %l.1 = load float, ptr %p.invar, align 4\n %add.1 = or disjoint i64 %iv.mul, 1\n %gep.dst.119 = getelementptr float, ptr %dst.1, i64 %add.1\n store float %l.1, ptr %gep.dst.119, align 4\n %add.2 = or disjoint i64 %iv.mul, 2\n %gep.dst.129 = getelementptr float, ptr %dst.1, i64 %add.2\n store float 0.000000e+00, ptr %gep.dst.129, align 4\n %add.3 = or disjoint i64 %iv.mul, 3\n %gep.dst.140 = getelementptr float, ptr %dst.1, i64 %add.3\n store float 0.000000e+00, ptr %gep.dst.140, align 4\n %l.2 = load float, ptr %p.invar, align 4\n %gep.dst.247 = getelementptr float, ptr %dst.2, i64 %iv.mul\n store float %l.2, ptr %gep.dst.247, align 4\n %gep.dst.255 = getelementptr float, ptr %dst.2, i64 %add.1\n store float 0.000000e+00, ptr %gep.dst.255, align 4\n %gep.dst.265 = getelementptr float, ptr %dst.2, i64 %add.2\n store float 0.000000e+00, ptr %gep.dst.265, align 4\n %gep.dst.276 = getelementptr float, ptr %dst.2, i64 %add.3\n store float 0.000000e+00, ptr %gep.dst.276, align 4\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 0\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "interleave_store_double_i64", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.15.0\"\n\ndefine void @interleave_store_double_i64(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.1 = getelementptr { double, i64 }, ptr %dst, i64 %iv, i32 1\n store i64 %iv, ptr %gep.1, align 8\n %gep.0 = getelementptr { double, i64 }, ptr %dst, i64 %iv\n store double 0.000000e+00, ptr %gep.0, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "interleave_store_i64_double_2", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.15.0\"\n\ndefine void @interleave_store_i64_double_2(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.0 = getelementptr { i64, double }, ptr %dst, i64 %iv\n store i64 %iv, ptr %gep.0, align 8\n %gep.1 = getelementptr { i64, double }, ptr %dst, i64 %iv, i32 1\n store double 0.000000e+00, ptr %gep.1, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "[clang] Assertion in `llvm::LoopVectorizationPlanner::computeBestVF()`", "body": "When building the METIS (v5.1.0) library (GKLib) an assertion is thrown during LoopVectorize. This error was introduced within the past few days (see git sha of broken and working versions below).\r\n\r\n#### Error:\r\n```shell\r\nclang: /mnt/thera/cicd/actions_work/ActionsRunner/ActionsRunner/trunk_repos/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7457: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0. Program arguments: /share/contrib-modules/Core/llvm/2024-10-16/bin/clang -I/home/bcornill/libraries/metis-5.1.0/GKlib/. -I/home/bcornill/libraries/metis-5.1.0/GKlib/test -DLINUX -D_FILE_OFFSET_BITS=64 -DNDEBUG -DNDEBUG2 -DHAVE_EXECINFO_H -DHAVE_GETLINE -O3 -MD -MT CMakeFiles/GKlib.dir/blas.c.o -MF CMakeFiles/GKlib.dir/blas.c.o.d -o CMakeFiles/GKlib.dir/blas.c.o -c /home/bcornill/libraries/metis-5.1.0/GKlib/blas.c\r\n1. parser at end of file\r\n2. Optimizer\r\n3. Running pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"/home/bcornill/libraries/metis-5.1.0/GKlib/blas.c\"\r\n4. Running pass \"loop-vectorize\" on function \"gk_dargmax_n\"\r\n #0 0x00007fa15911cc8b llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMSupport.so.20.0git+0x1c2c8b)\r\n #1 0x00007fa15911a664 llvm::sys::CleanupOnSignal(unsigned long) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMSupport.so.20.0git+0x1c0664)\r\n #2 0x00007fa1590227f8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\r\n #3 0x00007fa1615e0dd0 __restore_rt sigaction.c:0:0\r\n #4 0x00007fa15821e70f raise (/lib64/libc.so.6+0x3770f)\r\n #5 0x00007fa158208b25 abort (/lib64/libc.so.6+0x21b25)\r\n #6 0x00007fa1582089f9 _nl_load_domain.cold.0 loadmsgcat.c:0:0\r\n #7 0x00007fa158216cc6 .annobin___GI___assert_fail.end assert.c:0:0\r\n #8 0x00007fa15db3af1b llvm::LoopVectorizationPlanner::computeBestVF() (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMVectorize.so.20.0git+0xb5f1b)\r\n #9 0x00007fa15db52d46 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMVectorize.so.20.0git+0xcdd46)\r\n#10 0x00007fa15db55ed1 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMVectorize.so.20.0git+0xd0ed1)\r\n#11 0x00007fa15db56517 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMVectorize.so.20.0git+0xd1517)\r\n#12 0x00007fa155d90a9e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/../lib/libLLVMPasses.so.20.0git+0x76a9e)\r\n#13 0x00007fa15998eab7 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMCore.so.20.0git+0x323ab7)\r\n#14 0x00007fa16434e17e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMX86CodeGen.so.20.0git+0xa917e)\r\n#15 0x00007fa15998ce55 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMCore.so.20.0git+0x321e55)\r\n#16 0x00007fa16434e56e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMX86CodeGen.so.20.0git+0xa956e)\r\n#17 0x00007fa15998d9bc llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMCore.so.20.0git+0x3229bc)\r\n#18 0x00007fa15faab288 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#19 0x00007fa15faaeb9c clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libclangCodeGen.so.20.0git+0x111b9c)\r\n#20 0x00007fa15ffdd293 clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libclangCodeGen.so.20.0git+0x640293)\r\n#21 0x00007fa150c5592c clang::ParseAST(clang::Sema&, bool, bool) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/../lib/libclangParse.so.20.0git+0x3c92c)\r\n#22 0x00007fa15d75dee9 clang::FrontendAction::Execute() (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libclangFrontend.so.20.0git+0x138ee9)\r\n#23 0x00007fa15d6ceb05 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libclangFrontend.so.20.0git+0xa9b05)\r\n#24 0x00007fa1613cad83 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libclangFrontendTool.so.20.0git+0x4d83)\r\n#25 0x00000000004141c7 cc1_main(llvm::ArrayRef, char const*, void*) (/share/contrib-modules/Core/llvm/2024-10-16/bin/clang+0x4141c7)\r\n#26 0x000000000040b96a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#27 0x00007fa15d135049 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\r\n#28 0x00007fa159022cb4 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libLLVMSupport.so.20.0git+0xc8cb4)\r\n#29 0x00007fa15d13589e clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\r\n#30 0x00007fa15d0f22fc clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libclangDriver.so.20.0git+0x922fc)\r\n#31 0x00007fa15d0f33fd clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libclangDriver.so.20.0git+0x933fd)\r\n#32 0x00007fa15d1039c4 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/../lib/libclangDriver.so.20.0git+0xa39c4)\r\n#33 0x0000000000410b5b clang_main(int, char**, llvm::ToolContext const&) (/share/contrib-modules/Core/llvm/2024-10-16/bin/clang+0x410b5b)\r\n#34 0x000000000040b074 main (/share/contrib-modules/Core/llvm/2024-10-16/bin/clang+0x40b074)\r\n#35 0x00007fa15820a6a3 __libc_start_main (/lib64/libc.so.6+0x236a3)\r\n#36 0x000000000040b0be _start (/share/contrib-modules/Core/llvm/2024-10-16/bin/clang+0x40b0be)\r\nclang: error: clang frontend command failed with exit code 134 (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project ae68d532f810e217c747b10b26aeea3bb84c3844)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin\r\nBuild config: +assertions\r\nConfiguration file: /share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/clang.cfg\r\nclang: note: diagnostic msg:\r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclang: note: diagnostic msg: /tmp/blas-9fe25d.c\r\nclang: note: diagnostic msg: /tmp/blas-9fe25d.sh\r\nclang: note: diagnostic msg:\r\n\r\n********************\r\n```\r\n#### Reproducer:\r\n[gklib-blas.tar.gz](https://github.com/user-attachments/files/17437550/gklib-blas.tar.gz)\r\n\r\n#### Broken Versions:\r\n```shell\r\n>> clang --version\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project ae68d532f810e217c747b10b26aeea3bb84c3844)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin\r\nBuild config: +assertions\r\nConfiguration file: /share/contrib-modules/Core/llvm/2024-10-16_20.0-0/bin/clang.cfg\r\n\r\n>> clang --version\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project af90e7c5161de9a36af768dd5c9d73464e0eed64)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /share/contrib-modules/Core/llvm/2024-10-18_20.0-0/bin\r\nBuild config: +assertions\r\nConfiguration file: /share/contrib-modules/Core/llvm/2024-10-18_20.0-0/bin/clang.cfg\r\n```\r\n\r\n#### Working version:\r\n```shell\r\n>> clang --version\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project 48deb3568eb2452ff385b04b8f71c34121f47387)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /share/contrib-modules/Core/llvm/2024-10-13_20.0-0/bin\r\nBuild config: +assertions\r\nConfiguration file: /share/contrib-modules/Core/llvm/2024-10-13_20.0-0/bin/clang.cfg\r\n```", "author": "bcornille", "labels": ["vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "Thanks, taking a look!"}]}, "verified": true} {"bug_id": "112577", "issue_url": "https://github.com/llvm/llvm-project/issues/112577", "bug_type": "miscompilation", "base_commit": "76196998e25b98d81abc437708622261810782ca", "knowledge_cutoff": "2024-10-16T16:30:23Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "709abacdc350d63c61888607edb28ce272daa0a0", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[17224, 17232]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::collectValuesToDemote"]}}, "patch": "commit 709abacdc350d63c61888607edb28ce272daa0a0\nAuthor: Alexey Bataev \nDate: Fri Oct 18 13:54:30 2024 -0700\n\n [SLP]Check that operand of abs does not overflow before making it part of minbitwidth transformation\n \n Need to check that the operand of the abs intrinsic can be safely\n truncated before making it part of the minbitwidth transformation.\n \n Fixes #112577\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex e7b52fbdcc3b..e1aa6127ac03 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -17224,9 +17224,25 @@ bool BoUpSLP::collectValuesToDemote(\n MaskedValueIsZero(I->getOperand(1), Mask, SimplifyQuery(*DL)));\n });\n };\n+ auto AbsChecker = [&](unsigned BitWidth, unsigned OrigBitWidth) {\n+ assert(BitWidth <= OrigBitWidth && \"Unexpected bitwidths!\");\n+ return all_of(E.Scalars, [&](Value *V) {\n+ auto *I = cast(V);\n+ unsigned SignBits = OrigBitWidth - BitWidth;\n+ APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, BitWidth - 1);\n+ unsigned Op0SignBits =\n+ ComputeNumSignBits(I->getOperand(0), *DL, 0, AC, nullptr, DT);\n+ return SignBits <= Op0SignBits &&\n+ ((SignBits != Op0SignBits &&\n+ !isKnownNonNegative(I->getOperand(0), SimplifyQuery(*DL))) ||\n+ MaskedValueIsZero(I->getOperand(0), Mask, SimplifyQuery(*DL)));\n+ });\n+ };\n if (ID != Intrinsic::abs) {\n Operands.push_back(getOperandEntry(&E, 1));\n CallChecker = CompChecker;\n+ } else {\n+ CallChecker = AbsChecker;\n }\n InstructionCost BestCost =\n std::numeric_limits::max();\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/abs-overflow-incorrect-minbws.ll", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test(i32 %n) {\nentry:\n %n1 = add i32 %n, 1\n %zn1 = zext nneg i32 %n1 to i64\n %m1 = mul nuw nsw i64 %zn1, 273837369\n %a1 = call i64 @llvm.abs.i64(i64 %m1, i1 true)\n %t1 = trunc i64 %a1 to i32\n %n2 = add i32 %n, 2\n %zn2 = zext nneg i32 %n2 to i64\n %m2 = mul nuw nsw i64 %zn2, 273837369\n %a2 = call i64 @llvm.abs.i64(i64 %m2, i1 true)\n %t2 = trunc i64 %a2 to i32\n %res1 = add i32 %t1, %t2\n ret i32 %res1\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i64 @llvm.abs.i64(i64, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[SLP] SLP Vectorizer miscompile from reducing abs bitwidth", "body": "Bisected miscompile to this change https://github.com/llvm/llvm-project/commit/4a1c53f9fabbc18d436dcd4d5b572b82656fbbf9\r\n\r\nReduced internal test to this example:\r\nhttps://godbolt.org/z/njWYP9exd\r\n```llvm\r\ndefine dso_local noundef i32 @test(i32 %n) local_unnamed_addr #0 {\r\nentry:\r\n %n1 = add i32 %n, 1\r\n %zn1 = zext nneg i32 %n1 to i64\r\n %m1 = mul nuw nsw i64 %zn1, 273837369\r\n %a1 = call noundef i64 @llvm.abs.i64(i64 %m1, i1 true)\r\n %t1 = trunc i64 %a1 to i32\r\n %n2 = add i32 %n, 2\r\n %zn2 = zext nneg i32 %n2 to i64\r\n %m2 = mul nuw nsw i64 %zn2, 273837369\r\n %a2 = call noundef i64 @llvm.abs.i64(i64 %m2, i1 true)\r\n %t2 = trunc i64 %a2 to i32\r\n\r\n %res1 = add i32 %t1, %t2\r\n ret i32 %res1\r\n}\r\n```\r\nRunning slp vectorizer on it gives\r\n```llvm\r\ndefine dso_local noundef i32 @test(i32 %n) local_unnamed_addr {\r\n %0 = insertelement <2 x i32> poison, i32 %n, i32 0\r\n %1 = shufflevector <2 x i32> %0, <2 x i32> poison, <2 x i32> zeroinitializer\r\n %2 = add <2 x i32> %1, \r\n %3 = mul <2 x i32> %2, \r\n %4 = call <2 x i32> @llvm.abs.v2i32(<2 x i32> %3, i1 false)\r\n %5 = extractelement <2 x i32> %4, i32 0\r\n %6 = extractelement <2 x i32> %4, i32 1\r\n %res1 = add i32 %5, %6\r\n ret i32 %res1\r\n}\r\n```\r\nwhich is incorrect. Counterexample by alive2:\r\nhttps://alive2.llvm.org/ce/z/cEcY-u", "author": "danilaml", "labels": ["miscompilation", "release:backport", "llvm:SLPVectorizer"], "comments": [{"author": "AZero13", "body": "Do we plan on backporting this to 19.x @alexey-bataev "}]}, "verified": true} {"bug_id": "88297", "issue_url": "https://github.com/llvm/llvm-project/issues/88297", "bug_type": "crash", "base_commit": "9ec6c5d26321f5d32d97218f850ae7cafda32b2e", "knowledge_cutoff": "2024-04-10T16:52:41Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "34777c238b117b52dd41a9d12e8b54fb83677a12", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanValue.h": [[361, 368], [376, 382]]}, "bug_location_funcname": {}}, "patch": "commit 34777c238b117b52dd41a9d12e8b54fb83677a12\nAuthor: Florian Hahn \nDate: Tue Apr 16 21:24:24 2024 +0100\n\n [VPlan] Don't mark VPBlendRecipe as phi-like.\n \n VPBlendRecipes don't get lowered to phis and usually do not appear at\n the beginning of blocks, due to their masks appearing before them.\n \n This effectively relaxes an over-eager verifier message.\n \n Fixes https://github.com/llvm/llvm-project/issues/88297.\n Fixes https://github.com/llvm/llvm-project/issues/88804.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanValue.h b/llvm/lib/Transforms/Vectorize/VPlanValue.h\nindex da3a768552fc..3f8d4f4fe7d6 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanValue.h\n+++ b/llvm/lib/Transforms/Vectorize/VPlanValue.h\n@@ -361,8 +361,8 @@ public:\n VPWidenMemoryInstructionSC,\n VPWidenSC,\n VPWidenSelectSC,\n- // START: Phi-like recipes. Need to be kept together.\n VPBlendSC,\n+ // START: Phi-like recipes. Need to be kept together.\n VPWidenPHISC,\n VPPredInstPHISC,\n // START: SubclassID for recipes that inherit VPHeaderPHIRecipe.\n@@ -376,7 +376,7 @@ public:\n VPReductionPHISC,\n // END: SubclassID for recipes that inherit VPHeaderPHIRecipe\n // END: Phi-like recipes\n- VPFirstPHISC = VPBlendSC,\n+ VPFirstPHISC = VPWidenPHISC,\n VPFirstHeaderPHISC = VPCanonicalIVPHISC,\n VPLastHeaderPHISC = VPReductionPHISC,\n VPLastPHISC = VPReductionPHISC,\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/blend-in-header.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "", "test_body": "\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\n\n; Test with blend recipe in header VPBB, from\n; https://github.com/llvm/llvm-project/issues/88297.\ndefine i64 @pr88297() {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 false, label %loop.latch, label %then\n\nthen:\n br label %loop.latch\n\nloop.latch:\n %r = phi i64 [ 1, %then ], [ 0, %loop.header ]\n %iv.next = add i32 %iv, 1\n %icmp = icmp sgt i32 %iv, 1000\n br i1 %icmp, label %exit, label %loop.header\n\nexit:\n %r.lcssa = phi i64 [ %r, %loop.latch ]\n ret i64 %r.lcssa\n}\n\ndefine i64 @pr88297_incoming_ops_reordered() {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 false, label %loop.latch, label %then\n\nthen:\n br label %loop.latch\n\nloop.latch:\n %r = phi i64 [ 0, %loop.header ], [ 1, %then ]\n %iv.next = add i32 %iv, 1\n %icmp = icmp sgt i32 %iv, 1000\n br i1 %icmp, label %exit, label %loop.header\n\nexit:\n %r.lcssa = phi i64 [ %r, %loop.latch ]\n ret i64 %r.lcssa\n}\n\ndefine i64 @invar_cond(i1 %c) {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %loop.latch, label %then\n\nthen:\n br label %loop.latch\n\nloop.latch:\n %r = phi i64 [ 1, %then ], [ 0, %loop.header ]\n %iv.next = add i32 %iv, 1\n %icmp = icmp sgt i32 %iv, 1000\n br i1 %icmp, label %exit, label %loop.header\n\nexit:\n %r.lcssa = phi i64 [ %r, %loop.latch ]\n ret i64 %r.lcssa\n}\n\ndefine i64 @invar_cond_incoming_ops_reordered(i1 %c) {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %loop.latch, label %then\n\nthen:\n br label %loop.latch\n\nloop.latch:\n %r = phi i64 [ 0, %loop.header ], [ 1, %then ]\n %iv.next = add i32 %iv, 1\n %icmp = icmp sgt i32 %iv, 1000\n br i1 %icmp, label %exit, label %loop.header\n\nexit:\n %r.lcssa = phi i64 [ %r, %loop.latch ]\n ret i64 %r.lcssa\n}\n;.\n;."}]}], "issue": {"title": "Assertion `verifyVPlanIsValid(*Plan) && \"VPlan is invalid\"' failed", "body": "The following test fails when run opt with -passes loop-vectorize\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat() gc \"statepoint-example\" {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb2, %bb\r\n %phi = phi i32 [ %add, %bb2 ], [ 0, %bb ]\r\n br i1 false, label %bb2, label %bb4\r\n\r\nbb2: ; preds = %bb4, %bb1\r\n %phi3 = phi i64 [ 1, %bb4 ], [ 0, %bb1 ]\r\n %add = add i32 %phi, 1\r\n %icmp = icmp sgt i32 %phi, 0\r\n br i1 %icmp, label %bb5, label %bb1\r\n\r\nbb4: ; preds = %bb1\r\n br label %bb2\r\n\r\nbb5: ; preds = %bb2\r\n %phi6 = phi i64 [ %phi3, %bb2 ]\r\n ret void\r\n}\r\n```\r\nReproducer https://godbolt.org/z/9YW1o7qo6\r\nStack dump:\r\n```\r\nFound non-header PHI recipe in header VPBB: BLEND ir<%phi3> = ir<1> ir<0>/ir\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:8555: void llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount): Assertion `verifyVPlanIsValid(*Plan) && \"VPlan is invalid\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes loop-vectorize \r\n #0 0x0000000004d05ee8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d05ee8)\r\n #1 0x0000000004d0363c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f7623242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007f76232969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007f7623242476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007f76232287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007f762322871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007f7623239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000003de70b8 llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3de70b8)\r\n #9 0x0000000003dee731 llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3dee731)\r\n#10 0x0000000003df18fe llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3df18fe)\r\n#11 0x0000000003df4c6d llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3df4c6d)\r\n#12 0x0000000003df5e7b llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3df5e7b)\r\n#13 0x0000000002d67f0e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2d67f0e)\r\n#14 0x0000000000db3554 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdb3554)\r\n#15 0x0000000004b1663e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4b1663e)\r\n#16 0x0000000000d97cce llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xd97cce)\r\n#17 0x0000000004b1454c llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4b1454c)\r\n#18 0x00000000008c74f2 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8c74f2)\r\n#19 0x00000000008bad45 optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8bad45)\r\n#20 0x00007f7623229d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#21 0x00007f7623229e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#22 0x00000000008b1aee _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8b1aee)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "121581", "issue_url": "https://github.com/llvm/llvm-project/issues/121581", "bug_type": "miscompilation", "base_commit": "95c5c5d4badf7c2128d098be325356e15c2197be", "knowledge_cutoff": "2025-01-03T16:31:11Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "fac46469977da9c4e9c6eeaac21103c971190577", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": [[747, 752], [764, 770], [804, 810]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": ["InstCombinerImpl::foldGEPICmp"]}}, "patch": "commit fac46469977da9c4e9c6eeaac21103c971190577\nAuthor: Yingwei Zheng \nDate: Sat Jan 4 17:23:57 2025 +0800\n\n [InstCombine] Check no wrap flags before folding icmp of GEPs with same indices (#121628)\n \n Alive2: https://alive2.llvm.org/ce/z/Dr3Sbe\n Closes https://github.com/llvm/llvm-project/issues/121581.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\nindex d6fdade25559..8b23583c5106 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n@@ -747,6 +747,8 @@ Instruction *InstCombinerImpl::foldGEPICmp(GEPOperator *GEPLHS, Value *RHS,\n ConstantExpr::getPointerBitCastOrAddrSpaceCast(\n cast(RHS), Base->getType()));\n } else if (GEPOperator *GEPRHS = dyn_cast(RHS)) {\n+ GEPNoWrapFlags NW = GEPLHS->getNoWrapFlags() & GEPRHS->getNoWrapFlags();\n+\n // If the base pointers are different, but the indices are the same, just\n // compare the base pointer.\n if (PtrBase != GEPRHS->getOperand(0)) {\n@@ -764,7 +766,8 @@ Instruction *InstCombinerImpl::foldGEPICmp(GEPOperator *GEPLHS, Value *RHS,\n \n // If all indices are the same, just compare the base pointers.\n Type *BaseType = GEPLHS->getOperand(0)->getType();\n- if (IndicesTheSame && CmpInst::makeCmpResultType(BaseType) == I.getType())\n+ if (IndicesTheSame &&\n+ CmpInst::makeCmpResultType(BaseType) == I.getType() && CanFold(NW))\n return new ICmpInst(Cond, GEPLHS->getOperand(0), GEPRHS->getOperand(0));\n \n // If we're comparing GEPs with two base pointers that only differ in type\n@@ -804,7 +807,6 @@ Instruction *InstCombinerImpl::foldGEPICmp(GEPOperator *GEPLHS, Value *RHS,\n return transformToIndexedCompare(GEPLHS, RHS, Cond, DL, *this);\n }\n \n- GEPNoWrapFlags NW = GEPLHS->getNoWrapFlags() & GEPRHS->getNoWrapFlags();\n if (GEPLHS->getNumOperands() == GEPRHS->getNumOperands() &&\n GEPLHS->getSourceElementType() == GEPRHS->getSourceElementType()) {\n // If the GEPs only differ by one index, compare it.\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/icmp-gep.ll", "commands": ["opt -passes=instcombine -S < %s"], "tests": [{"test_name": "pointer_icmp_aligned_with_offset_negative", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @pointer_icmp_aligned_with_offset_negative(ptr align 8 %a, ptr align 8 %a2) {\n %gep = getelementptr i8, ptr %a, i64 8\n %cmp = icmp eq ptr %gep, %a2\n ret i1 %cmp\n}\n"}, {"test_name": "gep_diff_base_same_indices_nuw_nusw", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @gep_diff_base_same_indices_nuw_nusw(ptr %x, ptr %y, i64 %z) {\n %gep1 = getelementptr nuw i8, ptr %x, i64 %z\n %gep2 = getelementptr nusw i8, ptr %y, i64 %z\n %cmp = icmp ult ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}, {"test_name": "gep_diff_base_same_indices", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @gep_diff_base_same_indices(ptr %x, ptr %y, i64 %z) {\n %gep1 = getelementptr i8, ptr %x, i64 %z\n %gep2 = getelementptr i8, ptr %y, i64 %z\n %cmp = icmp ult ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}, {"test_name": "gep_diff_base_same_indices_nusw", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @gep_diff_base_same_indices_nusw(ptr %x, ptr %y, i64 %z) {\n %gep1 = getelementptr nusw i8, ptr %x, i64 %z\n %gep2 = getelementptr nusw i8, ptr %y, i64 %z\n %cmp = icmp ult ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}, {"test_name": "gep_diff_base_same_indices_nuw", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @gep_diff_base_same_indices_nuw(ptr %x, ptr %y, i64 %z) {\n %gep1 = getelementptr nuw i8, ptr %x, i64 %z\n %gep2 = getelementptr nuw i8, ptr %y, i64 %z\n %cmp = icmp ult ptr %gep1, %gep2\n ret i1 %cmp\n}\n"}]}], "issue": {"title": "[InstCombine] Incorrect fold of the comparison of GEPs", "body": "Alive2: https://alive2.llvm.org/ce/z/HRsqeW\n```\ndefine i1 @test(ptr %x, ptr %y) {\n %gep1 = getelementptr i8, ptr %x, i32 1\n %gep2 = getelementptr i8, ptr %y, i32 1\n %cmp = icmp ult ptr %gep1, %gep2\n ret i1 %cmp\n}\n```\n```\n\n----------------------------------------\ndefine i1 @test(ptr %x, ptr %y) {\n#0:\n %gep1 = gep ptr %x, 1 x i32 1\n %gep2 = gep ptr %y, 1 x i32 1\n %cmp = icmp ult ptr %gep1, %gep2\n ret i1 %cmp\n}\n=>\ndefine i1 @test(ptr %x, ptr %y) {\n#0:\n %cmp = icmp ult ptr %x, %y\n ret i1 %cmp\n}\nTransformation doesn't verify!\n\nERROR: Value mismatch\n\nExample:\nptr %x = pointer(non-local, block_id=0, offset=-1) / Address=#xf\nptr %y = null\n\nSource:\nptr %gep1 = null\nptr %gep2 = pointer(non-local, block_id=0, offset=1) / Address=#x1\ni1 %cmp = #x1 (1)\n\nSOURCE MEMORY STATE\n===================\nNON-LOCAL BLOCKS:\nBlock 0 >\tsize: 0\talign: 1\talloc type: 0\talive: false\taddress: 0\nBlock 1 >\tsize: 1\talign: 2\talloc type: 0\talive: true\taddress: 2\nBlock 2 >\tsize: 0\talign: 1\talloc type: 0\talive: true\taddress: 3\n\nTarget:\ni1 %cmp = #x0 (0)\nSource value: #x1 (1)\nTarget value: #x0 (0)\n\nSummary:\n 0 correct transformations\n 1 incorrect transformations\n 0 failed-to-prove transformations\n 0 Alive2 errors\n```", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "91127", "issue_url": "https://github.com/llvm/llvm-project/issues/91127", "bug_type": "miscompilation", "base_commit": "30367cb5982dfdab2655401f020711311e7d78b9", "knowledge_cutoff": "2024-05-05T15:35:24Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "d3dad7a74be50d90754bf5c072e1a2330e92712c", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": [[1510, 1515]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": ["InstCombinerImpl::foldICmpTruncWithTruncOrExt"]}}, "patch": "commit d3dad7a74be50d90754bf5c072e1a2330e92712c\nAuthor: Yingwei Zheng \nDate: Mon May 6 16:30:07 2024 +0800\n\n [InstCombine] Fix miscompilation caused by #90436 (#91133)\n \n Proof: https://alive2.llvm.org/ce/z/iRnJ4i\n \n Fixes https://github.com/llvm/llvm-project/issues/91127.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\nindex c60a290ce72e..7092fb5e509b 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n@@ -1510,6 +1510,7 @@ InstCombinerImpl::foldICmpTruncWithTruncOrExt(ICmpInst &Cmp,\n std::swap(X, Y);\n Pred = Cmp.getSwappedPredicate(Pred);\n }\n+ YIsSExt = !(NoWrapFlags & TruncInst::NoUnsignedWrap);\n }\n // Try to match icmp (trunc nuw X), (zext Y)\n else if (!Cmp.isSigned() &&\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/icmp-of-trunc-ext.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "trunc_equality_both_sext", "test_body": "define i1 @trunc_equality_both_sext(i32 %x, i8 %y) {\n %xt = trunc nuw nsw i32 %x to i16\n %ye = sext i8 %y to i16\n %c = icmp ne i16 %xt, %ye\n ret i1 %c\n}\n"}, {"test_name": "test_slt_nuw", "test_body": "define i1 @test_slt_nuw(i32 %x, i16 %y) {\n %conv1 = trunc nuw nsw i32 %x to i8\n %conv2 = trunc nuw nsw i16 %y to i8\n %cond = icmp slt i8 %conv1, %conv2\n ret i1 %cond\n}\n"}, {"test_name": "test_slt", "test_body": "define i1 @test_slt(i32 %x, i16 %y) {\n %conv1 = trunc nsw i32 %x to i8\n %conv2 = trunc nsw i16 %y to i8\n %cond = icmp slt i8 %conv1, %conv2\n ret i1 %cond\n}\n"}, {"test_name": "test_ult_nuw", "test_body": "define i1 @test_ult_nuw(i32 %x, i16 %y) {\n %conv1 = trunc nuw nsw i32 %x to i8\n %conv2 = trunc nuw nsw i16 %y to i8\n %cond = icmp ult i8 %conv1, %conv2\n ret i1 %cond\n}\n"}, {"test_name": "test_ult", "test_body": "define i1 @test_ult(i32 %x, i16 %y) {\n %conv1 = trunc nsw i32 %x to i8\n %conv2 = trunc nsw i16 %y to i8\n %cond = icmp ult i8 %conv1, %conv2\n ret i1 %cond\n}\n"}, {"test_name": "test_eq1", "test_body": "define i1 @test_eq1(i32 %x, i16 %y) {\n %conv1 = trunc nsw i32 %x to i8\n %conv2 = trunc nsw i16 %y to i8\n %cond = icmp eq i8 %conv1, %conv2\n ret i1 %cond\n}\n"}, {"test_name": "test_eq2", "test_body": "define i1 @test_eq2(i32 %x, i16 %y) {\n %conv1 = trunc nsw i32 %x to i8\n %conv2 = trunc nsw i16 %y to i8\n %cond = icmp eq i8 %conv2, %conv1\n ret i1 %cond\n}\n"}]}], "issue": {"title": "[InstCombine] Miscompilation which converts `icmp eq (trunc nsw i64 X), (trunc nsw i32 Y)` into `icmp eq (zext i32 Y), X`", "body": "Reduced test case: https://alive2.llvm.org/ce/z/MLgA5x\r\n```\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i32 @src() {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %if.end, %entry\r\n %storemerge = phi i32 [ 3, %entry ], [ %dec, %if.end ]\r\n %h = phi i32 [ 1, %entry ], [ %hres, %if.end ]\r\n %e = phi i32 [ 0, %entry ], [ -1, %if.end ]\r\n %tobool.not = icmp eq i32 %storemerge, 0\r\n br i1 %tobool.not, label %for.end9, label %for.body\r\n\r\nfor.body: ; preds = %for.cond\r\n %cmp = icmp eq i32 %h, 5\r\n %conv1 = zext i1 %cmp to i64\r\n %conv2 = sext i32 %e to i64\r\n %cond.i = call i64 @llvm.umax.i64(i64 %conv1, i64 %conv2)\r\n %conv3 = trunc nsw i64 %cond.i to i16\r\n %conv4 = trunc nsw i32 %e to i16\r\n %sub.i = sub i16 %conv3, %conv4\r\n %tobool6.not = icmp eq i16 %sub.i, 0\r\n br i1 %tobool6.not, label %if.end, label %if.then\r\n\r\nif.then: ; preds = %for.body\r\n br label %if.end\r\n\r\nif.end: ; preds = %if.then, %for.body\r\n %hres = phi i32 [ 0, %if.then ], [ %h, %for.body ]\r\n %dec = add nsw i32 %storemerge, -1\r\n br label %for.cond\r\n\r\nfor.end9: ; preds = %for.cond\r\n ret i32 %h\r\n}\r\n\r\ndefine i32 @tgt() {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %if.end, %entry\r\n %storemerge = phi i32 [ 3, %entry ], [ %dec, %if.end ]\r\n %h = phi i32 [ 1, %entry ], [ %hres, %if.end ]\r\n %e = phi i32 [ 0, %entry ], [ -1, %if.end ]\r\n %tobool.not = icmp eq i32 %storemerge, 0\r\n br i1 %tobool.not, label %for.end9, label %for.body\r\n\r\nfor.body: ; preds = %for.cond\r\n %cmp = icmp eq i32 %h, 5\r\n %conv1 = zext i1 %cmp to i64\r\n %conv2 = sext i32 %e to i64\r\n %cond.i = call i64 @llvm.umax.i64(i64 %conv1, i64 %conv2)\r\n %0 = zext i32 %e to i64\r\n %tobool6.not = icmp eq i64 %cond.i, %0\r\n br i1 %tobool6.not, label %if.end, label %if.then\r\n\r\nif.then: ; preds = %for.body\r\n br label %if.end\r\n\r\nif.end: ; preds = %if.then, %for.body\r\n %hres = phi i32 [ 0, %if.then ], [ %h, %for.body ]\r\n %dec = add nsw i32 %storemerge, -1\r\n br label %for.cond\r\n\r\nfor.end9: ; preds = %for.cond\r\n ret i32 %h\r\n}\r\n```\r\n\r\nllvm version: de9b386f84b58ad0ffc12e221bc6d9161ca5b62d", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "dtcxzyw", "body": "Reduced test case: https://alive2.llvm.org/ce/z/ndR8PQ\r\n```\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i1 @src(i64 %x, i32 %y) {\r\n %conv1 = trunc nsw i64 %x to i16\r\n %conv2 = trunc nsw i32 %y to i16\r\n %cond = icmp eq i16 %conv1, %conv2\r\n ret i1 %cond\r\n}\r\n\r\ndefine i1 @tgt(i64 %x, i32 %y) {\r\n %1 = zext i32 %y to i64\r\n %cond = icmp eq i64 %1, %x\r\n ret i1 %cond\r\n}\r\n```\r\n\r\nIt should be caused by https://github.com/llvm/llvm-project/pull/90436.\r\n\r\nI will post a fix later. cc @nikic\r\n"}, {"author": "dtcxzyw", "body": "POC without DataLayout:\r\n```\r\ndefine i1 @test_eq(i32 %x, i16 %y) {\r\n %conv1 = trunc nsw i32 %x to i8\r\n %conv2 = trunc nsw i16 %y to i8\r\n %cond = icmp eq i8 %conv1, %conv2\r\n ret i1 %cond\r\n}\r\n\r\ndefine i1 @test_ult(i32 %x, i16 %y) {\r\n %conv1 = trunc nsw i32 %x to i8\r\n %conv2 = trunc nsw i16 %y to i8\r\n %cond = icmp ult i8 %conv1, %conv2\r\n ret i1 %cond\r\n}\r\n\r\ndefine i1 @test_slt(i32 %x, i16 %y) {\r\n %conv1 = trunc nsw i32 %x to i8\r\n %conv2 = trunc nsw i16 %y to i8\r\n %cond = icmp slt i8 %conv1, %conv2\r\n ret i1 %cond\r\n}\r\n```"}]}, "verified": true} {"bug_id": "77553", "issue_url": "https://github.com/llvm/llvm-project/issues/77553", "bug_type": "crash", "base_commit": "c9c8f0c2fcf3b25ec310a75216f1d5b582ec343f", "knowledge_cutoff": "2024-01-10T04:18:48Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "66eedd1dd370d22ddf994540c20848618d64d1a6", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[1704, 1714]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["InstCombinerImpl::foldSelectInstWithICmp"]}}, "patch": "commit 66eedd1dd370d22ddf994540c20848618d64d1a6\nAuthor: hanbeom \nDate: Thu Jan 11 17:34:30 2024 +0900\n\n [InstCombine] Fix worklist management in select fold (#77738)\n \n `InstCombine` uses `Worklist` to manage change history. `setOperand`,\n which was previously used to change the `Select` Instruction, does not,\n so it is `run` twice, which causes an `LLVM ERROR`.\n \n This problem is resolved by changing `setOperand` to `replaceOperand` as\n the change history will be registered in the Worklist.\n \n Fixes #77553.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex ab55f235920a..21bfc91148bf 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -1704,11 +1704,11 @@ Instruction *InstCombinerImpl::foldSelectInstWithICmp(SelectInst &SI,\n if (CmpRHS != CmpLHS && isa(CmpRHS) && !isa(CmpLHS)) {\n if (CmpLHS == TrueVal && Pred == ICmpInst::ICMP_EQ) {\n // Transform (X == C) ? X : Y -> (X == C) ? C : Y\n- SI.setOperand(1, CmpRHS);\n+ replaceOperand(SI, 1, CmpRHS);\n Changed = true;\n } else if (CmpLHS == FalseVal && Pred == ICmpInst::ICMP_NE) {\n // Transform (X != C) ? Y : X -> (X != C) ? Y : C\n- SI.setOperand(2, CmpRHS);\n+ replaceOperand(SI, 2, CmpRHS);\n Changed = true;\n }\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/select.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "src_select_xxory_eq0_xorxy_y", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine i32 @src_select_xxory_eq0_xorxy_y(i32 %x, i32 %y) {\n %xor = xor i32 %x, %y\n %xor0 = icmp eq i32 %xor, 0\n %cond = select i1 %xor0, i32 %xor, i32 %y\n ret i32 %cond\n}\n"}]}], "issue": {"title": "LLVM ERROR while InstCombine", "body": "```\r\n$ bllvm0/bin/clang --version\r\nclang version 18.0.0git (git@github.com:ParkHanbum/llvm-project.git a9f39ff2b628e38826d5b95c1e8ae3cb7c692de9)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /home/m/bllvm0/bin\r\n```\r\nmy clang version.\r\n\r\nIR that caused the error\r\n```\r\ndefine i32 @no_trans_select_xor_eq0_xor_and(i32 %x, i32 %y) {\r\n; CHECK-LABEL: @no_trans_select_xor_eq0_xor_and(\r\n; CHECK-NEXT: [[XOR:%.*]] = xor i32 [[X:%.*]], [[Y:%.*]]\r\n; CHECK-NEXT: [[XOR0:%.*]] = icmp eq i32 [[XOR]], 0\r\n; CHECK-NEXT: [[AND:%.*]] = and i32 [[X]], [[Y]]\r\n; CHECK-NEXT: [[COND:%.*]] = select i1 [[XOR0]], i32 [[XOR]], i32 [[AND]]\r\n; CHECK-NEXT: ret i32 [[COND]]\r\n;\r\n %xor = xor i32 %x, %y\r\n %xor0 = icmp eq i32 %xor, 0\r\n %and = and i32 %x, %y\r\n %cond = select i1 %xor0, i32 %xor, i32 %and\r\n ret i32 %cond\r\n}\r\n```\r\n\r\nand Debug message :\r\n```\r\n$ bllvm0/bin/opt -passes=instcombine temp.ll -debug\r\nINSTCOMBINE ITERATION #1 on no_trans_select_xor_eq0_xor_and\r\nADD: ret i32 %cond\r\nADD: %cond = select i1 %xor0, i32 %xor, i32 %and\r\nADD: %and = and i32 %x, %y\r\nADD: %xor0 = icmp eq i32 %xor, 0\r\nADD: %xor = xor i32 %x, %y\r\nIC: Visiting: %xor = xor i32 %x, %y\r\nIC: Visiting: %xor0 = icmp eq i32 %xor, 0\r\nIC: Visiting: %and = and i32 %x, %y\r\nIC: Visiting: %cond = select i1 %xor0, i32 %xor, i32 %and\r\nIC: Mod = %cond = select i1 %xor0, i32 %xor, i32 %and\r\n New = %cond = select i1 %xor0, i32 0, i32 %and\r\nADD: %cond = select i1 %xor0, i32 0, i32 %and\r\nIC: Visiting: %cond = select i1 %xor0, i32 0, i32 %and\r\nIC: Visiting: ret i32 %cond\r\n\r\n\r\nINSTCOMBINE ITERATION #2 on no_trans_select_xor_eq0_xor_and\r\nADD: ret i32 %cond\r\nADD: %cond = select i1 %xor0, i32 0, i32 %and\r\nADD: %and = and i32 %x, %y\r\nADD: %xor0 = icmp eq i32 %xor, 0\r\nADD: %xor = xor i32 %x, %y\r\nIC: Visiting: %xor = xor i32 %x, %y\r\nIC: Visiting: %xor0 = icmp eq i32 %xor, 0\r\nIC: Old = %xor0 = icmp eq i32 %xor, 0\r\n New = = icmp eq i32 %x, %y\r\nADD: %xor0 = icmp eq i32 %x, %y\r\nIC: ERASE %1 = icmp eq i32 %xor, 0\r\nADD DEFERRED: %xor = xor i32 %x, %y\r\nIC: ERASE %xor = xor i32 %x, %y\r\nIC: Visiting: %xor0 = icmp eq i32 %x, %y\r\nIC: Visiting: %and = and i32 %x, %y\r\nIC: Visiting: %cond = select i1 %xor0, i32 0, i32 %and\r\nIC: Visiting: ret i32 %cond\r\nLLVM ERROR: Instruction Combining did not reach a fixpoint after 1 iterations\r\n\r\n```\r\n\r\nIf time isn't tight, I'd like to take on this issue.\r\n\r\n", "author": "ParkHanbum", "labels": ["llvm:instcombine", "crash-on-valid"], "comments": [{"author": "nikic", "body": "The problem is likely that https://github.com/llvm/llvm-project/blob/084f1c2ee074a5ac8186ea4b5b181b48bf4621b6/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp#L1704-L1714 calls setOperand() instead of replaceOperand()."}, {"author": "ParkHanbum", "body": "@nikic can I try this?"}, {"author": "nikic", "body": "Sure."}]}, "verified": true} {"bug_id": "115465", "issue_url": "https://github.com/llvm/llvm-project/issues/115465", "bug_type": "miscompilation", "base_commit": "d822c099eeacc69f6bf834a6373a41d0c9f84a3e", "knowledge_cutoff": "2024-11-08T11:25:34Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "27bf45aa36386136db179c494358670a994a98a5", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp": [[2904, 2910]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp": ["InstCombinerImpl::visitShuffleVectorInst"]}}, "patch": "commit 27bf45aa36386136db179c494358670a994a98a5\nAuthor: Yingwei Zheng \nDate: Sun Nov 10 17:07:25 2024 +0800\n\n [InstCombine] Fix poison safety of folding shufflevector into select (#115483)\n \n We are allowed to fold shufflevector into select iff the condition is\n guaranteed not to be poison or the RHS is a poison.\n Alive2: https://alive2.llvm.org/ce/z/28zEWR\n \n Closes https://github.com/llvm/llvm-project/issues/115465.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\nindex 454fe5a91d37..ede89b099e8d 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp\n@@ -2904,7 +2904,9 @@ Instruction *InstCombinerImpl::visitShuffleVectorInst(ShuffleVectorInst &SVI) {\n if (auto *SI = dyn_cast(LHS)) {\n // We cannot do this fold for elementwise select since ShuffleVector is\n // not elementwise.\n- if (SI->getCondition()->getType()->isIntegerTy()) {\n+ if (SI->getCondition()->getType()->isIntegerTy() &&\n+ (isa(RHS) ||\n+ isGuaranteedNotToBePoison(SI->getCondition()))) {\n if (Instruction *I = FoldOpIntoSelect(SVI, SI))\n return I;\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/vec_shuffle.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "shuf_cmp_may_be_poison", "test_body": "define <4 x i8> @shuf_cmp_may_be_poison(<4 x i8> %x, <4 x i8> %y, i1 %cmp) {\n %sel = select i1 %cmp, <4 x i8> %y, <4 x i8> \n %shuf = shufflevector <4 x i8> %sel, <4 x i8> , <4 x i32> \n ret <4 x i8> %shuf\n}\n"}, {"test_name": "shuf_same_length_vec_select", "test_body": "define <4 x i32> @shuf_same_length_vec_select(<4 x i1> %cond) {\n %sel = select <4 x i1> %cond, <4 x i32> , <4 x i32> \n %shuf = shufflevector <4 x i32> %sel, <4 x i32> , <4 x i32> \n ret <4 x i32> %shuf\n}\n"}]}], "issue": {"title": "[InstCombine] folded select produces different result on a blended vector", "body": "Alive2 report: https://alive2.llvm.org/ce/z/73v-ug\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine <4 x i8> @sel_shuf_commute3.2(<4 x i8> %x, <4 x i8> %y, i1 %cmp) {\r\n#0:\r\n %blend = shufflevector <4 x i8> { 0, 0, 0, 0 }, <4 x i8> %y, 0, 5, 2, 3\r\n %r = select i1 %cmp, <4 x i8> %y, <4 x i8> %blend\r\n %#1 = select <4 x i1> { 0, 1, 0, 1 }, <4 x i8> { 0, 1, 2, 3 }, <4 x i8> %r\r\n ret <4 x i8> %#1\r\n}\r\n=>\r\ndefine <4 x i8> @sel_shuf_commute3.2(<4 x i8> %x, <4 x i8> %y, i1 %cmp) {\r\n#0:\r\n %#1 = shufflevector <4 x i8> %y, <4 x i8> { poison, 1, poison, 3 }, 0, 5, 2, 7\r\n %#2 = select i1 %cmp, <4 x i8> %#1, <4 x i8> { 0, 1, 0, 3 }\r\n ret <4 x i8> %#2\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\n<4 x i8> %x = < poison, poison, poison, poison >\r\n<4 x i8> %y = < poison, poison, poison, poison >\r\ni1 %cmp = poison\r\n\r\nSource:\r\n<4 x i8> %blend = < #x00 (0), poison, #x00 (0), #x00 (0) >\r\n<4 x i8> %r = < poison, poison, poison, poison >\r\n<4 x i8> %#1 = < poison, #x01 (1), poison, #x03 (3) >\r\n\r\nTarget:\r\n<4 x i8> %#1 = < poison, #x01 (1), poison, #x03 (3) >\r\n<4 x i8> %#2 = < poison, poison, poison, poison >\r\nSource value: < poison, #x01 (1), poison, #x03 (3) >\r\nTarget value: < poison, poison, poison, poison >\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```\r\n", "author": "bongjunj", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "87410", "issue_url": "https://github.com/llvm/llvm-project/issues/87410", "bug_type": "crash", "base_commit": "0bbe953aa3289a32cd816647820c8676bb3a61bc", "knowledge_cutoff": "2024-04-02T21:05:28Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "c83698367125703827f1b739393f006c399cb213", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[8233, 8238]], "llvm/lib/Transforms/Vectorize/VPlan.h": [[1932, 1945], [1949, 1963]], "llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[1513, 1518]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["VPRecipeBuilder::tryToBlend"], "llvm/lib/Transforms/Vectorize/VPlan.h": ["VPBlendRecipe", "clone", "getIncomingValue", "getMask", "getNumIncomingValues"], "llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPBlendRecipe::print"]}}, "patch": "commit c83698367125703827f1b739393f006c399cb213\nAuthor: Florian Hahn \nDate: Tue Apr 9 11:14:05 2024 +0100\n\n [VPlan] Remove unused first mask op from VPBlendRecipe. (#87770)\n \n VPBlendRecipe does not use the first mask operand. Removing it allows\n VPlan-based DCE to remove unused mask computations.\n \n This also fixes #87410, where unused Not VPInstructions are considered\n having only their first lane demanded, but some of their operands\n providing a vector value due to other users.\n \n Fixes https://github.com/llvm/llvm-project/issues/87410\n \n PR: https://github.com/llvm/llvm-project/pull/87770\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex fd54faf17ca3..5535cc55e932 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -8221,6 +8221,8 @@ VPBlendRecipe *VPRecipeBuilder::tryToBlend(PHINode *Phi,\n // builder. At this point we generate the predication tree. There may be\n // duplications since this is a simple recursive scan, but future\n // optimizations will clean it up.\n+ // TODO: At the moment the first mask is always skipped, but it would be\n+ // better to skip the most expensive mask.\n SmallVector OperandsWithMask;\n \n for (unsigned In = 0; In < NumIncoming; In++) {\n@@ -8233,6 +8235,8 @@ VPBlendRecipe *VPRecipeBuilder::tryToBlend(PHINode *Phi,\n \"Distinct incoming values with one having a full mask\");\n break;\n }\n+ if (In == 0)\n+ continue;\n OperandsWithMask.push_back(EdgeMask);\n }\n return new VPBlendRecipe(Phi, OperandsWithMask);\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h\nindex 5f6334b974cf..5dc905a3c407 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlan.h\n+++ b/llvm/lib/Transforms/Vectorize/VPlan.h\n@@ -1932,14 +1932,12 @@ public:\n class VPBlendRecipe : public VPSingleDefRecipe {\n public:\n /// The blend operation is a User of the incoming values and of their\n- /// respective masks, ordered [I0, M0, I1, M1, ...]. Note that a single value\n- /// might be incoming with a full mask for which there is no VPValue.\n+ /// respective masks, ordered [I0, I1, M1, I2, M2, ...]. Note that the first\n+ /// incoming value does not have a mask associated.\n VPBlendRecipe(PHINode *Phi, ArrayRef Operands)\n : VPSingleDefRecipe(VPDef::VPBlendSC, Operands, Phi, Phi->getDebugLoc()) {\n- assert(Operands.size() > 0 &&\n- ((Operands.size() == 1) || (Operands.size() % 2 == 0)) &&\n- \"Expected either a single incoming value or a positive even number \"\n- \"of operands\");\n+ assert((Operands.size() + 1) % 2 == 0 &&\n+ \"Expected an odd number of operands\");\n }\n \n VPRecipeBase *clone() override {\n@@ -1949,15 +1947,20 @@ public:\n \n VP_CLASSOF_IMPL(VPDef::VPBlendSC)\n \n- /// Return the number of incoming values, taking into account that a single\n+ /// Return the number of incoming values, taking into account that the first\n /// incoming value has no mask.\n unsigned getNumIncomingValues() const { return (getNumOperands() + 1) / 2; }\n \n /// Return incoming value number \\p Idx.\n- VPValue *getIncomingValue(unsigned Idx) const { return getOperand(Idx * 2); }\n+ VPValue *getIncomingValue(unsigned Idx) const {\n+ return Idx == 0 ? getOperand(0) : getOperand(Idx * 2 - 1);\n+ }\n \n /// Return mask number \\p Idx.\n- VPValue *getMask(unsigned Idx) const { return getOperand(Idx * 2 + 1); }\n+ VPValue *getMask(unsigned Idx) const {\n+ assert(Idx > 0 && \"First index has no mask associated.\");\n+ return getOperand(Idx * 2);\n+ }\n \n /// Generate the phi/select nodes.\n void execute(VPTransformState &State) override;\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex 2438e4dae3eb..625319954e9b 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -1513,6 +1513,8 @@ void VPBlendRecipe::print(raw_ostream &O, const Twine &Indent,\n for (unsigned I = 0, E = getNumIncomingValues(); I < E; ++I) {\n O << \" \";\n getIncomingValue(I)->printAsOperand(O, SlotTracker);\n+ if (I == 0)\n+ continue;\n O << \"/\";\n getMask(I)->printAsOperand(O, SlotTracker);\n }\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "test_not_first_lane_only_constant", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @test_not_first_lane_only_constant(ptr %A, ptr noalias %B) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i16 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %gep.A = getelementptr inbounds i16, ptr %A, i16 %iv\n br i1 false, label %loop.latch, label %else.1\n\nelse.1: ; preds = %loop.header\n br i1 false, label %then.2, label %else.2\n\nthen.2: ; preds = %else.1\n br label %else.2\n\nelse.2: ; preds = %then.2, %else.1\n br label %loop.latch\n\nloop.latch: ; preds = %else.2, %loop.header\n %merge = phi ptr [ %B, %else.2 ], [ poison, %loop.header ]\n %l = load i16, ptr %merge, align 2\n %iv.next = add i16 %iv, 1\n store i16 %l, ptr %gep.A, align 2\n %c.2 = icmp eq i16 %iv.next, 1000\n br i1 %c.2, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}, {"test_name": "test_not_first_lane_only_wide_compare", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @test_not_first_lane_only_wide_compare(ptr %A, ptr noalias %B, i16 %x, i16 %y) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i16 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %gep.A = getelementptr inbounds i16, ptr %A, i16 %iv\n %l.0 = load i16, ptr %gep.A, align 2\n %c.0 = icmp ult i16 %l.0, %x\n br i1 %c.0, label %loop.latch, label %else.1\n\nelse.1: ; preds = %loop.header\n %c.1 = icmp ult i16 %l.0, %y\n br i1 %c.1, label %then.2, label %else.2\n\nthen.2: ; preds = %else.1\n br label %else.2\n\nelse.2: ; preds = %then.2, %else.1\n br label %loop.latch\n\nloop.latch: ; preds = %else.2, %loop.header\n %merge = phi ptr [ %B, %else.2 ], [ poison, %loop.header ]\n %l = load i16, ptr %merge, align 2\n %iv.next = add i16 %iv, 1\n store i16 %l, ptr %gep.A, align 2\n %c.2 = icmp eq i16 %iv.next, 1000\n br i1 %c.2, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}], "issue": {"title": "[LoopVectorize][VPlan] Assertion \"Trying to access a single scalar per part but has multiple scalars per part.\" failed.", "body": "Reduced LLVM IR\r\n```llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main() #0 {\r\nentry:\r\n br label %for.body.i\r\n\r\nfor.body.i: ; preds = %if.end.i, %entry\r\n %q.0139.i = phi i16 [ 0, %entry ], [ %add.i, %if.end.i ]\r\n %idxprom.i = sext i16 0 to i64\r\n br i1 false, label %if.end.i, label %if.then.i\r\n\r\nif.then.i: ; preds = %for.body.i\r\n br i1 false, label %cond.end41.i, label %cond.end41.i\r\n\r\ncond.end41.i: ; preds = %if.then.i, %if.then.i\r\n br label %if.end.i\r\n\r\nif.end.i: ; preds = %cond.end41.i, %for.body.i\r\n %.in.i = phi ptr [ null, %cond.end41.i ], [ poison, %for.body.i ]\r\n %0 = load i16, ptr %.in.i, align 2\r\n %add.i = add i16 %q.0139.i, 1\r\n %tobool.not.i = icmp eq i16 %add.i, 0\r\n br i1 %tobool.not.i, label %m.exit, label %for.body.i\r\n\r\nm.exit: ; preds = %if.end.i\r\n %.lcssa = phi i16 [ %0, %if.end.i ]\r\n ret i32 0\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\r\n```\r\n\r\nBacktrace:\r\n```\r\nopt: /scratch/tc-testing/tc-apr-2/llvm/llvm/lib/Transforms/Vectorize/VPlan.cpp:249: llvm::Value* llvm::VPTransformState::get(llvm::VPValue*, unsigned int, bool): Assertion `(VF.isScalar() || Def->isLiveIn() || (hasScalarValue(Def, VPIteration(Part, 0)) && Data.PerPartScalars[Def][Part].size() == 1)) && \"Trying to access a single scalar per part but has multiple scalars \" \"per part.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt --passes=loop-vectorize reduced.ll\r\n #0 0x00005ed60f56eb60 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2d84b60)\r\n #1 0x00005ed60f56bf6f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2d81f6f)\r\n #2 0x00005ed60f56c0c5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007f973de42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007f973de969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007f973de969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007f973de969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007f973de42476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007f973de287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007f973de2871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007f973de39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00005ed60e6cbb72 llvm::VPTransformState::get(llvm::VPValue*, unsigned int, bool) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1ee1b72)\r\n#12 0x00005ed60e6fa9a6 llvm::VPInstruction::generatePerPart(llvm::VPTransformState&, unsigned int) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1f109a6)\r\n#13 0x00005ed60e6fbbdf llvm::VPInstruction::execute(llvm::VPTransformState&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1f11bdf)\r\n#14 0x00005ed60e6d74ff llvm::VPBasicBlock::execute(llvm::VPTransformState*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1eed4ff)\r\n#15 0x00005ed60e6cd9e0 llvm::VPRegionBlock::execute(llvm::VPTransformState*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1ee39e0)\r\n#16 0x00005ed60e6d90e6 llvm::VPlan::execute(llvm::VPTransformState*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1eef0e6)\r\n#17 0x00005ed60e5d76f3 llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1ded6f3)\r\n#18 0x00005ed60e5e9718 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1dff718)\r\n#19 0x00005ed60e5eb73e llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1e0173e)\r\n#20 0x00005ed60e5ec88d llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1e0288d)\r\n#21 0x00005ed60d4f3276 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd09276)\r\n#22 0x00005ed60f395141 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2bab141)\r\n#23 0x00005ed60d4eabd6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd00bd6)\r\n#24 0x00005ed60f393e3b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2ba9e3b)\r\n#25 0x00005ed60d4f2dc6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd08dc6)\r\n#26 0x00005ed60f391cb1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2ba7cb1)\r\n#27 0x00005ed60cd6cce5 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x582ce5)\r\n#28 0x00005ed60cd5f316 optMain (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x575316)\r\n#29 0x00007f973de29d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#30 0x00007f973de29e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#31 0x00007f973de29e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#32 0x00005ed60cd54f95 _start (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x56af95)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt\r\n```\r\nGodbolt: https://godbolt.org/z/jYWea8z4q\r\n\r\nFound via fuzzer.", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "88849", "issue_url": "https://github.com/llvm/llvm-project/issues/88849", "bug_type": "crash", "base_commit": "6b98ab95f0d36705b5a1fc1e755c992ba2329c89", "knowledge_cutoff": "2024-04-16T07:36:13Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "b8741cc185e89a95c9161218084f7c8b5578cf96", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlan.cpp": [[254, 259]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlan.cpp": ["VPTransformState::get"]}}, "patch": "commit b8741cc185e89a95c9161218084f7c8b5578cf96\nAuthor: Florian Hahn \nDate: Fri Jul 19 10:33:07 2024 +0100\n\n [VPlan] Relax assertion retrieving a scalar from VPTransformState::get.\n \n The current assertion VPTransformState::get when retrieving a single\n scalar only does not account for cases where a def has multiple users,\n some demanding all scalar lanes, some demanding only a single scalar.\n \n For an example, see the modified test case. Relax the assertion by also\n allowing requesting scalar lanes only when the Def doesn't have only its\n first lane used.\n \n Fixes https://github.com/llvm/llvm-project/issues/88849.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp\nindex 83a035fb4df8..58de6256900f 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp\n@@ -254,6 +254,7 @@ Value *VPTransformState::get(VPValue *Def, const VPIteration &Instance) {\n Value *VPTransformState::get(VPValue *Def, unsigned Part, bool NeedsScalar) {\n if (NeedsScalar) {\n assert((VF.isScalar() || Def->isLiveIn() || hasVectorValue(Def, Part) ||\n+ !vputils::onlyFirstLaneUsed(Def) ||\n (hasScalarValue(Def, VPIteration(Part, 0)) &&\n Data.PerPartScalars[Def][Part].size() == 1)) &&\n \"Trying to access a single scalar per part but has multiple scalars \"\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/SystemZ/scalar-steps-with-users-demanding-all-lanes-and-first-lane-only.ll", "commands": ["opt -p loop-vectorize -mtriple=s390x-unknown-linux -mcpu=z16 -S %s"], "tests": [{"test_name": "test_scalar_iv_steps_used_by_replicate_and_first_lane_only_vpinst", "test_body": "target datalayout = \"E-m:e-i1:8:16-i8:8:16-i64:64-f128:64-v128:64-a:8:16-n32:64\"\n\n@src = external global [8 x i32], align 4\n\ndefine void @test_scalar_iv_steps_used_by_replicate_and_first_lane_only_vpinst(ptr noalias %dst, ptr noalias %src.1) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %mul.iv = mul nsw i64 %iv, 4\n %gep.src.1 = getelementptr inbounds i8, ptr %src.1, i64 %mul.iv\n %l.1 = load i8, ptr %gep.src.1, align 1\n %c = icmp eq i8 %l.1, 0\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n %iv.or = or disjoint i64 %iv, 4\n %gep.src = getelementptr inbounds [8 x i32], ptr @src, i64 0, i64 %iv.or\n %l.2 = load i32, ptr %gep.src, align 4\n store i32 %l.2, ptr %dst, align 4\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %iv.next = add nuw nsw i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 4\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}], "issue": {"title": "LoopVectorizer asserts: Assertion `(VF.isScalar() || Def->isLiveIn() || (hasScalarValue(Def, VPIteration(Part, 0)) && Data.PerPartScalars[Def][Part].size() == 1)) && \"Trying to access a single scalar per part but has multiple scalars \" \"per part.", "body": "[tc_crash35_aftercreduce.ll.tar.gz](https://github.com/llvm/llvm-project/files/14988465/tc_crash35_aftercreduce.ll.tar.gz)\r\n\r\nopt -mtriple=s390x-unknown-linux -mcpu=z16 -O3 tc_crash35_aftercreduce.ll -o /dev/null\r\n\r\nVPlan.cpp:253: llvm::Value* llvm::VPTransformState::get(llvm::VPValue*, unsigned int, bool): Assertion `(VF.isScalar() || Def->isLiveIn() || (hasScalarValue(Def, VPIteration(Part, 0)) && Data.PerPartScalars[Def][Part].size() == 1)) && \"Trying to access a single scalar per part but has multiple scalars \" \"per part.\"' failed.\r\n\r\n@fhahn ", "author": "JonPsson1", "labels": ["vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "Should be fixed now, sorry for the delay @JonPsson1 !"}]}, "verified": true} {"bug_id": "95016", "issue_url": "https://github.com/llvm/llvm-project/issues/95016", "bug_type": "crash", "base_commit": "1846523bb77275de954ac573110171bd39bfa930", "knowledge_cutoff": "2024-06-10T18:08:58Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "380beaec8633bad0148aec02f03a85d9a59b2a2d", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[835, 845]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["getSameOpcode"]}}, "patch": "commit 380beaec8633bad0148aec02f03a85d9a59b2a2d\nAuthor: Gabriel Baraldi \nDate: Tue Jul 2 09:15:51 2024 -0300\n\n Fix potential crash in SLPVectorizer caused by missing check (#95937)\n \n I'm not super familiar with this code, but it seems that we were just\n missing a check.\n \n The original code that triggered this did not have uselistorders but\n llvm-reduce created them and it reproduces the same issue in a way more\n compact way.\n \n Fixes https://github.com/llvm/llvm-project/issues/95016\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex c4876bb92a1c..868e9e2687f5 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -835,11 +835,11 @@ static InstructionsState getSameOpcode(ArrayRef VL,\n auto *CallBase = cast(IBase);\n if (Call->getCalledFunction() != CallBase->getCalledFunction())\n return InstructionsState(VL[BaseIndex], nullptr, nullptr);\n- if (Call->hasOperandBundles() &&\n+ if (Call->hasOperandBundles() && (!CallBase->hasOperandBundles() ||\n !std::equal(Call->op_begin() + Call->getBundleOperandsStartIndex(),\n Call->op_begin() + Call->getBundleOperandsEndIndex(),\n CallBase->op_begin() +\n- CallBase->getBundleOperandsStartIndex()))\n+ CallBase->getBundleOperandsStartIndex())))\n return InstructionsState(VL[BaseIndex], nullptr, nullptr);\n Intrinsic::ID ID = getVectorIntrinsicIDForCall(Call, &TLI);\n if (ID != BaseID)\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/AArch64/uselistorder.ll", "commands": ["opt < %s -passes=slp-vectorizer -S -pass-remarks-missed=slp-vectorizer 2>&1"], "tests": [{"test_name": "", "test_body": "\ntarget datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64-unknown-linux-gnu\"\n\n; This test has UB but the crash in #95016 only happens with it\ndefine void @uselistorder_test() {\n;\n %max1 = call double @llvm.maximum.f64(double 0.000000e+00, double 0.000000e+00) [ \"a_list\"(ptr null) ]\n %add1 = fadd double %max1, 0.000000e+00\n %mul1 = fmul double 0.000000e+00, %add1\n %mul2 = fmul double %mul1, 0.000000e+00\n %sel1 = select i1 false, double 0.000000e+00, double %mul2\n %max2 = call double @llvm.maximum.f64(double 0.000000e+00, double 0.000000e+00)\n %add2 = fadd double %max2, 0.000000e+00\n %mul3 = fmul double 0.000000e+00, %add2\n %mul4 = fmul double %mul3, 0.000000e+00\n %sel2 = select i1 false, double 0.000000e+00, double %mul4\n %mul5 = fmul double %sel2, 0.000000e+00\n %add3 = fadd double 0.000000e+00, %mul5\n %gep1 = getelementptr { double, [1 x [2 x double]] }, ptr null, i64 0, i32 1\n store double %add3, ptr %gep1, align 8\n %mul6 = fmul double %sel1, 0.000000e+00\n %add4 = fadd double %mul6, 0.000000e+00\n store double %add4, ptr null, align 8\n ret void\n}\n\ndeclare double @llvm.maximum.f64(double, double) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }"}]}], "issue": {"title": "SLPVectorizer (hasOperandBundles() && \"Don't call otherwise!\") assertion trigger with verified module", "body": "SLPVectorize asserts while optimizing the following module\r\n```llvm-ir\r\ntarget datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128-ni:10:11:12:13\"\r\ntarget triple = \"arm64-apple-darwin23.5.0\"\r\n\r\ndefine swiftcc void @julia_multiparameter_a_res_17823() {\r\ntop:\r\n %0 = call double @llvm.maximum.f64(double 0.000000e+00, double 0.000000e+00) [ \"jl_roots\"(ptr null) ]\r\n %1 = fadd double %0, 0.000000e+00\r\n %2 = fmul double 0.000000e+00, %1\r\n %3 = fmul double %2, 0.000000e+00\r\n %4 = select i1 false, double 0.000000e+00, double %3\r\n %5 = call double @llvm.maximum.f64(double 0.000000e+00, double 0.000000e+00)\r\n %6 = fadd double %5, 0.000000e+00\r\n %7 = fmul double 0.000000e+00, %6\r\n %8 = fmul double %7, 0.000000e+00\r\n %9 = select i1 false, double 0.000000e+00, double %8\r\n %10 = fmul double %9, 0.000000e+00\r\n %11 = fadd double 0.000000e+00, %10\r\n %12 = getelementptr { double, [1 x [2 x double]] }, ptr null, i64 0, i32 1\r\n store double %11, ptr %12, align 8\r\n %13 = fmul double %4, 0.000000e+00\r\n %14 = fadd double %13, 0.000000e+00\r\n store double %14, ptr null, align 8\r\n ret void\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare double @llvm.maximum.f64(double, double) #0\r\n\r\n; uselistorder directives\r\nuselistorder ptr @llvm.maximum.f64, { 1, 0 }\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\n\r\nThis also seems to trigger on godbolt on whatever version of trunk they have there\r\nhttps://godbolt.org/z/favErWG5K\r\n\r\n\r\nstacktrace:\r\n> #0 0x0000000004e258c8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e258c8)\r\n #1 0x0000000004e2301c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007c7148042520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007c71480969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007c7148042476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007c71480287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007c714802871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007c7148039e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000003ee5e8f (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3ee5e8f)\r\n #9 0x0000000003f5e546 getSameOpcode(llvm::ArrayRef, llvm::TargetLibraryInfo const&, unsigned int) (.constprop.1) SLPVectorizer.cpp:0:0\r\n#10 0x0000000003f67928 llvm::slpvectorizer::BoUpSLP::LookAheadHeuristics::getShallowScore(llvm::Value*, llvm::Value*, llvm::Instruction*, llvm::Instruction*, llvm::ArrayRef) const (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f67928)\r\n#11 0x0000000003f67de0 llvm::slpvectorizer::BoUpSLP::LookAheadHeuristics::getScoreAtLevelRec(llvm::Value*, llvm::Value*, llvm::Instruction*, llvm::Instruction*, int, llvm::ArrayRef) const (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f67de0)\r\n#12 0x0000000003f6aaf5 llvm::slpvectorizer::BoUpSLP::VLOperands::getBestOperand(unsigned int, int, int, llvm::ArrayRef, llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f6aaf5)\r\n#13 0x0000000003f6bfcb llvm::slpvectorizer::BoUpSLP::VLOperands::reorder() (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f6bfcb)\r\n#14 0x0000000003f6c4f6 llvm::slpvectorizer::BoUpSLP::reorderInputsAccordingToOpcode(llvm::ArrayRef, llvm::SmallVectorImpl&, llvm::SmallVectorImpl&, llvm::slpvectorizer::BoUpSLP const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f6c4f6)\r\n#15 0x0000000003f702f4 llvm::slpvectorizer::BoUpSLP::buildTree_rec(llvm::ArrayRef, unsigned int, llvm::slpvectorizer::BoUpSLP::EdgeInfo const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f702f4)\r\n#16 0x0000000003f70392 llvm::slpvectorizer::BoUpSLP::buildTree_rec(llvm::ArrayRef, unsigned int, llvm::slpvectorizer::BoUpSLP::EdgeInfo const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f70392)\r\n#17 0x0000000003f70362 llvm::slpvectorizer::BoUpSLP::buildTree_rec(llvm::ArrayRef, unsigned int, llvm::slpvectorizer::BoUpSLP::EdgeInfo const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f70362)\r\n#18 0x0000000003f7022b llvm::slpvectorizer::BoUpSLP::buildTree_rec(llvm::ArrayRef, unsigned int, llvm::slpvectorizer::BoUpSLP::EdgeInfo const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f7022b)\r\n#19 0x0000000003f70362 llvm::slpvectorizer::BoUpSLP::buildTree_rec(llvm::ArrayRef, unsigned int, llvm::slpvectorizer::BoUpSLP::EdgeInfo const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f70362)\r\n#20 0x0000000003f70362 llvm::slpvectorizer::BoUpSLP::buildTree_rec(llvm::ArrayRef, unsigned int, llvm::slpvectorizer::BoUpSLP::EdgeInfo const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f70362)\r\n#21 0x0000000003f6fe47 llvm::slpvectorizer::BoUpSLP::buildTree_rec(llvm::ArrayRef, unsigned int, llvm::slpvectorizer::BoUpSLP::EdgeInfo const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f6fe47)\r\n#22 0x0000000003f71b59 llvm::slpvectorizer::BoUpSLP::buildTree(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f71b59)\r\n#23 0x0000000003f94622 llvm::SLPVectorizerPass::vectorizeStoreChain(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, unsigned int, unsigned int, unsigned int&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f94622)\r\n#24 0x0000000003f95fb0 llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::'lambda'(std::set, llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::StoreDistCompare, std::allocator>> const&)::operator()(std::set, llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::StoreDistCompare, std::allocator>> const&) const SLPVectorizer.cpp:0:0\r\n#25 0x0000000003f97e88 llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f97e88)\r\n#26 0x0000000003f985b3 llvm::SLPVectorizerPass::vectorizeStoreChains(llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f985b3)\r\n#27 0x0000000003f9970f llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#28 0x0000000003f9a163 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f9a163)\r\n#29 0x0000000002e1013e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2e1013e)\r\n#30 0x0000000004c2e64c llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c2e64c)\r\n#31 0x0000000000db3bce llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdb3bce)\r\n#32 0x0000000004c2cf8e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c2cf8e)\r\n#33 0x0000000000db563e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdb563e)\r\n#34 0x0000000004c2ae9c llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c2ae9c)\r\n#35 0x00000000008d7b32 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d7b32)\r\n#36 0x00000000008cacdc optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8cacdc)\r\n#37 0x00007c7148029d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#38 0x00007c7148029e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#39 0x00000000008c212e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8c212e)", "author": "gbaraldi", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "121745", "issue_url": "https://github.com/llvm/llvm-project/issues/121745", "bug_type": "miscompilation", "base_commit": "3874c64418d2a7e36eab9af9253d905b48b36078", "knowledge_cutoff": "2025-01-06T10:26:21Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "f9369cc602272796c15de1065a782f812e791df3", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": [[562, 582], [1086, 1103]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": ["VPlanTransforms::adjustFixedOrderRecurrences", "VPlanTransforms::clearReductionWrapFlags", "collectUsersRecursively", "createScalarIVSteps", "legalizeAndOptimizeInductions"]}}, "patch": "commit f9369cc602272796c15de1065a782f812e791df3\nAuthor: Florian Hahn \nDate: Mon Jan 6 22:40:41 2025 +0000\n\n [VPlan] Make sure last IV increment value is available if needed.\n \n Legalize extract-from-ends using uniform VPReplicateRecipe of wide\n inductions to use regular VPReplicateRecipe, so the correct end value\n is available.\n \n Fixes https://github.com/llvm/llvm-project/issues/121745.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\nindex 395287bde76f..3e3f5adf73a0 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n@@ -562,21 +562,63 @@ createScalarIVSteps(VPlan &Plan, InductionDescriptor::InductionKind Kind,\n return Builder.createScalarIVSteps(InductionOpcode, FPBinOp, BaseIV, Step);\n }\n \n+static SmallVector collectUsersRecursively(VPValue *V) {\n+ SetVector Users(V->user_begin(), V->user_end());\n+ for (unsigned I = 0; I != Users.size(); ++I) {\n+ VPRecipeBase *Cur = cast(Users[I]);\n+ if (isa(Cur))\n+ continue;\n+ for (VPValue *V : Cur->definedValues())\n+ Users.insert(V->user_begin(), V->user_end());\n+ }\n+ return Users.takeVector();\n+}\n+\n /// Legalize VPWidenPointerInductionRecipe, by replacing it with a PtrAdd\n /// (IndStart, ScalarIVSteps (0, Step)) if only its scalar values are used, as\n /// VPWidenPointerInductionRecipe will generate vectors only. If some users\n /// require vectors while other require scalars, the scalar uses need to extract\n /// the scalars from the generated vectors (Note that this is different to how\n-/// int/fp inductions are handled). Also optimize VPWidenIntOrFpInductionRecipe,\n-/// if any of its users needs scalar values, by providing them scalar steps\n-/// built on the canonical scalar IV and update the original IV's users. This is\n-/// an optional optimization to reduce the needs of vector extracts.\n+/// int/fp inductions are handled). Legalize extract-from-ends using uniform\n+/// VPReplicateRecipe of wide inductions to use regular VPReplicateRecipe, so\n+/// the correct end value is available. Also optimize\n+/// VPWidenIntOrFpInductionRecipe, if any of its users needs scalar values, by\n+/// providing them scalar steps built on the canonical scalar IV and update the\n+/// original IV's users. This is an optional optimization to reduce the needs of\n+/// vector extracts.\n static void legalizeAndOptimizeInductions(VPlan &Plan) {\n+ using namespace llvm::VPlanPatternMatch;\n SmallVector ToRemove;\n VPBasicBlock *HeaderVPBB = Plan.getVectorLoopRegion()->getEntryBasicBlock();\n bool HasOnlyVectorVFs = !Plan.hasVF(ElementCount::getFixed(1));\n VPBuilder Builder(HeaderVPBB, HeaderVPBB->getFirstNonPhi());\n for (VPRecipeBase &Phi : HeaderVPBB->phis()) {\n+ auto *PhiR = dyn_cast(&Phi);\n+ if (!PhiR)\n+ break;\n+\n+ // Check if any uniform VPReplicateRecipes using the phi recipe are used by\n+ // ExtractFromEnd. Those must be replaced by a regular VPReplicateRecipe to\n+ // ensure the final value is available.\n+ // TODO: Remove once uniformity analysis is done on VPlan.\n+ for (VPUser *U : collectUsersRecursively(PhiR)) {\n+ auto *ExitIRI = dyn_cast(U);\n+ VPValue *Op;\n+ if (!ExitIRI || !match(ExitIRI->getOperand(0),\n+ m_VPInstruction(\n+ m_VPValue(Op), m_VPValue())))\n+ continue;\n+ auto *RepR = dyn_cast(Op);\n+ if (!RepR || !RepR->isUniform())\n+ continue;\n+ assert(!RepR->isPredicated() && \"RepR must not be predicated\");\n+ Instruction *I = RepR->getUnderlyingInstr();\n+ auto *Clone =\n+ new VPReplicateRecipe(I, RepR->operands(), /*IsUniform*/ false);\n+ Clone->insertAfter(RepR);\n+ RepR->replaceAllUsesWith(Clone);\n+ }\n+\n // Replace wide pointer inductions which have only their scalars used by\n // PtrAdd(IndStart, ScalarIVSteps (0, Step)).\n if (auto *PtrIV = dyn_cast(&Phi)) {\n@@ -1086,18 +1128,6 @@ bool VPlanTransforms::adjustFixedOrderRecurrences(VPlan &Plan,\n return true;\n }\n \n-static SmallVector collectUsersRecursively(VPValue *V) {\n- SetVector Users(V->user_begin(), V->user_end());\n- for (unsigned I = 0; I != Users.size(); ++I) {\n- VPRecipeBase *Cur = cast(Users[I]);\n- if (isa(Cur))\n- continue;\n- for (VPValue *V : Cur->definedValues())\n- Users.insert(V->user_begin(), V->user_end());\n- }\n- return Users.takeVector();\n-}\n-\n void VPlanTransforms::clearReductionWrapFlags(VPlan &Plan) {\n for (VPRecipeBase &R :\n Plan.getVectorLoopRegion()->getEntryBasicBlock()->phis()) {\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/iv_outside_user.ll", "commands": ["opt -S -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 < %s", "opt -S -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=1 < %s"], "tests": [{"test_name": "test_iv_uniform_with_outside_use_scev_simplification", "test_body": "define i32 @test_iv_uniform_with_outside_use_scev_simplification(ptr %dst) {\nentry:\n %step.1 = sext i8 0 to i32\n %step.2 = add nsw i32 %step.1, 1\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.dst = getelementptr inbounds i16, ptr %dst, i32 %iv\n store i16 0, ptr %gep.dst, align 2\n %iv.next = add i32 %step.2, %iv\n %cmp.i = icmp slt i32 %iv.next, 8\n br i1 %cmp.i, label %loop, label %e.exit\n\ne.exit: ; preds = %loop\n %res = phi i32 [ %iv.next, %loop ]\n ret i32 %res\n}\n", "additional_args": "-src-unroll=8 -tgt-unroll=8"}]}], "issue": {"title": "[clang] Miscompilation at -O2/3", "body": "This code prints 8 at `-O0/1` and prints 1 at `-O2/3`:\n\n```c\nint printf(const char *, ...);\nstatic char a;\nstatic char *b = &a;\nstatic int c;\nshort d;\nvoid e() {\n short f[8];\n char **g[] = {&b, &b};\n c = 0;\n for (; c < 8; c = 81 + 462704684 + *b - 462704765 + c + 1)\n f[c] = 0;\n d = f[5];\n}\nint main() {\n e();\n printf(\"%d\\n\", c);\n}\n```\n\nCompiler Explorer: https://godbolt.org/z/MEKz8oncP\n\nBisected to https://github.com/llvm/llvm-project/commit/df4a615c988f3ae56f7e68a7df86acb60f16493a, which was committed by @fhahn ", "author": "cardigan1008", "labels": ["miscompilation", "vectorizers"], "comments": [{"author": "dtcxzyw", "body": "Reduced testcase: https://alive2.llvm.org/ce/z/6Bg63G\n```\n; bin/opt -passes=loop-vectorize test.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i32 @src() {\nentry:\n %f.i = alloca [8 x i16], align 16\n %conv.i = sext i8 0 to i32\n %invariant.op.i = add nsw i32 %conv.i, 1\n br label %for.body.i\n\nfor.body.i: ; preds = %for.body.i, %entry\n %0 = phi i32 [ 0, %entry ], [ %add2.reass.i, %for.body.i ]\n %idxprom.i = sext i32 %0 to i64\n %arrayidx.i = getelementptr [8 x i16], ptr %f.i, i64 0, i64 %idxprom.i\n store i16 0, ptr %arrayidx.i, align 2\n %add2.reass.i = add i32 %invariant.op.i, %0\n %cmp.i = icmp slt i32 %add2.reass.i, 8\n br i1 %cmp.i, label %for.body.i, label %e.exit\n\ne.exit: ; preds = %for.body.i\n %add2.reass.i.lcssa = phi i32 [ %add2.reass.i, %for.body.i ]\n ret i32 %add2.reass.i.lcssa\n}\n```"}, {"author": "fhahn", "body": "Looking into it now, thanks"}, {"author": "yaoxinliu", "body": "Thank God! clang19 works as expected."}]}, "verified": true} {"bug_id": "117517", "issue_url": "https://github.com/llvm/llvm-project/issues/117517", "bug_type": "crash", "base_commit": "7e3187e12a2cef8b2552e08ce9423aca9c09d813", "knowledge_cutoff": "2024-11-25T05:18:47Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "57bbdbd7ae3698a274edd4dd6ef1b53d9129e552", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[1143, 1151]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["addMask"]}}, "patch": "commit 57bbdbd7ae3698a274edd4dd6ef1b53d9129e552\nAuthor: Alexey Bataev \nDate: Mon Nov 25 07:56:45 2024 -0800\n\n [SLP]Relax assertion in mask combine for non-power-of-2 number of elements\n \n The nodes may contain non-power-of-2 number of elements. Need to relax\n the assertion to avoid possible compiler crash\n \n Fixes #117517\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex d033b7c2ef4a..73218f377a06 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -1143,9 +1143,7 @@ static void addMask(SmallVectorImpl &Mask, ArrayRef SubMask,\n assert(\n (!ExtendingManyInputs || SubMask.size() > Mask.size() ||\n // Check if input scalars were extended to match the size of other node.\n- (SubMask.size() == Mask.size() &&\n- std::all_of(std::next(Mask.begin(), Mask.size() / 2), Mask.end(),\n- [](int Idx) { return Idx == PoisonMaskElem; }))) &&\n+ (SubMask.size() == Mask.size() && Mask.back() == PoisonMaskElem)) &&\n \"SubMask with many inputs support must be larger than the mask.\");\n if (Mask.empty()) {\n Mask.append(SubMask.begin(), SubMask.end());\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/non-power-2-num-elems-reused.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -slp-threshold=-10 < %s"], "tests": [{"test_name": "", "test_body": "\ndefine i64 @test() {\n;\nentry:\n %xor148.2.i = xor i32 0, 0\n %conv193.i = zext i32 %xor148.2.i to i64\n %conv193.1.i = zext i32 %xor148.2.i to i64\n %or194.1.i = or i64 %conv193.i, %conv193.1.i\n %xor148.2.i.1 = xor i32 0, 0\n %conv193.i.1 = zext i32 %xor148.2.i.1 to i64\n %or194.i.1 = or i64 %or194.1.i, %conv193.i.1\n %conv193.1.i.1 = zext i32 %xor148.2.i.1 to i64\n %or194.1.i.1 = or i64 %or194.i.1, %conv193.1.i.1\n %xor148.2.i.2 = xor i32 0, 0\n %conv193.i.2 = zext i32 %xor148.2.i.2 to i64\n %or194.i.2 = or i64 %or194.1.i.1, %conv193.i.2\n %conv193.1.i.2 = zext i32 %xor148.2.i.2 to i64\n %or194.1.i.2 = or i64 %or194.i.2, %conv193.1.i.2\n %xor148.2.i.3 = xor i32 0, 0\n %conv193.i.3 = zext i32 %xor148.2.i.3 to i64\n %or194.i.3 = or i64 %or194.1.i.2, %conv193.i.3\n %conv193.1.i.3 = zext i32 %xor148.2.i.3 to i64\n %or194.1.i.3 = or i64 %or194.i.3, %conv193.1.i.3\n %xor148.2.i.4 = xor i32 0, 0\n %conv193.i.4 = zext i32 %xor148.2.i.4 to i64\n %or194.i.4 = or i64 %or194.1.i.3, %conv193.i.4\n %conv193.1.i.4 = zext i32 %xor148.2.i.4 to i64\n %or194.1.i.4 = or i64 %or194.i.4, %conv193.1.i.4\n %xor148.2.i.5 = xor i32 0, 0\n %conv193.i.5 = zext i32 %xor148.2.i.5 to i64\n %or194.i.5 = or i64 %or194.1.i.4, %conv193.i.5\n %conv193.1.i.5 = zext i32 %xor148.2.i.5 to i64\n %or194.1.i.5 = or i64 %or194.i.5, %conv193.1.i.5\n %xor148.2.i.6 = xor i32 0, 0\n %conv193.i.6 = zext i32 %xor148.2.i.6 to i64\n %or194.i.6 = or i64 %or194.1.i.5, %conv193.i.6\n %or54.i.i.6 = or i32 %xor148.2.i.6, 0\n %conv193.1.i.6 = zext i32 %or54.i.i.6 to i64\n %xor148.2.i.7 = xor i32 0, 0\n %conv193.i.7 = zext i32 %xor148.2.i.7 to i64\n %0 = or i64 %or194.i.6, %conv193.i.7\n %conv193.1.i.7 = zext i32 %xor148.2.i.7 to i64\n %1 = or i64 %0, %conv193.1.i.7\n %or194.1.i.7 = or i64 %1, %conv193.1.i.6\n ret i64 %or194.1.i.7\n}"}]}], "issue": {"title": "[SLPVectorizer] Assertion \"SubMask with many inputs support must be larger than the mask.\" failed.", "body": "Reproducer: https://godbolt.org/z/Mzbe1G875\r\n```\r\n; bin/opt -passes=slp-vectorizer test.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i64 @test() {\r\nentry:\r\n %xor148.2.i = xor i32 0, 0\r\n %conv193.i = zext i32 %xor148.2.i to i64\r\n %conv193.1.i = zext i32 %xor148.2.i to i64\r\n %or194.1.i = or i64 %conv193.i, %conv193.1.i\r\n %xor148.2.i.1 = xor i32 0, 0\r\n %conv193.i.1 = zext i32 %xor148.2.i.1 to i64\r\n %or194.i.1 = or i64 %or194.1.i, %conv193.i.1\r\n %conv193.1.i.1 = zext i32 %xor148.2.i.1 to i64\r\n %or194.1.i.1 = or i64 %or194.i.1, %conv193.1.i.1\r\n %xor148.2.i.2 = xor i32 0, 0\r\n %conv193.i.2 = zext i32 %xor148.2.i.2 to i64\r\n %or194.i.2 = or i64 %or194.1.i.1, %conv193.i.2\r\n %conv193.1.i.2 = zext i32 %xor148.2.i.2 to i64\r\n %or194.1.i.2 = or i64 %or194.i.2, %conv193.1.i.2\r\n %xor148.2.i.3 = xor i32 0, 0\r\n %conv193.i.3 = zext i32 %xor148.2.i.3 to i64\r\n %or194.i.3 = or i64 %or194.1.i.2, %conv193.i.3\r\n %conv193.1.i.3 = zext i32 %xor148.2.i.3 to i64\r\n %or194.1.i.3 = or i64 %or194.i.3, %conv193.1.i.3\r\n %xor148.2.i.4 = xor i32 0, 0\r\n %conv193.i.4 = zext i32 %xor148.2.i.4 to i64\r\n %or194.i.4 = or i64 %or194.1.i.3, %conv193.i.4\r\n %conv193.1.i.4 = zext i32 %xor148.2.i.4 to i64\r\n %or194.1.i.4 = or i64 %or194.i.4, %conv193.1.i.4\r\n %xor148.2.i.5 = xor i32 0, 0\r\n %conv193.i.5 = zext i32 %xor148.2.i.5 to i64\r\n %or194.i.5 = or i64 %or194.1.i.4, %conv193.i.5\r\n %conv193.1.i.5 = zext i32 %xor148.2.i.5 to i64\r\n %or194.1.i.5 = or i64 %or194.i.5, %conv193.1.i.5\r\n %xor148.2.i.6 = xor i32 0, 0\r\n %conv193.i.6 = zext i32 %xor148.2.i.6 to i64\r\n %or194.i.6 = or i64 %or194.1.i.5, %conv193.i.6\r\n %or54.i.i.6 = or i32 %xor148.2.i.6, 0\r\n %conv193.1.i.6 = zext i32 %or54.i.i.6 to i64\r\n %xor148.2.i.7 = xor i32 0, 0\r\n %conv193.i.7 = zext i32 %xor148.2.i.7 to i64\r\n %0 = or i64 %or194.i.6, %conv193.i.7\r\n %conv193.1.i.7 = zext i32 %xor148.2.i.7 to i64\r\n %1 = or i64 %0, %conv193.1.i.7\r\n %or194.1.i.7 = or i64 %1, %conv193.1.i.6\r\n ret i64 %or194.1.i.7\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:1143: void addMask(llvm::SmallVectorImpl&, llvm::ArrayRef, bool): Assertion `(!ExtendingManyInputs || SubMask.size() > Mask.size() || (SubMask.size() == Mask.size() && std::all_of(std::next(Mask.begin(), Mask.size() / 2), Mask.end(), [](int Idx) { return Idx == PoisonMaskElem; }))) && \"SubMask with many inputs support must be larger than the mask.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"test\"\r\n #0 0x00000000051b7378 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x51b7378)\r\n #1 0x00000000051b4d6c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x0000751221842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007512218969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x0000751221842476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007512218287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000075122182871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x0000751221839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x00000000041e91f7 llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::finalize(llvm::ArrayRef, llvm::ArrayRef>, llvm::ArrayRef, unsigned int, llvm::function_ref&)>) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41e91f7)\r\n #9 0x00000000042256f9 llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::processBuildVector, llvm::slpvectorizer::BoUpSLP, llvm::SmallPtrSetImpl>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::TargetTransformInfo&, llvm::ArrayRef&, llvm::slpvectorizer::BoUpSLP&, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42256f9)\r\n#10 0x000000000422f707 llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x422f707)\r\n#11 0x00000000042636a1 llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42636a1)\r\n#12 0x00000000042691b0 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#13 0x000000000426c681 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x426c681)\r\n#14 0x0000000004270deb llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\r\n#15 0x00000000042758a6 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42758a6)\r\n#16 0x000000000427bf86 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#17 0x000000000427ca9b llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x427ca9b)\r\n#18 0x0000000002fcc5ce llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2fcc5ce)\r\n#19 0x0000000004fb6a68 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb6a68)\r\n#20 0x0000000000e3dd4e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3dd4e)\r\n#21 0x0000000004fb548e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb548e)\r\n#22 0x0000000000e3d3fe llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3d3fe)\r\n#23 0x0000000004fb4ec0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb4ec0)\r\n#24 0x000000000090ce7a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x90ce7a)\r\n#25 0x00000000008ff519 optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ff519)\r\n#26 0x0000751221829d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#27 0x0000751221829e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#28 0x00000000008f6d6e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f6d6e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "78621", "issue_url": "https://github.com/llvm/llvm-project/issues/78621", "bug_type": "miscompilation", "base_commit": "90ba33099cbb17e7c159e9ebc5a512037db99d6d", "knowledge_cutoff": "2024-01-18T20:57:06Z", "lit_test_dir": ["llvm/test/Transforms/ConstraintElimination"], "hints": {"fix_commit": "3d91d9613e294b242d853039209b40a0cb7853f2", "components": ["ConstraintElimination"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/ConstraintElimination.cpp": [[1061, 1071]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/ConstraintElimination.cpp": ["State::addInfoFor"]}}, "patch": "commit 3d91d9613e294b242d853039209b40a0cb7853f2\nAuthor: Florian Hahn \nDate: Wed Jan 24 14:25:54 2024 +0000\n\n [ConstraintElim] Make sure min/max intrinsic results are not poison.\n \n The result of umin may be poison and in that case the added constraints\n are not be valid in contexts where poison doesn't cause UB. Only queue\n facts for min/max intrinsics if the result is guaranteed to not be\n poison.\n \n This could be improved in the future, by only adding the fact when\n solving conditions using the result value.\n \n Fixes https://github.com/llvm/llvm-project/issues/78621.\n\ndiff --git a/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp b/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\nindex 8f09569d0d9c..7b672e89b67a 100644\n--- a/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\n+++ b/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\n@@ -1061,11 +1061,16 @@ void State::addInfoFor(BasicBlock &BB) {\n FactOrCheck::getCheck(DT.getNode(&BB), cast(&I)));\n break;\n // Enqueue the intrinsics to add extra info.\n- case Intrinsic::abs:\n case Intrinsic::umin:\n case Intrinsic::umax:\n case Intrinsic::smin:\n case Intrinsic::smax:\n+ // TODO: Check if it is possible to instead only added the min/max facts\n+ // when simplifying uses of the min/max intrinsics.\n+ if (!isGuaranteedNotToBePoison(&I))\n+ break;\n+ [[fallthrough]];\n+ case Intrinsic::abs:\n WorkList.push_back(FactOrCheck::getInstFact(DT.getNode(&BB), &I));\n break;\n }\n", "tests": [{"file": "llvm/test/Transforms/ConstraintElimination/minmax.ll", "commands": ["opt -passes=constraint-elimination -S %s"], "tests": [{"test_name": "smin_branchless", "test_body": "define i1 @smin_branchless(i32 %x, i32 %y) {\nentry:\n %min = call i32 @llvm.smin.i32(i32 %x, i32 %y)\n %cmp1 = icmp sle i32 %min, %x\n %cmp2 = icmp sgt i32 %min, %x\n %ret = xor i1 %cmp1, %cmp2\n ret i1 %ret\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.smin.i32(i32, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}, {"file": "llvm/test/Transforms/ConstraintElimination/umin-result-may-be-poison.ll", "commands": ["opt -p constraint-elimination -S %s"], "tests": [{"test_name": "umin_poison_call_before_UB", "test_body": "define i1 @umin_poison_call_before_UB(i32 %arg) {\n %icmp = icmp slt i32 %arg, 0\n %shl = shl nuw nsw i32 %arg, 3\n %min = call i32 @llvm.umin.i32(i32 %shl, i32 80)\n call void @fn()\n call void @noundef(i32 noundef %min)\n %cmp2 = shl nuw nsw i32 %arg, 3\n ret i1 %icmp\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.umin.i32(i32, i32) #0\n\ndeclare void @noundef(i32 noundef)\n\ndeclare void @fn()\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "umin_not_used", "test_body": "define i1 @umin_not_used(i32 %arg) {\n %icmp = icmp slt i32 %arg, 0\n %shl = shl nuw nsw i32 %arg, 3\n %1 = call i32 @llvm.umin.i32(i32 %shl, i32 80)\n %cmp2 = shl nuw nsw i32 %arg, 3\n ret i1 %icmp\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.umin.i32(i32, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "umin_poison_is_UB_via_call", "test_body": "define i1 @umin_poison_is_UB_via_call(i32 %arg) {\n %icmp = icmp slt i32 %arg, 0\n %shl = shl nuw nsw i32 %arg, 3\n %min = call i32 @llvm.umin.i32(i32 %shl, i32 80)\n call void @noundef(i32 noundef %min)\n %cmp2 = shl nuw nsw i32 %arg, 3\n ret i1 %icmp\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.umin.i32(i32, i32) #0\n\ndeclare void @noundef(i32 noundef)\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[ConstraintElimination] A miscompile in presence of `shl nuw nsw` and `llvm.umin`", "body": "The following commit: https://github.com/llvm/llvm-project/commit/71f56e49ceca75dbf82cbb9537c2545c2d2e51c9\r\nTriggered this miscompile: https://alive2.llvm.org/ce/z/oBg-u8\r\n\r\nNow `opt -passes=constraint-elimination -S` turns\r\n```\r\ndefine i1 @test(i32 %arg) {\r\n %icmp = icmp slt i32 %arg, 0\r\n %shl = shl nuw nsw i32 %arg, 3\r\n %call4 = call i32 @llvm.umin.i32(i32 %shl, i32 80)\r\n ret i1 %icmp\r\n}\r\n\r\ndeclare i32 @llvm.umin.i32(i32, i32) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\ninto\r\n```\r\ndefine i1 @test(i32 %arg) {\r\n %shl = shl nuw nsw i32 %arg, 3\r\n %call4 = call i32 @llvm.umin.i32(i32 %shl, i32 80)\r\n ret i1 false\r\n}\r\n\r\ndeclare i32 @llvm.umin.i32(i32, i32) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\n\r\nGiven this debug output:\r\n```\r\nopt -passes=constraint-elimination before-constr-elim-renamed.ll -S -debug-only=constraint-elimination,constraint-system\r\nProcessing fact to add to the system: icmp ule i32 %call4, %shl\r\nAdding 'icmp ule i32 %call4, %shl'\r\n constraint: -8 * %arg + %call4 <= 0\r\n\r\n---\r\n-1 * %arg <= 0\r\n-8 * %arg + %call4 <= 0\r\n-1 * %call4 <= 0\r\n8 * %arg <= -1\r\nunsat\r\nAdding 'icmp sge i32 %call4, 0'\r\n constraint: -1 * %call4 <= 0\r\n\r\nAdding 'icmp sle i32 %call4, %shl'\r\n constraint: -8 * %arg + %call4 <= 0\r\n\r\nProcessing fact to add to the system: icmp ule i32 %call4, 80\r\nAdding 'icmp ule i32 %call4, 80'\r\n constraint: %call4 <= 80\r\n\r\nAdding 'icmp sge i32 %call4, 0'\r\n constraint: -1 * %call4 <= 0\r\n\r\nAdding 'icmp sle i32 %call4, 80'\r\n constraint: %call4 <= 80\r\n\r\nTop of stack : 0 1\r\nCB: 0 1\r\nProcessing condition to simplify: %icmp = icmp slt i32 %arg, 0\r\nChecking %icmp = icmp slt i32 %arg, 0\r\n---\r\n-1 * %call4 <= 0\r\n-8 * %arg + %call4 <= 0\r\n-1 * %call4 <= 0\r\n%call4 <= 80\r\n-1 * %arg <= 0\r\nsat\r\n---\r\n-1 * %call4 <= 0\r\n-8 * %arg + %call4 <= 0\r\n-1 * %call4 <= 0\r\n%call4 <= 80\r\n%arg <= -1\r\nunsat\r\nCondition icmp sge i32 %arg, 0 implied by dominating constraints\r\n-1 * %call4 <= 0\r\n-8 * %arg + %call4 <= 0\r\n-1 * %call4 <= 0\r\n%call4 <= 80\r\n```\r\nI'm not 100% sure if the aforementioned patch causes the miscompile, of if it merely triggers it on this specific reproducer. The fact that something like `-1 * %call4 <= 0` is reported as a \"dominating constraint\" seems suspicious to a person who's not familiar with the logic of ConstraintElimination and it doesn't look directly related to that patch. However, if I replace the `shl nuw nsw i32 %arg, 3\r\n` with `mul nuw nsw i32 %arg, 8`, the miscompile doesn't happen: https://alive2.llvm.org/ce/z/EL2bCh.", "author": "DaniilSuchkov", "labels": ["miscompilation", "release:backport", "llvm:transforms"], "comments": [{"author": "nikic", "body": "From the debug log, I suspect that we're using the unsigned constraint system to prove a signed fact for some reason.\r\n\r\ncc @fhahn "}, {"author": "nikic", "body": "Okay, this is an interesting case.\r\n\r\nWe are adding an `icmp ule i32 %call4, %shl` fact, which get transferred as `icmp sge i32 %call4, 0` and `icmp sle i32 %call4, %shl` to the signed system, because `%shl` is known non-negative.\r\n\r\nThen we are checking whether a solution of these constraints plus `%arg <= -1` exists, which it does not.\r\n\r\nThis is \"correct\" in the sense that *if* that umin call returns a non-poison result, then `icmp slt i32 %arg, 0` is indeed known to be false. But we obviously can't actually make that assumption.\r\n\r\n@dtcxzyw @fhahn This makes me think that the way we are currently adding facts for MinMaxIntrinsic (and I guess abs as well) is pretty fundamentally incorrect, because it effectively imposes a constraint that the umin operation returns a well-defined value. Any thoughts on how to fix this without dropping support for them entirely?"}, {"author": "nikic", "body": "(The difference between shl and mul in the reproducer is because ValueTracking apparently currently doesn't know that mul nuw nsw is non-negative, but knows that shl nuw nsw is.)"}, {"author": "dtcxzyw", "body": "> @dtcxzyw @fhahn This makes me think that the way we are currently adding facts for MinMaxIntrinsic (and I guess abs as well) is pretty fundamentally incorrect, because it effectively imposes a constraint that the umin operation returns a well-defined value. Any thoughts on how to fix this without dropping support for them entirely?\r\n\r\nCan we conservatively check these intrinsics using `isGuaranteedNotToBePoison`?\r\n"}, {"author": "nikic", "body": "If it's not too complicated, I think a better solution would be to only add the intrinsic constraints when the result variable is added to the constraint system."}, {"author": "fhahn", "body": "> If it's not too complicated, I think a better solution would be to only add the intrinsic constraints when the result variable is added to the constraint system.\r\n\r\nOne way to do that would be to check the uses of the intrinsic and find all the branches that it feeds, and add it as fact for the successors."}, {"author": "nikic", "body": "@dtcxzyw Do you plan to submit a PR for this issue?"}, {"author": "dtcxzyw", "body": "> @dtcxzyw Do you plan to submit a PR for this issue?\r\n\r\nI will post a patch later."}, {"author": "fhahn", "body": "I've pushed a fix that only adds the facts if the result is guaranteed to not be poison so the mis-compile is fixed. I also added a TODO to improve things if possible.\r\n\r\nReopening so we can pick the fix for the release branch"}, {"author": "nikic", "body": "@fhahn Why does your fix only do this for min/max but not abs? Can't it have the same problem?"}, {"author": "fhahn", "body": "> @fhahn Why does your fix only do this for min/max but not abs? Can't it have the same problem?\r\n\r\nI wasn't able to come up with a problematic test case so far for `abs`, but there's a few more things to try."}, {"author": "DaniilSuchkov", "body": "@fhahn I'm not familiar with how this pass works, so I have a question: will this pass still use min/max/etc. to infer facts about the arguments of those intrinsics? I'm just a bit confused by the terminology: I see that it adds \"constraints\" based on things like `icmp`, which on their own don't actually constrain anything unless you branch on the result (or do something similar)."}, {"author": "nikic", "body": "> @fhahn I'm not familiar with how this pass works, so I have a question: will this pass still use min/max/etc. to infer facts about the arguments of those intrinsics? I'm just a bit confused by the terminology: I see that it adds \"constraints\" based on things like `icmp`, which on their own don't actually constrain anything unless you branch on the result (or do something similar).\r\n\r\nConstraints for icmps will only be added in branches that the icmp dominates (or code that an assume with an icmp dominates). The intrinsic handling was a bit special in that constraints are added unconditionally."}]}, "verified": true} {"bug_id": "97590", "issue_url": "https://github.com/llvm/llvm-project/issues/97590", "bug_type": "crash", "base_commit": "ed51908cec879c9dff435abdc70d8b03afc35c07", "knowledge_cutoff": "2024-07-03T15:09:36Z", "lit_test_dir": ["llvm/test/Transforms/Inline"], "hints": {"fix_commit": "be7239e5a60927f5b4932f995dc4b57423ea8534", "components": ["InlineCost"], "bug_location_lineno": {"llvm/lib/Analysis/InlineCost.cpp": [[2644, 2651]]}, "bug_location_funcname": {"llvm/lib/Analysis/InlineCost.cpp": ["CallAnalyzer::stripAndComputeInBoundsConstantOffsets"]}}, "patch": "commit be7239e5a60927f5b4932f995dc4b57423ea8534\nAuthor: Yingwei Zheng \nDate: Tue Jul 9 15:08:04 2024 +0800\n\n [Inline] Remove bitcast handling in `CallAnalyzer::stripAndComputeInBoundsConstantOffsets` (#97988)\n \n As we are now using opaque pointers, bitcast handling is no longer\n needed.\n \n Closes https://github.com/llvm/llvm-project/issues/97590.\n\ndiff --git a/llvm/lib/Analysis/InlineCost.cpp b/llvm/lib/Analysis/InlineCost.cpp\nindex 94aa8ecf6f55..345e5a019520 100644\n--- a/llvm/lib/Analysis/InlineCost.cpp\n+++ b/llvm/lib/Analysis/InlineCost.cpp\n@@ -2644,8 +2644,6 @@ ConstantInt *CallAnalyzer::stripAndComputeInBoundsConstantOffsets(Value *&V) {\n if (!GEP->isInBounds() || !accumulateGEPOffset(*GEP, Offset))\n return nullptr;\n V = GEP->getPointerOperand();\n- } else if (Operator::getOpcode(V) == Instruction::BitCast) {\n- V = cast(V)->getOperand(0);\n } else if (GlobalAlias *GA = dyn_cast(V)) {\n if (GA->isInterposable())\n break;\n", "tests": [{"file": "llvm/test/Transforms/Inline/pr97590.ll", "commands": ["opt -S -passes=inline < %s"], "tests": [{"test_name": "", "test_body": "\n; Check that we don't crash when computing the\n; base pointer of `bitcast <1 x ptr> to ptr`\n\ndefine void @caller(<1 x ptr> %x) {\n;\nentry:\n %p = bitcast <1 x ptr> %x to ptr\n call void @callee(ptr %p)\n ret void\n}\n\ndefine void @callee(ptr %p) {\n;\nentry:\n ret void\n}"}]}], "issue": {"title": "[Inline] Assertion `V->getType()->isPointerTy() && \"Unexpected operand type!\"' failed.", "body": "Reproducer: https://godbolt.org/z/qGc1Kx5YK\r\n```\r\n; bin/opt -passes=inline reduced.ll -S\r\ndefine i32 @func_120(i1 %tobool626, ptr %g_231, ptr %l_506, i1 %tobool724, ptr %p) {\r\nentry:\r\n %0 = bitcast i1 %tobool626 to <1 x i1>\r\n call void @llvm.masked.store.v1i64.p0(<1 x i64> zeroinitializer, ptr %g_231, i32 8, <1 x i1> %0)\r\n %1 = xor i1 %tobool626, true\r\n %2 = bitcast i1 %1 to <1 x i1>\r\n %3 = call <1 x ptr> @llvm.masked.load.v1p0.p0(ptr %g_231, i32 8, <1 x i1> %2, <1 x ptr> poison)\r\n %4 = bitcast <1 x ptr> %3 to ptr\r\n br i1 %tobool626, label %common.ret, label %if.else647\r\n\r\ncommon.ret: ; preds = %entry, %lor.end739\r\n ret i32 0\r\n\r\nif.else647: ; preds = %entry\r\n br i1 %tobool724, label %land.rhs725, label %lor.end739\r\n\r\nland.rhs725: ; preds = %if.else647\r\n %call726 = load i32, ptr %p, align 4\r\n br label %lor.end739\r\n\r\nlor.end739: ; preds = %land.rhs725, %if.else647\r\n %call745 = call i32 @func_136(ptr %4)\r\n br label %common.ret\r\n}\r\n\r\ndefine i32 @func_136(ptr nocapture readnone %p_137) {\r\nentry:\r\n ret i32 0\r\n}\r\n\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/lib/Analysis/InlineCost.cpp:2656: llvm::ConstantInt* {anonymous}::CallAnalyzer::stripAndComputeInBoundsConstantOffsets(llvm::Value*&): Assertion `V->getType()->isPointerTy() && \"Unexpected operand type!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=inline \r\n1.\tRunning pass \"cgscc(inline)\" on module \"\"\r\n #0 0x0000000004e7cf38 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7cf38)\r\n #1 0x0000000004e7a69c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000077b96f042520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000077b96f0969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000077b96f042476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000077b96f0287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000077b96f02871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000077b96f039e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x00000000043ea53a (anonymous namespace)::CallAnalyzer::analyze() (.part.0) InlineCost.cpp:0:0\r\n #9 0x00000000043eb650 llvm::getInlineCost(llvm::CallBase&, llvm::Function*, llvm::InlineParams const&, llvm::TargetTransformInfo&, llvm::function_ref, llvm::function_ref, llvm::function_ref, llvm::ProfileSummaryInfo*, llvm::OptimizationRemarkEmitter*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x43eb650)\r\n#10 0x00000000043ebba6 llvm::getInlineCost(llvm::CallBase&, llvm::InlineParams const&, llvm::TargetTransformInfo&, llvm::function_ref, llvm::function_ref, llvm::function_ref, llvm::ProfileSummaryInfo*, llvm::OptimizationRemarkEmitter*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x43ebba6)\r\n#11 0x00000000043ed0ab llvm::InlineCost llvm::function_ref::callback_fn&, llvm::InlineParams const&)::'lambda2'(llvm::CallBase&)>(long, llvm::CallBase&) InlineAdvisor.cpp:0:0\r\n#12 0x00000000043f310d llvm::shouldInline(llvm::CallBase&, llvm::function_ref, llvm::OptimizationRemarkEmitter&, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x43f310d)\r\n#13 0x00000000043f4af5 getDefaultInlineAdvice(llvm::CallBase&, llvm::AnalysisManager&, llvm::InlineParams const&) InlineAdvisor.cpp:0:0\r\n#14 0x00000000043f4d3b llvm::DefaultInlineAdvisor::getAdviceImpl(llvm::CallBase&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x43f4d3b)\r\n#15 0x00000000043eda71 llvm::InlineAdvisor::getAdvice(llvm::CallBase&, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x43eda71)\r\n#16 0x000000000371e48d llvm::InlinerPass::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x371e48d)\r\n#17 0x0000000002e127ae llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2e127ae)\r\n#18 0x000000000431385b llvm::PassManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x431385b)\r\n#19 0x0000000002e1154e llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>, llvm::AnalysisManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2e1154e)\r\n#20 0x00000000043155b9 llvm::ModuleToPostOrderCGSCCPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x43155b9)\r\n#21 0x0000000002e1152e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2e1152e)\r\n#22 0x0000000004c780c0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c780c0)\r\n#23 0x00000000008d8702 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d8702)\r\n#24 0x00000000008cb60c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8cb60c)\r\n#25 0x000077b96f029d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#26 0x000077b96f029e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#27 0x00000000008c2a6e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8c2a6e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\nhttps://github.com/llvm/llvm-project/blob/f1a8f94bba8b090a0d667065e4b8b4fc66b6a5cc/llvm/lib/Analysis/InlineCost.cpp#L2647-L2648\r\nshould be removed as we use opaque pointers now.\r\n", "author": "dtcxzyw", "labels": ["crash-on-valid", "llvm:analysis"], "comments": [{"author": "dtcxzyw", "body": "Related patch: #96878 "}, {"author": "dtcxzyw", "body": "Reduced reproducer: https://godbolt.org/z/rbjK6ETEW\r\n```\r\n; bin/opt -passes=inline reduced.ll -S\r\ndefine void @caller(<1 x ptr> %x) {\r\nentry:\r\n %p = bitcast <1 x ptr> %x to ptr\r\n call void @callee(ptr %p)\r\n ret void\r\n}\r\n\r\ndefine void @callee(ptr %p) {\r\nentry:\r\n ret void\r\n}\r\n```"}]}, "verified": true} {"bug_id": "94328", "issue_url": "https://github.com/llvm/llvm-project/issues/94328", "bug_type": "crash", "base_commit": "098bd842a7e50853fa231f8b73c24ec5006fe063", "knowledge_cutoff": "2024-06-04T09:16:11Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "008df3cf85e9bb1532c079bfd7a7a00e90e0a3c6", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[3914, 3932], [9516, 9521]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::collectLoopUniforms", "VPReplicateRecipe::execute"]}}, "patch": "commit 008df3cf85e9bb1532c079bfd7a7a00e90e0a3c6\nAuthor: Florian Hahn \nDate: Fri Jul 19 12:02:25 2024 +0100\n\n [LV] Check isPredInst instead of isScalarWithPred in uniform analysis. (#98892)\n \n Any instruction marked as uniform will result in a uniform\n VPReplicateRecipe. If it requires predication, it will be placed in a\n replicate region, even if isScalarWithPredication returns false.\n \n Check isPredicatedInst instead of isScalarWithPredication to avoid\n generating uniform VPReplicateRecipes placed inside a replicate region.\n This fixes an assertion when using scalable VFs.\n \n Fixes https://github.com/llvm/llvm-project/issues/80416.\n Fixes https://github.com/llvm/llvm-project/issues/94328.\n Fixes https://github.com/llvm/llvm-project/issues/99625.\n \n PR: https://github.com/llvm/llvm-project/pull/98892\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex fbca4cdcbcfc..cceed75aa50b 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -3914,19 +3914,19 @@ void LoopVectorizationCostModel::collectLoopUniforms(ElementCount VF) {\n SetVector Worklist;\n \n // Add uniform instructions demanding lane 0 to the worklist. Instructions\n- // that are scalar with predication must not be considered uniform after\n+ // that require predication must not be considered uniform after\n // vectorization, because that would create an erroneous replicating region\n // where only a single instance out of VF should be formed.\n- // TODO: optimize such seldom cases if found important, see PR40816.\n auto addToWorklistIfAllowed = [&](Instruction *I) -> void {\n if (isOutOfScope(I)) {\n LLVM_DEBUG(dbgs() << \"LV: Found not uniform due to scope: \"\n << *I << \"\\n\");\n return;\n }\n- if (isScalarWithPredication(I, VF)) {\n- LLVM_DEBUG(dbgs() << \"LV: Found not uniform being ScalarWithPredication: \"\n- << *I << \"\\n\");\n+ if (isPredicatedInst(I)) {\n+ LLVM_DEBUG(\n+ dbgs() << \"LV: Found not uniform due to requiring predication: \" << *I\n+ << \"\\n\");\n return;\n }\n LLVM_DEBUG(dbgs() << \"LV: Found uniform instruction: \" << *I << \"\\n\");\n@@ -9516,6 +9516,8 @@ void VPInterleaveRecipe::execute(VPTransformState &State) {\n void VPReplicateRecipe::execute(VPTransformState &State) {\n Instruction *UI = getUnderlyingInstr();\n if (State.Instance) { // Generate a single instance.\n+ assert((State.VF.isScalar() || !isUniform()) &&\n+ \"uniform recipe shouldn't be predicated\");\n assert(!State.VF.isScalable() && \"Can't scalarize a scalable vector\");\n State.ILV->scalarizeInstruction(UI, this, *State.Instance, State);\n // Insert scalar instance packing it into a vector.\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/divs-with-scalable-vfs.ll", "commands": ["opt -p loop-vectorize -mtriple aarch64 -mcpu=neoverse-v1 -S %s"], "tests": [{"test_name": "udiv_urem_feeding_gep", "test_body": "define void @udiv_urem_feeding_gep(i64 %x, ptr %dst, i64 %N) {\nentry:\n %mul.1.i = mul i64 %x, %x\n %mul.2.i = mul i64 %mul.1.i, %x\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %div.i = udiv i64 %iv, %mul.2.i\n %rem.i = urem i64 %iv, %mul.2.i\n %div.1.i = udiv i64 %rem.i, %mul.1.i\n %rem.1.i = urem i64 %rem.i, %mul.1.i\n %div.2.i = udiv i64 %rem.1.i, %x\n %rem.2.i = urem i64 %rem.1.i, %x\n %mul.i = mul i64 %x, %div.i\n %add.i = add i64 %mul.i, %div.1.i\n %mul.1.i9 = mul i64 %add.i, %x\n %add.1.i = add i64 %mul.1.i9, %div.2.i\n %mul.2.i11 = mul i64 %add.1.i, %x\n %add.2.i = add i64 %mul.2.i11, %rem.2.i\n %sext.i = shl i64 %add.2.i, 32\n %conv6.i = ashr i64 %sext.i, 32\n %gep = getelementptr i64, ptr %dst, i64 %conv6.i\n store i64 %div.i, ptr %gep, align 4\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv, %N\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "sdiv_feeding_gep_predicated", "test_body": "define void @sdiv_feeding_gep_predicated(ptr %dst, i32 %x, i64 %M, i64 %conv6, i64 %N) {\nentry:\n %conv61 = zext i32 %x to i64\n br label %loop\n\nloop: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %c = icmp ule i64 %iv, %M\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop\n %div18 = sdiv i64 %M, %conv6\n %conv20 = trunc i64 %div18 to i32\n %mul30 = mul i64 %div18, %conv61\n %sub31 = sub i64 %iv, %mul30\n %conv34 = trunc i64 %sub31 to i32\n %mul35 = mul i32 %x, %conv20\n %add36 = add i32 %mul35, %conv34\n %idxprom = sext i32 %add36 to i64\n %gep = getelementptr double, ptr %dst, i64 %idxprom\n store double 0.000000e+00, ptr %gep, align 8\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}, {"test_name": "sdiv_feeding_gep", "test_body": "define void @sdiv_feeding_gep(ptr %dst, i32 %x, i64 %M, i64 %conv6, i64 %N) {\nentry:\n %conv61 = zext i32 %x to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %div18 = sdiv i64 %M, %conv6\n %conv20 = trunc i64 %div18 to i32\n %mul30 = mul i64 %div18, %conv61\n %sub31 = sub i64 %iv, %mul30\n %conv34 = trunc i64 %sub31 to i32\n %mul35 = mul i32 %x, %conv20\n %add36 = add i32 %mul35, %conv34\n %idxprom = sext i32 %add36 to i64\n %gep = getelementptr double, ptr %dst, i64 %idxprom\n store double 0.000000e+00, ptr %gep, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "[VPlan] Report \"Assertion `!State->VF.isScalable() && \"VF is assumed to be non scalable.\"' failed\"", "body": "The IR is put at the end.\r\n\r\nCompile command is `opt -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue`\r\n\r\nThe error is\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/VPlan.cpp:734: virtual void llvm::VPRegionBlock::execute(llvm::VPTransformState*): Assertion `!State->VF.isScalable() && \"VF is assumed to be non scalable.\"' failed.\r\n```\r\n\r\nIt can be seen at https://godbolt.org/z/s4bqzdKPP\r\n\r\n```\r\n; ModuleID = 'test.cpp'\r\nsource_filename = \"test.cpp\"\r\ntarget datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"aarch64-unknown-linux-gnu\"\r\n\r\n%struct.ident_t = type { i32, i32, i32, i32, ptr }\r\n\r\n@0 = private unnamed_addr constant [23 x i8] c\";unknown;unknown;0;0;;\\00\", align 1\r\n@1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 514, i32 0, i32 22, ptr @0 }, align 8\r\n@2 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 22, ptr @0 }, align 8\r\n\r\n; Function Attrs: mustprogress nounwind uwtable vscale_range(1,16)\r\ndefine dso_local void @_Z4testiiPdS_(i32 noundef %nx, i32 noundef %ik, ptr noundef %out, ptr noundef %rspace) local_unnamed_addr #0 {\r\nentry:\r\n %nx.addr = alloca i32, align 4\r\n %ik.addr = alloca i32, align 4\r\n %out.addr = alloca ptr, align 8\r\n %rspace.addr = alloca ptr, align 8\r\n store i32 %nx, ptr %nx.addr, align 4\r\n store i32 %ik, ptr %ik.addr, align 4\r\n store ptr %out, ptr %out.addr, align 8\r\n store ptr %rspace, ptr %rspace.addr, align 8\r\n call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr nonnull @2, i32 4, ptr nonnull @_Z4testiiPdS_.omp_outlined, ptr nonnull %nx.addr, ptr nonnull %ik.addr, ptr nonnull %out.addr, ptr nonnull %rspace.addr)\r\n ret void\r\n}\r\n\r\n; Function Attrs: alwaysinline norecurse nounwind uwtable vscale_range(1,16)\r\ndefine internal void @_Z4testiiPdS_.omp_outlined(ptr noalias nocapture noundef readonly %.global_tid., ptr noalias nocapture noundef readnone %.bound_tid., ptr noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) %nx, ptr noalias nocapture noundef nonnull readonly align 4 dereferenceable(4) %ik, ptr noalias nocapture noundef nonnull readonly align 8 dereferenceable(8) %out, ptr noalias nocapture noundef nonnull readonly align 8 dereferenceable(8) %rspace) #1 {\r\nentry:\r\n %.omp.lb = alloca i64, align 8\r\n %.omp.ub = alloca i64, align 8\r\n %.omp.stride = alloca i64, align 8\r\n %.omp.is_last = alloca i32, align 4\r\n %0 = load i32, ptr %nx, align 4\r\n %1 = load i32, ptr %ik, align 4\r\n %cmp = icmp sgt i32 %0, 0\r\n %cmp8 = icmp sgt i32 %1, 0\r\n %or.cond = select i1 %cmp, i1 %cmp8, i1 false\r\n br i1 %or.cond, label %omp.precond.then, label %omp.precond.end\r\n\r\nomp.precond.then: ; preds = %entry\r\n %conv = zext i32 %0 to i64\r\n %conv6 = zext i32 %1 to i64\r\n %mul = mul nuw nsw i64 %conv6, %conv\r\n %sub7 = add nsw i64 %mul, -1\r\n call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %.omp.lb) #3\r\n store i64 0, ptr %.omp.lb, align 8\r\n call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %.omp.ub) #3\r\n store i64 %sub7, ptr %.omp.ub, align 8\r\n call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %.omp.stride) #3\r\n store i64 1, ptr %.omp.stride, align 8\r\n call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %.omp.is_last) #3\r\n store i32 0, ptr %.omp.is_last, align 4\r\n %2 = load i32, ptr %.global_tid., align 4\r\n call void @__kmpc_for_static_init_8(ptr nonnull @1, i32 %2, i32 33, ptr nonnull %.omp.is_last, ptr nonnull %.omp.lb, ptr nonnull %.omp.ub, ptr nonnull %.omp.stride, i64 1, i64 512)\r\n %3 = load i64, ptr %.omp.ub, align 8\r\n %cond60 = call i64 @llvm.smin.i64(i64 %3, i64 %sub7)\r\n store i64 %cond60, ptr %.omp.ub, align 8\r\n %4 = load i64, ptr %.omp.lb, align 8\r\n %cmp12.not61 = icmp sgt i64 %4, %cond60\r\n br i1 %cmp12.not61, label %omp.dispatch.end, label %omp.inner.for.cond.preheader.lr.ph\r\n\r\nomp.inner.for.cond.preheader.lr.ph: ; preds = %omp.precond.then\r\n br label %omp.inner.for.cond.preheader\r\n\r\nomp.inner.for.cond.preheader: ; preds = %omp.inner.for.cond.preheader.lr.ph, %omp.dispatch.inc\r\n %5 = phi i64 [ %4, %omp.inner.for.cond.preheader.lr.ph ], [ %add42, %omp.dispatch.inc ]\r\n %cond62 = phi i64 [ %cond60, %omp.inner.for.cond.preheader.lr.ph ], [ %cond, %omp.dispatch.inc ]\r\n %smax = call i64 @llvm.smax.i64(i64 %cond62, i64 %5)\r\n %6 = add i64 %smax, 1\r\n %7 = load ptr, ptr %rspace, align 8\r\n %8 = load ptr, ptr %out, align 8\r\n br label %omp.inner.for.body\r\n\r\nomp.inner.for.body: ; preds = %omp.inner.for.cond.preheader, %omp.inner.for.body\r\n %.omp.iv.059 = phi i64 [ %5, %omp.inner.for.cond.preheader ], [ %add41, %omp.inner.for.body ]\r\n %div18 = sdiv i64 %.omp.iv.059, %conv6\r\n %conv20 = trunc i64 %div18 to i32\r\n %mul30 = mul nsw i64 %div18, %conv6\r\n %sub31 = sub nsw i64 %.omp.iv.059, %mul30\r\n %conv34 = trunc i64 %sub31 to i32\r\n %mul35 = mul nsw i32 %1, %conv20\r\n %add36 = add nsw i32 %mul35, %conv34\r\n %idxprom = sext i32 %add36 to i64\r\n %arrayidx = getelementptr inbounds double, ptr %7, i64 %idxprom\r\n %9 = load double, ptr %arrayidx, align 8\r\n %arrayidx40 = getelementptr inbounds double, ptr %8, i64 %idxprom\r\n store double %9, ptr %arrayidx40, align 8\r\n %add41 = add i64 %.omp.iv.059, 1\r\n %exitcond = icmp ne i64 %add41, %6\r\n br i1 %exitcond, label %omp.inner.for.body, label %omp.dispatch.inc\r\n\r\nomp.dispatch.inc: ; preds = %omp.inner.for.body\r\n %10 = load i64, ptr %.omp.stride, align 8\r\n %add42 = add nsw i64 %10, %5\r\n store i64 %add42, ptr %.omp.lb, align 8\r\n %add43 = add nsw i64 %10, %cond62\r\n %cond = call i64 @llvm.smin.i64(i64 %add43, i64 %sub7)\r\n store i64 %cond, ptr %.omp.ub, align 8\r\n %cmp12.not = icmp sgt i64 %add42, %cond\r\n br i1 %cmp12.not, label %omp.dispatch.cond.omp.dispatch.end_crit_edge, label %omp.inner.for.cond.preheader\r\n\r\nomp.dispatch.cond.omp.dispatch.end_crit_edge: ; preds = %omp.dispatch.inc\r\n br label %omp.dispatch.end\r\n\r\nomp.dispatch.end: ; preds = %omp.dispatch.cond.omp.dispatch.end_crit_edge, %omp.precond.then\r\n call void @__kmpc_for_static_fini(ptr nonnull @1, i32 %2)\r\n call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %.omp.is_last) #3\r\n call void @llvm.lifetime.end.p0(i64 8, ptr nonnull %.omp.stride) #3\r\n call void @llvm.lifetime.end.p0(i64 8, ptr nonnull %.omp.ub) #3\r\n call void @llvm.lifetime.end.p0(i64 8, ptr nonnull %.omp.lb) #3\r\n br label %omp.precond.end\r\n\r\nomp.precond.end: ; preds = %omp.dispatch.end, %entry\r\n ret void\r\n}\r\n\r\n; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)\r\ndeclare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #2\r\n\r\n; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)\r\ndeclare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #2\r\n\r\n; Function Attrs: nounwind\r\ndeclare void @__kmpc_for_static_init_8(ptr, i32, i32, ptr, ptr, ptr, ptr, i64, i64) local_unnamed_addr #3\r\n\r\n; Function Attrs: nounwind\r\ndeclare void @__kmpc_for_static_fini(ptr, i32) local_unnamed_addr #3\r\n\r\n; Function Attrs: nounwind\r\ndeclare void @__kmpc_fork_call(ptr, i32, ptr, ...) local_unnamed_addr #3\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i64 @llvm.smin.i64(i64, i64) #4\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i64 @llvm.smax.i64(i64, i64) #4\r\n\r\nattributes #0 = { mustprogress nounwind uwtable vscale_range(1,16) \"frame-pointer\"=\"non-leaf\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"generic\" \"target-features\"=\"+fp-armv8,+fullfp16,+neon,+sve,+v8a,-fmv\" }\r\nattributes #1 = { alwaysinline norecurse nounwind uwtable vscale_range(1,16) \"frame-pointer\"=\"non-leaf\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"generic\" \"target-features\"=\"+fp-armv8,+fullfp16,+neon,+sve,+v8a,-fmv\" }\r\nattributes #2 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }\r\nattributes #3 = { nounwind }\r\nattributes #4 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n\r\n```", "author": "eastB233", "labels": ["vectorizers", "crash"], "comments": [{"author": "eastB233", "body": "Simpler test case is\r\n```c\r\nextern const int npy;\r\nextern const int nx;\r\ndouble* rspace;\r\n\r\nvoid recip2real(double* out, const double factor)\r\n{\r\n#ifdef _OPENMP\r\n#pragma omp parallel for collapse(2)\r\n#endif\r\n for (int ix = 0; ix < nx; ++ix)\r\n {\r\n for (int ipy = 0; ipy < npy; ++ipy) {\r\n out[ix * npy + ipy] += factor * rspace[ix * npy + ipy];\r\n }\r\n }\r\n}\r\n```\r\n\r\nCommand is `clang -O3 -march=armv9-a -fopenmp -mllvm -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -S`\r\n\r\nIt can be seen at https://godbolt.org/z/6MGsecEr5"}, {"author": "eastB233", "body": "I don't know much about VPlan, and as far as I understand by tracing the code,\r\n\r\n### first, it fails at the following VPlan\r\n```\r\nVPlan 'Initial VPlan for VF={vscale x 1,vscale x 2},UF>=1' {\r\nvp<%2> = original trip-count\r\n\r\nph:\r\n EMIT vp<%2> = EXPAND SCEV (1 + (-1 * %5) + ((-1 + ((zext i32 %0 to i64) * (sext i32 %1 to i64))) smin %4))\r\nNo successors\r\n\r\nvector.ph:\r\n EMIT vp<%3> = TC > VF ? TC - VF : 0 vp<%2>\r\n EMIT vp<%4> = VF * Part + ir<0>\r\n EMIT vp<%5> = active lane mask vp<%4>, vp<%2>\r\nSuccessor(s): vector loop\r\n\r\n vector loop: {\r\n vector.body:\r\n EMIT vp<%6> = CANONICAL-INDUCTION\r\n ACTIVE-LANE-MASK-PHI vp<%7> = phi vp<%5>, vp<%27>\r\n vp<%8> = DERIVED-IV ir<%5> + vp<%6> * ir<1>\r\n vp<%9> = SCALAR-STEPS vp<%8>, ir<1>\r\n Successor(s): pred.sdiv\r\n\r\n pred.sdiv: {\r\n pred.sdiv.entry:\r\n BRANCH-ON-MASK vp<%7>\r\n Successor(s): pred.sdiv.if, pred.sdiv.continue\r\n\r\n pred.sdiv.if:\r\n CLONE ir<%div24> = sdiv vp<%9>, ir<%conv6>\r\n Successor(s): pred.sdiv.continue\r\n\r\n pred.sdiv.continue:\r\n PHI-PREDICATED-INSTRUCTION vp<%11> = ir<%div24>\r\n No successors\r\n }\r\n Successor(s): omp.inner.for.body.0\r\n\r\n omp.inner.for.body.0:\r\n CLONE ir<%conv26> = trunc vp<%11>\r\n CLONE ir<%mul36> = mul nsw vp<%11>, ir<%conv6>\r\n CLONE ir<%sub37> = sub nsw vp<%9>, ir<%mul36>\r\n CLONE ir<%conv40> = trunc ir<%sub37>\r\n CLONE ir<%mul41> = mul nsw ir<%1>, ir<%conv26>\r\n CLONE ir<%add42> = add nsw ir<%mul41>, ir<%conv40>\r\n CLONE ir<%idxprom> = sext ir<%add42>\r\n CLONE ir<%arrayidx> = getelementptr inbounds ir<%6>, ir<%idxprom>\r\n WIDEN ir<%8> = load ir<%arrayidx>, vp<%7>\r\n WIDEN ir<%mul43> = fmul contract ir<%2>, ir<%8>\r\n CLONE ir<%arrayidx47> = getelementptr inbounds ir<%7>, ir<%idxprom>\r\n WIDEN ir<%9> = load ir<%arrayidx47>, vp<%7>\r\n WIDEN ir<%add48> = fadd contract ir<%mul43>, ir<%9>\r\n WIDEN store ir<%arrayidx47>, ir<%add48>, vp<%7>\r\n EMIT vp<%25> = VF * UF + vp<%6>\r\n EMIT vp<%26> = VF * Part + vp<%6>\r\n EMIT vp<%27> = active lane mask vp<%26>, vp<%3>\r\n EMIT vp<%28> = not vp<%27>\r\n EMIT branch-on-cond vp<%28>\r\n No successors\r\n}\r\nSuccessor(s): middle.block\r\n```\r\n\r\nVPRegion `pred.sdiv` fails at assertion\r\n```c++\r\nvoid VPRegionBlock::execute(VPTransformState *State) {\r\n...\r\n if (!isReplicator()) {\r\n...\r\n return;\r\n }\r\n...\r\n for (...) {\r\n assert(!State->VF.isScalable() && \"VF is assumed to be non scalable.\");\r\n }\r\n}\r\n```\r\nI think VPRegion `pred.sdiv` should have `isReplicator() == false` or `pred.sdiv` just should not exist.\r\n\r\n### second, I find VPRegion `pred.sdiv` is splitted from `WorkList`\r\n```c++\r\nstatic void addReplicateRegions(VPlan &Plan) {\r\n SmallVector WorkList;\r\n for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly(\r\n vp_depth_first_deep(Plan.getEntry()))) {\r\n for (VPRecipeBase &R : *VPBB)\r\n if (auto *RepR = dyn_cast(&R)) {\r\n if (RepR->isPredicated())\r\n WorkList.push_back(RepR);\r\n }\r\n }\r\n...\r\n}\r\n```\r\nby VPRecipe `CLONE ir<%div24> = sdiv ir<%.omp.iv.065>, ir<%conv6>, vp<%7>`, where the instruction is `%div24 = sdiv i64 %.omp.iv.065, %conv6`\r\nI think this VPRecipe should have `isPredicated() == false` here, so it will not be splitted.\r\n\r\n### third, I find this VPRecipe is created here\r\n```c++\r\nVPRecipeOrVPValueTy VPRecipeBuilder::handleReplication(...) {\r\n bool IsUniform = LoopVectorizationPlanner::getDecisionAndClampRange(\r\n [&](ElementCount VF) { return CM.isUniformAfterVectorization(I, VF); },\r\n Range);\r\n\r\n bool IsPredicated = CM.isPredicatedInst(I);\r\n...\r\n VPValue *BlockInMask = nullptr;\r\n if (!IsPredicated) {\r\n // Finalize the recipe for Instr, first if it is not predicated.\r\n LLVM_DEBUG(dbgs() << \"LV: Scalarizing:\" << *I << \"\\n\");\r\n } else {\r\n LLVM_DEBUG(dbgs() << \"LV: Scalarizing and predicating:\" << *I << \"\\n\");\r\n // Instructions marked for predication are replicated and a mask operand is\r\n // added initially. Masked replicate recipes will later be placed under an\r\n // if-then construct to prevent side-effects. Generate recipes to compute\r\n // the block mask for this region.\r\n BlockInMask = createBlockInMask(I->getParent(), Plan);\r\n }\r\n\r\n auto *Recipe = new VPReplicateRecipe(I, Plan.mapToVPValues(I->operands()),\r\n IsUniform, BlockInMask);\r\n return toVPRecipeResult(Recipe);\r\n}\r\n```\r\nI notice that instruction `I` (`%div24 = sdiv i64 %.omp.iv.065, %conv6`) do not need to vectorize, because function `isScalarAfterVectorization` returns true and `I` is just used to calculate the index.\r\nIt seems reasonable that a `scalar` instruction does not need `Predicated`.\r\n\r\n### So I try the following modification,\r\n```diff\r\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\r\nindex 907b8ce002e8..76a5704a61c5 100644\r\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\r\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\r\n@@ -9819,7 +9819,9 @@ VPRecipeOrVPValueTy VPRecipeBuilder::handleReplication(Instruction *I,\r\n [&](ElementCount VF) { return CM.isUniformAfterVectorization(I, VF); },\r\n Range);\r\n\r\n- bool IsPredicated = CM.isPredicatedInst(I);\r\n+ bool IsPredicated = LoopVectorizationPlanner::getDecisionAndClampRange(\r\n+ [&](ElementCount VF) { return CM.isPredicatedInst(I) && !CM.isScalarAfterVectorization(I, VF); },\r\n+ Range);\r\n\r\n // Even if the instruction is not marked as uniform, there are certain\r\n // intrinsic calls that can be effectively treated as such, so we check for\r\n```\r\n\r\nJust my guess, I'm not sure if it is correct direction."}, {"author": "eastB233", "body": "Ping @fhahn "}, {"author": "eastB233", "body": "I think I misunderstand something, and the changes above may be wrong.\r\n\r\nAnd I have another way. In instruction `%div24 = sdiv i64 %.omp.iv.065, %conv6`, `%conv6` is invariant in loop, so it seems we do not need `Predicated`.\r\n\r\nI try the following patch,\r\n```diff\r\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\r\nindex c7c19ef456c7..f294703e1529 100644\r\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\r\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\r\n@@ -3856,21 +3856,21 @@ bool LoopVectorizationCostModel::isPredicatedInst(Instruction *I) const {\r\n !Legal->blockNeedsPredication(I->getParent()))\r\n return false;\r\n return true;\r\n }\r\n case Instruction::UDiv:\r\n case Instruction::SDiv:\r\n case Instruction::SRem:\r\n case Instruction::URem:\r\n // TODO: We can use the loop-preheader as context point here and get\r\n // context sensitive reasoning\r\n- return !isSafeToSpeculativelyExecute(I);\r\n+ return !isSafeToSpeculativelyExecute(I) && !Legal->isInvariant(I->getOperand(1));\r\n case Instruction::Call:\r\n return Legal->isMaskRequired(I);\r\n }\r\n }\r\n\r\n std::pair\r\n LoopVectorizationCostModel::getDivRemSpeculationCost(Instruction *I,\r\n ElementCount VF) const {\r\n assert(I->getOpcode() == Instruction::UDiv ||\r\n I->getOpcode() == Instruction::SDiv ||\r\n\r\n```"}, {"author": "eastB233", "body": "ping @sdesmalen-arm @davemgreen @paulwalker-arm "}, {"author": "fhahn", "body": "@eastB233 unfortunately I don't think this change is correct, e.g. consider https://github.com/llvm/llvm-project/blob/967eba07549d64f15e7a91e798aa46214704f62b/llvm/test/Transforms/LoopVectorize/X86/divs-with-tail-folding.ll#L251 when the sdiv/udiv is executed conditionally in the loop"}]}, "verified": true} {"bug_id": "118467", "issue_url": "https://github.com/llvm/llvm-project/issues/118467", "bug_type": "miscompilation", "base_commit": "73731d6873b6fb0757c3065aaf2452eaccd0eebc", "knowledge_cutoff": "2024-12-03T10:44:21Z", "lit_test_dir": ["llvm/test/Transforms/AggressiveInstCombine"], "hints": {"fix_commit": "f68b0e36997322eeda8fd199ea80deb1b49c5410", "components": ["AggressiveInstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp": [[803, 810]]}, "bug_location_funcname": {"llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp": ["foldConsecutiveLoads"]}}, "patch": "commit f68b0e36997322eeda8fd199ea80deb1b49c5410\nAuthor: Antonio Frighetto \nDate: Wed Dec 4 10:15:11 2024 +0100\n\n [AggressiveInstCombine] Use APInt and avoid truncation when folding loads\n \n A miscompilation issue has been addressed with improved handling.\n \n Fixes: https://github.com/llvm/llvm-project/issues/118467.\n\ndiff --git a/llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp b/llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp\nindex b5b561797f75..45ee2d472a11 100644\n--- a/llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp\n+++ b/llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp\n@@ -803,8 +803,7 @@ static bool foldConsecutiveLoads(Instruction &I, const DataLayout &DL,\n APInt Offset1(DL.getIndexTypeSizeInBits(Load1Ptr->getType()), 0);\n Load1Ptr = Load1Ptr->stripAndAccumulateConstantOffsets(\n DL, Offset1, /* AllowNonInbounds */ true);\n- Load1Ptr = Builder.CreatePtrAdd(Load1Ptr,\n- Builder.getInt32(Offset1.getZExtValue()));\n+ Load1Ptr = Builder.CreatePtrAdd(Load1Ptr, Builder.getInt(Offset1));\n }\n // Generate wider load.\n NewLoad = Builder.CreateAlignedLoad(WiderType, Load1Ptr, LI1->getAlign(),\n", "tests": [{"file": "llvm/test/Transforms/AggressiveInstCombine/X86/or-load.ll", "commands": ["opt < %s -passes=aggressive-instcombine -mtriple x86_64-none-eabi -mattr=avx2 -data-layout=\"e-n64\" -S", "opt < %s -passes=aggressive-instcombine -mtriple x86_64-none-eabi -mattr=avx2 -data-layout=\"E-n64\" -S"], "tests": [{"test_name": "loadcombine_consecutive_idx_64", "test_body": "define i32 @loadcombine_consecutive_idx_64(ptr %data) {\nentry:\n %arrayidx = getelementptr inbounds nuw i8, ptr %data, i64 2149675577\n %val = load i8, ptr %arrayidx, align 1\n %conv = zext i8 %val to i32\n %arrayidx.2 = getelementptr inbounds nuw i8, ptr %data, i64 2149675576\n %val.2 = load i8, ptr %arrayidx.2, align 1\n %conv.2 = zext i8 %val.2 to i32\n %shl = shl nuw nsw i32 %conv, 8\n %or = or disjoint i32 %shl, %conv.2\n ret i32 %or\n}\n"}]}], "issue": {"title": "Invalid LLVM IR code generated on x86-64 with a very simple sample (crash the generated executable code)", "body": "Below is a very simple program that's loading 16 bits table value into a 32 bits variable from a 8 bits table.\nOn **x86-64** with **-O2** or **-O3** flags, this program works fine with LLVM 16.0.0 and below, but since LLVM 17.0.0, the LLVM IR and the generated assembly code are using a wrong calculated offset on the table (doesn't occurs on LLVM armv8 whatever the version).\nAs you can see below, the offset in the table which is supposed to be **2149675576** (**0x80217238**) has its 32 upper bits inverted for **-2145291720** (**0xFFFFFFFF80217238**).\n\n```c\n\n#include \n\n// Load 16 bits into 32 bits value from data offset 0x80217238\n\nuint32_t Test(const uint8_t* data)\n{ \n uint32_t a, b, c; \n\tb = 0xFFFF8022 << 16;\n b += 0xFFFFE808; \n a = data[b + 0xFFFF8A31];\n c = data[b + 0xFFFF8A30];\n c &= ~0x0000FF00;\n c |= ((a << 8) | (a >> 24)) & 0x0000FF00;\n\treturn c;\n}\n\n```\n\nLLVM IR with LLVM 16.0.0 (from Compiler Explorer)\n```ll\ndefine dso_local noundef i32 @Test(unsigned char const*)(ptr nocapture noundef readonly %data) local_unnamed_addr {\nentry:\n %arrayidx4 = getelementptr inbounds i8, ptr %data, i64 2149675576\n %0 = load i16, ptr %arrayidx4, align 1\n %1 = zext i16 %0 to i32\n ret i32 %1\n}\n\ndeclare void @llvm.dbg.value(metadata, metadata, metadata) #1\n```\nLLVM IR with LLVM 19.1.0 and trunk (from Compiler Explorer)\n```ll\ndefine dso_local noundef range(i32 0, 65536) i32 @Test(unsigned char const*)(ptr nocapture noundef readonly %data) local_unnamed_addr {\nentry:\n %0 = getelementptr i8, ptr %data, i64 -2145291720\n %1 = load i16, ptr %0, align 1\n %2 = zext i16 %1 to i32\n ret i32 %2\n}\n```\n\n\n\n\n\n", "author": "dje64240", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "topperc", "body": "opt-bisect suggests AggressiveInstCominer. There were two separate loads that got merged. The resulting GEP had an index of `i32 -2145291720` with trunk and `i64 2149675576` with clang 16. Those would be the same number if GEP zero extended indices instead of sign extending."}, {"author": "antoniofrighetto", "body": "Alive2: https://alive2.llvm.org/ce/z/G5CgkD."}, {"author": "AZero13", "body": "> opt-bisect suggests AggressiveInstCominer. There were two separate loads that got merged. The resulting GEP had an index of `i32 -2145291720` with trunk and `i64 2149675576` with clang 16. Those would be the same number if GEP zero extended indices instead of sign extending.\n\nSee my PR for 19.x "}]}, "verified": true} {"bug_id": "104714", "issue_url": "https://github.com/llvm/llvm-project/issues/104714", "bug_type": "crash", "base_commit": "a2d8743cc86f96f6b1cbd85798328bd3fb2bf4de", "knowledge_cutoff": "2024-08-18T15:54:27Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "cb4efe1d078144a72306732a56afea3885650e8d", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[7237, 7242], [7292, 7298]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationPlanner::computeBestVF", "LoopVectorizationPlanner::cost"]}}, "patch": "commit cb4efe1d078144a72306732a56afea3885650e8d\nAuthor: Florian Hahn \nDate: Thu Aug 22 21:38:06 2024 +0100\n\n [VPlan] Don't trigger VF assertion if VPlan has extra simplifications.\n \n There are cases where VPlans contain some simplifications that are very\n hard to accurately account for up-front in the legacy cost model. Those\n cases are caused by un-simplified inputs, which trigger the assert\n ensuring both the legacy and VPlan-based cost model agree on the VF.\n \n To avoid false positives due to missed simplifications in general, only\n trigger the assert if the chosen VPlan doesn't contain any additional\n simplifications.\n \n Fixes https://github.com/llvm/llvm-project/issues/104714.\n Fixes https://github.com/llvm/llvm-project/issues/105713.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 8e9324ba718b..32e8f3312578 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -7237,6 +7237,56 @@ InstructionCost LoopVectorizationPlanner::cost(VPlan &Plan,\n return Cost;\n }\n \n+/// Return true if the original loop \\ TheLoop contains any instructions that do\n+/// not have corresponding recipes in \\p Plan and are not marked to be ignored\n+/// in \\p CostCtx. This means the VPlan contains simplification that the legacy\n+/// cost-model did not account for.\n+static bool\n+planContainsAdditionalSimplifications(VPlan &Plan, ElementCount VF,\n+ VPCostContext &CostCtx, Loop *TheLoop,\n+ LoopVectorizationCostModel &CM) {\n+ // First collect all instructions for the recipes in Plan.\n+ auto GetInstructionForCost = [](const VPRecipeBase *R) -> Instruction * {\n+ if (auto *S = dyn_cast(R))\n+ return dyn_cast_or_null(S->getUnderlyingValue());\n+ if (auto *WidenMem = dyn_cast(R))\n+ return &WidenMem->getIngredient();\n+ return nullptr;\n+ };\n+\n+ DenseSet SeenInstrs;\n+ auto Iter = vp_depth_first_deep(Plan.getEntry());\n+ for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly(Iter)) {\n+ for (VPRecipeBase &R : *VPBB) {\n+ if (auto *IR = dyn_cast(&R)) {\n+ auto *IG = IR->getInterleaveGroup();\n+ unsigned NumMembers = IG->getNumMembers();\n+ for (unsigned I = 0; I != NumMembers; ++I) {\n+ if (Instruction *M = IG->getMember(I))\n+ SeenInstrs.insert(M);\n+ }\n+ continue;\n+ }\n+ if (Instruction *UI = GetInstructionForCost(&R))\n+ SeenInstrs.insert(UI);\n+ }\n+ }\n+\n+ // Return true if the loop contains any instructions that are not also part of\n+ // the VPlan or are skipped for VPlan-based cost computations. This indicates\n+ // that the VPlan contains extra simplifications.\n+ return any_of(\n+ TheLoop->blocks(), [&SeenInstrs, VF, &CostCtx, &CM](BasicBlock *BB) {\n+ return any_of(*BB, [&SeenInstrs, VF, &CostCtx, &CM](Instruction &I) {\n+ if (isa(&I))\n+ return false;\n+ return !SeenInstrs.contains(&I) &&\n+ !CostCtx.skipCostComputation(&I, true) &&\n+ !CM.canTruncateToMinimalBitwidth(&I, VF);\n+ });\n+ });\n+}\n+\n VectorizationFactor LoopVectorizationPlanner::computeBestVF() {\n if (VPlans.empty())\n return VectorizationFactor::Disabled();\n@@ -7292,7 +7342,20 @@ VectorizationFactor LoopVectorizationPlanner::computeBestVF() {\n // cost-model and will be retired once the VPlan-based cost-model is\n // stabilized.\n VectorizationFactor LegacyVF = selectVectorizationFactor();\n- assert(BestFactor.Width == LegacyVF.Width &&\n+ VPlan &BestPlan = getPlanFor(BestFactor.Width);\n+\n+ // Pre-compute the cost and use it to check if BestPlan contains any\n+ // simplifications not accounted for in the legacy cost model. If that's the\n+ // case, don't trigger the assertion, as the extra simplifications may cause a\n+ // different VF to be picked by the VPlan-based cost model.\n+ LLVMContext &LLVMCtx = OrigLoop->getHeader()->getContext();\n+ VPCostContext CostCtx(CM.TTI, *CM.TLI, Legal->getWidestInductionType(),\n+ LLVMCtx, CM);\n+ precomputeCosts(BestPlan, BestFactor.Width, CostCtx);\n+ assert((BestFactor.Width == LegacyVF.Width ||\n+ planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width),\n+ BestFactor.Width, CostCtx,\n+ OrigLoop, CM)) &&\n \" VPlan cost model and legacy cost model disagreed\");\n assert((BestFactor.Width.isScalar() || BestFactor.ScalarCost > 0) &&\n \"when vectorizing, the scalar cost must be computed.\");\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/blocks-with-dead-instructions.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "empty_block_with_phi_2", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @empty_block_with_phi_2(ptr %src, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i32 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i32 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %loop.latch, label %else\n\nelse: ; preds = %loop.header\n br label %loop.latch\n\nloop.latch: ; preds = %else, %loop.header\n %p = phi i16 [ %l, %loop.header ], [ 99, %else ]\n store i16 %p, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 1\n %0 = icmp eq i64 %iv.next, %N\n br i1 %0, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "dead_load_in_block", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @dead_load_in_block(ptr %dst, ptr %src, i8 %N, i64 %x) #0 {\nentry:\n %N.ext = zext i8 %N to i64\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %l.0 = load i32, ptr %src, align 4\n %c.0 = icmp eq i32 %l.0, 0\n br i1 %c.0, label %loop.latch, label %then\n\nthen: ; preds = %loop.header\n %gep.src.x = getelementptr i32, ptr %src, i64 %x\n %l.dead = load i32, ptr %gep.src.x, align 4\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %gep.dst = getelementptr i32, ptr %dst, i64 %iv\n store i32 0, ptr %gep.dst, align 4\n %iv.next = add i64 %iv, 3\n %cmp = icmp ult i64 %iv, %N.ext\n br i1 %cmp, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "multiple_blocks_with_dead_insts_3", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @multiple_blocks_with_dead_insts_3(ptr %src) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 0, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %loop.header\n %dead.gep.1 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nelse: ; preds = %loop.header\n %dead.gep.2 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else, %then\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "multiple_blocks_with_dead_insts_4", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @multiple_blocks_with_dead_insts_4(ptr %src, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %loop.header\n br label %then.1\n\nthen.1: ; preds = %then\n %dead.gep.1 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nelse: ; preds = %loop.header\n %dead.gep.2 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else, %then.1\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "empty_block_with_phi_1", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @empty_block_with_phi_1(ptr %src, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i32 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i32 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %p = phi i16 [ %l, %loop.header ], [ 99, %then ]\n store i16 %p, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 1\n %0 = icmp eq i64 %iv.next, %N\n br i1 %0, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "block_with_dead_inst_2", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @block_with_dead_inst_2(ptr %src) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 0, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %loop.latch, label %else\n\nelse: ; preds = %loop.header\n %dead.gep = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else, %loop.header\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "multiple_blocks_with_dead_inst_multiple_successors_6", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @multiple_blocks_with_dead_inst_multiple_successors_6(ptr %src, i1 %ic, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %loop.header\n br i1 %ic, label %then.1, label %else\n\nthen.1: ; preds = %then\n %dead.gep.1 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nelse: ; preds = %then, %loop.header\n br label %else.2\n\nelse.2: ; preds = %else\n %dead.gep.2 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else.2, %then.1\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "multiple_blocks_with_dead_inst_multiple_successors_5", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @multiple_blocks_with_dead_inst_multiple_successors_5(ptr %src) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %loop.header\n br label %then.1\n\nthen.1: ; preds = %then\n %dead.gep.1 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nelse: ; preds = %loop.header\n br label %else.2\n\nelse.2: ; preds = %else\n %dead.gep.2 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else.2, %then.1\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "block_with_dead_inst_1", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @block_with_dead_inst_1(ptr %src, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n %dead.gep = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %0 = icmp eq i64 %iv.next, %N\n br i1 %0, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}]}], "issue": {"title": "[VPlan] Assertion `VF.Width == BestVF && \"VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Tested using d082f1f37d8cb7a0c6875537ba873a631b154d53\r\n\r\nC Testcase:\r\n```c\r\nint a;\r\nextern int b[];\r\nsigned char c;\r\nint d[1];\r\nvoid e() {\r\n for (short f = 2; f < c; f += 3) {\r\n a = ({ d[f - 1]; }) ? d[f] : 0;\r\n b[f] = 0;\r\n }\r\n}\r\n```\r\nCommand:\r\n```\r\n/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/clang -march=rv64gcv_zvl256b -O3 red.c -o /dev/null\r\n```\r\n\r\nTestcase:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @e(i8 %0, i64 %indvars.iv) #0 {\r\nentry:\r\n %1 = zext i8 %0 to i64\r\n br label %for.body\r\n\r\nfor.cond.for.cond.cleanup_crit_edge: ; preds = %cond.end\r\n ret void\r\n\r\nfor.body: ; preds = %cond.end, %entry\r\n %indvars.iv1 = phi i64 [ 0, %entry ], [ %indvars.iv.next, %cond.end ]\r\n %sub = add i64 0, 0\r\n %arrayidx = getelementptr [1 x i32], ptr null, i64 0, i64 %sub\r\n %2 = load i32, ptr %arrayidx, align 4\r\n %tobool.not = icmp eq i32 %2, 0\r\n br i1 %tobool.not, label %cond.end, label %cond.true\r\n\r\ncond.true: ; preds = %for.body\r\n %arrayidx5 = getelementptr [1 x i32], ptr null, i64 0, i64 %indvars.iv\r\n %3 = load i32, ptr %arrayidx5, align 4\r\n br label %cond.end\r\n\r\ncond.end: ; preds = %cond.true, %for.body\r\n %arrayidx7 = getelementptr [0 x i32], ptr null, i64 0, i64 %indvars.iv1\r\n store i32 0, ptr %arrayidx7, align 4\r\n %indvars.iv.next = add i64 %indvars.iv1, 3\r\n %cmp = icmp ult i64 %indvars.iv1, %1\r\n br i1 %cmp, label %for.body, label %for.cond.for.cond.cleanup_crit_edge\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl256b,+zvl32b,+zvl64b,-b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10037: bool llvm::LoopVectorizePass::processLoop(llvm::Loop*): Assertion `VF.Width == BestVF && \"VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"e\"\r\n #0 0x000063e889a673c0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2ff63c0)\r\n #1 0x000063e889a647df llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2ff37df)\r\n #2 0x000063e889a64935 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007a5322242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007a53222969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007a53222969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007a53222969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007a5322242476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007a53222287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007a532222871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007a5322239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x000063e888a5ba9a llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1feaa9a)\r\n#12 0x000063e888a5df51 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fecf51)\r\n#13 0x000063e888a5e5c7 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fed5c7)\r\n#14 0x000063e8878c6096 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe55096)\r\n#15 0x000063e88987deef llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e0ceef)\r\n#16 0x000063e8878c91d6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe581d6)\r\n#17 0x000063e88987cc6b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e0bc6b)\r\n#18 0x000063e8878c54c6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe544c6)\r\n#19 0x000063e88987accd llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e09ccd)\r\n#20 0x000063e8870f3eb6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x682eb6)\r\n#21 0x000063e8870e5c41 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x674c41)\r\n#22 0x00007a5322229d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#23 0x00007a5322229e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#24 0x00007a5322229e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#25 0x000063e8870dbff5 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x66aff5)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\nGodbolt: https://godbolt.org/z/v1nbzWqMb\r\n\r\nFound via fuzzer.\r\n\r\ncc @fhahn ", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": [{"author": "patrick-rivos", "body": "Linking some meta-discussion about the assert here: https://github.com/llvm/llvm-project/issues/100591"}, {"author": "zhendongsu", "body": "Some additional reproducers on x86_64-linux-gnu in case they are of help:\r\n\r\n```\r\n[519] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 410f751144e8b2e9574f03e0d0fb8560fe3cb797)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[520] % \r\n[520] % clangtk -Os small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7288: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `BestFactor.Width == LegacyVF.Width && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/WAIT/20240822-clangtk-m64-Os-Wall-Wextra-pipe-fPIC-build-182745/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/WAIT/20240822-clangtk-m64-Os-Wall-Wextra-pipe-fPIC-build-182745/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -Os -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-e51115.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4.\tRunning pass \"loop-vectorize\" on function \"main\"\r\n #0 0x00005559685ec3ef llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x42ea3ef)\r\n #1 0x00005559685e9934 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f0380f76420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007f03809ad00b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007f038098c859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007f038098c729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007f038098c729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007f038099dfd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x0000555969ef4948 llvm::LoopVectorizationPlanner::computeBestVF() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5bf2948)\r\n #9 0x0000555969f0f9ea llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c0d9ea)\r\n#10 0x0000555969f12691 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c10691)\r\n#11 0x0000555969f12d43 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c10d43)\r\n#12 0x0000555969a9d326 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x579b326)\r\n#13 0x0000555967f872a1 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c852a1)\r\n#14 0x0000555965ae2ae6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x17e0ae6)\r\n#15 0x0000555967f85bed llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c83bed)\r\n#16 0x0000555965ae19e6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x17df9e6)\r\n#17 0x0000555967f83cf1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c81cf1)\r\n#18 0x00005559688a0c29 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#19 0x00005559688a3d83 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x45a1d83)\r\n#20 0x0000555968f440fa clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4c420fa)\r\n#21 0x000055596ad8938c clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x6a8738c)\r\n#22 0x0000555968f44598 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4c42598)\r\n#23 0x00005559691dec59 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4edcc59)\r\n#24 0x00005559691603be clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4e5e3be)\r\n#25 0x00005559692cb886 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4fc9886)\r\n#26 0x00005559656b5dd5 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13b3dd5)\r\n#27 0x00005559656aee7a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#28 0x00005559656b284e clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13b084e)\r\n#29 0x00005559655a31bb main (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x12a11bb)\r\n#30 0x00007f038098e083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#31 0x00005559656ae90e _start (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13ac90e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 410f751144e8b2e9574f03e0d0fb8560fe3cb797)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-1b0f28.c\r\nclangtk: note: diagnostic msg: /tmp/small-1b0f28.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[521] % \r\n[521] % cat small.c\r\nchar a;\r\nint b, *c = &b, d[3], e;\r\nint main() {\r\n int f, g[3] = {0,0,0};\r\n g[e] = *c = 0;\r\n for (; *c < 3; (*c)++) {\r\n f = a ? b % a : 0;\r\n g[2] &= d[*c] & f;\r\n }\r\n return 0;\r\n}\r\n```\r\n\r\n```\r\n[537] % clangtk -O3 small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7288: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `BestFactor.Width == LegacyVF.Width && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/WAIT/20240814-clangtk-m64-O3-build-085604/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/WAIT/20240814-clangtk-m64-O3-build-085604/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O3 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-83d1cf.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4.\tRunning pass \"loop-vectorize\" on function \"main\"\r\n #0 0x0000558427a123ef llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x42ea3ef)\r\n #1 0x0000558427a0f934 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007fa32f70a420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007fa32f14100b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007fa32f120859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007fa32f120729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007fa32f120729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007fa32f131fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x000055842931a948 llvm::LoopVectorizationPlanner::computeBestVF() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5bf2948)\r\n #9 0x00005584293359ea llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c0d9ea)\r\n#10 0x0000558429338691 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c10691)\r\n#11 0x0000558429338d43 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c10d43)\r\n#12 0x0000558428ec3326 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x579b326)\r\n#13 0x00005584273ad2a1 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c852a1)\r\n#14 0x0000558424f08ae6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x17e0ae6)\r\n#15 0x00005584273abbed llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c83bed)\r\n#16 0x0000558424f079e6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x17df9e6)\r\n#17 0x00005584273a9cf1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c81cf1)\r\n#18 0x0000558427cc6c29 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#19 0x0000558427cc9d83 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x45a1d83)\r\n#20 0x000055842836a0fa clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4c420fa)\r\n#21 0x000055842a1af38c clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x6a8738c)\r\n#22 0x000055842836a598 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4c42598)\r\n#23 0x0000558428604c59 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4edcc59)\r\n#24 0x00005584285863be clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4e5e3be)\r\n#25 0x00005584286f1886 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4fc9886)\r\n#26 0x0000558424adbdd5 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13b3dd5)\r\n#27 0x0000558424ad4e7a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#28 0x0000558424ad884e clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13b084e)\r\n#29 0x00005584249c91bb main (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x12a11bb)\r\n#30 0x00007fa32f122083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#31 0x0000558424ad490e _start (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13ac90e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 410f751144e8b2e9574f03e0d0fb8560fe3cb797)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-aeee73.c\r\nclangtk: note: diagnostic msg: /tmp/small-aeee73.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[538] % \r\n[538] % cat small.c\r\nint a[1], *d, e, f[1];\r\nstatic int b, c, *volatile g = &c;\r\nvoid h(int i) { b = b >> 1 ^ a[b ^ i & 1]; }\r\nvoid j(long i) {\r\n b = b >> 1 ^ b;\r\n b = b >> 1 ^ a[b ^ 1];\r\n h(i >> 1);\r\n h(i >> 3);\r\n h(i >> 2);\r\n h(i >> 4);\r\n h(i >> 5);\r\n}\r\nstatic void m() {\r\n int l;\r\n while (e) {\r\n c = 0;\r\n for (; c < 5; c++) {\r\n l = 0;\r\n for (; l < 7; l++)\r\n if (*d)\r\n j(f[c * l]);\r\n }\r\n }\r\n}\r\nint main() {\r\n if (e) {\r\n m();\r\n g;\r\n }\r\n return 0;\r\n}\r\n```\r\n"}, {"author": "fhahn", "body": "@zhendongsu thanks for the report, it was the same issue as the test case @sjoerdmeijer shared at https://github.com/llvm/llvm-project/issues/100591, which was a different underlying issue to the one for this issue.\r\n\r\nBoth should be fixed now, please let me know if you find any other instances of this triggering."}, {"author": "zhendongsu", "body": "Hi @fhahn, here is another related instance that still triggers:\r\n```\r\n[512] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git f22b1da8791edd557ce34c87190e329df2e1c892)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[513] % \r\n[513] % clangtk -Os small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7359: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), BestFactor.Width, CostCtx, OrigLoop, CM)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20240825-clangtk-m64-Os-Wall-Wextra-pipe-fPIC-build-043108/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20240825-clangtk-m64-Os-Wall-Wextra-pipe-fPIC-build-043108/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -Os -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-d7ba19.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4.\tRunning pass \"loop-vectorize\" on function \"main\"\r\n #0 0x0000564ef7a4babf llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x42ddabf)\r\n #1 0x0000564ef7a49004 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f89634f9420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007f8962f3000b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007f8962f0f859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007f8962f0f729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007f8962f0f729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007f8962f20fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x0000564ef9358b42 llvm::LoopVectorizationPlanner::computeBestVF() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5beab42)\r\n #9 0x0000564ef937504a llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c0704a)\r\n#10 0x0000564ef9377c81 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c09c81)\r\n#11 0x0000564ef9378333 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c0a333)\r\n#12 0x0000564ef8f026e6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x57946e6)\r\n#13 0x0000564ef73e68d1 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c788d1)\r\n#14 0x0000564ef4f4ff46 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x17e1f46)\r\n#15 0x0000564ef73e521d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c7721d)\r\n#16 0x0000564ef4f4ee46 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x17e0e46)\r\n#17 0x0000564ef73e3321 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c75321)\r\n#18 0x0000564ef7d04bb1 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#19 0x0000564ef7d07d73 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4599d73)\r\n#20 0x0000564ef83a840c clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4c3a40c)\r\n#21 0x0000564efa1f004c clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x6a8204c)\r\n#22 0x0000564ef83a8828 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4c3a828)\r\n#23 0x0000564ef8643cb9 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4ed5cb9)\r\n#24 0x0000564ef85c543e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4e5743e)\r\n#25 0x0000564ef8730876 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4fc2876)\r\n#26 0x0000564ef4b21e55 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13b3e55)\r\n#27 0x0000564ef4b1aefa ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#28 0x0000564ef4b1e8ce clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13b08ce)\r\n#29 0x0000564ef4a0f1bb main (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x12a11bb)\r\n#30 0x00007f8962f11083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#31 0x0000564ef4b1a98e _start (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13ac98e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git f22b1da8791edd557ce34c87190e329df2e1c892)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-6fc973.c\r\nclangtk: note: diagnostic msg: /tmp/small-6fc973.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[514] % \r\n[514] % cat small.c\r\nint a, d;\r\nstatic int b = 1, c;\r\nvoid e() {\r\n while (a)\r\n b--;\r\n}\r\nint main() {\r\n for (d = 0; d < 3; d++)\r\n if (a)\r\n for (c = 0; c < 2; c++)\r\n if (b)\r\n break;\r\n return 0;\r\n}\r\n```\r\n\r\nCompiler Explorer: https://godbolt.org/z/zanjboPYo"}, {"author": "zhendongsu", "body": "Hi @fhahn, another reproducer that triggers at -O{s,2,3}:\r\n```\r\n[520] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 7e6b1504c7cf6976ac8e9012c4513ffa258bd8eb)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[521] % \r\n[521] % clangtk -O3 small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7359: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), BestFactor.Width, CostCtx, OrigLoop, CM)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20240826-clangtk-m64-Os-build-191303/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20240826-clangtk-m64-Os-build-191303/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O3 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-bd6e8d.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4.\tRunning pass \"loop-vectorize\" on function \"main\"\r\n #0 0x0000561e24237f5f llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x42f4f5f)\r\n #1 0x0000561e242354a4 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f7461dc8420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007f74617ff00b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007f74617de859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007f74617de729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007f74617de729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007f74617effd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x0000561e25b45232 llvm::LoopVectorizationPlanner::computeBestVF() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c02232)\r\n #9 0x0000561e25b616ca llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c1e6ca)\r\n#10 0x0000561e25b64301 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c21301)\r\n#11 0x0000561e25b649b3 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5c219b3)\r\n#12 0x0000561e256ef0a6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x57ac0a6)\r\n#13 0x0000561e23bd2da1 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c8fda1)\r\n#14 0x0000561e21724f86 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x17e1f86)\r\n#15 0x0000561e23bd16ed llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c8e6ed)\r\n#16 0x0000561e21723e86 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x17e0e86)\r\n#17 0x0000561e23bcf7f1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3c8c7f1)\r\n#18 0x0000561e244f0f41 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#19 0x0000561e244f4103 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x45b1103)\r\n#20 0x0000561e24b9480c clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4c5180c)\r\n#21 0x0000561e269dcc5c clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x6a99c5c)\r\n#22 0x0000561e24b94c28 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4c51c28)\r\n#23 0x0000561e24e30189 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4eed189)\r\n#24 0x0000561e24db162e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4e6e62e)\r\n#25 0x0000561e24f1cde6 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4fd9de6)\r\n#26 0x0000561e212f6ed5 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13b3ed5)\r\n#27 0x0000561e212eff7a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#28 0x0000561e212f394e clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13b094e)\r\n#29 0x0000561e211e41eb main (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x12a11eb)\r\n#30 0x00007f74617e0083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#31 0x0000561e212efa0e _start (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13aca0e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 7e6b1504c7cf6976ac8e9012c4513ffa258bd8eb)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-796071.c\r\nclangtk: note: diagnostic msg: /tmp/small-796071.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[522] % \r\n[522] % cat small.c\r\nchar a;\r\nstatic char b, *c = &b, **d = &c;\r\nint e;\r\nunsigned f;\r\nchar g(int h) { return a << h; }\r\nint main() {\r\n char ***i[2] = {&d, &d};\r\n int *j;\r\n for (; f < 3; f++) {\r\n j = &e;\r\n *j = g(**d);\r\n }\r\n return 0;\r\n}\r\n```\r\n\r\nCompiler Explorer: https://godbolt.org/z/7Y5PzTx4P"}]}, "verified": true} {"bug_id": "97330", "issue_url": "https://github.com/llvm/llvm-project/issues/97330", "bug_type": "miscompilation", "base_commit": "fdcfb277465e4530e5837fb8a95031794e58cb9b", "knowledge_cutoff": "2024-07-01T17:37:19Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "b558ac0eef57a3737b1e27844115fa91e0b32582", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineInternal.h": [[545, 554]], "llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": [[91, 98], [124, 173], [1105, 1117], [1119, 1131]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": ["InstCombinerImpl::SimplifyDemandedBits", "InstCombinerImpl::SimplifyDemandedUseBits"]}}, "patch": "commit b558ac0eef57a3737b1e27844115fa91e0b32582\nAuthor: Nikita Popov \nDate: Tue Jul 2 10:08:45 2024 +0200\n\n [InstCombine] Fix context for multi-use demanded bits simplification\n \n When simplifying a multi-use root value, the demanded bits were\n reset to full, but we also need to reset the context extract. To\n make this convenient (without requiring by-value passing of\n SimplifyQuery), move the logic that that handles constants and\n dispatches to SimplifyDemandedUseBits/SimplifyMultipleUseDemandedBits\n into SimplifyDemandedBits. The SimplifyDemandedInstructionBits\n caller starts with full demanded bits and an appropriate context\n anyway.\n \n The different context instruction does mean that the ephemeral\n value protection no longer triggers in some cases, as the changes\n to assume tests show.\n \n An alternative, which I will explore in a followup, is to always\n use SimplifyMultipleUseDemandedBits() -- the previous root special\n case is only really intended for SimplifyDemandedInstructionBits(),\n which now no longer shares this code path.\n \n Fixes https://github.com/llvm/llvm-project/issues/97330.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\nindex 318c455fd7ef..64fbcc80e0ed 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\n@@ -545,10 +545,11 @@ public:\n ConstantInt *&Less, ConstantInt *&Equal,\n ConstantInt *&Greater);\n \n- /// Attempts to replace V with a simpler value based on the demanded\n+ /// Attempts to replace I with a simpler value based on the demanded\n /// bits.\n- Value *SimplifyDemandedUseBits(Value *V, APInt DemandedMask, KnownBits &Known,\n- unsigned Depth, const SimplifyQuery &Q);\n+ Value *SimplifyDemandedUseBits(Instruction *I, const APInt &DemandedMask,\n+ KnownBits &Known, unsigned Depth,\n+ const SimplifyQuery &Q);\n using InstCombiner::SimplifyDemandedBits;\n bool SimplifyDemandedBits(Instruction *I, unsigned Op,\n const APInt &DemandedMask, KnownBits &Known,\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\nindex 6cf2e71363ab..b1d03786f321 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n@@ -91,8 +91,47 @@ bool InstCombinerImpl::SimplifyDemandedBits(Instruction *I, unsigned OpNo,\n KnownBits &Known, unsigned Depth,\n const SimplifyQuery &Q) {\n Use &U = I->getOperandUse(OpNo);\n- Value *NewVal = SimplifyDemandedUseBits(U.get(), DemandedMask, Known,\n- Depth, Q);\n+ Value *V = U.get();\n+ if (isa(V)) {\n+ llvm::computeKnownBits(V, Known, Depth, Q);\n+ return false;\n+ }\n+\n+ Known.resetAll();\n+ if (DemandedMask.isZero()) {\n+ // Not demanding any bits from V.\n+ replaceUse(U, UndefValue::get(V->getType()));\n+ return true;\n+ }\n+\n+ if (Depth == MaxAnalysisRecursionDepth)\n+ return false;\n+\n+ Instruction *VInst = dyn_cast(V);\n+ if (!VInst) {\n+ llvm::computeKnownBits(V, Known, Depth, Q);\n+ return false;\n+ }\n+\n+ Value *NewVal;\n+ if (VInst->hasOneUse()) {\n+ // If the instruction has one use, we can directly simplify it.\n+ NewVal = SimplifyDemandedUseBits(VInst, DemandedMask, Known, Depth, Q);\n+ } else if (Depth != 0) {\n+ // If there are multiple uses of this instruction and we aren't at the root,\n+ // then we can simplify VInst to some other value, but not modify the\n+ // instruction.\n+ NewVal =\n+ SimplifyMultipleUseDemandedBits(VInst, DemandedMask, Known, Depth, Q);\n+ } else {\n+ // If this is the root being simplified, allow it to have multiple uses,\n+ // just set the DemandedMask to all bits and reset the context instruction.\n+ // This allows visitTruncInst (for example) to simplify the operand of a\n+ // trunc without duplicating all the SimplifyDemandedUseBits() logic.\n+ NewVal =\n+ SimplifyDemandedUseBits(VInst, APInt::getAllOnes(Known.getBitWidth()),\n+ Known, Depth, Q.getWithInstruction(VInst));\n+ }\n if (!NewVal) return false;\n if (Instruction* OpInst = dyn_cast(U))\n salvageDebugInfo(*OpInst);\n@@ -124,50 +163,21 @@ bool InstCombinerImpl::SimplifyDemandedBits(Instruction *I, unsigned OpNo,\n /// operands based on the information about what bits are demanded. This returns\n /// some other non-null value if it found out that V is equal to another value\n /// in the context where the specified bits are demanded, but not for all users.\n-Value *InstCombinerImpl::SimplifyDemandedUseBits(Value *V, APInt DemandedMask,\n+Value *InstCombinerImpl::SimplifyDemandedUseBits(Instruction *I,\n+ const APInt &DemandedMask,\n KnownBits &Known,\n unsigned Depth,\n const SimplifyQuery &Q) {\n- assert(V != nullptr && \"Null pointer of Value???\");\n+ assert(I != nullptr && \"Null pointer of Value???\");\n assert(Depth <= MaxAnalysisRecursionDepth && \"Limit Search Depth\");\n uint32_t BitWidth = DemandedMask.getBitWidth();\n- Type *VTy = V->getType();\n+ Type *VTy = I->getType();\n assert(\n (!VTy->isIntOrIntVectorTy() || VTy->getScalarSizeInBits() == BitWidth) &&\n Known.getBitWidth() == BitWidth &&\n \"Value *V, DemandedMask and Known must have same BitWidth\");\n \n- if (isa(V)) {\n- llvm::computeKnownBits(V, Known, Depth, Q);\n- return nullptr;\n- }\n-\n- Known.resetAll();\n- if (DemandedMask.isZero()) // Not demanding any bits from V.\n- return UndefValue::get(VTy);\n-\n- if (Depth == MaxAnalysisRecursionDepth)\n- return nullptr;\n-\n- Instruction *I = dyn_cast(V);\n- if (!I) {\n- llvm::computeKnownBits(V, Known, Depth, Q);\n- return nullptr; // Only analyze instructions.\n- }\n-\n- // If there are multiple uses of this value and we aren't at the root, then\n- // we can't do any simplifications of the operands, because DemandedMask\n- // only reflects the bits demanded by *one* of the users.\n- if (Depth != 0 && !I->hasOneUse())\n- return SimplifyMultipleUseDemandedBits(I, DemandedMask, Known, Depth, Q);\n-\n KnownBits LHSKnown(BitWidth), RHSKnown(BitWidth);\n- // If this is the root being simplified, allow it to have multiple uses,\n- // just set the DemandedMask to all bits so that we can try to simplify the\n- // operands. This allows visitTruncInst (for example) to simplify the\n- // operand of a trunc without duplicating all the logic below.\n- if (Depth == 0 && !V->hasOneUse())\n- DemandedMask.setAllBits();\n \n // Update flags after simplifying an operand based on the fact that some high\n // order bits are not demanded.\n@@ -1105,13 +1115,13 @@ Value *InstCombinerImpl::SimplifyDemandedUseBits(Value *V, APInt DemandedMask,\n }\n \n if (!KnownBitsComputed)\n- llvm::computeKnownBits(V, Known, Depth, Q);\n+ llvm::computeKnownBits(I, Known, Depth, Q);\n break;\n }\n }\n \n- if (V->getType()->isPointerTy()) {\n- Align Alignment = V->getPointerAlignment(DL);\n+ if (I->getType()->isPointerTy()) {\n+ Align Alignment = I->getPointerAlignment(DL);\n Known.Zero.setLowBits(Log2(Alignment));\n }\n \n@@ -1119,13 +1129,14 @@ Value *InstCombinerImpl::SimplifyDemandedUseBits(Value *V, APInt DemandedMask,\n // constant. We can't directly simplify pointers as a constant because of\n // pointer provenance.\n // TODO: We could return `(inttoptr const)` for pointers.\n- if (!V->getType()->isPointerTy() && DemandedMask.isSubsetOf(Known.Zero | Known.One))\n+ if (!I->getType()->isPointerTy() &&\n+ DemandedMask.isSubsetOf(Known.Zero | Known.One))\n return Constant::getIntegerValue(VTy, Known.One);\n \n if (VerifyKnownBits) {\n- KnownBits ReferenceKnown = llvm::computeKnownBits(V, Depth, Q);\n+ KnownBits ReferenceKnown = llvm::computeKnownBits(I, Depth, Q);\n if (Known != ReferenceKnown) {\n- errs() << \"Mismatched known bits for \" << *V << \" in \"\n+ errs() << \"Mismatched known bits for \" << *I << \" in \"\n << I->getFunction()->getName() << \"\\n\";\n errs() << \"computeKnownBits(): \" << ReferenceKnown << \"\\n\";\n errs() << \"SimplifyDemandedBits(): \" << Known << \"\\n\";\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/assume-inseltpoison.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "PR40940", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\ndefine i32 @PR40940(<4 x i8> %x) {\n %shuf = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> \n %t2 = bitcast <4 x i8> %shuf to i32\n %t3 = icmp ult i32 %t2, 65536\n call void @llvm.assume(i1 %t3)\n ret i32 %t2\n}\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}]}, {"file": "llvm/test/Transforms/InstCombine/assume.ll", "commands": ["opt < %s -passes=instcombine -S", "opt < %s -passes=instcombine --enable-knowledge-retention -S", "opt < %s -passes=instcombine -S --try-experimental-debuginfo-iterators", "opt < %s -passes=instcombine --enable-knowledge-retention -S --try-experimental-debuginfo-iterators"], "tests": [{"test_name": "PR40940", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\ndefine i32 @PR40940(<4 x i8> %x) {\n %shuf = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> \n %t2 = bitcast <4 x i8> %shuf to i32\n %t3 = icmp ult i32 %t2, 65536\n call void @llvm.assume(i1 %t3)\n ret i32 %t2\n}\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n\n!llvm.dbg.cu = !{}\n!llvm.module.flags = !{!0, !1, !2, !3}\n\n!0 = !{i32 2, !\"Dwarf Version\", i32 4}\n!1 = !{i32 2, !\"Debug Info Version\", i32 3}\n!2 = !{i32 1, !\"wchar_size\", i32 4}\n!3 = !{i32 8, !\"PIC Level\", i32 2}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/known-bits.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "simplifydemanded_context", "test_body": "define i8 @simplifydemanded_context(i8 %x, i8 %y) {\n %and1 = and i8 %x, 1\n call void @dummy()\n %x.lobits = and i8 %x, 3\n %precond = icmp eq i8 %x.lobits, 0\n call void @llvm.assume(i1 %precond)\n %and2 = and i8 %and1, %y\n ret i8 %and2\n}\n\ndeclare void @dummy()\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "pr97330", "test_body": "define i16 @pr97330(i1 %c, ptr %p1, ptr %p2) {\nentry:\n %v = load i64, ptr %p1, align 8\n %conv = trunc i64 %v to i16\n br i1 %c, label %exit, label %if\n\nif: ; preds = %entry\n %cmp = icmp ne i16 %conv, 1\n %conv2 = zext i1 %cmp to i32\n store i32 %conv2, ptr %p2, align 4\n %cmp2 = icmp eq i64 %v, 1\n call void @llvm.assume(i1 %cmp2)\n unreachable\n\nexit: ; preds = %entry\n ret i16 %conv\n}\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}]}], "issue": {"title": "[InstCombine] Miscompilation with llvm.assume in unreachable blocks", "body": "Reproducer: https://alive2.llvm.org/ce/z/zKvpXV\r\n```\r\n; bin/opt -passes=instcombine reduced.ll -S\r\ndefine i16 @src(i16 %g, ptr %e, ptr %d) {\r\nentry:\r\n %0 = load i64, ptr %d, align 8\r\n %conv = trunc i64 %0 to i16\r\n %tobool.not.i = icmp eq i16 %g, 0\r\n br i1 %tobool.not.i, label %i.exit, label %for.cond.preheader.i\r\n\r\nfor.cond.preheader.i:\r\n %cmp5.i = icmp ne i16 %g, %conv\r\n %conv6.i = zext i1 %cmp5.i to i32\r\n store i32 %conv6.i, ptr %e, align 4\r\n %cmp7.i = icmp eq i64 %0, 1\r\n call void @llvm.assume(i1 %cmp7.i)\r\n unreachable\r\n\r\ni.exit:\r\n ret i16 %conv\r\n}\r\n```\r\n```\r\ndefine i16 @tgt(i16 %g, ptr %e, ptr %d) {\r\nentry:\r\n %tobool.not.i = icmp eq i16 %g, 0\r\n br i1 %tobool.not.i, label %i.exit, label %for.cond.preheader.i\r\n\r\nfor.cond.preheader.i: ; preds = %entry\r\n unreachable\r\n\r\ni.exit: ; preds = %entry\r\n ret i16 1\r\n}\r\n```\r\n\r\nllvm version: a8e1c3e1239604ac787b6a2d39b5278ddec8aa8a\r\n\r\ncc @nikic", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "dtcxzyw", "body": "Reverting https://github.com/llvm/llvm-project/commit/11484cb817bcc2a6e2ef9572be982a1a5a4964ec fixes this issue.\r\n"}, {"author": "nikic", "body": "I've applied two fixes for this. https://github.com/llvm/llvm-project/commit/86b37944a70229b07626e63bdb9a46b4bc3d1460 tries to stick closer to the current logic, while https://github.com/llvm/llvm-project/commit/05670b42f5b45710bfdba48dcb7e8c30c8c7478f is a more thorough fix on top of that."}, {"author": "dtcxzyw", "body": "Confirmed that the original case has been fixed. Thank you!\r\n"}, {"author": "chapuni", "body": "@nikic Fixed, thanks!"}]}, "verified": true} {"bug_id": "82052", "issue_url": "https://github.com/llvm/llvm-project/issues/82052", "bug_type": "miscompilation", "base_commit": "a45df47375e50914900dcc07abd2fa67bfa0dd3b", "knowledge_cutoff": "2024-02-16T22:14:29Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "d2a26a7bd5fc7cc5752337b7f4f999642feb37dc", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": [[1870, 1883]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": ["InstCombinerImpl::SimplifyDemandedVectorElts"]}}, "patch": "commit d2a26a7bd5fc7cc5752337b7f4f999642feb37dc\nAuthor: Antonio Frighetto \nDate: Sun Feb 18 18:10:41 2024 +0100\n\n [InstCombine] Do not perform binop-of-shuffle when mask is poison\n \n A miscompilation issue has been addressed with refined checking.\n Shuffle masks operand may be turned into `poison` if this does\n not lead to observable changes. This however may not guarantee\n binop to binop-of-shuffle replacement to be sound anymore.\n \n Fixes: https://github.com/llvm/llvm-project/issues/82052.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\nindex 5f13454089e5..97ae980a7cba 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n@@ -1870,14 +1870,16 @@ Value *InstCombinerImpl::SimplifyDemandedVectorElts(Value *V,\n Value *ShufOp = MatchShufAsOp0 ? X : Y;\n Value *OtherOp = MatchShufAsOp0 ? Y : X;\n for (User *U : OtherOp->users()) {\n- auto Shuf = m_Shuffle(m_Specific(ShufOp), m_Value(), m_ZeroMask());\n+ ArrayRef Mask;\n+ auto Shuf = m_Shuffle(m_Specific(ShufOp), m_Value(), m_Mask(Mask));\n if (BO->isCommutative()\n ? match(U, m_c_BinOp(Opcode, Shuf, m_Specific(OtherOp)))\n : MatchShufAsOp0\n ? match(U, m_BinOp(Opcode, Shuf, m_Specific(OtherOp)))\n : match(U, m_BinOp(Opcode, m_Specific(OtherOp), Shuf)))\n- if (DT.dominates(U, I))\n- return U;\n+ if (match(Mask, m_ZeroMask()) && Mask[0] != PoisonMaskElem)\n+ if (DT.dominates(U, I))\n+ return U;\n }\n return nullptr;\n };\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/vec_demanded_elts.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "common_binop_demand_via_extelt_op0_mismatch_elt1", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\n\ndeclare void @use(<2 x i4>)\n\ndefine i4 @common_binop_demand_via_extelt_op0_mismatch_elt1(<2 x i4> %x, <2 x i4> %y) {\n %xshuf = shufflevector <2 x i4> %x, <2 x i4> poison, <2 x i32> zeroinitializer\n %b_xshuf_y = sub <2 x i4> %xshuf, %y\n %b_xy = sub nsw <2 x i4> %x, %y\n %b_xy0 = extractelement <2 x i4> %b_xy, i32 1\n call void @use(<2 x i4> %b_xshuf_y)\n ret i4 %b_xy0\n}\n"}, {"test_name": "common_binop_demand_via_splat_mask_poison", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine <2 x i8> @common_binop_demand_via_splat_mask_poison(<2 x i8> %x, <2 x i8> %y) {\n %ysplat = shufflevector <2 x i8> %y, <2 x i8> poison, <2 x i32> \n %vv = add <2 x i8> %x, %ysplat\n %m = add <2 x i8> %x, %y\n %msplat = shufflevector <2 x i8> %m, <2 x i8> poison, <2 x i32> zeroinitializer\n %res = add <2 x i8> %vv, %msplat\n ret <2 x i8> %res\n}\n"}, {"test_name": "common_binop_demand_via_splat_mask_poison_2", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine <2 x i8> @common_binop_demand_via_splat_mask_poison_2(<2 x i8> %x, <2 x i8> %y) {\n %ysplat = shufflevector <2 x i8> %y, <2 x i8> poison, <2 x i32> \n %vv = add <2 x i8> %x, %ysplat\n %m = add <2 x i8> %x, %y\n %msplat = shufflevector <2 x i8> %m, <2 x i8> %y, <2 x i32> \n %res = add <2 x i8> %vv, %msplat\n ret <2 x i8> %res\n}\n"}, {"test_name": "common_binop_demand_via_splat_mask_poison_3", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine <2 x i8> @common_binop_demand_via_splat_mask_poison_3(<2 x i8> %x, <2 x i8> %y) {\n %ysplat = shufflevector <2 x i8> %y, <2 x i8> poison, <2 x i32> \n %vv = add <2 x i8> %x, %ysplat\n %m = add <2 x i8> %x, %y\n %msplat = shufflevector <2 x i8> %m, <2 x i8> poison, <2 x i32> zeroinitializer\n %res = add <2 x i8> %vv, %msplat\n ret <2 x i8> %res\n}\n"}]}], "issue": {"title": "[InstCombine] Replacement of binop with binop-of-shuffle incorrectly adds extra poison value", "body": "The following IR is a minimal repro when run with `opt -passes=instcombine`:\r\n```llvm\r\ndefine i8 @src(<2 x i8> %x, <2 x i8> %y) {\r\n %xsplat = shufflevector <2 x i8> %x, <2 x i8> poison, <2 x i32> \r\n %vv = mul <2 x i8> %xsplat, %y\r\n %m = mul <2 x i8> %x, %y ; <<<< Note that %m is later replaced with %vv in the output\r\n %msplat = shufflevector <2 x i8> %m, <2 x i8> poison, <2 x i32> \r\n %res = add <2 x i8> %vv, %msplat\r\n %vget_lane = extractelement <2 x i8> %res, i64 1\r\n ret i8 %vget_lane\r\n}\r\n```\r\nWhich on 17.x or trunk will give:\r\n```llvm\r\ndefine i8 @src(<2 x i8> %x, <2 x i8> %y) {\r\n %xsplat = shufflevector <2 x i8> %x, <2 x i8> poison, <2 x i32> \r\n %vv = mul <2 x i8> %xsplat, %y\r\n %msplat = shufflevector <2 x i8> %vv, <2 x i8> poison, <2 x i32> \r\n %res = add <2 x i8> %vv, %msplat\r\n %vget_lane = extractelement <2 x i8> %res, i64 1\r\n ret i8 %vget_lane\r\n}\r\n```\r\nHowever, this is not correct, because the added poisoned values end up observed in the final output. First, in InstCombine, the first element in each mask for the shuffles are first replaced with `poison` due to not being observed in the final `extractelement`. After that we run into this code in `InstCombineSimplifyDemanded.cpp`:\r\nhttps://github.com/llvm/llvm-project/blob/d8278b682386f51dfba204849c624672a3df40c7/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp#L1866-L1880\r\n\r\nThis is what replaces the `mul <2 x i8> %x, %y` with `mul <2 x i8> %xsplat, %y` (`m_ZeroMask` also matches poison). However in doing so, it introduces the extra poison from `%xsplat`, meaning that `%msplat` has both its elements poisoned, which ends up propagating to `%vget_lane`.\r\n\r\nHere is a Godbolt link showing the test case on 16.x, 17.x, and trunk:\r\nhttps://godbolt.org/z/5eeT7x4s3\r\n\r\nAnd also an Alive2 example showing the bad transformation with a counter-example:\r\nhttps://alive2.llvm.org/ce/z/RpBJEN\r\n\r\nThe code was first introduced in 3b090ff2bdb2828b53d6fec9c06b9d52fed09f42, which a bisect shows is when this first started repro'ing\r\n\r\nI have verified that this still repros on latest trunk, 2de269a641e4ffbb7a44e559c4c0a91bb66df823\r\n\r\nFor priority/triage: this bug was found by a fuzzer meant to test SIMD codegen. It was not from manually written code", "author": "Benjins", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "Benjins", "body": "A slightly smaller test case:\r\nhttps://alive2.llvm.org/ce/z/6AMPWC\r\n```llvm\r\ndefine <2 x i8> @src(<2 x i8> %x, <2 x i8> %y) {\r\n %xsplat = shufflevector <2 x i8> %x, <2 x i8> poison, <2 x i32> \r\n %vv = mul <2 x i8> %xsplat, %y\r\n %m = mul <2 x i8> %x, %y\r\n %msplat = shufflevector <2 x i8> %m, <2 x i8> poison, <2 x i32> \r\n %res = add <2 x i8> %vv, %msplat\r\n ret <2 x i8> %res\r\n}\r\n```\r\nThis doesn't extract the vector, and instead manually specifies the poisonous elements in the shuffles. As Alive notes, the output for `src(<0, 0>, <0, 0>)` goes from `` to ``"}]}, "verified": true} {"bug_id": "114371", "issue_url": "https://github.com/llvm/llvm-project/issues/114371", "bug_type": "crash", "base_commit": "71cfa381ef8c4fe659c67e8b2901d767e10f2aff", "knowledge_cutoff": "2024-10-31T07:26:11Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "e7080fd735d02590cdce1bef97a7f64a34145b15", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[16127, 16137]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree"]}}, "patch": "commit e7080fd735d02590cdce1bef97a7f64a34145b15\nAuthor: Alexey Bataev \nDate: Thu Oct 31 09:26:53 2024 -0700\n\n [SLP]Extra check if the intruction matked for removal, must be replaced in reduction ops\n \n If the instruction is vectorized and it is a part of the reduced values\n gather/buildvector node, it should replaced in reduced operation\n instructions before removal properly, to avoid compiler crash.\n \n Fixes #114371\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 268546fe99e1..593964177ad7 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -16127,11 +16127,13 @@ BoUpSLP::vectorizeTree(const ExtraValueToDebugLocsMap &ExternallyUsedValues,\n if (IE->Idx != 0 &&\n !(VectorizableTree.front()->isGather() &&\n !IE->UserTreeIndices.empty() &&\n- any_of(IE->UserTreeIndices,\n- [&](const EdgeInfo &EI) {\n- return EI.UserTE == VectorizableTree.front().get() &&\n- EI.EdgeIdx == UINT_MAX;\n- })) &&\n+ (ValueToGatherNodes.lookup(I).contains(\n+ VectorizableTree.front().get()) ||\n+ any_of(IE->UserTreeIndices,\n+ [&](const EdgeInfo &EI) {\n+ return EI.UserTE == VectorizableTree.front().get() &&\n+ EI.EdgeIdx == UINT_MAX;\n+ }))) &&\n !(GatheredLoadsEntriesFirst.has_value() &&\n IE->Idx >= *GatheredLoadsEntriesFirst &&\n VectorizableTree.front()->isGather() &&\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/external-reduced-value-vectorized.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test(ptr %c, i16 %a, i16 %0) {\nentry:\n %tobool = icmp ne i16 %a, 0\n %1 = zext i1 %tobool to i16\n %cmp3 = icmp ugt i16 %0, %1\n %2 = and i1 %tobool, %cmp3\n %tobool.1 = icmp ne i16 %a, 0\n %3 = zext i1 %tobool.1 to i16\n %cmp3.1 = icmp ugt i16 %0, %3\n %4 = and i1 %tobool.1, %cmp3.1\n %5 = select i1 %2, i1 %4, i1 false\n %tobool.2 = icmp ne i16 %a, 0\n %6 = zext i1 %tobool.2 to i16\n %cmp3.2 = icmp ugt i16 %0, %6\n %7 = and i1 %tobool.2, %cmp3.2\n %8 = select i1 %5, i1 %7, i1 false\n %tobool.3 = icmp ne i16 %a, 0\n %9 = zext i1 %tobool.3 to i16\n %cmp3.3 = icmp ugt i16 %a, %9\n %10 = icmp ult i16 %a, -2\n %11 = and i1 %10, %cmp3.3\n %12 = select i1 %8, i1 %11, i1 false\n %13 = zext i1 %12 to i32\n store i32 %13, ptr %c, align 4\n ret i32 0\n}\n"}]}], "issue": {"title": "clang crashes on valid code at -O{2,3} on x86_64-linux-gnu: Assertion `(I->use_empty() || all_of(I->uses(), [&](Use &U) { return isDeleted( cast(U.getUser())); })) && \"trying to erase instruction with users.\"' failed ", "body": "It appears to be a recent regression as it doesn't reproduce with 19.1.0 and earlier.\r\n\r\nCompiler Explorer: https://godbolt.org/z/5Tca1zsE8\r\n\r\n```\r\n[507] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git fdc78120bde1426b223196bec645b6f27fc60f0e)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[508] % \r\n[508] % clangtk -O2 small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:2748: void llvm::slpvectorizer::BoUpSLP::removeInstructionsAndOperands(llvm::ArrayRef) [with T = llvm::Instruction]: Assertion `(I->use_empty() || all_of(I->uses(), [&](Use &U) { return isDeleted( cast(U.getUser())); })) && \"trying to erase instruction with users.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20241031-clangtk-m64-O3-build-070121/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20241031-clangtk-m64-O3-build-070121/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-2a4e26.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4.\tRunning pass \"slp-vectorizer\" on function \"main\"\r\n #0 0x000055e84bc23a6f llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4428a6f)\r\n #1 0x000055e84bc21274 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007efe85a5c420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007efe8549300b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007efe85472859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007efe85472729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007efe85472729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007efe85483fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x000055e84d68990f void llvm::slpvectorizer::BoUpSLP::removeInstructionsAndOperands(llvm::ArrayRef) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5e8e90f)\r\n #9 0x000055e84d6e6d28 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::SmallDenseSet> const&, llvm::Instruction*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5eebd28)\r\n#10 0x000055e84d6f8d01 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#11 0x000055e84d6fbb89 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::SmallVectorImpl&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5f00b89)\r\n#12 0x000055e84d7004ab llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\r\n#13 0x000055e84d704d60 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5f09d60)\r\n#14 0x000055e84d70b5a0 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#15 0x000055e84d70c273 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5f11273)\r\n#16 0x000055e84d17e016 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5983016)\r\n#17 0x000055e84b5aa8b1 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3daf8b1)\r\n#18 0x000055e84909b826 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x18a0826)\r\n#19 0x000055e84b5a921d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3dae21d)\r\n#20 0x000055e84909a226 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x189f226)\r\n#21 0x000055e84b5a8c51 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3dadc51)\r\n#22 0x000055e84bedd2d3 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#23 0x000055e84bee0aca clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x46e5aca)\r\n#24 0x000055e84c5df37c clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4de437c)\r\n#25 0x000055e84e4e0fac clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x6ce5fac)\r\n#26 0x000055e84c5df798 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4de4798)\r\n#27 0x000055e84c8a7649 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x50ac649)\r\n#28 0x000055e84c823e7e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5028e7e)\r\n#29 0x000055e84c993f46 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5198f46)\r\n#30 0x000055e848c544f5 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x14594f5)\r\n#31 0x000055e848c4c01a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#32 0x000055e848c4ffd7 clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1454fd7)\r\n#33 0x000055e848b3a78b main (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x133f78b)\r\n#34 0x00007efe85474083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#35 0x000055e848c4baae _start (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1450aae)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git fdc78120bde1426b223196bec645b6f27fc60f0e)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-8056bb.c\r\nclangtk: note: diagnostic msg: /tmp/small-8056bb.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[509] % \r\n[509] % cat small.c\r\nshort a;\r\nint b, d;\r\nunsigned short c;\r\nint main() {\r\n b = 0;\r\n for (; b < 5; b++) {\r\n a--;\r\n d |= (3 - (((!a) < 1) < c && a)) & c;\r\n }\r\n return 0;\r\n}\r\n```\r\n\r\n", "author": "zhendongsu", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "102187", "issue_url": "https://github.com/llvm/llvm-project/issues/102187", "bug_type": "crash", "base_commit": "4b6e3e38c46f8fedb92a30b29cc9654450244e2d", "knowledge_cutoff": "2024-08-06T17:31:46Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "2601d6f189cf1d4e4e77a7cba83e5d32c87bf079", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[14049, 14060], [14080, 14090]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree"]}}, "patch": "commit 2601d6f189cf1d4e4e77a7cba83e5d32c87bf079\nAuthor: Alexey Bataev \nDate: Tue Aug 6 12:27:36 2024 -0700\n\n [SLP]Fix PR102187: do not insert extractelement before landingpad instruction.\n \n Landingpad instruction must be the very first instruction after the phi\n nodes, so need to inser extractelement/shuffles after this instruction.\n \n Fixes https://github.com/llvm/llvm-project/issues/102187\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 9589d306581e..5c8873853a73 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -14049,12 +14049,19 @@ Value *BoUpSLP::vectorizeTree(\n \"ExternallyUsedValues map or remain as scalar in vectorized \"\n \"instructions\");\n if (auto *VecI = dyn_cast(Vec)) {\n- if (auto *PHI = dyn_cast(VecI))\n- Builder.SetInsertPoint(PHI->getParent(),\n- PHI->getParent()->getFirstNonPHIIt());\n- else\n+ if (auto *PHI = dyn_cast(VecI)) {\n+ if (PHI->getParent()->isLandingPad())\n+ Builder.SetInsertPoint(\n+ PHI->getParent(),\n+ std::next(\n+ PHI->getParent()->getLandingPadInst()->getIterator()));\n+ else\n+ Builder.SetInsertPoint(PHI->getParent(),\n+ PHI->getParent()->getFirstNonPHIIt());\n+ } else {\n Builder.SetInsertPoint(VecI->getParent(),\n std::next(VecI->getIterator()));\n+ }\n } else {\n Builder.SetInsertPoint(&F->getEntryBlock(), F->getEntryBlock().begin());\n }\n@@ -14080,11 +14087,18 @@ Value *BoUpSLP::vectorizeTree(\n auto VecIt = VectorCasts.find(Key);\n if (VecIt == VectorCasts.end()) {\n IRBuilderBase::InsertPointGuard Guard(Builder);\n- if (auto *IVec = dyn_cast(Vec))\n- Builder.SetInsertPoint(\n- IVec->getParent()->getFirstNonPHIOrDbgOrLifetime());\n- else if (auto *IVec = dyn_cast(Vec))\n+ if (auto *IVec = dyn_cast(Vec)) {\n+ if (IVec->getParent()->isLandingPad())\n+ Builder.SetInsertPoint(IVec->getParent(),\n+ std::next(IVec->getParent()\n+ ->getLandingPadInst()\n+ ->getIterator()));\n+ else\n+ Builder.SetInsertPoint(\n+ IVec->getParent()->getFirstNonPHIOrDbgOrLifetime());\n+ } else if (auto *IVec = dyn_cast(Vec)) {\n Builder.SetInsertPoint(IVec->getNextNonDebugInstruction());\n+ }\n Vec = Builder.CreateIntCast(\n Vec,\n getWidenedType(\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/extractelement-phi-in-landingpad.ll", "commands": ["opt -S --passes=slp-vectorizer -slp-threshold=-99999 -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define void @test() personality ptr null {\nbb:\n invoke void null()\n to label %bb65 unwind label %bb4\n\nbb2: ; No predecessors!\n invoke void null()\n to label %bb65 unwind label %bb4\n\nbb4: ; preds = %bb2, %bb\n %phi5 = phi i32 [ 0, %bb ], [ 0, %bb2 ]\n %phi6 = phi i32 [ 0, %bb ], [ 0, %bb2 ]\n %landingpad = landingpad { ptr, i32 }\n cleanup\n call void null(i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5, i32 %phi5)\n ret void\n\nbb65: ; preds = %bb2, %bb\n ret void\n}\n"}]}], "issue": {"title": "LLVM ERROR: Broken module found, compilation aborted!", "body": "To reproduce run opt with \r\n[Test.ll.tgz](https://github.com/user-attachments/files/16513795/Test.ll.tgz)\r\nusing `-passes slp-vectorizer -slp-threshold=-99999` \r\n\r\nReproducer https://godbolt.org/z/f4szGsMYd\r\n\r\n```\r\nThe unwind destination does not have an exception handling instruction!\r\n %invoke = invoke i64 null(ptr addrspace(1) null, i32 0, i64 0, i32 0)\r\n to label %bb65 unwind label %bb4\r\nThe unwind destination does not have an exception handling instruction!\r\n %invoke3 = invoke i64 null(ptr addrspace(1) null, i32 0, i64 0, i32 0)\r\n to label %bb65 unwind label %bb4\r\nLandingPadInst not the first non-PHI instruction in the block.\r\n %landingpad = landingpad { ptr, i32 }\r\n cleanup\r\nLLVM ERROR: Broken module found, compilation aborted!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes slp-vectorizer -slp-threshold=-99999 \r\n1.\tRunning pass \"verify\" on module \"\"\r\n #0 0x0000000004f1d2d8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f1d2d8)\r\n #1 0x0000000004f1aa4c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x0000788c1ca42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x0000788c1ca969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x0000788c1ca42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x0000788c1ca287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00000000007b40fd llvm::json::operator==(llvm::json::Value const&, llvm::json::Value const&) (.cold) JSON.cpp:0:0\r\n #7 0x0000000004e571e8 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e571e8)\r\n #8 0x0000000004d53550 llvm::VerifierPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d53550)\r\n #9 0x00000000008d8a7e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d8a7e)\r\n#10 0x0000000004d17270 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d17270)\r\n#11 0x00000000008e37f2 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e37f2)\r\n#12 0x00000000008d671c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d671c)\r\n#13 0x0000788c1ca29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#14 0x0000788c1ca29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#15 0x00000000008cdb7e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8cdb7e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "104422", "issue_url": "https://github.com/llvm/llvm-project/issues/104422", "bug_type": "miscompilation", "base_commit": "13a6a7975256ebdbce85f1174ae2eec735fa0d7a", "knowledge_cutoff": "2024-08-15T09:29:27Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "56140a8258a3498cfcd9f0f05c182457d43cbfd2", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[15663, 15669]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::collectValuesToDemote"]}}, "patch": "commit 56140a8258a3498cfcd9f0f05c182457d43cbfd2\nAuthor: Alexey Bataev \nDate: Thu Aug 15 07:57:37 2024 -0700\n\n [SLP]Fix PR104422: Wrong value truncation\n \n The minbitwidth restrictions can be skipped only for immediate reduced\n values, for other nodes still need to check if external users allow\n bitwidth reduction.\n \n Fixes https://github.com/llvm/llvm-project/issues/104422\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 81841a8f6928..87b4ed599b5b 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -15663,7 +15663,8 @@ bool BoUpSLP::collectValuesToDemote(\n if (any_of(E.Scalars, [&](Value *V) {\n return !all_of(V->users(), [=](User *U) {\n return getTreeEntry(U) ||\n- (UserIgnoreList && UserIgnoreList->contains(U)) ||\n+ (E.Idx == 0 && UserIgnoreList &&\n+ UserIgnoreList->contains(U)) ||\n (!isa(U) && U->getType()->isSized() &&\n !U->getType()->isScalableTy() &&\n DL->getTypeSizeInBits(U->getType()) <= BitWidth);\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/operand-is-reduced-val.ll", "commands": ["opt -S --passes=slp-vectorizer < %s -slp-threshold=-10"], "tests": [{"test_name": "src", "test_body": "define i64 @src(i32 %a) {\nentry:\n %0 = sext i32 %a to i64\n %1 = add nsw i64 %0, 4294967297\n %2 = sext i32 %a to i64\n %3 = add nsw i64 %2, 4294967297\n %4 = add i64 %3, %1\n %5 = and i64 %3, 1\n %6 = add i64 %4, %5\n %7 = sext i32 %a to i64\n %8 = add nsw i64 %7, 4294967297\n %9 = add i64 %8, %6\n %10 = and i64 %8, 1\n %11 = add i64 %9, %10\n %12 = sext i32 %a to i64\n %13 = add nsw i64 %12, 4294967297\n %14 = add i64 %13, %11\n %15 = and i64 %13, 1\n %16 = add i64 %14, %15\n %17 = sext i32 %a to i64\n %18 = add nsw i64 %17, 4294967297\n %19 = add i64 %18, %16\n %20 = and i64 %18, 1\n %21 = add i64 %19, %20\n ret i64 %21\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Wrong value truncation ", "body": "- There seems to be an issue with SLPVectorizer. In the following example, a 2-byte wide value is incorrectly truncated:\r\n\r\n```\r\ndefine i64 @test(ptr addrspace(1) %arr, i32 %a) {\r\nentry:\r\n br label %loop\r\n\r\nloop:\r\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\r\n %local_1_7 = phi i64 [ 0, %entry ], [ %4, %loop ]\r\n %0 = sext i32 %a to i64\r\n %1 = add nsw i64 %0, 4294967297\r\n %2 = add i64 %1, %local_1_7\r\n %3 = and i64 %1, 1\r\n %4 = add i64 %2, %3\r\n %iv.next = add nuw nsw i64 %iv, 1\r\n %5 = icmp ugt i64 %iv, 355\r\n br i1 %5, label %exit, label %loop\r\n\r\nexit:\r\n ret i64 %4\r\n}\r\n```\r\n- In the loop we summarize the constant 4294967297 (`0x1'0000'0001`), but after running SLPVectorizer over the unrolled loop, the constant is truncated to `1`:\r\n```\r\nloop: \r\n %iv = phi i64 [ 0, %entry ], [ %iv.next.16, %loop ]\r\n %local_1_7 = phi i64 [ 0, %entry ], [ %op.rdx3, %loop ]\r\n %2 = add <16 x i32> %1, \r\n %3 = sext <16 x i32> %2 to <16 x i64>\r\n...\r\n```\r\n- Command to reproduce: `opt -passes=loop-unroll,slp-vectorizer -unroll-count=17`\r\n- https://godbolt.org/z/zcPefTPzW\r\n- Looks like the issue was caused by the https://github.com/llvm/llvm-project/commit/7f2167868d8c1cedd3915883412b9c787a2f01db\r\n\r\n@alexey-bataev Could you please take a look?\r\n", "author": "aleks-tmb", "labels": ["miscompilation", "release:backport", "llvm:SLPVectorizer"], "comments": [{"author": "nikic", "body": "This probably needs an LLVM 19 backport?\r\n\r\n/cherry-pick 65ac12d3c9877ecf5b97552364e7eead887d94eb 56140a8258a3498cfcd9f0f05c182457d43cbfd2"}]}, "verified": true} {"bug_id": "112078", "issue_url": "https://github.com/llvm/llvm-project/issues/112078", "bug_type": "miscompilation", "base_commit": "39aae575744de7ae77e755142e0746a5b473b3ed", "knowledge_cutoff": "2024-10-12T04:08:35Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "9edc454ee601e04500529c98b753d3bd8f427d01", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": [[935, 943], [945, 955], [3362, 3370]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": ["InstCombinerImpl::foldAndOrOfICmps", "foldIsPowerOf2OrZero"]}}, "patch": "commit 9edc454ee601e04500529c98b753d3bd8f427d01\nAuthor: Yingwei Zheng \nDate: Mon Oct 14 20:52:55 2024 +0800\n\n [InstCombine] Drop range attributes in `foldIsPowerOf2OrZero` (#112178)\n \n Closes https://github.com/llvm/llvm-project/issues/112078.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\nindex 453071f3f982..64bee4ab974e 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n@@ -935,9 +935,11 @@ static Value *foldSignedTruncationCheck(ICmpInst *ICmp0, ICmpInst *ICmp1,\n \n /// Fold (icmp eq ctpop(X) 1) | (icmp eq X 0) into (icmp ult ctpop(X) 2) and\n /// fold (icmp ne ctpop(X) 1) & (icmp ne X 0) into (icmp ugt ctpop(X) 1).\n-/// Also used for logical and/or, must be poison safe.\n+/// Also used for logical and/or, must be poison safe if range attributes are\n+/// dropped.\n static Value *foldIsPowerOf2OrZero(ICmpInst *Cmp0, ICmpInst *Cmp1, bool IsAnd,\n- InstCombiner::BuilderTy &Builder) {\n+ InstCombiner::BuilderTy &Builder,\n+ InstCombinerImpl &IC) {\n CmpInst::Predicate Pred0, Pred1;\n Value *X;\n if (!match(Cmp0, m_ICmp(Pred0, m_Intrinsic(m_Value(X)),\n@@ -945,11 +947,19 @@ static Value *foldIsPowerOf2OrZero(ICmpInst *Cmp0, ICmpInst *Cmp1, bool IsAnd,\n !match(Cmp1, m_ICmp(Pred1, m_Specific(X), m_ZeroInt())))\n return nullptr;\n \n- Value *CtPop = Cmp0->getOperand(0);\n- if (IsAnd && Pred0 == ICmpInst::ICMP_NE && Pred1 == ICmpInst::ICMP_NE)\n+ auto *CtPop = cast(Cmp0->getOperand(0));\n+ if (IsAnd && Pred0 == ICmpInst::ICMP_NE && Pred1 == ICmpInst::ICMP_NE) {\n+ // Drop range attributes and re-infer them in the next iteration.\n+ CtPop->dropPoisonGeneratingAnnotations();\n+ IC.addToWorklist(CtPop);\n return Builder.CreateICmpUGT(CtPop, ConstantInt::get(CtPop->getType(), 1));\n- if (!IsAnd && Pred0 == ICmpInst::ICMP_EQ && Pred1 == ICmpInst::ICMP_EQ)\n+ }\n+ if (!IsAnd && Pred0 == ICmpInst::ICMP_EQ && Pred1 == ICmpInst::ICMP_EQ) {\n+ // Drop range attributes and re-infer them in the next iteration.\n+ CtPop->dropPoisonGeneratingAnnotations();\n+ IC.addToWorklist(CtPop);\n return Builder.CreateICmpULT(CtPop, ConstantInt::get(CtPop->getType(), 2));\n+ }\n \n return nullptr;\n }\n@@ -3362,9 +3372,9 @@ Value *InstCombinerImpl::foldAndOrOfICmps(ICmpInst *LHS, ICmpInst *RHS,\n /*IsLogical*/ false, Builder, Q))\n return V;\n \n- if (Value *V = foldIsPowerOf2OrZero(LHS, RHS, IsAnd, Builder))\n+ if (Value *V = foldIsPowerOf2OrZero(LHS, RHS, IsAnd, Builder, *this))\n return V;\n- if (Value *V = foldIsPowerOf2OrZero(RHS, LHS, IsAnd, Builder))\n+ if (Value *V = foldIsPowerOf2OrZero(RHS, LHS, IsAnd, Builder, *this))\n return V;\n \n // TODO: One of these directions is fine with logical and/or, the other could\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/ispow2.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "is_pow2or0_decrement_op", "test_body": "define i1 @is_pow2or0_decrement_op(i8 %x) {\n %dec = add i8 %x, -1\n %and = and i8 %dec, %x\n %cmp = icmp eq i8 %and, 0\n ret i1 %cmp\n}\n"}, {"test_name": "is_pow2_negate_op_vec", "test_body": "define <2 x i1> @is_pow2_negate_op_vec(<2 x i32> %x) {\n %neg = sub <2 x i32> zeroinitializer, %x\n %and = and <2 x i32> %neg, %x\n %cmp = icmp eq <2 x i32> %and, %x\n %notzero = icmp ne <2 x i32> %x, zeroinitializer\n %r = and <2 x i1> %cmp, %notzero\n ret <2 x i1> %r\n}\n"}, {"test_name": "is_pow2_ctpop_wrong_cmp_op1_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_wrong_cmp_op1_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 3\n %notzero = icmp ne i32 %x, 0\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_negate_op", "test_body": "define i1 @is_pow2_negate_op(i32 %x) {\n %neg = sub i32 0, %x\n %and = and i32 %neg, %x\n %cmp = icmp eq i32 %and, %x\n %notzero = icmp ne i32 %x, 0\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n"}, {"test_name": "not_pow2_nz_known_bits", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @not_pow2_nz_known_bits(i32 %xin) {\n %x = or i32 %xin, 1\n %cnt = call i32 @llvm.ctpop.i32(i32 %x)\n %r = icmp ne i32 %cnt, 1\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_nz_known_bits", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_nz_known_bits(i32 %xin) {\n %x = or i32 %xin, 64\n %cnt = call i32 @llvm.ctpop.i32(i32 %x)\n %r = icmp eq i32 %cnt, 1\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_power2_or_zero_with_range", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_power2_or_zero_with_range(i32 %x) {\n %ctpop = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %ctpop, 1\n %notzero = icmp eq i32 %x, 0\n %res = select i1 %notzero, i1 true, i1 %cmp\n ret i1 %res\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_is_p2_or_z", "test_body": "define i1 @blsmsk_is_p2_or_z(i32 %xx, i32 %yy) {\n %x = or i32 %xx, %yy\n %xm1 = add i32 %x, -1\n %y = xor i32 %x, %xm1\n %r = icmp uge i32 %x, %y\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2nor0_ctpop_commute_vec", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>) #0\n\ndefine <2 x i1> @isnot_pow2nor0_ctpop_commute_vec(<2 x i8> %x) {\n %t0 = tail call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %x)\n %cmp = icmp ne <2 x i8> %t0, splat (i8 1)\n %notzero = icmp ne <2 x i8> %x, zeroinitializer\n %r = and <2 x i1> %notzero, %cmp\n ret <2 x i1> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_ctpop_wrong_pred1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2or0_ctpop_wrong_pred1(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 1\n %iszero = icmp eq i32 %x, 0\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_non_zero_ne_1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #1\n\ndefine i1 @is_pow2_non_zero_ne_1(i32 %x) {\n %notzero = icmp ne i32 %x, 0\n call void @llvm.assume(i1 %notzero)\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 1\n ret i1 %cmp\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\nattributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "blsmsk_is_p2_or_z_fail", "test_body": "define i1 @blsmsk_is_p2_or_z_fail(i32 %xx, i32 %yy) {\n %x = or i32 %xx, %yy\n %xm1 = add i32 %x, -1\n %y = xor i32 %x, %xm1\n %r = icmp ugt i32 %x, %y\n ret i1 %r\n}\n"}, {"test_name": "blsmsk_is_p2_or_z_uge_xy_fail", "test_body": "define i1 @blsmsk_is_p2_or_z_uge_xy_fail(i8 %xx, i8 %yy) {\n %x = or i8 %xx, %yy\n %xm1 = add i8 %x, -1\n %y = xor i8 %x, %xm1\n %r = icmp uge i8 %x, %y\n ret i1 %r\n}\n"}, {"test_name": "is_pow2or0_ctpop_logical_extra_uses", "test_body": "declare void @use(i32)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use_i1(i1)\n\ndefine i1 @is_pow2or0_ctpop_logical_extra_uses(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n call void @use(i32 %t0)\n %cmp = icmp eq i32 %t0, 1\n call void @use_i1(i1 %cmp)\n %iszero = icmp eq i32 %x, 0\n call void @use_i1(i1 %iszero)\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop_commute_vec", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>) #0\n\ndefine <2 x i1> @isnot_pow2_ctpop_commute_vec(<2 x i8> %x) {\n %t0 = tail call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %x)\n %cmp = icmp ugt <2 x i8> %t0, splat (i8 1)\n %iszero = icmp eq <2 x i8> %x, zeroinitializer\n %r = or <2 x i1> %cmp, %iszero\n ret <2 x i1> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_isnt_p2_or_z_fail_wrong_add", "test_body": "define i1 @blsmsk_isnt_p2_or_z_fail_wrong_add(i32 %x, i32 %z) {\n %xm1 = add i32 %z, -1\n %y = xor i32 %x, %xm1\n %r = icmp ult i32 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "is_pow2_ctpop_wrong_pred2", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_wrong_pred2(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n %cmp2 = icmp sgt i32 %x, 0\n %r = and i1 %cmp2, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_ctpop_swap_cmp", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2or0_ctpop_swap_cmp(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 1\n %iszero = icmp eq i32 %x, 0\n %r = or i1 %cmp, %iszero\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_ctpop_commute_vec_wrong_pred3", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>) #0\n\ndefine <2 x i1> @is_pow2or0_ctpop_commute_vec_wrong_pred3(<2 x i8> %x) {\n %t0 = tail call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %x)\n %cmp = icmp eq <2 x i8> %t0, splat (i8 1)\n %iszero = icmp eq <2 x i8> %x, zeroinitializer\n %r = and <2 x i1> %cmp, %iszero\n ret <2 x i1> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_decrement_op_logical", "test_body": "define i1 @is_pow2_decrement_op_logical(i8 %x) {\n %dec = add i8 %x, -1\n %and = and i8 %dec, %x\n %cmp = icmp eq i8 %and, 0\n %notzero = icmp ne i8 %x, 0\n %r = select i1 %cmp, i1 %notzero, i1 false\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2or0_negate_op_vec", "test_body": "define <2 x i1> @isnot_pow2or0_negate_op_vec(<2 x i32> %x) {\n %neg = sub <2 x i32> zeroinitializer, %x\n %and = and <2 x i32> %neg, %x\n %cmp = icmp ne <2 x i32> %and, %x\n ret <2 x i1> %cmp\n}\n"}, {"test_name": "is_pow2or0_ctpop_commute_vec", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>) #0\n\ndefine <2 x i1> @is_pow2or0_ctpop_commute_vec(<2 x i8> %x) {\n %t0 = tail call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %x)\n %cmp = icmp eq <2 x i8> %t0, splat (i8 1)\n %iszero = icmp eq <2 x i8> %x, zeroinitializer\n %r = or <2 x i1> %iszero, %cmp\n ret <2 x i1> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_negate_op_logical", "test_body": "define i1 @is_pow2_negate_op_logical(i32 %x) {\n %neg = sub i32 0, %x\n %and = and i32 %neg, %x\n %cmp = icmp eq i32 %and, %x\n %notzero = icmp ne i32 %x, 0\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2nor0_ctpop", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2nor0_ctpop(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 1\n %notzero = icmp ne i32 %x, 0\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_isnt_p2_or_z_ult_xy_fail", "test_body": "define i1 @blsmsk_isnt_p2_or_z_ult_xy_fail(i8 %xx, i8 %yy) {\n %x = or i8 %xx, %yy\n %xm1 = add i8 %x, -1\n %y = xor i8 %x, %xm1\n %r = icmp ult i8 %x, %y\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2nor0_wrong_pred3_ctpop_commute_vec", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>) #0\n\ndefine <2 x i1> @isnot_pow2nor0_wrong_pred3_ctpop_commute_vec(<2 x i8> %x) {\n %t0 = tail call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %x)\n %cmp = icmp ne <2 x i8> %t0, splat (i8 1)\n %notzero = icmp ne <2 x i8> %x, zeroinitializer\n %r = or <2 x i1> %cmp, %notzero\n ret <2 x i1> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_negate_op_vec", "test_body": "define <2 x i1> @is_pow2or0_negate_op_vec(<2 x i32> %x) {\n %neg = sub <2 x i32> zeroinitializer, %x\n %and = and <2 x i32> %neg, %x\n %cmp = icmp eq <2 x i32> %and, %x\n ret <2 x i1> %cmp\n}\n"}, {"test_name": "isnot_pow2or0_negate_op", "test_body": "define i1 @isnot_pow2or0_negate_op(i32 %x) {\n %neg = sub i32 0, %x\n %and = and i32 %neg, %x\n %cmp = icmp ne i32 %and, %x\n ret i1 %cmp\n}\n"}, {"test_name": "not_pow2_or_z_known_bits_fail_wrong_cmp", "test_body": "define <2 x i1> @not_pow2_or_z_known_bits_fail_wrong_cmp(<2 x i32> %xin) {\n %x = or <2 x i32> %xin, splat (i32 64)\n %cnt = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %x)\n %r = icmp ugt <2 x i32> %cnt, splat (i32 2)\n ret <2 x i1> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i32> @llvm.ctpop.v2i32(<2 x i32>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_ctpop_wrong_cmp_op2", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_wrong_cmp_op2(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n %notzero = icmp ne i32 %x, 1\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_ctpop_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n %notzero = icmp ne i32 %x, 0\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_decrement_op_logical", "test_body": "define i1 @isnot_pow2_decrement_op_logical(i8 %x) {\n %dec = add i8 %x, -1\n %and = and i8 %dec, %x\n %cmp = icmp ne i8 %and, 0\n %iszero = icmp eq i8 %x, 0\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2_negate_op_logical", "test_body": "define i1 @isnot_pow2_negate_op_logical(i32 %x) {\n %neg = sub i32 0, %x\n %and = and i32 %neg, %x\n %cmp = icmp ne i32 %and, %x\n %iszero = icmp eq i32 %x, 0\n %r = select i1 %cmp, i1 true, i1 %iszero\n ret i1 %r\n}\n"}, {"test_name": "is_pow2or0_ctpop", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2or0_ctpop(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 1\n %iszero = icmp eq i32 %x, 0\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_is_p2_or_z_uge_yx", "test_body": "define i1 @blsmsk_is_p2_or_z_uge_yx(i8 %xx, i8 %yy) {\n %x = or i8 %xx, %yy\n %xm1 = add i8 %x, -1\n %y = xor i8 %x, %xm1\n %r = icmp uge i8 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "has_single_bit_inv", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @has_single_bit_inv(i32 %x) {\nentry:\n %cmp1 = icmp eq i32 %x, 0\n %popcnt = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)\n %cmp2 = icmp ugt i32 %popcnt, 1\n %sel = select i1 %cmp1, i1 true, i1 %cmp2\n ret i1 %sel\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_negate_op", "test_body": "define i1 @is_pow2or0_negate_op(i32 %x) {\n %neg = sub i32 0, %x\n %and = and i32 %neg, %x\n %cmp = icmp eq i32 %and, %x\n ret i1 %cmp\n}\n"}, {"test_name": "blsmsk_isnt_p2_or_z_ugt_yx_fail", "test_body": "define i1 @blsmsk_isnt_p2_or_z_ugt_yx_fail(i8 %xx, i8 %yy) {\n %x = or i8 %xx, %yy\n %xm1 = add i8 %x, -1\n %y = xor i8 %x, %xm1\n %r = icmp ugt i8 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "blsmsk_isnt_p2_or_z_fail", "test_body": "define i1 @blsmsk_isnt_p2_or_z_fail(i32 %x) {\n %xm1 = add i32 %x, -1\n %y = xor i32 %x, %xm1\n %r = icmp ule i32 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "is_pow2_decrement_op_vec", "test_body": "define <2 x i1> @is_pow2_decrement_op_vec(<2 x i8> %x) {\n %dec = add <2 x i8> %x, splat (i8 -1)\n %and = and <2 x i8> %dec, %x\n %cmp = icmp eq <2 x i8> %and, zeroinitializer\n %notzero = icmp ne <2 x i8> %x, zeroinitializer\n %r = and <2 x i1> %notzero, %cmp\n ret <2 x i1> %r\n}\n"}, {"test_name": "isnot_pow2nor0_ctpop_extra_uses", "test_body": "declare void @use(i32)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use_i1(i1)\n\ndefine i1 @isnot_pow2nor0_ctpop_extra_uses(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n call void @use(i32 %t0)\n %cmp = icmp ne i32 %t0, 1\n call void @use_i1(i1 %cmp)\n %notzero = icmp ne i32 %x, 0\n call void @use_i1(i1 %notzero)\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2nor0_ctpop_wrong_pred1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2nor0_ctpop_wrong_pred1(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 1\n %notzero = icmp ne i32 %x, 0\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2_ctpop(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n %iszero = icmp eq i32 %x, 0\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_isnt_p2_or_z_fail_multiuse", "test_body": "declare void @use.i32(i32)\n\ndefine i1 @blsmsk_isnt_p2_or_z_fail_multiuse(i32 %x) {\n %xm1 = add i32 %x, -1\n %y = xor i32 %x, %xm1\n call void @use.i32(i32 %y)\n %r = icmp ult i32 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "is_pow2or0_ctpop_wrong_pred2_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2or0_ctpop_wrong_pred2_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 1\n %iszero = icmp ne i32 %x, 0\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_ctpop_extra_uses", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use_i1(i1)\n\ndefine i1 @is_pow2_ctpop_extra_uses(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n call void @use_i1(i1 %cmp)\n %notzero = icmp ne i32 %x, 0\n call void @use_i1(i1 %notzero)\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_isnt_p2_or_z_ult_yx", "test_body": "define i1 @blsmsk_isnt_p2_or_z_ult_yx(i8 %xx, i8 %yy) {\n %x = or i8 %xx, %yy\n %xm1 = add i8 %x, -1\n %y = xor i8 %x, %xm1\n %r = icmp ult i8 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "is_pow2or0_ctpop_wrong_cmp_op1_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2or0_ctpop_wrong_cmp_op1_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 3\n %iszero = icmp eq i32 %x, 0\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_isnt_p2_or_z_fail_bad_xor", "test_body": "define i1 @blsmsk_isnt_p2_or_z_fail_bad_xor(i32 %x, i32 %z) {\n %xm1 = add i32 %x, -1\n %y = xor i32 %z, %xm1\n %r = icmp ult i32 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "is_pow2or0_negate_op_extra_use2", "test_body": "declare void @use(i32)\n\ndefine i1 @is_pow2or0_negate_op_extra_use2(i32 %x) {\n %neg = sub i32 0, %x\n %and = and i32 %neg, %x\n call void @use(i32 %and)\n %cmp = icmp eq i32 %and, %x\n ret i1 %cmp\n}\n"}, {"test_name": "isnot_pow2_negate_op", "test_body": "define i1 @isnot_pow2_negate_op(i32 %x) {\n %neg = sub i32 0, %x\n %and = and i32 %neg, %x\n %cmp = icmp ne i32 %and, %x\n %iszero = icmp eq i32 %x, 0\n %r = or i1 %cmp, %iszero\n ret i1 %r\n}\n"}, {"test_name": "is_pow2or0_negate_op_extra_use1", "test_body": "declare void @use(i32)\n\ndefine i1 @is_pow2or0_negate_op_extra_use1(i32 %x) {\n %neg = sub i32 0, %x\n call void @use(i32 %neg)\n %and = and i32 %neg, %x\n %cmp = icmp eq i32 %and, %x\n ret i1 %cmp\n}\n"}, {"test_name": "isnot_pow2_ctpop_wrong_cmp_op1_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2_ctpop_wrong_cmp_op1_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 2\n %iszero = icmp eq i32 %x, 0\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop_wrong_cmp_op2_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2_ctpop_wrong_cmp_op2_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n %iszero = icmp eq i32 %x, 1\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_ctpop_wrong_cmp_op1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_wrong_cmp_op1(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 3\n %notzero = icmp ne i32 %x, 0\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_ctpop_commute_vec_wrong_cmp_op1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>) #0\n\ndefine <2 x i1> @is_pow2or0_ctpop_commute_vec_wrong_cmp_op1(<2 x i8> %x) {\n %t0 = tail call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %x)\n %cmp = icmp eq <2 x i8> %t0, \n %iszero = icmp eq <2 x i8> %x, zeroinitializer\n %r = or <2 x i1> %cmp, %iszero\n ret <2 x i1> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop_wrong_pred2_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2_ctpop_wrong_pred2_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n %cmp2 = icmp slt i32 %x, 0\n %r = select i1 %cmp2, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_ctpop_extra_uses_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use_i1(i1)\n\ndefine i1 @is_pow2_ctpop_extra_uses_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n call void @use_i1(i1 %cmp)\n %notzero = icmp ne i32 %x, 0\n call void @use_i1(i1 %notzero)\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2or0_negate_op_commute3", "test_body": "define i1 @isnot_pow2or0_negate_op_commute3(i32 %p) {\n %x = urem i32 42, %p\n %neg = sub i32 0, %x\n %and = and i32 %x, %neg\n %cmp = icmp ne i32 %x, %and\n ret i1 %cmp\n}\n"}, {"test_name": "is_pow2_ctpop_wrong_cmp_op2_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_wrong_cmp_op2_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n %notzero = icmp ne i32 %x, 1\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_non_zero_eq_1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #1\n\ndefine i1 @is_pow2_non_zero_eq_1(i32 %x) {\n %notzero = icmp ne i32 %x, 0\n call void @llvm.assume(i1 %notzero)\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 1\n ret i1 %cmp\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\nattributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "isnot_pow2nor0_ctpop_logical_extra_uses", "test_body": "declare void @use(i32)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use_i1(i1)\n\ndefine i1 @isnot_pow2nor0_ctpop_logical_extra_uses(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n call void @use(i32 %t0)\n %cmp = icmp ne i32 %t0, 1\n call void @use_i1(i1 %cmp)\n %notzero = icmp ne i32 %x, 0\n call void @use_i1(i1 %notzero)\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_ctpop_wrong_pred2", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2or0_ctpop_wrong_pred2(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 1\n %iszero = icmp ne i32 %x, 0\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop_wrong_cmp_op2", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2_ctpop_wrong_cmp_op2(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n %iszero = icmp eq i32 %x, 1\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_ctpop_wrong_pred1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_wrong_pred1(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 2\n %notzero = icmp ne i32 %x, 0\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop_extra_uses_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use_i1(i1)\n\ndefine i1 @isnot_pow2_ctpop_extra_uses_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n call void @use_i1(i1 %cmp)\n %iszero = icmp eq i32 %x, 0\n call void @use_i1(i1 %iszero)\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_ctpop_extra_uses", "test_body": "declare void @use(i32)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use_i1(i1)\n\ndefine i1 @is_pow2or0_ctpop_extra_uses(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n call void @use(i32 %t0)\n %cmp = icmp eq i32 %t0, 1\n call void @use_i1(i1 %cmp)\n %iszero = icmp eq i32 %x, 0\n call void @use_i1(i1 %iszero)\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_non_zero_ult_2", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #1\n\ndefine i1 @is_pow2_non_zero_ult_2(i32 %x) {\n %notzero = icmp ne i32 %x, 0\n call void @llvm.assume(i1 %notzero)\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n ret i1 %cmp\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\nattributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "is_pow2_ctpop_commute_vec", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>) #0\n\ndefine <2 x i1> @is_pow2_ctpop_commute_vec(<2 x i8> %x) {\n %t0 = tail call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %x)\n %cmp = icmp ult <2 x i8> %t0, splat (i8 2)\n %notzero = icmp ne <2 x i8> %x, zeroinitializer\n %r = and <2 x i1> %cmp, %notzero\n ret <2 x i1> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_decrement_op", "test_body": "define i1 @isnot_pow2_decrement_op(i8 %x) {\n %dec = add i8 %x, -1\n %and = and i8 %dec, %x\n %cmp = icmp ne i8 %and, 0\n %iszero = icmp eq i8 %x, 0\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2nor0_ctpop_wrong_cmp_op1_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2nor0_ctpop_wrong_cmp_op1_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 5\n %notzero = icmp ne i32 %x, 0\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_ctpop_wrong_pred1_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_wrong_pred1_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 2\n %notzero = icmp ne i32 %x, 0\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2or0_decrement_op_vec", "test_body": "define <2 x i1> @isnot_pow2or0_decrement_op_vec(<2 x i8> %x) {\n %dec = add <2 x i8> %x, splat (i8 -1)\n %and = and <2 x i8> %dec, %x\n %cmp = icmp ne <2 x i8> %and, zeroinitializer\n ret <2 x i1> %cmp\n}\n"}, {"test_name": "is_pow2_ctpop_wrong_pred2_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop_wrong_pred2_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n %cmp2 = icmp sgt i32 %x, 0\n %r = select i1 %cmp2, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_isnt_p2_or_z_ugt_xy", "test_body": "define i1 @blsmsk_isnt_p2_or_z_ugt_xy(i8 %xx, i8 %yy) {\n %x = or i8 %xx, %yy\n %xm1 = add i8 %x, -1\n %y = xor i8 %x, %xm1\n %r = icmp ugt i8 %x, %y\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2nor0_ctpop_commute_vec_wrong_cmp_op1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>) #0\n\ndefine <2 x i1> @isnot_pow2nor0_ctpop_commute_vec_wrong_cmp_op1(<2 x i8> %x) {\n %t0 = tail call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %x)\n %cmp = icmp ne <2 x i8> %t0, \n %notzero = icmp ne <2 x i8> %x, zeroinitializer\n %r = and <2 x i1> %cmp, %notzero\n ret <2 x i1> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2nor0_ctpop_wrong_pred2_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2nor0_ctpop_wrong_pred2_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 1\n %notzero = icmp eq i32 %x, 0\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_negate_op_commute1", "test_body": "define i1 @is_pow2or0_negate_op_commute1(i32 %p) {\n %x = srem i32 42, %p\n %neg = sub i32 0, %x\n %and = and i32 %x, %neg\n %cmp = icmp eq i32 %and, %x\n ret i1 %cmp\n}\n"}, {"test_name": "is_pow2_or_z_known_bits", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_or_z_known_bits(i32 %xin) {\n %x = or i32 %xin, -2147483648\n %cnt = call i32 @llvm.ctpop.i32(i32 %x)\n %r = icmp ult i32 %cnt, 2\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2or0_decrement_op_vec", "test_body": "define <2 x i1> @is_pow2or0_decrement_op_vec(<2 x i8> %x) {\n %dec = add <2 x i8> %x, splat (i8 -1)\n %and = and <2 x i8> %dec, %x\n %cmp = icmp eq <2 x i8> %and, zeroinitializer\n ret <2 x i1> %cmp\n}\n"}, {"test_name": "is_pow2_decrement_op", "test_body": "define i1 @is_pow2_decrement_op(i8 %x) {\n %dec = add i8 %x, -1\n %and = and i8 %dec, %x\n %cmp = icmp eq i8 %and, 0\n %notzero = icmp ne i8 %x, 0\n %r = and i1 %cmp, %notzero\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2_ctpop_wrong_pred2", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2_ctpop_wrong_pred2(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n %cmp2 = icmp slt i32 %x, 0\n %r = or i1 %cmp2, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2_ctpop_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n %iszero = icmp eq i32 %x, 0\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2or0_negate_op_commute2", "test_body": "define i1 @isnot_pow2or0_negate_op_commute2(i32 %p) {\n %x = urem i32 42, %p\n %neg = sub i32 0, %x\n %and = and i32 %neg, %x\n %cmp = icmp ne i32 %x, %and\n ret i1 %cmp\n}\n"}, {"test_name": "blsmsk_is_p2_or_z_fail_bad_cmp", "test_body": "define i1 @blsmsk_is_p2_or_z_fail_bad_cmp(i32 %x, i32 %z) {\n %xm1 = add i32 %x, -1\n %y = xor i32 %x, %xm1\n %r = icmp uge i32 %y, %z\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2_decrement_op_vec", "test_body": "define <2 x i1> @isnot_pow2_decrement_op_vec(<2 x i8> %x) {\n %dec = add <2 x i8> %x, splat (i8 -1)\n %and = and <2 x i8> %dec, %x\n %cmp = icmp ne <2 x i8> %and, zeroinitializer\n %iszero = icmp eq <2 x i8> %x, zeroinitializer\n %r = or <2 x i1> %cmp, %iszero\n ret <2 x i1> %r\n}\n"}, {"test_name": "is_pow2or0_ctpop_wrong_cmp_op1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2or0_ctpop_wrong_cmp_op1(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 2\n %iszero = icmp eq i32 %x, 0\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_is_p2_or_z_ule_yx_fail", "test_body": "define i1 @blsmsk_is_p2_or_z_ule_yx_fail(i8 %xx, i8 %yy) {\n %x = or i8 %xx, %yy\n %xm1 = add i8 %x, -1\n %y = xor i8 %x, %xm1\n %r = icmp ule i8 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "not_pow2_or_z_known_bits", "test_body": "define <2 x i1> @not_pow2_or_z_known_bits(<2 x i32> %xin) {\n %x = or <2 x i32> %xin, splat (i32 64)\n %cnt = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %x)\n %r = icmp ugt <2 x i32> %cnt, splat (i32 1)\n ret <2 x i1> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i32> @llvm.ctpop.v2i32(<2 x i32>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_negate_op_vec", "test_body": "define <2 x i1> @isnot_pow2_negate_op_vec(<2 x i32> %x) {\n %neg = sub <2 x i32> zeroinitializer, %x\n %and = and <2 x i32> %neg, %x\n %cmp = icmp ne <2 x i32> %and, %x\n %iszero = icmp eq <2 x i32> %x, zeroinitializer\n %r = or <2 x i1> %iszero, %cmp\n ret <2 x i1> %r\n}\n"}, {"test_name": "is_pow2or0_ctpop_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2or0_ctpop_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 1\n %iszero = icmp eq i32 %x, 0\n %r = select i1 %iszero, i1 true, i1 %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop_extra_uses", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use_i1(i1)\n\ndefine i1 @isnot_pow2_ctpop_extra_uses(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n call void @use_i1(i1 %cmp)\n %iszero = icmp eq i32 %x, 0\n call void @use_i1(i1 %iszero)\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_power2_or_zero_inv_with_range", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_power2_or_zero_inv_with_range(i32 %x) {\n %ctpop = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %ctpop, 1\n %notzero = icmp ne i32 %x, 0\n %res = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %res\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2nor0_ctpop_wrong_pred2", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2nor0_ctpop_wrong_pred2(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp eq i32 %t0, 1\n %notzero = icmp eq i32 %x, 0\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "has_single_bit", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @has_single_bit(i32 %x) {\nentry:\n %cmp1 = icmp ne i32 %x, 0\n %popcnt = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)\n %cmp2 = icmp ult i32 %popcnt, 2\n %sel = select i1 %cmp1, i1 %cmp2, i1 false\n ret i1 %sel\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_fail_pr63327", "test_body": "define i1 @is_pow2_fail_pr63327(i32 %x) {\n %nx = sub i32 0, %x\n %x_and_nx = and i32 %x, %nx\n %r = icmp sge i32 %x_and_nx, %x\n ret i1 %r\n}\n"}, {"test_name": "is_pow2_ctpop", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @is_pow2_ctpop(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ult i32 %t0, 2\n %notzero = icmp ne i32 %x, 0\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "is_pow2_nz_known_bits_fail_multiuse", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndeclare void @use.i32(i32)\n\ndefine i1 @is_pow2_nz_known_bits_fail_multiuse(i32 %xin) {\n %x = or i32 %xin, 64\n %cnt = call i32 @llvm.ctpop.i32(i32 %x)\n call void @use.i32(i32 %cnt)\n %r = icmp eq i32 %cnt, 1\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_isnt_p2_or_z", "test_body": "define i1 @blsmsk_isnt_p2_or_z(i32 %x) {\n %xm1 = add i32 %x, -1\n %y = xor i32 %x, %xm1\n %r = icmp ult i32 %y, %x\n ret i1 %r\n}\n"}, {"test_name": "isnot_pow2or0_decrement_op", "test_body": "define i1 @isnot_pow2or0_decrement_op(i8 %x) {\n %dec = add i8 %x, -1\n %and = and i8 %dec, %x\n %cmp = icmp ne i8 %and, 0\n ret i1 %cmp\n}\n"}, {"test_name": "not_pow2_nz_known_bits_fail_not_p2_test", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @not_pow2_nz_known_bits_fail_not_p2_test(i32 %xin) {\n %x = or i32 %xin, 1\n %cnt = call i32 @llvm.ctpop.i32(i32 %x)\n %r = icmp ne i32 %cnt, 2\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2_ctpop_wrong_cmp_op1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2_ctpop_wrong_cmp_op1(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 2\n %iszero = icmp eq i32 %x, 0\n %r = or i1 %iszero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "blsmsk_is_p2_or_z_ule_xy", "test_body": "define i1 @blsmsk_is_p2_or_z_ule_xy(i8 %xx, i8 %yy) {\n %x = or i8 %xx, %yy\n %xm1 = add i8 %x, -1\n %y = xor i8 %x, %xm1\n %r = icmp ule i8 %x, %y\n ret i1 %r\n}\n"}, {"test_name": "is_pow2_non_zero_ugt_1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #1\n\ndefine i1 @is_pow2_non_zero_ugt_1(i32 %x) {\n %notzero = icmp ne i32 %x, 0\n call void @llvm.assume(i1 %notzero)\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ugt i32 %t0, 1\n ret i1 %cmp\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\nattributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "isnot_pow2nor0_ctpop_swap_cmp", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2nor0_ctpop_swap_cmp(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 1\n %notzero = icmp ne i32 %x, 0\n %r = and i1 %cmp, %notzero\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2nor0_ctpop_logical", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2nor0_ctpop_logical(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 1\n %notzero = icmp ne i32 %x, 0\n %r = select i1 %notzero, i1 %cmp, i1 false\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "isnot_pow2nor0_ctpop_wrong_cmp_op1", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @isnot_pow2nor0_ctpop_wrong_cmp_op1(i32 %x) {\n %t0 = tail call i32 @llvm.ctpop.i32(i32 %x)\n %cmp = icmp ne i32 %t0, 4\n %notzero = icmp ne i32 %x, 0\n %r = and i1 %notzero, %cmp\n ret i1 %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[InstCombine] `range` attribute should be dropped in `foldIsPowerOf2OrZero`", "body": "Reproducer: https://alive2.llvm.org/ce/z/u62KKP (Please use latest alive2 with https://github.com/AliveToolkit/alive2/pull/1097)\r\n```\r\ndefine i1 @src(i32 %x) {\r\n %t0 = tail call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)\r\n %cmp = icmp ne i32 %t0, 1\r\n %notzero = icmp ne i32 %x, 0\r\n %r = select i1 %notzero, i1 %cmp, i1 false\r\n ret i1 %r\r\n}\r\n\r\ndefine i1 @tgt(i32 %x) {\r\n %t0 = tail call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)\r\n %r = icmp ugt i32 %t0, 1\r\n ret i1 %r\r\n}\r\n```\r\n```\r\n----------------------------------------\r\ndefine i1 @src(i32 %x) {\r\n#0:\r\n %t0 = ctpop i32 %x\r\n %#range_0_%t0 = !range i32 %t0, i32 1, i32 33\r\n %cmp = icmp ne i32 %#range_0_%t0, 1\r\n %notzero = icmp ne i32 %x, 0\r\n %r = select i1 %notzero, i1 %cmp, i1 0\r\n ret i1 %r\r\n}\r\n=>\r\ndefine i1 @tgt(i32 %x) {\r\n#0:\r\n %t0 = ctpop i32 %x\r\n %#range_0_%t0 = !range i32 %t0, i32 1, i32 33\r\n %r = icmp ugt i32 %#range_0_%t0, 1\r\n ret i1 %r\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\ni32 %x = #x00000000 (0)\r\n\r\nSource:\r\ni32 %t0 = #x00000000 (0)\r\ni32 %#range_0_%t0 = poison\r\ni1 %cmp = poison\r\ni1 %notzero = #x0 (0)\r\ni1 %r = #x0 (0)\r\n\r\nTarget:\r\ni32 %t0 = #x00000000 (0)\r\ni32 %#range_0_%t0 = poison\r\ni1 %r = poison\r\nSource value: #x0 (0)\r\nTarget value: poison\r\n```", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "87856", "issue_url": "https://github.com/llvm/llvm-project/issues/87856", "bug_type": "crash", "base_commit": "b17d44558ba4c30a3005089b334f68593d6a9c7c", "knowledge_cutoff": "2024-04-06T02:19:05Z", "lit_test_dir": ["llvm/test/Transforms/Attributor"], "hints": {"fix_commit": "2d61692d4b75b7d92b44bfafb1011c1723955151", "components": ["Attributor"], "bug_location_lineno": {"llvm/lib/Transforms/IPO/AttributorAttributes.cpp": [[5899, 5908]]}, "bug_location_funcname": {"llvm/lib/Transforms/IPO/AttributorAttributes.cpp": ["AANoCaptureImpl::updateImpl"]}}, "patch": "commit 2d61692d4b75b7d92b44bfafb1011c1723955151\nAuthor: Matt Arsenault \nDate: Thu May 23 19:31:33 2024 +0200\n\n Attributor: Do not assume function context in AANoCapture (#91462)\n \n If the calling function has the null_pointer_is_valid attribute, somehow\n a null constant reaches here. I'm not sure why exactly, it doesn't\n happen for other types of constants.\n \n Fixes #87856\n\ndiff --git a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp\nindex 41b66aafe7d3..1b3bf3c732ed 100644\n--- a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp\n+++ b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp\n@@ -5690,6 +5690,9 @@ bool AANoCapture::isImpliedByIR(Attributor &A, const IRPosition &IRP,\n return V.use_empty();\n \n // You cannot \"capture\" null in the default address space.\n+ //\n+ // FIXME: This should use NullPointerIsDefined to account for the function\n+ // attribute.\n if (isa(V) || (isa(V) &&\n V.getType()->getPointerAddressSpace() == 0)) {\n return true;\n@@ -5899,10 +5902,13 @@ ChangeStatus AANoCaptureImpl::updateImpl(Attributor &A) {\n \n const Function *F =\n isArgumentPosition() ? IRP.getAssociatedFunction() : IRP.getAnchorScope();\n- assert(F && \"Expected a function!\");\n- const IRPosition &FnPos = IRPosition::function(*F);\n+\n+ // TODO: Is the checkForAllUses below useful for constants?\n+ if (!F)\n+ return indicatePessimisticFixpoint();\n \n AANoCapture::StateType T;\n+ const IRPosition &FnPos = IRPosition::function(*F);\n \n // Readonly means we cannot capture through memory.\n bool IsKnown;\n", "tests": [{"file": "llvm/test/Transforms/Attributor/issue87856.ll", "commands": ["opt -S -passes=attributor < %s"], "tests": [{"test_name": "", "test_body": "\ndefine void @null_ptr_is_valid_call_with_null() #0 {\n;\n call void @store_as0(ptr null)\n ret void\n}\n\ndefine void @null_ptr_is_valid_call_with_undef() #0 {\n;\n call void @store_as0(ptr undef)\n ret void\n}\n\ndefine void @store_as0(ptr %0) {\n;\n store i16 0, ptr %0, align 2\n ret void\n}\n\ndefine void @call_store_as1() {\n;\n call void @store_as1(ptr addrspace(1) null)\n ret void\n}\n\ndefine void @store_as1(ptr addrspace(1) %arg) {\n;\n store i16 0, ptr addrspace(1) %arg, align 2\n ret void\n}\n\nattributes #0 = { null_pointer_is_valid }\n;.\n;."}]}], "issue": {"title": "[attributor] Crash on `null_pointer_is_valid` function", "body": "With\r\n\r\n```\r\ntarget datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"aarch64-none-linux-android28\"\r\n\r\n; Function Attrs: null_pointer_is_valid\r\ndefine i64 @foo() #0 {\r\n call fastcc void @bar(ptr null)\r\n ret i64 0\r\n}\r\n\r\ndefine fastcc void @bar(ptr %0) {\r\n store i16 0, ptr %0, align 2\r\n ret void\r\n}\r\n\r\nattributes #0 = { null_pointer_is_valid }\r\n```\r\n\r\nand running\r\n\r\n```\r\nopt -passes=attributor file.ll\r\n```\r\n\r\nwe crash with\r\n\r\n```\r\nopt: llvm-project/llvm/lib/Transforms/IPO/AttributorAttributes.cpp:5901: virtual ChangeStatus (anonymous namespace)::AANoCaptureImpl::updateImpl(Attributor &): Assertion `F && \"Expected a function!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: opt -passes=attributor reduced.ll\r\n #0 0x000000000165bf38 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (llvm-project/build/Release/bin/opt+0x165bf38)\r\n #1 0x0000000001659a40 llvm::sys::RunSignalHandlers() (llvm-project/build/Release/bin/opt+0x1659a40)\r\n #2 0x000000000165c938 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007f3ec923e6f0 __restore_rt (/lib64/libc.so.6+0x3e6f0)\r\n #4 0x00007f3ec928b94c __pthread_kill_implementation (/lib64/libc.so.6+0x8b94c)\r\n #5 0x00007f3ec923e646 gsignal (/lib64/libc.so.6+0x3e646)\r\n #6 0x00007f3ec92287f3 abort (/lib64/libc.so.6+0x287f3)\r\n #7 0x00007f3ec922871b _nl_load_domain.cold (/lib64/libc.so.6+0x2871b)\r\n #8 0x00007f3ec9237386 (/lib64/libc.so.6+0x37386)\r\n #9 0x0000000002f2a65b (anonymous namespace)::AANoCaptureImpl::updateImpl(llvm::Attributor&) AttributorAttributes.cpp:0:0\r\n#10 0x0000000002ec8492 llvm::AbstractAttribute::update(llvm::Attributor&) (llvm-project/build/Release/bin/opt+0x2ec8492)\r\n#11 0x0000000002ed3b83 llvm::Attributor::updateAA(llvm::AbstractAttribute&) (llvm-project/build/Release/bin/opt+0x2ed3b83)\r\n#12 0x0000000002eec289 llvm::AANoCapture const* llvm::Attributor::getOrCreateAAFor(llvm::IRPosition, llvm::AbstractAttribute const*, llvm::DepClassTy, bool, bool) Attributor.cpp:0:0\r\n#13 0x0000000002f1ab57 (anonymous namespace)::AANoAliasCallSiteArgument::updateImpl(llvm::Attributor&) AttributorAttributes.cpp:0:0\r\n#14 0x0000000002ec8492 llvm::AbstractAttribute::update(llvm::Attributor&) (llvm-project/build/Release/bin/opt+0x2ec8492)\r\n#15 0x0000000002ed3b83 llvm::Attributor::updateAA(llvm::AbstractAttribute&) (llvm-project/build/Release/bin/opt+0x2ed3b83)\r\n#16 0x0000000002efba19 llvm::AANoAlias const* llvm::Attributor::getOrCreateAAFor(llvm::IRPosition, llvm::AbstractAttribute const*, llvm::DepClassTy, bool, bool) Attributor.cpp:0:0\r\n#17 0x0000000002ede880 void llvm::Attributor::checkAndQueryIRAttr<(llvm::Attribute::AttrKind)20, llvm::AANoAlias>(llvm::IRPosition const&, llvm::AttributeSet) (llvm-project/build/Release/bin/opt+0x2ede880)\r\n#18 0x0000000002edd545 llvm::Attributor::identifyDefaultAbstractAttributes(llvm::Function&) (llvm-project/build/Release/bin/opt+0x2edd545)\r\n#19 0x0000000002ee0800 runAttributorOnFunctions(llvm::InformationCache&, llvm::SetVector, llvm::DenseSet>, 0u>&, llvm::AnalysisGetter&, llvm::CallGraphUpdater&, bool, bool) Attributor.cpp:0:0\r\n#20 0x0000000002edfc41 llvm::AttributorPass::run(llvm::Module&, llvm::AnalysisManager&) (llvm-project/build/Release/bin/opt+0x2edfc41)\r\n#21 0x0000000002b6fcdd llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) PassBuilderPipelines.cpp:0:0\r\n#22 0x0000000001853e56 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (llvm-project/build/Release/bin/opt+0x1853e56)\r\n#23 0x0000000002afaa4b llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (llvm-project/build/Release/bin/opt+0x2afaa4b)\r\n#24 0x0000000001622f36 optMain (llvm-project/build/Release/bin/opt+0x1622f36)\r\n```", "author": "lanza", "labels": ["crash-on-valid", "llvm::Attributor"], "comments": [{"author": "lanza", "body": "https://github.com/llvm/llvm-project/blob/0f52f4ddd909eb38f2a691ffed8469263fe5f635/llvm/lib/Transforms/IPO/AttributorAttributes.cpp#L5901"}, {"author": "dtcxzyw", "body": "cc @arsenm "}, {"author": "arsenm", "body": "Just returning on the null function seems to work, but I don't understand exactly why it reaches here. AANoCapture::isImpliedByIR is incorrectly reimplementing NullPointerIsDefined but fixing that doesn't seem to help any "}]}, "verified": true} {"bug_id": "126974", "issue_url": "https://github.com/llvm/llvm-project/issues/126974", "bug_type": "miscompilation", "base_commit": "dab9156923133b4ce3c40efcae4f80b0d720e72f", "knowledge_cutoff": "2025-02-12T21:21:07Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "29f3a352068ce562bcb65e18a676c82a9991583c", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": [[5637, 5642]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": ["InstCombinerImpl::foldICmpWithMinMax"]}}, "patch": "commit 29f3a352068ce562bcb65e18a676c82a9991583c\nAuthor: Yingwei Zheng \nDate: Sun Feb 16 20:18:29 2025 +0800\n\n [InstCombine] Do not keep samesign when speculatively executing icmps (#127007)\n \n Closes https://github.com/llvm/llvm-project/issues/126974.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\nindex 96d6db2ba5bf..76020d2b1dbf 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n@@ -5637,6 +5637,11 @@ Instruction *InstCombinerImpl::foldICmpWithMinMax(Instruction &I,\n return false;\n return std::nullopt;\n };\n+ // Remove samesign here since it is illegal to keep it when we speculatively\n+ // execute comparisons. For example, `icmp samesign ult umax(X, -46), -32`\n+ // cannot be decomposed into `(icmp samesign ult X, -46) or (icmp samesign ult\n+ // -46, -32)`. `X` is allowed to be non-negative here.\n+ Pred = static_cast(Pred);\n auto CmpXZ = IsCondKnownTrue(simplifyICmpInst(Pred, X, Z, Q));\n auto CmpYZ = IsCondKnownTrue(simplifyICmpInst(Pred, Y, Z, Q));\n if (!CmpXZ.has_value() && !CmpYZ.has_value())\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/umax-icmp.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "pr126974", "test_body": "define i1 @pr126974(i8 %x) {\nentry:\n %cond = icmp sgt i8 %x, -2\n br i1 %cond, label %if.then, label %if.else\n\nif.then: ; preds = %entry\n %umax = call i8 @llvm.umax.i8(i8 %x, i8 -46)\n %cmp = icmp samesign ult i8 %umax, -32\n ret i1 %cmp\n\nif.else: ; preds = %entry\n ret i1 false\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i8 @llvm.umax.i8(i8, i8) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[RISC-V] Miscompile using rv64gcv", "body": "Testcase:\n```c\nint a[][21][21];\nvoid b(int c, int d, int e, short f, _Bool g, char h, char i, int j, int k,\n int l, unsigned char m, int n, char o, int p, long long q[],\n unsigned short r[][10], unsigned s[][10],\n unsigned long long t[][10][10][10], unsigned long long u[][10][10][10],\n int w[][10][10][10][10], _Bool x) {\n for (signed z = 0; z < (char)m + 3; z = 4)\n for (short aa = ({\n __typeof__(0) ab = 173;\n __typeof__(0) ac = ({\n __typeof__(0) ab = m;\n ab;\n });\n ab > ac ? ab : ac;\n }) -\n 171;\n aa < 9; aa = k)\n a[z][aa][aa] = ~0;\n}\ntypedef int ad;\nint printf(const char *, ...);\nlong long ae;\nvoid af(long long *ae, int v) { *ae ^= v; }\nint c;\nint d;\nint e;\nshort f;\n_Bool g;\nchar h;\nchar i;\nint j;\nint k = 2805750831;\nint l;\nchar m;\nint n;\nchar o;\nint p;\nlong long q[0];\nunsigned short r[0][10];\nunsigned s[0][10];\nunsigned long long t[0][10][10][10];\nunsigned long long u[0][10][10][10];\nint w[0][10][10][10][10];\n_Bool x;\nint a[21][21][21];\nvoid ag() {\n for (ad y = 0; y < 10; ++y)\n for (ad ah = 0; ah < 10; ++ah)\n for (ad ai = 0; ai < 10; ++ai)\n af(&ae, a[y][ah][ai]);\n}\nint main() {\n b(c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, w, x);\n ag();\n printf(\"%llu\\n\", ae);\n}\n```\n\nCommands:\n```\n# riscv \n$ QEMU_CPU=rv64,vlen=128,rvv_ta_all_1s=true,rvv_ma_all_1s=true,v=true,vext_spec=v1.0,zve32f=true,zve64f=true timeout --verbose -k 0.1 4 /scratch/ewlu/daily-upstream-build/build-gcv/bin/qemu-riscv64 user-config.out 1\n0\n\n# x86\n$ ./native.out 1\n18446744073709551615\n```\n\nGodbolt: https://godbolt.org/z/85rz18rxY\n\nBisected to cfee344dda7394631f2177a15e56cfeee1d61fc4 as the first bad commit\n\nI tried reducing the testcase down farther but was unsuccessful\n\nFound via fuzzer. First detected 01/24/2025", "author": "ewlu", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "dzaima", "body": "Some manual simplification, reproduces on x86-64 `-O3` too:\n\n```c\n#include\n\nvoid f(int more_than_20, unsigned char zero) {\n int i = 0;\n while (i < (signed char)zero + 2) {\n int max = zero < 210 ? 210 : zero; // 210; has to be >127\n int j = max - 200; // 10\n while (j < 20) {\n printf(\"I should be printed once\\n\");\n j = more_than_20;\n }\n i = 2; // can be anything >2\n }\n}\n\nint main() {\n f(50, 0);\n}\n```\n\nhttps://godbolt.org/z/qfn9PT6r5"}, {"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/adparc"}, {"author": "dtcxzyw", "body": "Further reduced: https://alive2.llvm.org/ce/z/5da83j"}]}, "verified": true} {"bug_id": "107139", "issue_url": "https://github.com/llvm/llvm-project/issues/107139", "bug_type": "crash", "base_commit": "5d17293caaf0f62ea94fecc137b9b6f07c659dac", "knowledge_cutoff": "2024-09-03T18:28:59Z", "lit_test_dir": ["llvm/test/Transforms/Coroutines"], "hints": {"fix_commit": "853bff2122e1f42398587f76234c05d56f16318e", "components": ["Coroutines"], "bug_location_lineno": {"llvm/lib/Transforms/Coroutines/CoroSplit.cpp": [[2080, 2091], [2105, 2117], [2200, 2205], [2211, 2217], [2223, 2236]]}, "bug_location_funcname": {"llvm/lib/Transforms/Coroutines/CoroSplit.cpp": ["CoroSplitPass::run", "replacePrepare", "splitCoroutine", "updateCallGraphAfterCoroutineSplit"]}}, "patch": "commit 853bff2122e1f42398587f76234c05d56f16318e\nAuthor: Yuxuan Chen \nDate: Thu Sep 12 10:45:20 2024 -0700\n\n [Coroutines] properly update CallGraph in CoroSplit (#107935)\n \n Fixes https://github.com/llvm/llvm-project/issues/107139.\n \n We weren't updating the call graph properly in CoroSplit. This crash is\n due to the await_suspend() function calling the coroutine, forming a\n multi-node SCC. The issue bisected to\n https://github.com/llvm/llvm-project/pull/79712 but I think this is red\n herring. We haven't been properly updating the call graph.\n \n Added an example of such code as a test case.\n\ndiff --git a/llvm/lib/Transforms/Coroutines/CoroSplit.cpp b/llvm/lib/Transforms/Coroutines/CoroSplit.cpp\nindex dc3829d7f28e..8ea460badaad 100644\n--- a/llvm/lib/Transforms/Coroutines/CoroSplit.cpp\n+++ b/llvm/lib/Transforms/Coroutines/CoroSplit.cpp\n@@ -2080,12 +2080,13 @@ splitCoroutine(Function &F, SmallVectorImpl &Clones,\n return Shape;\n }\n \n-static void updateCallGraphAfterCoroutineSplit(\n+static LazyCallGraph::SCC &updateCallGraphAfterCoroutineSplit(\n LazyCallGraph::Node &N, const coro::Shape &Shape,\n const SmallVectorImpl &Clones, LazyCallGraph::SCC &C,\n LazyCallGraph &CG, CGSCCAnalysisManager &AM, CGSCCUpdateResult &UR,\n FunctionAnalysisManager &FAM) {\n \n+ auto *CurrentSCC = &C;\n if (!Clones.empty()) {\n switch (Shape.ABI) {\n case coro::ABI::Switch:\n@@ -2105,13 +2106,16 @@ static void updateCallGraphAfterCoroutineSplit(\n }\n \n // Let the CGSCC infra handle the changes to the original function.\n- updateCGAndAnalysisManagerForCGSCCPass(CG, C, N, AM, UR, FAM);\n+ CurrentSCC = &updateCGAndAnalysisManagerForCGSCCPass(CG, *CurrentSCC, N, AM,\n+ UR, FAM);\n }\n \n // Do some cleanup and let the CGSCC infra see if we've cleaned up any edges\n // to the split functions.\n postSplitCleanup(N.getFunction());\n- updateCGAndAnalysisManagerForFunctionPass(CG, C, N, AM, UR, FAM);\n+ CurrentSCC = &updateCGAndAnalysisManagerForFunctionPass(CG, *CurrentSCC, N,\n+ AM, UR, FAM);\n+ return *CurrentSCC;\n }\n \n /// Replace a call to llvm.coro.prepare.retcon.\n@@ -2200,6 +2204,7 @@ PreservedAnalyses CoroSplitPass::run(LazyCallGraph::SCC &C,\n if (Coroutines.empty() && PrepareFns.empty())\n return PreservedAnalyses::all();\n \n+ auto *CurrentSCC = &C;\n // Split all the coroutines.\n for (LazyCallGraph::Node *N : Coroutines) {\n Function &F = N->getFunction();\n@@ -2211,7 +2216,8 @@ PreservedAnalyses CoroSplitPass::run(LazyCallGraph::SCC &C,\n coro::Shape Shape =\n splitCoroutine(F, Clones, FAM.getResult(F),\n OptimizeFrame, MaterializableCallback);\n- updateCallGraphAfterCoroutineSplit(*N, Shape, Clones, C, CG, AM, UR, FAM);\n+ CurrentSCC = &updateCallGraphAfterCoroutineSplit(\n+ *N, Shape, Clones, *CurrentSCC, CG, AM, UR, FAM);\n \n auto &ORE = FAM.getResult(F);\n ORE.emit([&]() {\n@@ -2223,14 +2229,14 @@ PreservedAnalyses CoroSplitPass::run(LazyCallGraph::SCC &C,\n \n if (!Shape.CoroSuspends.empty()) {\n // Run the CGSCC pipeline on the original and newly split functions.\n- UR.CWorklist.insert(&C);\n+ UR.CWorklist.insert(CurrentSCC);\n for (Function *Clone : Clones)\n UR.CWorklist.insert(CG.lookupSCC(CG.get(*Clone)));\n }\n }\n \n for (auto *PrepareFn : PrepareFns) {\n- replaceAllPrepares(PrepareFn, CG, C);\n+ replaceAllPrepares(PrepareFn, CG, *CurrentSCC);\n }\n \n return PreservedAnalyses::none();\n", "tests": [{"file": "llvm/test/Transforms/Coroutines/gh107139-split-in-scc.ll", "commands": ["opt < %s -passes='cgscc(coro-split)' -S"], "tests": [{"test_name": "", "test_body": "; Verify that we don't crash on mutually recursive coroutines\n\ntarget triple = \"x86_64-redhat-linux-gnu\"\n\ndefine void @foo() presplitcoroutine personality ptr null {\nentry:\n\n %0 = call token @llvm.coro.id(i32 0, ptr null, ptr null, ptr null)\n %1 = call ptr @llvm.coro.begin(token %0, ptr null)\n %2 = call token @llvm.coro.save(ptr null)\n %3 = call i8 @llvm.coro.suspend(token none, i1 false)\n %4 = call token @llvm.coro.save(ptr null)\n call void @llvm.coro.await.suspend.void(ptr null, ptr null, ptr @bar)\n ret void\n}\n\ndefine void @bar(ptr %0, ptr %1) {\nentry:\n call void @foo()\n ret void\n}\n\n\ndeclare token @llvm.coro.id(i32, ptr readnone, ptr nocapture readonly, ptr) #0\ndeclare ptr @llvm.coro.begin(token, ptr writeonly) nounwind\ndeclare token @llvm.coro.save(ptr) nomerge nounwind\ndeclare void @llvm.coro.await.suspend.void(ptr, ptr, ptr)\ndeclare i8 @llvm.coro.suspend(token, i1) nounwind\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(argmem: read) }"}]}], "issue": {"title": "\"Invalidated the current SCC\" assertion failure after #79712", "body": "For the source file in https://gist.github.com/smeenai/fcd14380467e813a28f1ef8c5484b775 (reduced from a large example that uses libunifex; it's probably possible to reduce this further but I'm not at all familiar with coroutines), the following command succeeds with Clang 18 but fails with Clang 19 and trunk:\r\n\r\n```\r\n$ clang -c -o /dev/null coroutine_crash.cpp -std=c++20 -w\r\nclang: llvm/lib/Analysis/CGSCCPassManager.cpp:1167: LazyCallGraph::SCC &updateCGAndAnalysisManagerForPass(llvm::LazyCallGraph &, LazyCallGraph::SCC &, LazyCallGraph::Node &, llvm::CGSCCAnalysisManager &, llvm::CGSCCUpdateResult &, llvm::FunctionAnalysisManager &, bool): Assertion `!UR.InvalidatedSCCs.count(C) && \"Invalidated the current SCC!\"' failed.\r\n```\r\n\r\nI bisected this to #79712. CC @fpasserby, @ChuanqiXu9, and @yuxuanchen1997 ", "author": "smeenai", "labels": ["crash", "coroutines", "llvm:transforms"], "comments": [{"author": "yuxuanchen1997", "body": "I took a look at this and I think it was due to incorrect call graph updates. `updateCGAndAnalysisManagerFor(CGSCC|Function)Pass` seems to return an updated SCC but we discarded the updated SCC.\r\n\r\nSee example usage: https://github.com/llvm/llvm-project/blob/main/llvm/lib/Analysis/CGSCCPassManager.cpp#L561-L562\r\nSee implementation: https://github.com/llvm/llvm-project/blob/main/llvm/lib/Analysis/CGSCCPassManager.cpp#L1170-L1175\r\n\r\nI have a patch https://github.com/llvm/llvm-project/pull/107935 but still need some time for a test case. "}, {"author": "yuxuanchen1997", "body": "This happens when foo has is not a single node SCC:\r\n```\r\ndefine void @foo() presplitcoroutine personality ptr null {\r\nentry:\r\n %0 = call token @llvm.coro.id(i32 0, ptr null, ptr null, ptr null)\r\n %1 = call ptr @llvm.coro.begin(token %0, ptr null)\r\n %2 = call token @llvm.coro.save(ptr null)\r\n %3 = call i8 @llvm.coro.suspend(token none, i1 false)\r\n %4 = call token @llvm.coro.save(ptr null)\r\n call void @llvm.coro.await.suspend.void(ptr null, ptr null, ptr @bar)\r\n ret void\r\n}\r\n\r\ndefine void @bar(ptr %0, ptr %1) {\r\nentry:\r\n call void @foo()\r\n ret void\r\n}\r\n\r\ndeclare token @llvm.coro.id(i32, ptr readnone, ptr nocapture readonly, ptr) #0\r\ndeclare ptr @llvm.coro.begin(token, ptr writeonly) nounwind\r\ndeclare token @llvm.coro.save(ptr) nomerge nounwind\r\ndeclare void @llvm.coro.await.suspend.void(ptr, ptr, ptr)\r\ndeclare i8 @llvm.coro.suspend(token, i1) nounwind\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(argmem: read) }\r\n```\r\n\r\nRepro the crash with `clang -O0 bad.ll`."}, {"author": "yuxuanchen1997", "body": "The 3rd argument to `@llvm.coro.await.suspend` is the `await_suspend` function in C++. This happens when `await_suspend()` calls the same coroutine from its body. "}]}, "verified": true} {"bug_id": "87378", "issue_url": "https://github.com/llvm/llvm-project/issues/87378", "bug_type": "crash", "base_commit": "d8db13edd60b195e6692deb04d63e633319e486a", "knowledge_cutoff": "2024-04-02T17:38:53Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "6261c53c6f9c82bfca5df2ed5917876d748ed079", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h": [[175, 181]], "llvm/lib/Transforms/Vectorize/VPlan.h": [[913, 918], [920, 928], [1016, 1021], [1221, 1226]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h": ["createOr", "createSelect"], "llvm/lib/Transforms/Vectorize/VPlan.h": ["GEPFlagsTy", "VPInstruction", "VPRecipeWithIRFlags"]}}, "patch": "commit 6261c53c6f9c82bfca5df2ed5917876d748ed079\nAuthor: Florian Hahn \nDate: Tue Apr 2 21:48:35 2024 +0100\n\n [VPlan] Make sure OR VPInstructions are treated as disjoint ops.\n \n Make sure that VPInstructions with OR opcodes are properly registered as\n disjoint ops.\n \n Fixes https://github.com/llvm/llvm-project/issues/87378.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h b/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h\nindex 5d03b66b0ce3..ebca2d855a46 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h\n@@ -175,7 +175,10 @@ public:\n \n VPValue *createOr(VPValue *LHS, VPValue *RHS, DebugLoc DL = {},\n const Twine &Name = \"\") {\n- return createInstruction(Instruction::BinaryOps::Or, {LHS, RHS}, DL, Name);\n+\n+ return tryInsertInstruction(new VPInstruction(\n+ Instruction::BinaryOps::Or, {LHS, RHS},\n+ VPRecipeWithIRFlags::DisjointFlagsTy(false), DL, Name));\n }\n \n VPValue *createSelect(VPValue *Cond, VPValue *TrueVal, VPValue *FalseVal,\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h\nindex 3baca43f7276..707a826ecdc2 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlan.h\n+++ b/llvm/lib/Transforms/Vectorize/VPlan.h\n@@ -913,6 +913,11 @@ public:\n WrapFlagsTy(bool HasNUW, bool HasNSW) : HasNUW(HasNUW), HasNSW(HasNSW) {}\n };\n \n+ struct DisjointFlagsTy {\n+ char IsDisjoint : 1;\n+ DisjointFlagsTy(bool IsDisjoint) : IsDisjoint(IsDisjoint) {}\n+ };\n+\n protected:\n struct GEPFlagsTy {\n char IsInBounds : 1;\n@@ -920,9 +925,6 @@ protected:\n };\n \n private:\n- struct DisjointFlagsTy {\n- char IsDisjoint : 1;\n- };\n struct ExactFlagsTy {\n char IsExact : 1;\n };\n@@ -1016,6 +1018,12 @@ public:\n : VPSingleDefRecipe(SC, Operands, DL), OpType(OperationType::FPMathOp),\n FMFs(FMFs) {}\n \n+ template \n+ VPRecipeWithIRFlags(const unsigned char SC, IterT Operands,\n+ DisjointFlagsTy DisjointFlags, DebugLoc DL = {})\n+ : VPSingleDefRecipe(SC, Operands, DL), OpType(OperationType::DisjointOp),\n+ DisjointFlags(DisjointFlags) {}\n+\n protected:\n template \n VPRecipeWithIRFlags(const unsigned char SC, IterT Operands,\n@@ -1221,6 +1229,14 @@ public:\n : VPRecipeWithIRFlags(VPDef::VPInstructionSC, Operands, WrapFlags, DL),\n Opcode(Opcode), Name(Name.str()) {}\n \n+ VPInstruction(unsigned Opcode, std::initializer_list Operands,\n+ DisjointFlagsTy DisjointFlag, DebugLoc DL = {},\n+ const Twine &Name = \"\")\n+ : VPRecipeWithIRFlags(VPDef::VPInstructionSC, Operands, DisjointFlag, DL),\n+ Opcode(Opcode), Name(Name.str()) {\n+ assert(Opcode == Instruction::Or && \"only OR opcodes can be disjoint\");\n+ }\n+\n VPInstruction(unsigned Opcode, std::initializer_list Operands,\n FastMathFlags FMFs, DebugLoc DL = {}, const Twine &Name = \"\");\n \n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/pr87378-vpinstruction-or-drop-poison-generating-flags.ll", "commands": ["opt -p loop-vectorize -mattr=\"+v\" -S %s"], "tests": [{"test_name": "", "test_body": "\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\n; Test case for https://github.com/llvm/llvm-project/issues/87378.\ndefine void @pr87378_vpinstruction_or_drop_poison_generating_flags(ptr %arg, i64 %a, i64 %b, i64 %c) {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %c.1 = icmp ule i64 %iv, %a\n br i1 %c.1, label %then.1, label %else.1\n\nthen.1:\n %c.2 = icmp ule i64 %iv, %b\n br i1 %c.2, label %else.1, label %merge\n\nelse.1:\n %c.3 = icmp ule i64 %iv, %c\n br i1 %c.3, label %then.2, label %loop.latch\n\nthen.2:\n br label %merge\n\nmerge:\n %idx = phi i64 [ poison, %then.1 ], [ %iv, %then.2 ]\n %getelementptr = getelementptr i16, ptr %arg, i64 %idx\n store i16 0, ptr %getelementptr, align 2\n br label %loop.latch\n\nloop.latch:\n %iv.next = add i64 %iv, 1\n %icmp = icmp eq i64 %iv, 1000\n br i1 %icmp, label %exit, label %loop.header\n\nexit:\n ret void\n}\n;.\n;."}]}], "issue": {"title": "[LoopVectorize] Assertion `OpType == OperationType::DisjointOp && \"recipe cannot have a disjoing flag\"' failed.", "body": "Reduced LLVM IR:\r\n``` llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main() #0 {\r\nentry:\r\n br label %for.body300.us.us.i\r\n\r\nfor.body300.us.us.i: ; preds = %if.end559.us.us.i, %entry\r\n br label %for.body326.us.us.i\r\n\r\nfor.body326.us.us.i: ; preds = %if.end559.us.us.i, %for.body300.us.us.i\r\n %indvars.iv18 = phi i64 [ 0, %for.body300.us.us.i ], [ %indvars.iv.next19, %if.end559.us.us.i ]\r\n br i1 false, label %cond.end339.us.us.i, label %cond.end354.us.us.i\r\n\r\ncond.end339.us.us.i: ; preds = %for.body326.us.us.i\r\n br i1 false, label %cond.end339.us.us.i.if.then357.us.us.i_crit_edge, label %cond.end339.us.us.i.cond.end354.us.us.i_crit_edge\r\n\r\ncond.end339.us.us.i.cond.end354.us.us.i_crit_edge: ; preds = %cond.end339.us.us.i\r\n br label %cond.end354.us.us.i\r\n\r\ncond.end339.us.us.i.if.then357.us.us.i_crit_edge: ; preds = %cond.end339.us.us.i\r\n br label %if.then357.us.us.i\r\n\r\ncond.end354.us.us.i: ; preds = %cond.end339.us.us.i.cond.end354.us.us.i_crit_edge, %for.body326.us.us.i\r\n br i1 false, label %cond.end354.us.us.i.if.then357.us.us.i_crit_edge, label %if.end559.us.us.i\r\n\r\ncond.end354.us.us.i.if.then357.us.us.i_crit_edge: ; preds = %cond.end354.us.us.i\r\n br label %if.then357.us.us.i\r\n\r\nif.then357.us.us.i: ; preds = %cond.end354.us.us.i.if.then357.us.us.i_crit_edge, %cond.end339.us.us.i.if.then357.us.us.i_crit_edge\r\n %.pre1143.i7 = phi i64 [ poison, %cond.end339.us.us.i.if.then357.us.us.i_crit_edge ], [ %indvars.iv18, %cond.end354.us.us.i.if.then357.us.us.i_crit_edge ]\r\n %arrayidx429.us.us.i = getelementptr [14 x i16], ptr null, i64 0, i64 %.pre1143.i7\r\n store i16 0, ptr %arrayidx429.us.us.i, align 2\r\n br label %if.end559.us.us.i\r\n\r\nif.end559.us.us.i: ; preds = %if.then357.us.us.i, %cond.end354.us.us.i\r\n %indvars.iv.next19 = add i64 %indvars.iv18, 1\r\n %0 = trunc i64 %indvars.iv18 to i32\r\n %cmp323.us.us.i = icmp slt i32 %0, 1\r\n br i1 %cmp323.us.us.i, label %for.body326.us.us.i, label %for.body300.us.us.i\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl256b,+zvl32b,+zvl64b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zaamo,-experimental-zabha,-experimental-zalasr,-experimental-zalrsc,-experimental-zcmop,-experimental-zfbfmin,-experimental-zicfilp,-experimental-zicfiss,-experimental-zimop,-experimental-ztso,-experimental-zvfbfmin,-experimental-zvfbfwma,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smepmp,-ssaia,-ssccptr,-sscofpmf,-sscounterenw,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-za128rs,-za64rs,-zacas,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmp,-zcmt,-zdinx,-zfa,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-zmmul,-zvbb,-zvbc,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nBacktrace:\r\n```\r\nopt: /scratch/tc-testing/tc-apr-2/llvm/llvm/lib/Transforms/Vectorize/VPlan.h:1131: bool llvm::VPRecipeWithIRFlags::isDisjoint() const: Assertion `OpType == OperationType::DisjointOp && \"recipe cannot have a disjoing flag\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt reduced.ll --passes=loop-vectorize\r\n #0 0x00005bd204995b60 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2d84b60)\r\n #1 0x00005bd204992f6f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2d81f6f)\r\n #2 0x00005bd2049930c5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x000076eef1a42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x000076eef1a969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x000076eef1a969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x000076eef1a969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x000076eef1a42476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x000076eef1a287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x000076eef1a2871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x000076eef1a39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00005bd203b25ea2 llvm::VPlanTransforms::dropPoisonGeneratingRecipes(llvm::VPlan&, llvm::function_ref)::'lambda'(llvm::VPRecipeBase*)::operator()(llvm::VPRecipeBase*) const VPlanTransforms.cpp:0:0\r\n#12 0x00005bd203b2c77d llvm::VPlanTransforms::dropPoisonGeneratingRecipes(llvm::VPlan&, llvm::function_ref) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1f1b77d)\r\n#13 0x00005bd203a05435 llvm::LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(llvm::VFRange&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1df4435)\r\n#14 0x00005bd203a05c9e llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1df4c9e)\r\n#15 0x00005bd203a0c319 llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1dfb319)\r\n#16 0x00005bd203a0f8a9 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1dfe8a9)\r\n#17 0x00005bd203a1273e llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1e0173e)\r\n#18 0x00005bd203a1388d llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x1e0288d)\r\n#19 0x00005bd20291a276 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd09276)\r\n#20 0x00005bd2047bc141 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2bab141)\r\n#21 0x00005bd202911bd6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd00bd6)\r\n#22 0x00005bd2047bae3b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2ba9e3b)\r\n#23 0x00005bd202919dc6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0xd08dc6)\r\n#24 0x00005bd2047b8cb1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x2ba7cb1)\r\n#25 0x00005bd202193ce5 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x582ce5)\r\n#26 0x00005bd202186316 optMain (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x575316)\r\n#27 0x000076eef1a29d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#28 0x000076eef1a29e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#29 0x000076eef1a29e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#30 0x00005bd20217bf95 _start (/scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt+0x56af95)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-apr-2/build-rv64gcv/build-llvm-linux/bin/opt reduced.l\r\n```\r\n\r\nGodbolt: https://godbolt.org/z/jj39h3z7M\r\n\r\nMinor spelling nit: `disjoing -> disjoint`\r\n\r\nFound via fuzzer.\r\n\r\nLikely caused/exposed by https://github.com/llvm/llvm-project/commit/6ef829941b38f7e8a28c4cba1ff25cd0ae9f7d3d https://github.com/llvm/llvm-project/pull/83821 @fhahn ", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "113143", "issue_url": "https://github.com/llvm/llvm-project/issues/113143", "bug_type": "crash", "base_commit": "42ba452aa94e4da277842d8990ad958a6256e558", "knowledge_cutoff": "2024-10-21T10:11:54Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "9e03920cbf946e7ba282e99213707643a23ae5fb", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[12466, 12472]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::isGatherShuffledSingleRegisterEntry"]}}, "patch": "commit 9e03920cbf946e7ba282e99213707643a23ae5fb\nAuthor: Alexey Bataev \nDate: Mon Oct 21 08:57:36 2024 -0700\n\n [SLP]Ignore root gather node, when searching for reuses\n \n Root gather/buildvector node should be ignored when SLP vectorizer tries\n to find matching gather nodes, vectorized earlier. This node is\n definitely the last one in the pipeline and it does not have users. It\n may cause the compiler crash\n \n Fixes #113143\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 1098bf578d2d..a11e3f3815cb 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -12466,7 +12466,7 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(\n // Build a list of tree entries where V is used.\n SmallPtrSet VToTEs;\n for (const TreeEntry *TEPtr : ValueToGatherNodes.find(V)->second) {\n- if (TEPtr == TE)\n+ if (TEPtr == TE || TEPtr->Idx == 0)\n continue;\n assert(any_of(TEPtr->Scalars,\n [&](Value *V) { return GatheredScalars.contains(V); }) &&\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/root-gather-reused-scalar.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define void @test(ptr %a, i32 %0, i32 %1, i1 %cmp1) {\nentry:\n %2 = load i32, ptr %a, align 4\n %tobool10.not = icmp eq i32 %0, 0\n %cmp4.3 = icmp ne i32 %1, 0\n %3 = and i1 %cmp4.3, %tobool10.not\n %cmp2.2 = xor i1 %tobool10.not, true\n %conv3.2 = zext i1 %cmp2.2 to i32\n %cmp4.2 = icmp ne i32 %2, %conv3.2\n %cmp2.1 = xor i1 %cmp1, true\n %conv3.1 = zext i1 %cmp2.1 to i32\n %cmp4.1 = icmp ne i32 %2, %conv3.1\n %4 = select i1 %3, i1 %cmp4.2, i1 false\n %5 = select i1 %4, i1 %cmp4.1, i1 false\n %and.3 = zext i1 %5 to i32\n store i32 %and.3, ptr %a, align 4\n ret void\n}\n"}]}], "issue": {"title": "clang crashes on valid code at -O2 on x86_64-linux-gnu: Assertion `TEPtr->UserTreeIndices.size() == 1 && \"Expected only single user of a gather node.\"' failed ", "body": "It appears to be a recent regression as it doesn't reproduce with 19.1.0 and earlier. \r\n\r\nCompiler Explorer: https://godbolt.org/z/a1W3c5ard\r\n\r\n```\r\n[523] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git df8b785838a2db01b4d056e603f7317209accefb)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[524] % \r\n[524] % clangtk -O2 small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:12559: std::optional llvm::slpvectorizer::BoUpSLP::isGatherShuffledSingleRegisterEntry(const llvm::slpvectorizer::BoUpSLP::TreeEntry*, llvm::ArrayRef, llvm::MutableArrayRef, llvm::SmallVectorImpl&, unsigned int, bool): Assertion `TEPtr->UserTreeIndices.size() == 1 && \"Expected only single user of a gather node.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20241020-clangtk-m64-O2-build-231331/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20241020-clangtk-m64-O2-build-231331/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-572b89.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4.\tRunning pass \"slp-vectorizer\" on function \"main\"\r\n #0 0x0000560576c8b9af llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x43f29af)\r\n #1 0x0000560576c89184 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f4623c27420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007f462365e00b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007f462363d859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007f462363d729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007f462363d729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007f462364efd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x0000560578707edc llvm::slpvectorizer::BoUpSLP::isGatherShuffledSingleRegisterEntry(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::MutableArrayRef, llvm::SmallVectorImpl&, unsigned int, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5e6eedc)\r\n #9 0x0000560578708cd8 llvm::slpvectorizer::BoUpSLP::isGatherShuffledEntry(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallVectorImpl&, llvm::SmallVectorImpl>&, unsigned int, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5e6fcd8)\r\n#10 0x000056057870a6f3 llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::processBuildVector, llvm::slpvectorizer::BoUpSLP, llvm::SmallPtrSetImpl>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::TargetTransformInfo&, llvm::ArrayRef&, llvm::slpvectorizer::BoUpSLP&, llvm::SmallPtrSetImpl&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5e716f3)\r\n#11 0x000056057870cde7 llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5e73de7)\r\n#12 0x0000560578745f69 llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5eacf69)\r\n#13 0x000056057874b57b (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#14 0x000056057874eb79 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::SmallVectorImpl&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5eb5b79)\r\n#15 0x000056057875349b llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\r\n#16 0x0000560578757d50 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5ebed50)\r\n#17 0x000056057875e590 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#18 0x000056057875f263 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5ec6263)\r\n#19 0x00005605781e0e66 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5947e66)\r\n#20 0x0000560576614891 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3d7b891)\r\n#21 0x000056057410e656 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1875656)\r\n#22 0x00005605766131fd llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3d7a1fd)\r\n#23 0x000056057410d196 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1874196)\r\n#24 0x0000560576612c31 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3d79c31)\r\n#25 0x0000560576f44dd3 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#26 0x0000560576f483a8 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x46af3a8)\r\n#27 0x000056057764462c clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4dab62c)\r\n#28 0x000056057951629c clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x6c7d29c)\r\n#29 0x0000560577644a48 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4daba48)\r\n#30 0x000056057790d3d9 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x50743d9)\r\n#31 0x00005605778895ce clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4ff05ce)\r\n#32 0x00005605779f9f26 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5160f26)\r\n#33 0x0000560573cd1fe5 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1438fe5)\r\n#34 0x0000560573cc9afa ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#35 0x0000560573cce52e clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x143552e)\r\n#36 0x0000560573bba67b main (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x132167b)\r\n#37 0x00007f462363f083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#38 0x0000560573cc958e _start (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x143058e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git df8b785838a2db01b4d056e603f7317209accefb)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-d99d4e.c\r\nclangtk: note: diagnostic msg: /tmp/small-d99d4e.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[525] % \r\n[525] % cat small.c\r\nint a, b, c, d, e, f;\r\nint main() {\r\n for (f = 0; f < 4; f++) {\r\n d &= (((a > 0) != b) != c) && b;\r\n for (b = 0; b < 1; b++)\r\n if (e)\r\n break;\r\n }\r\n return 0;\r\n}\r\n```\r\n\r\n", "author": "zhendongsu", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "107036", "issue_url": "https://github.com/llvm/llvm-project/issues/107036", "bug_type": "crash", "base_commit": "70a19adbc60c738903bbbb8e6d5ef2d41b681089", "knowledge_cutoff": "2024-09-03T01:22:22Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "f381cd069965dabfeb277f30a4e532d7fd498f6e", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[16055, 16069]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::computeMinimumValueSizes"]}}, "patch": "commit f381cd069965dabfeb277f30a4e532d7fd498f6e\nAuthor: Alexey Bataev \nDate: Tue Sep 3 04:52:47 2024 -0700\n\n [SLP]Fix PR107036: Check if the type of the user is sizable before requesting its size.\n \n Only some instructions should be considered as potentially reducing the\n size of the operands types, not all instructions should be considered.\n \n Fixes https://github.com/llvm/llvm-project/issues/107036\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 58137cd6f543..f58803fc56a2 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -16055,15 +16055,16 @@ void BoUpSLP::computeMinimumValueSizes() {\n const TreeEntry *UserTE = E.UserTreeIndices.back().UserTE;\n if (TE == UserTE || !TE)\n return false;\n+ if (!isa(U) ||\n+ !isa(UserTE->getMainOp()))\n+ return true;\n unsigned UserTESz = DL->getTypeSizeInBits(\n UserTE->Scalars.front()->getType());\n auto It = MinBWs.find(TE);\n if (It != MinBWs.end() && It->second.first > UserTESz)\n return true;\n- // The size of icmp is always 1 and should not be\n- // considered.\n- if (TE->getOpcode() == Instruction::ICmp)\n- return true;\n return DL->getTypeSizeInBits(U->getType()) > UserTESz;\n }));\n })) {\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/minbw-user-non-sizable.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s -slp-threshold=-100"], "tests": [{"test_name": "test", "test_body": "define void @test(ptr %i) {\nbb:\n %i1 = getelementptr i8, ptr %i, i64 4\n br label %bb2\n\nbb2: ; preds = %bb2, %bb\n %i3 = phi i32 [ %i6, %bb2 ], [ 0, %bb ]\n %i4 = phi i32 [ %i8, %bb2 ], [ 0, %bb ]\n store i32 %i3, ptr %i, align 4\n store i32 %i4, ptr %i1, align 4\n %i5 = trunc i32 0 to i1\n %i6 = select i1 %i5, i32 0, i32 0\n %i7 = trunc i32 %i4 to i1\n %i8 = select i1 %i7, i32 0, i32 0\n br label %bb2\n}\n"}]}], "issue": {"title": "[SLP] Cannot getTypeInfo() on a type that is unsized!", "body": "I tried the following IR with SLP:\r\n\r\n```llvm\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @foo(ptr %i) {\r\nbb:\r\n %i1 = getelementptr i8, ptr %i, i64 4\r\n br label %bb2\r\n\r\nbb2: ; preds = %bb2, %bb\r\n %i3 = phi i32 [ %i6, %bb2 ], [ 0, %bb ]\r\n %i4 = phi i32 [ %i8, %bb2 ], [ 0, %bb ]\r\n store i32 %i3, ptr %i\r\n store i32 %i4, ptr %i1\r\n %i5 = trunc i32 0 to i1\r\n %i6 = select i1 %i5, i32 0, i32 0\r\n %i7 = trunc i32 %i4 to i1\r\n %i8 = select i1 %i7, i32 0, i32 0\r\n br label %bb2\r\n}\r\n```\r\n\r\nIt failed with an error output:`Ty->isSized() && \"Cannot getTypeInfo() on a type that is unsized!\"' failed`. It's getting the size returned by the `store i32 %i4, ptr %i1` instruction.​\r\n\r\nhttps://llvm.godbolt.org/z/fMx9qMcfK\r\n\r\ncc f6e01b9ece1e73f6eda6e1dbff3aa72e917f4007 @alexey-bataev\r\n", "author": "DianQK", "labels": ["release:backport", "llvm:SLPVectorizer", "crash-on-valid"], "comments": [{"author": "dtcxzyw", "body": "https://github.com/llvm/llvm-project/blob/2cbd1bc830861bc08f78fb6cc11747b82f66c4c6/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp#L16067\r\n\r\n`U` is `store i32 %i4, ptr %i1`.\r\n"}, {"author": "DianQK", "body": "Reverting f6e01b9ece1e73f6eda6e1dbff3aa72e917f4007 seems reasonable to me. What do you think? @alexey-bataev ^^"}, {"author": "alexey-bataev", "body": "will fix it asap today"}]}, "verified": true} {"bug_id": "127128", "issue_url": "https://github.com/llvm/llvm-project/issues/127128", "bug_type": "crash", "base_commit": "1ff5f328d9824694cc356ebf78adad8816a6de86", "knowledge_cutoff": "2025-02-13T20:59:01Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "ac217ee389d63124432e5e6890851a678f7a676b", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[13206, 13213]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::isGatherShuffledSingleRegisterEntry"]}}, "patch": "commit ac217ee389d63124432e5e6890851a678f7a676b\nAuthor: Alexey Bataev \nDate: Thu Feb 13 14:19:51 2025 -0800\n\n [SLP] Check for PHI nodes (potentially cycles!) when checking dependencies\n \n When checking for dependecies for gather nodes with users with the same\n last instruction, cannot rely on the index order, if there is (even\n potential!) cycle in the graph, which may cause order not work correctly\n and cause compiler crash.\n \n Fixes #127128\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex fb72a112b9ac..780cafe91d6b 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -13206,8 +13206,16 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(\n continue;\n // If the user instruction is used for some reason in different\n // vectorized nodes - make it depend on index.\n+ // If any vector node is PHI node, this dependency might not work\n+ // because of cycle dependencies, so disable it.\n if (TEUseEI.UserTE != UseEI.UserTE &&\n- TEUseEI.UserTE->Idx < UseEI.UserTE->Idx)\n+ (TEUseEI.UserTE->Idx < UseEI.UserTE->Idx ||\n+ any_of(\n+ VectorizableTree,\n+ [](const std::unique_ptr &TE) {\n+ return TE->State == TreeEntry::Vectorize &&\n+ TE->getOpcode() == Instruction::PHI;\n+ })))\n continue;\n }\n \n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/delayed-gather-emission.ll", "commands": ["opt -passes=slp-vectorizer -mtriple=x86_64 -mattr=+avx2 -S < %s"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nentry:\n %.pre.pre = load float, ptr poison, align 4\n br label %bb1\n\nbb1: ; preds = %bb2, %entry\n %.pre = phi float [ %.pre.pre, %entry ], [ %i2, %bb2 ]\n %foxtrot.0 = phi float [ undef, %entry ], [ %gulf.0, %bb2 ]\n br label %bb2\n\nbb2: ; preds = %bb2, %bb1\n %i = phi float [ %.pre, %bb1 ], [ %i2, %bb2 ]\n %gulf.0 = phi float [ %foxtrot.0, %bb1 ], [ %div, %bb2 ]\n %div = fdiv float %i, %gulf.0\n %i1 = load float, ptr poison, align 4\n %div1 = fdiv float %gulf.0, %i1\n %mul = fmul float %div, %div1\n tail call void @foo(float %mul)\n %i2 = load float, ptr poison, align 4\n %tobool = fcmp une float %i2, 0.000000e+00\n br i1 %tobool, label %bb1, label %bb2\n}\n\ndeclare void @foo(float)\n"}]}, {"file": "llvm/test/Transforms/SLPVectorizer/X86/phi-node-with-cycle.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -mcpu=haswell < %s"], "tests": [{"test_name": "", "test_body": "\ndefine void @test(float %0) {\n;\n %2 = fdiv float 0.000000e+00, 0.000000e+00\n %3 = fdiv float 0.000000e+00, 0.000000e+00\n %4 = fdiv float %0, 0.000000e+00\n br label %5\n\n5:\n %6 = fmul float %4, 0.000000e+00\n %7 = fsub float 0.000000e+00, %6\n %8 = fmul float %3, 0.000000e+00\n %9 = fsub float 0.000000e+00, %8\n br label %11\n\n10:\n br label %11\n\n11:\n %12 = phi float [ %7, %5 ], [ 0.000000e+00, %10 ]\n %13 = phi float [ %9, %5 ], [ 0.000000e+00, %10 ]\n br label %14\n\n14:\n %15 = fmul float %2, 0.000000e+00\n %16 = fsub float %12, %15\n %17 = fmul float %4, 0.000000e+00\n %18 = fsub float %13, %17\n %19 = fadd float %16, %18\n %20 = call float @llvm.fabs.f32(float %19)\n ret void\n}\n"}]}], "issue": {"title": "SLPVectorizer miscompilation", "body": "```llvm\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:10:11:12:13\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine swiftcc { ptr addrspace(10), i8 } @julia__dir2ori_404(float %0) {\n %2 = fdiv float 0.000000e+00, 0.000000e+00\n %3 = fdiv float 0.000000e+00, 0.000000e+00\n %4 = fdiv float %0, 0.000000e+00\n br label %5\n\n5: ; preds = %1\n %6 = fmul float %4, 0.000000e+00\n %7 = fsub float 0.000000e+00, %6\n %8 = fmul float %3, 0.000000e+00\n %9 = fsub float 0.000000e+00, %8\n br label %11\n\n10: ; No predecessors!\n br label %11\n\n11: ; preds = %10, %5\n %12 = phi float [ %7, %5 ], [ 0.000000e+00, %10 ]\n %13 = phi float [ %9, %5 ], [ 0.000000e+00, %10 ]\n br label %14\n\n14: ; preds = %11\n %15 = fmul float %2, 0.000000e+00\n %16 = fsub float %12, %15\n %17 = fmul float %4, 0.000000e+00\n %18 = fsub float %13, %17\n %19 = fadd float %16, %18\n %20 = call float @llvm.fabs.f32(float %19)\n ret { ptr addrspace(10), i8 } zeroinitializer\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.fabs.f32(float) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n```\nproduces \n```\nInstruction does not dominate all uses!\n %4 = insertelement <2 x float> , float %0, i32 0\n %2 = shufflevector <2 x float> %4, <2 x float> , <2 x i32> \nLLVM ERROR: Broken module found, compilation aborted!\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\nStack dump:\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer -mcpu=haswell \n1.\tRunning pass \"verify\" on module \"\"\n #0 0x00000000054edec8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x54edec8)\n #1 0x00000000054eb884 SignalHandler(int, siginfo_t*, void*) Signals.cpp:0:0\n #2 0x00007faaea842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #3 0x00007faaea8969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #4 0x00007faaea842476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #5 0x00007faaea8287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #6 0x000000000082790d llvm::json::operator==(llvm::json::Value const&, llvm::json::Value const&) (.cold) JSON.cpp:0:0\n #7 0x000000000542b6e8 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x542b6e8)\n #8 0x0000000005329278 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x5329278)\n #9 0x000000000095605e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x95605e)\n#10 0x00000000052e7b50 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x52e7b50)\n#11 0x000000000096174a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x96174a)\n#12 0x0000000000953c7c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x953c7c)\n#13 0x00007faaea829d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#14 0x00007faaea829e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#15 0x000000000094b555 _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x94b555)\nProgram terminated with signal: SIGSEGV\nCompiler returned: 139\n```\non 19 and trunk.\nGodbolt: https://godbolt.org/z/qxKP6198T", "author": "Zentrik", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": [{"author": "Zentrik", "body": "Bisected thanks to manyclangs to 66775f8ccdcc8264ef349518e1c59d96d4227823 (8e2b2c4181506efc5b9321c203dd107bbd63392b seemed to be good) @alexey-bataev"}]}, "verified": true} {"bug_id": "109333", "issue_url": "https://github.com/llvm/llvm-project/issues/109333", "bug_type": "crash", "base_commit": "c498af71d73e96e95730f3b3caaea08d3edbe62c", "knowledge_cutoff": "2024-09-19T20:50:34Z", "lit_test_dir": ["llvm/test/Transforms/LoopUnroll"], "hints": {"fix_commit": "5bcc82d43388bb0daa122d5fe7ecda5eca27fc16", "components": ["LoopPeel"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/LoopPeel.cpp": [[859, 865]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/LoopPeel.cpp": ["cloneLoopBlocks"]}}, "patch": "commit 5bcc82d43388bb0daa122d5fe7ecda5eca27fc16\nAuthor: Nikita Popov \nDate: Fri Sep 20 16:57:46 2024 +0200\n\n [LoopPeel] Fix LCSSA phi node invalidation\n \n In the test case, the BECount of the second loop uses %load,\n but we only have an LCSSA phi node for %add, so that is what\n gets invalidated. Use the forgetLcssaPhiWithNewPredecessor()\n API instead, which will invalidate the roots of the expression\n instead.\n \n Fixes https://github.com/llvm/llvm-project/issues/109333.\n\ndiff --git a/llvm/lib/Transforms/Utils/LoopPeel.cpp b/llvm/lib/Transforms/Utils/LoopPeel.cpp\nindex 5d7c0d947fac..760f1619e030 100644\n--- a/llvm/lib/Transforms/Utils/LoopPeel.cpp\n+++ b/llvm/lib/Transforms/Utils/LoopPeel.cpp\n@@ -859,7 +859,7 @@ static void cloneLoopBlocks(\n if (LatchInst && L->contains(LatchInst))\n LatchVal = VMap[LatchVal];\n PHI.addIncoming(LatchVal, cast(VMap[Edge.first]));\n- SE.forgetValue(&PHI);\n+ SE.forgetLcssaPhiWithNewPredecessor(L, &PHI);\n }\n \n // LastValueMap is updated with the values for the current loop\n", "tests": [{"file": "llvm/test/Transforms/LoopUnroll/pr109333.ll", "commands": ["opt -S -passes=\"print,loop-unroll\" -unroll-runtime < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "define void @test(i1 %c, ptr %p) {\nentry:\n br label %loop.1\n\nloop.1: ; preds = %loop.1, %entry\n %phi = phi ptr [ null, %entry ], [ %p, %loop.1 ]\n %load = load i64, ptr %p, align 8\n %add = add i64 %load, 1\n br i1 %c, label %if, label %loop.1\n\nif: ; preds = %loop.1\n %add.lcssa = phi i64 [ %add, %loop.1 ]\n %gep = getelementptr i64, ptr %p, i64 %add.lcssa\n br label %loop.2\n\nloop.2: ; preds = %loop.2, %if\n %iv = phi ptr [ %p, %if ], [ %iv.next, %loop.2 ]\n %iv.next = getelementptr i8, ptr %iv, i64 8\n %icmp = icmp eq ptr %iv, %gep\n br i1 %icmp, label %exit, label %loop.2\n\nexit: ; preds = %loop.2\n ret void\n}\n"}]}], "issue": {"title": "LTO error: Instruction does not dominate all uses", "body": "clang crashes while linking\r\n```\r\nclang: error: unable to execute command: Aborted (core dumped)\r\nclang: error: linker command failed due to signal (use -v to see invocation)\r\n```\r\nI followed the instructions [here](https://llvm.org/docs/HowToSubmitABug.html) and narrowed it down to an LTO bug. File to reproduce is attached. [a.out.0.2.internalize-reduced.bc.tar.gz](https://github.com/user-attachments/files/17066318/a.out.0.2.internalize-reduced.bc.tar.gz)\r\n```\r\n$ opt \"-passes=lto\" a.out.0.2.internalize-reduced.bc\r\nWARNING: You're attempting to print out a bitcode file.\r\nThis is inadvisable as it may cause display problems. If\r\nyou REALLY want to taste LLVM bitcode first-hand, you\r\ncan force output with the `-f' option.\r\n\r\nInstruction does not dominate all uses!\r\n %.lcssa16 = phi i64 [ %12, %9 ]\r\n %27 = add i64 %.lcssa16, 2\r\nLLVM ERROR: Broken module found, compilation aborted!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: opt -passes=lto reduced-redacted.bc #0 0x00007e899efc4d90 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) /usr/src/debug/llvm/llvm-18.1.8.src/lib/Support/Unix/Signals.inc:723:22\r\n #1 0x00007e899efc1d4d llvm::sys::RunSignalHandlers() /usr/src/debug/llvm/llvm-18.1.8.src/lib/Support/Signals.cpp:105:20\r\n #2 0x00007e899efc1d4d SignalHandler /usr/src/debug/llvm/llvm-18.1.8.src/lib/Support/Unix/Signals.inc:403:31\r\n #3 0x00007e899e24c1d0 (/usr/lib/libc.so.6+0x3d1d0)\r\n #4 0x00007e899e2a53f4 __pthread_kill_implementation /usr/src/debug/glibc/glibc/nptl/pthread_kill.c:44:76\r\n #5 0x00007e899e24c120 raise /usr/src/debug/glibc/glibc/signal/../sysdeps/posix/raise.c:27:6\r\n #6 0x00007e899e2334c3 abort /usr/src/debug/glibc/glibc/stdlib/abort.c:81:7\r\n #7 0x00007e899ecaba1b std::mutex::lock() /usr/include/c++/14.1.1/bits/std_mutex.h:117:22\r\n #8 0x00007e899ecaba1b std::lock_guard::lock_guard(std::mutex&) /usr/include/c++/14.1.1/bits/std_mutex.h:250:23\r\n #9 0x00007e899ecaba1b llvm::install_bad_alloc_error_handler(void (*)(void*, char const*, bool), void*) /usr/src/debug/llvm/llvm-18.1.8.src/lib/Support/ErrorHandling.cpp:131:61\r\n#10 0x00007e899eec070e (/usr/lib/libLLVM.so.18.1+0x6c070e)\r\n#11 0x00007e899f21aa1b (/usr/lib/libLLVM.so.18.1+0xa1aa1b)\r\n#12 0x000055a86e4acf57 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) /usr/src/debug/llvm/llvm-18.1.8.src/include/llvm/IR/PassManagerInternal.h:90:3\r\n#13 0x00007e899f1d767e llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) /usr/src/debug/llvm/llvm-18.1.8.src/incl\r\nude/llvm/IR/PassManager.h:547:20\r\n#14 0x000055a86e4b7052 llvm::SmallPtrSetImplBase::isSmall() const /usr/src/debug/llvm/llvm-18.1.8.src/include/llvm/ADT/SmallPtrSet.h:195:33\r\n#15 0x000055a86e4b7052 llvm::SmallPtrSetImplBase::~SmallPtrSetImplBase() /usr/src/debug/llvm/llvm-18.1.8.src/include/llvm/ADT/SmallPtrSet.h:83:17\r\n#16 0x000055a86e4b7052 llvm::SmallPtrSetImpl::~SmallPtrSetImpl() /usr/src/debug/llvm/llvm-18.1.8.src/include/llvm/ADT/SmallPtrSet.h:345:7\r\n#17 0x000055a86e4b7052 llvm::SmallPtrSet::~SmallPtrSet() /usr/src/debug/llvm/llvm-18.1.8.src/include/llvm/ADT/SmallPtrSet.h:451:7\r\n#18 0x000055a86e4b7052 llvm::PreservedAnalyses::~PreservedAnalyses() /usr/src/debug/llvm/llvm-18.1.8.src/include/llvm/IR/PassManager.h:172:7\r\n#19 0x000055a86e4b7052 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutpu\r\ntFile*, llvm::StringRef, llvm::ArrayRef, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) /usr/src/debug/llvm/llvm-18.1.8\r\n.src/tools/opt/NewPMDriver.cpp:527:10\r\n#20 0x000055a86e4ab1ed std::__cxx11::basic_string, std::allocator>::~basic_string() /usr/include/c++/14.1.1/bits/basic_string.h:809:19\r\n#21 0x000055a86e4ab1ed main /usr/src/debug/llvm/llvm-18.1.8.src/tools/opt/opt.cpp:747:3\r\n#22 0x00007e899e234e08 __libc_start_call_main /usr/src/debug/glibc/glibc/csu/../sysdeps/nptl/libc_start_call_main.h:74:3\r\n#23 0x00007e899e234ecc call_init /usr/src/debug/glibc/glibc/csu/../csu/libc-start.c:128:20\r\n#24 0x00007e899e234ecc __libc_start_main /usr/src/debug/glibc/glibc/csu/../csu/libc-start.c:347:5\r\n#25 0x000055a86e4abfd5 _start (/usr/bin/opt+0x1cfd5)\r\nAborted (core dumped)\r\n```\r\nThis happened after adding `-fsplit-lto-unit` to some parts of my code base.\r\n", "author": "googlebleh", "labels": ["loopoptim", "release:backport", "llvm:crash"], "comments": [{"author": "googlebleh", "body": "While bisecting using this reduced test case, I was able to reproduce this crash on several versions of LLVM between 15 and 18.1.8. I was also able to reproduce on the tip of `main` (e439fdf4ea0dbc6f001428f4d4956700bf26bb97). However, this crash did not reproduce on 15.0.7. The commit on 15.0.7 that \"fixed\" the issue is\r\n```\r\ncommit f3c5289e78462fb96015f79c954d95a0d527ba55 \r\nAuthor: Martin Storsjö \r\nDate: Wed Oct 5 14:44:21 2022 +0300 \r\n \r\n Revert \"Recommit \"[SCEV] Look through single value PHIs.\" (take 3)\" \r\n \r\n This reverts commit 20d798bd47ec5191de1b2a8a031da06a04e612e1. \r\n \r\n This commit caused crashes in some cases, see github issue #58152. \r\n This is fixed on main, but backporting it requires multiple \r\n nontrivial cherrypicks. \r\n \r\n Updating llvm/test/Transforms/LoopVectorize/create-induction-resume.ll \r\n with update_test_checks.py, so this isn't an exact automatic revert, \r\n as that test case was added after the reverted commit. \r\n \r\n This fixes #58152 for the release branch. \r\n \r\n llvm/lib/Analysis/ScalarEvolution.cpp | 7 ++-\r\n llvm/test/Analysis/DependenceAnalysis/lcssa.ll | 2 +- \r\n llvm/test/Analysis/ScalarEvolution/cycled_phis.ll | 4 +- \r\n .../ScalarEvolution/incorrect-exit-count.ll | 2 +- \r\n .../Analysis/ScalarEvolution/solve-quadratic-i1.ll | 4 +- \r\n .../ScalarEvolution/solve-quadratic-overflow.ll | 6 +-- \r\n llvm/test/Analysis/ScalarEvolution/trivial-phis.ll | 2 +- \r\n llvm/test/Transforms/LoopStrengthReduce/funclet.ll | 40 +++++++++-------\r\n .../LoopVectorize/create-induction-resume.ll | 24 ++++------\r\n llvm/test/Transforms/LoopVectorize/pr45259.ll | 55 +++++++++++-----------\r\n 10 files changed, 75 insertions(+), 71 deletions(-)\r\n```\r\nspecifically, this (almost) 1-line diff\r\n```diff\r\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\r\nindex 4b2db80bc1ec..3b73f9511eca 100644\r\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\r\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\r\n@@ -6023,8 +6023,13 @@ const SCEV *ScalarEvolution::createNodeForPHI(PHINode *PN) {\r\n if (const SCEV *S = createAddRecFromPHI(PN))\r\n return S;\r\n \r\n+ // If the PHI has a single incoming value, follow that value, unless the\r\n+ // PHI's incoming blocks are in a different loop, in which case doing so\r\n+ // risks breaking LCSSA form. Instcombine would normally zap these, but\r\n+ // it doesn't have DominatorTree information, so it may miss cases.\r\n if (Value *V = simplifyInstruction(PN, {getDataLayout(), &TLI, &DT, &AC}))\r\n- return getSCEV(V);\r\n+ if (LI.replacementPreservesLCSSAForm(PN, V))\r\n+ return getSCEV(V);\r\n \r\n if (const SCEV *S = createNodeFromSelectLikePHI(PN))\r\n return S;\r\n```\r\n"}, {"author": "nikic", "body": "Reduced:\r\n```llvm\r\n; RUN: opt -passes=\"print,loop-unroll\" -unroll-runtime\r\ndefine void @test(i1 %c, ptr %p) {\r\nentry:\r\n br label %loop\r\n\r\nloop:\r\n %phi = phi ptr [ null, %entry ], [ %p, %loop ]\r\n %load = load i64, ptr %p, align 8\r\n %add = add i64 %load, 1\r\n br i1 %c, label %bb2, label %loop\r\n\r\nbb2:\r\n %add.lcssa = phi i64 [ %add, %loop ]\r\n %gep = getelementptr i64, ptr %p, i64 %add.lcssa\r\n br label %loop2\r\n\r\nloop2:\r\n %iv = phi ptr [ %p, %bb2 ], [ %iv.next, %loop2 ]\r\n %iv.next = getelementptr i8, ptr %iv, i64 8\r\n %icmp = icmp eq ptr %iv, %gep\r\n br i1 %icmp, label %exit, label %loop2\r\n\r\nexit:\r\n ret void\r\n}\r\n```"}, {"author": "nikic", "body": "From a cursory look, the problem here is that the BECount of `loop2` depends on `%load`, and after peeling it should instead depend on a phi node, but instead still depends on `%load`. As the BECount does not dominate the loop, SCEV is invalid.\r\n\r\nThis issue sounds very familiar, we've fixed variants of this in the past."}]}, "verified": true} {"bug_id": "124387", "issue_url": "https://github.com/llvm/llvm-project/issues/124387", "bug_type": "miscompilation", "base_commit": "77c325b646301e394bcd89c2980b4c2da8af49cd", "knowledge_cutoff": "2025-01-25T03:24:37Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "2131115be5b9d8b39af80973d9b64c0adc41d38d", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": [[1039, 1049]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": ["InstCombinerImpl::SimplifyDemandedUseBits"]}}, "patch": "commit 2131115be5b9d8b39af80973d9b64c0adc41d38d\nAuthor: Fangrui Song \nDate: Sat Jan 25 13:35:11 2025 -0800\n\n [InstCombine] Drop Range attribute when simplifying 'fshl' based on demanded bits (#124429)\n \n When simplifying operands based on demanded bits, the return value range\n of llvm.fshl might change. Keeping the Range attribute might cause\n llvm.fshl to generate a poison and lead to miscompile. Drop the Range\n attribute similar to `dropPosonGeneratingFlags` elsewhere.\n \n Fix #124387\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\nindex 943598a30f04..2c8939b5a051 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n@@ -1039,11 +1039,14 @@ Value *InstCombinerImpl::SimplifyDemandedUseBits(Instruction *I,\n APInt DemandedMaskLHS(DemandedMask.lshr(ShiftAmt));\n APInt DemandedMaskRHS(DemandedMask.shl(BitWidth - ShiftAmt));\n if (I->getOperand(0) != I->getOperand(1)) {\n- if (SimplifyDemandedBits(I, 0, DemandedMaskLHS, LHSKnown,\n- Depth + 1, Q) ||\n+ if (SimplifyDemandedBits(I, 0, DemandedMaskLHS, LHSKnown, Depth + 1,\n+ Q) ||\n SimplifyDemandedBits(I, 1, DemandedMaskRHS, RHSKnown, Depth + 1,\n- Q))\n+ Q)) {\n+ // Range attribute may no longer hold.\n+ I->dropPoisonGeneratingReturnAttributes();\n return I;\n+ }\n } else { // fshl is a rotate\n // Avoid converting rotate into funnel shift.\n // Only simplify if one operand is constant.\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/fsh.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "fshl_range_trunc", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.fshl.i32(i32, i32, i32) #0\n\ndefine i8 @fshl_range_trunc(i1 %x) {\n %zext = zext i1 %x to i32\n %or = or disjoint i32 %zext, -2\n %fshl = call range(i32 -4, 2) i32 @llvm.fshl.i32(i32 %or, i32 %or, i32 1)\n %tr = trunc nsw i32 %fshl to i8\n ret i8 %tr\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[clang] Miscompile with -O2/3/s", "body": "This code prints random value at `-O2/3/s` and -1 at `-O0/1`:\n\n```c\nint printf(const char *, ...);\nint a, b;\nvoid c(char d) { a = d; }\nint e(int d) {\n if (d < 0)\n return 1;\n return 0;\n}\nint f() {\n if (b)\n return 0;\n return 1;\n}\nint g(int d) {\n int h = 0;\n if (3 + d)\n h = f() - 1 - 1;\n return e(h) + h + h;\n}\nint main() {\n int i = g(0);\n c(i);\n printf(\"%d\\n\", a);\n}\n```\n\nCompiler Explorer: https://godbolt.org/z/KfWsKxcvE\n\nIt starts from x86_64 clang 19.1.0. ", "author": "cardigan1008", "labels": ["miscompilation", "llvm:instcombine", "regression:19"], "comments": [{"author": "MaskRay", "body": "InstCombine miscompiles \n\n```llvm\ndefine i32 @src(i32 %x) local_unnamed_addr {\nentry:\n %cmp = icmp eq i32 %x, 0\n %zext = zext i1 %cmp to i32\n %or = or disjoint i32 %zext, -2\n %fshl = call range(i32 -4, 2) i32 @llvm.fshl.i32(i32 %or, i32 %or, i32 1)\n %tr = trunc nsw i32 %fshl to i8\n %res = sext i8 %tr to i32\n ret i32 %res\n}\n```\n\nedit: `g` has a Range return attribute, which is propagated to the call site in `main` after inlining.\nInstCombine `InstCombinerImpl::SimplifyDemandedUseBits` simplifies the operand, but keeps the no-longer-valid Range return attribute.\n\nThen, CorrelatedValuePropagationPass added nsw, and the value eventually becomes poison."}, {"author": "nikic", "body": "Slightly reduced: https://alive2.llvm.org/ce/z/3Kf7-2\n\nI assume the problem is that we don't clear the return range attribute in SimplifyDemanded."}]}, "verified": true} {"bug_id": "91691", "issue_url": "https://github.com/llvm/llvm-project/issues/91691", "bug_type": "miscompilation", "base_commit": "1fadb2b0c881ced247931f442fdee6c4ed96dccb", "knowledge_cutoff": "2024-05-10T02:43:43Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "b5f4210e9f51f938ae517f219f04f9ab431a2684", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[3343, 3349], [3366, 3371], [3379, 3384], [3448, 3461]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["foldBitCeil", "isSafeToRemoveBitCeilSelect"]}}, "patch": "commit b5f4210e9f51f938ae517f219f04f9ab431a2684\nAuthor: Yingwei Zheng \nDate: Mon May 13 14:27:59 2024 +0800\n\n [InstCombine] Drop nuw flag when CtlzOp is a sub nuw (#91776)\n \n See the following case:\n ```\n define i32 @src1(i32 %x) {\n %dec = sub nuw i32 -2, %x\n %ctlz = tail call i32 @llvm.ctlz.i32(i32 %dec, i1 false)\n %sub = sub nsw i32 32, %ctlz\n %shl = shl i32 1, %sub\n %ugt = icmp ult i32 %x, -2\n %sel = select i1 %ugt, i32 %shl, i32 1\n ret i32 %sel\n }\n \n define i32 @tgt1(i32 %x) {\n %dec = sub nuw i32 -2, %x\n %ctlz = tail call i32 @llvm.ctlz.i32(i32 %dec, i1 false)\n %sub = sub nsw i32 32, %ctlz\n %and = and i32 %sub, 31\n %shl = shl nuw i32 1, %and\n ret i32 %shl\n }\n ```\n `nuw` in `%dec` should be dropped after the select instruction is\n eliminated.\n \n Alive2: https://alive2.llvm.org/ce/z/7S9529\n \n Fixes https://github.com/llvm/llvm-project/issues/91691.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex ee090e012508..a3ddb402bf66 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -3343,7 +3343,8 @@ Instruction *InstCombinerImpl::foldSelectOfBools(SelectInst &SI) {\n // pattern.\n static bool isSafeToRemoveBitCeilSelect(ICmpInst::Predicate Pred, Value *Cond0,\n const APInt *Cond1, Value *CtlzOp,\n- unsigned BitWidth) {\n+ unsigned BitWidth,\n+ bool &ShouldDropNUW) {\n // The challenge in recognizing std::bit_ceil(X) is that the operand is used\n // for the CTLZ proper and select condition, each possibly with some\n // operation like add and sub.\n@@ -3366,6 +3367,8 @@ static bool isSafeToRemoveBitCeilSelect(ICmpInst::Predicate Pred, Value *Cond0,\n ConstantRange CR = ConstantRange::makeExactICmpRegion(\n CmpInst::getInversePredicate(Pred), *Cond1);\n \n+ ShouldDropNUW = false;\n+\n // Match the operation that's used to compute CtlzOp from CommonAncestor. If\n // CtlzOp == CommonAncestor, return true as no operation is needed. If a\n // match is found, execute the operation on CR, update CR, and return true.\n@@ -3379,6 +3382,7 @@ static bool isSafeToRemoveBitCeilSelect(ICmpInst::Predicate Pred, Value *Cond0,\n return true;\n }\n if (match(CtlzOp, m_Sub(m_APInt(C), m_Specific(CommonAncestor)))) {\n+ ShouldDropNUW = true;\n CR = ConstantRange(*C).sub(CR);\n return true;\n }\n@@ -3448,14 +3452,20 @@ static Instruction *foldBitCeil(SelectInst &SI, IRBuilderBase &Builder) {\n Pred = CmpInst::getInversePredicate(Pred);\n }\n \n+ bool ShouldDropNUW;\n+\n if (!match(FalseVal, m_One()) ||\n !match(TrueVal,\n m_OneUse(m_Shl(m_One(), m_OneUse(m_Sub(m_SpecificInt(BitWidth),\n m_Value(Ctlz)))))) ||\n !match(Ctlz, m_Intrinsic(m_Value(CtlzOp), m_Zero())) ||\n- !isSafeToRemoveBitCeilSelect(Pred, Cond0, Cond1, CtlzOp, BitWidth))\n+ !isSafeToRemoveBitCeilSelect(Pred, Cond0, Cond1, CtlzOp, BitWidth,\n+ ShouldDropNUW))\n return nullptr;\n \n+ if (ShouldDropNUW)\n+ cast(CtlzOp)->setHasNoUnsignedWrap(false);\n+\n // Build 1 << (-CTLZ & (BitWidth-1)). The negation likely corresponds to a\n // single hardware instruction as opposed to BitWidth - CTLZ, where BitWidth\n // is an integer constant. Masking with BitWidth-1 comes free on some\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/bit_ceil.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "pr91691", "test_body": "define i32 @pr91691(i32 %0) {\n %2 = sub nuw i32 -2, %0\n %3 = tail call i32 @llvm.ctlz.i32(i32 %2, i1 false)\n %4 = sub i32 32, %3\n %5 = shl i32 1, %4\n %6 = icmp ult i32 %0, -2\n %7 = select i1 %6, i32 %5, i32 1\n ret i32 %7\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctlz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "bit_ceil_v4i32", "test_body": "define <4 x i32> @bit_ceil_v4i32(<4 x i32> %x) {\n %dec = add <4 x i32> %x, splat (i32 -1)\n %ctlz = tail call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %dec, i1 false)\n %sub = sub <4 x i32> splat (i32 32), %ctlz\n %shl = shl <4 x i32> splat (i32 1), %sub\n %ugt = icmp ugt <4 x i32> %x, splat (i32 1)\n %sel = select <4 x i1> %ugt, <4 x i32> %shl, <4 x i32> splat (i32 1)\n ret <4 x i32> %sel\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i32> @llvm.ctlz.v4i32(<4 x i32>, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "pr91691_keep_nsw", "test_body": "define i32 @pr91691_keep_nsw(i32 %0) {\n %2 = sub nsw i32 -2, %0\n %3 = tail call i32 @llvm.ctlz.i32(i32 %2, i1 false)\n %4 = sub i32 32, %3\n %5 = shl i32 1, %4\n %6 = icmp ult i32 %0, -2\n %7 = select i1 %6, i32 %5, i32 1\n ret i32 %7\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctlz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "instcombine removes a select, making code more poisonous", "body": "https://alive2.llvm.org/ce/z/c__jy8\r\n\r\nthis function:\r\n```llvm\r\ndefine i32 @f(i32 %0) {\r\n %2 = sub nuw i32 -2, %0\r\n %3 = tail call i32 @llvm.ctlz.i32(i32 %2, i1 false)\r\n %4 = sub i32 32, %3\r\n %5 = shl i32 1, %4\r\n %6 = icmp ult i32 %0, -2\r\n %7 = select i1 %6, i32 %5, i32 1\r\n ret i32 %7\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i32 @llvm.ctlz.i32(i32, i1 immarg) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\n\r\nis getting rewritten to not have the select, but the select was blocking a poison value. the bad thing happens when -1 is passed as an argument, see the Alive link for a detailed execution trace\r\n\r\n```llvm\r\ndefine i32 @f(i32 %0) {\r\n %2 = sub nuw i32 -2, %0\r\n %3 = tail call range(i32 0, 33) i32 @llvm.ctlz.i32(i32 %2, i1 false)\r\n %4 = sub nsw i32 0, %3\r\n %5 = and i32 %4, 31\r\n %6 = shl nuw i32 1, %5\r\n ret i32 %6\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i32 @llvm.ctlz.i32(i32, i1 immarg) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\n\r\ncc @nunoplopes ", "author": "regehr", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "106626", "issue_url": "https://github.com/llvm/llvm-project/issues/106626", "bug_type": "crash", "base_commit": "e51fc36c385d756ccbc2be8c47c52af207c3aead", "knowledge_cutoff": "2024-08-29T20:51:42Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "cc943a67d114e28c28f561c3b1a48ff2003264ce", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[3237, 3251]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["findLaneForValue"]}}, "patch": "commit cc943a67d114e28c28f561c3b1a48ff2003264ce\nAuthor: Alexey Bataev \nDate: Thu Aug 29 15:05:09 2024 -0700\n\n [SLP]Fix PR106626: trye several attempts for lookup values, if not found.\n \n If the value is used in Scalar several times, the first attempt to find\n its position in the node (if ReuseShuffleIndices and ReorderIndices not\n empty) may fail. In this case need to find another copy of the same\n value and try again.\n Fixes https://github.com/llvm/llvm-project/issues/106626\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 775fa9ba75cf..edb2567fa057 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -3237,15 +3237,25 @@ private:\n /// When ReuseReorderShuffleIndices is empty it just returns position of \\p\n /// V within vector of Scalars. Otherwise, try to remap on its reuse index.\n int findLaneForValue(Value *V) const {\n- unsigned FoundLane = std::distance(Scalars.begin(), find(Scalars, V));\n- assert(FoundLane < Scalars.size() && \"Couldn't find extract lane\");\n- if (!ReorderIndices.empty())\n- FoundLane = ReorderIndices[FoundLane];\n- assert(FoundLane < Scalars.size() && \"Couldn't find extract lane\");\n- if (!ReuseShuffleIndices.empty()) {\n- FoundLane = std::distance(ReuseShuffleIndices.begin(),\n- find(ReuseShuffleIndices, FoundLane));\n+ unsigned FoundLane = getVectorFactor();\n+ for (auto *It = find(Scalars, V), *End = Scalars.end(); It != End;\n+ std::advance(It, 1)) {\n+ if (*It != V)\n+ continue;\n+ FoundLane = std::distance(Scalars.begin(), It);\n+ assert(FoundLane < Scalars.size() && \"Couldn't find extract lane\");\n+ if (!ReorderIndices.empty())\n+ FoundLane = ReorderIndices[FoundLane];\n+ assert(FoundLane < Scalars.size() && \"Couldn't find extract lane\");\n+ if (ReuseShuffleIndices.empty())\n+ break;\n+ if (auto *RIt = find(ReuseShuffleIndices, FoundLane);\n+ RIt != ReuseShuffleIndices.end()) {\n+ FoundLane = std::distance(ReuseShuffleIndices.begin(), RIt);\n+ break;\n+ }\n }\n+ assert(FoundLane < getVectorFactor() && \"Unable to find given value.\");\n return FoundLane;\n }\n \n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/AArch64/reused-scalar-repeated-in-node.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=aarch64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nentry:\n br label %bb61\n\nbb61: ; preds = %entry\n br label %bb64\n\nbb62: ; No predecessors!\n br i1 poison, label %bb63, label %bb64\n\nbb63: ; preds = %bb62\n br label %bb64\n\nbb64: ; preds = %bb63, %bb62, %bb61\n %i = phi nsz float [ poison, %bb61 ], [ poison, %bb63 ], [ poison, %bb62 ]\n %i65 = phi nsz float [ poison, %bb61 ], [ poison, %bb63 ], [ poison, %bb62 ]\n %i66 = load float, ptr poison, align 16\n %i67 = load float, ptr poison, align 4\n %i68 = load float, ptr poison, align 8\n %i69 = load float, ptr poison, align 4\n %i70 = load float, ptr poison, align 4\n %i71 = load float, ptr poison, align 16\n %i72 = load float, ptr poison, align 4\n %i73 = load float, ptr poison, align 8\n %i74 = load float, ptr poison, align 4\n %i75 = load float, ptr poison, align 16\n %i76 = load float, ptr poison, align 4\n br i1 poison, label %bb167, label %bb77\n\nbb77: ; preds = %bb64\n br label %bb78\n\nbb78: ; preds = %bb78, %bb77\n %i79 = phi nsz float [ %i66, %bb77 ], [ %i103, %bb78 ]\n %i80 = phi nsz float [ %i67, %bb77 ], [ %i104, %bb78 ]\n %i81 = phi nsz float [ %i68, %bb77 ], [ %i105, %bb78 ]\n %i82 = phi nsz float [ poison, %bb77 ], [ %i106, %bb78 ]\n %i83 = phi nsz float [ poison, %bb77 ], [ %i123, %bb78 ]\n %i84 = phi nsz float [ %i69, %bb77 ], [ %i124, %bb78 ]\n %i85 = phi nsz float [ poison, %bb77 ], [ %i125, %bb78 ]\n %i86 = phi nsz float [ %i70, %bb77 ], [ %i126, %bb78 ]\n %i87 = fmul fast float %i79, poison\n %i88 = fmul fast float %i80, poison\n %i89 = fmul fast float %i81, poison\n %i90 = fmul fast float %i82, poison\n %i91 = fmul fast float %i83, poison\n %i92 = fadd fast float %i91, %i87\n %i93 = fmul fast float %i84, poison\n %i94 = fadd fast float %i93, %i88\n %i95 = fmul fast float %i85, poison\n %i96 = fadd fast float %i95, %i89\n %i97 = fmul fast float %i86, poison\n %i98 = fadd fast float %i97, %i90\n %i99 = fadd fast float %i92, poison\n %i100 = fadd fast float %i94, poison\n %i101 = fadd fast float %i96, poison\n %i102 = fadd fast float %i98, poison\n %i103 = fadd fast float %i99, poison\n %i104 = fadd fast float %i100, poison\n %i105 = fadd fast float %i101, poison\n %i106 = fadd fast float %i102, poison\n %i107 = fmul fast float %i79, poison\n %i108 = fmul fast float %i80, poison\n %i109 = fmul fast float %i81, poison\n %i110 = fmul fast float %i82, poison\n %i111 = fmul fast float %i83, poison\n %i112 = fadd fast float %i111, %i107\n %i113 = fmul fast float %i84, poison\n %i114 = fadd fast float %i113, %i108\n %i115 = fmul fast float %i85, poison\n %i116 = fadd fast float %i115, %i109\n %i117 = fmul fast float %i86, poison\n %i118 = fadd fast float %i117, %i110\n %i119 = fadd fast float %i112, poison\n %i120 = fadd fast float %i114, poison\n %i121 = fadd fast float %i116, poison\n %i122 = fadd fast float %i118, poison\n %i123 = fadd fast float %i119, poison\n %i124 = fadd fast float %i120, poison\n %i125 = fadd fast float %i121, poison\n %i126 = fadd fast float %i122, poison\n %i127 = fmul fast float %i79, %i\n %i128 = fmul fast float %i80, %i\n %i129 = fmul fast float %i81, %i\n %i130 = fmul fast float %i82, %i\n %i131 = fmul fast float %i83, %i65\n %i132 = fadd fast float %i131, %i127\n %i133 = fmul fast float %i84, %i65\n %i134 = fadd fast float %i133, %i128\n %i135 = fmul fast float %i85, %i65\n %i136 = fadd fast float %i135, %i129\n %i137 = fmul fast float %i86, %i65\n %i138 = fadd fast float %i137, %i130\n %i139 = fadd fast float %i132, poison\n %i140 = fadd fast float %i134, poison\n %i141 = fadd fast float %i136, poison\n %i142 = fadd fast float %i138, poison\n %i143 = fadd fast float %i139, poison\n %i144 = fadd fast float %i140, poison\n %i145 = fadd fast float %i141, poison\n %i146 = fadd fast float %i142, poison\n %i147 = fmul fast float %i79, poison\n %i148 = fmul fast float %i80, poison\n %i149 = fmul fast float %i81, poison\n %i150 = fmul fast float %i82, poison\n %i151 = fmul fast float %i83, poison\n %i152 = fadd fast float %i151, %i147\n %i153 = fmul fast float %i84, poison\n %i154 = fadd fast float %i153, %i148\n %i155 = fmul fast float %i85, poison\n %i156 = fadd fast float %i155, %i149\n %i157 = fmul fast float %i86, poison\n %i158 = fadd fast float %i157, %i150\n %i159 = fadd fast float %i152, poison\n %i160 = fadd fast float %i154, poison\n %i161 = fadd fast float %i156, poison\n %i162 = fadd fast float %i158, poison\n %i163 = fadd fast float %i159, poison\n %i164 = fadd fast float %i160, poison\n %i165 = fadd fast float %i161, poison\n %i166 = fadd fast float %i162, poison\n br i1 poison, label %bb78, label %bb167\n\nbb167: ; preds = %bb78, %bb64\n %i168 = phi nsz float [ %i76, %bb64 ], [ %i166, %bb78 ]\n %i169 = phi nsz float [ poison, %bb64 ], [ %i165, %bb78 ]\n %i170 = phi nsz float [ poison, %bb64 ], [ %i164, %bb78 ]\n %i171 = phi nsz float [ %i75, %bb64 ], [ %i163, %bb78 ]\n %i172 = phi nsz float [ %i74, %bb64 ], [ %i146, %bb78 ]\n %i173 = phi nsz float [ %i73, %bb64 ], [ %i145, %bb78 ]\n %i174 = phi nsz float [ %i72, %bb64 ], [ %i144, %bb78 ]\n %i175 = phi nsz float [ %i71, %bb64 ], [ %i143, %bb78 ]\n %i176 = phi nsz float [ %i70, %bb64 ], [ %i126, %bb78 ]\n %i177 = phi nsz float [ poison, %bb64 ], [ %i125, %bb78 ]\n %i178 = phi nsz float [ %i69, %bb64 ], [ %i124, %bb78 ]\n %i179 = phi nsz float [ poison, %bb64 ], [ %i123, %bb78 ]\n %i180 = phi nsz float [ poison, %bb64 ], [ %i106, %bb78 ]\n %i181 = phi nsz float [ %i68, %bb64 ], [ %i105, %bb78 ]\n %i182 = phi nsz float [ %i67, %bb64 ], [ %i104, %bb78 ]\n %i183 = phi nsz float [ %i66, %bb64 ], [ %i103, %bb78 ]\n store float %i182, ptr poison, align 1\n store float %i174, ptr poison, align 1\n br i1 poison, label %bb186, label %bb184\n\nbb184: ; preds = %bb167\n br label %bb185\n\nbb185: ; preds = %bb185, %bb184\n br i1 poison, label %bb185, label %bb186\n\nbb186: ; preds = %bb185, %bb167\n %i187 = phi nsz float [ %i178, %bb167 ], [ poison, %bb185 ]\n ret void\n}\n"}]}], "issue": {"title": "[SLP] Assertion \"All elements in mask must be less than CommonVF.\" failed", "body": "[PR77529](https://github.com/llvm/llvm-project/pull/77529) trigger assertion on \"All elements in mask must be less than CommonVF.\" failed.\r\n\r\n@alexey-bataev \r\n\r\nTo reproduce, run with: opt -S -passes=slp-vectorizer t.ll -o o.ll\r\nUsing reduced t.ll below\r\n```\r\ntarget datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-Fn32\"\r\ntarget triple = \"aarch64-unknown-linux-gnu\"\r\n\r\ndefine void @test() {\r\nentry:\r\n br label %bb61\r\n\r\nbb61:\r\n br label %bb64\r\n\r\nbb62:\r\n br i1 poison, label %bb63, label %bb64\r\n\r\nbb63:\r\n br label %bb64\r\n\r\nbb64:\r\n %i = phi nsz float [ poison, %bb61 ], [ poison, %bb63 ], [ poison, %bb62 ]\r\n %i65 = phi nsz float [ poison, %bb61 ], [ poison, %bb63 ], [ poison, %bb62 ]\r\n %i66 = load float, ptr poison, align 16\r\n %i67 = load float, ptr poison, align 4\r\n %i68 = load float, ptr poison, align 8\r\n %i69 = load float, ptr poison, align 4\r\n %i70 = load float, ptr poison, align 4\r\n %i71 = load float, ptr poison, align 16\r\n %i72 = load float, ptr poison, align 4\r\n %i73 = load float, ptr poison, align 8\r\n %i74 = load float, ptr poison, align 4\r\n %i75 = load float, ptr poison, align 16\r\n %i76 = load float, ptr poison, align 4\r\n br i1 poison, label %bb167, label %bb77\r\n\r\nbb77: ; preds = %bb64\r\n br label %bb78\r\n\r\nbb78: ; preds = %bb78, %bb77\r\n %i79 = phi nsz float [ %i66, %bb77 ], [ %i103, %bb78 ]\r\n %i80 = phi nsz float [ %i67, %bb77 ], [ %i104, %bb78 ]\r\n %i81 = phi nsz float [ %i68, %bb77 ], [ %i105, %bb78 ]\r\n %i82 = phi nsz float [ poison, %bb77 ], [ %i106, %bb78 ]\r\n %i83 = phi nsz float [ poison, %bb77 ], [ %i123, %bb78 ]\r\n %i84 = phi nsz float [ %i69, %bb77 ], [ %i124, %bb78 ]\r\n %i85 = phi nsz float [ poison, %bb77 ], [ %i125, %bb78 ]\r\n %i86 = phi nsz float [ %i70, %bb77 ], [ %i126, %bb78 ]\r\n %i87 = fmul fast float %i79, poison\r\n %i88 = fmul fast float %i80, poison\r\n %i89 = fmul fast float %i81, poison\r\n %i90 = fmul fast float %i82, poison\r\n %i91 = fmul fast float %i83, poison\r\n %i92 = fadd fast float %i91, %i87\r\n %i93 = fmul fast float %i84, poison\r\n %i94 = fadd fast float %i93, %i88\r\n %i95 = fmul fast float %i85, poison\r\n %i96 = fadd fast float %i95, %i89\r\n %i97 = fmul fast float %i86, poison\r\n %i98 = fadd fast float %i97, %i90\r\n %i99 = fadd fast float %i92, poison\r\n %i100 = fadd fast float %i94, poison\r\n %i101 = fadd fast float %i96, poison\r\n %i102 = fadd fast float %i98, poison\r\n %i103 = fadd fast float %i99, poison\r\n %i104 = fadd fast float %i100, poison\r\n %i105 = fadd fast float %i101, poison\r\n %i106 = fadd fast float %i102, poison\r\n %i107 = fmul fast float %i79, poison\r\n %i108 = fmul fast float %i80, poison\r\n %i109 = fmul fast float %i81, poison\r\n %i110 = fmul fast float %i82, poison\r\n %i111 = fmul fast float %i83, poison\r\n %i112 = fadd fast float %i111, %i107\r\n %i113 = fmul fast float %i84, poison\r\n %i114 = fadd fast float %i113, %i108\r\n %i115 = fmul fast float %i85, poison\r\n %i116 = fadd fast float %i115, %i109\r\n %i117 = fmul fast float %i86, poison\r\n %i118 = fadd fast float %i117, %i110\r\n %i119 = fadd fast float %i112, poison\r\n %i120 = fadd fast float %i114, poison\r\n %i121 = fadd fast float %i116, poison\r\n %i122 = fadd fast float %i118, poison\r\n %i123 = fadd fast float %i119, poison\r\n %i124 = fadd fast float %i120, poison\r\n %i125 = fadd fast float %i121, poison\r\n %i126 = fadd fast float %i122, poison\r\n %i127 = fmul fast float %i79, %i\r\n %i128 = fmul fast float %i80, %i\r\n %i129 = fmul fast float %i81, %i\r\n %i130 = fmul fast float %i82, %i\r\n %i131 = fmul fast float %i83, %i65\r\n %i132 = fadd fast float %i131, %i127\r\n %i133 = fmul fast float %i84, %i65\r\n %i134 = fadd fast float %i133, %i128\r\n %i135 = fmul fast float %i85, %i65\r\n %i136 = fadd fast float %i135, %i129\r\n %i137 = fmul fast float %i86, %i65\r\n %i138 = fadd fast float %i137, %i130\r\n %i139 = fadd fast float %i132, poison\r\n %i140 = fadd fast float %i134, poison\r\n %i141 = fadd fast float %i136, poison\r\n %i142 = fadd fast float %i138, poison\r\n %i143 = fadd fast float %i139, poison\r\n %i144 = fadd fast float %i140, poison\r\n %i145 = fadd fast float %i141, poison\r\n %i146 = fadd fast float %i142, poison\r\n %i147 = fmul fast float %i79, poison\r\n %i148 = fmul fast float %i80, poison\r\n %i149 = fmul fast float %i81, poison\r\n %i150 = fmul fast float %i82, poison\r\n %i151 = fmul fast float %i83, poison\r\n %i152 = fadd fast float %i151, %i147\r\n %i153 = fmul fast float %i84, poison\r\n %i154 = fadd fast float %i153, %i148\r\n %i155 = fmul fast float %i85, poison\r\n %i156 = fadd fast float %i155, %i149\r\n %i157 = fmul fast float %i86, poison\r\n %i158 = fadd fast float %i157, %i150\r\n %i159 = fadd fast float %i152, poison\r\n %i160 = fadd fast float %i154, poison\r\n %i161 = fadd fast float %i156, poison\r\n %i162 = fadd fast float %i158, poison\r\n %i163 = fadd fast float %i159, poison\r\n %i164 = fadd fast float %i160, poison\r\n %i165 = fadd fast float %i161, poison\r\n %i166 = fadd fast float %i162, poison\r\n br i1 poison, label %bb78, label %bb167\r\n\r\nbb167: ; preds = %bb78, %bb64\r\n %i168 = phi nsz float [ %i76, %bb64 ], [ %i166, %bb78 ]\r\n %i169 = phi nsz float [ poison, %bb64 ], [ %i165, %bb78 ]\r\n %i170 = phi nsz float [ poison, %bb64 ], [ %i164, %bb78 ]\r\n %i171 = phi nsz float [ %i75, %bb64 ], [ %i163, %bb78 ]\r\n %i172 = phi nsz float [ %i74, %bb64 ], [ %i146, %bb78 ]\r\n %i173 = phi nsz float [ %i73, %bb64 ], [ %i145, %bb78 ]\r\n %i174 = phi nsz float [ %i72, %bb64 ], [ %i144, %bb78 ]\r\n %i175 = phi nsz float [ %i71, %bb64 ], [ %i143, %bb78 ]\r\n %i176 = phi nsz float [ %i70, %bb64 ], [ %i126, %bb78 ]\r\n %i177 = phi nsz float [ poison, %bb64 ], [ %i125, %bb78 ]\r\n %i178 = phi nsz float [ %i69, %bb64 ], [ %i124, %bb78 ]\r\n %i179 = phi nsz float [ poison, %bb64 ], [ %i123, %bb78 ]\r\n %i180 = phi nsz float [ poison, %bb64 ], [ %i106, %bb78 ]\r\n %i181 = phi nsz float [ %i68, %bb64 ], [ %i105, %bb78 ]\r\n %i182 = phi nsz float [ %i67, %bb64 ], [ %i104, %bb78 ]\r\n %i183 = phi nsz float [ %i66, %bb64 ], [ %i103, %bb78 ]\r\n store float %i182, ptr poison, align 1\r\n store float %i174, ptr poison, align 1\r\n br i1 poison, label %bb186, label %bb184\r\n\r\nbb184: ; preds = %bb167\r\n br label %bb185\r\n\r\nbb185: ; preds = %bb185, %bb184\r\n br i1 poison, label %bb185, label %bb186\r\n\r\nbb186: ; preds = %bb185, %bb167\r\n %i187 = phi nsz float [ %i178, %bb167 ], [ poison, %bb185 ]\r\n ret void\r\n}\r\n```", "author": "huihzhang", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "111606", "issue_url": "https://github.com/llvm/llvm-project/issues/111606", "bug_type": "crash", "base_commit": "374fffe015b48676fb3785167d363454a2f4dd1f", "knowledge_cutoff": "2024-10-08T23:44:36Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "01cbbc52dc95fe5d9e30ecbfb2ec8c7c1e0199c4", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[645, 652]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPInstruction::generate"]}}, "patch": "commit 01cbbc52dc95fe5d9e30ecbfb2ec8c7c1e0199c4\nAuthor: Florian Hahn \nDate: Wed Oct 9 13:18:53 2024 +0100\n\n [VPlan] Request lane 0 for pointer arg in PtrAdd.\n \n After 7f74651, the pointer operand may be replicated of a PtrAdd. Instead\n of requesting a single scalar, request lane 0, which correctly handles the\n case when there is a scalar-per-lane.\n \n Fixes https://github.com/llvm/llvm-project/issues/111606.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex 123eb0d8d71a..2468616be0bd 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -645,8 +645,8 @@ Value *VPInstruction::generate(VPTransformState &State) {\n case VPInstruction::PtrAdd: {\n assert(vputils::onlyFirstLaneUsed(this) &&\n \"can only generate first lane for PtrAdd\");\n- Value *Ptr = State.get(getOperand(0), /* IsScalar */ true);\n- Value *Addend = State.get(getOperand(1), /* IsScalar */ true);\n+ Value *Ptr = State.get(getOperand(0), VPLane(0));\n+ Value *Addend = State.get(getOperand(1), VPLane(0));\n return isInBounds() ? Builder.CreateInBoundsPtrAdd(Ptr, Addend, Name)\n : Builder.CreatePtrAdd(Ptr, Addend, Name);\n }\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/interleave-ptradd-with-replicated-operand.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -S %s"], "tests": [{"test_name": "test_interleave_ptradd_with_replicated_op", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine ptr @test_interleave_ptradd_with_replicated_op(ptr %m) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %ptr.iv = phi ptr [ %m, %entry ], [ %ptr.iv.next, %loop ]\n %iv = phi i32 [ 1, %entry ], [ %iv.next, %loop ]\n %ptr.iv.next = getelementptr i8, ptr %ptr.iv, i64 8\n %p.4 = getelementptr i8, ptr %ptr.iv, i64 4\n %l.1 = load i32, ptr %p.4, align 4\n %p.0 = getelementptr i8, ptr %ptr.iv, i64 0\n %l.2 = load i32, ptr %p.0, align 4\n %add = add i32 %l.1, %l.2\n store i32 %add, ptr %ptr.iv, align 4\n %iv.next = add i32 %iv, 1\n %tobool.not = icmp eq i32 %iv, 100\n br i1 %tobool.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret ptr %p.4\n}\n\nattributes #0 = { \"target-cpu\"=\"znver2\" }\n"}]}], "issue": {"title": "A/F: \"Trying to access a single scalar per part but has multiple scalars \" \"per part.\"' failed after 7f74651", "body": "Recently one of our internal tests started hitting an assertion failure when compiling which I bisected back to commit 7f74651. \r\n\r\nHere is a reduced repro from the crashing code:\r\n```c\r\ntypedef float a;\r\ntypedef struct {\r\n a b, c;\r\n} d;\r\ntypedef struct {\r\n d e;\r\n a f;\r\n} g;\r\n#define h() \\\r\n i = j->e.b; \\\r\n k = l->e.c; \\\r\n l->f = k;\r\na i, k;\r\nint m;\r\ng *l;\r\nvoid n() {\r\n g *j = l + 1;\r\n for (; m; ++m, ++l, ++j) {\r\n h()\r\n }\r\n}\r\n```\r\n\r\nWhen compiled with `-O2 -g -ffast-math -march=znver2` using a compiler built from 7f74651 you can see the crash:\r\n```\r\n$ ~/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang -c -O2 -g -ffast-math -march=znver2 repro.c\r\nclang: /home/dyung/src/upstream/llvm_clean_git/llvm/lib/Transforms/Vectorize/VPlan.cpp:258: llvm::Value* llvm::VPTransformState::get(llvm::VPValue*, bool): Assertion `(VF.isScalar() || Def->isLiveIn() || hasVectorValue(Def) || !vputils::onlyFirstLaneUsed(Def) || (hasScalarValue(Def, VPLane(0)) && Data.VPV2Scalars[Def].size() == 1)) && \"Trying to access a single scalar per part but has multiple scalars \" \"per part.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0. Program arguments: /home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang -c -O2 -g -ffast-math -march=znver2 repro2.c\r\n1. parser at end of file\r\n2. Optimizer\r\n3. Running pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"repro2.c\"\r\n4. Running pass \"loop-vectorize\" on function \"n\"\r\n #0 0x000055d0e9ad0e1f llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x235be1f)\r\n #1 0x000055d0e9ace96c llvm::sys::CleanupOnSignal(unsigned long) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x235996c)\r\n #2 0x000055d0e9a20cc8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\r\n #3 0x00007f44c03e9420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #4 0x00007f44bfeb600b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #5 0x00007f44bfe95859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #6 0x00007f44bfe95729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #7 0x00007f44bfe95729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #8 0x00007f44bfea6fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #9 0x000055d0eb66055a llvm::VPTransformState::get(llvm::VPValue*, bool) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3eeb55a)\r\n#10 0x000055d0eb6926a9 llvm::VPInstruction::generate(llvm::VPTransformState&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3f1d6a9)\r\n#11 0x000055d0eb695d01 llvm::VPInstruction::execute(llvm::VPTransformState&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3f20d01)\r\n#12 0x000055d0eb65fb77 llvm::VPBasicBlock::executeRecipes(llvm::VPTransformState*, llvm::BasicBlock*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3eeab77)\r\n#13 0x000055d0eb66283e llvm::VPBasicBlock::execute(llvm::VPTransformState*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3eed83e)\r\n#14 0x000055d0eb6645c8 llvm::VPRegionBlock::execute(llvm::VPTransformState*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3eef5c8)\r\n#15 0x000055d0eb660b8e llvm::VPlan::execute(llvm::VPTransformState*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3eebb8e)\r\n#16 0x000055d0eb5362cb llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3dc12cb)\r\n#17 0x000055d0eb54baf4 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3dd6af4)\r\n#18 0x000055d0eb54dac1 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3dd8ac1)\r\n#19 0x000055d0eb54e133 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3dd9133)\r\n#20 0x000055d0eaf6f396 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x37fa396)\r\n#21 0x000055d0e9589031 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x1e14031)\r\n#22 0x000055d0e86a7026 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0xf32026)\r\n#23 0x000055d0e958799d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x1e1299d)\r\n#24 0x000055d0e86a79e6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0xf329e6)\r\n#25 0x000055d0e95873d1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x1e123d1)\r\n#26 0x000055d0e9d65d23 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#27 0x000055d0e9d692f8 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x25f42f8)\r\n#28 0x000055d0ea4492ac clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x2cd42ac)\r\n#29 0x000055d0ec3dbb9c clang::ParseAST(clang::Sema&, bool, bool) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x4c66b9c)\r\n#30 0x000055d0ea4496c8 clang::CodeGenAction::ExecuteAction() (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x2cd46c8)\r\n#31 0x000055d0ea7119a9 clang::FrontendAction::Execute() (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x2f9c9a9)\r\n#32 0x000055d0ea68d99e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x2f1899e)\r\n#33 0x000055d0ea7fe466 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x3089466)\r\n#34 0x000055d0e862a295 cc1_main(llvm::ArrayRef, char const*, void*) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0xeb5295)\r\n#35 0x000055d0e862201a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#36 0x000055d0ea48f17d void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\r\n#37 0x000055d0e9a211d0 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x22ac1d0)\r\n#38 0x000055d0ea48f79f clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\r\n#39 0x000055d0ea4538f4 clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x2cde8f4)\r\n#40 0x000055d0ea454a0d clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x2cdfa0d)\r\n#41 0x000055d0ea45cb55 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0x2ce7b55)\r\n#42 0x000055d0e862722b clang_main(int, char**, llvm::ToolContext const&) (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0xeb222b)\r\n#43 0x000055d0e855712b main (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0xde212b)\r\n#44 0x00007f44bfe97083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#45 0x000055d0e8621aae _start (/home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin/clang+0xeacaae)\r\nclang: error: clang frontend command failed with exit code 134 (use -v to see invocation)\r\nclang version 20.0.0 (https://github.com/llvm/llvm-project.git 7f74651837b8f95996ea4c774b2e604020446951)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /home/dyung/src/upstream/7f74651837b8f95996ea4c774b2e604020446951-linux/bin\r\nBuild config: +assertions\r\n```\r\n\r\nHere is a link to godbolt also showing the failure: https://godbolt.org/z/bnPM9qqYo", "author": "dyung", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "116809", "issue_url": "https://github.com/llvm/llvm-project/issues/116809", "bug_type": "crash", "base_commit": "0a1795f781163a8e7b72fee492f87ab27d5e3995", "knowledge_cutoff": "2024-11-19T14:26:43Z", "lit_test_dir": ["llvm/test/Transforms/LICM"], "hints": {"fix_commit": "18b02bbf441660683df7f3925946984203d49bab", "components": ["LICM", "MemorySSAUpdater"], "bug_location_lineno": {"llvm/include/llvm/Analysis/MemorySSAUpdater.h": [[190, 196]], "llvm/lib/Analysis/MemorySSAUpdater.cpp": [[1403, 1411]], "llvm/lib/Transforms/Scalar/LICM.cpp": [[1465, 1472]]}, "bug_location_funcname": {"llvm/lib/Analysis/MemorySSAUpdater.cpp": ["MemorySSAUpdater::createMemoryAccessBefore", "MemorySSAUpdater::createMemoryAccessInBB"], "llvm/lib/Transforms/Scalar/LICM.cpp": ["cloneInstructionInExitBlock"]}}, "patch": "commit 18b02bbf441660683df7f3925946984203d49bab\nAuthor: DianQK \nDate: Wed Nov 20 19:52:51 2024 +0800\n\n [LICM] allow MemoryAccess creation failure (#116813)\n \n Fixes #116809.\n \n After running some passes (SimpleLoopUnswitch, LoopInstSimplify, etc.),\n MemorySSA might be outdated, and the instruction `I` may have become a\n non-memory touching instruction.\n \n LICM has already handled this, but it does not pass\n `CreationMustSucceed=false` to `createDefinedAccess`.\n\ndiff --git a/llvm/include/llvm/Analysis/MemorySSAUpdater.h b/llvm/include/llvm/Analysis/MemorySSAUpdater.h\nindex d4da3ef1146d..055feceefb05 100644\n--- a/llvm/include/llvm/Analysis/MemorySSAUpdater.h\n+++ b/llvm/include/llvm/Analysis/MemorySSAUpdater.h\n@@ -190,7 +190,8 @@ public:\n /// inaccessible and it *must* have removeMemoryAccess called on it.\n MemoryAccess *createMemoryAccessInBB(Instruction *I, MemoryAccess *Definition,\n const BasicBlock *BB,\n- MemorySSA::InsertionPlace Point);\n+ MemorySSA::InsertionPlace Point,\n+ bool CreationMustSucceed = true);\n \n /// Create a MemoryAccess in MemorySSA before an existing MemoryAccess.\n ///\ndiff --git a/llvm/lib/Analysis/MemorySSAUpdater.cpp b/llvm/lib/Analysis/MemorySSAUpdater.cpp\nindex aa550f0b6a7b..f672bd0e1e13 100644\n--- a/llvm/lib/Analysis/MemorySSAUpdater.cpp\n+++ b/llvm/lib/Analysis/MemorySSAUpdater.cpp\n@@ -1403,9 +1403,11 @@ void MemorySSAUpdater::changeToUnreachable(const Instruction *I) {\n \n MemoryAccess *MemorySSAUpdater::createMemoryAccessInBB(\n Instruction *I, MemoryAccess *Definition, const BasicBlock *BB,\n- MemorySSA::InsertionPlace Point) {\n- MemoryUseOrDef *NewAccess = MSSA->createDefinedAccess(I, Definition);\n- MSSA->insertIntoListsForBlock(NewAccess, BB, Point);\n+ MemorySSA::InsertionPlace Point, bool CreationMustSucceed) {\n+ MemoryUseOrDef *NewAccess = MSSA->createDefinedAccess(\n+ I, Definition, /*Template=*/nullptr, CreationMustSucceed);\n+ if (NewAccess)\n+ MSSA->insertIntoListsForBlock(NewAccess, BB, Point);\n return NewAccess;\n }\n \ndiff --git a/llvm/lib/Transforms/Scalar/LICM.cpp b/llvm/lib/Transforms/Scalar/LICM.cpp\nindex fa04ced7182d..94bfe44a847a 100644\n--- a/llvm/lib/Transforms/Scalar/LICM.cpp\n+++ b/llvm/lib/Transforms/Scalar/LICM.cpp\n@@ -1465,8 +1465,11 @@ static Instruction *cloneInstructionInExitBlock(\n \n if (MSSAU.getMemorySSA()->getMemoryAccess(&I)) {\n // Create a new MemoryAccess and let MemorySSA set its defining access.\n+ // After running some passes, MemorySSA might be outdated, and the\n+ // instruction `I` may have become a non-memory touching instruction.\n MemoryAccess *NewMemAcc = MSSAU.createMemoryAccessInBB(\n- New, nullptr, New->getParent(), MemorySSA::Beginning);\n+ New, nullptr, New->getParent(), MemorySSA::Beginning,\n+ /*CreationMustSucceed=*/false);\n if (NewMemAcc) {\n if (auto *MemDef = dyn_cast(NewMemAcc))\n MSSAU.insertDef(MemDef, /*RenameUses=*/true);\n", "tests": [{"file": "llvm/test/Transforms/LICM/PR116813-memoryssa-outdated.ll", "commands": ["opt -passes='loop-mssa(simple-loop-unswitch,licm)' -verify-memoryssa -S < %s"], "tests": [{"test_name": "", "test_body": "\n; Check that running LICM after SimpleLoopUnswitch does not result in a crash.\n\ndefine i32 @foo(i1 %arg, ptr %arg1) {\n;\nstart:\n br label %loop\n\nloop: ; preds = %loop, %bb\n %i = select i1 %arg, ptr %arg1, ptr @bar\n %i3 = call i32 %i()\n br i1 %arg, label %loop, label %ret\n\nret: ; preds = %loop\n ret i32 %i3\n}\n\ndeclare i32 @bar() nounwind willreturn memory(none)"}]}], "issue": {"title": "[SimpleLoopUnswitch][LICM] Assertion `NewAccess != nullptr && \"Tried to create a memory access for a \" \"non-memory touching instruction\"'", "body": "The following IR triggers the assertion mentioned in the title:\r\n\r\n```llvm\r\n; opt -passes='loop-mssa(simple-loop-unswitch,licm)'\r\ndefine i32 @foo(i1 %arg, ptr %arg1) {\r\nbb:\r\n br label %bb2\r\n\r\nbb2: ; preds = %bb2, %bb\r\n %i = select i1 %arg, ptr %arg1, ptr @bar\r\n %i3 = call i32 %i()\r\n br i1 %arg, label %bb2, label %bb4\r\n\r\nbb4: ; preds = %bb2\r\n ret i32 %i3\r\n}\r\n\r\ndeclare i32 @bar() nounwind willreturn memory(none)\r\n```\r\n\r\nhttps://llvm.godbolt.org/z/bj79jxo8x", "author": "DianQK", "labels": ["crash", "llvm:analysis", "llvm:transforms"], "comments": []}, "verified": true} {"bug_id": "119002", "issue_url": "https://github.com/llvm/llvm-project/issues/119002", "bug_type": "crash", "base_commit": "6797b0f0c002c9f3dda7a2b502ddf33e2c8f6777", "knowledge_cutoff": "2024-12-06T17:29:17Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "376dad72abb4960cd93e37e3e98de3950b5c3dfe", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[16215, 16220]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree"]}}, "patch": "commit 376dad72abb4960cd93e37e3e98de3950b5c3dfe\nAuthor: Alexey Bataev \nDate: Fri Dec 6 13:39:05 2024 -0800\n\n [SLP]Move resulting vector before inert point, if the late generated buildvector fully matched\n \n If the perfect diamond match was detected for the postponed buildvectors\n and the vector for the previous node comes after the current node, need\n to move the vector register before the current inserting point to\n prevent compiler crash.\n \n Fixes #119002\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 51841a842ce0..50590e51ae0e 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -16215,6 +16215,11 @@ BoUpSLP::vectorizeTree(const ExtraValueToDebugLocsMap &ExternallyUsedValues,\n }\n Builder.SetCurrentDebugLocation(UserI->getDebugLoc());\n Value *Vec = vectorizeTree(TE, /*PostponedPHIs=*/false);\n+ if (auto *VecI = dyn_cast(Vec);\n+ VecI && VecI->getParent() == Builder.GetInsertBlock() &&\n+ Builder.GetInsertPoint()->comesBefore(VecI))\n+ VecI->moveBeforePreserving(*Builder.GetInsertBlock(),\n+ Builder.GetInsertPoint());\n if (Vec->getType() != PrevVec->getType()) {\n assert(Vec->getType()->isIntOrIntVectorTy() &&\n PrevVec->getType()->isIntOrIntVectorTy() &&\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/perfect-matched-reused-bv.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f < %s"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb4, %bb\n %phi = phi i32 [ 0, %bb ], [ %add6, %bb4 ]\n %phi2 = phi i32 [ 0, %bb ], [ %add, %bb4 ]\n %or = or i32 %phi2, 0\n %or3 = or i32 %phi, 0\n br i1 false, label %bb7, label %bb4\n\nbb4: ; preds = %bb1\n %add = add i32 0, 0\n %add5 = add i32 0, 0\n %add6 = add i32 %phi, 0\n br i1 false, label %bb7, label %bb1\n\nbb7: ; preds = %bb4, %bb1\n %phi8 = phi i32 [ %or, %bb1 ], [ %add5, %bb4 ]\n %phi9 = phi i32 [ %or3, %bb1 ], [ %add6, %bb4 ]\n ret void\n}\n"}]}], "issue": {"title": "Instruction does not dominate all uses! LLVM ERROR: Broken module found, compilation aborted! (yet another)", "body": "To reproduce run the test below with -passes slp-vectorizer\n```\n; ModuleID = './reduced.ll'\nsource_filename = \"./reduced.ll\"\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @wombat() #0 gc \"statepoint-example\" {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb4, %bb\n %phi = phi i32 [ 0, %bb ], [ %add6, %bb4 ]\n %phi2 = phi i32 [ 0, %bb ], [ %add, %bb4 ]\n %or = or i32 %phi2, 0\n %or3 = or i32 %phi, 0\n br i1 false, label %bb7, label %bb4\n\nbb4: ; preds = %bb1\n %add = add i32 0, 0\n %add5 = add i32 0, 0\n %add6 = add i32 %phi, 0\n br i1 false, label %bb7, label %bb1\n\nbb7: ; preds = %bb4, %bb1\n %phi8 = phi i32 [ %or, %bb1 ], [ %add5, %bb4 ]\n %phi9 = phi i32 [ %or3, %bb1 ], [ %add6, %bb4 ]\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,-amx-fp8,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,-movrs,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-avx10.2-256,-gfni,-avxvnniint16,-amx-fp16,-zu,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,-nf,-amx-tf32,-amx-avx512,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-amx-transpose,-avx10.2-512,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-amx-movrs,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f\" }\n```\nReproducer: https://godbolt.org/z/Kqddzds1c\n\nStack dump:\n```\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes slp-vectorizer \n1.\tRunning pass \"verify\" on module \"\"\n #0 0x00000000051b4a48 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x51b4a48)\n #1 0x00000000051b243c SignalHandler(int) Signals.cpp:0:0\n #2 0x00007c6fb5442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #3 0x00007c6fb54969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #4 0x00007c6fb5442476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #5 0x00007c6fb54287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #6 0x00000000007e2469 llvm::json::operator==(llvm::json::Value const&, llvm::json::Value const&) (.cold) JSON.cpp:0:0\n #7 0x00000000050f32d8 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50f32d8)\n #8 0x0000000004ff2498 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ff2498)\n #9 0x00000000009057ae llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x9057ae)\n#10 0x0000000004fb1d10 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb1d10)\n#11 0x0000000000910d7a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x910d7a)\n#12 0x00000000009033f9 optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x9033f9)\n#13 0x00007c6fb5429d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#14 0x00007c6fb5429e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#15 0x00000000008facde _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8facde)\nProgram terminated with signal: SIGSEGV\nCompiler returned: 139\n```", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "84025", "issue_url": "https://github.com/llvm/llvm-project/issues/84025", "bug_type": "miscompilation", "base_commit": "f51ade25b9205efee09a4915031848cebe772805", "knowledge_cutoff": "2024-03-05T15:07:55Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "e96c0c1d5e0a9916098b1a31acb006ea6c1108fb", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp": [[2167, 2180]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp": ["collectInsertionElements"]}}, "patch": "commit e96c0c1d5e0a9916098b1a31acb006ea6c1108fb\nAuthor: Quentin Dian \nDate: Wed Mar 6 06:16:28 2024 +0800\n\n [InstCombine] Fix shift calculation in InstCombineCasts (#84027)\n \n Fixes #84025.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp\nindex 33ed1d557537..45afa6363ae0 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp\n@@ -2167,14 +2167,14 @@ static bool collectInsertionElements(Value *V, unsigned Shift,\n Type *ElementIntTy = IntegerType::get(C->getContext(), ElementSize);\n \n for (unsigned i = 0; i != NumElts; ++i) {\n- unsigned ShiftI = Shift + i * ElementSize;\n+ unsigned ShiftI = i * ElementSize;\n Constant *Piece = ConstantFoldBinaryInstruction(\n Instruction::LShr, C, ConstantInt::get(C->getType(), ShiftI));\n if (!Piece)\n return false;\n \n Piece = ConstantExpr::getTrunc(Piece, ElementIntTy);\n- if (!collectInsertionElements(Piece, ShiftI, Elements, VecEltTy,\n+ if (!collectInsertionElements(Piece, ShiftI + Shift, Elements, VecEltTy,\n isBigEndian))\n return false;\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/bitcast.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "bitcast_from_single_element_pointer_vector_to_pointer", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\ntarget triple = \"x86_64-apple-darwin10.0.0\"\n\ndefine ptr @bitcast_from_single_element_pointer_vector_to_pointer(<1 x ptr> %ptrvec) {\n %ptr = bitcast <1 x ptr> %ptrvec to ptr\n ret ptr %ptr\n}\n"}, {"test_name": "bitcast_shl", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\ntarget triple = \"x86_64-apple-darwin10.0.0\"\n\ndefine <4 x i32> @bitcast_shl(i32 %arg) {\n %i = zext i32 %arg to i64\n %i1 = shl i64 %i, 32\n %i2 = or i64 %i1, 65\n %i3 = zext i64 %i2 to i128\n %i4 = shl i128 %i3, 64\n %i5 = bitcast i128 %i4 to <4 x i32>\n ret <4 x i32> %i5\n}\n"}, {"test_name": "select_bitcast_unsized_pointer", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\ntarget triple = \"x86_64-apple-darwin10.0.0\"\n\ndeclare void @f1()\n\ndeclare void @f2()\n\ndefine ptr @select_bitcast_unsized_pointer(i1 %c) {\n %s = select i1 %c, ptr @f1, ptr @f2\n ret ptr %s\n}\n"}]}], "issue": {"title": "InstCombine calculated a wrong `insertelement` instruction", "body": "I tried this IR:\r\n\r\n```llvm\r\ndefine <4 x i32> @src(i32 %arg) {\r\n %i = zext i32 %arg to i64\r\n %i1 = shl i64 %i, 32\r\n %i2 = or i64 %i1, 65\r\n %i3 = zext i64 %i2 to i128\r\n %i4 = shl i128 %i3, 64\r\n %i5 = bitcast i128 %i4 to <4 x i32>\r\n ret <4 x i32> %i5\r\n}\r\n\r\ndefine <4 x i32> @tgt(i32 %arg) {\r\n %i5 = insertelement <4 x i32> , i32 %arg, i64 3\r\n ret <4 x i32> %i5\r\n}\r\n```\r\n\r\nI expect `src` to be:\r\n\r\n```llvm\r\ndefine <4 x i32> @src(i32 %arg) {\r\n %i5 = insertelement <4 x i32> , i32 %arg, i64 3\r\n ret <4 x i32> %i5\r\n}\r\n```\r\n\r\nBut I'm getting the same instructions as `tgt`.\r\n\r\nGodbolt: https://llvm.godbolt.org/z/KnMEY4rTq\r\nAlive2: https://alive2.llvm.org/ce/z/7FIJNi\r\nFrom: https://github.com/rust-lang/rust/issues/121996", "author": "DianQK", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "cbeuw", "body": "Looks like this bug is almost 11 years old: https://github.com/llvm/llvm-project/commit/feb34713d531d1f48c37116e8cab84814202d057! Thanks for fixing"}, {"author": "DianQK", "body": "> Looks like this bug is almost 11 years old: [feb3471](https://github.com/llvm/llvm-project/commit/feb34713d531d1f48c37116e8cab84814202d057)! Thanks for fixing\r\n\r\nBased on my experience, we likely have many more latent bugs. :)"}]}, "verified": true} {"bug_id": "90900", "issue_url": "https://github.com/llvm/llvm-project/issues/90900", "bug_type": "crash", "base_commit": "311339e25cd4d431902f93b1cbc6b67ac3fd2abf", "knowledge_cutoff": "2024-05-02T19:52:37Z", "lit_test_dir": ["llvm/test/Transforms/Reg2Mem"], "hints": {"fix_commit": "fdaad738753cde2bba6480c2ee5d1e9fb45064b9", "components": ["DemoteRegToStack"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/DemoteRegToStack.cpp": [[50, 55], [102, 110]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/DemoteRegToStack.cpp": ["llvm::DemoteRegToStack"]}}, "patch": "commit fdaad738753cde2bba6480c2ee5d1e9fb45064b9\nAuthor: XChy \nDate: Thu May 16 20:13:39 2024 +0800\n\n [Reg2Mem] Handle CallBr instructions (#90953)\n \n Fixes #90900\n\ndiff --git a/llvm/lib/Transforms/Utils/DemoteRegToStack.cpp b/llvm/lib/Transforms/Utils/DemoteRegToStack.cpp\nindex b2a88eadd3de..f03de2f25a41 100644\n--- a/llvm/lib/Transforms/Utils/DemoteRegToStack.cpp\n+++ b/llvm/lib/Transforms/Utils/DemoteRegToStack.cpp\n@@ -50,6 +50,15 @@ AllocaInst *llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,\n assert(BB && \"Unable to split critical edge.\");\n (void)BB;\n }\n+ } else if (CallBrInst *CBI = dyn_cast(&I)) {\n+ for (int i = 0; i < CBI->getNumSuccessors(); i++) {\n+ auto *Succ = CBI->getSuccessor(i);\n+ if (!Succ->getSinglePredecessor()) {\n+ assert(isCriticalEdge(II, i) && \"Expected a critical edge!\");\n+ BasicBlock *BB = SplitCriticalEdge(II, i);\n+ assert(BB && \"Unable to split critical edge.\");\n+ }\n+ }\n }\n \n // Change all of the users of the instruction to read from the stack slot.\n@@ -102,9 +111,14 @@ AllocaInst *llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,\n new StoreInst(&I, Slot, Handler->getFirstInsertionPt());\n return Slot;\n }\n+ } else if (InvokeInst *II = dyn_cast(&I)) {\n+ InsertPt = II->getNormalDest()->getFirstInsertionPt();\n+ } else if (CallBrInst *CBI = dyn_cast(&I)) {\n+ for (BasicBlock *Succ : successors(CBI))\n+ new StoreInst(CBI, Slot, Succ->getFirstInsertionPt());\n+ return Slot;\n } else {\n- InvokeInst &II = cast(I);\n- InsertPt = II.getNormalDest()->getFirstInsertionPt();\n+ llvm_unreachable(\"Unsupported terminator for Reg2Mem\");\n }\n \n new StoreInst(&I, Slot, InsertPt);\n", "tests": [{"file": "llvm/test/Transforms/Reg2Mem/callbr-crash.ll", "commands": ["opt -passes=reg2mem -S < %s"], "tests": [{"test_name": "", "test_body": "\ndefine void @crash() {\n;\nentry:\n %a = callbr i64 asm \"\", \"=r,r,!i\"(i64 0)\n to label %then [label %else]\n\nthen:\n %b = inttoptr i64 %a to ptr\n br label %else\n\nelse:\n ret void\n}"}]}], "issue": {"title": "Reg2Mem pass triggered assertion `isa(V al) && \"cast() argument of incompatible type!\"", "body": "https://godbolt.org/z/oxW8zbKr6\r\n`opt --passes=reg2mem reduced.ll`\r\n```llvm\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\n%struct.__large_struct = type { [100 x i64] }\r\n\r\ndefine fastcc i64 @__se_sys_pselect6() {\r\n %1 = callbr i64 asm sideeffect \"\\0A1:\\09movq $1,$0\\0A .pushsection \\22__ex_table\\22,\\22a\\22\\0A .balign 4\\0A .long (1b) - .\\0A .long (${2:l}) - .\\0A .long 3 \\0A .popsection\\0A\", \"=r,*m,!i,~{dirflag},~{fpsr},~{flags}\"(ptr elementtype(%struct.__large_struct) null)\r\n to label %2 [label %common.ret]\r\n\r\n2: ; preds = %0\r\n %3 = inttoptr i64 %1 to ptr\r\n %4 = callbr i64 asm sideeffect \"\\0A1:\\09movq $1,$0\\0A .pushsection \\22__ex_table\\22,\\22a\\22\\0A .balign 4\\0A .long (1b) - .\\0A .long (${2:l}) - .\\0A .long 3 \\0A .popsection\\0A\", \"=r,*m,!i,~{dirflag},~{fpsr},~{flags}\"(ptr elementtype(%struct.__large_struct) null)\r\n to label %common.ret [label %common.ret]\r\n\r\ncommon.ret: ; preds = %2, %2, %0\r\n ret i64 0\r\n}\r\n```", "author": "XChy", "labels": ["crash", "llvm:transforms"], "comments": []}, "verified": true} {"bug_id": "107015", "issue_url": "https://github.com/llvm/llvm-project/issues/107015", "bug_type": "crash", "base_commit": "50a02e7c68f964c5d8338369746c849ed8d3bef4", "knowledge_cutoff": "2024-09-02T18:45:19Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "954ed05c1001489802ae43d4bea403181570a799", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[8393, 8398]], "llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h": [[174, 184]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["VPRecipeBuilder::tryToWiden"], "llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h": ["getVPValueOrAddLiveIn"]}}, "patch": "commit 954ed05c1001489802ae43d4bea403181570a799\nAuthor: Florian Hahn \nDate: Mon Sep 2 21:25:30 2024 +0100\n\n [VPlan] Simplify MUL operands at recipe construction.\n \n This moves the logic to create simplified operands using SCEV to MUL\n recipe creation. This is needed to match the behavior of the legacy's cost\n model. TODOs are to extend to other opcodes and move to a transform.\n \n Note that this also restricts the number of SCEV simplifications we\n apply to more precisely match the cases handled by the legacy cost\n model.\n \n Fixes https://github.com/llvm/llvm-project/issues/107015.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex a8722db654f5..a64848c1fdce 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -8393,6 +8393,20 @@ VPWidenRecipe *VPRecipeBuilder::tryToWiden(Instruction *I,\n case Instruction::Sub:\n case Instruction::Xor:\n case Instruction::Freeze:\n+ if (I->getOpcode() == Instruction::Mul) {\n+ // Simplify operands of multiplications using SCEV. This is needed at the\n+ // moment to match the behavior of the legacy cost-model.\n+ // TODO: Generalize to any opcode and move to VPlan transformation.\n+ SmallVector NewOps(Operands);\n+ ScalarEvolution &SE = *PSE.getSE();\n+ for (unsigned I = 0; I < Operands.size(); ++I) {\n+ Value *V = NewOps[I]->getUnderlyingValue();\n+ if (!isa(V) && SE.isSCEVable(V->getType()))\n+ if (auto *C = dyn_cast(PSE.getSE()->getSCEV(V)))\n+ NewOps[I] = Plan.getOrAddLiveIn(C->getValue());\n+ }\n+ return new VPWidenRecipe(*I, make_range(NewOps.begin(), NewOps.end()));\n+ }\n return new VPWidenRecipe(*I, make_range(Operands.begin(), Operands.end()));\n };\n }\ndiff --git a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h\nindex 0fbdeeddaf4a..25b8bf3e089e 100644\n--- a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h\n+++ b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h\n@@ -174,11 +174,6 @@ public:\n if (auto *R = Ingredient2Recipe.lookup(I))\n return R->getVPSingleValue();\n }\n- ScalarEvolution &SE = *PSE.getSE();\n- if (!isa(V) && SE.isSCEVable(V->getType()))\n- if (auto *C = dyn_cast(PSE.getSE()->getSCEV(V)))\n- return Plan.getOrAddLiveIn(C->getValue());\n-\n return Plan.getOrAddLiveIn(V);\n }\n };\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/mul-simplification.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "mul_select_operand_known_1_via_scev", "test_body": "target triple = \"arm64-apple-macosx\"\n\ndefine i64 @mul_select_operand_known_1_via_scev() {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %red = phi i64 [ 12, %entry ], [ %red.next, %loop ]\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %0 = and i32 %iv, 1\n %cmp1.i = icmp eq i32 %0, 0\n %narrow.i = select i1 %cmp1.i, i32 1, i32 %iv\n %mul = zext nneg i32 %narrow.i to i64\n %red.next = mul nsw i64 %red, %mul\n %iv.next = add nuw nsw i32 %iv, 1\n %ec = icmp eq i32 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n %res = phi i64 [ %red.next, %loop ]\n ret i64 %res\n}\n"}]}], "issue": {"title": "[clang] Crash with assertion failed in Vectorize", "body": "The compiler crashed with -O2 in this code:\r\n\r\n```c\r\n#include\r\nstatic uint8_t b[][1] = {127};\r\nlong a(int c, int e) {\r\n long d;\r\n for (; c <= e; c++)\r\n if (c % 2 == 0)\r\n ;\r\n else\r\n d *= c;\r\n return d;\r\n}\r\nint64_t f();\r\nint16_t g() {\r\n int32_t h = 1653000360;\r\n f(a(h - 1653000360, b[0][0] - 126));\r\n uint8_t *i[] = {b, b};\r\n}\r\n```\r\n\r\nThe crash is:\r\n```\r\nclang: /root/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7369: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\n```\r\n\r\nThe details can be found here: https://godbolt.org/z/eeWW6raPb", "author": "cardigan1008", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "83931", "issue_url": "https://github.com/llvm/llvm-project/issues/83931", "bug_type": "crash", "base_commit": "d95a0d7c0ff324c1e84606d26591416e6bf02984", "knowledge_cutoff": "2024-03-05T00:39:01Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "d51fcd4ed86ac6075c8a25b053c2b66051feaf62", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": [[6544, 6549]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": ["InstCombiner::getFlippedStrictnessPredicateAndConstant"]}}, "patch": "commit d51fcd4ed86ac6075c8a25b053c2b66051feaf62\nAuthor: Yingwei Zheng \nDate: Tue Mar 5 21:08:15 2024 +0800\n\n [InstCombine] Handle scalable splat in `getFlippedStrictnessPredicateAndConstant` (#83980)\n \n This patch adds support for canonicalization of icmp with a scalable\n splat. Some optimizations assume that `icmp pred X, APInt C` is in\n canonical form.\n \n Fixes https://github.com/llvm/llvm-project/issues/83931.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\nindex 55207a8343d0..1cebab8203ea 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n@@ -6544,6 +6544,13 @@ InstCombiner::getFlippedStrictnessPredicateAndConstant(CmpInst::Predicate Pred,\n if (!SafeReplacementConstant)\n SafeReplacementConstant = CI;\n }\n+ } else if (isa(C->getType())) {\n+ // Handle scalable splat\n+ Value *SplatC = C->getSplatValue();\n+ auto *CI = dyn_cast_or_null(SplatC);\n+ // Bail out if the constant can't be safely incremented/decremented.\n+ if (!CI || !ConstantIsOk(CI))\n+ return std::nullopt;\n } else {\n // ConstantExpr?\n return std::nullopt;\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/pr83931.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "dont_crash", "test_body": "define @dont_crash( %x) {\nentry:\n %div = sdiv %x, splat (i64 309383)\n %ret = icmp sge %div, zeroinitializer\n ret %ret\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/select.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "scalable_sign_bits", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine @scalable_sign_bits( %x) {\n %a = sext %x to \n %b = shl %a, splat (i32 16)\n ret %b\n}\n"}, {"test_name": "scalable_non_zero", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine @scalable_non_zero( %x) {\n %a = or %x, splat (i32 1)\n %b = add %a, splat (i32 -1)\n %cmp = icmp ult %b, splat (i32 56)\n ret %cmp\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/vscale_cmp.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "sge", "test_body": "define @sge( %x) {\n %cmp = icmp sge %x, zeroinitializer\n ret %cmp\n}\n"}]}], "issue": {"title": "[InstCombine][RISC-V] UNREACHABLE executed at InstCombineCompares.cpp:2788", "body": "Testcase:\r\n```c\r\nchar a;\r\nint b;\r\nlong *c[5][5];\r\nint main() {\r\n for (int d = -2041027590; d; d++)\r\n a |= b <= c[3][d][8] / 309383;\r\n}\r\n```\r\n\r\nReduced LLVM IR:\r\n```llvm ir\r\n; ModuleID = 'reduced.ll'\r\nsource_filename = \"red.c\"\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\n; Function Attrs: nofree norecurse nosync nounwind memory(readwrite, argmem: read, inaccessiblemem: none) uwtable vscale_range(2,1024)\r\ndefine noundef signext i32 @main() #0 {\r\nentry:\r\n %0 = getelementptr inbounds i8, zeroinitializer, i64 64\r\n %wide.masked.gather = call @llvm.masked.gather.nxv2i64.nxv2p0( %0, i32 8, shufflevector ( insertelement ( poison, i1 true, i64 0), poison, zeroinitializer), poison)\r\n %1 = sdiv %wide.masked.gather, shufflevector ( insertelement ( poison, i64 309383, i64 0), poison, zeroinitializer)\r\n %2 = icmp sge %1, zeroinitializer\r\n %3 = zext %2 to \r\n %4 = call i8 @llvm.vector.reduce.or.nxv2i8( %3)\r\n store i8 %4, ptr null, align 1\r\n ret i32 0\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(read)\r\ndeclare @llvm.masked.gather.nxv2i64.nxv2p0(, i32 immarg, , ) #1\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i8 @llvm.vector.reduce.or.nxv2i8() #2\r\n\r\n; uselistorder directives\r\nuselistorder poison, { 1, 2, 0 }\r\n\r\nattributes #0 = { nofree norecurse nosync nounwind memory(readwrite, argmem: read, inaccessiblemem: none) uwtable vscale_range(2,1024) \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"generic-rv64\" \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v\" }\r\nattributes #1 = { nocallback nofree nosync nounwind willreturn memory(read) }\r\nattributes #2 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\n\r\nBacktrace:\r\n```\r\n> /scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt -passes=\"instcombine\" reduced.ll --disable-output\r\nUnhandled icmp predicate!\r\nUNREACHABLE executed at /scratch/tc-testing/tc-feb-22-llvm/llvm/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp:2788!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt -passes=instcombine reduced.ll --disable-output\r\n #0 0x00005565be2171a0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x2ce31a0)\r\n #1 0x00005565be2145af llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x2ce05af)\r\n #2 0x00005565be214705 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007f8b9f842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007f8b9f8969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007f8b9f8969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007f8b9f8969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007f8b9f842476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007f8b9f8287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00005565be14ca5e (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x2c18a5e)\r\n#10 0x00005565bd1c1139 llvm::InstCombinerImpl::foldICmpDivConstant(llvm::ICmpInst&, llvm::BinaryOperator*, llvm::APInt const&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x1c8d139)\r\n#11 0x00005565bd1d710e llvm::InstCombinerImpl::foldICmpBinOpWithConstant(llvm::ICmpInst&, llvm::BinaryOperator*, llvm::APInt const&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x1ca310e)\r\n#12 0x00005565bd1d7393 llvm::InstCombinerImpl::foldICmpInstWithConstant(llvm::ICmpInst&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x1ca3393)\r\n#13 0x00005565bd1e46eb llvm::InstCombinerImpl::visitICmpInst(llvm::ICmpInst&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x1cb06eb)\r\n#14 0x00005565bd1457a5 llvm::InstCombinerImpl::run() (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x1c117a5)\r\n#15 0x00005565bd146e85 combineInstructionsOverFunction(llvm::Function&, llvm::InstructionWorklist&, llvm::AAResults*, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::ProfileSummaryInfo*, llvm::LoopInfo*, llvm::InstCombineOptions const&) InstructionCombining.cpp:0:0\r\n#16 0x00005565bd147ef5 llvm::InstCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x1c13ef5)\r\n#17 0x00005565bc1e5c96 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0xcb1c96)\r\n#18 0x00005565be044751 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x2b10751)\r\n#19 0x00005565bc1dd5b6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0xca95b6)\r\n#20 0x00005565be04344b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x2b0f44b)\r\n#21 0x00005565bc1e5746 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0xcb1746)\r\n#22 0x00005565be0412c1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x2b0d2c1)\r\n#23 0x00005565bba77e55 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x543e55)\r\n#24 0x00005565bba6a296 optMain (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x536296)\r\n#25 0x00007f8b9f829d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#26 0x00007f8b9f829e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#27 0x00007f8b9f829e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#28 0x00005565bba5f5a5 _start (/scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt+0x52b5a5)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-feb-22-llvm/build/build-llvm-linux/bin/opt reduced.ll\r\n```\r\n\r\nGodbolt: https://godbolt.org/z/9qdYfPTrE\r\n\r\nAssert being tripped: https://github.com/llvm/llvm-project/blob/main/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp#L2788\r\n\r\nFound via fuzzer.", "author": "patrick-rivos", "labels": ["crash", "llvm:instcombine"], "comments": [{"author": "dtcxzyw", "body": "Reduced test case: https://godbolt.org/z/5qocbWsW3\r\n```\r\ndefine @test( %p) {\r\nentry:\r\n %1 = sdiv %p, splat (i64 309383)\r\n %2 = icmp sge %1, zeroinitializer\r\n ret %2\r\n}\r\n\r\n```"}]}, "verified": true} {"bug_id": "111042", "issue_url": "https://github.com/llvm/llvm-project/issues/111042", "bug_type": "crash", "base_commit": "67c0846357bcd6faca713315380f9981a805a6e5", "knowledge_cutoff": "2024-10-03T18:44:33Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "68210c7c269b7b7325dedaea7cea9fe9d979fab5", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[2185, 2190]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPPredInstPHIRecipe::execute"]}}, "patch": "commit 68210c7c269b7b7325dedaea7cea9fe9d979fab5\nAuthor: Florian Hahn \nDate: Sat Oct 5 19:15:05 2024 +0100\n\n [VPlan] Only generate first lane for VPPredInstPHI if no others used.\n \n IF only the first lane of the result is used, only generate the first\n lane.\n \n Fixes https://github.com/llvm/llvm-project/issues/111042.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex 0d092b9c10ac..376b3d33be76 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -2185,6 +2185,9 @@ void VPPredInstPHIRecipe::execute(VPTransformState &State) {\n // predicated iteration inserts its generated value in the correct vector.\n State.reset(getOperand(0), VPhi);\n } else {\n+ if (vputils::onlyFirstLaneUsed(this) && !State.Lane->isFirstLane())\n+ return;\n+\n Type *PredInstType = getOperand(0)->getUnderlyingValue()->getType();\n PHINode *Phi = State.Builder.CreatePHI(PredInstType, 2);\n Phi->addIncoming(PoisonValue::get(ScalarPredInst->getType()),\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/divs-with-tail-folding.ll", "commands": ["opt -p loop-vectorize -mtriple x86_64 -prefer-predicate-over-epilogue=predicate-dont-vectorize -mcpu=skylake-avx512 -S %s"], "tests": [{"test_name": "sdiv_feeding_gep_predicated", "test_body": "define void @sdiv_feeding_gep_predicated(ptr %dst, i32 %x, i64 %M, i64 %conv6, i64 %N) {\nentry:\n %conv61 = zext i32 %x to i64\n br label %loop\n\nloop: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %c = icmp ule i64 %iv, %M\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop\n %div18 = sdiv i64 %M, %conv6\n %conv20 = trunc i64 %div18 to i32\n %mul30 = mul i64 %div18, %conv61\n %sub31 = sub i64 %iv, %mul30\n %conv34 = trunc i64 %sub31 to i32\n %mul35 = mul i32 %x, %conv20\n %add36 = add i32 %mul35, %conv34\n %idxprom = sext i32 %add36 to i64\n %gep = getelementptr double, ptr %dst, i64 %idxprom\n store double 0.000000e+00, ptr %gep, align 8\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/drop-poison-generating-flags.ll", "commands": ["opt %s -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S"], "tests": [{"test_name": "pr70590_recipe_without_underlying_instr", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n@c = external global [5 x i8]\n\ndefine void @pr70590_recipe_without_underlying_instr(i64 %n, ptr noalias %dst) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %inc, %loop.latch ]\n %cmp = icmp eq i64 %iv, %n\n br i1 %cmp, label %loop.latch, label %then\n\nthen: ; preds = %loop.header\n %rem = srem i64 3, 0\n %add3 = add i64 %rem, -3\n %add5 = add i64 %iv, %add3\n %gep = getelementptr [5 x i8], ptr @c, i64 0, i64 %add5\n %l = load i8, ptr %gep, align 1\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %sr = phi i8 [ 0, %loop.header ], [ %l, %then ]\n %gep.dst = getelementptr i8, ptr %dst, i64 %iv\n store i8 %sr, ptr %gep.dst, align 4\n %inc = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %inc, 4\n br i1 %exitcond.not, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/replicate-recipe-with-only-first-lane-used.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=2 -S %s"], "tests": [{"test_name": "replicate_udiv_with_only_first_lane_used", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @replicate_udiv_with_only_first_lane_used(i32 %x, ptr %dst, i64 %d) {\nentry:\n %c = icmp eq i32 %x, 10\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 true, label %loop.latch, label %else\n\nelse: ; preds = %loop.header\n %div.i = udiv i64 99, %d\n br label %loop.latch\n\nloop.latch: ; preds = %else, %loop.header\n %retval.0.i = phi i64 [ %div.i, %else ], [ 0, %loop.header ]\n %gep = getelementptr i16, ptr %dst, i64 %retval.0.i\n store i16 0, ptr %gep, align 2\n %iv.next = add i64 %iv, 1\n %cmp = icmp ult i64 %iv.next, 101\n br i1 %cmp, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}], "issue": {"title": "[clang] Assertion failed in VPlan.cpp:258", "body": "When I compiled the code with -O2 flag, it crashed:\r\n\r\n```c\r\n#include \r\nstatic uint8_t c[] = {1};\r\nvoid a();\r\nlong b(int h) {\r\n long d = 0;\r\n if (h < 0)\r\n return -1;\r\n int e = 0;\r\n for (int f = 0; f <= h; f++) {\r\n e++;\r\n d += f;\r\n }\r\n long a = d / e;\r\n return a;\r\n}\r\nvoid g();\r\nuint8_t i() {\r\n uint16_t k[4];\r\n int f;\r\n for (; g + f < 4; f++)\r\n k[b(c[0] - 1)] = 5;\r\n for (; f; a(k))\r\n ;\r\n uint8_t *l[3][3], j;\r\n l[0][j] = c;\r\n}\r\n```\r\n\r\nThe crash is:\r\n```\r\nclang: /root/llvm-project/llvm/lib/Transforms/Vectorize/VPlan.cpp:258: llvm::Value* llvm::VPTransformState::get(llvm::VPValue*, bool): Assertion `(VF.isScalar() || Def->isLiveIn() || hasVectorValue(Def) || !vputils::onlyFirstLaneUsed(Def) || (hasScalarValue(Def, VPLane(0)) && Data.VPV2Scalars[Def].size() == 1)) && \"Trying to access a single scalar per part but has multiple scalars \" \"per part.\"' failed.\r\n```\r\n\r\nDetails can be found here: https://godbolt.org/z/fTrKKha5E", "author": "cardigan1008", "labels": ["vectorizers", "crash", "regression:19"], "comments": [{"author": "shafik", "body": "Looks like a clang-19 regression: https://godbolt.org/z/6d3edKn86"}]}, "verified": true} {"bug_id": "89959", "issue_url": "https://github.com/llvm/llvm-project/issues/89959", "bug_type": "crash", "base_commit": "bc8a4ea11070d06374b403cd09b771a99cc6ba1a", "knowledge_cutoff": "2024-04-24T17:43:58Z", "lit_test_dir": ["llvm/test/Transforms/IRCE"], "hints": {"fix_commit": "22da5a6e34ed6146752b24d9156a678b50fddaef", "components": ["InductiveRangeCheckElimination"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp": [[279, 284]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp": ["InductiveRangeCheck::parseRangeCheckICmp"]}}, "patch": "commit 22da5a6e34ed6146752b24d9156a678b50fddaef\nAuthor: Yingwei Zheng \nDate: Fri Apr 26 16:25:33 2024 +0800\n\n [IRCE] Skip icmp ptr in `InductiveRangeCheck::parseRangeCheckICmp` (#89967)\n \n Fixes https://github.com/llvm/llvm-project/issues/89959.\n\ndiff --git a/llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp b/llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp\nindex 9df28747570c..104e8ceb7967 100644\n--- a/llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp\n+++ b/llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp\n@@ -279,6 +279,9 @@ bool InductiveRangeCheck::parseRangeCheckICmp(Loop *L, ICmpInst *ICI,\n Value *LHS = ICI->getOperand(0);\n Value *RHS = ICI->getOperand(1);\n \n+ if (!LHS->getType()->isIntegerTy())\n+ return false;\n+\n // Canonicalize to the `Index Pred Invariant` comparison\n if (IsLoopInvariant(LHS)) {\n std::swap(LHS, RHS);\n", "tests": [{"file": "llvm/test/Transforms/IRCE/pr89959.ll", "commands": ["opt -passes=irce -S < %s 2>&1"], "tests": [{"test_name": "", "test_body": "\n; Make sure we don't crash.\ndefine void @pr89959() {\n;\ntop:\n br label %L3\n\nL3:\n %value_phi = phi ptr [ null, %top ], [ %0, %L13 ]\n %0 = getelementptr i8, ptr %value_phi, i64 8\n %.not = icmp ule ptr %value_phi, null\n br i1 %.not, label %L13, label %L15\n\nL13:\n br label %L3\n\nL15:\n ret void\n}"}]}], "issue": {"title": "Assertion in Scalar Evolution when running IRCE", "body": "The following IR crashes with \r\n```llvm\r\ndefine void @foo() {\r\ntop:\r\n br label %L3\r\n\r\nL3: ; preds = %L13, %top\r\n %value_phi = phi ptr [ null, %top ], [ %0, %L13 ]\r\n %0 = getelementptr i8, ptr %value_phi, i64 8\r\n %.not = icmp ule ptr %value_phi, null\r\n br i1 %.not, label %L13, label %L15\r\n\r\nL13: ; preds = %L3\r\n br label %L3\r\n\r\nL15: ; preds = %L3\r\n ret void\r\n}\r\n```\r\n\r\n`opt --passes=irce` \r\n\r\nhttps://godbolt.org/z/1sbvaT7r1 ", "author": "gbaraldi", "labels": ["llvm:SCEV", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "107051", "issue_url": "https://github.com/llvm/llvm-project/issues/107051", "bug_type": "crash", "base_commit": "fcb7b390ccd5b4cfc71f13b5e16a846f3f400c10", "knowledge_cutoff": "2024-09-03T06:19:24Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "3bd161e98d89d31696002994771b7761f1c74859", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[1290, 1297], [6194, 6199], [6206, 6219], [6229, 6248]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::setVectorizedCallDecision", "collectUniformsAndScalars"]}}, "patch": "commit 3bd161e98d89d31696002994771b7761f1c74859\nAuthor: Florian Hahn \nDate: Tue Sep 3 21:06:31 2024 +0100\n\n [LV] Honor forced scalars in setVectorizedCallDecision.\n \n Similarly to dd94537b4, setVectorizedCallDecision also did not consider\n ForcedScalars. This lead to VPlans not reflecting the decision by the\n legacy cost model (cost computation would use scalar cost, VPlan would\n have VPWidenCallRecipe).\n \n To fix this, check if the call has been forced to scalar in\n setVectorizedCallDecision.\n \n Note that this requires moving setVectorizedCallDecision after\n collectLoopUniforms (which sets ForcedScalars). collectLoopUniforms does\n not depend on call decisions and can safely be moved.\n \n Fixes https://github.com/llvm/llvm-project/issues/107051.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 17050b2b433c..0200525a718d 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -1290,8 +1290,8 @@ public:\n if (VF.isScalar() || Uniforms.contains(VF))\n return;\n setCostBasedWideningDecision(VF);\n- setVectorizedCallDecision(VF);\n collectLoopUniforms(VF);\n+ setVectorizedCallDecision(VF);\n collectLoopScalars(VF);\n }\n \n@@ -6194,6 +6194,7 @@ void LoopVectorizationCostModel::setVectorizedCallDecision(ElementCount VF) {\n assert(!VF.isScalar() &&\n \"Trying to set a vectorization decision for a scalar VF\");\n \n+ auto ForcedScalar = ForcedScalars.find(VF);\n for (BasicBlock *BB : TheLoop->blocks()) {\n // For each instruction in the old loop.\n for (Instruction &I : *BB) {\n@@ -6206,14 +6207,37 @@ void LoopVectorizationCostModel::setVectorizedCallDecision(ElementCount VF) {\n InstructionCost VectorCost = InstructionCost::getInvalid();\n InstructionCost IntrinsicCost = InstructionCost::getInvalid();\n TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n-\n Function *ScalarFunc = CI->getCalledFunction();\n Type *ScalarRetTy = CI->getType();\n SmallVector Tys, ScalarTys;\n- bool MaskRequired = Legal->isMaskRequired(CI);\n for (auto &ArgOp : CI->args())\n ScalarTys.push_back(ArgOp->getType());\n \n+ // Estimate cost of scalarized vector call. The source operands are\n+ // assumed to be vectors, so we need to extract individual elements from\n+ // there, execute VF scalar calls, and then gather the result into the\n+ // vector return value.\n+ InstructionCost ScalarCallCost =\n+ TTI.getCallInstrCost(ScalarFunc, ScalarRetTy, ScalarTys, CostKind);\n+\n+ // Compute costs of unpacking argument values for the scalar calls and\n+ // packing the return values to a vector.\n+ InstructionCost ScalarizationCost =\n+ getScalarizationOverhead(CI, VF, CostKind);\n+\n+ ScalarCost = ScalarCallCost * VF.getKnownMinValue() + ScalarizationCost;\n+ // Honor ForcedScalars decision.\n+ // TODO: For calls, it might still be more profitable to widen. Use\n+ // VPlan-based cost model to compare different options.\n+ if (VF.isVector() && ForcedScalar != ForcedScalars.end() &&\n+ ForcedScalar->second.contains(CI)) {\n+ setCallWideningDecision(CI, VF, CM_Scalarize, nullptr,\n+ Intrinsic::not_intrinsic, std::nullopt,\n+ ScalarCost);\n+ continue;\n+ }\n+\n+ bool MaskRequired = Legal->isMaskRequired(CI);\n // Compute corresponding vector type for return value and arguments.\n Type *RetTy = ToVectorTy(ScalarRetTy, VF);\n for (Type *ScalarTy : ScalarTys)\n@@ -6229,20 +6253,6 @@ void LoopVectorizationCostModel::setVectorizedCallDecision(ElementCount VF) {\n continue;\n }\n \n- // Estimate cost of scalarized vector call. The source operands are\n- // assumed to be vectors, so we need to extract individual elements from\n- // there, execute VF scalar calls, and then gather the result into the\n- // vector return value.\n- InstructionCost ScalarCallCost =\n- TTI.getCallInstrCost(ScalarFunc, ScalarRetTy, ScalarTys, CostKind);\n-\n- // Compute costs of unpacking argument values for the scalar calls and\n- // packing the return values to a vector.\n- InstructionCost ScalarizationCost =\n- getScalarizationOverhead(CI, VF, CostKind);\n-\n- ScalarCost = ScalarCallCost * VF.getKnownMinValue() + ScalarizationCost;\n-\n // Find the cost of vectorizing the call, if we can find a suitable\n // vector variant of the function.\n bool UsesMask = false;\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/call-costs.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "call_forced_scalar", "test_body": "target triple = \"arm64-apple-macosx11.0.0\"\n\ndefine void @call_forced_scalar(ptr %src.1, ptr %src.2, ptr noalias %dst.1, ptr noalias %dst.2) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %0 = load i32, ptr %src.1, align 4\n %smax = tail call i32 @llvm.smax.i32(i32 %0, i32 0)\n %umin = tail call i32 @llvm.umin.i32(i32 %smax, i32 1)\n %umin.ext = zext i32 %umin to i64\n %gep.src.2 = getelementptr i8, ptr %src.2, i64 %umin.ext\n %1 = load i8, ptr %gep.src.2, align 1\n %l.ext = zext i8 %1 to i32\n %mul = mul i32 3, %l.ext\n store i32 %mul, ptr %dst.1, align 4\n %gep.dst.2 = getelementptr i32, ptr %dst.2, i64 %iv\n store i32 0, ptr %gep.dst.2, align 4\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 0\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.smax.i32(i32, i32) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.umin.i32(i32, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "call_scalarized", "test_body": "target triple = \"arm64-apple-macosx11.0.0\"\n\ndefine void @call_scalarized(ptr noalias %src, ptr noalias %dst, double %0) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 100, %entry ], [ %iv.next, %loop.latch ]\n %iv.next = add i64 %iv, -1\n %gep.src = getelementptr double, ptr %src, i64 %iv.next\n %l = load double, ptr %gep.src, align 8\n %cmp295 = fcmp ugt double %0, 0.000000e+00\n %cmp299 = fcmp ugt double %l, 0.000000e+00\n %or.cond = or i1 %cmp295, %cmp299\n br i1 %or.cond, label %loop.latch, label %then\n\nthen: ; preds = %loop.header\n %sqrt = call double @llvm.sqrt.f64(double %l)\n %gep.dst = getelementptr double, ptr %dst, i64 %iv.next\n store double %sqrt, ptr %gep.dst, align 8\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %tobool.not = icmp eq i64 %iv.next, 0\n br i1 %tobool.not, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare double @llvm.sqrt.f64(double) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "fshl_operand_first_order_recurrence", "test_body": "target triple = \"arm64-apple-macosx11.0.0\"\n\ndefine void @fshl_operand_first_order_recurrence(ptr %dst, ptr noalias %src) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %recur = phi i64 [ 0, %entry ], [ %l, %loop ]\n %gep.src = getelementptr inbounds i64, ptr %src, i64 %iv\n %l = load i64, ptr %gep.src, align 8\n %or = tail call i64 @llvm.fshl.i64(i64 1, i64 %recur, i64 1)\n %gep.dst = getelementptr inbounds i64, ptr %dst, i64 %iv\n store i64 %or, ptr %gep.dst, align 8\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv, 100\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i64 @llvm.fshl.i64(i64, i64, i64) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "powi_call", "test_body": "target triple = \"arm64-apple-macosx11.0.0\"\n\ndefine void @powi_call(ptr %P) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr inbounds double, ptr %P, i64 %iv\n %l = load double, ptr %gep, align 8\n %powi = tail call double @llvm.powi.f64.i32(double %l, i32 3)\n store double %powi, ptr %gep, align 8\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare double @llvm.powi.f64.i32(double, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[clang] Assertion failed in Vectorize", "body": "I compiled this code with -Os and a crash appeared in assertion failure:\r\n\r\n```c\r\n#include \r\nint32_t *c;\r\nint8_t ****d;\r\nuint64_t g;\r\nint a();\r\nlong b(long, long h, long p3) {\r\n long e;\r\n int f = 0;\r\n if (h) {\r\n e = h;\r\n f = 1;\r\n }\r\n if (e > p3)\r\n f = 2;\r\n switch (f) {\r\n case 1:\r\n e++;\r\n case 2:\r\n e--;\r\n }\r\n if (e < h)\r\n e = h;\r\n return e;\r\n}\r\n\r\nuint8_t div_func_uint8_t_u_u(uint8_t ui1, uint8_t ui2)\r\n{\r\n return (ui1 / ui2);\r\n}\r\n\r\nuint64_t j() {\r\n uint16_t k[8];\r\n int i;\r\n for (i = 0; *c + ****d + (int)g + i < 8; i++)\r\n k[b(0, g, 0)] = 6;\r\n *c = div_func_uint8_t_u_u(k[6], a());\r\n}\r\n```\r\n\r\nThe crash is:\r\n```\r\nclang: /root/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7369: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\n```\r\n\r\nDetails can be found here: https://godbolt.org/z/T81adKPY5\r\n\r\n", "author": "cardigan1008", "labels": ["vectorizers", "crash"], "comments": [{"author": "patrick-rivos", "body": "Unreduced LLVM IR:\r\n```llvm ir\r\n; ModuleID = '/app/example.c'\r\nsource_filename = \"/app/example.c\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\n@c = dso_local local_unnamed_addr global ptr null, align 8\r\n@d = dso_local local_unnamed_addr global ptr null, align 8\r\n@g = dso_local local_unnamed_addr global i64 0, align 8\r\n\r\n; Function Attrs: mustprogress nofree norecurse nosync nounwind optsize willreturn memory(none) uwtable\r\ndefine dso_local i64 @b(i64 noundef %0, i64 noundef %h, i64 noundef %p3) local_unnamed_addr #0 {\r\nentry:\r\n %tobool.not = icmp ne i64 %h, 0\r\n %spec.select = zext i1 %tobool.not to i32\r\n %cmp = icmp sgt i64 %h, %p3\r\n %f.1 = select i1 %cmp, i32 2, i32 %spec.select\r\n switch i32 %f.1, label %sw.epilog [\r\n i32 1, label %sw.bb\r\n i32 2, label %sw.bb3\r\n ]\r\n\r\nsw.bb: ; preds = %entry\r\n %inc = add nsw i64 %h, 1\r\n br label %sw.bb3\r\n\r\nsw.bb3: ; preds = %entry, %sw.bb\r\n %e.1 = phi i64 [ %h, %entry ], [ %inc, %sw.bb ]\r\n %dec = add nsw i64 %e.1, -1\r\n br label %sw.epilog\r\n\r\nsw.epilog: ; preds = %sw.bb3, %entry\r\n %e.2 = phi i64 [ %h, %entry ], [ %dec, %sw.bb3 ]\r\n %spec.select14 = tail call i64 @llvm.smax.i64(i64 %e.2, i64 %h)\r\n ret i64 %spec.select14\r\n}\r\n\r\n; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)\r\ndeclare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1\r\n\r\n; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)\r\ndeclare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1\r\n\r\n; Function Attrs: mustprogress nofree norecurse nosync nounwind optsize willreturn memory(none) uwtable\r\ndefine dso_local noundef zeroext i8 @div_func_uint8_t_u_u(i8 noundef zeroext %ui1, i8 noundef zeroext %ui2) local_unnamed_addr #0 {\r\nentry:\r\n %0 = udiv i8 %ui1, %ui2\r\n ret i8 %0\r\n}\r\n\r\n; Function Attrs: nounwind optsize uwtable\r\ndefine dso_local i64 @j() local_unnamed_addr #2 {\r\nentry:\r\n %k = alloca [8 x i16], align 16\r\n call void @llvm.lifetime.start.p0(i64 16, ptr nonnull %k) #5\r\n %0 = load ptr, ptr @c, align 8, !tbaa !6\r\n %1 = load i32, ptr %0, align 4, !tbaa !10\r\n %2 = load ptr, ptr @d, align 8, !tbaa !6\r\n %3 = load ptr, ptr %2, align 8, !tbaa !6\r\n %4 = load ptr, ptr %3, align 8, !tbaa !6\r\n %5 = load ptr, ptr %4, align 8, !tbaa !6\r\n %6 = load i8, ptr %5, align 1, !tbaa !12\r\n %conv = sext i8 %6 to i32\r\n %7 = load i64, ptr @g, align 8, !tbaa !13\r\n %conv1 = trunc i64 %7 to i32\r\n %invariant.op = add i32 %conv, %conv1\r\n %add3.reass12 = add i32 %1, %invariant.op\r\n %cmp13 = icmp slt i32 %add3.reass12, 8\r\n br i1 %cmp13, label %for.body.lr.ph, label %for.end\r\n\r\nfor.body.lr.ph: ; preds = %entry\r\n %tobool.not.i = icmp ne i64 %7, 0\r\n %spec.select.i = zext i1 %tobool.not.i to i32\r\n %cmp.i = icmp sgt i64 %7, 0\r\n %f.1.i = select i1 %cmp.i, i32 2, i32 %spec.select.i\r\n %inc.i = add nsw i64 %7, 1\r\n br label %for.body\r\n\r\nfor.body: ; preds = %for.body.lr.ph, %b.exit\r\n %i.014 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %b.exit ]\r\n switch i32 %f.1.i, label %b.exit [\r\n i32 1, label %sw.bb.i\r\n i32 2, label %sw.bb3.i\r\n ]\r\n\r\nsw.bb.i: ; preds = %for.body\r\n br label %sw.bb3.i\r\n\r\nsw.bb3.i: ; preds = %sw.bb.i, %for.body\r\n %e.1.i = phi i64 [ %7, %for.body ], [ %inc.i, %sw.bb.i ]\r\n %dec.i = add nsw i64 %e.1.i, -1\r\n br label %b.exit\r\n\r\nb.exit: ; preds = %for.body, %sw.bb3.i\r\n %e.2.i = phi i64 [ %7, %for.body ], [ %dec.i, %sw.bb3.i ]\r\n %spec.select14.i = tail call i64 @llvm.smax.i64(i64 %e.2.i, i64 %7)\r\n %arrayidx = getelementptr inbounds [8 x i16], ptr %k, i64 0, i64 %spec.select14.i\r\n store i16 6, ptr %arrayidx, align 2, !tbaa !15\r\n %inc = add nuw nsw i32 %i.014, 1\r\n %add = add i32 %1, %inc\r\n %add3.reass = add i32 %add, %invariant.op\r\n %cmp = icmp slt i32 %add3.reass, 8\r\n br i1 %cmp, label %for.body, label %for.end.loopexit, !llvm.loop !17\r\n\r\nfor.end.loopexit: ; preds = %b.exit\r\n %arrayidx5.phi.trans.insert = getelementptr inbounds i8, ptr %k, i64 12\r\n %.pre = load i16, ptr %arrayidx5.phi.trans.insert, align 4, !tbaa !15\r\n %8 = trunc i16 %.pre to i8\r\n br label %for.end\r\n\r\nfor.end: ; preds = %for.end.loopexit, %entry\r\n %conv6 = phi i8 [ %8, %for.end.loopexit ], [ undef, %entry ]\r\n %call7 = tail call i32 (...) @a() #6\r\n %conv8 = trunc i32 %call7 to i8\r\n %9 = udiv i8 %conv6, %conv8\r\n %conv10 = zext i8 %9 to i32\r\n %10 = load ptr, ptr @c, align 8, !tbaa !6\r\n store i32 %conv10, ptr %10, align 4, !tbaa !10\r\n call void @llvm.lifetime.end.p0(i64 16, ptr nonnull %k) #5\r\n ret i64 undef\r\n}\r\n\r\n; Function Attrs: optsize\r\ndeclare i32 @a(...) local_unnamed_addr #3\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i64 @llvm.smax.i64(i64, i64) #4\r\n\r\nattributes #0 = { mustprogress nofree norecurse nosync nounwind optsize willreturn memory(none) uwtable \"min-legal-vector-width\"=\"0\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\r\nattributes #1 = { mustprogress nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }\r\nattributes #2 = { nounwind optsize uwtable \"min-legal-vector-width\"=\"0\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\r\nattributes #3 = { optsize \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\r\nattributes #4 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\nattributes #5 = { nounwind }\r\nattributes #6 = { nounwind optsize }\r\n\r\n!llvm.module.flags = !{!0, !1, !2, !3, !4}\r\n!llvm.ident = !{!5}\r\n\r\n!0 = !{i32 7, !\"Dwarf Version\", i32 4}\r\n!1 = !{i32 1, !\"wchar_size\", i32 4}\r\n!2 = !{i32 8, !\"PIC Level\", i32 2}\r\n!3 = !{i32 7, !\"PIE Level\", i32 2}\r\n!4 = !{i32 7, !\"uwtable\", i32 2}\r\n!5 = !{!\"clang version 20.0.0git (https://github.com/llvm/llvm-project.git b6597f521d8a040f2b9fee54b3f89c380de8e432)\"}\r\n!6 = !{!7, !7, i64 0}\r\n!7 = !{!\"any pointer\", !8, i64 0}\r\n!8 = !{!\"omnipotent char\", !9, i64 0}\r\n!9 = !{!\"Simple C/C++ TBAA\"}\r\n!10 = !{!11, !11, i64 0}\r\n!11 = !{!\"int\", !8, i64 0}\r\n!12 = !{!8, !8, i64 0}\r\n!13 = !{!14, !14, i64 0}\r\n!14 = !{!\"long\", !8, i64 0}\r\n!15 = !{!16, !16, i64 0}\r\n!16 = !{!\"short\", !8, i64 0}\r\n!17 = distinct !{!17, !18}\r\n!18 = !{!\"llvm.loop.mustprogress\"}\r\n```\r\n\r\nhttps://godbolt.org/z/Yo1crEjGG\r\n\r\ncc @fhahn "}]}, "verified": true} {"bug_id": "99701", "issue_url": "https://github.com/llvm/llvm-project/issues/99701", "bug_type": "crash", "base_commit": "9a258664024d12a06ba8eb9344e270a9bb5f5d87", "knowledge_cutoff": "2024-07-19T20:27:53Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "ba8126b6fef79bd344a247f6291aaec7b67bdff0", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[137, 142], [6681, 6686], [6690, 6695], [6717, 6722]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::collectValuesToIgnore"]}}, "patch": "commit ba8126b6fef79bd344a247f6291aaec7b67bdff0\nAuthor: Florian Hahn \nDate: Wed Jul 24 09:31:32 2024 +0100\n\n [LV] Mark dead instructions in loop as free.\n \n Update collectValuesToIgnore to also ignore dead instructions in the\n loop. Such instructions will be removed by VPlan-based DCE and won't be\n considered by the VPlan-based cost model.\n \n This closes a gap between the legacy and VPlan-based cost model. In\n practice with the default pipelines, there shouldn't be any dead\n instructions in loops reaching LoopVectorize, but it is easy to generate\n such cases by hand or automatically via fuzzers.\n \n Fixes https://github.com/llvm/llvm-project/issues/99701.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 8fa1a57177d9..e4477f96aa62 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -137,6 +137,7 @@\n #include \"llvm/Support/raw_ostream.h\"\n #include \"llvm/Transforms/Utils/BasicBlockUtils.h\"\n #include \"llvm/Transforms/Utils/InjectTLIMappings.h\"\n+#include \"llvm/Transforms/Utils/Local.h\"\n #include \"llvm/Transforms/Utils/LoopSimplify.h\"\n #include \"llvm/Transforms/Utils/LoopUtils.h\"\n #include \"llvm/Transforms/Utils/LoopVersioning.h\"\n@@ -6681,6 +6682,7 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n CodeMetrics::collectEphemeralValues(TheLoop, AC, ValuesToIgnore);\n \n SmallVector DeadInterleavePointerOps;\n+ SmallVector DeadOps;\n for (BasicBlock *BB : TheLoop->blocks())\n for (Instruction &I : *BB) {\n // Find all stores to invariant variables. Since they are going to sink\n@@ -6690,6 +6692,17 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n Legal->isInvariantAddressOfReduction(SI->getPointerOperand()))\n ValuesToIgnore.insert(&I);\n \n+ if (VecValuesToIgnore.contains(&I) || ValuesToIgnore.contains(&I))\n+ continue;\n+\n+ // Add instructions that would be trivially dead and are only used by\n+ // values already ignored to DeadOps to seed worklist.\n+ if (wouldInstructionBeTriviallyDead(&I, TLI) &&\n+ all_of(I.users(), [this](User *U) {\n+ return VecValuesToIgnore.contains(U) || ValuesToIgnore.contains(U);\n+ }))\n+ DeadOps.push_back(&I);\n+\n // For interleave groups, we only create a pointer for the start of the\n // interleave group. Queue up addresses of group members except the insert\n // position for further processing.\n@@ -6717,6 +6730,29 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n DeadInterleavePointerOps.append(Op->op_begin(), Op->op_end());\n }\n \n+ // Mark ops that would be trivially dead and are only used by ignored\n+ // instructions as free.\n+ for (unsigned I = 0; I != DeadOps.size(); ++I) {\n+ auto *Op = dyn_cast(DeadOps[I]);\n+ // Skip any op that shouldn't be considered dead.\n+ if (!Op || !TheLoop->contains(Op) ||\n+ !wouldInstructionBeTriviallyDead(Op, TLI) ||\n+ any_of(Op->users(), [this](User *U) {\n+ return !VecValuesToIgnore.contains(U) && !ValuesToIgnore.contains(U);\n+ }))\n+ continue;\n+\n+ // If all of Op's users are in ValuesToIgnore, add it to ValuesToIgnore\n+ // which applies for both scalar and vector versions. Otherwise it is only\n+ // dead in vector versions, so only add it to VecValuesToIgnore.\n+ if (all_of(Op->users(),\n+ [this](User *U) { return ValuesToIgnore.contains(U); }))\n+ ValuesToIgnore.insert(Op);\n+\n+ VecValuesToIgnore.insert(Op);\n+ DeadOps.append(Op->op_begin(), Op->op_end());\n+ }\n+\n // Ignore type-promoting instructions we identified during reduction\n // detection.\n for (const auto &Reduction : Legal->getReductionVars()) {\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll", "commands": ["opt -p loop-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S %s"], "tests": [{"test_name": "", "test_body": "\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\n; Test with a dead load in the loop, from\n; https://github.com/llvm/llvm-project/issues/99701\ndefine void @dead_load(ptr %p, i16 %start) {\n;\nentry:\n %start.ext = sext i16 %start to i64\n br label %loop\n\nloop:\n %iv = phi i64 [ %start.ext, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr i16, ptr %p, i64 %iv\n store i16 0, ptr %gep, align 2\n %l = load i16, ptr %gep, align 2\n %iv.next = add i64 %iv, 3\n %cmp = icmp slt i64 %iv, 111\n br i1 %cmp, label %loop, label %exit\n\nexit:\n ret void\n}\n;.\n;."}]}], "issue": {"title": "[VPlan] Assertion `VF.Width == Width && \"VPlan cost model and legacy cost model disagreed\"' failed.", "body": "I have a few finds by the fuzzer that affect LLVM (tested using a1d77caaabbb5279b734c061dab36b2138ec476d).\r\nI think they're all duplicates but I'm attaching them all in a .zip so it can be confirmed.\r\n\r\nTestcase:\r\n```llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @e(i64 %0) #0 {\r\nentry:\r\n br label %for.body\r\n\r\nfor.cond.for.cond.cleanup_crit_edge: ; preds = %for.body\r\n ret void\r\n\r\nfor.body: ; preds = %for.body, %entry\r\n %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]\r\n %arrayidx = getelementptr i16, ptr null, i64 %indvars.iv\r\n %1 = load i16, ptr %arrayidx, align 2\r\n %arrayidx5 = getelementptr [0 x i8], ptr null, i64 0, i64 %indvars.iv\r\n store i8 0, ptr %arrayidx5, align 1\r\n %indvars.iv.next = add nsw i64 %indvars.iv, 4\r\n %cmp = icmp ult i64 %indvars.iv, %0\r\n br i1 %cmp, label %for.body, label %for.cond.for.cond.cleanup_crit_edge\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zacas,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n\r\n```\r\n\r\nCommand/backtrace:\r\n```bash\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10444: bool llvm::LoopVectorizePass::processLoop(llvm::Loop*): Assertion `VF.Width == Width && \"VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"e\"\r\n #0 0x000055b365fe4a60 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f0ea60)\r\n #1 0x000055b365fe1e7f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f0be7f)\r\n #2 0x000055b365fe1fd5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007612a3042520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007612a30969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007612a30969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007612a30969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007612a3042476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007612a30287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007612a302871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007612a3039e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x000055b365005ba8 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f2fba8)\r\n#12 0x000055b365008761 llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f32761)\r\n#13 0x000055b365008ee0 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f32ee0)\r\n#14 0x000055b363e81296 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdab296)\r\n#15 0x000055b365dfd4be llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d274be)\r\n#16 0x000055b363e84426 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdae426)\r\n#17 0x000055b365dfc24b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d2624b)\r\n#18 0x000055b363e80776 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdaa776)\r\n#19 0x000055b365dfa39d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d2439d)\r\n#20 0x000055b3636cbdc6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5f5dc6)\r\n#21 0x000055b3636bda11 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5e7a11)\r\n#22 0x00007612a3029d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#23 0x00007612a3029e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#24 0x00007612a3029e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#25 0x000055b3636b3855 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5dd855)\r\nzsh: IOT instruction /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\n\r\nGodbolt: https://godbolt.org/z/a7KPnMh94\r\n\r\nZip with all finds reduced: \r\n[finds.zip](https://github.com/user-attachments/files/16316789/finds.zip)\r\n\r\nRelated: https://github.com/llvm/llvm-project/pull/92555 @fhahn \r\n\r\nFound via fuzzer.", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": [{"author": "JonPsson1", "body": "I'm seeing this as well with a multitude of hits with csmith. One reduced test case:\r\n\r\nclang -O3 -march=z16 crash0.i -o a.out -w -mllvm -disable-licm-promotion -mllvm -force-target-instruction-cost=1 -mllvm -unroll-count=4\r\nLoopVectorize.cpp:10071: bool llvm::LoopVectorizePass::processLoop(llvm::Loop*): Assertion `VF.Width == Width && \"VPlan cost model and legacy cost model disagreed\"' failed.\r\n#9 0x0000000004749e30 llvm::LoopVectorizePass::processLoop()\r\n\r\n[tc_lv.i.tar.gz](https://github.com/user-attachments/files/16337463/tc_lv.i.tar.gz)\r\n\r\n"}, {"author": "fhahn", "body": "Thanks @patrick-rivos, it looks like all test cases had dead instructions, which would be considered by the legacy cost model but not the VPlan-based one (because VPlan-based DCE would clean them up before). Should be fixed now.\r\n\r\n@JonPsson1's case was a different issue and should also be fixed."}]}, "verified": true} {"bug_id": "122496", "issue_url": "https://github.com/llvm/llvm-project/issues/122496", "bug_type": "miscompilation", "base_commit": "b4ce29ab31b29ca926704c160e3909298ddf2b2b", "knowledge_cutoff": "2025-01-10T16:59:46Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "8df64ed77727ab9b7540819f2fe64379e88a50be", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[3806, 3812]], "llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": [[621, 648]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::collectLoopUniforms"], "llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": ["legalizeAndOptimizeInductions"]}}, "patch": "commit 8df64ed77727ab9b7540819f2fe64379e88a50be\nAuthor: Florian Hahn \nDate: Sun Jan 12 22:03:21 2025 +0000\n\n [LV] Don't consider IV increments uniform if exit value is used outside.\n \n In some cases, there might be a chain of uniform instructions producing\n the exit value. To generate correct code in all cases, consider the IV\n increment not uniform, if there are users outside the loop.\n \n Instead, let VPlan narrow the IV, if possible using the logic from\n 3ff1d01985752.\n \n Test case from #122602 verified with Alive2:\n https://alive2.llvm.org/ce/z/bA4EGj\n \n Fixes https://github.com/llvm/llvm-project/issues/122496.\n Fixes https://github.com/llvm/llvm-project/issues/122602.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex b017b61a45a0..d32a463a996c 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -3806,7 +3806,7 @@ void LoopVectorizationCostModel::collectLoopUniforms(ElementCount VF) {\n // uniform after vectorization.\n bool UniformIndUpdate = all_of(IndUpdate->users(), [&](User *U) -> bool {\n auto *I = cast(U);\n- return I == Ind || !TheLoop->contains(I) || Worklist.count(I) ||\n+ return I == Ind || Worklist.count(I) ||\n IsVectorizedMemAccessUse(I, IndUpdate);\n });\n if (!UniformIndUpdate)\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\nindex f440bf2eb022..545d277d7aa0 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n@@ -621,28 +621,6 @@ static void legalizeAndOptimizeInductions(VPlan &Plan) {\n Def->replaceAllUsesWith(Clone);\n }\n \n- // Check if any uniform VPReplicateRecipes using the phi recipe are used by\n- // ExtractFromEnd. Those must be replaced by a regular VPReplicateRecipe to\n- // ensure the final value is available.\n- // TODO: Remove once uniformity analysis is done on VPlan.\n- for (VPUser *U : Users) {\n- auto *ExitIRI = dyn_cast(U);\n- VPValue *Op;\n- if (!ExitIRI || !match(ExitIRI->getOperand(0),\n- m_VPInstruction(\n- m_VPValue(Op), m_VPValue())))\n- continue;\n- auto *RepR = dyn_cast(Op);\n- if (!RepR || !RepR->isUniform())\n- continue;\n- assert(!RepR->isPredicated() && \"RepR must not be predicated\");\n- Instruction *I = RepR->getUnderlyingInstr();\n- auto *Clone =\n- new VPReplicateRecipe(I, RepR->operands(), /*IsUniform*/ false);\n- Clone->insertAfter(RepR);\n- RepR->replaceAllUsesWith(Clone);\n- }\n-\n // Replace wide pointer inductions which have only their scalars used by\n // PtrAdd(IndStart, ScalarIVSteps (0, Step)).\n if (auto *PtrIV = dyn_cast(&Phi)) {\n", "tests": [{"file": "", "commands": ["opt -passes=loop-vectorize -S %s"], "tests": [{"test_name": "", "test_body": "; bin/opt -passes=loop-vectorize test.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@f = global [8 x [1 x i32]] zeroinitializer\n\ndefine i32 @main() {\nentry:\n br label %for.body6.i\n\nfor.body6.i: ; preds = %for.body6.i, %entry\n %conv34.i = phi i32 [ %conv.i, %for.body6.i ], [ 0, %entry ]\n %storemerge3133.i = phi i16 [ %inc.i, %for.body6.i ], [ 0, %entry ]\n %idxprom.i = zext i32 %conv34.i to i64\n %arrayidx.i = getelementptr [8 x [1 x i32]], ptr @f, i64 0, i64 %idxprom.i\n store i32 0, ptr %arrayidx.i, align 4\n %inc.i = add i16 %storemerge3133.i, 1\n %conv.i = zext i16 %inc.i to i32\n %cmp4.i = icmp ult i16 %storemerge3133.i, 3\n br i1 %cmp4.i, label %for.body6.i, label %for.cond11thread-pre-split.i\n\nfor.cond11thread-pre-split.i: ; preds = %for.body6.i\n %0 = add i32 %conv.i, -9\n %cmp24.i = icmp ult i32 %0, -6\n br i1 %cmp24.i, label %for.cond23.i, label %for.end27.i\n\nfor.cond23.i: ; preds = %for.cond23.i, %for.cond11thread-pre-split.i\n br label %for.cond23.i\n\nfor.end27.i: ; preds = %for.cond11thread-pre-split.i\n ret i32 0\n}", "lli_expected_out": ""}]}], "issue": {"title": "[clang] Miscompilation at -O2/3", "body": "This code prints 0 at `-O0/1` and triggers SIGKILL at `-O2/3`:\n\n```c\nint printf(const char *, ...);\nint a;\nshort b, c;\nlong e;\nint f[8][1];\nunsigned g;\nint h(int i) {\n long d = 0;\n for (; (a -= i) >= 0; d += 6)\n ;\n return d;\n}\nvoid j() {\n g = 0;\n for (; h(90) + g <= 0; g++) {\n int k = -1;\n b = 0;\n for (; k + g - -1 + b <= 3; b++)\n f[b + 3][0] = c;\n for (; b + g - 3 + e <= 8; e++)\n ;\n for (; e <= 3;)\n ;\n }\n}\nint main() {\n j();\n printf(\"%d\\n\", f[0][0]);\n}\n```\n\nCompiler Explorer: https://godbolt.org/z/3x8Yc3fnW\n\nIt seems to be a recent regression. ", "author": "cardigan1008", "labels": ["regression", "miscompilation", "vectorizers"], "comments": [{"author": "fhahn", "body": "Bisecting now"}, {"author": "dtcxzyw", "body": "Reproducer: https://godbolt.org/z/6bzT9rs3K\n```\n; bin/opt -passes=loop-vectorize test.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@f = global [8 x [1 x i32]] zeroinitializer\n\ndefine i32 @main() {\nentry:\n br label %for.body6.i\n\nfor.body6.i: ; preds = %for.body6.i, %entry\n %conv34.i = phi i32 [ %conv.i, %for.body6.i ], [ 0, %entry ]\n %storemerge3133.i = phi i16 [ %inc.i, %for.body6.i ], [ 0, %entry ]\n %idxprom.i = zext i32 %conv34.i to i64\n %arrayidx.i = getelementptr [8 x [1 x i32]], ptr @f, i64 0, i64 %idxprom.i\n store i32 0, ptr %arrayidx.i, align 4\n %inc.i = add i16 %storemerge3133.i, 1\n %conv.i = zext i16 %inc.i to i32\n %cmp4.i = icmp ult i16 %storemerge3133.i, 3\n br i1 %cmp4.i, label %for.body6.i, label %for.cond11thread-pre-split.i\n\nfor.cond11thread-pre-split.i: ; preds = %for.body6.i\n %0 = add i32 %conv.i, -9\n %cmp24.i = icmp ult i32 %0, -6\n br i1 %cmp24.i, label %for.cond23.i, label %for.end27.i\n\nfor.cond23.i: ; preds = %for.cond23.i, %for.cond11thread-pre-split.i\n br label %for.cond23.i\n\nfor.end27.i: ; preds = %for.cond11thread-pre-split.i\n ret i32 0\n}\n```\n[llubi](https://github.com/dtcxzyw/llvm-ub-aware-interpreter) output:\n\nBefore:\n```\nEntering function main\n br label %for.body6.i jump to %for.body6.i\n phi i32 %conv34.i -> i32 0\n phi i16 %storemerge3133.i -> i16 0\n %idxprom.i = zext i32 %conv34.i to i64 -> i64 0\n %arrayidx.i = getelementptr [8 x [1 x i32]], ptr @f, i64 0, i64 %idxprom.i -> Ptr 16[@f]\n store i32 0, ptr %arrayidx.i, align 4\n %inc.i = add i16 %storemerge3133.i, 1 -> i16 1\n %conv.i = zext i16 %inc.i to i32 -> i32 1\n %cmp4.i = icmp ult i16 %storemerge3133.i, 3 -> T\n br i1 %cmp4.i, label %for.body6.i, label %for.cond11thread-pre-split.i jump to %for.body6.i\n phi i32 %conv34.i -> i32 1\n phi i16 %storemerge3133.i -> i16 1\n %idxprom.i = zext i32 %conv34.i to i64 -> i64 1\n %arrayidx.i = getelementptr [8 x [1 x i32]], ptr @f, i64 0, i64 %idxprom.i -> Ptr 20[@f + 4]\n store i32 0, ptr %arrayidx.i, align 4\n %inc.i = add i16 %storemerge3133.i, 1 -> i16 2\n %conv.i = zext i16 %inc.i to i32 -> i32 2\n %cmp4.i = icmp ult i16 %storemerge3133.i, 3 -> T\n br i1 %cmp4.i, label %for.body6.i, label %for.cond11thread-pre-split.i jump to %for.body6.i\n phi i32 %conv34.i -> i32 2\n phi i16 %storemerge3133.i -> i16 2\n %idxprom.i = zext i32 %conv34.i to i64 -> i64 2\n %arrayidx.i = getelementptr [8 x [1 x i32]], ptr @f, i64 0, i64 %idxprom.i -> Ptr 24[@f + 8]\n store i32 0, ptr %arrayidx.i, align 4\n %inc.i = add i16 %storemerge3133.i, 1 -> i16 3\n %conv.i = zext i16 %inc.i to i32 -> i32 3\n %cmp4.i = icmp ult i16 %storemerge3133.i, 3 -> T\n br i1 %cmp4.i, label %for.body6.i, label %for.cond11thread-pre-split.i jump to %for.body6.i\n phi i32 %conv34.i -> i32 3\n phi i16 %storemerge3133.i -> i16 3\n %idxprom.i = zext i32 %conv34.i to i64 -> i64 3\n %arrayidx.i = getelementptr [8 x [1 x i32]], ptr @f, i64 0, i64 %idxprom.i -> Ptr 28[@f + 12]\n store i32 0, ptr %arrayidx.i, align 4\n %inc.i = add i16 %storemerge3133.i, 1 -> i16 4\n %conv.i = zext i16 %inc.i to i32 -> i32 4\n %cmp4.i = icmp ult i16 %storemerge3133.i, 3 -> F\n br i1 %cmp4.i, label %for.body6.i, label %for.cond11thread-pre-split.i jump to %for.cond11thread-pre-split.i\n %0 = add i32 %conv.i, -9 -> i32 -5\n %cmp24.i = icmp ult i32 %0, -6 -> F\n br i1 %cmp24.i, label %for.cond23.i, label %for.end27.i jump to %for.end27.i\n ret i32 0\nExiting function main\n```\nAfter:\n```\nEntering function main\n br i1 false, label %scalar.ph, label %vector.ph jump to %vector.ph\n br label %vector.body jump to %vector.body\n %0 = zext i32 0 to i64 -> i64 0\n %1 = getelementptr [8 x [1 x i32]], ptr @f, i64 0, i64 %0 -> Ptr 16[@f]\n %2 = getelementptr i32, ptr %1, i32 0 -> Ptr 16[@f]\n store <4 x i32> zeroinitializer, ptr %2, align 4\n %3 = add i16 0, 1 -> i16 1\n %4 = zext i16 %3 to i32 -> i32 1\n %5 = zext i16 %3 to i32 -> i32 1\n %6 = zext i16 %3 to i32 -> i32 1\n %7 = zext i16 %3 to i32 -> i32 1\n br label %middle.block jump to %middle.block\n br i1 true, label %for.cond11thread-pre-split.i, label %scalar.ph jump to %for.cond11thread-pre-split.i\n phi i32 %conv.i.lcssa -> i32 1\n %8 = add i32 %conv.i.lcssa, -9 -> i32 -8\n %cmp24.i = icmp ult i32 %8, -6 -> T\n br i1 %cmp24.i, label %for.cond23.i.preheader, label %for.end27.i jump to %for.cond23.i.preheader\n br label %for.cond23.i jump to %for.cond23.i\n br label %for.cond23.i jump to %for.cond23.i\n br label %for.cond23.i jump to %for.cond23.i\n ...\n```\n"}]}, "verified": true} {"bug_id": "106257", "issue_url": "https://github.com/llvm/llvm-project/issues/106257", "bug_type": "crash", "base_commit": "a4989cd603b8e8185e35e3c2b7b48b422d4898be", "knowledge_cutoff": "2024-08-27T17:46:38Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "4b84288f00f3e112045bbc4be4f1539147574ead", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[8648, 8653]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["addUsersInExitBlock"]}}, "patch": "commit 4b84288f00f3e112045bbc4be4f1539147574ead\nAuthor: Florian Hahn \nDate: Wed Aug 28 19:12:04 2024 +0100\n\n [VPlan] Pass live-ins used as exit values straight to live-out.\n \n Live-ins that are used as exit values don't need to be extracted, they\n can be passed through directly. This fixes a crash when trying to\n extract from a live-in.\n \n Fixes https://github.com/llvm/llvm-project/issues/106257.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 4c68a95b9c26..cb104c4ed2d0 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -8648,6 +8648,11 @@ addUsersInExitBlock(VPlan &Plan,\n \n // Introduce VPUsers modeling the exit values.\n for (const auto &[ExitPhi, V] : ExitingValuesToFix) {\n+ // Pass live-in values used by exit phis directly through to the live-out.\n+ if (V->isLiveIn()) {\n+ Plan.addLiveOut(ExitPhi, V);\n+ continue;\n+ }\n VPValue *Ext = B.createNaryOp(\n VPInstruction::ExtractFromEnd,\n {V, Plan.getOrAddLiveIn(ConstantInt::get(\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/extract-from-end-vector-constant.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -S %s"], "tests": [{"test_name": "exit_value_vector_live_in", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine <2 x i64> @exit_value_vector_live_in(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next.1, %loop ]\n %gep = getelementptr i16, ptr %dst, i64 %iv\n store i16 0, ptr %gep, align 2\n %iv.next.1 = add i64 %iv, 1\n %ec = icmp ult i64 %iv, 1000\n br i1 %ec, label %loop, label %exit\n\nexit: ; preds = %loop\n %res = phi <2 x i64> [ zeroinitializer, %loop ]\n ret <2 x i64> %res\n}\n"}, {"test_name": "exit_value_scalar_live_in", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine i64 @exit_value_scalar_live_in(ptr %dst, i64 %in) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next.1, %loop ]\n %gep = getelementptr i16, ptr %dst, i64 %iv\n store i16 0, ptr %gep, align 2\n %iv.next.1 = add i64 %iv, 1\n %ec = icmp ult i64 %iv, 1000\n br i1 %ec, label %loop, label %exit\n\nexit: ; preds = %loop\n %res = phi i64 [ %in, %loop ]\n ret i64 %res\n}\n"}]}], "issue": {"title": "[LoopVectorize] Assertion `(GeneratedValue->getType()->isVectorTy() == !GeneratesPerFirstLaneOnly || State.VF.isScalar()) && \"scalar value but not only first lane defined\"' failed.", "body": "Testcase:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine fastcc void @g() #0 {\r\nfor.cond1.preheader.us.preheader:\r\n br label %for.cond.cleanup20.us.us.1\r\n\r\nfor.cond.cleanup20.us.us.1: ; preds = %for.cond.cleanup20.us.us.1, %for.cond1.preheader.us.preheader\r\n %indvars.iv.1 = phi i64 [ 0, %for.cond1.preheader.us.preheader ], [ %indvars.iv.next.1, %for.cond.cleanup20.us.us.1 ]\r\n %arrayidx.us.us.1 = getelementptr [0 x i16], ptr null, i64 0, i64 %indvars.iv.1\r\n store i16 0, ptr %arrayidx.us.us.1, align 2\r\n %indvars.iv.next.1 = add i64 %indvars.iv.1, 1\r\n %cmp12.us.us.1 = icmp ult i64 %indvars.iv.1, 1\r\n br i1 %cmp12.us.us.1, label %for.cond.cleanup20.us.us.1, label %for.cond.cleanup14.us.us.1\r\n\r\nfor.cond.cleanup14.us.us.1: ; preds = %for.cond.cleanup20.us.us.1\r\n %.lcssa.1 = phi <2 x i64> [ zeroinitializer, %for.cond.cleanup20.us.us.1 ]\r\n ret void\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\nhttps://godbolt.org/z/aoe8vdc1f\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize -S reduced.ll\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp:732: virtual void llvm::VPInstruction::execute(llvm::VPTransformState&): Assertion `(GeneratedValue->getType()->isVectorTy() == !GeneratesPerFirstLaneOnly || State.VF.isScalar()) && \"scalar value but not only first lane defined\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize -S reduced.ll\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"g\"\r\n #0 0x0000644a7eb61500 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x300e500)\r\n #1 0x0000644a7eb5e91f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x300b91f)\r\n #2 0x0000644a7eb5ea75 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x0000790e6c842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x0000790e6c8969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x0000790e6c8969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x0000790e6c8969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x0000790e6c842476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x0000790e6c8287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x0000790e6c82871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x0000790e6c839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x0000644a7dc8e57a llvm::VPInstruction::execute(llvm::VPTransformState&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x213b57a)\r\n#12 0x0000644a7dc4f617 llvm::VPBasicBlock::executeRecipes(llvm::VPTransformState*, llvm::BasicBlock*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20fc617)\r\n#13 0x0000644a7dc4f910 llvm::VPIRBasicBlock::execute(llvm::VPTransformState*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20fc910)\r\n#14 0x0000644a7dc51c2d llvm::VPlan::execute(llvm::VPTransformState*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20fec2d)\r\n#15 0x0000644a7db37af9 llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fe4af9)\r\n#16 0x0000644a7db4afce llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ff7fce)\r\n#17 0x0000644a7db4d101 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffa101)\r\n#18 0x0000644a7db4d777 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffa777)\r\n#19 0x0000644a7c9b0a46 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5da46)\r\n#20 0x0000644a7e96c95f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1995f)\r\n#21 0x0000644a7c9b3bd6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe60bd6)\r\n#22 0x0000644a7e96b6db llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e186db)\r\n#23 0x0000644a7c9afe76 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5ce76)\r\n#24 0x0000644a7e96971d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1671d)\r\n#25 0x0000644a7c1d9e36 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x686e36)\r\n#26 0x0000644a7c1cbbc1 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x678bc1)\r\n#27 0x0000790e6c829d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#28 0x0000790e6c829e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#29 0x0000790e6c829e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#30 0x0000644a7c1c1f75 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x66ef75)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\n\r\nFound via fuzzer", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "85536", "issue_url": "https://github.com/llvm/llvm-project/issues/85536", "bug_type": "miscompilation", "base_commit": "f84980570d3f85bdf5c9432647c05bae04a735a0", "knowledge_cutoff": "2024-03-16T15:33:21Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "252d01952c087cf0d141f7f281cf60efeb98be41", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstructionCombining.cpp": [[1650, 1655]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstructionCombining.cpp": ["foldOperationIntoSelectOperand"]}}, "patch": "commit 252d01952c087cf0d141f7f281cf60efeb98be41\nAuthor: Yingwei Zheng \nDate: Sun Mar 17 14:15:27 2024 +0800\n\n [InstCombine] Drop UB-implying attrs/metadata after speculating an instruction (#85542)\n \n When speculating an instruction in `InstCombinerImpl::FoldOpIntoSelect`,\n the call may result in undefined behavior. This patch drops all\n UB-implying attrs/metadata to fix this.\n \n Fixes #85536.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\nindex edb046defbc1..5d3e41fb5f29 100644\n--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\n@@ -1650,6 +1650,7 @@ static Value *foldOperationIntoSelectOperand(Instruction &I, SelectInst *SI,\n Value *NewOp, InstCombiner &IC) {\n Instruction *Clone = I.clone();\n Clone->replaceUsesOfWith(SI, NewOp);\n+ Clone->dropUBImplyingAttrsAndMetadata();\n IC.InsertNewInstBefore(Clone, SI->getIterator());\n return Clone;\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/intrinsic-select.ll", "commands": ["opt -passes=instcombine -S < %s"], "tests": [{"test_name": "test_drop_noundef", "test_body": "define i8 @test_drop_noundef(i1 %cond, i8 %val) {\nentry:\n %sel = select i1 %cond, i8 -1, i8 %val\n %ret = call noundef i8 @llvm.smin.i8(i8 %sel, i8 0)\n ret i8 %ret\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i8 @llvm.smin.i8(i8, i8) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "pr85536", "test_body": "define i1 @pr85536(i32 %a) {\nentry:\n %cmp1 = icmp ugt i32 %a, 30\n %shl1 = shl nsw i32 -1, %a\n %zext = zext i32 %shl1 to i64\n %shl2 = shl i64 %zext, 48\n %shr = ashr exact i64 %shl2, 48\n %sel = select i1 %cmp1, i64 -1, i64 %shr\n %smin = call noundef i64 @llvm.smin.i64(i64 %sel, i64 0)\n %masked = and i64 %smin, 65535\n %ret = icmp eq i64 %masked, 0\n ret i1 %ret\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i64 @llvm.smin.i64(i64, i64) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "vec_to_scalar_select_vector", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.vector.reduce.add.v2i32(<2 x i32>) #0\n\ndefine i32 @vec_to_scalar_select_vector(<2 x i1> %b) {\n %s = select <2 x i1> %b, <2 x i32> , <2 x i32> \n %c = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %s)\n ret i32 %c\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[InstCombine] Miscompile at -O1", "body": "Reduced test case: https://godbolt.org/z/z1a3665hd\r\n```\r\n#include \r\n#include \r\nstatic uint16_t\r\n(safe_lshift_func_uint16_t_u_s)(uint16_t left, int right )\r\n{\r\n return\r\n ((((int)right) < 0) || (((int)right) >= 32) || (left > ((65535) >> ((int)right)))) ?\r\n ((left)) :\r\n (left << ((int)right));\r\n}\r\nstatic int16_t\r\n(safe_unary_minus_func_int16_t_s)(int16_t si )\r\n{\r\n return\r\n -si;\r\n}\r\nstatic int32_t\r\n(safe_lshift_func_int32_t_s_u)(int32_t left, unsigned int right )\r\n{\r\n return\r\n ((left < 0) || (((unsigned int)right) >= 32) || (left > ((2147483647) >> ((unsigned int)right)))) ?\r\n ((left)) :\r\n (left << ((unsigned int)right));\r\n}\r\nlong smin(long d, long p) { return d < p ? d : p; }\r\nstruct e { uint32_t f; } static g[] = {1, 36};\r\nint64_t h, i;\r\nuint8_t j(uint64_t m) {\r\n if (safe_lshift_func_uint16_t_u_s(\r\n smin(m, safe_unary_minus_func_int16_t_s(\r\n safe_lshift_func_int32_t_s_u(1, g[1].f))),\r\n 3))\r\n h = 0;\r\n return m;\r\n}\r\nint8_t k() {\r\n j(0);\r\n struct e *l[] = {&g[1], &g[1], &g[1], &g[1], &g[1],\r\n &g[1], &g[1], &g[1], &g[1]};\r\n return i;\r\n}\r\nint main() {\r\n printf(\"%d\\n\", k());\r\n return 0;\r\n}\r\n```\r\n\r\n```\r\n> gcc -O0 -fsanitize=address,undefined -w test.c && ./a.out\r\n0\r\n> clang -O3 -w test.c && ./a.out\r\n; no output\r\n```\r\n\r\nReduced LLVM IR: https://alive2.llvm.org/ce/z/hWuhFG\r\n```\r\n%struct.e = type { i32 }\r\n\r\n@g = internal global [2 x %struct.e] [%struct.e { i32 1 }, %struct.e { i32 36 }], align 4\r\n@h = dso_local local_unnamed_addr global i64 0, align 8\r\n@i = dso_local local_unnamed_addr global i64 0, align 8\r\n\r\ndefine i8 @src() {\r\nentry:\r\n %0 = load i32, ptr getelementptr inbounds ([2 x %struct.e], ptr @g, i64 0, i64 1), align 4\r\n %or.cond.i.i = icmp ult i32 %0, 31\r\n %shl.i.neg.i = shl nsw i32 -1, %0\r\n %.neg.i = zext i32 %shl.i.neg.i to i64\r\n %sext.i = shl i64 %.neg.i, 48\r\n %1 = ashr exact i64 %sext.i, 48\r\n %2 = call i64 @llvm.smin.i64(i64 %1, i64 0)\r\n %3 = and i64 %2, 65535\r\n %tobool.not9.i1 = icmp eq i64 %3, 0\r\n %tobool.not9.i = and i1 %or.cond.i.i, %tobool.not9.i1\r\n br i1 %tobool.not9.i, label %j.exit, label %if.then.i\r\n\r\nif.then.i:\r\n store i64 0, ptr @h, align 8\r\n br label %j.exit\r\n\r\nj.exit:\r\n %4 = load i64, ptr @i, align 8\r\n %conv = trunc i64 %4 to i8\r\n ret i8 %conv\r\n}\r\n\r\n==>\r\n\r\ndefine i8 @tgt() {\r\n ret i8 poison\r\n}\r\n```\r\n\r\nLLVM version: 74d1a40915834cbf0629f8d34a7265734d4d9073\r\n\r\ncc @nikic ", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "dtcxzyw", "body": "Reduced IR: https://alive2.llvm.org/ce/z/Gv9zkm\r\n```\r\n@g = internal global [1 x i32] [i32 1], align 4\r\n\r\ndefine i32 @k() {\r\nentry:\r\n %ret = load i32, ptr getelementptr inbounds ([1 x i32], ptr @g, i64 0), align 4\r\n ret i32 %ret\r\n}\r\n\r\n==>\r\ndefine i32 @k() {\r\nentry:\r\n ret i32 1\r\n}\r\n```\r\n"}, {"author": "nikic", "body": "@dtcxzyw That last transform is correct for GlobalOpt, alive2 does not support IPO transforms."}, {"author": "dtcxzyw", "body": "> @dtcxzyw That last transform is correct for GlobalOpt, alive2 does not support IPO transforms.\r\n\r\nOh, sorry about my mistake. I will provide a godbolt link.\r\n\r\n\r\n"}, {"author": "dtcxzyw", "body": "The root cause: https://alive2.llvm.org/ce/z/ThJs3Y\r\n```\r\ndefine i1 @src(i32 %a) {\r\nentry:\r\n %or.cond.i.i = icmp ugt i32 %a, 30\r\n %shl.i.neg.i = shl nsw i32 -1, %a\r\n %.neg.i = zext i32 %shl.i.neg.i to i64\r\n %sext.i = shl i64 %.neg.i, 48\r\n %1 = ashr exact i64 %sext.i, 48\r\n %conv2.i = select i1 %or.cond.i.i, i64 -1, i64 %1\r\n %cond.i8.i = call noundef i64 @llvm.smin.i64(i64 0, i64 %conv2.i)\r\n %2 = and i64 %cond.i8.i, 65535\r\n %tobool.not9.i = icmp eq i64 %2, 0\r\n ret i1 %tobool.not9.i\r\n}\r\n\r\n==> \r\n\r\ndefine i1 @tgt(i32 %a) {\r\nentry:\r\n %or.cond.i.i = icmp ult i32 %a, 31\r\n %shl.i.neg.i = shl nsw i32 -1, %a\r\n %.neg.i = zext i32 %shl.i.neg.i to i64\r\n %sext.i = shl i64 %.neg.i, 48\r\n %0 = ashr exact i64 %sext.i, 48\r\n %1 = call noundef i64 @llvm.smin.i64(i64 %0, i64 0)\r\n %2 = and i64 %1, 65535\r\n %tobool.not9.i1 = icmp eq i64 %2, 0\r\n %tobool.not9.i = and i1 %or.cond.i.i, %tobool.not9.i1\r\n ret i1 %tobool.not9.i\r\n}\r\n```"}, {"author": "dtcxzyw", "body": "We forget to drop the `noundef` attribute for `smin` :(\r\n"}, {"author": "dtcxzyw", "body": "Minimal reproducer: https://alive2.llvm.org/ce/z/aBubZ4\r\n```\r\ndefine i8 @src(i1 %cond, i8 %val) {\r\nentry:\r\n %sel = select i1 %cond, i8 -1, i8 %val\r\n %ret = call noundef i8 @llvm.smin.i8(i8 %sel, i8 0)\r\n ret i8 %ret\r\n}\r\n\r\n==>\r\n\r\ndefine i8 @tgt(i1 %cond, i8 %val) {\r\n %min = call noundef i8 @llvm.smin.i8(i8 %val, i8 0)\r\n %ret = select i1 %cond, i8 -1, i8 %min\r\n ret i8 %ret\r\n}\r\n```\r\n\r\n"}, {"author": "nikic", "body": "@dtcxzyw We're probably missing a call to dropUBImplyingAttrsAndMetadata in FoldOpIntoSelect."}]}, "verified": true} {"bug_id": "124526", "issue_url": "https://github.com/llvm/llvm-project/issues/124526", "bug_type": "crash", "base_commit": "e14962a39cc6476bebba65e5639b74b0318c7fc3", "knowledge_cutoff": "2025-01-27T10:24:37Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "212f344b84b400c0a9dedfa3c1ec6af9d9d30223", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstructionCombining.cpp": [[732, 738]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstructionCombining.cpp": ["tryFactorization"]}}, "patch": "commit 212f344b84b400c0a9dedfa3c1ec6af9d9d30223\nAuthor: Nikita Popov \nDate: Mon Jan 27 16:24:15 2025 +0100\n\n [InstCombine] Handle constant expression result in tryFactorization()\n \n If IRBuilder folds the result to a constant expression, don't try\n to set nowrap flags on it.\n \n Fixes https://github.com/llvm/llvm-project/issues/124526.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\nindex cad17c511b6d..a64c188575e6 100644\n--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\n@@ -732,7 +732,7 @@ static Value *tryFactorization(BinaryOperator &I, const SimplifyQuery &SQ,\n RetVal->takeName(&I);\n \n // Try to add no-overflow flags to the final value.\n- if (isa(RetVal)) {\n+ if (isa(RetVal)) {\n bool HasNSW = false;\n bool HasNUW = false;\n if (isa(&I)) {\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/add2.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "mul_add_to_mul_9", "test_body": "define i16 @mul_add_to_mul_9(i16 %a) {\n %mul1 = mul nsw i16 %a, 16384\n %mul2 = mul nsw i16 %a, 16384\n %add = add nsw i16 %mul1, %mul2\n ret i16 %add\n}\n"}, {"test_name": "shl_add_to_shl_constexpr", "test_body": "@g = external global i8\n\ndefine i32 @shl_add_to_shl_constexpr() {\n %shl = shl i32 ptrtoint (ptr @g to i32), 1\n %add = add i32 %shl, %shl\n ret i32 %add\n}\n"}]}], "issue": {"title": "[clang] Crash at -O2: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.", "body": "This code crashes at `-O2`:\n\n```c\nint a, b;\nvoid c();\nint d(int e) {\n if (e)\n return 3;\n return 2;\n}\nint f(int e) {\n int i = d(e) + e;\n return i;\n}\nint j(int e) {\n static char g;\n char *h = &g;\n a = h - \"\";\n b = a << e;\n return f(b - 7) + 303 + b;\n}\nvoid k() {\n int l = j(4);\n c(l);\n}\n```\n\nCompiler Explorer: https://godbolt.org/z/4h9PrKG8z\n\nCrash:\n\n```console\nclang: /root/llvm-project/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = llvm::Instruction; From = llvm::Value]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\n```\n\nBacktrace:\n\n```console\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/clang -gdwarf-4 -g -o /app/output.s -mllvm --x86-asm-syntax=intel -fno-verbose-asm -S --gcc-toolchain=/opt/compiler-explorer/gcc-snapshot -fcolor-diagnostics -fno-crash-diagnostics -O2 -Wextra -Wall \n1.\t parser at end of file\n2.\tOptimizer\n3.\tRunning pass \"require,function(invalidate),require,cgscc(devirt<4>(inline,function-attrs,openmp-opt-cgscc,function(sroa,early-cse,speculative-execution,jump-threading,correlated-propagation,simplifycfg,instcombine,aggressive-instcombine,libcalls-shrinkwrap,tailcallelim,simplifycfg,reassociate,constraint-elimination,loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-unswitch),simplifycfg,instcombine,loop(loop-idiom,indvars,extra-simple-loop-unswitch-passes,loop-deletion,loop-unroll-full),sroa,vector-combine,mldst-motion,gvn<>,sccp,bdce,instcombine,jump-threading,correlated-propagation,adce,memcpyopt,dse,move-auto-init,loop-mssa(licm),coro-elide,simplifycfg,instcombine),function-attrs,function(require),coro-split,coro-annotation-elide)),function(invalidate),cgscc(devirt<4>())\" on module \"\"\n4.\tRunning pass \"cgscc(devirt<4>(inline,function-attrs,openmp-opt-cgscc,function(sroa,early-cse,speculative-execution,jump-threading,correlated-propagation,simplifycfg,instcombine,aggressive-instcombine,libcalls-shrinkwrap,tailcallelim,simplifycfg,reassociate,constraint-elimination,loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-unswitch),simplifycfg,instcombine,loop(loop-idiom,indvars,extra-simple-loop-unswitch-passes,loop-deletion,loop-unroll-full),sroa,vector-combine,mldst-motion,gvn<>,sccp,bdce,instcombine,jump-threading,correlated-propagation,adce,memcpyopt,dse,move-auto-init,loop-mssa(licm),coro-elide,simplifycfg,instcombine),function-attrs,function(require),coro-split,coro-annotation-elide))\" on module \"\"\n5.\tRunning pass \"instcombine\" on function \"k\"\n #0 0x0000000003def1b8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3def1b8)\n #1 0x0000000003dece74 llvm::sys::CleanupOnSignal(unsigned long) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3dece74)\n #2 0x0000000003d39368 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\n #3 0x0000789749042520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #4 0x00007897490969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #5 0x0000789749042476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #6 0x00007897490287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #7 0x000078974902871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #8 0x0000789749039e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #9 0x0000000000d8d74a (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xd8d74a)\n#10 0x0000000003935620 tryFactorization(llvm::BinaryOperator&, llvm::SimplifyQuery const&, llvm::IRBuilder&, llvm::Instruction::BinaryOps, llvm::Value*, llvm::Value*, llvm::Value*, llvm::Value*) InstructionCombining.cpp:0:0\n#11 0x0000000003935ce2 llvm::InstCombinerImpl::tryFactorizationFolds(llvm::BinaryOperator&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3935ce2)\n#12 0x0000000003935d67 llvm::InstCombinerImpl::foldUsingDistributiveLaws(llvm::BinaryOperator&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3935d67)\n#13 0x0000000003972ab2 llvm::InstCombinerImpl::visitAdd(llvm::BinaryOperator&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3972ab2)\n#14 0x00000000039586c1 llvm::InstCombinerImpl::run() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x39586c1)\n#15 0x000000000395a066 combineInstructionsOverFunction(llvm::Function&, llvm::InstructionWorklist&, llvm::AAResults*, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::BranchProbabilityInfo*, llvm::ProfileSummaryInfo*, llvm::InstCombineOptions const&) (.isra.0) InstructionCombining.cpp:0:0\n#16 0x000000000395b52c llvm::InstCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x395b52c)\n#17 0x0000000004094a6e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4094a6e)\n#18 0x0000000003796c50 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3796c50)\n#19 0x000000000114aa4e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x114aa4e)\n#20 0x0000000002c71112 llvm::CGSCCToFunctionPassAdaptor::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2c71112)\n#21 0x000000000114aede llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x114aede)\n#22 0x0000000002c68502 llvm::PassManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2c68502)\n#23 0x000000000548dd6e llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>, llvm::AnalysisManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x548dd6e)\n#24 0x0000000002c6dbd6 llvm::DevirtSCCRepeatedPass::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2c6dbd6)\n#25 0x000000000548dd8e llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x548dd8e)\n#26 0x0000000002c6ba86 llvm::ModuleToPostOrderCGSCCPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2c6ba86)\n#27 0x000000000548dd4e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x548dd4e)\n#28 0x0000000003794f70 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3794f70)\n#29 0x00000000056e3783 llvm::ModuleInlinerWrapperPass::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x56e3783)\n#30 0x000000000532b69e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x532b69e)\n#31 0x0000000003794f70 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3794f70)\n#32 0x00000000040a7734 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\n#33 0x00000000040ab13e clang::emitBackendOutput(clang::CompilerInstance&, clang::CodeGenOptions&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x40ab13e)\n#34 0x000000000477edb0 clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x477edb0)\n#35 0x00000000067c259c clang::ParseAST(clang::Sema&, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x67c259c)\n#36 0x000000000477f1b8 clang::CodeGenAction::ExecuteAction() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x477f1b8)\n#37 0x0000000004a44575 clang::FrontendAction::Execute() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4a44575)\n#38 0x00000000049c72ae clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x49c72ae)\n#39 0x0000000004b31ede clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4b31ede)\n#40 0x0000000000cfcd0f cc1_main(llvm::ArrayRef, char const*, void*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xcfcd0f)\n#41 0x0000000000cf47aa ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\n#42 0x00000000047c5ae9 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\n#43 0x0000000003d39814 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3d39814)\n#44 0x00000000047c60df clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\n#45 0x0000000004788f4d clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4788f4d)\n#46 0x0000000004789fce clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4789fce)\n#47 0x0000000004791b05 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4791b05)\n#48 0x0000000000cf9b23 clang_main(int, char**, llvm::ToolContext const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xcf9b23)\n#49 0x0000000000bc2a84 main (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xbc2a84)\n#50 0x0000789749029d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#51 0x0000789749029e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#52 0x0000000000cf4255 _start (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xcf4255)\n```", "author": "cardigan1008", "labels": ["crash", "llvm:instcombine", "regression:19"], "comments": [{"author": "shafik", "body": "Looks like a clang-19 regression: https://godbolt.org/z/EvWdTGY6E"}, {"author": "nikic", "body": "```llvm\n; RUN: opt -passes=instcombine < %s\n@g = external global i8\n\ndefine i32 @test() {\n %shl = shl i32 ptrtoint (ptr @g to i32), 1\n %add1 = add i32 %shl, -1\n %add2 = add i32 %shl, 2\n %add3 = add i32 %add1, %add2\n ret i32 %add3 \n} \n```"}, {"author": "nikic", "body": "Further reduced:\n```llvm\n@g = external global i8\n\ndefine i32 @test() {\n %shl = shl i32 ptrtoint (ptr @g to i32), 1\n %add = add i32 %shl, %shl\n ret i32 %add \n} \n```"}]}, "verified": true} {"bug_id": "119063", "issue_url": "https://github.com/llvm/llvm-project/issues/119063", "bug_type": "crash", "base_commit": "118f7b95c8546826f222c13daea47cde4aa97bc9", "knowledge_cutoff": "2024-12-07T08:46:55Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "f4f6566e44566f3d8cf9517767d457227125ca93", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": [[3622, 3628]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": ["InstCombinerImpl::foldICmpBinOpEqualityWithConstant"]}}, "patch": "commit f4f6566e44566f3d8cf9517767d457227125ca93\nAuthor: Yingwei Zheng \nDate: Sun Dec 8 13:21:34 2024 +0800\n\n [InstCombine] Fix type mismatch in `foldICmpBinOpEqualityWithConstant` (#119068)\n \n Closes https://github.com/llvm/llvm-project/issues/119063.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\nindex b694fde30927..54053c4c9e28 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n@@ -3622,7 +3622,8 @@ Instruction *InstCombinerImpl::foldICmpBinOpEqualityWithConstant(\n m_OneUse(m_c_Or(m_CombineAnd(m_Value(Sel),\n m_Select(m_Value(Cond), m_Value(TV),\n m_Value(FV))),\n- m_Value(Other))))) {\n+ m_Value(Other)))) &&\n+ Cond->getType() == Cmp.getType()) {\n const SimplifyQuery Q = SQ.getWithInstruction(&Cmp);\n // Easy case is if eq/ne matches whether 0 is trueval/falseval.\n if (Pred == ICmpInst::ICMP_EQ\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/icmp-or-of-select-with-zero.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "pr119063", "test_body": "define <4 x i1> @pr119063(<4 x i32> %x, i1 %cond) {\nentry:\n %sel = select i1 %cond, <4 x i32> splat (i32 1), <4 x i32> zeroinitializer\n %or = or <4 x i32> %sel, %x\n %cmp = icmp ne <4 x i32> %or, zeroinitializer\n ret <4 x i1> %cmp\n}\n"}, {"test_name": "src_tv_ne_invert", "test_body": "declare void @use.i8(i8)\n\ndeclare void @use.i1(i1)\n\ndefine i1 @src_tv_ne_invert(i1 %c1, i8 %a, i8 %b, i8 %x, i8 %yy) {\n %not_c0 = icmp ugt i8 %a, %b\n call void @use.i1(i1 %not_c0)\n %c0 = xor i1 %not_c0, true\n %y = add nuw i8 %yy, 1\n %sel = select i1 %c0, i8 0, i8 %y\n %cc = or i1 %c0, %c1\n %sel_other = select i1 %cc, i8 %y, i8 %b\n %selx = or i8 %sel, %x\n %r = icmp ne i8 %selx, 0\n call void @use.i8(i8 %sel)\n call void @use.i8(i8 %sel_other)\n ret i1 %r\n}\n"}]}], "issue": {"title": "[InstCombine] Assertion `S1->getType() == S2->getType() && \"Cannot create binary operator with two operands of differing type!\"' failed.", "body": "This code crashes at O2/3:\n```c\nint a, b;\nstatic int c;\nstatic int *d = &c;\nshort e;\nchar g;\nunsigned short h;\nint *i;\nint j(int k) {\n if (k & a)\n return 1;\n return 0;\n}\nint l(int k) {\n int f = j(k);\n return f;\n}\nvoid n() { int *m[] = {0, &d}; }\nvoid o() {\n for (; g; g++) {\n int *p = &b;\n short *q = &e;\n int r = l(*d + 5);\n *p = 0;\n *q = *i || (unsigned short)(r + 22158 + 36566 + *d - 1641080165) + h;\n }\n}\n```\n\nCompiler Explorer: https://godbolt.org/z/n7GPEKsxb\n\nBisected to https://github.com/llvm/llvm-project/commit/4eb98384099cbd1d901d8ef6a210af95d9c36adf, which was committed by @fhahn \n\nCrash: \n```\nclang: /root/llvm-project/llvm/lib/IR/Instructions.cpp:2642: static llvm::BinaryOperator* llvm::BinaryOperator::Create(llvm::Instruction::BinaryOps, llvm::Value*, llvm::Value*, const llvm::Twine&, llvm::InsertPosition): Assertion `S1->getType() == S2->getType() && \"Cannot create binary operator with two operands of differing type!\"' failed.\n```\n\nBacktrace:\n```\nStack dump:\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/clang -gdwarf-4 -g -o /app/output.s -mllvm --x86-asm-syntax=intel -fno-verbose-asm -S --gcc-toolchain=/opt/compiler-explorer/gcc-snapshot -fcolor-diagnostics -fno-crash-diagnostics -O2 -Wall -Wextra \n1.\t parser at end of file\n2.\tOptimizer\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"\"\n4.\tRunning pass \"instcombine\" on function \"o\"\n #0 0x0000000003c05148 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3c05148)\n #1 0x0000000003c02e54 llvm::sys::CleanupOnSignal(unsigned long) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3c02e54)\n #2 0x0000000003b503b8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\n #3 0x00007bfb65442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #4 0x00007bfb654969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #5 0x00007bfb65442476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #6 0x00007bfb654287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #7 0x00007bfb6542871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #8 0x00007bfb65439e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #9 0x000000000351a1f3 (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x351a1f3)\n#10 0x0000000003523d7c (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3523d7c)\n#11 0x00000000037edc06 llvm::InstCombinerImpl::foldICmpBinOpEqualityWithConstant(llvm::ICmpInst&, llvm::BinaryOperator*, llvm::APInt const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x37edc06)\n#12 0x0000000003810ac9 llvm::InstCombinerImpl::foldICmpInstWithConstant(llvm::ICmpInst&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3810ac9)\n#13 0x0000000003819aad llvm::InstCombinerImpl::visitICmpInst(llvm::ICmpInst&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3819aad)\n#14 0x0000000003775251 llvm::InstCombinerImpl::run() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3775251)\n#15 0x0000000003776bc6 combineInstructionsOverFunction(llvm::Function&, llvm::InstructionWorklist&, llvm::AAResults*, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::BranchProbabilityInfo*, llvm::ProfileSummaryInfo*, llvm::InstCombineOptions const&) (.isra.0) InstructionCombining.cpp:0:0\n#16 0x000000000377805c llvm::InstCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x377805c)\n#17 0x0000000003ea39ae llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3ea39ae)\n#18 0x00000000035b5920 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35b5920)\n#19 0x00000000010f19ce llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x10f19ce)\n#20 0x00000000035b423b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35b423b)\n#21 0x00000000010f105e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x10f105e)\n#22 0x00000000035b3c40 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35b3c40)\n#23 0x0000000003eb5912 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\n#24 0x0000000003eb918d clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3eb918d)\n#25 0x00000000045875de clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x45875de)\n#26 0x000000000654152c clang::ParseAST(clang::Sema&, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x654152c)\n#27 0x00000000045879b8 clang::CodeGenAction::ExecuteAction() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x45879b8)\n#28 0x0000000004842b39 clang::FrontendAction::Execute() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4842b39)\n#29 0x00000000047c27ce clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x47c27ce)\n#30 0x000000000492db9e clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x492db9e)\n#31 0x0000000000cb21bf cc1_main(llvm::ArrayRef, char const*, void*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xcb21bf)\n#32 0x0000000000ca9d5a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\n#33 0x00000000045cad09 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\n#34 0x0000000003b50864 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3b50864)\n#35 0x00000000045cb2ff clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\n#36 0x000000000459149d clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x459149d)\n#37 0x000000000459258d clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x459258d)\n#38 0x0000000004599945 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4599945)\n#39 0x0000000000caf003 clang_main(int, char**, llvm::ToolContext const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xcaf003)\n#40 0x0000000000b82f64 main (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xb82f64)\n#41 0x00007bfb65429d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#42 0x00007bfb65429e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#43 0x0000000000ca9805 _start (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xca9805)\n```", "author": "cardigan1008", "labels": ["llvm:instcombine", "crash-on-valid"], "comments": [{"author": "dtcxzyw", "body": "Reduced reproducer: https://godbolt.org/z/najxdPz5h\n"}]}, "verified": true} {"bug_id": "112350", "issue_url": "https://github.com/llvm/llvm-project/issues/112350", "bug_type": "miscompilation", "base_commit": "35e937b4de1890186347a382f7727ba86441dbda", "knowledge_cutoff": "2024-10-15T12:03:18Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "aad3a1630e385a6834f92a5f1d6045451ba21e4e", "components": ["ValueTracking"], "bug_location_lineno": {"llvm/lib/Analysis/ValueTracking.cpp": [[8522, 8527], [8530, 8535]]}, "bug_location_funcname": {"llvm/lib/Analysis/ValueTracking.cpp": ["llvm::isKnownInversion"]}}, "patch": "commit aad3a1630e385a6834f92a5f1d6045451ba21e4e\nAuthor: Yingwei Zheng \nDate: Thu Oct 17 00:27:21 2024 +0800\n\n [ValueTracking] Respect `samesign` flag in `isKnownInversion` (#112390)\n \n In https://github.com/llvm/llvm-project/pull/93591 we introduced\n `isKnownInversion` and assumes `X` is poison implies `Y` is poison\n because they share common operands. But after introducing `samesign`\n this assumption no longer hold if `X` is an icmp has `samesign` flag.\n \n Alive2 link: https://alive2.llvm.org/ce/z/rj3EwQ (Please run it locally\n with this patch and https://github.com/AliveToolkit/alive2/pull/1098).\n \n This approach is the most conservative way in my mind to address this\n problem. If `X` has `samesign` flag, it will check if `Y` also has this\n flag and make sure constant RHS operands have the same sign.\n \n Fixes https://github.com/llvm/llvm-project/issues/112350.\n\ndiff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp\nindex eb8d17044a17..e9ed8b3c862b 100644\n--- a/llvm/lib/Analysis/ValueTracking.cpp\n+++ b/llvm/lib/Analysis/ValueTracking.cpp\n@@ -8522,6 +8522,10 @@ bool llvm::isKnownInversion(const Value *X, const Value *Y) {\n !match(Y, m_c_ICmp(Pred2, m_Specific(A), m_Value(C))))\n return false;\n \n+ // They must both have samesign flag or not.\n+ if (cast(X)->hasSameSign() != cast(Y)->hasSameSign())\n+ return false;\n+\n if (B == C)\n return Pred1 == ICmpInst::getInversePredicate(Pred2);\n \n@@ -8530,6 +8534,11 @@ bool llvm::isKnownInversion(const Value *X, const Value *Y) {\n if (!match(B, m_APInt(RHSC1)) || !match(C, m_APInt(RHSC2)))\n return false;\n \n+ // Sign bits of two RHSCs should match.\n+ if (cast(X)->hasSameSign() &&\n+ RHSC1->isNonNegative() != RHSC2->isNonNegative())\n+ return false;\n+\n const auto CR1 = ConstantRange::makeExactICmpRegion(Pred1, *RHSC1);\n const auto CR2 = ConstantRange::makeExactICmpRegion(Pred2, *RHSC2);\n \n", "tests": [{"file": "llvm/test/Transforms/InstCombine/select-cmp.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "test_select_inverse_nonconst4", "test_body": "define i1 @test_select_inverse_nonconst4(i64 %x, i64 %y, i64 %z, i1 %cond) {\n %cmp1 = icmp ult i64 %x, %y\n %cmp2 = icmp uge i64 %z, %y\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\n ret i1 %sel\n}\n"}, {"test_name": "test_select_inverse_samesign_both", "test_body": "define i1 @test_select_inverse_samesign_both(i64 %x, i64 %y, i1 %cond) {\n %cmp1 = icmp samesign ult i64 %x, %y\n %cmp2 = icmp samesign uge i64 %x, %y\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\n ret i1 %sel\n}\n"}, {"test_name": "test_select_inverse_samesign_false_arm_rhsc_same_sign", "test_body": "define i1 @test_select_inverse_samesign_false_arm_rhsc_same_sign(i64 %x, i64 %y, i1 %cond) {\n %cmp1 = icmp ult i64 %x, 11\n %cmp2 = icmp samesign ugt i64 %x, 10\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\n ret i1 %sel\n}\n"}, {"test_name": "test_select_inverse_samesign_true_arm", "test_body": "define i1 @test_select_inverse_samesign_true_arm(i64 %x, i64 %y, i1 %cond) {\n %cmp1 = icmp samesign ult i64 %x, %y\n %cmp2 = icmp uge i64 %x, %y\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\n ret i1 %sel\n}\n"}, {"test_name": "test_select_inverse_samesign_true_arm_rhsc_same_sign", "test_body": "define i1 @test_select_inverse_samesign_true_arm_rhsc_same_sign(i64 %x, i64 %y, i1 %cond) {\n %cmp1 = icmp samesign ult i64 %x, 11\n %cmp2 = icmp ugt i64 %x, 10\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\n ret i1 %sel\n}\n"}, {"test_name": "test_select_inverse_samesign_both_rhsc_diff_sign", "test_body": "define i1 @test_select_inverse_samesign_both_rhsc_diff_sign(i64 %x, i64 %y, i1 %cond) {\n %cmp1 = icmp samesign slt i64 %x, 0\n %cmp2 = icmp samesign sgt i64 %x, -1\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\n ret i1 %sel\n}\n"}, {"test_name": "test_select_inverse_samesign_both_rhsc_same_sign", "test_body": "define i1 @test_select_inverse_samesign_both_rhsc_same_sign(i64 %x, i64 %y, i1 %cond) {\n %cmp1 = icmp samesign ult i64 %x, 11\n %cmp2 = icmp samesign ugt i64 %x, 10\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\n ret i1 %sel\n}\n"}, {"test_name": "test_select_inverse_samesign_false_arm", "test_body": "define i1 @test_select_inverse_samesign_false_arm(i64 %x, i64 %y, i1 %cond) {\n %cmp1 = icmp ult i64 %x, %y\n %cmp2 = icmp samesign uge i64 %x, %y\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\n ret i1 %sel\n}\n"}, {"test_name": "sel_icmp_two_cmp", "test_body": "define i1 @sel_icmp_two_cmp(i1 %c, i32 %a1, i32 %a2, i32 %a3, i32 %a4) {\n %v1 = call i8 @llvm.ucmp.i8.i32(i32 %a1, i32 %a2)\n %v2 = call i8 @llvm.scmp.i8.i32(i32 %a3, i32 %a4)\n %sel = select i1 %c, i8 %v1, i8 %v2\n %cmp = icmp sle i8 %sel, 0\n ret i1 %cmp\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i8 @llvm.scmp.i8.i32(i32, i32) #0\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i8 @llvm.ucmp.i8.i32(i32, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[InstCombine] `samesign` flag should be dropped", "body": "Reproducer:\r\n```\r\ndefine i1 @src(i64 %x, i64 %y, i1 %cond) {\r\n %cmp1 = icmp ult i64 %x, %y\r\n %cmp2 = icmp samesign uge i64 %x, %y\r\n %sel = select i1 %cond, i1 %cmp1, i1 %cmp2\r\n ret i1 %sel\r\n}\r\n\r\ndefine i1 @tgt(i64 %x, i64 %y, i1 %cond) {\r\n %cmp2 = icmp samesign uge i64 %x, %y\r\n %sel = xor i1 %cond, %cmp2\r\n ret i1 %sel\r\n}\r\n```\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:analysis"], "comments": [{"author": "antoniofrighetto", "body": "Don't we miss checking the flag in `hasPoisonGeneratingFlags`?"}, {"author": "dtcxzyw", "body": "> Don't we miss checking the flag in `hasPoisonGeneratingFlags`?\r\n\r\nYeah. I am preparing a patch.\r\n"}, {"author": "dtcxzyw", "body": "But adding support in `hasPoisonGeneratingFlags` doesn't address this issue :(\r\n"}, {"author": "antoniofrighetto", "body": "> But adding support in `hasPoisonGeneratingFlags` doesn't address this issue :(\r\n\r\nI assume we may need to `dropPoisonGeneratingFlags` in InstCombineCompares, where needed?"}]}, "verified": true} {"bug_id": "108620", "issue_url": "https://github.com/llvm/llvm-project/issues/108620", "bug_type": "crash", "base_commit": "03618ceabd003be94190b068db11b607fd1108ef", "knowledge_cutoff": "2024-09-13T18:15:02Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "1e3536ef3141d6429f3616af624b81b6d6ab2959", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[1283, 1293], [14222, 14235], [14373, 14378], [14426, 14436], [14530, 14537], [14757, 14762], [17660, 17666], [17956, 17961], [18024, 18031]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree", "tryToReduce"]}}, "patch": "commit 1e3536ef3141d6429f3616af624b81b6d6ab2959\nAuthor: Alexey Bataev \nDate: Fri Sep 13 15:40:22 2024 -0700\n\n [SLP]Fix PR108620: Need to check, if the reduced value was transformed\n \n Before trying to include the scalar into the list of\n ExternallyUsedValues, need to check, if it was transformed in previous\n iteration and use the transformed value, not the original one, to avoid\n compiler crash when building external uses.\n \n Fixes https://github.com/llvm/llvm-project/issues/108620\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 0afe02fc08ff..5f2bf082fb87 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -1283,11 +1283,8 @@ public:\n /// Vectorize the tree but with the list of externally used values \\p\n /// ExternallyUsedValues. Values in this MapVector can be replaced but the\n /// generated extractvalue instructions.\n- /// \\param ReplacedExternals containd list of replaced external values\n- /// {scalar, replace} after emitting extractelement for external uses.\n Value *\n vectorizeTree(const ExtraValueToDebugLocsMap &ExternallyUsedValues,\n- SmallVectorImpl> &ReplacedExternals,\n Instruction *ReductionRoot = nullptr);\n \n /// \\returns the cost incurred by unwanted spills and fills, caused by\n@@ -14222,14 +14219,12 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n \n Value *BoUpSLP::vectorizeTree() {\n ExtraValueToDebugLocsMap ExternallyUsedValues;\n- SmallVector> ReplacedExternals;\n- return vectorizeTree(ExternallyUsedValues, ReplacedExternals);\n+ return vectorizeTree(ExternallyUsedValues);\n }\n \n-Value *BoUpSLP::vectorizeTree(\n- const ExtraValueToDebugLocsMap &ExternallyUsedValues,\n- SmallVectorImpl> &ReplacedExternals,\n- Instruction *ReductionRoot) {\n+Value *\n+BoUpSLP::vectorizeTree(const ExtraValueToDebugLocsMap &ExternallyUsedValues,\n+ Instruction *ReductionRoot) {\n // All blocks must be scheduled before any instructions are inserted.\n for (auto &BSIter : BlocksSchedules) {\n scheduleBlock(BSIter.second.get());\n@@ -14373,6 +14368,7 @@ Value *BoUpSLP::vectorizeTree(\n SmallDenseSet UsedInserts;\n DenseMap, Value *> VectorCasts;\n SmallDenseSet ScalarsWithNullptrUser;\n+ SmallDenseSet IgnoredExtracts;\n // Extract all of the elements with the external uses.\n for (const auto &ExternalUse : ExternalUses) {\n Value *Scalar = ExternalUse.Scalar;\n@@ -14426,11 +14422,16 @@ Value *BoUpSLP::vectorizeTree(\n if (ReplaceInst) {\n // Leave the instruction as is, if it cheaper extracts and all\n // operands are scalar.\n- auto *CloneInst = Inst->clone();\n- CloneInst->insertBefore(Inst);\n- if (Inst->hasName())\n- CloneInst->takeName(Inst);\n- Ex = CloneInst;\n+ if (auto *EE = dyn_cast(Inst)) {\n+ IgnoredExtracts.insert(EE);\n+ Ex = EE;\n+ } else {\n+ auto *CloneInst = Inst->clone();\n+ CloneInst->insertBefore(Inst);\n+ if (Inst->hasName())\n+ CloneInst->takeName(Inst);\n+ Ex = CloneInst;\n+ }\n } else if (auto *ES = dyn_cast(Scalar);\n ES && isa(Vec)) {\n Value *V = ES->getVectorOperand();\n@@ -14530,8 +14531,12 @@ Value *BoUpSLP::vectorizeTree(\n }\n Value *NewInst = ExtractAndExtendIfNeeded(Vec);\n // Required to update internally referenced instructions.\n- Scalar->replaceAllUsesWith(NewInst);\n- ReplacedExternals.emplace_back(Scalar, NewInst);\n+ if (Scalar != NewInst) {\n+ assert((!isa(Scalar) ||\n+ !IgnoredExtracts.contains(cast(Scalar))) &&\n+ \"Extractelements should not be replaced.\");\n+ Scalar->replaceAllUsesWith(NewInst);\n+ }\n continue;\n }\n \n@@ -14757,6 +14762,9 @@ Value *BoUpSLP::vectorizeTree(\n if (Entry->getOpcode() == Instruction::GetElementPtr &&\n !isa(Scalar))\n continue;\n+ if (auto *EE = dyn_cast(Scalar);\n+ EE && IgnoredExtracts.contains(EE))\n+ continue;\n #ifndef NDEBUG\n Type *Ty = Scalar->getType();\n if (!Ty->isVoidTy()) {\n@@ -17660,7 +17668,6 @@ public:\n // because of the vectorization.\n DenseMap TrackedVals(ReducedVals.size() *\n ReducedVals.front().size());\n- SmallVector> ReplacedExternals;\n \n // The compare instruction of a min/max is the insertion point for new\n // instructions and may be replaced with a new compare instruction.\n@@ -17956,6 +17963,8 @@ public:\n if (Cnt >= Pos && Cnt < Pos + ReduxWidth)\n continue;\n Value *RdxVal = Candidates[Cnt];\n+ if (auto It = TrackedVals.find(RdxVal); It != TrackedVals.end())\n+ RdxVal = It->second;\n if (!Visited.insert(RdxVal).second)\n continue;\n // Check if the scalar was vectorized as part of the vectorization\n@@ -18024,8 +18033,8 @@ public:\n InsertPt = GetCmpForMinMaxReduction(RdxRootInst);\n \n // Vectorize a tree.\n- Value *VectorizedRoot = V.vectorizeTree(LocalExternallyUsedValues,\n- ReplacedExternals, InsertPt);\n+ Value *VectorizedRoot =\n+ V.vectorizeTree(LocalExternallyUsedValues, InsertPt);\n \n Builder.SetInsertPoint(InsertPt);\n \n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/reduced-val-extracted-and-externally-used.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define void @test(i32 %arg) {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb1, %bb\n %phi = phi i32 [ 0, %bb ], [ %add27, %bb1 ]\n %phi2 = phi i32 [ 0, %bb ], [ %add24, %bb1 ]\n %phi3 = phi i32 [ 0, %bb ], [ %xor26, %bb1 ]\n %add = add i32 %phi2, 0\n %add4 = add i32 %phi2, 0\n %xor = xor i32 %add, %add4\n %add5 = add i32 %phi, 0\n %add6 = add i32 %phi2, 0\n %add7 = add i32 %phi2, 0\n %xor8 = xor i32 %add6, %xor\n %xor9 = xor i32 %xor8, %add5\n %xor10 = xor i32 %xor9, %add7\n %add11 = add i32 %phi, 0\n %add12 = add i32 %phi2, 0\n %add13 = add i32 %phi2, 0\n %xor14 = xor i32 %add12, %xor10\n %xor15 = xor i32 %xor14, %add11\n %xor16 = xor i32 %xor15, %add13\n %add17 = add i32 %phi, 0\n %add18 = add i32 %phi2, 0\n %add19 = add i32 %phi2, 0\n %xor20 = xor i32 %add18, %xor16\n %xor21 = xor i32 %xor20, %add17\n %xor22 = xor i32 %xor21, %add19\n %add23 = add i32 %phi2, 0\n %add24 = add i32 %arg, 0\n %xor25 = xor i32 %add23, %xor22\n %xor26 = xor i32 %xor25, %add24\n %add27 = add i32 1, 0\n %icmp = icmp ult i32 %add27, 0\n br label %bb1\n}\n"}]}], "issue": {"title": "Assertion `(I->use_empty() || all_of(I->uses(), [&](Use &U) { return isDeleted( cast(U.getUser())); })) && \"trying to erase instruction with users.\"' failed.", "body": "To reproduce run the following test with -passes=slp-vectorizer\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat(i32 %arg) gc \"statepoint-example\" {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb1, %bb\r\n %phi = phi i32 [ 0, %bb ], [ %add27, %bb1 ]\r\n %phi2 = phi i32 [ 0, %bb ], [ %add24, %bb1 ]\r\n %phi3 = phi i32 [ 0, %bb ], [ %xor26, %bb1 ]\r\n %add = add i32 %phi2, 0\r\n %add4 = add i32 %phi2, 0\r\n %xor = xor i32 %add, %add4\r\n %add5 = add i32 %phi, 0\r\n %add6 = add i32 %phi2, 0\r\n %add7 = add i32 %phi2, 0\r\n %xor8 = xor i32 %add6, %xor\r\n %xor9 = xor i32 %xor8, %add5\r\n %xor10 = xor i32 %xor9, %add7\r\n %add11 = add i32 %phi, 0\r\n %add12 = add i32 %phi2, 0\r\n %add13 = add i32 %phi2, 0\r\n %xor14 = xor i32 %add12, %xor10\r\n %xor15 = xor i32 %xor14, %add11\r\n %xor16 = xor i32 %xor15, %add13\r\n %add17 = add i32 %phi, 0\r\n %add18 = add i32 %phi2, 0\r\n %add19 = add i32 %phi2, 0\r\n %xor20 = xor i32 %add18, %xor16\r\n %xor21 = xor i32 %xor20, %add17\r\n %xor22 = xor i32 %xor21, %add19\r\n %add23 = add i32 %phi2, 0\r\n %add24 = add i32 %arg, 0\r\n %xor25 = xor i32 %add23, %xor22\r\n %xor26 = xor i32 %xor25, %add24\r\n %add27 = add i32 1, 0\r\n %icmp = icmp ult i32 %add27, 0\r\n br label %bb1\r\n}\r\n```\r\nReproducer: https://godbolt.org/z/77hKre61P\r\n\r\nStack dump:\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:2663: void llvm::slpvectorizer::BoUpSLP::removeInstructionsAndOperands(llvm::ArrayRef) [with T = llvm::Instruction]: Assertion `(I->use_empty() || all_of(I->uses(), [&](Use &U) { return isDeleted( cast(U.getUser())); })) && \"trying to erase instruction with users.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"wombat\"\r\n #0 0x00000000050106b8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50106b8)\r\n #1 0x000000000500ddfc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x0000737fffa42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x0000737fffa969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x0000737fffa42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x0000737fffa287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x0000737fffa2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x0000737fffa39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x00000000040b91cf void llvm::slpvectorizer::BoUpSLP::removeInstructionsAndOperands(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40b91cf)\r\n #9 0x000000000410b80a llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::MapVector, llvm::DenseMap, llvm::detail::DenseMapPair>, llvm::SmallVector>, 0u>> const&, llvm::SmallVectorImpl>&, llvm::Instruction*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x410b80a)\r\n#10 0x0000000004115bed (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#11 0x0000000004118321 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4118321)\r\n#12 0x000000000411cb18 llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x411cb18)\r\n#13 0x00000000041212fa llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41212fa)\r\n#14 0x0000000004127719 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#15 0x00000000041283cb llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41283cb)\r\n#16 0x0000000002efda4e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2efda4e)\r\n#17 0x0000000004e12948 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e12948)\r\n#18 0x0000000000e06f8e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe06f8e)\r\n#19 0x0000000004e1136e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e1136e)\r\n#20 0x0000000000e0679e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe0679e)\r\n#21 0x0000000004e10da0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e10da0)\r\n#22 0x00000000008fb642 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fb642)\r\n#23 0x00000000008ee4cc optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ee4cc)\r\n#24 0x0000737fffa29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#25 0x0000737fffa29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#26 0x00000000008e5f9e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e5f9e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "91369", "issue_url": "https://github.com/llvm/llvm-project/issues/91369", "bug_type": "miscompilation", "base_commit": "9a28814f59e8f52cc63ae3d17023cee8348d9b53", "knowledge_cutoff": "2024-05-07T18:12:26Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "082c81ae4ab9db6bb0acd52098288223dd58501a", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[8841, 8848]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes"]}}, "patch": "commit 082c81ae4ab9db6bb0acd52098288223dd58501a\nAuthor: Florian Hahn \nDate: Tue May 7 21:31:40 2024 +0100\n\n [LV] Properly extend versioned constant strides.\n \n We only version unknown strides to 1. If the original type is i1, then\n the sign of the extension matters. Properly extend the stride value\n before replacing it.\n \n Fixes https://github.com/llvm/llvm-project/issues/91369.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 3be0102bea3e..261933966b74 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -8841,8 +8841,10 @@ LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {\n VPValue *StrideVPV = Plan->getLiveIn(U);\n if (!StrideVPV)\n continue;\n- VPValue *CI = Plan->getOrAddLiveIn(ConstantInt::get(\n- U->getType(), ScevStride->getAPInt().getSExtValue()));\n+ unsigned BW = U->getType()->getScalarSizeInBits();\n+ APInt C = isa(U) ? ScevStride->getAPInt().sext(BW)\n+ : ScevStride->getAPInt().zext(BW);\n+ VPValue *CI = Plan->getOrAddLiveIn(ConstantInt::get(U->getType(), C));\n StrideVPV->replaceAllUsesWith(CI);\n }\n }\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/version-stride-with-integer-casts.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=4 -S %s"], "tests": [{"test_name": "zext_of_i1_stride", "test_body": "target datalayout = \"p:8:8:8\"\n\n; Function Attrs: mustprogress\ndefine void @zext_of_i1_stride(i1 %g, ptr %dst) #0 {\nentry:\n %g.16 = zext i1 %g to i16\n %g.64 = zext i1 %g to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr inbounds i16, ptr %dst, i64 %iv\n store i16 %g.16, ptr %gep, align 2\n %iv.next = add nuw nsw i64 %iv, %g.64\n %cmp = icmp ult i64 %iv.next, 16\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { mustprogress }\n", "additional_args": "-src-unroll=32 -tgt-unroll=32"}]}], "issue": {"title": "[LoopVectorize] Sign-extension miscompile", "body": "C Testcase:\r\n```c\r\nlong a, d;\r\nshort b[20];\r\nshort c[20];\r\n_Bool e[20];\r\nint f(_Bool g) {\r\n for (long h = 0; h < 16; h += g)\r\n b[h] = (short) g;\r\n for (int i = 0; i < 12; i++)\r\n for (int j = d; j; j++)\r\n c[i] = (short) a < e[i];\r\n return (int) b[0];\r\n}\r\nint main() {\r\n __builtin_printf(\"%X\\n\", f(1));\r\n}\r\n```\r\nGodbolt: https://godbolt.org/z/3E8frs1Me\r\n\r\nWith loop-vectorize: `FFFFFFFF`\r\nWithout loop-vectorize: `1`\r\n\r\nReduced LLVM IR:\r\n```llvm ir\r\n; ModuleID = 'red.ll'\r\nsource_filename = \"red.c\"\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\n@b = global [20 x i16] zeroinitializer, align 2\r\n@.str = constant [4 x i8] c\"%X\\0A\\00\"\r\n\r\n; Function Attrs: nofree norecurse nosync nounwind memory(readwrite, argmem: none, inaccessiblemem: none) uwtable vscale_range(2,1024)\r\ndefine i32 @f(i1 %g) #0 {\r\nentry:\r\n %conv = zext i1 %g to i16\r\n %zext.g.i64 = zext i1 %g to i64\r\n br label %for.body\r\n\r\nfor.cond3.preheader: ; preds = %for.body\r\n %0 = load i64, ptr @b, align 2\r\n %zero = and i64 %0, 0\r\n %true = icmp eq i64 %zero, 0\r\n %2 = load i16, ptr @b, align 2\r\n %conv26 = sext i16 %2 to i32\r\n ret i32 %conv26\r\n\r\nfor.body: ; preds = %for.body, %entry\r\n %phi.0 = phi i64 [ 0, %entry ], [ %add, %for.body ]\r\n %idx.0 = getelementptr inbounds [20 x i16], ptr @b, i64 0, i64 %phi.0\r\n store i16 %conv, ptr %idx.0, align 2\r\n %add = add nuw nsw i64 %phi.0, %zext.g.i64\r\n %cmp = icmp ult i64 %add, 16\r\n br i1 %cmp, label %for.body, label %for.cond3.preheader, !llvm.loop !0\r\n}\r\n\r\n; Function Attrs: nofree nounwind uwtable vscale_range(2,1024)\r\ndefine noundef signext i32 @main() #1 {\r\nentry:\r\n %call = tail call signext i32 @f(i1 noundef zeroext true)\r\n %call1 = tail call signext i32 (ptr, ...) @printf(ptr noundef nonnull @.str, i32 noundef signext %call) #3\r\n ret i32 0\r\n}\r\n\r\n; Function Attrs: nofree nounwind\r\ndeclare noundef signext i32 @printf(ptr nocapture noundef readonly, ...) #2\r\n\r\nattributes #0 = { nofree norecurse nosync nounwind memory(readwrite, argmem: none, inaccessiblemem: none) uwtable vscale_range(2,1024) \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"generic-rv64\" \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zaamo,-experimental-zabha,-experimental-zalasr,-experimental-zalrsc,-experimental-zfbfmin,-experimental-zicfilp,-experimental-zicfiss,-experimental-ztso,-experimental-zvfbfmin,-experimental-zvfbfwma,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smepmp,-smstateen,-ssaia,-ssccptr,-sscofpmf,-sscounterenw,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-za128rs,-za64rs,-zacas,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-zmmul,-zvbb,-zvbc,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\nattributes #1 = { nofree nounwind uwtable vscale_range(2,1024) \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"generic-rv64\" \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zaamo,-experimental-zabha,-experimental-zalasr,-experimental-zalrsc,-experimental-zfbfmin,-experimental-zicfilp,-experimental-zicfiss,-experimental-ztso,-experimental-zvfbfmin,-experimental-zvfbfwma,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smepmp,-smstateen,-ssaia,-ssccptr,-sscofpmf,-sscounterenw,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-za128rs,-za64rs,-zacas,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-zmmul,-zvbb,-zvbc,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\nattributes #2 = { nofree nounwind \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"generic-rv64\" \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zaamo,-experimental-zabha,-experimental-zalasr,-experimental-zalrsc,-experimental-zfbfmin,-experimental-zicfilp,-experimental-zicfiss,-experimental-ztso,-experimental-zvfbfmin,-experimental-zvfbfwma,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smepmp,-smstateen,-ssaia,-ssccptr,-sscofpmf,-sscounterenw,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-za128rs,-za64rs,-zacas,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-zmmul,-zvbb,-zvbc,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\nattributes #3 = { nounwind }\r\n\r\n!0 = distinct !{!0, !1}\r\n!1 = !{!\"llvm.loop.mustprogress\"}\r\n```\r\nLoopVectorize pass.\r\n\r\nLooks like a sext/zext related issue.", "author": "patrick-rivos", "labels": ["miscompilation", "vectorizers"], "comments": []}, "verified": true} {"bug_id": "91178", "issue_url": "https://github.com/llvm/llvm-project/issues/91178", "bug_type": "miscompilation", "base_commit": "7098cd215b95286794d9e0c822e8323ad0509750", "knowledge_cutoff": "2024-05-06T09:44:48Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine", "llvm/test/Transforms/PGOProfile"], "hints": {"fix_commit": "d085b42cbbefe79a41113abcd2b1e1f2a203acef", "components": ["InstructionSimplify"], "bug_location_lineno": {"llvm/lib/Analysis/InstructionSimplify.cpp": [[4312, 4317]]}, "bug_location_funcname": {"llvm/lib/Analysis/InstructionSimplify.cpp": ["simplifyWithOpReplaced"]}}, "patch": "commit d085b42cbbefe79a41113abcd2b1e1f2a203acef\nAuthor: Yingwei Zheng \nDate: Wed May 8 10:04:09 2024 +0800\n\n [InstSimplify] Do not simplify freeze in `simplifyWithOpReplaced` (#91215)\n \n See the LangRef:\n > All uses of a value returned by the same ‘freeze’ instruction are\n guaranteed to always observe the same value, while different ‘freeze’\n instructions may yield different values.\n \n It is incorrect to replace freezes with the simplified value.\n \n Proof:\n https://alive2.llvm.org/ce/z/3Dn9Cd\n https://alive2.llvm.org/ce/z/Qyh5h6\n \n Fixes https://github.com/llvm/llvm-project/issues/91178\n\ndiff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp\nindex 4061dae83c10..37a7259a5cd0 100644\n--- a/llvm/lib/Analysis/InstructionSimplify.cpp\n+++ b/llvm/lib/Analysis/InstructionSimplify.cpp\n@@ -4312,6 +4312,10 @@ static Value *simplifyWithOpReplaced(Value *V, Value *Op, Value *RepOp,\n if (match(I, m_Intrinsic()))\n return nullptr;\n \n+ // Don't simplify freeze.\n+ if (isa(I))\n+ return nullptr;\n+\n // Replace Op with RepOp in instruction operands.\n SmallVector NewOps;\n bool AnyReplaced = false;\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/icmp.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "icmp_freeze_sext", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @icmp_freeze_sext(i16 %x, i16 %y) {\n %cmp1 = icmp uge i16 %x, %y\n %ext = sext i1 %cmp1 to i16\n %ext.fr = freeze i16 %ext\n %cmp2 = icmp uge i16 %ext.fr, %y\n ret i1 %cmp2\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/select.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "test_replace_freeze_oneuse", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine i8 @test_replace_freeze_oneuse(i1 %x, i8 %y) {\n %ext = zext i1 %x to i8\n %shl = shl nuw i8 %ext, %y\n %shl.fr = freeze i8 %shl\n %sel = select i1 %x, i8 0, i8 %shl.fr\n ret i8 %sel\n}\n"}, {"test_name": "sequence_select_with_same_cond_extra_use", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndeclare void @use32(i32)\n\ndefine i32 @sequence_select_with_same_cond_extra_use(i1 %c1, i1 %c2) {\n %s1 = select i1 %c1, i32 23, i32 45\n call void @use32(i32 %s1)\n %s2 = select i1 %c2, i32 666, i32 %s1\n %s3 = select i1 %c1, i32 789, i32 %s2\n ret i32 %s3\n}\n"}, {"test_name": "test_replace_freeze_multiuse", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine i8 @test_replace_freeze_multiuse(i1 %x, i8 %y) {\n %ext = zext i1 %x to i8\n %shl = shl nuw i8 %ext, %y\n %shl.fr = freeze i8 %shl\n %sel = select i1 %x, i8 0, i8 %shl.fr\n %add = add i8 %shl.fr, %sel\n ret i8 %add\n}\n"}]}], "issue": {"title": "[InstCombine] Miscompilation which simplifies a use of a multi-use freeze", "body": "Reduced test case: https://alive2.llvm.org/ce/z/_JyUGw\r\n```\r\ndefine i8 @func78(i1 %0, i8 %1) {\r\n %3 = zext i1 %0 to i8\r\n %4 = shl i8 %3, %1\r\n %5 = freeze i8 %4\r\n %6 = select i1 %0, i8 0, i8 %5\r\n %7 = call i8 @llvm.fshr.i8(i8 0, i8 %5, i8 %6)\r\n ret i8 %7\r\n}\r\n```\r\n\r\n```\r\n----------------------------------------\r\ndefine i8 @func78(i1 %#0, i8 %#1) {\r\n#2:\r\n %#3 = zext i1 %#0 to i8\r\n %#4 = shl i8 %#3, %#1\r\n %#5 = freeze i8 %#4\r\n %#6 = select i1 %#0, i8 0, i8 %#5\r\n %#7 = fshr i8 0, i8 %#5, i8 %#6\r\n ret i8 %#7\r\n}\r\n=>\r\ndefine i8 @func78(i1 %#0, i8 %#1) nofree noundef willreturn memory(none) {\r\n#2:\r\n %#3 = zext i1 %#0 to i8\r\n %#4 = shl nuw i8 %#3, %#1\r\n %#5 = freeze i8 %#4\r\n ret i8 %#5\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Value mismatch\r\n\r\nExample:\r\ni1 %#0 = #x0 (0)\r\ni8 %#1 = #x08 (8)\r\n\r\nSource:\r\ni8 %#3 = #x00 (0)\r\ni8 %#4 = poison\r\ni8 %#5 = #x00 (0)\r\ni8 %#6 = #x00 (0)\r\ni8 %#7 = #x00 (0)\r\n\r\nTarget:\r\ni8 %#3 = #x00 (0)\r\ni8 %#4 = poison\r\ni8 %#5 = #x9d (157, -99)\r\nSource value: #x00 (0)\r\nTarget value: #x9d (157, -99)\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "89672", "issue_url": "https://github.com/llvm/llvm-project/issues/89672", "bug_type": "miscompilation", "base_commit": "b64e483785bfef5ec4977988543ed5cfaf62f306", "knowledge_cutoff": "2024-04-22T21:35:16Z", "lit_test_dir": ["llvm/test/Transforms/SimplifyCFG"], "hints": {"fix_commit": "883887493c882d656d5da100ee637a348e81357c", "components": ["SimplifyCFG"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/SimplifyCFG.cpp": [[2888, 2894], [2896, 2902]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/SimplifyCFG.cpp": ["isSafeToSpeculateStore"]}}, "patch": "commit 883887493c882d656d5da100ee637a348e81357c\nAuthor: Nikita Popov \nDate: Tue Apr 23 12:36:35 2024 +0900\n\n [SimplifyCFG] Check alignment when speculating stores\n \n When speculating a store based on a preceding load/store, we need\n to ensure that the speculated store does not have a higher\n alignment (which might only be guaranteed by the branch condition).\n \n There are various ways in which this could be strengthened (we\n could get or enforce the alignment), but for now just do the\n simple check against the preceding load/store.\n \n Fixes https://github.com/llvm/llvm-project/issues/89672.\n\ndiff --git a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp\nindex da3eb229c041..0826d748ba0d 100644\n--- a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp\n+++ b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp\n@@ -2888,7 +2888,8 @@ static Value *isSafeToSpeculateStore(Instruction *I, BasicBlock *BrBB,\n // simple, to avoid introducing a spurious non-atomic write after an\n // atomic write.\n if (SI->getPointerOperand() == StorePtr &&\n- SI->getValueOperand()->getType() == StoreTy && SI->isSimple())\n+ SI->getValueOperand()->getType() == StoreTy && SI->isSimple() &&\n+ SI->getAlign() >= StoreToHoist->getAlign())\n // Found the previous store, return its value operand.\n return SI->getValueOperand();\n return nullptr; // Unknown store.\n@@ -2896,7 +2897,7 @@ static Value *isSafeToSpeculateStore(Instruction *I, BasicBlock *BrBB,\n \n if (auto *LI = dyn_cast(&CurI)) {\n if (LI->getPointerOperand() == StorePtr && LI->getType() == StoreTy &&\n- LI->isSimple()) {\n+ LI->isSimple() && LI->getAlign() >= StoreToHoist->getAlign()) {\n // Local objects (created by an `alloca` instruction) are always\n // writable, so once we are past a read from a location it is valid to\n // also write to that same location.\n", "tests": [{"file": "llvm/test/Transforms/SimplifyCFG/speculate-store.ll", "commands": ["opt -passes=simplifycfg -simplifycfg-require-and-preserve-domtree=1 -S < %s"], "tests": [{"test_name": "wrong_align_store", "test_body": "define void @wrong_align_store(ptr %A, i32 %B, i32 %C, i32 %D) {\nentry:\n store i32 %B, ptr %A, align 4\n %cmp = icmp sgt i32 %D, 42\n br i1 %cmp, label %if.then, label %ret.end\n\nif.then: ; preds = %entry\n store i32 %C, ptr %A, align 8\n br label %ret.end\n\nret.end: ; preds = %if.then, %entry\n ret void\n}\n"}, {"test_name": "wrong_align_load", "test_body": "define void @wrong_align_load(i32 %C, i32 %D) {\nentry:\n %A = alloca i32, align 4\n %0 = load i32, ptr %A, align 4\n %cmp = icmp sgt i32 %D, 42\n br i1 %cmp, label %if.then, label %ret.end\n\nif.then: ; preds = %entry\n store i32 %C, ptr %A, align 8\n br label %ret.end\n\nret.end: ; preds = %if.then, %entry\n ret void\n}\n"}]}], "issue": {"title": "coalescing of redundant vector stores isn't preserving alignment correctly", "body": "https://alive2.llvm.org/ce/z/-qQphe\r\n\r\noptimizing this code:\r\n```llvm\r\ndefine i32 @f(ptr %0, i1 %1) {\r\n store <2 x i64> zeroinitializer, ptr %0, align 8\r\n br i1 %1, label %4, label %3\r\n\r\n3: ; preds = %2\r\n store <2 x i64> zeroinitializer, ptr %0, align 16\r\n br label %4\r\n\r\n4: ; preds = %3, %2\r\n ret i32 0\r\n}\r\n```\r\n\r\nis mostly doing what we expect, but the coalesced store should retain the smaller alignnment value of the two, not the larger:\r\n```lllvm\r\ndefine noundef i32 @f(ptr nocapture writeonly %0, i1 %1) local_unnamed_addr #0 {\r\n store <2 x i64> zeroinitializer, ptr %0, align 16\r\n ret i32 0\r\n}\r\n\r\nattributes #0 = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: write) }\r\n```\r\n\r\ncc @nunoplopes @hatsunespica", "author": "regehr", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "nikic", "body": "The issue is introduced by SimplifyCFG: https://alive2.llvm.org/ce/z/zf-j74"}, {"author": "nikic", "body": "Specifically this is isSafeToSpeculateStore(), which looks for a preceding store to check whether it is safe to speculate, but fails to check alignment."}]}, "verified": true} {"bug_id": "97586", "issue_url": "https://github.com/llvm/llvm-project/issues/97586", "bug_type": "crash", "base_commit": "be7239e5a60927f5b4932f995dc4b57423ea8534", "knowledge_cutoff": "2024-07-03T14:47:17Z", "lit_test_dir": ["llvm/test/Transforms/LoopUnroll"], "hints": {"fix_commit": "cff8d716bdf017a2af8eb8623257fd33ee43f30e", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[8412, 8418]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::visitAndClearUsers"]}}, "patch": "commit cff8d716bdf017a2af8eb8623257fd33ee43f30e\nAuthor: v01dXYZ <14996868+v01dXYZ@users.noreply.github.com>\nDate: Tue Jul 9 09:14:33 2024 +0200\n\n [SCEV] forgetValue: support (with-overflow-inst op0, op1) (#98015)\n \n The use-def walk in forgetValue() was skipping instructions with\n non-SCEVable types. However, SCEV may look past with.overflow\n intrinsics returning aggregates.\n \n Fixes #97586.\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex 430e1c6d8f8c..51cffac80876 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -8412,7 +8412,7 @@ void ScalarEvolution::visitAndClearUsers(\n SmallVectorImpl &ToForget) {\n while (!Worklist.empty()) {\n Instruction *I = Worklist.pop_back_val();\n- if (!isSCEVable(I->getType()))\n+ if (!isSCEVable(I->getType()) && !isa(I))\n continue;\n \n ValueExprMapType::iterator It =\n", "tests": [{"file": "llvm/test/Transforms/LoopUnroll/peel-loop-scev-invalidate-with-overflow-inst.ll", "commands": ["opt < %s -S -passes='print,loop-unroll,print' 2>&1"], "tests": [{"test_name": "", "test_body": ";\n; This test ensures that (extractvalue 0 (with-overflow-inst op0, op1))\n; is invalidated by LoopPeel when the operands of with-overflow-inst\n; are changed.\n;\n; In the following case, LoopPeel modifies the CFG into another one\n; with %bb7 not dominating %bb2 and %bb3 although %extractvalue is\n; still the step for the %bb3 loop. %call has been modified and uses\n; different operands but the SCEV value for %extractvalue has not been\n; invalidated and still refers to %load in its SCEV operands\n; (SCEV(%extractvalue) := -2 + -2 * %load).\n;\n; When LoopUnroll tries to compute the SCEV for the %bb3 Phi, the\n; stale data for %extractvalue is used whereas %load is not available\n; in %bb3 which is wrong.\n;\n; for more details and nice pictures: https://github.com/llvm/llvm-project/issues/97586\n;\n; Although the reason for the bug was in forgetValue, it is still relevant to\n; test if LoopPeel invalidates %extractvalue after changing %call.\n;\n; forgetValue only walks the users, so calling it on the IV Phis does not\n; invalidate %extractvalue (thus forgetLoop does not invalidate it too).\n; It has to be done by LoopPeel itself.\n\n\ndefine void @loop_peeling_smul_with_overflow() {\n; before loop-unroll\n; after loop-unroll\n;\nbb:\n br label %bb1\n\nbb1: ; preds = %bb3, %bb\n %phi = phi i32 [ 0, %bb ], [ %phi4, %bb3 ]\n br label %bb5\n\nbb2: ; preds = %bb7\n %call = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 %add8, i32 -2)\n %extractvalue = extractvalue { i32, i1 } %call, 0\n br label %bb3\n\nbb3: ; preds = %bb3, %bb2\n %phi4 = phi i32 [ %add, %bb3 ], [ 0, %bb2 ]\n %add = add i32 %extractvalue, %phi4\n br i1 false, label %bb3, label %bb1\n\nbb5: ; preds = %bb7, %bb1\n %phi6 = phi i32 [ 1, %bb1 ], [ 0, %bb7 ]\n %icmp = icmp eq i32 %phi, 0\n br i1 %icmp, label %bb9, label %bb7\n\nbb7: ; preds = %bb5\n %load = load i32, ptr addrspace(1) null, align 4\n %add8 = add i32 %load, 1\n br i1 false, label %bb2, label %bb5\n\nbb9: ; preds = %bb5\n ret void\n}"}]}], "issue": {"title": "Assertion Failure in `getSCEV` during loop unroll", "body": "```\r\n; ModuleID = ''\r\nsource_filename = \"\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @ham() {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb3, %bb\r\n %phi = phi i32 [ 0, %bb ], [ %phi4, %bb3 ]\r\n br label %bb5\r\n\r\nbb2: ; preds = %bb7\r\n %call = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 %add8, i32 -2)\r\n %extractvalue = extractvalue { i32, i1 } %call, 0\r\n br label %bb3\r\n\r\nbb3: ; preds = %bb3, %bb2\r\n %phi4 = phi i32 [ %add, %bb3 ], [ 0, %bb2 ]\r\n %add = add i32 %extractvalue, %phi4\r\n br i1 false, label %bb3, label %bb1\r\n\r\nbb5: ; preds = %bb7, %bb1\r\n %phi6 = phi i32 [ 1, %bb1 ], [ 0, %bb7 ]\r\n %icmp = icmp eq i32 %phi, 0\r\n br i1 %icmp, label %bb9, label %bb7\r\n\r\nbb7: ; preds = %bb5\r\n %load = load i32, ptr addrspace(1) null, align 4\r\n %add8 = add i32 %load, 1\r\n br i1 false, label %bb2, label %bb5\r\n\r\nbb9: ; preds = %bb5\r\n ret void\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare { i32, i1 } @llvm.smul.with.overflow.i32(i32, i32) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\ncrashes with the below backtrace when run with `opt --passes=\"print,loop-unroll \"` \r\n```\r\nopt: /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:3678: const llvm::SCEV* llvm::ScalarEvolution::getAddRecExpr(llvm::SmallVectorImpl&, const llvm::Loop*, llvm::SCEV::NoWrapFlags): Assertion `isAvailableAtLoopEntry(Operands[i], L) && \"SCEVAddRecExpr operand is not available at loop entry!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /home/manishkh/Workspace7/orcaInstall/installDA/bin/opt --passes=print,loop-unroll reduced2.ll -disable-output\r\n #0 0x00007f272f49aa72 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) /home/manishkh/Workspace7/orca/llvm/lib/Support/Unix/Signals.inc:723:0\r\n #1 0x00007f272f49aeaf PrintStackTraceSignalHandler(void*) /home/manishkh/Workspace7/orca/llvm/lib/Support/Unix/Signals.inc:798:0\r\n #2 0x00007f272f498793 llvm::sys::RunSignalHandlers() /home/manishkh/Workspace7/orca/llvm/lib/Support/Signals.cpp:105:0\r\n #3 0x00007f272f49a459 SignalHandler(int) /home/manishkh/Workspace7/orca/llvm/lib/Support/Unix/Signals.inc:413:0\r\n #4 0x00007f2738fd7630 __restore_rt sigaction.c:0:0\r\n #5 0x00007f272b893387 raise (/lib64/libc.so.6+0x36387)\r\n #6 0x00007f272b894a78 abort (/lib64/libc.so.6+0x37a78)\r\n #7 0x00007f272b88c1a6 __assert_fail_base (/lib64/libc.so.6+0x2f1a6)\r\n #8 0x00007f272b88c252 (/lib64/libc.so.6+0x2f252)\r\n #9 0x00007f2732031da6 llvm::ScalarEvolution::getAddRecExpr(llvm::SmallVectorImpl&, llvm::Loop const*, llvm::SCEV::NoWrapFlags) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:3677:0\r\n#10 0x00007f2732031b29 llvm::ScalarEvolution::getAddRecExpr(llvm::SCEV const*, llvm::SCEV const*, llvm::Loop const*, llvm::SCEV::NoWrapFlags) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:3660:0\r\n#11 0x00007f2732039c11 llvm::ScalarEvolution::createSimpleAffineAddRec(llvm::PHINode*, llvm::Value*, llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:5776:0\r\n#12 0x00007f2732039f65 llvm::ScalarEvolution::createAddRecFromPHI(llvm::PHINode*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:5831:0\r\n#13 0x00007f273203aadf llvm::ScalarEvolution::createNodeForPHI(llvm::PHINode*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:6025:0\r\n#14 0x00007f273204562a llvm::ScalarEvolution::createSCEV(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8069:0\r\n#15 0x00007f2732042338 llvm::ScalarEvolution::createSCEVIter(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:7447:0\r\n#16 0x00007f2732035a05 llvm::ScalarEvolution::getSCEV(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:4572:0\r\n#17 0x00007f273203ab71 llvm::ScalarEvolution::createNodeForPHI(llvm::PHINode*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:6029:0\r\n#18 0x00007f273204562a llvm::ScalarEvolution::createSCEV(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8069:0\r\n#19 0x00007f2732042338 llvm::ScalarEvolution::createSCEVIter(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:7447:0\r\n#20 0x00007f2732035a05 llvm::ScalarEvolution::getSCEV(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:4572:0\r\n#21 0x00007f2732039fcf llvm::ScalarEvolution::createAddRecFromPHI(llvm::PHINode*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:5840:0\r\n#22 0x00007f273203aadf llvm::ScalarEvolution::createNodeForPHI(llvm::PHINode*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:6025:0\r\n#23 0x00007f273204562a llvm::ScalarEvolution::createSCEV(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8069:0\r\n#24 0x00007f2732042338 llvm::ScalarEvolution::createSCEVIter(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:7447:0\r\n#25 0x00007f2732035a05 llvm::ScalarEvolution::getSCEV(llvm::Value*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:4572:0\r\n#26 0x00007f273204a26a llvm::ScalarEvolution::computeExitLimitFromICmp(llvm::Loop const*, llvm::ICmpInst*, bool, bool, bool) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:9117:0\r\n#27 0x00007f273204971e llvm::ScalarEvolution::computeExitLimitFromCondImpl(llvm::ScalarEvolution::ExitLimitCache&, llvm::Loop const*, llvm::Value*, bool, bool, bool) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8975:0\r\n#28 0x00007f273204955b llvm::ScalarEvolution::computeExitLimitFromCondCached(llvm::ScalarEvolution::ExitLimitCache&, llvm::Loop const*, llvm::Value*, bool, bool, bool) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8958:0\r\n#29 0x00007f27320491a4 llvm::ScalarEvolution::computeExitLimitFromCond(llvm::Loop const*, llvm::Value*, bool, bool, bool) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8914:0\r\n#30 0x00007f2732048f49 llvm::ScalarEvolution::computeExitLimit(llvm::Loop const*, llvm::BasicBlock*, bool) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8888:0\r\n#31 0x00007f27320488db llvm::ScalarEvolution::computeBackedgeTakenCount(llvm::Loop const*, bool) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8797:0\r\n#32 0x00007f27320469fb llvm::ScalarEvolution::getBackedgeTakenInfo(llvm::Loop const*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8358:0\r\n#33 0x00007f27320464ab llvm::ScalarEvolution::getExitCount(llvm::Loop const*, llvm::BasicBlock const*, llvm::ScalarEvolution::ExitCountKind) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8270:0\r\n#34 0x00007f27320460c6 llvm::ScalarEvolution::getSmallConstantTripCount(llvm::Loop const*, llvm::BasicBlock const*) /home/manishkh/Workspace7/orca/llvm/lib/Analysis/ScalarEvolution.cpp:8203:0\r\n#35 0x00007f273146141a tryToUnrollLoop(llvm::Loop*, llvm::DominatorTree&, llvm::LoopInfo*, llvm::ScalarEvolution&, llvm::TargetTransformInfo const&, llvm::AssumptionCache&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::ProfileSummaryInfo*, bool, int, bool, bool, bool, std::optional, std::optional, std::optional, std::optional, std::optional, std::optional, std::optional, std::optional, llvm::AAResults*) /home/manishkh/Workspace7/orca/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp:1259:0\r\n#36 0x00007f2731462e2c llvm::LoopUnrollPass::run(llvm::Function&, llvm::AnalysisManager&) /home/manishkh/Workspace7/orca/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp:1658:0\r\n#37 0x00007f27339727ac llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) /home/manishkh/Workspace7/orca/llvm/include/llvm/IR/PassManagerInternal.h:91:0\r\n#38 0x00007f272f831f82 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) /home/manishkh/Workspace7/orca/llvm/include/llvm/IR/PassManager.h:228:0\r\n#39 0x00007f27331d1e32 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) /home/manishkh/Workspace7/orca/llvm/include/llvm/IR/PassManagerInternal.h:91:0\r\n#40 0x00007f272f830d12 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) /home/manishkh/Workspace7/orca/llvm/lib/IR/PassManager.cpp:128:0\r\n#41 0x00007f27331d1d96 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) /home/manishkh/Workspace7/orca/llvm/include/llvm/IR/PassManagerInternal.h:91:0\r\n#42 0x00007f272f831c5a llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) /home/manishkh/Workspace7/orca/llvm/include/llvm/IR/PassManager.h:228:0\r\n#43 0x00000000004acc3e llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) /home/manishkh/Workspace7/orca/llvm/tools/opt/NewPMDriver.cpp:559:0\r\n#44 0x0000000000477a33 optMain /home/manishkh/Workspace7/orca/llvm/tools/opt/optdriver.cpp:756:0\r\n#45 0x000000000047540f main /home/manishkh/Workspace7/orca/llvm/tools/opt/opt.cpp:25:0\r\n#46 0x00007f272b87f555 __libc_start_main (/lib64/libc.so.6+0x22555)\r\n#47 0x00000000004752c9 _start (/home/manishkh/Workspace7/orcaInstall/installDA/bin/opt+0x4752c9)\r\n```\r\n\r\nas demonstrated by this godbolt example: https://godbolt.org/z/be3a4Yr9e", "author": "Nirhar", "labels": ["llvm:crash", "llvm:SCEV"], "comments": [{"author": "Nirhar", "body": "I did some preliminary analysis on this bug, and here is what I found:\r\nThe IR at the point of crash looks like this:\r\n```\r\ndefine zing void @ham() {\r\nbb:\r\n br label %bb1\r\n\r\nbb1.loopexit: ; preds = %bb3\r\n %phi4.lcssa = phi i32 [ %phi4, %bb3 ]\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb1.loopexit, %bb\r\n %phi = phi i32 [ 0, %bb ], [ %phi4.lcssa, %bb1.loopexit ]\r\n br label %bb5.peel.begin\r\n\r\nbb5.peel.begin: ; preds = %bb1\r\n br label %bb5.peel\r\n\r\nbb5.peel: ; preds = %bb5.peel.begin, exiting\r\n %icmp.peel = icmp eq i32 %phi, 0\r\n br i1 %icmp.peel, label %bb9.loopexit2, label %bb7.peel\r\n\r\nbb7.peel: ; preds = %bb5.peel\r\n %load.peel = load i32, ptr addrspace(1) null, align 4\r\n %add8.peel = add i32 %load.peel, 1\r\n br i1 false, label %bb2, label %bb5.peel.next\r\n\r\nbb5.peel.next: ; preds = %bb7.peel\r\n br label %bb5.peel.next1\r\n\r\nbb5.peel.next1: ; preds = %bb5.peel.next\r\n br label %bb1.peel.newph\r\n\r\nbb1.peel.newph: ; preds = %bb5.peel.next1\r\n br label %bb5\r\n\r\nbb2.loopexit: ; preds = %bb7\r\n %add8.lcssa.ph = phi i32 [ %add8, %bb7 ]\r\n br label %bb2\r\n\r\nbb2: ; preds = %bb2.loopexit, %bb7.peel\r\n %add8.lcssa = phi i32 [ %add8.peel, %bb7.peel ], [ %add8.lcssa.ph, %bb2.loopexit ]\r\n %call = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 %add8.lcssa, i32 -2)\r\n %extractvalue = extractvalue { i32, i1 } %call, 0\r\n br label %bb3\r\n\r\nbb3: ; preds = %bb3, %bb2\r\n %phi4 = phi i32 [ %add, %bb3 ], [ 0, %bb2 ]\r\n %add = add i32 %extractvalue, %phi4\r\n br i1 false, label %bb3, label %bb1.loopexit\r\n\r\nbb5: ; preds = %bb7, %bb1.peel.newph, exiting\r\n br i1 false, label %bb9.loopexit, label %bb7\r\n\r\nbb7: ; preds = %bb5\r\n %load = load i32, ptr addrspace(1) null, align 4 \r\n %add8 = add i32 %load, 1\r\n br i1 false, label %bb2.loopexit, label %bb5, !llvm.loop !0\r\n\r\nbb9.loopexit: ; preds = %bb5\r\n br label %bb9\r\n\r\nbb9.loopexit2: ; preds = %bb5.peel\r\n br label %bb9\r\n\r\nbb9: ; preds = %bb9.loopexit2, %bb9.loopexit\r\n ret void\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare { i32, i1 } @llvm.smul.with.overflow.i32(i32, i32) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\nWe have 3 loops in this this IR(1 outer and 2 inner, out of which the bb5-bb7 loop has been peeled.):\r\n```\r\nLoop at depth 1 containing: %bb1
,%bb5,%bb7,%bb2,%bb3,%bb1.loopexit,%bb5.peel.begin,%bb5.peel.next,%bb1.peel.newph,%bb5.peel,%bb7.peel,%bb5.peel.next1\r\n Loop at depth 2 containing: %bb5
,%bb7\r\n Loop at depth 2 containing: %bb3
\r\n``` \r\nHere is the CFG for this IR:\r\n![Screenshot 2024-07-03 at 8 24 52 PM](https://github.com/llvm/llvm-project/assets/46352931/02e57740-9ef6-4945-868e-62a645325552)\r\n\r\nThe assertion fails because SCEV expects bb7 to properly dominate bb3, which it does in the old IR, but after peeling, it does not. This seems to arise from the fact that SCEVs in %bb2 have not been invalidated after the peeling. I see that the SCEV for `%extractvalue` comes from the ValueExprMap instead of being recomputed, following the peel, and the use of this invalid SCEV is what I think is causing the problem.\r\n\r\n\r\n"}, {"author": "Nirhar", "body": "Perhaps these updates to SE are not enough?\r\nhttps://github.com/llvm/llvm-project/blob/86d456ad8f2aef6edcc36d44d26119025331981e/llvm/lib/Transforms/Utils/LoopPeel.cpp#L1088-L1090"}, {"author": "Nirhar", "body": "@nikic @fhahn @RKSimon is `forgetTopmostLoop` supposed to forget SCEV of all SSA values inside the loop(ie, header, latch and other blocks) ? I see that its implementation does not do that currently."}, {"author": "nikic", "body": "@Nirhar Generally no. It forgets information about induction variables (things based on loop header phis), as well as information pertaining to the loop (such as BECounts, exit values etc), but not of literally everything defined in the loop."}, {"author": "nikic", "body": "From a super quick look, the issue here might be that the exit phi forgetValue() calls don't invalidate enough values due to the condition at https://github.com/llvm/llvm-project/blob/e55585fd7bddf5bb3824a53cbe2971206d3c20c6/llvm/lib/Analysis/ScalarEvolution.cpp#L8415, which means that we will stop at the aggregate results of with.overflow intrinsics, even though SCEV can look through them. It might be that they need special handling."}, {"author": "v01dXYZ", "body": "When replacing the `smul.with.overflow + extractvalue` by a `mul`, it doesn't abort.\r\nWhen removing the SCEVable check, it doesn't abort.\r\n\r\nThe function that allows to traverse unSCEVable when creating the SCEV is `MatchBinaryOp` (cf `createSimpleAffineRec`). We could make this function returns a subset of the use-def to SCEVable or have another function that do the reverse with a def-use matching."}, {"author": "Nirhar", "body": "I think that the issue might be something else. From my analysis, here is how the control flow goes:\r\n1. When we hit `forgetTopmostLoop`, we try to forget the SCEV values of the loop:\r\n```\r\nLoop at depth 1 containing: %bb1
,%bb5,%bb7,%bb2,%bb3,%bb1.loopexit,%bb5.peel.begin,%bb5.peel.next,%bb1.peel.newph,%bb5.peel,%bb7.peel,%bb5.peel.next1\r\n Loop at depth 2 containing: %bb5
,%bb7\r\n Loop at depth 2 containing: %bb3
\r\n```\r\nWe first take the phi node in the header `%phi = phi i32 [ 0, %bb ], [ %phi4.lcssa, %bb1.loopexit ]` and start the DFS from there. From here, we go to the following nodes:(` %icmp.peel = icmp eq i32 %phi, 0` , ` br i1 %icmp.peel, label %bb9.loopexit2, label %bb7.peel` and now non existent `%icmp = icmp eq i32 %phi, 0` , `br i1 %icmp, label %bb9, label %bb7`. The DFS for the outer loop ends here.\r\n\r\n2. Next we traverse the inner loop:\r\n```\r\nLoop at depth 2 containing: %bb3
\r\n```\r\nHere we start with phi node in the header `%phi4 = phi i32 [ %add, %bb3 ], [ 0, %bb2 ]` and then traverse through the nodes: `%add = add i32 %extractvalue, %phi4` and `%phi4.lcssa = phi i32 [ %phi4, %bb3 ]`, after which we quit traversal.\r\n\r\n3. Finally we visit the inner loop:\r\n```\r\nLoop at depth 2 containing: %bb5
,%bb7\r\n```\r\nWe start and end DFS at (now non-existent)`%phi6 = phi i32 [ 0, %bb1.peel.newph ], [ 0, %bb7 ]`.\r\n\r\n**_TLDR_**; Throughout the entire traversal, we never visit the basic block bb2, and never invalidate its SCEV values. We don't even visit smul intrinsic call to forget its SCEV. Perhaps one needs to consider use-def chains to forget SCEV values, similar to the now considered def-use chain.\r\n\r\nNote: The IR at the point of crash is slightly different from the IR after the previous loop-peel. I've used the term `now non-existent` to mean that that node is not present in the IR at the crash point."}, {"author": "nikic", "body": "@Nirhar The forgetTopmostLoop() call is not what is responsible for the invalidation here, it's the forgetValue() call on the LCSSA phis in https://github.com/llvm/llvm-project/blob/27ccc8835e5163484234549fa6128eeb00d7432d/llvm/lib/Transforms/Utils/LoopPeel.cpp#L862"}, {"author": "Nirhar", "body": "Aah sorry, I missed that. Thanks for pointing it out!"}]}, "verified": true} {"bug_id": "76713", "issue_url": "https://github.com/llvm/llvm-project/issues/76713", "bug_type": "miscompilation", "base_commit": "71bcef0b0bf55a96f85b2f323b0beb13ad8e2caa", "knowledge_cutoff": "2024-01-02T12:49:15Z", "lit_test_dir": ["llvm/test/Transforms/ConstraintElimination"], "hints": {"fix_commit": "3c127e83c07c3791e86413c22a414a030d4302e3", "components": ["ConstraintElimination"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/ConstraintElimination.cpp": [[366, 371], [569, 578]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/ConstraintElimination.cpp": ["add", "decompose", "mul"]}}, "patch": "commit 3c127e83c07c3791e86413c22a414a030d4302e3\nAuthor: Florian Hahn \nDate: Tue Jan 2 22:05:57 2024 +0000\n\n [ConstraintElim] Replace NUWSub decomp with recursive decomp of ops.\n \n The current patterns for NUWSub decompositions do not handle negative\n constants correctly at the moment (causing #76713).\n \n Replace the incorrect pattern by more general code that recursively\n decomposes the operands and then combines the results. This is already\n done in most other places that handle operators like add/mul.\n \n This means we fall back to the general constant handling code (fixes the\n mis-compile) while also being able to support reasoning about\n decomposable expressions in the SUB operands.\n \n Fixes https://github.com/llvm/llvm-project/issues/76713.\n\ndiff --git a/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp b/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\nindex 49ac1e96e255..06c87bd6dc37 100644\n--- a/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\n+++ b/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\n@@ -366,6 +366,13 @@ struct Decomposition {\n append_range(Vars, Other.Vars);\n }\n \n+ void sub(const Decomposition &Other) {\n+ Decomposition Tmp = Other;\n+ Tmp.mul(-1);\n+ add(Tmp.Offset);\n+ append_range(Vars, Tmp.Vars);\n+ }\n+\n void mul(int64_t Factor) {\n Offset = multiplyWithOverflow(Offset, Factor);\n for (auto &Var : Vars)\n@@ -569,10 +576,12 @@ static Decomposition decompose(Value *V,\n return Result;\n }\n \n- if (match(V, m_NUWSub(m_Value(Op0), m_ConstantInt(CI))) && canUseSExt(CI))\n- return {-1 * CI->getSExtValue(), {{1, Op0}}};\n- if (match(V, m_NUWSub(m_Value(Op0), m_Value(Op1))))\n- return {0, {{1, Op0}, {-1, Op1}}};\n+ if (match(V, m_NUWSub(m_Value(Op0), m_Value(Op1)))) {\n+ auto ResA = decompose(Op0, Preconditions, IsSigned, DL);\n+ auto ResB = decompose(Op1, Preconditions, IsSigned, DL);\n+ ResA.sub(ResB);\n+ return ResA;\n+ }\n \n return {V, IsKnownNonNegative};\n }\n", "tests": [{"file": "llvm/test/Transforms/ConstraintElimination/sub-nuw.ll", "commands": ["opt -passes=constraint-elimination -S %s"], "tests": [{"test_name": "pr76713", "test_body": "define i1 @pr76713(i16 %i1, i16 %i3) {\nentry:\n %c1 = icmp ult i16 %i1, -1\n %c2 = icmp uge i16 %i1, -3\n %c3 = icmp ult i16 %i3, 2\n %and = and i1 %c1, %c2\n %and.2 = and i1 %and, %c3\n br i1 %and, label %then, label %else\n\nthen: ; preds = %entry\n %sub = sub nuw nsw i16 %i1, -3\n %arrayidx.idx = mul nuw nsw i16 %i3, 4\n %i6 = add nuw nsw i16 %arrayidx.idx, %sub\n %c4 = icmp ult i16 12, %i6\n ret i1 %c4\n\nelse: ; preds = %entry\n ret i1 false\n}\n"}, {"test_name": "sub_nuw_chained_negative_constants", "test_body": "declare void @use(i1)\n\ndefine void @sub_nuw_chained_negative_constants(i8 %a) {\nentry:\n %sub1 = sub nuw i8 %a, 10\n %sub2 = sub nuw i8 %sub1, -126\n %c.1 = icmp ugt i8 %sub2, 20\n br i1 %c.1, label %exit.1, label %exit.2\n\nexit.1: ; preds = %entry\n %c.2 = icmp ugt i8 %a, -96\n call void @use(i1 %c.2)\n %c.3 = icmp ugt i8 %a, -95\n call void @use(i1 %c.3)\n ret void\n\nexit.2: ; preds = %entry\n %c.4 = icmp ugt i8 %a, -96\n call void @use(i1 %c.4)\n %c.5 = icmp ugt i8 %a, -95\n call void @use(i1 %c.5)\n ret void\n}\n"}, {"test_name": "sub_nuw_neg_i16", "test_body": "define i1 @sub_nuw_neg_i16(i16 %a) {\nentry:\n %neg2 = sub nuw i16 %a, -305\n %c.1 = icmp ugt i16 0, %neg2\n br i1 %c.1, label %exit.1, label %exit.2\n\nexit.1: ; preds = %entry\n %c.2 = icmp ugt i16 %a, 0\n ret i1 %c.2\n\nexit.2: ; preds = %entry\n %c.3 = icmp ugt i16 %a, 0\n ret i1 %c.3\n}\n"}, {"test_name": "sub_nuw_chained_positive_constants", "test_body": "declare void @use(i1)\n\ndefine void @sub_nuw_chained_positive_constants(i16 %a) {\nentry:\n %sub1 = sub nuw i16 %a, 10\n %sub2 = sub nuw i16 %sub1, 20\n %c.1 = icmp ugt i16 %sub2, 90\n br i1 %c.1, label %exit.1, label %exit.2\n\nexit.1: ; preds = %entry\n %c.2 = icmp ugt i16 %a, 120\n call void @use(i1 %c.2)\n %c.3 = icmp ugt i16 %a, 121\n call void @use(i1 %c.3)\n ret void\n\nexit.2: ; preds = %entry\n %c.4 = icmp ugt i16 %a, 120\n call void @use(i1 %c.4)\n %c.5 = icmp ugt i16 %a, 121\n call void @use(i1 %c.5)\n ret void\n}\n"}]}], "issue": {"title": "Miscompile with opt -passes=\"constraint-elimination\"", "body": "llvm commit: 5842dfe34d89\r\nReproduce with: ```opt -passes=\"constraint-elimination\" bbi-90501.ll -S -o -```\r\n[bbi-90501.ll.gz](https://github.com/llvm/llvm-project/files/13810077/bbi-90501.ll.gz)\r\n\r\nThere are two nested loops in the input that should finish and then we should end up at\r\n```\r\nbb.2: ; preds = %bb.1\r\n ret i16 0\r\n```\r\nbut with constraint-elimination the condition in bb.5 is changed to\r\n```\r\n %i10 = or i1 true, %i9\r\n br i1 %i10, label %bb.7, label %bb.6\r\n```\r\nso we will go to\r\n```\r\nbb.7: ; preds = %bb.5\r\n ret i16 1\r\n```\r\ninstead.", "author": "mikaelholmen", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "mikaelholmen", "body": "This starts happening with e6a1657fa30c747f4412fc47f567660ebe861a9e\r\n```\r\n[ConstraintElim] Add A < B if A is an increasing phi for A != B.\r\n```\r\n@fhahn : any idea what's up?"}, {"author": "fhahn", "body": "Looks like an issue with `sub` decomposition with negative constants, preparing a fix."}, {"author": "mikaelholmen", "body": "Thanks for the quick fix! I've verified it solves the problem I saw."}, {"author": "fhahn", "body": "@mikaelholmen thanks for the reproducer!"}, {"author": "fhahn", "body": "(appreciated as always, thanks for your extensive testing)"}]}, "verified": true} {"bug_id": "116775", "issue_url": "https://github.com/llvm/llvm-project/issues/116775", "bug_type": "crash", "base_commit": "bdf00e2216280edef1ec91ccc07987db92197b59", "knowledge_cutoff": "2024-11-19T09:53:12Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "03d8831fa8ef5b7e32172c718b550a454645faea", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": [[997, 1003]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": ["InstCombinerImpl::SimplifyDemandedUseBits"]}}, "patch": "commit 03d8831fa8ef5b7e32172c718b550a454645faea\nAuthor: Yingwei Zheng \nDate: Tue Nov 19 22:17:24 2024 +0800\n\n [InstCombine] Handle constant GEP expr in `SimplifyDemandedUseBits` (#116794)\n \n Closes https://github.com/llvm/llvm-project/issues/116775.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\nindex 5eb807dcb76c..392c5c78345c 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n@@ -997,7 +997,7 @@ Value *InstCombinerImpl::SimplifyDemandedUseBits(Instruction *I,\n uint64_t MaskedGEPIndex = HighBitsGEPIndex | MaskedLowBitsGEPIndex;\n \n if (MaskedGEPIndex != GEPIndex) {\n- auto *GEP = cast(II->getArgOperand(0));\n+ auto *GEP = cast(II->getArgOperand(0));\n Builder.SetInsertPoint(I);\n Type *GEPIndexType =\n DL.getIndexType(GEP->getPointerOperand()->getType());\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/ptrmask.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "ptrmask_demandedbits_constantexpr", "test_body": "target datalayout = \"p1:64:64:64:32\"\n\n@GC_arrays = external global { i8, i8, i64 }\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare ptr @llvm.ptrmask.p0.i64(ptr, i64) #0\n\ndefine ptr @ptrmask_demandedbits_constantexpr() {\nentry:\n %aligned_result = call ptr @llvm.ptrmask.p0.i64(ptr getelementptr inbounds (i8, ptr @GC_arrays, i64 1), i64 -8)\n ret ptr %aligned_result\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "ptrmask_is_useless_fail1", "test_body": "target datalayout = \"p1:64:64:64:32\"\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare ptr @llvm.ptrmask.p0.i64(ptr, i64) #0\n\ndefine ptr @ptrmask_is_useless_fail1(i64 %i, i64 %m) {\n %m0 = and i64 %m, 127\n %i0 = and i64 %i, 511\n %p0 = inttoptr i64 %i0 to ptr\n %r = call ptr @llvm.ptrmask.p0.i64(ptr %p0, i64 %m0)\n ret ptr %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "Segfault when building https://github.com/ivmai/bdwgc@4fe4238", "body": "Attempting to build [https://github.com/ivmai/bdwgc@4fe4238](https://github.com/ivmai/bdwgc/tree/4fe42389509ff059bad8d4c955a570b34a6bda4b) using\r\n```shell\r\n$ ./autogen.sh\r\n$ ./configure --with-libatomic-ops --enable-cplusplus --enable-large-config\r\n$ make\r\n```\r\nresults in\r\n```text\r\nlibtool: compile: clang -std=gnu23 -DHAVE_CONFIG_H -I./include -I./include -DNDEBUG -fexceptions -DGC_VISIBILITY_HIDDEN_SET -fvisibility=hidden -Wall -Wextra -Wpedantic -Wno-long-long -O2 -march=native -Wno-frame-address -MT extra/gc.lo -MD -MP -MF extra/.deps/gc.Tpo -c extra/gc.c -fno-common -DPIC -o extra/.libs/gc.o\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: clang -std=gnu23 -DHAVE_CONFIG_H -I./include -I./include -DNDEBUG -fexceptions -DGC_VISIBILITY_HIDDEN_SET -fvisibility=hidden -Wall -Wextra -Wpedantic -Wno-long-long -O2 -march=native -Wno-frame-address -MT extra/gc.lo -MD -MP -MF extra/.deps/gc.Tpo -c extra/gc.c -fno-common -DPIC -o extra/.libs/gc.o\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"require,function(invalidate),require,cgscc(devirt<4>(inline,function-attrs,openmp-opt-cgscc,function(sroa,early-cse,speculative-execution,jump-threading,correlated-propagation,simplifycfg,instcombine,aggressive-instcombine,libcalls-shrinkwrap,tailcallelim,simplifycfg,reassociate,constraint-elimination,loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-unswitch),simplifycfg,instcombine,loop(loop-idiom,indvars,simple-loop-unswitch,loop-deletion,loop-unroll-full),sroa,vector-combine,mldst-motion,gvn<>,sccp,bdce,instcombine,jump-threading,correlated-propagation,adce,memcpyopt,dse,move-auto-init,loop-mssa(licm),coro-elide,simplifycfg,instcombine),function-attrs,function(require),coro-split,coro-annotation-elide)),function(invalidate),cgscc(devirt<4>())\" on module \"extra/gc.c\"\r\n4.\tRunning pass \"cgscc(devirt<4>(inline,function-attrs,openmp-opt-cgscc,function(sroa,early-cse,speculative-execution,jump-threading,correlated-propagation,simplifycfg,instcombine,aggressive-instcombine,libcalls-shrinkwrap,tailcallelim,simplifycfg,reassociate,constraint-elimination,loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-unswitch),simplifycfg,instcombine,loop(loop-idiom,indvars,simple-loop-unswitch,loop-deletion,loop-unroll-full),sroa,vector-combine,mldst-motion,gvn<>,sccp,bdce,instcombine,jump-threading,correlated-propagation,adce,memcpyopt,dse,move-auto-init,loop-mssa(licm),coro-elide,simplifycfg,instcombine),function-attrs,function(require),coro-split,coro-annotation-elide))\" on module \"extra/gc.c\"\r\n5.\tRunning pass \"instcombine\" on function \"GC_push_finalizer_structures\"\r\n #0 0x0000000112ecda98 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x3299a98)\r\n #1 0x0000000112eccb47 llvm::sys::RunSignalHandlers() (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x3298b47)\r\n #2 0x0000000112eccde1 llvm::sys::CleanupOnSignal(unsigned long) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x3298de1)\r\n #3 0x0000000112df2644 CrashRecoverySignalHandler(int) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x31be644)\r\n #4 0x00007ff81abe4fdd (/usr/lib/system/libsystem_platform.dylib+0x7ff80046afdd)\r\n #5 0x0000000112e1d131 computeForAddCarry(llvm::KnownBits const&, llvm::KnownBits const&, bool, bool) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x31e9131)\r\n #6 0x0000000112318cdd llvm::TargetFolder::FoldGEP(llvm::Type*, llvm::Value*, llvm::ArrayRef, llvm::GEPNoWrapFlags) const (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x26e4cdd)\r\n #7 0x00000001102d2177 llvm::IRBuilderBase::CreateGEP(llvm::Type*, llvm::Value*, llvm::ArrayRef, llvm::Twine const&, llvm::GEPNoWrapFlags) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x69e177)\r\n #8 0x000000011209545a llvm::InstCombinerImpl::SimplifyDemandedUseBits(llvm::Instruction*, llvm::APInt const&, llvm::KnownBits&, unsigned int, llvm::SimplifyQuery const&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x246145a)\r\n #9 0x0000000112097524 llvm::InstCombinerImpl::SimplifyDemandedInstructionBits(llvm::Instruction&, llvm::KnownBits&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x2463524)\r\n#10 0x0000000111ffc58b llvm::InstCombinerImpl::visitCallInst(llvm::CallInst&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x23c858b)\r\n#11 0x0000000111f85788 llvm::InstCombinerImpl::run() (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x2351788)\r\n#12 0x0000000111f765e2 combineInstructionsOverFunction(llvm::Function&, llvm::InstructionWorklist&, llvm::AAResults*, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::BranchProbabilityInfo*, llvm::ProfileSummaryInfo*, llvm::InstCombineOptions const&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x23425e2)\r\n#13 0x0000000111fc1b85 llvm::InstCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x238db85)\r\n#14 0x0000000110596272 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x962272)\r\n#15 0x0000000112b6736a llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x2f3336a)\r\n#16 0x000000011003f522 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x40b522)\r\n#17 0x00000001122fe7db llvm::CGSCCToFunctionPassAdaptor::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x26ca7db)\r\n#18 0x0000000110b41712 llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0xf0d712)\r\n#19 0x00000001122fa573 llvm::PassManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x26c6573)\r\n#20 0x0000000110b3e332 llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>, llvm::AnalysisManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0xf0a332)\r\n#21 0x00000001122fcc0e llvm::DevirtSCCRepeatedPass::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x26c8c0e)\r\n#22 0x0000000110b413a2 llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0xf0d3a2)\r\n#23 0x00000001122fbe83 llvm::ModuleToPostOrderCGSCCPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x26c7e83)\r\n#24 0x0000000110b3e012 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0xf0a012)\r\n#25 0x0000000112b65a22 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x2f31a22)\r\n#26 0x0000000110d25d8b llvm::ModuleInlinerWrapperPass::run(llvm::Module&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x10f1d8b)\r\n#27 0x0000000110594012 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x960012)\r\n#28 0x0000000112b65a22 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x2f31a22)\r\n#29 0x000000010788b478 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::__1::unique_ptr>&, std::__1::unique_ptr>&, clang::BackendConsumer*) (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x172d478)\r\n#30 0x0000000107883b08 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::__1::unique_ptr>, clang::BackendConsumer*) (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x1725b08)\r\n#31 0x0000000107d037d3 clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x1ba57d3)\r\n#32 0x000000010635df1d clang::ParseAST(clang::Sema&, bool, bool) (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x1fff1d)\r\n#33 0x000000010877f3d7 clang::FrontendAction::Execute() (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x26213d7)\r\n#34 0x00000001086e0500 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x2582500)\r\n#35 0x00000001087f60e9 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x26980e9)\r\n#36 0x000000010128d4f9 cc1_main(llvm::ArrayRef, char const*, void*) (/opt/llvm-project@HEAD/bin/clang-20+0x1000084f9)\r\n#37 0x0000000101289f88 ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) (/opt/llvm-project@HEAD/bin/clang-20+0x100004f88)\r\n#38 0x00000001082b552e void llvm::function_ref::callback_fn>, std::__1::basic_string, std::__1::allocator>*, bool*) const::$_0>(long) (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x215752e)\r\n#39 0x0000000112df288c llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/opt/llvm-project@HEAD/lib/libLLVM.dylib+0x31be88c)\r\n#40 0x00000001082b5431 clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__1::basic_string, std::__1::allocator>*, bool*) const (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x2157431)\r\n#41 0x0000000108262131 clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x2104131)\r\n#42 0x00000001082623bf clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x21043bf)\r\n#43 0x00000001082b00c0 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/opt/llvm-project@HEAD/lib/libclang-cpp.dylib+0x21520c0)\r\n#44 0x00000001012892b3 clang_main(int, char**, llvm::ToolContext const&) (/opt/llvm-project@HEAD/bin/clang-20+0x1000042b3)\r\n#45 0x00000001012984cb main (/opt/llvm-project@HEAD/bin/clang-20+0x1000134cb)\r\n#46 0x00007ff81a82a345 \r\nclang: error: clang frontend command failed with exit code 139 (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git c25c6c32494c8d1038438b6208d42ba40f25270e)\r\nTarget: x86_64-apple-darwin23.6.0\r\nThread model: posix\r\nInstalledDir: /opt/llvm-project@HEAD/bin\r\nclang: note: diagnostic msg: \r\n********************\r\n```\r\n\r\nFind attached the preprocessed source code at [gc-add69e.c.gz](https://github.com/user-attachments/files/17813033/gc-add69e.c.gz).", "author": "d125q", "labels": ["llvm:instcombine", "crash-on-valid"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://godbolt.org/z/Ts478aen6\r\n```\r\n; bin/opt -passes=instcombine test.ll -S\r\n@GC_arrays = external global { i8, i8, i64 }\r\n\r\ndefine ptr @GC_push_finalizer_structures() {\r\nentry:\r\n %aligned_result.i = call ptr @llvm.ptrmask.p0.i64(ptr getelementptr inbounds (i8, ptr @GC_arrays, i64 1), i64 -8)\r\n ret ptr %aligned_result.i\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = llvm::GetElementPtrInst; From = llvm::Value]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=instcombine \r\n1.\tRunning pass \"function(instcombine)\" on module \"\"\r\n2.\tRunning pass \"instcombine\" on function \"GC_push_finalizer_structures\"\r\n #0 0x0000000005194b38 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x5194b38)\r\n #1 0x000000000519252c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000074b7d8442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000074b7d84969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000074b7d8442476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000074b7d84287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000074b7d842871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000074b7d8439e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x00000000040d9310 decltype(auto) llvm::cast(llvm::Value*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40d9310)\r\n #9 0x00000000040e2f9e llvm::InstCombinerImpl::SimplifyDemandedUseBits(llvm::Instruction*, llvm::APInt const&, llvm::KnownBits&, unsigned int, llvm::SimplifyQuery const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40e2f9e)\r\n#10 0x00000000040e6082 llvm::InstCombinerImpl::SimplifyDemandedInstructionBits(llvm::Instruction&, llvm::KnownBits&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40e6082)\r\n#11 0x00000000040355a6 llvm::InstCombinerImpl::visitCallInst(llvm::CallInst&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40355a6)\r\n#12 0x0000000003fe2311 llvm::InstCombinerImpl::run() (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3fe2311)\r\n#13 0x0000000003fe3cc6 combineInstructionsOverFunction(llvm::Function&, llvm::InstructionWorklist&, llvm::AAResults*, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::BranchProbabilityInfo*, llvm::ProfileSummaryInfo*, llvm::InstCombineOptions const&) (.isra.0) InstructionCombining.cpp:0:0\r\n#14 0x0000000003fe504e llvm::InstCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3fe504e)\r\n#15 0x0000000002c5c24e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2c5c24e)\r\n#16 0x0000000004f949e8 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f949e8)\r\n#17 0x0000000000e4027e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe4027e)\r\n#18 0x0000000004f9340e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f9340e)\r\n#19 0x0000000000e3f92e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3f92e)\r\n#20 0x0000000004f92e40 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f92e40)\r\n#21 0x0000000000909fda llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x909fda)\r\n#22 0x00000000008fc70a optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fc70a)\r\n#23 0x000074b7d8429d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#24 0x000074b7d8429e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#25 0x00000000008f3e9e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f3e9e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```"}]}, "verified": true} {"bug_id": "112666", "issue_url": "https://github.com/llvm/llvm-project/issues/112666", "bug_type": "miscompilation", "base_commit": "d6d73ec89e493c69cf24dc3a710d861e2ce08acb", "knowledge_cutoff": "2024-10-17T06:28:53Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "ff07df6620c32571c7e13ff96ec7976c63ed0ab8", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp": [[333, 338]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp": ["Negator::visitImpl"]}}, "patch": "commit ff07df6620c32571c7e13ff96ec7976c63ed0ab8\nAuthor: Yingwei Zheng \nDate: Fri Nov 8 16:20:04 2024 +0800\n\n [InstCombine] Drop nsw in negation of select (#112893)\n \n Closes https://github.com/llvm/llvm-project/issues/112666 and\n https://github.com/llvm/llvm-project/issues/114181.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp b/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\nindex 9bd848552615..2210336d92bf 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\n@@ -333,6 +333,17 @@ std::array Negator::getSortedOperandsOfBinOp(Instruction *I) {\n NewSelect->swapValues();\n // Don't swap prof metadata, we didn't change the branch behavior.\n NewSelect->setName(I->getName() + \".neg\");\n+ // Poison-generating flags should be dropped\n+ Value *TV = NewSelect->getTrueValue();\n+ Value *FV = NewSelect->getFalseValue();\n+ if (match(TV, m_Neg(m_Specific(FV))))\n+ cast(TV)->dropPoisonGeneratingFlags();\n+ else if (match(FV, m_Neg(m_Specific(TV))))\n+ cast(FV)->dropPoisonGeneratingFlags();\n+ else {\n+ cast(TV)->dropPoisonGeneratingFlags();\n+ cast(FV)->dropPoisonGeneratingFlags();\n+ }\n Builder.Insert(NewSelect);\n return NewSelect;\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/sub-of-negatible.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "negate_select_of_op_vs_negated_op_nsw_commuted", "test_body": "define i8 @negate_select_of_op_vs_negated_op_nsw_commuted(i8 %x, i8 %y, i1 %c) {\n %t0 = sub nsw i8 0, %x\n %t1 = select i1 %c, i8 %x, i8 %t0\n %t2 = sub i8 %y, %t1\n ret i8 %t2\n}\n"}, {"test_name": "negate_select_of_op_vs_negated_op_nsw", "test_body": "define i8 @negate_select_of_op_vs_negated_op_nsw(i8 %x, i8 %y, i1 %c) {\n %t0 = sub nsw i8 0, %x\n %t1 = select i1 %c, i8 %t0, i8 %x\n %t2 = sub i8 %y, %t1\n ret i8 %t2\n}\n"}, {"test_name": "dont_negate_ordinary_select", "test_body": "define i8 @dont_negate_ordinary_select(i8 %x, i8 %y, i8 %z, i1 %c) {\n %t0 = select i1 %c, i8 %x, i8 %y\n %t1 = sub i8 %z, %t0\n ret i8 %t1\n}\n"}, {"test_name": "negate_select_of_op_vs_negated_op_nsw_xyyx", "test_body": "define i8 @negate_select_of_op_vs_negated_op_nsw_xyyx(i8 %x, i8 %y, i8 %z, i1 %c) {\n %sub1 = sub nsw i8 %x, %y\n %sub2 = sub nsw i8 %y, %x\n %t1 = select i1 %c, i8 %sub1, i8 %sub2\n %t2 = sub i8 %z, %t1\n ret i8 %t2\n}\n"}, {"test_name": "negate_select_of_op_vs_negated_op", "test_body": "declare void @use8(i8)\n\ndefine i8 @negate_select_of_op_vs_negated_op(i8 %x, i8 %y, i1 %c) {\n %t0 = sub i8 0, %x\n call void @use8(i8 %t0)\n %t1 = select i1 %c, i8 %t0, i8 %x, !prof !0\n %t2 = sub i8 %y, %t1\n ret i8 %t2\n}\n\n!0 = !{!\"branch_weights\", i32 40, i32 1}\n"}]}], "issue": {"title": "[InstCombine] `nsw` flag should be dropped in Negator", "body": "Reproducer: https://alive2.llvm.org/ce/z/ZFF_QK\r\n```\r\ndefine i8 @src(i8 %x, i8 %y, i1 %c) {\r\n %t0 = sub nsw i8 0, %x\r\n %t1 = select i1 %c, i8 %t0, i8 %x\r\n %t2 = sub i8 %y, %t1\r\n ret i8 %t2\r\n}\r\n\r\ndefine i8 @tgt(i8 %x, i8 %y, i1 %c) {\r\n %t0 = sub nsw i8 0, %x\r\n %1 = select i1 %c, i8 %x, i8 %t0\r\n %t2 = add i8 %1, %y\r\n ret i8 %t2\r\n}\r\n```\r\n```\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\ni8 %x = #x80 (128, -128)\r\ni8 %y = #x00 (0)\r\ni1 %c = #x0 (0)\r\n\r\nSource:\r\ni8 %t0 = poison\r\ni8 %t1 = #x80 (128, -128)\r\ni8 %t2 = #x80 (128, -128)\r\n\r\nTarget:\r\ni8 %t0 = poison\r\ni8 %#1 = poison\r\ni8 %t2 = poison\r\nSource value: #x80 (128, -128)\r\nTarget value: poison\r\n```", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "AZero13", "body": "Is this an issue in 19.x or just main?"}, {"author": "AZero13", "body": "Confirmed this issue exists in 19.x"}]}, "verified": true} {"bug_id": "105988", "issue_url": "https://github.com/llvm/llvm-project/issues/105988", "bug_type": "miscompilation", "base_commit": "4b0c0ec6b8065e611640f44adce94e2da12b3a3a", "knowledge_cutoff": "2024-08-25T14:54:29Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "e1d22512906e69846c8f6a2d29b30832b7c12b46", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[15975, 15980]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::computeMinimumValueSizes"]}}, "patch": "commit e1d22512906e69846c8f6a2d29b30832b7c12b46\nAuthor: Alexey Bataev \nDate: Mon Aug 26 11:30:13 2024 -0700\n\n [SLP]Fix minbitwidth analysis for gather nodes with icmp users.\n \n If the node is not in MinBWs container and the user node is icmp node,\n the compiler should not check the type size of the user instruction, it\n is always 1 and is not good for actual bitwidth analysis.\n \n Fixes https://github.com/llvm/llvm-project/issues/105988\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex def73e8d0c0d..ed47ed661ab9 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -15975,6 +15975,10 @@ void BoUpSLP::computeMinimumValueSizes() {\n auto It = MinBWs.find(TE);\n if (It != MinBWs.end() && It->second.first > UserTESz)\n return true;\n+ // The size of icmp is always 1 and should not be\n+ // considered.\n+ if (TE->getOpcode() == Instruction::ICmp)\n+ return true;\n return DL->getTypeSizeInBits(U->getType()) > UserTESz;\n }));\n })) {\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/gather-with-cmp-user.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define i1 @test(i32 %g, i16 %d) {\nentry:\n %0 = and i16 %d, 1\n %xor.i.i = xor i32 %g, 1\n %conv1.i.i = trunc i32 %xor.i.i to i8\n %notsub.i = add i8 %conv1.i.i, -1\n %cmp.i.i = icmp sgt i8 %notsub.i, -3\n %conv3.i.i = zext i1 %cmp.i.i to i32\n %cmp4.i.i = icmp sgt i32 %xor.i.i, %conv3.i.i\n %conv1.1.i.i = trunc i32 %g to i8\n %notsub25.i = add i8 %conv1.1.i.i, -1\n %cmp.1.i.i = icmp sgt i8 %notsub25.i, -3\n %conv3.1.i.i = zext i1 %cmp.1.i.i to i32\n %cmp4.1.i.i = icmp sgt i32 %g, %conv3.1.i.i\n %notsub26.i = add i8 %conv1.1.i.i, -9\n %cmp.i17.i = icmp sgt i8 %notsub26.i, -3\n %conv3.i18.i = zext i1 %cmp.i17.i to i32\n %cmp4.i19.i = icmp sgt i32 %g, %conv3.i18.i\n %notsub27.i = add i8 %conv1.i.i, -9\n %cmp.1.i22.i = icmp sgt i8 %notsub27.i, -3\n %conv3.1.i23.i = zext i1 %cmp.1.i22.i to i32\n %cmp4.1.i24.i = icmp sgt i32 %xor.i.i, %conv3.1.i23.i\n %1 = and i1 %cmp4.i19.i, %cmp4.1.i24.i\n %2 = and i1 %cmp4.i.i, %1\n %3 = and i1 %cmp4.1.i.i, %2\n ret i1 %3\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Miscompile at O3", "body": "Reproducer: https://gist.github.com/dtcxzyw/117dc735db97658ea852d30fd896242a\r\n```console\r\n> bin/clang -O0 test.c -w -I/usr/include/csmith && ./a.out\r\nchecksum = BDCB7A78\r\n> bin/clang -O3 test.c -w -mllvm -inline-threshold=100000 -I/usr/include/csmith && ./a.out\r\nchecksum = 9198FC41\r\n```\r\n\r\nllvm version: 1193f7d6487d2d94009f8d8d27da3907136482b9\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/GCXoe9\r\n```\r\n; bin/opt -passes=slp-vectorizer test.ll -S\r\n\r\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i1 @main(i32 %g.promoted.i.i, i16 %d.promoted.i.i) {\r\nentry:\r\n %0 = and i16 %d.promoted.i.i, 1\r\n %xor.i.i = xor i32 %g.promoted.i.i, 1\r\n %conv1.i.i = trunc i32 %xor.i.i to i8\r\n %notsub.i = add i8 %conv1.i.i, -1\r\n %cmp.i.i = icmp sgt i8 %notsub.i, -3\r\n %conv3.i.i = zext i1 %cmp.i.i to i32\r\n %cmp4.i.i = icmp sgt i32 %xor.i.i, %conv3.i.i\r\n %conv1.1.i.i = trunc i32 %g.promoted.i.i to i8\r\n %notsub25.i = add i8 %conv1.1.i.i, -1\r\n %cmp.1.i.i = icmp sgt i8 %notsub25.i, -3\r\n %conv3.1.i.i = zext i1 %cmp.1.i.i to i32\r\n %cmp4.1.i.i = icmp sgt i32 %g.promoted.i.i, %conv3.1.i.i\r\n %notsub26.i = add i8 %conv1.1.i.i, -9\r\n %cmp.i17.i = icmp sgt i8 %notsub26.i, -3\r\n %conv3.i18.i = zext i1 %cmp.i17.i to i32\r\n %cmp4.i19.i = icmp sgt i32 %g.promoted.i.i, %conv3.i18.i\r\n %notsub27.i = add i8 %conv1.i.i, -9\r\n %cmp.1.i22.i = icmp sgt i8 %notsub27.i, -3\r\n %conv3.1.i23.i = zext i1 %cmp.1.i22.i to i32\r\n %cmp4.1.i24.i = icmp sgt i32 %xor.i.i, %conv3.1.i23.i\r\n %1 = and i1 %cmp4.i19.i, %cmp4.1.i24.i\r\n %2 = and i1 %cmp4.i.i, %1\r\n %3 = and i1 %cmp4.1.i.i, %2\r\n ret i1 %3\r\n}\r\n```\r\nOutput:\r\n```\r\ndefine i1 @main(i32 %g.promoted.i.i, i16 %d.promoted.i.i) {\r\nentry:\r\n %0 = and i16 %d.promoted.i.i, 1\r\n %xor.i.i = xor i32 %g.promoted.i.i, 1\r\n %1 = trunc i32 %g.promoted.i.i to i8\r\n %2 = insertelement <2 x i8> poison, i8 %1, i32 0\r\n %3 = trunc i32 %xor.i.i to i8\r\n %4 = insertelement <2 x i8> %2, i8 %3, i32 1\r\n %5 = shufflevector <2 x i8> %4, <2 x i8> poison, <4 x i32> \r\n %6 = add <4 x i8> %5, \r\n %7 = icmp sgt <4 x i8> %6, \r\n %8 = zext <4 x i1> %7 to <4 x i8>\r\n %9 = sext <2 x i8> %4 to <2 x i32>\r\n %10 = shufflevector <2 x i32> %9, <2 x i32> poison, <4 x i32> \r\n %11 = zext <4 x i8> %8 to <4 x i32>\r\n %12 = icmp sgt <4 x i32> %10, %11\r\n %13 = call i1 @llvm.vector.reduce.and.v4i1(<4 x i1> %12)\r\n ret i1 %13\r\n}\r\n```\r\n\r\ncc @alexey-bataev "}]}, "verified": true} {"bug_id": "100298", "issue_url": "https://github.com/llvm/llvm-project/issues/100298", "bug_type": "miscompilation", "base_commit": "75642a00e15b722bdfb90726be31f1c8adaeb0c5", "knowledge_cutoff": "2024-07-24T04:19:53Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "59eae919c938f890e9b9b4be8a3fa3cb1b11ed89", "components": ["ValueTracking"], "bug_location_lineno": {"llvm/include/llvm/Analysis/SimplifyQuery.h": [[130, 135]], "llvm/lib/Analysis/ValueTracking.cpp": [[1435, 1441], [1468, 1474], [1546, 1552], [2329, 2335], [2943, 2949], [3509, 3515], [4001, 4007], [5909, 5918]]}, "bug_location_funcname": {"llvm/include/llvm/Analysis/SimplifyQuery.h": ["getWithCondContext"], "llvm/lib/Analysis/ValueTracking.cpp": ["ComputeNumSignBitsImpl", "computeKnownBitsFromOperator", "computeKnownFPClass", "isKnownNonZeroFromOperator", "isKnownToBeAPowerOfTwo", "isNonEqualPHIs"]}}, "patch": "commit a87fbeb3a77a53ded341277c5b326f7696d47594\nAuthor: Yingwei Zheng \nDate: Wed Jul 24 20:06:36 2024 +0800\n\n [ValueTracking] Don't use CondContext in dataflow analysis of phi nodes (#100316)\n \n See the following case:\n ```\n define i16 @pr100298() {\n entry:\n br label %for.inc\n \n for.inc:\n %indvar = phi i32 [ -15, %entry ], [ %mask, %for.inc ]\n %add = add nsw i32 %indvar, 9\n %mask = and i32 %add, 65535\n %cmp1 = icmp ugt i32 %mask, 5\n br i1 %cmp1, label %for.inc, label %for.end\n \n for.end:\n %conv = trunc i32 %add to i16\n %cmp2 = icmp ugt i32 %mask, 3\n %shl = shl nuw i16 %conv, 14\n %res = select i1 %cmp2, i16 %conv, i16 %shl\n ret i16 %res\n }\n ```\n \n When computing knownbits of `%shl` with `%cmp2=false`, we cannot use\n this condition in the analysis of `%mask (%for.inc -> %for.inc)`.\n \n Fixes https://github.com/llvm/llvm-project/issues/100298.\n \n (cherry picked from commit 59eae919c938f890e9b9b4be8a3fa3cb1b11ed89)\n\ndiff --git a/llvm/include/llvm/Analysis/SimplifyQuery.h b/llvm/include/llvm/Analysis/SimplifyQuery.h\nindex a560744f0122..e8f43c8c2e91 100644\n--- a/llvm/include/llvm/Analysis/SimplifyQuery.h\n+++ b/llvm/include/llvm/Analysis/SimplifyQuery.h\n@@ -130,6 +130,12 @@ struct SimplifyQuery {\n Copy.CC = &CC;\n return Copy;\n }\n+\n+ SimplifyQuery getWithoutCondContext() const {\n+ SimplifyQuery Copy(*this);\n+ Copy.CC = nullptr;\n+ return Copy;\n+ }\n };\n \n } // end namespace llvm\ndiff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp\nindex 40fe1ffe13f1..4b77c0046cc7 100644\n--- a/llvm/lib/Analysis/ValueTracking.cpp\n+++ b/llvm/lib/Analysis/ValueTracking.cpp\n@@ -1435,7 +1435,7 @@ static void computeKnownBitsFromOperator(const Operator *I,\n // inferred hold at original context instruction. TODO: It may be\n // correct to use the original context. IF warranted, explore and\n // add sufficient tests to cover.\n- SimplifyQuery RecQ = Q;\n+ SimplifyQuery RecQ = Q.getWithoutCondContext();\n RecQ.CxtI = P;\n computeKnownBits(R, DemandedElts, Known2, Depth + 1, RecQ);\n switch (Opcode) {\n@@ -1468,7 +1468,7 @@ static void computeKnownBitsFromOperator(const Operator *I,\n // phi. This is important because that is where the value is actually\n // \"evaluated\" even though it is used later somewhere else. (see also\n // D69571).\n- SimplifyQuery RecQ = Q;\n+ SimplifyQuery RecQ = Q.getWithoutCondContext();\n \n unsigned OpNum = P->getOperand(0) == R ? 0 : 1;\n Instruction *RInst = P->getIncomingBlock(OpNum)->getTerminator();\n@@ -1546,7 +1546,7 @@ static void computeKnownBitsFromOperator(const Operator *I,\n // phi. This is important because that is where the value is actually\n // \"evaluated\" even though it is used later somewhere else. (see also\n // D69571).\n- SimplifyQuery RecQ = Q;\n+ SimplifyQuery RecQ = Q.getWithoutCondContext();\n RecQ.CxtI = P->getIncomingBlock(u)->getTerminator();\n \n Known2 = KnownBits(BitWidth);\n@@ -2329,7 +2329,7 @@ bool isKnownToBeAPowerOfTwo(const Value *V, bool OrZero, unsigned Depth,\n // it is an induction variable where in each step its value is a power of\n // two.\n auto *PN = cast(I);\n- SimplifyQuery RecQ = Q;\n+ SimplifyQuery RecQ = Q.getWithoutCondContext();\n \n // Check if it is an induction variable and always power of two.\n if (isPowerOfTwoRecurrence(PN, OrZero, Depth, RecQ))\n@@ -2943,7 +2943,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,\n return true;\n \n // Check if all incoming values are non-zero using recursion.\n- SimplifyQuery RecQ = Q;\n+ SimplifyQuery RecQ = Q.getWithoutCondContext();\n unsigned NewDepth = std::max(Depth, MaxAnalysisRecursionDepth - 1);\n return llvm::all_of(PN->operands(), [&](const Use &U) {\n if (U.get() == PN)\n@@ -3509,7 +3509,7 @@ static bool isNonEqualPHIs(const PHINode *PN1, const PHINode *PN2,\n if (UsedFullRecursion)\n return false;\n \n- SimplifyQuery RecQ = Q;\n+ SimplifyQuery RecQ = Q.getWithoutCondContext();\n RecQ.CxtI = IncomBB->getTerminator();\n if (!isKnownNonEqual(IV1, IV2, DemandedElts, Depth + 1, RecQ))\n return false;\n@@ -4001,7 +4001,7 @@ static unsigned ComputeNumSignBitsImpl(const Value *V,\n \n // Take the minimum of all incoming values. This can't infinitely loop\n // because of our depth threshold.\n- SimplifyQuery RecQ = Q;\n+ SimplifyQuery RecQ = Q.getWithoutCondContext();\n Tmp = TyBits;\n for (unsigned i = 0, e = NumIncomingValues; i != e; ++i) {\n if (Tmp == 1) return Tmp;\n@@ -5909,10 +5909,10 @@ void computeKnownFPClass(const Value *V, const APInt &DemandedElts,\n // Recurse, but cap the recursion to two levels, because we don't want\n // to waste time spinning around in loops. We need at least depth 2 to\n // detect known sign bits.\n- computeKnownFPClass(\n- IncValue, DemandedElts, InterestedClasses, KnownSrc,\n- PhiRecursionLimit,\n- Q.getWithInstruction(P->getIncomingBlock(U)->getTerminator()));\n+ computeKnownFPClass(IncValue, DemandedElts, InterestedClasses, KnownSrc,\n+ PhiRecursionLimit,\n+ Q.getWithoutCondContext().getWithInstruction(\n+ P->getIncomingBlock(U)->getTerminator()));\n \n if (First) {\n Known = KnownSrc;\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/pr100298.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "pr100298", "test_body": "define i16 @pr100298() {\nentry:\n br label %for.inc\n\nfor.inc: ; preds = %for.inc, %entry\n %indvar = phi i32 [ -15, %entry ], [ %mask, %for.inc ]\n %add = add nsw i32 %indvar, 9\n %mask = and i32 %add, 65535\n %cmp1 = icmp ugt i32 %mask, 5\n br i1 %cmp1, label %for.inc, label %for.end\n\nfor.end: ; preds = %for.inc\n %conv = trunc i32 %add to i16\n %cmp2 = icmp ugt i32 %mask, 3\n %shl = shl nuw i16 %conv, 14\n %res = select i1 %cmp2, i16 %conv, i16 %shl\n ret i16 %res\n}\n"}]}], "issue": {"title": "Miscompilation at O3", "body": "```\r\n#include \"csmith.h\"\r\nuint32_t a;\r\nint16_t c;\r\nvoid d() {\r\n for (a = -15; a >= 6; a = safe_add_func_uint16_t_u_u(a, 9))\r\n ;\r\n c = safe_lshift_func_int16_t_s_u(safe_lshift_func_uint16_t_u_s(a,14), 0);\r\n}\r\nint main() {\r\n d();\r\n printf(\"%d\\n\", c);\r\n return 0;\r\n}\r\n```\r\n\r\n```\r\ndtcxzyw@dtcxzyw:~/WorkSpace/Projects/compilers/LLVM/llvm-build$ gcc -O0 test.c -fsanitize=undefined -I/usr/include/csmith -w\r\ndtcxzyw@dtcxzyw:~/WorkSpace/Projects/compilers/LLVM/llvm-build$ ./a.out \r\n-16384\r\ndtcxzyw@dtcxzyw:~/WorkSpace/Projects/compilers/LLVM/llvm-build$ bin/clang -O3 test.c -I/usr/include/csmith -w\r\ndtcxzyw@dtcxzyw:~/WorkSpace/Projects/compilers/LLVM/llvm-build$ ./a.out \r\n0\r\n```\r\n\r\nllvm version: 2d6ff0c533aa1b0fe6d91764846f636c4a4e911c", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:analysis"], "comments": [{"author": "dtcxzyw", "body": "Reduced testcase: https://alive2.llvm.org/ce/z/9fExrA\r\n```\r\n; bin/opt -passes=instcombine test.ll -S\r\ndefine i16 @src() {\r\nentry:\r\n br label %for.inc\r\n\r\nfor.inc:\r\n %storemerge6 = phi i32 [ -15, %entry ], [ %conv1, %for.inc ]\r\n %add.i = add nsw i32 %storemerge6, 9\r\n %conv1 = and i32 %add.i, 65535\r\n %cmp = icmp ugt i32 %conv1, 5\r\n br i1 %cmp, label %for.inc, label %for.end\r\n\r\nfor.end:\r\n %conv2 = trunc nuw nsw i32 %conv1 to i16\r\n %cmp3.i = icmp ugt i32 %conv1, 3\r\n %0 = shl nuw i16 %conv2, 14\r\n %cond.in.i = select i1 %cmp3.i, i16 %conv2, i16 %0\r\n ret i16 %cond.in.i\r\n}\r\n```"}]}, "verified": true} {"bug_id": "96294", "issue_url": "https://github.com/llvm/llvm-project/issues/96294", "bug_type": "crash", "base_commit": "5262865aac683b72f3e66de7a122e0c455ab6b9b", "knowledge_cutoff": "2024-06-21T10:15:21Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "fdb9f96fa2a926425bdf8315048db7623d63547d", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6723, 6736], [6771, 6776], [6781, 6788]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::collectValuesToIgnore"]}}, "patch": "commit fdb9f96fa2a926425bdf8315048db7623d63547d\nAuthor: Florian Hahn \nDate: Sun Aug 4 20:54:25 2024 +0100\n\n [LV] Consider earlier stores to invariant reduction address as dead.\n \n For invariant stores to an address of a reduction, only the latest store\n will be generated outside the loop. Consider earlier stores as dead.\n \n This fixes a difference between the legacy and VPlan-based cost model.\n \n Fixes https://github.com/llvm/llvm-project/issues/96294.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 40d10395de17..1eca9b10fad5 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6723,14 +6723,18 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n return RequiresScalarEpilogue &&\n !TheLoop->contains(cast(U)->getParent());\n };\n+ MapVector> DeadInvariantStoreOps;\n for (BasicBlock *BB : TheLoop->blocks())\n for (Instruction &I : *BB) {\n // Find all stores to invariant variables. Since they are going to sink\n // outside the loop we do not need calculate cost for them.\n StoreInst *SI;\n if ((SI = dyn_cast(&I)) &&\n- Legal->isInvariantAddressOfReduction(SI->getPointerOperand()))\n+ Legal->isInvariantAddressOfReduction(SI->getPointerOperand())) {\n ValuesToIgnore.insert(&I);\n+ auto I = DeadInvariantStoreOps.insert({SI->getPointerOperand(), {}});\n+ I.first->second.push_back(SI->getValueOperand());\n+ }\n \n if (VecValuesToIgnore.contains(&I) || ValuesToIgnore.contains(&I))\n continue;\n@@ -6771,6 +6775,10 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n DeadInterleavePointerOps.append(Op->op_begin(), Op->op_end());\n }\n \n+ for (const auto &[_, Ops] : DeadInvariantStoreOps) {\n+ for (Value *Op : ArrayRef(Ops).drop_back())\n+ DeadOps.push_back(Op);\n+ }\n // Mark ops that would be trivially dead and are only used by ignored\n // instructions as free.\n BasicBlock *Header = TheLoop->getHeader();\n@@ -6781,8 +6789,8 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n (isa(Op) && Op->getParent() == Header) ||\n !wouldInstructionBeTriviallyDead(Op, TLI) ||\n any_of(Op->users(), [this, IsLiveOutDead](User *U) {\n- return !VecValuesToIgnore.contains(U) && ValuesToIgnore.contains(U) &&\n- !IsLiveOutDead(U);\n+ return !VecValuesToIgnore.contains(U) &&\n+ !ValuesToIgnore.contains(U) && !IsLiveOutDead(U);\n }))\n continue;\n \n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/cost-model.ll", "commands": ["opt < %s -passes=loop-vectorize -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -S"], "tests": [{"test_name": "reduction_store", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine void @reduction_store(ptr noalias %src, ptr %dst, i1 %x) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %red = phi i32 [ 0, %entry ], [ %red.next, %loop ]\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src = getelementptr inbounds i32, ptr %src, i32 %iv\n %l = load i32, ptr %gep.src, align 4\n %l.and = and i32 %l, 3\n store i32 %l.and, ptr %dst, align 4\n %x.ext = zext i1 %x to i64\n %lshr = lshr i64 %x.ext, 12\n %t = trunc i64 %lshr to i32\n %red.next = and i32 %red, %t\n store i32 %red.next, ptr %dst, align 4\n %iv.next = add i32 %iv, 1\n %ec = icmp eq i32 %iv, 29\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"znver3\" }\n"}]}], "issue": {"title": "llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10353: bool llvm::LoopVectorizePass::processLoop(llvm::Loop*): Assertion `VF.Width == Width && \"VPlan cost model and legacy cost model disagreed\"' failed.", "body": "[calcc-0ebd65.c.gz](https://github.com/user-attachments/files/15926158/calcc-0ebd65.c.gz)\r\n[calcc-0ebd65.sh.gz](https://github.com/user-attachments/files/15926159/calcc-0ebd65.sh.gz)\r\n\r\nStack backtrace is \r\n\r\n#11 0x0000000003adfddd llvm::LoopVectorizePass::processLoop(llvm::Loop*) /home/dcb40b/llvm/trunk/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10353:9\r\n#12 0x0000000003ae347f llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) /home/dcb40b/llvm/trunk/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10446:27\r\n#13 0x0000000003ae4382 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) /home/dcb40b/llvm/trunk/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10486:5\r\n#14 0x00000000034fbd6d llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) /home/dcb40b/llvm/trunk/llvm/include/llvm/IR/PassManagerInternal.h:91:3\r\n#15 0x0000000001ae5b21 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) /home/dcb40b/llvm/trunk/llvm/include/llvm/IR/PassManagerImpl.h:53:18\r\n", "author": "dcb314", "labels": ["vectorizers", "crash-on-valid"], "comments": [{"author": "dcb314", "body": "The crash first seems to occur sometime between git hash fd88089151e66a4cb1d90aaa224e4cb4e7a748f4 from yesterday 20240620\r\nand 225d8fc8eb24fb797154c1ef6dcbe5ba033142da from today,\r\nwhich is 134 commits.\r\n"}, {"author": "fhahn", "body": "Thanks for the report, caused by https://github.com/llvm/llvm-project/pull/92555 which has been reverted for now. I'll add a reduced test based on the reproducer."}, {"author": "fhahn", "body": "https://github.com/llvm/llvm-project/pull/92555 has been recommitted with a fix. Please let me know if you see any other issues!"}, {"author": "dcb314", "body": "It has come back:\r\n\r\nclang: /home/dcb40b/llvm/trunk/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10377: bool llvm::LoopVectorizePass::processLoop(llvm::Loop*): Assertion `VF.Width == Width && \"VPlan cost model and legacy cost model disagreed\"' failed.\r\n\r\n[in-4e39b8.sh.gz](https://github.com/user-attachments/files/16445872/in-4e39b8.sh.gz)\r\n[Uploading in-4e39b8.c.gz…]()\r\n"}, {"author": "dcb314", "body": "Seems to have come back sometime between d99efd53f5103df4db517f499fcdd26cd3b080a3\r\n(dated 20240710) and 4710e0f498cb661ca17c99cb174616102fcad923, dated 20240711.\r\n\r\n"}, {"author": "fhahn", "body": "@dcb314 it looks like there has been an issue with uploading the `.c` file. Could you re-upload it again?"}, {"author": "dcb314", "body": "My apologies for the failed upload.\r\nHere is another try:\r\n\r\n[in-4e39b8.c.gz](https://github.com/user-attachments/files/16455342/in-4e39b8.c.gz)\r\n"}]}, "verified": true} {"bug_id": "116375", "issue_url": "https://github.com/llvm/llvm-project/issues/116375", "bug_type": "crash", "base_commit": "07507cb5919cae0ae880bfee538ebc993b97dd6c", "knowledge_cutoff": "2024-11-15T11:48:16Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "320038579d3c23b78f99618b71640f51423fe321", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[3317, 3322], [3325, 3331]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPFirstOrderRecurrencePHIRecipe::computeCost"]}}, "patch": "commit 320038579d3c23b78f99618b71640f51423fe321\nAuthor: Florian Hahn \nDate: Thu Nov 21 21:11:20 2024 +0000\n\n [VPlan] Return cost of PHI for scalar VFs in computeCost for FORs.\n \n This fixes a crash when the VF is scalar.\n \n Fixes https://github.com/llvm/llvm-project/issues/116375.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex 71aca3be9e5d..24cf4666c62c 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -3317,6 +3317,10 @@ void VPFirstOrderRecurrencePHIRecipe::execute(VPTransformState &State) {\n InstructionCost\n VPFirstOrderRecurrencePHIRecipe::computeCost(ElementCount VF,\n VPCostContext &Ctx) const {\n+ TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n+ if (VF.isScalar())\n+ return Ctx.TTI.getCFInstrCost(Instruction::PHI, CostKind);\n+\n if (VF.isScalable() && VF.getKnownMinValue() == 1)\n return InstructionCost::getInvalid();\n \n@@ -3325,7 +3329,6 @@ VPFirstOrderRecurrencePHIRecipe::computeCost(ElementCount VF,\n Type *VectorTy =\n ToVectorTy(Ctx.Types.inferScalarType(this->getVPSingleValue()), VF);\n \n- TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n return Ctx.TTI.getShuffleCost(TargetTransformInfo::SK_Splice,\n cast(VectorTy), Mask, CostKind,\n VF.getKnownMinValue() - 1);\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/invalid-costs.ll", "commands": ["opt -passes=\"loop-vectorize\" -pass-remarks-output=/dev/stderr -S %s"], "tests": [{"test_name": "", "test_body": "\n; REMARKS: the cost-model indicates that vectorization is not beneficial\n\n; Test for https://github.com/llvm/llvm-project/issues/116375.\ndefine void @test_i24_load_for(ptr noalias %src, ptr %dst) {\n;\nentry:\n br label %loop\n\nloop:\n %iv = phi i16 [ 0, %entry ], [ %iv.next, %loop ]\n %for = phi i24 [ 0, %entry ], [ %for.next, %loop ]\n %iv.next = add i16 %iv, 1\n %gep.src = getelementptr inbounds i24, ptr %src, i16 %iv\n %for.next = load i24, ptr %gep.src, align 1\n %gep.dst = getelementptr inbounds i24, ptr %dst, i16 %iv\n store i24 %for, ptr %gep.dst\n %ec = icmp eq i16 %iv.next, 1000\n br i1 %ec, label %exit, label %loop\n\nexit:\n ret void\n}"}]}], "issue": {"title": "opt -passes=\"loop-vectorize\" -pass-remarks-output=tmp.opts crashes with 'Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.'", "body": "llvm commit: 10b048c8922d\r\nReproduce with: ```opt -passes=\"loop-vectorize\" bbi-101140.ll -o /dev/null -pass-remarks-output=tmp.opts```\r\nResult:\r\n```\r\nopt: ../include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From *) [To = llvm::VectorType, From = llvm::Type]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: build-all/bin/opt -passes=loop-vectorize bbi-101140.ll -o /dev/null -pass-remarks-output=tmp.opts\r\n1.\tRunning pass \"function(loop-vectorize)\" on module \"bbi-101140.ll\"\r\n2.\tRunning pass \"loop-vectorize\" on function \"main\"\r\n #0 0x000055b814f24938 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (build-all/bin/opt+0x4435938)\r\n #1 0x000055b814f223fe llvm::sys::RunSignalHandlers() (build-all/bin/opt+0x44333fe)\r\n #2 0x000055b814f2516d SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007f5d1b5dbcf0 __restore_rt (/lib64/libpthread.so.0+0x12cf0)\r\n #4 0x00007f5d19194acf raise (/lib64/libc.so.6+0x4eacf)\r\n #5 0x00007f5d19167ea5 abort (/lib64/libc.so.6+0x21ea5)\r\n #6 0x00007f5d19167d79 _nl_load_domain.cold.0 (/lib64/libc.so.6+0x21d79)\r\n #7 0x00007f5d1918d426 (/lib64/libc.so.6+0x47426)\r\n #8 0x000055b816474525 llvm::VPFirstOrderRecurrencePHIRecipe::computeCost(llvm::ElementCount, llvm::VPCostContext&) const (build-all/bin/opt+0x5985525)\r\n #9 0x000055b816461f4c llvm::VPRecipeBase::cost(llvm::ElementCount, llvm::VPCostContext&) (build-all/bin/opt+0x5972f4c)\r\n#10 0x000055b8163faa3a llvm::LoopVectorizationPlanner::emitInvalidCostRemarks(llvm::OptimizationRemarkEmitter*) (build-all/bin/opt+0x590ba3a)\r\n#11 0x000055b816427189 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (build-all/bin/opt+0x5938189)\r\n#12 0x000055b81642cd0b llvm::LoopVectorizePass::runImpl(llvm::Function&) (build-all/bin/opt+0x593dd0b)\r\n#13 0x000055b81642d568 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (build-all/bin/opt+0x593e568)\r\n#14 0x000055b8162e33dd llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) PassBuilderPipelines.cpp:0:0\r\n#15 0x000055b81512d317 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (build-all/bin/opt+0x463e317)\r\n#16 0x000055b8162dcf6d llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) PassBuilderPipelines.cpp:0:0\r\n#17 0x000055b815131ea6 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (build-all/bin/opt+0x4642ea6)\r\n#18 0x000055b8162d6a0d llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) PassBuilderPipelines.cpp:0:0\r\n#19 0x000055b81512c047 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (build-all/bin/opt+0x463d047)\r\n#20 0x000055b81627ab63 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (build-all/bin/opt+0x578bb63)\r\n#21 0x000055b814eebd0a optMain (build-all/bin/opt+0x43fcd0a)\r\n#22 0x00007f5d19180d85 __libc_start_main (/lib64/libc.so.6+0x3ad85)\r\n#23 0x000055b814ee5b2e _start (build-all/bin/opt+0x43f6b2e)\r\nAbort (core dumped)\r\n```\r\n[bbi-101140.ll.gz](https://github.com/user-attachments/files/17774956/bbi-101140.ll.gz)\r\n", "author": "mikaelholmen", "labels": ["vectorizers", "crash"], "comments": [{"author": "mikaelholmen", "body": "This starts happening with https://github.com/llvm/llvm-project/commit/680901ed8010319843cd81275b845d682f77e27f\r\n```\r\n[VPlan] Implement VPHeaderPHIRecipe::computeCost.\r\n\r\nFill out computeCost implementations for various header PHI recipes,\r\nmatching the legacy cost model for now.\r\n```"}, {"author": "mikaelholmen", "body": "Thanks!"}]}, "verified": true} {"bug_id": "106655", "issue_url": "https://github.com/llvm/llvm-project/issues/106655", "bug_type": "crash", "base_commit": "2d5613afec0f4afeeb03cfd4edac556a65ad0eaf", "knowledge_cutoff": "2024-08-30T03:04:59Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "87a988e881ac92e3d87aae01dc632f33c1fb36aa", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[13137, 13143], [13197, 13203], [13249, 13255], [13259, 13265], [13474, 13480], [13518, 13524], [13571, 13577], [13593, 13599], [13611, 13617], [13655, 13661], [13688, 13694], [13780, 13786], [13794, 13800], [13859, 13865], [13941, 13947], [14039, 14044]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree"]}}, "patch": "commit 87a988e881ac92e3d87aae01dc632f33c1fb36aa\nAuthor: Alexey Bataev \nDate: Fri Aug 30 05:16:57 2024 -0700\n\n [SLP]Fix PR106655: Use FinalShuffle for alternate cast nodes.\n \n Need to use FinalShuffle function for all vectorized results to\n correctly produce vectorized value.\n \n Fixes https://github.com/llvm/llvm-project/issues/106655\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex edb2567fa057..345b01b82c6a 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -13137,7 +13137,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n }\n \n bool IsReverseOrder = isReverseOrder(E->ReorderIndices);\n- auto FinalShuffle = [&](Value *V, const TreeEntry *E, VectorType *VecTy) {\n+ auto FinalShuffle = [&](Value *V, const TreeEntry *E) {\n ShuffleInstructionBuilder ShuffleBuilder(ScalarTy, Builder, *this);\n if (E->getOpcode() == Instruction::Store &&\n E->State == TreeEntry::Vectorize) {\n@@ -13197,7 +13197,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n PH->getParent()->getFirstInsertionPt());\n Builder.SetCurrentDebugLocation(PH->getDebugLoc());\n \n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n if (PostponedPHIs)\n@@ -13249,7 +13249,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n if (const TreeEntry *TE = getTreeEntry(V))\n V = TE->VectorizedValue;\n setInsertPointAfterBundle(E);\n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n E->VectorizedValue = V;\n return V;\n }\n@@ -13259,7 +13259,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n Value *Ptr = LI->getPointerOperand();\n LoadInst *V = Builder.CreateAlignedLoad(VecTy, Ptr, LI->getAlign());\n Value *NewV = propagateMetadata(V, E->Scalars);\n- NewV = FinalShuffle(NewV, E, VecTy);\n+ NewV = FinalShuffle(NewV, E);\n E->VectorizedValue = NewV;\n return NewV;\n }\n@@ -13474,7 +13474,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n Value *V = (VecOpcode != ShuffleOrOp && VecOpcode == Instruction::BitCast)\n ? InVec\n : Builder.CreateCast(VecOpcode, InVec, VecTy);\n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n ++NumVectorInstructions;\n@@ -13518,7 +13518,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n propagateIRFlags(V, E->Scalars, VL0);\n // Do not cast for cmps.\n VecTy = cast(V->getType());\n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n ++NumVectorInstructions;\n@@ -13571,7 +13571,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n assert(getNumElements(Cond->getType()) == TrueNumElements &&\n \"Cannot vectorize Instruction::Select\");\n Value *V = Builder.CreateSelect(Cond, True, False);\n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n ++NumVectorInstructions;\n@@ -13593,7 +13593,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n if (auto *I = dyn_cast(V))\n V = propagateMetadata(I, E->Scalars);\n \n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n ++NumVectorInstructions;\n@@ -13611,7 +13611,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n }\n \n Value *V = Builder.CreateFreeze(Op);\n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n ++NumVectorInstructions;\n@@ -13655,7 +13655,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n auto *CI = dyn_cast(Op);\n return CI && CI->getValue().countr_one() >= It->second.first;\n })) {\n- V = FinalShuffle(I == 0 ? RHS : LHS, E, VecTy);\n+ V = FinalShuffle(I == 0 ? RHS : LHS, E);\n E->VectorizedValue = V;\n ++NumVectorInstructions;\n return V;\n@@ -13688,7 +13688,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n I->setHasNoUnsignedWrap(/*b=*/false);\n }\n \n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n ++NumVectorInstructions;\n@@ -13780,7 +13780,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n }\n Value *V = propagateMetadata(NewLI, E->Scalars);\n \n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n E->VectorizedValue = V;\n ++NumVectorInstructions;\n return V;\n@@ -13794,7 +13794,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n if (VecValue->getType() != VecTy)\n VecValue =\n Builder.CreateIntCast(VecValue, VecTy, GetOperandSignedness(0));\n- VecValue = FinalShuffle(VecValue, E, VecTy);\n+ VecValue = FinalShuffle(VecValue, E);\n \n Value *Ptr = SI->getPointerOperand();\n Instruction *ST;\n@@ -13859,7 +13859,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n V = propagateMetadata(I, GEPs);\n }\n \n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n ++NumVectorInstructions;\n@@ -13941,7 +13941,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n Value *V = Builder.CreateCall(CF, OpVecs, OpBundles);\n \n propagateIRFlags(V, E->Scalars, VL0);\n- V = FinalShuffle(V, E, VecTy);\n+ V = FinalShuffle(V, E);\n \n E->VectorizedValue = V;\n ++NumVectorInstructions;\n@@ -14039,6 +14039,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n \"Expected same type as operand.\");\n if (auto *I = dyn_cast(LHS))\n LHS = propagateMetadata(I, E->Scalars);\n+ LHS = FinalShuffle(LHS, E);\n E->VectorizedValue = LHS;\n ++NumVectorInstructions;\n return LHS;\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/resized-alt-shuffle-after-minbw.ll", "commands": ["opt -S --passes=slp-vectorizer -slp-vectorize-hor=false < %s"], "tests": [{"test_name": "func", "test_body": "define void @func(i32 %0) {\n %2 = shl i32 %0, 0\n %3 = sext i32 %2 to i64\n %4 = shl i32 0, 0\n %5 = sext i32 %4 to i64\n %6 = or i32 0, 0\n %7 = or i32 0, 0\n %8 = zext i32 %6 to i64\n %9 = zext i32 %7 to i64\n %10 = zext i32 0 to i64\n %11 = zext i32 0 to i64\n %12 = zext i32 0 to i64\n %13 = zext i32 0 to i64\n %14 = zext i32 0 to i64\n %15 = zext i32 0 to i64\n %16 = zext i32 0 to i64\n %17 = zext i32 0 to i64\n %18 = zext i32 0 to i64\n %19 = zext i32 0 to i64\n %20 = zext i32 0 to i64\n %21 = zext i32 0 to i64\n %22 = zext i32 0 to i64\n %23 = zext i32 0 to i64\n %24 = zext i32 0 to i64\n %25 = zext i32 0 to i64\n %26 = zext i32 0 to i64\n %27 = or i64 %3, 0\n %28 = or i64 %3, %8\n %29 = or i64 %3, %9\n %30 = or i64 %3, %10\n %31 = or i64 %3, %11\n %32 = or i64 %3, %12\n %33 = or i64 %3, %13\n %34 = or i64 %3, %14\n %35 = or i64 %3, %15\n %36 = or i64 %3, %16\n %37 = or i64 %3, %17\n %38 = or i64 %3, %18\n %39 = or i64 %3, %19\n %40 = or i64 %3, %20\n %41 = or i64 %3, %21\n %42 = or i64 %3, %22\n %43 = or i64 %3, %23\n %44 = or i64 %3, %24\n %45 = or i64 %3, %25\n %46 = or i64 %3, 0\n %47 = or i64 %3, 0\n %48 = or i64 %3, 0\n %49 = or i64 %3, 0\n %50 = or i64 %3, 0\n %51 = or i64 %3, 0\n %52 = or i64 %3, 0\n %53 = or i64 %3, 0\n %54 = or i64 %3, 0\n %55 = or i64 %3, 0\n %56 = or i64 %3, 0\n %57 = or i64 %3, 0\n %58 = or i64 %3, 0\n %59 = icmp slt i64 %28, 0\n %60 = icmp slt i64 %29, 0\n %61 = icmp slt i64 %30, 0\n %62 = icmp slt i64 %31, 0\n %63 = icmp slt i64 %32, 0\n %64 = icmp slt i64 %33, 0\n %65 = icmp slt i64 %34, 0\n %66 = icmp slt i64 %35, 0\n %67 = icmp slt i64 %36, 0\n %68 = icmp slt i64 %37, 0\n %69 = icmp slt i64 %38, 0\n %70 = icmp slt i64 %39, 0\n %71 = icmp slt i64 %40, 0\n %72 = icmp slt i64 %41, 0\n %73 = icmp slt i64 %42, 0\n %74 = icmp slt i64 %43, 0\n %75 = icmp slt i64 %44, 0\n %76 = icmp slt i64 %45, 0\n %77 = icmp slt i64 %46, 0\n %78 = icmp slt i64 %47, 0\n %79 = icmp slt i64 %48, 0\n %80 = icmp slt i64 %49, 0\n %81 = icmp slt i64 %50, 0\n %82 = icmp slt i64 %51, 0\n %83 = icmp slt i64 %52, 0\n %84 = icmp slt i64 %53, 0\n %85 = icmp slt i64 %54, 0\n %86 = icmp slt i64 %55, 0\n %87 = icmp slt i64 %56, 0\n %88 = icmp slt i64 %57, 0\n %89 = icmp slt i64 %58, 0\n %90 = and i1 false, %59\n %91 = and i1 false, %60\n %92 = and i1 false, %61\n %93 = and i1 false, %62\n %94 = and i1 false, %63\n %95 = and i1 false, %64\n %96 = and i1 false, %65\n %97 = and i1 false, %66\n %98 = and i1 false, %67\n %99 = and i1 false, %68\n %100 = and i1 false, %69\n %101 = and i1 false, %70\n %102 = and i1 false, %71\n %103 = and i1 false, %72\n %104 = and i1 false, %73\n %105 = and i1 false, %74\n %106 = and i1 false, %75\n %107 = and i1 false, %76\n %108 = icmp eq i32 %2, 0\n %109 = and i1 false, %77\n %110 = and i1 false, %78\n %111 = and i1 false, %79\n %112 = and i1 false, %80\n %113 = and i1 false, %81\n %114 = and i1 false, %82\n %115 = and i1 false, %83\n %116 = and i1 false, %84\n %117 = and i1 false, %85\n %118 = and i1 false, %86\n %119 = or i64 %5, %26\n %120 = getelementptr float, ptr addrspace(1) null, i64 %119\n %121 = icmp slt i64 %119, 0\n ret void\n}\n"}]}], "issue": {"title": "[SLP] Assertion `I >= 0 && I < (NumOpElts * 2) && \"Out-of-bounds shuffle mask element\"' failed.", "body": "Live link: https://godbolt.org/z/nePMscjGr\r\n\r\nTest case:\r\n\r\n```ll\r\ndefine void @func(i32 %0) {\r\n %2 = shl i32 %0, 0\r\n %3 = sext i32 %2 to i64\r\n %4 = shl i32 0, 0\r\n %5 = sext i32 %4 to i64\r\n %6 = or i32 0, 0\r\n %7 = or i32 0, 0\r\n %8 = zext i32 %6 to i64\r\n %9 = zext i32 %7 to i64\r\n %10 = zext i32 0 to i64\r\n %11 = zext i32 0 to i64\r\n %12 = zext i32 0 to i64\r\n %13 = zext i32 0 to i64\r\n %14 = zext i32 0 to i64\r\n %15 = zext i32 0 to i64\r\n %16 = zext i32 0 to i64\r\n %17 = zext i32 0 to i64\r\n %18 = zext i32 0 to i64\r\n %19 = zext i32 0 to i64\r\n %20 = zext i32 0 to i64\r\n %21 = zext i32 0 to i64\r\n %22 = zext i32 0 to i64\r\n %23 = zext i32 0 to i64\r\n %24 = zext i32 0 to i64\r\n %25 = zext i32 0 to i64\r\n %26 = zext i32 0 to i64\r\n %27 = or i64 %3, 0\r\n %28 = or i64 %3, %8\r\n %29 = or i64 %3, %9\r\n %30 = or i64 %3, %10\r\n %31 = or i64 %3, %11\r\n %32 = or i64 %3, %12\r\n %33 = or i64 %3, %13\r\n %34 = or i64 %3, %14\r\n %35 = or i64 %3, %15\r\n %36 = or i64 %3, %16\r\n %37 = or i64 %3, %17\r\n %38 = or i64 %3, %18\r\n %39 = or i64 %3, %19\r\n %40 = or i64 %3, %20\r\n %41 = or i64 %3, %21\r\n %42 = or i64 %3, %22\r\n %43 = or i64 %3, %23\r\n %44 = or i64 %3, %24\r\n %45 = or i64 %3, %25\r\n %46 = or i64 %3, 0\r\n %47 = or i64 %3, 0\r\n %48 = or i64 %3, 0\r\n %49 = or i64 %3, 0\r\n %50 = or i64 %3, 0\r\n %51 = or i64 %3, 0\r\n %52 = or i64 %3, 0\r\n %53 = or i64 %3, 0\r\n %54 = or i64 %3, 0\r\n %55 = or i64 %3, 0\r\n %56 = or i64 %3, 0\r\n %57 = or i64 %3, 0\r\n %58 = or i64 %3, 0\r\n %59 = icmp slt i64 %28, 0\r\n %60 = icmp slt i64 %29, 0\r\n %61 = icmp slt i64 %30, 0\r\n %62 = icmp slt i64 %31, 0\r\n %63 = icmp slt i64 %32, 0\r\n %64 = icmp slt i64 %33, 0\r\n %65 = icmp slt i64 %34, 0\r\n %66 = icmp slt i64 %35, 0\r\n %67 = icmp slt i64 %36, 0\r\n %68 = icmp slt i64 %37, 0\r\n %69 = icmp slt i64 %38, 0\r\n %70 = icmp slt i64 %39, 0\r\n %71 = icmp slt i64 %40, 0\r\n %72 = icmp slt i64 %41, 0\r\n %73 = icmp slt i64 %42, 0\r\n %74 = icmp slt i64 %43, 0\r\n %75 = icmp slt i64 %44, 0\r\n %76 = icmp slt i64 %45, 0\r\n %77 = icmp slt i64 %46, 0\r\n %78 = icmp slt i64 %47, 0\r\n %79 = icmp slt i64 %48, 0\r\n %80 = icmp slt i64 %49, 0\r\n %81 = icmp slt i64 %50, 0\r\n %82 = icmp slt i64 %51, 0\r\n %83 = icmp slt i64 %52, 0\r\n %84 = icmp slt i64 %53, 0\r\n %85 = icmp slt i64 %54, 0\r\n %86 = icmp slt i64 %55, 0\r\n %87 = icmp slt i64 %56, 0\r\n %88 = icmp slt i64 %57, 0\r\n %89 = icmp slt i64 %58, 0\r\n %90 = and i1 false, %59\r\n %91 = and i1 false, %60\r\n %92 = and i1 false, %61\r\n %93 = and i1 false, %62\r\n %94 = and i1 false, %63\r\n %95 = and i1 false, %64\r\n %96 = and i1 false, %65\r\n %97 = and i1 false, %66\r\n %98 = and i1 false, %67\r\n %99 = and i1 false, %68\r\n %100 = and i1 false, %69\r\n %101 = and i1 false, %70\r\n %102 = and i1 false, %71\r\n %103 = and i1 false, %72\r\n %104 = and i1 false, %73\r\n %105 = and i1 false, %74\r\n %106 = and i1 false, %75\r\n %107 = and i1 false, %76\r\n %108 = icmp eq i32 %2, 0\r\n %109 = and i1 false, %77\r\n %110 = and i1 false, %78\r\n %111 = and i1 false, %79\r\n %112 = and i1 false, %80\r\n %113 = and i1 false, %81\r\n %114 = and i1 false, %82\r\n %115 = and i1 false, %83\r\n %116 = and i1 false, %84\r\n %117 = and i1 false, %85\r\n %118 = and i1 false, %86\r\n %119 = or i64 %5, %26\r\n %120 = getelementptr float, ptr addrspace(1) null, i64 %119\r\n %121 = icmp slt i64 %119, 0\r\n ret void\r\n}\r\n```\r\n\r\nRunning via `opt --passes=slp-vectorizer -slp-vectorize-hor=false`:\r\n\r\n```sh\r\nopt: /root/llvm-project/llvm/lib/IR/Instructions.cpp:1853: bool isSingleSourceMaskImpl(llvm::ArrayRef, int): Assertion `I >= 0 && I < (NumOpElts * 2) && \"Out-of-bounds shuffle mask element\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S --passes=slp-vectorizer -slp-vectorize-hor=false \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"func\"\r\n #0 0x0000000004fa4138 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fa4138)\r\n #1 0x0000000004fa18ac SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000073f701e42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000073f701e969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000073f701e42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000073f701e287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000073f701e2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000073f701e39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000004d097f6 isSingleSourceMaskImpl(llvm::ArrayRef, int) Instructions.cpp:0:0\r\n #9 0x0000000004d1097a llvm::ShuffleVectorInst::isExtractSubvectorMask(llvm::ArrayRef, int, int&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d1097a)\r\n#10 0x000000000400fc53 (anonymous namespace)::BaseShuffleAnalysis::isIdentityMask(llvm::ArrayRef, llvm::FixedVectorType const*, bool) SLPVectorizer.cpp:0:0\r\n#11 0x0000000004043781 (anonymous namespace)::BaseShuffleAnalysis::peekThroughShuffles(llvm::Value*&, llvm::SmallVectorImpl&, bool) SLPVectorizer.cpp:0:0\r\n#12 0x000000000405bd21 llvm::Value* (anonymous namespace)::BaseShuffleAnalysis::createShuffle(llvm::Value*, llvm::Value*, llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP::ShuffleInstructionBuilder::ShuffleIRBuilder&) SLPVectorizer.cpp:0:0\r\n#13 0x000000000405c50a llvm::slpvectorizer::BoUpSLP::ShuffleInstructionBuilder::createShuffle(llvm::Value*, llvm::Value*, llvm::ArrayRef) (.isra.0) SLPVectorizer.cpp:0:0\r\n#14 0x000000000405ce73 llvm::slpvectorizer::BoUpSLP::ShuffleInstructionBuilder::finalize(llvm::ArrayRef, llvm::ArrayRef>, unsigned int, llvm::function_ref&)>) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x405ce73)\r\n#15 0x000000000405d3d9 llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool)::'lambda1'(llvm::Value*, llvm::ArrayRef)::operator()(llvm::Value*, llvm::ArrayRef) const SLPVectorizer.cpp:0:0\r\n#16 0x0000000004095c0a llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4095c0a)\r\n#17 0x0000000004097032 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4097032)\r\n#18 0x0000000004095aa8 llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4095aa8)\r\n#19 0x0000000004097294 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4097294)\r\n#20 0x00000000040b31cc llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::MapVector, llvm::DenseMap, llvm::detail::DenseMapPair>, llvm::SmallVector>, 0u>> const&, llvm::SmallVectorImpl>&, llvm::Instruction*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40b31cc)\r\n#21 0x00000000040b6d45 llvm::slpvectorizer::BoUpSLP::vectorizeTree() (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40b6d45)\r\n#22 0x00000000040c3b2a llvm::SLPVectorizerPass::tryToVectorizeList(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40c3b2a)\r\n#23 0x00000000040c78a7 bool llvm::SLPVectorizerPass::vectorizeCmpInsts>(llvm::iterator_range>, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40c78a7)\r\n#24 0x00000000040cb5d0 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40cb5d0)\r\n#25 0x00000000040d1d49 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#26 0x00000000040d29fb llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40d29fb)\r\n...\r\n```\r\n\r\nBisected to f3d2609af3031ddb54030548e86335f295cf49ca / https://github.com/llvm/llvm-project/pull/104144", "author": "rupprecht", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": [{"author": "dklimkin", "body": "Confirmed this resolved the issue. Thanks!"}]}, "verified": true} {"bug_id": "123639", "issue_url": "https://github.com/llvm/llvm-project/issues/123639", "bug_type": "miscompilation", "base_commit": "5e4c34a9b67d9ce15d14f83bcea96ccb69de1eae", "knowledge_cutoff": "2025-01-20T16:05:11Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "7d01a8f2b9ac28ffe73bef4b513d383d3edf34b5", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[13205, 13213], [14497, 14503]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::isGatherShuffledSingleRegisterEntry", "add"]}}, "patch": "commit 7d01a8f2b9ac28ffe73bef4b513d383d3edf34b5\nAuthor: Alexey Bataev \nDate: Mon Jan 20 14:22:20 2025 -0800\n\n [SLP]Fix vector factor for repeated node for bv\n \n When adding a node vector, when it is used already in the shuffle for\n buildvector, need to calculate vector factor from all vector, not only\n this single vector, to avoid incorrect result. Also, need to increase\n stability of the reused entries detection to avoid mismatch in cost\n estimation/codegen.\n \n Fixes #123639\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex fc3afed391a0..ad4855d90874 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -13205,9 +13205,12 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(\n VTE = *MIt;\n }\n }\n- Instruction &LastBundleInst = getLastInstructionInBundle(VTE);\n- if (&LastBundleInst == TEInsertPt || !CheckOrdering(&LastBundleInst))\n- continue;\n+ if (none_of(TE->CombinedEntriesWithIndices,\n+ [&](const auto &P) { return P.first == VTE->Idx; })) {\n+ Instruction &LastBundleInst = getLastInstructionInBundle(VTE);\n+ if (&LastBundleInst == TEInsertPt || !CheckOrdering(&LastBundleInst))\n+ continue;\n+ }\n VToTEs.insert(VTE);\n }\n if (VToTEs.empty())\n@@ -14497,7 +14500,9 @@ public:\n break;\n }\n }\n- int VF = getVF(V1);\n+ unsigned VF = 0;\n+ for (Value *V : InVectors)\n+ VF = std::max(VF, getVF(V));\n for (unsigned Idx = 0, Sz = CommonMask.size(); Idx < Sz; ++Idx)\n if (Mask[Idx] != PoisonMaskElem && CommonMask[Idx] == PoisonMaskElem)\n CommonMask[Idx] = Mask[Idx] + (It == InVectors.begin() ? 0 : VF);\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/multi-node-reuse-in-bv.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @main() {\nentry:\n %.neg15 = sub i32 0, 0\n %0 = trunc i64 1 to i32\n %.neg.1 = sub i32 0, 0\n %.neg15.1 = sub i32 0, %0\n %.neg16.1 = add i32 %.neg.1, %.neg15.1\n %1 = add i32 %.neg16.1, %.neg15\n %2 = trunc i64 0 to i32\n %.neg.2 = sub i32 0, %0\n %.neg15.2 = sub i32 0, %2\n %.neg16.2 = add i32 %.neg.2, %.neg15.2\n %3 = add i32 %.neg16.2, %1\n %4 = trunc i64 0 to i32\n %.neg.3 = sub i32 0, %2\n %.neg15.3 = sub i32 0, %4\n %.neg16.3 = add i32 %.neg.3, %.neg15.3\n %5 = add i32 %.neg16.3, %3\n %6 = trunc i64 0 to i32\n %.neg.4 = sub i32 0, %4\n %.neg15.4 = sub i32 0, %6\n %.neg16.4 = add i32 %.neg.4, %.neg15.4\n %7 = add i32 %.neg16.4, %5\n %.neg.5 = sub i32 0, %6\n %.neg15.5 = sub i32 0, 0\n %.neg16.5 = add i32 %.neg.5, %.neg15.5\n %8 = add i32 %.neg16.5, %7\n %.neg15.6 = sub i32 0, 0\n %.neg16.6 = add i32 0, %.neg15.6\n %9 = add i32 %.neg16.6, %8\n %.neg.7 = sub i32 0, 0\n %.neg15.7 = sub i32 0, 0\n %.neg16.7 = add i32 %.neg.7, %.neg15.7\n %10 = add i32 %.neg16.7, %9\n %11 = trunc i64 0 to i32\n %.neg.8 = sub i32 0, 0\n %.neg15.8 = sub i32 0, %11\n %.neg16.8 = add i32 %.neg.8, %.neg15.8\n %12 = add i32 %.neg16.8, %10\n %13 = trunc i64 0 to i32\n %.neg.9 = sub i32 0, %11\n %.neg15.9 = sub i32 0, %13\n %.neg16.9 = add i32 %.neg.9, %.neg15.9\n %14 = add i32 %.neg16.9, %12\n %15 = trunc i64 0 to i32\n %.neg.10 = sub i32 0, %13\n %.neg15.10 = sub i32 0, %15\n %.neg16.10 = add i32 %.neg.10, %.neg15.10\n %16 = add i32 %.neg16.10, %14\n %17 = trunc i64 0 to i32\n %.neg.11 = sub i32 0, %15\n %.neg15.11 = sub i32 0, %17\n %.neg16.11 = add i32 %.neg.11, %.neg15.11\n %18 = add i32 %.neg16.11, %16\n %19 = trunc i64 0 to i32\n %.neg.12 = sub i32 0, %17\n %.neg15.12 = sub i32 0, %19\n %.neg16.12 = add i32 %.neg.12, %.neg15.12\n %20 = add i32 %.neg16.12, %18\n %.neg.13 = sub i32 0, %19\n %.neg15.13 = sub i32 0, 0\n %.neg16.13 = add i32 %.neg.13, %.neg15.13\n %21 = add i32 %.neg16.13, %20\n %.neg.14 = sub i32 0, 0\n %.neg15.14 = sub i32 0, 0\n %.neg16.14 = add i32 %.neg.14, %.neg15.14\n %22 = add i32 %.neg16.14, %21\n %.neg.15 = sub i32 0, 0\n %.neg15.15 = sub i32 0, 0\n %.neg16.15 = add i32 %.neg.15, %.neg15.15\n %23 = add i32 %.neg16.15, %22\n %.neg.16 = sub i32 0, 0\n %.neg15.16 = sub i32 0, 0\n %.neg16.16 = add i32 %.neg.16, %.neg15.16\n %24 = add i32 %.neg16.16, %23\n %res = xor i32 %24, -2\n ret i32 %res\n}\n", "lli_expected_out": ""}]}], "issue": {"title": "SLP Vectorizer miscompilation", "body": "During our local testing, we identified a miscompilation where the SLP Vectorizer appears to be the cause. Here is the reproducer we managed to prepare:\nhttps://godbolt.org/z/9nePcq7Go\n\nThe issue can be demonstrated using the `lli` tool. Unfortunately, I cannot use alive2 due to unknown libcall errors, such as @llvm.vector.insert.v32i32.v16i32 not being recognized by alive2\n\n```\n$ bin/opt -passes=slp-vectorizer -mattr=+avx2 src.ll -S > tgt.ll\n$ bin/lli src.ll\nResult: -256\n$ bin/lli tgt.ll\nResult: -272\n```\n\nThe pseudocode represented by the IR is\n```\nt = 0\nfor (i = 0; i < 16; i++) {\n s -= (t + i)\n t = i\n}\n```\n```\ns -= (1 + 0) + (2 + 1) + (3 + 2) + (4 + 3) + ... + (14 + 15) + (15 + 16)\ns = -256\n```\nThis indicates that the SLP Vectorizer pass produces incorrect IR.\n\nThe issue became reproducible after patch `[SLP] Match poison as instruction with the same opcode` made by @alexey-bataev \nhttps://github.com/llvm/llvm-project/commit/b8703369daf777706196ff914c0376c27adde3cf\n\nI'll provide more details on how the patch affects the issue as soon as I receive them.", "author": "aleks-tmb", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": [{"author": "alexey-bataev", "body": "No related directly to this patch, reveals the bug that can be triggered even without patch. I reduced the reproducer, will commit fix soon"}]}, "verified": true} {"bug_id": "88796", "issue_url": "https://github.com/llvm/llvm-project/issues/88796", "bug_type": "crash", "base_commit": "9ef28cf88ca6e45c3ecb75c649463f8797db68d2", "knowledge_cutoff": "2024-04-15T21:04:00Z", "lit_test_dir": ["llvm/test/Transforms/VectorCombine"], "hints": {"fix_commit": "57b9c15227ec15a5e2abf4587d7d0ad536cff9e6", "components": ["VectorCombine"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": [[1961, 1977]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": ["VectorCombine::foldTruncFromReductions"]}}, "patch": "commit 57b9c15227ec15a5e2abf4587d7d0ad536cff9e6\nAuthor: Ramkumar Ramachandra \nDate: Wed May 8 09:47:55 2024 +0100\n\n VectorCombine: fix logical error after m_Trunc match (#91201)\n \n The matcher m_Trunc() matches an Operator with a given Opcode, which\n could either be an Instruction or ConstExpr.\n VectorCombine::foldTruncFromReductions() incorrectly assumes that the\n pattern matched is always an Instruction, and attempts a cast. Fix this.\n \n Fixes #88796.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\nindex bbb70134870a..8573a8adf53b 100644\n--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n@@ -1961,17 +1961,17 @@ bool VectorCombine::foldTruncFromReductions(Instruction &I) {\n if (!match(ReductionSrc, m_OneUse(m_Trunc(m_Value(TruncSrc)))))\n return false;\n \n- auto *Trunc = cast(ReductionSrc);\n auto *TruncSrcTy = cast(TruncSrc->getType());\n auto *ReductionSrcTy = cast(ReductionSrc->getType());\n Type *ResultTy = I.getType();\n \n TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n- InstructionCost OldCost =\n- TTI.getCastInstrCost(Instruction::Trunc, ReductionSrcTy, TruncSrcTy,\n- TTI::CastContextHint::None, CostKind, Trunc) +\n- TTI.getArithmeticReductionCost(ReductionOpc, ReductionSrcTy, std::nullopt,\n- CostKind);\n+ InstructionCost OldCost = TTI.getArithmeticReductionCost(\n+ ReductionOpc, ReductionSrcTy, std::nullopt, CostKind);\n+ if (auto *Trunc = dyn_cast(ReductionSrc))\n+ OldCost +=\n+ TTI.getCastInstrCost(Instruction::Trunc, ReductionSrcTy, TruncSrcTy,\n+ TTI::CastContextHint::None, CostKind, Trunc);\n InstructionCost NewCost =\n TTI.getArithmeticReductionCost(ReductionOpc, TruncSrcTy, std::nullopt,\n CostKind) +\n", "tests": [{"file": "llvm/test/Transforms/VectorCombine/pr88796.ll", "commands": ["opt -passes=vector-combine -S %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test() {\nentry:\n %0 = tail call i16 @llvm.vector.reduce.and.nxv8i16( trunc ( splat (i32 268435456) to ))\n ret i32 0\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i16 @llvm.vector.reduce.and.nxv8i16() #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[VectorCombine] Assertion 'isa(Val) && \"cast() argument of incompatible type!\"' failed.", "body": "Reduced LLVM IR:\r\n```llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main() #0 {\r\nentry:\r\n %0 = tail call i16 @llvm.vector.reduce.and.nxv8i16( trunc ( shufflevector ( insertelement ( poison, i32 268435456, i64 0), poison, zeroinitializer) to ))\r\n ret i32 0\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i16 @llvm.vector.reduce.and.nxv8i16() #1\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl256b,+zvl32b,+zvl64b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zaamo,-experimental-zabha,-experimental-zalasr,-experimental-zalrsc,-experimental-zfbfmin,-experimental-zicfilp,-experimental-zicfiss,-experimental-ztso,-experimental-zvfbfmin,-experimental-zvfbfwma,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smepmp,-ssaia,-ssccptr,-sscofpmf,-sscounterenw,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-za128rs,-za64rs,-zacas,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-zmmul,-zvbb,-zvbc,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\nattributes #1 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\nGodbolt: https://godbolt.org/z/jrEv178Ev\r\n\r\nBacktrace:\r\n```\r\nopt: /scratch/tc-testing/tc-apr-15/llvm/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = llvm::CastInst; From = llvm::Value]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt --passes=vector-combine reduced.ll\r\n #0 0x00005eb857c5a1b0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x2da91b0)\r\n #1 0x00005eb857c575bf llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x2da65bf)\r\n #2 0x00005eb857c57715 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x000071f7d0e42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x000071f7d0e969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x000071f7d0e969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x000071f7d0e969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x000071f7d0e42476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x000071f7d0e287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x000071f7d0e2871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x000071f7d0e39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00005eb856da36eb (anonymous namespace)::VectorCombine::run()::'lambda'(llvm::Instruction&)::operator()(llvm::Instruction&) const VectorCombine.cpp:0:0\r\n#12 0x00005eb856da4290 llvm::VectorCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x1ef3290)\r\n#13 0x00005eb855bc7936 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0xd16936)\r\n#14 0x00005eb857a7f451 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x2bce451)\r\n#15 0x00005eb855bbec26 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0xd0dc26)\r\n#16 0x00005eb857a7e12b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x2bcd12b)\r\n#17 0x00005eb855bc6ee6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0xd15ee6)\r\n#18 0x00005eb857a7bf61 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x2bcaf61)\r\n#19 0x00005eb855440205 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x58f205)\r\n#20 0x00005eb855432836 optMain (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x581836)\r\n#21 0x000071f7d0e29d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#22 0x000071f7d0e29e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#23 0x000071f7d0e29e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#24 0x00005eb8554284b5 _start (/scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/opt+0x5774b5)\r\nzsh: IOT instruction /scratch/tc-testing/tc-apr-15/build-rv64gcv/build-llvm-linux/bin/op\r\n```\r\n\r\nFound via fuzzer.", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "120615", "issue_url": "https://github.com/llvm/llvm-project/issues/120615", "bug_type": "crash", "base_commit": "e3fe41cdf5583d3a2f7454c76fa5cadccdccaf22", "knowledge_cutoff": "2024-12-19T18:18:18Z", "lit_test_dir": ["llvm/test/Analysis/ScalarEvolution"], "hints": {"fix_commit": "f035351af785b7349ab7bcd55149c781ceca24cb", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[15765, 15770]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::LoopGuards::collectFromBlock"]}}, "patch": "commit f035351af785b7349ab7bcd55149c781ceca24cb\nAuthor: Julian Nagele \nDate: Tue Dec 31 10:24:48 2024 +0100\n\n [SCEV] Make sure starting block is marked as visited when recursively collecting loop guards. (#120749)\n \n When `collectFromBlock` is called without a predecessor (in particular\n for loops that don't have a unique predecessor outside the loop) we\n never start climbing the predecessor chain, and thus don't mark the\n starting block as visited.\n \n Fixes https://github.com/llvm/llvm-project/issues/120615.\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex 8ab56025546e..b5668a14a4a2 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -15765,6 +15765,7 @@ void ScalarEvolution::LoopGuards::collectFromBlock(\n // original header.\n // TODO: share this logic with isLoopEntryGuardedByCond.\n unsigned NumCollectedConditions = 0;\n+ VisitedBlocks.insert(Block);\n std::pair Pair(Pred, Block);\n for (; Pair.first;\n Pair = SE.getPredecessorWithUniqueSuccessorForBB(Pair.first)) {\n", "tests": [{"file": "llvm/test/Analysis/ScalarEvolution/backedge-taken-count-guard-info-with-multiple-predecessors.ll", "commands": ["opt < %s -disable-output -passes=nary-reassociate --scalar-evolution-use-expensive-range-sharpening 2>&1"], "tests": [{"test_name": "pr120615", "test_body": "define void @pr120615() {\nentry:\n br label %header\n\nbb: ; No predecessors!\n br label %header\n\nheader: ; preds = %header, %bb, %entry\n %0 = phi i32 [ %1, %header ], [ 0, %bb ], [ 0, %entry ]\n %1 = add i32 %0, 1\n %icmp = icmp slt i32 %0, 0\n br i1 %icmp, label %header, label %exit\n\nexit: ; preds = %header\n ret void\n}\n"}]}], "issue": {"title": "[SCEV] Segfault in SCEV LoopGuards", "body": "For the following IR:\n```llvm\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @foo() {\n.split.us120:\n br label %.split.us120.split.split.us.split.split.us.split.split.us.split\n\n.split.us120.split.split.us.split.split.us.split.split.us.split: ; preds = %.noexc7.us.us.us.us.us, %.split.us120\n br label %.noexc7.us.us.us.us.us\n\n.lr.ph.us.us.us.us407: ; No predecessors!\n switch i32 0, label %.split142.us.split.us.split.us [\n i32 0, label %.split160.us.split.us.split.us\n i32 1, label %.noexc7.us.us.us.us.us\n ]\n\n.noexc7.us.us.us.us.us: ; preds = %.noexc7.us.us.us.us.us, %.lr.ph.us.us.us.us407, %.split.us120.split.split.us.split.split.us.split.split.us.split\n %0 = phi i32 [ %1, %.noexc7.us.us.us.us.us ], [ 0, %.lr.ph.us.us.us.us407 ], [ 0, %.split.us120.split.split.us.split.split.us.split.split.us.split ]\n %1 = add i32 %0, 1\n %.not.i3.us.us.us.us384.us = icmp slt i32 %0, 0\n br i1 %.not.i3.us.us.us.us384.us, label %.noexc7.us.us.us.us.us, label %.split.us120.split.split.us.split.split.us.split.split.us.split\n\n.split142.us.split.us.split.us: ; preds = %.lr.ph.us.us.us.us407\n ret void\n\n.split160.us.split.us.split.us: ; preds = %.lr.ph.us.us.us.us407\n ret void\n}\n```\n`opt` crashes when run using `-passes=nary-reassociate --scalar-evolution-use-expensive-range-sharpening`\n\ngodbolt: https://godbolt.org/z/xPv4TMMo8\n\nBacktrace (truncated due to length limits):\n```\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\nStack dump:\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=nary-reassociate --scalar-evolution-use-expensive-range-sharpening \n1.\tRunning pass \"function(nary-reassociate)\" on module \"\"\n2.\tRunning pass \"nary-reassociate\" on function \"foo\"\n #0 0x0000000005257198 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x5257198)\n #1 0x0000000005254b9c SignalHandler(int) Signals.cpp:0:0\n #2 0x00007f9ab9c42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #3 0x00000000051640cb llvm::hash_value(llvm::APInt const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x51640cb)\n #4 0x00000000051641a9 llvm::DenseMapInfo::getHashValue(llvm::APInt const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x51641a9)\n #5 0x0000000004ed726b bool llvm::DenseMapBase>, llvm::DenseMapInfo, llvm::detail::DenseMapPair>>>, llvm::APInt, std::unique_ptr>, llvm::DenseMapInfo, llvm::detail::DenseMapPair>>>::LookupBucketFor(llvm::APInt const&, llvm::detail::DenseMapPair>>*&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed726b)\n #6 0x0000000004edb996 llvm::ConstantInt::get(llvm::LLVMContext&, llvm::APInt const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4edb996)\n #7 0x0000000004eeba30 llvm::ConstantInt::get(llvm::Type*, llvm::APInt const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4eeba30)\n #8 0x0000000004eb6172 llvm::ConstantFoldBinaryInstruction(unsigned int, llvm::Constant*, llvm::Constant*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4eb6172)\n #9 0x0000000004ee9f2e llvm::ConstantExpr::get(unsigned int, llvm::Constant*, llvm::Constant*, unsigned int, llvm::Type*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ee9f2e)\n #10 0x00000000049427fb llvm::ScalarEvolution::getNegativeSCEV(llvm::SCEV const*, llvm::SCEV::NoWrapFlags) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x49427fb)\n #11 0x00000000049429c4 llvm::ScalarEvolution::getMinusSCEV(llvm::SCEV const*, llvm::SCEV const*, llvm::SCEV::NoWrapFlags, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x49429c4)\n #12 0x0000000004950e7a llvm::ScalarEvolution::LoopGuards::collectFromBlock(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::BasicBlock const*, llvm::BasicBlock const*, llvm::SmallPtrSetImpl&, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4950e7a)\n #13 0x00000000049524c2 llvm::ScalarEvolution::LoopGuards::collectFromPHI(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::PHINode const&, llvm::SmallPtrSetImpl&, llvm::SmallDenseMap, llvm::detail::DenseMapPair>&, unsigned int)::'lambda'(unsigned int)::operator()(unsigned int) const ScalarEvolution.cpp:0:0\n #14 0x0000000004952702 llvm::ScalarEvolution::LoopGuards::collectFromPHI(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::PHINode const&, llvm::SmallPtrSetImpl&, llvm::SmallDenseMap, llvm::detail::DenseMapPair>&, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4952702)\n #15 0x0000000004951839 llvm::ScalarEvolution::LoopGuards::collectFromBlock(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::BasicBlock const*, llvm::BasicBlock const*, llvm::SmallPtrSetImpl&, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4951839)\n #16 0x0000000004952c7d llvm::ScalarEvolution::applyLoopGuards(llvm::SCEV const*, llvm::Loop const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4952c7d)\n #17 0x0000000004952fb0 llvm::ScalarEvolution::getRangeForAffineNoSelfWrappingAR(llvm::SCEVAddRecExpr const*, llvm::SCEV const*, unsigned int, llvm::ScalarEvolution::RangeSignHint) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4952fb0)\n #18 0x000000000492888d llvm::ScalarEvolution::getRangeRef(llvm::SCEV const*, llvm::ScalarEvolution::RangeSignHint, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x492888d)\n #19 0x0000000004942b52 llvm::ScalarEvolution::isKnownPredicateViaConstantRanges(llvm::CmpInst::Predicate, llvm::SCEV const*, llvm::SCEV const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4942b52)\n #20 0x0000000004942e18 llvm::ScalarEvolution::isKnownViaNonRecursiveReasoning(llvm::CmpInst::Predicate, llvm::SCEV const*, llvm::SCEV const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4942e18)\n #21 0x0000000004943799 llvm::ScalarEvolution::getMinMaxExpr(llvm::SCEVTypes, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4943799)\n #22 0x0000000004943f1a llvm::ScalarEvolution::getSMinExpr(llvm::SCEV const*, llvm::SCEV const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4943f1a)\n #23 0x0000000004951128 llvm::ScalarEvolution::LoopGuards::collectFromBlock(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::BasicBlock const*, llvm::BasicBlock const*, llvm::SmallPtrSetImpl&, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4951128)\n...\n#255 0x00000000049524c2 llvm::ScalarEvolution::LoopGuards::collectFromPHI(llvm::ScalarEvolution&, llvm::ScalarEvolution::LoopGuards&, llvm::PHINode const&, llvm::SmallPtrSetImpl&, llvm::SmallDenseMap, llvm::detail::DenseMapPair>&, unsigned int)::'lambda'(unsigned int)::operator()(unsigned int) const ScalarEvolution.cpp:0:0\nProgram terminated with signal: SIGSEGV\nCompiler returned: 139\n```\n\nStack exhaustion via some recursion error maybe?", "author": "danilaml", "labels": ["llvm:SCEV", "crash-on-valid"], "comments": [{"author": "danilaml", "body": "@juliannagele possibly due to https://github.com/llvm/llvm-project/pull/113915 (from backtrace)"}, {"author": "fhahn", "body": "Likely the same as https://github.com/llvm/llvm-project/issues/120237"}, {"author": "juliannagele", "body": "Related, but slightly different, should be fixed here https://github.com/llvm/llvm-project/pull/120749."}]}, "verified": true} {"bug_id": "83127", "issue_url": "https://github.com/llvm/llvm-project/issues/83127", "bug_type": "hang", "base_commit": "7a0acccd81df268dc7ad4c0358c42552789f19b4", "knowledge_cutoff": "2024-02-27T12:20:45Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "9f45c5e1a65a1abf4920b617d36ed05e73c04bea", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[1285, 1291], [1303, 1309]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["InstCombinerImpl::foldSelectValueEquivalence"]}}, "patch": "commit 9f45c5e1a65a1abf4920b617d36ed05e73c04bea\nAuthor: Nikita Popov \nDate: Wed Mar 6 09:33:51 2024 +0100\n\n [InstCombine] Fix infinite loop in select equivalence fold (#84036)\n \n When replacing with a non-constant, it's possible that the result of the\n simplification is actually more complicated than the original, and may\n result in an infinite combine loop.\n \n Mitigate the issue by requiring that either the replacement or\n simplification result is constant, which should ensure that it's\n simpler. While this check is crude, it does not appear to cause\n optimization regressions in real-world code in practice.\n \n Fixes https://github.com/llvm/llvm-project/issues/83127.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex c47bc33df070..aee18d770f72 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -1285,7 +1285,11 @@ Instruction *InstCombinerImpl::foldSelectValueEquivalence(SelectInst &Sel,\n isGuaranteedNotToBeUndefOrPoison(CmpRHS, SQ.AC, &Sel, &DT)) {\n if (Value *V = simplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, SQ,\n /* AllowRefinement */ true))\n- return replaceOperand(Sel, Swapped ? 2 : 1, V);\n+ // Require either the replacement or the simplification result to be a\n+ // constant to avoid infinite loops.\n+ // FIXME: Make this check more precise.\n+ if (isa(CmpRHS) || isa(V))\n+ return replaceOperand(Sel, Swapped ? 2 : 1, V);\n \n // Even if TrueVal does not simplify, we can directly replace a use of\n // CmpLHS with CmpRHS, as long as the instruction is not used anywhere\n@@ -1303,7 +1307,8 @@ Instruction *InstCombinerImpl::foldSelectValueEquivalence(SelectInst &Sel,\n isGuaranteedNotToBeUndefOrPoison(CmpLHS, SQ.AC, &Sel, &DT))\n if (Value *V = simplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, SQ,\n /* AllowRefinement */ true))\n- return replaceOperand(Sel, Swapped ? 2 : 1, V);\n+ if (isa(CmpLHS) || isa(V))\n+ return replaceOperand(Sel, Swapped ? 2 : 1, V);\n \n auto *FalseInst = dyn_cast(FalseVal);\n if (!FalseInst)\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/select.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "select_replacement_shift_noundef", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine i8 @select_replacement_shift_noundef(i8 %x, i8 %y, i8 %z) {\n %shr = lshr exact i8 %x, 1\n call void @use_i8(i8 noundef %shr)\n %cmp = icmp eq i8 %shr, %y\n %shl = shl i8 %y, 1\n %sel = select i1 %cmp, i8 %shl, i8 %z\n ret i8 %sel\n}\n\ndeclare void @use_i8(i8)\n"}, {"test_name": "select_replacement_loop4", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine i16 @select_replacement_loop4(i16 noundef %p_12) {\n %cmp1 = icmp ult i16 %p_12, 2\n %and1 = and i16 %p_12, 1\n %and2 = select i1 %cmp1, i16 %and1, i16 0\n %cmp2 = icmp eq i16 %and2, %p_12\n %and3 = select i1 %cmp2, i16 %and1, i16 0\n ret i16 %and3\n}\n"}, {"test_name": "select_replacement_loop3", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine i8 @select_replacement_loop3(i32 noundef %x) {\n %trunc = trunc i32 %x to i8\n %rev = call i8 @llvm.bitreverse.i8(i8 %trunc)\n %ext = zext i8 %rev to i32\n %cmp = icmp eq i32 %ext, %x\n %sel = select i1 %cmp, i8 %trunc, i8 0\n ret i8 %sel\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i8 @llvm.bitreverse.i8(i8) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "select_replacement_sub", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine i8 @select_replacement_sub(i8 %x, i8 %y, i8 %z) {\n %cmp = icmp eq i8 %x, %y\n %sub = sub i8 %x, %y\n %sel = select i1 %cmp, i8 %sub, i8 %z\n ret i8 %sel\n}\n"}, {"test_name": "select_replacement_gep_inbounds", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine ptr @select_replacement_gep_inbounds(ptr %base, i64 %offset) {\n %cmp = icmp eq i64 %offset, 0\n %gep = getelementptr inbounds i8, ptr %base, i64 %offset\n %sel = select i1 %cmp, ptr %base, ptr %gep\n ret ptr %sel\n}\n"}, {"test_name": "select_replacement_loop2", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine i32 @select_replacement_loop2(i32 %arg, i32 %arg2) {\n %div = udiv i32 %arg, %arg2\n %mul = mul nsw i32 %div, %arg2\n %cmp = icmp eq i32 %mul, %arg\n %sel = select i1 %cmp, i32 %div, i32 undef\n ret i32 %sel\n}\n"}]}], "issue": {"title": "[InstCombine] InstCombine gets stuck when simplifying selects", "body": "Reduced test case: https://godbolt.org/z/ds9eq96db\r\n```\r\ndefine i16 @func(i16 noundef %p_12) {\r\nentry:\r\n %cmp1 = icmp ult i16 %p_12, 2\r\n %and1 = and i16 %p_12, 1\r\n %and2 = select i1 %cmp1, i16 %and1, i16 0\r\n %cmp2 = icmp eq i16 %and2, %p_12\r\n %and3 = select i1 %cmp2, i16 %and1, i16 0\r\n ret i16 %and3\r\n}\r\n```\r\n```\r\nADD: %and3 = select i1 %cmp2, i16 %and2, i16 0\r\nADD: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nADD: %and1 = and i16 %p_12, 1\r\nIC: Visiting: %and1 = and i16 %p_12, 1\r\nIC: Visiting: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nIC: Visiting: %and3 = select i1 %cmp2, i16 %and2, i16 0\r\nADD DEFERRED: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nADD DEFERRED: %cmp2 = icmp eq i16 %and2, %p_12\r\nIC: Mod = %and3 = select i1 %cmp2, i16 %and2, i16 0\r\n New = %and3 = select i1 %cmp2, i16 %and1, i16 0\r\nADD: %and3 = select i1 %cmp2, i16 %and1, i16 0\r\nADD: %cmp2 = icmp eq i16 %and2, %p_12\r\nADD: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nIC: Visiting: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nIC: Visiting: %cmp2 = icmp eq i16 %and2, %p_12\r\nIC: Visiting: %and3 = select i1 %cmp2, i16 %and1, i16 0\r\nADD DEFERRED: %and1 = and i16 %p_12, 1\r\nADD DEFERRED: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nIC: Mod = %and3 = select i1 %cmp2, i16 %and1, i16 0\r\n New = %and3 = select i1 %cmp2, i16 %and2, i16 0\r\n\r\n\r\nADD: %and3 = select i1 %cmp2, i16 %and2, i16 0\r\nADD: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nADD: %and1 = and i16 %p_12, 1\r\nIC: Visiting: %and1 = and i16 %p_12, 1\r\nIC: Visiting: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nIC: Visiting: %and3 = select i1 %cmp2, i16 %and2, i16 0\r\nADD DEFERRED: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nADD DEFERRED: %cmp2 = icmp eq i16 %and2, %p_12\r\nIC: Mod = %and3 = select i1 %cmp2, i16 %and2, i16 0\r\n New = %and3 = select i1 %cmp2, i16 %and1, i16 0\r\nADD: %and3 = select i1 %cmp2, i16 %and1, i16 0\r\nADD: %cmp2 = icmp eq i16 %and2, %p_12\r\nADD: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nIC: Visiting: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nIC: Visiting: %cmp2 = icmp eq i16 %and2, %p_12\r\nIC: Visiting: %and3 = select i1 %cmp2, i16 %and1, i16 0\r\nADD DEFERRED: %and1 = and i16 %p_12, 1\r\nADD DEFERRED: %and2 = select i1 %cmp1, i16 %and1, i16 0\r\nIC: Mod = %and3 = select i1 %cmp2, i16 %and1, i16 0\r\n New = %and3 = select i1 %cmp2, i16 %and2, i16 0\r\n\r\n...\r\n```", "author": "dtcxzyw", "labels": ["llvm:instcombine", "llvm:hang"], "comments": [{"author": "dtcxzyw", "body": "Another reproducer: https://godbolt.org/z/7MEYTe9cq\r\n```\r\ndeclare void @use(i32,i32)\r\n\r\ndefine void @func(ptr %p) {\r\n %1 = load i32, ptr %p, align 4\r\n %2 = sext i32 %1 to i64\r\n %3 = icmp slt i64 %2, 2\r\n %4 = zext i1 %3 to i32\r\n %5 = and i32 %4, %1\r\n %6 = icmp eq i32 %5, %1\r\n %7 = zext i1 %6 to i32\r\n %8 = and i32 %7, %1\r\n call void @use(i32 %8, i32 noundef %1)\r\n ret void\r\n}\r\n```\r\n"}, {"author": "dtcxzyw", "body": "Third reproducer: https://godbolt.org/z/8f8b4xved\r\n```\r\ndefine i8 @func(i32 noundef %p_38) {\r\nentry:\r\n %conv393 = trunc i32 %p_38 to i8\r\n %0 = call i8 @llvm.bitreverse.i8(i8 %conv393)\r\n %conv394 = zext i8 %0 to i32\r\n %cmp395 = icmp eq i32 %conv394, %p_38\r\n %conv397 = zext i1 %cmp395 to i8\r\n %mul.i = mul i8 %conv393, %conv397\r\n ret i8 %mul.i\r\n}\r\n```"}, {"author": "dtcxzyw", "body": "> Third reproducer: https://godbolt.org/z/8f8b4xved\r\n> \r\n> ```\r\n> define i8 @func(i32 noundef %p_38) {\r\n> entry:\r\n> %conv393 = trunc i32 %p_38 to i8\r\n> %0 = call i8 @llvm.bitreverse.i8(i8 %conv393)\r\n> %conv394 = zext i8 %0 to i32\r\n> %cmp395 = icmp eq i32 %conv394, %p_38\r\n> %conv397 = zext i1 %cmp395 to i8\r\n> %mul.i = mul i8 %conv393, %conv397\r\n> ret i8 %mul.i\r\n> }\r\n> ```\r\n\r\nA simpler testcase:\r\n```\r\ndefine i8 @func(i32 noundef %x) {\r\nentry:\r\n %trunc = trunc i32 %x to i8\r\n %rev = call i8 @llvm.bitreverse.i8(i8 %trunc)\r\n %ext = zext i8 %rev to i32\r\n %cmp = icmp eq i32 %ext, %x\r\n %sel = select i1 %cmp, i8 %trunc, i8 0\r\n ret i8 %sel\r\n}\r\n```\r\n\r\n@nikic I have no idea how to fix it in `foldSelectValueEquivalence`/`simplifyWithOpReplaced`. `simplifyWithOpReplaced` may replace the input `x` with a more complicated value `f(x)`.\r\n"}, {"author": "dtcxzyw", "body": "@nikic Ping."}]}, "verified": true} {"bug_id": "80836", "issue_url": "https://github.com/llvm/llvm-project/issues/80836", "bug_type": "crash", "base_commit": "ea2d9383a23ca17b9240ad64c2adc5f2b5a73dc0", "knowledge_cutoff": "2024-02-06T14:26:26Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "1c10821022f1799452065fb57474e894e2562b7f", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[2010, 2025]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["getCost"]}}, "patch": "commit 1c10821022f1799452065fb57474e894e2562b7f\nAuthor: David Sherwood <57997763+david-arm@users.noreply.github.com>\nDate: Wed Feb 14 16:07:51 2024 +0000\n\n [LoopVectorize] Fix divide-by-zero bug (#80836) (#81721)\n \n When attempting to use the estimated trip count to refine the costs of\n the runtime memory checks we should also check for sane trip counts to\n prevent divide-by-zero faults on some platforms.\n \n Fixes #80836\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex b39bf21cacdb..98b177cf5d2d 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -2010,16 +2010,18 @@ public:\n BestTripCount = *EstimatedTC;\n }\n \n+ BestTripCount = std::max(BestTripCount, 1U);\n InstructionCost NewMemCheckCost = MemCheckCost / BestTripCount;\n \n // Let's ensure the cost is always at least 1.\n NewMemCheckCost = std::max(*NewMemCheckCost.getValue(),\n (InstructionCost::CostType)1);\n \n- LLVM_DEBUG(dbgs()\n- << \"We expect runtime memory checks to be hoisted \"\n- << \"out of the outer loop. Cost reduced from \"\n- << MemCheckCost << \" to \" << NewMemCheckCost << '\\n');\n+ if (BestTripCount > 1)\n+ LLVM_DEBUG(dbgs()\n+ << \"We expect runtime memory checks to be hoisted \"\n+ << \"out of the outer loop. Cost reduced from \"\n+ << MemCheckCost << \" to \" << NewMemCheckCost << '\\n');\n \n MemCheckCost = NewMemCheckCost;\n }\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/low_trip_memcheck_cost.ll", "commands": ["opt -p loop-vectorize -debug-only=loop-vectorize -S -disable-output < %s 2>&1"], "tests": [{"test_name": "outer_known_tc3_full_range_checks", "test_body": "target triple = \"aarch64-unknown-linux-gnu\"\n\ndefine void @outer_known_tc3_full_range_checks(ptr nocapture noundef %dst, ptr nocapture noundef readonly %src, i64 noundef %n) {\nentry:\n br label %outer.loop\n\nouter.loop: ; preds = %inner.exit, %entry\n %outer.iv = phi i64 [ 0, %entry ], [ %outer.iv.next, %inner.exit ]\n %0 = mul nsw i64 %outer.iv, %n\n br label %inner.loop\n\ninner.loop: ; preds = %inner.loop, %outer.loop\n %iv.inner = phi i64 [ 0, %outer.loop ], [ %iv.inner.next, %inner.loop ]\n %1 = add nuw nsw i64 %iv.inner, %0\n %arrayidx.us = getelementptr inbounds i32, ptr %src, i64 %1\n %2 = load i32, ptr %arrayidx.us, align 4\n %arrayidx8.us = getelementptr inbounds i32, ptr %dst, i64 %1\n %3 = load i32, ptr %arrayidx8.us, align 4\n %add9.us = add nsw i32 %3, %2\n store i32 %add9.us, ptr %arrayidx8.us, align 4\n %iv.inner.next = add nuw nsw i64 %iv.inner, 1\n %inner.exit.cond = icmp eq i64 %iv.inner.next, %n\n br i1 %inner.exit.cond, label %inner.exit, label %inner.loop\n\ninner.exit: ; preds = %inner.loop\n %outer.iv.next = add nuw nsw i64 %outer.iv, 1\n %outer.exit.cond = icmp eq i64 %outer.iv.next, 3\n br i1 %outer.exit.cond, label %outer.exit, label %outer.loop\n\nouter.exit: ; preds = %inner.exit\n ret void\n}\n"}, {"test_name": "outer_pgo_minus1", "test_body": "target triple = \"aarch64-unknown-linux-gnu\"\n\ndefine void @outer_pgo_minus1(ptr nocapture noundef %a, ptr nocapture noundef readonly %b, i64 noundef %m, i64 noundef %n) {\nentry:\n br label %outer.loop\n\nouter.loop: ; preds = %inner.exit, %entry\n %outer.iv = phi i64 [ %outer.iv.next, %inner.exit ], [ 0, %entry ]\n %mul.us = mul nsw i64 %outer.iv, %n\n br label %inner.loop\n\ninner.loop: ; preds = %inner.loop, %outer.loop\n %inner.iv = phi i64 [ 0, %outer.loop ], [ %inner.iv.next, %inner.loop ]\n %add.us = add nuw nsw i64 %inner.iv, %mul.us\n %arrayidx.us = getelementptr inbounds i8, ptr %b, i64 %add.us\n %0 = load i8, ptr %arrayidx.us, align 1\n %arrayidx7.us = getelementptr inbounds i8, ptr %a, i64 %add.us\n %1 = load i8, ptr %arrayidx7.us, align 1\n %add9.us = add i8 %1, %0\n store i8 %add9.us, ptr %arrayidx7.us, align 1\n %inner.iv.next = add nuw nsw i64 %inner.iv, 1\n %exitcond.not = icmp eq i64 %inner.iv.next, %n\n br i1 %exitcond.not, label %inner.exit, label %inner.loop\n\ninner.exit: ; preds = %inner.loop\n %outer.iv.next = add nuw nsw i64 %outer.iv, 1\n %exitcond26.not = icmp eq i64 %outer.iv.next, %m\n br i1 %exitcond26.not, label %outer.exit, label %outer.loop, !prof !0\n\nouter.exit: ; preds = %inner.exit\n ret void\n}\n\n!0 = !{!\"branch_weights\", i32 1, i32 -1}\n"}]}], "issue": {"title": "Unhandled Floating point exception with loop-vectorize ", "body": "Run opt with -passes=loop-vectorize \r\n\r\nhttps://godbolt.org/z/s3PWY3vhE\r\n\r\nTest.ll\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine ptr addrspace(1) @wombat(i64 %arg, ptr addrspace(1) %arg1) gc \"statepoint-example\" {\r\nbb:\r\n br label %bb2\r\n\r\nbb2: ; preds = %bb4, %bb\r\n br label %bb3\r\n\r\nbb3: ; preds = %bb3, %bb2\r\n %phi = phi i64 [ 0, %bb2 ], [ %add, %bb3 ]\r\n %add = add i64 %phi, 1\r\n %load = load i8, ptr addrspace(1) %arg1, align 1\r\n %shl = shl i64 0, 0\r\n store i16 0, ptr addrspace(1) null, align 2\r\n %icmp = icmp ult i64 %phi, %arg\r\n br i1 %icmp, label %bb3, label %bb4\r\n\r\nbb4: ; preds = %bb3\r\n br i1 false, label %bb5, label %bb2, !prof !0\r\n\r\nbb5: ; preds = %bb4\r\n ret ptr addrspace(1) null\r\n}\r\n\r\n!0 = !{!\"branch_weights\", i32 1, i32 -1}\r\n```", "author": "TatyanaDoubts", "labels": ["vectorizers", "crash-on-valid"], "comments": [{"author": "DaniilSuchkov", "body": "It started failing after 962fbafecf4730ba84a3b9fd7a662a5c30bb2c7c\r\n```\r\nAuthor: David Sherwood <57997763+david-arm@users.noreply.github.com>\r\nDate: Fri Jan 26 14:43:48 2024 +0000\r\n\r\n [LoopVectorize] Refine runtime memory check costs when there is an outer loop (#76034)\r\n\r\n When we generate runtime memory checks for an inner loop it's\r\n possible that these checks are invariant in the outer loop and\r\n so will get hoisted out. In such cases, the effective cost of\r\n the checks should reduce to reflect the outer loop trip count.\r\n\r\n This fixes a 25% performance regression introduced by commit\r\n\r\n 49b0e6dcc296792b577ae8f0f674e61a0929b99d\r\n\r\n when building the SPEC2017 x264 benchmark with PGO, where we\r\n decided the inner loop trip count wasn't high enough to warrant\r\n the (incorrect) high cost of the runtime checks. Also, when\r\n runtime memory checks consist entirely of diff checks these are\r\n likely to be outer loop invariant.\r\n```\r\nWe have a division by 0 here https://github.com/llvm/llvm-project/commit/962fbafecf4730ba84a3b9fd7a662a5c30bb2c7c#diff-da321d454a7246f8ae276bf1db2782bf26b5210b8133cb59e4d7fd45d0905decR2114 because `getLoopEstimatedTripCount(OuterLoop)` just above that line returns 0."}, {"author": "david-arm", "body": "OK thanks for raising this issue - I'll take a look this morning!"}, {"author": "david-arm", "body": "I have a fix in review - https://github.com/llvm/llvm-project/pull/81721"}]}, "verified": true} {"bug_id": "98139", "issue_url": "https://github.com/llvm/llvm-project/issues/98139", "bug_type": "miscompilation", "base_commit": "6aefa918768c834ec21e0b599fa14d03b9ce3dc0", "knowledge_cutoff": "2024-07-09T09:49:19Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "722151664ecfd53295392ad875ff9bfdcd1deaaf", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": [[411, 417]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": ["InstCombinerImpl::SimplifyDemandedUseBits"]}}, "patch": "commit 722151664ecfd53295392ad875ff9bfdcd1deaaf\nAuthor: Yingwei Zheng \nDate: Tue Jul 9 22:04:55 2024 +0800\n\n [InstCombine] Fix typo in `adjustKnownBitsForSelectArm` (#98155)\n \n Fixes https://github.com/llvm/llvm-project/issues/98139.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\nindex 98f085a2c756..081e783c964f 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n@@ -411,7 +411,7 @@ Value *InstCombinerImpl::SimplifyDemandedUseBits(Instruction *I,\n // Only known if known in both the LHS and RHS.\n adjustKnownBitsForSelectArm(LHSKnown, I->getOperand(0), I->getOperand(1),\n /*Invert=*/false, Depth, Q);\n- adjustKnownBitsForSelectArm(LHSKnown, I->getOperand(0), I->getOperand(2),\n+ adjustKnownBitsForSelectArm(RHSKnown, I->getOperand(0), I->getOperand(2),\n /*Invert=*/true, Depth, Q);\n Known = LHSKnown.intersectWith(RHSKnown);\n break;\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/pr98139.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "pr98139", "test_body": "define i8 @pr98139(i16 %a, i16 %b, i1 %cond) {\n %sel = select i1 %cond, i16 1, i16 254\n %cond1 = icmp ne i16 %b, 0\n %mask = and i16 %a, 255\n %cond2 = icmp ne i16 %mask, 255\n %cond3 = or i1 %cond1, %cond2\n %sel2 = select i1 %cond3, i16 %sel, i16 255\n %sub = sub i16 %sel2, %a\n %trunc = trunc i16 %sub to i8\n ret i8 %trunc\n}\n"}]}], "issue": {"title": "wrong code at -O{s,2,3} on x86_64-linux-gnu", "body": "It appears to be a recent regression as it doesn't reproduce with 18.1.0 and earlier.\r\n\r\nCompiler Explorer: https://godbolt.org/z/o595xsMdb\r\n\r\n```\r\n[512] % clangtk -v\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git 765e2f9a8de27cc8fd8c75540844e9630d8229ad)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[513] % \r\n[513] % clangtk -O1 small.c; ./a.out\r\n1\r\n[514] % clangtk -O3 small.c; ./a.out\r\n-2\r\n1\r\n[515] % cat small.c\r\nint printf(const char *, ...);\r\nint a, b, c, e;\r\nchar d;\r\nint main() {\r\n int f = 1;\r\n unsigned g = 1;\r\n for (; c < 2; c++) {\r\n if (g)\r\n b = 1;\r\n char h = f;\r\n f = ~h;\r\n d = ~b - ~g * (a || f);\r\n g = ~g;\r\n if (g < 1)\r\n break;\r\n if (d)\r\n printf(\"%d\\n\", g);\r\n f = e;\r\n }\r\n return 0;\r\n}\r\n```", "author": "zhendongsu", "labels": ["miscompilation", "confirmed", "llvm:instcombine"], "comments": [{"author": "dtcxzyw", "body": "There is a bug in `SimplifyDemandedInstructionBits` :)\r\nI am working to provide a minimal reproducer.\r\n"}, {"author": "dtcxzyw", "body": "@nikic https://github.com/llvm/llvm-project/blob/a9183b8899f2c02ba3710b27c0cbdde6831a627b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp#L411-L416\r\n\r\nI will post a fix later.\r\n"}, {"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/GAUsBC\r\n```\r\ndefine i8 @src(i16 %a, i16 %b, i1 %cond) {\r\n %sel = select i1 %cond, i16 1, i16 254\r\n %cond1 = icmp ne i16 %b, 0\r\n %mask = and i16 %a, 255\r\n %cond2 = icmp ne i16 %mask, 255\r\n %cond3 = or i1 %cond1, %cond2\r\n %sel2 = select i1 %cond3, i16 %sel, i16 255\r\n %sub = sub i16 %sel2, %a\r\n %trunc = trunc i16 %sub to i8\r\n ret i8 %trunc\r\n}\r\n\r\ndefine i8 @tgt(i16 %a1, i16 %a2, i1 %cond) {\r\n %1 = trunc i16 %a1 to i8\r\n %conv6 = xor i8 %1, -1\r\n ret i8 %conv6\r\n}\r\n```"}]}, "verified": true} {"bug_id": "97475", "issue_url": "https://github.com/llvm/llvm-project/issues/97475", "bug_type": "miscompilation", "base_commit": "b77e734e4e6c8f5e016ba3ac49526862e6039482", "knowledge_cutoff": "2024-07-02T20:25:03Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "3969d2c3b5f42e4a180f5205efa780b0f950d733", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[4049, 4055]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["InstCombinerImpl::visitSelectInst"]}}, "patch": "commit 3969d2c3b5f42e4a180f5205efa780b0f950d733\nAuthor: Nikita Popov \nDate: Wed Jul 3 09:47:01 2024 +0200\n\n [InstCombine] Disable select known bits fold for vectors\n \n This is not safe if the simplification ends up looking through\n lane-crossing operations. For now, we don't have a good way to\n limit this in computeKnownBits(), so just disable vector handling\n entirely.\n \n Fixes https://github.com/llvm/llvm-project/issues/97475.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex 736013395e8c..394dfca262e1 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -4049,7 +4049,9 @@ Instruction *InstCombinerImpl::visitSelectInst(SelectInst &SI) {\n if (CondVal->getType() == SI.getType() && isKnownInversion(FalseVal, TrueVal))\n return BinaryOperator::CreateXor(CondVal, FalseVal);\n \n- if (SelType->isIntOrIntVectorTy() &&\n+ // For vectors, this transform is only safe if the simplification does not\n+ // look through any lane-crossing operations. For now, limit to scalars only.\n+ if (SelType->isIntegerTy() &&\n (!isa(TrueVal) || !isa(FalseVal))) {\n // Try to simplify select arms based on KnownBits implied by the condition.\n CondContext CC(CondVal);\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/select-binop-cmp.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "select_xor_icmp_vec_bad", "test_body": "define <2 x i8> @select_xor_icmp_vec_bad(<2 x i8> %x, <2 x i8> %y, <2 x i8> %z) {\n %A = icmp eq <2 x i8> %x, \n %B = xor <2 x i8> %x, %z\n %C = select <2 x i1> %A, <2 x i8> %B, <2 x i8> %y\n ret <2 x i8> %C\n}\n"}, {"test_name": "vec_select_no_equivalence", "test_body": "define <2 x i32> @vec_select_no_equivalence(<2 x i32> %x) {\n %x10 = shufflevector <2 x i32> %x, <2 x i32> undef, <2 x i32> \n %cond = icmp eq <2 x i32> %x, zeroinitializer\n %s = select <2 x i1> %cond, <2 x i32> %x10, <2 x i32> %x\n ret <2 x i32> %s\n}\n"}]}], "issue": {"title": "integer vector miscompile from instcombine", "body": "I believe this is being mis-optimized:\r\n```lllvm\r\ndefine <4 x i32> @f(<4 x i32> %0) {\r\n %2 = icmp eq <4 x i32> %0, zeroinitializer\r\n %3 = shufflevector <4 x i32> %0, <4 x i32> zeroinitializer, <4 x i32> \r\n %4 = select <4 x i1> %2, <4 x i32> %3, <4 x i32> zeroinitializer\r\n ret <4 x i32> %4\r\n}\r\n```\r\nI'll just give Alive's reasoning, which is pretty easy to follow and seems correct:\r\n```\r\nregehr@ohm:~$ ~/alive2-regehr/build/alive-tv -passes=instcombine foo.ll\r\n\r\n----------------------------------------\r\ndefine <4 x i32> @f(<4 x i32> %#0) {\r\n#1:\r\n %#2 = icmp eq <4 x i32> %#0, { 0, 0, 0, 0 }\r\n %#3 = shufflevector <4 x i32> %#0, <4 x i32> { 0, 0, 0, 0 }, 1, 1, 0, 3\r\n %#4 = select <4 x i1> %#2, <4 x i32> %#3, <4 x i32> { 0, 0, 0, 0 }\r\n ret <4 x i32> %#4\r\n}\r\n=>\r\ndefine <4 x i32> @f(<4 x i32> %#0) {\r\n#1:\r\n ret <4 x i32> { 0, 0, 0, 0 }\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Value mismatch\r\n\r\nExample:\r\n<4 x i32> %#0 = < #x00000000 (0), #x00000001 (1), #x00000000 (0), poison >\r\n\r\nSource:\r\n<4 x i1> %#2 = < #x1 (1), #x0 (0), #x1 (1), poison >\r\n<4 x i32> %#3 = < #x00000001 (1), #x00000001 (1), #x00000000 (0), poison >\r\n<4 x i32> %#4 = < #x00000001 (1), #x00000000 (0), #x00000000 (0), poison >\r\n\r\nTarget:\r\nSource value: < #x00000001 (1), #x00000000 (0), #x00000000 (0), poison >\r\nTarget value: < #x00000000 (0), #x00000000 (0), #x00000000 (0), #x00000000 (0) >\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\nregehr@ohm:~$ \r\n```\r\ncc @nunoplopes ", "author": "regehr", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "nikic", "body": "Probably caused by https://github.com/llvm/llvm-project/commit/77eb05683082dd3751ccfab963f5160f1852058d. Fell into the vector select trap again."}]}, "verified": true} {"bug_id": "126085", "issue_url": "https://github.com/llvm/llvm-project/issues/126085", "bug_type": "crash", "base_commit": "6575154b6eca80097d77db69ce1ee60c72854ee6", "knowledge_cutoff": "2025-02-06T15:41:55Z", "lit_test_dir": ["llvm/test/Transforms/VectorCombine"], "hints": {"fix_commit": "eb2b453eb73dfe0b8cf516dfb4d568c0b858d1fa", "components": ["VectorCombine"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": [[3210, 3216]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": ["VectorCombine::foldInsExtVectorToShuffle"]}}, "patch": "commit eb2b453eb73dfe0b8cf516dfb4d568c0b858d1fa\nAuthor: Simon Pilgrim \nDate: Thu Feb 6 16:40:48 2025 +0000\n\n [VectorCombine] foldInsExtVectorToShuffle - ensure we call getShuffleCost with the input operand type, not the result\n \n Typo in #121216\n \n Fixes #126085\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\nindex 4f46b621c8a7..746742e14d08 100644\n--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n@@ -3210,7 +3210,7 @@ bool VectorCombine::foldInsExtVectorToShuffle(Instruction &I) {\n ExtToVecMask[0] = ExtIdx;\n // Add cost for expanding or narrowing\n NewCost = TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc,\n- DstVecTy, ExtToVecMask, CostKind);\n+ SrcVecTy, ExtToVecMask, CostKind);\n NewCost += TTI.getShuffleCost(SK, DstVecTy, Mask, CostKind);\n }\n \n", "tests": [{"file": "llvm/test/Transforms/VectorCombine/X86/pr126085.ll", "commands": ["opt < %s -passes=vector-combine -S -mtriple=x86_64--"], "tests": [{"test_name": "", "test_body": "\ndefine i32 @test(ptr %a0) {\n;\n %load = load <16 x i8>, ptr %a0, align 1\n %shuf = shufflevector <16 x i8> %load, <16 x i8> poison, <4 x i32> \n %elt = extractelement <16 x i8> %load, i64 11\n %ins = insertelement <4 x i8> %shuf, i8 %elt, i64 1\n %res = bitcast <4 x i8> %ins to i32\n ret i32 %res\n}"}]}], "issue": {"title": "[VectorCombine] Assertion `I >= 0 && I < (NumOpElts * 2) && \"Out-of-bounds shuffle mask element\"' failed.", "body": "Reproducer:\n```\n; bin/opt -passes=vector-combine reduced.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i32 @test(ptr readonly captures(none) dereferenceable(16) %0) {\n %2 = load <16 x i8>, ptr %0, align 1\n %.sroa.01.2.vec.insert = shufflevector <16 x i8> %2, <16 x i8> poison, <4 x i32> \n %3 = extractelement <16 x i8> %2, i64 11\n %.sroa.01.3.vec.insert = insertelement <4 x i8> %.sroa.01.2.vec.insert, i8 %3, i64 1\n %4 = bitcast <4 x i8> %.sroa.01.3.vec.insert to i32\n ret i32 %4\n}\n```\n```\nopt: /home/dtcxzyw/WorkSpace/Projects/compilers/llvm-project/llvm/lib/IR/Instructions.cpp:1890: bool isSingleSourceMaskImpl(llvm::ArrayRef, int): Assertion `I >= 0 && I < (NumOpElts * 2) && \"Out-of-bounds shuffle mask element\"' failed.\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\nStack dump:\n0. Program arguments: bin/opt -passes=vector-combine reduced.ll -S\n1. Running pass \"function(vector-combine)\" on module \"reduced.ll\"\n2. Running pass \"vector-combine\" on function \"test\"\n #0 0x00007f4379a17152 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.21.0git+0x217152)\n #1 0x00007f4379a1403f llvm::sys::RunSignalHandlers() (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.21.0git+0x21403f)\n #2 0x00007f4379a1417c SignalHandler(int) Signals.cpp:0:0\n #3 0x00007f4379442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #4 0x00007f43794969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\n #5 0x00007f43794969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\n #6 0x00007f43794969fc pthread_kill ./nptl/pthread_kill.c:89:10\n #7 0x00007f4379442476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\n #8 0x00007f43794287f3 abort ./stdlib/abort.c:81:7\n #9 0x00007f437942871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\n#10 0x00007f4379439e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n#11 0x00007f4370e71a16 isSingleSourceMaskImpl(llvm::ArrayRef, int) Instructions.cpp:0:0\n#12 0x00007f4370e7c3dc llvm::ShuffleVectorInst::isReverseMask(llvm::ArrayRef, int) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.21.0git+0x27c3dc)\n#13 0x00007f43788933f2 llvm::X86TTIImpl::getShuffleCost(llvm::TargetTransformInfo::ShuffleKind, llvm::VectorType*, llvm::ArrayRef, llvm::TargetTransformInfo::TargetCostKind, int, llvm::VectorType*, llvm::ArrayRef, llvm::Instruction const*) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.21.0git+0x4933f2)\n#14 0x00007f43716ddf0c llvm::TargetTransformInfo::getShuffleCost(llvm::TargetTransformInfo::ShuffleKind, llvm::VectorType*, llvm::ArrayRef, llvm::TargetTransformInfo::TargetCostKind, int, llvm::VectorType*, llvm::ArrayRef, llvm::Instruction const*) const (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMAnalysis.so.21.0git+0x4ddf0c)\n#15 0x00007f437224f6e7 (anonymous namespace)::VectorCombine::foldInsExtVectorToShuffle(llvm::Instruction&) VectorCombine.cpp:0:0\n#16 0x00007f4372256116 (anonymous namespace)::VectorCombine::run()::'lambda'(llvm::Instruction&)::operator()(llvm::Instruction&) const VectorCombine.cpp:0:0\n#17 0x00007f437225787f llvm::VectorCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.21.0git+0x25787f)\n#18 0x00007f4373f225c5 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMPasses.so.21.0git+0x1225c5)\n#19 0x00007f4370f10cad llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.21.0git+0x310cad)\n#20 0x00007f43784dc235 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.21.0git+0xdc235)\n#21 0x00007f4370f0ee75 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.21.0git+0x30ee75)\n#22 0x00007f43784dcbf5 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.21.0git+0xdcbf5)\n#23 0x00007f4370f0fb2a llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.21.0git+0x30fb2a)\n#24 0x00007f4379ba3dc6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.21.0git+0x2fdc6)\n#25 0x00007f4379baf5c2 optMain (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.21.0git+0x3b5c2)\n#26 0x00007f4379429d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\n#27 0x00007f4379429e40 call_init ./csu/../csu/libc-start.c:128:20\n#28 0x00007f4379429e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\n#29 0x00006293bc5a5095 _start (bin/opt+0x1095)\nAborted (core dumped)\n```\n\nTriggered by https://github.com/llvm/llvm-project/pull/121216. cc @ParkHanbum @RKSimon \nReported by https://github.com/dtcxzyw/llvm-opt-benchmark/issues/2093\n\n", "author": "dtcxzyw", "labels": ["crash-on-valid", "llvm:transforms"], "comments": [{"author": "RKSimon", "body": "Found the problem - fix incoming"}]}, "verified": true} {"bug_id": "107171", "issue_url": "https://github.com/llvm/llvm-project/issues/107171", "bug_type": "crash", "base_commit": "34f2c9a9ce73a61b27d75dab7e1eed256491afcc", "knowledge_cutoff": "2024-09-04T00:23:05Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "3fe6a064f15cd854fd497594cc20e8b680cd2133", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6606, 6614]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInstructionCost"]}}, "patch": "commit 3fe6a064f15cd854fd497594cc20e8b680cd2133\nAuthor: Florian Hahn \nDate: Wed Sep 4 20:50:05 2024 +0100\n\n [LV] Check if compare is truncated directly in getInstructionCost.\n \n The current check for truncated compares in getInstructionCost misses\n cases where either the first or both operands are constants.\n Check directly if the compare is marked for truncation. In that case,\n the minimum bitwidth is that of the operands.\n \n The patch also adds asserts to ensure that.\n \n This fixes a divergence between legacy and VPlan-based cost model, where\n the legacy cost model incorrectly estimated the cost of compares with\n truncated operands.\n \n Fixes https://github.com/llvm/llvm-project/issues/107171.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 0200525a718d..0ccf442dac99 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6606,9 +6606,20 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I,\n case Instruction::ICmp:\n case Instruction::FCmp: {\n Type *ValTy = I->getOperand(0)->getType();\n+\n Instruction *Op0AsInstruction = dyn_cast(I->getOperand(0));\n- if (canTruncateToMinimalBitwidth(Op0AsInstruction, VF))\n- ValTy = IntegerType::get(ValTy->getContext(), MinBWs[Op0AsInstruction]);\n+ (void)Op0AsInstruction;\n+ assert((!canTruncateToMinimalBitwidth(Op0AsInstruction, VF) ||\n+ canTruncateToMinimalBitwidth(I, VF)) &&\n+ \"truncating Op0 must imply truncating the compare\");\n+ if (canTruncateToMinimalBitwidth(I, VF)) {\n+ assert(!canTruncateToMinimalBitwidth(Op0AsInstruction, VF) ||\n+ MinBWs[I] == MinBWs[Op0AsInstruction] &&\n+ \"if both the operand and the compare are marked for \"\n+ \"truncation, they must have the same bitwidth\");\n+ ValTy = IntegerType::get(ValTy->getContext(), MinBWs[I]);\n+ }\n+\n VectorTy = ToVectorTy(ValTy, VF);\n return TTI.getCmpSelInstrCost(I->getOpcode(), VectorTy, nullptr,\n cast(I)->getPredicate(), CostKind,\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-cost.ll", "commands": ["opt -p loop-vectorize -mattr=+v -S %s"], "tests": [{"test_name": "test_pr98413_zext_removed", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @test_pr98413_zext_removed(ptr %src, ptr noalias %dst, i64 %x) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src = getelementptr inbounds i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep.src, align 8\n %ext.l = zext i16 %l to i64\n %and = and i64 %x, %ext.l\n %trunc.and = trunc i64 %and to i8\n %gep.dst = getelementptr inbounds i8, ptr %dst, i64 %iv\n store i8 %trunc.and, ptr %gep.dst, align 1\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv, 96\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "truncate_to_i1_used_by_branch", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @truncate_to_i1_used_by_branch(i8 %x, ptr %dst) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %f.039 = phi i8 [ 0, %entry ], [ %add, %loop.latch ]\n %0 = or i8 23, %x\n %extract.t = trunc i8 %0 to i1\n br i1 %extract.t, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n store i8 0, ptr %dst, align 1\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %add = add i8 %f.039, 1\n %conv = sext i8 %f.039 to i32\n %cmp = icmp slt i32 %conv, 1\n br i1 %cmp, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v,+zvl256b\" }\n"}, {"test_name": "icmp_ops_narrowed_to_i1", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine i8 @icmp_ops_narrowed_to_i1() #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i16 [ 0, %entry ], [ %iv.next, %loop ]\n %c = icmp eq i8 0, 0\n %ext = zext i1 %c to i64\n %shr = lshr i64 %ext, 1\n %trunc = trunc i64 %shr to i8\n %iv.next = add i16 %iv, 1\n %ec = icmp eq i16 %iv.next, 100\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i8 %trunc\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "test_pr98413_sext_removed", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @test_pr98413_sext_removed(ptr %src, ptr noalias %dst, i64 %x) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src = getelementptr inbounds i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep.src, align 8\n %ext.l = sext i16 %l to i64\n %and = and i64 %x, %ext.l\n %trunc.and = trunc i64 %and to i8\n %gep.dst = getelementptr inbounds i8, ptr %dst, i64 %iv\n store i8 %trunc.and, ptr %gep.dst, align 1\n %iv.next = add i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv, 96\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "[VPlan] Assertion \" VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Testing using 3bd161e98d89d31696002994771b7761f1c74859\r\n\r\nC testcase:\r\n```c\r\nchar a;\r\nunsigned b = 221;\r\nshort c = 35;\r\nchar d[6];\r\nint main() {\r\n d[0] = 4079751220931503459;\r\n for (short e = 3; e; e += 3)\r\n for (short f = 0; f < 1; f += -1832)\r\n for (long g = ({\r\n __typeof__(0) h = !d[f * 1];\r\n h;\r\n });\r\n g < c - 3; g += b - 219)\r\n a -= 8;\r\n}\r\n```\r\n`/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/clang -I/scratch/tc-testing/compiler-fuzz-ci/csmith-build/include -fsigned-char -fno-strict-aliasing -fwrapv -march=rv64gcv -flto -O3 -fuse-ld=lld red.c -o rv64gcv.out -w`\r\n\r\nLLVM IR:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main() #0 {\r\nentry:\r\n br label %for.body14.preheader48\r\n\r\nfor.cond.cleanup: ; preds = %for.body14.preheader48\r\n store i8 %1, ptr null, align 1\r\n ret i32 0\r\n\r\nfor.body14.preheader48: ; preds = %for.body14.preheader48, %entry\r\n %e.047 = phi i16 [ 0, %entry ], [ %add27, %for.body14.preheader48 ]\r\n %tobool6.not = icmp eq i8 0, 0\r\n %conv7 = zext i1 %tobool6.not to i64\r\n %0 = lshr i64 %conv7, 1\r\n %1 = trunc i64 %0 to i8\r\n %add27 = add i16 %e.047, 1\r\n %tobool.not = icmp eq i16 %add27, 0\r\n br i1 %tobool.not, label %for.cond.cleanup, label %for.body14.preheader48\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7389: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"main\"\r\n #0 0x000061eac73d9f50 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x3010f50)\r\n #1 0x000061eac73d735f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x300e35f)\r\n #2 0x000061eac73d74b5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007496e1242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007496e12969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007496e12969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007496e12969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007496e1242476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007496e12287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007496e122871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007496e1239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x000061eac63ad810 llvm::LoopVectorizationPlanner::computeBestVF() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fe4810)\r\n#12 0x000061eac63c45bb llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffb5bb)\r\n#13 0x000061eac63c75d1 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffe5d1)\r\n#14 0x000061eac63c7c47 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffec47)\r\n#15 0x000061eac5222626 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe59626)\r\n#16 0x000061eac71e5cef llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1ccef)\r\n#17 0x000061eac52257b6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5c7b6)\r\n#18 0x000061eac71e4a6b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1ba6b)\r\n#19 0x000061eac5221a56 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe58a56)\r\n#20 0x000061eac71e2aad llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e19aad)\r\n#21 0x000061eac4a4fde6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x686de6)\r\n#22 0x000061eac4a41b71 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x678b71)\r\n#23 0x00007496e1229d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#24 0x00007496e1229e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#25 0x00007496e1229e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#26 0x000061eac4a37f25 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x66ef25)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\nhttps://godbolt.org/z/x316voG8q\r\n\r\nFound via fuzzer\r\n\r\nRelated/prior issue: https://github.com/llvm/llvm-project/issues/106780\r\n\r\ncc @fhahn", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "77468", "issue_url": "https://github.com/llvm/llvm-project/issues/77468", "bug_type": "crash", "base_commit": "39b2104b4a4e0990eddc763eab99b28e8deab953", "knowledge_cutoff": "2024-01-09T13:19:28Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "59d6f033a25d161e494457e8bb6e30375eb7f40f", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": [[895, 901], [948, 953], [979, 995]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": ["VPlanTransforms::truncateToMinimalBitwidths"]}}, "patch": "commit 59d6f033a25d161e494457e8bb6e30375eb7f40f\nAuthor: Florian Hahn \nDate: Fri Jan 12 13:14:13 2024 +0000\n\n [VPlan] Support narrowing widened loads in truncateToMinimimalBitwidths.\n \n MinBWs may also contain widened load instructions, handle them by only\n narrowing their result.\n \n Fixes https://github.com/llvm/llvm-project/issues/77468\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\nindex 5c430620a2dc..b3694e74a385 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n@@ -895,7 +895,10 @@ void VPlanTransforms::truncateToMinimalBitwidths(\n vp_depth_first_deep(Plan.getVectorLoopRegion()))) {\n for (VPRecipeBase &R : make_early_inc_range(*VPBB)) {\n if (!isa(&R))\n+ VPWidenSelectRecipe, VPWidenMemoryInstructionRecipe>(&R))\n+ continue;\n+ if (isa(&R) &&\n+ cast(&R)->isStore())\n continue;\n \n VPValue *ResultVPV = R.getVPSingleValue();\n@@ -948,6 +951,23 @@ void VPlanTransforms::truncateToMinimalBitwidths(\n \n auto *NewResTy = IntegerType::get(Ctx, NewResSizeInBits);\n \n+ // Any wrapping introduced by shrinking this operation shouldn't be\n+ // considered undefined behavior. So, we can't unconditionally copy\n+ // arithmetic wrapping flags to VPW.\n+ if (auto *VPW = dyn_cast(&R))\n+ VPW->dropPoisonGeneratingFlags();\n+\n+ // Extend result to original width.\n+ auto *Ext = new VPWidenCastRecipe(Instruction::ZExt, ResultVPV, OldResTy);\n+ Ext->insertAfter(&R);\n+ ResultVPV->replaceAllUsesWith(Ext);\n+ Ext->setOperand(0, ResultVPV);\n+\n+ if (isa(&R)) {\n+ assert(!cast(&R)->isStore() && \"stores cannot be narrowed\");\n+ continue;\n+ }\n+\n // Shrink operands by introducing truncates as needed.\n unsigned StartIdx = isa(&R) ? 1 : 0;\n for (unsigned Idx = StartIdx; Idx != R.getNumOperands(); ++Idx) {\n@@ -979,17 +999,6 @@ void VPlanTransforms::truncateToMinimalBitwidths(\n }\n }\n \n- // Any wrapping introduced by shrinking this operation shouldn't be\n- // considered undefined behavior. So, we can't unconditionally copy\n- // arithmetic wrapping flags to VPW.\n- if (auto *VPW = dyn_cast(&R))\n- VPW->dropPoisonGeneratingFlags();\n-\n- // Extend result to original width.\n- auto *Ext = new VPWidenCastRecipe(Instruction::ZExt, ResultVPV, OldResTy);\n- Ext->insertAfter(&R);\n- ResultVPV->replaceAllUsesWith(Ext);\n- Ext->setOperand(0, ResultVPV);\n }\n }\n \n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/trunc-loads-p16.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "pr77468", "test_body": "target datalayout = \"p:16:16\"\n\ndefine void @pr77468(ptr noalias %src, ptr noalias %dst, i1 %x) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i16 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src = getelementptr i32, ptr %src, i16 %iv\n %l = load i32, ptr %gep.src, align 1\n %x.ext = zext i1 %x to i32\n %and = and i32 %x.ext, %l\n %gep.dst = getelementptr i16, ptr %dst, i16 %iv\n %t = trunc i32 %and to i16\n store i16 %t, ptr %gep.dst, align 2\n %iv.next = add i16 %iv, 1\n %exitcond.not = icmp eq i16 %iv.next, 100\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "Assertion `MinBWs.size() == NumProcessedRecipes && \"some entries in MinBWs haven't been processed\"' failed. for opt -passes=loop-vectorize", "body": "llvm commit: c1ed45a271145\r\nReproduce with: ```opt -passes=\"loop-vectorize\" bbi-90681_2.ll -S -o /dev/null -mtriple=x86_64```\r\nResult:\r\n```\r\nopt: ../lib/Transforms/Vectorize/VPlanTransforms.cpp:997: static void llvm::VPlanTransforms::truncateToMinimalBitwidths(llvm::VPlan &, const MapVector &, llvm::LLVMContext &): Assertion `MinBWs.size() == NumProcessedRecipes && \"some entries in MinBWs haven't been processed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: ../../main-github/llvm/build-all/bin/opt -passes=loop-vectorize bbi-90681_2.ll -S -o /dev/null -mtriple=x86_64\r\n #0 0x0000563bdd2b15d7 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (../../main-github/llvm/build-all/bin/opt+0x63c95d7)\r\n #1 0x0000563bdd2af12e llvm::sys::RunSignalHandlers() (../../main-github/llvm/build-all/bin/opt+0x63c712e)\r\n #2 0x0000563bdd2b1c9f SignalHandler(int) (../../main-github/llvm/build-all/bin/opt+0x63c9c9f)\r\n #3 0x00007f87a7de7630 __restore_rt (/lib64/libpthread.so.0+0xf630)\r\n #4 0x00007f87a552e387 raise (/lib64/libc.so.6+0x36387)\r\n #5 0x00007f87a552fa78 abort (/lib64/libc.so.6+0x37a78)\r\n #6 0x00007f87a55271a6 __assert_fail_base (/lib64/libc.so.6+0x2f1a6)\r\n #7 0x00007f87a5527252 (/lib64/libc.so.6+0x2f252)\r\n #8 0x0000563bdd584d86 (../../main-github/llvm/build-all/bin/opt+0x669cd86)\r\n #9 0x0000563bdd50f875 llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount) (../../main-github/llvm/build-all/bin/opt+0x6627875)\r\n#10 0x0000563bdd50eeec llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (../../main-github/llvm/build-all/bin/opt+0x6626eec)\r\n#11 0x0000563bdd522bc9 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (../../main-github/llvm/build-all/bin/opt+0x663abc9)\r\n#12 0x0000563bdd5289b3 llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (../../main-github/llvm/build-all/bin/opt+0x66409b3)\r\n#13 0x0000563bdd529318 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x6641318)\r\n#14 0x0000563bdd4cafad llvm::detail::PassModel >::run(llvm::Function&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x65e2fad)\r\n#15 0x0000563bdcce36d4 llvm::PassManager >::run(llvm::Function&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x5dfb6d4)\r\n#16 0x0000563bdb09c87d llvm::detail::PassModel >, llvm::PreservedAnalyses, llvm::AnalysisManager >::run(llvm::Function&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x41b487d)\r\n#17 0x0000563bdcce7abe llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x5dffabe)\r\n#18 0x0000563bdb09c61d llvm::detail::PassModel >::run(llvm::Module&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x41b461d)\r\n#19 0x0000563bdcce27f6 llvm::PassManager >::run(llvm::Module&, llvm::AnalysisManager&) (../../main-github/llvm/build-all/bin/opt+0x5dfa7f6)\r\n#20 0x0000563bdabb7143 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (../../main-github/llvm/build-all/bin/opt+0x3ccf143)\r\n#21 0x0000563bdabc5844 main (../../main-github/llvm/build-all/bin/opt+0x3cdd844)\r\n#22 0x00007f87a551a555 __libc_start_main (/lib64/libc.so.6+0x22555)\r\n#23 0x0000563bdabb12e9 _start (../../main-github/llvm/build-all/bin/opt+0x3cc92e9)\r\nAbort (core dumped)\r\n```\r\n[bbi-90681_2.ll.gz](https://github.com/llvm/llvm-project/files/13874170/bbi-90681_2.ll.gz)\r\n", "author": "mikaelholmen", "labels": ["vectorizers", "crash"], "comments": [{"author": "mikaelholmen", "body": "This starts happening with 70535f5e609f747c28\r\n```\r\n[VPlan] Replace IR based truncateToMinimalBitwidths with VPlan version.\r\n```\r\nPing @fhahn \r\n"}, {"author": "mikaelholmen", "body": "Thanks!"}]}, "verified": true} {"bug_id": "82877", "issue_url": "https://github.com/llvm/llvm-project/issues/82877", "bug_type": "hang", "base_commit": "f32c6b28b84130dc77e0e69d6d3b692aec933280", "knowledge_cutoff": "2024-02-24T13:05:43Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "abe4677d9f8ab82f722f29dbf57952052accd70b", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstructionCombining.cpp": [[2341, 2351], [2361, 2372], [2375, 2380]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstructionCombining.cpp": ["InstCombiner::getFreelyInvertedImpl"]}}, "patch": "commit abe4677d9f8ab82f722f29dbf57952052accd70b\nAuthor: Yingwei Zheng \nDate: Tue Mar 5 03:51:06 2024 +0800\n\n [InstCombine] Fix infinite loop due to incorrect `DoesConsume` (#82973)\n \n When a call to `getFreelyInvertedImpl` with a select/phi node fails,\n `DoesConsume` should not be changed.\n \n Fixes https://github.com/llvm/llvm-project/issues/82877.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\nindex 87c8dca7efed..80ce0c9275b2 100644\n--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\n@@ -2341,11 +2341,13 @@ Value *InstCombiner::getFreelyInvertedImpl(Value *V, bool WillInvertAllUses,\n !shouldAvoidAbsorbingNotIntoSelect(*cast(V));\n // Selects/min/max with invertible operands are freely invertible\n if (IsSelect || match(V, m_MaxOrMin(m_Value(A), m_Value(B)))) {\n+ bool LocalDoesConsume = DoesConsume;\n if (!getFreelyInvertedImpl(B, B->hasOneUse(), /*Builder*/ nullptr,\n- DoesConsume, Depth))\n+ LocalDoesConsume, Depth))\n return nullptr;\n if (Value *NotA = getFreelyInvertedImpl(A, A->hasOneUse(), Builder,\n- DoesConsume, Depth)) {\n+ LocalDoesConsume, Depth)) {\n+ DoesConsume = LocalDoesConsume;\n if (Builder != nullptr) {\n Value *NotB = getFreelyInvertedImpl(B, B->hasOneUse(), Builder,\n DoesConsume, Depth);\n@@ -2361,12 +2363,13 @@ Value *InstCombiner::getFreelyInvertedImpl(Value *V, bool WillInvertAllUses,\n }\n \n if (PHINode *PN = dyn_cast(V)) {\n+ bool LocalDoesConsume = DoesConsume;\n SmallVector, 8> IncomingValues;\n for (Use &U : PN->operands()) {\n BasicBlock *IncomingBlock = PN->getIncomingBlock(U);\n Value *NewIncomingVal = getFreelyInvertedImpl(\n U.get(), /*WillInvertAllUses=*/false,\n- /*Builder=*/nullptr, DoesConsume, MaxAnalysisRecursionDepth - 1);\n+ /*Builder=*/nullptr, LocalDoesConsume, MaxAnalysisRecursionDepth - 1);\n if (NewIncomingVal == nullptr)\n return nullptr;\n // Make sure that we can safely erase the original PHI node.\n@@ -2375,6 +2378,8 @@ Value *InstCombiner::getFreelyInvertedImpl(Value *V, bool WillInvertAllUses,\n if (Builder != nullptr)\n IncomingValues.emplace_back(NewIncomingVal, IncomingBlock);\n }\n+\n+ DoesConsume = LocalDoesConsume;\n if (Builder != nullptr) {\n IRBuilderBase::InsertPointGuard Guard(*Builder);\n Builder->SetInsertPoint(PN);\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/pr82877.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "func", "test_body": "define i64 @func(i32 %p, i1 %cmp1) {\nentry:\n %not = xor i32 %p, -1\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %p0 = phi i32 [ %not, %entry ], [ %conv, %for.body ]\n %cond = select i1 %cmp1, i32 0, i32 -1231558963\n %xor = xor i32 %cond, %p0\n %cmp2 = icmp ne i32 %xor, 1\n %conv = zext i1 %cmp2 to i32\n br i1 %cmp2, label %for.body, label %for.exit\n\nfor.exit: ; preds = %for.body\n ret i64 0\n}\n"}]}], "issue": {"title": "[InstCombine] InstCombine gets stuck when simplifying icmps ", "body": "Reduced test case: https://godbolt.org/z/dc6jnha5q\r\n```\r\ndefine i64 @func(i32 %p) {\r\nentry:\r\n %not = xor i32 %p, -1\r\n br label %for.body\r\n\r\nfor.body:\r\n %p0 = phi i32 [ %not, %entry ], [ %conv, %for.body ]\r\n %p1 = phi i32 [ 0, %entry ], [ %inc, %for.body ]\r\n %inc = add i32 %p1, 1\r\n %cmp1 = icmp ne i32 %inc, 0\r\n %cond = select i1 %cmp1, i32 0, i32 -1231558963\r\n %xor = xor i32 %cond, %p0\r\n %cmp2 = icmp ne i32 %xor, -2\r\n %conv = zext i1 %cmp2 to i32\r\n br i1 %cmp2, label %for.body, label %for.exit\r\n\r\nfor.exit:\r\n ret i64 0\r\n}\r\n```\r\n\r\n```\r\nIC: Visiting: %p0 = phi i32 [ %not, %entry ], [ %conv, %for.body ]\r\nIC: Visiting: %cmp2 = icmp ne i32 %1, -2\r\nADD DEFERRED: %2 = select i1 %cmp1.not, i32 1231558962, i32 -1\r\nADD DEFERRED: %3 = xor i32 %2, %p0\r\nIC: Old = %cmp2 = icmp ne i32 %1, -2\r\n New = = icmp ne i32 %3, 1\r\nADD: %cmp2 = icmp ne i32 %3, 1\r\nIC: ERASE %4 = icmp ne i32 %1, -2\r\nADD DEFERRED: %1 = xor i32 %0, %p0\r\nIC: ERASE %1 = xor i32 %0, %p0\r\nADD DEFERRED: %0 = select i1 %cmp1.not, i32 -1231558963, i32 0\r\nADD DEFERRED: %p0 = phi i32 [ %not, %entry ], [ %conv, %for.body ]\r\nADD: %p0 = phi i32 [ %not, %entry ], [ %conv, %for.body ]\r\nIC: ERASE %0 = select i1 %cmp1.not, i32 -1231558963, i32 0\r\nADD DEFERRED: %cmp1.not = icmp eq i32 %inc, 0\r\nADD: %cmp1.not = icmp eq i32 %inc, 0\r\nADD: %1 = xor i32 %0, %p0\r\nADD: %0 = select i1 %cmp1.not, i32 1231558962, i32 -1\r\nIC: Visiting: %0 = select i1 %cmp1.not, i32 1231558962, i32 -1\r\nIC: Visiting: %1 = xor i32 %0, %p0\r\nIC: Visiting: %cmp1.not = icmp eq i32 %inc, 0\r\nIC: Visiting: %p0 = phi i32 [ %not, %entry ], [ %conv, %for.body ]\r\nIC: Visiting: %cmp2 = icmp ne i32 %1, 1\r\nADD DEFERRED: %2 = select i1 %cmp1.not, i32 -1231558963, i32 0\r\nADD DEFERRED: %3 = xor i32 %2, %p0\r\nIC: Old = %cmp2 = icmp ne i32 %1, 1\r\n New = = icmp ne i32 %3, -2\r\nADD: %cmp2 = icmp ne i32 %3, -2\r\nIC: ERASE %4 = icmp ne i32 %1, 1\r\nADD DEFERRED: %1 = xor i32 %0, %p0\r\nIC: ERASE %1 = xor i32 %0, %p0\r\nADD DEFERRED: %0 = select i1 %cmp1.not, i32 1231558962, i32 -1\r\nADD DEFERRED: %p0 = phi i32 [ %not, %entry ], [ %conv, %for.body ]\r\nADD: %p0 = phi i32 [ %not, %entry ], [ %conv, %for.body ]\r\nIC: ERASE %0 = select i1 %cmp1.not, i32 1231558962, i32 -1\r\nADD DEFERRED: %cmp1.not = icmp eq i32 %inc, 0\r\nADD: %cmp1.not = icmp eq i32 %inc, 0\r\nADD: %1 = xor i32 %0, %p0\r\nADD: %0 = select i1 %cmp1.not, i32 -1231558963, i32 0\r\nIC: Visiting: %0 = select i1 %cmp1.not, i32 -1231558963, i32 0\r\nIC: Visiting: %1 = xor i32 %0, %p0\r\nIC: Visiting: %cmp1.not = icmp eq i32 %inc, 0\r\n\r\nRepeats multiple times...\r\n```", "author": "dtcxzyw", "labels": ["llvm:instcombine", "llvm:hang"], "comments": [{"author": "dtcxzyw", "body": "Another reproducer:\r\n```\r\n/*\r\n * This is a RANDOMLY GENERATED PROGRAM.\r\n *\r\n * Generator: csmith 2.4.0\r\n * Git version: 0ec6f1b\r\n * Options: --max-funcs 3 --max-block-depth 3 --quiet --no-packed-struct --no-bitfields --no-volatiles --no-volatile-pointers --output ./csmith2024-02-24@16:09/test34436.c\r\n * Seed: 3421191658\r\n */\r\n\r\n#include \"csmith.h\"\r\n\r\n\r\nstatic long __undefined;\r\n\r\n\r\n\r\nstatic int32_t g_11[2] = {(-1L),(-1L)};\r\nstatic uint16_t g_12 = 0x5EB7L;\r\nstatic uint64_t g_31 = 2UL;\r\nstatic uint64_t g_33[9][1][9] = {{{0x0273394210AB9CA2LL,0x0B9C45D188285970LL,0x0B9C45D188285970LL,0x0273394210AB9CA2LL,18446744073709551611UL,18446744073709551607UL,0x461D90BCC9DF78A4LL,0x0B9C45D188285970LL,18446744073709551607UL}},{{4UL,0x0B9C45D188285970LL,0xF205A4AAC46587A7LL,0x461D90BCC9DF78A4LL,18446744073709551614UL,18446744073709551614UL,0x461D90BCC9DF78A4LL,0xF205A4AAC46587A7LL,0x0B9C45D188285970LL}},{{0x461D90BCC9DF78A4LL,18446744073709551611UL,0xF205A4AAC46587A7LL,0xD17401D8B2D29C61LL,18446744073709551611UL,0x0B9C45D188285970LL,1UL,0xF205A4AAC46587A7LL,0xF205A4AAC46587A7LL}},{{4UL,18446744073709551614UL,0x0B9C45D188285970LL,0xD17401D8B2D29C61LL,0x0B9C45D188285970LL,18446744073709551614UL,4UL,0x0B9C45D188285970LL,0xF205A4AAC46587A7LL}},{{0x0273394210AB9CA2LL,18446744073709551611UL,18446744073709551607UL,0x461D90BCC9DF78A4LL,0x0B9C45D188285970LL,18446744073709551607UL,1UL,18446744073709551607UL,0xDAEE255298A7B986LL}},{{0xF205A4AAC46587A7LL,0xDAEE255298A7B986LL,0xDAEE255298A7B986LL,0xF205A4AAC46587A7LL,18446744073709551606UL,0x68F5B79C14114C53LL,18446744073709551607UL,0xDAEE255298A7B986LL,0x68F5B79C14114C53LL}},{{0x0B9C45D188285970LL,0xDAEE255298A7B986LL,1UL,18446744073709551607UL,18446744073709551609UL,18446744073709551609UL,18446744073709551607UL,1UL,0xDAEE255298A7B986LL}},{{18446744073709551607UL,18446744073709551606UL,1UL,18446744073709551611UL,18446744073709551606UL,0xDAEE255298A7B986LL,18446744073709551614UL,1UL,1UL}},{{0x0B9C45D188285970LL,18446744073709551609UL,0xDAEE255298A7B986LL,18446744073709551611UL,0xDAEE255298A7B986LL,18446744073709551609UL,0x0B9C45D188285970LL,0xDAEE255298A7B986LL,1UL}}};\r\nstatic const int32_t *g_35 = &g_11[1];\r\nstatic int64_t g_66[5][4] = {{(-9L),6L,6L,(-9L)},{6L,(-9L),6L,6L},{(-9L),(-9L),3L,(-9L)},{(-9L),6L,6L,(-9L)},{6L,(-9L),6L,6L}};\r\nstatic uint32_t g_94[3] = {0x3680C6D9L,0x3680C6D9L,0x3680C6D9L};\r\nstatic uint32_t g_149 = 5UL;\r\nstatic int64_t g_157[9] = {9L,9L,(-1L),9L,9L,(-1L),9L,9L,(-1L)};\r\nstatic int8_t g_165 = 8L;\r\nstatic int16_t g_168 = 0xD32BL;\r\nstatic int32_t *g_169 = &g_11[0];\r\nstatic int32_t *g_170 = (void*)0;\r\nstatic uint8_t g_185 = 246UL;\r\nstatic int32_t **g_256 = &g_169;\r\nstatic int32_t *** const g_255 = &g_256;\r\nstatic uint32_t g_306 = 0xF063AF95L;\r\nstatic int16_t g_311 = 7L;\r\nstatic int32_t g_313 = 0x472F43A5L;\r\nstatic const uint16_t g_341 = 0xB915L;\r\nstatic uint8_t g_346[6] = {0x3EL,0x3EL,0x3EL,0x3EL,0x3EL,0x3EL};\r\nstatic uint8_t g_366[6][2] = {{0x39L,0x39L},{0UL,0x39L},{0x39L,0UL},{0x39L,0x39L},{0UL,0x39L},{0x39L,0UL}};\r\nstatic const int64_t *g_373 = &g_157[0];\r\nstatic const int64_t **g_372 = &g_373;\r\nstatic uint32_t g_380 = 4294967293UL;\r\nstatic uint16_t g_407 = 0UL;\r\nstatic int8_t g_409 = (-2L);\r\nstatic int8_t *g_408 = &g_409;\r\nstatic const int8_t g_412 = 0xF0L;\r\nstatic const int8_t g_414 = 0x89L;\r\nstatic const int8_t g_416[5][10] = {{0L,0xDDL,0L,(-1L),(-1L),0L,0xDDL,0L,(-1L),(-1L)},{0L,0xDDL,0L,(-1L),(-1L),0L,0xDDL,0L,(-1L),(-1L)},{0L,0xDDL,0L,(-1L),(-1L),0L,0xDDL,0L,(-1L),(-1L)},{0L,0xDDL,0L,(-1L),(-1L),0L,0xDDL,0L,(-1L),(-1L)},{0L,0xDDL,0L,(-1L),(-1L),0L,0xDDL,0L,(-1L),(-1L)}};\r\nstatic const int8_t *g_415[9][1] = {{&g_416[2][2]},{&g_416[2][2]},{&g_416[2][2]},{&g_416[2][2]},{&g_416[2][2]},{&g_416[2][2]},{&g_416[2][2]},{&g_416[2][2]},{&g_416[2][2]}};\r\nstatic int8_t g_426 = (-6L);\r\nstatic int32_t g_457 = 0x1708815BL;\r\nstatic int16_t *g_466[10][10][2] = {{{&g_311,&g_168},{&g_311,(void*)0},{&g_311,&g_168},{&g_311,&g_311},{&g_168,&g_311},{(void*)0,&g_311},{&g_168,&g_311},{&g_311,&g_168},{&g_311,(void*)0},{&g_311,&g_168}},{{&g_311,&g_311},{&g_168,&g_311},{(void*)0,&g_311},{&g_168,&g_311},{&g_311,&g_168},{&g_311,(void*)0},{&g_311,&g_168},{&g_311,&g_311},{&g_168,&g_311},{(void*)0,&g_311}},{{&g_168,&g_311},{&g_311,&g_168},{&g_311,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168}},{{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168}},{{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0}},{{&g_168,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0}},{{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0}},{{(void*)0,(void*)0},{&g_168,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168},{&g_168,(void*)0}},{{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168}},{{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168},{&g_168,(void*)0},{(void*)0,(void*)0},{(void*)0,&g_168},{&g_168,&g_168},{(void*)0,(void*)0},{(void*)0,(void*)0},{&g_168,&g_168}}};\r\nstatic int32_t *g_596[1] = {&g_11[0]};\r\n\r\n\r\n\r\nstatic uint8_t func_1(void);\r\nstatic int32_t func_8(int32_t p_9);\r\nstatic int16_t func_15(int8_t p_16, uint32_t p_17, int32_t p_18);\r\n\r\n\r\n\r\n\r\n\r\nstatic uint8_t func_1(void)\r\n{ \r\n int32_t l_7 = (-8L);\r\n uint16_t *l_572 = &g_407;\r\n int32_t *l_573 = &g_11[1];\r\n int32_t *l_588 = &g_313;\r\n int8_t *l_595 = (void*)0;\r\n int32_t **l_597 = &g_596[0];\r\n (*l_573) = (safe_div_func_uint16_t_u_u(((*l_572) = (0xCDF98A59L && (((0x4EBB0CB8CE3E63E3LL > (safe_lshift_func_uint32_t_u_u(1UL, (~(l_7 , func_8(l_7)))))) > (((safe_mul_func_uint32_t_u_u((l_7 || (safe_add_func_uint8_t_u_u(((g_341 && (-1L)) , 0x01L), g_416[2][6]))), l_7)) & 0xDE15L) , l_7)) >= l_7))), 65527UL));\r\n (*l_597) = (((((*l_573) ^ (**g_372)) != (safe_div_func_int8_t_s_s((safe_rshift_func_uint64_t_u_u(0x1A63987AF980F8B3LL, (safe_rshift_func_uint8_t_u_u((((g_426 ^= ((safe_add_func_int64_t_s_s(((safe_add_func_uint32_t_u_u(((safe_sub_func_uint16_t_u_u((safe_mod_func_int8_t_s_s(0x95L, (((*l_588) = (*l_573)) ^ ((*l_573) != (safe_rshift_func_int32_t_s_s((*l_573), ((((safe_sub_func_int32_t_s_s(((((safe_mod_func_int8_t_s_s((*g_408), 3UL)) && 0x38L) || g_168) == (*l_573)), (*l_573))) && 249UL) & g_31) , g_457))))))), g_94[0])) , g_341), g_94[2])) ^ g_149), (*g_373))) > 0xF735A181L)) != (*l_573)) != (-1L)), g_185)))), (*l_573)))) ^ (-1L)) , g_596[0]);\r\n for (g_185 = 0; (g_185 == 37); g_185 = safe_add_func_uint32_t_u_u(g_185, 4))\r\n { \r\n uint32_t l_600 = 0x5F5A8D5BL;\r\n (**l_597) = l_600;\r\n }\r\n return g_426;\r\n}\r\n\r\n\r\n\r\n\r\nstatic int32_t func_8(int32_t p_9)\r\n{ \r\n int32_t *l_10[7][10][3] = {{{&g_11[0],&g_11[0],&g_11[1]},{&g_11[1],&g_11[0],(void*)0},{&g_11[0],&g_11[0],&g_11[1]},{&g_11[0],&g_11[0],&g_11[1]},{&g_11[0],(void*)0,(void*)0},{&g_11[0],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[0],&g_11[0],&g_11[1]},{(void*)0,&g_11[1],(void*)0},{&g_11[1],(void*)0,&g_11[1]}},{{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[0],&g_11[1],(void*)0},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[0],&g_11[0],&g_11[1]},{&g_11[0],&g_11[1],&g_11[0]},{(void*)0,&g_11[1],&g_11[0]},{&g_11[0],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]}},{{(void*)0,(void*)0,(void*)0},{&g_11[0],&g_11[1],&g_11[1]},{(void*)0,&g_11[1],(void*)0},{&g_11[0],&g_11[0],&g_11[1]},{(void*)0,&g_11[0],&g_11[1]},{&g_11[1],&g_11[1],&g_11[0]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[0],&g_11[1],&g_11[1]},{(void*)0,(void*)0,&g_11[0]},{&g_11[0],&g_11[1],&g_11[0]}},{{&g_11[0],(void*)0,&g_11[0]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[0],&g_11[1],&g_11[1]},{&g_11[1],&g_11[0],&g_11[0]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],&g_11[0],(void*)0},{&g_11[1],&g_11[0],&g_11[1]},{&g_11[1],&g_11[0],(void*)0},{&g_11[1],&g_11[1],&g_11[1]}},{{&g_11[0],&g_11[1],&g_11[1]},{&g_11[1],&g_11[0],&g_11[1]},{&g_11[1],&g_11[1],&g_11[0]},{&g_11[0],&g_11[1],&g_11[0]},{&g_11[1],(void*)0,&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],(void*)0,(void*)0},{&g_11[0],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]}},{{&g_11[0],&g_11[0],&g_11[1]},{&g_11[1],&g_11[0],&g_11[1]},{&g_11[1],&g_11[1],&g_11[0]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],(void*)0,&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[0],&g_11[1],(void*)0},{&g_11[1],&g_11[1],&g_11[1]}},{{&g_11[0],&g_11[0],&g_11[1]},{&g_11[0],&g_11[1],&g_11[0]},{(void*)0,&g_11[1],&g_11[0]},{&g_11[0],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{&g_11[1],&g_11[1],&g_11[1]},{(void*)0,(void*)0,(void*)0},{&g_11[0],&g_11[1],&g_11[1]},{(void*)0,&g_11[1],(void*)0},{&g_11[0],&g_11[0],&g_11[1]}}};\r\n int32_t **l_23 = &l_10[6][7][1];\r\n int32_t l_24[1];\r\n uint64_t *l_30 = &g_31;\r\n uint64_t *l_32 = &g_33[8][0][8];\r\n int16_t *l_310 = &g_311;\r\n int32_t *l_312 = &g_313;\r\n uint16_t *l_314 = &g_12;\r\n uint8_t * const l_319 = &g_185;\r\n int32_t *l_338[7];\r\n int8_t l_381 = 0x82L;\r\n int64_t *l_425 = &g_66[1][0];\r\n int64_t **l_424[6] = {&l_425,&l_425,&l_425,&l_425,&l_425,&l_425};\r\n int64_t ***l_423 = &l_424[2];\r\n int i, j, k;\r\n for (i = 0; i < 1; i++)\r\n l_24[i] = 1L;\r\n for (i = 0; i < 7; i++)\r\n l_338[i] = &g_313;\r\n --g_12;\r\n (*l_312) |= (((*l_310) |= func_15((((((*l_32) = ((*l_30) = (((safe_mod_func_uint64_t_u_u((((p_9 , (void*)0) == ((*l_23) = l_10[5][8][0])) && (g_11[1] <= 0x8543L)), l_24[0])) , (!(safe_div_func_uint16_t_u_u((g_12 , ((((((safe_sub_func_int64_t_s_s(((-1L) && (g_12 >= p_9)), p_9)) > 0xED293E35L) | 0x12CB5DCFL) > g_11[1]) , p_9) , p_9)), g_11[1])))) && g_11[0]))) , 0UL) != 0x52L) || g_12), p_9, g_12)) <= 0xA360L);\r\n (*l_23) = (void*)0;\r\n if (((p_9 && (((((*l_312) >= ((*l_314) = 65535UL)) , (safe_div_func_int32_t_s_s((safe_mod_func_uint16_t_u_u(((((l_319 == l_319) >= (safe_rshift_func_int16_t_s_s((((safe_mod_func_uint32_t_u_u(p_9, g_149)) , (void*)0) != l_30), p_9))) , g_31) != p_9), g_31)), p_9))) != g_157[0]) && p_9)) && 0L))\r\n { \r\n int8_t * const l_324[7] = {(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0};\r\n int8_t *l_326[9] = {&g_165,&g_165,&g_165,&g_165,&g_165,&g_165,&g_165,&g_165,&g_165};\r\n int8_t **l_325 = &l_326[6];\r\n int8_t *l_328 = &g_165;\r\n int8_t **l_327 = &l_328;\r\n int32_t l_333 = 1L;\r\n int32_t l_369 = 0x5EBB0ECBL;\r\n const int8_t *l_413 = &g_414;\r\n int32_t **l_417 = (void*)0;\r\n int32_t l_458[9][2][1] = {{{(-5L)},{(-5L)}},{{(-6L)},{(-5L)}},{{(-5L)},{(-6L)}},{{(-5L)},{(-5L)}},{{(-6L)},{(-5L)}},{{(-5L)},{(-6L)}},{{(-5L)},{(-5L)}},{{(-6L)},{(-5L)}},{{(-5L)},{(-6L)}}};\r\n int32_t l_496[6][7] = {{1L,0xA4F1B851L,0x37A88DA2L,0xA4F1B851L,1L,0L,0L},{1L,0xA4F1B851L,0x37A88DA2L,0xA4F1B851L,1L,0L,0L},{1L,0xA4F1B851L,0x37A88DA2L,0xA4F1B851L,1L,0L,0L},{1L,0xA4F1B851L,0x37A88DA2L,0xA4F1B851L,1L,0L,0L},{1L,0xA4F1B851L,0x37A88DA2L,0xA4F1B851L,1L,0L,0L},{1L,0xA4F1B851L,0x37A88DA2L,0xA4F1B851L,1L,0L,0L}};\r\n uint8_t l_499 = 4UL;\r\n int i, j, k;\r\n if (((((l_324[0] == ((*l_327) = ((*l_325) = &g_165))) , 0x77C2C50F563A9C1DLL) , ((*l_319) = (safe_mul_func_int32_t_s_s(p_9, ((safe_lshift_func_uint64_t_u_u(l_333, (--(*l_32)))) & (safe_mul_func_uint64_t_u_u(((void*)0 != l_338[2]), (safe_sub_func_int64_t_s_s(0x64D8F0F17592A3D5LL, ((0x23L ^ g_313) < g_12)))))))))) , g_341))\r\n { \r\n return p_9;\r\n }\r\n else\r\n { \r\n int64_t l_367 = 1L;\r\n int32_t l_368 = 3L;\r\n int64_t *l_399 = &g_157[0];\r\n int64_t **l_398[8] = {&l_399,&l_399,&l_399,&l_399,&l_399,&l_399,&l_399,&l_399};\r\n int64_t ***l_397 = &l_398[6];\r\n int32_t **l_420 = (void*)0;\r\n int32_t l_460[9][4][7] = {{{0x3D2FE6F7L,(-1L),0L,0x6944E926L,(-1L),(-1L),0x2AF43CFEL},{0x77DC1A4FL,7L,0L,0xB5C39BA6L,0x8A6F8D2AL,0x78B5A455L,6L},{0L,3L,0x3D2FE6F7L,0x27A80343L,0x38C9E105L,(-1L),(-1L)},{0xD73C9A77L,0xCE3D98E2L,0x38C9E105L,(-1L),8L,(-1L),0x38C9E105L}},{{6L,6L,0x835FBC13L,5L,0xCCB98A17L,5L,0x651280DBL},{8L,1L,(-1L),0L,(-1L),(-1L),0x571592F3L},{0L,0x10A7A682L,0xB9488037L,6L,0xCCB98A17L,0x2F71711FL,0x27A80343L},{(-1L),0x38C9E105L,1L,4L,8L,7L,(-1L)}},{{0x7639893BL,0L,0x8D6AEC21L,1L,0x38C9E105L,0x541A2144L,0x08FC8D1FL},{0x651280DBL,(-1L),0L,0xD73C9A77L,0x8A6F8D2AL,0x74F12713L,0L},{1L,0x74F12713L,1L,0x52681015L,(-1L),0L,0x34D5741FL},{0x6944E926L,0xB5C39BA6L,0x78B5A455L,0x92017A1BL,(-1L),0L,(-1L)}},{{0x787EDD2CL,0x0AF78BCDL,0xB5C39BA6L,5L,(-1L),0x74F12713L,0L},{0x74F12713L,(-10L),0x10A7A682L,1L,0L,0x541A2144L,3L},{0L,0xCCB98A17L,0x7639893BL,6L,0xA8583C15L,7L,0xA8583C15L},{0x8A6F8D2AL,0xC899F44EL,0xC899F44EL,0x8A6F8D2AL,0x88020F57L,0x2F71711FL,(-1L)}},{{0xA8583C15L,0x835FBC13L,0x2AF43CFEL,0L,0x787EDD2CL,0x77DC1A4FL,1L},{1L,5L,5L,0x73D9572DL,0L,0L,0x77DC1A4FL},{(-1L),0x7639893BL,0x2DCD775DL,2L,0L,0xC899F44EL,(-10L)},{(-1L),0xB9488037L,0x835FBC13L,1L,0x92017A1BL,4L,0xCE3D98E2L}},{{0x7BF1CC1CL,6L,(-1L),0x92017A1BL,0x8A6F8D2AL,(-1L),(-1L)},{4L,0x571592F3L,0xC899F44EL,8L,0L,0x835FBC13L,0L},{2L,0x77DC1A4FL,8L,0x7BF1CC1CL,0xCCB98A17L,0x74F12713L,0x08FC8D1FL},{2L,0x8A6F8D2AL,3L,0x571592F3L,(-1L),0L,1L}},{{4L,3L,0x92017A1BL,0x38C9E105L,0x7BF1CC1CL,0x88020F57L,5L},{0x7BF1CC1CL,(-1L),0L,0x78B5A455L,5L,(-1L),0x835FBC13L},{(-1L),0L,0L,0L,0L,(-1L),0L},{(-1L),0x3D2FE6F7L,1L,3L,0xC14CCFD7L,0L,0x2F71711FL}},{{1L,0x787EDD2CL,1L,0x651280DBL,8L,0x92017A1BL,0x7639893BL},{(-10L),0x3D2FE6F7L,1L,(-1L),0L,(-1L),1L},{0x52681015L,0L,2L,(-1L),0x88020F57L,0x0AF78BCDL,0L},{7L,(-1L),0x0AF78BCDL,1L,0xB9488037L,1L,0L}},{{(-1L),3L,0x2DFF457AL,0x787EDD2CL,1L,0L,0x8D6AEC21L},{0x27A80343L,0x8A6F8D2AL,6L,0x2DFF457AL,1L,8L,(-1L)},{(-1L),0x77DC1A4FL,6L,1L,(-1L),1L,0x34D5741FL},{6L,0x571592F3L,0x2DFF457AL,1L,0x78B5A455L,0x52681015L,0x3D2FE6F7L}}};\r\n uint8_t l_461 = 0xE6L;\r\n int i, j, k;\r\n for (g_185 = 0; (g_185 <= 2); g_185 += 1)\r\n { \r\n int32_t l_358[8][2] = {{0xC6A9D3CAL,(-5L)},{0x33E30AB4L,0x33E30AB4L},{(-5L),0xC6A9D3CAL},{0x356F701CL,0xC6A9D3CAL},{(-5L),0x33E30AB4L},{0x33E30AB4L,(-5L)},{0xC6A9D3CAL,0x356F701CL},{0xC6A9D3CAL,(-5L)}};\r\n uint8_t *l_364 = (void*)0;\r\n uint8_t *l_365[9][4][7] = {{{&g_185,&g_366[5][1],(void*)0,&g_185,&g_366[5][1],&g_366[5][1],(void*)0},{&g_185,&g_366[4][1],&g_185,&g_366[2][0],(void*)0,&g_366[5][1],&g_185},{(void*)0,&g_366[5][1],&g_366[1][0],&g_366[5][1],&g_366[1][0],&g_366[5][1],(void*)0},{&g_185,&g_366[5][1],(void*)0,&g_366[2][0],&g_185,&g_366[4][1],&g_185}},{{(void*)0,&g_366[5][1],&g_366[5][1],&g_185,(void*)0,&g_366[5][1],&g_185},{&g_366[5][1],&g_366[5][1],(void*)0,&g_185,&g_366[5][1],&g_366[5][1],&g_366[5][1]},{&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[1][1],&g_185},{&g_366[2][0],&g_185,&g_366[5][1],&g_366[4][1],&g_366[5][1],&g_185,&g_366[5][1]}},{{(void*)0,&g_366[5][1],&g_366[1][1],&g_366[5][1],&g_366[4][0],&g_366[5][1],(void*)0},{&g_366[2][0],&g_366[4][1],&g_366[5][1],&g_185,&g_366[5][1],&g_366[5][1],&g_185},{&g_366[5][1],&g_366[5][0],&g_366[5][1],&g_366[5][1],&g_366[0][1],&g_366[5][1],(void*)0},{&g_185,(void*)0,&g_366[5][1],&g_366[5][1],&g_366[2][0],&g_366[5][1],&g_366[5][1]}},{{&g_366[0][1],&g_185,&g_366[4][0],&g_366[5][1],&g_366[1][0],&g_366[5][1],&g_185},{&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_185,&g_185,&g_366[5][1],&g_366[5][1]},{&g_366[1][1],&g_366[1][1],&g_366[0][1],&g_185,&g_366[1][0],&g_366[5][1],&g_366[5][1]},{&g_366[5][1],&g_185,&g_366[2][0],&g_366[5][1],&g_366[2][0],&g_185,&g_366[5][1]}},{{&g_366[5][1],&g_366[5][1],&g_366[1][0],&g_185,&g_366[0][1],&g_366[1][1],&g_366[1][1]},{&g_366[5][1],&g_366[5][1],&g_185,&g_185,&g_366[5][1],&g_366[5][1],&g_366[5][1]},{&g_185,&g_366[5][1],&g_366[1][0],&g_366[5][1],&g_366[4][0],&g_185,&g_366[0][1]},{&g_366[5][1],&g_366[5][1],&g_366[2][0],&g_366[5][1],&g_366[5][1],(void*)0,&g_185}},{{(void*)0,&g_366[5][1],&g_366[0][1],&g_366[5][1],&g_366[5][1],&g_366[5][0],&g_366[5][1]},{&g_185,&g_366[5][1],&g_366[5][1],&g_185,&g_366[5][1],&g_366[4][1],&g_366[2][0]},{(void*)0,&g_366[5][1],&g_366[4][0],&g_366[5][1],&g_366[1][1],&g_366[5][1],(void*)0},{&g_366[5][1],&g_185,&g_366[5][1],&g_366[4][1],&g_366[5][1],&g_185,&g_366[2][0]}},{{&g_185,&g_366[1][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1]},{&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_185,&g_185},{&g_366[5][1],&g_185,&g_366[1][1],&g_366[5][1],(void*)0,&g_366[5][1],&g_366[0][1]},{&g_366[5][1],(void*)0,&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[4][1],&g_366[5][1]}},{{&g_366[1][1],&g_366[5][0],&g_366[5][1],&g_185,&g_366[5][1],&g_366[5][0],&g_366[1][1]},{&g_366[5][1],&g_366[4][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],(void*)0,&g_366[5][1]},{&g_366[0][1],&g_366[5][1],(void*)0,&g_366[5][1],&g_366[1][1],&g_185,&g_366[5][1]},{&g_185,&g_185,&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1]}},{{&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],&g_366[5][1],(void*)0,&g_366[5][1]},{&g_366[5][1],&g_366[5][1],&g_366[5][1],(void*)0,&g_185,&g_366[5][1],&g_185},{&g_366[0][1],&g_366[5][1],(void*)0,&g_185,&g_185,&g_366[1][1],&g_366[5][1]},{&g_366[5][1],(void*)0,&g_185,&g_366[5][1],&g_185,&g_185,&g_366[5][1]}}};\r\n const int64_t ***l_374 = &g_372;\r\n int32_t l_379 = 1L;\r\n int i, j, k;\r\n (***g_255) &= (p_9 , ((((safe_div_func_int32_t_s_s(((*l_312) = (safe_add_func_int16_t_s_s(p_9, (g_346[4] = ((*l_314) = 0x6B72L))))), ((safe_sub_func_int16_t_s_s(((g_380 = (safe_sub_func_uint64_t_u_u(((safe_rshift_func_uint64_t_u_s(((((!(safe_sub_func_int64_t_s_s((safe_mul_func_int16_t_s_s(0xB195L, l_358[2][0])), (safe_unary_minus_func_int8_t_s(g_157[5]))))) , ((safe_rshift_func_uint32_t_u_s((((safe_mod_func_uint64_t_u_u((l_358[7][1] = p_9), (((g_366[3][0]--) & (((((*l_374) = g_372) != ((safe_sub_func_int16_t_s_s((l_369 |= ((((**l_327) |= (safe_div_func_int16_t_s_s((g_157[7] , g_157[0]), 0x06ABL))) | g_31) , g_306)), p_9)) , (void*)0)) == 0UL) , p_9)) , p_9))) < g_341) >= g_33[1][0][8]), 17)) , 0x2338L)) ^ 65535UL) , g_94[1]), l_379)) & (*g_373)), 0x739DBB08F21F6146LL))) , g_31), 0L)) && l_381))) >= (-5L)) & 0L) ^ l_379));\r\n }\r\n for (g_165 = 0; (g_165 == (-6)); g_165 = safe_sub_func_int8_t_s_s(g_165, 3))\r\n { \r\n const int64_t ***l_396 = (void*)0;\r\n int64_t ****l_400 = (void*)0;\r\n int64_t ****l_401 = &l_397;\r\n int32_t l_406 = (-1L);\r\n const int8_t *l_411[4][10][5] = {{{&g_412,&g_412,&g_412,&g_412,&g_412},{(void*)0,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,(void*)0,&g_412,&g_412,&g_412},{&g_412,(void*)0,&g_412,(void*)0,&g_412},{&g_412,&g_412,&g_412,(void*)0,&g_412},{&g_412,&g_412,&g_412,&g_412,(void*)0},{&g_412,&g_412,&g_412,&g_412,&g_412},{(void*)0,&g_412,&g_412,&g_412,&g_412}},{{&g_412,&g_412,&g_412,&g_412,&g_412},{(void*)0,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,(void*)0,&g_412,&g_412,&g_412},{&g_412,(void*)0,&g_412,(void*)0,&g_412},{&g_412,&g_412,&g_412,(void*)0,&g_412},{&g_412,&g_412,&g_412,&g_412,(void*)0},{&g_412,&g_412,&g_412,&g_412,&g_412},{(void*)0,&g_412,&g_412,&g_412,&g_412}},{{&g_412,&g_412,&g_412,&g_412,&g_412},{(void*)0,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,(void*)0,&g_412,&g_412,&g_412},{&g_412,(void*)0,&g_412,(void*)0,&g_412},{&g_412,&g_412,&g_412,(void*)0,&g_412},{&g_412,&g_412,&g_412,&g_412,(void*)0},{&g_412,&g_412,&g_412,&g_412,&g_412},{(void*)0,&g_412,&g_412,&g_412,&g_412}},{{&g_412,&g_412,&g_412,&g_412,&g_412},{(void*)0,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,(void*)0},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,&g_412,&g_412,&g_412,(void*)0},{(void*)0,&g_412,&g_412,&g_412,&g_412},{(void*)0,&g_412,&g_412,(void*)0,&g_412},{&g_412,&g_412,&g_412,&g_412,&g_412},{&g_412,(void*)0,&g_412,&g_412,&g_412}}};\r\n const int8_t **l_410[5][8][6] = {{{&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[3][7][3],&l_411[2][7][3]},{&l_411[3][5][2],&l_411[2][7][3],&l_411[3][4][4],&l_411[2][7][3],&l_411[2][7][3],(void*)0},{(void*)0,&l_411[2][7][3],(void*)0,&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3]},{&l_411[3][5][2],&l_411[0][2][2],&l_411[2][7][3],&l_411[2][7][3],&l_411[3][3][1],&l_411[2][7][3]},{&l_411[2][7][3],&l_411[0][3][1],&l_411[0][7][3],&l_411[2][3][0],&l_411[2][7][3],&l_411[2][7][3]},{(void*)0,&l_411[2][8][1],&l_411[2][7][3],&l_411[2][7][3],(void*)0,&l_411[2][7][3]},{&l_411[2][7][3],&l_411[3][1][4],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3]},{&l_411[0][8][2],&l_411[2][7][3],&l_411[0][7][3],&l_411[2][7][3],&l_411[0][7][3],&l_411[2][7][3]}},{{&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],(void*)0,&l_411[2][7][3]},{&l_411[2][7][3],(void*)0,(void*)0,&l_411[0][2][2],&l_411[2][7][3],(void*)0},{&l_411[2][7][3],(void*)0,&l_411[3][4][4],&l_411[3][1][4],(void*)0,&l_411[2][7][3]},{&l_411[1][2][4],&l_411[2][7][3],&l_411[2][7][3],&l_411[0][3][1],&l_411[0][7][3],&l_411[2][3][0]},{&l_411[3][3][1],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3]},{&l_411[3][2][0],&l_411[3][1][4],&l_411[3][7][3],&l_411[2][7][3],(void*)0,&l_411[2][8][1]},{&l_411[3][2][0],&l_411[2][8][1],(void*)0,&l_411[2][7][3],&l_411[2][7][3],&l_411[3][6][3]},{&l_411[3][3][1],&l_411[0][3][1],&l_411[2][7][3],&l_411[0][3][1],&l_411[3][3][1],&l_411[2][7][3]}},{{&l_411[1][2][4],&l_411[0][2][2],&l_411[2][7][3],&l_411[3][1][4],&l_411[2][7][3],&l_411[1][1][3]},{&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[0][2][2],&l_411[2][7][3],&l_411[1][1][3]},{&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[3][7][3],&l_411[2][7][3]},{&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[3][6][3]},{&l_411[0][8][2],(void*)0,(void*)0,&l_411[2][7][3],&l_411[3][5][2],&l_411[2][8][1]},{&l_411[2][7][3],&l_411[2][7][3],&l_411[3][7][3],&l_411[2][7][3],&l_411[3][5][2],&l_411[2][7][3]},{(void*)0,(void*)0,&l_411[2][7][3],&l_411[2][3][0],&l_411[2][7][3],&l_411[2][3][0]},{&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[3][7][3],&l_411[2][7][3]}},{{&l_411[3][5][2],&l_411[2][7][3],&l_411[3][4][4],&l_411[2][7][3],&l_411[2][7][3],(void*)0},{(void*)0,&l_411[2][7][3],(void*)0,&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3]},{&l_411[3][5][2],&l_411[0][2][2],&l_411[2][7][3],&l_411[2][7][3],(void*)0,&l_411[2][7][3]},{(void*)0,&l_411[2][7][3],&l_411[2][7][3],&l_411[0][3][1],&l_411[1][2][4],(void*)0},{&l_411[3][5][2],&l_411[3][1][4],&l_411[1][2][4],&l_411[2][4][2],&l_411[2][7][3],&l_411[2][7][3]},{&l_411[3][3][1],&l_411[2][7][3],&l_411[1][2][4],&l_411[2][7][3],(void*)0,(void*)0},{&l_411[3][4][4],&l_411[2][7][3],&l_411[2][7][3],&l_411[1][1][3],&l_411[2][7][3],&l_411[2][7][3]},{&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][8][1],(void*)0,&l_411[2][7][3]}},{{&l_411[2][7][3],&l_411[2][3][0],&l_411[3][5][2],&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3]},{&l_411[2][7][3],&l_411[2][3][0],(void*)0,&l_411[2][7][3],(void*)0,&l_411[0][2][2]},{&l_411[2][7][3],&l_411[2][7][3],(void*)0,&l_411[2][7][3],&l_411[2][7][3],&l_411[0][3][1]},{(void*)0,&l_411[2][7][3],&l_411[3][3][1],&l_411[0][2][2],(void*)0,&l_411[2][8][1]},{&l_411[2][7][3],&l_411[2][7][3],&l_411[0][7][3],(void*)0,&l_411[2][7][3],&l_411[3][1][4]},{&l_411[2][7][3],&l_411[3][1][4],&l_411[2][7][3],&l_411[0][2][2],&l_411[1][2][4],&l_411[2][7][3]},{(void*)0,&l_411[2][7][3],&l_411[2][7][3],&l_411[2][7][3],(void*)0,&l_411[2][7][3]},{&l_411[2][7][3],&l_411[2][7][3],&l_411[3][7][3],&l_411[2][7][3],&l_411[2][7][3],(void*)0}}};\r\n int32_t **l_418 = &l_338[6];\r\n int32_t ***l_419[4][2] = {{&l_23,&l_418},{&l_23,&l_418},{&l_23,&l_418},{&l_23,&l_418}};\r\n int i, j, k;\r\n l_333 ^= (((safe_rshift_func_int64_t_s_u((p_9 , ((((((g_407 |= (l_369 = (safe_mod_func_int64_t_s_s(((0xA2936E82L == ((safe_add_func_uint32_t_u_u((safe_mod_func_uint8_t_u_u(l_368, (((((((p_9 , (safe_mod_func_int64_t_s_s((safe_mod_func_int8_t_s_s((l_396 != ((*l_401) = l_397)), (safe_lshift_func_uint32_t_u_s((g_311 || ((safe_lshift_func_uint8_t_u_u((p_9 <= p_9), 4)) | (*g_373))), 31)))), 1L))) && p_9) , &g_373) == &l_399) & (-2L)) , g_313) , g_149))), l_406)) | (-7L))) || g_33[2][0][6]), 0xE09499CAD8E44A95LL)))) ^ p_9) > (*g_169)) || p_9) , g_94[2]) >= (-1L))), 54)) , (*l_312)) | p_9);\r\n (*g_169) = (((*l_314) = (((g_408 = &g_165) != (g_415[6][0] = (l_413 = (void*)0))) <= ((((((*g_255) = l_417) != (l_420 = (l_23 = l_418))) | (l_368 ^= ((((*l_319) = ((((((p_9 | ((safe_mul_func_uint16_t_u_u(0x7263L, (&g_373 == ((**l_401) = (*l_397))))) < ((&g_372 != l_423) & 0x5FL))) && g_414) , p_9) , (*g_372)) != (void*)0) ^ g_185)) && g_426) >= 0x0347D94C0175571CLL))) & p_9) , g_366[5][1]))) == g_414);\r\n if (p_9)\r\n break;\r\n if ((*g_169))\r\n continue;\r\n }\r\n if ((((*l_425) = (p_9 , p_9)) ^ p_9))\r\n { \r\n int32_t *l_434 = &l_24[0];\r\n l_434 = ((safe_unary_minus_func_int64_t_s((safe_rshift_func_uint32_t_u_s((safe_add_func_int32_t_s_s((safe_lshift_func_uint16_t_u_u(5UL, 10)), 4294967293UL)), 24)))) , &g_313);\r\n (*l_23) = (*l_23);\r\n (*l_23) = l_434;\r\n (*l_23) = l_434;\r\n }\r\n else\r\n { \r\n int8_t ***l_437 = (void*)0;\r\n int8_t ***l_438[3];\r\n int32_t l_443 = (-7L);\r\n int32_t l_459[8][9][1] = {{{0x6734A3A9L},{0x17C83ADDL},{0x328560B7L},{1L},{0x220AB6B7L},{1L},{0x220AB6B7L},{1L},{0x328560B7L}},{{0x17C83ADDL},{0x6734A3A9L},{0x7340F128L},{6L},{0x8A5B444AL},{0x553E4A9CL},{6L},{0x71E29F3CL},{6L}},{{0x553E4A9CL},{0x8A5B444AL},{6L},{0x7340F128L},{0x6734A3A9L},{0x17C83ADDL},{0x328560B7L},{1L},{0x220AB6B7L}},{{1L},{0x220AB6B7L},{1L},{0x328560B7L},{0x17C83ADDL},{0x6734A3A9L},{0x7340F128L},{6L},{0x8A5B444AL}},{{0x553E4A9CL},{6L},{0x71E29F3CL},{6L},{0x553E4A9CL},{0x8A5B444AL},{6L},{0x7340F128L},{0x6734A3A9L}},{{0x17C83ADDL},{0x328560B7L},{1L},{0x220AB6B7L},{1L},{0x220AB6B7L},{1L},{0x328560B7L},{0x17C83ADDL}},{{0x6734A3A9L},{0x7340F128L},{6L},{0x8A5B444AL},{0x553E4A9CL},{6L},{0x71E29F3CL},{6L},{0x553E4A9CL}},{{0x8A5B444AL},{6L},{0x7340F128L},{0x6734A3A9L},{0x17C83ADDL},{0x328560B7L},{1L},{0x220AB6B7L},{1L}}};\r\n int16_t * const l_467 = &g_168;\r\n int i, j, k;\r\n for (i = 0; i < 3; i++)\r\n l_438[i] = &l_327;\r\n l_443 = (((safe_div_func_uint32_t_u_u((&g_415[6][0] == (l_327 = &g_408)), (l_333 = 1L))) & ((safe_lshift_func_int8_t_s_s((safe_sub_func_int16_t_s_s(l_443, ((~((!(l_417 != l_420)) | (0x38L ^ ((safe_sub_func_uint64_t_u_u((safe_sub_func_uint16_t_u_u((((safe_add_func_int8_t_s_s((((*l_314) = (safe_mul_func_int64_t_s_s(((*l_425) ^= (*g_373)), (g_457 = (safe_add_func_int64_t_s_s((((~l_443) | p_9) , (*g_373)), g_11[1])))))) , 9L), 0xE4L)) , g_346[4]) , 0x1284L), p_9)), 0x1264E6488AB867F8LL)) , 0xF8L)))) , g_380))), 1)) , (*g_408))) & (-8L));\r\n l_461--;\r\n (*l_312) ^= (((g_466[8][2][1] = &g_168) == l_467) < ((safe_rshift_func_int32_t_s_u((p_9 != (safe_add_func_int64_t_s_s((safe_rshift_func_int8_t_s_s((safe_rshift_func_int16_t_s_u((safe_rshift_func_int16_t_s_s(((safe_add_func_int32_t_s_s((safe_sub_func_uint16_t_u_u(p_9, ((++(*l_314)) & (~((safe_div_func_int32_t_s_s((safe_rshift_func_int16_t_s_s((((*l_30) = l_443) & (((safe_lshift_func_uint16_t_u_u((+((safe_mul_func_uint8_t_u_u((--(*l_319)), l_496[1][1])) | ((safe_add_func_uint8_t_u_u((((*l_32) = p_9) <= (l_326[7] == &g_346[2])), 0xA6L)) , p_9))), 13)) , p_9) < l_443)), p_9)), 9UL)) | g_11[1]))))), p_9)) , 0x32F2L), g_416[3][4])), g_66[3][3])), 2)), p_9))), 31)) == (*g_373)));\r\n (*g_169) ^= p_9;\r\n }\r\n }\r\n (*g_169) = p_9;\r\n ++l_499;\r\n }\r\n else\r\n { \r\n int64_t *l_512[8] = {(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0};\r\n int32_t l_513[4];\r\n int64_t l_524[6][10][4] = {{{0x4C3D815AE43813BALL,0xF0C6CFC406DFDF2ALL,0x082E0D4638A26EFBLL,1L},{0xC881B6CB4CC4A5C9LL,0xE4DE13E2542AFC13LL,(-5L),0x2E1C39099BE1D503LL},{(-2L),(-1L),0xC881B6CB4CC4A5C9LL,(-1L)},{0xCD7E5718D9CE495DLL,0x25989EA0395394D5LL,0x25989EA0395394D5LL,0xCD7E5718D9CE495DLL},{0xF0C6CFC406DFDF2ALL,0xF884BC59129141D1LL,(-1L),(-1L)},{9L,0xE1735034541A91DALL,1L,0x082E0D4638A26EFBLL},{(-2L),3L,0xC42F5F76941C79BDLL,0x082E0D4638A26EFBLL},{0x25989EA0395394D5LL,0xE1735034541A91DALL,1L,0x19EE989D2DF288E2LL},{0xCF630E151FD4A638LL,1L,0x4C3D815AE43813BALL,(-3L)},{(-1L),(-7L),9L,0xE4DE13E2542AFC13LL}},{{1L,0x19EE989D2DF288E2LL,1L,0xCF630E151FD4A638LL},{0L,(-5L),0xA72890E95F62B265LL,0xE1735034541A91DALL},{0xCD7E5718D9CE495DLL,0xA726B99D96E26521LL,0L,0x49C30984D15A5727LL},{9L,(-1L),0L,(-1L)},{0xF884BC59129141D1LL,0xCD7E5718D9CE495DLL,0xF884BC59129141D1LL,3L},{3L,0xC881B6CB4CC4A5C9LL,0xE1735034541A91DALL,7L},{(-1L),0x082E0D4638A26EFBLL,(-10L),0xC881B6CB4CC4A5C9LL},{1L,0x2E1C39099BE1D503LL,(-10L),(-1L)},{(-1L),(-1L),0xE1735034541A91DALL,1L},{3L,(-1L),0xF884BC59129141D1LL,0x5333F827537ACEF5LL}},{{0xF884BC59129141D1LL,0x5333F827537ACEF5LL,0L,(-7L)},{9L,0x25989EA0395394D5LL,0L,1L},{0xCD7E5718D9CE495DLL,1L,0xA72890E95F62B265LL,0x4C3D815AE43813BALL},{0L,(-2L),1L,0L},{1L,0xC42F5F76941C79BDLL,9L,9L},{(-1L),(-1L),0x4C3D815AE43813BALL,(-2L)},{0xCF630E151FD4A638LL,0x49C30984D15A5727LL,1L,1L},{0x25989EA0395394D5LL,0L,0xC42F5F76941C79BDLL,1L},{(-2L),0L,1L,1L},{0L,0x49C30984D15A5727LL,0xE4DE13E2542AFC13LL,(-2L)}},{{0xA726B99D96E26521LL,(-1L),(-7L),9L},{(-3L),0xC42F5F76941C79BDLL,0xF0C6CFC406DFDF2ALL,0L},{(-1L),(-2L),0x5333F827537ACEF5LL,0x4C3D815AE43813BALL},{0xF0C6CFC406DFDF2ALL,1L,0x49C30984D15A5727LL,1L},{(-1L),0x25989EA0395394D5LL,1L,(-7L)},{0L,0x5333F827537ACEF5LL,(-1L),0x5333F827537ACEF5LL},{0x49C30984D15A5727LL,(-1L),0x2E1C39099BE1D503LL,1L},{(-7L),(-1L),0L,(-1L)},{(-2L),0x2E1C39099BE1D503LL,(-2L),0xC881B6CB4CC4A5C9LL},{(-2L),0x082E0D4638A26EFBLL,0L,7L}},{{(-7L),0xC881B6CB4CC4A5C9LL,0x2E1C39099BE1D503LL,3L},{0x49C30984D15A5727LL,0xCD7E5718D9CE495DLL,(-1L),(-1L)},{0L,(-1L),1L,0x49C30984D15A5727LL},{(-1L),0xA726B99D96E26521LL,0x49C30984D15A5727LL,0xE1735034541A91DALL},{0xF0C6CFC406DFDF2ALL,(-5L),0x5333F827537ACEF5LL,0xCF630E151FD4A638LL},{(-1L),0x19EE989D2DF288E2LL,0xF0C6CFC406DFDF2ALL,0xE4DE13E2542AFC13LL},{(-3L),(-7L),(-7L),(-3L)},{0xA726B99D96E26521LL,1L,0xE4DE13E2542AFC13LL,0x19EE989D2DF288E2LL},{0L,0xE1735034541A91DALL,1L,0x082E0D4638A26EFBLL},{(-2L),3L,0xC42F5F76941C79BDLL,0x082E0D4638A26EFBLL}},{{0x25989EA0395394D5LL,0xE1735034541A91DALL,1L,0x19EE989D2DF288E2LL},{0xCF630E151FD4A638LL,1L,0x4C3D815AE43813BALL,(-3L)},{(-1L),(-7L),9L,0xE4DE13E2542AFC13LL},{0xE1735034541A91DALL,(-1L),(-1L),0L},{(-2L),0x5333F827537ACEF5LL,0x4C3D815AE43813BALL,3L},{(-3L),0x2E1C39099BE1D503LL,7L,9L},{0L,0xF884BC59129141D1LL,0x25989EA0395394D5LL,0xE4DE13E2542AFC13LL},{1L,(-3L),1L,(-10L)},{(-10L),0xF0C6CFC406DFDF2ALL,3L,(-1L)},{1L,0x49C30984D15A5727LL,1L,0xF0C6CFC406DFDF2ALL}}};\r\n int16_t l_549 = 8L;\r\n int8_t l_567 = 0x65L;\r\n int i, j, k;\r\n for (i = 0; i < 4; i++)\r\n l_513[i] = 0x50CAB1D2L;\r\n (*g_169) |= 1L;\r\n for (g_12 = 0; (g_12 != 9); g_12++)\r\n { \r\n for (g_426 = 17; (g_426 == (-2)); g_426 = safe_sub_func_uint16_t_u_u(g_426, 3))\r\n { \r\n uint8_t l_523[10][4][6] = {{{0xF4L,0x83L,0UL,0xF4L,0xF1L,0x11L},{255UL,0xF4L,0UL,0xF4L,255UL,0xB7L},{0xF4L,255UL,0xB7L,0xC9L,255UL,0x6DL},{0x83L,0xF4L,0xF1L,255UL,0xF1L,0x6DL}},{{0xF1L,0x83L,0xB7L,0x11L,0x11L,0xB7L},{0xF1L,0xF1L,0UL,255UL,253UL,0x11L},{0x83L,0xF1L,0UL,0xC9L,0x11L,0UL},{0xF4L,0x83L,0UL,0xF4L,0xF1L,0x11L}},{{255UL,0xF4L,0UL,0xF4L,255UL,0xB7L},{0xF4L,255UL,0xB7L,0xC9L,255UL,0x6DL},{0x83L,0xF4L,0xF1L,255UL,0xF1L,0x6DL},{0xF1L,0x83L,0xB7L,0x11L,0x11L,0xB7L}},{{0xF1L,0xF1L,0UL,255UL,253UL,0x11L},{0x83L,0xF1L,0UL,0xC9L,0x11L,0UL},{0xF4L,0x83L,0UL,0xF4L,0xF1L,0x11L},{255UL,0xF4L,0UL,0xF4L,255UL,0xB7L}},{{0xF4L,255UL,0xB7L,0xC9L,255UL,0x6DL},{0x83L,0xF4L,0xF1L,255UL,0xF1L,0x6DL},{0xF1L,0x83L,0xB7L,0x11L,0x11L,0xB7L},{0xF1L,0xF1L,0UL,255UL,253UL,0x11L}},{{0x83L,0xF1L,0UL,0xC9L,0x11L,0UL},{0xF4L,0x83L,0UL,0xF4L,0xF1L,0x11L},{255UL,0xF4L,0UL,0xF4L,255UL,0xB7L},{0xF4L,255UL,0xB7L,0xC9L,255UL,0x6DL}},{{0x83L,0xF4L,0xF1L,255UL,0xF1L,0x6DL},{0xF1L,0x83L,0xB7L,0x11L,0x11L,0xB7L},{0xF1L,0xF1L,0UL,255UL,253UL,0x11L},{0x83L,0x79L,255UL,255UL,0x80L,0xF1L}},{{8UL,1UL,255UL,8UL,0x79L,0xC9L},{253UL,8UL,0xF1L,8UL,253UL,0x11L},{8UL,253UL,0x11L,255UL,253UL,0x83L},{1UL,8UL,253UL,253UL,0x79L,0x83L}},{{0x79L,1UL,0x11L,0x80L,0x80L,0x11L},{0x79L,0x79L,0xF1L,253UL,0x5FL,0xC9L},{1UL,0x79L,255UL,255UL,0x80L,0xF1L},{8UL,1UL,255UL,8UL,0x79L,0xC9L}},{{253UL,8UL,0xF1L,8UL,253UL,0x11L},{8UL,253UL,0x11L,255UL,253UL,0x83L},{1UL,8UL,253UL,253UL,0x79L,0x83L},{0x79L,1UL,0x11L,0x80L,0x80L,0x11L}}};\r\n int i, j, k;\r\n (**g_256) = ((safe_div_func_int8_t_s_s(1L, ((p_9 , ((safe_sub_func_int32_t_s_s(p_9, ((safe_add_func_int16_t_s_s((l_512[7] == (void*)0), ((*l_310) = (g_165 | l_513[1])))) | (safe_add_func_uint64_t_u_u(p_9, (safe_sub_func_uint32_t_u_u((safe_sub_func_int8_t_s_s(((!(safe_add_func_uint64_t_u_u(((p_9 <= 0L) ^ g_33[8][0][8]), (-1L)))) < l_513[0]), g_149)), l_513[1]))))))) , (-1L))) && g_426))) ^ l_523[3][2][5]);\r\n return l_524[3][2][1];\r\n }\r\n for (g_426 = 0; (g_426 < 8); g_426++)\r\n { \r\n (**g_255) = &l_513[2];\r\n }\r\n }\r\n for (g_426 = 26; (g_426 == 5); g_426 = safe_sub_func_uint8_t_u_u(g_426, 8))\r\n { \r\n int32_t l_534 = (-1L);\r\n int32_t l_536 = 0xD26B6040L;\r\n uint64_t l_537 = 18446744073709551610UL;\r\n for (g_407 = 25; (g_407 < 30); g_407 = safe_add_func_uint64_t_u_u(g_407, 8))\r\n { \r\n uint16_t l_533 = 0xD203L;\r\n int32_t l_535 = (-4L);\r\n (**g_256) = (safe_add_func_uint16_t_u_u((l_533 != (((l_533 > p_9) > p_9) ^ l_533)), l_513[1]));\r\n l_537++;\r\n return l_535;\r\n }\r\n for (l_537 = 0; (l_537 >= 29); l_537++)\r\n { \r\n uint32_t *l_564[5][7][7] = {{{&g_94[2],&g_380,&g_149,&g_380,&g_94[0],&g_380,&g_149},{&g_149,&g_149,(void*)0,&g_94[2],&g_380,&g_94[2],&g_380},{&g_94[1],&g_149,&g_149,&g_94[2],&g_94[2],&g_380,&g_380},{&g_94[2],&g_380,&g_380,&g_94[0],&g_380,&g_94[2],&g_149},{&g_94[2],&g_94[2],&g_94[2],(void*)0,&g_94[0],(void*)0,&g_149},{&g_94[1],&g_149,&g_94[2],&g_94[0],&g_149,(void*)0,&g_149},{&g_149,&g_94[2],(void*)0,&g_94[2],&g_94[2],(void*)0,&g_149}},{{&g_94[2],&g_380,&g_94[2],&g_94[2],&g_94[0],(void*)0,&g_380},{&g_94[2],&g_94[2],&g_94[2],&g_380,&g_380,(void*)0,&g_380},{&g_380,&g_380,&g_380,&g_94[2],&g_380,(void*)0,&g_149},{&g_94[2],&g_94[2],&g_149,&g_380,&g_94[0],&g_94[2],&g_149},{&g_380,&g_149,(void*)0,&g_380,&g_94[2],&g_380,(void*)0},{&g_94[2],&g_94[2],&g_149,&g_94[2],&g_149,&g_94[2],&g_149},{&g_94[2],&g_380,&g_149,&g_380,&g_94[0],&g_380,&g_149}},{{&g_149,&g_149,(void*)0,&g_94[2],&g_380,&g_94[2],&g_380},{&g_94[1],&g_149,&g_149,&g_94[2],&g_94[2],&g_380,&g_380},{&g_94[2],&g_380,&g_94[2],&g_94[0],&g_94[0],&g_94[2],&g_149},{&g_380,&g_149,&g_380,&g_149,&g_380,&g_149,&g_94[2]},{&g_94[2],(void*)0,&g_149,&g_94[0],&g_94[1],&g_380,&g_94[2]},{&g_94[0],&g_380,&g_94[2],&g_94[2],&g_149,&g_149,&g_149},{&g_149,&g_94[2],&g_149,&g_380,&g_94[2],&g_149,&g_94[2]}},{{&g_94[2],(void*)0,&g_380,&g_94[2],(void*)0,&g_380,&g_94[1]},{&g_380,&g_94[2],&g_94[2],&g_94[2],(void*)0,&g_149,&g_380},{&g_149,&g_380,&g_94[2],&g_149,&g_94[2],&g_94[2],&g_94[2]},{&g_380,(void*)0,&g_94[2],&g_149,&g_149,&g_380,&g_149},{&g_94[2],&g_149,&g_380,&g_94[2],&g_94[1],&g_94[2],&g_94[2]},{&g_149,&g_380,&g_380,&g_94[2],&g_380,&g_94[2],&g_380},{&g_94[0],&g_94[0],&g_94[2],&g_380,&g_94[0],&g_94[2],&g_94[1]}},{{&g_94[2],&g_94[0],&g_94[2],&g_94[2],&g_94[2],&g_380,&g_94[2]},{&g_380,&g_380,&g_94[2],&g_94[0],&g_94[0],&g_94[2],&g_149},{&g_380,&g_149,&g_380,&g_149,&g_380,&g_149,&g_94[2]},{&g_94[2],(void*)0,&g_149,&g_94[0],&g_94[1],&g_380,&g_94[2]},{&g_94[0],&g_380,&g_94[2],&g_94[2],&g_149,&g_149,&g_149},{&g_149,&g_94[2],&g_149,&g_380,&g_94[2],&g_149,&g_94[2]},{&g_94[2],(void*)0,&g_380,&g_94[2],(void*)0,&g_380,&g_94[1]}}};\r\n int32_t l_565 = 0x56C2B96EL;\r\n uint16_t l_566[10][7] = {{0x136DL,0UL,0x0D3AL,0UL,0x136DL,0xF559L,4UL},{0UL,0UL,0xE1E7L,0x5533L,2UL,0x5533L,0xE1E7L},{4UL,4UL,0UL,0x0D3AL,0xFD06L,0xFB9BL,0xF306L},{0UL,0x5533L,1UL,1UL,0x5533L,0UL,2UL},{0x136DL,0UL,0UL,0xDCD5L,0xFD06L,0xFD06L,0xDCD5L},{2UL,0x92D3L,2UL,65535UL,2UL,0xE80DL,0UL},{0UL,0UL,0x136DL,0xFB9BL,0x136DL,0UL,0UL},{1UL,0x5533L,0UL,2UL,0x852DL,0xE80DL,0x852DL},{0UL,4UL,4UL,4UL,0xF559L,0x136DL,0UL},{0x92D3L,0xE80DL,2UL,0x5533L,0x5533L,2UL,0xE80DL}};\r\n int i, j, k;\r\n l_567 = (safe_div_func_int32_t_s_s(((*l_312) ^= (!(**g_372))), (safe_lshift_func_int32_t_s_s((p_9 == ((((((*l_314) ^= l_549) || (((--(*l_30)) && ((*g_373) >= (safe_sub_func_uint8_t_u_u(p_9, ((*l_319) = (((safe_mod_func_int64_t_s_s(((*l_425) = (safe_mod_func_uint16_t_u_u(((l_536 = ((safe_mul_func_uint16_t_u_u(g_346[3], (((((safe_sub_func_int8_t_s_s((safe_add_func_uint32_t_u_u((&g_380 != l_564[2][0][2]), ((0xDD250ABB124CD325LL < g_409) >= (*g_373)))), g_407)) , (*g_373)) < l_565) , &g_185) != (void*)0))) , p_9)) < p_9), p_9))), 0xEEC46540531036A0LL)) == 4294967288UL) < g_412)))))) != 0x6A948BE95CCE685BLL)) != p_9) , l_566[0][3]) , g_366[3][0])), p_9))));\r\n (**g_256) |= ((void*)0 != &g_407);\r\n if (p_9)\r\n break;\r\n }\r\n }\r\n }\r\n return p_9;\r\n}\r\n\r\n\r\n\r\n\r\nstatic int16_t func_15(int8_t p_16, uint32_t p_17, int32_t p_18)\r\n{ \r\n uint16_t l_43[8][6] = {{4UL,0x864FL,65527UL,0x864FL,4UL,65527UL},{0x864FL,4UL,65527UL,0x11FBL,0x11FBL,65527UL},{0x11FBL,0x11FBL,65527UL,4UL,0x864FL,65527UL},{4UL,0x864FL,65527UL,0x864FL,4UL,65527UL},{0x864FL,4UL,65527UL,0x11FBL,0x11FBL,65527UL},{0x11FBL,0x11FBL,65527UL,4UL,0x864FL,65527UL},{4UL,0x864FL,65527UL,0x864FL,4UL,65527UL},{0x864FL,4UL,65527UL,0x11FBL,0x11FBL,65527UL}};\r\n int32_t l_54 = 1L;\r\n int32_t l_55 = 0x1814A667L;\r\n int32_t l_57 = 0xF053B50BL;\r\n int32_t l_58[8];\r\n int64_t *l_83 = &g_66[3][3];\r\n int32_t l_106[2];\r\n int32_t *l_111 = (void*)0;\r\n int32_t **l_110[9] = {(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0};\r\n int32_t ***l_109 = &l_110[8];\r\n const int32_t l_145 = (-1L);\r\n int64_t *l_156 = &g_157[0];\r\n int8_t *l_164 = &g_165;\r\n uint16_t l_166 = 4UL;\r\n int16_t *l_167 = &g_168;\r\n int64_t *l_171 = (void*)0;\r\n int64_t *l_172 = &g_157[8];\r\n uint32_t l_196 = 4294967295UL;\r\n uint16_t l_199 = 4UL;\r\n int8_t l_288 = 0L;\r\n int8_t l_291 = (-5L);\r\n uint32_t l_302 = 18446744073709551615UL;\r\n int i, j;\r\n for (i = 0; i < 8; i++)\r\n l_58[i] = 0xE0FBC7DBL;\r\n for (i = 0; i < 2; i++)\r\n l_106[i] = (-1L);\r\nlbl_244:\r\n for (g_31 = 0; (g_31 <= 0); g_31 += 1)\r\n { \r\n int32_t l_52[5] = {0x50B10B9BL,0x50B10B9BL,0x50B10B9BL,0x50B10B9BL,0x50B10B9BL};\r\n uint16_t l_108 = 1UL;\r\n int i;\r\n if (g_11[(g_31 + 1)])\r\n break;\r\n if (g_11[(g_31 + 1)])\r\n break;\r\n for (p_18 = 0; (p_18 <= 0); p_18 += 1)\r\n { \r\n int32_t *l_38 = &g_11[(g_31 + 1)];\r\n int32_t l_53 = 0x2DAC39F6L;\r\n int32_t l_56[8][5][2] = {{{0xEDC4196DL,0x55470A06L},{0xFFFF0E59L,0L},{6L,0xEDC4196DL},{(-1L),9L},{2L,0x62CF0DA5L}},{{0x2A43FB2FL,0xFFFF0E59L},{0x7A36831AL,0x7A36831AL},{(-1L),6L},{0x28A4BD3CL,0xEF314FBCL},{0xFB4D153BL,0L}},{{0x1831F773L,0xFB4D153BL},{9L,0x0066401EL},{9L,0xFB4D153BL},{0x1831F773L,0L},{0xFB4D153BL,0xEF314FBCL}},{{0x28A4BD3CL,6L},{(-1L),0x7A36831AL},{0x7A36831AL,0xFFFF0E59L},{0x2A43FB2FL,0x62CF0DA5L},{2L,9L}},{{(-1L),0xEDC4196DL},{6L,0L},{0xFFFF0E59L,0x55470A06L},{0xEDC4196DL,0x55470A06L},{0xFFFF0E59L,0L}},{{6L,0xEDC4196DL},{(-1L),9L},{2L,0x62CF0DA5L},{0x2A43FB2FL,0xFFFF0E59L},{0x7A36831AL,0x7A36831AL}},{{(-1L),6L},{0x28A4BD3CL,0xEF314FBCL},{0xFB4D153BL,0L},{0x1831F773L,0xFB4D153BL},{9L,0x0066401EL}},{{9L,0xFB4D153BL},{0x1831F773L,0L},{0xFB4D153BL,0xEF314FBCL},{0x28A4BD3CL,6L},{(-1L),0x7A36831AL}}};\r\n uint8_t l_59[3];\r\n int i, j, k;\r\n for (i = 0; i < 3; i++)\r\n l_59[i] = 0x9FL;\r\n if (p_18)\r\n { \r\n const int32_t *l_34 = &g_11[0];\r\n int32_t *l_37 = &g_11[1];\r\n int32_t **l_36[5][1][4] = {{{&l_37,&l_37,&l_37,&l_37}},{{&l_37,&l_37,&l_37,&l_37}},{{&l_37,&l_37,&l_37,&l_37}},{{&l_37,&l_37,&l_37,&l_37}},{{&l_37,&l_37,&l_37,&l_37}}};\r\n int32_t ***l_39 = &l_36[3][0][1];\r\n int i, j, k;\r\n g_35 = l_34;\r\n g_11[g_31] = 0x10944B0CL;\r\n l_38 = &g_11[1];\r\n g_11[g_31] = (((*l_39) = (void*)0) == &l_38);\r\n }\r\n else\r\n { \r\n int32_t *l_40 = &g_11[0];\r\n int32_t *l_41 = &g_11[(g_31 + 1)];\r\n int32_t *l_42[3][7] = {{&g_11[(g_31 + 1)],&g_11[(g_31 + 1)],&g_11[1],&g_11[1],&g_11[(g_31 + 1)],&g_11[(g_31 + 1)],&g_11[1]},{&g_11[1],&g_11[(g_31 + 1)],&g_11[1],&g_11[(g_31 + 1)],&g_11[1],&g_11[(g_31 + 1)],&g_11[1]},{&g_11[(g_31 + 1)],&g_11[1],&g_11[1],&g_11[(g_31 + 1)],&g_11[(g_31 + 1)],&g_11[1],&g_11[1]}};\r\n int i, j;\r\n l_43[0][0]--;\r\n }\r\n for (p_17 = 0; (p_17 <= 0); p_17 += 1)\r\n { \r\n int32_t *l_46 = &g_11[1];\r\n int32_t *l_47 = &g_11[1];\r\n int32_t *l_48 = &g_11[1];\r\n int32_t *l_49 = &g_11[1];\r\n int32_t *l_50 = &g_11[0];\r\n int32_t *l_51[3][10] = {{(void*)0,(void*)0,&g_11[1],(void*)0,(void*)0,&g_11[1],(void*)0,(void*)0,&g_11[1],(void*)0},{(void*)0,&g_11[1],&g_11[1],(void*)0,&g_11[1],&g_11[1],(void*)0,&g_11[1],&g_11[1],(void*)0},{&g_11[1],(void*)0,&g_11[1],&g_11[1],(void*)0,&g_11[1],&g_11[1],(void*)0,&g_11[1],&g_11[1]}};\r\n int32_t **l_62[7][7] = {{&l_38,&l_38,&l_47,&l_38,&l_38,&l_47,&l_38},{(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0},{&l_51[1][6],&l_38,&l_51[1][6],&l_51[1][6],&l_38,&l_51[1][6],&l_51[1][6]},{(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0},{&l_38,&l_51[1][6],&l_51[1][6],&l_38,&l_51[1][6],&l_51[1][6],&l_38},{(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0,(void*)0},{&l_38,&l_38,&l_47,&l_38,&l_38,&l_47,&l_38}};\r\n int i, j;\r\n --l_59[0];\r\n g_35 = l_51[1][6];\r\n }\r\n if ((*l_38))\r\n continue;\r\n for (l_53 = 0; (l_53 <= 1); l_53 += 1)\r\n { \r\n int i, j, k;\r\n return l_56[(g_31 + 6)][(l_53 + 2)][p_18];\r\n }\r\n for (l_54 = 0; (l_54 >= 0); l_54 -= 1)\r\n { \r\n int64_t *l_65 = &g_66[3][3];\r\n int64_t **l_84 = (void*)0;\r\n int64_t **l_85 = &l_65;\r\n int64_t *l_87[8] = {&g_66[2][3],&g_66[4][1],&g_66[4][1],&g_66[2][3],&g_66[4][1],&g_66[4][1],&g_66[2][3],&g_66[4][1]};\r\n int64_t **l_86 = &l_87[7];\r\n uint32_t *l_93[3];\r\n int32_t l_104 = 0L;\r\n int32_t l_105[1];\r\n uint32_t l_107 = 0UL;\r\n int32_t l_112 = 0x2EFCB156L;\r\n int i;\r\n for (i = 0; i < 3; i++)\r\n l_93[i] = &g_94[2];\r\n for (i = 0; i < 1; i++)\r\n l_105[i] = 0L;\r\n l_112 &= ((((safe_mul_func_uint8_t_u_u(((((*l_65) &= 1L) , (safe_div_func_int32_t_s_s((safe_div_func_uint32_t_u_u((((((((safe_rshift_func_uint32_t_u_s(g_11[(g_31 + 1)], ((safe_div_func_int8_t_s_s(((((safe_mul_func_int64_t_s_s(((safe_sub_func_int16_t_s_s(p_16, (p_16 <= ((!(safe_mul_func_int64_t_s_s(((*l_83) = (~(l_105[0] ^= ((((*l_85) = l_83) == ((*l_86) = (void*)0)) >= ((((((+(safe_mul_func_uint8_t_u_u(0x1FL, (safe_sub_func_uint32_t_u_u((g_94[2]--), (+(safe_mul_func_int64_t_s_s(((safe_mod_func_uint32_t_u_u(g_11[1], (safe_sub_func_uint16_t_u_u(((g_12 <= g_11[(g_31 + 1)]) >= g_12), 7L)))) != p_16), l_55)))))))) == l_104) , 0x02DFE65AL) , p_17) , g_12) ^ 0xA833846322F5C3AALL))))), g_31))) < l_106[0])))) , l_107), 0UL)) , l_104) & 0x93ACL) > p_16), g_11[1])) , l_108))) , 4L) <= l_58[7]) , l_109) != (void*)0) & p_16) < p_16), 4294967291UL)), p_17))) & p_16), g_12)) != l_52[3]) , p_17) ^ 0x0A3500CD5182C662LL);\r\n }\r\n }\r\n }\r\n for (p_17 = 1; (p_17 <= 5); p_17 += 1)\r\n { \r\n int64_t * const *l_114 = (void*)0;\r\n int32_t l_115 = 0xC0B326B5L;\r\n int32_t l_122 = (-1L);\r\n int32_t l_128 = 3L;\r\n int32_t l_129 = (-6L);\r\n int32_t ***l_144 = &l_110[3];\r\n for (g_12 = 0; (g_12 <= 5); g_12 += 1)\r\n { \r\n int i, j;\r\n if (l_43[g_12][g_12])\r\n break;\r\n for (p_18 = 0; (p_18 <= 8); p_18 += 1)\r\n { \r\n return p_17;\r\n }\r\n for (g_31 = 0; (g_31 <= 7); g_31 += 1)\r\n { \r\n int i;\r\n l_58[g_31] = (~p_18);\r\n if (p_16)\r\n continue;\r\n g_11[1] = (((0x33A8E694L & p_18) , 0xE34AC405L) & (((void*)0 == l_114) , 0x7FEED684L));\r\n }\r\n }\r\n if (p_16)\r\n { \r\n uint64_t l_116 = 0xDCA698BDAB2CB753LL;\r\n int32_t l_119 = (-7L);\r\n int32_t l_123 = 0xDAE3C54FL;\r\n int32_t l_125 = 0x5C04B26AL;\r\n int32_t l_127[4][3] = {{1L,1L,0x4C425AE9L},{1L,1L,0x4C425AE9L},{1L,1L,0x4C425AE9L},{1L,1L,0x4C425AE9L}};\r\n uint32_t *l_146 = (void*)0;\r\n uint32_t *l_147 = (void*)0;\r\n uint32_t *l_148 = &g_149;\r\n int i, j;\r\n l_115 = l_115;\r\n for (p_18 = 1; (p_18 <= 5); p_18 += 1)\r\n { \r\n int16_t l_120 = 0xDB9CL;\r\n int32_t l_121 = 0x49656563L;\r\n int32_t l_124[1][4] = {{0x080E9152L,0x080E9152L,0x080E9152L,0x080E9152L}};\r\n int32_t l_126 = (-6L);\r\n uint16_t l_130 = 0x6C88L;\r\n int i, j;\r\n --l_116;\r\n l_130++;\r\n }\r\n g_11[0] = ((l_127[1][0] = ((1L ^ (safe_mod_func_int16_t_s_s(((g_94[1] , (safe_add_func_int16_t_s_s(((l_83 != &l_116) < ((((+((*l_148) ^= (g_94[1] = ((safe_mul_func_int32_t_s_s(((((safe_mul_func_int32_t_s_s((g_11[1] , (g_33[8][0][8] > (((((safe_mod_func_uint8_t_u_u(0xB6L, ((g_31 <= (((l_115 >= 1UL) < g_94[2]) < p_16)) , p_17))) , 0xB692CA01L) , &l_110[8]) == l_144) | g_94[2]))), g_31)) , (void*)0) != (void*)0) , p_17), l_145)) | 0x7FL)))) != 1L) != l_123) && p_17)), p_18))) || p_16), g_12))) || (-1L))) == 4294967292UL);\r\n }\r\n else\r\n { \r\n int32_t **l_150 = (void*)0;\r\n l_150 = (*l_144);\r\n }\r\n return g_33[6][0][7];\r\n }\r\n if (((g_66[3][3] && 0x44L) | (((*l_172) = (&g_11[0] == (g_170 = (g_169 = ((!(safe_lshift_func_int8_t_s_u(((safe_mul_func_uint8_t_u_u((g_12 , (((l_83 == (l_156 = l_83)) || g_94[1]) == (((*l_167) = (safe_div_func_uint16_t_u_u((0x63AD2DFDL > (safe_mul_func_int8_t_s_s(((*l_164) ^= (safe_add_func_uint64_t_u_u(g_33[2][0][2], 0UL))), l_166))), g_33[8][0][8]))) <= g_11[1]))), p_17)) >= (-1L)), 6))) , &g_11[1]))))) , 0L)))\r\n { \r\n uint16_t l_173 = 1UL;\r\n uint32_t *l_186[9];\r\n int64_t *l_188[5] = {&g_66[3][3],&g_66[3][3],&g_66[3][3],&g_66[3][3],&g_66[3][3]};\r\n int32_t l_193 = 1L;\r\n int32_t ***l_257 = &g_256;\r\n uint64_t *l_276[9] = {&g_33[7][0][8],&g_33[7][0][8],&g_33[7][0][8],&g_33[7][0][8],&g_33[7][0][8],&g_33[7][0][8],&g_33[7][0][8],&g_33[7][0][8],&g_33[7][0][8]};\r\n int32_t l_283 = (-9L);\r\n int32_t l_287 = 0L;\r\n int32_t l_292 = 0x1C44CF0EL;\r\n int32_t l_293 = 0x11BC2BF2L;\r\n int32_t l_295 = 0xE1623BE8L;\r\n int32_t l_296[1][9] = {{0xD646E06EL,0xD646E06EL,0x34B75A98L,0xD646E06EL,0xD646E06EL,0x34B75A98L,0xD646E06EL,0xD646E06EL,0x34B75A98L}};\r\n uint16_t l_299 = 0x0AFDL;\r\n uint16_t l_305 = 0xFAD6L;\r\n int i, j;\r\n for (i = 0; i < 9; i++)\r\n l_186[i] = &g_149;\r\nlbl_267:\r\n ++l_173;\r\n for (l_55 = 0; (l_55 <= 5); l_55 += 1)\r\n { \r\n uint8_t *l_184 = &g_185;\r\n int64_t * const l_187 = &g_66[3][3];\r\n int64_t **l_191 = &l_172;\r\n const uint8_t l_192 = 0x7EL;\r\n int32_t l_236 = 1L;\r\n int i;\r\n if ((safe_div_func_uint8_t_u_u((safe_mod_func_int32_t_s_s((*g_170), ((safe_mod_func_uint8_t_u_u((((*l_184) = (l_164 != &p_16)) || (((((((l_186[6] = l_111) != &g_94[2]) , l_187) == l_188[3]) == (safe_rshift_func_uint32_t_u_u(g_33[2][0][5], (((*l_191) = &g_157[3]) != &g_66[3][3])))) > 0xB3L) != l_173)), 3L)) , p_17))), l_192)))\r\n { \r\n int32_t l_194 = 0xBE4B794FL;\r\n int32_t l_195[8] = {0xFA8EE88EL,0xFA8EE88EL,0x2CC03310L,0xFA8EE88EL,0xFA8EE88EL,0x2CC03310L,0xFA8EE88EL,0xFA8EE88EL};\r\n int i;\r\n ++l_196;\r\n g_170 = &g_11[1];\r\n return g_33[6][0][6];\r\n }\r\n else\r\n { \r\n --l_199;\r\n }\r\n for (g_165 = 0; (g_165 <= 8); g_165 += 1)\r\n { \r\n int i, j;\r\n (*g_170) ^= l_43[l_55][l_55];\r\n return g_157[(l_55 + 1)];\r\n }\r\n for (g_12 = 1; (g_12 <= 5); g_12 += 1)\r\n { \r\n uint64_t l_239 = 0xCCD539717E3722A4LL;\r\n int i, j;\r\n (*g_169) = (safe_mod_func_int8_t_s_s((safe_mul_func_int8_t_s_s((safe_div_func_int32_t_s_s((safe_sub_func_uint32_t_u_u((safe_mod_func_uint16_t_u_u((safe_rshift_func_uint16_t_u_u(((l_193 = (((safe_mod_func_uint64_t_u_u(p_16, (safe_sub_func_uint8_t_u_u(((*l_184)++), ((((safe_unary_minus_func_uint32_t_u((((safe_mul_func_int16_t_s_s((0L ^ l_173), ((((safe_sub_func_uint16_t_u_u((safe_add_func_int64_t_s_s(p_17, (((safe_sub_func_uint16_t_u_u(p_18, (l_43[(g_12 + 1)][g_12] |= (((safe_add_func_int16_t_s_s(((g_33[0][0][2]++) || (!l_173)), ((l_236 = 0x9C1AF257L) >= (*g_169)))) < (safe_lshift_func_int32_t_s_u((*g_169), 5))) || g_33[8][0][8])))) < l_193) ^ g_11[1]))), g_168)) , g_149) , g_66[3][3]) , 0x3A3CL))) & l_239) , p_17))) < l_239) > l_193) >= 18446744073709551607UL))))) , (-1L)) != g_12)) ^ 0x81AF9209L), 2)), g_165)), (*g_169))), g_157[2])), g_165)), g_94[2]));\r\n (*g_170) = p_18;\r\n }\r\n for (l_173 = 1; (l_173 <= 5); l_173 += 1)\r\n { \r\n l_193 = (-1L);\r\n return g_33[8][0][8];\r\n }\r\n }\r\n for (g_165 = 0; (g_165 >= (-25)); g_165--)\r\n { \r\n uint8_t *l_251[10][10][2] = {{{&g_185,(void*)0},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,(void*)0},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,(void*)0},{&g_185,&g_185}},{{(void*)0,&g_185},{&g_185,(void*)0},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,(void*)0},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185}},{{&g_185,&g_185},{&g_185,&g_185},{&g_185,(void*)0},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,(void*)0},{&g_185,&g_185}},{{&g_185,&g_185},{(void*)0,&g_185},{&g_185,(void*)0},{(void*)0,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{(void*)0,&g_185}},{{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{(void*)0,(void*)0},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,&g_185},{&g_185,(void*)0}},{{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,(void*)0},{&g_185,(void*)0},{&g_185,(void*)0},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,&g_185}},{{&g_185,&g_185},{&g_185,(void*)0},{&g_185,&g_185},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,(void*)0},{(void*)0,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185}},{{&g_185,&g_185},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{(void*)0,(void*)0},{&g_185,&g_185}},{{(void*)0,&g_185},{&g_185,&g_185},{&g_185,(void*)0},{&g_185,&g_185},{&g_185,&g_185},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,(void*)0},{&g_185,(void*)0},{&g_185,(void*)0}},{{&g_185,&g_185},{(void*)0,&g_185},{&g_185,&g_185},{&g_185,&g_185},{&g_185,(void*)0},{&g_185,&g_185},{&g_185,&g_185},{(void*)0,&g_185},{&g_185,(void*)0},{(void*)0,&g_185}}};\r\n int32_t l_252[2];\r\n int32_t l_266 = (-9L);\r\n int64_t *l_274 = &g_157[7];\r\n int32_t l_289[7] = {0xA2D0A057L,0xA2D0A057L,0xA2D0A057L,0xA2D0A057L,0xA2D0A057L,0xA2D0A057L,0xA2D0A057L};\r\n int i, j, k;\r\n for (i = 0; i < 2; i++)\r\n l_252[i] = 0L;\r\n for (p_16 = (-8); (p_16 > (-2)); p_16++)\r\n { \r\n if (g_168)\r\n goto lbl_244;\r\n }\r\n if (((safe_sub_func_int16_t_s_s((((safe_add_func_int16_t_s_s((safe_lshift_func_uint8_t_u_s((l_252[1] ^= (g_149 , p_16)), (((p_16 > (((0xA6BC4429L & (g_255 != l_257)) ^ (((safe_mul_func_int64_t_s_s((g_33[8][0][8] || ((safe_add_func_uint64_t_u_u((safe_mul_func_uint64_t_u_u(((safe_add_func_int8_t_s_s(g_12, l_266)) , 0x221D217DB600C3A3LL), 0xF75A03B73B3FD032LL)), g_33[8][0][8])) | g_157[0])), 0x36BB0D7DC54451A0LL)) != 0xE35EL) <= g_149)) < 4294967293UL)) | p_16) | l_266))), g_33[3][0][6])) > 1UL) , g_33[1][0][6]), g_66[1][1])) | p_16))\r\n { \r\n if ((***g_255))\r\n break;\r\n }\r\n else\r\n { \r\n uint16_t l_270[10] = {0x99BCL,0x1F3FL,0x2820L,0x2820L,0x1F3FL,0x99BCL,0x1F3FL,0x2820L,0x2820L,0x1F3FL};\r\n int64_t **l_275 = &l_188[4];\r\n int32_t l_284 = (-1L);\r\n int32_t l_285 = 1L;\r\n int32_t l_286 = 0x8B011B8DL;\r\n int32_t l_290 = 1L;\r\n int32_t l_294 = 0x4EF4180CL;\r\n int32_t l_297 = (-1L);\r\n int32_t l_298[9] = {9L,9L,9L,9L,9L,9L,9L,9L,9L};\r\n int i;\r\n if (l_55)\r\n goto lbl_267;\r\n (***g_255) = (g_31 == g_66[3][3]);\r\n (***l_257) ^= ((safe_add_func_int32_t_s_s(l_270[0], (((safe_sub_func_int32_t_s_s((0x1737AA7C1C7358A4LL && g_157[3]), (((safe_unary_minus_func_int16_t_s(((((*l_275) = l_274) != (void*)0) != ((void*)0 != l_276[5])))) <= (safe_rshift_func_int32_t_s_s((safe_div_func_uint64_t_u_u((safe_add_func_uint16_t_u_u(g_33[8][0][8], g_94[2])), l_270[0])), l_252[1]))) & g_168))) , 4294967295UL) , p_16))) , p_16);\r\n --l_299;\r\n }\r\n l_302++;\r\n g_306 &= (l_305 = (***l_257));\r\n }\r\n }\r\n else\r\n { \r\n uint8_t l_307[9][7] = {{0x0DL,0x0DL,1UL,0xE6L,1UL,1UL,1UL},{0xE6L,255UL,255UL,0xE6L,0UL,246UL,0xE6L},{246UL,1UL,0UL,0UL,0UL,0x22L,1UL},{0UL,255UL,0x0DL,246UL,246UL,0x0DL,255UL},{0UL,1UL,0x22L,0UL,9UL,9UL,0UL},{0x22L,255UL,0x22L,9UL,255UL,1UL,1UL},{255UL,0UL,0x0DL,0UL,255UL,0x0DL,246UL},{246UL,1UL,9UL,246UL,9UL,1UL,246UL},{0x22L,246UL,1UL,9UL,246UL,9UL,1UL}};\r\n int i, j;\r\n --l_307[0][5];\r\n }\r\n return p_16;\r\n}\r\n\r\n\r\n\r\n\r\n\r\nint main (int argc, char* argv[])\r\n{\r\n int i, j, k;\r\n int print_hash_value = 0;\r\n if (argc == 2 && strcmp(argv[1], \"1\") == 0) print_hash_value = 1;\r\n platform_main_begin();\r\n crc32_gentab();\r\n func_1();\r\n for (i = 0; i < 2; i++)\r\n {\r\n transparent_crc(g_11[i], \"g_11[i]\", print_hash_value);\r\n if (print_hash_value) printf(\"index = [%d]\\n\", i);\r\n\r\n }\r\n transparent_crc(g_12, \"g_12\", print_hash_value);\r\n transparent_crc(g_31, \"g_31\", print_hash_value);\r\n for (i = 0; i < 9; i++)\r\n {\r\n for (j = 0; j < 1; j++)\r\n {\r\n for (k = 0; k < 9; k++)\r\n {\r\n transparent_crc(g_33[i][j][k], \"g_33[i][j][k]\", print_hash_value);\r\n if (print_hash_value) printf(\"index = [%d][%d][%d]\\n\", i, j, k);\r\n\r\n }\r\n }\r\n }\r\n for (i = 0; i < 5; i++)\r\n {\r\n for (j = 0; j < 4; j++)\r\n {\r\n transparent_crc(g_66[i][j], \"g_66[i][j]\", print_hash_value);\r\n if (print_hash_value) printf(\"index = [%d][%d]\\n\", i, j);\r\n\r\n }\r\n }\r\n for (i = 0; i < 3; i++)\r\n {\r\n transparent_crc(g_94[i], \"g_94[i]\", print_hash_value);\r\n if (print_hash_value) printf(\"index = [%d]\\n\", i);\r\n\r\n }\r\n transparent_crc(g_149, \"g_149\", print_hash_value);\r\n for (i = 0; i < 9; i++)\r\n {\r\n transparent_crc(g_157[i], \"g_157[i]\", print_hash_value);\r\n if (print_hash_value) printf(\"index = [%d]\\n\", i);\r\n\r\n }\r\n transparent_crc(g_165, \"g_165\", print_hash_value);\r\n transparent_crc(g_168, \"g_168\", print_hash_value);\r\n transparent_crc(g_185, \"g_185\", print_hash_value);\r\n transparent_crc(g_306, \"g_306\", print_hash_value);\r\n transparent_crc(g_311, \"g_311\", print_hash_value);\r\n transparent_crc(g_313, \"g_313\", print_hash_value);\r\n transparent_crc(g_341, \"g_341\", print_hash_value);\r\n for (i = 0; i < 6; i++)\r\n {\r\n transparent_crc(g_346[i], \"g_346[i]\", print_hash_value);\r\n if (print_hash_value) printf(\"index = [%d]\\n\", i);\r\n\r\n }\r\n for (i = 0; i < 6; i++)\r\n {\r\n for (j = 0; j < 2; j++)\r\n {\r\n transparent_crc(g_366[i][j], \"g_366[i][j]\", print_hash_value);\r\n if (print_hash_value) printf(\"index = [%d][%d]\\n\", i, j);\r\n\r\n }\r\n }\r\n transparent_crc(g_380, \"g_380\", print_hash_value);\r\n transparent_crc(g_407, \"g_407\", print_hash_value);\r\n transparent_crc(g_409, \"g_409\", print_hash_value);\r\n transparent_crc(g_412, \"g_412\", print_hash_value);\r\n transparent_crc(g_414, \"g_414\", print_hash_value);\r\n for (i = 0; i < 5; i++)\r\n {\r\n for (j = 0; j < 10; j++)\r\n {\r\n transparent_crc(g_416[i][j], \"g_416[i][j]\", print_hash_value);\r\n if (print_hash_value) printf(\"index = [%d][%d]\\n\", i, j);\r\n\r\n }\r\n }\r\n transparent_crc(g_426, \"g_426\", print_hash_value);\r\n transparent_crc(g_457, \"g_457\", print_hash_value);\r\n platform_main_end(crc32_context ^ 0xFFFFFFFFUL, print_hash_value);\r\n return 0;\r\n}\r\n\r\n/************************ statistics *************************\r\nXXX max struct depth: 0\r\nbreakdown:\r\n depth: 0, occurrence: 138\r\nXXX total union variables: 0\r\n\r\nXXX max expression depth: 48\r\nbreakdown:\r\n depth: 1, occurrence: 114\r\n depth: 2, occurrence: 28\r\n depth: 3, occurrence: 1\r\n depth: 4, occurrence: 1\r\n depth: 5, occurrence: 1\r\n depth: 6, occurrence: 2\r\n depth: 16, occurrence: 1\r\n depth: 17, occurrence: 2\r\n depth: 18, occurrence: 2\r\n depth: 19, occurrence: 1\r\n depth: 24, occurrence: 1\r\n depth: 25, occurrence: 1\r\n depth: 28, occurrence: 1\r\n depth: 29, occurrence: 1\r\n depth: 30, occurrence: 1\r\n depth: 31, occurrence: 1\r\n depth: 32, occurrence: 1\r\n depth: 34, occurrence: 3\r\n depth: 38, occurrence: 1\r\n depth: 43, occurrence: 1\r\n depth: 48, occurrence: 1\r\n\r\nXXX total number of pointers: 101\r\n\r\nXXX times a variable address is taken: 515\r\nXXX times a pointer is dereferenced on RHS: 36\r\nbreakdown:\r\n depth: 1, occurrence: 32\r\n depth: 2, occurrence: 2\r\n depth: 3, occurrence: 2\r\nXXX times a pointer is dereferenced on LHS: 69\r\nbreakdown:\r\n depth: 1, occurrence: 59\r\n depth: 2, occurrence: 7\r\n depth: 3, occurrence: 3\r\nXXX times a pointer is compared with null: 5\r\nXXX times a pointer is compared with address of another variable: 5\r\nXXX times a pointer is compared with another pointer: 3\r\nXXX times a pointer is qualified to be dereferenced: 1932\r\n\r\nXXX max dereference level: 3\r\nbreakdown:\r\n level: 0, occurrence: 0\r\n level: 1, occurrence: 190\r\n level: 2, occurrence: 35\r\n level: 3, occurrence: 30\r\nXXX number of pointers point to pointers: 36\r\nXXX number of pointers point to scalars: 65\r\nXXX number of pointers point to structs: 0\r\nXXX percent of pointers has null in alias set: 38.6\r\nXXX average alias set size: 1.44\r\n\r\nXXX times a non-volatile is read: 427\r\nXXX times a non-volatile is write: 241\r\nXXX times a volatile is read: 0\r\nXXX times read thru a pointer: 0\r\nXXX times a volatile is write: 0\r\nXXX times written thru a pointer: 0\r\nXXX times a volatile is available for access: 0\r\nXXX percentage of non-volatile access: 100\r\n\r\nXXX forward jumps: 0\r\nXXX backward jumps: 2\r\n\r\nXXX stmts: 111\r\nXXX max block depth: 3\r\nbreakdown:\r\n depth: 0, occurrence: 13\r\n depth: 1, occurrence: 17\r\n depth: 2, occurrence: 28\r\n depth: 3, occurrence: 53\r\n\r\nXXX percentage a fresh-made variable is used: 16.5\r\nXXX percentage an existing variable is used: 83.5\r\nXXX total OOB instances added: 0\r\n********************* end of statistics **********************/\r\n\r\n```"}]}, "verified": true} {"bug_id": "106417", "issue_url": "https://github.com/llvm/llvm-project/issues/106417", "bug_type": "crash", "base_commit": "66927fb95abef9327b453d7213c5df7d641269be", "knowledge_cutoff": "2024-08-28T16:21:38Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "c4906588ce47de33d59bcd95f3e82ce2c3e61c23", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[7113, 7119], [7121, 7126]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationPlanner::precomputeCosts"]}}, "patch": "commit c4906588ce47de33d59bcd95f3e82ce2c3e61c23\nAuthor: Florian Hahn \nDate: Thu Aug 29 21:19:59 2024 +0100\n\n [VPlan] Use skipCostComputation when pre-computing induction costs.\n \n This ensures we skip any instructions identified to be ignored by the\n legacy cost model as well. Fixes a divergence between legacy and\n VPlan-based cost model.\n \n Fixes https://github.com/llvm/llvm-project/issues/106417.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 4cc75e2e7546..6babfd1eee91 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -7113,7 +7113,7 @@ LoopVectorizationPlanner::precomputeCosts(VPlan &Plan, ElementCount VF,\n IVInsts.push_back(CI);\n }\n for (Instruction *IVInst : IVInsts) {\n- if (!CostCtx.SkipCostComputation.insert(IVInst).second)\n+ if (CostCtx.skipCostComputation(IVInst, VF.isVector()))\n continue;\n InstructionCost InductionCost = CostCtx.getLegacyCost(IVInst, VF);\n LLVM_DEBUG({\n@@ -7121,6 +7121,7 @@ LoopVectorizationPlanner::precomputeCosts(VPlan &Plan, ElementCount VF,\n << \": induction instruction \" << *IVInst << \"\\n\";\n });\n Cost += InductionCost;\n+ CostCtx.SkipCostComputation.insert(IVInst);\n }\n }\n \n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/induction-costs.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "skip_free_iv_truncate", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @skip_free_iv_truncate(i16 %x, ptr %A) #0 {\nentry:\n %x.i32 = sext i16 %x to i32\n %x.i64 = sext i16 %x to i64\n %invariant.gep = getelementptr i8, ptr %A, i64 -8\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %x.i64, %entry ], [ %iv.next, %loop ]\n %iv.conv = phi i32 [ %x.i32, %entry ], [ %3, %loop ]\n %gep.i64 = getelementptr i64, ptr %A, i64 %iv\n %0 = load i64, ptr %gep.i64, align 8\n %1 = sext i32 %iv.conv to i64\n %gep.conv = getelementptr i64, ptr %invariant.gep, i64 %1\n %2 = load i64, ptr %gep.conv, align 8\n %gep.i16 = getelementptr i16, ptr %A, i64 %iv\n store i16 0, ptr %gep.i16, align 2\n %iv.next = add i64 %iv, 3\n %3 = trunc i64 %iv.next to i32\n %c = icmp slt i64 %iv, 99\n br i1 %c, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v,+zvl256b\" }\n"}]}], "issue": {"title": "[VPlan] Assertion \" VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Tested using 190449a5d2a9ddfc2180dce11714a4443f0f29f0\r\n\r\nTestcase:\r\n```c\r\nchar a;\r\nextern short b[][6][6];\r\nshort c;\r\nlong d;\r\nlong *e, *f;\r\nvoid g(char h[][6]) {\r\n for (short i = d + 9; i < 4; i += 3) {\r\n a = ({ f[i]; }) < f[i - 1];\r\n for (int j = c; j; j = ({ h[5][i + 2]; }) ? e[i - 1] : 0)\r\n ;\r\n b[i][i][i] = 0;\r\n }\r\n}\r\n```\r\nhttps://godbolt.org/z/EGdEPbKad\r\n\r\nIR testcase:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @g(i16 %0, i64 %indvars.iv52) #0 {\r\nentry:\r\n %conv146 = sext i16 %0 to i32\r\n %1 = sext i16 %0 to i64\r\n %invariant.gep = getelementptr i8, ptr null, i64 -8\r\n br label %for.body.us\r\n\r\nfor.body.us: ; preds = %for.body.us, %entry\r\n %indvars.iv521 = phi i64 [ %1, %entry ], [ %indvars.iv.next53, %for.body.us ]\r\n %conv149.us = phi i32 [ %conv146, %entry ], [ %5, %for.body.us ]\r\n %arrayidx.us = getelementptr i64, ptr null, i64 %indvars.iv521\r\n %2 = load i64, ptr %arrayidx.us, align 8\r\n %3 = sext i32 %conv149.us to i64\r\n %gep = getelementptr i64, ptr %invariant.gep, i64 %3\r\n %4 = load i64, ptr %gep, align 8\r\n %arrayidx31.us = getelementptr [0 x [6 x [6 x i16]]], ptr null, i64 0, i64 %indvars.iv521, i64 %indvars.iv52, i64 %indvars.iv52\r\n store i16 0, ptr %arrayidx31.us, align 2\r\n %indvars.iv.next53 = add i64 %indvars.iv521, 3\r\n %5 = trunc i64 %indvars.iv.next53 to i32\r\n %cmp.us = icmp slt i64 %indvars.iv521, 0\r\n br i1 %cmp.us, label %for.body.us, label %for.cond.for.cond.cleanup_crit_edge.split.us\r\n\r\nfor.cond.for.cond.cleanup_crit_edge.split.us: ; preds = %for.body.us\r\n ret void\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl256b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize -S reduced.ll\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7359: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), BestFactor.Width, CostCtx, OrigLoop, CM)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize -S reduced.ll\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"g\"\r\n #0 0x0000640a195798f0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x30178f0)\r\n #1 0x0000640a19576d0f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x3014d0f)\r\n #2 0x0000640a19576e65 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007aac19442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007aac194969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007aac194969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007aac194969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007aac19442476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007aac194287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007aac1942871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007aac19439e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x0000640a18546ac9 llvm::LoopVectorizationPlanner::computeBestVF() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fe4ac9)\r\n#12 0x0000640a1855cedb llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffaedb)\r\n#13 0x0000640a1855fef1 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffdef1)\r\n#14 0x0000640a18560567 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffe567)\r\n#15 0x0000640a173c0a96 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5ea96)\r\n#16 0x0000640a19384d4f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e22d4f)\r\n#17 0x0000640a173c3c26 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe61c26)\r\n#18 0x0000640a19383acb llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e21acb)\r\n#19 0x0000640a173bfec6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5dec6)\r\n#20 0x0000640a19381b0d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1fb0d)\r\n#21 0x0000640a16be9da6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x687da6)\r\n#22 0x0000640a16bdbb31 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x679b31)\r\n#23 0x00007aac19429d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#24 0x00007aac19429e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#25 0x00007aac19429e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#26 0x0000640a16bd1ee5 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x66fee5)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\nhttps://godbolt.org/z/W3KqbcKrb\r\n\r\nFound via fuzzer\r\nRelated/prior issue: #104714\r\n\r\ncc @fhahn ", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "121459", "issue_url": "https://github.com/llvm/llvm-project/issues/121459", "bug_type": "miscompilation", "base_commit": "68d265666e708bad1c63b419b6275aaba1a7dcd2", "knowledge_cutoff": "2025-01-02T09:03:32Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "a4d92400a6db9566d84cb4b900149e36e117f452", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstructionCombining.cpp": [[2782, 2787], [2838, 2843], [2847, 2852]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstructionCombining.cpp": ["foldGEPOfPhi"]}}, "patch": "commit a4d92400a6db9566d84cb4b900149e36e117f452\nAuthor: Yingwei Zheng \nDate: Fri Jan 3 23:19:57 2025 +0800\n\n [InstCombine] Fix GEPNoWrapFlags propagation in `foldGEPOfPhi` (#121572)\n \n Closes https://github.com/llvm/llvm-project/issues/121459.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\nindex 934156f04f7f..f63de1f0d410 100644\n--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp\n@@ -2782,6 +2782,7 @@ static Instruction *foldGEPOfPhi(GetElementPtrInst &GEP, PHINode *PN,\n // loop iteration).\n if (Op1 == &GEP)\n return nullptr;\n+ GEPNoWrapFlags NW = Op1->getNoWrapFlags();\n \n int DI = -1;\n \n@@ -2838,6 +2839,8 @@ static Instruction *foldGEPOfPhi(GetElementPtrInst &GEP, PHINode *PN,\n }\n }\n }\n+\n+ NW &= Op2->getNoWrapFlags();\n }\n \n // If not all GEPs are identical we'll have to create a new PHI node.\n@@ -2847,6 +2850,8 @@ static Instruction *foldGEPOfPhi(GetElementPtrInst &GEP, PHINode *PN,\n return nullptr;\n \n auto *NewGEP = cast(Op1->clone());\n+ NewGEP->setNoWrapFlags(NW);\n+\n if (DI == -1) {\n // All the GEPs feeding the PHI are identical. Clone one down into our\n // BB so that it can be merged with the current GEP.\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/opaque-ptr.ll", "commands": ["opt -S -passes='instcombine' < %s"], "tests": [{"test_name": "gep_of_phi_of_gep_different_type", "test_body": "define ptr @gep_of_phi_of_gep_different_type(i1 %c, ptr %p) {\n br i1 %c, label %if, label %else\n\nif: ; preds = %0\n %gep1 = getelementptr i32, ptr %p, i64 1\n br label %join\n\nelse: ; preds = %0\n %gep2 = getelementptr i64, ptr %p, i64 2\n br label %join\n\njoin: ; preds = %else, %if\n %phi = phi ptr [ %gep1, %if ], [ %gep2, %else ]\n %gep = getelementptr i32, ptr %phi, i64 1\n ret ptr %gep\n}\n"}, {"test_name": "gep_of_phi_of_gep_flags2", "test_body": "define ptr @gep_of_phi_of_gep_flags2(i1 %c, ptr %p) {\n br i1 %c, label %if, label %else\n\nif: ; preds = %0\n %gep1 = getelementptr nuw i32, ptr %p, i64 1\n br label %join\n\nelse: ; preds = %0\n %gep2 = getelementptr nuw i32, ptr %p, i64 2\n br label %join\n\njoin: ; preds = %else, %if\n %phi = phi ptr [ %gep1, %if ], [ %gep2, %else ]\n %gep = getelementptr i32, ptr %phi, i64 1\n ret ptr %gep\n}\n"}, {"test_name": "gep_of_phi_of_gep_flags1", "test_body": "define ptr @gep_of_phi_of_gep_flags1(i1 %c, ptr %p) {\n br i1 %c, label %if, label %else\n\nif: ; preds = %0\n %gep1 = getelementptr inbounds i32, ptr %p, i64 1\n br label %join\n\nelse: ; preds = %0\n %gep2 = getelementptr i32, ptr %p, i64 2\n br label %join\n\njoin: ; preds = %else, %if\n %phi = phi ptr [ %gep1, %if ], [ %gep2, %else ]\n %gep = getelementptr i32, ptr %phi, i64 1\n ret ptr %gep\n}\n"}]}], "issue": {"title": "[InstCombine] GEPNoWrapFlags is propagated incorrectly", "body": "Alive2: https://alive2.llvm.org/ce/z/3ME7S7\n```\ndefine ptr @src(i1 %c, ptr %p) {\n br i1 %c, label %if, label %else\n\nif:\n %gep1 = getelementptr inbounds i32, ptr %p, i64 1\n br label %join\n\nelse:\n %gep2 = getelementptr i64, ptr %p, i64 2\n br label %join\n\njoin:\n %phi = phi ptr [ %gep1, %if ], [ %gep2, %else ]\n %gep = getelementptr i32, ptr %phi, i64 1\n ret ptr %gep\n}\n```\n```\n\n----------------------------------------\ndefine ptr @src(i1 %c, ptr %p) {\n#0:\n br i1 %c, label %if, label %else\n\nelse:\n %gep2 = gep ptr %p, 8 x i64 2\n br label %join\n\nif:\n %gep1 = gep inbounds ptr %p, 4 x i64 1\n br label %join\n\njoin:\n %phi = phi ptr [ %gep1, %if ], [ %gep2, %else ]\n %gep = gep ptr %phi, 4 x i64 1\n ret ptr %gep\n}\n=>\ndefine ptr @src(i1 %c, ptr %p) {\n#0:\n br i1 %c, label %if, label %else\n\nelse:\n br label %join\n\nif:\n br label %join\n\njoin:\n %#1 = phi i64 [ 4, %if ], [ 16, %else ]\n %#2 = gep inbounds nuw ptr %p, 1 x i64 %#1\n %gep = gep ptr %#2, 1 x i64 4\n ret ptr %gep\n}\nTransformation doesn't verify!\n\nERROR: Target is more poisonous than source\n\nExample:\ni1 %c = #x0 (0)\nptr %p = null\n\nSource:\n >> Jump to %else\nptr %gep2 = pointer(non-local, block_id=0, offset=16) / Address=#x0000000000000010\n >> Jump to %join\nptr %phi = pointer(non-local, block_id=0, offset=16) / Address=#x0000000000000010\nptr %gep = pointer(non-local, block_id=0, offset=20) / Address=#x0000000000000014\n\nSOURCE MEMORY STATE\n===================\nNON-LOCAL BLOCKS:\nBlock 0 >\tsize: 0\talign: 1\talloc type: 0\talive: false\taddress: 0\nBlock 1 >\tsize: 0\talign: 1\talloc type: 0\talive: true\taddress: 1\n\nTarget:\n >> Jump to %else\n >> Jump to %join\ni64 %#1 = #x0000000000000010 (16)\nptr %#2 = poison\nptr %gep = poison\nSource value: pointer(non-local, block_id=0, offset=20) / Address=#x0000000000000014\nTarget value: poison\n\nSummary:\n 0 correct transformations\n 1 incorrect transformations\n 0 failed-to-prove transformations\n 0 Alive2 errors\n```\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "109468", "issue_url": "https://github.com/llvm/llvm-project/issues/109468", "bug_type": "crash", "base_commit": "51039101cf32591782ef564a108db71150a3b7c3", "knowledge_cutoff": "2024-09-20T20:22:11Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "a068b974b199b0e7350da2d9506adc7df3995ce3", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlan.h": [[2709, 2714], [2787, 2792]], "llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[2267, 2272], [2363, 2368]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPWidenLoadEVLRecipe::execute", "VPWidenLoadEVLRecipe::print", "VPWidenStoreEVLRecipe::execute", "VPWidenStoreEVLRecipe::print"]}}, "patch": "commit a068b974b199b0e7350da2d9506adc7df3995ce3\nAuthor: Elvis Wang \nDate: Thu Sep 26 07:10:25 2024 +0800\n\n [VPlan] Implement VPWidenLoad/StoreEVLRecipe::computeCost(). (#109644)\n \n Currently the EVL recipes transfer the tail masking to the EVL.\n But in the legacy cost model, the mask exist and will calculate the\n instruction cost of the mask.\n To fix the difference between the VPlan-based cost model and the legacy\n cost model, we always calculate the instruction cost for the mask in the\n EVL recipes.\n \n Note that we should remove the mask cost in the EVL recipes when we\n don't need to compare to the legacy cost model.\n \n This patch also fixes #109468.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h\nindex bbcfaf9e19cd..23a24ce293ef 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlan.h\n+++ b/llvm/lib/Transforms/Vectorize/VPlan.h\n@@ -2709,6 +2709,10 @@ struct VPWidenLoadEVLRecipe final : public VPWidenMemoryRecipe, public VPValue {\n /// Generate the wide load or gather.\n void execute(VPTransformState &State) override;\n \n+ /// Return the cost of this VPWidenLoadEVLRecipe.\n+ InstructionCost computeCost(ElementCount VF,\n+ VPCostContext &Ctx) const override;\n+\n #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)\n /// Print the recipe.\n void print(raw_ostream &O, const Twine &Indent,\n@@ -2787,6 +2791,10 @@ struct VPWidenStoreEVLRecipe final : public VPWidenMemoryRecipe {\n /// Generate the wide store or scatter.\n void execute(VPTransformState &State) override;\n \n+ /// Return the cost of this VPWidenStoreEVLRecipe.\n+ InstructionCost computeCost(ElementCount VF,\n+ VPCostContext &Ctx) const override;\n+\n #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)\n /// Print the recipe.\n void print(raw_ostream &O, const Twine &Indent,\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex dacba152611c..9a0aefb898e5 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -2267,6 +2267,31 @@ void VPWidenLoadEVLRecipe::execute(VPTransformState &State) {\n State.set(this, Res);\n }\n \n+InstructionCost VPWidenLoadEVLRecipe::computeCost(ElementCount VF,\n+ VPCostContext &Ctx) const {\n+ if (!Consecutive || IsMasked)\n+ return VPWidenMemoryRecipe::computeCost(VF, Ctx);\n+\n+ // We need to use the getMaskedMemoryOpCost() instead of getMemoryOpCost()\n+ // here because the EVL recipes using EVL to replace the tail mask. But in the\n+ // legacy model, it will always calculate the cost of mask.\n+ // TODO: Using getMemoryOpCost() instead of getMaskedMemoryOpCost when we\n+ // don't need to compare to the legacy cost model.\n+ Type *Ty = ToVectorTy(getLoadStoreType(&Ingredient), VF);\n+ const Align Alignment =\n+ getLoadStoreAlignment(const_cast(&Ingredient));\n+ unsigned AS =\n+ getLoadStoreAddressSpace(const_cast(&Ingredient));\n+ TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n+ InstructionCost Cost = Ctx.TTI.getMaskedMemoryOpCost(\n+ Ingredient.getOpcode(), Ty, Alignment, AS, CostKind);\n+ if (!Reverse)\n+ return Cost;\n+\n+ return Cost + Ctx.TTI.getShuffleCost(TargetTransformInfo::SK_Reverse,\n+ cast(Ty), {}, CostKind, 0);\n+}\n+\n #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)\n void VPWidenLoadEVLRecipe::print(raw_ostream &O, const Twine &Indent,\n VPSlotTracker &SlotTracker) const {\n@@ -2363,6 +2388,31 @@ void VPWidenStoreEVLRecipe::execute(VPTransformState &State) {\n State.addMetadata(NewSI, SI);\n }\n \n+InstructionCost VPWidenStoreEVLRecipe::computeCost(ElementCount VF,\n+ VPCostContext &Ctx) const {\n+ if (!Consecutive || IsMasked)\n+ return VPWidenMemoryRecipe::computeCost(VF, Ctx);\n+\n+ // We need to use the getMaskedMemoryOpCost() instead of getMemoryOpCost()\n+ // here because the EVL recipes using EVL to replace the tail mask. But in the\n+ // legacy model, it will always calculate the cost of mask.\n+ // TODO: Using getMemoryOpCost() instead of getMaskedMemoryOpCost when we\n+ // don't need to compare to the legacy cost model.\n+ Type *Ty = ToVectorTy(getLoadStoreType(&Ingredient), VF);\n+ const Align Alignment =\n+ getLoadStoreAlignment(const_cast(&Ingredient));\n+ unsigned AS =\n+ getLoadStoreAddressSpace(const_cast(&Ingredient));\n+ TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;\n+ InstructionCost Cost = Ctx.TTI.getMaskedMemoryOpCost(\n+ Ingredient.getOpcode(), Ty, Alignment, AS, CostKind);\n+ if (!Reverse)\n+ return Cost;\n+\n+ return Cost + Ctx.TTI.getShuffleCost(TargetTransformInfo::SK_Reverse,\n+ cast(Ty), {}, CostKind, 0);\n+}\n+\n #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)\n void VPWidenStoreEVLRecipe::print(raw_ostream &O, const Twine &Indent,\n VPSlotTracker &SlotTracker) const {\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-uniform-store.ll", "commands": ["opt < %s --prefer-predicate-over-epilogue=predicate-dont-vectorize --passes=loop-vectorize -mcpu=sifive-p470 -mattr=+v,+f -force-tail-folding-style=data-with-evl -S"], "tests": [{"test_name": "lshift_significand", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @lshift_significand(i32 %n, ptr nocapture writeonly %dst) {\nentry:\n %cmp1.peel = icmp eq i32 %n, 0\n %spec.select = select i1 %cmp1.peel, i64 2, i64 0\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %spec.select, %entry ], [ %iv.next, %loop ]\n %0 = sub nuw nsw i64 1, %iv\n %arrayidx13 = getelementptr i64, ptr %dst, i64 %0\n store i64 0, ptr %arrayidx13, align 8\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 3\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "[LV][VPlan] Crash due to disagreements on the VPlan cost v.s. legacy cost model", "body": "Given this input:\r\n``` llvm\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @lshift_significand(i32 %n, ptr %0) {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %for.cond, %entry\r\n %i.0 = phi i32 [ 0, %entry ], [ 2, %for.cond ]\r\n %add = or i32 %n, %i.0\r\n %cmp1 = icmp eq i32 %add, 0\r\n br i1 %cmp1, label %for.cond, label %for.cond7\r\n\r\nfor.cond7: ; preds = %for.body9, %for.cond\r\n %i.1 = phi i32 [ %i.0, %for.cond ], [ %inc15, %for.body9 ]\r\n %cmp8 = icmp ult i32 %i.1, 3\r\n br i1 %cmp8, label %for.body9, label %for.end16\r\n\r\nfor.body9: ; preds = %for.cond7\r\n %sub11 = sub nuw i32 1, %i.1\r\n %idxprom12 = zext i32 %sub11 to i64\r\n %arrayidx13 = getelementptr [3 x i64], ptr %0, i64 0, i64 %idxprom12\r\n store i64 0, ptr %arrayidx13, align 8\r\n %inc15 = add i32 %i.1, 1\r\n br label %for.cond7\r\n\r\nfor.end16: ; preds = %for.cond7\r\n ret void\r\n}\r\n```\r\nAnd this command:\r\n``` bash\r\nopt -mcpu=sifive-p470 -O3 -prefer-predicate-over-epilogue=predicate-dont-vectorize -force-tail-folding-style=data-with-evl input.ll -disable-output\r\n```\r\nWe'll get the following assertion:\r\n```\r\nllvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7386: VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\n```\r\n\r\n\r\nPreliminary investigation shows that this is caused by the disagreement on (reverse) widen store's cost. First, this is the trace for VPlan's cost model:\r\n```\r\nLV: Scalar loop costs: 4.\r\nCost of 1 for VF vscale x 1: induction instruction %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\r\nCost of 0 for VF vscale x 1: induction instruction %indvars.iv = phi i64 [ %spec.select, %entry ], [ %indvars.iv.next, %for.body9 ]\r\nCost of 1 for VF vscale x 1: exit condition instruction %exitcond.not = icmp eq i64 %indvars.iv.next, 3\r\nCost of 0 for VF vscale x 1: EMIT vp<%4> = CANONICAL-INDUCTION ir<0>, vp<%12>\r\nCost of 0 for VF vscale x 1: EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<%5> = phi ir<0>, vp<%11>\r\nCost of 0 for VF vscale x 1: EMIT vp<%6> = EXPLICIT-VECTOR-LENGTH vp<%5>, vp<%3>\r\nCost of 0 for VF vscale x 1: vp<%7> = DERIVED-IV ir<%spec.select> + vp<%5> * ir<1>\r\nCost of 0 for VF vscale x 1: vp<%8> = SCALAR-STEPS vp<%7>, ir<1>\r\nCost of 1 for VF vscale x 1: CLONE ir<%1> = sub nuw nsw ir<1>, vp<%8>\r\nCost of 0 for VF vscale x 1: CLONE ir<%arrayidx13> = getelementptr ir<%0>, ir<0>, ir<%1>\r\nCost of 0 for VF vscale x 1: vp<%9> = vector-pointer (reverse) ir<%arrayidx13>\r\nCost of 8 for VF vscale x 1: WIDEN vp.store vp<%9>, ir<0>, vp<%6>\r\nCost of 0 for VF vscale x 1: SCALAR-CAST vp<%10> = zext vp<%6> to i64\r\nCost of 0 for VF vscale x 1: EMIT vp<%11> = add vp<%10>, vp<%5>\r\nCost of 0 for VF vscale x 1: EMIT vp<%12> = add vp<%4>, vp<%0>\r\nCost of 0 for VF vscale x 1: EMIT branch-on-count vp<%12>, vp<%1>\r\nCost of 0 for VF vscale x 1: vector loop backedge\r\nCost for VF vscale x 1: 11\r\nCost of 1 for VF vscale x 2: induction instruction %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\r\nCost of 0 for VF vscale x 2: induction instruction %indvars.iv = phi i64 [ %spec.select, %entry ], [ %indvars.iv.next, %for.body9 ]\r\nCost of 1 for VF vscale x 2: exit condition instruction %exitcond.not = icmp eq i64 %indvars.iv.next, 3\r\nCost of 0 for VF vscale x 2: EMIT vp<%4> = CANONICAL-INDUCTION ir<0>, vp<%12>\r\nCost of 0 for VF vscale x 2: EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<%5> = phi ir<0>, vp<%11>\r\nCost of 0 for VF vscale x 2: EMIT vp<%6> = EXPLICIT-VECTOR-LENGTH vp<%5>, vp<%3>\r\nCost of 0 for VF vscale x 2: vp<%7> = DERIVED-IV ir<%spec.select> + vp<%5> * ir<1>\r\nCost of 0 for VF vscale x 2: vp<%8> = SCALAR-STEPS vp<%7>, ir<1>\r\nCost of 1 for VF vscale x 2: CLONE ir<%1> = sub nuw nsw ir<1>, vp<%8>\r\nCost of 0 for VF vscale x 2: CLONE ir<%arrayidx13> = getelementptr ir<%0>, ir<0>, ir<%1>\r\nCost of 0 for VF vscale x 2: vp<%9> = vector-pointer (reverse) ir<%arrayidx13>\r\nCost of 14 for VF vscale x 2: WIDEN vp.store vp<%9>, ir<0>, vp<%6>\r\nCost of 0 for VF vscale x 2: SCALAR-CAST vp<%10> = zext vp<%6> to i64\r\nCost of 0 for VF vscale x 2: EMIT vp<%11> = add vp<%10>, vp<%5>\r\nCost of 0 for VF vscale x 2: EMIT vp<%12> = add vp<%4>, vp<%0>\r\nCost of 0 for VF vscale x 2: EMIT branch-on-count vp<%12>, vp<%1>\r\nCost of 0 for VF vscale x 2: vector loop backedge\r\nCost for VF vscale x 2: 17\r\n```\r\nWith VPlan's cost model, we will eventually choose scalar loop, because when VF=vscale x 1, the final cost is `ceil(11 / 2) = 6`; when VF=vscale x 2, the final cost is `ceil(17 / 4) = 5`. Both of them are larger than the scalar cost, 4.\r\n\r\nWhile with the legacy cost model:\r\n```\r\nLV: Found an estimated cost of 0 for VF 1 For instruction: %indvars.iv = phi i64 [ %spec.select, %entry ], [ %indvars.iv.next, %for.body9 ]\r\nLV: Found an estimated cost of 1 for VF 1 For instruction: %1 = sub nuw nsw i64 1, %indvars.iv\r\nLV: Found an estimated cost of 0 for VF 1 For instruction: %arrayidx13 = getelementptr [3 x i64], ptr %0, i64 0, i64 %1\r\nLV: Found an estimated cost of 1 for VF 1 For instruction: store i64 0, ptr %arrayidx13, align 8\r\nLV: Found an estimated cost of 1 for VF 1 For instruction: %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\r\nLV: Found an estimated cost of 1 for VF 1 For instruction: %exitcond.not = icmp eq i64 %indvars.iv.next, 3\r\nLV: Found an estimated cost of 0 for VF 1 For instruction: br i1 %exitcond.not, label %for.end16, label %for.body9\r\nLV: Scalar loop costs: 4.\r\nLV: Found an estimated cost of 0 for VF vscale x 1 For instruction: %indvars.iv = phi i64 [ %spec.select, %entry ], [ %indvars.iv.next, %for.body9 ]\r\nLV: Found an estimated cost of 1 for VF vscale x 1 For instruction: %1 = sub nuw nsw i64 1, %indvars.iv\r\nLV: Found an estimated cost of 0 for VF vscale x 1 For instruction: %arrayidx13 = getelementptr [3 x i64], ptr %0, i64 0, i64 %1\r\nLV: Found an estimated cost of 7 for VF vscale x 1 For instruction: store i64 0, ptr %arrayidx13, align 8\r\nLV: Found an estimated cost of 1 for VF vscale x 1 For instruction: %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\r\nLV: Found an estimated cost of 1 for VF vscale x 1 For instruction: %exitcond.not = icmp eq i64 %indvars.iv.next, 3\r\nLV: Found an estimated cost of 0 for VF vscale x 1 For instruction: br i1 %exitcond.not, label %for.end16, label %for.body9\r\nLV: Vector loop of width vscale x 1 costs: 5 (assuming a minimum vscale of 2).\r\nLV: Found an estimated cost of 0 for VF vscale x 2 For instruction: %indvars.iv = phi i64 [ %spec.select, %entry ], [ %indvars.iv.next, %for.body9 ]\r\nLV: Found an estimated cost of 1 for VF vscale x 2 For instruction: %1 = sub nuw nsw i64 1, %indvars.iv\r\nLV: Found an estimated cost of 0 for VF vscale x 2 For instruction: %arrayidx13 = getelementptr [3 x i64], ptr %0, i64 0, i64 %1\r\nLV: Found an estimated cost of 13 for VF vscale x 2 For instruction: store i64 0, ptr %arrayidx13, align 8\r\nLV: Found an estimated cost of 1 for VF vscale x 2 For instruction: %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\r\nLV: Found an estimated cost of 1 for VF vscale x 2 For instruction: %exitcond.not = icmp eq i64 %indvars.iv.next, 3\r\nLV: Found an estimated cost of 0 for VF vscale x 2 For instruction: br i1 %exitcond.not, label %for.end16, label %for.body9\r\nLV: Vector loop of width vscale x 2 costs: 4 (assuming a minimum vscale of 2).\r\nLV: Selecting VF: vscale x 2.\r\n```\r\nWe will eventually choose VF=vscale x 2.\r\n\r\nThe key difference is the cost of `store i64 0, ptr %arrayidx13, align 8` v.s. cost of `WIDEN vp.store vp<%9>, ir<0>, vp<%6>`, where the latter is larger than the former by 1.\r\n\r\nIn both cases (the store instruction in VPlan cost model v.s. legacy cost model) their costs are computed by the base cost of store + the cost of shuffle due to being reverse store. I used debugger to confirm that they used the same shuffle cost, which means that for some reason, VPlan's cost model yield a slightly higher cost for store than that in the legacy cost model.\r\n", "author": "mshockwave", "labels": ["vectorizers", "crash"], "comments": [{"author": "ElvisWang123", "body": "It seems that the legacy cost model using `getMaskedMemoryOpCost` to query the instruction cost. But in the VPlan-based cost model using `getMemoryOpCost`. \r\n\r\nIn this case, the mask in the `VPWidenStoreEVLRecipe` is false since it is using the EVL recipe.\r\nThe legacy cost model will not check the tail folding type so it will using `getMaskedMemoryOpCost` instead of `getMemoryOpCost` . And the `MaskedMemoryCost` is slightly smaller than non-mask version."}, {"author": "ElvisWang123", "body": "Open a PR to fix this issue. #109644 \r\n\r\nWe remove the shuffle cost when the value stored by reversed vector is loop invariant."}, {"author": "mshockwave", "body": "CC @fhahn "}]}, "verified": true} {"bug_id": "112076", "issue_url": "https://github.com/llvm/llvm-project/issues/112076", "bug_type": "miscompilation", "base_commit": "8c56dd3040f295874e3d5742b5dfc109adf1f728", "knowledge_cutoff": "2024-10-12T03:31:00Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "2c094ac761912eea0d7e8ccb140bc647b5378bdf", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[3466, 3472], [3501, 3506], [4148, 4154]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["InstCombinerImpl::visitSelectInst", "foldBitCeil"]}}, "patch": "commit 2c094ac761912eea0d7e8ccb140bc647b5378bdf\nAuthor: Yingwei Zheng \nDate: Wed Nov 20 21:15:26 2024 +0800\n\n [InstCombine] Drop range attributes in `foldBitCeil` (#116641)\n \n Closes https://github.com/llvm/llvm-project/issues/112076\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex 28dbca027037..2526ce7704ab 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -3466,7 +3466,8 @@ static bool isSafeToRemoveBitCeilSelect(ICmpInst::Predicate Pred, Value *Cond0,\n // Note that the select is optimized away while the shift count is masked with\n // 31. We handle some variations of the input operand like std::bit_ceil(X +\n // 1).\n-static Instruction *foldBitCeil(SelectInst &SI, IRBuilderBase &Builder) {\n+static Instruction *foldBitCeil(SelectInst &SI, IRBuilderBase &Builder,\n+ InstCombinerImpl &IC) {\n Type *SelType = SI.getType();\n unsigned BitWidth = SelType->getScalarSizeInBits();\n \n@@ -3501,6 +3502,10 @@ static Instruction *foldBitCeil(SelectInst &SI, IRBuilderBase &Builder) {\n // single hardware instruction as opposed to BitWidth - CTLZ, where BitWidth\n // is an integer constant. Masking with BitWidth-1 comes free on some\n // hardware as part of the shift instruction.\n+\n+ // Drop range attributes and re-infer them in the next iteration.\n+ cast(Ctlz)->dropPoisonGeneratingAnnotations();\n+ IC.addToWorklist(cast(Ctlz));\n Value *Neg = Builder.CreateNeg(Ctlz);\n Value *Masked =\n Builder.CreateAnd(Neg, ConstantInt::get(SelType, BitWidth - 1));\n@@ -4148,7 +4153,7 @@ Instruction *InstCombinerImpl::visitSelectInst(SelectInst &SI) {\n if (sinkNotIntoOtherHandOfLogicalOp(SI))\n return &SI;\n \n- if (Instruction *I = foldBitCeil(SI, Builder))\n+ if (Instruction *I = foldBitCeil(SI, Builder, *this))\n return I;\n \n if (Instruction *I = foldSelectToCmp(SI))\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/bit_ceil.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "test_drop_range_attr", "test_body": "define i32 @test_drop_range_attr(i32 %x) {\n %ctlz = call range(i32 1, 33) i32 @llvm.ctlz.i32(i32 %x, i1 false)\n %sub = sub i32 32, %ctlz\n %shl = shl i32 1, %sub\n %dec = add i32 %x, -1\n %ult = icmp ult i32 %dec, -2\n %sel = select i1 %ult, i32 %shl, i32 1\n ret i32 %sel\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctlz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "pr91691_keep_nsw", "test_body": "define i32 @pr91691_keep_nsw(i32 %0) {\n %2 = sub nsw i32 -2, %0\n %3 = tail call i32 @llvm.ctlz.i32(i32 %2, i1 false)\n %4 = sub i32 32, %3\n %5 = shl i32 1, %4\n %6 = icmp ult i32 %0, -2\n %7 = select i1 %6, i32 %5, i32 1\n ret i32 %7\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctlz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[InstCombine] `range` attribute should be dropped in `foldBitCeil`", "body": "Reproducer: https://alive2.llvm.org/ce/z/NQTmkh (Please use latest alive2 with https://github.com/AliveToolkit/alive2/pull/1097)\r\n```\r\ndefine i32 @src(i32 %x) {\r\n %ctlz = call range(i32 1, 33) i32 @llvm.ctlz.i32(i32 %x, i1 false)\r\n %sub = sub i32 32, %ctlz\r\n %shl = shl i32 1, %sub\r\n %dec = add i32 %x, -1\r\n %ult = icmp ult i32 %dec, -2\r\n %sel = select i1 %ult, i32 %shl, i32 1\r\n ret i32 %sel\r\n}\r\n\r\ndefine i32 @tgt(i32 %x) {\r\n %ctlz = call range(i32 1, 33) i32 @llvm.ctlz.i32(i32 %x, i1 false)\r\n %1 = sub nsw i32 0, %ctlz\r\n %2 = and i32 %1, 31\r\n %sel = shl nuw i32 1, %2\r\n ret i32 %sel\r\n}\r\n```\r\n```\r\n----------------------------------------\r\ndefine i32 @src(i32 %x) {\r\n#0:\r\n %ctlz = ctlz i32 %x, 0\r\n %#range_0_%ctlz = !range i32 %ctlz, i32 1, i32 33\r\n %sub = sub i32 32, %#range_0_%ctlz\r\n %shl = shl i32 1, %sub\r\n %dec = add i32 %x, 4294967295\r\n %ult = icmp ult i32 %dec, 4294967294\r\n %sel = select i1 %ult, i32 %shl, i32 1\r\n ret i32 %sel\r\n}\r\n=>\r\ndefine i32 @tgt(i32 %x) {\r\n#0:\r\n %ctlz = ctlz i32 %x, 0\r\n %#range_0_%ctlz = !range i32 %ctlz, i32 1, i32 33\r\n %#1 = sub nsw i32 0, %#range_0_%ctlz\r\n %#2 = and i32 %#1, 31\r\n %sel = shl nuw i32 1, %#2\r\n ret i32 %sel\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\ni32 %x = #xffffffff (4294967295, -1)\r\n\r\nSource:\r\ni32 %ctlz = #x00000000 (0)\r\ni32 %#range_0_%ctlz = poison\r\ni32 %sub = poison\r\ni32 %shl = poison\r\ni32 %dec = #xfffffffe (4294967294, -2)\r\ni1 %ult = #x0 (0)\r\ni32 %sel = #x00000001 (1)\r\n\r\nTarget:\r\ni32 %ctlz = #x00000000 (0)\r\ni32 %#range_0_%ctlz = poison\r\ni32 %#1 = poison\r\ni32 %#2 = poison\r\ni32 %sel = poison\r\nSource value: #x00000001 (1)\r\nTarget value: poison\r\n```\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "89516", "issue_url": "https://github.com/llvm/llvm-project/issues/89516", "bug_type": "miscompilation", "base_commit": "18bb175428f520aaa4a5e388bd3b680a1a7c60c0", "knowledge_cutoff": "2024-04-20T22:58:16Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "6309440c218778db027306826993e484eab2be17", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[2720, 2726]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["foldSelectWithSRem"]}}, "patch": "commit 6309440c218778db027306826993e484eab2be17\nAuthor: Yingwei Zheng \nDate: Sun Apr 21 22:41:32 2024 +0800\n\n [InstCombine] Fix unexpected overwriting in `foldSelectWithSRem` (#89539)\n \n Fixes #89516\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex 0262af28068b..73600206a55c 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -2720,7 +2720,7 @@ static Instruction *foldSelectWithSRem(SelectInst &SI, InstCombinerImpl &IC,\n // %cnd = icmp slt i32 %rem, 0\n // %add = add i32 %rem, %n\n // %sel = select i1 %cnd, i32 %add, i32 %rem\n- if (match(TrueVal, m_Add(m_Value(RemRes), m_Value(Remainder))) &&\n+ if (match(TrueVal, m_Add(m_Specific(RemRes), m_Value(Remainder))) &&\n match(RemRes, m_SRem(m_Value(Op), m_Specific(Remainder))) &&\n IC.isKnownToBeAPowerOfTwo(Remainder, /*OrZero*/ true) &&\n FalseVal == RemRes)\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/select-divrem.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "rem_euclid_pow2_false_arm_folded", "test_body": "define i32 @rem_euclid_pow2_false_arm_folded(i32 %n) {\n %rem = srem i32 %n, 2\n %nonneg = icmp sge i32 %rem, 0\n %res = select i1 %nonneg, i32 %rem, i32 1\n ret i32 %res\n}\n"}, {"test_name": "pr89516", "test_body": "define i8 @pr89516(i8 %n, i8 %x) {\n %cond = icmp slt i8 %x, 0\n %pow2 = shl nuw i8 1, %n\n %srem = srem i8 1, %pow2\n %add = add nuw i8 %srem, %pow2\n %res = select i1 %cond, i8 %add, i8 %srem\n ret i8 %res\n}\n"}]}], "issue": {"title": "miscompile of vanilla integer code by InstCombine", "body": "https://alive2.llvm.org/ce/z/jRfVLr\r\n\r\nthis function:\r\n```llvm\r\ndefine i8 @f(i8 %0, i8 %1) {\r\n %3 = icmp slt i8 %1, 0\r\n %4 = shl i8 1, %0\r\n %5 = srem i8 1, %4\r\n %6 = add i8 %5, %4\r\n %7 = select i1 %3, i8 %6, i8 %5\r\n ret i8 %7\r\n}\r\n```\r\nis getting mis-optimized to:\r\n```llvm\r\ndefine i8 @f(i8 %0, i8 %1) {\r\n %3 = icmp ne i8 %0, 0\r\n %4 = zext i1 %3 to i8\r\n ret i8 %4\r\n}\r\n```\r\nhere I'll just give Alive's work:\r\n```\r\nExample:\r\ni8 %#0 = #x00 (0)\r\ni8 %#1 = #xff (255, -1)\r\n\r\nSource:\r\ni1 %#3 = #x1 (1)\r\ni8 %#4 = #x01 (1)\r\ni8 %#5 = #x00 (0)\r\ni8 %#6 = #x01 (1)\r\ni8 %#7 = #x01 (1)\r\n\r\nTarget:\r\ni1 %#3 = #x0 (0)\r\ni8 %#4 = #x00 (0)\r\nSource value: #x01 (1)\r\nTarget value: #x00 (0)\r\n```\r\n\r\ncc @nunoplopes @hatsunespica", "author": "regehr", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "102279", "issue_url": "https://github.com/llvm/llvm-project/issues/102279", "bug_type": "crash", "base_commit": "f7e1efe8bcea14686d1c303e058f687b6ca1c79f", "knowledge_cutoff": "2024-08-07T07:28:15Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "441f94f4bdf6e2b9747ec12194c595098d9c3a5b", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[17336, 17362]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["tryToReduce"]}}, "patch": "commit 441f94f4bdf6e2b9747ec12194c595098d9c3a5b\nAuthor: Alexey Bataev \nDate: Wed Aug 7 04:09:55 2024 -0700\n\n [SLP]Fix PR102279: check the tracked values for extractelements, not the original values\n \n If the reduced value was replaced by the extractelement instruction\n during vectorization and we attempt to check if this is so, need to\n check the tracked value, not the original (deleted) instruction.\n Otherwise, the compiler may crash\n \n Fixes https://github.com/llvm/llvm-project/issues/102279\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex c9335b7a89e1..1fee06d145b9 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -17336,27 +17336,23 @@ public:\n // Try to handle shuffled extractelements.\n if (S.getOpcode() == Instruction::ExtractElement && !S.isAltShuffle() &&\n I + 1 < E) {\n- InstructionsState NextS = getSameOpcode(ReducedVals[I + 1], TLI);\n- if (NextS.getOpcode() == Instruction::ExtractElement &&\n- !NextS.isAltShuffle()) {\n- SmallVector CommonCandidates(Candidates);\n- for (Value *RV : ReducedVals[I + 1]) {\n- Value *RdxVal = TrackedVals.find(RV)->second;\n- // Check if the reduction value was not overriden by the\n- // extractelement instruction because of the vectorization and\n- // exclude it, if it is not compatible with other values.\n- if (auto *Inst = dyn_cast(RdxVal))\n- if (!NextS.getOpcode() || !NextS.isOpcodeOrAlt(Inst))\n- continue;\n- CommonCandidates.push_back(RdxVal);\n- TrackedToOrig.try_emplace(RdxVal, RV);\n- }\n- SmallVector Mask;\n- if (isFixedVectorShuffle(CommonCandidates, Mask)) {\n- ++I;\n- Candidates.swap(CommonCandidates);\n- ShuffledExtracts = true;\n- }\n+ SmallVector CommonCandidates(Candidates);\n+ for (Value *RV : ReducedVals[I + 1]) {\n+ Value *RdxVal = TrackedVals.find(RV)->second;\n+ // Check if the reduction value was not overriden by the\n+ // extractelement instruction because of the vectorization and\n+ // exclude it, if it is not compatible with other values.\n+ auto *Inst = dyn_cast(RdxVal);\n+ if (!Inst)\n+ continue;\n+ CommonCandidates.push_back(RdxVal);\n+ TrackedToOrig.try_emplace(RdxVal, RV);\n+ }\n+ SmallVector Mask;\n+ if (isFixedVectorShuffle(CommonCandidates, Mask)) {\n+ ++I;\n+ Candidates.swap(CommonCandidates);\n+ ShuffledExtracts = true;\n }\n }\n \n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/reduced-value-replace-extractelement.ll", "commands": ["opt -S --passes=slp-vectorizer -slp-threshold=-99999 < %s -mtriple=x86_64-unknown-linux-gnu"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb1, %bb\n %phi = phi i32 [ 0, %bb ], [ %mul9, %bb1 ]\n %phi2 = phi i32 [ 0, %bb ], [ 0, %bb1 ]\n %trunc = trunc i64 0 to i32\n %mul = mul i32 0, %trunc\n %mul3 = mul i32 %trunc, %phi\n %mul4 = mul i32 %mul3, %mul\n %mul5 = mul i32 %mul4, %mul\n %trunc6 = trunc i64 0 to i32\n %mul7 = mul i32 0, %trunc6\n %mul8 = mul i32 %mul5, %mul7\n %mul9 = mul i32 %mul8, %mul7\n br label %bb1\n}\n"}]}, {"file": "llvm/test/Transforms/SLPVectorizer/X86/reduction-gather-non-scheduled-extracts.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-sie-ps5 < %s"], "tests": [{"test_name": "tes", "test_body": "define void @tes() {\nentry:\n %0 = extractelement <2 x i1> zeroinitializer, i64 0\n %1 = extractelement <2 x i1> zeroinitializer, i64 0\n %2 = fcmp ole <2 x double> zeroinitializer, zeroinitializer\n %3 = extractelement <2 x i1> %2, i64 0\n %4 = extractelement <2 x i1> zeroinitializer, i64 0\n br label %5\n\n5: ; preds = %entry\n %6 = select i1 false, i1 false, i1 false\n %7 = select i1 %6, i1 %0, i1 false\n %8 = select i1 %7, i1 %1, i1 false\n %9 = select i1 %8, i1 false, i1 false\n %10 = select i1 %9, i1 %3, i1 false\n %11 = select i1 %10, i1 %4, i1 false\n br i1 %11, label %12, label %13\n\n12: ; preds = %5\n ret void\n\n13: ; preds = %5\n ret void\n}\n"}]}], "issue": {"title": "Crash in getSameOpcode(llvm::ArrayRef, llvm::TargetLibraryInfo const&, unsigned int)", "body": " To reproduce run the test below with `-passes slp-vectorizer -slp-threshold=-99999`\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat() #0 gc \"statepoint-example\" {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb1, %bb\r\n %phi = phi i32 [ 0, %bb ], [ %mul9, %bb1 ]\r\n %phi2 = phi i32 [ 0, %bb ], [ 0, %bb1 ]\r\n %trunc = trunc i64 0 to i32\r\n %mul = mul i32 0, %trunc\r\n %mul3 = mul i32 %trunc, %phi\r\n %mul4 = mul i32 %mul3, %mul\r\n %mul5 = mul i32 %mul4, %mul\r\n %trunc6 = trunc i64 0 to i32\r\n %mul7 = mul i32 0, %trunc6\r\n %mul8 = mul i32 %mul5, %mul7\r\n %mul9 = mul i32 %mul8, %mul7\r\n br label %bb1\r\n}\r\n\r\n```\r\nReproducer: https://godbolt.org/z/sqEov8n9h\r\n\r\nStack dump:\r\n```\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes slp-vectorizer -slp-threshold=-99999 \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"wombat\"\r\n #0 0x0000000004f2ff18 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f2ff18)\r\n #1 0x0000000004f2d68c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007a3cdfe42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000000000403fb8a getSameOpcode(llvm::ArrayRef, llvm::TargetLibraryInfo const&, unsigned int) (.constprop.1) SLPVectorizer.cpp:0:0\r\n #4 0x000000000406b0a4 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n #5 0x000000000406dab4 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x406dab4)\r\n #6 0x00000000040720f8 llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40720f8)\r\n #7 0x00000000040763be llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40763be)\r\n #8 0x000000000407c648 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n #9 0x000000000407d11b llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x407d11b)\r\n#10 0x0000000002e7f61e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2e7f61e)\r\n#11 0x0000000004d2d348 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d2d348)\r\n#12 0x0000000000de6cce llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xde6cce)\r\n#13 0x0000000004d2bd3e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d2bd3e)\r\n#14 0x0000000000de731e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xde731e)\r\n#15 0x0000000004d29ea0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d29ea0)\r\n#16 0x00000000008e4312 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e4312)\r\n#17 0x00000000008d723c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d723c)\r\n#18 0x00007a3cdfe29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#19 0x00007a3cdfe29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#20 0x00000000008ce6ae _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ce6ae)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "crash"], "comments": [{"author": "RKSimon", "body": "CC @alexey-bataev "}]}, "verified": true} {"bug_id": "79742", "issue_url": "https://github.com/llvm/llvm-project/issues/79742", "bug_type": "crash", "base_commit": "a9689c6029e3078e09e43a4efb2b2ced98e9020f", "knowledge_cutoff": "2024-01-28T09:40:40Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "17fb3e82f6c950267bb01cc2fd2a84b0d9e9d0d8", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": [[1031, 1037], [1040, 1047]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": ["VPlanTransforms::truncateToMinimalBitwidths"]}}, "patch": "commit 17fb3e82f6c950267bb01cc2fd2a84b0d9e9d0d8\nAuthor: Florian Hahn \nDate: Tue Apr 23 11:50:25 2024 +0100\n\n [VPlan] Skip extending ICmp results in trunateToMinimalBitwidth.\n \n Results of icmp don't need extending after truncating their operands, as\n the result will always be i1. Skip them during extending.\n \n Fixes https://github.com/llvm/llvm-project/issues/79742\n Fixes https://github.com/llvm/llvm-project/issues/85185\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\nindex 007dc3f89b3f..9580dc4a27f6 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n@@ -1031,7 +1031,9 @@ void VPlanTransforms::truncateToMinimalBitwidths(\n if (auto *VPW = dyn_cast(&R))\n VPW->dropPoisonGeneratingFlags();\n \n- if (OldResSizeInBits != NewResSizeInBits) {\n+ using namespace llvm::VPlanPatternMatch;\n+ if (OldResSizeInBits != NewResSizeInBits &&\n+ !match(&R, m_Binary(m_VPValue(), m_VPValue()))) {\n // Extend result to original width.\n auto *Ext =\n new VPWidenCastRecipe(Instruction::ZExt, ResultVPV, OldResTy);\n@@ -1040,8 +1042,9 @@ void VPlanTransforms::truncateToMinimalBitwidths(\n Ext->setOperand(0, ResultVPV);\n assert(OldResSizeInBits > NewResSizeInBits && \"Nothing to shrink?\");\n } else\n- assert(cast(&R)->getOpcode() == Instruction::ICmp &&\n- \"Only ICmps should not need extending the result.\");\n+ assert(\n+ match(&R, m_Binary(m_VPValue(), m_VPValue())) &&\n+ \"Only ICmps should not need extending the result.\");\n \n assert(!isa(&R) && \"stores cannot be narrowed\");\n if (isa(&R))\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/trunc-extended-icmps.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "test_icmp_and_op_zext", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine i32 @test_icmp_and_op_zext(ptr %dst, i64 %a) {\nentry:\n %and = and i64 %a, 7304878031173690989\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i16 [ 1, %entry ], [ %iv.next, %loop ]\n %c = icmp ne i64 %and, 0\n %c.ext = zext i1 %c to i64\n %or = or i64 %and, %c.ext\n %or.trunc = trunc i64 %or to i8\n %gep = getelementptr i8, ptr %dst, i16 %iv\n store i8 %or.trunc, ptr %gep, align 1\n %iv.next = add i16 %iv, 1\n %ec = icmp eq i16 %iv.next, 1000\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i32 0\n}\n"}, {"test_name": "ext_cmp", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @ext_cmp(ptr %src.1, ptr %src.2, ptr noalias %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src.1 = getelementptr inbounds i16, ptr %src.1, i64 %iv\n %i2 = load i16, ptr %gep.src.1, align 2\n %i3 = sext i16 %i2 to i32\n %c.1 = icmp sgt i32 0, %i3\n %gep.src.2 = getelementptr inbounds i8, ptr %src.2, i64 %iv\n %i4 = load i8, ptr %gep.src.2, align 2\n %i5 = zext i8 %i4 to i32\n %i6 = select i1 %c.1, i32 0, i32 %i5\n %i7 = and i32 %i6, 0\n %i8 = trunc nuw nsw i32 %i7 to i16\n %gep.dst = getelementptr inbounds i16, ptr %dst, i64 %iv\n store i16 %i8, ptr %gep.dst, align 2\n %iv.next = add nsw i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "test_icmp_constant_op_zext", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine i32 @test_icmp_constant_op_zext(ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i16 [ 1, %entry ], [ %iv.next, %loop ]\n %c = icmp ne i64 7304878031173690989, 0\n %c.ext = zext i1 %c to i64\n %or = or i64 7304878031173690989, %c.ext\n %or.trunc = trunc i64 %or to i8\n %gep = getelementptr i8, ptr %dst, i16 %iv\n store i8 %or.trunc, ptr %gep, align 1\n %iv.next = add i16 %iv, 1\n %ec = icmp eq i16 %iv.next, 1000\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i32 0\n}\n"}]}], "issue": {"title": "llvm crash on loop-vectorize: Assertion `OldResSizeInBits > NewResSizeInBits && \"Nothing to shrink?\"' failed.", "body": "llvm crashes on the following IR.\r\n\r\nBisected to 822c749aec69ace75bd3923a4c850e2fd2e34fb1, which was committed by @fhahn \r\n\r\nCompiler explorer: https://godbolt.org/z/rWj9dhvPo\r\n\r\n```llvm\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @i(i64 %0) {\r\nentry:\r\n %conv = sext i16 1 to i32\r\n br label %for.body\r\n\r\nfor.body: ; preds = %for.body, %entry\r\n %1 = phi i32 [ 0, %entry ], [ %inc, %for.body ]\r\n %tobool.not.i = icmp eq i64 %0, 0\r\n %..i = select i1 %tobool.not.i, i32 0, i32 0\r\n %cmp = icmp eq i32 %..i, 0\r\n %conv1 = zext i1 %tobool.not.i to i32\r\n %or = or i32 %conv1, %conv\r\n %cmp2 = icmp slt i32 %or, 7\r\n %conv4 = zext i1 %cmp2 to i64\r\n store i64 %conv4, ptr null, align 8\r\n %inc = add i32 %1, 1\r\n %tobool.not = icmp eq i32 %inc, 0\r\n br i1 %tobool.not, label %for.cond.for.end_crit_edge, label %for.body\r\n\r\nfor.cond.for.end_crit_edge: ; preds = %for.body\r\n ret void\r\n}\r\n```\r\n\r\n```console\r\n% opt -passes=loop-vectorize reduced.ll\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/opt -disable-output -passes=loop-vectorize reduced.ll\r\n #0 0x00007f5302fedeff llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x902eff)\r\n #1 0x00007f5302feb654 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f5302300090 (/lib/x86_64-linux-gnu/libc.so.6+0x43090)\r\n #3 0x00007f5304a458f0 llvm::InnerLoopVectorizer::truncateToMinimalBitwidths(llvm::VPTransformState&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x235a8f0)\r\n #4 0x00007f5304a49d95 llvm::InnerLoopVectorizer::fixVectorizedLoop(llvm::VPTransformState&, llvm::VPlan&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x235ed95)\r\n #5 0x00007f5304a4abb1 llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair>*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x235fbb1)\r\n #6 0x00007f5304a4d564 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x2362564)\r\n #7 0x00007f5304a4fbdd llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x2364bdd)\r\n #8 0x00007f5304a50e20 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x2365e20)\r\n #9 0x00007f5305e0d916 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x3722916)\r\n#10 0x00007f53031fd3fd llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0xb123fd)\r\n#11 0x00007f5305e069f6 llvm::detail::PassModel>, llvm::PreservedAnalyses, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x371b9f6)\r\n#12 0x00007f53031fbc8d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0xb10c8d)\r\n#13 0x00007f5305e0d696 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0x3722696)\r\n#14 0x00007f53031f9c0d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/../lib/libLLVM-18git.so+0xb0ec0d)\r\n#15 0x0000555c5281fcb5 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/opt+0x25cb5)\r\n#16 0x0000555c52813876 main (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/opt+0x19876)\r\n#17 0x00007f53022e1083 __libc_start_main /build/glibc-wuryBv/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#18 0x0000555c5281451e _start (/zdata/shaoli/compilers/ccbuilder-compilers/clang-822c749aec69ace75bd3923a4c850e2fd2e34fb1/bin/opt+0x1a51e)\r\n[1] 1975861 segmentation fault -disable-output -passes=loop-vectorize reduced.ll\r\n```\r\n\r\n\r\nThis IR is reduced from \r\n```C\r\nlong a, b;\r\nint c, d, f;\r\nstatic long *e = &b;\r\nstatic short g = 5;\r\nint h() {\r\n if (a)\r\n return 0;\r\n return 5;\r\n}\r\nvoid i() {\r\n int j;\r\n long **k = &e;\r\n for (; d; d++) {\r\n j = h();\r\n *e = 6 >= (g | (j + c == 0));\r\n }\r\n j = &k || 1;\r\n short *l[6];\r\n l[f] = &g;\r\n}\r\nint main() {}\r\n```", "author": "shao-hua-li", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "114905", "issue_url": "https://github.com/llvm/llvm-project/issues/114905", "bug_type": "miscompilation", "base_commit": "1b476ecdcf4b544af1436341fc923c0b73793cbe", "knowledge_cutoff": "2024-11-05T01:11:52Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "889215a30ed60474e573f9632d1fa362dfa1b04e", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[19821, 19841]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["tryToReduce"]}}, "patch": "commit 889215a30ed60474e573f9632d1fa362dfa1b04e\nAuthor: Alexey Bataev \nDate: Thu Dec 26 05:09:51 2024 -0800\n\n [SLP]Followup fix for the poisonous logical op in reductions\n \n If the VectorizedTree still may generate poisonous value, but it is not\n the original operand of the reduction op, need to check if Res still the\n operand, to generate correct code.\n \n Fixes #114905\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 57f3016fbe1e..e9fc89fa242a 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -19821,21 +19821,21 @@ public:\n Builder.SetCurrentDebugLocation(\n cast(ReductionOps.front().front())->getDebugLoc());\n if (AnyBoolLogicOp) {\n-\n- if (auto It = ReducedValsToOps.find(VectorizedTree);\n- It == ReducedValsToOps.end() ||\n+ auto It = ReducedValsToOps.find(VectorizedTree);\n+ auto It1 = ReducedValsToOps.find(Res);\n+ if ((It == ReducedValsToOps.end() && It1 == ReducedValsToOps.end()) ||\n isGuaranteedNotToBePoison(VectorizedTree, AC) ||\n- any_of(It->getSecond(), [&](Instruction *I) {\n- return isBoolLogicOp(I) &&\n- getRdxOperand(I, 0) == VectorizedTree;\n- })) {\n+ (It != ReducedValsToOps.end() &&\n+ any_of(It->getSecond(), [&](Instruction *I) {\n+ return isBoolLogicOp(I) &&\n+ getRdxOperand(I, 0) == VectorizedTree;\n+ }))) {\n ;\n- } else if (auto It = ReducedValsToOps.find(Res);\n- It == ReducedValsToOps.end() ||\n- isGuaranteedNotToBePoison(Res, AC) ||\n- any_of(It->getSecond(), [&](Instruction *I) {\n+ } else if (isGuaranteedNotToBePoison(Res, AC) ||\n+ (It1 != ReducedValsToOps.end() &&\n+ any_of(It1->getSecond(), [&](Instruction *I) {\n return isBoolLogicOp(I) && getRdxOperand(I, 0) == Res;\n- })) {\n+ }))) {\n std::swap(VectorizedTree, Res);\n } else {\n VectorizedTree = Builder.CreateFreeze(VectorizedTree);\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/reduction-logical.ll", "commands": ["opt < %s -passes=slp-vectorizer -mtriple=x86_64-- -S", "opt < %s -passes=slp-vectorizer -mtriple=x86_64-- -mattr=avx512vl -S"], "tests": [{"test_name": "logical_or_icmp_extra_op", "test_body": "define i1 @logical_or_icmp_extra_op(<4 x i32> %x, <4 x i32> %y, i1 %c) {\n %x0 = extractelement <4 x i32> %x, i32 0\n %x1 = extractelement <4 x i32> %x, i32 1\n %x2 = extractelement <4 x i32> %x, i32 2\n %x3 = extractelement <4 x i32> %x, i32 3\n %y0 = extractelement <4 x i32> %y, i32 0\n %y1 = extractelement <4 x i32> %y, i32 1\n %y2 = extractelement <4 x i32> %y, i32 2\n %y3 = extractelement <4 x i32> %y, i32 3\n %d0 = icmp slt i32 %x0, %y0\n %d1 = icmp slt i32 %x1, %y1\n %d2 = icmp slt i32 %x2, %y2\n %d3 = icmp slt i32 %x3, %y3\n %s3 = select i1 %c, i1 true, i1 %c\n %s4 = select i1 %s3, i1 true, i1 %d0\n %s5 = select i1 %s4, i1 true, i1 %d1\n %s6 = select i1 %s5, i1 true, i1 %d2\n %s7 = select i1 %s6, i1 true, i1 %d3\n ret i1 %s7\n}\n"}, {"test_name": "logical_and_icmp_extra_op", "test_body": "define i1 @logical_and_icmp_extra_op(<4 x i32> %x, <4 x i32> %y, i1 %c) {\n %x0 = extractelement <4 x i32> %x, i32 0\n %x1 = extractelement <4 x i32> %x, i32 1\n %x2 = extractelement <4 x i32> %x, i32 2\n %x3 = extractelement <4 x i32> %x, i32 3\n %y0 = extractelement <4 x i32> %y, i32 0\n %y1 = extractelement <4 x i32> %y, i32 1\n %y2 = extractelement <4 x i32> %y, i32 2\n %y3 = extractelement <4 x i32> %y, i32 3\n %d0 = icmp slt i32 %x0, %y0\n %d1 = icmp slt i32 %x1, %y1\n %d2 = icmp slt i32 %x2, %y2\n %d3 = icmp slt i32 %x3, %y3\n %s3 = select i1 %c, i1 %c, i1 false\n %s4 = select i1 %s3, i1 %d0, i1 false\n %s5 = select i1 %s4, i1 %d1, i1 false\n %s6 = select i1 %s5, i1 %d2, i1 false\n %s7 = select i1 %s6, i1 %d3, i1 false\n ret i1 %s7\n}\n"}]}, {"file": "llvm/test/Transforms/SLPVectorizer/logical-ops-poisonous-repeated.ll", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "test", "test_body": "define i1 @test(<4 x i32> %x) {\n %x0 = extractelement <4 x i32> %x, i32 0\n %x1 = extractelement <4 x i32> %x, i32 -1\n %x2 = extractelement <4 x i32> %x, i32 2\n %x3 = extractelement <4 x i32> %x, i32 3\n %1 = icmp ugt i32 %x0, 0\n %c1 = icmp slt i32 %x1, 0\n %c2 = icmp sgt i32 %x2, 0\n %c3 = icmp slt i32 %x3, 0\n %s1 = select i1 %1, i1 %c1, i1 false\n %s2 = select i1 %s1, i1 %c3, i1 false\n %s3 = select i1 %s2, i1 %c3, i1 false\n ret i1 %s3\n}\n"}]}], "issue": {"title": "wrong folding of `select`s in slp-vectorizer", "body": "Alive2 report: https://alive2.llvm.org/ce/z/jEopx9\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine i1 @logical_and_icmp_diff_preds.2(<4 x i32> %x) {\r\n#0:\r\n %x0 = extractelement <4 x i32> %x, i32 0\r\n %x1 = extractelement <4 x i32> %x, i32 4294967295\r\n %x3 = extractelement <4 x i32> %x, i32 3\r\n %#1 = icmp ugt i32 %x0, 0\r\n %c1 = icmp slt i32 %x1, 0\r\n %c3 = icmp slt i32 %x3, 0\r\n %s1 = select i1 %#1, i1 %c1, i1 0\r\n %s2 = select i1 %s1, i1 %c3, i1 0\r\n %s3 = select i1 %s2, i1 %c3, i1 0\r\n ret i1 %s3\r\n}\r\n=>\r\ndefine i1 @logical_and_icmp_diff_preds.2(<4 x i32> %x) {\r\n#0:\r\n %x0 = extractelement <4 x i32> %x, i32 0\r\n %x1 = extractelement <4 x i32> %x, i32 4294967295\r\n %x3 = extractelement <4 x i32> %x, i32 3\r\n %#1 = icmp ugt i32 %x0, 0\r\n %c1 = icmp slt i32 %x1, 0\r\n %c3 = icmp slt i32 %x3, 0\r\n %op.rdx = select i1 %c3, i1 %c1, i1 0\r\n %op.rdx1 = select i1 %op.rdx, i1 %#1, i1 0\r\n ret i1 %op.rdx1\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\n<4 x i32> %x = < #x00000000 (0), #x00000000 (0), #x00000000 (0), #xffffffff (4294967295, -1) >\r\n\r\nSource:\r\ni32 %x0 = #x00000000 (0)\r\ni32 %x1 = poison\r\ni32 %x3 = #xffffffff (4294967295, -1)\r\ni1 %#1 = #x0 (0)\r\ni1 %c1 = poison\r\ni1 %c3 = #x1 (1)\r\ni1 %s1 = #x0 (0)\r\ni1 %s2 = #x0 (0)\r\ni1 %s3 = #x0 (0)\r\n\r\nTarget:\r\ni32 %x0 = #x00000000 (0)\r\ni32 %x1 = poison\r\ni32 %x3 = #xffffffff (4294967295, -1)\r\ni1 %#1 = #x0 (0)\r\ni1 %c1 = poison\r\ni1 %c3 = #x1 (1)\r\ni1 %op.rdx = poison\r\ni1 %op.rdx1 = poison\r\nSource value: #x0 (0)\r\nTarget value: poison\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": [{"author": "nunoplopes", "body": "Alive2 is still not happy with the fix:\n```llvm\ndefine i1 @test(<4 x i32> %x) {\n %x0 = extractelement <4 x i32> %x, i32 0\n %x1 = extractelement <4 x i32> %x, i32 4294967295\n %x3 = extractelement <4 x i32> %x, i32 3\n %#1 = icmp ugt i32 %x0, 0\n %c1 = icmp slt i32 %x1, 0\n %c3 = icmp slt i32 %x3, 0\n %s1 = select i1 %#1, i1 %c1, i1 0\n %s2 = select i1 %s1, i1 %c3, i1 0\n %s3 = select i1 %s2, i1 %c3, i1 0\n ret i1 %s3\n}\n=>\ndefine i1 @test(<4 x i32> %x) {\n %x0 = extractelement <4 x i32> %x, i32 0\n %x1 = extractelement <4 x i32> %x, i32 4294967295\n %x3 = extractelement <4 x i32> %x, i32 3\n %#1 = icmp ugt i32 %x0, 0\n %c1 = icmp slt i32 %x1, 0\n %c3 = icmp slt i32 %x3, 0\n %#2 = freeze i1 %c3\n %op.rdx = select i1 %#2, i1 %c1, i1 0\n %op.rdx1 = select i1 %op.rdx, i1 %#1, i1 0\n ret i1 %op.rdx1\n}\nTransformation doesn't verify!\n\nERROR: Target is more poisonous than source\n\nExample:\n<4 x i32> %x = < #x00000000 (0), poison, poison, #xffffffff (4294967295, -1) >\n\nSource:\ni32 %x0 = #x00000000 (0)\ni32 %x1 = poison\ni32 %x3 = #xffffffff (4294967295, -1)\ni1 %#1 = #x0 (0)\ni1 %c1 = poison\ni1 %c3 = #x1 (1)\ni1 %s1 = #x0 (0)\ni1 %s2 = #x0 (0)\ni1 %s3 = #x0 (0)\n\nTarget:\ni32 %x0 = #x00000000 (0)\ni32 %x1 = poison\ni32 %x3 = #xffffffff (4294967295, -1)\ni1 %#1 = #x0 (0)\ni1 %c1 = poison\ni1 %c3 = #x1 (1)\ni1 %#2 = #x1 (1)\ni1 %op.rdx = poison\ni1 %op.rdx1 = poison\nSource value: #x0 (0)\nTarget value: poison\n```"}]}, "verified": true} {"bug_id": "108697", "issue_url": "https://github.com/llvm/llvm-project/issues/108697", "bug_type": "crash", "base_commit": "848cb21ddca5dc4d8c52e487264936be8f1bfe8d", "knowledge_cutoff": "2024-09-14T13:01:28Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "9de327c94d0c995803b6485fb28ac4cad1e53bbe", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6551, 6564]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInstructionCost"]}}, "patch": "commit 9de327c94d0c995803b6485fb28ac4cad1e53bbe\nAuthor: Florian Hahn \nDate: Wed Oct 2 20:16:41 2024 +0100\n\n [LV] Generalize predication checks from 2c8836c899 for operands.\n \n This fixes another case where the VPlan-based and legacy cost models\n disagree. If any of the operands is predicated, it can't be trivially\n hoisted and we should consider the cost for evaluating it each loop\n iteration.\n \n Fixes https://github.com/llvm/llvm-project/issues/108697.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex de3b981a4fe3..6e082b1c134d 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6551,14 +6551,17 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I,\n Op2 = cast(PSE.getSCEV(Op2))->getValue();\n }\n auto Op2Info = TTI.getOperandInfo(Op2);\n- auto IsInvariant = [this](Value *Op) {\n+ std::function IsInvariant =\n+ [this, &IsInvariant](Value *Op) -> bool {\n if (!Legal->isInvariant(Op))\n return false;\n- // Consider Op2 invariant, if it is not a predicated instruction in the\n- // loop. In that case, it is not trivially hoistable.\n+ // Consider Op2invariant, if it or its operands aren't predicated\n+ // instruction in the loop. In that case, it is not trivially hoistable.\n return !isa(Op) ||\n !TheLoop->contains(cast(Op)) ||\n- !isPredicatedInst(cast(Op));\n+ (!isPredicatedInst(cast(Op)) &&\n+ all_of(cast(Op)->operands(),\n+ [&IsInvariant](Value *Op) { return IsInvariant(Op); }));\n };\n if (Op2Info.Kind == TargetTransformInfo::OK_AnyValue && IsInvariant(Op2))\n Op2Info.Kind = TargetTransformInfo::OK_UniformValue;\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/predicated-instruction-cost.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "predicated_urem_shl_cost", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @predicated_urem_shl_cost(ptr %A, i32 %x, i1 %c) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 1, %entry ], [ %iv.next, %loop.latch ]\n %gep = getelementptr inbounds i32, ptr %A, i32 %iv\n %l = load i32, ptr %gep, align 4\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n %rem = urem i32 2, %x\n %shl = shl i32 %l, %rem\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %p = phi i32 [ 0, %loop.header ], [ %shl, %then ]\n store i32 %p, ptr %gep, align 4\n %iv.next = add i32 %iv, 1\n %ec = icmp eq i32 %iv, 0\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}, {"test_name": "test_wide_shift_uses_predicated_invariant_instruction", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @test_wide_shift_uses_predicated_invariant_instruction(i32 %d, i1 %c, ptr %dst) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %loop.latch, label %else\n\nelse: ; preds = %loop.header\n %rem = urem i32 100, %d\n %sext = shl i32 %rem, 12\n %shl.i = shl i32 999, %sext\n br label %loop.latch\n\nloop.latch: ; preds = %else, %loop.header\n %p = phi i32 [ %shl.i, %else ], [ 0, %loop.header ]\n %gep = getelementptr inbounds i32, ptr %dst, i32 %iv\n store i32 %p, ptr %gep, align 4\n %iv.next = add i32 %iv, 1\n %ec = icmp eq i32 %iv.next, 100\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}], "issue": {"title": "clang crashes on valid code at -O{2,3} on x86_64-linux-gnu: Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed ", "body": "Compiler Explorer: https://godbolt.org/z/9K4a3Pr7j\r\n\r\n```\r\n[508] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 223e2efa5e886502a9467b7ef700ebce9b7886e8)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[509] % \r\n[509] % clangtk -O2 small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7390: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /local/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20240914-clangtk-m64-O3-build-061621/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20240914-clangtk-m64-O3-build-061621/delta -resource-dir /local/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/11/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-34578e.o -x c small.c\r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4.\tRunning pass \"loop-vectorize\" on function \"main\"\r\n #0 0x000055ba1845c4ff llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x43484ff)\r\n #1 0x000055ba18459a34 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007fbbfabd4420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #3 0x00007fbbfa60b00b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007fbbfa5ea859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007fbbfa5ea729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007fbbfa5ea729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007fbbfa5fbfd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x000055ba19db5fcb llvm::LoopVectorizationPlanner::computeBestVF() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5ca1fcb)\r\n #9 0x000055ba19dd169a llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5cbd69a)\r\n#10 0x000055ba19dd42d1 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5cc02d1)\r\n#11 0x000055ba19dd4943 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5cc0943)\r\n#12 0x000055ba19946a06 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x5832a06)\r\n#13 0x000055ba17dec0d1 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3cd80d1)\r\n#14 0x000055ba15926546 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1812546)\r\n#15 0x000055ba17deaa3d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3cd6a3d)\r\n#16 0x000055ba15925216 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x1811216)\r\n#17 0x000055ba17dea471 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x3cd6471)\r\n#18 0x000055ba18713673 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#19 0x000055ba187167f8 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x46027f8)\r\n#20 0x000055ba18decf5c clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4cd8f5c)\r\n#21 0x000055ba1ac8966c clang::ParseAST(clang::Sema&, bool, bool) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x6b7566c)\r\n#22 0x000055ba18ded378 clang::CodeGenAction::ExecuteAction() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4cd9378)\r\n#23 0x000055ba1908a659 clang::FrontendAction::Execute() (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4f76659)\r\n#24 0x000055ba1900c07e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x4ef807e)\r\n#25 0x000055ba191772b6 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x50632b6)\r\n#26 0x000055ba154f2b15 cc1_main(llvm::ArrayRef, char const*, void*) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13deb15)\r\n#27 0x000055ba154ebbba ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#28 0x000055ba154ef58e clang_main(int, char**, llvm::ToolContext const&) (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13db58e)\r\n#29 0x000055ba153dd43b main (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x12c943b)\r\n#30 0x00007fbbfa5ec083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#31 0x000055ba154eb64e _start (/local/suz-local/software/local/clang-trunk/bin/clang-20+0x13d764e)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 223e2efa5e886502a9467b7ef700ebce9b7886e8)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-682958.c\r\nclangtk: note: diagnostic msg: /tmp/small-682958.sh\r\nclangtk: note: diagnostic msg: \r\n\r\n********************\r\n[510] % \r\n[510] % cat small.c\r\nshort a, b, c, d;\r\nunsigned short e;\r\nint main() {\r\n for (; d; d++) {\r\n e = a + b;\r\n c = b > 3 || e > b ? e : e << b;\r\n c = c ? 2 / c : 0;\r\n b = 1;\r\n }\r\n return 0;\r\n}\r\n```", "author": "zhendongsu", "labels": ["vectorizers", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "116249", "issue_url": "https://github.com/llvm/llvm-project/issues/116249", "bug_type": "miscompilation", "base_commit": "c1f6cb74634509d0e4204dadd46566185fa33e2b", "knowledge_cutoff": "2024-11-14T15:57:03Z", "lit_test_dir": ["llvm/test/Transforms/IndVarSimplify"], "hints": {"fix_commit": "40a647fc7dc6048c92e2d580b61f5feca0785980", "components": ["IndVarSimplify"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/IndVarSimplify.cpp": [[1530, 1535]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/IndVarSimplify.cpp": ["IndVarSimplify::canonicalizeExitCondition"]}}, "patch": "commit 40a647fc7dc6048c92e2d580b61f5feca0785980\nAuthor: Yingwei Zheng \nDate: Fri Nov 15 09:15:37 2024 +0800\n\n [IndVarSimplify] Drop samesign flags after narrowing compares (#116263)\n \n Samesign flag cannot be preserved after narrowing the compare since the\n position of the sign bit is changed.\n \n Closes https://github.com/llvm/llvm-project/issues/116249.\n\ndiff --git a/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp b/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp\nindex 2b2d516a7079..8a3e0bc3eb97 100644\n--- a/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp\n+++ b/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp\n@@ -1530,6 +1530,8 @@ bool IndVarSimplify::canonicalizeExitCondition(Loop *L) {\n L->getLoopPreheader()->getTerminator()->getIterator());\n ICmp->setOperand(Swapped ? 1 : 0, LHSOp);\n ICmp->setOperand(Swapped ? 0 : 1, NewRHS);\n+ // Samesign flag cannot be preserved after narrowing the compare.\n+ ICmp->setSameSign(false);\n if (LHS->use_empty())\n DeadInsts.push_back(LHS);\n };\n", "tests": [{"file": "llvm/test/Transforms/IndVarSimplify/finite-exit-comparisons.ll", "commands": ["opt -passes=indvars -S < %s -indvars-predicate-loops=0"], "tests": [{"test_name": "test_drop_icmp_samesign", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i8 @test_drop_icmp_samesign(i1 %cond, i32 range(i32 0, 32) %x) {\nentry:\n br i1 %cond, label %for.body, label %else\n\nelse: ; preds = %entry\n %call1 = call i8 @callee()\n br label %exit\n\nfor.body: ; preds = %for.body, %entry\n %call2 = call i8 @callee()\n %ext = zext i8 %call2 to i32\n %cond2 = icmp samesign ugt i32 %x, %ext\n br i1 %cond2, label %for.body, label %exit\n\nexit: ; preds = %for.body, %else\n %res = phi i8 [ %call1, %else ], [ %call2, %for.body ]\n ret i8 %res\n}\n\ndeclare i8 @callee()\n"}]}], "issue": {"title": "[IndVarSimplify] samesign flag should be dropped after narrowing", "body": "Found with https://github.com/dtcxzyw/llvm-ub-aware-interpreter.\r\nReproducer: https://alive2.llvm.org/ce/z/fcnvLv\r\n```\r\n; bin/opt -passes=indvars test.ll -S\r\ndefine i8 @test(i1 %cond, i32 range(i32 0, 32) %x) {\r\nentry:\r\n br i1 %cond, label %for.body, label %else\r\n\r\nelse:\r\n %call1 = call i8 @callee()\r\n br label %exit\r\n\r\nfor.body:\r\n %call2 = call i8 @callee()\r\n %ext = zext i8 %call2 to i32\r\n %cond2 = icmp samesign ugt i32 %x, %ext\r\n br i1 %cond2, label %for.body, label %exit\r\n\r\nexit:\r\n %res = phi i8 [ %call1, %else ], [ %call2, %for.body ]\r\n ret i8 %res\r\n}\r\n```\r\n```\r\n\r\n----------------------------------------\r\ndeclare i8 @callee()\r\n\r\ndefine i8 @test(i1 %cond, i32 %x) {\r\ninit:\r\n %#range_0_%x = !range i32 %x, i32 0, i32 32\r\n br label %entry\r\n\r\nentry:\r\n br i1 %cond, label %for.body, label %else\r\n\r\nelse:\r\n %call1 = call i8 @callee()\r\n br label %exit\r\n\r\nfor.body:\r\n %call2 = call i8 @callee()\r\n %ext = zext i8 %call2 to i32\r\n %cond2 = icmp samesign ugt i32 %#range_0_%x, %ext\r\n br i1 %cond2, label %for.body, label %exit\r\n\r\nexit:\r\n %res = phi i8 [ %call1, %else ], [ %call2, %for.body ]\r\n ret i8 %res\r\n}\r\n=>\r\ndeclare i8 @callee()\r\n\r\ndefine i8 @test(i1 %cond, i32 %x) {\r\ninit:\r\n %#range_0_%x = !range i32 %x, i32 0, i32 32\r\n br label %entry\r\n\r\nentry:\r\n br i1 %cond, label %for.body.preheader, label %else\r\n\r\nelse:\r\n %call1 = call i8 @callee()\r\n br label %exit\r\n\r\nfor.body.preheader:\r\n %#0 = trunc i32 %#range_0_%x to i8\r\n br label %for.body\r\n\r\nfor.body:\r\n %call2 = call i8 @callee()\r\n %cond2 = icmp samesign ugt i8 %#0, %call2\r\n br i1 %cond2, label %for.body, label %exit.loopexit\r\n\r\nexit.loopexit:\r\n %call2.lcssa = phi i8 [ %call2, %for.body ]\r\n br label %exit\r\n\r\nexit:\r\n %res = phi i8 [ %call1, %else ], [ %call2.lcssa, %exit.loopexit ]\r\n ret i8 %res\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Source is more defined than target\r\n\r\nExample:\r\ni1 %cond = #x1 (1)\r\ni32 %x = #x00000000 (0)\r\n\r\nSource:\r\ni32 %#range_0_%x = #x00000000 (0)\r\n >> Jump to %entry\r\n >> Jump to %for.body\r\ni8 %call2 = #x80 (128, -128)\r\ni32 %ext = #x00000080 (128)\r\ni1 %cond2 = #x0 (0)\r\n >> Jump to %exit\r\ni8 %res = #x80 (128, -128)\r\n\r\nSOURCE MEMORY STATE\r\n===================\r\nNON-LOCAL BLOCKS:\r\nBlock 0 >\tsize: 0\talign: 1\talloc type: 0\talive: false\taddress: 0\r\nBlock 1 >\tsize: 4\talign: 1\talloc type: 0\talive: true\taddress: 1\r\n\r\nTarget:\r\ni32 %#range_0_%x = #x00000000 (0)\r\n >> Jump to %entry\r\n >> Jump to %for.body.preheader\r\ni8 %#0 = #x00 (0)\r\n >> Jump to %for.body\r\ni8 %call2 = #x80 (128, -128)\r\ni1 %cond2 = poison\r\nUB triggered on br\r\n\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:transforms"], "comments": []}, "verified": true} {"bug_id": "97452", "issue_url": "https://github.com/llvm/llvm-project/issues/97452", "bug_type": "crash", "base_commit": "5c204b1d2619cbad7b7ad228b62feae08781a6db", "knowledge_cutoff": "2024-07-02T17:49:01Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "2b3b405b09a0d965a4aff1f92958418ddbf1e7f6", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6813, 6818]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInstructionCost"]}}, "patch": "commit 2b3b405b09a0d965a4aff1f92958418ddbf1e7f6\nAuthor: Florian Hahn \nDate: Thu Jul 4 11:44:50 2024 +0100\n\n [LV] Don't vectorize first-order recurrence with VF \n \n The assertion added as part of https://github.com/llvm/llvm-project/pull/93395\n surfaced cases where first-order recurrences are vectorized with\n . If vscale is 1, then we are unable to extract the\n penultimate value (second to last lane). Previously this case got\n mis-compiled, trying to extract from an invalid lane (-1)\n https://llvm.godbolt.org/z/3adzYYcf9.\n \n Fixes https://github.com/llvm/llvm-project/issues/97452.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex f9c0c66e6e0a..d7b0240fd8a8 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6813,6 +6813,11 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I, ElementCount VF,\n \n // First-order recurrences are replaced by vector shuffles inside the loop.\n if (VF.isVector() && Legal->isFixedOrderRecurrence(Phi)) {\n+ // For , if vscale = 1 we are unable to extract the\n+ // penultimate value of the recurrence.\n+ // TODO: Consider vscale_range info.\n+ if (VF.isScalable() && VF.getKnownMinValue() == 1)\n+ return InstructionCost::getInvalid();\n SmallVector Mask(VF.getKnownMinValue());\n std::iota(Mask.begin(), Mask.end(), VF.getKnownMinValue() - 1);\n return TTI.getShuffleCost(TargetTransformInfo::SK_Splice,\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/first-order-recurrence-scalable-vf1.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "", "test_body": "\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\n; Make sure we do not pick as VF for a loop with a\n; first-order recurrence.\ndefine i64 @pr97452_scalable_vf1_for(ptr %src) #0 {\n;\nentry:\n br label %loop\n\nloop:\n %for = phi i64 [ 0, %entry ], [ %l, %loop ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.next = add i64 %iv, 1\n %gep = getelementptr inbounds i64, ptr %src, i64 %iv\n %l = load i64, ptr %gep, align 8\n %ec = icmp eq i64 %iv, 22\n br i1 %ec, label %exit, label %loop\n\nexit:\n %res = phi i64 [ %for, %loop ]\n ret i64 %res\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v,+zvl128b,+zvl256b\" }\n;.\n;."}]}, {"file": "llvm/test/Transforms/LoopVectorize/first-order-recurrence-scalable-vf1.ll", "commands": ["opt -p loop-vectorize -scalable-vectorization=on -force-vector-width=1 -force-target-supports-scalable-vectors=true -S %s"], "tests": [{"test_name": "", "test_body": "\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\n\ndefine i64 @pr97452_scalable_vf1_for_live_out(ptr %src) {\n;\nentry:\n br label %loop\n\nloop:\n %for = phi i64 [ 0, %entry ], [ %l, %loop ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.next = add i64 %iv, 1\n %gep = getelementptr inbounds i64, ptr %src, i64 %iv\n %l = load i64, ptr %gep, align 8\n %ec = icmp eq i64 %iv, 22\n br i1 %ec, label %exit, label %loop\n\nexit:\n %res = phi i64 [ %for, %loop ]\n ret i64 %res\n}\n\n\ndefine void @pr97452_scalable_vf1_for_no_live_out(ptr %src, ptr noalias %dst) {\n;\nentry:\n br label %loop\n\nloop:\n %for = phi i64 [ 0, %entry ], [ %l, %loop ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.next = add i64 %iv, 1\n %gep = getelementptr inbounds i64, ptr %src, i64 %iv\n %l = load i64, ptr %gep, align 8\n %gep.dst = getelementptr inbounds i64, ptr %dst, i64 %iv\n store i64 %l, ptr %gep.dst\n %ec = icmp eq i64 %iv, 22\n br i1 %ec, label %exit, label %loop\n\nexit:\n ret void\n}"}]}], "issue": {"title": "[LoopVectorize] Assertion `Offset <= State.VF.getKnownMinValue() && \"invalid offset to extract from\"' failed.", "body": "Reduced testcase:\r\n```llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main() #0 {\r\nentry:\r\n br label %for.body\r\n\r\nfor.cond.cleanup: ; preds = %for.body\r\n %.lcssa17 = phi i64 [ %0, %for.body ]\r\n ret i32 0\r\n\r\nfor.body: ; preds = %for.body, %entry\r\n %0 = phi i64 [ 0, %entry ], [ %1, %for.body ]\r\n %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]\r\n %indvars.iv.next = add i64 %indvars.iv, 1\r\n %1 = load i64, ptr null, align 8\r\n %exitcond.not = icmp eq i64 %indvars.iv, 22\r\n br i1 %exitcond.not, label %for.cond.cleanup, label %for.body\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+v,+zvl128b,+zvl256b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```bash\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll\r\nWARNING: You're attempting to print out a bitcode file.\r\nThis is inadvisable as it may cause display problems. If\r\nyou REALLY want to taste LLVM bitcode first-hand, you\r\ncan force output with the `-f' option.\r\n\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp:569: llvm::Value* llvm::VPInstruction::generatePerPart(llvm::VPTransformState&, unsigned int): Assertion `Offset <= State.VF.getKnownMinValue() && \"invalid offset to extract from\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"main\"\r\n #0 0x000055972ae01800 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2ed2800)\r\n #1 0x000055972adfec0f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2ecfc0f)\r\n #2 0x000055972adfed65 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x0000728161442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007281614969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007281614969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007281614969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x0000728161442476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007281614287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x000072816142871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x0000728161439e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x0000559729f62d80 llvm::VPInstruction::generatePerPart(llvm::VPTransformState&, unsigned int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2033d80)\r\n#12 0x0000559729f63155 llvm::VPInstruction::execute(llvm::VPTransformState&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2034155)\r\n#13 0x0000559729f2de47 llvm::VPBasicBlock::executeRecipes(llvm::VPTransformState*, llvm::BasicBlock*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffee47)\r\n#14 0x0000559729f2e141 llvm::VPIRBasicBlock::execute(llvm::VPTransformState*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fff141)\r\n#15 0x0000559729f30286 llvm::VPlan::execute(llvm::VPTransformState*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2001286)\r\n#16 0x0000559729e2380a llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ef480a)\r\n#17 0x0000559729e34a7c llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f05a7c)\r\n#18 0x0000559729e36f7e llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f07f7e)\r\n#19 0x0000559729e380a0 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f090a0)\r\n#20 0x0000559728cc6dc6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xd97dc6)\r\n#21 0x000055972ac1ce46 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2cede46)\r\n#22 0x0000559728cc9f56 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xd9af56)\r\n#23 0x000055972ac1bbcb llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2cecbcb)\r\n#24 0x0000559728cc6526 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xd97526)\r\n#25 0x000055972ac19d02 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2cead02)\r\n#26 0x0000559728520136 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5f1136)\r\n#27 0x0000559728511d51 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5e2d51)\r\n#28 0x0000728161429d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#29 0x0000728161429e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#30 0x0000728161429e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#31 0x0000559728507b85 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5d8b85)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\nGodbolt: https://godbolt.org/z/ojz8sMa1r\r\n\r\nFound via fuzzer", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": [{"author": "svs-quic", "body": "Bisected to #93395 \r\ncc: @fhahn "}, {"author": "fhahn", "body": "Thanks for the test case. Should be fixed now. The assertion highlighted a case where we previously miscompiled for the case where VF is ; in that case we are unable to extract the penultimate value (second-to-last) if vscale = 1"}]}, "verified": true} {"bug_id": "95547", "issue_url": "https://github.com/llvm/llvm-project/issues/95547", "bug_type": "miscompilation", "base_commit": "7767f0d47428db66d65b07b35aa52f0507df71f9", "knowledge_cutoff": "2024-06-14T13:54:00Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "534f8569a3c9fccfd5cbc5f632b63ad0cf711098", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp": [[291, 300]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp": ["canEvaluateTruncated"]}}, "patch": "commit 534f8569a3c9fccfd5cbc5f632b63ad0cf711098\nAuthor: Nikita Popov \nDate: Mon Jun 17 15:36:55 2024 +0200\n\n [InstCombine] Don't preserve context across div\n \n We can't preserve the context across a non-speculatable instruction,\n as this might introduce a trap. Alternatively, we could also\n insert all the replacement instruction at the use-site, but that\n would be a more intrusive change for the sake of this edge case.\n \n Fixes https://github.com/llvm/llvm-project/issues/95547.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp\nindex 1b4c319032ca..6a6bdba3fa70 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp\n@@ -291,10 +291,12 @@ static bool canEvaluateTruncated(Value *V, Type *Ty, InstCombinerImpl &IC,\n uint32_t BitWidth = Ty->getScalarSizeInBits();\n assert(BitWidth < OrigBitWidth && \"Unexpected bitwidths!\");\n APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, BitWidth);\n- if (IC.MaskedValueIsZero(I->getOperand(0), Mask, 0, CxtI) &&\n- IC.MaskedValueIsZero(I->getOperand(1), Mask, 0, CxtI)) {\n- return canEvaluateTruncated(I->getOperand(0), Ty, IC, CxtI) &&\n- canEvaluateTruncated(I->getOperand(1), Ty, IC, CxtI);\n+ // Do not preserve the original context instruction. Simplifying div/rem\n+ // based on later context may introduce a trap.\n+ if (IC.MaskedValueIsZero(I->getOperand(0), Mask, 0, I) &&\n+ IC.MaskedValueIsZero(I->getOperand(1), Mask, 0, I)) {\n+ return canEvaluateTruncated(I->getOperand(0), Ty, IC, I) &&\n+ canEvaluateTruncated(I->getOperand(1), Ty, IC, I);\n }\n break;\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/trunc.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "trunc_nuw_xor_vector", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine <2 x i1> @trunc_nuw_xor_vector(<2 x i8> %x, <2 x i8> %y) {\n %xor = xor <2 x i8> %x, %y\n %r = trunc nuw <2 x i8> %xor to <2 x i1>\n ret <2 x i1> %r\n}\n"}, {"test_name": "pr95547", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndeclare void @use.i8(i8)\n\ndefine void @pr95547(i32 %x) {\n %x.trunc = trunc i32 %x to i16\n %div = udiv i16 11, %x.trunc\n %cmp = icmp ult i32 %x, 256\n br i1 %cmp, label %loop, label %exit\n\nloop: ; preds = %loop, %0\n %trunc = trunc i16 %div to i8\n call void @use.i8(i8 %trunc)\n br label %loop\n\nexit: ; preds = %0\n ret void\n}\n"}]}], "issue": {"title": "[InstCombine] Wrong value truncation due to incorrect analysis of iv known bits", "body": "```llvm\r\ndefine void @src(i32 %0, ptr %p, i1 %cnd) {\r\nentry:\r\n br i1 %cnd, label %header, label %exit\r\n\r\nheader: ; preds = %latch, %entry\r\n %iv = phi i32 [ %iv.next, %latch ], [ 256, %entry ]\r\n %iv.trunc = trunc i32 %iv to i16\r\n %div = udiv i16 11, %iv.trunc\r\n br i1 %cnd, label %loop1, label %loop0\r\n\r\nloop0: ; preds = %header\r\n %load = load i32, ptr %p, align 4\r\n br label %loop1\r\n\r\nloop1: ; preds = %loop0, %header\r\n %v = phi i32 [ %load, %loop0 ], [ 0, %header ]\r\n store i32 %v, ptr %p, align 4\r\n %guard = icmp ugt i32 %iv, 10\r\n br i1 %guard, label %latch, label %loop2\r\n\r\nloop2: ; preds = %loop1\r\n %div.trunc = trunc i16 %div to i8\r\n store i8 %div.trunc, ptr %p, align 1\r\n br label %latch\r\n\r\nlatch: ; preds = %loop2, %loop1\r\n %iv.next = add nuw nsw i32 %iv, 1\r\n %exitcond = icmp ugt i32 %iv, 300\r\n br i1 %exitcond, label %exit, label %header\r\n\r\nexit: ; preds = %latch\r\n ret void\r\n}\r\n```\r\nIf we run `instcombine` pass on this example, it will convert `trunc i32 %iv to i16` into `trunc i32 %iv to i8` to avoid cast:\r\nhttps://godbolt.org/z/ze4E45ThT\r\n\r\nBut this transformation is wrong, because on the first iteration `%iv` is equal to 256, and the 8 bits trunc makes it equal to `0`. This give us a division by zero on the next instruction: `%div = udiv i16 11, %iv.trunc`.\r\n\r\nIt seems that the problematic patch is https://github.com/llvm/llvm-project/pull/73662\r\n\r\n@nikic Could you please take a look?", "author": "aleks-tmb", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "antoniofrighetto", "body": "May be false positive in Alive2 as well?"}, {"author": "dtcxzyw", "body": "> May be false positive in Alive2 as well?\r\n\r\nYou need `-src-unroll` and `-tgt-unroll` for loops :)\r\nhttps://alive2.llvm.org/ce/z/Vh4zAU"}, {"author": "antoniofrighetto", "body": "Just out of curiosity, did you fine-tune the iterations in a particular way, or some trial-and-error? (I had tried a few ones before and didn’t manage to have the UB triggered)"}, {"author": "dtcxzyw", "body": "> Just out of curiosity, did you fine-tune the iterations in a particular way, or some trial-and-error? (I had tried a few ones before and didn’t manage to have the UB triggered)\r\n\r\nI just full unrolled the loop :)\r\n```\r\n%iv.next = add nuw nsw i32 %iv, 1\r\n%exitcond = icmp ugt i32 %iv, 300\r\n```"}, {"author": "antoniofrighetto", "body": "> I just full unrolled the loop :)\r\n\r\nDefinitely makes sense, thanks!"}, {"author": "nikic", "body": "Reduced test case:\r\n```llvm\r\ndefine void @src(i32 %x) {\r\n %x.trunc = trunc i32 %x to i16\r\n %div = udiv i16 11, %x.trunc\r\n %cmp = icmp ult i32 %x, 256\r\n br i1 %cmp, label %loop, label %exit\r\n\r\nloop:\r\n %trunc = trunc i16 %div to i8\r\n call void @use(i8 %trunc)\r\n br label %loop\r\n\r\nexit:\r\n ret void\r\n}\r\n\r\ndeclare void @use(i8)\r\n```\r\n\r\nThe issue is that canEvaluateTruncated() preserves context instruction across non-speculatable operations."}]}, "verified": true} {"bug_id": "112460", "issue_url": "https://github.com/llvm/llvm-project/issues/112460", "bug_type": "miscompilation", "base_commit": "65cf7afb6d9d8c6137b90d909ee4fcf251439f48", "knowledge_cutoff": "2024-10-16T01:34:25Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "e56e9dd8adca2e86f22783bf5e745ee1ba7ead5f", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[14953, 14958], [17095, 17100]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::collectValuesToDemote", "BoUpSLP::vectorizeTree"]}}, "patch": "commit e56e9dd8adca2e86f22783bf5e745ee1ba7ead5f\nAuthor: Alexey Bataev \nDate: Fri Oct 18 13:32:41 2024 -0700\n\n [SLP]Fix minbitwidth emission and analysis for freeze instruction\n \n Need to add minbw emission and analysis for freeze instruction to fix\n incorrect signedness propagation.\n \n Fixes #112460\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 53632efe913e..e7b52fbdcc3b 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -14953,6 +14953,12 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {\n return E->VectorizedValue;\n }\n \n+ if (Op->getType() != VecTy) {\n+ assert((It != MinBWs.end() || getOperandEntry(E, 0)->isGather() ||\n+ MinBWs.contains(getOperandEntry(E, 0))) &&\n+ \"Expected item in MinBWs.\");\n+ Op = Builder.CreateIntCast(Op, VecTy, GetOperandSignedness(0));\n+ }\n Value *V = Builder.CreateFreeze(Op);\n V = FinalShuffle(V, E);\n \n@@ -17095,6 +17101,8 @@ bool BoUpSLP::collectValuesToDemote(\n return TryProcessInstruction(\n BitWidth, {getOperandEntry(&E, 0), getOperandEntry(&E, 1)});\n }\n+ case Instruction::Freeze:\n+ return TryProcessInstruction(BitWidth, getOperandEntry(&E, 0));\n case Instruction::Shl: {\n // If we are truncating the result of this SHL, and if it's a shift of an\n // inrange amount, we can always perform a SHL in a smaller type.\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/freeze-signedness-missed.ll", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test(i1 %.b, i8 %conv18, i32 %k.promoted61) {\n %not..b79 = xor i1 %.b, true\n %1 = zext i1 %not..b79 to i8\n %cmp.i51 = icmp eq i8 %1, 0\n %cond.i55 = freeze i8 %1\n %.cmp = icmp ugt i8 %cond.i55, %conv18\n %.urem = select i1 %.cmp, i8 0, i8 %cond.i55\n %2 = sub nuw i8 %conv18, %.urem\n %cond.in.i = select i1 %cmp.i51, i8 %conv18, i8 %2\n %not..b80 = xor i1 %.b, true\n %3 = zext i1 %not..b80 to i8\n %cmp.i51.1 = icmp eq i8 %3, 0\n %cond.i55.1 = freeze i8 %3\n %.cmp.1 = icmp ugt i8 %cond.i55.1, %conv18\n %.urem.1 = select i1 %.cmp.1, i8 0, i8 %cond.i55.1\n %4 = sub nuw i8 %conv18, %.urem.1\n %cond.in.i.1 = select i1 %cmp.i51.1, i8 %conv18, i8 %4\n %not..b81 = xor i1 %.b, true\n %5 = zext i1 %not..b81 to i8\n %cmp.i51.2 = icmp eq i8 %5, 0\n %cond.i55.2 = freeze i8 %5\n %.cmp.2 = icmp ugt i8 %cond.i55.2, %conv18\n %.urem.2 = select i1 %.cmp.2, i8 0, i8 %cond.i55.2\n %6 = sub nuw i8 %conv18, %.urem.2\n %cond.in.i.2 = select i1 %cmp.i51.2, i8 %conv18, i8 %6\n %not..b = xor i1 %.b, true\n %7 = zext i1 %not..b to i8\n %cmp.i51.3 = icmp eq i8 %7, 0\n %cond.i55.3 = freeze i8 %7\n %.cmp.3 = icmp ugt i8 %cond.i55.3, %conv18\n %.urem.3 = select i1 %.cmp.3, i8 0, i8 %cond.i55.3\n %8 = sub nuw i8 %conv18, %.urem.3\n %cond.in.i.3 = select i1 %cmp.i51.3, i8 %conv18, i8 %8\n %conv26 = zext nneg i8 %cond.in.i to i32\n %or = or i32 %k.promoted61, %conv26\n %conv26.1 = zext nneg i8 %cond.in.i.1 to i32\n %or.1 = or i32 %or, %conv26.1\n %conv26.2 = zext nneg i8 %cond.in.i.2 to i32\n %or.2 = or i32 %or.1, %conv26.2\n %conv26.3 = zext nneg i8 %cond.in.i.3 to i32\n %or.3 = or i32 %or.2, %conv26.3\n ret i32 %or.3\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Miscompilation at -O2", "body": "Reproducer: https://alive2.llvm.org/ce/z/Tw0oUm\r\n```\r\n; bin/opt -passes=slp-vectorizer test.ll -S\r\ndefine i32 @src(i1 %.b, i8 %conv18, i32 %k.promoted61) {\r\n %not..b79 = xor i1 %.b, true\r\n %3 = zext i1 %not..b79 to i8\r\n %cmp.i51 = icmp eq i8 %3, 0\r\n %cond.i55.frozen = freeze i8 %3\r\n %.cmp = icmp ugt i8 %cond.i55.frozen, %conv18\r\n %.urem = select i1 %.cmp, i8 0, i8 %cond.i55.frozen\r\n %4 = sub nuw i8 %conv18, %.urem\r\n %cond.in.i = select i1 %cmp.i51, i8 %conv18, i8 %4\r\n %not..b80 = xor i1 %.b, true\r\n %5 = zext i1 %not..b80 to i8\r\n %cmp.i51.1 = icmp eq i8 %5, 0\r\n %cond.i55.frozen.1 = freeze i8 %5\r\n %.cmp.1 = icmp ugt i8 %cond.i55.frozen.1, %conv18\r\n %.urem.1 = select i1 %.cmp.1, i8 0, i8 %cond.i55.frozen.1\r\n %6 = sub nuw i8 %conv18, %.urem.1\r\n %cond.in.i.1 = select i1 %cmp.i51.1, i8 %conv18, i8 %6\r\n %not..b81 = xor i1 %.b, true\r\n %7 = zext i1 %not..b81 to i8\r\n %cmp.i51.2 = icmp eq i8 %7, 0\r\n %cond.i55.frozen.2 = freeze i8 %7\r\n %.cmp.2 = icmp ugt i8 %cond.i55.frozen.2, %conv18\r\n %.urem.2 = select i1 %.cmp.2, i8 0, i8 %cond.i55.frozen.2\r\n %8 = sub nuw i8 %conv18, %.urem.2\r\n %cond.in.i.2 = select i1 %cmp.i51.2, i8 %conv18, i8 %8\r\n %not..b = xor i1 %.b, true\r\n %9 = zext i1 %not..b to i8\r\n %cmp.i51.3 = icmp eq i8 %9, 0\r\n %cond.i55.frozen.3 = freeze i8 %9\r\n %.cmp.3 = icmp ugt i8 %cond.i55.frozen.3, %conv18\r\n %.urem.3 = select i1 %.cmp.3, i8 0, i8 %cond.i55.frozen.3\r\n %10 = sub nuw i8 %conv18, %.urem.3\r\n %cond.in.i.3 = select i1 %cmp.i51.3, i8 %conv18, i8 %10\r\n %conv26 = zext nneg i8 %cond.in.i to i32\r\n %or = or i32 %k.promoted61, %conv26\r\n %conv26.1 = zext nneg i8 %cond.in.i.1 to i32\r\n %or.1 = or i32 %or, %conv26.1\r\n %conv26.2 = zext nneg i8 %cond.in.i.2 to i32\r\n %or.2 = or i32 %or.1, %conv26.2\r\n %conv26.3 = zext nneg i8 %cond.in.i.3 to i32\r\n %or.3 = or i32 %or.2, %conv26.3\r\n ret i32 %or.3\r\n}\r\n```\r\n```\r\ndefine i32 @src(i1 %.b, i8 %conv18, i32 %k.promoted61) {\r\n %1 = insertelement <4 x i1> poison, i1 %.b, i32 0\r\n %2 = shufflevector <4 x i1> %1, <4 x i1> poison, <4 x i32> zeroinitializer\r\n %3 = xor <4 x i1> %2, \r\n %4 = zext <4 x i1> %3 to <4 x i8>\r\n %5 = icmp eq <4 x i8> %4, zeroinitializer\r\n %6 = freeze <4 x i1> %3\r\n %7 = sext <4 x i1> %6 to <4 x i8>\r\n %8 = insertelement <4 x i8> poison, i8 %conv18, i32 0\r\n %9 = shufflevector <4 x i8> %8, <4 x i8> poison, <4 x i32> zeroinitializer\r\n %10 = icmp ugt <4 x i8> %7, %9\r\n %11 = select <4 x i1> %10, <4 x i8> zeroinitializer, <4 x i8> %7\r\n %12 = sub nuw <4 x i8> %9, %11\r\n %13 = select <4 x i1> %5, <4 x i8> %9, <4 x i8> %12\r\n %14 = zext <4 x i8> %13 to <4 x i32>\r\n %15 = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> %14)\r\n %op.rdx = or i32 %15, %k.promoted61\r\n ret i32 %op.rdx\r\n}\r\n```\r\n```\r\nTransformation doesn't verify!\r\n\r\nERROR: Value mismatch\r\n\r\nExample:\r\ni1 %.b = #x0 (0)\r\ni8 %conv18 = #x01 (1)\r\ni32 %k.promoted61 = #x00000000 (0)\r\n\r\nSource:\r\ni1 %not..b79 = #x1 (1)\r\ni8 %#1 = #x01 (1)\r\ni1 %cmp.i51 = #x0 (0)\r\ni8 %cond.i55.frozen = #x01 (1)\r\ni1 %.cmp = #x0 (0)\r\ni8 %.urem = #x01 (1)\r\ni8 %#2 = #x00 (0)\r\ni8 %cond.in.i = #x00 (0)\r\ni1 %not..b80 = #x1 (1)\r\ni8 %#3 = #x01 (1)\r\ni1 %cmp.i51.1 = #x0 (0)\r\ni8 %cond.i55.frozen.1 = #x01 (1)\r\ni1 %.cmp.1 = #x0 (0)\r\ni8 %.urem.1 = #x01 (1)\r\ni8 %#4 = #x00 (0)\r\ni8 %cond.in.i.1 = #x00 (0)\r\ni1 %not..b81 = #x1 (1)\r\ni8 %#5 = #x01 (1)\r\ni1 %cmp.i51.2 = #x0 (0)\r\ni8 %cond.i55.frozen.2 = #x01 (1)\r\ni1 %.cmp.2 = #x0 (0)\r\ni8 %.urem.2 = #x01 (1)\r\ni8 %#6 = #x00 (0)\r\ni8 %cond.in.i.2 = #x00 (0)\r\ni1 %not..b = #x1 (1)\r\ni8 %#7 = #x01 (1)\r\ni1 %cmp.i51.3 = #x0 (0)\r\ni8 %cond.i55.frozen.3 = #x01 (1)\r\ni1 %.cmp.3 = #x0 (0)\r\ni8 %.urem.3 = #x01 (1)\r\ni8 %#8 = #x00 (0)\r\ni8 %cond.in.i.3 = #x00 (0)\r\ni32 %conv26 = #x00000000 (0)\r\ni32 %or = #x00000000 (0)\r\ni32 %conv26.1 = #x00000000 (0)\r\ni32 %or.1 = #x00000000 (0)\r\ni32 %conv26.2 = #x00000000 (0)\r\ni32 %or.2 = #x00000000 (0)\r\ni32 %conv26.3 = #x00000000 (0)\r\ni32 %or.3 = #x00000000 (0)\r\n\r\nTarget:\r\n<4 x i1> %#1 = < #x0 (0), poison, poison, poison >\r\n<4 x i1> %#2 = < #x0 (0), #x0 (0), #x0 (0), #x0 (0) >\r\n<4 x i1> %#3 = < #x1 (1), #x1 (1), #x1 (1), #x1 (1) >\r\n<4 x i8> %#4 = < #x01 (1), #x01 (1), #x01 (1), #x01 (1) >\r\n<4 x i1> %#5 = < #x0 (0), #x0 (0), #x0 (0), #x0 (0) >\r\n<4 x i1> %#6 = < #x1 (1), #x1 (1), #x1 (1), #x1 (1) >\r\n<4 x i8> %#7 = < #xff (255, -1), #xff (255, -1), #xff (255, -1), #xff (255, -1) >\r\n<4 x i8> %#8 = < #x01 (1), poison, poison, poison >\r\n<4 x i8> %#9 = < #x01 (1), #x01 (1), #x01 (1), #x01 (1) >\r\n<4 x i1> %#10 = < #x1 (1), #x1 (1), #x1 (1), #x1 (1) >\r\n<4 x i8> %#11 = < #x00 (0), #x00 (0), #x00 (0), #x00 (0) >\r\n<4 x i8> %#12 = < #x01 (1), #x01 (1), #x01 (1), #x01 (1) >\r\n<4 x i8> %#13 = < #x01 (1), #x01 (1), #x01 (1), #x01 (1) >\r\n<4 x i32> %#14 = < #x00000001 (1), #x00000001 (1), #x00000001 (1), #x00000001 (1) >\r\ni32 %#15 = #x00000001 (1)\r\ni32 %op.rdx = #x00000001 (1)\r\nSource value: #x00000000 (0)\r\nTarget value: #x00000001 (1)\r\n```\r\n\r\nllvm version: 69f7758ddba662b63667507f2c472c008909dd7e\r\ncc @alexey-bataev ", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:SLPVectorizer"], "comments": []}, "verified": true} {"bug_id": "122913", "issue_url": "https://github.com/llvm/llvm-project/issues/122913", "bug_type": "crash", "base_commit": "195a1fc5b05d7a42b2e3fa383edb9a7e8b34a9c5", "knowledge_cutoff": "2025-01-14T15:11:18Z", "lit_test_dir": ["llvm/test/Analysis/ScalarEvolution"], "hints": {"fix_commit": "137d706739653304294adef84ed758e3e498d975", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[15328, 15333]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::LoopGuards::collect"]}}, "patch": "commit 137d706739653304294adef84ed758e3e498d975\nAuthor: Julian Nagele \nDate: Wed Jan 22 18:36:37 2025 +0000\n\n [SCEV] Do not attempt to collect loop guards for loops without predecessor. (#123662)\n \n Attempting to collect loop guards for loops without a predecessor can\n lead to non-terminating recursion trying to construct a SCEV.\n \n Fixes https://github.com/llvm/llvm-project/issues/122913.\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex 7673c3548175..210c7cab965e 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -15328,6 +15328,8 @@ ScalarEvolution::LoopGuards::collect(const Loop *L, ScalarEvolution &SE) {\n BasicBlock *Header = L->getHeader();\n BasicBlock *Pred = L->getLoopPredecessor();\n LoopGuards Guards(SE);\n+ if (!Pred)\n+ return Guards;\n SmallPtrSet VisitedBlocks;\n collectFromBlock(SE, Guards, Header, Pred, VisitedBlocks);\n return Guards;\n", "tests": [{"file": "llvm/test/Analysis/ScalarEvolution/backedge-taken-count-guard-info-with-multiple-predecessors.ll", "commands": ["opt < %s -disable-output -passes=nary-reassociate --scalar-evolution-use-expensive-range-sharpening 2>&1"], "tests": [{"test_name": "pr122913", "test_body": "target triple = \"x86_64-unknown-linux-gnu\"\n\ndefine ptr @f(i32 %0) {\n switch i32 0, label %bb4 [\n i32 1, label %bb4\n i32 2, label %bb4\n i32 3, label %bb4\n i32 4, label %bb1\n i32 5, label %bb4\n i32 6, label %bb4\n ]\n\nbb: ; No predecessors!\n switch i32 0, label %bb4 [\n i32 0, label %bb4\n i32 1, label %bb1\n ]\n\nbb1: ; preds = %bb2, %bb, %1\n %2 = phi i32 [ %3, %bb2 ], [ 0, %bb ], [ 0, %1 ]\n switch i32 %0, label %bb3 [\n i32 0, label %bb2\n i32 1, label %bb2\n i32 2, label %bb2\n ]\n\nbb2: ; preds = %bb1, %bb1, %bb1\n %3 = add i32 %2, 1\n %4 = icmp ult i32 %0, 0\n br i1 %4, label %bb1, label %bb4\n\nbb3: ; preds = %bb1\n unreachable\n\nbb4: ; preds = %bb2, %bb, %bb, %1, %1, %1, %1, %1, %1\n ret ptr null\n}"}]}], "issue": {"title": "[SCEV] Another SEGV/stack overflow in LoopGuards", "body": "Similar to https://github.com/llvm/llvm-project/issues/120615. Looks like the fix wasn't a complete one. Here is an example:\n\n```llvm\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine ptr @f(i32 %0) {\n switch i32 0, label %bb4 [\n i32 1, label %bb4\n i32 2, label %bb4\n i32 3, label %bb4\n i32 4, label %bb1\n i32 5, label %bb4\n i32 6, label %bb4\n ]\n\nbb: ; No predecessors!\n switch i32 0, label %bb4 [\n i32 0, label %bb4\n i32 1, label %bb1\n ]\n\nbb1: ; preds = %bb2, %bb, %1\n %2 = phi i32 [ %3, %bb2 ], [ 0, %bb ], [ 0, %1 ]\n switch i32 %0, label %bb3 [\n i32 0, label %bb2\n i32 1, label %bb2\n i32 2, label %bb2\n ]\n\nbb2: ; preds = %bb1, %bb1, %bb1\n %3 = add i32 %2, 1\n %4 = icmp ult i32 %0, 0\n br i1 %4, label %bb1, label %bb4\n\nbb3: ; preds = %bb1\n unreachable\n\nbb4: ; preds = %bb2, %bb, %bb, %1, %1, %1, %1, %1, %1\n ret ptr null\n}\n```\nCrashes with the same command line `opt -passes=nary-reassociate --scalar-evolution-use-expensive-range-sharpening`\ngodbolt: https://godbolt.org/z/4d3jo8jTz", "author": "danilaml", "labels": ["llvm:SCEV", "crash-on-valid"], "comments": [{"author": "juliannagele", "body": "Thanks, checking!"}, {"author": "danilaml", "body": "@juliannagele Thanks for looking into it! Do you need any additional info?"}]}, "verified": true} {"bug_id": "113462", "issue_url": "https://github.com/llvm/llvm-project/issues/113462", "bug_type": "crash", "base_commit": "3605d9a456185f4af78c01a2684b822b57bca9b0", "knowledge_cutoff": "2024-10-23T14:33:27Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "b65b2b4ab60763515694c740935989f908a03312", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[9840, 9852]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["computeExtractCost"]}}, "patch": "commit b65b2b4ab60763515694c740935989f908a03312\nAuthor: Alexey Bataev \nDate: Wed Oct 23 12:03:38 2024 -0700\n\n [SLP]Expand vector to the whole register size in extracts adjustment\n \n Need to expand the number of elements to the whole register to correctly\n process estimation and avoid compiler crash.\n \n Fixes #113462\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex d824c40c401d..889c4d94ba5e 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -9840,13 +9840,14 @@ class BoUpSLP::ShuffleCostEstimator : public BaseShuffleAnalysis {\n ::getShuffleCost(TTI, *RegShuffleKind,\n getWidenedType(ScalarTy, EltsPerVector), SubMask);\n }\n+ const unsigned BaseVF = getFullVectorNumberOfElements(\n+ *R.TTI, VL.front()->getType(), alignTo(NumElts, EltsPerVector));\n for (unsigned Idx : Indices) {\n- assert((Idx + EltsPerVector) <= alignTo(NumElts, EltsPerVector) &&\n+ assert((Idx + EltsPerVector) <= BaseVF &&\n \"SK_ExtractSubvector index out of range\");\n- Cost += ::getShuffleCost(\n- TTI, TTI::SK_ExtractSubvector,\n- getWidenedType(ScalarTy, alignTo(NumElts, EltsPerVector)), {},\n- CostKind, Idx, getWidenedType(ScalarTy, EltsPerVector));\n+ Cost += ::getShuffleCost(TTI, TTI::SK_ExtractSubvector,\n+ getWidenedType(ScalarTy, BaseVF), {}, CostKind,\n+ Idx, getWidenedType(ScalarTy, EltsPerVector));\n }\n // Second attempt to check, if just a permute is better estimated than\n // subvector extract.\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/partial-register-extract.ll", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndeclare void @callee(i32, i64)\n\ndefine i32 @main(i32 %g_274.promoted2477.i.i.i, ptr %p) {\nentry:\n %g_274.promoted2477.i.i.i1 = load i32, ptr %p, align 4\n br i1 false, label %for.inc347.i.i.i, label %for.cond159.preheader.i.i.i\n\nfor.cond159.preheader.i.i.i: ; preds = %entry\n %tobool239.not.us.i.i.i.not = icmp eq i32 %g_274.promoted2477.i.i.i1, 0\n %tobool239.not.us.i.1.i.i = icmp eq i32 %g_274.promoted2477.i.i.i1, 0\n %tobool239.not.us.i.3.i.i = icmp eq i32 %g_274.promoted2477.i.i.i1, 0\n %0 = or i1 %tobool239.not.us.i.3.i.i, %tobool239.not.us.i.1.i.i\n %tobool239.not.us.i.4.i.i = icmp eq i32 %g_274.promoted2477.i.i.i1, 0\n %inc34.i21362404.us.v.i.4.i.i = select i1 %tobool239.not.us.i.4.i.i, i64 0, i64 0\n %tobool239.not.us.i.5.i.i = icmp eq i32 0, 0\n %inc34.i21362404.us.v.i.5.i.i = select i1 %tobool239.not.us.i.5.i.i, i64 0, i64 0\n %inc34.i21362404.us.i.5.i.i = or i64 %inc34.i21362404.us.v.i.4.i.i, %inc34.i21362404.us.v.i.5.i.i\n %1 = or i1 %tobool239.not.us.i.5.i.i, %tobool239.not.us.i.4.i.i\n %inc34.i21362404.us.v.i.6.i.i = select i1 false, i64 0, i64 0\n %inc34.i21362404.us.i.6.i.i = or i64 %inc34.i21362404.us.i.5.i.i, %inc34.i21362404.us.v.i.6.i.i\n %tobool239.not.us.i.7.i.i = icmp eq i32 0, 0\n %inc34.i21362404.us.v.i.7.i.i = select i1 false, i64 0, i64 0\n %inc34.i21362404.us.i.7.i.i = or i64 %inc34.i21362404.us.i.6.i.i, %inc34.i21362404.us.v.i.7.i.i\n %inc34.i21362404.us.v.i.8.i.i = select i1 false, i64 0, i64 0\n %inc34.i21362404.us.i.8.i.i = or i64 %inc34.i21362404.us.i.7.i.i, %inc34.i21362404.us.v.i.8.i.i\n %inc34.i21362404.us.v.i.9.i.i = select i1 false, i64 0, i64 0\n %inc34.i21362404.us.i.9.i.i = or i64 %inc34.i21362404.us.i.8.i.i, %inc34.i21362404.us.v.i.9.i.i\n %tobool239.not.us.i.10.i.i = icmp eq i32 0, 0\n %tobool239.not.us.i.11.i.i = icmp eq i32 0, 0\n %2 = or i1 %tobool239.not.us.i.11.i.i, %tobool239.not.us.i.10.i.i\n %tobool239.not.us.i.12.i.i = icmp eq i32 %g_274.promoted2477.i.i.i, 0\n %3 = or i1 %tobool239.not.us.i.12.i.i, %2\n %tobool239.not.us.i.13.i.i = icmp eq i32 0, 0\n %tobool239.not.us.i.14.i.i = icmp eq i32 0, 0\n %4 = or i1 %tobool239.not.us.i.14.i.i, %tobool239.not.us.i.13.i.i\n %tobool239.not.us.i.16.i.i = icmp eq i32 0, 0\n %tobool239.not.us.i.17.i.i = icmp eq i32 0, 0\n %5 = or i1 %tobool239.not.us.i.17.i.i, %tobool239.not.us.i.16.i.i\n %tobool239.not.us.i.18.i.i = icmp eq i32 0, 0\n %6 = or i1 %tobool239.not.us.i.18.i.i, %5\n %tobool239.not.us.i.19.i.i = icmp eq i32 0, 0\n %tobool239.not.us.i.20.i.i = icmp eq i32 0, 0\n %7 = or i1 %tobool239.not.us.i.20.i.i, %tobool239.not.us.i.19.i.i\n %tobool239.not.us.i.21.i.i = icmp eq i32 0, 0\n %8 = or i1 %tobool239.not.us.i.21.i.i, %7\n %tobool239.not.us.i.22.i.i = icmp eq i32 0, 0\n %tobool239.not.us.i.23.i.i = icmp eq i32 0, 0\n %9 = or i1 %tobool239.not.us.i.23.i.i, %tobool239.not.us.i.22.i.i\n %tobool239.not.us.i.24.i.i = icmp eq i32 0, 0\n %10 = or i1 %tobool239.not.us.i.24.i.i, %9\n %11 = select i1 %10, i1 true, i1 %8\n %12 = select i1 %11, i1 true, i1 %6\n %13 = select i1 %12, i1 true, i1 %4\n %14 = select i1 %13, i1 true, i1 %3\n %15 = select i1 %14, i1 true, i1 %tobool239.not.us.i.7.i.i\n %16 = select i1 %15, i1 true, i1 %1\n %17 = or i1 %0, %tobool239.not.us.i.i.i.not\n %18 = select i1 %16, i1 true, i1 %17\n %and252.us.i.24.i.i = select i1 %18, i32 0, i32 0\n br label %for.inc347.i.i.i\n\nfor.inc347.i.i.i: ; preds = %for.cond159.preheader.i.i.i, %entry\n %g_202.promoted.i.i.i = phi i32 [ %and252.us.i.24.i.i, %for.cond159.preheader.i.i.i ], [ 0, %entry ]\n %.us-phi24612468.i.i.i = phi i64 [ %inc34.i21362404.us.i.9.i.i, %for.cond159.preheader.i.i.i ], [ 0, %entry ]\n call void @callee(i32 %g_202.promoted.i.i.i, i64 %.us-phi24612468.i.i.i)\n ret i32 0\n}"}]}], "issue": {"title": "[SLPVectorizer] Assertion `(Idx + EltsPerVector) <= alignTo(NumElts, EltsPerVector) && \"SK_ExtractSubvector index out of range\"' failed.", "body": "Reproducer: https://godbolt.org/z/fxYWYnWKs\r\n```\r\n; bin/opt -passes=slp-vectorizer reduced.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndeclare void @callee(i32, i64)\r\n\r\ndefine i32 @main(i32 %g_274.promoted2477.i.i.i, ptr %p) {\r\nentry:\r\n %g_274.promoted2477.i.i.i1 = load i32, ptr %p, align 4\r\n br i1 false, label %for.inc347.i.i.i, label %for.cond159.preheader.i.i.i\r\n\r\nfor.cond159.preheader.i.i.i: ; preds = %entry\r\n %tobool239.not.us.i.i.i.not = icmp eq i32 %g_274.promoted2477.i.i.i1, 0\r\n %tobool239.not.us.i.1.i.i = icmp eq i32 %g_274.promoted2477.i.i.i1, 0\r\n %tobool239.not.us.i.3.i.i = icmp eq i32 %g_274.promoted2477.i.i.i1, 0\r\n %0 = or i1 %tobool239.not.us.i.3.i.i, %tobool239.not.us.i.1.i.i\r\n %tobool239.not.us.i.4.i.i = icmp eq i32 %g_274.promoted2477.i.i.i1, 0\r\n %inc34.i21362404.us.v.i.4.i.i = select i1 %tobool239.not.us.i.4.i.i, i64 0, i64 0\r\n %tobool239.not.us.i.5.i.i = icmp eq i32 0, 0\r\n %inc34.i21362404.us.v.i.5.i.i = select i1 %tobool239.not.us.i.5.i.i, i64 0, i64 0\r\n %inc34.i21362404.us.i.5.i.i = or i64 %inc34.i21362404.us.v.i.4.i.i, %inc34.i21362404.us.v.i.5.i.i\r\n %1 = or i1 %tobool239.not.us.i.5.i.i, %tobool239.not.us.i.4.i.i\r\n %inc34.i21362404.us.v.i.6.i.i = select i1 false, i64 0, i64 0\r\n %inc34.i21362404.us.i.6.i.i = or i64 %inc34.i21362404.us.i.5.i.i, %inc34.i21362404.us.v.i.6.i.i\r\n %tobool239.not.us.i.7.i.i = icmp eq i32 0, 0\r\n %inc34.i21362404.us.v.i.7.i.i = select i1 false, i64 0, i64 0\r\n %inc34.i21362404.us.i.7.i.i = or i64 %inc34.i21362404.us.i.6.i.i, %inc34.i21362404.us.v.i.7.i.i\r\n %inc34.i21362404.us.v.i.8.i.i = select i1 false, i64 0, i64 0\r\n %inc34.i21362404.us.i.8.i.i = or i64 %inc34.i21362404.us.i.7.i.i, %inc34.i21362404.us.v.i.8.i.i\r\n %inc34.i21362404.us.v.i.9.i.i = select i1 false, i64 0, i64 0\r\n %inc34.i21362404.us.i.9.i.i = or i64 %inc34.i21362404.us.i.8.i.i, %inc34.i21362404.us.v.i.9.i.i\r\n %tobool239.not.us.i.10.i.i = icmp eq i32 0, 0\r\n %tobool239.not.us.i.11.i.i = icmp eq i32 0, 0\r\n %2 = or i1 %tobool239.not.us.i.11.i.i, %tobool239.not.us.i.10.i.i\r\n %tobool239.not.us.i.12.i.i = icmp eq i32 %g_274.promoted2477.i.i.i, 0\r\n %3 = or i1 %tobool239.not.us.i.12.i.i, %2\r\n %tobool239.not.us.i.13.i.i = icmp eq i32 0, 0\r\n %tobool239.not.us.i.14.i.i = icmp eq i32 0, 0\r\n %4 = or i1 %tobool239.not.us.i.14.i.i, %tobool239.not.us.i.13.i.i\r\n %tobool239.not.us.i.16.i.i = icmp eq i32 0, 0\r\n %tobool239.not.us.i.17.i.i = icmp eq i32 0, 0\r\n %5 = or i1 %tobool239.not.us.i.17.i.i, %tobool239.not.us.i.16.i.i\r\n %tobool239.not.us.i.18.i.i = icmp eq i32 0, 0\r\n %6 = or i1 %tobool239.not.us.i.18.i.i, %5\r\n %tobool239.not.us.i.19.i.i = icmp eq i32 0, 0\r\n %tobool239.not.us.i.20.i.i = icmp eq i32 0, 0\r\n %7 = or i1 %tobool239.not.us.i.20.i.i, %tobool239.not.us.i.19.i.i\r\n %tobool239.not.us.i.21.i.i = icmp eq i32 0, 0\r\n %8 = or i1 %tobool239.not.us.i.21.i.i, %7\r\n %tobool239.not.us.i.22.i.i = icmp eq i32 0, 0\r\n %tobool239.not.us.i.23.i.i = icmp eq i32 0, 0\r\n %9 = or i1 %tobool239.not.us.i.23.i.i, %tobool239.not.us.i.22.i.i\r\n %tobool239.not.us.i.24.i.i = icmp eq i32 0, 0\r\n %10 = or i1 %tobool239.not.us.i.24.i.i, %9\r\n %11 = select i1 %10, i1 true, i1 %8\r\n %12 = select i1 %11, i1 true, i1 %6\r\n %13 = select i1 %12, i1 true, i1 %4\r\n %14 = select i1 %13, i1 true, i1 %3\r\n %15 = select i1 %14, i1 true, i1 %tobool239.not.us.i.7.i.i\r\n %16 = select i1 %15, i1 true, i1 %1\r\n %17 = or i1 %0, %tobool239.not.us.i.i.i.not\r\n %18 = select i1 %16, i1 true, i1 %17\r\n %and252.us.i.24.i.i = select i1 %18, i32 0, i32 0\r\n br label %for.inc347.i.i.i\r\n\r\nfor.inc347.i.i.i: ; preds = %for.cond159.preheader.i.i.i, %entry\r\n %g_202.promoted.i.i.i = phi i32 [ %and252.us.i.24.i.i, %for.cond159.preheader.i.i.i ], [ 0, %entry ]\r\n %.us-phi24612468.i.i.i = phi i64 [ %inc34.i21362404.us.i.9.i.i, %for.cond159.preheader.i.i.i ], [ 0, %entry ]\r\n call void @callee(i32 %g_202.promoted.i.i.i, i64 %.us-phi24612468.i.i.i)\r\n ret i32 0\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:9844: llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::computeExtractCost(llvm::ArrayRef, llvm::ArrayRef, llvm::ArrayRef >, unsigned int): Assertion `(Idx + EltsPerVector) <= alignTo(NumElts, EltsPerVector) && \"SK_ExtractSubvector index out of range\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"main\"\r\n #0 0x00000000050d8608 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50d8608)\r\n #1 0x00000000050d5fbc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007ed1b9642520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007ed1b96969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007ed1b9642476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007ed1b96287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007ed1b962871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007ed1b9639e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x000000000413c75f llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::computeExtractCost(llvm::ArrayRef, llvm::ArrayRef, llvm::ArrayRef>, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x413c75f)\r\n #9 0x0000000004140d8b llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::adjustExtracts(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::MutableArrayRef, llvm::ArrayRef>, unsigned int, bool&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4140d8b)\r\n#10 0x000000000417b42a llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::processBuildVector, llvm::slpvectorizer::BoUpSLP, llvm::SmallPtrSetImpl>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::TargetTransformInfo&, llvm::ArrayRef&, llvm::slpvectorizer::BoUpSLP&, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x417b42a)\r\n#11 0x000000000417cae7 llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x417cae7)\r\n#12 0x00000000041b6e11 llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41b6e11)\r\n#13 0x00000000041bc5df (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#14 0x00000000041bfb71 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41bfb71)\r\n#15 0x00000000041c431b llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\r\n#16 0x00000000041c856e llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41c856e)\r\n#17 0x00000000041cf231 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#18 0x00000000041cfedb llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41cfedb)\r\n#19 0x0000000002f64f0e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f64f0e)\r\n#20 0x0000000004ed9788 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed9788)\r\n#21 0x0000000000e2773e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe2773e)\r\n#22 0x0000000004ed81ae llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed81ae)\r\n#23 0x0000000000e26e8e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe26e8e)\r\n#24 0x0000000004ed7be0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed7be0)\r\n#25 0x000000000090a07a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x90a07a)\r\n#26 0x00000000008fce7c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fce7c)\r\n#27 0x00007ed1b9629d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#28 0x00007ed1b9629e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#29 0x00000000008f493e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f493e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\n\r\nllvm version: 8a9921f5692ab33451d11454b40a023ca0965a69\r\n", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "89338", "issue_url": "https://github.com/llvm/llvm-project/issues/89338", "bug_type": "miscompilation", "base_commit": "ab1d988e1483895a71dcd257bdcaf6fdb0a83d62", "knowledge_cutoff": "2024-04-19T01:55:15Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "46957a138dea339799059ea5bf032e40042ff185", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp": [[1988, 1994]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp": ["InstCombinerImpl::visitCallInst"]}}, "patch": "commit 46957a138dea339799059ea5bf032e40042ff185\nAuthor: Nikita Popov \nDate: Fri Apr 19 14:00:49 2024 +0900\n\n [InstCombine] Fix incorrect fshr to fshl transform\n \n This transform is only valid if the (modular) shift amount is not\n zero.\n \n Proof: https://alive2.llvm.org/ce/z/WBxn-x\n \n Fixes https://github.com/llvm/llvm-project/issues/89338.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\nindex aa547565bd08..e7a2f54f8680 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\n@@ -1988,7 +1988,10 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {\n // is not entirely arbitrary. For historical reasons, the backend may\n // recognize rotate left patterns but miss rotate right patterns.\n if (IID == Intrinsic::fshr) {\n- // fshr X, Y, C --> fshl X, Y, (BitWidth - C)\n+ // fshr X, Y, C --> fshl X, Y, (BitWidth - C) if C is not zero.\n+ if (!isKnownNonZero(ShAmtC, SQ.getWithInstruction(II)))\n+ return nullptr;\n+\n Constant *LeftShiftC = ConstantExpr::getSub(WidthC, ShAmtC);\n Module *Mod = II->getModule();\n Function *Fshl = Intrinsic::getDeclaration(Mod, Intrinsic::fshl, Ty);\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/fsh.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "fsh_unary_shuffle_ops_partial_widening", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i32> @llvm.fshr.v2i32(<2 x i32>, <2 x i32>, <2 x i32>) #0\n\ndefine <2 x i32> @fsh_unary_shuffle_ops_partial_widening(<3 x i32> %x, <2 x i32> %y, <2 x i32> %z) {\n %a = shufflevector <3 x i32> %x, <3 x i32> poison, <2 x i32> \n %b = shufflevector <2 x i32> %y, <2 x i32> poison, <2 x i32> \n %c = shufflevector <2 x i32> %z, <2 x i32> poison, <2 x i32> \n %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c)\n ret <2 x i32> %r\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "fshr_vec_zero_elem", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i32> @llvm.fshr.v2i32(<2 x i32>, <2 x i32>, <2 x i32>) #0\n\ndefine <2 x i32> @fshr_vec_zero_elem(<2 x i32> %x, <2 x i32> %y) {\n %fsh = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i32> )\n ret <2 x i32> %fsh\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "instcombine is wrong about a vector fshr->shl transformation", "body": "Instcombine seems to be mis-optimizing this function:\r\n```llvm\r\ndefine i32 @f(<4 x i8> %0) {\r\n %2 = call <4 x i8> @llvm.fshr.v4i8(<4 x i8> %0, <4 x i8> zeroinitializer, <4 x i8> )\r\n %3 = bitcast <4 x i8> %2 to i32\r\n ret i32 %3\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare <4 x i8> @llvm.fshr.v4i8(<4 x i8>, <4 x i8>, <4 x i8>) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\nthe result is:\r\n```llvm\r\ndefine i32 @f(<4 x i8> %0) {\r\n %2 = shl <4 x i8> %0, \r\n %3 = bitcast <4 x i8> %2 to i32\r\n ret i32 %3\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare <4 x i8> @llvm.fshr.v4i8(<4 x i8>, <4 x i8>, <4 x i8>) #0\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare <4 x i8> @llvm.fshl.v4i8(<4 x i8>, <4 x i8>, <4 x i8>) #0\r\n\r\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\n\r\nAlive says:\r\n```\r\nERROR: Value mismatch\r\n\r\nExample:\r\n<4 x i8> %#0 = < #x00 (0), #x01 (1), #x00 (0), #x00 (0) >\r\n\r\nSource:\r\n<4 x i8> %#2 = < #x00 (0), #x00 (0), #x00 (0), #x00 (0) >\r\ni32 %#3 = #x00000000 (0)\r\n\r\nTarget:\r\n<4 x i8> %#2 = < #x00 (0), #x01 (1), #x00 (0), #x00 (0) >\r\ni32 %#3 = #x00000100 (256)\r\nSource value: #x00000000 (0)\r\nTarget value: #x00000100 (256)\r\n```\r\n\r\nhttps://alive2.llvm.org/ce/z/s4UpPe\r\n\r\ncc @hatsunespica\r\n\r\n", "author": "regehr", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "nikic", "body": "Smaller test case: https://alive2.llvm.org/ce/z/hScyGk"}, {"author": "nikic", "body": "I believe the root cause is the fshr to fshl canonicalization: https://alive2.llvm.org/ce/z/NQM3tR\r\n\r\nWe're performing this transform: https://alive2.llvm.org/ce/z/6UUEsS But we don't verify the pre-condition that the (modular) shift amount is not zero!\r\n\r\nFor non-vectors this is a non-issue because we'd fold a zero shift earlier, but for vectors it is an issue."}]}, "verified": true} {"bug_id": "110787", "issue_url": "https://github.com/llvm/llvm-project/issues/110787", "bug_type": "crash", "base_commit": "8805ff72a04308562ea9e962415d238dd356be69", "knowledge_cutoff": "2024-10-02T04:16:44Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "4dede756f25b153a81ea883ef84064f5a1eaaec2", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[18848, 18853], [18877, 18883]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["tryToReduce"]}}, "patch": "commit 4dede756f25b153a81ea883ef84064f5a1eaaec2\nAuthor: Alexey Bataev \nDate: Wed Oct 2 05:50:20 2024 -0700\n\n [SLP]Transform nodes before building externally used values\n \n transformNodes function may create new vector nodes, so the reduced\n values might be vectorized later. Need to build the list of the\n externally used values after the transformNodes() function call to avoid\n compiler crash.\n \n Fixe #110787\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 15e798bd6c98..ce1e5753c0c8 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -18848,6 +18848,7 @@ public:\n ++SameValuesCounter.try_emplace(OrigV).first->second;\n }\n }\n+ V.transformNodes();\n SmallPtrSet VLScalars(VL.begin(), VL.end());\n // Gather externally used values.\n SmallPtrSet Visited;\n@@ -18877,7 +18878,6 @@ public:\n for (Value *RdxVal : VL)\n if (RequiredExtract.contains(RdxVal))\n LocalExternallyUsedValues[RdxVal];\n- V.transformNodes();\n V.buildExternalUses(LocalExternallyUsedValues);\n \n V.computeMinimumValueSizes();\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/reduced-val-vectorized-in-transform.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test(i1 %cond) {\nentry:\n br label %bb\n\nbb: ; preds = %bb, %entry\n %p1 = phi i32 [ %or92, %bb ], [ 0, %entry ]\n %p2 = phi i32 [ %0, %bb ], [ 0, %entry ]\n %p3 = phi i32 [ %4, %bb ], [ 0, %entry ]\n %0 = or i32 1, 0\n %or8.i = or i32 0, 0\n %or9.i = or i32 0, 0\n %or91 = or i32 %p1, 0\n %or12.i = or i32 %p2, 0\n %or92 = or i32 1, 0\n %1 = xor i32 %or91, %or12.i\n %2 = xor i32 %1, %or9.i\n %3 = xor i32 %2, %or8.i\n %4 = xor i32 %3, %or92\n br i1 %cond, label %exit, label %bb\n\nexit: ; preds = %bb\n ret i32 %4\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Assertion `(I->use_empty() || all_of(I->uses(), [&](Use &U) { return isDeleted( cast(U.getUser())); })) && \"trying to erase instruction with users.\"' failed.", "body": "Reproducer: https://godbolt.org/z/b8cz56M9z\r\n```\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i32 @test(i1 %cond) {\r\nentry:\r\n br label %for.end14.i.i.i.us.i\r\n\r\nfor.end14.i.i.i.us.i:\r\n %.promoted10921110.i.i.i.us.i = phi i32 [ %inc229.1.2.i.i.i.us.us.12.i, %for.end14.i.i.i.us.i ], [ 0, %entry ]\r\n %.promoted1106.i.i.i.us.i = phi i32 [ %0, %for.end14.i.i.i.us.i ], [ 0, %entry ]\r\n %g_276.promoted956960.i.i.i.us.i = phi i32 [ %4, %for.end14.i.i.i.us.i ], [ 0, %entry ]\r\n %0 = or i32 1, 0\r\n %inc229.1974.i.i.i.us.us.8.i = or i32 0, 0\r\n %inc229.1974.i.i.i.us.us.9.i = or i32 0, 0\r\n %inc229.1.2.i.i.i.us.us.9.i = or i32 %.promoted10921110.i.i.i.us.i, 0\r\n %inc229.1974.i.i.i.us.us.12.i = or i32 %.promoted1106.i.i.i.us.i, 0\r\n %inc229.1.2.i.i.i.us.us.12.i = or i32 1, 0\r\n %1 = xor i32 %inc229.1.2.i.i.i.us.us.9.i, %inc229.1974.i.i.i.us.us.12.i\r\n %2 = xor i32 %1, %inc229.1974.i.i.i.us.us.9.i\r\n %3 = xor i32 %2, %inc229.1974.i.i.i.us.us.8.i\r\n %4 = xor i32 %3, %inc229.1.2.i.i.i.us.us.12.i\r\n br i1 %cond, label %exit, label %for.end14.i.i.i.us.i\r\n\r\nexit:\r\n ret i32 %4\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:2720: void llvm::slpvectorizer::BoUpSLP::removeInstructionsAndOperands(llvm::ArrayRef) [with T = llvm::Instruction]: Assertion `(I->use_empty() || all_of(I->uses(), [&](Use &U) { return isDeleted( cast(U.getUser())); })) && \"trying to erase instruction with users.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"test\"\r\n #0 0x000000000507e338 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x507e338)\r\n #1 0x000000000507bcec SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f54a0842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007f54a08969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007f54a0842476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007f54a08287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007f54a082871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007f54a0839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000004102d6f void llvm::slpvectorizer::BoUpSLP::removeInstructionsAndOperands(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4102d6f)\r\n #9 0x000000000415dea2 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::MapVector, llvm::DenseMap, llvm::detail::DenseMapPair>, llvm::SmallVector>, 0u>> const&, llvm::Instruction*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x415dea2)\r\n#10 0x0000000004167819 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#11 0x000000000416a302 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x416a302)\r\n#12 0x000000000416ebb8 llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x416ebb8)\r\n#13 0x0000000004172da1 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4172da1)\r\n#14 0x00000000041796b1 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#15 0x000000000417a33b llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x417a33b)\r\n#16 0x0000000002f33c8e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f33c8e)\r\n#17 0x0000000004e81158 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e81158)\r\n#18 0x0000000000e15bde llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe15bde)\r\n#19 0x0000000004e7fb7e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7fb7e)\r\n#20 0x0000000000e1538e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe1538e)\r\n#21 0x0000000004e7f5b0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e7f5b0)\r\n#22 0x0000000000904102 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x904102)\r\n#23 0x00000000008f6f8c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f6f8c)\r\n#24 0x00007f54a0829d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#25 0x00007f54a0829e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#26 0x00000000008eea5e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8eea5e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\nllvm version: e379094328e49731a606304f7e3559d4f1fa96f9", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": [{"author": "alexey-bataev", "body": "Fixed in 4dede756f25b153a81ea883ef84064f5a1eaaec2"}]}, "verified": true} {"bug_id": "113423", "issue_url": "https://github.com/llvm/llvm-project/issues/113423", "bug_type": "miscompilation", "base_commit": "a1c6dc223ee1eedc049890676992b508ebd6b623", "knowledge_cutoff": "2024-10-23T07:13:21Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "cf1963afad335cf74a9411f106d1f2fe80dbed2f", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[529, 537], [564, 569]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["InstCombinerImpl::foldSelectIntoOp"]}}, "patch": "commit cf1963afad335cf74a9411f106d1f2fe80dbed2f\nAuthor: Yingwei Zheng \nDate: Thu Oct 31 23:26:45 2024 +0800\n\n [InstCombine] Fix FMF propagation in `foldSelectIntoOp` (#114356)\n \n Closes https://github.com/llvm/llvm-project/issues/113423.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex c5f39a4c381e..9e193e7faa8a 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -529,9 +529,6 @@ Instruction *InstCombinerImpl::foldSelectIntoOp(SelectInst &SI, Value *TrueVal,\n if (!OpToFold)\n return nullptr;\n \n- // TODO: We probably ought to revisit cases where the select and FP\n- // instructions have different flags and add tests to ensure the\n- // behaviour is correct.\n FastMathFlags FMF;\n if (isa(&SI))\n FMF = SI.getFastMathFlags();\n@@ -564,6 +561,8 @@ Instruction *InstCombinerImpl::foldSelectIntoOp(SelectInst &SI, Value *TrueVal,\n BinaryOperator *BO =\n BinaryOperator::Create(TVI->getOpcode(), FalseVal, NewSel);\n BO->copyIRFlags(TVI);\n+ if (isa(&SI))\n+ BO->andIRFlags(NewSel);\n return BO;\n };\n \n", "tests": [{"file": "llvm/test/Transforms/InstCombine/fold-select-fmul-if-zero.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "fmul_by_self_if_0_oeq_zero_f32_fmul_nnan_ninf_nsz", "test_body": "define float @fmul_by_self_if_0_oeq_zero_f32_fmul_nnan_ninf_nsz(float %x) {\n %x.is.zero = fcmp oeq float %x, 0.000000e+00\n %scaled.x = fmul nnan ninf nsz float %x, %x\n %scaled.if.denormal = select nnan i1 %x.is.zero, float %scaled.x, float %x\n ret float %scaled.if.denormal\n}\n"}]}], "issue": {"title": "Wrong folding of `select` into floating number ops", "body": "https://github.com/llvm/llvm-project/blob/ae618d360456c5ccf6c8cf2294e708ac6625950e/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp#L559-L568\r\n\r\nThe compiler explorer's alive2 times out for this case.\r\nBut I post a proof that the optimization actually happens: https://alive2.llvm.org/ce/z/7MRSu5\r\n\r\n\r\nThis is the Alive2 report on my machine:\r\n```llvm\r\n----------------------------------------\r\ndefine float @fmul_by_self_if_0_oeq_zero_f32_fmul_nnan_ninf_nsz.2(float %x) {\r\n#0:\r\n %#1 = fcmp ule float %x, 0.000000\r\n %scaled.x = fmul nnan ninf nsz float %x, %x\r\n %scaled.if.denormal = select nnan i1 %#1, float %x, float %scaled.x\r\n ret float %scaled.if.denormal\r\n}\r\n=>\r\ndefine float @fmul_by_self_if_0_oeq_zero_f32_fmul_nnan_ninf_nsz.2(float %x) {\r\n#0:\r\n %#1 = fcmp ule float %x, 0.000000\r\n %scaled.x = select nnan i1 %#1, float 1.000000, float %x\r\n %scaled.if.denormal = fmul nnan ninf nsz float %scaled.x, %x\r\n ret float %scaled.if.denormal\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\nfloat %x = #xff800000 (-oo)\r\n\r\nSource:\r\ni1 %#1 = #x1 (1)\r\nfloat %scaled.x = poison\r\nfloat %scaled.if.denormal = #xff800000 (-oo)\r\n\r\nTarget:\r\ni1 %#1 = #x1 (1)\r\nfloat %scaled.x = #x3f800000 (1)\r\nfloat %scaled.if.denormal = poison\r\nSource value: #xff800000 (-oo)\r\nTarget value: poison\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:instcombine", "floating-point"], "comments": []}, "verified": true} {"bug_id": "121110", "issue_url": "https://github.com/llvm/llvm-project/issues/121110", "bug_type": "miscompilation", "base_commit": "f68dbbbd57dd0947730300d1e827ad16c2dfffb5", "knowledge_cutoff": "2024-12-25T13:57:11Z", "lit_test_dir": ["llvm/test/Transforms/VectorCombine"], "hints": {"fix_commit": "f2f02b21cd581057e3c9b4a7a27e0014eeb9ba15", "components": ["VectorCombine"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": [[1669, 1675], [1677, 1684], [1727, 1740], [1750, 1759]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": ["VectorCombine::foldShuffleOfBinops"]}}, "patch": "commit f2f02b21cd581057e3c9b4a7a27e0014eeb9ba15\nAuthor: Simon Pilgrim \nDate: Sat Dec 28 09:21:22 2024 +0000\n\n [VectorCombine] foldShuffleOfBinops - only accept exact matching cmp predicates\n \n m_SpecificCmp allowed equivalent predicate+flags which don't necessarily work after being folded from \"shuffle (cmpop), (cmpop)\" into \"cmpop (shuffle), (shuffle)\"\n \n Fixes #121110\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\nindex ecbc13d489eb..2460ccc61d84 100644\n--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n@@ -1669,7 +1669,8 @@ bool VectorCombine::foldShuffleOfBinops(Instruction &I) {\n \n Value *X, *Y, *Z, *W;\n bool IsCommutative = false;\n- CmpPredicate Pred = CmpInst::BAD_ICMP_PREDICATE;\n+ CmpPredicate PredLHS = CmpInst::BAD_ICMP_PREDICATE;\n+ CmpPredicate PredRHS = CmpInst::BAD_ICMP_PREDICATE;\n if (match(LHS, m_BinOp(m_Value(X), m_Value(Y))) &&\n match(RHS, m_BinOp(m_Value(Z), m_Value(W)))) {\n auto *BO = cast(LHS);\n@@ -1677,8 +1678,9 @@ bool VectorCombine::foldShuffleOfBinops(Instruction &I) {\n if (llvm::is_contained(OldMask, PoisonMaskElem) && BO->isIntDivRem())\n return false;\n IsCommutative = BinaryOperator::isCommutative(BO->getOpcode());\n- } else if (match(LHS, m_Cmp(Pred, m_Value(X), m_Value(Y))) &&\n- match(RHS, m_SpecificCmp(Pred, m_Value(Z), m_Value(W)))) {\n+ } else if (match(LHS, m_Cmp(PredLHS, m_Value(X), m_Value(Y))) &&\n+ match(RHS, m_Cmp(PredRHS, m_Value(Z), m_Value(W))) &&\n+ (CmpInst::Predicate)PredLHS == (CmpInst::Predicate)PredRHS) {\n IsCommutative = cast(LHS)->isCommutative();\n } else\n return false;\n@@ -1727,14 +1729,14 @@ bool VectorCombine::foldShuffleOfBinops(Instruction &I) {\n TTI.getShuffleCost(SK0, BinOpTy, NewMask0, CostKind, 0, nullptr, {X, Z}) +\n TTI.getShuffleCost(SK1, BinOpTy, NewMask1, CostKind, 0, nullptr, {Y, W});\n \n- if (Pred == CmpInst::BAD_ICMP_PREDICATE) {\n+ if (PredLHS == CmpInst::BAD_ICMP_PREDICATE) {\n NewCost +=\n TTI.getArithmeticInstrCost(LHS->getOpcode(), ShuffleDstTy, CostKind);\n } else {\n auto *ShuffleCmpTy =\n FixedVectorType::get(BinOpTy->getElementType(), ShuffleDstTy);\n NewCost += TTI.getCmpSelInstrCost(LHS->getOpcode(), ShuffleCmpTy,\n- ShuffleDstTy, Pred, CostKind);\n+ ShuffleDstTy, PredLHS, CostKind);\n }\n \n LLVM_DEBUG(dbgs() << \"Found a shuffle feeding two binops: \" << I\n@@ -1750,10 +1752,10 @@ bool VectorCombine::foldShuffleOfBinops(Instruction &I) {\n \n Value *Shuf0 = Builder.CreateShuffleVector(X, Z, NewMask0);\n Value *Shuf1 = Builder.CreateShuffleVector(Y, W, NewMask1);\n- Value *NewBO = Pred == CmpInst::BAD_ICMP_PREDICATE\n+ Value *NewBO = PredLHS == CmpInst::BAD_ICMP_PREDICATE\n ? Builder.CreateBinOp(\n cast(LHS)->getOpcode(), Shuf0, Shuf1)\n- : Builder.CreateCmp(Pred, Shuf0, Shuf1);\n+ : Builder.CreateCmp(PredLHS, Shuf0, Shuf1);\n \n // Intersect flags from the old binops.\n if (auto *NewInst = dyn_cast(NewBO)) {\n", "tests": [{"file": "llvm/test/Transforms/VectorCombine/X86/shuffle-of-cmps.ll", "commands": ["opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=sse2", "opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=sse4.2", "opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=avx2", "opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=avx512vl"], "tests": [{"test_name": "shuf_icmp_ugt_v4i32_use", "test_body": "declare void @use(<4 x i1>)\n\ndefine <4 x i32> @shuf_icmp_ugt_v4i32_use(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z, <4 x i32> %w) {\n %c0 = icmp ugt <4 x i32> %x, %y\n %c1 = icmp ugt <4 x i32> %z, %w\n call void @use(<4 x i1> %c0)\n %s = shufflevector <4 x i1> %c0, <4 x i1> %c1, <4 x i32> \n %r = sext <4 x i1> %s to <4 x i32>\n ret <4 x i32> %r\n}\n"}, {"test_name": "PR121110", "test_body": "define <2 x i1> @PR121110() {\n %ugt = icmp samesign ugt <2 x i32> zeroinitializer, zeroinitializer\n %sgt = icmp sgt <2 x i32> zeroinitializer, \n %res = shufflevector <2 x i1> %ugt, <2 x i1> %sgt, <2 x i32> \n ret <2 x i1> %res\n}\n"}]}], "issue": {"title": "[clang] Miscompilation at -Os", "body": "This code prints 9 at `-Os` and 0 at `-O0/1/2/3`:\n\n```c\nint printf(const char *, ...);\nchar a = 9, c, d;\nint b, e;\nunsigned short f;\nchar *g = &a;\nvoid h(int i) {\n for (; b; b++)\n c &= 0 <= i;\n}\nstatic short j(unsigned long i) {\n int k;\n for (; e + d + 4 > 0;) {\n k = i + 49;\n h(k + i - 52 + i);\n *g = 0;\n return 0;\n }\n return 0;\n}\nint main() {\n j(6 < (unsigned short)(f - 7) + f);\n printf(\"%d\\n\", a);\n}\n```\n\nCompiler Explorer: https://godbolt.org/z/Mo5TEKh8r\n\nBisected to https://github.com/llvm/llvm-project/commit/5287299f8809ae927a0acafb179c4b37ce9ff21d, which was committed by @RKSimon ", "author": "cardigan1008", "labels": ["miscompilation", "vectorizers"], "comments": [{"author": "RKSimon", "body": "I'm not convinced the patch is responsible, it's more likely exposed something else. I'll take a look after the Christmas break."}, {"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/jdkwFh\n```\n; bin/opt -passes=vector-combine test.ll -S\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine <2 x i1> @src() {\nentry:\n %0 = icmp samesign ugt <2 x i32> zeroinitializer, zeroinitializer\n %1 = icmp sgt <2 x i32> zeroinitializer, \n %2 = shufflevector <2 x i1> %0, <2 x i1> %1, <2 x i32> \n ret <2 x i1> %2\n}\n```\n```\ndefine <2 x i1> @tgt() {\nentry:\n ret <2 x i1> zeroinitializer\n}\n```\n```\n\n----------------------------------------\ndefine <2 x i1> @src() {\nentry:\n %#0 = icmp samesign ugt <2 x i32> { 0, 0 }, { 0, 0 }\n %#1 = icmp sgt <2 x i32> { 0, 0 }, { 6, 4294967292 }\n %#2 = shufflevector <2 x i1> %#0, <2 x i1> %#1, 0, 3\n ret <2 x i1> %#2\n}\n=>\ndefine <2 x i1> @tgt() {\nentry:\n ret <2 x i1> { 0, 0 }\n}\nTransformation doesn't verify!\n\nERROR: Value mismatch\n\nExample:\n\nSource:\n<2 x i1> %#0 = < #x0 (0), #x0 (0) >\n<2 x i1> %#1 = < #x0 (0), #x1 (1) >\n<2 x i1> %#2 = < #x0 (0), #x1 (1) >\n\nTarget:\nSource value: < #x0 (0), #x1 (1) >\nTarget value: < #x0 (0), #x0 (0) >\n```\n"}]}, "verified": true} {"bug_id": "119893", "issue_url": "https://github.com/llvm/llvm-project/issues/119893", "bug_type": "crash", "base_commit": "cd093c2e1bac35dd2c6b914d0b64ce56683cb50a", "knowledge_cutoff": "2024-12-13T16:00:39Z", "lit_test_dir": ["llvm/test/Transforms/EarlyCSE"], "hints": {"fix_commit": "5528388e3664c6d7d292f20a739f1bf1c8ef768d", "components": ["EarlyCSE", "IR"], "bug_location_lineno": {"llvm/include/llvm/IR/CmpPredicate.h": [[71, 83]], "llvm/lib/IR/Instructions.cpp": [[3946, 3955]], "llvm/lib/Transforms/Scalar/EarlyCSE.cpp": [[290, 296]]}, "bug_location_funcname": {"llvm/lib/IR/Instructions.cpp": ["CmpPredicate::getSwapped", "llvm::hash_value"], "llvm/lib/Transforms/Scalar/EarlyCSE.cpp": ["getHashValueImpl"]}}, "patch": "commit 5528388e3664c6d7d292f20a739f1bf1c8ef768d\nAuthor: Ramkumar Ramachandra \nDate: Fri Dec 13 22:06:39 2024 +0000\n\n EarlyCSE: fix CmpPredicate duplicate-hashing (#119902)\n \n Strip hash_value() for CmpPredicate, as different callers have different\n hashing use-cases. In this case, there is just one caller, namely\n EarlyCSE, which calls hash_combine() on a CmpPredicate, which used to\n call hash_combine() on a CmpInst::Predicate prior to 4a0d53a\n (PatternMatch: migrate to CmpPredicate). This has uncovered a bug where\n two icmp instructions differing in just the fact that one of them has\n the samesign flag on it are hashed differently, leading to divergent\n hashing, and a crash. Fix this crash by dropping samesign information on\n icmp instructions before hashing them, preserving the former behavior.\n \n Fixes #119893.\n\ndiff --git a/llvm/include/llvm/IR/CmpPredicate.h b/llvm/include/llvm/IR/CmpPredicate.h\nindex ce78e4311f9f..9aa1449465f5 100644\n--- a/llvm/include/llvm/IR/CmpPredicate.h\n+++ b/llvm/include/llvm/IR/CmpPredicate.h\n@@ -71,13 +71,7 @@ public:\n \n /// Get the swapped predicate of a CmpInst.\n static CmpPredicate getSwapped(const CmpInst *Cmp);\n-\n- /// Provided to facilitate storing a CmpPredicate in data structures that\n- /// require hashing.\n- friend hash_code hash_value(const CmpPredicate &Arg); // NOLINT\n };\n-\n-[[nodiscard]] hash_code hash_value(const CmpPredicate &Arg);\n } // namespace llvm\n \n #endif\ndiff --git a/llvm/lib/IR/Instructions.cpp b/llvm/lib/IR/Instructions.cpp\nindex d1da02c744f1..2d6fe40f4c1d 100644\n--- a/llvm/lib/IR/Instructions.cpp\n+++ b/llvm/lib/IR/Instructions.cpp\n@@ -3946,10 +3946,6 @@ CmpPredicate CmpPredicate::getSwapped(const CmpInst *Cmp) {\n return getSwapped(get(Cmp));\n }\n \n-hash_code llvm::hash_value(const CmpPredicate &Arg) { // NOLINT\n- return hash_combine(Arg.Pred, Arg.HasSameSign);\n-}\n-\n //===----------------------------------------------------------------------===//\n // SwitchInst Implementation\n //===----------------------------------------------------------------------===//\ndiff --git a/llvm/lib/Transforms/Scalar/EarlyCSE.cpp b/llvm/lib/Transforms/Scalar/EarlyCSE.cpp\nindex 682c5c3d8c63..3a0ae6b01a11 100644\n--- a/llvm/lib/Transforms/Scalar/EarlyCSE.cpp\n+++ b/llvm/lib/Transforms/Scalar/EarlyCSE.cpp\n@@ -290,7 +290,8 @@ static unsigned getHashValueImpl(SimpleValue Val) {\n Pred = CmpInst::getInversePredicate(Pred);\n std::swap(A, B);\n }\n- return hash_combine(Inst->getOpcode(), Pred, X, Y, A, B);\n+ return hash_combine(Inst->getOpcode(),\n+ static_cast(Pred), X, Y, A, B);\n }\n \n if (CastInst *CI = dyn_cast(Inst))\n", "tests": [{"file": "llvm/test/Transforms/EarlyCSE/pr119893.ll", "commands": ["opt -passes=early-cse -S %s"], "tests": [{"test_name": "samesign_hash_bug", "test_body": "define i32 @samesign_hash_bug(i16 %v) {\n %zext = zext i16 %v to i32\n %icmp.samesign = icmp samesign ugt i32 %zext, 31\n %select.icmp.samesign = select i1 %icmp.samesign, i32 0, i32 1\n %ashr = ashr i32 0, %select.icmp.samesign\n %icmp = icmp ugt i32 %zext, 31\n %select.icmp = select i1 %icmp, i32 1, i32 0\n %ret = add i32 %ashr, %select.icmp\n ret i32 %ret\n}\n"}]}], "issue": {"title": "[EarlyCSE] Assertion `!FoundVal && \"Key already in new map?\"' failed.", "body": "Reproducer:\n```\n; bin/opt -passes=early-cse reduced.ll -S\n\ndefine i32 @func_207(i16 %p_208.coerce, i32 %conv, i32 %0, i64 %1, ptr %p) {\nentry:\n %conv1 = zext i16 %p_208.coerce to i32\n %conv31 = sext i32 %conv to i64\n %cmp.i = icmp eq i64 %conv31, 0\n %xor = xor i32 %conv, 1\n store i32 %xor, ptr %p, align 4\n %cmp2.i = icmp samesign ugt i32 %conv1, 31\n %shr.i601 = select i1 %cmp2.i, i32 0, i32 %conv1\n %cond.i602 = ashr i32 1, %shr.i601\n %conv6.i603 = trunc i32 %cond.i602 to i16\n %conv245 = trunc i16 %p_208.coerce to i8\n %mul.i628 = mul i8 -107, %conv245\n %conv247 = sext i8 %mul.i628 to i64\n %cond.i629 = call i64 @llvm.smin.i64(i64 0, i64 %conv247)\n %cmp249 = icmp slt i64 %cond.i629, 0\n %conv251 = zext i1 %cmp249 to i64\n %cmp1.i630 = icmp ugt i32 %conv1, 31\n %narrow.i631 = select i1 %cmp1.i630, i32 0, i32 %conv1\n %shr.i632 = zext i32 %narrow.i631 to i64\n %cond.i633 = lshr i64 %conv251, %shr.i632\n %cmp258 = icmp slt i16 %p_208.coerce, 0\n %2 = zext i1 %cmp258 to i16\n %cmp261 = icmp ugt i16 1, %2\n %conv263 = zext i1 %cmp261 to i64\n %cmp344 = icmp eq i16 %p_208.coerce, 0\n %conv345 = zext i1 %cmp344 to i32\n store i32 %conv345, ptr %p, align 4\n %conv351 = sext i32 %0 to i64\n %sub.i641 = call i64 @llvm.ucmp.i64.i64(i64 0, i64 %conv351)\n %conv353 = trunc i64 %sub.i641 to i16\n %3 = mul i16 %conv353, -1\n %conv355 = zext i16 %3 to i64\n %cmp356 = icmp sle i64 1, %conv355\n %conv357 = zext i1 %cmp356 to i32\n %conv359 = trunc i32 %conv357 to i8\n %conv.i650 = sext i8 %conv359 to i32\n %4 = icmp ugt i32 %conv, 0\n %shr.i652 = lshr i32 1, %conv\n %cmp9.i = icmp slt i32 %shr.i652, 1\n %or.cond.i = select i1 %4, i1 false, i1 %cmp9.i\n %shl.i653 = shl i32 %conv.i650, 1\n %5 = trunc i32 %shl.i653 to i8\n %cond.i654 = select i1 %or.cond.i, i8 0, i8 %5\n %conv3612 = sext i8 %cond.i654 to i32\n %conv3623 = trunc i64 %1 to i32\n %6 = or i32 1, %conv3612\n %or.cond.i655 = icmp slt i32 %6, 0\n %cmp3.i = icmp sgt i32 %conv3623, 0\n %or.cond4.i = or i1 %cmp3.i, %or.cond.i655\n %shr.i656 = select i1 %or.cond4.i, i32 0, i32 1\n %cond.i657 = ashr i32 %conv, %shr.i656\n %cmp.i658 = icmp slt i32 %cond.i657, 0\n %shr.i660 = select i1 %cmp.i658, i32 0, i32 1\n %cond.i661 = ashr i32 1, %shr.i660\n %conv365 = trunc i32 %cond.i661 to i16\n %add.i662 = or i16 1, %conv365\n %conv368 = sext i16 %add.i662 to i32\n ret i32 %conv368\n}\n```\n```\nopt: /data/zyw/llvm-project/llvm/include/llvm/ADT/DenseMap.h:419: void llvm::DenseMapBase::moveFromOldBuckets(BucketT*, BucketT*) [with DerivedT = llvm::DenseMap<{anonymous}::SimpleValue, llvm::ScopedHashTableVal<{anonymous}::SimpleValue, llvm::Value*>*, llvm::DenseMapInfo<{anonymous}::SimpleValue>, llvm::detail::DenseMapPair<{anonymous}::SimpleValue, llvm::ScopedHashTableVal<{anonymous}::SimpleValue, llvm::Value*>*> >; KeyT = {anonymous}::SimpleValue; ValueT = llvm::ScopedHashTableVal<{anonymous}::SimpleValue, llvm::Value*>*; KeyInfoT = llvm::DenseMapInfo<{anonymous}::SimpleValue>; BucketT = llvm::detail::DenseMapPair<{anonymous}::SimpleValue, llvm::ScopedHashTableVal<{anonymous}::SimpleValue, llvm::Value*>*>]: Assertion `!FoundVal && \"Key already in new map?\"' failed.\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\nStack dump:\n0. Program arguments: /data/zyw/llvm-build/bin/opt -passes=early-cse reduced.ll\n1. Running pass \"function(early-cse<>)\" on module \"reduced.ll\"\n2. Running pass \"early-cse<>\" on function \"func_207\"\n #0 0x00007ffff7def832 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/data/zyw/llvm-build/bin/../lib/libLLVMSupport.so.20.0git+0x1ef832)\n #1 0x00007ffff7dec9ef llvm::sys::RunSignalHandlers() (/data/zyw/llvm-build/bin/../lib/libLLVMSupport.so.20.0git+0x1ec9ef)\n #2 0x00007ffff7decb35 SignalHandler(int) Signals.cpp:0:0\n #3 0x00007ffff7842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #4 0x00007ffff78969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #5 0x00007ffff7842476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #6 0x00007ffff78287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #7 0x00007ffff782871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #8 0x00007ffff7839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #9 0x00007ffff4efa75e llvm::DenseMap<(anonymous namespace)::SimpleValue, llvm::ScopedHashTableVal<(anonymous namespace)::SimpleValue, llvm::Value*>*, llvm::DenseMapInfo<(anonymous namespace)::SimpleValue, void>, llvm::detail::DenseMapPair<(anonymous namespace)::SimpleValue, llvm::ScopedHashTableVal<(anonymous namespace)::SimpleValue, llvm::Value*>*>>::grow(unsigned int) EarlyCSE.cpp:0:0\n#10 0x00007ffff4efbd5a llvm::ScopedHashTable<(anonymous namespace)::SimpleValue, llvm::Value*, llvm::DenseMapInfo<(anonymous namespace)::SimpleValue, void>, llvm::RecyclingAllocator, llvm::ScopedHashTableVal<(anonymous namespace)::SimpleValue, llvm::Value*>, 32ul, 8ul>>::insertIntoScope(llvm::ScopedHashTableScope<(anonymous namespace)::SimpleValue, llvm::Value*, llvm::DenseMapInfo<(anonymous namespace)::SimpleValue, void>, llvm::RecyclingAllocator, llvm::ScopedHashTableVal<(anonymous namespace)::SimpleValue, llvm::Value*>, 32ul, 8ul>>*, (anonymous namespace)::SimpleValue const&, llvm::Value* const&) EarlyCSE.cpp:0:0\n#11 0x00007ffff4efdf19 (anonymous namespace)::EarlyCSE::processNode(llvm::DomTreeNodeBase*) (.isra.0) EarlyCSE.cpp:0:0\n#12 0x00007ffff4efeef5 (anonymous namespace)::EarlyCSE::run() EarlyCSE.cpp:0:0\n#13 0x00007ffff4f00796 llvm::EarlyCSEPass::run(llvm::Function&, llvm::AnalysisManager&) (/data/zyw/llvm-build/bin/../lib/../lib/libLLVMScalarOpts.so.20.0git+0x100796)\n#14 0x00007ffff6684345 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/data/zyw/llvm-build/bin/../lib/../lib/libLLVMPasses.so.20.0git+0x84345)\n#15 0x00007ffff38ee63d llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/data/zyw/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x2ee63d)\n#16 0x00007ffff70a9445 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/data/zyw/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.20.0git+0xa9445)\n#17 0x00007ffff38ec628 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/data/zyw/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x2ec628)\n#18 0x00007ffff70a9e05 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/data/zyw/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.20.0git+0xa9e05)\n#19 0x00007ffff38ed28d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/data/zyw/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x2ed28d)\n#20 0x00007ffff7f9b1eb llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/data/zyw/llvm-build/bin/../lib/libLLVMOptDriver.so.20.0git+0x241eb)\n#21 0x00007ffff7fa5619 optMain (/data/zyw/llvm-build/bin/../lib/libLLVMOptDriver.so.20.0git+0x2e619)\n#22 0x00007ffff7829d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#23 0x00007ffff7829e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#24 0x0000555555555095 _start (/data/zyw/llvm-build/bin/opt+0x1095)\nAborted (core dumped)\n```", "author": "dtcxzyw", "labels": ["crash-on-valid", "llvm:transforms"], "comments": [{"author": "dtcxzyw", "body": "Reverting 4a0d53a0b0a58a3c6980a7c551357ac71ba3db10 fixes this issue.\ncc @artagnon @nikic Can you double check the implementation of `hash_value(const CmpPredicate&)`?\n"}, {"author": "nikic", "body": "Yeah, the implementation is incorrect for EarlyCSE (which should not take into account poison generating flags). I think we should remove the hash_value implementation and make the caller handler it appropriately."}, {"author": "artagnon", "body": "Not sure I understand. The backtrace points to EarlyCSE.cpp:1558, and the Instruction it crashes on is an `or`. With surrounding context it is:\n\nhttps://github.com/llvm/llvm-project/blob/main/llvm/lib/Transforms/Scalar/EarlyCSE.cpp#L1392-L1560"}, {"author": "artagnon", "body": "So I think this would fix the issue, although I'm not sure I understand the underlying issue:\n\n```diff\ndiff --git a/llvm/lib/Transforms/Scalar/EarlyCSE.cpp b/llvm/lib/Transforms/Scalar/EarlyCSE.cpp\nindex 682c5c3d8c63..8e87f8b9a457 100644\n--- a/llvm/lib/Transforms/Scalar/EarlyCSE.cpp\n+++ b/llvm/lib/Transforms/Scalar/EarlyCSE.cpp\n@@ -290,7 +290,7 @@ static unsigned getHashValueImpl(SimpleValue Val) {\n Pred = CmpInst::getInversePredicate(Pred);\n std::swap(A, B);\n }\n- return hash_combine(Inst->getOpcode(), Pred, X, Y, A, B);\n+ return hash_combine(Inst->getOpcode(), static_cast(Pred), X, Y, A, B);\n }\n\n if (CastInst *CI = dyn_cast(Inst))\n```"}, {"author": "artagnon", "body": "We hash two equal predicates, one with with samesign information (leading to a divergent hash), leading to a divergent hash_combine call, and hence a duplicate insertion into DenseMap:\n\n```cpp\n SelectPatternFlavor SPF;\n Value *Cond, *A, *B;\n if (matchSelectWithOptionalNotCond(Inst, Cond, A, B, SPF)) {\n // Hash min/max (cmp + select) to allow for commuted operands.\n // Min/max may also have non-canonical compare predicate (eg, the compare for\n // smin may use 'sgt' rather than 'slt'), and non-canonical operands in the\n // compare.\n // TODO: We should also detect FP min/max.\n if (SPF == SPF_SMIN || SPF == SPF_SMAX ||\n SPF == SPF_UMIN || SPF == SPF_UMAX) {\n if (A > B)\n std::swap(A, B);\n return hash_combine(Inst->getOpcode(), SPF, A, B);\n }\n\n // Hash general selects to allow matching commuted true/false operands.\n\n // If we do not have a compare as the condition, just hash in the condition.\n CmpPredicate Pred;\n Value *X, *Y;\n if (!match(Cond, m_Cmp(Pred, m_Value(X), m_Value(Y))))\n return hash_combine(Inst->getOpcode(), Cond, A, B);\n\n // Similar to cmp normalization (above) - canonicalize the predicate value:\n // select (icmp Pred, X, Y), A, B --> select (icmp InvPred, X, Y), B, A\n if (CmpInst::getInversePredicate(Pred) < Pred) {\n Pred = CmpInst::getInversePredicate(Pred);\n std::swap(A, B);\n }\n return hash_combine(Inst->getOpcode(), static_cast(Pred), X, Y, A, B); /// fixed\n }\n```\n\nThese are the duplicates:\n\n```llvm\n %cmp2.i = icmp samesign ugt i32 %conv1, 31\n %cmp1.i630 = icmp ugt i32 %conv1, 31\n```"}]}, "verified": true} {"bug_id": "89390", "issue_url": "https://github.com/llvm/llvm-project/issues/89390", "bug_type": "miscompilation", "base_commit": "d674f45d51bffbba474b12e07f7d57a2390d2f31", "knowledge_cutoff": "2024-04-19T14:44:07Z", "lit_test_dir": ["llvm/test/Transforms/VectorCombine"], "hints": {"fix_commit": "4cc9c6d98dfef90d1ffa69977d13ffa2894a10f7", "components": ["VectorCombine"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": [[1405, 1410]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": ["VectorCombine::foldShuffleOfBinops"]}}, "patch": "commit 4cc9c6d98dfef90d1ffa69977d13ffa2894a10f7\nAuthor: Simon Pilgrim \nDate: Mon Apr 22 09:00:20 2024 +0100\n\n [VectorCombine] foldShuffleOfBinops - don't fold shuffle(divrem(x,y),divrem(z,w)) if mask contains poison\n \n Fixes #89390\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\nindex 4918cee1fa82..f23b10540338 100644\n--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n@@ -1405,6 +1405,11 @@ bool VectorCombine::foldShuffleOfBinops(Instruction &I) {\n B0->getOpcode() != B1->getOpcode() || B0->getType() != VecTy)\n return false;\n \n+ // Don't introduce poison into div/rem.\n+ if (any_of(Mask, [](int M) { return M == PoisonMaskElem; }) &&\n+ B0->isIntDivRem())\n+ return false;\n+\n // Try to replace a binop with a shuffle if the shuffle is not costly.\n // The new shuffle will choose from a single, common operand, so it may be\n // cheaper than the existing two-operand shuffle.\n", "tests": [{"file": "llvm/test/Transforms/VectorCombine/X86/shuffle-of-binops.ll", "commands": ["opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=SSE2", "opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=AVX2"], "tests": [{"test_name": "shuf_and_v16i16_yy_expensive_shuf", "test_body": "define <16 x i16> @shuf_and_v16i16_yy_expensive_shuf(<16 x i16> %x, <16 x i16> %y, <16 x i16> %z) {\n %b0 = and <16 x i16> %x, %y\n %b1 = and <16 x i16> %y, %z\n %r = shufflevector <16 x i16> %b0, <16 x i16> %b1, <16 x i32> \n ret <16 x i16> %r\n}\n"}, {"test_name": "shuf_srem_v4i32_poison", "test_body": "define <4 x i32> @shuf_srem_v4i32_poison(<4 x i32> %a0, <4 x i32> %a1) {\n %srem0 = srem <4 x i32> %a1, %a0\n %srem1 = srem <4 x i32> splat (i32 1), %a0\n %r = shufflevector <4 x i32> %srem0, <4 x i32> %srem1, <4 x i32> \n ret <4 x i32> %r\n}\n"}]}], "issue": {"title": "VectorCombine transform leading to srem by poison", "body": "https://alive2.llvm.org/ce/z/HRAkAK\r\n\r\nthis:\r\n```llvm\r\ndefine <4 x i32> @f(<4 x i32> %0, <4 x i32> %1) {\r\n %3 = srem <4 x i32> %1, %0\r\n %4 = srem <4 x i32> , %0\r\n %5 = shufflevector <4 x i32> %3, <4 x i32> %4, <4 x i32> \r\n ret <4 x i32> %5\r\n}\r\n```\r\nis getting vector-combined to this:\r\n```llvm\r\ndefine <4 x i32> @f(<4 x i32> %0, <4 x i32> %1) {\r\n %3 = shufflevector <4 x i32> %1, <4 x i32> , <4 x i32> \r\n %4 = shufflevector <4 x i32> %0, <4 x i32> poison, <4 x i32> \r\n %5 = srem <4 x i32> %3, %4\r\n ret <4 x i32> %5\r\n}\r\n```\r\n\r\nbut that poison in lane 1 is now problematic, this gives us full UB when it hits the srem\r\n\r\ncc @nunoplopes @hatsunespica\r\n", "author": "regehr", "labels": ["miscompilation", "vectorizers"], "comments": [{"author": "RKSimon", "body": "I'll take a look at this"}]}, "verified": true} {"bug_id": "93096", "issue_url": "https://github.com/llvm/llvm-project/issues/93096", "bug_type": "miscompilation", "base_commit": "300e5b91144249a5b142b02d6331fcb33e13cf1d", "knowledge_cutoff": "2024-05-22T21:03:20Z", "lit_test_dir": ["llvm/test/Transforms/SCCP"], "hints": {"fix_commit": "27392a35ef052aa515f38359de89aced11a4363d", "components": ["SCCPSolver"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/SCCPSolver.cpp": [[43, 49], [1297, 1303], [1329, 1336], [1534, 1541], [1818, 1824], [1863, 1869]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/SCCPSolver.cpp": ["SCCPInstVisitor::handleCallResult", "SCCPInstVisitor::handleExtractOfWithOverflow", "SCCPInstVisitor::visitBinaryOperator", "SCCPInstVisitor::visitCastInst", "getConstantRange"]}}, "patch": "commit 27392a35ef052aa515f38359de89aced11a4363d\nAuthor: Nikita Popov \nDate: Thu May 23 15:14:34 2024 +0200\n\n [SCCP] Don't allow undef ranges when performing operations (#93163)\n \n When performing some range operation (e.g. and) on a constant range that\n includes undef, we currently just ignore the undef value, which is\n obviously incorrect. Instead, we can do one of two things:\n * Say that the result range also includes undef.\n * Treat undef as a full range.\n \n This patch goes with the second approach -- I'd expect it to be a bit\n better overall, e.g. it allows preserving the fact that a zext of a\n range with undef isn't a full range.\n \n Fixes https://github.com/llvm/llvm-project/issues/93096.\n\ndiff --git a/llvm/lib/Transforms/Utils/SCCPSolver.cpp b/llvm/lib/Transforms/Utils/SCCPSolver.cpp\nindex ce40e8b31b76..4f36bac11e34 100644\n--- a/llvm/lib/Transforms/Utils/SCCPSolver.cpp\n+++ b/llvm/lib/Transforms/Utils/SCCPSolver.cpp\n@@ -43,7 +43,7 @@ static ValueLatticeElement::MergeOptions getMaxWidenStepsOpts() {\n }\n \n static ConstantRange getConstantRange(const ValueLatticeElement &LV, Type *Ty,\n- bool UndefAllowed = true) {\n+ bool UndefAllowed) {\n assert(Ty->isIntOrIntVectorTy() && \"Should be int or int vector\");\n if (LV.isConstantRange(UndefAllowed))\n return LV.getConstantRange();\n@@ -1297,7 +1297,8 @@ void SCCPInstVisitor::visitCastInst(CastInst &I) {\n \n if (I.getDestTy()->isIntegerTy() && I.getSrcTy()->isIntOrIntVectorTy()) {\n auto &LV = getValueState(&I);\n- ConstantRange OpRange = getConstantRange(OpSt, I.getSrcTy());\n+ ConstantRange OpRange =\n+ getConstantRange(OpSt, I.getSrcTy(), /*UndefAllowed=*/false);\n \n Type *DestTy = I.getDestTy();\n // Vectors where all elements have the same known constant range are treated\n@@ -1329,8 +1330,8 @@ void SCCPInstVisitor::handleExtractOfWithOverflow(ExtractValueInst &EVI,\n return; // Wait to resolve.\n \n Type *Ty = LHS->getType();\n- ConstantRange LR = getConstantRange(L, Ty);\n- ConstantRange RR = getConstantRange(R, Ty);\n+ ConstantRange LR = getConstantRange(L, Ty, /*UndefAllowed=*/false);\n+ ConstantRange RR = getConstantRange(R, Ty, /*UndefAllowed=*/false);\n if (Idx == 0) {\n ConstantRange Res = LR.binaryOp(WO->getBinaryOp(), RR);\n mergeInValue(&EVI, ValueLatticeElement::getRange(Res));\n@@ -1534,8 +1535,10 @@ void SCCPInstVisitor::visitBinaryOperator(Instruction &I) {\n return markOverdefined(&I);\n \n // Try to simplify to a constant range.\n- ConstantRange A = getConstantRange(V1State, I.getType());\n- ConstantRange B = getConstantRange(V2State, I.getType());\n+ ConstantRange A =\n+ getConstantRange(V1State, I.getType(), /*UndefAllowed=*/false);\n+ ConstantRange B =\n+ getConstantRange(V2State, I.getType(), /*UndefAllowed=*/false);\n \n auto *BO = cast(&I);\n ConstantRange R = ConstantRange::getEmpty(I.getType()->getScalarSizeInBits());\n@@ -1818,7 +1821,8 @@ void SCCPInstVisitor::handleCallResult(CallBase &CB) {\n \n // Combine range info for the original value with the new range from the\n // condition.\n- auto CopyOfCR = getConstantRange(CopyOfVal, CopyOf->getType());\n+ auto CopyOfCR = getConstantRange(CopyOfVal, CopyOf->getType(),\n+ /*UndefAllowed=*/true);\n auto NewCR = ImposedCR.intersectWith(CopyOfCR);\n // If the existing information is != x, do not use the information from\n // a chained predicate, as the != x information is more likely to be\n@@ -1863,7 +1867,8 @@ void SCCPInstVisitor::handleCallResult(CallBase &CB) {\n const ValueLatticeElement &State = getValueState(Op);\n if (State.isUnknownOrUndef())\n return;\n- OpRanges.push_back(getConstantRange(State, Op->getType()));\n+ OpRanges.push_back(\n+ getConstantRange(State, Op->getType(), /*UndefAllowed=*/false));\n }\n \n ConstantRange Result =\n", "tests": [{"file": "llvm/test/Transforms/SCCP/ip-add-range-to-call.ll", "commands": ["opt -passes=ipsccp -S %s"], "tests": [{"test_name": "caller5", "test_body": "declare hidden i32 @callee5(i32, i32)\n\ndefine i32 @caller5() {\n %c1 = call i32 @callee5(i32 10, i32 100)\n %c2 = call i32 @callee5(i32 20, i32 200)\n %a = add i32 %c1, %c2\n ret i32 %a\n}\n"}]}, {"file": "llvm/test/Transforms/SCCP/range-with-undef.ll", "commands": ["opt -S -passes=ipsccp < %s"], "tests": [{"test_name": "test_with_overflow", "test_body": "define i9 @test_with_overflow(i1 %cond, i8 %a) {\nentry:\n br i1 %cond, label %if, label %join\n\nif: ; preds = %entry\n %a.ext = zext i8 %a to i16\n br label %join\n\njoin: ; preds = %if, %entry\n %phi = phi i16 [ undef, %entry ], [ %a.ext, %if ]\n %wo = call { i16, i1 } @llvm.uadd.with.overflow.i16(i16 %phi, i16 1)\n %add = extractvalue { i16, i1 } %wo, 0\n %trunc = trunc i16 %add to i9\n ret i9 %trunc\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare { i16, i1 } @llvm.uadd.with.overflow.i16(i16, i16) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "test_binop", "test_body": "define i8 @test_binop(i1 %cond, i8 %a) {\nentry:\n br i1 %cond, label %if, label %join\n\nif: ; preds = %entry\n %a.ext = zext i8 %a to i16\n br label %join\n\njoin: ; preds = %if, %entry\n %phi = phi i16 [ undef, %entry ], [ %a.ext, %if ]\n %and = and i16 %phi, -1\n %trunc = trunc i16 %and to i8\n ret i8 %trunc\n}\n"}, {"test_name": "test_cast", "test_body": "define i8 @test_cast(i1 %cond, i8 %a) {\nentry:\n br i1 %cond, label %if, label %join\n\nif: ; preds = %entry\n %a.ext = zext i8 %a to i16\n br label %join\n\njoin: ; preds = %if, %entry\n %phi = phi i16 [ undef, %entry ], [ %a.ext, %if ]\n %zext = zext i16 %phi to i32\n %trunc = trunc i32 %zext to i8\n ret i8 %trunc\n}\n"}, {"test_name": "test_intrin", "test_body": "define i8 @test_intrin(i1 %cond, i8 %a) {\nentry:\n br i1 %cond, label %if, label %join\n\nif: ; preds = %entry\n %a.ext = zext i8 %a to i16\n br label %join\n\njoin: ; preds = %if, %entry\n %phi = phi i16 [ undef, %entry ], [ %a.ext, %if ]\n %umax = call i16 @llvm.umax.i16(i16 %phi, i16 42)\n %trunc = trunc i16 %umax to i8\n ret i8 %trunc\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i16 @llvm.umax.i16(i16, i16) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "nuw/nsw should not be added to trunc instruction when there are undef bits in the source value", "body": "Run ipsccp on the code\r\n```\r\n; opt -passes=ipsccp -S test.ll\r\n\r\ndefine void @foo(i1 %cond1, i32 %p1) {\r\nentry:\r\n br label %loop\r\n\r\nloop:\r\n %phi1 = phi i64 [undef, %entry], [%phi2, %latch]\r\n br i1 %cond1, label %then2, label %else2\r\n\r\nthen2:\r\n %and = and i64 %phi1, -1095216660481 ; 0xFFFFFF00FFFFFFFF\r\n br label %latch\r\n\r\nelse2:\r\n %ext = zext i32 %p1 to i64 \r\n %or = or i64 %ext, 4294967296 ; 0x100000000\r\n br label %latch\r\n\r\nlatch:\r\n %phi2 = phi i64 [%and, %then2], [%or, %else2]\r\n %shr = lshr i64 %phi2, 32\r\n %res = trunc i64 %shr to i8\r\n %cond2 = call i1 @bar(i8 %res)\r\n br i1 %cond2, label %loop, label %exit\r\n\r\nexit:\r\n ret void\r\n}\r\n\r\ndeclare i1 @bar(i8)\r\n```\r\nLLVM adds nuw/nsw flags to the trunc instruction\r\n```\r\n%res = trunc nuw nsw i64 %shr to i8\r\n```\r\nIt means all high bits of %shr are 0. \r\nBut %phi1 may be undef, bits [32..39] of %and are cleared, bits[40..63] can still be undef, and later it is moved to bits[8..31] of %shr, so not all high bits of %shr are 0.\r\n", "author": "weiguozhi", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "nikic", "body": "We do exclude ranges with undef here: https://github.com/llvm/llvm-project/blob/3061fed2f05cde574c84a26c702b1e7932fa4d7e/llvm/lib/Transforms/Utils/SCCPSolver.cpp#L118-L119 Maybe the undef flag doesn't get propagated properly somewhere..."}, {"author": "nikic", "body": "Simplified test case:\r\n```llvm\r\ndefine i8 @src(i1 %cond, i8 %a) {\r\nentry: \r\n br i1 %cond, label %if, label %join\r\n \r\nif:\r\n %a.ext = zext i8 %a to i16 \r\n br label %join \r\n\r\njoin:\r\n %phi = phi i16 [ undef, %entry ], [ %a.ext, %if ] \r\n %and = and i16 %phi, u0x0000ffff \r\n %trunc = trunc i16 %and to i8\r\n ret i8 %trunc\r\n} \r\n```\r\nThis is indeed an issue with the propagation of the undef flag."}]}, "verified": true} {"bug_id": "109835", "issue_url": "https://github.com/llvm/llvm-project/issues/109835", "bug_type": "crash", "base_commit": "79ecb814d0c929a66ad92c7b3e91191f01247ac1", "knowledge_cutoff": "2024-09-24T17:26:42Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "cc01112660499a4db74bc3ee24b6f60b03d88dbd", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[11786, 11792], [11794, 11799]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::getTreeCost"]}}, "patch": "commit cc01112660499a4db74bc3ee24b6f60b03d88dbd\nAuthor: Han-Kuan Chen \nDate: Tue Oct 1 19:15:58 2024 +0800\n\n [SLP][REVEC] getTypeSizeInBits should apply to scalar type instead of FixedVectorType. (#110610)\n \n reference: https://github.com/llvm/llvm-project/issues/109835\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 68bf5c52814f..54a1b99606bc 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -11786,7 +11786,7 @@ InstructionCost BoUpSLP::getTreeCost(ArrayRef VectorizedVals) {\n auto BWIt = MinBWs.find(&Root);\n if (BWIt != MinBWs.end()) {\n Type *DstTy = Root.Scalars.front()->getType();\n- unsigned OriginalSz = DL->getTypeSizeInBits(DstTy);\n+ unsigned OriginalSz = DL->getTypeSizeInBits(DstTy->getScalarType());\n unsigned SrcSz =\n ReductionBitWidth == 0 ? BWIt->second.first : ReductionBitWidth;\n if (OriginalSz != SrcSz) {\n@@ -11794,6 +11794,10 @@ InstructionCost BoUpSLP::getTreeCost(ArrayRef VectorizedVals) {\n if (OriginalSz > SrcSz)\n Opcode = BWIt->second.second ? Instruction::SExt : Instruction::ZExt;\n Type *SrcTy = IntegerType::get(DstTy->getContext(), SrcSz);\n+ if (auto *VecTy = dyn_cast(DstTy)) {\n+ assert(SLPReVec && \"Only supported by REVEC.\");\n+ SrcTy = getWidenedType(SrcTy, VecTy->getNumElements());\n+ }\n Cost += TTI->getCastInstrCost(Opcode, DstTy, SrcTy,\n TTI::CastContextHint::None,\n TTI::TCK_RecipThroughput);\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/SystemZ/revec-fix-109835.ll", "commands": ["opt -mtriple=s390x-unknown-linux -mcpu=z16 -passes=slp-vectorizer -S -slp-revec %s"], "tests": [{"test_name": "func_1", "test_body": "@g_3_1_0 = external dso_local global i32, align 4\n@g_7 = external dso_local global [5 x i32], align 4\n\ndefine void @func_1() {\n %.pre = load i32, ptr @g_7, align 4\n %1 = load <4 x i32>, ptr getelementptr inbounds nuw (i8, ptr @g_7, i64 4), align 4\n br label %.loopexit\n\n.loopexit: ; preds = %.loopexit, %0\n %2 = phi i32 [ %op.rdx15, %.loopexit ], [ %.pre, %0 ]\n %3 = phi <4 x i32> [ %71, %.loopexit ], [ %1, %0 ]\n %4 = load volatile i32, ptr @g_3_1_0, align 4\n %5 = load volatile i32, ptr @g_3_1_0, align 4\n %6 = load volatile i32, ptr @g_3_1_0, align 4\n %7 = load volatile i32, ptr @g_3_1_0, align 4\n %8 = load volatile i32, ptr @g_3_1_0, align 4\n %9 = load volatile i32, ptr @g_3_1_0, align 4\n %10 = load volatile i32, ptr @g_3_1_0, align 4\n %11 = load volatile i32, ptr @g_3_1_0, align 4\n %12 = load volatile i32, ptr @g_3_1_0, align 4\n %13 = load volatile i32, ptr @g_3_1_0, align 4\n %14 = load volatile i32, ptr @g_3_1_0, align 4\n %15 = load volatile i32, ptr @g_3_1_0, align 4\n %16 = load volatile i32, ptr @g_3_1_0, align 4\n %17 = load volatile i32, ptr @g_3_1_0, align 4\n %18 = load volatile i32, ptr @g_3_1_0, align 4\n %19 = load volatile i32, ptr @g_3_1_0, align 4\n %20 = insertelement <4 x i32> poison, i32 %4, i32 0\n %21 = insertelement <4 x i32> %20, i32 %9, i32 1\n %22 = insertelement <4 x i32> %21, i32 %14, i32 2\n %23 = insertelement <4 x i32> %22, i32 %19, i32 3\n %24 = icmp eq <4 x i32> %23, zeroinitializer\n %25 = load volatile i32, ptr @g_3_1_0, align 4\n %26 = load volatile i32, ptr @g_3_1_0, align 4\n %27 = load volatile i32, ptr @g_3_1_0, align 4\n %28 = load volatile i32, ptr @g_3_1_0, align 4\n %29 = load volatile i32, ptr @g_3_1_0, align 4\n %.not2.410 = icmp eq i32 %29, 0\n %30 = zext i1 %.not2.410 to i32\n %31 = zext <4 x i1> %24 to <4 x i32>\n %32 = call i32 @llvm.vector.reduce.xor.v4i32(<4 x i32> %31)\n %op.rdx = xor i32 %32, %30\n %op.rdx15 = xor i32 %op.rdx, %2\n %33 = load volatile i32, ptr @g_3_1_0, align 4\n %34 = load volatile i32, ptr @g_3_1_0, align 4\n %35 = load volatile i32, ptr @g_3_1_0, align 4\n %36 = load volatile i32, ptr @g_3_1_0, align 4\n %37 = insertelement <4 x i32> poison, i32 %5, i32 0\n %38 = insertelement <4 x i32> %37, i32 %6, i32 1\n %39 = insertelement <4 x i32> %38, i32 %7, i32 2\n %40 = insertelement <4 x i32> %39, i32 %8, i32 3\n %41 = icmp eq <4 x i32> %40, zeroinitializer\n %42 = zext <4 x i1> %41 to <4 x i32>\n %43 = xor <4 x i32> %3, %42\n %44 = insertelement <4 x i32> poison, i32 %10, i32 0\n %45 = insertelement <4 x i32> %44, i32 %11, i32 1\n %46 = insertelement <4 x i32> %45, i32 %12, i32 2\n %47 = insertelement <4 x i32> %46, i32 %13, i32 3\n %48 = icmp eq <4 x i32> %47, zeroinitializer\n %49 = zext <4 x i1> %48 to <4 x i32>\n %50 = xor <4 x i32> %43, %49\n %51 = insertelement <4 x i32> poison, i32 %15, i32 0\n %52 = insertelement <4 x i32> %51, i32 %16, i32 1\n %53 = insertelement <4 x i32> %52, i32 %17, i32 2\n %54 = insertelement <4 x i32> %53, i32 %18, i32 3\n %55 = icmp eq <4 x i32> %54, zeroinitializer\n %56 = zext <4 x i1> %55 to <4 x i32>\n %57 = xor <4 x i32> %50, %56\n %58 = insertelement <4 x i32> poison, i32 %25, i32 0\n %59 = insertelement <4 x i32> %58, i32 %26, i32 1\n %60 = insertelement <4 x i32> %59, i32 %27, i32 2\n %61 = insertelement <4 x i32> %60, i32 %28, i32 3\n %62 = icmp eq <4 x i32> %61, zeroinitializer\n %63 = zext <4 x i1> %62 to <4 x i32>\n %64 = xor <4 x i32> %57, %63\n %65 = insertelement <4 x i32> poison, i32 %33, i32 0\n %66 = insertelement <4 x i32> %65, i32 %34, i32 1\n %67 = insertelement <4 x i32> %66, i32 %35, i32 2\n %68 = insertelement <4 x i32> %67, i32 %36, i32 3\n %69 = icmp eq <4 x i32> %68, zeroinitializer\n %70 = zext <4 x i1> %69 to <4 x i32>\n %71 = xor <4 x i32> %64, %70\n br label %.loopexit\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.vector.reduce.xor.v4i32(<4 x i32>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[SLP vectorizer] Assertion fails with slp-revec", "body": "[tc_slp_revec.tar.gz](https://github.com/user-attachments/files/17118250/tc_slp_revec.tar.gz)\r\n\r\nopt -mtriple=systemz-unknown -mcpu=z16 -O3 -o out.ll tc_crash22_aftercreduce.ll -slp-revec\r\n\r\nType.cpp:691: static llvm::FixedVectorType* llvm::FixedVectorType::get(llvm::Type*, unsigned int): Assertion `isValidElementType(\r\nElementType) && \"Element type of a VectorType must \" \"be an integer, floating point, or \" \"pointer type.\"' failed.\r\n\r\n#9 0x000002aa076e014e llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::finalize\r\n\r\n", "author": "JonPsson1", "labels": ["llvm:SLPVectorizer", "crash"], "comments": [{"author": "HanKuanChen", "body": "The fix is landing."}, {"author": "JonPsson1", "body": "Still failing, but this time I see another assertion:\r\n\r\n[tc_slp_revec.tar.gz](https://github.com/user-attachments/files/17192435/tc_slp_revec.tar.gz)\r\n\r\nopt -mtriple=s390x-unknown-linux -mcpu=z16 -O3 tc_crash38_aftercreduce.ll -o /dev/null -slp-revec\r\n\r\nAssertion `!Dst->isVectorTy()' failed.\r\n14 0x00000000030504c2 llvm::SLPVectorizerPass::vectorizeHorReduction\r\n\r\n"}, {"author": "HanKuanChen", "body": "The fix is landing."}]}, "verified": true} {"bug_id": "89500", "issue_url": "https://github.com/llvm/llvm-project/issues/89500", "bug_type": "miscompilation", "base_commit": "a1b1c4a6d1d52916c5d885170a5f54632d579cdc", "knowledge_cutoff": "2024-04-20T12:43:21Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "7339f7ba3053db7595ece1ca5f49bd2e4c3c8305", "components": ["InstCombine"], "bug_location_lineno": {"llvm/include/llvm/IR/PatternMatch.h": [[350, 357], [374, 380], [389, 400], [484, 489], [2596, 2601]], "llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[1722, 1732]]}, "bug_location_funcname": {"llvm/include/llvm/IR/PatternMatch.h": ["isValue", "m_AllOnes", "m_Not", "m_c_SMin", "match"], "llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["foldSelectICmpEq"]}}, "patch": "commit 7339f7ba3053db7595ece1ca5f49bd2e4c3c8305\nAuthor: Nikita Popov \nDate: Wed Apr 24 10:57:17 2024 +0900\n\n [InstCombine] Fix poison propagation in select of bitwise fold (#89701)\n \n We're replacing the select with the false value here, but it may be more\n poisonous if m_Not contains poison elements. Fix this by introducing a\n m_NotForbidPoison matcher and using it here.\n \n Fixes https://github.com/llvm/llvm-project/issues/89500.\n\ndiff --git a/llvm/include/llvm/IR/PatternMatch.h b/llvm/include/llvm/IR/PatternMatch.h\nindex 1fee1901fabb..0b13b4aad9c3 100644\n--- a/llvm/include/llvm/IR/PatternMatch.h\n+++ b/llvm/include/llvm/IR/PatternMatch.h\n@@ -350,8 +350,9 @@ template inline constantint_match m_ConstantInt() {\n \n /// This helper class is used to match constant scalars, vector splats,\n /// and fixed width vectors that satisfy a specified predicate.\n-/// For fixed width vector constants, poison elements are ignored.\n-template \n+/// For fixed width vector constants, poison elements are ignored if AllowPoison\n+/// is true.\n+template \n struct cstval_pred_ty : public Predicate {\n template bool match(ITy *V) {\n if (const auto *CV = dyn_cast(V))\n@@ -374,7 +375,7 @@ struct cstval_pred_ty : public Predicate {\n Constant *Elt = C->getAggregateElement(i);\n if (!Elt)\n return false;\n- if (isa(Elt))\n+ if (AllowPoison && isa(Elt))\n continue;\n auto *CV = dyn_cast(Elt);\n if (!CV || !this->isValue(CV->getValue()))\n@@ -389,12 +390,13 @@ struct cstval_pred_ty : public Predicate {\n };\n \n /// specialization of cstval_pred_ty for ConstantInt\n-template \n-using cst_pred_ty = cstval_pred_ty;\n+template \n+using cst_pred_ty = cstval_pred_ty;\n \n /// specialization of cstval_pred_ty for ConstantFP\n template \n-using cstfp_pred_ty = cstval_pred_ty;\n+using cstfp_pred_ty = cstval_pred_ty;\n \n /// This helper class is used to match scalar and vector constants that\n /// satisfy a specified predicate, and bind them to an APInt.\n@@ -484,6 +486,10 @@ inline cst_pred_ty m_AllOnes() {\n return cst_pred_ty();\n }\n \n+inline cst_pred_ty m_AllOnesForbidPoison() {\n+ return cst_pred_ty();\n+}\n+\n struct is_maxsignedvalue {\n bool isValue(const APInt &C) { return C.isMaxSignedValue(); }\n };\n@@ -2596,6 +2602,13 @@ m_Not(const ValTy &V) {\n return m_c_Xor(m_AllOnes(), V);\n }\n \n+template \n+inline BinaryOp_match, ValTy, Instruction::Xor,\n+ true>\n+m_NotForbidPoison(const ValTy &V) {\n+ return m_c_Xor(m_AllOnesForbidPoison(), V);\n+}\n+\n /// Matches an SMin with LHS and RHS in either order.\n template \n inline MaxMin_match\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex 73600206a55c..117eb7a1dcc9 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -1722,11 +1722,11 @@ static Instruction *foldSelectICmpEq(SelectInst &SI, ICmpInst *ICI,\n return match(CmpRHS, m_Zero()) && match(FalseVal, matchInner);\n \n if (NotMask == NotInner) {\n- return match(FalseVal,\n- m_c_BinOp(OuterOpc, m_Not(matchInner), m_Specific(CmpRHS)));\n+ return match(FalseVal, m_c_BinOp(OuterOpc, m_NotForbidPoison(matchInner),\n+ m_Specific(CmpRHS)));\n } else if (NotMask == NotRHS) {\n- return match(FalseVal,\n- m_c_BinOp(OuterOpc, matchInner, m_Not(m_Specific(CmpRHS))));\n+ return match(FalseVal, m_c_BinOp(OuterOpc, matchInner,\n+ m_NotForbidPoison(m_Specific(CmpRHS))));\n } else {\n return match(FalseVal,\n m_c_BinOp(OuterOpc, matchInner, m_Specific(CmpRHS)));\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/select.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "src_and_eq_C_xor_OrAndNotC_vec_poison", "test_body": "target datalayout = \"e-p:64:64-p1:16:16-p2:32:32:32-p3:64:64:64\"\n\ndefine <2 x i32> @src_and_eq_C_xor_OrAndNotC_vec_poison(<2 x i32> %0, <2 x i32> %1, <2 x i32> %2) {\nentry:\n %and = and <2 x i32> %1, %0\n %cmp = icmp eq <2 x i32> %and, %2\n %xor = xor <2 x i32> %1, %0\n %or = or <2 x i32> %1, %0\n %not = xor <2 x i32> %2, \n %and1 = and <2 x i32> %or, %not\n %cond = select <2 x i1> %cmp, <2 x i32> %xor, <2 x i32> %and1\n ret <2 x i32> %cond\n}\n"}]}], "issue": {"title": "[InstCombine] Miscompilation with poison vectors in `src_and_eq_C_xor_OrAndNotC` ", "body": "Alive2: https://alive2.llvm.org/ce/z/yB3oAx\r\n```ll\r\ndefine <2 x i32> @src_and_eq_C_xor_OrAndNotC(<2 x i32> %0, <2 x i32> %1, <2 x i32> %2) {\r\nentry:\r\n %and = and <2 x i32> %1, %0\r\n %cmp = icmp eq <2 x i32> %and, %2\r\n %xor = xor <2 x i32> %1, %0\r\n %or = or <2 x i32> %1, %0\r\n %not = xor <2 x i32> %2, \r\n %and1 = and <2 x i32> %or, %not\r\n %cond = select <2 x i1> %cmp, <2 x i32> %xor, <2 x i32> %and1\r\n ret <2 x i32> %cond\r\n}\r\n```\r\n```ll\r\n\r\n----------------------------------------\r\ndefine <2 x i32> @src_and_eq_C_xor_OrAndNotC(<2 x i32> %#0, <2 x i32> %#1, <2 x i32> %#2) {\r\nentry:\r\n %and = and <2 x i32> %#1, %#0\r\n %cmp = icmp eq <2 x i32> %and, %#2\r\n %xor = xor <2 x i32> %#1, %#0\r\n %or = or <2 x i32> %#1, %#0\r\n %not = xor <2 x i32> %#2, { 4294967295, poison }\r\n %and1 = and <2 x i32> %or, %not\r\n %cond = select <2 x i1> %cmp, <2 x i32> %xor, <2 x i32> %and1\r\n ret <2 x i32> %cond\r\n}\r\n=>\r\ndefine <2 x i32> @src_and_eq_C_xor_OrAndNotC(<2 x i32> %#0, <2 x i32> %#1, <2 x i32> %#2) nofree willreturn memory(none) {\r\nentry:\r\n %or = or <2 x i32> %#1, %#0\r\n %not = xor <2 x i32> %#2, { 4294967295, poison }\r\n %and1 = and <2 x i32> %or, %not\r\n ret <2 x i32> %and1\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\n<2 x i32> %#0 = < poison, #x00000000 (0) >\r\n<2 x i32> %#1 = < poison, #x00000000 (0) >\r\n<2 x i32> %#2 = < poison, #x00000000 (0) >\r\n\r\nSource:\r\n<2 x i32> %and = < poison, #x00000000 (0) >\r\n<2 x i1> %cmp = < poison, #x1 (1) >\r\n<2 x i32> %xor = < poison, #x00000000 (0) >\r\n<2 x i32> %or = < poison, #x00000000 (0) >\r\n<2 x i32> %not = < poison, poison >\r\n<2 x i32> %and1 = < poison, poison >\r\n<2 x i32> %cond = < poison, #x00000000 (0) >\r\n\r\nTarget:\r\n<2 x i32> %or = < poison, #x00000000 (0) >\r\n<2 x i32> %not = < poison, poison >\r\n<2 x i32> %and1 = < poison, poison >\r\nSource value: < poison, #x00000000 (0) >\r\nTarget value: < poison, poison >\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```\r\n\r\nFound with https://github.com/dtcxzyw/llvm-tools.\r\ncc @nikic", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "nikic", "body": "Looks like a recent issue introduced with https://github.com/llvm/llvm-project/pull/73362."}, {"author": "ParkHanbum", "body": "I'll not forget to create a testcase with poison.\r\n"}]}, "verified": true} {"bug_id": "100591", "issue_url": "https://github.com/llvm/llvm-project/issues/100591", "bug_type": "crash", "base_commit": "71801707e33c235656b172fa7dfb8662473a95c2", "knowledge_cutoff": "2024-07-25T16:31:43Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "e9e3a183d6d8d8729223b7131aa57fd9f02f0440", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6730, 6738], [6765, 6770], [6789, 6796]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::collectValuesToIgnore"]}}, "patch": "commit e9e3a183d6d8d8729223b7131aa57fd9f02f0440\nAuthor: Florian Hahn \nDate: Sun Aug 18 12:51:17 2024 +0100\n\n [LV] Don't cost branches and conditions to empty blocks.\n \n Update the legacy cost model skip branches with successors blocks\n that are empty or only contain dead instructions, together with their\n conditions. Such branches and conditions won't result in any\n generated code and will be cleaned up by VPlan transforms.\n \n This fixes a difference between the legacy and VPlan-based cost model.\n \n When running LV in its usual pipeline position, such dead blocks should\n already have been cleaned up, but they might be generated manually or by\n fuzzers.\n \n Fixes https://github.com/llvm/llvm-project/issues/100591.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 27e99694aafd..55c0ba3dd8f9 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6730,9 +6730,12 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n return RequiresScalarEpilogue &&\n !TheLoop->contains(cast(U)->getParent());\n };\n+\n+ LoopBlocksDFS DFS(TheLoop);\n+ DFS.perform(LI);\n MapVector> DeadInvariantStoreOps;\n- for (BasicBlock *BB : TheLoop->blocks())\n- for (Instruction &I : *BB) {\n+ for (BasicBlock *BB : reverse(make_range(DFS.beginRPO(), DFS.endRPO())))\n+ for (Instruction &I : reverse(*BB)) {\n // Find all stores to invariant variables. Since they are going to sink\n // outside the loop we do not need calculate cost for them.\n StoreInst *SI;\n@@ -6765,6 +6768,13 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n Value *PointerOp = getLoadStorePointerOperand(&I);\n DeadInterleavePointerOps.push_back(PointerOp);\n }\n+\n+ // Queue branches for analysis. They are dead, if their successors only\n+ // contain dead instructions.\n+ if (auto *Br = dyn_cast(&I)) {\n+ if (Br->isConditional())\n+ DeadOps.push_back(&I);\n+ }\n }\n \n // Mark ops feeding interleave group members as free, if they are only used\n@@ -6789,8 +6799,36 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {\n // Mark ops that would be trivially dead and are only used by ignored\n // instructions as free.\n BasicBlock *Header = TheLoop->getHeader();\n+\n+ // Returns true if the block contains only dead instructions. Such blocks will\n+ // be removed by VPlan-to-VPlan transforms and won't be considered by the\n+ // VPlan-based cost model, so skip them in the legacy cost-model as well.\n+ auto IsEmptyBlock = [this](BasicBlock *BB) {\n+ return all_of(*BB, [this](Instruction &I) {\n+ return ValuesToIgnore.contains(&I) || VecValuesToIgnore.contains(&I) ||\n+ (isa(&I) && !cast(&I)->isConditional());\n+ });\n+ };\n for (unsigned I = 0; I != DeadOps.size(); ++I) {\n auto *Op = dyn_cast(DeadOps[I]);\n+\n+ // Check if the branch should be considered dead.\n+ if (auto *Br = dyn_cast_or_null(Op)) {\n+ BasicBlock *ThenBB = Br->getSuccessor(0);\n+ BasicBlock *ElseBB = Br->getSuccessor(1);\n+ bool ThenEmpty = IsEmptyBlock(ThenBB);\n+ bool ElseEmpty = IsEmptyBlock(ElseBB);\n+ if ((ThenEmpty && ElseEmpty) ||\n+ (ThenEmpty && ThenBB->getSingleSuccessor() == ElseBB &&\n+ ElseBB->phis().empty()) ||\n+ (ElseEmpty && ElseBB->getSingleSuccessor() == ThenBB &&\n+ ThenBB->phis().empty())) {\n+ VecValuesToIgnore.insert(Br);\n+ DeadOps.push_back(Br->getCondition());\n+ }\n+ continue;\n+ }\n+\n // Skip any op that shouldn't be considered dead.\n if (!Op || !TheLoop->contains(Op) ||\n (isa(Op) && Op->getParent() == Header) ||\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/blocks-with-dead-instructions.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "block_with_dead_inst_1", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @block_with_dead_inst_1(ptr %src, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n %dead.gep = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %0 = icmp eq i64 %iv.next, %N\n br i1 %0, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "multiple_blocks_with_dead_insts_4", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @multiple_blocks_with_dead_insts_4(ptr %src, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %loop.header\n br label %then.1\n\nthen.1: ; preds = %then\n %dead.gep.1 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nelse: ; preds = %loop.header\n %dead.gep.2 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else, %then.1\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "empty_block_with_phi_1", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @empty_block_with_phi_1(ptr %src, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i32 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i32 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %p = phi i16 [ %l, %loop.header ], [ 99, %then ]\n store i16 %p, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 1\n %0 = icmp eq i64 %iv.next, %N\n br i1 %0, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "multiple_blocks_with_dead_insts_3", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @multiple_blocks_with_dead_insts_3(ptr %src) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 0, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %loop.header\n %dead.gep.1 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nelse: ; preds = %loop.header\n %dead.gep.2 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else, %then\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "multiple_blocks_with_dead_inst_multiple_successors_5", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @multiple_blocks_with_dead_inst_multiple_successors_5(ptr %src) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %loop.header\n br label %then.1\n\nthen.1: ; preds = %then\n %dead.gep.1 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nelse: ; preds = %loop.header\n br label %else.2\n\nelse.2: ; preds = %else\n %dead.gep.2 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else.2, %then.1\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "multiple_blocks_with_dead_inst_multiple_successors_6", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @multiple_blocks_with_dead_inst_multiple_successors_6(ptr %src, i1 %ic, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %then, label %else\n\nthen: ; preds = %loop.header\n br i1 %ic, label %then.1, label %else\n\nthen.1: ; preds = %then\n %dead.gep.1 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nelse: ; preds = %then, %loop.header\n br label %else.2\n\nelse.2: ; preds = %else\n %dead.gep.2 = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else.2, %then.1\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "empty_block_with_phi_2", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @empty_block_with_phi_2(ptr %src, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i32 [ 1, %entry ], [ %xor, %loop.latch ]\n %xor = xor i32 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %loop.latch, label %else\n\nelse: ; preds = %loop.header\n br label %loop.latch\n\nloop.latch: ; preds = %else, %loop.header\n %p = phi i16 [ %l, %loop.header ], [ 99, %else ]\n store i16 %p, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 1\n %0 = icmp eq i64 %iv.next, %N\n br i1 %0, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "block_with_dead_inst_2", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @block_with_dead_inst_2(ptr %src) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %xor1315 = phi i16 [ 0, %entry ], [ %xor, %loop.latch ]\n %xor = xor i16 0, 0\n %gep = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep, align 2\n %c = icmp eq i16 %l, 0\n br i1 %c, label %loop.latch, label %else\n\nelse: ; preds = %loop.header\n %dead.gep = getelementptr i64, ptr %src, i64 %iv\n br label %loop.latch\n\nloop.latch: ; preds = %else, %loop.header\n store i16 %xor, ptr %gep, align 2\n %iv.next = add nsw i64 %iv, 3\n %ec = icmp eq i64 %iv.next, 1000\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}]}], "issue": {"title": "[VPlan] Assertion `VF.Width == Width && \"VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Tested using 51d4980a133db12888207698e39c469cb7055cac\r\n\r\nTestcase:\r\n```llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @f() #0 {\r\nentry:\r\n br label %for.body\r\n\r\nfor.cond.cleanup: ; preds = %cond.end7\r\n ret void\r\n\r\nfor.body: ; preds = %cond.end7, %entry\r\n %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %cond.end7 ]\r\n %xor1315 = phi i32 [ 0, %entry ], [ %xor, %cond.end7 ]\r\n %xor = xor i32 0, 0\r\n %arrayidx = getelementptr [1 x i16], ptr null, i64 0, i64 %indvars.iv\r\n %0 = load i16, ptr %arrayidx, align 2\r\n %tobool1.not = icmp eq i16 %0, 0\r\n br i1 %tobool1.not, label %cond.false, label %cond.end7\r\n\r\ncond.false: ; preds = %for.body\r\n %arrayidx3 = getelementptr [1 x i64], ptr null, i64 0, i64 %indvars.iv\r\n br label %cond.end7\r\n\r\ncond.end7: ; preds = %cond.false, %for.body\r\n %indvars.iv.next = add nsw i64 %indvars.iv, 3\r\n %1 = icmp eq i64 %indvars.iv.next, 0\r\n br i1 %1, label %for.cond.cleanup, label %for.body\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nBacktrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10123: bool llvm::LoopVectorizePass::processLoop(llvm::Loop*): Assertion `VF.Width == Width && \"VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"f\"\r\n #0 0x00005e0a0dd8f3b0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f1f3b0)\r\n #1 0x00005e0a0dd8c7cf llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f1c7cf)\r\n #2 0x00005e0a0dd8c925 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x000073a243a42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x000073a243a969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x000073a243a969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x000073a243a969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x000073a243a42476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x000073a243a287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x000073a243a2871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x000073a243a39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00005e0a0cdaadd8 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f3add8)\r\n#12 0x00005e0a0cdad991 llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f3d991)\r\n#13 0x00005e0a0cdae110 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f3e110)\r\n#14 0x00005e0a0bc1fda6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdafda6)\r\n#15 0x00005e0a0dba750e llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d3750e)\r\n#16 0x00005e0a0bc22f36 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdb2f36)\r\n#17 0x00005e0a0dba629b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d3629b)\r\n#18 0x00005e0a0bc1f286 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdaf286)\r\n#19 0x00005e0a0dba43ed llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d343ed)\r\n#20 0x00005e0a0b4686b6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5f86b6)\r\n#21 0x00005e0a0b45a2f1 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5ea2f1)\r\n#22 0x000073a243a29d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#23 0x000073a243a29e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#24 0x000073a243a29e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#25 0x00005e0a0b450185 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5e0185)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\n\r\nGodbolt: https://godbolt.org/z/TnYreKbqn\r\n\r\n@fhahn \r\n\r\nFound via fuzzer.", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "Another case with dead instructions that get simplified by VPlan-DCE :/"}, {"author": "patrick-rivos", "body": "Just to add a bit of context here: I'm hitting this assertion >300 times a day so I've allowlisted it in the fuzzer I run. There are likely other interesting cases where this assertion triggers but I don't have the bandwidth to deduplicate this failure to pull out those cases. :-("}, {"author": "preames", "body": "> Just to add a bit of context here: I'm hitting this assertion >300 times a day so I've allowlisted it in the fuzzer I run. There are likely other interesting cases where this assertion triggers but I don't have the bandwidth to deduplicate this failure to pull out those cases. :-(\r\n\r\n@fhahn This assertion is triggering in a large number of fuzzer runs, and we have some reason to believe there are other failures be masked by this one. (We have at least one instance of this found so far by accident.) This needs to be either fixed or reverted so that tree is green. "}, {"author": "fhahn", "body": "Will look into a fix now that I'm back. The crash should not reproduce with `opt -passes='instcombine,simplifycfg,loop-vectorize'` for all instances of the same underlying issue."}, {"author": "nikic", "body": "@fhahn I feel like at this point this assertion is doing a lot more harm than good. It's very fragile, and doesn't seem to help in finding any genuine bugs, just incidental differences in simplification power."}, {"author": "artagnon", "body": "I was also looking into this bug briefly, but couldn't quite figure out why we didn't choose to rip out the legacy costing."}, {"author": "sjoerdmeijer", "body": "If e9e3a183d6d8d8729223b7131aa57fd9f02f0440 was supposed to fix this, then I can confirm I am still seeing this assert. \r\n\r\nIf this is indeed something else or new, then I agree with @nikic that this assert is too fragile."}, {"author": "fhahn", "body": "The assertion originally surfaced real differences that needed fixing; the fuzzer-generated issues over the last weeks mostly were due to unsimplified input, which is very difficult to completely match without actually iteratively simplifying. Seeing the recent https://github.com/llvm/llvm-project/issues/104714 with a C reproducer it was a bit surprising that dead loads (and subsequent dead blocks/conditions) survive until LoopVectorize in the default pipeline.\r\n\r\nThere will be a number of code changes over the next few weeks/months which may introduce unintended divergences the assert is supposed to catch. For that reason it would be beneficial to keep the assert.\r\n\r\nIdeally we would only check if no extra simplifications have been applied on the selected VPlan, which should be possible to check once #103033 lands.\r\n\r\nWe could also guard it with an option, similar to InstCombine's fix point iteration assertion, which can also trigger asserts by default on valid IR input.\r\n\r\n> If [e9e3a18](https://github.com/llvm/llvm-project/commit/e9e3a183d6d8d8729223b7131aa57fd9f02f0440) was supposed to fix this, then I can confirm I am still seeing this assert.\r\n> \r\n\r\nCould you share the reproducer somewhere? Would be good to check if this is due to the simplification issue mentioned above or a genuine divergence"}, {"author": "sjoerdmeijer", "body": "> Could you share the reproducer somewhere? Would be good to check if this is due to the simplification issue mentioned above or a genuine divergence\r\n\r\nThis runs in the same assert with `opt -S -passes=loop-vectorize`: \r\n\r\n```\r\ntarget datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-Fn32\"\r\ntarget triple = \"aarch64-unknown-linux-gnu\"\r\n\r\ndefine void @foo() #0 {\r\nentry:\r\n br label %for.body315.us.us\r\n\r\nfor.body315.us.us: ; preds = %cond.end388.us.us, %entry\r\n %indvars.iv10421 = phi i64 [ %indvars.iv.next1043, %cond.end388.us.us ], [ 0, %entry ]\r\n br i1 false, label %cond.end388.us.us, label %cond.true380.us.us\r\n\r\ncond.true380.us.us: ; preds = %for.body315.us.us\r\n %0 = load i32, ptr null, align 4\r\n br label %cond.end388.us.us\r\n\r\ncond.end388.us.us: ; preds = %cond.true380.us.us, %for.body315.us.us\r\n %indvars.iv.next1043 = add i64 %indvars.iv10421, 1\r\n %cmp313.us.us = icmp slt i64 %indvars.iv10421, 1\r\n br i1 %cmp313.us.us, label %for.body315.us.us, label %for.cond311.for.cond.cleanup314_crit_edge.loopexit, !llvm.loop !0\r\n\r\nfor.cond311.for.cond.cleanup314_crit_edge.loopexit: ; preds = %cond.end388.us.us\r\n ret void\r\n}\r\n\r\nattributes #0 = { \"target-cpu\"=\"neoverse-v2\" }\r\n\r\n!0 = distinct !{!0, !1, !2, !3}\r\n!1 = !{!\"llvm.loop.mustprogress\"}\r\n!2 = !{!\"llvm.loop.vectorize.predicate.enable\", i1 true}\r\n!3 = !{!\"llvm.loop.vectorize.enable\", i1 true}\r\n```"}]}, "verified": true} {"bug_id": "98675", "issue_url": "https://github.com/llvm/llvm-project/issues/98675", "bug_type": "crash", "base_commit": "22b7b84860d39da71964c9b329937f2ee1d875ba", "knowledge_cutoff": "2024-07-12T18:26:47Z", "lit_test_dir": ["llvm/test/Transforms/MemCpyOpt"], "hints": {"fix_commit": "99685a54d1322694847f59386a548b3696a43fc3", "components": ["MemCpyOptimizer"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp": [[1188, 1196]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp": ["MemCpyOptPass::processMemCpyMemCpyDependence"]}}, "patch": "commit 99685a54d1322694847f59386a548b3696a43fc3\nAuthor: Yingwei Zheng \nDate: Sat Jul 13 04:27:07 2024 +0800\n\n [MemCpyOpt] Use `dyn_cast` to fix assertion failure in `processMemCpyMemCpyDependence` (#98686)\n \n Fixes https://github.com/llvm/llvm-project/issues/98675.\n\ndiff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\nindex 1c65219585e5..4063762c88a2 100644\n--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\n+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\n@@ -1188,9 +1188,9 @@ bool MemCpyOptPass::processMemCpyMemCpyDependence(MemCpyInst *M,\n if (MDestOffset == MForwardOffset)\n CopySource = M->getDest();\n else {\n- NewCopySource = cast(Builder.CreateInBoundsPtrAdd(\n- CopySource, Builder.getInt64(MForwardOffset)));\n- CopySource = NewCopySource;\n+ CopySource = Builder.CreateInBoundsPtrAdd(\n+ CopySource, Builder.getInt64(MForwardOffset));\n+ NewCopySource = dyn_cast(CopySource);\n }\n // We need to update `MCopyLoc` if an offset exists.\n MCopyLoc = MCopyLoc.getWithNewPtr(CopySource);\n", "tests": [{"file": "llvm/test/Transforms/MemCpyOpt/memcpy-memcpy-offset.ll", "commands": ["opt < %s -passes=memcpyopt -S -verify-memoryssa"], "tests": [{"test_name": "do_not_forward_offset_and_store", "test_body": "define void @do_not_forward_offset_and_store(ptr %src, ptr %dest) {\n %cpy_tmp = alloca [9 x i8], align 1\n call void @llvm.memcpy.p0.p0.i64(ptr align 1 %cpy_tmp, ptr align 1 %src, i64 7, i1 false)\n %src_offset = getelementptr inbounds i8, ptr %src, i64 1\n store i8 1, ptr %src_offset, align 1\n %cpy_tmp_offset = getelementptr inbounds i8, ptr %cpy_tmp, i64 1\n call void @llvm.memcpy.p0.p0.i64(ptr align 1 %dest, ptr align 1 %cpy_tmp_offset, i64 5, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}, {"test_name": "pr98675", "test_body": "@buf = external global [32 x i8]\n\ndefine void @pr98675(ptr noalias %p1, ptr noalias %p2) {\n call void @llvm.memcpy.p0.p0.i64(ptr %p1, ptr @buf, i64 26, i1 false)\n %gep = getelementptr i8, ptr %p1, i64 10\n call void @llvm.memmove.p0.p0.i64(ptr %p2, ptr %gep, i64 1, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memmove.p0.p0.i64(ptr nocapture writeonly, ptr nocapture readonly, i64, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}]}], "issue": {"title": "[MemCpyOpt] Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.", "body": "Detected by https://github.com/dtcxzyw/llvm-opt-benchmark/issues/935.\r\n\r\nReproducer: \r\n```\r\n; bin/opt -passes=memcpyopt -S test.ll\r\n@b = external global [32 x i8]\r\n\r\ndefine void @_Z15TestBasicStringIN5eastl12basic_stringIc17CountingAllocatorEEEiv(ptr %call.i.i.i12866) {\r\n %a = alloca [32 x i8], align 1\r\n call void @llvm.memcpy.p0.p0.i64(ptr %call.i.i.i12866, ptr @b, i64 26, i1 false)\r\n %.sroa.gep22184 = getelementptr i8, ptr %call.i.i.i12866, i64 10\r\n call void @llvm.memmove.p0.p0.i64(ptr %a, ptr %.sroa.gep22184, i64 1, i1 false)\r\n ret void\r\n}\r\n```\r\n```\r\nopt: /home/dtcxzyw/WorkSpace/Projects/compilers/llvm-project/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = Instruction; From = Value]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: bin/opt -passes=memcpyopt reduced.ll\r\n1. Running pass \"function(memcpyopt)\" on module \"reduced.ll\"\r\n2. Running pass \"memcpyopt\" on function \"_Z15TestBasicStringIN5eastl12basic_stringIc17CountingAllocatorEEEiv\"\r\n #0 0x00007e04ebe0c9c2 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.19.0git+0x20c9c2)\r\n #1 0x00007e04ebe0988f llvm::sys::RunSignalHandlers() (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.19.0git+0x20988f)\r\n #2 0x00007e04ebe099d5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007e04eb842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007e04eb8969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007e04eb8969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007e04eb8969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007e04eb842476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007e04eb8287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007e04eb82871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007e04eb839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00007e04e5ab5dfa llvm::MemCpyOptPass::processMemCpyMemCpyDependence(llvm::MemCpyInst*, llvm::MemCpyInst*, llvm::BatchAAResults&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMScalarOpts.so.19.0git+0x2b5dfa)\r\n#12 0x00007e04e5abd2aa llvm::MemCpyOptPass::processMemCpy(llvm::MemCpyInst*, llvm::ilist_iterator_w_bits, false, false>&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMScalarOpts.so.19.0git+0x2bd2aa)\r\n#13 0x00007e04e5abd9c4 llvm::MemCpyOptPass::iterateOnFunction(llvm::Function&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMScalarOpts.so.19.0git+0x2bd9c4)\r\n#14 0x00007e04e5abdcce llvm::MemCpyOptPass::runImpl(llvm::Function&, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::AssumptionCache*, llvm::DominatorTree*, llvm::PostDominatorTree*, llvm::MemorySSA*) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMScalarOpts.so.19.0git+0x2bdcce)\r\n#15 0x00007e04e5abe15c llvm::MemCpyOptPass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMScalarOpts.so.19.0git+0x2be15c)\r\n#16 0x00007e04e74cde55 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMPasses.so.19.0git+0xcde55)\r\n#17 0x00007e04e474b7b2 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.19.0git+0x34b7b2)\r\n#18 0x00007e04eaad5065 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.19.0git+0xd5065)\r\n#19 0x00007e04e474a40e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.19.0git+0x34a40e)\r\n#20 0x00007e04eaad5985 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.19.0git+0xd5985)\r\n#21 0x00007e04e47483da llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.19.0git+0x3483da)\r\n#22 0x00007e04ec103267 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.19.0git+0x2d267)\r\n#23 0x00007e04ec10f212 optMain (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.19.0git+0x39212)\r\n#24 0x00007e04eb829d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#25 0x00007e04eb829e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#26 0x00007e04eb829e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#27 0x00005df852081095 _start (bin/opt+0x1095)\r\nAborted (core dumped)\r\n```\r\n\r\nThis issue is caused by https://github.com/llvm/llvm-project/commit/fa2421392802025e842454a31ed50dbfab643130.\r\nhttps://github.com/llvm/llvm-project/blob/d1dc5047f1127d64eb82eec976fabaa2d9d20c11/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp#L1191-L1192\r\n`Builder.CreateInBoundsPtrAdd` may return `GetElementPtrConstantExpr`.\r\n\r\ncc @nikic @DianQK \r\n\r\n", "author": "dtcxzyw", "labels": ["crash-on-valid", "llvm:transforms"], "comments": [{"author": "DianQK", "body": "It looks similar to https://lab.llvm.org/buildbot/#/builders/41/builds/587. I really shouldn't have merged the code before going to bed :3"}]}, "verified": true} {"bug_id": "98660", "issue_url": "https://github.com/llvm/llvm-project/issues/98660", "bug_type": "crash", "base_commit": "efde640cdfede834fa79e0eea69f82fb769d6beb", "knowledge_cutoff": "2024-07-12T16:45:01Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "fc9cd3272b50f4ee9f18c4ab82c278bbb014d99f", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[8693, 8698]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["addUsersInExitBlock"]}}, "patch": "commit fc9cd3272b50f4ee9f18c4ab82c278bbb014d99f\nAuthor: Florian Hahn \nDate: Sun Jul 14 20:49:03 2024 +0100\n\n [VPlan] Don't add live-outs for IV phis.\n \n Resume and exit values for inductions are currently still created\n outside of VPlan and independent of the induction recipes. Don't add\n live-outs for now, as the additional unneeded users can pessimize other\n anlysis.\n \n Fixes https://github.com/llvm/llvm-project/issues/98660.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 7d37d67cde29..5520baef7152 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -8693,6 +8693,14 @@ static void addUsersInExitBlock(VPBasicBlock *HeaderVPBB, Loop *OrigLoop,\n Value *IncomingValue =\n ExitPhi.getIncomingValueForBlock(ExitingBB);\n VPValue *V = Builder.getVPValueOrAddLiveIn(IncomingValue, Plan);\n+ // Exit values for inductions are computed and updated outside of VPlan and\n+ // independent of induction recipes.\n+ // TODO: Compute induction exit values in VPlan, use VPLiveOuts to update\n+ // live-outs.\n+ if ((isa(V) &&\n+ !cast(V)->getTruncInst()) ||\n+ isa(V))\n+ continue;\n Plan.addLiveOut(&ExitPhi, V);\n }\n }\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/sve-live-out-pointer-induction.ll", "commands": ["opt -passes=loop-vectorize -mtriple=aarch64-unknown -mattr=+sve -S %s"], "tests": [{"test_name": "test", "test_body": "define ptr @test(ptr %start.1, ptr %start.2, ptr %end) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv.1 = phi ptr [ %start.1, %entry ], [ %iv.1.next, %loop ]\n %iv.2 = phi ptr [ %start.2, %entry ], [ %iv.2.next, %loop ]\n store i64 0, ptr %iv.2, align 4\n %iv.2.next = getelementptr inbounds ptr, ptr %iv.2, i64 1\n %iv.1.next = getelementptr inbounds ptr, ptr %iv.1, i64 1\n %cmp.i.i.not.i = icmp eq ptr %iv.2.next, %end\n br i1 %cmp.i.i.not.i, label %exit, label %loop\n\nexit: ; preds = %loop\n %res.lcssa = phi ptr [ %iv.1, %loop ]\n ret ptr %res.lcssa\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/ephemeral-recipes.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "ephemeral_load_and_compare_iv_used_outside", "test_body": "target datalayout = \"e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx\"\n\ndefine i32 @ephemeral_load_and_compare_iv_used_outside(ptr %start, ptr %end) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi ptr [ %start, %entry ], [ %iv.next, %loop ]\n %iv.next = getelementptr nusw i8, ptr %iv, i64 -8\n %l1 = load i32, ptr %iv, align 4\n %cmp = icmp ne i32 %l1, 0\n call void @llvm.assume(i1 %cmp)\n %cmp.not = icmp eq ptr %iv, %end\n br i1 %cmp.not, label %exit, label %loop\n\nexit: ; preds = %loop\n %final.load = load i32, ptr %iv, align 4\n ret i32 %final.load\n}\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #1\n\nattributes #0 = { \"target-cpu\"=\"skylake-avx512\" }\nattributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/X86/iv-live-outs.ll", "commands": ["opt -p loop-vectorize -mcpu=skylake-avx512 -S %s"], "tests": [{"test_name": "", "test_body": "\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i64 @test_pr98660(ptr %dst, i64 %N) {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %or = or disjoint i64 %iv, 1\n %gep = getelementptr i32, ptr %dst, i64 %or\n %l = load i32, ptr %gep\n %c = icmp eq i32 %l, 0\n br i1 %c, label %then, label %loop.latch\n\nthen:\n store i32 0, ptr %gep, align 4\n br label %loop.latch\n\nloop.latch:\n %iv.next = add i64 %iv, 1\n %ec = icmp ult i64 %iv, %N\n br i1 %ec, label %loop.header, label %exit\n\nexit:\n %ret = phi i64 [ %iv, %loop.latch ]\n ret i64 %ret\n}\n;.\n;."}]}, {"file": "llvm/test/Transforms/LoopVectorize/iv_outside_user.ll", "commands": ["opt -S -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 < %s", "opt -S -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=1 < %s"], "tests": [{"test_name": "iv_2_dead_in_loop_only_used_outside", "test_body": "define i32 @iv_2_dead_in_loop_only_used_outside(ptr %ptr) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.2 = phi i32 [ 0, %entry ], [ %iv.2.next, %loop ]\n %iv.next = add nuw i64 %iv, 1\n %iv.2.next = add nuw i32 %iv.2, 2\n %gep.ptr = getelementptr inbounds i64, ptr %ptr, i64 %iv\n store i64 %iv, ptr %gep.ptr, align 4\n %exitcond = icmp ugt i64 %iv, 1000\n br i1 %exitcond, label %exit, label %loop\n\nexit: ; preds = %loop\n %iv.2.lcssa = phi i32 [ %iv.2, %loop ]\n ret i32 %iv.2.lcssa\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/no-fold-tail-by-masking-iv-external-uses.ll", "commands": ["opt < %s -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S"], "tests": [{"test_name": "test", "test_body": "define i32 @test(ptr %arr, i64 %n) {\nentry:\n %cmp1 = icmp ugt i64 %n, 1\n br i1 %cmp1, label %preheader, label %done\n\npreheader: ; preds = %entry\n br label %loop\n\nloop: ; preds = %loop, %preheader\n %conv = phi i64 [ %conv2, %loop ], [ 1, %preheader ]\n %i = phi i8 [ %inc, %loop ], [ 1, %preheader ]\n %sub = add nsw i64 %conv, -1\n %ptr = getelementptr inbounds i32, ptr %arr, i64 %sub\n store i32 65, ptr %ptr, align 4\n %inc = add i8 %i, 1\n %conv2 = zext i8 %inc to i64\n %cmp2 = icmp ult i64 %conv2, %n\n br i1 %cmp2, label %loop, label %load_val, !llvm.loop !0\n\nload_val: ; preds = %loop\n %final = phi i64 [ %conv, %loop ]\n %ptr2 = getelementptr inbounds i32, ptr %arr, i64 %final\n %val = load i32, ptr %ptr2, align 4\n br label %done\n\ndone: ; preds = %load_val, %entry\n %value = phi i32 [ %val, %load_val ], [ 0, %entry ]\n ret i32 %value\n}\n\n!0 = distinct !{!0, !1, !2, !3}\n!1 = !{!\"llvm.loop.unroll.disable\"}\n!2 = !{!\"llvm.loop.vectorize.predicate.enable\", i1 true}\n!3 = !{!\"llvm.loop.vectorize.enable\", i1 true}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/pr58811-scev-expansion.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "test1_pr58811", "test_body": "define void @test1_pr58811() {\nentry:\n br label %loop.1.preheader\n\nloop.1.preheader: ; preds = %unreachable.bb, %entry\n %iv.1.ph = phi i32 [ %sub93.2, %unreachable.bb ], [ 0, %entry ]\n br label %loop.1\n\nloop.1: ; preds = %loop.1, %loop.1.preheader\n %iv.1 = phi i32 [ %iv.1.next, %loop.1 ], [ %iv.1.ph, %loop.1.preheader ]\n %iv.2 = phi i32 [ %iv.2.next, %loop.1 ], [ 0, %loop.1.preheader ]\n %iv.2.next = add i32 %iv.2, 1\n %iv.1.next = add i32 %iv.2, %iv.1\n br i1 false, label %loop.1, label %loop.2.preheader\n\nloop.2.preheader: ; preds = %loop.1\n %iv.1.lcssa = phi i32 [ %iv.1, %loop.1 ]\n br label %loop.2\n\nloop.2: ; preds = %loop.2, %loop.2.preheader\n %iv.3 = phi i16 [ %iv.3.next, %loop.2 ], [ 0, %loop.2.preheader ]\n %iv.4 = phi i32 [ %iv.4.next, %loop.2 ], [ 0, %loop.2.preheader ]\n %iv.4.next = sub i32 %iv.4, %iv.1.lcssa\n %iv.3.next = add i16 %iv.3, 1\n %cmp88.1 = icmp ult i16 %iv.3, 198\n br i1 %cmp88.1, label %loop.2, label %loop.3.preheader\n\nloop.3.preheader: ; preds = %loop.2\n %iv.4.lcssa = phi i32 [ %iv.4, %loop.2 ]\n br label %loop.3\n\nloop.3: ; preds = %loop.3, %loop.3.preheader\n %iv.5 = phi i32 [ %sub93.2, %loop.3 ], [ 0, %loop.3.preheader ]\n %sub93.2 = sub i32 %iv.5, %iv.4.lcssa\n br label %loop.3\n\nunreachable.bb: ; No predecessors!\n br label %loop.1.preheader\n}\n"}, {"test_name": "test2_pr58811", "test_body": "define void @test2_pr58811() {\nentry:\n br label %loop.1.header\n\nloop.1.header: ; preds = %loop.4, %entry\n %p.1 = phi i32 [ 0, %entry ], [ %sub93.2, %loop.4 ]\n br label %loop.2\n\nloop.2: ; preds = %loop.2, %loop.1.header\n %iv.2 = phi i32 [ %p.1, %loop.1.header ], [ %add101, %loop.2 ]\n %iv.3 = phi i32 [ 0, %loop.1.header ], [ %sub93, %loop.2 ]\n %sub93 = add i32 %iv.3, 1\n %add101 = add i32 %iv.3, %iv.2\n br i1 false, label %loop.2, label %loop.3\n\nloop.3: ; preds = %loop.3, %loop.2\n %iv.4 = phi i16 [ 0, %loop.2 ], [ %inc.1, %loop.3 ]\n %iv.5 = phi i32 [ 0, %loop.2 ], [ %sub93.1, %loop.3 ]\n %sub93.1 = sub i32 %iv.5, %iv.2\n %inc.1 = add i16 %iv.4, 1\n %cmp88.1 = icmp ult i16 %iv.4, 198\n br i1 %cmp88.1, label %loop.3, label %loop.4\n\nloop.4: ; preds = %loop.4, %loop.3\n %iv.6 = phi i32 [ 0, %loop.3 ], [ %sub93.2, %loop.4 ]\n %sub93.2 = sub i32 %iv.6, %iv.5\n br i1 false, label %loop.4, label %loop.1.header\n}\n"}, {"test_name": "test3_pr58811", "test_body": "define void @test3_pr58811() {\nentry:\n br label %loop.1.header\n\nloop.1.header: ; preds = %loop.1.latch, %entry\n %p.1 = phi i32 [ 0, %entry ], [ %sub93.2, %loop.1.latch ]\n %rem85 = urem i32 1, %p.1\n br label %loop.2\n\nloop.2: ; preds = %loop.2, %loop.1.header\n %p.2 = phi i32 [ 1, %loop.1.header ], [ 0, %loop.2 ]\n %add101 = add i32 %rem85, %p.2\n br i1 false, label %loop.2, label %loop.3\n\nloop.3: ; preds = %loop.3, %loop.2\n %iv.3 = phi i16 [ 0, %loop.2 ], [ %inc.1, %loop.3 ]\n %iv.4 = phi i32 [ 0, %loop.2 ], [ %sub93.1, %loop.3 ]\n %sub93.1 = sub i32 %iv.4, %add101\n %inc.1 = add i16 %iv.3, 1\n %cmp88.1 = icmp ult i16 %iv.3, 198\n br i1 %cmp88.1, label %loop.3, label %loop.4\n\nloop.4: ; preds = %loop.4, %loop.3\n %iv.5 = phi i32 [ 0, %loop.3 ], [ %sub93.2, %loop.4 ]\n %sub93.2 = sub i32 %iv.5, %iv.4\n br label %loop.4\n\nloop.1.latch: ; No predecessors!\n br label %loop.1.header\n}\n"}]}], "issue": {"title": "Assertion `(VF.isScalar() || Def->isLiveIn() || hasVectorValue(Def, Part) || (hasScalarValue(Def, VPIteration(Part, 0)) && Data.PerPartScalars[Def][Part].size() == 1)) && \"Trying to access a single scalar per part but has multiple scalars \" \"per part.\"' failed.", "body": "To reproduce run opt opt with -passes loop-vectorize with the following test:\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat(i64 %arg) #0 gc \"statepoint-example\" {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb3, %bb\r\n %phi = phi i64 [ 0, %bb ], [ %add, %bb3 ]\r\n br i1 false, label %bb2, label %bb3\r\n\r\nbb2: ; preds = %bb1\r\n %or = or disjoint i64 %phi, 0\r\n %getelementptr = getelementptr i32, ptr addrspace(1) null, i64 %or\r\n store i32 0, ptr addrspace(1) %getelementptr, align 4\r\n br label %bb3\r\n\r\nbb3: ; preds = %bb2, %bb1\r\n %add = add i64 %phi, 1\r\n %icmp = icmp ult i64 %phi, %arg\r\n br i1 %icmp, label %bb1, label %bb4\r\n\r\nbb4: ; preds = %bb3\r\n %phi5 = phi i64 [ %phi, %bb3 ]\r\n ret void\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-gfni,-avxvnniint16,-amx-fp16,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f\" }\r\n```\r\nReproducer:\r\nhttps://godbolt.org/z/Ghz3147nj\r\n\r\nStack dump:\r\n```\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes loop-vectorize \r\n1.\tRunning pass \"function(loop-vectorize)\" on module \"\"\r\n2.\tRunning pass \"loop-vectorize\" on function \"wombat\"\r\n #0 0x0000000004e9b948 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e9b948)\r\n #1 0x0000000004e990bc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000070edf8442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000070edf84969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000070edf8442476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000070edf84287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000070edf842871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000070edf8439e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x00000000040395f7 llvm::VPTransformState::get(llvm::VPValue*, unsigned int, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40395f7)\r\n #9 0x0000000004056ece llvm::VPInstruction::generatePerPart(llvm::VPTransformState&, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4056ece)\r\n#10 0x00000000040595ee llvm::VPInstruction::execute(llvm::VPTransformState&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40595ee)\r\n#11 0x0000000004030a1f llvm::VPBasicBlock::executeRecipes(llvm::VPTransformState*, llvm::BasicBlock*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4030a1f)\r\n#12 0x00000000040320e8 llvm::VPBasicBlock::execute(llvm::VPTransformState*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x40320e8)\r\n#13 0x0000000004030612 llvm::VPRegionBlock::execute(llvm::VPTransformState*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4030612)\r\n#14 0x0000000004039946 llvm::VPlan::execute(llvm::VPTransformState*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4039946)\r\n#15 0x0000000003f1d475 llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f1d475)\r\n#16 0x0000000003f332ea llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f332ea)\r\n#17 0x0000000003f35a29 llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f35a29)\r\n#18 0x0000000003f3614b llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3f3614b)\r\n#19 0x0000000002e212be llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2e212be)\r\n#20 0x0000000004c9f2d8 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c9f2d8)\r\n#21 0x0000000000dc9bee llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdc9bee)\r\n#22 0x0000000004c9dd46 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c9dd46)\r\n#23 0x0000000000dca1fe llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdca1fe)\r\n#24 0x0000000004c9bf10 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c9bf10)\r\n#25 0x00000000008db1d2 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8db1d2)\r\n#26 0x00000000008ce0fc optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ce0fc)\r\n#27 0x000070edf8429d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#28 0x000070edf8429e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#29 0x00000000008c556e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8c556e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "Interesting, let me take a look"}]}, "verified": true} {"bug_id": "91883", "issue_url": "https://github.com/llvm/llvm-project/issues/91883", "bug_type": "crash", "base_commit": "90109d444839683b09f0aafdc50b749cb4b3203b", "knowledge_cutoff": "2024-05-12T12:03:48Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "67d840b60fbd75ca1b52d77bd3353771ec853735", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlan.cpp": [[246, 252]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlan.cpp": ["VPTransformState::get"]}}, "patch": "commit 67d840b60fbd75ca1b52d77bd3353771ec853735\nAuthor: Florian Hahn \nDate: Tue May 14 19:10:48 2024 +0100\n\n [VPlan] Relax over-aggressive assertion in VPTransformState::get().\n \n There are cases where a vector value has some users that demand the\n the single scalar value only (NeedsScalar), while other users demand the\n vector value (see attached test cases). In those cases, the NeedsScalar\n users should only demand the first lane.\n \n Fixes https://github.com/llvm/llvm-project/issues/91883.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp\nindex 999236ae8489..27f8e239b1c0 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp\n@@ -246,7 +246,7 @@ Value *VPTransformState::get(VPValue *Def, const VPIteration &Instance) {\n \n Value *VPTransformState::get(VPValue *Def, unsigned Part, bool NeedsScalar) {\n if (NeedsScalar) {\n- assert((VF.isScalar() || Def->isLiveIn() ||\n+ assert((VF.isScalar() || Def->isLiveIn() || hasVectorValue(Def, Part) ||\n (hasScalarValue(Def, VPIteration(Part, 0)) &&\n Data.PerPartScalars[Def][Part].size() == 1)) &&\n \"Trying to access a single scalar per part but has multiple scalars \"\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/widened-value-used-as-scalar-and-first-lane.ll", "commands": ["opt -S -passes=loop-vectorize -mcpu=skylake-avx512 -mtriple=x86_64-apple-macosx -S %s"], "tests": [{"test_name": "", "test_body": "\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\n\n; Test cases based on https://github.com/llvm/llvm-project/issues/91883.\ndefine void @iv.4_used_as_vector_and_first_lane(ptr %src, ptr noalias %dst) {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %g.src = getelementptr inbounds i64, ptr %src, i64 %iv\n %l = load i64, ptr %g.src\n %iv.4 = add nuw nsw i64 %iv, 4\n %c = icmp ule i64 %l, 128\n br i1 %c, label %loop.then, label %loop.latch\n\nloop.then:\n %or = or disjoint i64 %iv.4, 1\n %g.dst = getelementptr inbounds i64, ptr %dst, i64 %or\n store i64 %iv.4, ptr %g.dst, align 4\n br label %loop.latch\n\nloop.latch:\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond = icmp eq i64 %iv.next, 32\n br i1 %exitcond, label %exit, label %loop.header\n\nexit:\n ret void\n}\n\ndefine void @iv.4_used_as_first_lane(ptr %src, ptr noalias %dst) {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %g.src = getelementptr inbounds i64, ptr %src, i64 %iv\n %l = load i64, ptr %g.src\n %iv.4 = add nuw nsw i64 %iv, 4\n %c = icmp ule i64 %l, 128\n br i1 %c, label %loop.then, label %loop.latch\n\nloop.then:\n %or = or disjoint i64 %iv.4, 1\n %g.dst = getelementptr inbounds i64, ptr %dst, i64 %or\n store i64 %l, ptr %g.dst, align 4\n br label %loop.latch\n\nloop.latch:\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond = icmp eq i64 %iv.next, 32\n br i1 %exitcond, label %exit, label %loop.header\n\nexit:\n ret void\n}\n;.\n;."}]}], "issue": {"title": "New LLVM vectorizer crash in assertion", "body": "@fhahn hits an assertion introduced in https://github.com/llvm/llvm-project/pull/80271\r\n\r\nInput file: https://gist.github.com/cheshire/e504c63db9bdb300fbdcc1df2e183bd8\r\n\r\nRepro:\r\n\r\n```\r\nopt -S -passes=loop-vectorize -mattr=+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-avx512pf,-tsxldtrk,-ptwrite,-widekl,-sm3,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-avx512er,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-gfni,-avxvnniint16,-amx-fp16,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-prefetchwt1,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a -mcpu=skylake-avx512 /tmp/minimized_crash.ll\r\n```", "author": "cheshire", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "98435", "issue_url": "https://github.com/llvm/llvm-project/issues/98435", "bug_type": "miscompilation", "base_commit": "0e124537aabea0d18e030226a3ca4c6f494c1322", "knowledge_cutoff": "2024-07-11T06:06:08Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "9304af3927caecdb43d3a9b5d16c6a5b7a6b5594", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": [[1735, 1751]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": ["InstCombinerImpl::SimplifyDemandedVectorElts"]}}, "patch": "commit 9304af3927caecdb43d3a9b5d16c6a5b7a6b5594\nAuthor: Jorge Botto <23462171+jf-botto@users.noreply.github.com>\nDate: Fri Aug 9 11:52:56 2024 +0100\n\n [InstCombine] Fixing wrong select folding in vectors with undef elements (#102244)\n \n This PR fixes https://github.com/llvm/llvm-project/issues/98435.\n `SimplifyDemandedVectorElts` mishandles the undef by assuming that\n !isNullValue() means the condition is true.\n \n By preventing any value that we're not certain equals 1 or 0, it avoids\n having to make any particular choice by not demanding bits from a\n particular branch with potentially picking a wrong value.\n \n Proof: https://alive2.llvm.org/ce/z/r8CmEu\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\nindex c494fec84c1e..153d8c238ed4 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n@@ -1735,17 +1735,12 @@ Value *InstCombinerImpl::SimplifyDemandedVectorElts(Value *V,\n APInt DemandedLHS(DemandedElts), DemandedRHS(DemandedElts);\n if (auto *CV = dyn_cast(Sel->getCondition())) {\n for (unsigned i = 0; i < VWidth; i++) {\n- // isNullValue() always returns false when called on a ConstantExpr.\n- // Skip constant expressions to avoid propagating incorrect information.\n Constant *CElt = CV->getAggregateElement(i);\n- if (isa(CElt))\n- continue;\n- // TODO: If a select condition element is undef, we can demand from\n- // either side. If one side is known undef, choosing that side would\n- // propagate undef.\n+\n+ // isNullValue() always returns false when called on a ConstantExpr.\n if (CElt->isNullValue())\n DemandedLHS.clearBit(i);\n- else\n+ else if (CElt->isOneValue())\n DemandedRHS.clearBit(i);\n }\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/pr98435.ll", "commands": ["opt -S -passes=instcombine < %s 2>&1"], "tests": [{"test_name": "pr98435", "test_body": "define <2 x i1> @pr98435(<2 x i1> %val) {\n %val1 = select <2 x i1> , <2 x i1> splat (i1 true), <2 x i1> %val\n ret <2 x i1> %val1\n}\n"}]}], "issue": {"title": "Wrong select folding for undef conditions by InstCombine", "body": "https://github.com/llvm/llvm-project/blob/7eae9bb856135136cddc4208a2b1546e9db44c9c/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp#L3582-L3585\r\n\r\nAlive2 report: https://alive2.llvm.org/ce/z/Qtu72J\r\n\r\n\r\n```ll\r\n----------------------------------------\r\ndefine <2 x i1> @fun0(<2 x i1> %val0, <2 x i1> %val1) {\r\n#0:\r\n %val2 = select <2 x i1> %val0, <2 x i1> %val0, <2 x i1> %val0\r\n %val3 = select <2 x i1> { undef, 1 }, <2 x i1> { 1, 1 }, <2 x i1> %val2\r\n ret <2 x i1> %val3\r\n}\r\n=>\r\ndefine <2 x i1> @fun0(<2 x i1> %val0, <2 x i1> %val1) {\r\n#0:\r\n ret <2 x i1> { poison, 1 }\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\n<2 x i1> %val0 = < #x0 (0), #x0 (0) >\r\n<2 x i1> %val1 = < #x0 (0), #x0 (0) >\r\n\r\nSource:\r\n<2 x i1> %val2 = < #x0 (0), #x0 (0) >\r\n<2 x i1> %val3 = < #x0 (0)\t[based on undef value], #x1 (1) >\r\n\r\nTarget:\r\nSource value: < #x0 (0), #x1 (1) >\r\nTarget value: < poison, #x1 (1) >\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "jf-botto", "body": "Would this issue be to fix InstCombine to not try to optimise it or get InstCombine to pick `<2 x i1> { 1, 1 }` because `undef` let's it pick either values and picking a constant vector is better than picking a variable?"}, {"author": "nikic", "body": "The problematic step seems to be this one: https://alive2.llvm.org/ce/z/C2waWU\r\n\r\nI don't think the quoted code is responsible for this, as it is guarded by `!isa(CondVal)`."}, {"author": "jf-botto", "body": "I think the problem is with how `SimplifyDemandedVectorElts` handles the undef: https://github.com/llvm/llvm-project/blob/af80d3a248101d6f5d9d5e229c7899136b8ce0b8/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp#L1721-L1760"}, {"author": "nikic", "body": "@jf-botto Agree. It assumes that !isNullValue() means \"is true\", which is not the case."}]}, "verified": true} {"bug_id": "105120", "issue_url": "https://github.com/llvm/llvm-project/issues/105120", "bug_type": "crash", "base_commit": "6ec3130a38e6982a61e7fa74bd5223c95c0bb918", "knowledge_cutoff": "2024-08-20T17:04:34Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "e31252bf54dedadfe78b36d07ea6084156faa38a", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[7227, 7232], [7238, 7247]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::buildTree_rec"]}}, "patch": "commit e31252bf54dedadfe78b36d07ea6084156faa38a\nAuthor: Alexey Bataev \nDate: Wed Aug 21 11:47:00 2024 -0700\n\n [SLP]Fix PR105120: fix the order of phi nodes vectorization.\n \n The operands of the phi nodes should be vectorized in the same order, in\n which they were created, otherwise the compiler may crash when trying\n to correctly build dependency for nodes with non-schedulable\n instructions for gather/buildvector nodes.\n \n Fixes https://github.com/llvm/llvm-project/issues/105120\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex dee6d688b1b9..848e0de20e7b 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -7227,6 +7227,22 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n \n unsigned ShuffleOrOp = S.isAltShuffle() ?\n (unsigned) Instruction::ShuffleVector : S.getOpcode();\n+ auto CreateOperandNodes = [&](TreeEntry *TE, const auto &Operands) {\n+ // Postpone PHI nodes creation\n+ SmallVector PHIOps;\n+ for (unsigned I : seq(Operands.size())) {\n+ ArrayRef Op = Operands[I];\n+ if (Op.empty())\n+ continue;\n+ InstructionsState S = getSameOpcode(Op, *TLI);\n+ if (S.getOpcode() != Instruction::PHI || S.isAltShuffle())\n+ buildTree_rec(Op, Depth + 1, {TE, I});\n+ else\n+ PHIOps.push_back(I);\n+ }\n+ for (unsigned I : PHIOps)\n+ buildTree_rec(Operands[I], Depth + 1, {TE, I});\n+ };\n switch (ShuffleOrOp) {\n case Instruction::PHI: {\n auto *PH = cast(VL0);\n@@ -7238,10 +7254,12 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n // Keeps the reordered operands to avoid code duplication.\n PHIHandler Handler(*DT, PH, VL);\n Handler.buildOperands();\n- for (unsigned I : seq(0, PH->getNumOperands()))\n+ for (unsigned I : seq(PH->getNumOperands()))\n TE->setOperand(I, Handler.getOperands(I));\n- for (unsigned I : seq(0, PH->getNumOperands()))\n- buildTree_rec(Handler.getOperands(I), Depth + 1, {TE, I});\n+ SmallVector> Operands(PH->getNumOperands());\n+ for (unsigned I : seq(PH->getNumOperands()))\n+ Operands[I] = Handler.getOperands(I);\n+ CreateOperandNodes(TE, Operands);\n return;\n }\n case Instruction::ExtractValue:\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/phi-nodes-as-operand-reorder.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -slp-threshold=-99999 < %s"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb3, %bb\n %phi = phi i32 [ 0, %bb ], [ %add5, %bb3 ]\n %phi2 = phi i32 [ 0, %bb ], [ %add, %bb3 ]\n br i1 false, label %bb6, label %bb3\n\nbb3: ; preds = %bb1\n %add = add i32 0, 0\n %add4 = add i32 0, 0\n %add5 = add i32 %phi, 0\n br i1 false, label %bb6, label %bb1\n\nbb6: ; preds = %bb3, %bb1\n %phi7 = phi i32 [ %phi2, %bb1 ], [ %add4, %bb3 ]\n %phi8 = phi i32 [ %phi, %bb1 ], [ %add5, %bb3 ]\n ret void\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Instruction does not dominate all uses! LLVM ERROR: Broken module found, compilation aborted!", "body": "To reproduce run the test below with -passes slp-vectorizer -slp-threshold=-99999\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat() #0 gc \"statepoint-example\" {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb3, %bb\r\n %phi = phi i32 [ 0, %bb ], [ %add5, %bb3 ]\r\n %phi2 = phi i32 [ 0, %bb ], [ %add, %bb3 ]\r\n br i1 false, label %bb6, label %bb3\r\n\r\nbb3: ; preds = %bb1\r\n %add = add i32 0, 0\r\n %add4 = add i32 0, 0\r\n %add5 = add i32 %phi, 0\r\n br i1 false, label %bb6, label %bb1\r\n\r\nbb6: ; preds = %bb3, %bb1\r\n %phi7 = phi i32 [ %phi2, %bb1 ], [ %add4, %bb3 ]\r\n %phi8 = phi i32 [ %phi, %bb1 ], [ %add5, %bb3 ]\r\n ret void\r\n}\r\n```\r\nReproducer: https://godbolt.org/z/W96s5a1cb\r\n\r\n```\r\nInstruction does not dominate all uses!\r\n %2 = shufflevector <2 x i32> %0, <2 x i32> , <2 x i32> \r\n %1 = add <2 x i32> zeroinitializer, %2\r\nLLVM ERROR: Broken module found, compilation aborted!\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes slp-vectorizer -slp-threshold=-99999 \r\n1.\tRunning pass \"verify\" on module \"\"\r\n #0 0x0000000004f84b88 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f84b88)\r\n #1 0x0000000004f822fc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000074dcb3642520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000074dcb36969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000074dcb3642476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000074dcb36287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00000000007b7cc5 llvm::json::operator==(llvm::json::Value const&, llvm::json::Value const&) (.cold) JSON.cpp:0:0\r\n #7 0x0000000004ebea18 (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ebea18)\r\n #8 0x0000000004dbae20 llvm::VerifierPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4dbae20)\r\n #9 0x00000000008dbbde llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8dbbde)\r\n#10 0x0000000004d7ebb0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d7ebb0)\r\n#11 0x00000000008e6912 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e6912)\r\n#12 0x00000000008d97ec optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d97ec)\r\n#13 0x000074dcb3629d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#14 0x000074dcb3629e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#15 0x00000000008d127e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d127e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "llvm:crash"], "comments": [{"author": "alexey-bataev", "body": "Yes"}]}, "verified": true} {"bug_id": "105722", "issue_url": "https://github.com/llvm/llvm-project/issues/105722", "bug_type": "crash", "base_commit": "84497c6f4f6c79b0d8c38da666724eed7e9e8db5", "knowledge_cutoff": "2024-08-22T19:30:39Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "533e6bbd0d344a710c491a9eb0ce0ba0852b08cb", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h": [[13, 18], [173, 178]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h": ["getVPValueOrAddLiveIn"]}}, "patch": "commit 533e6bbd0d344a710c491a9eb0ce0ba0852b08cb\nAuthor: Florian Hahn \nDate: Mon Aug 26 09:15:58 2024 +0100\n\n [VPlan] Simplify live-ins if they are SCEVConstant.\n \n The legacy cost model in some parts checks if any of the operands are\n constants via SCEV. Update VPlan construction to replace live-ins that\n are constants via SCEV with such constants. This means VPlans (and\n codegen) reflects what we computing the cost of and removes another case\n where the legacy and VPlan cost model diverged.\n \n Fixes https://github.com/llvm/llvm-project/issues/105722.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h\nindex 0b27933925d8..0fbdeeddaf4a 100644\n--- a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h\n+++ b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h\n@@ -13,6 +13,7 @@\n #include \"VPlan.h\"\n #include \"llvm/ADT/DenseMap.h\"\n #include \"llvm/ADT/PointerUnion.h\"\n+#include \"llvm/Analysis/ScalarEvolutionExpressions.h\"\n #include \"llvm/IR/IRBuilder.h\"\n \n namespace llvm {\n@@ -173,6 +174,11 @@ public:\n if (auto *R = Ingredient2Recipe.lookup(I))\n return R->getVPSingleValue();\n }\n+ ScalarEvolution &SE = *PSE.getSE();\n+ if (!isa(V) && SE.isSCEVable(V->getType()))\n+ if (auto *C = dyn_cast(PSE.getSE()->getSCEV(V)))\n+ return Plan.getOrAddLiveIn(C->getValue());\n+\n return Plan.getOrAddLiveIn(V);\n }\n };\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/cost-model.ll", "commands": ["opt < %s -passes=loop-vectorize -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -S"], "tests": [{"test_name": "cost_assume", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i64 @cost_assume(ptr %end, i64 %N) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %0 = phi i64 [ 0, %entry ], [ %1, %loop ]\n %1 = add i64 %0, 1\n %iv.next = add nsw i64 %iv, 1\n %c = icmp ne i64 %N, 0\n tail call void @llvm.assume(i1 %c)\n %gep = getelementptr nusw [9 x i8], ptr null, i64 %iv.next\n %ec = icmp eq ptr %gep, %end\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i64 %1\n}\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "cost_model_1", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\n@c = external global [2048 x i32], align 16\n@b = external global [2048 x i32], align 16\n@d = external global [2048 x i32], align 16\n@a = external global [2048 x i32], align 16\n\n; Function Attrs: noinline nounwind ssp uwtable\ndefine void @cost_model_1() #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]\n %0 = shl nsw i64 %indvars.iv, 1\n %arrayidx = getelementptr inbounds [2048 x i32], ptr @c, i64 0, i64 %0\n %1 = load i32, ptr %arrayidx, align 8\n %idxprom1 = sext i32 %1 to i64\n %arrayidx2 = getelementptr inbounds [2048 x i32], ptr @b, i64 0, i64 %idxprom1\n %2 = load i32, ptr %arrayidx2, align 4\n %arrayidx4 = getelementptr inbounds [2048 x i32], ptr @d, i64 0, i64 %indvars.iv\n %3 = load i32, ptr %arrayidx4, align 4\n %idxprom5 = sext i32 %3 to i64\n %arrayidx6 = getelementptr inbounds [2048 x i32], ptr @a, i64 0, i64 %idxprom5\n store i32 %2, ptr %arrayidx6, align 4\n %indvars.iv.next = add i64 %indvars.iv, 1\n %lftr.wideiv = trunc i64 %indvars.iv.next to i32\n %exitcond = icmp eq i32 %lftr.wideiv, 256\n br i1 %exitcond, label %for.end, label %for.body\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { noinline nounwind ssp uwtable }\n"}, {"test_name": "avx512_cond_load_cost", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i64 @avx512_cond_load_cost(ptr %src, i32 %a, i64 %b, i32 %c, i32 %d) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %c.1 = icmp slt i32 %iv, 0\n br i1 %c.1, label %if.then, label %loop.latch\n\nif.then: ; preds = %loop.header\n %0 = urem i32 %a, %c\n %mul = sub i32 0, %0\n %div = udiv i32 %c, %d\n %or = or i32 %div, %mul\n %ext = sext i32 %or to i64\n %gep = getelementptr { i64, i64, i64 }, ptr %src, i64 %ext, i32 2\n %l = load i64, ptr %gep, align 8\n %or.2 = or i64 %l, %b\n br label %loop.latch\n\nloop.latch: ; preds = %if.then, %loop.header\n %res = phi i64 [ 0, %loop.header ], [ %or.2, %if.then ]\n %iv.next = add i32 %iv, 1\n %ec = icmp ult i32 %iv, %c\n br i1 %ec, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret i64 %res\n}\n\nattributes #0 = { \"target-features\"=\"+avx512bw,+avx512cd,+avx512dq,+avx512f,+avx512vl\" }\n"}, {"test_name": "multi_exit", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine void @multi_exit(ptr %dst, ptr %src.1, ptr %src.2, i64 %A, i64 %B) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop.latch, %entry\n %iv.1.wide = phi i64 [ 0, %entry ], [ %iv.1.next.wide, %loop.latch ]\n %iv.1 = phi i32 [ 0, %entry ], [ %iv.1.next, %loop.latch ]\n %ec.1 = icmp ult i64 %iv.1.wide, %A\n br i1 %ec.1, label %loop.latch, label %exit\n\nloop.latch: ; preds = %loop\n %l.1 = load i64, ptr %src.1, align 8\n %l.2 = load i64, ptr %src.2, align 8\n %cmp55.us = icmp eq i64 %l.1, 0\n %cmp.i.us = icmp ne i64 %l.2, 0\n %and = and i1 %cmp.i.us, %cmp55.us\n %ext = zext i1 %and to i8\n store i8 %ext, ptr %dst, align 1\n %iv.1.next = add i32 %iv.1, 1\n %iv.1.next.wide = zext i32 %iv.1.next to i64\n %ec.2 = icmp ult i64 %iv.1.next.wide, %B\n br i1 %ec.2, label %loop, label %exit\n\nexit: ; preds = %loop.latch, %loop\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"penryn\" }\n"}, {"test_name": "cost_duplicate_recipe_for_sinking", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine void @cost_duplicate_recipe_for_sinking(ptr %A, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %iv.shl = shl nsw i64 %iv, 2\n %gep.0 = getelementptr nusw double, ptr %A, i64 %iv.shl\n %l = load double, ptr %gep.0, align 8\n %c = fcmp oeq double %l, 0.000000e+00\n br i1 %c, label %if.then, label %loop.latch\n\nif.then: ; preds = %loop.header\n %gep.1 = getelementptr double, ptr %A, i64 %iv.shl\n store double 0.000000e+00, ptr %gep.1, align 8\n br label %loop.latch\n\nloop.latch: ; preds = %if.then, %loop.header\n %iv.next = add nsw i64 %iv, 1\n %ec = icmp eq i64 %iv, %N\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"znver3\" }\n"}, {"test_name": "any_of_cost", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i1 @any_of_cost(ptr %start, ptr %end) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %any.of = phi i1 [ false, %entry ], [ %any.of.next, %loop ]\n %ptr.iv = phi ptr [ %start, %entry ], [ %ptr.iv.next, %loop ]\n %gep = getelementptr i8, ptr %ptr.iv, i64 8\n %l = load ptr, ptr %gep, align 8\n %cmp13.not.not = icmp eq ptr %l, null\n %any.of.next = select i1 %cmp13.not.not, i1 %any.of, i1 false\n %ptr.iv.next = getelementptr inbounds i8, ptr %ptr.iv, i64 40\n %cmp.not = icmp eq ptr %ptr.iv, %end\n br i1 %cmp.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i1 %any.of.next\n}\n\nattributes #0 = { \"target-cpu\"=\"penryn\" }\n"}, {"test_name": "live_in_known_1_via_scev", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i64 @live_in_known_1_via_scev() {\nentry:\n %sel = select i1 false, i32 3, i32 0\n br label %ph\n\nph: ; preds = %entry\n %p = phi i32 [ 1, %entry ]\n %N = add nuw nsw i32 %sel, 6\n %p.ext = zext nneg i32 %p to i64\n br label %loop\n\nloop: ; preds = %loop, %ph\n %iv = phi i32 [ 0, %ph ], [ %iv.next, %loop ]\n %red = phi i64 [ 3, %ph ], [ %red.mul, %loop ]\n %red.mul = mul nsw i64 %red, %p.ext\n %iv.next = add nuw nsw i32 %iv, 1\n %ec = icmp eq i32 %iv.next, %N\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n %res = phi i64 [ %red.mul, %loop ]\n ret i64 %res\n}\n"}, {"test_name": "reduction_store", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine void @reduction_store(ptr noalias %src, ptr %dst, i1 %x) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %red = phi i32 [ 0, %entry ], [ %red.next, %loop ]\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src = getelementptr inbounds i32, ptr %src, i32 %iv\n %l = load i32, ptr %gep.src, align 4\n %l.and = and i32 %l, 3\n store i32 %l.and, ptr %dst, align 4\n %x.ext = zext i1 %x to i64\n %lshr = lshr i64 %x.ext, 12\n %t = trunc i64 %lshr to i32\n %red.next = and i32 %red, %t\n store i32 %red.next, ptr %dst, align 4\n %iv.next = add i32 %iv, 1\n %ec = icmp eq i32 %iv, 29\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"znver3\" }\n"}, {"test_name": "PR27826", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine float @PR27826(ptr nocapture readonly %a, ptr nocapture readonly %b, i32 %n) {\nentry:\n %cmp = icmp sgt i32 %n, 0\n br i1 %cmp, label %preheader, label %for.end\n\npreheader: ; preds = %entry\n %t0 = sext i32 %n to i64\n br label %for\n\nfor: ; preds = %for, %preheader\n %indvars.iv = phi i64 [ 0, %preheader ], [ %indvars.iv.next, %for ]\n %s.02 = phi float [ 0.000000e+00, %preheader ], [ %add4, %for ]\n %arrayidx = getelementptr inbounds float, ptr %a, i64 %indvars.iv\n %t1 = load float, ptr %arrayidx, align 4\n %arrayidx3 = getelementptr inbounds float, ptr %b, i64 %indvars.iv\n %t2 = load float, ptr %arrayidx3, align 4\n %add = fadd fast float %t1, %s.02\n %add4 = fadd fast float %add, %t2\n %indvars.iv.next = add nuw nsw i64 %indvars.iv, 32\n %cmp1 = icmp slt i64 %indvars.iv.next, %t0\n br i1 %cmp1, label %for, label %loopexit\n\nloopexit: ; preds = %for\n %add4.lcssa = phi float [ %add4, %for ]\n br label %for.end\n\nfor.end: ; preds = %loopexit, %entry\n %s.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %add4.lcssa, %loopexit ]\n ret float %s.0.lcssa\n}\n"}]}], "issue": {"title": "[clang] Assertion failed in Vectorize", "body": "To reproduce, compile it with -O2: clang -O2 example.c\r\n\r\n```c\r\n#include \r\nint a;\r\nlong b;\r\nstatic int32_t c;\r\nvoid d(int e, int h) {\r\n int f = a = 1;\r\n for (; a <= e; a++)\r\n f *= a;\r\n a = 0;\r\n for (; a < h; a++)\r\n b *= f;\r\n}\r\nlong g() {}\r\nint32_t i() {\r\n for (; g();)\r\n for (c = 3;;)\r\n ;\r\n}\r\nuint16_t j() { d(c + 1, c + 6); }\r\n```\r\n\r\nThe error message is as follows:\r\n\r\n```\r\nllvm/lib/Transforms/Vectorize/LoopVectorize.cpp:9985: bool llvm::LoopVectorizePass::processLoop(llvm::Loop *): Assertion `VF.Width == BestVF && \"VPlan cost model and legacy cost model disagreed\"' failed.\r\n```", "author": "cardigan1008", "labels": ["vectorizers", "confirmed", "crash"], "comments": [{"author": "EugeneZelenko", "body": "See https://godbolt.org/z/jc8qqo8Ta"}, {"author": "shafik", "body": "Looks like a recent regression since it does not crash on trunk: https://godbolt.org/z/sKjdvMq54\r\n\r\nAssertion:\r\n\r\n```console\r\nclang++: /root/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7288: \r\nllvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF():\r\nAssertion `BestFactor.Width == LegacyVF.Width && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\n```\r\n\r\nBacktrace: \r\n\r\n```console\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/clang++ -gdwarf-4 -g -o /app/output.s -mllvm --x86-asm-syntax=intel -fno-verbose-asm -S --gcc-toolchain=/opt/compiler-explorer/gcc-snapshot -fcolor-diagnostics -fno-crash-diagnostics -x c -O2 \r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"\"\r\n4.\tRunning pass \"loop-vectorize\" on function \"j\"\r\n #0 0x0000000003b02648 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x3b02648)\r\n #1 0x0000000003b0033c llvm::sys::CleanupOnSignal(unsigned long) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x3b0033c)\r\n #2 0x0000000003a492b8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\r\n #3 0x000078485ee42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x000078485ee969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #5 0x000078485ee42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #6 0x000078485ee287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #7 0x000078485ee2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #8 0x000078485ee39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #9 0x00000000054df418 llvm::LoopVectorizationPlanner::computeBestVF() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x54df418)\r\n#10 0x00000000054fa07a llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x54fa07a)\r\n#11 0x00000000054fcd39 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x54fcd39)\r\n#12 0x00000000054fd3c3 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x54fd3c3)\r\n#13 0x00000000050ac09e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x50ac09e)\r\n#14 0x00000000034cac68 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x34cac68)\r\n#15 0x00000000010bce6e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x10bce6e)\r\n#16 0x00000000034c966e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x34c966e)\r\n#17 0x00000000010bc75e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x10bc75e)\r\n#18 0x00000000034c7800 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x34c7800)\r\n#19 0x0000000003da8895 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#20 0x0000000003dab9fd clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x3dab9fd)\r\n#21 0x00000000044196ac clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x44196ac)\r\n#22 0x000000000651f80c clang::ParseAST(clang::Sema&, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x651f80c)\r\n#23 0x0000000004419b18 clang::CodeGenAction::ExecuteAction() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x4419b18)\r\n#24 0x00000000046a79f9 clang::FrontendAction::Execute() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x46a79f9)\r\n#25 0x000000000462b30e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x462b30e)\r\n#26 0x000000000479104e clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x479104e)\r\n#27 0x0000000000ca5b0f cc1_main(llvm::ArrayRef, char const*, void*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0xca5b0f)\r\n#28 0x0000000000c9ed5a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#29 0x000000000445d1d9 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\r\n#30 0x0000000003a49764 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x3a49764)\r\n#31 0x000000000445d7cf clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\r\n#32 0x0000000004423365 clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x4423365)\r\n#33 0x0000000004423dcd clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x4423dcd)\r\n#34 0x000000000442b9e5 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0x442b9e5)\r\n#35 0x0000000000ca2f05 clang_main(int, char**, llvm::ToolContext const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0xca2f05)\r\n#36 0x0000000000b776e4 main (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0xb776e4)\r\n#37 0x000078485ee29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#38 0x000078485ee29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#39 0x0000000000c9e80e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+++0xc9e80e)\r\nclang++: error: clang frontend command failed with exit code 134 (use -v to see invocation)\r\nCompiler returned: 134\r\n```"}, {"author": "cardigan1008", "body": "The crash persists in the above cases even with the applied patches. "}, {"author": "fhahn", "body": "@cardigan1008 I can't reproduce the crash locally with latest main. If it still reproduces on current main for you, could you share the crashing IR, using `-mllvm -print-on-crash -mllvm -print-module-scope`?"}, {"author": "cardigan1008", "body": "@fhahn Apologies for the confusion. I pulled the latest version, and this case is now working. However, I encountered a similar crash with another case that I'm currently fuzzing. I'll open a separate issue as soon as I've finished reducing it."}, {"author": "cardigan1008", "body": "Hi @fhahn , I have submitted another issue with the same crash in #106248 . "}]}, "verified": true} {"bug_id": "106248", "issue_url": "https://github.com/llvm/llvm-project/issues/106248", "bug_type": "crash", "base_commit": "616f7d3d4f6d9bea6f776e357c938847e522a681", "knowledge_cutoff": "2024-08-27T17:02:16Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "0a272d3a1703415abca42dc2e2cc2b57cb30734e", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6529, 6534]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInstructionCost"]}}, "patch": "commit 0a272d3a1703415abca42dc2e2cc2b57cb30734e\nAuthor: Florian Hahn \nDate: Thu Aug 29 12:08:27 2024 +0100\n\n [LV] Use SCEV to analyze second operand for cost query.\n \n Improve operand analysis using SCEV for cost purposes. This fixes a\n divergence between legacy and VPlan-based cost-modeling after\n 533e6bbd0d34.\n \n Fixes https://github.com/llvm/llvm-project/issues/106248.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 56f51e14a6eb..4cc75e2e7546 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6529,6 +6529,10 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I,\n // Certain instructions can be cheaper to vectorize if they have a constant\n // second vector operand. One example of this are shifts on x86.\n Value *Op2 = I->getOperand(1);\n+ if (!isa(Op2) && PSE.getSE()->isSCEVable(Op2->getType()) &&\n+ isa(PSE.getSCEV(Op2))) {\n+ Op2 = cast(PSE.getSCEV(Op2))->getValue();\n+ }\n auto Op2Info = TTI.getOperandInfo(Op2);\n if (Op2Info.Kind == TargetTransformInfo::OK_AnyValue &&\n Legal->isInvariant(Op2))\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/cost-constant-known-via-scev.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "test_foldable_live_in_via_scev", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i64 @test_foldable_live_in_via_scev() {\nentry:\n %conv = zext i16 -6 to i64\n %add = add nsw i64 %conv, -65528\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 1, %entry ], [ %iv.next, %loop ]\n %red = phi i64 [ 1, %entry ], [ %mul, %loop ]\n %mul = mul nsw i64 %red, %add\n %iv.next = add nsw i32 %iv, 1\n %ec = icmp eq i32 %iv.next, 100\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n %ret = phi i64 [ %mul, %loop ]\n ret i64 %ret\n}\n"}]}], "issue": {"title": "[clang] Assertion failed in Vectorize", "body": "In #105722 , we resolved the assertion failure in vectorization. However, I met this crash again with the following case:\r\n\r\n```c\r\n#include \r\nint a;\r\nlong b;\r\nstruct e {\r\n uint16_t f;\r\n uint64_t g;\r\n int32_t h\r\n} static i = {65530};\r\nvoid c(int d) {\r\n for (; a; a++)\r\n b *= d;\r\n}\r\nuint32_t j(struct e, uint16_t);\r\nuint64_t k() { j(i, 0); }\r\nuint32_t j(struct e, uint16_t) { c(i.f - 65528); }\r\n```\r\nI compiled it with -O2 flag and it crashed. But when I compiled with -O1/-O0, it worked.\r\n\r\nThe crash is: clang++: /root/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7359: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), BestFactor.Width, CostCtx, OrigLoop, CM)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\n\r\nThe details can be found here: https://godbolt.org/z/8x48xsKs6", "author": "cardigan1008", "labels": ["vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "Thanks for the report @cardigan1008, should be fixed now"}]}, "verified": true} {"bug_id": "126012", "issue_url": "https://github.com/llvm/llvm-project/issues/126012", "bug_type": "miscompilation", "base_commit": "52a02b6d1e0c6b492495ff79a3a06ce93e6180b8", "knowledge_cutoff": "2025-02-06T05:45:23Z", "lit_test_dir": ["llvm/test/Transforms/IndVarSimplify"], "hints": {"fix_commit": "7aed53eb1982113e825534f0f66d0a0e46e7a5ed", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[12400, 12405]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::isImpliedViaMerge"]}}, "patch": "commit 7aed53eb1982113e825534f0f66d0a0e46e7a5ed\nAuthor: Nikita Popov \nDate: Mon Feb 10 10:07:21 2025 +0100\n\n [ScalarEvolution] Handle addrec incoming value in isImpliedViaMerge() (#126236)\n \n The code already guards against values coming from a previous iteration\n using properlyDominates(). However, addrecs are considered to properly\n dominate the loop they are defined in.\n \n Handle this special case separately, by checking for expressions that\n have computable loop evolution (this should cover cases like a zext of\n an addrec as well).\n \n I considered changing the definition of properlyDominates() instead, but\n decided against it. The current definition is useful in other context,\n e.g. when deciding whether an expression is safe to expand in a given\n block.\n \n Fixes https://github.com/llvm/llvm-project/issues/126012.\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex f89887118d8d..46a5c44f4e41 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -12400,6 +12400,12 @@ bool ScalarEvolution::isImpliedViaMerge(CmpPredicate Pred, const SCEV *LHS,\n // iteration of a loop.\n if (!properlyDominates(L, LBB))\n return false;\n+ // Addrecs are considered to properly dominate their loop, so are missed\n+ // by the previous check. Discard any values that have computable\n+ // evolution in this loop.\n+ if (auto *Loop = LI.getLoopFor(LBB))\n+ if (hasComputableLoopEvolution(L, Loop))\n+ return false;\n if (!ProvedEasily(L, RHS))\n return false;\n }\n", "tests": [{"file": "llvm/test/Transforms/IndVarSimplify/pr126012.ll", "commands": ["opt -S -passes=indvars < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test() {\nentry:\n br label %for.preheader\n\nfor.preheader: ; preds = %for.inc, %entry\n %indvar1 = phi i32 [ 0, %entry ], [ %phi, %for.inc ]\n %indvar2 = phi i32 [ 1, %entry ], [ %indvar3, %for.inc ]\n %indvar3 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]\n %cond1 = icmp eq i32 %indvar3, 0\n br i1 %cond1, label %for.inc, label %for.end\n\nfor.end: ; preds = %for.preheader\n %cmp = icmp sgt i32 %indvar2, 0\n %ext = zext i1 %cmp to i32\n br label %for.inc\n\nfor.inc: ; preds = %for.end, %for.preheader\n %phi = phi i32 [ %ext, %for.end ], [ 0, %for.preheader ]\n %inc = add i32 %indvar3, 1\n %exitcond = icmp eq i32 %indvar3, 2\n br i1 %exitcond, label %for.exit, label %for.preheader\n\nfor.exit: ; preds = %for.inc\n ret i32 %indvar1\n}\n", "additional_args": "-src-unroll=4 -tgt-unroll=4"}]}], "issue": {"title": "[IndVarSimplify] Miscompilation at O2", "body": "Reproducer: https://alive2.llvm.org/ce/z/UTnMBe\n```\n; bin/opt -passes=indvars reduced.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@.str = constant [4 x i8] c\"%d\\0A\\00\"\n\ndefine i32 @main() {\nentry:\n br label %for.cond1.preheader.i.i\n\nfor.cond1.preheader.i.i: ; preds = %for.inc18.i.i, %entry\n %.pre.i.i = phi i64 [ 0, %entry ], [ %1, %for.inc18.i.i ]\n %j.sroa.0.043.i.i = phi i64 [ 1, %entry ], [ %storemerge42.i.i, %for.inc18.i.i ]\n %storemerge42.i.i = phi i64 [ 0, %entry ], [ %inc22.i.i, %for.inc18.i.i ]\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i\n\nfor.end.i.i: ; preds = %for.cond1.preheader.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0\n %dec.348.i.i = add i64 %.pre.i.i, 1\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0\n br label %for.inc18.i.i\n\nfor.inc18.i.i: ; preds = %for.end.i.i, %for.cond1.preheader.i.i\n %1 = phi i64 [ %spec.select.i.i, %for.end.i.i ], [ 0, %for.cond1.preheader.i.i ]\n %inc22.i.i = add i64 %storemerge42.i.i, 1\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i\n\ng.exit: ; preds = %for.inc18.i.i\n %conv = trunc i64 %.pre.i.i to i32\n %call1 = call i32 (ptr, ...) @printf(ptr @.str, i32 %conv)\n ret i32 0\n}\n\ndeclare i32 @printf(ptr, ...)\n```\nBefore:\n```\nEntering function main\n br label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 0\n phi i64 %j.sroa.0.043.i.i -> i64 1\n phi i64 %storemerge42.i.i -> i64 0\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> T\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 0\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 1\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 0\n phi i64 %j.sroa.0.043.i.i -> i64 0\n phi i64 %storemerge42.i.i -> i64 1\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> F\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 1\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 0\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 0\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 2\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 0\n phi i64 %j.sroa.0.043.i.i -> i64 1\n phi i64 %storemerge42.i.i -> i64 2\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> T\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 1\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 1\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 1\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 3\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 1\n phi i64 %j.sroa.0.043.i.i -> i64 2\n phi i64 %storemerge42.i.i -> i64 3\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> T\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 2\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 2\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 2\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 4\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 2\n phi i64 %j.sroa.0.043.i.i -> i64 3\n phi i64 %storemerge42.i.i -> i64 4\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> T\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 3\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 3\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 3\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 5\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 3\n phi i64 %j.sroa.0.043.i.i -> i64 4\n phi i64 %storemerge42.i.i -> i64 5\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> T\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 4\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 4\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 4\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 6\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 4\n phi i64 %j.sroa.0.043.i.i -> i64 5\n phi i64 %storemerge42.i.i -> i64 6\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> T\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 5\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 5\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 5\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 7\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 5\n phi i64 %j.sroa.0.043.i.i -> i64 6\n phi i64 %storemerge42.i.i -> i64 7\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> T\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 6\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 6\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 6\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 8\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 6\n phi i64 %j.sroa.0.043.i.i -> i64 7\n phi i64 %storemerge42.i.i -> i64 8\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> T\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 7\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 7\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 7\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 9\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 7\n phi i64 %j.sroa.0.043.i.i -> i64 8\n phi i64 %storemerge42.i.i -> i64 9\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %0 = icmp sgt i64 %j.sroa.0.043.i.i, 0 -> T\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 8\n %spec.select.i.i = select i1 %0, i64 %dec.348.i.i, i64 0 -> i64 8\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %1 -> i64 8\n %inc22.i.i = add i64 %storemerge42.i.i, 1 -> i64 10\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> T\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %g.exit\n %conv = trunc i64 %.pre.i.i to i32 -> i32 7\n %call1 = call i32 (ptr, ...) @printf(ptr @.str, i32 %conv)\n Printf: 7\n -> i32 2\n ret i32 0\nExiting function main\n```\nAfter:\n```\n; ModuleID = 'reduced.ll'\nsource_filename = \"reduced.ll\"\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@.str = constant [4 x i8] c\"%d\\0A\\00\"\n\ndefine i32 @main() {\nentry:\n br label %for.cond1.preheader.i.i\n\nfor.cond1.preheader.i.i: ; preds = %for.inc18.i.i, %entry\n %.pre.i.i = phi i64 [ 0, %entry ], [ %0, %for.inc18.i.i ]\n %storemerge42.i.i = phi i64 [ 0, %entry ], [ %inc22.i.i, %for.inc18.i.i ]\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i\n\nfor.end.i.i: ; preds = %for.cond1.preheader.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0\n br label %for.inc18.i.i\n\nfor.inc18.i.i: ; preds = %for.end.i.i, %for.cond1.preheader.i.i\n %0 = phi i64 [ %spec.select.i.i, %for.end.i.i ], [ 0, %for.cond1.preheader.i.i ]\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i\n\ng.exit: ; preds = %for.inc18.i.i\n %.pre.i.i.lcssa = phi i64 [ %.pre.i.i, %for.inc18.i.i ]\n %conv = trunc i64 %.pre.i.i.lcssa to i32\n %call1 = call i32 (ptr, ...) @printf(ptr @.str, i32 %conv)\n ret i32 0\n}\n\ndeclare i32 @printf(ptr, ...)\n```\n```\nEntering function main\n br label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 0\n phi i64 %storemerge42.i.i -> i64 0\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> T\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 0\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 1\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 0\n phi i64 %storemerge42.i.i -> i64 1\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 1\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 1\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 1\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 2\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 1\n phi i64 %storemerge42.i.i -> i64 2\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 2\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 2\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 2\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 3\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 2\n phi i64 %storemerge42.i.i -> i64 3\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 3\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 3\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 3\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 4\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 3\n phi i64 %storemerge42.i.i -> i64 4\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 4\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 4\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 4\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 5\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 4\n phi i64 %storemerge42.i.i -> i64 5\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 5\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 5\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 5\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 6\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 5\n phi i64 %storemerge42.i.i -> i64 6\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 6\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 6\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 6\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 7\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 6\n phi i64 %storemerge42.i.i -> i64 7\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 7\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 7\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 7\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 8\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 7\n phi i64 %storemerge42.i.i -> i64 8\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 8\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 8\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 8\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 9\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> F\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %for.cond1.preheader.i.i\n phi i64 %.pre.i.i -> i64 8\n phi i64 %storemerge42.i.i -> i64 9\n %tobool.not.i.i = icmp eq i64 %storemerge42.i.i, 0 -> F\n br i1 %tobool.not.i.i, label %for.inc18.i.i, label %for.end.i.i jump to %for.end.i.i\n %dec.348.i.i = add i64 %.pre.i.i, 1 -> i64 9\n %spec.select.i.i = select i1 true, i64 %dec.348.i.i, i64 0 -> i64 9\n br label %for.inc18.i.i jump to %for.inc18.i.i\n phi i64 %0 -> i64 9\n %inc22.i.i = add nuw nsw i64 %storemerge42.i.i, 1 -> i64 10\n %exitcond50.not.i.i = icmp eq i64 %storemerge42.i.i, 9 -> T\n br i1 %exitcond50.not.i.i, label %g.exit, label %for.cond1.preheader.i.i jump to %g.exit\n phi i64 %.pre.i.i.lcssa -> i64 8\n %conv = trunc i64 %.pre.i.i.lcssa to i32 -> i32 8\n %call1 = call i32 (ptr, ...) @printf(ptr @.str, i32 %conv)\n Printf: 8\n -> i32 2\n ret i32 0\nExiting function main\n```\nllvm version: 52fc6ffcda0895c0c7b976ad1f5cb5a282b571d2", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:SCEV"], "comments": [{"author": "dtcxzyw", "body": "Reduced reproducer: https://alive2.llvm.org/ce/z/_x96gs\n```\n; bin/opt -passes=indvars reduced.ll -S --debug\ndefine i32 @src() {\nentry:\n br label %for.preheader\n\nfor.preheader:\n %indvar1 = phi i32 [ 0, %entry ], [ %phi, %for.inc ]\n %indvar2 = phi i32 [ 1, %entry ], [ %indvar3, %for.inc ]\n %indvar3 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]\n %cond1 = icmp eq i32 %indvar3, 0\n br i1 %cond1, label %for.inc, label %for.end\n\nfor.end:\n %cmp = icmp sgt i32 %indvar2, 0\n %ext = zext i1 %cmp to i32\n br label %for.inc\n\nfor.inc:\n %phi = phi i32 [ %ext, %for.end ], [ 0, %for.preheader ]\n %inc = add i32 %indvar3, 1\n %exitcond = icmp eq i32 %indvar3, 2\n br i1 %exitcond, label %for.exit, label %for.preheader\n\nfor.exit:\n ret i32 %indvar1\n}\n\n```\n```\nINDVARS: Eliminated comparison: %cmp = icmp sgt i32 %indvar2, 0\n\ndefine i32 @src() {\nentry:\n br label %for.preheader\n\nfor.preheader: ; preds = %for.inc, %entry\n %indvar1 = phi i32 [ 0, %entry ], [ %phi, %for.inc ]\n %indvar3 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]\n %cond1 = icmp eq i32 %indvar3, 0\n br i1 %cond1, label %for.inc, label %for.end\n\nfor.end: ; preds = %for.preheader\n %ext = zext i1 true to i32\n br label %for.inc\n\nfor.inc: ; preds = %for.end, %for.preheader\n %phi = phi i32 [ %ext, %for.end ], [ 0, %for.preheader ]\n %inc = add nuw nsw i32 %indvar3, 1\n %exitcond = icmp eq i32 %indvar3, 2\n br i1 %exitcond, label %for.exit, label %for.preheader\n\nfor.exit: ; preds = %for.inc\n %indvar1.lcssa = phi i32 [ %indvar1, %for.inc ]\n ret i32 %indvar1.lcssa\n}\n```\n`ScalarEvolution::isImpliedViaMerge(sgt, %indvar2, 0, {0,+,1}<%for.preheader>(= %indvar3), 0)` wrongly returns true.\nI can't quite understand why `%indvar3` properly dominates `%for.preheader`: https://github.com/llvm/llvm-project/blob/26ecddb05d13c101ccd840a6710eb5f8b82de841/llvm/lib/Analysis/ScalarEvolution.cpp#L14131-L14141\n\n@nikic @fhahn Can you help fix this issue?\n\n\n"}, {"author": "nikic", "body": "Related previous fix: https://reviews.llvm.org/D101829"}]}, "verified": true} {"bug_id": "95520", "issue_url": "https://github.com/llvm/llvm-project/issues/95520", "bug_type": "crash", "base_commit": "e7acb3792b8ba971bda2c9564ed26c2a60a2ddbc", "knowledge_cutoff": "2024-06-14T09:26:34Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "40a72f8cc414726a8be234a260650fd62354da21", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlan.cpp": [[231, 236]], "llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[577, 583]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlan.cpp": ["VPTransformState::get"], "llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPInstruction::generatePerPart"]}}, "patch": "commit 40a72f8cc414726a8be234a260650fd62354da21\nAuthor: Florian Hahn \nDate: Fri Jun 14 22:16:51 2024 +0100\n\n [VPlan] Support extracting any lane of uniform value.\n \n If the value we are extracting a lane from is uniform, only the first\n lane will be set. Return lane 0 for any requested lane.\n \n This fixes a crash when trying to extract the last lane for a\n first-order recurrence resume value.\n \n Fixes https://github.com/llvm/llvm-project/issues/95520.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp\nindex f17be451e684..348a2be5072b 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp\n@@ -231,6 +231,11 @@ Value *VPTransformState::get(VPValue *Def, const VPIteration &Instance) {\n return Data\n .PerPartScalars[Def][Instance.Part][Instance.Lane.mapToCacheIndex(VF)];\n }\n+ if (!Instance.Lane.isFirstLane() &&\n+ vputils::isUniformAfterVectorization(Def) &&\n+ hasScalarValue(Def, {Instance.Part, VPLane::getFirstLane()})) {\n+ return Data.PerPartScalars[Def][Instance.Part][0];\n+ }\n \n assert(hasVectorValue(Def, Instance.Part));\n auto *VecPart = Data.PerPartOutput[Def][Instance.Part];\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex 7a482455473e..a3ff6395bb39 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -577,7 +577,8 @@ Value *VPInstruction::generatePerPart(VPTransformState &State, unsigned Part) {\n // When loop is unrolled without vectorizing, retrieve UF - Offset.\n Res = State.get(getOperand(0), State.UF - Offset);\n }\n- Res->setName(Name);\n+ if (isa(Res))\n+ Res->setName(Name);\n return Res;\n }\n case VPInstruction::LogicalAnd: {\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll", "commands": ["opt -p loop-vectorize -S %s", "opt -p loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -S %s"], "tests": [{"test_name": "chained_recurrences", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"arm64-apple-macosx14.0.0\"\n\ndefine i32 @chained_recurrences(i32 %x, i64 %y, ptr %src.1, i32 %z, ptr %src.2) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %0 = phi i32 [ 0, %entry ], [ %3, %loop ]\n %1 = phi i32 [ 0, %entry ], [ %0, %loop ]\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %sum.red = phi i32 [ 0, %entry ], [ %red.2, %loop ]\n %2 = add i64 %y, 1\n %gep.1 = getelementptr i32, ptr %src.1, i64 %2\n %3 = load i32, ptr %gep.1, align 4\n %or3 = or i32 %1, %x\n %iv.next = add i64 %iv, 1\n %shr = lshr i32 %x, 1\n %4 = shl i32 %or3, 1\n %5 = or i32 %4, 2\n %shl19 = shl i32 %x, 1\n %6 = or i32 %shr, %shl19\n %7 = or i32 %6, %5\n %8 = or i32 %7, %x\n %or20 = or i32 %z, %x\n %not = and i32 %or20, 1\n %and = xor i32 %not, 1\n %idx.ext.1 = zext i32 %and to i64\n %gep.2 = getelementptr i32, ptr %src.2, i64 %idx.ext.1\n %9 = load i32, ptr %gep.2, align 4\n %shr24 = lshr i32 %8, 1\n %idx.ext.2 = zext i32 %shr24 to i64\n %gep.3 = getelementptr i32, ptr %src.2, i64 %idx.ext.2\n %10 = load i32, ptr %gep.3, align 4\n %red.1 = or i32 %9, %sum.red\n %red.2 = or i32 %red.1, %10\n %ec = icmp eq i64 %iv, %y\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i32 %red.2\n}\n\nattributes #0 = { \"target-features\"=\"+sve\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/first-order-recurrence-chains.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "test_resinking_required", "test_body": "define double @test_resinking_required(ptr %p, ptr noalias %a, ptr noalias %b) {\nEntry:\n br label %Loop\n\nLoop: ; preds = %Loop, %Entry\n %for.1 = phi double [ %l1, %Loop ], [ 0.000000e+00, %Entry ]\n %for.2 = phi double [ %l2, %Loop ], [ 0.000000e+00, %Entry ]\n %for.3 = phi double [ %for.2, %Loop ], [ 0.000000e+00, %Entry ]\n %iv = phi i64 [ %iv.next, %Loop ], [ 0, %Entry ]\n %USE_2_FORS = fdiv double %for.3, %for.1\n %div = fdiv double 0.000000e+00, %for.1\n %l1 = load double, ptr %a, align 8\n %iv.next = add nuw nsw i64 %iv, 1\n %l2 = load double, ptr %b, align 8\n store double %div, ptr %p, align 8\n %cond = icmp eq i64 %iv.next, 0\n br i1 %cond, label %End, label %Loop\n\nEnd: ; preds = %Loop\n %res.1 = fadd double %for.1, %for.2\n %res.2 = fadd double %res.1, %for.3\n ret double %res.2\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/first-order-recurrence-complex.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "instruction_with_2_FOR_operands_and_multiple_other_uses_chain", "test_body": "define void @instruction_with_2_FOR_operands_and_multiple_other_uses_chain(ptr noalias %dst.1, ptr noalias %dst.2, ptr noalias %dst.3, ptr noalias %for.ptr.1, ptr noalias %for.ptr.2) {\nbb:\n br label %loop\n\nloop: ; preds = %loop, %bb\n %for.1 = phi float [ 0.000000e+00, %bb ], [ %for.1.next, %loop ]\n %for.2 = phi float [ 0.000000e+00, %bb ], [ %for.2.next, %loop ]\n %iv = phi i64 [ 0, %bb ], [ %iv.next, %loop ]\n %for.1.use.1 = fmul fast float %for.1, 2.000000e+00\n %for.1.use.c = fmul fast float %for.1.use.1, 2.000000e+00\n %used.by.both = fmul fast float %for.1.use.c, %for.2\n %for.2.next = load float, ptr %for.ptr.2, align 4\n %for.1.use.3 = fadd fast float %for.1, 1.000000e+00\n %iv.next = add nuw nsw i64 %iv, 1\n %for.1.next = load float, ptr %for.ptr.1, align 4\n %gep.dst.1 = getelementptr inbounds float, ptr %dst.1, i64 %iv\n store float %used.by.both, ptr %gep.dst.1, align 4\n %gep.dst.2 = getelementptr inbounds float, ptr %dst.2, i64 %iv\n store float %for.1.use.1, ptr %gep.dst.2, align 4\n %gep.dst.3 = getelementptr inbounds float, ptr %dst.3, i64 %iv\n store float %for.1.use.3, ptr %gep.dst.3, align 4\n %ec = icmp slt i64 %iv, 1000\n br i1 %ec, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "instruction_with_2_FOR_operands_and_multiple_other_uses", "test_body": "define void @instruction_with_2_FOR_operands_and_multiple_other_uses(ptr noalias %dst.1, ptr noalias %dst.2, ptr noalias %dst.3, ptr noalias %for.ptr.1, ptr noalias %for.ptr.2) {\nbb:\n br label %loop\n\nloop: ; preds = %loop, %bb\n %for.1 = phi float [ 0.000000e+00, %bb ], [ %for.1.next, %loop ]\n %for.2 = phi float [ 0.000000e+00, %bb ], [ %for.2.next, %loop ]\n %iv = phi i64 [ 0, %bb ], [ %iv.next, %loop ]\n %for.1.use.1 = fmul fast float %for.1, 2.000000e+00\n %used.by.both = fmul fast float %for.1, %for.2\n %for.2.next = load float, ptr %for.ptr.2, align 4\n %for.1.use.3 = fadd fast float %for.1, 1.000000e+00\n %iv.next = add nuw nsw i64 %iv, 1\n %for.1.next = load float, ptr %for.ptr.1, align 4\n %gep.dst.1 = getelementptr inbounds float, ptr %dst.1, i64 %iv\n store float %used.by.both, ptr %gep.dst.1, align 4\n %gep.dst.2 = getelementptr inbounds float, ptr %dst.2, i64 %iv\n store float %for.1.use.1, ptr %gep.dst.2, align 4\n %gep.dst.3 = getelementptr inbounds float, ptr %dst.3, i64 %iv\n store float %for.1.use.3, ptr %gep.dst.3, align 4\n %ec = icmp slt i64 %iv, 1000\n br i1 %ec, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "instruction_with_2_FOR_operands", "test_body": "define void @instruction_with_2_FOR_operands(ptr noalias %A, ptr noalias %B, ptr noalias %C) {\nbb:\n br label %bb13\n\nbb13: ; preds = %bb13, %bb\n %tmp37 = phi float [ %tmp60, %bb13 ], [ 0.000000e+00, %bb ]\n %tmp27 = phi float [ %tmp49, %bb13 ], [ 1.000000e+00, %bb ]\n %iv = phi i64 [ %iv.next, %bb13 ], [ 0, %bb ]\n %tmp38 = fmul fast float %tmp37, %tmp27\n %iv.next = add nuw nsw i64 %iv, 1\n %gep = getelementptr inbounds float, ptr %C, i64 %iv\n %tmp49 = load float, ptr %A, align 4\n %tmp60 = load float, ptr %B, align 4\n store float %tmp38, ptr %gep, align 4\n %tmp12 = icmp slt i64 %iv, 1000\n br i1 %tmp12, label %bb13, label %bb74\n\nbb74: ; preds = %bb13\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/first-order-recurrence-multiply-recurrences.ll", "commands": [], "tests": [{"test_name": "test_pr54233_for_depend_on_each_other", "test_body": "define void @test_pr54233_for_depend_on_each_other(ptr noalias %a, ptr noalias %b) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %for.1 = phi i32 [ 0, %entry ], [ %for.1.next, %loop ]\n %for.2 = phi i32 [ 0, %entry ], [ %for.2.next, %loop ]\n %or = or i32 %for.2, 10\n %shl = shl i32 %for.2, %for.1\n %xor = xor i32 %shl, 255\n %and = and i32 %xor, %or\n %for.1.next = xor i32 12, %for.2\n %for.2.next = load i32, ptr %b, align 4\n %a.gep = getelementptr inbounds i32, ptr %a, i64 %iv\n store i32 %and, ptr %a.gep, align 4\n %iv.next = add nuw i64 %iv, 1\n %exitcond = icmp eq i64 %iv, 1000\n br i1 %exitcond, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "test_pr54223_sink_after_insertion_order", "test_body": "define void @test_pr54223_sink_after_insertion_order(ptr noalias %a, ptr noalias %b, ptr noalias %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %for.1 = phi float [ 0.000000e+00, %entry ], [ %for.1.next, %loop ]\n %for.2 = phi float [ 0.000000e+00, %entry ], [ %for.2.next, %loop ]\n %neg = fneg float %for.2\n %muladd = call float @llvm.fmuladd.f32(float %for.1, float %neg, float 0.000000e+00)\n %dst.gep = getelementptr inbounds float, ptr %dst, i64 %iv\n %iv.next = add nuw nsw i64 %iv, 1\n %for.1.next = load float, ptr %a, align 4\n %for.2.next = load float, ptr %b, align 4\n store float %muladd, ptr %dst.gep, align 4\n %exitcond.not = icmp eq i64 %iv.next, 10000\n br i1 %exitcond.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare float @llvm.fmuladd.f32(float, float, float) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll", "commands": ["opt < %s -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=2 -force-widen-divrem-via-safe-divisor=0 -S", "opt < %s -passes=loop-vectorize -force-vector-width=1 -force-vector-interleave=2 -force-widen-divrem-via-safe-divisor=0 -S", "opt < %s -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -force-widen-divrem-via-safe-divisor=0 -S"], "tests": [{"test_name": "recurence_uniform_load", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\n\ndefine i32 @recurence_uniform_load(ptr %src, ptr noalias %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %phi = phi i64 [ 0, %entry ], [ %add, %loop ]\n %recur = phi i32 [ 0, %entry ], [ %load, %loop ]\n %add = add i64 %phi, 1\n %load = load i32, ptr %src, align 4\n %icmp = icmp ult i64 %phi, 1\n br i1 %icmp, label %loop, label %exit\n\nexit: ; preds = %loop\n ret i32 0\n}\n"}]}], "issue": {"title": "Assertion `hasVectorValue(Def, Instance.Part)' failed.", "body": "To reproduce run the following test opt with -passes loop-vectorize:\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i32 @wombat() gc \"statepoint-example\" {\r\nbb:\r\n br label %bb2\r\n\r\nbb1: ; preds = %bb2\r\n ret i32 0\r\n\r\nbb2: ; preds = %bb2, %bb\r\n %phi = phi i64 [ 0, %bb ], [ %add, %bb2 ]\r\n %phi3 = phi i32 [ 0, %bb ], [ %load, %bb2 ]\r\n %add = add i64 %phi, 1\r\n %load = load i32, ptr addrspace(1) null, align 4\r\n %icmp = icmp ult i64 %phi, 1\r\n br i1 %icmp, label %bb2, label %bb1\r\n}\r\n```\r\nReproducer: https://godbolt.org/z/cv7Tf3vba\r\n\r\nStack dump:\r\n```\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes loop-vectorize \r\n #0 0x0000000004e41688 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e41688)\r\n #1 0x0000000004e3eddc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007ec2abe42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007ec2abe969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007ec2abe42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007ec2abe287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007ec2abe2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007ec2abe39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000003fecdce (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3fecdce)\r\n #9 0x0000000004008abe llvm::VPInstruction::generatePerPart(llvm::VPTransformState&, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4008abe)\r\n#10 0x000000000400ae5e llvm::VPInstruction::execute(llvm::VPTransformState&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x400ae5e)\r\n#11 0x0000000003fe5fbf llvm::VPBasicBlock::executeRecipes(llvm::VPTransformState*, llvm::BasicBlock*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3fe5fbf)\r\n#12 0x0000000003fe9cb0 llvm::VPBasicBlock::execute(llvm::VPTransformState*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3fe9cb0)\r\n#13 0x0000000003feda3e llvm::VPlan::execute(llvm::VPTransformState*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3feda3e)\r\n#14 0x0000000003eda4e5 llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3eda4e5)\r\n#15 0x0000000003eef18b llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3eef18b)\r\n#16 0x0000000003ef1a4d llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3ef1a4d)\r\n#17 0x0000000003ef2c5b llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3ef2c5b)\r\n#18 0x0000000002e0679e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2e0679e)\r\n#19 0x0000000004c49a6c llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c49a6c)\r\n#20 0x0000000000dbabee llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdbabee)\r\n#21 0x0000000004c483ae llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c483ae)\r\n#22 0x0000000000dbc65e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdbc65e)\r\n#23 0x0000000004c462bc llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c462bc)\r\n#24 0x00000000008db912 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8db912)\r\n#25 0x00000000008ceabc optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ceabc)\r\n#26 0x00007ec2abe29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#27 0x00007ec2abe29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#28 0x00000000008c5f0e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8c5f0e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\n\r\n", "author": "TatyanaDoubts", "labels": ["vectorizers", "crash-on-valid"], "comments": [{"author": "dtcxzyw", "body": "cc @fhahn "}]}, "verified": true} {"bug_id": "89669", "issue_url": "https://github.com/llvm/llvm-project/issues/89669", "bug_type": "miscompilation", "base_commit": "d97cdd7d088b11b45038112a5f7fe5cbf60a6461", "knowledge_cutoff": "2024-04-22T21:27:36Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "a1b1c4a6d1d52916c5d885170a5f54632d579cdc", "components": ["InstCombine", "ValueTracking"], "bug_location_lineno": {"llvm/include/llvm/Analysis/ValueTracking.h": [[131, 137]], "llvm/lib/Analysis/ValueTracking.cpp": [[8042, 8058]], "llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp": [[320, 326]]}, "bug_location_funcname": {"llvm/lib/Analysis/ValueTracking.cpp": ["llvm::isKnownNegation", "matchMinMax"], "llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp": ["Negator::visitImpl"]}}, "patch": "commit a1b1c4a6d1d52916c5d885170a5f54632d579cdc\nAuthor: Nikita Popov \nDate: Wed Apr 24 10:56:26 2024 +0900\n\n [InstCombine] Fix miscompile in negation of select (#89698)\n \n Swapping the operands of a select is not valid if one hand is more\n poisonous that the other, because the negation zero contains poison\n elements.\n \n Fix this by adding an extra parameter to isKnownNegation() to forbid\n poison elements.\n \n I've implemented this using manual checks to avoid needing four variants\n for the NeedsNSW/AllowPoison combinations. Maybe there is a better way\n to do this...\n \n Fixes https://github.com/llvm/llvm-project/issues/89669.\n\ndiff --git a/llvm/include/llvm/Analysis/ValueTracking.h b/llvm/include/llvm/Analysis/ValueTracking.h\nindex a2fa8f6064e1..571e44cdac26 100644\n--- a/llvm/include/llvm/Analysis/ValueTracking.h\n+++ b/llvm/include/llvm/Analysis/ValueTracking.h\n@@ -131,7 +131,8 @@ bool isKnownNonZero(const Value *V, const SimplifyQuery &Q, unsigned Depth = 0);\n /// Currently can recoginze Value pair:\n /// 1: if X = sub (0, Y) or Y = sub (0, X)\n /// 2: if X = sub (A, B) and Y = sub (B, A)\n-bool isKnownNegation(const Value *X, const Value *Y, bool NeedNSW = false);\n+bool isKnownNegation(const Value *X, const Value *Y, bool NeedNSW = false,\n+ bool AllowPoison = true);\n \n /// Returns true if the give value is known to be non-negative.\n bool isKnownNonNegative(const Value *V, const SimplifyQuery &SQ,\ndiff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp\nindex 21e3f8a4cc52..7501f78ca23b 100644\n--- a/llvm/lib/Analysis/ValueTracking.cpp\n+++ b/llvm/lib/Analysis/ValueTracking.cpp\n@@ -8042,17 +8042,27 @@ static SelectPatternResult matchMinMax(CmpInst::Predicate Pred,\n return {SPF_UNKNOWN, SPNB_NA, false};\n }\n \n-bool llvm::isKnownNegation(const Value *X, const Value *Y, bool NeedNSW) {\n+bool llvm::isKnownNegation(const Value *X, const Value *Y, bool NeedNSW,\n+ bool AllowPoison) {\n assert(X && Y && \"Invalid operand\");\n \n- // X = sub (0, Y) || X = sub nsw (0, Y)\n- if ((!NeedNSW && match(X, m_Sub(m_ZeroInt(), m_Specific(Y)))) ||\n- (NeedNSW && match(X, m_NSWNeg(m_Specific(Y)))))\n+ auto IsNegationOf = [&](const Value *X, const Value *Y) {\n+ if (!match(X, m_Neg(m_Specific(Y))))\n+ return false;\n+\n+ auto *BO = cast(X);\n+ if (NeedNSW && !BO->hasNoSignedWrap())\n+ return false;\n+\n+ auto *Zero = cast(BO->getOperand(0));\n+ if (!AllowPoison && !Zero->isNullValue())\n+ return false;\n+\n return true;\n+ };\n \n- // Y = sub (0, X) || Y = sub nsw (0, X)\n- if ((!NeedNSW && match(Y, m_Sub(m_ZeroInt(), m_Specific(X)))) ||\n- (NeedNSW && match(Y, m_NSWNeg(m_Specific(X)))))\n+ // X = -Y or Y = -X\n+ if (IsNegationOf(X, Y) || IsNegationOf(Y, X))\n return true;\n \n // X = sub (A, B), Y = sub (B, A) || X = sub nsw (A, B), Y = sub nsw (B, A)\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp b/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\nindex d697f361dec0..ed2a98ba4ae4 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\n@@ -320,7 +320,8 @@ std::array Negator::getSortedOperandsOfBinOp(Instruction *I) {\n return NegatedPHI;\n }\n case Instruction::Select: {\n- if (isKnownNegation(I->getOperand(1), I->getOperand(2))) {\n+ if (isKnownNegation(I->getOperand(1), I->getOperand(2), /*NeedNSW=*/false,\n+ /*AllowPoison=*/false)) {\n // Of one hand of select is known to be negation of another hand,\n // just swap the hands around.\n auto *NewSelect = cast(I->clone());\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/sub-of-negatible.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "negate_select_of_negation_poison", "test_body": "define <2 x i32> @negate_select_of_negation_poison(<2 x i1> %c, <2 x i32> %x) {\n %neg = sub <2 x i32> , %x\n %sel = select <2 x i1> %c, <2 x i32> %neg, <2 x i32> %x\n %neg2 = sub <2 x i32> zeroinitializer, %sel\n ret <2 x i32> %neg2\n}\n"}, {"test_name": "dont_negate_ordinary_select", "test_body": "define i8 @dont_negate_ordinary_select(i8 %x, i8 %y, i8 %z, i1 %c) {\n %t0 = select i1 %c, i8 %x, i8 %y\n %t1 = sub i8 %z, %t0\n ret i8 %t1\n}\n"}]}], "issue": {"title": "miscompile of vector double-negation + select by InstCombine", "body": "https://alive2.llvm.org/ce/z/N1RVAr\r\n\r\nthis:\r\n```llvm\r\ndefine <4 x i32> @f(<4 x i32> %0) {\r\n %2 = sub <4 x i32> zeroinitializer, %0\r\n %3 = select <4 x i1> , <4 x i32> %2, <4 x i32> %0\r\n %4 = sub <4 x i32> zeroinitializer, %3\r\n ret <4 x i32> %4\r\n}\r\n```\r\nis being optimized to:\r\n```llvm\r\ndefine <4 x i32> @f(<4 x i32> %0) {\r\n ret <4 x i32> %0\r\n}\r\n```\r\nhere's Alive's work explaining why this is wrong:\r\n```\r\nERROR: Value mismatch\r\n\r\nExample:\r\n<4 x i32> %#0 = < #x00000001 (1), poison, poison, poison >\r\n\r\nSource:\r\n<4 x i32> %#2 = < #xffffffff (4294967295, -1), poison, poison, poison >\r\n<4 x i32> %#3 = < #x00000001 (1), poison, poison, poison >\r\n<4 x i32> %#4 = < #xffffffff (4294967295, -1), poison, poison, poison >\r\n\r\nTarget:\r\nSource value: < #xffffffff (4294967295, -1), poison, poison, poison >\r\nTarget value: < #x00000001 (1), poison, poison, poison >\r\n```\r\n\r\ncc @nunoplopes @hatsunespica\r\n\r\n", "author": "regehr", "labels": ["miscompilation", "llvm:instcombine"], "comments": [{"author": "topperc", "body": "I think the problem is this code in Negator::visitImpl.\r\n\r\n```\r\n case Instruction::Select: {\r\n if (isKnownNegation(I->getOperand(1), I->getOperand(2))) { \r\n // Of one hand of select is known to be negation of another hand, \r\n // just swap the hands around. \r\n auto *NewSelect = cast(I->clone()); \r\n // Just swap the operands of the select. \r\n NewSelect->swapValues(); \r\n // Don't swap prof metadata, we didn't change the branch behavior. \r\n NewSelect->setName(I->getName() + \".neg\"); \r\n Builder.Insert(NewSelect); \r\n return NewSelect; \r\n } \r\n```\r\n\r\nThe IR at the time of that code running was\r\n```\r\n%2 = sub <4 x i32> , %0\r\n%3 = select <4 x i1> , <4 x i32> %2, <4 x i32> %0\r\n%4 = sub <4 x i32> zeroinitializer, %3\r\n```\r\n\r\nSo operand element 0 of %2 is poison. This occurred because element 0 isn't selected so SimplifiedDemandedVectorElts replaced with poison.\r\n\r\nOnce Negator swaps the operands, now element 1 of %2 is never selected so we end up with `sub <4 x i32> poison, %0`\r\n\r\nIt doesn't look like swapping the operands of the select is poison safe.\r\n\r\nCC @nikic "}, {"author": "nikic", "body": "Yeah, that analysis sounds right to me. With that in mind, a more direct test case would be:\r\n```llvm\r\ndefine <2 x i32> @f(<2 x i1> %c, <2 x i32> %x) {\r\n %neg = sub <2 x i32> , %x\r\n %sel = select <2 x i1> %c, <2 x i32> %neg, <2 x i32> %x\r\n %neg2 = sub <2 x i32> zeroinitializer, %sel\r\n ret <2 x i32> %neg2\r\n}\r\n```"}, {"author": "regehr", "body": "off topic but do we want an llvm-reduce pass that shrinks vectors? and maybe also shrinks lane sizes? if so I could work on that sort of thing this summer"}, {"author": "XChy", "body": "> off topic but do we want an llvm-reduce pass that shrinks vectors? and maybe also shrinks lane sizes? if so I could work on that sort of thing this summer\r\n\r\nSounds great to me. I think some SLPVectorizer testcases that involve 16 or even 32 lanes should be further reduced."}]}, "verified": true} {"bug_id": "78049", "issue_url": "https://github.com/llvm/llvm-project/issues/78049", "bug_type": "crash", "base_commit": "de8f782355842a5d3ad44ad8861c53e3ef6fa73a", "knowledge_cutoff": "2024-01-13T16:38:44Z", "lit_test_dir": ["llvm/test/Transforms/MoveAutoInit"], "hints": {"fix_commit": "26d3cd1d07ae3fc8312feb2ab3a3a86414f28c70", "components": ["MoveAutoInit"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/MoveAutoInit.cpp": [[164, 169], [178, 186]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/MoveAutoInit.cpp": ["runMoveAutoInit"]}}, "patch": "commit 26d3cd1d07ae3fc8312feb2ab3a3a86414f28c70\nAuthor: XChy \nDate: Tue Jan 16 18:45:44 2024 +0800\n\n [MoveAutoInit] Ignore unreachable basicblocks and handle catchswitch (#78232)\n \n Fixes #78049\n This patch has done:\n - Ignore unreachable predecessors when looking for nearest common\n dominator.\n - Check catchswitch with `getFirstNonPHI`, instead of\n `getFirstInsertionPt`. The latter skips EHPad.\n\ndiff --git a/llvm/lib/Transforms/Utils/MoveAutoInit.cpp b/llvm/lib/Transforms/Utils/MoveAutoInit.cpp\nindex a977ad87b79f..9a5dba219cee 100644\n--- a/llvm/lib/Transforms/Utils/MoveAutoInit.cpp\n+++ b/llvm/lib/Transforms/Utils/MoveAutoInit.cpp\n@@ -164,6 +164,9 @@ static bool runMoveAutoInit(Function &F, DominatorTree &DT, MemorySSA &MSSA) {\n if (TransitiveSuccessors.count(Pred))\n continue;\n \n+ if (!DT.isReachableFromEntry(Pred))\n+ continue;\n+\n DominatingPredecessor =\n DominatingPredecessor\n ? DT.findNearestCommonDominator(DominatingPredecessor, Pred)\n@@ -178,9 +181,10 @@ static bool runMoveAutoInit(Function &F, DominatorTree &DT, MemorySSA &MSSA) {\n \n // CatchSwitchInst blocks can only have one instruction, so they are not\n // good candidates for insertion.\n- while (isa(UsersDominator->getFirstInsertionPt())) {\n+ while (isa(UsersDominator->getFirstNonPHI())) {\n for (BasicBlock *Pred : predecessors(UsersDominator))\n- UsersDominator = DT.findNearestCommonDominator(UsersDominator, Pred);\n+ if (DT.isReachableFromEntry(Pred))\n+ UsersDominator = DT.findNearestCommonDominator(UsersDominator, Pred);\n }\n \n // We finally found a place where I can be moved while not introducing extra\n", "tests": [{"file": "llvm/test/Transforms/MoveAutoInit/catchswitch.ll", "commands": ["opt < %s -S -passes='move-auto-init' -verify-memoryssa"], "tests": [{"test_name": "test", "test_body": "declare void @dummy()\n\ndeclare void @dummy1()\n\ndefine void @test() personality ptr @dummy {\nentry:\n %p = alloca [2 x i16], i32 0, align 2\n store i32 0, ptr %p, align 2, !annotation !0\n br label %middle\n\nmiddle: ; preds = %entry\n %call = invoke ptr @dummy()\n to label %clean unwind label %catchbb\n\nclean: ; preds = %middle\n ret void\n\ncatchbb: ; preds = %middle\n %cs = catchswitch within none [label %pad, label %pad1] unwind to caller\n\npad: ; preds = %catchbb\n %c = catchpad within %cs [i32 0]\n call void @dummy1()\n ret void\n\npad1: ; preds = %catchbb\n %c1 = catchpad within %cs [i32 0]\n call void @dummy1()\n ret void\n}\n\n!0 = !{!\"auto-init\"}\n"}]}, {"file": "llvm/test/Transforms/MoveAutoInit/loop.ll", "commands": ["opt < %s -S -passes='move-auto-init' -verify-memoryssa"], "tests": [{"test_name": "foo", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @foo(i32 %x) {\nentry:\n %buffer = alloca [80 x i32], align 16\n call void @llvm.memset.p0.i64(ptr align 16 %buffer, i8 -86, i64 320, i1 false), !annotation !0\n br label %do.body\n\ndo.body: ; preds = %do.cond, %entry\n %x.addr.0 = phi i32 [ %x, %entry ], [ %dec, %do.cond ]\n %arrayidx = getelementptr inbounds [80 x i32], ptr %buffer, i64 0, i64 0\n call void @dump(ptr %arrayidx)\n br label %do.cond\n\ndo.cond: ; preds = %do.body\n %dec = add nsw i32 %x.addr.0, -1\n %tobool = icmp ne i32 %x.addr.0, 0\n br i1 %tobool, label %do.body, label %do.end\n\ndo.end: ; preds = %do.cond\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\ndeclare void @dump(ptr)\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\n\n!0 = !{!\"auto-init\"}\n"}, {"test_name": "bar", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\ndeclare void @dump(ptr)\n\ndefine void @bar(i32 %x, i32 %y) {\nentry:\n %buffer = alloca [80 x i32], align 16\n call void @llvm.memset.p0.i64(ptr align 16 %buffer, i8 -86, i64 320, i1 false), !annotation !0\n %tobool = icmp ne i32 %y, 0\n br i1 %tobool, label %if.then, label %if.end\n\nif.then: ; preds = %entry\n %add = add nsw i32 %x, %y\n br label %do.body\n\ndo.body: ; preds = %do.cond, %if.then\n %x.addr.0 = phi i32 [ %add, %if.then ], [ %dec, %do.cond ]\n %arrayidx = getelementptr inbounds [80 x i32], ptr %buffer, i64 0, i64 0\n call void @dump(ptr %arrayidx)\n br label %do.cond\n\ndo.cond: ; preds = %do.body\n %dec = add nsw i32 %x.addr.0, -1\n %tobool1 = icmp ne i32 %x.addr.0, 0\n br i1 %tobool1, label %do.body, label %do.end\n\ndo.end: ; preds = %do.cond\n br label %if.end\n\nif.end: ; preds = %do.end, %entry\n ret void\n}\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\n\n!0 = !{!\"auto-init\"}\n"}]}], "issue": {"title": "[MoveAutoInit] Assertion `NodeB && \"B must be in the tree\"' failed.", "body": "opt -O3 -sroa-skip-mem2reg -enable-dfa-jump-thread tc_dt.ll\r\n[tc_dt.tar.gz](https://github.com/llvm/llvm-project/files/13928836/tc_dt.tar.gz)\r\n\r\nGenericDomTree.h:501: NodeT* llvm::DominatorTreeBase::findNearestCommonDominator(NodeT*, NodeT*) const [with NodeT = llvm::BasicBlock; \r\nbool IsPostDom = false]: Assertion `NodeB && \"B must be in the tree\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\n\r\nReduced test case attached, but have not looked into it. Adding people high in the log...\r\n", "author": "JonPsson1", "labels": ["crash", "llvm:transforms"], "comments": [{"author": "bjope", "body": "Reduced the test case and pipeline a bit more:\r\n\r\nopt -S -o - -passes='dfa-jump-threading,move-auto-init' reduced.ll \r\n\r\nWith reduced.ll being\r\n```\r\ndefine void @func_1() {\r\nentry:\r\n %l_2586 = alloca [2 x i16], i32 0, align 2\r\n store i32 0, ptr %l_2586, align 2, !annotation !0\r\n br label %lbl_2844.preheader\r\n\r\nlbl_2844.preheader: ; preds = %entry\r\n br label %lbl_2844\r\n\r\nlbl_2844: ; preds = %cleanup1473, %lbl_2844.preheader\r\n call void null()\r\n br label %cleanup1473\r\n\r\nfor.end516: ; No predecessors!\r\n br label %cleanup1473\r\n\r\ncleanup1473: ; preds = %for.end516, %lbl_2844\r\n %cleanup.dest.slot.0 = phi i32 [ 0, %for.end516 ], [ 17, %lbl_2844 ]\r\n switch i32 %cleanup.dest.slot.0, label %cleanup1829 [\r\n i32 0, label %if.end1787\r\n i32 17, label %lbl_2844\r\n ]\r\n\r\nif.end1787: ; preds = %cleanup1473\r\n %cmp1789.not = icmp ult ptr null, %l_2586\r\n unreachable\r\n\r\ncleanup1829: ; preds = %cleanup1473\r\n ret void\r\n}\r\n\r\n!0 = !{!\"auto-init\"}\r\n\r\n```"}, {"author": "XChy", "body": "It seems that DFAJumpThreading broke the dominator tree analysis."}, {"author": "XChy", "body": "> It seems that DFAJumpThreading broke the dominator tree analysis.\r\n\r\nIn fact, *MoveAutoInit* should be to blame. This pass doesn't handle unreachable predecessors correctly. A reduced version without `DFAJumpThreading`: \r\n```llvm\r\ndeclare void @dummy()\r\n\r\ndefine void @func_1() {\r\nentry:\r\n %p = alloca [2 x i16], i32 0, align 2\r\n store i32 0, ptr %p, align 2, !annotation !0\r\n br i1 true, label %loop, label %usebb\r\n\r\nloop:\r\n call void @dummy()\r\n br label %loop\r\n\r\na:\r\n br label %loop\r\n\r\nb:\r\n br label %loop\r\n\r\nusebb:\r\n %use_p = icmp ult ptr null, %p\r\n ret void\r\n}\r\n\r\n!0 = !{!\"auto-init\"}\r\n```\r\n\r\nAnd after looking into the source, I found that this pass doesn't handle `catchswitch` well too. For example:\r\n```llvm\r\ndeclare void @dummy()\r\n\r\ndeclare void @dummy1()\r\n\r\ndefine void @func_1() personality ptr @dummy {\r\nentry:\r\n %p = alloca [2 x i16], i32 0, align 2\r\n store i32 0, ptr %p, align 2, !annotation !0\r\n %call = invoke ptr @dummy() to label %clean unwind label %catchbb\r\n\r\nunreachable:\r\n %call1 = invoke ptr @dummy() to label %clean unwind label %catchbb\r\n\r\nclean:\r\n ret void\r\n\r\ncatchbb:\r\n %cs = catchswitch within none [label %pad, label %pad1] unwind to caller\r\n\r\npad:\r\n %c = catchpad within %cs [ptr null, i32 0, ptr null]\r\n call void @dummy1()\r\n ret void\r\n\r\npad1:\r\n %c1 = catchpad within %cs [ptr null, i32 0, ptr null]\r\n call void @dummy1()\r\n ret void\r\n}\r\n\r\n!0 = !{!\"auto-init\"}\r\n```\r\n"}]}, "verified": true} {"bug_id": "106667", "issue_url": "https://github.com/llvm/llvm-project/issues/106667", "bug_type": "crash", "base_commit": "5500e21942f7047344b6fee62d3e08c0ba2f9182", "knowledge_cutoff": "2024-08-30T05:59:16Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "a4aa6bc8fc2130761b8db5db4748059127662785", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[2864, 2869], [6964, 6969], [7095, 7149], [9362, 9383], [12521, 12530], [12532, 12640]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::buildTree_rec", "BoUpSLP::getOperandEntry", "BoUpSLP::vectorizeOperand"]}}, "patch": "commit a4aa6bc8fc2130761b8db5db4748059127662785\nAuthor: Alexey Bataev \nDate: Fri Aug 30 10:17:31 2024 -0700\n\n [SLP]Fix PR106667: carefully look for operand nodes.\n \n If the operand node has the same scalars as one of the vectorized nodes,\n the compiler could miss this and incorrectly request minbitwidth data\n for the wrong node. It may lead to a compiler crash, because the\n vectorized node might have different minbw result.\n \n Fixes https://github.com/llvm/llvm-project/issues/106667\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 4c0a1c4c094b..e9785ef9ded2 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -2864,6 +2864,12 @@ private:\n /// avoid issues with def-use order.\n Value *vectorizeTree(TreeEntry *E, bool PostponedPHIs);\n \n+ TreeEntry *getMatchedVectorizedOperand(const TreeEntry *E, unsigned NodeIdx);\n+ const TreeEntry *getMatchedVectorizedOperand(const TreeEntry *E,\n+ unsigned NodeIdx) const {\n+ return const_cast(this)->getMatchedVectorizedOperand(E, NodeIdx);\n+ }\n+\n /// Vectorize a single entry in the tree, the \\p Idx-th operand of the entry\n /// \\p E.\n /// \\param PostponedPHIs true, if need to postpone emission of phi nodes to\n@@ -6964,6 +6970,55 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n }\n }\n \n+ // Check if this is a duplicate of another entry.\n+ if (TreeEntry *E = getTreeEntry(S.OpValue)) {\n+ LLVM_DEBUG(dbgs() << \"SLP: \\tChecking bundle: \" << *S.OpValue << \".\\n\");\n+ if (!E->isSame(VL)) {\n+ auto It = MultiNodeScalars.find(S.OpValue);\n+ if (It != MultiNodeScalars.end()) {\n+ auto *TEIt = find_if(It->getSecond(),\n+ [&](TreeEntry *ME) { return ME->isSame(VL); });\n+ if (TEIt != It->getSecond().end())\n+ E = *TEIt;\n+ else\n+ E = nullptr;\n+ } else {\n+ E = nullptr;\n+ }\n+ }\n+ if (!E) {\n+ if (!doesNotNeedToBeScheduled(S.OpValue)) {\n+ LLVM_DEBUG(dbgs() << \"SLP: Gathering due to partial overlap.\\n\");\n+ if (TryToFindDuplicates(S))\n+ newTreeEntry(VL, std::nullopt /*not vectorized*/, S, UserTreeIdx,\n+ ReuseShuffleIndices);\n+ return;\n+ }\n+ SmallPtrSet Nodes;\n+ Nodes.insert(getTreeEntry(S.OpValue));\n+ for (const TreeEntry *E : MultiNodeScalars.lookup(S.OpValue))\n+ Nodes.insert(E);\n+ SmallPtrSet Values(VL.begin(), VL.end());\n+ if (any_of(Nodes, [&](const TreeEntry *E) {\n+ return all_of(E->Scalars,\n+ [&](Value *V) { return Values.contains(V); });\n+ })) {\n+ LLVM_DEBUG(dbgs() << \"SLP: Gathering due to full overlap.\\n\");\n+ if (TryToFindDuplicates(S))\n+ newTreeEntry(VL, std::nullopt /*not vectorized*/, S, UserTreeIdx,\n+ ReuseShuffleIndices);\n+ return;\n+ }\n+ } else {\n+ // Record the reuse of the tree node. FIXME, currently this is only used\n+ // to properly draw the graph rather than for the actual vectorization.\n+ E->UserTreeIndices.push_back(UserTreeIdx);\n+ LLVM_DEBUG(dbgs() << \"SLP: Perfect diamond merge at \" << *S.OpValue\n+ << \".\\n\");\n+ return;\n+ }\n+ }\n+\n // Gather if we hit the RecursionMaxDepth, unless this is a load (or z/sext of\n // a load), in which case peek through to include it in the tree, without\n // ballooning over-budget.\n@@ -7095,55 +7150,6 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n // We now know that this is a vector of instructions of the same type from\n // the same block.\n \n- // Check if this is a duplicate of another entry.\n- if (TreeEntry *E = getTreeEntry(S.OpValue)) {\n- LLVM_DEBUG(dbgs() << \"SLP: \\tChecking bundle: \" << *S.OpValue << \".\\n\");\n- if (!E->isSame(VL)) {\n- auto It = MultiNodeScalars.find(S.OpValue);\n- if (It != MultiNodeScalars.end()) {\n- auto *TEIt = find_if(It->getSecond(),\n- [&](TreeEntry *ME) { return ME->isSame(VL); });\n- if (TEIt != It->getSecond().end())\n- E = *TEIt;\n- else\n- E = nullptr;\n- } else {\n- E = nullptr;\n- }\n- }\n- if (!E) {\n- if (!doesNotNeedToBeScheduled(S.OpValue)) {\n- LLVM_DEBUG(dbgs() << \"SLP: Gathering due to partial overlap.\\n\");\n- if (TryToFindDuplicates(S))\n- newTreeEntry(VL, std::nullopt /*not vectorized*/, S, UserTreeIdx,\n- ReuseShuffleIndices);\n- return;\n- }\n- SmallPtrSet Nodes;\n- Nodes.insert(getTreeEntry(S.OpValue));\n- for (const TreeEntry *E : MultiNodeScalars.lookup(S.OpValue))\n- Nodes.insert(E);\n- SmallPtrSet Values(VL.begin(), VL.end());\n- if (any_of(Nodes, [&](const TreeEntry *E) {\n- return all_of(E->Scalars,\n- [&](Value *V) { return Values.contains(V); });\n- })) {\n- LLVM_DEBUG(dbgs() << \"SLP: Gathering due to full overlap.\\n\");\n- if (TryToFindDuplicates(S))\n- newTreeEntry(VL, std::nullopt /*not vectorized*/, S, UserTreeIdx,\n- ReuseShuffleIndices);\n- return;\n- }\n- } else {\n- // Record the reuse of the tree node. FIXME, currently this is only used\n- // to properly draw the graph rather than for the actual vectorization.\n- E->UserTreeIndices.push_back(UserTreeIdx);\n- LLVM_DEBUG(dbgs() << \"SLP: Perfect diamond merge at \" << *S.OpValue\n- << \".\\n\");\n- return;\n- }\n- }\n-\n // Check that none of the instructions in the bundle are already in the tree.\n for (Value *V : VL) {\n if ((!IsScatterVectorizeUserTE && !isa(V)) ||\n@@ -9362,22 +9368,8 @@ public:\n \n const BoUpSLP::TreeEntry *BoUpSLP::getOperandEntry(const TreeEntry *E,\n unsigned Idx) const {\n- Value *Op = E->getOperand(Idx).front();\n- if (const TreeEntry *TE = getTreeEntry(Op)) {\n- if (find_if(TE->UserTreeIndices, [&](const EdgeInfo &EI) {\n- return EI.EdgeIdx == Idx && EI.UserTE == E;\n- }) != TE->UserTreeIndices.end())\n- return TE;\n- auto MIt = MultiNodeScalars.find(Op);\n- if (MIt != MultiNodeScalars.end()) {\n- for (const TreeEntry *TE : MIt->second) {\n- if (find_if(TE->UserTreeIndices, [&](const EdgeInfo &EI) {\n- return EI.EdgeIdx == Idx && EI.UserTE == E;\n- }) != TE->UserTreeIndices.end())\n- return TE;\n- }\n- }\n- }\n+ if (const TreeEntry *VE = getMatchedVectorizedOperand(E, Idx))\n+ return VE;\n const auto *It =\n find_if(VectorizableTree, [&](const std::unique_ptr &TE) {\n return TE->isGather() &&\n@@ -12521,10 +12513,9 @@ public:\n }\n };\n \n-Value *BoUpSLP::vectorizeOperand(TreeEntry *E, unsigned NodeIdx,\n- bool PostponedPHIs) {\n- ValueList &VL = E->getOperand(NodeIdx);\n- const unsigned VF = VL.size();\n+BoUpSLP::TreeEntry *BoUpSLP::getMatchedVectorizedOperand(const TreeEntry *E,\n+ unsigned NodeIdx) {\n+ ArrayRef VL = E->getOperand(NodeIdx);\n InstructionsState S = getSameOpcode(VL, *TLI);\n // Special processing for GEPs bundle, which may include non-gep values.\n if (!S.getOpcode() && VL.front()->getType()->isPointerTy()) {\n@@ -12532,109 +12523,113 @@ Value *BoUpSLP::vectorizeOperand(TreeEntry *E, unsigned NodeIdx,\n if (It != VL.end())\n S = getSameOpcode(*It, *TLI);\n }\n- if (S.getOpcode()) {\n- auto CheckSameVE = [&](const TreeEntry *VE) {\n- return VE->isSame(VL) &&\n- (any_of(VE->UserTreeIndices,\n- [E, NodeIdx](const EdgeInfo &EI) {\n- return EI.UserTE == E && EI.EdgeIdx == NodeIdx;\n- }) ||\n- any_of(VectorizableTree,\n- [E, NodeIdx, VE](const std::unique_ptr &TE) {\n- return TE->isOperandGatherNode({E, NodeIdx}) &&\n- VE->isSame(TE->Scalars);\n- }));\n+ if (!S.getOpcode())\n+ return nullptr;\n+ auto CheckSameVE = [&](const TreeEntry *VE) {\n+ return VE->isSame(VL) &&\n+ (any_of(VE->UserTreeIndices,\n+ [E, NodeIdx](const EdgeInfo &EI) {\n+ return EI.UserTE == E && EI.EdgeIdx == NodeIdx;\n+ }) ||\n+ any_of(VectorizableTree,\n+ [E, NodeIdx, VE](const std::unique_ptr &TE) {\n+ return TE->isOperandGatherNode(\n+ {const_cast(E), NodeIdx}) &&\n+ VE->isSame(TE->Scalars);\n+ }));\n+ };\n+ TreeEntry *VE = getTreeEntry(S.OpValue);\n+ if (VE && CheckSameVE(VE))\n+ return VE;\n+ auto It = MultiNodeScalars.find(S.OpValue);\n+ if (It != MultiNodeScalars.end()) {\n+ auto *I = find_if(It->getSecond(), [&](const TreeEntry *TE) {\n+ return TE != VE && CheckSameVE(TE);\n+ });\n+ if (I != It->getSecond().end())\n+ return *I;\n+ }\n+ return nullptr;\n+}\n+\n+Value *BoUpSLP::vectorizeOperand(TreeEntry *E, unsigned NodeIdx,\n+ bool PostponedPHIs) {\n+ ValueList &VL = E->getOperand(NodeIdx);\n+ const unsigned VF = VL.size();\n+ if (TreeEntry *VE = getMatchedVectorizedOperand(E, NodeIdx)) {\n+ auto FinalShuffle = [&](Value *V, ArrayRef Mask) {\n+ // V may be affected by MinBWs.\n+ // We want ShuffleInstructionBuilder to correctly support REVEC. The key\n+ // factor is the number of elements, not their type.\n+ Type *ScalarTy = cast(V->getType())->getElementType();\n+ unsigned NumElements = getNumElements(VL.front()->getType());\n+ ShuffleInstructionBuilder ShuffleBuilder(\n+ NumElements != 1 ? FixedVectorType::get(ScalarTy, NumElements)\n+ : ScalarTy,\n+ Builder, *this);\n+ ShuffleBuilder.add(V, Mask);\n+ SmallVector> SubVectors(\n+ E->CombinedEntriesWithIndices.size());\n+ transform(E->CombinedEntriesWithIndices, SubVectors.begin(),\n+ [&](const auto &P) {\n+ return std::make_pair(VectorizableTree[P.first].get(),\n+ P.second);\n+ });\n+ return ShuffleBuilder.finalize(std::nullopt, SubVectors);\n };\n- TreeEntry *VE = getTreeEntry(S.OpValue);\n- bool IsSameVE = VE && CheckSameVE(VE);\n- if (!IsSameVE) {\n- auto It = MultiNodeScalars.find(S.OpValue);\n- if (It != MultiNodeScalars.end()) {\n- auto *I = find_if(It->getSecond(), [&](const TreeEntry *TE) {\n- return TE != VE && CheckSameVE(TE);\n- });\n- if (I != It->getSecond().end()) {\n- VE = *I;\n- IsSameVE = true;\n- }\n- }\n- }\n- if (IsSameVE) {\n- auto FinalShuffle = [&](Value *V, ArrayRef Mask) {\n- // V may be affected by MinBWs.\n- // We want ShuffleInstructionBuilder to correctly support REVEC. The key\n- // factor is the number of elements, not their type.\n- Type *ScalarTy = cast(V->getType())->getElementType();\n- unsigned NumElements = getNumElements(VL.front()->getType());\n- ShuffleInstructionBuilder ShuffleBuilder(\n- NumElements != 1 ? FixedVectorType::get(ScalarTy, NumElements)\n- : ScalarTy,\n- Builder, *this);\n- ShuffleBuilder.add(V, Mask);\n- SmallVector> SubVectors(\n- E->CombinedEntriesWithIndices.size());\n- transform(E->CombinedEntriesWithIndices, SubVectors.begin(),\n- [&](const auto &P) {\n- return std::make_pair(VectorizableTree[P.first].get(),\n- P.second);\n- });\n- return ShuffleBuilder.finalize(std::nullopt, SubVectors);\n- };\n- Value *V = vectorizeTree(VE, PostponedPHIs);\n- if (VF * getNumElements(VL[0]->getType()) !=\n- cast(V->getType())->getNumElements()) {\n- if (!VE->ReuseShuffleIndices.empty()) {\n- // Reshuffle to get only unique values.\n- // If some of the scalars are duplicated in the vectorization\n- // tree entry, we do not vectorize them but instead generate a\n- // mask for the reuses. But if there are several users of the\n- // same entry, they may have different vectorization factors.\n- // This is especially important for PHI nodes. In this case, we\n- // need to adapt the resulting instruction for the user\n- // vectorization factor and have to reshuffle it again to take\n- // only unique elements of the vector. Without this code the\n- // function incorrectly returns reduced vector instruction with\n- // the same elements, not with the unique ones.\n-\n- // block:\n- // %phi = phi <2 x > { .., %entry} {%shuffle, %block}\n- // %2 = shuffle <2 x > %phi, poison, <4 x > <1, 1, 0, 0>\n- // ... (use %2)\n- // %shuffle = shuffle <2 x> %2, poison, <2 x> {2, 0}\n- // br %block\n- SmallVector Mask(VF, PoisonMaskElem);\n- for (auto [I, V] : enumerate(VL)) {\n- if (isa(V))\n- continue;\n- Mask[I] = VE->findLaneForValue(V);\n- }\n- V = FinalShuffle(V, Mask);\n- } else {\n- assert(VF < cast(V->getType())->getNumElements() &&\n- \"Expected vectorization factor less \"\n- \"than original vector size.\");\n- SmallVector UniformMask(VF, 0);\n- std::iota(UniformMask.begin(), UniformMask.end(), 0);\n- V = FinalShuffle(V, UniformMask);\n+ Value *V = vectorizeTree(VE, PostponedPHIs);\n+ if (VF * getNumElements(VL[0]->getType()) !=\n+ cast(V->getType())->getNumElements()) {\n+ if (!VE->ReuseShuffleIndices.empty()) {\n+ // Reshuffle to get only unique values.\n+ // If some of the scalars are duplicated in the vectorization\n+ // tree entry, we do not vectorize them but instead generate a\n+ // mask for the reuses. But if there are several users of the\n+ // same entry, they may have different vectorization factors.\n+ // This is especially important for PHI nodes. In this case, we\n+ // need to adapt the resulting instruction for the user\n+ // vectorization factor and have to reshuffle it again to take\n+ // only unique elements of the vector. Without this code the\n+ // function incorrectly returns reduced vector instruction with\n+ // the same elements, not with the unique ones.\n+\n+ // block:\n+ // %phi = phi <2 x > { .., %entry} {%shuffle, %block}\n+ // %2 = shuffle <2 x > %phi, poison, <4 x > <1, 1, 0, 0>\n+ // ... (use %2)\n+ // %shuffle = shuffle <2 x> %2, poison, <2 x> {2, 0}\n+ // br %block\n+ SmallVector Mask(VF, PoisonMaskElem);\n+ for (auto [I, V] : enumerate(VL)) {\n+ if (isa(V))\n+ continue;\n+ Mask[I] = VE->findLaneForValue(V);\n }\n- }\n- // Need to update the operand gather node, if actually the operand is not a\n- // vectorized node, but the buildvector/gather node, which matches one of\n- // the vectorized nodes.\n- if (find_if(VE->UserTreeIndices, [&](const EdgeInfo &EI) {\n- return EI.UserTE == E && EI.EdgeIdx == NodeIdx;\n- }) == VE->UserTreeIndices.end()) {\n- auto *It = find_if(\n- VectorizableTree, [&](const std::unique_ptr &TE) {\n- return TE->isGather() &&\n- TE->UserTreeIndices.front().UserTE == E &&\n- TE->UserTreeIndices.front().EdgeIdx == NodeIdx;\n- });\n- assert(It != VectorizableTree.end() && \"Expected gather node operand.\");\n- (*It)->VectorizedValue = V;\n- }\n- return V;\n+ V = FinalShuffle(V, Mask);\n+ } else {\n+ assert(VF < cast(V->getType())->getNumElements() &&\n+ \"Expected vectorization factor less \"\n+ \"than original vector size.\");\n+ SmallVector UniformMask(VF, 0);\n+ std::iota(UniformMask.begin(), UniformMask.end(), 0);\n+ V = FinalShuffle(V, UniformMask);\n+ }\n+ }\n+ // Need to update the operand gather node, if actually the operand is not a\n+ // vectorized node, but the buildvector/gather node, which matches one of\n+ // the vectorized nodes.\n+ if (find_if(VE->UserTreeIndices, [&](const EdgeInfo &EI) {\n+ return EI.UserTE == E && EI.EdgeIdx == NodeIdx;\n+ }) == VE->UserTreeIndices.end()) {\n+ auto *It =\n+ find_if(VectorizableTree, [&](const std::unique_ptr &TE) {\n+ return TE->isGather() && TE->UserTreeIndices.front().UserTE == E &&\n+ TE->UserTreeIndices.front().EdgeIdx == NodeIdx;\n+ });\n+ assert(It != VectorizableTree.end() && \"Expected gather node operand.\");\n+ (*It)->VectorizedValue = V;\n }\n+ return V;\n }\n \n // Find the corresponding gather entry and vectorize it.\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/multi-nodes-bv-vectorized.ll", "commands": ["opt -passes=slp-vectorizer -S -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define void @test(ptr %p) {\nentry:\n %conv548.2.i.13 = zext i32 0 to i64\n %and551.2.i.13 = and i64 0, %conv548.2.i.13\n %conv548.3.i.13 = zext i32 0 to i64\n %and551.3.i.13 = and i64 0, %conv548.3.i.13\n %0 = trunc i64 %and551.2.i.13 to i32\n %conv54.2.i.14 = and i32 %0, 0\n %conv548.2.i.14 = zext i32 %conv54.2.i.14 to i64\n %and551.2.i.14 = and i64 %and551.2.i.13, %conv548.2.i.14\n %1 = trunc i64 %and551.3.i.13 to i32\n %conv54.3.i.14 = and i32 %1, 0\n %conv548.3.i.14 = zext i32 %conv54.3.i.14 to i64\n %and551.3.i.14 = and i64 %and551.3.i.13, %conv548.3.i.14\n %and551.2.i.15 = and i64 %and551.2.i.14, 0\n %and551.3.i.15 = and i64 %and551.3.i.14, 0\n %and551.2.i.16 = and i64 %and551.2.i.15, 0\n %and551.3.i.16 = and i64 %and551.3.i.15, 0\n %and551.2.i.17 = and i64 %and551.2.i.16, 0\n %and551.3.i.17 = and i64 %and551.3.i.16, 0\n %and551.2.i.18 = and i64 %and551.2.i.17, 0\n %and551.3.i.18 = and i64 %and551.3.i.17, 0\n %and551.2.i.19 = and i64 %and551.2.i.18, 0\n %and551.3.i.19 = and i64 %and551.3.i.18, 0\n %and551.2.i.20 = and i64 %and551.2.i.19, 0\n %and551.3.i.20 = and i64 %and551.3.i.19, 0\n %and551.2.i.21 = and i64 %and551.2.i.20, 0\n %and551.3.i.21 = and i64 %and551.3.i.20, 0\n %gep1 = getelementptr inbounds i8, ptr %p, i64 16\n %gep2 = getelementptr inbounds i8, ptr %p, i64 24\n store i64 %and551.2.i.21, ptr %gep1, align 16\n store i64 %and551.3.i.21, ptr %gep2, align 8\n ret void\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Assertion `CastInst::castIsValid(opc, C, Ty) && \"Invalid constantexpr cast!\"' failed.", "body": "Reproducer: https://godbolt.org/z/PeoMv3Tsn\r\n```\r\n; bin/opt -passes=slp-vectorizer reduced.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @main(ptr %p) {\r\nentry:\r\n %conv548.2.i.13 = zext i32 0 to i64\r\n %and551.2.i.13 = and i64 0, %conv548.2.i.13\r\n %conv548.3.i.13 = zext i32 0 to i64\r\n %and551.3.i.13 = and i64 0, %conv548.3.i.13\r\n %0 = trunc i64 %and551.2.i.13 to i32\r\n %conv54.2.i.14 = and i32 %0, 0\r\n %conv548.2.i.14 = zext i32 %conv54.2.i.14 to i64\r\n %and551.2.i.14 = and i64 %and551.2.i.13, %conv548.2.i.14\r\n %1 = trunc i64 %and551.3.i.13 to i32\r\n %conv54.3.i.14 = and i32 %1, 0\r\n %conv548.3.i.14 = zext i32 %conv54.3.i.14 to i64\r\n %and551.3.i.14 = and i64 %and551.3.i.13, %conv548.3.i.14\r\n %and551.2.i.15 = and i64 %and551.2.i.14, 0\r\n %and551.3.i.15 = and i64 %and551.3.i.14, 0\r\n %and551.2.i.16 = and i64 %and551.2.i.15, 0\r\n %and551.3.i.16 = and i64 %and551.3.i.15, 0\r\n %and551.2.i.17 = and i64 %and551.2.i.16, 0\r\n %and551.3.i.17 = and i64 %and551.3.i.16, 0\r\n %and551.2.i.18 = and i64 %and551.2.i.17, 0\r\n %and551.3.i.18 = and i64 %and551.3.i.17, 0\r\n %and551.2.i.19 = and i64 %and551.2.i.18, 0\r\n %and551.3.i.19 = and i64 %and551.3.i.18, 0\r\n %and551.2.i.20 = and i64 %and551.2.i.19, 0\r\n %and551.3.i.20 = and i64 %and551.3.i.19, 0\r\n %and551.2.i.21 = and i64 %and551.2.i.20, 0\r\n %and551.3.i.21 = and i64 %and551.3.i.20, 0\r\n %gep1 = getelementptr inbounds i8, ptr %p, i64 16\r\n %gep2 = getelementptr inbounds i8, ptr %p, i64 24\r\n store i64 %and551.2.i.21, ptr %gep1, align 16\r\n store i64 %and551.3.i.21, ptr %gep2, align 8\r\n ret void\r\n}\r\n```\r\n```\r\nopt: /home/dtcxzyw/WorkSpace/Projects/compilers/llvm-project/llvm/lib/IR/Constants.cpp:2202: static llvm::Constant* llvm::ConstantExpr::getCast(unsigned int, llvm::Constant*, llvm::Type*, bool): Assertion `CastInst::castIsValid(opc, C, Ty) && \"Invalid constantexpr cast!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: bin/opt -passes=slp-vectorizer reduced.ll -S\r\n1. Running pass \"function(slp-vectorizer)\" on module \"reduced.ll\"\r\n2. Running pass \"slp-vectorizer\" on function \"main\"\r\n #0 0x00007c5d3e8138d2 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.20.0git+0x2138d2)\r\n #1 0x00007c5d3e81079f llvm::sys::RunSignalHandlers() (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.20.0git+0x21079f)\r\n #2 0x00007c5d3e8108e5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007c5d3e242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007c5d3e2969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007c5d3e2969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007c5d3e2969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007c5d3e242476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007c5d3e2287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007c5d3e22871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007c5d3e239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00007c5d36fb0c4d llvm::ConstantExpr::getCast(unsigned int, llvm::Constant*, llvm::Type*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x1b0c4d)\r\n#12 0x00007c5d394573b5 llvm::IRBuilderBase::CreateCast(llvm::Instruction::CastOps, llvm::Value*, llvm::Type*, llvm::Twine const&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x573b5)\r\n#13 0x00007c5d395ab173 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1ab173)\r\n#14 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#15 0x00007c5d395a7a13 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a13)\r\n#16 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#17 0x00007c5d395a7cbb llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7cbb)\r\n#18 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#19 0x00007c5d395a7a41 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a41)\r\n#20 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#21 0x00007c5d395a7a13 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a13)\r\n#22 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#23 0x00007c5d395a7a13 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a13)\r\n#24 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#25 0x00007c5d395a7a13 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a13)\r\n#26 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#27 0x00007c5d395a7a13 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a13)\r\n#28 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#29 0x00007c5d395a7a13 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a13)\r\n#30 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#31 0x00007c5d395a7a13 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a13)\r\n#32 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#33 0x00007c5d395a7a13 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a7a13)\r\n#34 0x00007c5d395acf1e llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1acf1e)\r\n#35 0x00007c5d395a8f65 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1a8f65)\r\n#36 0x00007c5d395ae278 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::MapVector, llvm::DenseMap, llvm::detail::DenseMapPair>, llvm::SmallVector>, 0u>> const&, llvm::SmallVectorImpl>&, llvm::Instruction*) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1ae278)\r\n#37 0x00007c5d395b207c llvm::slpvectorizer::BoUpSLP::vectorizeTree() (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1b207c)\r\n#38 0x00007c5d395c7cac llvm::SLPVectorizerPass::vectorizeStoreChain(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, unsigned int, unsigned int, unsigned int&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1c7cac)\r\n#39 0x00007c5d395c9637 llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::'lambda'(std::set, llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::StoreDistCompare, std::allocator>> const&)::operator()(std::set, llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&)::StoreDistCompare, std::allocator>> const&) const SLPVectorizer.cpp:0:0\r\n#40 0x00007c5d395cb92c llvm::SLPVectorizerPass::vectorizeStores(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, llvm::DenseSet, llvm::DenseMapInfo, void>>&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1cb92c)\r\n#41 0x00007c5d395cc20f llvm::SLPVectorizerPass::vectorizeStoreChains(llvm::slpvectorizer::BoUpSLP&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1cc20f)\r\n#42 0x00007c5d395cd30f llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1cd30f)\r\n#43 0x00007c5d395cdf86 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1cdf86)\r\n#44 0x00007c5d39ed4015 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMPasses.so.20.0git+0xd4015)\r\n#45 0x00007c5d371546ed llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x3546ed)\r\n#46 0x00007c5d3d4d7f35 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.20.0git+0xd7f35)\r\n#47 0x00007c5d37153256 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x353256)\r\n#48 0x00007c5d3d4d88f5 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.20.0git+0xd88f5)\r\n#49 0x00007c5d3715121a llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x35121a)\r\n#50 0x00007c5d3eb12fc7 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.20.0git+0x2cfc7)\r\n#51 0x00007c5d3eb1e962 optMain (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.20.0git+0x38962)\r\n#52 0x00007c5d3e229d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#53 0x00007c5d3e229e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#54 0x00007c5d3e229e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#55 0x00005949b291d095 _start (bin/opt+0x1095)\r\nAborted (core dumped)\r\n```\r\n\r\ncc @alexey-bataev ", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": [{"author": "dtcxzyw", "body": "Related PR: https://github.com/llvm/llvm-project/pull/103216\r\n"}]}, "verified": true} {"bug_id": "108698", "issue_url": "https://github.com/llvm/llvm-project/issues/108698", "bug_type": "miscompilation", "base_commit": "ba8e4246e2f17030788e8a4954bf5c290332206f", "knowledge_cutoff": "2024-09-14T13:16:53Z", "lit_test_dir": ["llvm/test/Transforms/VectorCombine"], "hints": {"fix_commit": "87663fdab9d0e7bcc0b963ea078da9e2eb574908", "components": ["VectorCombine"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": [[2597, 2607], [2628, 2634]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": ["VectorCombine::shrinkType"]}}, "patch": "commit 87663fdab9d0e7bcc0b963ea078da9e2eb574908\nAuthor: Yingwei Zheng \nDate: Sun Sep 15 18:38:06 2024 +0800\n\n [VectorCombine] Don't shrink lshr if the shamt is not less than bitwidth (#108705)\n \n Consider the following case:\n ```\n define <2 x i32> @test(<2 x i64> %vec.ind16, <2 x i32> %broadcast.splat20) {\n %19 = icmp eq <2 x i64> %vec.ind16, zeroinitializer\n %20 = zext <2 x i1> %19 to <2 x i32>\n %21 = lshr <2 x i32> %20, %broadcast.splat20\n ret <2 x i32> %21\n }\n ```\n After https://github.com/llvm/llvm-project/pull/104606, we shrink the\n lshr into:\n ```\n define <2 x i32> @test(<2 x i64> %vec.ind16, <2 x i32> %broadcast.splat20) {\n %1 = icmp eq <2 x i64> %vec.ind16, zeroinitializer\n %2 = trunc <2 x i32> %broadcast.splat20 to <2 x i1>\n %3 = lshr <2 x i1> %1, %2\n %4 = zext <2 x i1> %3 to <2 x i32>\n ret <2 x i32> %4\n }\n ```\n It is incorrect since `lshr i1 X, 1` returns `poison`.\n This patch adds additional check on the shamt operand. The lshr will get\n shrunk iff we ensure that the shamt is less than bitwidth of the smaller\n type. As `computeKnownBits(&I, *DL).countMaxActiveBits() > BW` always\n evaluates to true for `lshr(zext(X), Y)`, this check will only apply to\n bitwise logical instructions.\n \n Alive2: https://alive2.llvm.org/ce/z/j_RmTa\n Fixes https://github.com/llvm/llvm-project/issues/108698.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\nindex d7afe2f426d3..58701bfa60a3 100644\n--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n@@ -2597,11 +2597,19 @@ bool VectorCombine::shrinkType(llvm::Instruction &I) {\n auto *SmallTy = cast(ZExted->getType());\n unsigned BW = SmallTy->getElementType()->getPrimitiveSizeInBits();\n \n- // Check that the expression overall uses at most the same number of bits as\n- // ZExted\n- KnownBits KB = computeKnownBits(&I, *DL);\n- if (KB.countMaxActiveBits() > BW)\n- return false;\n+ if (I.getOpcode() == Instruction::LShr) {\n+ // Check that the shift amount is less than the number of bits in the\n+ // smaller type. Otherwise, the smaller lshr will return a poison value.\n+ KnownBits ShAmtKB = computeKnownBits(I.getOperand(1), *DL);\n+ if (ShAmtKB.getMaxValue().uge(BW))\n+ return false;\n+ } else {\n+ // Check that the expression overall uses at most the same number of bits as\n+ // ZExted\n+ KnownBits KB = computeKnownBits(&I, *DL);\n+ if (KB.countMaxActiveBits() > BW)\n+ return false;\n+ }\n \n // Calculate costs of leaving current IR as it is and moving ZExt operation\n // later, along with adding truncates if needed\n@@ -2628,7 +2636,7 @@ bool VectorCombine::shrinkType(llvm::Instruction &I) {\n return false;\n \n // Check if we can propagate ZExt through its other users\n- KB = computeKnownBits(UI, *DL);\n+ KnownBits KB = computeKnownBits(UI, *DL);\n if (KB.countMaxActiveBits() > BW)\n return false;\n \n", "tests": [{"file": "llvm/test/Transforms/VectorCombine/AArch64/shrink-types.ll", "commands": ["opt -passes=vector-combine -S %s"], "tests": [{"test_name": "pr108698", "test_body": "target triple = \"aarch64\"\n\ndefine <2 x i32> @pr108698(<2 x i64> %x, <2 x i32> %y) {\n %cmp = icmp eq <2 x i64> %x, zeroinitializer\n %ext = zext <2 x i1> %cmp to <2 x i32>\n %lshr = lshr <2 x i32> %ext, %y\n ret <2 x i32> %lshr\n}\n"}]}], "issue": {"title": "wrong code at -O{2,3} on x86_64-linux-gnu", "body": "It appears to be a recent regression as it doesn't reproduce with 18.1.0 and earlier. \r\n\r\nCompiler Explorer: https://godbolt.org/z/EGasYh8x1\r\n\r\n```\r\n[691] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 223e2efa5e886502a9467b7ef700ebce9b7886e8)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[692] % \r\n[692] % clangtk -O1 small.c; ./a.out\r\n[693] % \r\n[693] % clangtk -O2 small.c\r\n[694] % ./a.out\r\nAborted\r\n[695] % \r\n[695] % cat small.c\r\nchar a[8];\r\nint b = 1, c;\r\nint main() {\r\n for (; c < 8; c++)\r\n a[c] = !c >> b;\r\n if (a[0] != 0)\r\n __builtin_abort();\r\n return 0;\r\n}\r\n```", "author": "zhendongsu", "labels": ["miscompilation", "vectorizers"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/5SkPZS\r\n```\r\n; bin/opt -passes=vector-combine test.ll -S\r\n\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine <2 x i32> @test(<2 x i64> %vec.ind16, <2 x i32> %broadcast.splat20) {\r\n %19 = icmp eq <2 x i64> %vec.ind16, zeroinitializer\r\n %20 = zext <2 x i1> %19 to <2 x i32>\r\n %21 = lshr <2 x i32> %20, %broadcast.splat20\r\n ret <2 x i32> %21\r\n}\r\n```\r\n```\r\ndefine <2 x i32> @test(<2 x i64> %vec.ind16, <2 x i32> %broadcast.splat20) {\r\n %1 = icmp eq <2 x i64> %vec.ind16, zeroinitializer\r\n %2 = trunc <2 x i32> %broadcast.splat20 to <2 x i1>\r\n %3 = lshr <2 x i1> %1, %2\r\n %4 = zext <2 x i1> %3 to <2 x i32>\r\n ret <2 x i32> %4\r\n}\r\n```"}, {"author": "dtcxzyw", "body": "Related PR: https://github.com/llvm/llvm-project/pull/104606\r\n"}]}, "verified": true} {"bug_id": "81561", "issue_url": "https://github.com/llvm/llvm-project/issues/81561", "bug_type": "crash", "base_commit": "e79ad7bb94611666a23764459098574bc0394d56", "knowledge_cutoff": "2024-02-13T01:36:19Z", "lit_test_dir": ["llvm/test/DebugInfo/Generic"], "hints": {"fix_commit": "97088b2ab2184ad4bd64f59fba0b92b70468b10d", "components": ["ValueMapper"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/ValueMapper.cpp": [[552, 557]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/ValueMapper.cpp": ["Mapper::remapDPValue"]}}, "patch": "commit 97088b2ab2184ad4bd64f59fba0b92b70468b10d\nAuthor: Orlando Cazalet-Hyams \nDate: Tue Feb 13 11:35:28 2024 +0000\n\n [RemoveDIs][ValueMapper] Remap DIAssignIDs in DPValues (#81595)\n \n Fix crash raised in comments for 5c9f7682b090124d9a8b69f92d3f7c269dca25fc\n\ndiff --git a/llvm/lib/Transforms/Utils/ValueMapper.cpp b/llvm/lib/Transforms/Utils/ValueMapper.cpp\nindex a8ae3ee3b251..93a4c829df06 100644\n--- a/llvm/lib/Transforms/Utils/ValueMapper.cpp\n+++ b/llvm/lib/Transforms/Utils/ValueMapper.cpp\n@@ -552,6 +552,7 @@ void Mapper::remapDPValue(DPValue &V) {\n V.setKillAddress();\n else if (NewAddr)\n V.setAddress(NewAddr);\n+ V.setAssignId(cast(mapMetadata(V.getAssignID())));\n }\n \n // Find Value operands and remap those.\n", "tests": [{"file": "llvm/test/DebugInfo/Generic/ipsccp-remap-assign-id.ll", "commands": ["opt -passes=ipsccp %s -S -o -", "opt --try-experimental-debuginfo-iterators -passes=ipsccp %s -S -o -"], "tests": [{"test_name": "", "test_body": "\n;; Check the dbg.assign DIAssignID operand gets remapped after cloning.\n\n;\n\n; Function Attrs: mustprogress nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare void @llvm.dbg.declare(metadata, metadata, metadata)\n\ndefine void @inv_txfm_add_dct_dct_4x4_c() {\nentry:\n call void @inv_txfm_add_c(ptr @dav1d_inv_dct4_1d_c)\n ret void\n}\n\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg)\n\n; Function Attrs: noinline\ndefine void @inv_txfm_add_c(ptr %first_1d_fn) #2 {\nentry:\n %tmp = alloca [4096 x i32], i32 0, align 16, !DIAssignID !5\n tail call void @llvm.dbg.assign(metadata i1 undef, metadata !6, metadata !DIExpression(), metadata !5, metadata ptr %tmp, metadata !DIExpression()), !dbg !16\n call void @llvm.memset.p0.i64(ptr %tmp, i8 0, i64 0, i1 false), !DIAssignID !17\n call void %first_1d_fn(ptr null, i64 0, i32 0, i32 0)\n ret void\n}\n\ndeclare void @dav1d_inv_dct4_1d_c(ptr, i64, i32, i32)\n\ndeclare void @llvm.dbg.assign(metadata, metadata, metadata, metadata, metadata, metadata)\n\nattributes #2 = { noinline }\n\n!llvm.dbg.cu = !{!0}\n!llvm.module.flags = !{!3, !4}\n\n!0 = distinct !DICompileUnit(language: DW_LANG_C11, file: !1, isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !2, splitDebugInlining: false, nameTableKind: GNU)\n!1 = !DIFile(filename: \"itx_tmpl.c\", directory: \".\")\n!2 = !{}\n!3 = !{i32 2, !\"Debug Info Version\", i32 3}\n!4 = !{i32 7, !\"debug-info-assignment-tracking\", i1 true}\n!5 = distinct !DIAssignID()\n!6 = !DILocalVariable(name: \"tmp\", scope: !7, file: !1, line: 78, type: !10)\n!7 = distinct !DISubprogram(name: \"inv_txfm_add_c\", scope: !1, file: !1, line: 41, type: !8, scopeLine: 45, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)\n!8 = distinct !DISubroutineType(types: !9)\n!9 = !{null}\n!10 = !DICompositeType(tag: DW_TAG_array_type, baseType: !11, size: 131072, elements: !2)\n!11 = !DIDerivedType(tag: DW_TAG_typedef, name: \"int32_t\", file: !12, line: 26, baseType: !13)\n!12 = !DIFile(filename: \"stdint-intn.h\", directory: \".\")\n!13 = !DIDerivedType(tag: DW_TAG_typedef, name: \"__int32_t\", file: !14, line: 41, baseType: !15)\n!14 = !DIFile(filename: \"types.h\", directory: \".\")\n!15 = !DIBasicType(name: \"int\", size: 32, encoding: DW_ATE_signed)\n!16 = !DILocation(line: 0, scope: !7)\n!17 = distinct !DIAssignID()"}]}], "issue": {"title": "Clang frontend command failed to build lib/vsprintf.c in Linux kernel", "body": "```\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: clang -Wp,-MMD,lib/.vsprintf.o.d -nostdinc -I./arch/x86/include -I./arch/x86/include/generated -I./include -I./arch/x86/include/uapi -I./arch/x86/include/generated/uapi -I./include/uapi -I./include/generated/uapi -include ./include/linux/compiler-version.h -include ./include/linux/kconfig.h -include ./include/linux/compiler_types.h -D__KERNEL__ --target=x86_64-linux-gnu -fintegrated-as -Werror=unknown-warning-option -Werror=ignored-optimization-argument -Werror=option-ignored -Werror=unused-command-line-argument -fmacro-prefix-map=./= -std=gnu11 -fshort-wchar -funsigned-char -fno-common -fno-PIE -fno-strict-aliasing -mno-sse -mno-mmx -mno-sse2 -mno-3dnow -mno-avx -fcf-protection=none -m64 -falign-loops=1 -mno-80387 -mno-fp-ret-in-387 -mstack-alignment=8 -mskip-rax-setup -mtune=generic -mno-red-zone -mcmodel=kernel -Wno-sign-compare -fno-asynchronous-unwind-tables -mretpoline-external-thunk -mindirect-branch-cs-prefix -mfunction-return=thunk-extern -mharden-sls=all -fpatchable-function-entry=16,16 -fno-delete-null-pointer-checks -O2 -fstack-protector-strong -fno-omit-frame-pointer -fno-optimize-sibling-calls -ftrivial-auto-var-init=zero -fno-stack-clash-protection -fzero-call-used-regs=used-gpr -DCC_USING_NOP_MCOUNT -DCC_USING_FENTRY -falign-functions=16 -fstrict-flex-arrays=3 -fno-strict-overflow -fno-stack-check -Wall -Wundef -Werror=implicit-function-declaration -Werror=implicit-int -Werror=return-type -Werror=strict-prototypes -Wno-format-security -Wno-trigraphs -Wno-frame-address -Wno-address-of-packed-member -Wmissing-declarations -Wmissing-prototypes -Wframe-larger-than=1024 -Wno-gnu -Wno-unused-but-set-variable -Wno-unused-const-variable -Wvla -Wno-pointer-sign -Wcast-function-type -Wimplicit-fallthrough -Werror=date-time -Werror=incompatible-pointer-types -Wenum-conversion -Wno-unused-but-set-variable -Wno-unused-const-variable -Wno-format-overflow -Wno-format-truncation -Wno-pointer-to-enum-cast -Wno-tautological-constant-out-of-range-compare -Wno-unaligned-access -Wno-cast-function-type-strict -Wno-missing-field-initializers -Wno-type-limits -Wno-shift-negative-value -Wno-initializer-overrides -Wno-sign-compare -g -gdwarf-5 -fsanitize=array-bounds -fsanitize=shift -fsanitize=bool -fsanitize=enum -DKBUILD_MODFILE=\\\"lib/vsprintf\\\" -DKBUILD_BASENAME=\\\"vsprintf\\\" -DKBUILD_MODNAME=\\\"vsprintf\\\" -D__KBUILD_MODNAME=kmod_vsprintf -c -o lib/vsprintf.o lib/vsprintf.c\r\n1.\t parser at end of file\r\n2.\tCode generation\r\n3.\tRunning pass 'Function Pass Manager' on module 'lib/vsprintf.c'.\r\n4.\tRunning pass 'Assignment Tracking Analysis' on function '@ip4_string'\r\n #0 0x000055fc39a6be88 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x2d82e88)\r\n #1 0x000055fc39a69c6e llvm::sys::RunSignalHandlers() (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x2d80c6e)\r\n #2 0x000055fc399e7729 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\r\n #3 0x00007f7639a42990 (/lib/x86_64-linux-gnu/libc.so.6+0x42990)\r\n #4 0x000055fc39484775 (anonymous namespace)::AssignmentTrackingLowering::run(FunctionVarLocsBuilder*) AssignmentTrackingAnalysis.cpp:0:0\r\n #5 0x000055fc3947f3fe analyzeFunction(llvm::Function&, llvm::DataLayout const&, FunctionVarLocsBuilder*) AssignmentTrackingAnalysis.cpp:0:0\r\n #6 0x000055fc3947fdcd llvm::AssignmentTrackingAnalysis::runOnFunction(llvm::Function&) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x2796dcd)\r\n #7 0x000055fc395bab96 llvm::FPPassManager::runOnFunction(llvm::Function&) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x28d1b96)\r\n #8 0x000055fc395c1a23 llvm::FPPassManager::runOnModule(llvm::Module&) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x28d8a23)\r\n #9 0x000055fc395bb807 llvm::legacy::PassManagerImpl::run(llvm::Module&) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x28d2807)\r\n#10 0x000055fc39c161b6 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x2f2d1b6)\r\n#11 0x000055fc3a09fed5 clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x33b6ed5)\r\n#12 0x000055fc3b7248f6 clang::ParseAST(clang::Sema&, bool, bool) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x4a3b8f6)\r\n#13 0x000055fc3a316d86 clang::FrontendAction::Execute() (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x362dd86)\r\n#14 0x000055fc3a28c654 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x35a3654)\r\n#15 0x000055fc3a3d293e clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x36e993e)\r\n#16 0x000055fc37b20c3c cc1_main(llvm::ArrayRef, char const*, void*) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0xe37c3c)\r\n#17 0x000055fc37b1dc71 ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#18 0x000055fc3a0e9eb9 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::$_0>(long) Job.cpp:0:0\r\n#19 0x000055fc399e750c llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x2cfe50c)\r\n#20 0x000055fc3a0e9906 clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x3400906)\r\n#21 0x000055fc3a0abd57 clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x33c2d57)\r\n#22 0x000055fc3a0ac1fe clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x33c31fe)\r\n#23 0x000055fc3a0ca5df clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0x33e15df)\r\n#24 0x000055fc37b1d338 clang_main(int, char**, llvm::ToolContext const&) (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0xe34338)\r\n#25 0x000055fc37b2b8d7 main (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0xe428d7)\r\n#26 0x00007f7639a28150 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:74:3\r\n#27 0x00007f7639a28209 call_init ./csu/../csu/libc-start.c:128:20\r\n#28 0x00007f7639a28209 __libc_start_main ./csu/../csu/libc-start.c:347:5\r\n#29 0x000055fc37b1a4b5 _start (/home/eve/Downloads/llvm-project/build2/bin/clang-19+0xe314b5)\r\nclang: error: clang frontend command failed with exit code 139 (use -v to see invocation)\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git 3bdc4c702dfdd1f2fb513943f034d49bb79ded29)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /home/eve/Downloads/llvm-project/build2/bin\r\nclang: note: diagnostic msg: \r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclang: note: diagnostic msg: /tmp/vsprintf-e230c3.c\r\nclang: note: diagnostic msg: /tmp/vsprintf-e230c3.sh\r\nclang: note: diagnostic msg: \r\n\r\n********************\r\n\r\n```\r\n[vsprintf-e230c3.zip](https://github.com/llvm/llvm-project/files/14254556/vsprintf-e230c3.zip)\r\n\r\nThis issue occurs at commit:\r\n```\r\ncommit 3bdc4c702dfdd1f2fb513943f034d49bb79ded29 (HEAD -> main, origin/main, origin/HEAD)\r\nAuthor: Kevin Frei \r\nDate: Mon Feb 12 16:57:02 2024 -0800\r\n```\r\n\r\nThis issue does not occur at commit:\r\n```\r\ncommit 63b414e4977d6e19f05947c88f57cd127fa328e3\r\nAuthor: Vlad Serebrennikov \r\nDate: Sun Feb 11 15:01:18 2024 +0300\r\n```\r\nEdit: this issue also occurs at commit:\r\n```\r\ncommit 13d60ce2f262ef9055389908b63824e53b3054a1 (HEAD)\r\nAuthor: Amir Ayupov \r\nDate: Mon Feb 12 14:51:15 2024 -0800\r\n```\r\nThis issue does not occur at commit:\r\n```\r\ncommit f5fd0deb2371d0bae3bef2563f50e005a140fc6d (HEAD)\r\nAuthor: Joseph Huber \r\nDate: Mon Feb 12 08:14:54 2024 -0600\r\n```", "author": "eve2023", "labels": ["crash", "llvm:transforms"], "comments": [{"author": "ms178", "body": "CC'ing @nathanchance for awareness."}, {"author": "nathanchance", "body": "Caused by 5c9f7682b090124d9a8b69f92d3f7c269dca25fc, which was reverted in bc66e0cf9feb8f3c38fa2e33850b6c6e5477fb21 and forward fixed by 97088b2ab2184ad4bd64f59fba0b92b70468b10d. I can confirm that there is no crash at 5e3c7e3aa48356a62a4b70d5d9d3e4ddd055a390."}]}, "verified": true} {"bug_id": "107198", "issue_url": "https://github.com/llvm/llvm-project/issues/107198", "bug_type": "crash", "base_commit": "c2b92a4250b3f514685676ba8985ea73450f14d3", "knowledge_cutoff": "2024-09-04T08:17:58Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "d65ff3e9364536f9e0bd5f1c1bace626c256a2ad", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[7780, 7785]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::canMapToVector"]}}, "patch": "commit d65ff3e9364536f9e0bd5f1c1bace626c256a2ad\nAuthor: Alexey Bataev \nDate: Wed Sep 4 05:12:17 2024 -0700\n\n [SLP]Fix PR107198: add a check for empty complex type\n \n Need to check if the complex type is empty before trying to dig in,\n trying to find vectorizable type\n \n Fixes https://github.com/llvm/llvm-project/issues/107198\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 19b95cf473e9..a2af7f4e1b01 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -7780,6 +7780,8 @@ unsigned BoUpSLP::canMapToVector(Type *T) const {\n Type *EltTy = T;\n \n while (isa(EltTy)) {\n+ if (EltTy->isEmptyTy())\n+ return 0;\n if (auto *ST = dyn_cast(EltTy)) {\n // Check that struct is homogeneous.\n for (const auto *Ty : ST->elements())\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/empty-struct.ll", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "test", "test_body": "define { { {}, {}, {}, {}, {}, {}, {} } } @test({ {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, { {}, {}, {}, {}, {}, {}, {} }, { {} } } %0) {\n %2 = extractvalue { {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, { {}, {}, {}, {}, {}, {}, {} }, { {} } } %0, 18\n %3 = insertvalue { { {}, {}, {}, {}, {}, {}, {} } } undef, { {}, {}, {}, {}, {}, {}, {} } %2, 0\n ret { { {}, {}, {}, {}, {}, {}, {} } } %3\n}\n"}]}], "issue": {"title": "Crash on slp-vectorizer", "body": "When running `/opt/llvm-19/bin/opt -O3 -S bug.ll` opt crashes with the following:\r\n\r\n(tested on 19 rc2 and rc4 and 18)\r\n```\r\n/opt/llvm-19/bin/llvm-config --version\r\n19.1.0-rc2\r\n\r\n/opt/llvm-19/bin/opt: WARNING: failed to create target machine for 'arm64-apple-darwin23.6.0': unable to get target for 'arm64-apple-darwin23.6.0', see --version and --triple.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/llvm-19/bin/opt -O3 -S bug.ll\r\n1.\tRunning pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"lol.ll\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"bug\"\r\n #0 0x00005558779163f7 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) /home/admin/repos/llvm-project/llvm/lib/Support/Unix/Signals.inc:723:13\r\n #1 0x0000555877914570 llvm::sys::RunSignalHandlers() /home/admin/repos/llvm-project/llvm/lib/Support/Signals.cpp:106:18\r\n #2 0x0000555877916d9a SignalHandler(int) /home/admin/repos/llvm-project/llvm/lib/Support/Unix/Signals.inc:413:1\r\n #3 0x00007fb7b7e5b050 (/lib/x86_64-linux-gnu/libc.so.6+0x3c050)\r\n #4 0x0000555878d08ae0 llvm::slpvectorizer::BoUpSLP::canMapToVector(llvm::Type*) const /home/admin/repos/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:0:0\r\n #5 0x0000555878d3ca29 llvm::SLPVectorizerPass::vectorizeInsertValueInst(llvm::InsertValueInst*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, bool) /home/admin/repos/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:18336:8\r\n #6 0x0000555878d3d459 llvm::SLPVectorizerPass::vectorizeInserts(llvm::SmallSetVector&, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) /home/admin/repos/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:0:11\r\n #7 0x0000555878d3e35b llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&)::$_172::operator()(bool) const /home/admin/repos/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:18847:20\r\n #8 0x0000555878d34039 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) /home/admin/repos/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:18963:18\r\n #9 0x0000555878d31968 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) /home/admin/repos/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:15865:13\r\n#10 0x0000555878d31431 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) /home/admin/repos/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:15796:18\r\n#11 0x0000555878921dfd llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) /home/admin/repos/llvm-project/llvm/include/llvm/IR/PassManagerInternal.h:90:5\r\n#12 0x0000555877ab570d llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) /home/admin/repos/llvm-project/llvm/include/llvm/IR/PassManagerImpl.h:85:8\r\n#13 0x00005558785d887d llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) /home/admin/repos/llvm-project/llvm/include/llvm/IR/PassManagerInternal.h:90:5\r\n#14 0x0000555877ab94f0 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) /home/admin/repos/llvm-project/llvm/lib/IR/PassManager.cpp:129:23\r\n#15 0x00005558785d86cd llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) /home/admin/repos/llvm-project/llvm/include/llvm/IR/PassManagerInternal.h:90:5\r\n#16 0x0000555877ab44ed llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) /home/admin/repos/llvm-project/llvm/include/llvm/IR/PassManagerImpl.h:85:8\r\n#17 0x00005558788ca8c9 llvm::SmallPtrSetImplBase::isSmall() const /home/admin/repos/llvm-project/llvm/include/llvm/ADT/SmallPtrSet.h:197:33\r\n#18 0x00005558788ca8c9 llvm::SmallPtrSetImplBase::~SmallPtrSetImplBase() /home/admin/repos/llvm-project/llvm/include/llvm/ADT/SmallPtrSet.h:84:10\r\n#19 0x00005558788ca8c9 llvm::PreservedAnalyses::~PreservedAnalyses() /home/admin/repos/llvm-project/llvm/include/llvm/IR/Analysis.h:111:7\r\n#20 0x00005558788ca8c9 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) /home/admin/repos/llvm-project/llvm/tools/opt/NewPMDriver.cpp:547:3\r\n#21 0x00005558778e5779 optMain /home/admin/repos/llvm-project/llvm/tools/opt/optdriver.cpp:739:12\r\n#22 0x00007fb7b7e4624a (/lib/x86_64-linux-gnu/libc.so.6+0x2724a)\r\n#23 0x00007fb7b7e46305 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x27305)\r\n#24 0x00005558778dfd21 _start (/opt/llvm-19/bin/opt+0xf17d21)\r\nSegmentation fault\r\n```\r\n\r\n```llvmir\r\n; ModuleID = 'LLVMDialectModule'\r\nsource_filename = \"LLVMDialectModule\"\r\ntarget datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"arm64-apple-darwin23.6.0\"\r\n\r\ndefine { { {}, {}, {}, {}, {}, {}, {} } } @\"bug\"({ {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, { {}, {}, {}, {}, {}, {}, {} }, { {} } } %0) {\r\n %2 = extractvalue { {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}, { {}, {}, {}, {}, {}, {}, {} }, { {} } } %0, 18\r\n %3 = insertvalue { { {}, {}, {}, {}, {}, {}, {} } } undef, { {}, {}, {}, {}, {}, {}, {} } %2, 0\r\n ret { { {}, {}, {}, {}, {}, {}, {} } } %3\r\n}\r\n\r\n!llvm.module.flags = !{!0}\r\n\r\n!0 = !{i32 2, !\"Debug Info Version\", i32 3}\r\n```", "author": "edg-l", "labels": ["llvm:SLPVectorizer", "crash"], "comments": [{"author": "edg-l", "body": "one question, will this be included in llvm 19?"}, {"author": "alexey-bataev", "body": "Not sure, it is not an immediate regression from 18 to 19, and it is too late, I think, to include it into 19"}]}, "verified": true} {"bug_id": "125301", "issue_url": "https://github.com/llvm/llvm-project/issues/125301", "bug_type": "crash", "base_commit": "4f381d0be845a8702c89a0a7f4bd9188e1ff60a3", "knowledge_cutoff": "2025-01-31T21:43:25Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "75b922dccfc35ec25a520b1941e6682a300802b8", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlan.h": [[1058, 1063]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlan.h": ["classof"]}}, "patch": "commit 75b922dccfc35ec25a520b1941e6682a300802b8\nAuthor: Florian Hahn \nDate: Sat Feb 1 21:40:58 2025 +0000\n\n [VPlan] Check VPWidenIntrinsicSC in VPRecipeWithIRFlags::classof.\n \n When VPWidenIntrinsicRecipe was changed to inhert from VPRecipeWithIRFlags,\n VPRecipeWithIRFlags::classof wasn't updated accordingly. Also check for\n VPWidenIntrinsicSC in VPRecipeWithIRFlags::classof.\n \n Fixes https://github.com/llvm/llvm-project/issues/125301.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h\nindex 459222234bc3..87ef2dc6d8ea 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlan.h\n+++ b/llvm/lib/Transforms/Vectorize/VPlan.h\n@@ -1058,6 +1058,7 @@ public:\n R->getVPDefID() == VPRecipeBase::VPWidenEVLSC ||\n R->getVPDefID() == VPRecipeBase::VPWidenGEPSC ||\n R->getVPDefID() == VPRecipeBase::VPWidenCastSC ||\n+ R->getVPDefID() == VPRecipeBase::VPWidenIntrinsicSC ||\n R->getVPDefID() == VPRecipeBase::VPReplicateSC ||\n R->getVPDefID() == VPRecipeBase::VPReverseVectorPointerSC ||\n R->getVPDefID() == VPRecipeBase::VPVectorPointerSC;\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/drop-poison-generating-flags.ll", "commands": ["opt -p loop-vectorize -mcpu=neoverse-v2 -force-vector-width=4 -S %s"], "tests": [{"test_name": "", "test_body": "\ntarget triple = \"aarch64-unknown-linux\"\n\n; Test case where we visit a VPWidenIntrinsic (for @llvm.fabs) with nnan flags.\n; For https://github.com/llvm/llvm-project/issues/125301.\ndefine void @check_widen_intrinsic_with_nnan(ptr noalias %dst.0, ptr noalias %dst.1, ptr noalias %src.1, ptr %src.2) {\n;\nentry:\n br label %loop.header\n\nloop.header:\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %gep.src.1 = getelementptr inbounds double, ptr %src.1, i64 %iv\n %l.1 = load double, ptr %gep.src.1, align 8\n %abs = tail call nnan double @llvm.fabs.f64(double %l.1)\n %c.0 = fcmp olt double %abs, 1.000000e+00\n br i1 %c.0, label %then, label %else\n\nthen:\n %l.2 = load double, ptr %src.2, align 8\n %iv.sub.1 = add nsw i64 %iv, -1\n %gep.iv.sub.1 = getelementptr double, ptr %dst.0, i64 %iv.sub.1\n store double 0.000000e+00, ptr %gep.iv.sub.1, align 8\n %c.1 = fcmp oeq double %l.2, 0.000000e+00\n br i1 %c.1, label %merge, label %loop.latch\n\nelse:\n %iv.sub.2 = add nsw i64 %iv, -1\n %gep.iv.sub.2 = getelementptr double, ptr %dst.0, i64 %iv.sub.2\n store double 0.000000e+00, ptr %gep.iv.sub.2, align 8\n br label %merge\n\nmerge:\n %merge.iv = phi i64 [ %iv.sub.2, %else ], [ %iv.sub.1, %then ]\n %gep.dst.1 = getelementptr inbounds i32, ptr %dst.1, i64 %merge.iv\n store i32 10, ptr %gep.dst.1, align 4\n br label %loop.latch\n\nloop.latch:\n %iv.next = add nuw nsw i64 %iv, 1\n %exitcond.not = icmp eq i64 %iv.next, 1000\n br i1 %exitcond.not, label %exit, label %loop.header\n\nexit:\nret void\n}\n\ndeclare double @llvm.fabs.f64(double)"}]}], "issue": {"title": "[vplan] Assertion: found instruction with poison generating flags not covered by VPRecipeWithIRFlags", "body": "```\nclang++ -O3 -mcpu=neoverse-v2 -c reduced.ll\n\nclang++: llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp:1926: llvm::VPlanTransforms::dropPoisonGeneratingRecipes(llvm::VPlan&, const std::function&)::: Assertion `(!Instr || !Instr->hasPoisonGeneratingFlags()) && \"found instruction with poison generating flags not covered by \" \"VPRecipeWithIRFlags\"' failed.\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\nStack dump:\n0. Program arguments: clang++ -O3 -mcpu=neoverse-v2 -c reduced.ll\n1. Optimizer\n2. Running pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"reduced.ll\"\n3. Running pass \"loop-vectorize\" on function \"foo\"\n #0 0x0000aaaaaf09abe8 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int)\n #1 0x0000aaaaaf098908 llvm::sys::RunSignalHandlers()\n #2 0x0000aaaaaf098c88 llvm::sys::CleanupOnSignal(unsigned long)\n #3 0x0000aaaaaefee5bc CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\n #4 0x0000fffff7fb08dc (linux-vdso.so.1+0x8dc)\n #5 0x0000fffff7c2f200 __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\n #6 0x0000fffff7bea67c gsignal ./signal/../sysdeps/posix/raise.c:27:6\n #7 0x0000fffff7bd7130 abort ./stdlib/abort.c:81:7\n #8 0x0000fffff7be3fd0 __assert_fail_base ./assert/assert.c:89:7\n #9 0x0000fffff7be4040 __assert_perror_fail ./assert/assert-perr.c:31:1\n#10 0x0000aaaab0bfd210 llvm::VPlanTransforms::dropPoisonGeneratingRecipes(llvm::VPlan&, std::function const&)::'lambda'(llvm::VPRecipeBase*)::operator()(llvm::VPRecipeBase*) const (.isra.0) VPlanTransforms.cpp:0:0\n#11 0x0000aaaab0c065b4 llvm::VPlanTransforms::dropPoisonGeneratingRecipes(llvm::VPlan&, std::function const&) \n#12 0x0000aaaab0b99078 llvm::LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(llvm::VFRange&)\n#13 0x0000aaaab0b99634 llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount)\n#14 0x0000aaaab0b99c1c llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int)\n#15 0x0000aaaab0b9b05c llvm::LoopVectorizePass::processLoop(llvm::Loop*) \n#16 0x0000aaaab0b9d464 llvm::LoopVectorizePass::runImpl(llvm::Function&) \n#17 0x0000aaaab0b9da1c llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) \n#18 0x0000aaaab07856fc llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) PassBuilder.cpp:0:0\n#19 0x0000aaaaaeb3fef8 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&)\n#20 0x0000aaaaadbd51dc llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) NVPTXTargetMachine.cpp:0:0\n#21 0x0000aaaaaeb3e97c llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&)\n#22 0x0000aaaaadbd513c llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) NVPTXTargetMachine.cpp:0:0\n#23 0x0000aaaaaeb3e3a4 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) \n#24 0x0000aaaaafa2b598 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\n#25 0x0000aaaaafa2e258 clang::emitBackendOutput(clang::CompilerInstance&, clang::CodeGenOptions&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*)\n#26 0x0000aaaaafb69e8c clang::CodeGenAction::ExecuteAction()\n#27 0x0000aaaaafef684c clang::FrontendAction::Execute()\n#28 0x0000aaaaafe8bb30 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) \n#29 0x0000aaaaaffe49b0 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) \n#30 0x0000aaaaadb8f874 cc1_main(llvm::ArrayRef, char const*, void*)\n#31 0x0000aaaaadb87f34 ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\n#32 0x0000aaaaafcd7778 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\n#33 0x0000aaaaaefeeac0 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref)\n#34 0x0000aaaaafcd7d48 clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\n#35 0x0000aaaaafc9daf4 clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const\n#36 0x0000aaaaafc9e744 clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const \n#37 0x0000aaaaafca6268 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) \n#38 0x0000aaaaadb8c654 clang_main(int, char**, llvm::ToolContext const&)\n#39 0x0000aaaaadb9c06c main\n#40 0x0000fffff7bd73fc __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:74:3\n#41 0x0000fffff7bd74cc call_init ./csu/../csu/libc-start.c:128:20\n#42 0x0000fffff7bd74cc __libc_start_main ./csu/../csu/libc-start.c:379:5\n#43 0x0000aaaaadb87a24 _start\nclang++: error: clang frontend command failed with exit code 134 (use -v to see invocation)\nclang version 21.0.0git (https://github.com/llvm/llvm-project 82b923defe27cc46ecf3084f3a8f1c0d1c36199e)\nTarget: aarch64-unknown-linux-gnu\nThread model: posix\nInstalledDir: \nBuild config: +assertions\nclang++: note: diagnostic msg: Error generating preprocessed source(s) - no preprocessable inputs.\n```\n```\n# works ok\nclang++ -O3 -mcpu=neoverse-n1 -c reduced.ll\n```\n\n```\n; ModuleID = ''\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-Fn32\"\ntarget triple = \"aarch64-unknown-linux-gnu\"\n\ndefine void @foo(ptr %0, ptr %1, ptr %bar, i64 %2) {\n br label %4\n\n4: ; preds = %5, %3\n br label %5\n\n5: ; preds = %26, %4\n %6 = phi i32 [ %27, %26 ], [ 0, %4 ]\n %7 = phi i64 [ %28, %26 ], [ %2, %4 ]\n %8 = icmp sgt i64 %7, 0\n br i1 %8, label %9, label %4\n\n9: ; preds = %5\n %10 = load double, ptr %1, align 8\n %11 = call nnan double @llvm.fabs.f64(double %10)\n %12 = fcmp olt double %11, 1.000000e+00\n br i1 %12, label %13, label %21\n\n13: ; preds = %9\n %14 = load double, ptr %bar, align 8\n %15 = zext i32 %6 to i64\n %16 = add i64 %15, -1\n %17 = getelementptr double, ptr %0, i64 %16\n store double 0.000000e+00, ptr %17, align 8\n %18 = fcmp oeq double %14, 0.000000e+00\n br i1 %18, label %19, label %26\n\n19: ; preds = %13\n %20 = getelementptr i32, ptr null, i64 %16\n store i32 0, ptr %20, align 4\n br label %26\n\n21: ; preds = %9\n %22 = zext i32 %6 to i64\n %23 = add i64 %22, -1\n %24 = getelementptr double, ptr %0, i64 %23\n store double 0.000000e+00, ptr %24, align 8\n %25 = getelementptr i32, ptr null, i64 %23\n store i32 0, ptr %25, align 4\n br label %26\n\n26: ; preds = %21, %19, %13\n %27 = add nsw i32 %6, 1\n %28 = add i64 %7, -1\n br label %5\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare double @llvm.fabs.f64(double) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n```", "author": "ashermancinelli", "labels": ["release:backport", "vectorizers", "crash"], "comments": [{"author": "fhahn", "body": "@ashermancinelli it looks like `reduced.ll` is missing from the issue here? Could you attach it (or share a link to godbolt)?"}, {"author": "ashermancinelli", "body": "@fhahn Sorry about that! I updated the description."}]}, "verified": true} {"bug_id": "82884", "issue_url": "https://github.com/llvm/llvm-project/issues/82884", "bug_type": "miscompilation", "base_commit": "0c7a605ada6cb392e6e8c16dbccf2b7e59017399", "knowledge_cutoff": "2024-02-24T15:15:12Z", "lit_test_dir": ["llvm/test/Transforms/GVN"], "hints": {"fix_commit": "892b4beeac50920e630f10905b2916295e2eb6d8", "components": ["Local"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/Local.cpp": [[3369, 3379]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/Local.cpp": ["llvm::patchReplacementInstruction"]}}, "patch": "commit 892b4beeac50920e630f10905b2916295e2eb6d8\nAuthor: Yingwei Zheng \nDate: Mon Feb 26 15:55:56 2024 +0800\n\n [GVN] Drop nsw/nuw flags when replacing the result of a with.overflow intrinsic with a overflowing binary operator (#82935)\n \n Alive2: https://alive2.llvm.org/ce/z/gyL7mn\n Fixes https://github.com/llvm/llvm-project/issues/82884.\n\ndiff --git a/llvm/lib/Transforms/Utils/Local.cpp b/llvm/lib/Transforms/Utils/Local.cpp\nindex 1373f5f7f449..075eeb5b19fd 100644\n--- a/llvm/lib/Transforms/Utils/Local.cpp\n+++ b/llvm/lib/Transforms/Utils/Local.cpp\n@@ -3369,11 +3369,17 @@ void llvm::patchReplacementInstruction(Instruction *I, Value *Repl) {\n \n // Patch the replacement so that it is not more restrictive than the value\n // being replaced.\n+ WithOverflowInst *UnusedWO;\n+ // When replacing the result of a llvm.*.with.overflow intrinsic with a\n+ // overflowing binary operator, nuw/nsw flags may no longer hold.\n+ if (isa(ReplInst) &&\n+ match(I, m_ExtractValue<0>(m_WithOverflowInst(UnusedWO))))\n+ ReplInst->dropPoisonGeneratingFlags();\n // Note that if 'I' is a load being replaced by some operation,\n // for example, by an arithmetic operation, then andIRFlags()\n // would just erase all math flags from the original arithmetic\n // operation, which is clearly not wanted and not needed.\n- if (!isa(I))\n+ else if (!isa(I))\n ReplInst->andIRFlags(I);\n \n // FIXME: If both the original and replacement value are part of the\n", "tests": [{"file": "llvm/test/Transforms/GVN/pr82884.ll", "commands": ["opt -S -passes=gvn < %s"], "tests": [{"test_name": "pr82884", "test_body": "define i32 @pr82884(i32 %x) {\n %mul = mul nuw nsw i32 %x, %x\n call void @use(i32 %mul)\n %mul2 = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 %x, i32 %x)\n %ret = extractvalue { i32, i1 } %mul2, 0\n ret i32 %ret\n}\n\ndeclare void @use(i32)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare { i32, i1 } @llvm.smul.with.overflow.i32(i32, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[GVN] GVNPass forgets to remove poison generating flags", "body": "```llvm\r\n; ModuleID = '1851685_debug_2.e4807e3094afc004-cgu.0'\r\nsource_filename = \"1851685_debug_2.e4807e3094afc004-cgu.0\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\n%Adt58 = type { %Adt50 }\r\n%Adt50 = type { { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, { i16, i8, [5 x i8], i64 }, [8 x i32], { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr, [7 x i32], [3 x i32] }\r\n%Adt53 = type { { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] } }\r\n%Adt51 = type { { i64, i8, [7 x i8] }, { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, %Adt50, [8 x i32], i64, i64, [6 x i8], [10 x i8] }\r\n%Adt59 = type { { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, i64, ptr, i32, i16, [1 x i8], [1 x i8] }\r\n\r\n@vtable.0 = private unnamed_addr constant <{ ptr, [16 x i8], ptr, ptr, ptr }> <{ ptr @\"_ZN4core3ptr85drop_in_place$LT$std..rt..lang_start$LT$$LP$$RP$$GT$..$u7b$$u7b$closure$u7d$$u7d$$GT$17he9a486ff76ea726fE\", [16 x i8] c\"\\08\\00\\00\\00\\00\\00\\00\\00\\08\\00\\00\\00\\00\\00\\00\\00\", ptr @\"_ZN4core3ops8function6FnOnce40call_once$u7b$$u7b$vtable.shim$u7d$$u7d$17h91c73705548d2f78E\", ptr @\"_ZN3std2rt10lang_start28_$u7b$$u7b$closure$u7d$$u7d$17h1cb78ba075200b22E\", ptr @\"_ZN3std2rt10lang_start28_$u7b$$u7b$closure$u7d$$u7d$17h1cb78ba075200b22E\" }>, align 8\r\n@alloc_2cdd8ac758b4c372b56d8683306290e0 = private unnamed_addr constant <{ [15 x i8] }> <{ [15 x i8] c\"hi:%ld lo:%ld\\0A\\00\" }>, align 1\r\n\r\n; std::sys_common::backtrace::__rust_begin_short_backtrace\r\n; Function Attrs: noinline nonlazybind uwtable\r\ndefine internal void @_ZN3std10sys_common9backtrace28__rust_begin_short_backtrace17h176fc8086015470dE(ptr %f) unnamed_addr #0 {\r\nstart:\r\n; call core::ops::function::FnOnce::call_once\r\n call void @_ZN4core3ops8function6FnOnce9call_once17h23c7757b8dc992b5E(ptr %f)\r\n call void asm sideeffect \"\", \"~{memory}\"(), !srcloc !4\r\n ret void\r\n}\r\n\r\n; std::rt::lang_start::{{closure}}\r\n; Function Attrs: inlinehint nonlazybind uwtable\r\ndefine internal i32 @\"_ZN3std2rt10lang_start28_$u7b$$u7b$closure$u7d$$u7d$17h1cb78ba075200b22E\"(ptr align 8 %_1) unnamed_addr #2 {\r\nstart:\r\n %self = alloca i8, align 1\r\n %_4 = load ptr, ptr %_1, align 8, !nonnull !5, !noundef !5\r\n; call std::sys_common::backtrace::__rust_begin_short_backtrace\r\n call void @_ZN3std10sys_common9backtrace28__rust_begin_short_backtrace17h176fc8086015470dE(ptr %_4)\r\n; call <() as std::process::Termination>::report\r\n %0 = call i8 @\"_ZN54_$LT$$LP$$RP$$u20$as$u20$std..process..Termination$GT$6report17hb8b4bfb25c4e0e48E\"()\r\n store i8 %0, ptr %self, align 1\r\n %_6 = load i8, ptr %self, align 1, !noundef !5\r\n %_0 = zext i8 %_6 to i32\r\n ret i32 %_0\r\n}\r\n\r\n; core::ops::function::FnOnce::call_once{{vtable.shim}}\r\n; Function Attrs: inlinehint nonlazybind uwtable\r\ndefine internal i32 @\"_ZN4core3ops8function6FnOnce40call_once$u7b$$u7b$vtable.shim$u7d$$u7d$17h91c73705548d2f78E\"(ptr %_1) unnamed_addr #2 {\r\nstart:\r\n %_2 = alloca {}, align 1\r\n %0 = load ptr, ptr %_1, align 8, !nonnull !5, !noundef !5\r\n; call core::ops::function::FnOnce::call_once\r\n %_0 = call i32 @_ZN4core3ops8function6FnOnce9call_once17ha42f99eeb4429a31E(ptr %0)\r\n ret i32 %_0\r\n}\r\n\r\n; core::ops::function::FnOnce::call_once\r\n; Function Attrs: inlinehint nonlazybind uwtable\r\ndefine internal void @_ZN4core3ops8function6FnOnce9call_once17h23c7757b8dc992b5E(ptr %_1) unnamed_addr #2 {\r\nstart:\r\n %_2 = alloca {}, align 1\r\n call void %_1()\r\n ret void\r\n}\r\n\r\n; core::ops::function::FnOnce::call_once\r\n; Function Attrs: inlinehint nonlazybind uwtable\r\ndefine internal i32 @_ZN4core3ops8function6FnOnce9call_once17ha42f99eeb4429a31E(ptr %0) unnamed_addr #2 personality ptr @rust_eh_personality {\r\nstart:\r\n %1 = alloca { ptr, i32, [1 x i32] }, align 8\r\n %_2 = alloca {}, align 1\r\n %_1 = alloca ptr, align 8\r\n store ptr %0, ptr %_1, align 8\r\n; invoke std::rt::lang_start::{{closure}}\r\n %_0 = invoke i32 @\"_ZN3std2rt10lang_start28_$u7b$$u7b$closure$u7d$$u7d$17h1cb78ba075200b22E\"(ptr align 8 %_1)\r\n to label %bb1 unwind label %cleanup\r\n\r\nbb3: ; preds = %cleanup\r\n %2 = load ptr, ptr %1, align 8, !noundef !5\r\n %3 = getelementptr inbounds i8, ptr %1, i64 8\r\n %4 = load i32, ptr %3, align 8, !noundef !5\r\n %5 = insertvalue { ptr, i32 } poison, ptr %2, 0\r\n %6 = insertvalue { ptr, i32 } %5, i32 %4, 1\r\n resume { ptr, i32 } %6\r\n\r\ncleanup: ; preds = %start\r\n %7 = landingpad { ptr, i32 }\r\n cleanup\r\n %8 = extractvalue { ptr, i32 } %7, 0\r\n %9 = extractvalue { ptr, i32 } %7, 1\r\n store ptr %8, ptr %1, align 8\r\n %10 = getelementptr inbounds i8, ptr %1, i64 8\r\n store i32 %9, ptr %10, align 8\r\n br label %bb3\r\n\r\nbb1: ; preds = %start\r\n ret i32 %_0\r\n}\r\n\r\n; core::ptr::drop_in_place::{{closure}}>\r\n; Function Attrs: inlinehint nonlazybind uwtable\r\ndefine internal void @\"_ZN4core3ptr85drop_in_place$LT$std..rt..lang_start$LT$$LP$$RP$$GT$..$u7b$$u7b$closure$u7d$$u7d$$GT$17he9a486ff76ea726fE\"(ptr align 8 %_1) unnamed_addr #2 {\r\nstart:\r\n ret void\r\n}\r\n\r\n; core::hint::black_box\r\n; Function Attrs: inlinehint nonlazybind uwtable\r\ndefine internal void @_ZN4core4hint9black_box17h7d48bc7feb0af96eE(ptr sret({ i16, i8, [5 x i8], i64 }) align 8 %_0, ptr align 8 %dummy) unnamed_addr #2 {\r\nstart:\r\n call void @llvm.memcpy.p0.p0.i64(ptr align 8 %_0, ptr align 8 %dummy, i64 16, i1 false)\r\n call void asm sideeffect \"\", \"r,~{memory}\"(ptr %_0), !srcloc !4\r\n ret void\r\n}\r\n\r\n; core::hint::black_box\r\n; Function Attrs: inlinehint nonlazybind uwtable\r\ndefine internal void @_ZN4core4hint9black_box17ha89c3f9cb79c398cE() unnamed_addr #2 {\r\nstart:\r\n call void asm sideeffect \"\", \"~{memory}\"(), !srcloc !4\r\n ret void\r\n}\r\n\r\n; <() as std::process::Termination>::report\r\n; Function Attrs: inlinehint nonlazybind uwtable\r\ndefine internal i8 @\"_ZN54_$LT$$LP$$RP$$u20$as$u20$std..process..Termination$GT$6report17hb8b4bfb25c4e0e48E\"() unnamed_addr #2 {\r\nstart:\r\n ret i8 0\r\n}\r\n\r\n; _1851685_debug_2::enter\r\n; Function Attrs: nonlazybind uwtable\r\ndefine internal void @_ZN16_1851685_debug_25enter17h7e9cae7a334c5019E() unnamed_addr #1 {\r\nstart:\r\n %_3 = alloca [6 x i32], align 4\r\n %_2 = alloca { i16, i8, [5 x i8], i64 }, align 8\r\n %0 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_2, i32 0, i32 1\r\n store i8 0, ptr %0, align 2\r\n store i16 0, ptr %_2, align 8\r\n %1 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_2, i32 0, i32 3\r\n store i64 0, ptr %1, align 8\r\n %2 = getelementptr inbounds [6 x i32], ptr %_3, i64 0, i64 0\r\n call void @llvm.memset.p0.i64(ptr align 4 %2, i8 0, i64 24, i1 false)\r\n; call _1851685_debug_2::fn14\r\n call void @_ZN16_1851685_debug_24fn1417h1fc74d4c311bb23aE(ptr align 8 %_2, ptr align 4 %_3)\r\n ret void\r\n}\r\n\r\n; _1851685_debug_2::fn14\r\n; Function Attrs: nonlazybind uwtable\r\ndefine internal void @_ZN16_1851685_debug_24fn1417h1fc74d4c311bb23aE(ptr align 8 %_1, ptr align 4 %_2) unnamed_addr #1 {\r\nstart:\r\n %_16 = alloca i8, align 1\r\n %_15 = alloca %Adt58, align 16\r\n %_14 = alloca { i16, i8, [5 x i8], i64 }, align 8\r\n %_13 = alloca { { i16, i8, [5 x i8], i64 }, [6 x i32], i32, [1 x i32], [4 x i64] }, align 8\r\n %_12 = alloca %Adt53, align 16\r\n %_10 = alloca %Adt51, align 16\r\n %_7 = alloca %Adt59, align 8\r\n %_5 = alloca i128, align 16\r\n %0 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 1\r\n store i8 84, ptr %0, align 2\r\n store i16 -25440, ptr %_1, align 8\r\n %1 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 3\r\n store i64 -9223372036854775808, ptr %1, align 8\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb4, %start\r\n store double 0x41A15ED4AE000000, ptr %_7, align 8\r\n %2 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 4\r\n %3 = load i16, ptr %_1, align 8, !noundef !5\r\n %4 = urem i16 %3, -3470\r\n store i16 %4, ptr %2, align 4\r\n %5 = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 3029947793098001587, i64 -3638983068927244405)\r\n %6 = extractvalue { i64, i1 } %5, 0\r\n %7 = extractvalue { i64, i1 } %5, 1\r\n store i64 %6, ptr %_10, align 16\r\n %8 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %9 = zext i1 %7 to i8\r\n store i8 %9, ptr %8, align 8\r\n %10 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %11 = load i8, ptr %10, align 8, !range !6, !noundef !5\r\n %12 = trunc i8 %11 to i1\r\n %13 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %14 = load i8, ptr %13, align 8, !range !6, !noundef !5\r\n %15 = trunc i8 %14 to i1\r\n %_9 = xor i1 %12, %15\r\n %16 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 2\r\n store ptr %_10, ptr %16, align 8\r\n %17 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 1\r\n store i64 -1498547728408969452, ptr %17, align 8\r\n %18 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %19 = getelementptr inbounds %Adt50, ptr %18, i32 0, i32 2\r\n %20 = getelementptr inbounds [8 x i32], ptr %19, i64 0, i64 0\r\n store i32 168516996, ptr %20, align 16\r\n %21 = getelementptr inbounds [8 x i32], ptr %19, i64 0, i64 1\r\n store i32 245529160, ptr %21, align 4\r\n %22 = getelementptr inbounds [8 x i32], ptr %19, i64 0, i64 2\r\n store i32 1223136339, ptr %22, align 8\r\n %23 = getelementptr inbounds [8 x i32], ptr %19, i64 0, i64 3\r\n store i32 1622354325, ptr %23, align 4\r\n %24 = getelementptr inbounds [8 x i32], ptr %19, i64 0, i64 4\r\n store i32 -1002348909, ptr %24, align 16\r\n %25 = getelementptr inbounds [8 x i32], ptr %19, i64 0, i64 5\r\n store i32 1976405017, ptr %25, align 4\r\n %26 = getelementptr inbounds [8 x i32], ptr %19, i64 0, i64 6\r\n store i32 -1442509311, ptr %26, align 8\r\n %27 = getelementptr inbounds [8 x i32], ptr %19, i64 0, i64 7\r\n store i32 1165829790, ptr %27, align 4\r\n %28 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %29 = getelementptr inbounds %Adt50, ptr %28, i32 0, i32 3\r\n %30 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %29, i32 0, i32 2\r\n %31 = load double, ptr %_7, align 8, !noundef !5\r\n %32 = fneg double %31\r\n store double %32, ptr %30, align 16\r\n %33 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 3\r\n %34 = load i64, ptr %33, align 8, !noundef !5\r\n %_8 = xor i64 %34, -1\r\n store i64 7564290096487423208, ptr %_10, align 16\r\n %35 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %36 = zext i1 %_9 to i8\r\n store i8 %36, ptr %35, align 8\r\n %37 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %38 = getelementptr inbounds %Adt50, ptr %37, i32 0, i32 3\r\n %39 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %38, i32 0, i32 1\r\n store i64 %_8, ptr %39, align 16\r\n %40 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %41 = load i64, ptr %_10, align 16, !noundef !5\r\n store i64 %41, ptr %40, align 8\r\n %42 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %43 = getelementptr inbounds %Adt50, ptr %42, i32 0, i32 3\r\n %44 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %43, i32 0, i32 1\r\n %45 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %44, i32 0, i32 5\r\n; call _1851685_debug_2::fn15\r\n %46 = call i128 @_ZN16_1851685_debug_24fn1517hfe3f07cbb5faefb0E()\r\n store i128 %46, ptr %45, align 16\r\n %47 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %48 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %47, i32 0, i32 1\r\n %49 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %50 = load i64, ptr %49, align 8, !noundef !5\r\n %51 = trunc i64 %50 to i16\r\n store i16 %51, ptr %48, align 16\r\n %52 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 4\r\n %53 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 1\r\n %54 = load i64, ptr %53, align 8, !noundef !5\r\n store i64 %54, ptr %52, align 16\r\n %55 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %56 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %57 = getelementptr inbounds %Adt50, ptr %56, i32 0, i32 3\r\n %58 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %57, i32 0, i32 1\r\n %59 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %58, i32 0, i32 5\r\n %60 = load i128, ptr %59, align 16, !noundef !5\r\n %61 = ashr i128 %60, 0\r\n store i128 %61, ptr %55, align 16\r\n %62 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %63 = getelementptr inbounds %Adt50, ptr %62, i32 0, i32 3\r\n %64 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %63, i32 0, i32 2\r\n %65 = load double, ptr %64, align 16, !noundef !5\r\n %66 = fneg double %65\r\n store double %66, ptr %_7, align 8\r\n %67 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %68 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %67, i32 0, i32 2\r\n %69 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %70 = load i8, ptr %69, align 8, !range !6, !noundef !5\r\n %71 = trunc i8 %70 to i1\r\n %72 = xor i1 %71, true\r\n %73 = zext i1 %72 to i8\r\n store i8 %73, ptr %68, align 2\r\n %74 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %75 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %74, i32 0, i32 4\r\n %76 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %77 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %76, i32 0, i32 1\r\n %78 = load i16, ptr %77, align 16, !noundef !5\r\n %79 = getelementptr inbounds [1 x i16], ptr %75, i64 0, i64 0\r\n store i16 %78, ptr %79, align 4\r\n %80 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %81 = getelementptr inbounds %Adt50, ptr %80, i32 0, i32 1\r\n %82 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %81, i32 0, i32 1\r\n %83 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 1\r\n %84 = load i8, ptr %83, align 2, !noundef !5\r\n store i8 %84, ptr %82, align 2\r\n %85 = getelementptr inbounds [6 x i32], ptr %_2, i64 0, i64 0\r\n store i32 -154507050, ptr %85, align 4\r\n %86 = getelementptr inbounds [6 x i32], ptr %_2, i64 0, i64 1\r\n store i32 2086635898, ptr %86, align 4\r\n %87 = getelementptr inbounds [6 x i32], ptr %_2, i64 0, i64 2\r\n store i32 468582002, ptr %87, align 4\r\n %88 = getelementptr inbounds [6 x i32], ptr %_2, i64 0, i64 3\r\n store i32 -21184342, ptr %88, align 4\r\n %89 = getelementptr inbounds [6 x i32], ptr %_2, i64 0, i64 4\r\n store i32 -1173629817, ptr %89, align 4\r\n %90 = getelementptr inbounds [6 x i32], ptr %_2, i64 0, i64 5\r\n store i32 -2078649110, ptr %90, align 4\r\n %91 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n store i128 0, ptr %91, align 16\r\n %92 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %_12, i32 0, i32 2\r\n %93 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %94 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %93, i32 0, i32 2\r\n %95 = load i8, ptr %94, align 2, !range !6, !noundef !5\r\n %96 = trunc i8 %95 to i1\r\n %97 = zext i1 %96 to i8\r\n store i8 %97, ptr %92, align 2\r\n %98 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %99 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %98, i32 0, i32 1\r\n %100 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %101 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %100, i32 0, i32 1\r\n %102 = load i16, ptr %101, align 16, !noundef !5\r\n %103 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %104 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %103, i32 0, i32 1\r\n %105 = load i16, ptr %104, align 16, !noundef !5\r\n %106 = mul i16 %102, %105\r\n store i16 %106, ptr %99, align 16\r\n %107 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 3\r\n store i32 630181660, ptr %107, align 8\r\n %108 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 3\r\n %109 = load i32, ptr %108, align 8, !noundef !5\r\n %110 = uitofp i32 %109 to double\r\n store double %110, ptr %_7, align 8\r\n %_11 = sub i64 0, %_8\r\n %111 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %112 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %113 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %112, i32 0, i32 1\r\n %114 = load i16, ptr %113, align 16, !noundef !5\r\n %115 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %111, i32 0, i32 1\r\n store i16 %114, ptr %115, align 16\r\n %116 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %117 = load i128, ptr %116, align 16, !noundef !5\r\n store i128 %117, ptr %111, align 16\r\n %118 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %111, i32 0, i32 2\r\n %119 = zext i1 %_9 to i8\r\n store i8 %119, ptr %118, align 2\r\n %120 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %121 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %120, i32 0, i32 4\r\n %122 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %111, i32 0, i32 4\r\n %123 = load i16, ptr %121, align 4\r\n store i16 %123, ptr %122, align 4\r\n %124 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %125 = load i64, ptr %124, align 8, !noundef !5\r\n %126 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %127 = load i64, ptr %126, align 8, !noundef !5\r\n %128 = call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %125, i64 %127)\r\n %129 = extractvalue { i64, i1 } %128, 0\r\n %130 = extractvalue { i64, i1 } %128, 1\r\n store i64 %129, ptr %_10, align 16\r\n %131 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %132 = zext i1 %130 to i8\r\n store i8 %132, ptr %131, align 8\r\n br label %bb3\r\n\r\nbb3: ; preds = %bb6, %bb1\r\n %133 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %134 = getelementptr inbounds %Adt50, ptr %133, i32 0, i32 3\r\n %135 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %134, i32 0, i32 2\r\n %136 = load double, ptr %_7, align 8, !noundef !5\r\n %137 = fdiv double %136, 0x92917F04BFA33\r\n store double %137, ptr %135, align 16\r\n %138 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 5\r\n store i64 -6496701646011298800, ptr %138, align 8\r\n %139 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %140 = getelementptr inbounds %Adt50, ptr %139, i32 0, i32 1\r\n %141 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %140, i32 0, i32 1\r\n %142 = load i8, ptr %141, align 2, !noundef !5\r\n %143 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 1\r\n store i8 %142, ptr %143, align 2\r\n %144 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 4\r\n %145 = load i16, ptr %144, align 4, !noundef !5\r\n store i16 %145, ptr %_1, align 8\r\n %146 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 3\r\n store i64 %_11, ptr %146, align 8\r\n %147 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %148 = getelementptr inbounds %Adt50, ptr %147, i32 0, i32 3\r\n %149 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %148, i32 0, i32 1\r\n %150 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %149, i32 0, i32 2\r\n %151 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 3\r\n %152 = load i32, ptr %151, align 8, !noundef !5\r\n store i32 %152, ptr %150, align 16\r\n %153 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %154 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %153, i32 0, i32 2\r\n %155 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %156 = load i128, ptr %155, align 16, !noundef !5\r\n %157 = icmp sge i128 %156, 0\r\n %158 = zext i1 %157 to i8\r\n store i8 %158, ptr %154, align 2\r\n %159 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %160 = getelementptr inbounds %Adt50, ptr %159, i32 0, i32 1\r\n %161 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %160, i32 0, i32 3\r\n %162 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 3\r\n %163 = load i64, ptr %162, align 8, !noundef !5\r\n %164 = and i64 %163, %_8\r\n store i64 %164, ptr %161, align 8\r\n %165 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 6\r\n %166 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %167 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %166, i32 0, i32 2\r\n %168 = load i8, ptr %167, align 2, !range !6, !noundef !5\r\n %169 = trunc i8 %168 to i1\r\n %170 = getelementptr inbounds [6 x i8], ptr %165, i64 0, i64 0\r\n %171 = zext i1 %169 to i8\r\n store i8 %171, ptr %170, align 16\r\n %172 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %173 = load i8, ptr %172, align 8, !range !6, !noundef !5\r\n %174 = trunc i8 %173 to i1\r\n %175 = getelementptr inbounds [6 x i8], ptr %165, i64 0, i64 1\r\n %176 = zext i1 %174 to i8\r\n store i8 %176, ptr %175, align 1\r\n %177 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %_12, i32 0, i32 2\r\n %178 = load i8, ptr %177, align 2, !range !6, !noundef !5\r\n %179 = trunc i8 %178 to i1\r\n %180 = getelementptr inbounds [6 x i8], ptr %165, i64 0, i64 2\r\n %181 = zext i1 %179 to i8\r\n store i8 %181, ptr %180, align 2\r\n %182 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %_12, i32 0, i32 2\r\n %183 = load i8, ptr %182, align 2, !range !6, !noundef !5\r\n %184 = trunc i8 %183 to i1\r\n %185 = getelementptr inbounds [6 x i8], ptr %165, i64 0, i64 3\r\n %186 = zext i1 %184 to i8\r\n store i8 %186, ptr %185, align 1\r\n %187 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %188 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %187, i32 0, i32 2\r\n %189 = load i8, ptr %188, align 2, !range !6, !noundef !5\r\n %190 = trunc i8 %189 to i1\r\n %191 = getelementptr inbounds [6 x i8], ptr %165, i64 0, i64 4\r\n %192 = zext i1 %190 to i8\r\n store i8 %192, ptr %191, align 4\r\n %193 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %194 = load i8, ptr %193, align 8, !range !6, !noundef !5\r\n %195 = trunc i8 %194 to i1\r\n %196 = getelementptr inbounds [6 x i8], ptr %165, i64 0, i64 5\r\n %197 = zext i1 %195 to i8\r\n store i8 %197, ptr %196, align 1\r\n %198 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %199 = getelementptr inbounds %Adt50, ptr %198, i32 0, i32 3\r\n %200 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %199, i32 0, i32 1\r\n %201 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %200, i32 0, i32 5\r\n %202 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %203 = load i128, ptr %202, align 16, !noundef !5\r\n store i128 %203, ptr %201, align 16\r\n %204 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 2\r\n %205 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %206 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %205, i32 0, i32 1\r\n %207 = load i16, ptr %206, align 16, !noundef !5\r\n %208 = sub i16 0, %207\r\n store i16 %208, ptr %204, align 8\r\n %209 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %210 = getelementptr inbounds %Adt50, ptr %209, i32 0, i32 4\r\n %211 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 1\r\n store ptr %211, ptr %210, align 16\r\n %212 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 5\r\n %213 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 1\r\n %214 = load i8, ptr %213, align 2, !noundef !5\r\n %215 = getelementptr inbounds [1 x i8], ptr %212, i64 0, i64 0\r\n store i8 %214, ptr %215, align 2\r\n %216 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 3\r\n %217 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %218 = getelementptr inbounds %Adt50, ptr %217, i32 0, i32 2\r\n %219 = load <8 x i32>, ptr %218, align 16\r\n store <8 x i32> %219, ptr %216, align 16\r\n %220 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %221 = getelementptr inbounds %Adt50, ptr %220, i32 0, i32 3\r\n %222 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %221, i32 0, i32 1\r\n %223 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %222, i32 0, i32 3\r\n %224 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 5\r\n %225 = load i64, ptr %224, align 8, !noundef !5\r\n %226 = trunc i64 %225 to i32\r\n store i32 %226, ptr %223, align 4\r\n %227 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 3\r\n %228 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %229 = getelementptr inbounds %Adt50, ptr %228, i32 0, i32 1\r\n %230 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %229, i32 0, i32 3\r\n %231 = load i64, ptr %230, align 8, !noundef !5\r\n %232 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %233 = getelementptr inbounds %Adt50, ptr %232, i32 0, i32 3\r\n %234 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %233, i32 0, i32 1\r\n %235 = load i64, ptr %234, align 16, !noundef !5\r\n %236 = xor i64 %231, %235\r\n store i64 %236, ptr %227, align 8\r\n %237 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %238 = getelementptr inbounds %Adt50, ptr %237, i32 0, i32 3\r\n %239 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %238, i32 0, i32 2\r\n %240 = load double, ptr %_7, align 8, !noundef !5\r\n store double %240, ptr %239, align 16\r\n %241 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_14, i32 0, i32 1\r\n %242 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 1\r\n %243 = load i8, ptr %242, align 2, !noundef !5\r\n %244 = xor i8 %243, -1\r\n store i8 %244, ptr %241, align 2\r\n %245 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %246 = getelementptr inbounds %Adt50, ptr %245, i32 0, i32 1\r\n %247 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %248 = load i8, ptr %247, align 8, !range !6, !noundef !5\r\n %249 = trunc i8 %248 to i1\r\n %250 = zext i1 %249 to i16\r\n store i16 %250, ptr %246, align 16\r\n %251 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %252 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %253 = getelementptr inbounds %Adt50, ptr %252, i32 0, i32 3\r\n %254 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %253, i32 0, i32 1\r\n %255 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %254, i32 0, i32 3\r\n %256 = load i32, ptr %255, align 4, !noundef !5\r\n %257 = sext i32 %256 to i128\r\n store i128 %257, ptr %251, align 16\r\n %258 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %259 = getelementptr inbounds %Adt50, ptr %258, i32 0, i32 3\r\n %260 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %259, i32 0, i32 1\r\n %261 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %260, i32 0, i32 3\r\n %262 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %263 = getelementptr inbounds %Adt50, ptr %262, i32 0, i32 1\r\n %264 = load i16, ptr %263, align 16, !noundef !5\r\n %265 = zext i16 %264 to i32\r\n store i32 %265, ptr %261, align 4\r\n br label %bb4\r\n\r\nbb4: ; preds = %bb4, %bb3\r\n %266 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_14, i32 0, i32 1\r\n %267 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %268 = getelementptr inbounds %Adt50, ptr %267, i32 0, i32 3\r\n %269 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %268, i32 0, i32 1\r\n %270 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %269, i32 0, i32 3\r\n %271 = load i32, ptr %270, align 4, !noundef !5\r\n %272 = trunc i32 %271 to i8\r\n store i8 %272, ptr %266, align 2\r\n %273 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_14, i32 0, i32 1\r\n %274 = load i8, ptr %273, align 2, !noundef !5\r\n %275 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_13, i32 0, i32 1\r\n store i8 %274, ptr %275, align 2\r\n %276 = load i16, ptr %_1, align 8, !noundef !5\r\n store i16 %276, ptr %_13, align 8\r\n %277 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %278 = getelementptr inbounds %Adt50, ptr %277, i32 0, i32 1\r\n %279 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %278, i32 0, i32 3\r\n %280 = load i64, ptr %279, align 8, !noundef !5\r\n %281 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_13, i32 0, i32 3\r\n store i64 %280, ptr %281, align 8\r\n %282 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_14, i32 0, i32 3\r\n %283 = getelementptr inbounds %Adt59, ptr %_7, i32 0, i32 4\r\n %284 = load i16, ptr %283, align 4, !noundef !5\r\n %285 = zext i16 %284 to i64\r\n store i64 %285, ptr %282, align 8\r\n %286 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %287 = getelementptr inbounds %Adt50, ptr %286, i32 0, i32 3\r\n %288 = load i64, ptr %_10, align 16, !noundef !5\r\n %289 = sext i64 %288 to i128\r\n store i128 %289, ptr %287, align 16\r\n %290 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 1\r\n %291 = getelementptr inbounds { i128, i16, i8, [1 x i8], [1 x i16], [5 x i16] }, ptr %290, i32 0, i32 2\r\n %292 = load i8, ptr %291, align 2, !range !6, !noundef !5\r\n %293 = trunc i8 %292 to i1\r\n %294 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %295 = load i8, ptr %294, align 8, !range !6, !noundef !5\r\n %296 = trunc i8 %295 to i1\r\n %297 = and i1 %293, %296\r\n %298 = zext i1 %297 to i8\r\n store i8 %298, ptr %_16, align 1\r\n %299 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %300 = getelementptr inbounds { i64, i64, i64 }, ptr %299, i32 0, i32 2\r\n %301 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 3\r\n %302 = load i64, ptr %301, align 8, !noundef !5\r\n %303 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_14, i32 0, i32 3\r\n %304 = load i64, ptr %303, align 8, !noundef !5\r\n %305 = xor i64 %302, %304\r\n store i64 %305, ptr %300, align 8\r\n %306 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_13, i32 0, i32 1\r\n %307 = load i8, ptr %306, align 2, !noundef !5\r\n %308 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_14, i32 0, i32 1\r\n store i8 %307, ptr %308, align 2\r\n %309 = load i16, ptr %_1, align 8, !noundef !5\r\n store i16 %309, ptr %_14, align 8\r\n %310 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %311 = getelementptr inbounds { i64, i64, i64 }, ptr %310, i32 0, i32 2\r\n %312 = load i64, ptr %311, align 8, !noundef !5\r\n %313 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_14, i32 0, i32 3\r\n store i64 %312, ptr %313, align 8\r\n %314 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %315 = getelementptr inbounds %Adt50, ptr %314, i32 0, i32 5\r\n %316 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %317 = getelementptr inbounds %Adt50, ptr %316, i32 0, i32 3\r\n %318 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %317, i32 0, i32 1\r\n %319 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %318, i32 0, i32 3\r\n %320 = load i32, ptr %319, align 4, !noundef !5\r\n %321 = getelementptr inbounds [7 x i32], ptr %315, i64 0, i64 0\r\n store i32 %320, ptr %321, align 8\r\n %322 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %323 = getelementptr inbounds %Adt50, ptr %322, i32 0, i32 3\r\n %324 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %323, i32 0, i32 1\r\n %325 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %324, i32 0, i32 3\r\n %326 = load i32, ptr %325, align 4, !noundef !5\r\n %327 = getelementptr inbounds [7 x i32], ptr %315, i64 0, i64 1\r\n store i32 %326, ptr %327, align 4\r\n %328 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %329 = getelementptr inbounds %Adt50, ptr %328, i32 0, i32 3\r\n %330 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %329, i32 0, i32 1\r\n %331 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %330, i32 0, i32 3\r\n %332 = load i32, ptr %331, align 4, !noundef !5\r\n %333 = getelementptr inbounds [7 x i32], ptr %315, i64 0, i64 2\r\n store i32 %332, ptr %333, align 8\r\n %334 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %335 = getelementptr inbounds %Adt50, ptr %334, i32 0, i32 3\r\n %336 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %335, i32 0, i32 1\r\n %337 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %336, i32 0, i32 3\r\n %338 = load i32, ptr %337, align 4, !noundef !5\r\n %339 = getelementptr inbounds [7 x i32], ptr %315, i64 0, i64 3\r\n store i32 %338, ptr %339, align 4\r\n %340 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %341 = getelementptr inbounds %Adt50, ptr %340, i32 0, i32 3\r\n %342 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %341, i32 0, i32 1\r\n %343 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %342, i32 0, i32 3\r\n %344 = load i32, ptr %343, align 4, !noundef !5\r\n %345 = getelementptr inbounds [7 x i32], ptr %315, i64 0, i64 4\r\n store i32 %344, ptr %345, align 8\r\n %346 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %347 = getelementptr inbounds %Adt50, ptr %346, i32 0, i32 3\r\n %348 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %347, i32 0, i32 1\r\n %349 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %348, i32 0, i32 3\r\n %350 = load i32, ptr %349, align 4, !noundef !5\r\n %351 = getelementptr inbounds [7 x i32], ptr %315, i64 0, i64 5\r\n store i32 %350, ptr %351, align 4\r\n %352 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %353 = getelementptr inbounds %Adt50, ptr %352, i32 0, i32 3\r\n %354 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %353, i32 0, i32 1\r\n %355 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %354, i32 0, i32 3\r\n %356 = load i32, ptr %355, align 4, !noundef !5\r\n %357 = getelementptr inbounds [7 x i32], ptr %315, i64 0, i64 6\r\n store i32 %356, ptr %357, align 8\r\n %358 = getelementptr inbounds %Adt50, ptr %_15, i32 0, i32 3\r\n %359 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %358, i32 0, i32 1\r\n %360 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %359, i32 0, i32 5\r\n %361 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %362 = getelementptr inbounds %Adt50, ptr %361, i32 0, i32 3\r\n %363 = load i128, ptr %362, align 16, !noundef !5\r\n store i128 %363, ptr %360, align 16\r\n %364 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %365 = getelementptr inbounds %Adt50, ptr %364, i32 0, i32 3\r\n %366 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %365, i32 0, i32 1\r\n %367 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %366, i32 0, i32 1\r\n %368 = load double, ptr %_7, align 8, !noundef !5\r\n store double %368, ptr %367, align 8\r\n %369 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %370 = getelementptr inbounds %Adt50, ptr %369, i32 0, i32 3\r\n %371 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %370, i32 0, i32 2\r\n %372 = load double, ptr %371, align 16, !noundef !5\r\n store double %372, ptr %_7, align 8\r\n %373 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 1\r\n %374 = load i8, ptr %373, align 2, !noundef !5\r\n switch i8 %374, label %bb4 [\r\n i8 0, label %bb1\r\n i8 84, label %bb5\r\n ]\r\n\r\nbb5: ; preds = %bb4\r\n %375 = getelementptr inbounds %Adt50, ptr %_15, i32 0, i32 3\r\n %376 = getelementptr inbounds { i128, { i64, double, i32, i32, [2 x i32], i128 }, double, [6 x i32] }, ptr %375, i32 0, i32 1\r\n %377 = getelementptr inbounds { i64, double, i32, i32, [2 x i32], i128 }, ptr %376, i32 0, i32 5\r\n %378 = load i128, ptr %377, align 16, !noundef !5\r\n store i128 %378, ptr %_5, align 16\r\n %379 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %380 = load i64, ptr %379, align 8, !noundef !5\r\n store i64 %380, ptr %_10, align 16\r\n %381 = load i8, ptr %_16, align 1, !range !6, !noundef !5\r\n %382 = trunc i8 %381 to i1\r\n %383 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %384 = zext i1 %382 to i8\r\n store i8 %384, ptr %383, align 8\r\n %385 = load i64, ptr %_10, align 16, !noundef !5\r\n switch i64 %385, label %bb8 [\r\n i64 2, label %bb6\r\n i64 7564290096487423208, label %bb7\r\n ]\r\n\r\nbb8: ; preds = %bb7, %bb5\r\n ret void\r\n\r\nbb6: ; preds = %bb5\r\n %386 = getelementptr inbounds %Adt51, ptr %_10, i32 0, i32 2\r\n %387 = getelementptr inbounds %Adt50, ptr %386, i32 0, i32 1\r\n %388 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %387, i32 0, i32 1\r\n %389 = getelementptr inbounds { i16, i8, [5 x i8], i64 }, ptr %_1, i32 0, i32 1\r\n %390 = load i8, ptr %389, align 2, !noundef !5\r\n store i8 %390, ptr %388, align 2\r\n %391 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %392 = load i64, ptr %391, align 8, !noundef !5\r\n %393 = getelementptr inbounds { double, { i64, i64, i64 }, i16, i8, [5 x i8] }, ptr %_7, i32 0, i32 1\r\n %394 = load i64, ptr %393, align 8, !noundef !5\r\n %395 = call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %392, i64 %394)\r\n %396 = extractvalue { i64, i1 } %395, 0\r\n %397 = extractvalue { i64, i1 } %395, 1\r\n store i64 %396, ptr %_10, align 16\r\n %398 = getelementptr inbounds i8, ptr %_10, i64 8\r\n %399 = zext i1 %397 to i8\r\n store i8 %399, ptr %398, align 8\r\n br label %bb3\r\n\r\nbb7: ; preds = %bb5\r\n %400 = load i128, ptr %_5, align 16, !noundef !5\r\n; call _1851685_debug_2::bbox2\r\n call void @_ZN16_1851685_debug_25bbox217he0199f2e11de3e04E(i128 %400, ptr align 8 %_14)\r\n br label %bb8\r\n}\r\n\r\n; _1851685_debug_2::fn15\r\n; Function Attrs: nonlazybind uwtable\r\ndefine internal i128 @_ZN16_1851685_debug_24fn1517hfe3f07cbb5faefb0E() unnamed_addr #1 {\r\nstart:\r\n %_0 = alloca i128, align 16\r\n store i128 0, ptr %_0, align 16\r\n; call core::hint::black_box\r\n call void @_ZN4core4hint9black_box17ha89c3f9cb79c398cE()\r\n %0 = load i128, ptr %_0, align 16, !noundef !5\r\n ret i128 %0\r\n}\r\n\r\n; _1851685_debug_2::main\r\n; Function Attrs: nonlazybind uwtable\r\ndefine internal void @_ZN16_1851685_debug_24main17he76cece459490d8aE() unnamed_addr #1 {\r\nstart:\r\n; call _1851685_debug_2::enter\r\n call void @_ZN16_1851685_debug_25enter17h7e9cae7a334c5019E()\r\n ret void\r\n}\r\n\r\n; _1851685_debug_2::bbox2\r\n; Function Attrs: noinline nonlazybind uwtable\r\ndefine internal void @_ZN16_1851685_debug_25bbox217he0199f2e11de3e04E(i128 %val1, ptr align 8 %val2) unnamed_addr #0 {\r\nstart:\r\n %_6 = alloca { i16, i8, [5 x i8], i64 }, align 8\r\n %_5 = alloca { i16, i8, [5 x i8], i64 }, align 8\r\n; call _1851685_debug_2::print_i128\r\n call void @_ZN16_1851685_debug_210print_i12817h036b4f930abaf46cE(i128 %val1)\r\n call void @llvm.memcpy.p0.p0.i64(ptr align 8 %_6, ptr align 8 %val2, i64 16, i1 false)\r\n; call core::hint::black_box\r\n call void @_ZN4core4hint9black_box17h7d48bc7feb0af96eE(ptr sret({ i16, i8, [5 x i8], i64 }) align 8 %_5, ptr align 8 %_6)\r\n ret void\r\n}\r\n\r\n; _1851685_debug_2::print_i128\r\n; Function Attrs: nonlazybind uwtable\r\ndefine internal void @_ZN16_1851685_debug_210print_i12817h036b4f930abaf46cE(i128 %x) unnamed_addr #1 {\r\nstart:\r\n %_9 = lshr i128 %x, 64\r\n %_8 = trunc i128 %_9 to i64\r\n %_14 = trunc i128 %x to i64\r\n %_2 = call i32 (ptr, ...) @printf(ptr @alloc_2cdd8ac758b4c372b56d8683306290e0, i64 %_8, i64 %_14)\r\n ret void\r\n}\r\n\r\n; Function Attrs: nonlazybind uwtable\r\ndefine i32 @rust_eh_personality(i32, i32, i64, ptr, ptr) unnamed_addr #1 {\r\nstart:\r\n ret i32 0\r\n}\r\n\r\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\r\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #3\r\n\r\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\r\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #4\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare { i64, i1 } @llvm.ssub.with.overflow.i64(i64, i64) #5\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare { i64, i1 } @llvm.smul.with.overflow.i64(i64, i64) #5\r\n\r\n; Function Attrs: nonlazybind uwtable\r\ndeclare i32 @printf(ptr, ...) unnamed_addr #1\r\n\r\n; Function Attrs: nonlazybind\r\ndefine i32 @main(i32 %0, ptr %1) unnamed_addr #6 {\r\ntop:\r\n call void @_ZN16_1851685_debug_25enter17h7e9cae7a334c5019E()\r\n ret i32 0\r\n}\r\n\r\nattributes #0 = { noinline nonlazybind uwtable \"probe-stack\"=\"inline-asm\" \"target-cpu\"=\"x86-64\" }\r\nattributes #1 = { nonlazybind uwtable \"probe-stack\"=\"inline-asm\" \"target-cpu\"=\"x86-64\" }\r\nattributes #2 = { inlinehint nonlazybind uwtable \"probe-stack\"=\"inline-asm\" \"target-cpu\"=\"x86-64\" }\r\nattributes #3 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\r\nattributes #4 = { nocallback nofree nounwind willreturn memory(argmem: write) }\r\nattributes #5 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\nattributes #6 = { nonlazybind \"target-cpu\"=\"x86-64\" }\r\n\r\n!llvm.module.flags = !{!0, !1, !2}\r\n!llvm.ident = !{!3}\r\n\r\n!0 = !{i32 8, !\"PIC Level\", i32 2}\r\n!1 = !{i32 7, !\"PIE Level\", i32 2}\r\n!2 = !{i32 2, !\"RtLibUseGOT\", i32 1}\r\n!3 = !{!\"rustc version 1.78.0-dev\"}\r\n!4 = !{i32 753499}\r\n!5 = !{}\r\n!6 = !{i8 0, i8 2}\r\n\r\n```\r\n\r\nRight:\r\n```console\r\n$ clang -O1 1851685_debug_2.ll && ./a.out\r\nhi:-1 lo:-2580139605333929408\r\n```\r\n\r\nWrong:\r\n```console\r\n$ clang -O2 1851685_debug_2.ll && ./a.out\r\nhi:0 lo:-2580139605333929408\r\n```\r\n\r\nBisects down to \r\n```\r\nBISECT: running pass (504) InstCombinePass on _ZN16_1851685_debug_25enter17h7e9cae7a334c5019E\r\n```\r\n\r\nVersion:\r\n```\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git cd1d4d8dd31f527615de26f5b62d687c6b2982a6)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /opt/compiler-explorer/clang-trunk/bin\r\nCompiler returned: 0\r\n```\r\n\r\n\r\n
\r\nFrom Rust custom MIR\r\n\r\n```rust\r\n#![feature(custom_mir, core_intrinsics)]\r\nextern crate core;\r\nuse core::intrinsics::mir::*;\r\n\r\npub fn enter() {\r\n fn14((0, 0, 0), [0; 6]);\r\n}\r\n\r\n#[custom_mir(dialect = \"runtime\", phase = \"initial\")]\r\npub fn fn14(mut _3: (i8, u16, isize), mut _6: [i32; 6]) {\r\n mir! {\r\n let _1: u128;\r\n let _2: i128;\r\n let _4: (i128,);\r\n let _5: i128;\r\n let _9: Adt59;\r\n let _11: isize;\r\n let _12: bool;\r\n let _13: Adt51;\r\n let _15: isize;\r\n let _16: Adt53;\r\n let _21: (i32, [i32; 6], (i8, u16, isize), [usize; 4]);\r\n let _23: (i8, u16, isize);\r\n let _25: Adt58;\r\n let _26: Adt52;\r\n let _44: ();\r\n {\r\n _2 = 0;\r\n _5 = 0;\r\n _3 = (84_i8, 40096_u16, (-9223372036854775808_isize));\r\n Goto(bb2)\r\n }\r\n bb1 = {\r\n Return()\r\n }\r\n bb2 = {\r\n _9.fld3.1.0.0 = 145713751_u32 as f64;\r\n _9.fld5 = _3.1 % 62066_u16;\r\n _13.fld1 = Checked(3029947793098001587_i64 - (-3638983068927244405_i64));\r\n _12 = _13.fld1.1 ^ _13.fld1.1;\r\n _9.fld6 = core::ptr::addr_of!(_13.fld1);\r\n _9.fld4 = (16948196345300582164_u64,);\r\n _13.fld3.fld4 = [168516996_i32,245529160_i32,1223136339_i32,1622354325_i32,(-1002348909_i32),1976405017_i32,(-1442509311_i32),1165829790_i32];\r\n _13.fld3.fld6.1 = -_9.fld3.1.0.0;\r\n _11 = !_3.2;\r\n _13.fld1 = (7564290096487423208_i64, _12);\r\n _13.fld3.fld6.2.0 = _11;\r\n _9.fld3.2.0 = _13.fld1.0;\r\n Call(_13.fld3.fld6.2.4 = fn15(), ReturnTo(bb3), UnwindUnreachable())\r\n }\r\n bb3 = {\r\n _13.fld0.0 = _9.fld3.2.0 as i16;\r\n _13.fld2 = _9.fld4.0 as isize;\r\n _13.fld3.fld5.1.0 = _13.fld3.fld6.2.4.0 >> _2;\r\n _9.fld3.1.0.0 = -_13.fld3.fld6.1;\r\n _13.fld0.2 = !_13.fld1.1;\r\n _13.fld3.fld5.3 = [_13.fld0.0];\r\n Goto(bb4)\r\n }\r\n bb4 = {\r\n _13.fld3.fld3.0 = _3.0;\r\n _6 = [(-154507050_i32),2086635898_i32,468582002_i32,(-21184342_i32),(-1173629817_i32),(-2078649110_i32)];\r\n _13.fld3.fld5.1.0 = _5;\r\n _16.fld0.2 = _13.fld0.2;\r\n _13.fld3.fld5.0 = _13.fld0.0 * _13.fld0.0;\r\n _9.fld2 = 2797771210_u32 + 2127377746_u32;\r\n _9.fld3.1.0.0 = _9.fld2 as f64;\r\n _15 = -_11;\r\n _13.fld0 = (_13.fld3.fld5.0, _13.fld3.fld5.1, _12, _13.fld3.fld5.3);\r\n _13.fld1 = Checked(_9.fld3.2.0 * _9.fld3.2.0);\r\n Goto(bb5)\r\n }\r\n bb5 = {\r\n _13.fld3.fld6.1 = _9.fld3.1.0.0 / 0.000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000012739273109665954_f64;\r\n _13.fld4 = !6496701646011298799_usize;\r\n _3 = (_13.fld3.fld3.0, _9.fld5, _15);\r\n _13.fld3.fld6.2.1 = _9.fld2;\r\n _13.fld0.2 = _13.fld0.1.0 >= _5;\r\n _13.fld3.fld3.2 = _3.2 & _11;\r\n _13.fld6 = [_13.fld0.2,_13.fld1.1,_16.fld0.2,_16.fld0.2,_13.fld0.2,_13.fld1.1];\r\n _13.fld3.fld6.2.4 = _13.fld3.fld5.1;\r\n _9.fld3.0 = -_13.fld0.0;\r\n _13.fld3.fld0 = core::ptr::addr_of!(_9.fld4.0);\r\n _9.fld0 = [_3.0];\r\n _13.fld5 = _13.fld3.fld4;\r\n _13.fld3.fld6.2.3 = _13.fld4 as i32;\r\n _3.2 = _13.fld3.fld3.2 ^ _13.fld3.fld6.2.0;\r\n _13.fld3.fld6.1 = _9.fld3.1.0.0;\r\n _23.0 = !_3.0;\r\n _13.fld3.fld3.1 = _13.fld1.1 as u16;\r\n _13.fld3.fld5.1.0 = _13.fld3.fld6.2.3 as i128;\r\n _13.fld3.fld6.2.3 = _13.fld3.fld3.1 as i32;\r\n Goto(bb7)\r\n }\r\n bb7 = {\r\n _23.0 = _13.fld3.fld6.2.3 as i8;\r\n _21.2 = (_23.0, _3.1, _13.fld3.fld3.2);\r\n _23.2 = _9.fld5 as isize;\r\n _13.fld3.fld6.0 = _13.fld1.0 as i128;\r\n _26.fld0 = _13.fld0.2 & _13.fld1.1;\r\n _9.fld3.2.2 = _3.2 ^ _23.2;\r\n _23 = (_21.2.0, _3.1, _9.fld3.2.2);\r\n _13.fld3.fld2 = [_13.fld3.fld6.2.3,_13.fld3.fld6.2.3,_13.fld3.fld6.2.3,_13.fld3.fld6.2.3,_13.fld3.fld6.2.3,_13.fld3.fld6.2.3,_13.fld3.fld6.2.3];\r\n _25.fld3.fld6.2.4 = (_13.fld3.fld6.0,);\r\n _13.fld3.fld6.2.2 = _9.fld3.1.0.0;\r\n _9.fld3.1.0.0 = _13.fld3.fld6.1;\r\n match _3.0 {\r\n 0 => bb2,\r\n 84 => bb10,\r\n _ => bb7\r\n }\r\n }\r\n bb10 = {\r\n _4 = (_25.fld3.fld6.2.4.0,);\r\n _13.fld1 = (_9.fld3.2.0, _26.fld0);\r\n match _13.fld1.0 {\r\n 2 => bb14,\r\n 7564290096487423208 => bb18,\r\n _ => bb19\r\n }\r\n }\r\n bb14 = {\r\n _13.fld3.fld3.0 = _3.0;\r\n _13.fld1 = Checked(_9.fld3.2.0 * _9.fld3.2.0);\r\n Goto(bb5)\r\n }\r\n bb18 = {\r\n Call(_44 = bbox2(_4, Move(_23)), ReturnTo(bb19), UnwindUnreachable())\r\n }\r\n bb19 = {\r\n Return()\r\n }\r\n\r\n }\r\n}\r\n\r\n#[custom_mir(dialect = \"runtime\", phase = \"initial\")]\r\nfn fn15() -> (i128,) {\r\n mir! {\r\n let unit:();\r\n let _37: ();\r\n {\r\n RET = (0,);\r\n Call(_37 = core::hint::black_box(unit), ReturnTo(bb16), UnwindUnreachable())\r\n }\r\n bb16 = {\r\n Return()\r\n }\r\n\r\n }\r\n}\r\npub fn main() {\r\n enter();\r\n}\r\n\r\n#[derive(Debug)]\r\npub struct Adt49 {}\r\n#[derive(Debug)]\r\npub struct Adt50 {\r\n fld0: *const u64,\r\n fld2: [i32; 7],\r\n fld3: (i8, u16, isize),\r\n fld4: [i32; 8],\r\n fld5: (i16, (i128,), bool, [i16; 1]),\r\n fld6: (i128, f64, (isize, u32, f64, i32, (i128,)), [i32; 6]),\r\n}\r\n#[derive(Debug)]\r\npub struct Adt51 {\r\n fld0: (i16, (i128,), bool, [i16; 1]),\r\n fld1: (i64, bool),\r\n fld2: isize,\r\n fld3: Adt50,\r\n fld4: usize,\r\n fld5: [i32; 8],\r\n fld6: [bool; 6],\r\n}\r\n#[derive(Debug)]\r\npub struct Adt52 {\r\n fld0: bool,\r\n}\r\n#[derive(Debug)]\r\npub struct Adt53 {\r\n fld0: (i16, (i128,), bool, [i16; 1]),\r\n}\r\n#[derive(Debug)]\r\npub struct Adt58 {\r\n fld3: Adt50,\r\n}\r\n#[derive(Debug, Copy, Clone)]\r\npub struct Adt59 {\r\n fld0: [i8; 1],\r\n fld2: u32,\r\n fld3: (i16, ((f64,),), (i64, isize, isize), u8),\r\n fld4: (u64,),\r\n fld5: u16,\r\n fld6: *const (i64, bool),\r\n}\r\n\r\n#[inline(never)]\r\npub fn bbox2(val1: (i128,), val2: (i8, u16, isize)) {\r\n use core::hint::black_box;\r\n print_i128(val1.0);\r\n black_box(val2);\r\n}\r\n\r\n#[cfg(not(miri))]\r\nfn print_i128(x: i128) {\r\n extern \"C\" {\r\n fn printf(fmt: *const core::ffi::c_char, ...) -> core::ffi::c_int;\r\n }\r\n unsafe {\r\n printf(\r\n b\"hi:%ld lo:%ld\\n\\0\".as_ptr().cast(),\r\n (x as u128 >> 64) as i64,\r\n x as i64,\r\n );\r\n }\r\n}\r\n\r\n#[cfg(miri)]\r\nfn print_i128(x: i128) {\r\n println!(\"hi:{} lo:{}\", (x as u128 >> 64) as i64, x as i64);\r\n}\r\n\r\n```\r\n\r\n
\r\n", "author": "cbeuw", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://godbolt.org/z/xs1rhnMvb\r\nAlive2: https://alive2.llvm.org/ce/z/gyL7mn\r\n\r\nI will post a fix later.\r\n"}]}, "verified": true} {"bug_id": "102351", "issue_url": "https://github.com/llvm/llvm-project/issues/102351", "bug_type": "miscompilation", "base_commit": "203a2ca8cd6af505e11a38aebceeaf864271042c", "knowledge_cutoff": "2024-08-07T18:49:59Z", "lit_test_dir": ["llvm/test/Transforms/SimplifyCFG"], "hints": {"fix_commit": "3c9022c965b85951f30af140da591f819acef8a0", "components": ["Local"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/Local.cpp": [[1028, 1034], [1133, 1139]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/Local.cpp": ["CanRedirectPredsOfEmptyBBToSucc", "llvm::TryToSimplifyUncondBranchFromEmptyBlock"]}}, "patch": "commit 3c9022c965b85951f30af140da591f819acef8a0\nAuthor: AdityaK \nDate: Tue Sep 10 22:39:02 2024 -0700\n\n Bail out jump threading on indirect branches (#103688)\n \n The bug was introduced by\n https://github.com/llvm/llvm-project/pull/68473\n \n Fixes: #102351\n\ndiff --git a/llvm/lib/Transforms/Utils/Local.cpp b/llvm/lib/Transforms/Utils/Local.cpp\nindex d0669e44f821..c85c819263e2 100644\n--- a/llvm/lib/Transforms/Utils/Local.cpp\n+++ b/llvm/lib/Transforms/Utils/Local.cpp\n@@ -1028,7 +1028,14 @@ CanRedirectPredsOfEmptyBBToSucc(BasicBlock *BB, BasicBlock *Succ,\n if (!BB->hasNPredecessorsOrMore(2))\n return false;\n \n- // Get single common predecessors of both BB and Succ\n+ if (any_of(BBPreds, [](const BasicBlock *Pred) {\n+ return isa(Pred->begin()) &&\n+ isa(Pred->getTerminator());\n+ }))\n+ return false;\n+\n+ // Get the single common predecessor of both BB and Succ. Return false\n+ // when there are more than one common predecessors.\n for (BasicBlock *SuccPred : SuccPreds) {\n if (BBPreds.count(SuccPred)) {\n if (CommonPred)\n@@ -1133,7 +1140,7 @@ bool llvm::TryToSimplifyUncondBranchFromEmptyBlock(BasicBlock *BB,\n \n bool BBKillable = CanPropagatePredecessorsForPHIs(BB, Succ, BBPreds);\n \n- // Even if we can not fold bB into Succ, we may be able to redirect the\n+ // Even if we can not fold BB into Succ, we may be able to redirect the\n // predecessors of BB to Succ.\n bool BBPhisMergeable =\n BBKillable ||\n", "tests": [{"file": "llvm/test/Transforms/SimplifyCFG/switch-branch-fold-indirectbr-102351.ll", "commands": ["opt < %s -passes=simplifycfg -S"], "tests": [{"test_name": "main", "test_body": "@.str = private unnamed_addr constant [23 x i8] c\"OP_1:(instruction=%d)\\0A\\00\", align 1\n@.str.1 = private unnamed_addr constant [28 x i8] c\"TERMINATE:(instruction=%d)\\0A\\00\", align 1\n\n; Function Attrs: mustprogress norecurse uwtable\ndefine dso_local noundef i32 @main() #0 {\nentry:\n %bytecode = alloca [2 x ptr], align 16\n store ptr blockaddress(@main, %VM__OP_1), ptr %bytecode, align 16, !tbaa !5\n %arrayidx1 = getelementptr inbounds [2 x ptr], ptr %bytecode, i64 0, i64 1\n store ptr blockaddress(@main, %VM__TERMINATE), ptr %arrayidx1, align 8, !tbaa !5\n br label %while.body\n\nwhile.body: ; preds = %entry, %sw.epilog\n %state.0 = phi i32 [ 0, %entry ], [ %state.1, %sw.epilog ]\n %index.0 = phi i32 [ 0, %entry ], [ %index.2, %sw.epilog ]\n switch i32 %state.0, label %sw.epilog [\n i32 0, label %sw.bb\n i32 1, label %VM__OP_1\n i32 2, label %sw.bb4\n ]\n\nsw.bb: ; preds = %while.body\n br label %indirectgoto\n\nVM__OP_1: ; preds = %while.body, %indirectgoto\n %index.1 = phi i32 [ %index.3, %indirectgoto ], [ %index.0, %while.body ]\n br label %sw.epilog\n\nsw.bb4: ; preds = %while.body\n %call = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(1) @.str, i32 noundef %index.0)\n %inc = add nsw i32 %index.0, 1\n br label %indirectgoto\n\nsw.epilog: ; preds = %while.body, %VM__OP_1\n %state.1 = phi i32 [ %state.0, %while.body ], [ 2, %VM__OP_1 ]\n %index.2 = phi i32 [ %index.0, %while.body ], [ %index.1, %VM__OP_1 ]\n br label %while.body, !llvm.loop !9\n\nVM__TERMINATE: ; preds = %indirectgoto\n %call7 = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(1) @.str.1, i32 noundef %index.3)\n ret i32 0\n\nindirectgoto: ; preds = %sw.bb4, %sw.bb\n %index.3 = phi i32 [ %inc, %sw.bb4 ], [ %index.0, %sw.bb ]\n %idxprom.pn = sext i32 %index.3 to i64\n %indirect.goto.dest.in = getelementptr inbounds [2 x ptr], ptr %bytecode, i64 0, i64 %idxprom.pn\n %indirect.goto.dest = load ptr, ptr %indirect.goto.dest.in, align 8, !tbaa !5\n indirectbr ptr %indirect.goto.dest, [label %VM__OP_1, label %VM__TERMINATE]\n}\ndeclare i32 @printf(ptr noundef, ...) #1\n\nattributes #0 = { mustprogress norecurse uwtable \"min-legal-vector-width\"=\"0\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\nattributes #1 = { \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\n\n!llvm.module.flags = !{!0, !1, !2, !3}\n!llvm.ident = !{!4}\n\n!0 = !{i32 1, !\"wchar_size\", i32 4}\n!1 = !{i32 8, !\"PIC Level\", i32 2}\n!2 = !{i32 7, !\"PIE Level\", i32 2}\n!3 = !{i32 7, !\"uwtable\", i32 2}\n!4 = !{!\"clang version 18.0.0git (https://github.com/llvm/llvm-project.git 67782d2de5ea9c8653b8f0110237a3c355291c0e)\"}\n!5 = !{!6, !6, i64 0}\n!6 = !{!\"any pointer\", !7, i64 0}\n!7 = !{!\"omnipotent char\", !8, i64 0}\n!8 = !{!\"Simple C++ TBAA\"}\n!9 = !{!10, !10, i64 0}\n!10 = !{!\"int\", !7, i64 0}\n!11 = distinct !{!11, !12, !13}\n!12 = !{!\"llvm.loop.mustprogress\"}\n!13 = !{!\"llvm.loop.unroll.disable\"}", "lli_expected_out": "OP_1:(instruction=0)\nTERMINATE:(instruction=1)\n"}]}], "issue": {"title": "Miscompile in code with indirect gotos", "body": "The simplifycfg pass (incorrectly?) optimizes one of the case statements.\r\n\r\n```llvm\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\n@.str = private unnamed_addr constant [23 x i8] c\"OP_1:(instruction=%d)\\0A\\00\", align 1\r\n@.str.1 = private unnamed_addr constant [28 x i8] c\"TERMINATE:(instruction=%d)\\0A\\00\", align 1\r\n\r\n; Function Attrs: mustprogress norecurse uwtable\r\ndefine dso_local noundef i32 @main() #0 {\r\nentry:\r\n %bytecode = alloca [2 x ptr], align 16\r\n store ptr blockaddress(@main, %VM__OP_1), ptr %bytecode, align 16, !tbaa !5\r\n %arrayidx1 = getelementptr inbounds [2 x ptr], ptr %bytecode, i64 0, i64 1\r\n store ptr blockaddress(@main, %VM__TERMINATE), ptr %arrayidx1, align 8, !tbaa !5\r\n br label %while.body\r\n\r\nwhile.body: ; preds = %entry, %sw.epilog\r\n %state.0 = phi i32 [ 0, %entry ], [ %state.1, %sw.epilog ]\r\n %index.0 = phi i32 [ 0, %entry ], [ %index.2, %sw.epilog ]\r\n switch i32 %state.0, label %sw.epilog [\r\n i32 0, label %sw.bb\r\n i32 1, label %VM__OP_1\r\n i32 2, label %sw.bb4\r\n ]\r\n\r\nsw.bb: ; preds = %while.body\r\n br label %indirectgoto\r\n\r\nVM__OP_1: ; preds = %while.body, %indirectgoto\r\n %index.1 = phi i32 [ %index.3, %indirectgoto ], [ %index.0, %while.body ]\r\n br label %sw.epilog\r\n\r\nsw.bb4: ; preds = %while.body\r\n %call = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(1) @.str, i32 noundef %index.0)\r\n %inc = add nsw i32 %index.0, 1\r\n br label %indirectgoto\r\n\r\nsw.epilog: ; preds = %while.body, %VM__OP_1\r\n %state.1 = phi i32 [ %state.0, %while.body ], [ 2, %VM__OP_1 ]\r\n %index.2 = phi i32 [ %index.0, %while.body ], [ %index.1, %VM__OP_1 ]\r\n br label %while.body, !llvm.loop !9\r\n\r\nVM__TERMINATE: ; preds = %indirectgoto\r\n %call7 = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(1) @.str.1, i32 noundef %index.3)\r\n ret i32 0\r\n\r\nindirectgoto: ; preds = %sw.bb4, %sw.bb\r\n %index.3 = phi i32 [ %inc, %sw.bb4 ], [ %index.0, %sw.bb ]\r\n %idxprom.pn = sext i32 %index.3 to i64\r\n %indirect.goto.dest.in = getelementptr inbounds [2 x ptr], ptr %bytecode, i64 0, i64 %idxprom.pn\r\n %indirect.goto.dest = load ptr, ptr %indirect.goto.dest.in, align 8, !tbaa !5\r\n indirectbr ptr %indirect.goto.dest, [label %VM__OP_1, label %VM__TERMINATE]\r\n}\r\ndeclare i32 @printf(ptr noundef, ...) #1\r\n\r\nattributes #0 = { mustprogress norecurse uwtable \"min-legal-vector-width\"=\"0\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\r\nattributes #1 = { \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\r\n\r\n!llvm.module.flags = !{!0, !1, !2, !3}\r\n!llvm.ident = !{!4}\r\n\r\n!0 = !{i32 1, !\"wchar_size\", i32 4}\r\n!1 = !{i32 8, !\"PIC Level\", i32 2}\r\n!2 = !{i32 7, !\"PIE Level\", i32 2}\r\n!3 = !{i32 7, !\"uwtable\", i32 2}\r\n!4 = !{!\"clang version 18.0.0git (https://github.com/llvm/llvm-project.git 67782d2de5ea9c8653b8f0110237a3c355291c0e)\"}\r\n!5 = !{!6, !6, i64 0}\r\n!6 = !{!\"any pointer\", !7, i64 0}\r\n!7 = !{!\"omnipotent char\", !8, i64 0}\r\n!8 = !{!\"Simple C++ TBAA\"}\r\n!9 = !{!10, !10, i64 0}\r\n!10 = !{!\"int\", !7, i64 0}\r\n!11 = distinct !{!11, !12, !13}\r\n!12 = !{!\"llvm.loop.mustprogress\"}\r\n!13 = !{!\"llvm.loop.unroll.disable\"}\r\n```\r\n\r\n$ opt -passes=simplifycfg orig.ll -S -o out.ll\r\n```llvm\r\n; ModuleID = 'orig.ll'\r\nsource_filename = \"orig.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\n@.str = private unnamed_addr constant [23 x i8] c\"OP_1:(instruction=%d)\\0A\\00\", align 1\r\n@.str.1 = private unnamed_addr constant [28 x i8] c\"TERMINATE:(instruction=%d)\\0A\\00\", align 1\r\n\r\n; Function Attrs: mustprogress norecurse uwtable\r\ndefine dso_local noundef i32 @main() #0 {\r\nentry:\r\n %bytecode = alloca [2 x ptr], align 16\r\n store ptr blockaddress(@main, %VM__OP_1), ptr %bytecode, align 16, !tbaa !5\r\n %arrayidx1 = getelementptr inbounds [2 x ptr], ptr %bytecode, i64 0, i64 1\r\n store ptr blockaddress(@main, %VM__TERMINATE), ptr %arrayidx1, align 8, !tbaa !5\r\n br label %while.body\r\n\r\nwhile.body: ; preds = %sw.epilog, %entry\r\n %state.0 = phi i32 [ 0, %entry ], [ %state.1, %sw.epilog ]\r\n %index.0 = phi i32 [ 0, %entry ], [ %index.0, %sw.epilog ]\r\n switch i32 %state.0, label %sw.epilog [\r\n i32 0, label %VM__TERMINATE\r\n i32 1, label %VM__OP_1\r\n i32 2, label %sw.bb4\r\n ]\r\n\r\nVM__OP_1: ; preds = %while.body\r\n br label %sw.epilog\r\n\r\nsw.bb4: ; preds = %while.body\r\n %call = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(1) @.str, i32 noundef %index.0)\r\n %inc = add nsw i32 %index.0, 1\r\n br label %VM__TERMINATE\r\n\r\nsw.epilog: ; preds = %VM__OP_1, %while.body\r\n %state.1 = phi i32 [ %state.0, %while.body ], [ 2, %VM__OP_1 ]\r\n br label %while.body, !llvm.loop !9\r\n\r\nVM__TERMINATE: ; preds = %sw.bb4, %while.body\r\n %index.3 = phi i32 [ %inc, %sw.bb4 ], [ %index.0, %while.body ]\r\n %call7 = call i32 (ptr, ...) @printf(ptr noundef nonnull dereferenceable(1) @.str.1, i32 noundef %index.3)\r\n ret i32 0\r\n}\r\n\r\ndeclare i32 @printf(ptr noundef, ...) #1\r\n\r\nattributes #0 = { mustprogress norecurse uwtable \"min-legal-vector-width\"=\"0\" \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\r\nattributes #1 = { \"no-trapping-math\"=\"true\" \"stack-protector-buffer-size\"=\"8\" \"target-cpu\"=\"x86-64\" \"target-features\"=\"+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87\" \"tune-cpu\"=\"generic\" }\r\n\r\n!llvm.module.flags = !{!0, !1, !2, !3}\r\n!llvm.ident = !{!4}\r\n\r\n!0 = !{i32 1, !\"wchar_size\", i32 4}\r\n!1 = !{i32 8, !\"PIC Level\", i32 2}\r\n!2 = !{i32 7, !\"PIE Level\", i32 2}\r\n!3 = !{i32 7, !\"uwtable\", i32 2}\r\n!4 = !{!\"clang version 18.0.0git (https://github.com/llvm/llvm-project.git 67782d2de5ea9c8653b8f0110237a3c355291c0e)\"}\r\n!5 = !{!6, !6, i64 0}\r\n!6 = !{!\"any pointer\", !7, i64 0}\r\n!7 = !{!\"omnipotent char\", !8, i64 0}\r\n!8 = !{!\"Simple C++ TBAA\"}\r\n!9 = !{!10, !10, i64 0}\r\n!10 = !{!\"int\", !7, i64 0}\r\n```", "author": "hiraditya", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "hiraditya", "body": "Original C code\r\n\r\n```c\r\nextern int printf(const char *fmt, ...);\r\n\r\nint main() {\r\n void* bytecode[2];\r\n bytecode[0] = &&VM__OP_1;\r\n bytecode[1] = &&VM__TERMINATE;\r\n\r\n int state = 0;\r\n int index = 0;\r\n\r\n while (1) {\r\n switch (state) {\r\n case 0:\r\n goto *bytecode[index];\r\n case 1:\r\n // NOTE: THIS IS ONLY REACHABLE VIA INDIRECT GOTOS\r\n VM__OP_1:\r\n state = 2;\r\n break;\r\n case 2:\r\n printf(\"OP_1:(instruction=%d)\\n\", index);\r\n index++;\r\n goto *bytecode[index];\r\n }\r\n }\r\n\r\nVM__TERMINATE:\r\n printf(\"TERMINATE:(instruction=%d)\\n\", index);\r\n return 0;\r\n}\r\n```\r\n\r\nhttps://github.com/llvm/llvm-project/pull/76295 makes the miscompile go away(https://godbolt.org/z/d9bGnhhv6) but I think simplifycfg still has the bug.\r\n"}, {"author": "efriedma-quic", "body": "Looks like TryToSimplifyUncondBranchFromEmptyBlock is not handling indirectbr predecessors correctly: the indirectbr's successor list is rewritten so it has a successor that isn't address-taken. That edge is then proven impossible, so the indirectbr is completely eliminated."}]}, "verified": true} {"bug_id": "109510", "issue_url": "https://github.com/llvm/llvm-project/issues/109510", "bug_type": "crash", "base_commit": "eb6e7e8f89a3d5c1cbc9856774ca00208753fb12", "test_commit": "58e05779b483bb3d9f734bd09f5cb2241431a0b7", "knowledge_cutoff": "2024-09-21T04:01:16Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "bd8fe9972e3f17776e4e05e69e13ab8271d34132", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": [[1158, 1164], [1166, 1171]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp": ["VPlanTransforms::optimize"]}}, "patch": "commit bd8fe9972e3f17776e4e05e69e13ab8271d34132\nAuthor: Florian Hahn \nDate: Sat Sep 21 12:45:36 2024 +0100\n\n [VPlan] Mov licm to end of VPlan optimizations.\n \n This moves licm after expanding replicate regions. This fixes a crash\n when trying to hoist a predicated VPReplicateRecipes which later get\n expanded to replicate regions.\n \n Hoisting replicate regions out was not intended (see the discussion and\n at the review and comment on shallow traversal in licm()).\n \n Fixes https://github.com/llvm/llvm-project/issues/109510.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\nindex 06ac15f52fe1..fd6090affbbf 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp\n@@ -1158,7 +1158,6 @@ void VPlanTransforms::optimize(VPlan &Plan) {\n removeRedundantInductionCasts(Plan);\n \n simplifyRecipes(Plan);\n- licm(Plan);\n legalizeAndOptimizeInductions(Plan);\n removeDeadRecipes(Plan);\n \n@@ -1166,6 +1165,7 @@ void VPlanTransforms::optimize(VPlan &Plan) {\n \n removeRedundantExpandSCEVRecipes(Plan);\n mergeBlocksIntoPredecessors(Plan);\n+ licm(Plan);\n }\n \n // Add a VPActiveLaneMaskPHIRecipe and related recipes to \\p Plan and replace\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/invariant-replicate-region.ll", "commands": ["opt -p loop-vectorize -force-vector-width=4 -S %s"], "tests": [{"test_name": "test_invariant_replicate_region", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128-Fn32\"\ntarget triple = \"arm64-apple-macosx14.0.0\"\n\ndefine i32 @test_invariant_replicate_region(i32 %x, i1 %c) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %then, label %loop.latch\n\nthen: ; preds = %loop.header\n %rem.1 = urem i32 10, %x\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %res = phi i32 [ 0, %loop.header ], [ %rem.1, %then ]\n %iv.next = add i32 %iv, 1\n %ec = icmp eq i32 %iv, 99\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret i32 %res\n}\n"}]}], "issue": {"title": "A/F: verifyVPlanIsValid(*Plan) && \"VPlan is invalid\" after a861ed411a", "body": "We have an internal test that recently started to hit an assertion failure which I bisected back to commit a861ed411a359b7cb2b58d642e4a7f2fdcf92057. \r\n\r\nConsider the following code:\r\n```c\r\nint g_78, g_659, func_32_l_1852;\r\nlong g_128;\r\nunsigned func_32_p_33;\r\nvoid func_32() {\r\n int *l_1145 = &g_659;\r\n *l_1145 = 0;\r\n short __trans_tmp_11;\r\n for (; g_78; g_78 += 1) {\r\n func_32_l_1852 = 0;\r\n for (; func_32_l_1852 <= 3; func_32_l_1852 += 1) {\r\n int *l_1863 = &g_659;\r\n __trans_tmp_11 = *l_1863;\r\n if (10 % ((__trans_tmp_11 >= 0) - func_32_p_33))\r\n break;\r\n ++g_128;\r\n *l_1863 = func_32_p_33 <= --g_128;\r\n }\r\n }\r\n}\r\n```\r\n\r\nIf compiled with optimizations and a compiler built from a861ed411a359b7cb2b58d642e4a7f2fdcf92057 you can see it hits an assertion failure:\r\n```\r\n$ ~/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang -c -O2 repro.c\r\nVPIRBasicBlock can only be used as pre-header or a successor of middle-block at the moment!\r\nclang: /home/dyung/src/upstream/llvm_clean_git/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:8594: void llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount): Assertion `verifyVPlanIsValid(*Plan) && \"VPlan is invalid\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0. Program arguments: /home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang -c -O2 repro.c\r\n1. parser at end of file\r\n2. Optimizer\r\n3. Running pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"repro.c\"\r\n4. Running pass \"loop-vectorize\" on function \"func_32\"\r\n #0 0x00005572973a049f llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x233549f)\r\n #1 0x000055729739dfec llvm::sys::CleanupOnSignal(unsigned long) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2332fec)\r\n #2 0x00005572972f0418 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\r\n #3 0x00007f5d2a347420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #4 0x00007f5d29e1400b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #5 0x00007f5d29df3859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #6 0x00007f5d29df3729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #7 0x00007f5d29df3729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #8 0x00007f5d29e04fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #9 0x0000557298e08438 llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x3d9d438)\r\n#10 0x0000557298e0878f llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x3d9d78f)\r\n#11 0x0000557298e0a62e llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x3d9f62e)\r\n#12 0x0000557298e0d501 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x3da2501)\r\n#13 0x0000557298e0db73 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x3da2b73)\r\n#14 0x0000557298836146 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x37cb146)\r\n#15 0x0000557296e5bd41 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x1df0d41)\r\n#16 0x0000557295f84cc6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0xf19cc6)\r\n#17 0x0000557296e5a6ad llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x1def6ad)\r\n#18 0x0000557295f85686 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0xf1a686)\r\n#19 0x0000557296e5a0e1 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x1def0e1)\r\n#20 0x0000557297636863 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#21 0x0000557297639dbd clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x25cedbd)\r\n#22 0x0000557297d17c5c clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2cacc5c)\r\n#23 0x0000557299c7aa8c clang::ParseAST(clang::Sema&, bool, bool) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x4c0fa8c)\r\n#24 0x0000557297d18078 clang::CodeGenAction::ExecuteAction() (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2cad078)\r\n#25 0x0000557297fdf129 clang::FrontendAction::Execute() (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2f74129)\r\n#26 0x0000557297f5b6fe clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2ef06fe)\r\n#27 0x00005572980cba26 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x3060a26)\r\n#28 0x0000557295f09185 cc1_main(llvm::ArrayRef, char const*, void*) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0xe9e185)\r\n#29 0x0000557295f00f0a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#30 0x0000557297d5e0cd void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\r\n#31 0x00005572972f0920 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2285920)\r\n#32 0x0000557297d5e6ef clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\r\n#33 0x0000557297d22274 clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2cb7274)\r\n#34 0x0000557297d2338d clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2cb838d)\r\n#35 0x0000557297d2b4a5 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0x2cc04a5)\r\n#36 0x0000557295f0611b clang_main(int, char**, llvm::ToolContext const&) (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0xe9b11b)\r\n#37 0x0000557295e3609b main (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0xdcb09b)\r\n#38 0x00007f5d29df5083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#39 0x0000557295f0099e _start (/home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin/clang+0xe9599e)\r\nclang: error: clang frontend command failed with exit code 134 (use -v to see invocation)\r\nclang version 20.0.0 (https://github.com/llvm/llvm-project.git a861ed411a359b7cb2b58d642e4a7f2fdcf92057)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /home/dyung/src/upstream/a861ed411a359b7cb2b58d642e4a7f2fdcf92057-linux/bin\r\nBuild config: +assertions\r\n```\r\n\r\nBuilding with a compiler built from the commit that immediately preceded this one (37e5319a12ba47c18049728804d3d1e1b10c4eb4), you can see that the compilation succeeds.\r\n```\r\n$ ~/src/upstream/37e5319a12ba47c18049728804d3d1e1b10c4eb4-linux/bin/clang -c -O2 repro.c\r\n$\r\n```", "author": "dyung", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "93298", "issue_url": "https://github.com/llvm/llvm-project/issues/93298", "bug_type": "crash", "base_commit": "f31b197d9df141effd439de8be51ce24f3e8f200", "knowledge_cutoff": "2024-05-24T12:01:44Z", "lit_test_dir": ["llvm/test/Transforms/DeadStoreElimination"], "hints": {"fix_commit": "243ffbdf8b25285d04ee4393e86094312cb7c64f", "components": ["DeadStoreElimination"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp": [[2109, 2118]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp": ["eliminateRedundantStoresOfExistingValues"]}}, "patch": "commit 243ffbdf8b25285d04ee4393e86094312cb7c64f\nAuthor: eaeltsin <109593613+eaeltsin@users.noreply.github.com>\nDate: Mon May 27 09:26:44 2024 +0200\n\n [DSE] Check write location in IsRedundantStore (#93400)\n \n Fix https://github.com/llvm/llvm-project/issues/93298.\n\ndiff --git a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\nindex ed4212d29cef..172cce2cfa38 100644\n--- a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\n+++ b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\n@@ -2109,10 +2109,12 @@ struct DSEState {\n if (auto *MemSetI = dyn_cast(UpperInst)) {\n if (auto *SI = dyn_cast(DefInst)) {\n // MemSetInst must have a write location.\n- MemoryLocation UpperLoc = *getLocForWrite(UpperInst);\n+ auto UpperLoc = getLocForWrite(UpperInst);\n+ if (!UpperLoc)\n+ return false;\n int64_t InstWriteOffset = 0;\n int64_t DepWriteOffset = 0;\n- auto OR = isOverwrite(UpperInst, DefInst, UpperLoc, *MaybeDefLoc,\n+ auto OR = isOverwrite(UpperInst, DefInst, *UpperLoc, *MaybeDefLoc,\n InstWriteOffset, DepWriteOffset);\n Value *StoredByte = isBytewiseValue(SI->getValueOperand(), DL);\n return StoredByte && StoredByte == MemSetI->getOperand(1) &&\n", "tests": [{"file": "llvm/test/Transforms/DeadStoreElimination/simple.ll", "commands": ["opt < %s -passes=dse -S", "opt < %s -aa-pipeline=basic-aa -passes=dse -S"], "tests": [{"test_name": "test48", "test_body": "target datalayout = \"E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128\"\n\ndefine i32 @test48(ptr %P, ptr noalias %Q, ptr %R) {\n store i32 1, ptr %Q, align 4\n store i32 2, ptr %P, align 4\n store i32 3, ptr %Q, align 4\n %l = load i32, ptr %R, align 4\n ret i32 %l\n}\n"}, {"test_name": "test49", "test_body": "target datalayout = \"E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128\"\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\ndefine void @test49() {\nbb:\n call void @llvm.memset.p0.i64(ptr readonly null, i8 0, i64 0, i1 false)\n store ptr null, ptr null, align 8\n ret void\n}\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\n"}]}], "issue": {"title": "[DSE] crash on applying * to std::nullopt", "body": "[reduced5.zip](https://github.com/llvm/llvm-project/files/15433764/reduced5.zip)\r\n\r\nThe command\r\n```\r\nopt -passes=dse reduced5.ll\r\n```\r\nfails occasionally. Debug version reliably complains\r\n```\r\ninclude/c++/v1/optional:805: assertion this->has_value() failed: optional operator* called on a disengaged value\r\n```\r\nwith stack trace\r\n```\r\n frame #3: 0x00005555663c6a43 opt`std::__u::optional::operator*(this= Has Value=false ) && at optional:805:5\r\n frame #4: 0x00005555663c4d58 opt`(anonymous namespace)::DSEState::eliminateRedundantStoresOfExistingValues(this=0x00007fffffff9130)::'lambda'()::operator()() const at DeadStoreElimination.cpp:2119:39\r\n frame #5: 0x00005555663bf842 opt`(anonymous namespace)::DSEState::eliminateRedundantStoresOfExistingValues(this=0x00007fffffff9960) at DeadStoreElimination.cpp:2132:12\r\n frame #6: 0x00005555663b8956 opt`(anonymous namespace)::eliminateDeadStores(F=0x000033187fe3e008, AA=0x000033187feaafc8, MSSA=0x000033187fe603f0, DT=0x000033187fef6d78, PDT=0x000033187fed18d0, TLI=0x000033187fe0cd18, LI=0x000033187fe11cc8) at DeadStoreElimination.cpp:2332:23\r\n frame #7: 0x00005555663b6b3e opt`llvm::DSEPass::run(this=0x000033187fe08ad8, F=0x000033187fe3e008, AM=0x00007fffffffc1d8) at DeadStoreElimination.cpp:2355:18\r\n```\r\n\r\nThe local fix is\r\n```\r\n==== llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp#184\r\n# action=edit type=text\r\n--- llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\t2024-04-15 20:27:56.000000000 -0700\r\n+++ llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\t2024-05-24 04:36:17.000000000 -0700\r\n@@ -2109,10 +2109,12 @@\r\n if (auto *MemSetI = dyn_cast(UpperInst)) {\r\n if (auto *SI = dyn_cast(DefInst)) {\r\n // MemSetInst must have a write location.\r\n- MemoryLocation UpperLoc = *getLocForWrite(UpperInst);\r\n+ auto UpperLoc = getLocForWrite(UpperInst);\r\n+ if (!UpperLoc)\r\n+ return false;\r\n int64_t InstWriteOffset = 0;\r\n int64_t DepWriteOffset = 0;\r\n- auto OR = isOverwrite(UpperInst, DefInst, UpperLoc, *MaybeDefLoc,\r\n+ auto OR = isOverwrite(UpperInst, DefInst, *UpperLoc, *MaybeDefLoc,\r\n InstWriteOffset, DepWriteOffset);\r\n Value *StoredByte = isBytewiseValue(SI->getValueOperand(), DL);\r\n return StoredByte && StoredByte == MemSetI->getOperand(1) &&\r\n```\r\nwhich basically reverts a piece of https://github.com/llvm/llvm-project/commit/81d69e1bda9e4b6a83f29ba1f614e43ab4700972\r\n\r\nLooks like the actual problem happened earlier?\r\n\r\nPlease take a look, this is blocking us significantly.\r\n\r\n", "author": "eaeltsin", "labels": ["crash", "llvm:transforms"], "comments": [{"author": "eaeltsin", "body": "@nikic @fhahn FYI\r\n\r\nthe reproducer was cooked by @alinas \r\n"}]}, "verified": true} {"bug_id": "117518", "issue_url": "https://github.com/llvm/llvm-project/issues/117518", "bug_type": "crash", "base_commit": "3de21477c49172081e502b47d608e729915f0914", "knowledge_cutoff": "2024-11-25T05:31:23Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "f953b5eb72df77fc301aac210eab31c6270ff771", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[10875, 10883], [14372, 14378]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["finalize"]}}, "patch": "commit f953b5eb72df77fc301aac210eab31c6270ff771\nAuthor: Alexey Bataev \nDate: Mon Nov 25 08:29:22 2024 -0800\n\n [SLP]Relax assertion about subvectors mask size\n \n SubVectorsMask might be less than CommonMask, if the vectors with larger\n number of elements are permuted or reused elements are used. Need to\n consider this when estimation/building the vector to avoid compiler\n crash\n \n Fixes #117518\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 73218f377a06..34b80c1e3608 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -10875,9 +10875,10 @@ public:\n CommonMask[Idx] = Idx;\n // Add subvectors permutation cost.\n if (!SubVectorsMask.empty()) {\n- assert(SubVectorsMask.size() == CommonMask.size() &&\n+ assert(SubVectorsMask.size() <= CommonMask.size() &&\n \"Expected same size of masks for subvectors and common mask.\");\n- SmallVector SVMask(SubVectorsMask.begin(), SubVectorsMask.end());\n+ SmallVector SVMask(CommonMask.size(), PoisonMaskElem);\n+ copy(SubVectorsMask, SVMask.begin());\n for (auto [I1, I2] : zip(SVMask, CommonMask)) {\n if (I2 != PoisonMaskElem) {\n assert(I1 == PoisonMaskElem && \"Expected unused subvectors mask\");\n@@ -14372,7 +14373,8 @@ public:\n if (SubVectorsMask.empty()) {\n Vec = CreateSubVectors(Vec, CommonMask);\n } else {\n- SmallVector SVMask(SubVectorsMask.begin(), SubVectorsMask.end());\n+ SmallVector SVMask(CommonMask.size(), PoisonMaskElem);\n+ copy(SubVectorsMask, SVMask.begin());\n for (auto [I1, I2] : zip(SVMask, CommonMask)) {\n if (I2 != PoisonMaskElem) {\n assert(I1 == PoisonMaskElem && \"Expected unused subvectors mask\");\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/non-power-of-2-subvectors-insert.ll", "commands": ["opt -S --passes=slp-vectorizer -slp-threshold=-300 -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "", "test_body": "\ndefine void @test() {\n;\n %xor108.i.i.i = xor i64 0, 1\n %conv115.i.i.i = trunc i64 %xor108.i.i.i to i16\n %add.i.i.i.i = or i16 %conv115.i.i.i, 0\n %add.i.frozen.i.i.i = freeze i16 %add.i.i.i.i\n %.cmp.not.i.i.i = icmp eq i16 %add.i.frozen.i.i.i, 0\n %cond.i1002.1.i.i.i = lshr i64 0, 0\n %conv115.1.i.i.i = trunc i64 %cond.i1002.1.i.i.i to i16\n %add.i.1.i.i.i = or i16 %conv115.1.i.i.i, 0\n %add.i.frozen.1.i.i.i = freeze i16 %add.i.1.i.i.i\n %.cmp.not.1.i.i.i = icmp eq i16 %add.i.frozen.1.i.i.i, 0\n %cond.i1002.2.i.i.i = lshr i64 %xor108.i.i.i, 0\n %conv115.2.i.i.i = trunc i64 %cond.i1002.2.i.i.i to i16\n %add.i.2.i.i.i = or i16 %conv115.2.i.i.i, 0\n %add.i.frozen.2.i.i.i = freeze i16 %add.i.2.i.i.i\n %.cmp.not.2.i.i.i = icmp eq i16 %add.i.frozen.2.i.i.i, 0\n %cond.i1002.3.i.i.i = lshr i64 0, 0\n %conv115.3.i.i.i = trunc i64 %cond.i1002.3.i.i.i to i16\n %add.i.3.i.i.i = or i16 %conv115.3.i.i.i, 0\n %add.i.frozen.3.i.i.i = freeze i16 %add.i.3.i.i.i\n %.cmp.not.3.i.i.i = icmp eq i16 %add.i.frozen.3.i.i.i, 0\n %conv115.i.i.i.1 = trunc i64 %xor108.i.i.i to i16\n %add.i.i.i.i.1 = or i16 %conv115.i.i.i.1, 0\n %add.i.frozen.i.i.i.1 = freeze i16 %add.i.i.i.i.1\n %.cmp.not.i.i.i.1 = icmp eq i16 %add.i.frozen.i.i.i.1, 0\n %cond.i1002.1.i.i.i.1 = lshr i64 0, 0\n %conv115.1.i.i.i.1 = trunc i64 %cond.i1002.1.i.i.i.1 to i16\n %add.i.1.i.i.i.1 = or i16 %conv115.1.i.i.i.1, 0\n %add.i.frozen.1.i.i.i.1 = freeze i16 %add.i.1.i.i.i.1\n %.cmp.not.1.i.i.i.1 = icmp eq i16 %add.i.frozen.1.i.i.i.1, 0\n %cond.i1002.2.i.i.i.1 = lshr i64 0, 0\n %conv115.2.i.i.i.1 = trunc i64 %cond.i1002.2.i.i.i.1 to i16\n %add.i.2.i.i.i.1 = or i16 %conv115.2.i.i.i.1, 0\n %add.i.frozen.2.i.i.i.1 = freeze i16 %add.i.2.i.i.i.1\n %.cmp.not.2.i.i.i.1 = icmp eq i16 %add.i.frozen.2.i.i.i.1, 0\n %cond.i1002.3.i.i.i.1 = lshr i64 0, 0\n %conv115.3.i.i.i.1 = trunc i64 %cond.i1002.3.i.i.i.1 to i16\n %add.i.3.i.i.i.1 = or i16 %conv115.3.i.i.i.1, 0\n %add.i.frozen.3.i.i.i.1 = freeze i16 %add.i.3.i.i.i.1\n %.cmp.not.3.i.i.i.1 = icmp eq i16 %add.i.frozen.3.i.i.i.1, 0\n %conv115.i.i.i.2 = trunc i64 %xor108.i.i.i to i16\n %add.i.i.i.i.2 = or i16 %conv115.i.i.i.2, 0\n %add.i.frozen.i.i.i.2 = freeze i16 %add.i.i.i.i.2\n %.cmp.not.i.i.i.2 = icmp eq i16 %add.i.frozen.i.i.i.2, 0\n %cond.i1002.1.i.i.i.2 = lshr i64 0, 0\n %conv115.1.i.i.i.2 = trunc i64 %cond.i1002.1.i.i.i.2 to i16\n %add.i.1.i.i.i.2 = or i16 %conv115.1.i.i.i.2, 0\n %add.i.frozen.1.i.i.i.2 = freeze i16 %add.i.1.i.i.i.2\n %.cmp.not.1.i.i.i.2 = icmp eq i16 %add.i.frozen.1.i.i.i.2, 0\n %cond.i1002.2.i.i.i.2 = lshr i64 0, 0\n %conv115.2.i.i.i.2 = trunc i64 %cond.i1002.2.i.i.i.2 to i16\n %add.i.2.i.i.i.2 = or i16 %conv115.2.i.i.i.2, 0\n %add.i.frozen.2.i.i.i.2 = freeze i16 %add.i.2.i.i.i.2\n %.cmp.not.2.i.i.i.2 = icmp eq i16 %add.i.frozen.2.i.i.i.2, 0\n %cond.i1002.3.i.i.i.2 = lshr i64 0, 0\n %conv115.3.i.i.i.2 = trunc i64 %cond.i1002.3.i.i.i.2 to i16\n %add.i.3.i.i.i.2 = or i16 %conv115.3.i.i.i.2, 0\n %add.i.frozen.3.i.i.i.2 = freeze i16 %add.i.3.i.i.i.2\n %.cmp.not.3.i.i.i.2 = icmp eq i16 %add.i.frozen.3.i.i.i.2, 0\n %conv115.i.i.i.3 = trunc i64 %xor108.i.i.i to i16\n %add.i.i.i.i.3 = or i16 %conv115.i.i.i.3, 0\n %add.i.frozen.i.i.i.3 = freeze i16 %add.i.i.i.i.3\n %.cmp.not.i.i.i.3 = icmp eq i16 %add.i.frozen.i.i.i.3, 0\n %cond.i1002.1.i.i.i.3 = lshr i64 0, 0\n %conv115.1.i.i.i.3 = trunc i64 %cond.i1002.1.i.i.i.3 to i16\n %add.i.1.i.i.i.3 = or i16 %conv115.1.i.i.i.3, 0\n %add.i.frozen.1.i.i.i.3 = freeze i16 %add.i.1.i.i.i.3\n %.cmp.not.1.i.i.i.3 = icmp eq i16 %add.i.frozen.1.i.i.i.3, 0\n %cond.i1002.2.i.i.i.3 = lshr i64 0, 0\n %conv115.2.i.i.i.3 = trunc i64 %cond.i1002.2.i.i.i.3 to i16\n %add.i.2.i.i.i.3 = or i16 %conv115.2.i.i.i.3, 0\n %add.i.frozen.2.i.i.i.3 = freeze i16 %add.i.2.i.i.i.3\n %.cmp.not.2.i.i.i.3 = icmp eq i16 %add.i.frozen.2.i.i.i.3, 0\n %cond.i1002.3.i.i.i.3 = lshr i64 0, 0\n %conv115.3.i.i.i.3 = trunc i64 %cond.i1002.3.i.i.i.3 to i16\n %add.i.3.i.i.i.3 = or i16 %conv115.3.i.i.i.3, 0\n %add.i.frozen.3.i.i.i.3 = freeze i16 %add.i.3.i.i.i.3\n %.cmp.not.3.i.i.i.3 = icmp eq i16 %add.i.frozen.3.i.i.i.3, 0\n ret void\n}"}]}], "issue": {"title": "[SLPVectorizer] Assertion `SubVectorsMask.size() == CommonMask.size() && \"Expected same size of masks for subvectors and common mask.\"' failed.", "body": "Reproducer: https://godbolt.org/z/chKcKacPK\r\n```\r\n; bin/opt -passes=slp-vectorizer test.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @test() {\r\n %xor108.i.i.i = xor i64 0, 1\r\n %conv115.i.i.i = trunc i64 %xor108.i.i.i to i16\r\n %add.i.i.i.i = or i16 %conv115.i.i.i, 0\r\n %add.i.frozen.i.i.i = freeze i16 %add.i.i.i.i\r\n %.cmp.not.i.i.i = icmp eq i16 %add.i.frozen.i.i.i, 0\r\n %cond.i1002.1.i.i.i = lshr i64 0, 0\r\n %conv115.1.i.i.i = trunc i64 %cond.i1002.1.i.i.i to i16\r\n %add.i.1.i.i.i = or i16 %conv115.1.i.i.i, 0\r\n %add.i.frozen.1.i.i.i = freeze i16 %add.i.1.i.i.i\r\n %.cmp.not.1.i.i.i = icmp eq i16 %add.i.frozen.1.i.i.i, 0\r\n %cond.i1002.2.i.i.i = lshr i64 %xor108.i.i.i, 0\r\n %conv115.2.i.i.i = trunc i64 %cond.i1002.2.i.i.i to i16\r\n %add.i.2.i.i.i = or i16 %conv115.2.i.i.i, 0\r\n %add.i.frozen.2.i.i.i = freeze i16 %add.i.2.i.i.i\r\n %.cmp.not.2.i.i.i = icmp eq i16 %add.i.frozen.2.i.i.i, 0\r\n %cond.i1002.3.i.i.i = lshr i64 0, 0\r\n %conv115.3.i.i.i = trunc i64 %cond.i1002.3.i.i.i to i16\r\n %add.i.3.i.i.i = or i16 %conv115.3.i.i.i, 0\r\n %add.i.frozen.3.i.i.i = freeze i16 %add.i.3.i.i.i\r\n %.cmp.not.3.i.i.i = icmp eq i16 %add.i.frozen.3.i.i.i, 0\r\n %conv115.i.i.i.1 = trunc i64 %xor108.i.i.i to i16\r\n %add.i.i.i.i.1 = or i16 %conv115.i.i.i.1, 0\r\n %add.i.frozen.i.i.i.1 = freeze i16 %add.i.i.i.i.1\r\n %.cmp.not.i.i.i.1 = icmp eq i16 %add.i.frozen.i.i.i.1, 0\r\n %cond.i1002.1.i.i.i.1 = lshr i64 0, 0\r\n %conv115.1.i.i.i.1 = trunc i64 %cond.i1002.1.i.i.i.1 to i16\r\n %add.i.1.i.i.i.1 = or i16 %conv115.1.i.i.i.1, 0\r\n %add.i.frozen.1.i.i.i.1 = freeze i16 %add.i.1.i.i.i.1\r\n %.cmp.not.1.i.i.i.1 = icmp eq i16 %add.i.frozen.1.i.i.i.1, 0\r\n %cond.i1002.2.i.i.i.1 = lshr i64 0, 0\r\n %conv115.2.i.i.i.1 = trunc i64 %cond.i1002.2.i.i.i.1 to i16\r\n %add.i.2.i.i.i.1 = or i16 %conv115.2.i.i.i.1, 0\r\n %add.i.frozen.2.i.i.i.1 = freeze i16 %add.i.2.i.i.i.1\r\n %.cmp.not.2.i.i.i.1 = icmp eq i16 %add.i.frozen.2.i.i.i.1, 0\r\n %cond.i1002.3.i.i.i.1 = lshr i64 0, 0\r\n %conv115.3.i.i.i.1 = trunc i64 %cond.i1002.3.i.i.i.1 to i16\r\n %add.i.3.i.i.i.1 = or i16 %conv115.3.i.i.i.1, 0\r\n %add.i.frozen.3.i.i.i.1 = freeze i16 %add.i.3.i.i.i.1\r\n %.cmp.not.3.i.i.i.1 = icmp eq i16 %add.i.frozen.3.i.i.i.1, 0\r\n %conv115.i.i.i.2 = trunc i64 %xor108.i.i.i to i16\r\n %add.i.i.i.i.2 = or i16 %conv115.i.i.i.2, 0\r\n %add.i.frozen.i.i.i.2 = freeze i16 %add.i.i.i.i.2\r\n %.cmp.not.i.i.i.2 = icmp eq i16 %add.i.frozen.i.i.i.2, 0\r\n %cond.i1002.1.i.i.i.2 = lshr i64 0, 0\r\n %conv115.1.i.i.i.2 = trunc i64 %cond.i1002.1.i.i.i.2 to i16\r\n %add.i.1.i.i.i.2 = or i16 %conv115.1.i.i.i.2, 0\r\n %add.i.frozen.1.i.i.i.2 = freeze i16 %add.i.1.i.i.i.2\r\n %.cmp.not.1.i.i.i.2 = icmp eq i16 %add.i.frozen.1.i.i.i.2, 0\r\n %cond.i1002.2.i.i.i.2 = lshr i64 0, 0\r\n %conv115.2.i.i.i.2 = trunc i64 %cond.i1002.2.i.i.i.2 to i16\r\n %add.i.2.i.i.i.2 = or i16 %conv115.2.i.i.i.2, 0\r\n %add.i.frozen.2.i.i.i.2 = freeze i16 %add.i.2.i.i.i.2\r\n %.cmp.not.2.i.i.i.2 = icmp eq i16 %add.i.frozen.2.i.i.i.2, 0\r\n %cond.i1002.3.i.i.i.2 = lshr i64 0, 0\r\n %conv115.3.i.i.i.2 = trunc i64 %cond.i1002.3.i.i.i.2 to i16\r\n %add.i.3.i.i.i.2 = or i16 %conv115.3.i.i.i.2, 0\r\n %add.i.frozen.3.i.i.i.2 = freeze i16 %add.i.3.i.i.i.2\r\n %.cmp.not.3.i.i.i.2 = icmp eq i16 %add.i.frozen.3.i.i.i.2, 0\r\n %conv115.i.i.i.3 = trunc i64 %xor108.i.i.i to i16\r\n %add.i.i.i.i.3 = or i16 %conv115.i.i.i.3, 0\r\n %add.i.frozen.i.i.i.3 = freeze i16 %add.i.i.i.i.3\r\n %.cmp.not.i.i.i.3 = icmp eq i16 %add.i.frozen.i.i.i.3, 0\r\n %cond.i1002.1.i.i.i.3 = lshr i64 0, 0\r\n %conv115.1.i.i.i.3 = trunc i64 %cond.i1002.1.i.i.i.3 to i16\r\n %add.i.1.i.i.i.3 = or i16 %conv115.1.i.i.i.3, 0\r\n %add.i.frozen.1.i.i.i.3 = freeze i16 %add.i.1.i.i.i.3\r\n %.cmp.not.1.i.i.i.3 = icmp eq i16 %add.i.frozen.1.i.i.i.3, 0\r\n %cond.i1002.2.i.i.i.3 = lshr i64 0, 0\r\n %conv115.2.i.i.i.3 = trunc i64 %cond.i1002.2.i.i.i.3 to i16\r\n %add.i.2.i.i.i.3 = or i16 %conv115.2.i.i.i.3, 0\r\n %add.i.frozen.2.i.i.i.3 = freeze i16 %add.i.2.i.i.i.3\r\n %.cmp.not.2.i.i.i.3 = icmp eq i16 %add.i.frozen.2.i.i.i.3, 0\r\n %cond.i1002.3.i.i.i.3 = lshr i64 0, 0\r\n %conv115.3.i.i.i.3 = trunc i64 %cond.i1002.3.i.i.i.3 to i16\r\n %add.i.3.i.i.i.3 = or i16 %conv115.3.i.i.i.3, 0\r\n %add.i.frozen.3.i.i.i.3 = freeze i16 %add.i.3.i.i.i.3\r\n %.cmp.not.3.i.i.i.3 = icmp eq i16 %add.i.frozen.3.i.i.i.3, 0\r\n ret void\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:10880: llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::finalize(llvm::ArrayRef, llvm::ArrayRef >, llvm::ArrayRef, unsigned int, llvm::function_ref&)>): Assertion `SubVectorsMask.size() == CommonMask.size() && \"Expected same size of masks for subvectors and common mask.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"test\"\r\n #0 0x00000000051b7378 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x51b7378)\r\n #1 0x00000000051b4d6c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007ac42d642520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007ac42d6969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007ac42d642476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007ac42d6287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007ac42d62871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007ac42d639e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x00000000041e9132 llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::finalize(llvm::ArrayRef, llvm::ArrayRef>, llvm::ArrayRef, unsigned int, llvm::function_ref&)>) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41e9132)\r\n #9 0x00000000042256f9 llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::processBuildVector, llvm::slpvectorizer::BoUpSLP, llvm::SmallPtrSetImpl>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::TargetTransformInfo&, llvm::ArrayRef&, llvm::slpvectorizer::BoUpSLP&, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42256f9)\r\n#10 0x000000000422f707 llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x422f707)\r\n#11 0x00000000042636a1 llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42636a1)\r\n#12 0x000000000426d57c llvm::SLPVectorizerPass::tryToVectorizeList(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x426d57c)\r\n#13 0x000000000427168f bool llvm::SLPVectorizerPass::vectorizeCmpInsts>(llvm::iterator_range>, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x427168f)\r\n#14 0x0000000004271fdb llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&)::'lambda3'(bool)::operator()(bool) const SLPVectorizer.cpp:0:0\r\n#15 0x0000000004275785 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4275785)\r\n#16 0x000000000427bf86 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#17 0x000000000427ca9b llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x427ca9b)\r\n#18 0x0000000002fcc5ce llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2fcc5ce)\r\n#19 0x0000000004fb6a68 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb6a68)\r\n#20 0x0000000000e3dd4e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3dd4e)\r\n#21 0x0000000004fb548e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb548e)\r\n#22 0x0000000000e3d3fe llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3d3fe)\r\n#23 0x0000000004fb4ec0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fb4ec0)\r\n#24 0x000000000090ce7a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x90ce7a)\r\n#25 0x00000000008ff519 optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ff519)\r\n#26 0x00007ac42d629d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#27 0x00007ac42d629e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#28 0x00000000008f6d6e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f6d6e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "98753", "issue_url": "https://github.com/llvm/llvm-project/issues/98753", "bug_type": "miscompilation", "base_commit": "c30ce8b9d33d1050ead549705702c1472b7a7d3f", "knowledge_cutoff": "2024-07-13T17:35:41Z", "lit_test_dir": ["llvm/test/Transforms/InstSimplify"], "hints": {"fix_commit": "de29b850f03092195bf21f3a39402adb4ed3c216", "components": ["InstructionSimplify"], "bug_location_lineno": {"llvm/lib/Analysis/InstructionSimplify.cpp": [[1975, 1987], [4300, 4305], [4347, 4352], [4467, 4472], [4606, 4612]]}, "bug_location_funcname": {"llvm/lib/Analysis/InstructionSimplify.cpp": ["llvm::simplifyWithOpReplaced", "simplifyAndOrWithICmpEq", "simplifySelectWithICmpEq"]}}, "patch": "commit de29b850f03092195bf21f3a39402adb4ed3c216\nAuthor: Nikita Popov \nDate: Tue Jul 16 11:40:04 2024 +0200\n\n [InstSimplify] Fix simplifyAndOrWithICmpEq with undef refinement (#98898)\n \n The final case in Simplify (where Res == Absorber and the predicate is\n inverted) is not generally safe when the simplification is a refinement.\n In particular, we may simplify assuming a specific value for undef, but\n then chose a different one later.\n \n However, it *is* safe to refine poison in this context, unlike in the\n equivalent select folds. This is the reason why this fold did not use\n AllowRefinement=false in the first place, and using that option would\n introduce a lot of test regressions.\n \n This patch takes the middle path of disabling undef refinements in\n particular using the getWithoutUndef() SimplifyQuery option. However,\n this option doesn't actually work in this case, because the problematic\n fold is inside constant folding, and we currently don't propagate this\n option all the way from InstSimplify over ConstantFolding to\n ConstantFold. Work around this by explicitly checking for undef operands\n in simplifyWithOpReplaced().\n \n Finally, make sure that places where AllowRefinement=false also use\n Q.getWithoutUndef(). I don't have a specific test case for this (the\n original one does not work because we don't simplify selects with\n constant condition in this mode in the first place) but this seems like\n the correct thing to do to be conservative.\n \n Fixes https://github.com/llvm/llvm-project/issues/98753.\n\ndiff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp\nindex 242c200f7ef1..3a7ae577bb06 100644\n--- a/llvm/lib/Analysis/InstructionSimplify.cpp\n+++ b/llvm/lib/Analysis/InstructionSimplify.cpp\n@@ -1975,13 +1975,16 @@ static Value *simplifyAndOrWithICmpEq(unsigned Opcode, Value *Op0, Value *Op1,\n return nullptr;\n };\n \n- if (Value *Res =\n- simplifyWithOpReplaced(Op1, A, B, Q, /* AllowRefinement */ true,\n- /* DropFlags */ nullptr, MaxRecurse))\n+ // In the final case (Res == Absorber with inverted predicate), it is safe to\n+ // refine poison during simplification, but not undef. For simplicity always\n+ // disable undef-based folds here.\n+ if (Value *Res = simplifyWithOpReplaced(Op1, A, B, Q.getWithoutUndef(),\n+ /* AllowRefinement */ true,\n+ /* DropFlags */ nullptr, MaxRecurse))\n return Simplify(Res);\n- if (Value *Res =\n- simplifyWithOpReplaced(Op1, B, A, Q, /* AllowRefinement */ true,\n- /* DropFlags */ nullptr, MaxRecurse))\n+ if (Value *Res = simplifyWithOpReplaced(Op1, B, A, Q.getWithoutUndef(),\n+ /* AllowRefinement */ true,\n+ /* DropFlags */ nullptr, MaxRecurse))\n return Simplify(Res);\n \n return nullptr;\n@@ -4300,6 +4303,9 @@ static Value *simplifyWithOpReplaced(Value *V, Value *Op, Value *RepOp,\n bool AllowRefinement,\n SmallVectorImpl *DropFlags,\n unsigned MaxRecurse) {\n+ assert((AllowRefinement || !Q.CanUseUndef) &&\n+ \"If AllowRefinement=false then CanUseUndef=false\");\n+\n // Trivial replacement.\n if (V == Op)\n return RepOp;\n@@ -4347,6 +4353,11 @@ static Value *simplifyWithOpReplaced(Value *V, Value *Op, Value *RepOp,\n } else {\n NewOps.push_back(InstOp);\n }\n+\n+ // Bail out if any operand is undef and SimplifyQuery disables undef\n+ // simplification. Constant folding currently doesn't respect this option.\n+ if (isa(NewOps.back()) && !Q.CanUseUndef)\n+ return nullptr;\n }\n \n if (!AnyReplaced)\n@@ -4467,6 +4478,11 @@ Value *llvm::simplifyWithOpReplaced(Value *V, Value *Op, Value *RepOp,\n const SimplifyQuery &Q,\n bool AllowRefinement,\n SmallVectorImpl *DropFlags) {\n+ // If refinement is disabled, also disable undef simplifications (which are\n+ // always refinements) in SimplifyQuery.\n+ if (!AllowRefinement)\n+ return ::simplifyWithOpReplaced(V, Op, RepOp, Q.getWithoutUndef(),\n+ AllowRefinement, DropFlags, RecursionLimit);\n return ::simplifyWithOpReplaced(V, Op, RepOp, Q, AllowRefinement, DropFlags,\n RecursionLimit);\n }\n@@ -4606,7 +4622,7 @@ static Value *simplifySelectWithICmpEq(Value *CmpLHS, Value *CmpRHS,\n Value *TrueVal, Value *FalseVal,\n const SimplifyQuery &Q,\n unsigned MaxRecurse) {\n- if (simplifyWithOpReplaced(FalseVal, CmpLHS, CmpRHS, Q,\n+ if (simplifyWithOpReplaced(FalseVal, CmpLHS, CmpRHS, Q.getWithoutUndef(),\n /* AllowRefinement */ false,\n /* DropFlags */ nullptr, MaxRecurse) == TrueVal)\n return FalseVal;\n", "tests": [{"file": "llvm/test/Transforms/InstSimplify/and-or-implied-cond.ll", "commands": ["opt -S -passes=instsimplify < %s"], "tests": [{"test_name": "and_is_constant", "test_body": "define i1 @and_is_constant(ptr %arg, ptr %arg2) {\n %icmp = icmp eq ptr %arg, %arg2\n %call = call i1 @llvm.is.constant.i1(i1 %icmp)\n %and = and i1 %call, %icmp\n ret i1 %and\n}\n\n; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none)\ndeclare i1 @llvm.is.constant.i1(i1) #0\n\nattributes #0 = { convergent nocallback nofree nosync nounwind willreturn memory(none) }\n"}, {"test_name": "pr98753", "test_body": "define i1 @pr98753(i32 noundef %x, i32 %y) {\n %cmp1 = icmp ne i32 %x, 0\n %sel = select i1 %cmp1, i32 %y, i32 undef\n %cmp2 = icmp sgt i32 %sel, 0\n %and = and i1 %cmp1, %cmp2\n ret i1 %and\n}\n"}]}], "issue": {"title": "Miscompilation with libstdc++'s std::optional and -O1", "body": "https://godbolt.org/z/KjEreWf57\r\n\r\nWith Clang 18.1 and libstdc++, we get different behavior when compared to Clang 17 in the code below.\r\nEDIT: requires at least -O1.\r\n\r\n```cpp\r\n#include \r\n#include \r\n\r\n// defined in a separate compilation unit\r\nint takeIntRefAndReturn0(int&);\r\n\r\nstd::optional shouldReturnEmptyOptional() {\r\n int v = 5;\r\n if (takeIntRefAndReturn0(v))\r\n return v;\r\n return std::nullopt;\r\n}\r\n\r\nint main() {\r\n auto opt = shouldReturnEmptyOptional();\r\n if (opt && *opt > 0) {\r\n std::cout << \"SHOULD NOT BE PRINTED: *opt = \" << *opt << std::endl;\r\n return 1;\r\n }\r\n std::cout << \"SHOULD BE PRINTED\" << std::endl;\r\n}\r\n```\r\n\r\nWith Clang 17, we get `SHOULD BE PRINTED`, while with with Clang 18.1 we get `SHOULD NOT BE PRINTED: *opt = 5`.\r\n\r\nWith git-bisect, I found that this is caused by https://github.com/llvm/llvm-project/commit/060de415af335fdd82910f409e2be3b8457eaa43.\r\n\r\nAn isomorphic example to reproduce this (https://godbolt.org/z/9PsjY17sT):\r\n```cpp\r\nint takeIntRefReturn0(int &);\r\nvoid assertNotReached(int);\r\n\r\nstatic bool logicalAnd(bool a, bool b) { return a && b; }\r\n\r\nint main() {\r\n int v4;\r\n bool v3;\r\n {\r\n int v1 = 5;\r\n int v2 = takeIntRefReturn0(v1);\r\n v3 = v2 != 0;\r\n if (v3)\r\n v4 = v1;\r\n }\r\n // Move to a function so that && is not short cutted.\r\n // v4 will be undefined if v2 == 0, but v3 is false, so the branch shouldn't be entered.\r\n if (logicalAnd(v3, v4 > 0))\r\n assertNotReached(v4);\r\n\r\n return 0;\r\n}\r\n```\r\nNote in the generated LLVM IR that\r\n```llvm\r\n%6 = icmp sgt i32 %5, 0\r\n%7 = and i1 %3, %6\r\nbr i1 %7, label %8, label %9\r\n``` \r\nwas reduced to only\r\n```llvm\r\n%6 = icmp sgt i32 %5, 0\r\nbr i1 %6, label %7, label %8\r\n```", "author": "cadubentzen", "labels": ["regression", "miscompilation", "llvm:analysis"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/p4ZGxe\r\n```\r\n; bin/opt -passes=instcombine test.ll -S\r\ndefine i1 @src(i32 noundef %x, i32 %y) {\r\n %3 = icmp ne i32 %x, 0\r\n %5 = select i1 %3, i32 %y, i32 undef\r\n %6 = icmp sgt i32 %5, 0\r\n %7 = and i1 %3, %6\r\n br i1 %7, label %if.then, label %if.else\r\n\r\nif.then:\r\n call void @use(i32 noundef %5)\r\n ret i1 true\r\n\r\nif.else:\r\n ret i1 false\r\n}\r\n\r\ndefine i1 @tgt(i32 noundef %x, i32 %y) {\r\n %.not = icmp eq i32 %x, 0\r\n %1 = select i1 %.not, i32 undef, i32 %y\r\n %2 = icmp sgt i32 %1, 0\r\n br i1 %2, label %if.then, label %if.else\r\n\r\nif.then: ; preds = %0\r\n call void @use(i32 noundef %1)\r\n ret i1 true\r\n\r\nif.else: ; preds = %0\r\n ret i1 false\r\n}\r\n\r\ndeclare void @use(i32 %x)\r\n```\r\ncc @nikic"}, {"author": "nikic", "body": "A bit simpler using just instsimplify: https://alive2.llvm.org/ce/z/tnkf6A"}, {"author": "nikic", "body": "I expect the fix here is to perform replacements with `Q.getWithoutUndef()`. But probably not just here, but also in other places using simplifyWithOpReplaced(). (Possibly making it always use that mode.)"}, {"author": "cadubentzen", "body": "Thanks a lot for fixing this so quickly! I wonder if this should be included in 18.1.9?"}, {"author": "nikic", "body": "The LLVM 18 branch is already closed, there will be no further releases."}, {"author": "thesamesam", "body": "I'll add the milestone anyway as it's easier to track for downstream backports then."}]}, "verified": true} {"bug_id": "84709", "issue_url": "https://github.com/llvm/llvm-project/issues/84709", "bug_type": "crash", "base_commit": "4d0f79e346ceb0ddb25a94053c612a5b34a72100", "knowledge_cutoff": "2024-03-11T03:25:11Z", "lit_test_dir": ["llvm/test/Transforms/LoopStrengthReduce"], "hints": {"fix_commit": "beba307c5bc206168bdea3b893e02ea31579fe62", "components": ["LoopStrengthReduce"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp": [[6971, 6976], [6989, 6994]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp": ["ReduceLoopStrength"]}}, "patch": "commit beba307c5bc206168bdea3b893e02ea31579fe62\nAuthor: Nikita Popov \nDate: Tue Mar 12 16:23:25 2024 +0100\n\n [LSR] Clear SCEVExpander before deleting phi nodes\n \n Fixes https://github.com/llvm/llvm-project/issues/84709.\n\ndiff --git a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp\nindex 8b078ddc4e74..c4e1a0db8b32 100644\n--- a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp\n+++ b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp\n@@ -6971,6 +6971,7 @@ static bool ReduceLoopStrength(Loop *L, IVUsers &IU, ScalarEvolution &SE,\n Rewriter.setDebugType(DEBUG_TYPE);\n #endif\n unsigned numFolded = Rewriter.replaceCongruentIVs(L, &DT, DeadInsts, &TTI);\n+ Rewriter.clear();\n if (numFolded) {\n Changed = true;\n RecursivelyDeleteTriviallyDeadInstructionsPermissive(DeadInsts, &TLI,\n@@ -6989,6 +6990,7 @@ static bool ReduceLoopStrength(Loop *L, IVUsers &IU, ScalarEvolution &SE,\n SCEVExpander Rewriter(SE, DL, \"lsr\", true);\n int Rewrites = rewriteLoopExitValues(L, &LI, &TLI, &SE, &TTI, Rewriter, &DT,\n UnusedIndVarInLoop, DeadInsts);\n+ Rewriter.clear();\n if (Rewrites) {\n Changed = true;\n RecursivelyDeleteTriviallyDeadInstructionsPermissive(DeadInsts, &TLI,\n", "tests": [{"file": "llvm/test/Transforms/LoopStrengthReduce/pr84709.ll", "commands": ["opt -S -passes=loop-reduce < %s"], "tests": [{"test_name": "test", "test_body": "define i64 @test() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb5, %bb\n %phi = phi i8 [ %zext6, %bb5 ], [ 0, %bb ]\n br label %bb2\n\nbb2: ; preds = %bb2, %bb1\n %phi3 = phi i8 [ %add, %bb2 ], [ %phi, %bb1 ]\n %phi4 = phi i32 [ 0, %bb2 ], [ 1, %bb1 ]\n %add = add i8 %phi3, 1\n br i1 true, label %bb5, label %bb2\n\nbb5: ; preds = %bb2\n %zext = zext i8 %add to i32\n %icmp = icmp sge i32 %phi4, 0\n %zext6 = zext i1 %icmp to i8\n br label %bb1\n}\n"}]}], "issue": {"title": "[LSR] crashed on \"asserting value handle still pointed to this value!\"", "body": "https://godbolt.org/z/MWWcGdxT5\r\nLSR crashed when deleting dead PHIs ", "author": "coderchenlin", "labels": ["llvm:crash", "llvm:transforms"], "comments": [{"author": "coderchenlin", "body": "I will give a resolution to fix it later."}, {"author": "coderchenlin", "body": " #84777 For this case, I think we should not set the value with nullptr when the value's uses are not empty. "}, {"author": "nikic", "body": "Reduced:\r\n```llvm\r\ndefine i64 @test() {\r\nbb:\r\n br label %bb1\r\n\r\nbb1:\r\n %phi = phi i8 [ %zext6, %bb5 ], [ 0, %bb ]\r\n br label %bb2\r\n\r\nbb2:\r\n %phi3 = phi i8 [ %add, %bb2 ], [ %phi, %bb1 ]\r\n %phi4 = phi i32 [ 0, %bb2 ], [ 1, %bb1 ]\r\n %add = add i8 %phi3, 1\r\n br i1 true, label %bb5, label %bb2\r\n\r\nbb5:\r\n %zext = zext i8 %add to i32\r\n %icmp = icmp sge i32 %phi4, 0\r\n %zext6 = zext i1 %icmp to i8\r\n br label %bb1\r\n}\r\n```"}]}, "verified": true} {"bug_id": "91957", "issue_url": "https://github.com/llvm/llvm-project/issues/91957", "bug_type": "miscompilation", "base_commit": "c2a9a974ca85e4ac4509e368d4b9acae7e67bf71", "knowledge_cutoff": "2024-05-13T13:05:21Z", "lit_test_dir": ["llvm/test/Transforms/LICM"], "hints": {"fix_commit": "70091dc943ade280d75cea1e5ea5e93d9a8f934a", "components": ["LICM"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/LICM.cpp": [[2751, 2757], [2759, 2766]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/LICM.cpp": ["hoistMulAddAssociation"]}}, "patch": "commit 70091dc943ade280d75cea1e5ea5e93d9a8f934a\nAuthor: Antonio Frighetto \nDate: Sat May 18 16:52:17 2024 +0200\n\n [LICM] Invalidate cached SCEV results in `hoistMulAddAssociation`\n \n While reassociating expressions, LICM is required to invalidate SCEV\n results, as otherwise subsequent passes in the pipeline that leverage\n LICM foldings (e.g. IndVars), may reason on invalid expressions; thus\n miscompiling. This is achieved by rewriting the reassociable\n instruction from scratch.\n \n Fixes: https://github.com/llvm/llvm-project/issues/91957.\n\ndiff --git a/llvm/lib/Transforms/Scalar/LICM.cpp b/llvm/lib/Transforms/Scalar/LICM.cpp\nindex 6aa4188d1cc4..5eccf7b4adb6 100644\n--- a/llvm/lib/Transforms/Scalar/LICM.cpp\n+++ b/llvm/lib/Transforms/Scalar/LICM.cpp\n@@ -2751,7 +2751,7 @@ static bool hoistMulAddAssociation(Instruction &I, Loop &L,\n IRBuilder<> Builder(Preheader->getTerminator());\n for (auto *U : Changes) {\n assert(L.isLoopInvariant(U->get()));\n- Instruction *Ins = cast(U->getUser());\n+ auto *Ins = cast(U->getUser());\n Value *Mul;\n if (I.getType()->isIntOrIntVectorTy()) {\n Mul = Builder.CreateMul(U->get(), Factor, \"factor.op.mul\");\n@@ -2759,8 +2759,20 @@ static bool hoistMulAddAssociation(Instruction &I, Loop &L,\n Ins->dropPoisonGeneratingFlags();\n } else\n Mul = Builder.CreateFMulFMF(U->get(), Factor, Ins, \"factor.op.fmul\");\n- U->set(Mul);\n+\n+ // Rewrite the reassociable instruction.\n+ unsigned OpIdx = U->getOperandNo();\n+ auto *LHS = OpIdx == 0 ? Mul : Ins->getOperand(0);\n+ auto *RHS = OpIdx == 1 ? Mul : Ins->getOperand(1);\n+ auto *NewBO = BinaryOperator::Create(Ins->getOpcode(), LHS, RHS,\n+ Ins->getName() + \".reass\", Ins);\n+ NewBO->copyIRFlags(Ins);\n+ if (VariantOp == Ins)\n+ VariantOp = NewBO;\n+ Ins->replaceAllUsesWith(NewBO);\n+ eraseInstruction(*Ins, SafetyInfo, MSSAU);\n }\n+\n I.replaceAllUsesWith(VariantOp);\n eraseInstruction(I, SafetyInfo, MSSAU);\n return true;\n", "tests": [{"file": "llvm/test/Transforms/LICM/update-scev-after-hoist.ll", "commands": ["opt -S -passes='loop-unroll,loop-mssa(licm,indvars)' -unroll-count=4 < %s 2>&1"], "tests": [{"test_name": "main", "test_body": "define i16 @main() {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %mul = phi i16 [ 1, %entry ], [ %mul.n, %loop ]\n %div = phi i16 [ 32767, %entry ], [ %div.n, %loop ]\n %mul.n = mul i16 %mul, 2\n %div.n = sdiv i16 %div, 2\n %cmp = icmp sgt i16 %div, 0\n br i1 %cmp, label %loop, label %end\n\nend: ; preds = %loop\n ret i16 %mul\n}\n", "additional_args": "-src-unroll=16 -tgt-unroll=16"}]}], "issue": {"title": "Miscompile with opt -passes=\"loop-unroll,loop-mssa(licm,indvars)\" -unroll-count=4 ", "body": "llvm commit: e76b257483e6c\r\nReproduce with:\r\n```opt -passes=\"loop-unroll,loop-mssa(licm,indvars)\" -unroll-count=4 bbi-95405.ll -S -o -```\r\n\r\nThe input function returns 32768, but after running the passes as above we get\r\n```\r\ndefine i16 @foo() {\r\nentry:\r\n br label %loop\r\n\r\nloop: ; preds = %loop, %entry\r\n br i1 false, label %loop, label %end, !llvm.loop !0\r\n\r\nend: ; preds = %loop\r\n ret i16 8192\r\n}\r\n\r\n!0 = distinct !{!0, !1}\r\n!1 = !{!\"llvm.loop.unroll.disable\"}\r\n```\r\nSo now the function returns 8192 instead.\r\n\r\nIf I extract the IR after loop-unroll or licm and run the rest of the passes instead I get the correct result.\r\n\r\n[bbi-95405.ll.gz](https://github.com/llvm/llvm-project/files/15294896/bbi-95405.ll.gz)\r\n", "author": "mikaelholmen", "labels": ["miscompilation", "llvm:transforms"], "comments": [{"author": "mikaelholmen", "body": "This starts happening with https://github.com/llvm/llvm-project/pull/67736 , commit 2dd52046816\r\n```\r\nRecommit \"[LICM] Support integer mul/add in hoistFPAssociation. (#67736)\"\r\n\r\nWith a fix for build bot failure. I was accessing the type of a deleted\r\nInstruction.\r\n\r\nOriginal message:\r\n\r\nThe reassociation this is trying to repair can happen for integer types\r\ntoo.\r\n\r\nThis patch adds support for integer mul/add to hoistFPAssociation. The\r\nfunction has been renamed to hoistMulAddAssociation. I've used separate\r\nstatistics and limits for integer to allow tuning flexibility.\r\n```\r\n"}, {"author": "nikic", "body": "Without looking into this in detail, it's possible that the way the LICM code modifies instructions in place results incorrect cached SCEV results."}, {"author": "mikaelholmen", "body": "> Without looking into this in detail, it's possible that the way the LICM code modifies instructions in place results incorrect cached SCEV results.\r\n\r\nI added some printouts and see that when IndVarSimplify run rewriteLoopExitValues we do\r\n```\r\nconst SCEV *ExitValue = SE->getSCEVAtScope(Inst, L->getParentLoop());\r\n```\r\nfor Inst being\r\n```\r\n%mul.n = mul i16 %mul, 8\r\n```\r\nand then we get the SCEV\r\n```\r\n(2 * %mul)\r\n```\r\nwhich looks wrong.\r\n\r\nIf I split up the pipeline in several opt runs we instead get the SCEV\r\n```\r\n(8 * %mul)\r\n```\r\nwhich looks more reasonable (and then we also get the correct result).\r\nLICM changed\r\n```\r\n%mul.n = mul i16 %mul, 2\r\n```\r\nto\r\n```\r\n%mul.n = mul i16 %mul, 8\r\n```\r\nso yeah, it should perhaps do something (more) about SCEV."}, {"author": "mikaelholmen", "body": "@antoniofrighetto : nice that you work on this!\r\n\r\nBut oh it's annoying that I don't get any emails from github about that there was activity here with a linked PR with a fix etc even if I wrote this issue. :( I do get emails when there are comments but not when the issue was assigned or the PR was created. :("}, {"author": "antoniofrighetto", "body": "@mikaelholmen This is a bit inconvenient indeed, GH doesn’t seem to send notifications wrt that :( Closed this, thanks for reporting!"}, {"author": "mikaelholmen", "body": "> @mikaelholmen This is a bit inconvenient indeed, GH doesn’t seem to send notifications wrt that :( Closed this, thanks for reporting!\r\n\r\nThank you!"}]}, "verified": true} {"bug_id": "117308", "issue_url": "https://github.com/llvm/llvm-project/issues/117308", "bug_type": "crash", "base_commit": "6655c53ff02f0ffca0fcaa3261e14c2a6417d3da", "knowledge_cutoff": "2024-11-22T09:51:05Z", "lit_test_dir": ["llvm/test/Transforms/FunctionAttrs"], "hints": {"fix_commit": "5a90168fa34e15b544d7749ca4d247a16d360119", "components": ["ValueTracking"], "bug_location_lineno": {"llvm/lib/Analysis/ValueTracking.cpp": [[6886, 6892]]}, "bug_location_funcname": {"llvm/lib/Analysis/ValueTracking.cpp": ["llvm::getUnderlyingObjectAggressive"]}}, "patch": "commit 5a90168fa34e15b544d7749ca4d247a16d360119\nAuthor: Heejin Ahn \nDate: Wed Jan 15 11:53:51 2025 -0800\n\n [ValueTracking] Provide getUnderlyingObjectAggressive fallback (#123019)\n \n This callsite assumes `getUnderlyingObjectAggressive` returns a non-null\n pointer:\n \n https://github.com/llvm/llvm-project/blob/273a94b3d5a78cd9122c7b3bbb5d5a87147735d2/llvm/lib/Transforms/IPO/FunctionAttrs.cpp#L124\n \n But it can return null when there are cycles in the value chain so there\n is no more `Worklist` item anymore to explore, in which case it just\n returns `Object` at the end of the function without ever setting it:\n https://github.com/llvm/llvm-project/blob/9b5857a68381652dbea2a0c9efa734b6c4cf38c9/llvm/lib/Analysis/ValueTracking.cpp#L6866-L6867\n https://github.com/llvm/llvm-project/blob/9b5857a68381652dbea2a0c9efa734b6c4cf38c9/llvm/lib/Analysis/ValueTracking.cpp#L6889\n \n `getUnderlyingObject` does not seem to return null either judging by\n looking at its code and its callsites, so I think it is not likely to be\n the author's intention that `getUnderlyingObjectAggressive` returns\n null.\n \n So this checks whether `Object` is null at the end, and if so, falls\n back to the original first value.\n \n ---\n \n The test case here was reduced by bugpoint and further reduced manually,\n but I find it hard to reduce it further.\n \n To trigger this bug, the memory operation should not be reachable from\n the entry BB, because the `phi`s should form a cycle without introducing\n another value from the entry. I tried a minimal `phi` cycle with three\n BBs (entry BB + two BBs in a cycle), but it was skipped here:\n https://github.com/llvm/llvm-project/blob/273a94b3d5a78cd9122c7b3bbb5d5a87147735d2/llvm/lib/Transforms/IPO/FunctionAttrs.cpp#L121-L122\n To get the result that's not `ModRefInfo::NoModRef`, the length of `phi`\n chain needed to be greater than the `MaxLookup` value set in this\n function:\n \n https://github.com/llvm/llvm-project/blob/02403f4e450b86d93197dd34045ff40a34b21494/llvm/lib/Analysis/BasicAliasAnalysis.cpp#L744\n \n But just lengthening the `phi` chain to 8 didn't trigger the same error\n in `getUnderlyingObjectAggressive` because `getUnderlyingObject` here\n passes through a single-chain `phi`s so not all `phi`s end up in\n `Visited`:\n \n https://github.com/llvm/llvm-project/blob/9b5857a68381652dbea2a0c9efa734b6c4cf38c9/llvm/lib/Analysis/ValueTracking.cpp#L6863\n \n So I just submit here the smallest test case I managed to create.\n \n ---\n \n Fixes #117308 and fixes #122166.\n\ndiff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp\nindex 1bab9b32525c..1853b1c67159 100644\n--- a/llvm/lib/Analysis/ValueTracking.cpp\n+++ b/llvm/lib/Analysis/ValueTracking.cpp\n@@ -6886,7 +6886,7 @@ const Value *llvm::getUnderlyingObjectAggressive(const Value *V) {\n return FirstObject;\n } while (!Worklist.empty());\n \n- return Object;\n+ return Object ? Object : FirstObject;\n }\n \n /// This is the function that does the work of looking through basic\n", "tests": [{"file": "llvm/test/Transforms/FunctionAttrs/phi_cycle.ll", "commands": ["opt -passes=function-attrs -S < %s"], "tests": [{"test_name": "", "test_body": "\n; Regression test for a null-returning bug of getUnderlyingObjectAggressive().\n; This should not crash.\ndefine void @phi_cycle() {\nbb:\n unreachable\n\nbb1: ; preds = %bb17\n br label %bb2\n\nbb2: ; preds = %bb5, %bb1\n %phi = phi ptr [ %phi6, %bb1 ], [ %phi6, %bb5 ]\n br i1 poison, label %bb4, label %bb3\n\nbb3: ; preds = %bb2\n %getelementptr = getelementptr inbounds i8, ptr %phi, i32 poison\n br label %bb5\n\nbb4: ; preds = %bb2\n br label %bb7\n\nbb5: ; preds = %bb15, %bb3\n %phi6 = phi ptr [ %getelementptr, %bb3 ], [ %phi16, %bb15 ]\n br i1 poison, label %bb17, label %bb2\n\nbb7: ; preds = %bb15, %bb4\n %phi8 = phi ptr [ %phi, %bb4 ], [ %phi16, %bb15 ]\n br i1 poison, label %bb11, label %bb9\n\nbb9: ; preds = %bb7\n %getelementptr10 = getelementptr inbounds i8, ptr %phi8, i32 1\n store i8 poison, ptr %phi8, align 1\n br label %bb15\n\nbb11: ; preds = %bb7\n br i1 poison, label %bb13, label %bb12\n\nbb12: ; preds = %bb11\n br label %bb13\n\nbb13: ; preds = %bb12, %bb11\n %getelementptr14 = getelementptr inbounds i8, ptr %phi8, i32 poison\n br label %bb15\n\nbb15: ; preds = %bb13, %bb9\n %phi16 = phi ptr [ %getelementptr14, %bb13 ], [ %getelementptr10, %bb9 ]\n br i1 poison, label %bb5, label %bb7\n\nbb17: ; preds = %bb5\n br label %bb1\n}"}]}], "issue": {"title": "Crash on wasm-ld (failed Received SIGSEGV (-11))", "body": "I came across this error when trying to build wasm artifact of [this repository](https://github.com/daquexian/onnx-simplifier) ( branch `v0.4.36` ) using ninja.\r\n\r\nThe `emsdk` version that I used was 3.1.72.\r\n\r\n```\r\n: && /home/yogisalomo/emsdk/upstream/emscripten/em++ -O3 -DNDEBUG -s ALLOW_MEMORY_GROWTH=1 -s EXIT_RUNTIME=1 -s FORCE_FILESYSTEM=1 -s MODULARIZE=1 -s 'EXPORT_NAME=\"create_onnxsim\"' -s 'EXPORTED_RUNTIME_METHODS=[FS,ccall,cwrap,callMain]' -s EXPORTED_FUNCTIONS=[_main] CMakeFiles/onnxsim_bin.dir/onnxsim/bin/onnxsim_bin.cpp.o CMakeFiles/onnxsim_bin.dir/onnxsim/bin/onnxsim_option.cpp.o -o onnxsim.js libonnxsim.a third_party/onnx-optimizer/libonnx_optimizer.a libonnxruntime_webassembly.a && :\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /home/yogisalomo/emsdk/upstream/bin/wasm-ld -o onnxsim.wasm -L/home/yogisalomo/emsdk/upstream/emscripten/cache/sysroot/lib/wasm32-emscripten CMakeFiles/onnxsim_bin.dir/onnxsim/bin/onnxsim_bin.cpp.o CMakeFiles/onnxsim_bin.dir/onnxsim/bin/onnxsim_option.cpp.o libonnxsim.a third_party/onnx-optimizer/libonnx_optimizer.a libonnxruntime_webassembly.a -lGL-getprocaddr -lal -lhtml5 -lstubs -lc -ldlmalloc -lcompiler_rt -lc++-noexcept -lc++abi-noexcept -lsockets -mllvm -combiner-global-alias-analysis=false -mllvm -enable-emscripten-sjlj -mllvm -disable-lsr /tmp/tmpuew4mr6alibemscripten_js_symbols.so --strip-debug --export=_emscripten_stack_alloc --export=__get_temp_ret --export=__set_temp_ret --export=__funcs_on_exit --export=__wasm_call_ctors --export=emscripten_stack_get_current --export=_emscripten_stack_restore --export-if-defined=__start_em_asm --export-if-defined=__stop_em_asm --export-if-defined=__start_em_lib_deps --export-if-defined=__stop_em_lib_deps --export-if-defined=__start_em_js --export-if-defined=__stop_em_js --export-if-defined=main --export-if-defined=__main_argc_argv --export-if-defined=fflush --export-table -z stack-size=65536 --max-memory=2147483648 --initial-heap=16777216 --no-entry --table-base=1 --global-base=1024\r\n1. Running pass \"cgscc(function-attrs)\" on module \"ld-temp.o\"\r\n #0 0x00005a6692b63088 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x1083088)\r\n #1 0x00005a6692b603ee llvm::sys::RunSignalHandlers() (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x10803ee)\r\n #2 0x00005a6692b63c7f SignalHandler(int) Signals.cpp:0:0\r\n #3 0x000079f165242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00005a66941e9d3c checkFunctionMemoryAccess(llvm::Function&, bool, llvm::AAResults&, llvm::SmallSetVector const&) (.llvm.13958950707188126733) FunctionAttrs.cpp:0:0\r\n #5 0x00005a66941eb09a llvm::SmallSet> deriveAttrsInPostOrder&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&)::$_0&>(llvm::ArrayRef, llvm::PostOrderFunctionAttrsPass::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&)::$_0&, bool) FunctionAttrs.cpp:0:0\r\n #6 0x00005a66941ea819 llvm::PostOrderFunctionAttrsPass::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x270a819)\r\n #7 0x00005a6693a8370d llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) PassBuilder.cpp:0:0\r\n #8 0x00005a6694de42f3 llvm::ModuleToPostOrderCGSCCPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x33042f3)\r\n #9 0x00005a6693a605fd llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) PassBuilder.cpp:0:0\r\n#10 0x00005a6695695c9a llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x3bb5c9a)\r\n#11 0x00005a66939ba386 llvm::lto::opt(llvm::lto::Config const&, llvm::TargetMachine*, unsigned int, llvm::Module&, bool, llvm::ModuleSummaryIndex*, llvm::ModuleSummaryIndex const*, std::__2::vector> const&) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x1eda386)\r\n#12 0x00005a66939bc7e3 llvm::lto::backend(llvm::lto::Config const&, std::__2::function>> (unsigned int, llvm::Twine const&)>, unsigned int, llvm::Module&, llvm::ModuleSummaryIndex&) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x1edc7e3)\r\n#13 0x00005a66939a602f llvm::lto::LTO::runRegularLTO(std::__2::function>> (unsigned int, llvm::Twine const&)>) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x1ec602f)\r\n#14 0x00005a66939a52ad llvm::lto::LTO::run(std::__2::function>> (unsigned int, llvm::Twine const&)>, llvm::FileCache) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x1ec52ad)\r\n#15 0x00005a6693014dd3 lld::wasm::BitcodeCompiler::compile() (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x1534dd3)\r\n#16 0x00005a66930184b4 lld::wasm::SymbolTable::compileBitcodeFiles() (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x15384b4)\r\n#17 0x00005a6692ffe90c lld::wasm::(anonymous namespace)::LinkerDriver::linkerMain(llvm::ArrayRef) Driver.cpp:0:0\r\n#18 0x00005a6692ff88f8 lld::wasm::link(llvm::ArrayRef, llvm::raw_ostream&, llvm::raw_ostream&, bool, bool) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x15188f8)\r\n#19 0x00005a6692b660f0 lld::unsafeLldMain(llvm::ArrayRef, llvm::raw_ostream&, llvm::raw_ostream&, llvm::ArrayRef, bool) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0x10860f0)\r\n#20 0x00005a6692aa64d1 lld_main(int, char**, llvm::ToolContext const&) (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0xfc64d1)\r\n#21 0x00005a6692aa6ec4 main (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0xfc6ec4)\r\n#22 0x000079f165229d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#23 0x000079f165229e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#24 0x000079f165229e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#25 0x00005a6692a1336a _start (/home/yogisalomo/emsdk/upstream/bin/wasm-ld+0xf3336a)\r\nem++: error: '/home/yogisalomo/emsdk/upstream/bin/wasm-ld -o onnxsim.wasm -L/home/yogisalomo/emsdk/upstream/emscripten/cache/sysroot/lib/wasm32-emscripten CMakeFiles/onnxsim_bin.dir/onnxsim/bin/onnxsim_bin.cpp.o CMakeFiles/onnxsim_bin.dir/onnxsim/bin/onnxsim_option.cpp.o libonnxsim.a third_party/onnx-optimizer/libonnx_optimizer.a libonnxruntime_webassembly.a -lGL-getprocaddr -lal -lhtml5 -lstubs -lc -ldlmalloc -lcompiler_rt -lc++-noexcept -lc++abi-noexcept -lsockets -mllvm -combiner-global-alias-analysis=false -mllvm -enable-emscripten-sjlj -mllvm -disable-lsr /tmp/tmpuew4mr6alibemscripten_js_symbols.so --strip-debug --export=_emscripten_stack_alloc --export=__get_temp_ret --export=__set_temp_ret --export=__funcs_on_exit --export=__wasm_call_ctors --export=emscripten_stack_get_current --export=_emscripten_stack_restore --export-if-defined=__start_em_asm --export-if-defined=__stop_em_asm --export-if-defined=__start_em_lib_deps --export-if-defined=__stop_em_lib_deps --export-if-defined=__start_em_js --export-if-defined=__stop_em_js --export-if-defined=main --export-if-defined=__main_argc_argv --export-if-defined=fflush --export-table -z stack-size=65536 --max-memory=2147483648 --initial-heap=16777216 --no-entry --table-base=1 --global-base=1024' failed (received SIGSEGV (-11))\r\nninja: build stopped: subcommand failed.\r\n```\r\n\r\nPlease let me know if there's any other detail that I need to add to help solving this issue. Thank you.", "author": "yogisalomo", "labels": ["crash", "llvm:analysis"], "comments": [{"author": "aheejin", "body": "How can I reproduce it? I can't find any instructions in that repo. I cloned the repo and ran `./build_wasm.sh`, but it says\r\n```console\r\naheejin@aheejin:~/test/onnx-simplifier$ ./build_wasm.sh \r\n+ command -v emcmake\r\n+++ dirname -- ./build_wasm.sh\r\n++ cd -- .\r\n++ pwd\r\n+ SCRIPT_DIR=/usr/local/google/home/aheejin/test/onnx-simplifier\r\n+ WITH_NODE_RAW_FS=OFF\r\n+ cd /usr/local/google/home/aheejin/test/onnx-simplifier\r\n+ pushd third_party/onnxruntime/cmake/external/protobuf/cmake\r\n./build_wasm.sh: line 11: pushd: third_party/onnxruntime/cmake/external/protobuf/cmake: No such file or directory\r\n```"}, {"author": "yogisalomo", "body": "@aheejin thank you for responding. From the error message, seems like you didn't clone the repository recursively and therefore missing the files of the third_party repositories. Can you please try that?"}, {"author": "aheejin", "body": "Sorry for the long wait! In case you use emscripten releases, 4.0.1 release contains the fix: https://github.com/emscripten-core/emsdk/releases/tag/4.0.1"}]}, "verified": true} {"bug_id": "111887", "issue_url": "https://github.com/llvm/llvm-project/issues/111887", "bug_type": "crash", "base_commit": "195486950fa64938e62f6d85d31222fa41d0ee09", "knowledge_cutoff": "2024-10-10T18:31:43Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "4b5018d2311596778cade4db5177e2ab879cc218", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[1293, 1300], [6322, 6328], [18820, 18826], [18916, 18921], [18936, 18941], [19015, 19031], [19054, 19074]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::buildExternalUses", "tryToReduce"]}}, "patch": "commit 4b5018d2311596778cade4db5177e2ab879cc218\nAuthor: Alexey Bataev \nDate: Thu Oct 10 13:40:22 2024 -0700\n\n [SLP]Track repeated reduced value as it might be vectorized\n \n Need to track changes with the repeated reduced value, since it might be\n vectorized in the next attempt for reduction vectorization, to correctly\n generate the code and avoid compiler crash.\n \n Fixes #111887\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 94de520a2715..e2958c49b8ca 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -1293,8 +1293,7 @@ public:\n using InstrList = SmallVector;\n using ValueSet = SmallPtrSet;\n using StoreList = SmallVector;\n- using ExtraValueToDebugLocsMap =\n- MapVector>;\n+ using ExtraValueToDebugLocsMap = SmallDenseSet;\n using OrdersType = SmallVector;\n \n BoUpSLP(Function *Func, ScalarEvolution *Se, TargetTransformInfo *Tti,\n@@ -6322,7 +6321,7 @@ void BoUpSLP::buildExternalUses(\n continue;\n \n // Check if the scalar is externally used as an extra arg.\n- const auto *ExtI = ExternallyUsedValues.find(Scalar);\n+ const auto ExtI = ExternallyUsedValues.find(Scalar);\n if (ExtI != ExternallyUsedValues.end()) {\n int FoundLane = Entry->findLaneForValue(Scalar);\n LLVM_DEBUG(dbgs() << \"SLP: Need to extract: Extra arg from lane \"\n@@ -18820,7 +18819,7 @@ public:\n // List of the values that were reduced in other trees as part of gather\n // nodes and thus requiring extract if fully vectorized in other trees.\n SmallPtrSet RequiredExtract;\n- Value *VectorizedTree = nullptr;\n+ WeakTrackingVH VectorizedTree = nullptr;\n bool CheckForReusedReductionOps = false;\n // Try to vectorize elements based on their type.\n SmallVector States;\n@@ -18916,6 +18915,7 @@ public:\n bool SameScaleFactor = false;\n bool OptReusedScalars = IsSupportedHorRdxIdentityOp &&\n SameValuesCounter.size() != Candidates.size();\n+ BoUpSLP::ExtraValueToDebugLocsMap ExternallyUsedValues;\n if (OptReusedScalars) {\n SameScaleFactor =\n (RdxKind == RecurKind::Add || RdxKind == RecurKind::FAdd ||\n@@ -18936,6 +18936,7 @@ public:\n emitScaleForReusedOps(Candidates.front(), Builder, Cnt);\n VectorizedTree = GetNewVectorizedTree(VectorizedTree, RedVal);\n VectorizedVals.try_emplace(OrigV, Cnt);\n+ ExternallyUsedValues.insert(OrigV);\n continue;\n }\n }\n@@ -19015,17 +19016,18 @@ public:\n V.reorderBottomToTop(/*IgnoreReorder=*/true);\n // Keep extracted other reduction values, if they are used in the\n // vectorization trees.\n- BoUpSLP::ExtraValueToDebugLocsMap LocalExternallyUsedValues;\n+ BoUpSLP::ExtraValueToDebugLocsMap LocalExternallyUsedValues(\n+ ExternallyUsedValues);\n // The reduction root is used as the insertion point for new\n // instructions, so set it as externally used to prevent it from being\n // deleted.\n- LocalExternallyUsedValues[ReductionRoot];\n+ LocalExternallyUsedValues.insert(ReductionRoot);\n for (unsigned Cnt = 0, Sz = ReducedVals.size(); Cnt < Sz; ++Cnt) {\n if (Cnt == I || (ShuffledExtracts && Cnt == I - 1))\n continue;\n for (Value *V : ReducedVals[Cnt])\n if (isa(V))\n- LocalExternallyUsedValues[TrackedVals[V]];\n+ LocalExternallyUsedValues.insert(TrackedVals[V]);\n }\n if (!IsSupportedHorRdxIdentityOp) {\n // Number of uses of the candidates in the vector of values.\n@@ -19054,21 +19056,21 @@ public:\n // Check if the scalar was vectorized as part of the vectorization\n // tree but not the top node.\n if (!VLScalars.contains(RdxVal) && V.isVectorized(RdxVal)) {\n- LocalExternallyUsedValues[RdxVal];\n+ LocalExternallyUsedValues.insert(RdxVal);\n continue;\n }\n Value *OrigV = TrackedToOrig.at(RdxVal);\n unsigned NumOps =\n VectorizedVals.lookup(OrigV) + At(SameValuesCounter, OrigV);\n if (NumOps != ReducedValsToOps.at(OrigV).size())\n- LocalExternallyUsedValues[RdxVal];\n+ LocalExternallyUsedValues.insert(RdxVal);\n }\n // Do not need the list of reused scalars in regular mode anymore.\n if (!IsSupportedHorRdxIdentityOp)\n SameValuesCounter.clear();\n for (Value *RdxVal : VL)\n if (RequiredExtract.contains(RdxVal))\n- LocalExternallyUsedValues[RdxVal];\n+ LocalExternallyUsedValues.insert(RdxVal);\n V.buildExternalUses(LocalExternallyUsedValues);\n \n V.computeMinimumValueSizes();\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/RISCV/reduced-value-repeated-and-vectorized.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=riscv64-unknown-linux-gnu -mattr=+v < %s"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nentry:\n %0 = load i16, ptr null, align 2\n %1 = xor i16 %0, 0\n %2 = tail call i16 @llvm.smax.i16(i16 %1, i16 %0)\n %3 = tail call i16 @llvm.smax.i16(i16 0, i16 %2)\n %4 = load i16, ptr getelementptr inbounds (i8, ptr null, i64 6), align 2\n %5 = xor i16 %4, 0\n %6 = tail call i16 @llvm.smax.i16(i16 %5, i16 %0)\n %7 = tail call i16 @llvm.smax.i16(i16 %3, i16 %6)\n %8 = load i16, ptr getelementptr (i8, ptr null, i64 12), align 2\n %9 = xor i16 %8, 0\n %10 = tail call i16 @llvm.smax.i16(i16 %9, i16 %0)\n %11 = tail call i16 @llvm.smax.i16(i16 %7, i16 %10)\n %12 = load i16, ptr getelementptr (i8, ptr null, i64 18), align 2\n %13 = xor i16 %12, 0\n %14 = tail call i16 @llvm.smax.i16(i16 %13, i16 %0)\n %15 = tail call i16 @llvm.smax.i16(i16 %11, i16 %14)\n %16 = tail call i16 @llvm.smax.i16(i16 %15, i16 0)\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i16 @llvm.smax.i16(i16, i16) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[SLP-Vectorizer] Segfault in `HorizontalReduction::matchAssociativeReduction`", "body": "Testcase:\r\n```llvm ir\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main() #0 {\r\nfor.body.preheader:\r\n %0 = load i16, ptr null, align 2\r\n %1 = xor i16 %0, 0\r\n %2 = tail call i16 @llvm.smax.i16(i16 %1, i16 %0)\r\n %3 = tail call i16 @llvm.smax.i16(i16 0, i16 %2)\r\n %4 = load i16, ptr getelementptr inbounds (i8, ptr null, i64 6), align 2\r\n %5 = xor i16 %4, 0\r\n %6 = tail call i16 @llvm.smax.i16(i16 %5, i16 %0)\r\n %7 = tail call i16 @llvm.smax.i16(i16 %3, i16 %6)\r\n %8 = load i16, ptr getelementptr (i8, ptr null, i64 12), align 2\r\n %9 = xor i16 %8, 0\r\n %10 = tail call i16 @llvm.smax.i16(i16 %9, i16 %0)\r\n %11 = tail call i16 @llvm.smax.i16(i16 %7, i16 %10)\r\n %12 = load i16, ptr getelementptr (i8, ptr null, i64 18), align 2\r\n %13 = xor i16 %12, 0\r\n %14 = tail call i16 @llvm.smax.i16(i16 %13, i16 %0)\r\n %15 = tail call i16 @llvm.smax.i16(i16 %11, i16 %14)\r\n %16 = tail call i16 @llvm.smax.i16(i16 %15, i16 0)\r\n ret i32 0\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i16 @llvm.smax.i16(i16, i16) #1\r\n\r\n; uselistorder directives\r\nuselistorder ptr null, { 3, 2, 1, 0 }\r\nuselistorder ptr @llvm.smax.i16, { 8, 7, 6, 5, 4, 3, 2, 1, 0 }\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zacas,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\nattributes #1 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=slp-vectorizer reduced.ll\r\nWARNING: You're attempting to print out a bitcode file.\r\nThis is inadvisable as it may cause display problems. If\r\nyou REALLY want to taste LLVM bitcode first-hand, you\r\ncan force output with the `-f' option.\r\n\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=slp-vectorizer reduced.ll\r\n1. Running pass \"function(slp-vectorizer)\" on module \"reduced.ll\"\r\n2. Running pass \"slp-vectorizer\" on function \"main\"\r\n #0 0x0000602da32b35e0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x30fe5e0)\r\n #1 0x0000602da32b09ef llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x30fb9ef)\r\n #2 0x0000602da32b0b45 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007ea045842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x0000602da29cea64 llvm::getUnderlyingObject(llvm::Value const*, unsigned int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2819a64)\r\n #5 0x0000602da22d706d (anonymous namespace)::HorizontalReduction::matchAssociativeReduction(llvm::slpvectorizer::BoUpSLP&, llvm::Instruction*, llvm::ScalarEvolution&, llvm::DataLayout const&, llvm::TargetLibraryInfo const&)::'lambda0'(unsigned long, llvm::LoadInst*)::operator()(unsigned long, llvm::LoadInst*) const SLPVectorizer.cpp:0:0\r\n #6 0x0000602da22d797e generateKeySubkey(llvm::Value*, llvm::TargetLibraryInfo const*, llvm::function_ref, bool) (.constprop.0) SLPVectorizer.cpp:0:0\r\n #7 0x0000602da230d085 (anonymous namespace)::HorizontalReduction::matchAssociativeReduction(llvm::slpvectorizer::BoUpSLP&, llvm::Instruction*, llvm::ScalarEvolution&, llvm::DataLayout const&, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n #8 0x0000602da231f7cc llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*, llvm::SmallVectorImpl&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x216a7cc)\r\n #9 0x0000602da2323ec2 llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*) (.constprop.0) SLPVectorizer.cpp:0:0\r\n#10 0x0000602da2327dfb llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2172dfb)\r\n#11 0x0000602da232e400 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#12 0x0000602da232ef62 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2179f62)\r\n#13 0x0000602da1071366 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xebc366)\r\n#14 0x0000602da30bc88f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f0788f)\r\n#15 0x0000602da1073bb6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xebebb6)\r\n#16 0x0000602da30bac1b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f05c1b)\r\n#17 0x0000602da106fcc6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xebacc6)\r\n#18 0x0000602da30bb77d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f0677d)\r\n#19 0x0000602da088061b llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x6cb61b)\r\n#20 0x0000602da0872331 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x6bd331)\r\n#21 0x00007ea045829d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#22 0x00007ea045829e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#23 0x00007ea045829e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#24 0x0000602da0868855 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x6b3855)\r\nzsh: segmentation fault (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\n\r\nGodbolt: https://godbolt.org/z/nbar8q6ej\r\n\r\nFound via fuzzer (C program before reduction).", "author": "patrick-rivos", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "110193", "issue_url": "https://github.com/llvm/llvm-project/issues/110193", "bug_type": "crash", "base_commit": "7dfdca1961aadc75ca397818bfb9bd32f1879248", "knowledge_cutoff": "2024-09-27T00:54:58Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "af6354634d2cec14570108ee038ca4b18cf6856a", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[10340, 10352]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::getEntryCost"]}}, "patch": "commit af6354634d2cec14570108ee038ca4b18cf6856a\nAuthor: Alexey Bataev \nDate: Fri Sep 27 03:55:17 2024 -0700\n\n [SLP]Look for vector user when estimating the cost\n \n Need to find the first vector node user, not the very first user node at\n all. The very first user might be a gather, vectorized as clustered,\n which may cause compiler crash.\n \n Fixes https://github.com/llvm/llvm-project/issues/110193\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 48a8627ab63e..dee0b7e1f437 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -10340,13 +10340,16 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef VectorizedVals,\n InstructionCost VecCost = VectorCost(CommonCost);\n // Check if the current node must be resized, if the parent node is not\n // resized.\n- if (!UnaryInstruction::isCast(E->getOpcode()) && E->Idx != 0 &&\n+ if (It != MinBWs.end() && !UnaryInstruction::isCast(E->getOpcode()) &&\n+ E->Idx != 0 &&\n (E->getOpcode() != Instruction::Load ||\n !E->UserTreeIndices.empty())) {\n- const EdgeInfo &EI = E->UserTreeIndices.front();\n- if ((EI.UserTE->getOpcode() != Instruction::Select ||\n- EI.EdgeIdx != 0) &&\n- It != MinBWs.end()) {\n+ const EdgeInfo &EI =\n+ *find_if(E->UserTreeIndices, [](const EdgeInfo &EI) {\n+ return !EI.UserTE->isGather() || EI.EdgeIdx != UINT_MAX;\n+ });\n+ if (EI.UserTE->getOpcode() != Instruction::Select ||\n+ EI.EdgeIdx != 0) {\n auto UserBWIt = MinBWs.find(EI.UserTE);\n Type *UserScalarTy =\n EI.UserTE->getOperand(EI.EdgeIdx).front()->getType();\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/minbw-multiused-from-gather.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define i1 @test() {\nentry:\n %c18.i308.i.i = shl i32 0, 0\n %c19.i315.i.i = shl i32 0, 0\n %and.19.i316.i.i = and i32 %c18.i308.i.i, %c19.i315.i.i\n %c20.i322.i.i = shl i32 0, 0\n %and.20.i323.i.i = and i32 %and.19.i316.i.i, %c20.i322.i.i\n %c21.i329.i.i = shl i32 0, 0\n %and.21.i330.i.i = and i32 %and.20.i323.i.i, %c21.i329.i.i\n %0 = trunc i64 0 to i32\n %conv85.22.i333.i.i = or i32 0, %0\n %cmp3.i.22.i334.i.i = icmp ugt i32 %conv85.22.i333.i.i, 0\n %shl.i111.22.i335.i.i = select i1 %cmp3.i.22.i334.i.i, i32 0, i32 0\n %c22.i336.i.i = shl i32 %conv85.22.i333.i.i, %shl.i111.22.i335.i.i\n %and.22.i337.i.i = and i32 %and.21.i330.i.i, %c22.i336.i.i\n %1 = trunc i64 0 to i32\n %conv85.23.i340.i.i = or i32 0, %1\n %cmp3.i.23.i341.i.i = icmp ugt i32 %conv85.23.i340.i.i, 0\n %shl.i111.23.i342.i.i = select i1 %cmp3.i.23.i341.i.i, i32 0, i32 0\n %c23.i343.i.i = shl i32 %conv85.23.i340.i.i, %shl.i111.23.i342.i.i\n %and.23.i344.i.i = and i32 %and.22.i337.i.i, %c23.i343.i.i\n %2 = trunc i64 0 to i32\n %conv85.24.i347.i.i = or i32 0, %2\n %cmp3.i.24.i348.i.i = icmp ugt i32 %conv85.24.i347.i.i, 0\n %shl.i111.24.i349.i.i = select i1 %cmp3.i.24.i348.i.i, i32 0, i32 0\n %c24.i350.i.i = shl i32 %conv85.24.i347.i.i, %shl.i111.24.i349.i.i\n %and.24.i351.i.i = and i32 %and.23.i344.i.i, %c24.i350.i.i\n %3 = trunc i64 0 to i32\n %conv85.25.i354.i.i = or i32 0, %3\n %cmp3.i.25.i355.i.i = icmp ugt i32 %conv85.25.i354.i.i, 0\n %shl.i111.25.i356.i.i = select i1 %cmp3.i.25.i355.i.i, i32 0, i32 0\n %c25.i357.i.i = shl i32 %conv85.25.i354.i.i, %shl.i111.25.i356.i.i\n %and.25.i358.i.i = and i32 %and.24.i351.i.i, %c25.i357.i.i\n %conv109.i.i = trunc i32 %and.25.i358.i.i to i8\n %cmp.i.i54.i = icmp eq i8 %conv109.i.i, 0\n ret i1 %cmp.i.i54.i\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Assertion `OpIdx < Operands.size() && \"Off bounds\"' failed.", "body": "Reproducer: https://godbolt.org/z/6GbP5f7qa\r\n```\r\n; bin/opt -passes=slp-vectorizer reduced.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i1 @main() {\r\nentry:\r\n %cond.i112.18.i308.i.i = shl i32 0, 0\r\n %cond.i112.19.i315.i.i = shl i32 0, 0\r\n %and.19.i316.i.i = and i32 %cond.i112.18.i308.i.i, %cond.i112.19.i315.i.i\r\n %cond.i112.20.i322.i.i = shl i32 0, 0\r\n %and.20.i323.i.i = and i32 %and.19.i316.i.i, %cond.i112.20.i322.i.i\r\n %cond.i112.21.i329.i.i = shl i32 0, 0\r\n %and.21.i330.i.i = and i32 %and.20.i323.i.i, %cond.i112.21.i329.i.i\r\n %0 = trunc i64 0 to i32\r\n %conv85.22.i333.i.i = or i32 0, %0\r\n %cmp3.i.22.i334.i.i = icmp ugt i32 %conv85.22.i333.i.i, 0\r\n %shl.i111.22.i335.i.i = select i1 %cmp3.i.22.i334.i.i, i32 0, i32 0\r\n %cond.i112.22.i336.i.i = shl i32 %conv85.22.i333.i.i, %shl.i111.22.i335.i.i\r\n %and.22.i337.i.i = and i32 %and.21.i330.i.i, %cond.i112.22.i336.i.i\r\n %1 = trunc i64 0 to i32\r\n %conv85.23.i340.i.i = or i32 0, %1\r\n %cmp3.i.23.i341.i.i = icmp ugt i32 %conv85.23.i340.i.i, 0\r\n %shl.i111.23.i342.i.i = select i1 %cmp3.i.23.i341.i.i, i32 0, i32 0\r\n %cond.i112.23.i343.i.i = shl i32 %conv85.23.i340.i.i, %shl.i111.23.i342.i.i\r\n %and.23.i344.i.i = and i32 %and.22.i337.i.i, %cond.i112.23.i343.i.i\r\n %2 = trunc i64 0 to i32\r\n %conv85.24.i347.i.i = or i32 0, %2\r\n %cmp3.i.24.i348.i.i = icmp ugt i32 %conv85.24.i347.i.i, 0\r\n %shl.i111.24.i349.i.i = select i1 %cmp3.i.24.i348.i.i, i32 0, i32 0\r\n %cond.i112.24.i350.i.i = shl i32 %conv85.24.i347.i.i, %shl.i111.24.i349.i.i\r\n %and.24.i351.i.i = and i32 %and.23.i344.i.i, %cond.i112.24.i350.i.i\r\n %3 = trunc i64 0 to i32\r\n %conv85.25.i354.i.i = or i32 0, %3\r\n %cmp3.i.25.i355.i.i = icmp ugt i32 %conv85.25.i354.i.i, 0\r\n %shl.i111.25.i356.i.i = select i1 %cmp3.i.25.i355.i.i, i32 0, i32 0\r\n %cond.i112.25.i357.i.i = shl i32 %conv85.25.i354.i.i, %shl.i111.25.i356.i.i\r\n %and.25.i358.i.i = and i32 %and.24.i351.i.i, %cond.i112.25.i357.i.i\r\n %conv109.i.i = trunc i32 %and.25.i358.i.i to i8\r\n %cmp.i.i54.i = icmp eq i8 %conv109.i.i, 0\r\n ret i1 %cmp.i.i54.i\r\n}\r\n```\r\n```\r\nopt: /home/dtcxzyw/WorkSpace/Projects/compilers/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:3247: llvm::slpvectorizer::BoUpSLP::ValueList& llvm::slpvectorizer::BoUpSLP::TreeEntry::getOperand(unsigned int): Assertion `OpIdx < Operands.size() && \"Off bounds\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: bin/opt -passes=slp-vectorizer reduced.ll\r\n1. Running pass \"function(slp-vectorizer)\" on module \"reduced.ll\"\r\n2. Running pass \"slp-vectorizer\" on function \"main\"\r\n #0 0x000075e591a13b52 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.20.0git+0x213b52)\r\n #1 0x000075e591a10a1f llvm::sys::RunSignalHandlers() (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMSupport.so.20.0git+0x210a1f)\r\n #2 0x000075e591a10b65 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x000075e591442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x000075e5914969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x000075e5914969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x000075e5914969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x000075e591442476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x000075e5914287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x000075e59142871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x000075e591439e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x000075e58c574c14 (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x174c14)\r\n#12 0x000075e58c5bcafe llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1bcafe)\r\n#13 0x000075e58c5d0eed llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1d0eed)\r\n#14 0x000075e58c5d68cc (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#15 0x000075e58c5d997a llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*, llvm::SmallVectorImpl&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1d997a)\r\n#16 0x000075e58c5ddf28 llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1ddf28)\r\n#17 0x000075e58c5de0cb bool llvm::SLPVectorizerPass::vectorizeCmpInsts>(llvm::iterator_range>, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1de0cb)\r\n#18 0x000075e58c5e2299 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1e2299)\r\n#19 0x000075e58c5e8c0f llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1e8c0f)\r\n#20 0x000075e58c5e9876 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMVectorize.so.20.0git+0x1e9876)\r\n#21 0x000075e58ced5195 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMPasses.so.20.0git+0xd5195)\r\n#22 0x000075e58a15de5d llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x35de5d)\r\n#23 0x000075e5904d9d05 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.20.0git+0xd9d05)\r\n#24 0x000075e58a15bf96 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x35bf96)\r\n#25 0x000075e5904da6c5 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMX86CodeGen.so.20.0git+0xda6c5)\r\n#26 0x000075e58a15cb1a llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/../lib/libLLVMCore.so.20.0git+0x35cb1a)\r\n#27 0x000075e591c041c1 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.20.0git+0x2d1c1)\r\n#28 0x000075e591c0f944 optMain (/home/dtcxzyw/WorkSpace/Projects/compilers/LLVM/llvm-build/bin/../lib/libLLVMOptDriver.so.20.0git+0x38944)\r\n#29 0x000075e591429d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#30 0x000075e591429e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#31 0x000075e591429e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#32 0x00005b421cc0e095 _start (bin/opt+0x1095)\r\nAborted (core dumped)\r\n```\r\n\r\nllvm version: 2b84ef06ac55ac8de3c210d059ec3a3c96666a90", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "122583", "issue_url": "https://github.com/llvm/llvm-project/issues/122583", "bug_type": "miscompilation", "base_commit": "16e45b8fac797c6d4ba161228b54665492204a9d", "knowledge_cutoff": "2025-01-11T07:48:51Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "066b88879ab5c195e7e14609e546cc238c2f3bf3", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[2441, 2446]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["appendOperandsOfVL"]}}, "patch": "commit 066b88879ab5c195e7e14609e546cc238c2f3bf3\nAuthor: Alexey Bataev \nDate: Mon Jan 13 10:51:26 2025 -0800\n\n [SLP]Correctly set vector operand for extracts with poisons\n \n When extracts are vectorized and it has some poison values instead of\n instructions, need to correctly set the vectorized operand not as\n poison, but as a main vector operand of the main extract instruction.\n \n Fixes #122583\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 4b0ed5b30179..2742c3777c1e 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -2441,6 +2441,17 @@ public:\n // operations or alternating sequences (e.g., +, -), we can safely\n // tell the inverse operations by checking commutativity.\n if (isa(VL[Lane])) {\n+ if (auto *EI = dyn_cast(VL0)) {\n+ if (OpIdx == 0) {\n+ OpsVec[OpIdx][Lane] = {EI->getVectorOperand(), true, false};\n+ continue;\n+ }\n+ } else if (auto *EV = dyn_cast(VL0)) {\n+ if (OpIdx == 0) {\n+ OpsVec[OpIdx][Lane] = {EV->getAggregateOperand(), true, false};\n+ continue;\n+ }\n+ }\n OpsVec[OpIdx][Lane] = {\n PoisonValue::get(VL0->getOperand(OpIdx)->getType()), true,\n false};\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/extractelemets-extended-by-poison.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@j = global [4 x i64] zeroinitializer\n\ndefine i32 @main() {\nentry:\n %.pre.i = load i64, ptr getelementptr inbounds nuw (i8, ptr @j, i64 24), align 8\n %.pre50.i = load i64, ptr getelementptr inbounds nuw (i8, ptr @j, i64 16), align 16\n %.pre51.i = load i64, ptr getelementptr inbounds nuw (i8, ptr @j, i64 8), align 8\n %.pre52.i = load i64, ptr @j, align 16\n %0 = or i64 %.pre51.i, 0\n %1 = trunc i64 %.pre.i to i32\n %2 = add i32 %1, 0\n %3 = trunc i64 %.pre50.i to i32\n %4 = add i32 %3, 0\n %5 = trunc i64 %.pre51.i to i32\n %6 = add i32 %5, 0\n %7 = trunc i64 0 to i32\n %8 = add i32 %5, 0\n %9 = add i32 %7, 0\n %10 = add i32 %1, 0\n %11 = add i32 %3, 0\n %12 = add i32 %5, 0\n %13 = add i32 %7, 0\n %14 = trunc i64 %.pre.i to i32\n %15 = add i32 %14, 0\n %16 = trunc i64 %.pre50.i to i32\n %17 = add i32 %16, 0\n %18 = trunc i64 %.pre51.i to i32\n %19 = add i32 %18, 0\n %20 = trunc i64 %.pre52.i to i32\n %conv14.1.i = or i32 %9, %13\n %21 = or i32 %conv14.1.i, %6\n %22 = or i32 %21, %8\n %23 = or i32 %22, %12\n %24 = or i32 %23, %4\n %25 = or i32 %24, %11\n %26 = or i32 %25, %2\n %27 = or i32 %26, %10\n %28 = or i32 %27, %15\n %29 = or i32 %28, %17\n %30 = or i32 %29, %19\n %31 = add i32 %14, 0\n %32 = add i32 %16, 0\n %33 = add i32 %18, 0\n %34 = add i32 %20, 0\n %35 = add i32 %14, 0\n %36 = add i32 %16, 0\n %37 = add i32 %18, 0\n %38 = add i32 %20, 0\n %39 = add i32 %14, 0\n %40 = add i32 %16, 0\n %41 = add i32 %18, 0\n %42 = add i32 %20, 0\n %inc.3.3.i.1 = or i64 %.pre52.i, 0\n %conv14.i.1 = or i32 %38, %34\n %conv14.1.i.1 = or i32 %conv14.i.1, %42\n %conv14.3.i.1 = or i32 %conv14.1.i.1, %33\n %conv14.145.i.1 = or i32 %conv14.3.i.1, %37\n %conv14.1.1.i.1 = or i32 %conv14.145.i.1, %41\n %conv14.3.1.i.1 = or i32 %conv14.1.1.i.1, %32\n %conv14.247.i.1 = or i32 %conv14.3.1.i.1, %36\n %conv14.1.2.i.1 = or i32 %conv14.247.i.1, %40\n %conv14.3.2.i.1 = or i32 %conv14.1.2.i.1, %31\n %conv14.349.i.1 = or i32 %conv14.3.2.i.1, %35\n %conv14.1.3.i.1 = or i32 %conv14.349.i.1, %39\n %conv14.3.3.i.1 = or i32 %conv14.1.3.i.1, %30\n ret i32 %conv14.3.3.i.1\n}\n", "lli_expected_out": ""}]}], "issue": {"title": "[SLPVectorizer] Miscompilation", "body": "Reproducer: https://godbolt.org/z/7eEEeeKoo\nSorry, I cannot provide alive2 link since `llvm.vector.insert.v8i64.v4i64` is not supported.\n```\n; bin/opt -passes=slp-vectorizer reduced.ll -S -o opt.ll\n\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@j = global [4 x i64] zeroinitializer\n\ndefine i32 @main() {\nentry:\n %.pre.i = load i64, ptr getelementptr inbounds nuw (i8, ptr @j, i64 24), align 8\n %.pre50.i = load i64, ptr getelementptr inbounds nuw (i8, ptr @j, i64 16), align 16\n %.pre51.i = load i64, ptr getelementptr inbounds nuw (i8, ptr @j, i64 8), align 8\n %.pre52.i = load i64, ptr @j, align 16\n %0 = or i64 %.pre51.i, 0\n %1 = trunc i64 %.pre.i to i32\n %2 = add i32 %1, 0\n %3 = trunc i64 %.pre50.i to i32\n %4 = add i32 %3, 0\n %5 = trunc i64 %.pre51.i to i32\n %6 = add i32 %5, 0\n %7 = trunc i64 0 to i32\n %8 = add i32 %5, 0\n %9 = add i32 %7, 0\n %10 = add i32 %1, 0\n %11 = add i32 %3, 0\n %12 = add i32 %5, 0\n %13 = add i32 %7, 0\n %14 = trunc i64 %.pre.i to i32\n %15 = add i32 %14, 0\n %16 = trunc i64 %.pre50.i to i32\n %17 = add i32 %16, 0\n %18 = trunc i64 %.pre51.i to i32\n %19 = add i32 %18, 0\n %20 = trunc i64 %.pre52.i to i32\n %conv14.1.i = or i32 %9, %13\n %21 = or i32 %conv14.1.i, %6\n %22 = or i32 %21, %8\n %23 = or i32 %22, %12\n %24 = or i32 %23, %4\n %25 = or i32 %24, %11\n %26 = or i32 %25, %2\n %27 = or i32 %26, %10\n %28 = or i32 %27, %15\n %29 = or i32 %28, %17\n %30 = or i32 %29, %19\n %31 = add i32 %14, 0\n %32 = add i32 %16, 0\n %33 = add i32 %18, 0\n %34 = add i32 %20, 0\n %35 = add i32 %14, 0\n %36 = add i32 %16, 0\n %37 = add i32 %18, 0\n %38 = add i32 %20, 0\n %39 = add i32 %14, 0\n %40 = add i32 %16, 0\n %41 = add i32 %18, 0\n %42 = add i32 %20, 0\n %inc.3.3.i.1 = or i64 %.pre52.i, 0\n %conv14.i.1 = or i32 %38, %34\n %conv14.1.i.1 = or i32 %conv14.i.1, %42\n %conv14.3.i.1 = or i32 %conv14.1.i.1, %33\n %conv14.145.i.1 = or i32 %conv14.3.i.1, %37\n %conv14.1.1.i.1 = or i32 %conv14.145.i.1, %41\n %conv14.3.1.i.1 = or i32 %conv14.1.1.i.1, %32\n %conv14.247.i.1 = or i32 %conv14.3.1.i.1, %36\n %conv14.1.2.i.1 = or i32 %conv14.247.i.1, %40\n %conv14.3.2.i.1 = or i32 %conv14.1.2.i.1, %31\n %conv14.349.i.1 = or i32 %conv14.3.2.i.1, %35\n %conv14.1.3.i.1 = or i32 %conv14.349.i.1, %39\n %conv14.3.3.i.1 = or i32 %conv14.1.3.i.1, %30\n ret i32 %conv14.3.3.i.1\n}\n```\nOutput:\n```\nsource_filename = \"/app/example.ll\"\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i32 @main() {\n %0 = load <4 x i64>, ptr @j, align 16\n %1 = or i64 poison, 0\n %2 = shufflevector <4 x i64> %0, <4 x i64> poison, <8 x i32> \n %3 = shufflevector <4 x i64> %0, <4 x i64> poison, <8 x i32> \n %4 = shufflevector <8 x i64> %3, <8 x i64> , <8 x i32> \n %5 = call <8 x i64> @llvm.vector.insert.v8i64.v4i64(<8 x i64> %4, <4 x i64> %0, i64 0)\n %6 = trunc <8 x i64> %5 to <8 x i32>\n %7 = shufflevector <8 x i32> %6, <8 x i32> poison, <16 x i32> \n %8 = add <16 x i32> %7, zeroinitializer\n %9 = extractelement <4 x i64> %0, i32 0\n %inc.3.3.i.1 = or i64 %9, 0\n %10 = call i32 @llvm.vector.reduce.or.v16i32(<16 x i32> %8)\n %11 = call i32 @llvm.vector.reduce.or.v8i32(<8 x i32> poison)\n %op.rdx = or i32 %10, %11\n ret i32 %op.rdx\n}\n\ndeclare <8 x i64> @llvm.vector.insert.v8i64.v4i64(<8 x i64>, <4 x i64>, i64 immarg) #0\n\ndeclare i32 @llvm.vector.reduce.or.v16i32(<16 x i32>) #0\n\ndeclare i32 @llvm.vector.reduce.or.v8i32(<8 x i32>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n```\nlli output:\n```\n> bin/lli reduced.ll\n> echo $?\n0\n> bin/lli opt.ll\n> echo $?\n255\n```\n[llubi](https://github.com/dtcxzyw/llvm-ub-aware-interpreter) output:\nBefore:\n```\n> ./llubi reduced.ll --verbose\nEntering function main\n %0 = getelementptr inbounds nuw i8, ptr @j, i64 24 -> Ptr 40[@j + 24]\n %.pre.i = load i64, ptr %0, align 8 -> i64 0\n %1 = getelementptr inbounds nuw i8, ptr @j, i64 16 -> Ptr 32[@j + 16]\n %.pre50.i = load i64, ptr %1, align 16 -> i64 0\n %2 = getelementptr inbounds nuw i8, ptr @j, i64 8 -> Ptr 24[@j + 8]\n %.pre51.i = load i64, ptr %2, align 8 -> i64 0\n %.pre52.i = load i64, ptr @j, align 16 -> i64 0\n %3 = or i64 %.pre51.i, 0 -> i64 0\n %4 = trunc i64 %.pre.i to i32 -> i32 0\n %5 = add i32 %4, 0 -> i32 0\n %6 = trunc i64 %.pre50.i to i32 -> i32 0\n %7 = add i32 %6, 0 -> i32 0\n %8 = trunc i64 %.pre51.i to i32 -> i32 0\n %9 = add i32 %8, 0 -> i32 0\n %10 = trunc i64 0 to i32 -> i32 0\n %11 = add i32 %8, 0 -> i32 0\n %12 = add i32 %10, 0 -> i32 0\n %13 = add i32 %4, 0 -> i32 0\n %14 = add i32 %6, 0 -> i32 0\n %15 = add i32 %8, 0 -> i32 0\n %16 = add i32 %10, 0 -> i32 0\n %17 = trunc i64 %.pre.i to i32 -> i32 0\n %18 = add i32 %17, 0 -> i32 0\n %19 = trunc i64 %.pre50.i to i32 -> i32 0\n %20 = add i32 %19, 0 -> i32 0\n %21 = trunc i64 %.pre51.i to i32 -> i32 0\n %22 = add i32 %21, 0 -> i32 0\n %23 = trunc i64 %.pre52.i to i32 -> i32 0\n %conv14.1.i = or i32 %12, %16 -> i32 0\n %24 = or i32 %conv14.1.i, %9 -> i32 0\n %25 = or i32 %24, %11 -> i32 0\n %26 = or i32 %25, %15 -> i32 0\n %27 = or i32 %26, %7 -> i32 0\n %28 = or i32 %27, %14 -> i32 0\n %29 = or i32 %28, %5 -> i32 0\n %30 = or i32 %29, %13 -> i32 0\n %31 = or i32 %30, %18 -> i32 0\n %32 = or i32 %31, %20 -> i32 0\n %33 = or i32 %32, %22 -> i32 0\n %34 = add i32 %17, 0 -> i32 0\n %35 = add i32 %19, 0 -> i32 0\n %36 = add i32 %21, 0 -> i32 0\n %37 = add i32 %23, 0 -> i32 0\n %38 = add i32 %17, 0 -> i32 0\n %39 = add i32 %19, 0 -> i32 0\n %40 = add i32 %21, 0 -> i32 0\n %41 = add i32 %23, 0 -> i32 0\n %42 = add i32 %17, 0 -> i32 0\n %43 = add i32 %19, 0 -> i32 0\n %44 = add i32 %21, 0 -> i32 0\n %45 = add i32 %23, 0 -> i32 0\n %inc.3.3.i.1 = or i64 %.pre52.i, 0 -> i64 0\n %conv14.i.1 = or i32 %41, %37 -> i32 0\n %conv14.1.i.1 = or i32 %conv14.i.1, %45 -> i32 0\n %conv14.3.i.1 = or i32 %conv14.1.i.1, %36 -> i32 0\n %conv14.145.i.1 = or i32 %conv14.3.i.1, %40 -> i32 0\n %conv14.1.1.i.1 = or i32 %conv14.145.i.1, %44 -> i32 0\n %conv14.3.1.i.1 = or i32 %conv14.1.1.i.1, %35 -> i32 0\n %conv14.247.i.1 = or i32 %conv14.3.1.i.1, %39 -> i32 0\n %conv14.1.2.i.1 = or i32 %conv14.247.i.1, %43 -> i32 0\n %conv14.3.2.i.1 = or i32 %conv14.1.2.i.1, %34 -> i32 0\n %conv14.349.i.1 = or i32 %conv14.3.2.i.1, %38 -> i32 0\n %conv14.1.3.i.1 = or i32 %conv14.349.i.1, %42 -> i32 0\n %conv14.3.3.i.1 = or i32 %conv14.1.3.i.1, %33 -> i32 0\n ret i32 %conv14.3.3.i.1\nExiting function main\n```\nAfter:\n```\n> ./llubi opt.ll --verbose\nEntering function main\n %0 = load <4 x i64>, ptr @j, align 16 -> { i64 0, i64 0, i64 0, i64 0 }\n %1 = or i64 poison, 1 -> poison\n %2 = or i64 poison, 0 -> poison\n %3 = shufflevector <4 x i64> %0, <4 x i64> poison, <8 x i32> -> { poison, poison, poison, poison, poison, i64 0, poison, poison }\n %4 = shufflevector <4 x i64> %0, <4 x i64> poison, <8 x i32> -> { i64 0, i64 0, i64 0, i64 0, poison, poison, poison, poison }\n %5 = shufflevector <8 x i64> %4, <8 x i64> , <8 x i32> -> { poison, poison, poison, poison, i64 0, i64 0, poison, poison }\n %6 = call <8 x i64> @llvm.vector.insert.v8i64.v4i64(<8 x i64> %5, <4 x i64> %0, i64 0) -> { i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, poison, poison }\n %7 = trunc <8 x i64> %6 to <8 x i32> -> { i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, poison, poison }\n %8 = shufflevector <8 x i32> %7, <8 x i32> poison, <16 x i32> -> { i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0 }\n %9 = add <16 x i32> %8, zeroinitializer -> { i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0 }\n %10 = extractelement <4 x i64> %0, i32 0 -> i64 0\n %inc.3.3.i.1 = or i64 %10, 0 -> i64 0\n %11 = call i32 @llvm.vector.reduce.or.v16i32(<16 x i32> %9) -> i32 0\n %12 = call i32 @llvm.vector.reduce.or.v8i32(<8 x i32> poison) -> poison\n %op.rdx = or i32 %11, %12 -> poison\n ret i32 %op.rdx\nExiting function main\n\nUB triggered: Return a poison value\nExited with immediate UB.\nStacktrace:\n```\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:SLPVectorizer", "generated by fuzzer"], "comments": [{"author": "dtcxzyw", "body": "llvm version: 6329355860e9b66bc7ed68b46c166763e408d4cc"}]}, "verified": true} {"bug_id": "114860", "issue_url": "https://github.com/llvm/llvm-project/issues/114860", "bug_type": "crash", "base_commit": "7dffc96a54f90569d6226dd5713c80fc8f30c76f", "knowledge_cutoff": "2024-11-04T20:12:18Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "144bdf3eb7128518ed162c5a168e3ec90922cd9e", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[7521, 7526]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationPlanner::computeBestVF"]}}, "patch": "commit 144bdf3eb7128518ed162c5a168e3ec90922cd9e\nAuthor: Florian Hahn \nDate: Fri Nov 8 20:53:03 2024 +0000\n\n [VPlan] Also check if plan for best legacy VF contains simplifications.\n \n The plan for the VF chosen by the legacy cost model could also contain\n additional simplifications that cause cost differences. Also check if it\n contains simplifications.\n \n Fixes https://github.com/llvm/llvm-project/issues/114860.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex c07af8519049..82d85b112397 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -7521,6 +7521,8 @@ VectorizationFactor LoopVectorizationPlanner::computeBestVF() {\n precomputeCosts(BestPlan, BestFactor.Width, CostCtx);\n assert((BestFactor.Width == LegacyVF.Width ||\n planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width),\n+ CostCtx, OrigLoop) ||\n+ planContainsAdditionalSimplifications(getPlanFor(LegacyVF.Width),\n CostCtx, OrigLoop)) &&\n \" VPlan cost model and legacy cost model disagreed\");\n assert((BestFactor.Width.isScalar() || BestFactor.ScalarCost > 0) &&\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/select-invariant-cond-cost.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "test_invariant_cond_for_select", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"riscv64-unknown-linux-gnu\"\n\ndefine void @test_invariant_cond_for_select(ptr %dst, i8 %x) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %c.1 = icmp eq i8 %x, 0\n %c.2 = icmp sgt i64 %iv, 0\n %c.2.ext = zext i1 %c.2 to i64\n %sel = select i1 %c.1, i64 %c.2.ext, i64 0\n %sel.trunc = trunc i64 %sel to i8\n %gep = getelementptr inbounds i8, ptr %dst, i64 %iv\n store i8 %sel.trunc, ptr %gep, align 1\n %iv.next = add i64 %iv, 4\n %ec = icmp ult i64 %iv, 14\n br i1 %ec, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}]}], "issue": {"title": "[VPlan] Assertion \" VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Testcase:\r\n```c\r\nextern char a[];\r\nchar *b;\r\nint c = 40502824327348188;\r\nlong d[9][9];\r\nint main() {\r\n for (long e = 0; e < 9; ++e)\r\n for (long f = 0; f < 9; ++f)\r\n d[e][f] = 6032102684657866;\r\n for (unsigned char f = b[4]; f < 18; f += c - 216)\r\n a[f] = f % 2 ? d[f][f] : d[f][f] + 2036854775807 > 0;\r\n}\r\n```\r\n```\r\n/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/clang -I/scratch/tc-testing/compiler-fuzz-ci/csmith-build/include -fsigned-char -fno-strict-aliasing -fwrapv -march=rv64gcv -flto -O3 -fuse-ld=lld red.c -o rv64gcv.out\r\n```\r\n\r\n\r\nReduced testcase:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main() #0 {\r\nentry:\r\n br label %for.body15\r\n\r\nfor.cond.cleanup14.loopexit: ; preds = %for.body15\r\n ret i32 0\r\n\r\nfor.body15: ; preds = %for.body15, %entry\r\n %indvars.iv1 = phi i64 [ %indvars.iv.next, %for.body15 ], [ 0, %entry ]\r\n %tobool.not = icmp eq i8 0, 0\r\n %cmp24 = icmp sgt i64 %indvars.iv1, 0\r\n %conv26 = zext i1 %cmp24 to i64\r\n %cond = select i1 %tobool.not, i64 %conv26, i64 0\r\n %conv27 = trunc i64 %cond to i8\r\n store i8 %conv27, ptr null, align 1\r\n %indvars.iv.next = add i64 %indvars.iv1, 4\r\n %cmp12 = icmp ult i64 %indvars.iv1, 14\r\n br i1 %cmp12, label %for.body15, label %for.cond.cleanup14.loopexit\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-ssctr,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-sha,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smmpm,-smnpm,-smrnmi,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssnpm,-sspm,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-supm,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-svvptc,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zacas,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt --passes \"loop-vectorize\" reduced.ll -f\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7519: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt --passes loop-vectorize reduced.ll -f\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"main\"\r\n #0 0x00005f3d9fc9dc00 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x3169c00)\r\n #1 0x00005f3d9fc9b00f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x316700f)\r\n #2 0x00005f3d9fc9b165 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007abb28242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007abb282969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007abb282969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007abb282969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007abb28242476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007abb282287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007abb2822871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007abb28239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00005f3d9ebfb878 llvm::LoopVectorizationPlanner::computeBestVF() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20c7878)\r\n#12 0x00005f3d9ec12dec llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20dedec)\r\n#13 0x00005f3d9ec15f01 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20e1f01)\r\n#14 0x00005f3d9ec16577 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20e2577)\r\n#15 0x00005f3d9da0cba6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xed8ba6)\r\n#16 0x00005f3d9faa518f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f7118f)\r\n#17 0x00005f3d9da0fdd6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xedbdd6)\r\n#18 0x00005f3d9faa351b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f6f51b)\r\n#19 0x00005f3d9da0bc66 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xed7c66)\r\n#20 0x00005f3d9faa407d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f7007d)\r\n#21 0x00005f3d9d20ca8d llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x6d8a8d)\r\n#22 0x00005f3d9d1fe911 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x6ca911)\r\n#23 0x00007abb28229d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#24 0x00007abb28229e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#25 0x00007abb28229e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#26 0x00005f3d9d1f4e65 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x6c0e65)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\n\r\nFound via fuzzer on Oct 24th\r\n\r\nGodbolt: https://godbolt.org/z/zhseEjTso\r\n\r\ncc @fhahn ", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "113516", "issue_url": "https://github.com/llvm/llvm-project/issues/113516", "bug_type": "crash", "base_commit": "0aec4d2b78a08f5f3e4b0ad0340a99e486d00110", "knowledge_cutoff": "2024-10-24T03:05:37Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "d2e7ee77d33e8b3be3b1d4e9bc5bc4c60b62b554", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[16118, 16124]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree"]}}, "patch": "commit d2e7ee77d33e8b3be3b1d4e9bc5bc4c60b62b554\nAuthor: Alexey Bataev \nDate: Thu Oct 24 08:15:21 2024 -0700\n\n [SLP]Do not check for clustered loads only\n \n Since SLP support \"clusterization\" of the non-load instructions, the\n restriction for reduced values for loads only should be removed to avoid\n compiler crash.\n \n Fixes #113516\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex f76e913a57d5..012d85353f08 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -16118,7 +16118,7 @@ BoUpSLP::vectorizeTree(const ExtraValueToDebugLocsMap &ExternallyUsedValues,\n for (Instruction *I : RemovedInsts) {\n const TreeEntry *IE = getTreeEntry(I);\n if (IE->Idx != 0 &&\n- !(VectorizableTree.front()->isGather() && isa(I) &&\n+ !(VectorizableTree.front()->isGather() &&\n !IE->UserTreeIndices.empty() &&\n any_of(IE->UserTreeIndices,\n [&](const EdgeInfo &EI) {\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/non-load-reduced-as-part-of-bv.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "foo", "test_body": "define i1 @foo() {\nentry:\n %tobool.not.not509.i.1311.i.i = icmp ne i32 0, 0\n %cmp80.1319.i.i189 = icmp sgt i32 0, 0\n %tobool.not.not509.i.1.1.i.i = icmp ne i32 0, 0\n %cmp80.1.1.i.i190 = icmp sgt i32 0, 0\n %tobool.not.not509.i.2329.i.i = icmp ne i32 0, 0\n %storemerge.2333.i.i = select i1 %tobool.not.not509.i.2329.i.i, i32 0, i32 0\n %cmp80.2337.i.i192 = icmp sgt i32 0, 0\n %tobool.not.not509.i.1.2.i.i = icmp ne i32 %storemerge.2333.i.i, 0\n %cmp80.1.2.i.i193 = icmp sgt i32 0, 0\n %cmp80.1.2.i.i = select i1 %tobool.not.not509.i.1.2.i.i, i1 %cmp80.1.2.i.i193, i1 false\n %0 = select i1 %cmp80.1.2.i.i, i1 %tobool.not.not509.i.2329.i.i, i1 false\n %1 = select i1 %0, i1 %cmp80.2337.i.i192, i1 false\n %2 = select i1 %1, i1 false, i1 false\n %3 = select i1 %2, i1 %tobool.not.not509.i.1.1.i.i, i1 false\n %4 = select i1 %3, i1 %cmp80.1.1.i.i190, i1 false\n %5 = select i1 %4, i1 %tobool.not.not509.i.1311.i.i, i1 false\n %6 = select i1 %5, i1 %cmp80.1319.i.i189, i1 false\n ret i1 %6\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Assertion `(I->use_empty() || all_of(I->uses(), [&](Use &U) { return isDeleted( cast(U.getUser())); })) && \"trying to erase instruction with users.\"' failed.", "body": "Reduced reproducer: https://godbolt.org/z/azEc6zjod\r\n```\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i1 @foo() {\r\nentry:\r\n %tobool.not.not509.i.1311.i.i = icmp ne i32 0, 0\r\n %cmp80.1319.i.i189 = icmp sgt i32 0, 0\r\n %tobool.not.not509.i.1.1.i.i = icmp ne i32 0, 0\r\n %cmp80.1.1.i.i190 = icmp sgt i32 0, 0\r\n %tobool.not.not509.i.2329.i.i = icmp ne i32 0, 0\r\n %storemerge.2333.i.i = select i1 %tobool.not.not509.i.2329.i.i, i32 0, i32 0\r\n %cmp80.2337.i.i192 = icmp sgt i32 0, 0\r\n %tobool.not.not509.i.1.2.i.i = icmp ne i32 %storemerge.2333.i.i, 0\r\n %cmp80.1.2.i.i193 = icmp sgt i32 0, 0\r\n %cmp80.1.2.i.i = select i1 %tobool.not.not509.i.1.2.i.i, i1 %cmp80.1.2.i.i193, i1 false\r\n %0 = select i1 %cmp80.1.2.i.i, i1 %tobool.not.not509.i.2329.i.i, i1 false\r\n %1 = select i1 %0, i1 %cmp80.2337.i.i192, i1 false\r\n %2 = select i1 %1, i1 false, i1 false\r\n %3 = select i1 %2, i1 %tobool.not.not509.i.1.1.i.i, i1 false\r\n %4 = select i1 %3, i1 %cmp80.1.1.i.i190, i1 false\r\n %5 = select i1 %4, i1 %tobool.not.not509.i.1311.i.i, i1 false\r\n %6 = select i1 %5, i1 %cmp80.1319.i.i189, i1 false\r\n ret i1 %6\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:2748: void llvm::slpvectorizer::BoUpSLP::removeInstructionsAndOperands(llvm::ArrayRef) [with T = llvm::Instruction]: Assertion `(I->use_empty() || all_of(I->uses(), [&](Use &U) { return isDeleted( cast(U.getUser())); })) && \"trying to erase instruction with users.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"foo\"\r\n #0 0x00000000050d8608 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50d8608)\r\n #1 0x00000000050d5fbc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000071afd7642520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000071afd76969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000071afd7642476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000071afd76287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000071afd762871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000071afd7639e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x000000000414e3af void llvm::slpvectorizer::BoUpSLP::removeInstructionsAndOperands(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x414e3af)\r\n #9 0x00000000041aa335 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::SmallDenseSet> const&, llvm::Instruction*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41aa335)\r\n#10 0x00000000041bcd90 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#11 0x00000000041bfb71 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41bfb71)\r\n#12 0x00000000041c431b llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\r\n#13 0x00000000041c8ae8 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41c8ae8)\r\n#14 0x00000000041cf231 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#15 0x00000000041cfedb llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41cfedb)\r\n#16 0x0000000002f64f0e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f64f0e)\r\n#17 0x0000000004ed9788 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed9788)\r\n#18 0x0000000000e2773e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe2773e)\r\n#19 0x0000000004ed81ae llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed81ae)\r\n#20 0x0000000000e26e8e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe26e8e)\r\n#21 0x0000000004ed7be0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed7be0)\r\n#22 0x000000000090a07a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x90a07a)\r\n#23 0x00000000008fce7c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fce7c)\r\n#24 0x000071afd7629d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#25 0x000071afd7629e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#26 0x00000000008f493e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f493e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\nllvm version: 33363521ca24f912cc25530f6cecbca53acce8a3", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "107473", "issue_url": "https://github.com/llvm/llvm-project/issues/107473", "bug_type": "crash", "base_commit": "99fb1506a869fa5e82dbd36e1a63cd21450f1502", "knowledge_cutoff": "2024-09-05T21:51:36Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "e3c537ff903af9a92ff43bab6d21c0ea759d65e5", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[7314, 7322]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["planContainsAdditionalSimplifications"]}}, "patch": "commit e3c537ff903af9a92ff43bab6d21c0ea759d65e5\nAuthor: Florian Hahn \nDate: Tue Sep 10 21:37:12 2024 +0100\n\n [VPlan] Consider non-header phis in planContainsAdditionalSimp.\n \n Update planContainsAdditionalSimplifications to also check phis not in\n the loop header. This ensures we don't miss cases where VPBlendRecipes\n (which correspond to such phis) have been simplified.\n \n Fixes https://github.com/llvm/llvm-project/issues/107473.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 2be3b5775292..b821da03c16e 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -7314,9 +7314,10 @@ static bool planContainsAdditionalSimplifications(VPlan &Plan,\n // Return true if the loop contains any instructions that are not also part of\n // the VPlan or are skipped for VPlan-based cost computations. This indicates\n // that the VPlan contains extra simplifications.\n- return any_of(TheLoop->blocks(), [&SeenInstrs, &CostCtx](BasicBlock *BB) {\n- return any_of(*BB, [&SeenInstrs, &CostCtx](Instruction &I) {\n- if (isa(&I))\n+ return any_of(TheLoop->blocks(), [&SeenInstrs, &CostCtx,\n+ TheLoop](BasicBlock *BB) {\n+ return any_of(*BB, [&SeenInstrs, &CostCtx, TheLoop, BB](Instruction &I) {\n+ if (isa(&I) && BB == TheLoop->getHeader())\n return false;\n return !SeenInstrs.contains(&I) && !CostCtx.skipCostComputation(&I, true);\n });\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll", "commands": ["opt -p loop-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S %s"], "tests": [{"test_name": "dead_load", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @dead_load(ptr %p, i16 %start) {\nentry:\n %start.ext = sext i16 %start to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %start.ext, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr i16, ptr %p, i64 %iv\n store i16 0, ptr %gep, align 2\n %l = load i16, ptr %gep, align 2\n %iv.next = add i64 %iv, 3\n %cmp = icmp slt i64 %iv, 111\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "dead_live_out_due_to_scalar_epilogue_required", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine i8 @dead_live_out_due_to_scalar_epilogue_required(ptr %src, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %idxprom = sext i32 %iv to i64\n %gep.src = getelementptr i8, ptr %src, i64 %idxprom\n %l = load i8, ptr %gep.src, align 1\n %gep.dst = getelementptr i8, ptr %dst, i64 %idxprom\n store i8 0, ptr %gep.dst, align 1\n %iv.next = add i32 %iv, 4\n %cmp = icmp ult i32 %iv, 1001\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n %r = phi i8 [ %l, %loop ]\n ret i8 %r\n}\n"}, {"test_name": "test_phi_in_latch_redundant", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @test_phi_in_latch_redundant(ptr %dst, i32 %a) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 false, label %loop.latch, label %then\n\nthen: ; preds = %loop.header\n %not.a = xor i32 %a, -1\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %p = phi i32 [ %not.a, %then ], [ 0, %loop.header ]\n %gep = getelementptr i32, ptr %dst, i64 %iv\n store i32 %p, ptr %gep, align 4\n %iv.next = add i64 %iv, 9\n %ec = icmp slt i64 %iv, 322\n br i1 %ec, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}, {"test_name": "cost_of_exit_branch_and_cond_insts", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i16 @llvm.umax.i16(i16, i16) #0\n\ndefine i32 @cost_of_exit_branch_and_cond_insts(ptr %a, ptr %b, i1 %c, i16 %x) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %then, label %loop.exiting\n\nthen: ; preds = %loop.header\n %gep = getelementptr inbounds i32, ptr %b, i32 %iv\n store i1 false, ptr %a, align 1\n store i32 0, ptr %gep, align 4\n br label %loop.exiting\n\nloop.exiting: ; preds = %then, %loop.header\n %iv.next = add i32 %iv, 1\n %umax = tail call i16 @llvm.umax.i16(i16 %x, i16 111)\n %umax.ext = zext i16 %umax to i32\n %sub = sub i32 770, %umax.ext\n %ec = icmp slt i32 %iv, %sub\n br i1 %ec, label %loop.latch, label %exit\n\nloop.latch: ; preds = %loop.exiting\n br label %loop.header\n\nexit: ; preds = %loop.exiting\n br label %return\n\nreturn: ; preds = %exit\n ret i32 0\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[VPlan] Assertion \" VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Testing using 0c1500ef05e0a5b25cae79d2bd361dbc6e14e726\r\n\r\nC testcase:\r\n```\r\nint a;\r\nextern int b[][3][3][3][3][3];\r\n_Bool c;\r\nunsigned char d[1][13][13];\r\nunsigned e[1][13][13][13];\r\nint f[6];\r\nunsigned short g[1][13][13][13];\r\nunsigned h[1][13][13], i[1][13][13];\r\nshort j[1][13][13];\r\nunsigned k[3][13][13];\r\nunsigned short l[1][13];\r\nshort m[1][13];\r\nunsigned short n[1][13][13];\r\nunsigned o[1][13];\r\nvoid p(char q, int r, char s, unsigned char z[][13][13],\r\n unsigned aa[][13][13][13], int ab[], unsigned short ac[][13][13][13],\r\n unsigned ad[][13][13], short ae[][13][13], unsigned af[][13][13],\r\n unsigned short ag[][13], short ah[][13], unsigned short ai[][13][13],\r\n unsigned aj[][13][13], unsigned ak[][13]) {\r\n for (unsigned t = 0; t < q; t += 50)\r\n for (unsigned u = 0; u < 2; u = 6)\r\n for (short w = 0; w < 2; w += 42155)\r\n for (signed x = -15; x < r; x += s) {\r\n b[x][x][u][2][x][2] = ab[5] ? ~af[2][1][x] : 0;\r\n a = x ? c ? aj[2][u][x] : 0 : ae[2][w][4];\r\n }\r\n}\r\nint main() {\r\n f[5] = 8;\r\n for (long t = 0; t < 3; ++t)\r\n for (long v = 0; v < 3; ++v)\r\n for (long y = 0; y < 3; ++y)\r\n k[t][v][y] = 3004284279;\r\n p(8, 322, 9, d, e, f, g, h, j, k, l, m, n, i, o);\r\n}\r\n```\r\n\r\nLLVM IR:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main(i64 %indvars.iv) #0 {\r\nentry:\r\n br label %for.body17.us77.us.us.i\r\n\r\nfor.body17.us77.us.us.i: ; preds = %cond.end.us84.us.us.i, %entry\r\n %indvars.iv1 = phi i64 [ %indvars.iv.next, %cond.end.us84.us.us.i ], [ 0, %entry ]\r\n br i1 false, label %cond.end.us84.us.us.i, label %cond.true.us80.us.us.i\r\n\r\ncond.true.us80.us.us.i: ; preds = %for.body17.us77.us.us.i\r\n %not.us83.us.us.i = xor i32 0, 0\r\n br label %cond.end.us84.us.us.i\r\n\r\ncond.end.us84.us.us.i: ; preds = %cond.true.us80.us.us.i, %for.body17.us77.us.us.i\r\n %cond.us85.us.us.i = phi i32 [ %not.us83.us.us.i, %cond.true.us80.us.us.i ], [ 0, %for.body17.us77.us.us.i ]\r\n %arrayidx30.us87.us.us.i = getelementptr [0 x [3 x [3 x [3 x [3 x [3 x i32]]]]]], ptr null, i64 0, i64 %indvars.iv1, i64 %indvars.iv, i64 0, i64 2, i64 %indvars.iv, i64 2\r\n store i32 %cond.us85.us.us.i, ptr %arrayidx30.us87.us.us.i, align 4\r\n %indvars.iv.next = add i64 %indvars.iv1, 9\r\n %cmp14.us95.us.us.i = icmp slt i64 %indvars.iv1, 322\r\n br i1 %cmp14.us95.us.us.i, label %for.body17.us77.us.us.i, label %for.body17.us77.us.us.1.i.preheader\r\n\r\nfor.body17.us77.us.us.1.i.preheader: ; preds = %cond.end.us84.us.us.i\r\n ret i32 0\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7400: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"main\"\r\n #0 0x000059162be7f6e0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x30136e0)\r\n #1 0x000059162be7caef llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x3010aef)\r\n #2 0x000059162be7cc45 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007152dd642520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007152dd6969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007152dd6969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007152dd6969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007152dd642476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007152dd6287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007152dd62871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007152dd639e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x000059162ae52310 llvm::LoopVectorizationPlanner::computeBestVF() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1fe6310)\r\n#12 0x000059162ae6910b llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1ffd10b)\r\n#13 0x000059162ae6c121 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2000121)\r\n#14 0x000059162ae6c797 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2000797)\r\n#15 0x0000591629cc8756 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5c756)\r\n#16 0x000059162bc8b67f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1f67f)\r\n#17 0x0000591629ccb8e6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5f8e6)\r\n#18 0x000059162bc89a1b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1da1b)\r\n#19 0x0000591629cc7b36 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5bb36)\r\n#20 0x000059162bc8a57d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1e57d)\r\n#21 0x00005916294f41b6 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x6881b6)\r\n#22 0x00005916294e5f11 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x679f11)\r\n#23 0x00007152dd629d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#24 0x00007152dd629e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#25 0x00007152dd629e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#26 0x00005916294dc2c5 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x6702c5)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\nhttps://godbolt.org/z/qEEqe1exM\r\n\r\nFound via fuzzer\r\n\r\nRelated/prior issue: https://github.com/llvm/llvm-project/issues/107171\r\n\r\ncc @fhahn", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": [{"author": "patrick-rivos", "body": "Ping. The C fuzzer is hitting this assertion ~15 times a day since the runs starting sep 6th."}]}, "verified": true} {"bug_id": "116228", "issue_url": "https://github.com/llvm/llvm-project/issues/116228", "bug_type": "crash", "base_commit": "fbbea8929ffd2fae57a6013912455d566b884e12", "knowledge_cutoff": "2024-11-14T13:46:57Z", "lit_test_dir": ["llvm/test/Analysis/MemorySSA"], "hints": {"fix_commit": "46e04f7fe5c0f01ba452489bb966bdf2d560a63d", "components": ["MemorySSAUpdater"], "bug_location_lineno": {"llvm/include/llvm/Analysis/MemorySSAUpdater.h": [[260, 282]], "llvm/lib/Analysis/MemorySSAUpdater.cpp": [[565, 588], [593, 602], [615, 621], [668, 675], [692, 700], [716, 728], [741, 747]]}, "bug_location_funcname": {"llvm/lib/Analysis/MemorySSAUpdater.cpp": ["MemorySSAUpdater::cloneUsesAndDefs", "MemorySSAUpdater::privateUpdateExitBlocksForClonedLoop", "MemorySSAUpdater::updateForClonedBlockIntoPred", "MemorySSAUpdater::updateForClonedLoop", "getNewDefiningAccessForClone", "onlySingleValue"]}}, "patch": "commit 46e04f7fe5c0f01ba452489bb966bdf2d560a63d\nAuthor: Nikita Popov \nDate: Mon Dec 9 12:21:49 2024 +0100\n\n [MemorySSA] Handle MemoryDef optimized away during cloning (#117883)\n \n When determining the replacement access during cloning, we currently\n leave accesses for instructions that are not in the VMap alone. This is\n correct if the instruction is not in VMap because it hasn't been cloned,\n but not if it has been cloned and then removed. In that case, we should\n walk up to the defining access, like in other simplification cases.\n \n To distinguish the two cases, pass in a callback that queries whether\n the instruction is part of the cloned region.\n \n An alternative to this would be to delay removal of dead instructions in\n SimpleLoopUnswitch until after MSSA cloning.\n \n Fixes https://github.com/llvm/llvm-project/issues/116228.\n\ndiff --git a/llvm/include/llvm/Analysis/MemorySSAUpdater.h b/llvm/include/llvm/Analysis/MemorySSAUpdater.h\nindex 055feceefb05..b8e08f4b7842 100644\n--- a/llvm/include/llvm/Analysis/MemorySSAUpdater.h\n+++ b/llvm/include/llvm/Analysis/MemorySSAUpdater.h\n@@ -260,23 +260,32 @@ private:\n MemoryAccess *tryRemoveTrivialPhi(MemoryPhi *Phi, RangeType &Operands);\n void tryRemoveTrivialPhis(ArrayRef UpdatedPHIs);\n void fixupDefs(const SmallVectorImpl &);\n- // Clone all uses and defs from BB to NewBB given a 1:1 map of all\n- // instructions and blocks cloned, and a map of MemoryPhi : Definition\n- // (MemoryAccess Phi or Def). VMap maps old instructions to cloned\n- // instructions and old blocks to cloned blocks. MPhiMap, is created in the\n- // caller of this private method, and maps existing MemoryPhis to new\n- // definitions that new MemoryAccesses must point to. These definitions may\n- // not necessarily be MemoryPhis themselves, they may be MemoryDefs. As such,\n- // the map is between MemoryPhis and MemoryAccesses, where the MemoryAccesses\n- // may be MemoryPhis or MemoryDefs and not MemoryUses.\n- // If CloneWasSimplified = true, the clone was exact. Otherwise, assume that\n- // the clone involved simplifications that may have: (1) turned a MemoryUse\n- // into an instruction that MemorySSA has no representation for, or (2) turned\n- // a MemoryDef into a MemoryUse or an instruction that MemorySSA has no\n- // representation for. No other cases are supported.\n+ /// Clone all uses and defs from BB to NewBB given a 1:1 map of all\n+ /// instructions and blocks cloned, and a map of MemoryPhi : Definition\n+ /// (MemoryAccess Phi or Def).\n+ ///\n+ /// \\param VMap Maps old instructions to cloned instructions and old blocks\n+ /// to cloned blocks\n+ /// \\param MPhiMap, is created in the caller of this private method, and maps\n+ /// existing MemoryPhis to new definitions that new MemoryAccesses\n+ /// must point to. These definitions may not necessarily be MemoryPhis\n+ /// themselves, they may be MemoryDefs. As such, the map is between\n+ /// MemoryPhis and MemoryAccesses, where the MemoryAccesses may be\n+ /// MemoryPhis or MemoryDefs and not MemoryUses.\n+ /// \\param IsInClonedRegion Determines whether a basic block was cloned.\n+ /// References to accesses outside the cloned region will not be\n+ /// remapped.\n+ /// \\param CloneWasSimplified If false, the clone was exact. Otherwise,\n+ /// assume that the clone involved simplifications that may have:\n+ /// (1) turned a MemoryUse into an instruction that MemorySSA has no\n+ /// representation for, or (2) turned a MemoryDef into a MemoryUse or\n+ /// an instruction that MemorySSA has no representation for. No other\n+ /// cases are supported.\n void cloneUsesAndDefs(BasicBlock *BB, BasicBlock *NewBB,\n const ValueToValueMapTy &VMap, PhiToDefMap &MPhiMap,\n+ function_ref IsInClonedRegion,\n bool CloneWasSimplified = false);\n+\n template \n void privateUpdateExitBlocksForClonedLoop(ArrayRef ExitBlocks,\n Iter ValuesBegin, Iter ValuesEnd,\ndiff --git a/llvm/lib/Analysis/MemorySSAUpdater.cpp b/llvm/lib/Analysis/MemorySSAUpdater.cpp\nindex f672bd0e1e13..050b827388d3 100644\n--- a/llvm/lib/Analysis/MemorySSAUpdater.cpp\n+++ b/llvm/lib/Analysis/MemorySSAUpdater.cpp\n@@ -565,24 +565,26 @@ static MemoryAccess *onlySingleValue(MemoryPhi *MP) {\n return MA;\n }\n \n-static MemoryAccess *getNewDefiningAccessForClone(MemoryAccess *MA,\n- const ValueToValueMapTy &VMap,\n- PhiToDefMap &MPhiMap,\n- MemorySSA *MSSA) {\n+static MemoryAccess *getNewDefiningAccessForClone(\n+ MemoryAccess *MA, const ValueToValueMapTy &VMap, PhiToDefMap &MPhiMap,\n+ MemorySSA *MSSA, function_ref IsInClonedRegion) {\n MemoryAccess *InsnDefining = MA;\n if (MemoryDef *DefMUD = dyn_cast(InsnDefining)) {\n- if (!MSSA->isLiveOnEntryDef(DefMUD)) {\n- Instruction *DefMUDI = DefMUD->getMemoryInst();\n- assert(DefMUDI && \"Found MemoryUseOrDef with no Instruction.\");\n- if (Instruction *NewDefMUDI =\n- cast_or_null(VMap.lookup(DefMUDI))) {\n- InsnDefining = MSSA->getMemoryAccess(NewDefMUDI);\n- if (!InsnDefining || isa(InsnDefining)) {\n- // The clone was simplified, it's no longer a MemoryDef, look up.\n- InsnDefining = getNewDefiningAccessForClone(\n- DefMUD->getDefiningAccess(), VMap, MPhiMap, MSSA);\n- }\n- }\n+ if (MSSA->isLiveOnEntryDef(DefMUD))\n+ return DefMUD;\n+\n+ // If the MemoryDef is not part of the cloned region, leave it alone.\n+ Instruction *DefMUDI = DefMUD->getMemoryInst();\n+ assert(DefMUDI && \"Found MemoryUseOrDef with no Instruction.\");\n+ if (!IsInClonedRegion(DefMUDI->getParent()))\n+ return DefMUD;\n+\n+ auto *NewDefMUDI = cast_or_null(VMap.lookup(DefMUDI));\n+ InsnDefining = NewDefMUDI ? MSSA->getMemoryAccess(NewDefMUDI) : nullptr;\n+ if (!InsnDefining || isa(InsnDefining)) {\n+ // The clone was simplified, it's no longer a MemoryDef, look up.\n+ InsnDefining = getNewDefiningAccessForClone(\n+ DefMUD->getDefiningAccess(), VMap, MPhiMap, MSSA, IsInClonedRegion);\n }\n } else {\n MemoryPhi *DefPhi = cast(InsnDefining);\n@@ -593,10 +595,10 @@ static MemoryAccess *getNewDefiningAccessForClone(MemoryAccess *MA,\n return InsnDefining;\n }\n \n-void MemorySSAUpdater::cloneUsesAndDefs(BasicBlock *BB, BasicBlock *NewBB,\n- const ValueToValueMapTy &VMap,\n- PhiToDefMap &MPhiMap,\n- bool CloneWasSimplified) {\n+void MemorySSAUpdater::cloneUsesAndDefs(\n+ BasicBlock *BB, BasicBlock *NewBB, const ValueToValueMapTy &VMap,\n+ PhiToDefMap &MPhiMap, function_ref IsInClonedRegion,\n+ bool CloneWasSimplified) {\n const MemorySSA::AccessList *Acc = MSSA->getBlockAccesses(BB);\n if (!Acc)\n return;\n@@ -615,7 +617,7 @@ void MemorySSAUpdater::cloneUsesAndDefs(BasicBlock *BB, BasicBlock *NewBB,\n MemoryAccess *NewUseOrDef = MSSA->createDefinedAccess(\n NewInsn,\n getNewDefiningAccessForClone(MUD->getDefiningAccess(), VMap,\n- MPhiMap, MSSA),\n+ MPhiMap, MSSA, IsInClonedRegion),\n /*Template=*/CloneWasSimplified ? nullptr : MUD,\n /*CreationMustSucceed=*/false);\n if (NewUseOrDef)\n@@ -668,8 +670,13 @@ void MemorySSAUpdater::updateForClonedLoop(const LoopBlocksRPO &LoopBlocks,\n ArrayRef ExitBlocks,\n const ValueToValueMapTy &VMap,\n bool IgnoreIncomingWithNoClones) {\n- PhiToDefMap MPhiMap;\n+ SmallSetVector Blocks;\n+ for (BasicBlock *BB : concat(LoopBlocks, ExitBlocks))\n+ Blocks.insert(BB);\n \n+ auto IsInClonedRegion = [&](BasicBlock *BB) { return Blocks.contains(BB); };\n+\n+ PhiToDefMap MPhiMap;\n auto FixPhiIncomingValues = [&](MemoryPhi *Phi, MemoryPhi *NewPhi) {\n assert(Phi && NewPhi && \"Invalid Phi nodes.\");\n BasicBlock *NewPhiBB = NewPhi->getBlock();\n@@ -692,9 +699,10 @@ void MemorySSAUpdater::updateForClonedLoop(const LoopBlocksRPO &LoopBlocks,\n continue;\n \n // Determine incoming value and add it as incoming from IncBB.\n- NewPhi->addIncoming(\n- getNewDefiningAccessForClone(IncomingAccess, VMap, MPhiMap, MSSA),\n- IncBB);\n+ NewPhi->addIncoming(getNewDefiningAccessForClone(IncomingAccess, VMap,\n+ MPhiMap, MSSA,\n+ IsInClonedRegion),\n+ IncBB);\n }\n if (auto *SingleAccess = onlySingleValue(NewPhi)) {\n MPhiMap[Phi] = SingleAccess;\n@@ -716,13 +724,13 @@ void MemorySSAUpdater::updateForClonedLoop(const LoopBlocksRPO &LoopBlocks,\n MPhiMap[MPhi] = NewPhi;\n }\n // Update Uses and Defs.\n- cloneUsesAndDefs(BB, NewBlock, VMap, MPhiMap);\n+ cloneUsesAndDefs(BB, NewBlock, VMap, MPhiMap, IsInClonedRegion);\n };\n \n- for (auto *BB : llvm::concat(LoopBlocks, ExitBlocks))\n+ for (auto *BB : Blocks)\n ProcessBlock(BB);\n \n- for (auto *BB : llvm::concat(LoopBlocks, ExitBlocks))\n+ for (auto *BB : Blocks)\n if (MemoryPhi *MPhi = MSSA->getMemoryAccess(BB))\n if (MemoryAccess *NewPhi = MPhiMap.lookup(MPhi))\n FixPhiIncomingValues(MPhi, cast(NewPhi));\n@@ -741,7 +749,9 @@ void MemorySSAUpdater::updateForClonedBlockIntoPred(\n PhiToDefMap MPhiMap;\n if (MemoryPhi *MPhi = MSSA->getMemoryAccess(BB))\n MPhiMap[MPhi] = MPhi->getIncomingValueForBlock(P1);\n- cloneUsesAndDefs(BB, P1, VM, MPhiMap, /*CloneWasSimplified=*/true);\n+ cloneUsesAndDefs(\n+ BB, P1, VM, MPhiMap, [&](BasicBlock *CheckBB) { return BB == CheckBB; },\n+ /*CloneWasSimplified=*/true);\n }\n \n template \n", "tests": [{"file": "llvm/test/Analysis/MemorySSA/loop-rotate-update.ll", "commands": ["opt -disable-output -passes=\"loop-mssa(loop-rotate),print\" -verify-memoryssa < %s 2>&1"], "tests": [{"test_name": "", "test_body": "\n\n\n\ndefine void @test(ptr %p) {\nentry:\n br label %loop\n\nloop:\n store ptr null, ptr %p\n %val1 = load ptr, ptr %p\n %cmp = icmp eq ptr %val1, null\n br i1 %cmp, label %exit, label %loop.latch\n\nloop.latch:\n %val2 = load ptr, ptr %p\n br label %loop\n\nexit:\n ret void\n}"}]}, {"file": "llvm/test/Analysis/MemorySSA/pr116227.ll", "commands": ["opt -disable-output -passes=\"loop-mssa(simple-loop-unswitch),print\" -verify-memoryssa < %s 2>&1"], "tests": [{"test_name": "", "test_body": "\ndeclare ptr @malloc() allockind(\"alloc,uninitialized\")\n\n\n\n\ndefine void @test(i1 %arg) {\nentry:\n br label %for.body\n\nfor.body:\n %call.i = call ptr @malloc()\n %cmp.i = icmp ne ptr %call.i, null\n %or.cond.i = select i1 %cmp.i, i1 %arg, i1 false\n br i1 %or.cond.i, label %exit, label %for.body\n\nexit:\n ret void\n}\n\n\n\n\n\ndefine void @test_extra_defs(ptr %p, i1 %arg) {\nentry:\n store i8 1, ptr %p\n br label %for.body\n\nfor.body:\n store i8 2, ptr %p\n %call.i = call ptr @malloc()\n %cmp.i = icmp ne ptr %call.i, null\n %or.cond.i = select i1 %cmp.i, i1 %arg, i1 false\n br i1 %or.cond.i, label %exit, label %for.body\n\nexit:\n ret void\n}"}]}], "issue": {"title": "[clang] Crash at -O3: Assertion `MA->use_empty() && \"Trying to remove memory access that still has uses\"' failed", "body": "When I compiled the code with -O3, it crashed:\r\n\r\nBisected to https://github.com/llvm/llvm-project/commit/5a3e3675f6be3c49fb1a305202888cce4d916ba6, which was committed by @durin42 \r\n\r\nCompiler explorer: https://godbolt.org/z/3Wc1jdx6d\r\n\r\n```c\r\nstruct a {\r\n struct a *b\r\n};\r\nint c, g, i;\r\nvoid *malloc();\r\nint d() {\r\n struct a *e = malloc(c);\r\n if (e == 0)\r\n return 1;\r\n for (int f = 0; f < c; f++)\r\n e[f].b = 1 ? &e[1] : 0;\r\n}\r\nint h() {\r\n for (; g; g++)\r\n ;\r\n return 0;\r\n}\r\nvoid j() {\r\n for (;;)\r\n for (; h() + i; i++)\r\n d();\r\n}\r\nint main() {}\r\n```\r\nCrash is: \r\n\r\n```\r\nclang: /root/llvm-project/llvm/lib/Analysis/MemorySSA.cpp:1840: void llvm::MemorySSA::removeFromLookups(llvm::MemoryAccess*): Assertion `MA->use_empty() && \"Trying to remove memory access that still has uses\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/clang -gdwarf-4 -g -o /app/output.s -mllvm --x86-asm-syntax=intel -fno-verbose-asm -S --gcc-toolchain=/opt/compiler-explorer/gcc-snapshot -fcolor-diagnostics -fno-crash-diagnostics -O3 -Wall -Wextra \r\n1.\t parser at end of file\r\n2.\tOptimizer\r\n3.\tRunning pass \"require,function(invalidate),require,cgscc(devirt<4>(inline,function-attrs,argpromotion,openmp-opt-cgscc,function(sroa,early-cse,speculative-execution,jump-threading,correlated-propagation,simplifycfg,instcombine,aggressive-instcombine,libcalls-shrinkwrap,tailcallelim,simplifycfg,reassociate,constraint-elimination,loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-unswitch),simplifycfg,instcombine,loop(loop-idiom,indvars,simple-loop-unswitch,loop-deletion,loop-unroll-full),sroa,vector-combine,mldst-motion,gvn<>,sccp,bdce,instcombine,jump-threading,correlated-propagation,adce,memcpyopt,dse,move-auto-init,loop-mssa(licm),coro-elide,simplifycfg,instcombine),function-attrs,function(require),coro-split,coro-annotation-elide)),function(invalidate),cgscc(devirt<4>())\" on module \"\"\r\n4.\tRunning pass \"cgscc(devirt<4>(inline,function-attrs,argpromotion,openmp-opt-cgscc,function(sroa,early-cse,speculative-execution,jump-threading,correlated-propagation,simplifycfg,instcombine,aggressive-instcombine,libcalls-shrinkwrap,tailcallelim,simplifycfg,reassociate,constraint-elimination,loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-unswitch),simplifycfg,instcombine,loop(loop-idiom,indvars,simple-loop-unswitch,loop-deletion,loop-unroll-full),sroa,vector-combine,mldst-motion,gvn<>,sccp,bdce,instcombine,jump-threading,correlated-propagation,adce,memcpyopt,dse,move-auto-init,loop-mssa(licm),coro-elide,simplifycfg,instcombine),function-attrs,function(require),coro-split,coro-annotation-elide))\" on module \"\"\r\n5.\tRunning pass \"loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-unswitch)\" on function \"j\"\r\n #0 0x0000000003bfb838 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3bfb838)\r\n #1 0x0000000003bf953c llvm::sys::CleanupOnSignal(unsigned long) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3bf953c)\r\n #2 0x0000000003b46b38 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\r\n #3 0x00007afb46242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007afb462969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #5 0x00007afb46242476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #6 0x00007afb462287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #7 0x00007afb4622871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #8 0x00007afb46239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #9 0x0000000002bead0a llvm::MemorySSA::removeFromLookups(llvm::MemoryAccess*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2bead0a)\r\n#10 0x0000000002c03a6c llvm::MemorySSAUpdater::removeBlocks(llvm::SmallSetVector const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2c03a6c)\r\n#11 0x00000000084e163a deleteDeadBlocksFromLoop(llvm::Loop&, llvm::SmallVectorImpl&, llvm::DominatorTree&, llvm::LoopInfo&, llvm::MemorySSAUpdater*, llvm::ScalarEvolution*, llvm::LPMUpdater&) SimpleLoopUnswitch.cpp:0:0\r\n#12 0x00000000084ea50d unswitchNontrivialInvariants(llvm::Loop&, llvm::Instruction&, llvm::ArrayRef, llvm::IVConditionInfo&, llvm::DominatorTree&, llvm::LoopInfo&, llvm::AssumptionCache&, llvm::ScalarEvolution*, llvm::MemorySSAUpdater*, llvm::LPMUpdater&, bool, bool) SimpleLoopUnswitch.cpp:0:0\r\n#13 0x00000000084ec9a8 unswitchBestCondition(llvm::Loop&, llvm::DominatorTree&, llvm::LoopInfo&, llvm::AssumptionCache&, llvm::AAResults&, llvm::TargetTransformInfo&, llvm::ScalarEvolution*, llvm::MemorySSAUpdater*, llvm::LPMUpdater&) SimpleLoopUnswitch.cpp:0:0\r\n#14 0x00000000084f0679 llvm::SimpleLoopUnswitchPass::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x84f0679)\r\n#15 0x000000000525730e llvm::detail::PassModel, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x525730e)\r\n#16 0x0000000003a12986 std::optional llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::runSinglePass, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>, std::default_delete, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>>>>(llvm::Loop&, std::unique_ptr, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>, std::default_delete, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>>>&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&, llvm::PassInstrumentation&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3a12986)\r\n#17 0x0000000003a12b4e llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::runWithoutLoopNestPasses(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3a12b4e)\r\n#18 0x0000000003a13f04 llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3a13f04)\r\n#19 0x0000000003a14bfc llvm::FunctionToLoopPassAdaptor::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3a14bfc)\r\n#20 0x000000000112567e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x112567e)\r\n#21 0x00000000035b06d8 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35b06d8)\r\n#22 0x0000000001125c2e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x1125c2e)\r\n#23 0x0000000002ab3c12 llvm::CGSCCToFunctionPassAdaptor::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2ab3c12)\r\n#24 0x000000000112607e llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x112607e)\r\n#25 0x0000000002aaad52 llvm::PassManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2aaad52)\r\n#26 0x00000000052560ce llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>, llvm::AnalysisManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x52560ce)\r\n#27 0x0000000002ab0666 llvm::DevirtSCCRepeatedPass::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2ab0666)\r\n#28 0x00000000052560ee llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x52560ee)\r\n#29 0x0000000002aae3f6 llvm::ModuleToPostOrderCGSCCPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2aae3f6)\r\n#30 0x00000000052560ae llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x52560ae)\r\n#31 0x00000000035aeb30 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35aeb30)\r\n#32 0x0000000005476f75 llvm::ModuleInlinerWrapperPass::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x5476f75)\r\n#33 0x00000000050f42ce llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x50f42ce)\r\n#34 0x00000000035aeb30 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35aeb30)\r\n#35 0x0000000003ea671b (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#36 0x0000000003ea9f45 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3ea9f45)\r\n#37 0x0000000004575cfe clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4575cfe)\r\n#38 0x000000000673552c clang::ParseAST(clang::Sema&, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x673552c)\r\n#39 0x00000000045760e8 clang::CodeGenAction::ExecuteAction() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x45760e8)\r\n#40 0x00000000048312c9 clang::FrontendAction::Execute() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x48312c9)\r\n#41 0x00000000047b04be clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x47b04be)\r\n#42 0x000000000491c2ce clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x491c2ce)\r\n#43 0x0000000000ceb57f cc1_main(llvm::ArrayRef, char const*, void*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xceb57f)\r\n#44 0x0000000000ce323a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#45 0x00000000045b92e9 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\r\n#46 0x0000000003b46fe4 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3b46fe4)\r\n#47 0x00000000045b98df clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\r\n#48 0x000000000457fbcd clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x457fbcd)\r\n#49 0x0000000004580cbd clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4580cbd)\r\n#50 0x0000000004588075 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4588075)\r\n#51 0x0000000000ce8419 clang_main(int, char**, llvm::ToolContext const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xce8419)\r\n#52 0x0000000000bb4c14 main (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xbb4c14)\r\n#53 0x00007afb46229d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#54 0x00007afb46229e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#55 0x0000000000ce2cee _start (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xce2cee)\r\n```", "author": "cardigan1008", "labels": ["crash-on-valid", "llvm:analysis"], "comments": [{"author": "dtcxzyw", "body": "Reduced reproducer: https://godbolt.org/z/bK3z9nGsa\r\n```\r\n; bin/opt -passes=\"loop-mssa(licm,simple-loop-unswitch)\" reduced.ll -S\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\n; Function Attrs: allockind(\"alloc,uninitialized\")\r\ndeclare ptr @malloc() #0\r\n\r\ndefine void @j(i1 %cmp2.i) {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %for.cond1.for.cond_crit_edge, %entry\r\n br label %for.cond1\r\n\r\nfor.cond1: ; preds = %for.body.i, %for.body, %for.cond\r\n br i1 false, label %for.cond1.for.cond_crit_edge, label %for.body\r\n\r\nfor.cond1.for.cond_crit_edge: ; preds = %for.cond1\r\n br label %for.cond\r\n\r\nfor.body: ; preds = %for.cond1\r\n %call.i = call ptr @malloc()\r\n %cmp.i = icmp ne ptr %call.i, null\r\n %or.cond.i = select i1 %cmp.i, i1 %cmp2.i, i1 false\r\n br i1 %or.cond.i, label %for.body.i, label %for.cond1\r\n\r\nfor.body.i: ; preds = %for.body\r\n br label %for.cond1\r\n}\r\n\r\nattributes #0 = { allockind(\"alloc,uninitialized\") }\r\n```"}, {"author": "nikic", "body": "Can reduce this to:\n\n```llvm\n; RUN: opt -S -passes=\"loop-mssa(simple-loop-unswitch)\" -verify-memoryssa < %s\ndeclare ptr @malloc() allockind(\"alloc,uninitialized\")\n\ndefine void @test(i1 %arg) {\nentry:\n br label %for.body\n\nfor.body:\n %call.i = call ptr @malloc()\n %cmp.i = icmp ne ptr %call.i, null\n %or.cond.i = select i1 %cmp.i, i1 %arg, i1 false\n br i1 %or.cond.i, label %exit, label %for.body\n\nexit:\n ret void\n}\n```\n\n> opt: /home/npopov/repos/llvm-project/llvm/lib/Analysis/MemorySSA.cpp:2068: void llvm::MemorySSA::verifyOrderingDominationAndDefUses(IterT, VerificationLevel) const [IterT = llvm::iterator_range, false, false>>]: Assertion `dominates(MD, U) && \"Memory Def does not dominate it's uses\"' failed.\n"}, {"author": "nikic", "body": "We go from:\n```llvm\nMemorySSA for function: test\ndefine void @test(i1 %arg) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n; 2 = MemoryPhi({entry,liveOnEntry},{for.body,1})\n; 1 = MemoryDef(2)\n %call.i = call ptr @malloc()\n %cmp.i = icmp ne ptr %call.i, null\n %or.cond.i = select i1 %cmp.i, i1 %arg, i1 false\n br i1 %or.cond.i, label %exit, label %for.body\n\nexit: ; preds = %for.body\n ret void\n}\n```\nTo:\n```llvm\nMemorySSA for function: test\ndefine void @test(i1 %arg) {\nentry:\n %arg.fr = freeze i1 %arg\n br i1 %arg.fr, label %entry.split, label %entry.split.us\n\nentry.split.us: ; preds = %entry\n br label %for.body.us\n\nfor.body.us: ; preds = %for.body.us, %entry.split.us\n; 3 = MemoryPhi({entry.split.us,liveOnEntry},{for.body.us,1})\n br label %for.body.us\n\nentry.split: ; preds = %entry\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry.split\n; 2 = MemoryPhi({entry.split,liveOnEntry},{for.body,1})\n; 1 = MemoryDef(2)\n %call.i = call ptr @malloc()\n %cmp.i = icmp ne ptr %call.i, null\n %or.cond.i = select i1 %cmp.i, i1 true, i1 false\n br i1 %or.cond.i, label %exit, label %for.body\n\nexit: ; preds = %for.body\n ret void\n}\n```\n\nMemoryPhi 3 is broken.\n\nI think this is because we first delete the dead cloned instructions and then update MSSA and something gets confused in that case..."}]}, "verified": true} {"bug_id": "102994", "issue_url": "https://github.com/llvm/llvm-project/issues/102994", "bug_type": "hang", "base_commit": "15e915a44f0d0bf092214586d3ec86e2bb7636d7", "knowledge_cutoff": "2024-08-13T02:57:21Z", "lit_test_dir": ["llvm/test/Transforms/MemCpyOpt"], "hints": {"fix_commit": "378daa6c6fd10d3704be449f2fe9c55df522a6e9", "components": ["MemCpyOptimizer"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp": [[1193, 1198]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp": ["MemCpyOptPass::processMemCpyMemCpyDependence"]}}, "patch": "commit 378daa6c6fd10d3704be449f2fe9c55df522a6e9\nAuthor: Yingwei Zheng \nDate: Thu Aug 22 17:20:47 2024 +0800\n\n [MemCpyOpt] Avoid infinite loops in `MemCpyOptPass::processMemCpyMemCpyDependence` (#103218)\n \n Closes https://github.com/llvm/llvm-project/issues/102994.\n\ndiff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\nindex 1d779128e454..3f15fa2163d2 100644\n--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\n+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\n@@ -1193,6 +1193,10 @@ bool MemCpyOptPass::processMemCpyMemCpyDependence(MemCpyInst *M,\n CopySourceAlign = commonAlignment(*CopySourceAlign, MForwardOffset);\n }\n \n+ // Avoid infinite loops\n+ if (BAA.isMustAlias(M->getSource(), CopySource))\n+ return false;\n+\n // Verify that the copied-from memory doesn't change in between the two\n // transfers. For example, in:\n // memcpy(a <- b)\n", "tests": [{"file": "llvm/test/Transforms/MemCpyOpt/pr102994.ll", "commands": ["opt -S -passes=memcpyopt < %s"], "tests": [{"test_name": "func2", "test_body": "@g3 = external global i8, align 1\n\ndefine void @func2(ptr %p) {\nentry:\n %gep1 = getelementptr i8, ptr %p, i64 32\n %gep2 = getelementptr i8, ptr %p, i64 34\n %gep3 = getelementptr i8, ptr %p, i64 32\n call void @llvm.memcpy.p0.p0.i64(ptr %gep1, ptr %gep3, i64 32, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr @g3, ptr %gep2, i64 1, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}, {"test_name": "func", "test_body": "@g1 = external global i8\n@g2 = external global [64 x i8]\n\ndefine void @func() {\nentry:\n call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g2, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g2, i64 16), i64 20, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr @g1, ptr getelementptr inbounds (i8, ptr @g2, i64 24), i64 1, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}]}], "issue": {"title": "[MemcpyOpt] Infinite loops when forwarding memcpy on globals", "body": "Reproducer: https://godbolt.org/z/GKzGEKE18\r\n```\r\n; bin/opt -passes=memcpyopt reduced.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\n%struct.S1 = type { i64, i16, i32, %struct.S0, i8, i32, i8, i32, i8 }\r\n%struct.S0 = type { i32, i32, i32, i32, i16 }\r\n\r\n@g = external global i8\r\n@g_265 = external global %struct.S1\r\n\r\ndefine void @func() {\r\nentry:\r\n call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n call void @llvm.memcpy.p0.p0.i64(ptr @g, ptr getelementptr inbounds (i8, ptr @g_265, i64 24), i64 1, i1 false)\r\n ret void\r\n}\r\n```\r\n```\r\nFound roots: %entry \r\nClobber starting at access 1 = MemoryDef(liveOnEntry)\r\n for instruction call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n is 0 = MemoryDef(liveOnEntry)\r\nClobber starting at access 1 = MemoryDef(liveOnEntry)\r\n for instruction call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n is 1 = MemoryDef(liveOnEntry)\r\nClobber starting at access 1 = MemoryDef(liveOnEntry)\r\n for instruction call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n is 1 = MemoryDef(liveOnEntry)\r\nMemCpyOptPass: Forwarding memcpy->memcpy src:\r\n call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n call void @llvm.memcpy.p0.p0.i64(ptr @g, ptr getelementptr inbounds (i8, ptr @g_265, i64 24), i64 1, i1 false)\r\nClobber starting at access 1 = MemoryDef(liveOnEntry)\r\n for instruction call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n is 0 = MemoryDef(liveOnEntry)\r\nClobber starting at access 1 = MemoryDef(liveOnEntry)\r\n for instruction call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n is 1 = MemoryDef(liveOnEntry)\r\nClobber starting at access 1 = MemoryDef(liveOnEntry)\r\n for instruction call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n is 1 = MemoryDef(liveOnEntry)\r\nMemCpyOptPass: Forwarding memcpy->memcpy src:\r\n call void @llvm.memcpy.p0.p0.i64(ptr getelementptr inbounds (i8, ptr @g_265, i64 16), ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 20, i1 false)\r\n call void @llvm.memcpy.p0.p0.i64(ptr @g, ptr getelementptr inbounds (i8, ptr getelementptr inbounds nuw (i8, ptr @g_265, i64 16), i64 8), i64 1, i1 false)\r\n...\r\n```", "author": "dtcxzyw", "labels": ["llvm:hang", "llvm:transforms"], "comments": []}, "verified": true} {"bug_id": "108004", "issue_url": "https://github.com/llvm/llvm-project/issues/108004", "bug_type": "crash", "base_commit": "512cecad4c384c84b79fea050a755cb7e46c6ac5", "knowledge_cutoff": "2024-09-10T11:32:26Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "7858e14547c509c95503b74ff8ffc7bf2fc5b110", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[8663, 8668]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["collectUsersInExitBlock"]}}, "patch": "commit 7858e14547c509c95503b74ff8ffc7bf2fc5b110\nAuthor: Hari Limaye \nDate: Wed Sep 11 16:43:34 2024 +0100\n\n [LV] Amend check for IV increments in collectUsersInEntryBlock (#108020)\n \n The check for IV increments in collectUsersInEntryBlock currently\n triggers for exit-block PHIs which use the IV start value, resulting in\n us failing to add the input value for the middle block to these PHIs.\n \n Fix this by amending the check for IV increments to only include\n incoming values that are instructions inside the loop.\n \n Fixes #108004\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex b821da03c16e..640a7bf3d672 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -8663,6 +8663,7 @@ static MapVector collectUsersInExitBlock(\n !cast(V)->getTruncInst()) ||\n isa(V) ||\n (isa(IncomingValue) &&\n+ OrigLoop->contains(cast(IncomingValue)) &&\n any_of(IncomingValue->users(), [&Inductions](User *U) {\n auto *P = dyn_cast(U);\n return P && Inductions.contains(P);\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/use-iv-start-value.ll", "commands": ["opt < %s -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S"], "tests": [{"test_name": "foo", "test_body": "define i64 @foo(ptr %p1, ptr %p2, i64 %start, i64 %end) {\nentry:\n %start2 = and i64 %start, 12345\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %ind = phi i64 [ %start2, %entry ], [ %ind.next, %for.body ]\n %arrayidx1 = getelementptr inbounds i32, ptr %p1, i64 %ind\n %0 = load i32, ptr %arrayidx1, align 4\n %arrayidx2 = getelementptr inbounds i32, ptr %p2, i64 %ind\n %1 = load i32, ptr %arrayidx2, align 4\n %ind.next = add i64 %ind, 1\n %cmp = icmp ne i64 %ind.next, %end\n br i1 %cmp, label %for.body, label %exit\n\nexit: ; preds = %for.body\n %use = phi i64 [ %start2, %for.body ]\n ret i64 %use\n}\n"}]}], "issue": {"title": "[LoopVectorize] Assertion fails when IV start value is used by an exit block PHI", "body": "With the following IR:\r\n```\r\ndefine i64 @foo(i64 %start, i64 %end) {\r\nentry:\r\n %p1 = alloca [1024 x i32]\r\n %p2 = alloca [1024 x i32]\r\n call void @init_mem(ptr %p1, i64 1024)\r\n call void @init_mem(ptr %p2, i64 1024)\r\n %start2 = and i64 %start, 12345\r\n br label %for.body\r\n\r\nfor.body:\r\n %ind = phi i64 [ %ind.next, %for.body ], [ %start2, %entry ]\r\n %arrayidx1 = getelementptr inbounds i32, ptr %p1, i64 %ind\r\n %0 = load i32, ptr %arrayidx1, align 4\r\n %arrayidx2 = getelementptr inbounds i32, ptr %p2, i64 %ind\r\n %1 = load i32, ptr %arrayidx2, align 4\r\n %ind.next = add i64 %ind, 1\r\n %cmp = icmp ne i64 %ind.next, %end\r\n br i1 %cmp, label %for.body, label %exit\r\n\r\nexit:\r\n %use = phi i64 [ %start2, %for.body ]\r\n ret i64 %use\r\n}\r\n\r\ndeclare void @init_mem(ptr, i64)\r\n```\r\n\r\nLV crashes due to hitting the following assertion:\r\n```\r\nLV: Vectorizing: innermost loop.\r\nPHINode should have one entry for each predecessor of its parent basic block!\r\n %bob = phi i64 [ %start2, %for.body ]\r\nopt: /work/llvm-project-wt/nuw-clang-arr/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10244: bool llvm::LoopVectorizePass::processLoop(Loop *): Assertion `!verif\r\nyFunction(*L->getHeader()->getParent(), &dbgs())' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: ./build/bin/opt -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=4 -debug test.ll -o - -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"test.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"foo\"\r\n #0 0x0000aaaaaec431c0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (./build/bin/opt+0x41a31c0)\r\n #1 0x0000aaaaaec4102c llvm::sys::RunSignalHandlers() (./build/bin/opt+0x41a102c)\r\n #2 0x0000aaaaaec438dc SignalHandler(int) Signals.cpp:0:0\r\n #3 0x0000fffff7ffb9d0 (linux-vdso.so.1+0x9d0)\r\n #4 0x0000fffff7b5f200 __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x0000fffff7b1a67c gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #6 0x0000fffff7b07130 abort ./stdlib/abort.c:81:7\r\n #7 0x0000fffff7b13fd0 __assert_fail_base ./assert/assert.c:89:7\r\n #8 0x0000fffff7b14040 __assert_perror_fail ./assert/assert-perr.c:31:1\r\n #9 0x0000aaaaadee375c llvm::LoopVectorizePass::processLoop(llvm::Loop*) (./build/bin/opt+0x344375c)\r\n#10 0x0000aaaaadee72d0 llvm::LoopVectorizePass::runImpl(llvm::Function&) (./build/bin/opt+0x34472d0)\r\n#11 0x0000aaaaadee7ac0 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (./build/bin/opt+0x3447ac0)\r\n#12 0x0000aaaaaea9676c llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (./bui\r\nld/bin/opt+0x3ff676c)\r\n#13 0x0000aaaaaea9a1fc llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (./build/bin/opt+0x3ffa1fc)\r\n#14 0x0000aaaaaea958cc llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (./build/bin/o\r\npt+0x3ff58cc)\r\n#15 0x0000aaaaab5b3d88 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOu\r\ntputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::Outpu\r\ntKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (./build/bin/opt+0xb13d88)\r\n#16 0x0000aaaaab5a7fc0 optMain (./build/bin/opt+0xb07fc0)\r\n#17 0x0000fffff7b073fc __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:74:3\r\n#18 0x0000fffff7b074cc call_init ./csu/../csu/libc-start.c:128:20\r\n#19 0x0000fffff7b074cc __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#20 0x0000aaaaab5a1ef0 _start (./build/bin/opt+0xb01ef0)\r\n```\r\n\r\nSeems like we fail to properly handle fixing up the incoming value for this PHI from the middle block. ", "author": "hazzlim", "labels": ["vectorizers", "crash"], "comments": [{"author": "hazzlim", "body": "Note: I am actively looking into this and plan to put up a fix "}]}, "verified": true} {"bug_id": "113451", "issue_url": "https://github.com/llvm/llvm-project/issues/113451", "bug_type": "crash", "base_commit": "7a1036935f8122d9120df5ecf5afb5c927a9ea8b", "knowledge_cutoff": "2024-10-23T12:43:54Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "a3508e0246741851f06cb9ba86203be8d89febed", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[11588, 11594]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::isTreeTinyAndNotFullyVectorizable"]}}, "patch": "commit a3508e0246741851f06cb9ba86203be8d89febed\nAuthor: Alexey Bataev \nDate: Wed Oct 23 10:25:41 2024 -0700\n\n [SLP]Small buidlvector only graph should contains scalars from same block\n \n If the graph is small and has single buildvector node, all scalars\n instructions must be from the same basic block to prevent compiler\n crash.\n \n Fixes #113451\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 756b25ac9856..d824c40c401d 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -11588,7 +11588,8 @@ bool BoUpSLP::isTreeTinyAndNotFullyVectorizable(bool ForReduction) const {\n \n if (VectorizableTree.back()->isGather() &&\n VectorizableTree.back()->isAltShuffle() &&\n- VectorizableTree.back()->getVectorFactor() > 2)\n+ VectorizableTree.back()->getVectorFactor() > 2 &&\n+ allSameBlock(VectorizableTree.back()->Scalars))\n return false;\n \n assert(VectorizableTree.empty()\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/small-graph-diff-block-instructions.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512dq -slp-threshold=-99999 < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test(i32 %arg, i32 %arg1) {\nbb:\n %zext = zext i32 %arg1 to i64\n %zext2 = zext i32 %arg to i64\n %sext = sext i32 %arg to i64\n br label %bb3\n\nbb3: ; preds = %bb\n %getelementptr = getelementptr i64, ptr addrspace(1) null, i64 %zext\n %getelementptr4 = getelementptr i64, ptr addrspace(1) null, i64 %zext2\n %getelementptr5 = getelementptr i64, ptr addrspace(1) null, i64 %sext\n %zext6 = zext i32 0 to i64\n %getelementptr7 = getelementptr i64, ptr addrspace(1) null, i64 %zext6\n ret i32 0\n}\n"}]}], "issue": {"title": "Assertion `((GatheredLoadsEntriesFirst.has_value() && E->getOpcode() == Instruction::Load && E->isGather() && E->Idx < *GatheredLoadsEntriesFirst) || all_of(E->Scalars, [=](Value *V) -> bool { if (E->getOpcode() == Instruction::GetElementPtr && !isa(V)) return true; auto *I = cast(V); return !E->isOpcodeOrAlt(I) || I->getParent() == BB || isVectorLikeInstWithConstOps(I); })) && \"Expected gathered loads or GEPs or instructions from same basic \" \"block.\"' failed.", "body": "To reproduce the test below with -passes slp-vectorizer -slp-threshold=-99999 \r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i32 @wombat(i32 %arg, i32 %arg1) #0 gc \"statepoint-example\" {\r\nbb:\r\n %zext = zext i32 %arg1 to i64\r\n %zext2 = zext i32 %arg to i64\r\n %sext = sext i32 %arg to i64\r\n br label %bb3\r\n\r\nbb3: ; preds = %bb\r\n %getelementptr = getelementptr i64, ptr addrspace(1) null, i64 %zext\r\n %getelementptr4 = getelementptr i64, ptr addrspace(1) null, i64 %zext2\r\n %getelementptr5 = getelementptr i64, ptr addrspace(1) null, i64 %sext\r\n %zext6 = zext i32 0 to i64\r\n %getelementptr7 = getelementptr i64, ptr addrspace(1) null, i64 %zext6\r\n ret i32 0\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+prfchw,-cldemote,+avx,+aes,+sahf,+pclmul,-xop,+crc32,+xsaves,-avx512fp16,-usermsr,-sm4,-egpr,+sse4.1,-avx512ifma,+xsave,+sse4.2,-tsxldtrk,-sm3,-ptwrite,-widekl,+invpcid,+64bit,+xsavec,-avx10.1-512,-avx512vpopcntdq,+cmov,-avx512vp2intersect,+avx512cd,+movbe,-avxvnniint8,-ccmp,-amx-int8,-kl,-avx10.1-256,+evex512,-avxvnni,+rtm,+adx,+avx2,-hreset,-movdiri,-serialize,-sha512,-vpclmulqdq,+avx512vl,-uintr,-cf,+clflushopt,-raoint,-cmpccxadd,+bmi,-amx-tile,+sse,-avx10.2-256,-gfni,-avxvnniint16,-amx-fp16,-zu,-ndd,+xsaveopt,+rdrnd,+avx512f,-amx-bf16,-avx512bf16,-avx512vnni,-push2pop2,+cx8,+avx512bw,+sse3,+pku,-nf,+fsgsbase,-clzero,-mwaitx,-lwp,+lzcnt,-sha,-movdir64b,-ppx,-wbnoinvd,-enqcmd,-avx10.2-512,-avxneconvert,-tbm,-pconfig,-amx-complex,+ssse3,+cx16,+bmi2,+fma,+popcnt,-avxifma,+f16c,-avx512bitalg,-rdpru,+clwb,+mmx,+sse2,+rdseed,-avx512vbmi2,-prefetchi,-rdpid,-fma4,-avx512vbmi,-shstk,-vaes,-waitpkg,-sgx,+fxsr,+avx512dq,-sse4a,-avx512f\" }\r\n\r\n```\r\nReproducer: https://godbolt.org/z/8WhP4hY6G\r\n\r\nStack dump:\r\n```\r\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:13048: llvm::Instruction& llvm::slpvectorizer::BoUpSLP::getLastInstructionInBundle(const llvm::slpvectorizer::BoUpSLP::TreeEntry*): Assertion `((GatheredLoadsEntriesFirst.has_value() && E->getOpcode() == Instruction::Load && E->isGather() && E->Idx < *GatheredLoadsEntriesFirst) || all_of(E->Scalars, [=](Value *V) -> bool { if (E->getOpcode() == Instruction::GetElementPtr && !isa(V)) return true; auto *I = cast(V); return !E->isOpcodeOrAlt(I) || I->getParent() == BB || isVectorLikeInstWithConstOps(I); })) && \"Expected gathered loads or GEPs or instructions from same basic \" \"block.\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes slp-vectorizer -slp-threshold=-99999 \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"wombat\"\r\n #0 0x00000000050d8608 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x50d8608)\r\n #1 0x00000000050d5fbc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000070a067c42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000070a067c969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000070a067c42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000070a067c287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000070a067c2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000070a067c39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000004157831 llvm::slpvectorizer::BoUpSLP::getLastInstructionInBundle(llvm::slpvectorizer::BoUpSLP::TreeEntry const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4157831)\r\n #9 0x0000000004157890 llvm::slpvectorizer::BoUpSLP::setInsertPointAfterBundle(llvm::slpvectorizer::BoUpSLP::TreeEntry const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4157890)\r\n#10 0x00000000041820da llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41820da)\r\n#11 0x00000000041a7a5c llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::SmallDenseSet> const&, llvm::Instruction*) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41a7a5c)\r\n#12 0x00000000041ab4ad llvm::slpvectorizer::BoUpSLP::vectorizeTree() (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41ab4ad)\r\n#13 0x00000000041c0e45 llvm::SLPVectorizerPass::tryToVectorizeList(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41c0e45)\r\n#14 0x00000000041c2f7b llvm::SLPVectorizerPass::vectorizeGEPIndices(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41c2f7b)\r\n#15 0x00000000041cf257 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#16 0x00000000041cfedb llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41cfedb)\r\n#17 0x0000000002f64f0e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f64f0e)\r\n#18 0x0000000004ed9788 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed9788)\r\n#19 0x0000000000e2773e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe2773e)\r\n#20 0x0000000004ed81ae llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed81ae)\r\n#21 0x0000000000e26e8e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe26e8e)\r\n#22 0x0000000004ed7be0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4ed7be0)\r\n#23 0x000000000090a07a llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x90a07a)\r\n#24 0x00000000008fce7c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fce7c)\r\n#25 0x000070a067c29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#26 0x000070a067c29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#27 0x00000000008f493e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f493e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "116852", "issue_url": "https://github.com/llvm/llvm-project/issues/116852", "bug_type": "crash", "base_commit": "d8a1c6d72739c99f4b52a84e907b96b685b44c5e", "knowledge_cutoff": "2024-11-19T18:03:10Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "79682c4d57620e623fb30271cc8003d0c9e14a01", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[13614, 13620]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::gather"]}}, "patch": "commit 79682c4d57620e623fb30271cc8003d0c9e14a01\nAuthor: Alexey Bataev \nDate: Tue Nov 19 11:03:11 2024 -0800\n\n [SLP]Check if the buildvector root is not a part of the graph before deletion\n \n If the buildvector root has no uses, it might be still needed as a part\n of the graph, so need to check that it is not a part of the graph before\n deletion.\n \n Fixes #116852\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex dc0dffd9fcbf..e70627b6afc1 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -13614,7 +13614,10 @@ Value *BoUpSLP::gather(\n } else {\n Vec = CreateShuffle(Root, Vec, Mask);\n if (auto *OI = dyn_cast(OriginalRoot);\n- OI && OI->hasNUses(0))\n+ OI && OI->hasNUses(0) &&\n+ none_of(VectorizableTree, [&](const std::unique_ptr &TE) {\n+ return TE->VectorizedValue == OI;\n+ }))\n eraseInstruction(OI);\n }\n }\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/bv-root-part-of-graph.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "", "test_body": "\ndefine void @test() {\n;\nbb:\n br label %bb1\n\nbb1:\n %phi = phi i32 [ 0, %bb ], [ %or21, %bb1 ]\n %sitofp = sitofp i32 0 to float\n %fadd = fadd float %sitofp, %sitofp\n %fadd2 = fadd float %fadd, 0.000000e+00\n %fcmp = fcmp ogt float %fadd2, 0.000000e+00\n %select = select i1 %fcmp, i32 0, i32 0\n %select3 = select i1 false, i32 %select, i32 0\n %fadd4 = fadd float %sitofp, 0.000000e+00\n %fadd5 = fadd float %fadd4, 0.000000e+00\n %fcmp6 = fcmp ogt float %fadd5, 0.000000e+00\n %select7 = select i1 %fcmp6, i32 0, i32 0\n %select8 = select i1 false, i32 %select7, i32 0\n %or = or i32 %select3, %select8\n %sitofp9 = sitofp i32 0 to float\n %fmul = fmul float 0.000000e+00, 0.000000e+00\n %fadd10 = fadd float %sitofp9, %fmul\n %fadd11 = fadd float %fadd10, 0.000000e+00\n %fcmp12 = fcmp ogt float %fadd11, 0.000000e+00\n %select13 = select i1 %fcmp12, i32 0, i32 0\n %select14 = select i1 false, i32 %select13, i32 0\n %or15 = or i32 %select14, %or\n %fadd16 = fadd float %fmul, 0.000000e+00\n %fadd17 = fadd float %fadd16, 0.000000e+00\n %fcmp18 = fcmp ogt float %fadd17, 0.000000e+00\n %select19 = select i1 %fcmp18, i32 0, i32 0\n %select20 = select i1 false, i32 %select19, i32 0\n %or21 = or i32 %or15, %select20\n br label %bb1\n}"}]}], "issue": {"title": "Assertion `I->use_empty() && \"trying to erase instruction with users.\"' failed.", "body": "To reproduce run the test below with -passes=slp-vectorizer\r\n```\r\n; ModuleID = './reduced.ll'\r\nsource_filename = \"./reduced.ll\"\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @wombat() gc \"statepoint-example\" {\r\nbb:\r\n br label %bb1\r\n\r\nbb1: ; preds = %bb1, %bb\r\n %phi = phi i32 [ 0, %bb ], [ %or21, %bb1 ]\r\n %sitofp = sitofp i32 0 to float\r\n %fadd = fadd float %sitofp, %sitofp\r\n %fadd2 = fadd float %fadd, 0.000000e+00\r\n %fcmp = fcmp ogt float %fadd2, 0.000000e+00\r\n %select = select i1 %fcmp, i32 0, i32 0\r\n %select3 = select i1 false, i32 %select, i32 0\r\n %fadd4 = fadd float %sitofp, 0.000000e+00\r\n %fadd5 = fadd float %fadd4, 0.000000e+00\r\n %fcmp6 = fcmp ogt float %fadd5, 0.000000e+00\r\n %select7 = select i1 %fcmp6, i32 0, i32 0\r\n %select8 = select i1 false, i32 %select7, i32 0\r\n %or = or i32 %select3, %select8\r\n %sitofp9 = sitofp i32 0 to float\r\n %fmul = fmul float 0.000000e+00, 0.000000e+00\r\n %fadd10 = fadd float %sitofp9, %fmul\r\n %fadd11 = fadd float %fadd10, 0.000000e+00\r\n %fcmp12 = fcmp ogt float %fadd11, 0.000000e+00\r\n %select13 = select i1 %fcmp12, i32 0, i32 0\r\n %select14 = select i1 false, i32 %select13, i32 0\r\n %or15 = or i32 %select14, %or\r\n %fadd16 = fadd float %fmul, 0.000000e+00\r\n %fadd17 = fadd float %fadd16, 0.000000e+00\r\n %fcmp18 = fcmp ogt float %fadd17, 0.000000e+00\r\n %select19 = select i1 %fcmp18, i32 0, i32 0\r\n %select20 = select i1 false, i32 %select19, i32 0\r\n %or21 = or i32 %or15, %select20\r\n br label %bb1\r\n}\r\n```\r\nReproducer: https://godbolt.org/z/T8j1xEoKM\r\nStack dump:\r\n```\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"wombat\"\r\n #0 0x0000000005194b38 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x5194b38)\r\n #1 0x000000000519252c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007eacdf842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007eacdf8969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007eacdf842476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007eacdf8287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007eacdf82871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007eacdf839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x000000000419bb4f (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x419bb4f)\r\n #9 0x000000000425f0bf llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#10 0x000000000425fa7b llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x425fa7b)\r\n#11 0x0000000002fbe41e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2fbe41e)\r\n#12 0x0000000004f949e8 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f949e8)\r\n#13 0x0000000000e4027e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe4027e)\r\n#14 0x0000000004f9340e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f9340e)\r\n#15 0x0000000000e3f92e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3f92e)\r\n#16 0x0000000004f92e40 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f92e40)\r\n#17 0x0000000000909fda llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x909fda)\r\n#18 0x00000000008fc70a optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8fc70a)\r\n#19 0x00007eacdf829d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#20 0x00007eacdf829e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#21 0x00000000008f3e9e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f3e9e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\n", "author": "TatyanaDoubts", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "98665", "issue_url": "https://github.com/llvm/llvm-project/issues/98665", "bug_type": "crash", "base_commit": "662c6fc74c1ae7fdefd27524dfaeb7f3e9a1f553", "knowledge_cutoff": "2024-07-12T17:30:32Z", "lit_test_dir": ["llvm/test/Transforms/EarlyCSE"], "hints": {"fix_commit": "34bfed63313d1340378fc1be931253333db8c36c", "components": ["ConstantFold"], "bug_location_lineno": {"llvm/lib/Analysis/ConstantFolding.cpp": [[2754, 2780]]}, "bug_location_funcname": {"llvm/lib/Analysis/ConstantFolding.cpp": ["ConstantFoldIntrinsicCall2"]}}, "patch": "commit 34bfed63313d1340378fc1be931253333db8c36c\nAuthor: Yingwei Zheng \nDate: Mon Jul 15 13:00:25 2024 +0800\n\n [ConstantFold] Fix result type when folding powi.f16 (#98681)\n \n Fixes #98665.\n\ndiff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp\nindex 962880f68f07..6c52091cd5d7 100644\n--- a/llvm/lib/Analysis/ConstantFolding.cpp\n+++ b/llvm/lib/Analysis/ConstantFolding.cpp\n@@ -2754,27 +2754,28 @@ static Constant *ConstantFoldIntrinsicCall2(Intrinsic::ID IntrinsicID, Type *Ty,\n ((Mask & fcPosInf) && Op1V.isPosInfinity());\n return ConstantInt::get(Ty, Result);\n }\n+ case Intrinsic::powi: {\n+ int Exp = static_cast(Op2C->getSExtValue());\n+ switch (Ty->getTypeID()) {\n+ case Type::HalfTyID:\n+ case Type::FloatTyID: {\n+ APFloat Res(std::pow(Op1V.convertToFloat(), Exp));\n+ if (Ty->isHalfTy()) {\n+ bool Unused;\n+ Res.convert(APFloat::IEEEhalf(), APFloat::rmNearestTiesToEven,\n+ &Unused);\n+ }\n+ return ConstantFP::get(Ty->getContext(), Res);\n+ }\n+ case Type::DoubleTyID:\n+ return ConstantFP::get(Ty, std::pow(Op1V.convertToDouble(), Exp));\n+ default:\n+ return nullptr;\n+ }\n+ }\n default:\n break;\n }\n-\n- if (!Ty->isHalfTy() && !Ty->isFloatTy() && !Ty->isDoubleTy())\n- return nullptr;\n- if (IntrinsicID == Intrinsic::powi && Ty->isHalfTy())\n- return ConstantFP::get(\n- Ty->getContext(),\n- APFloat((float)std::pow((float)Op1V.convertToDouble(),\n- (int)Op2C->getZExtValue())));\n- if (IntrinsicID == Intrinsic::powi && Ty->isFloatTy())\n- return ConstantFP::get(\n- Ty->getContext(),\n- APFloat((float)std::pow((float)Op1V.convertToDouble(),\n- (int)Op2C->getZExtValue())));\n- if (IntrinsicID == Intrinsic::powi && Ty->isDoubleTy())\n- return ConstantFP::get(\n- Ty->getContext(),\n- APFloat((double)std::pow(Op1V.convertToDouble(),\n- (int)Op2C->getZExtValue())));\n }\n return nullptr;\n }\n", "tests": [{"file": "llvm/test/Transforms/EarlyCSE/math-2.ll", "commands": ["opt -passes=early-cse -earlycse-debug-hash -S -o - %s"], "tests": [{"test_name": "i_powi", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare double @llvm.powi.f64.i32(double, i32) #0\n\ndefine double @i_powi() {\n %res = tail call fast double @llvm.powi.f64.i32(double 1.000000e+00, i32 2)\n ret double %res\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "pr98665", "test_body": "define half @pr98665() {\n %x = call half @llvm.powi.f16.i32(half 0xH3C00, i32 1)\n ret half %x\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare half @llvm.powi.f16.i32(half, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "Misoptimization: `EarlyCSEPass` uses replaces `powi.f16` with `float` result", "body": "It looks like EarlyCSEPass is transformin the following:\r\n\r\n```llvm\r\n %_6 = alloca [48 x i8], align 8\r\n %_3 = alloca [2 x i8], align 2\r\n %0 = call half @llvm.powi.f16.i32(half 0xH3C00, i32 1) ; 0xH3C00 = 1.0f16\r\n store half %0, ptr %_3, align 2\r\n %1 = load half, ptr %_3, align 2\r\n %_4 = fcmp oeq half %1, 0xH3C00\r\n br i1 %_4, label %bb1, label %bb2\r\n```\r\n\r\nInto this:\r\n\r\n```llvm\r\n %_6 = alloca [48 x i8], align 8\r\n %_3 = alloca [2 x i8], align 2\r\n store float 1.000000e+00, ptr %_3, align 2\r\n %0 = load half, ptr %_3, align 2\r\n %_4 = fcmp oeq half %0, 0xH3C00\r\n br i1 %_4, label %bb1, label %bb2\r\n```\r\n\r\nAnd later InstCombine folds further into:\r\n\r\n```llvm\r\n %_6 = alloca [48 x i8], align 8\r\n %_3 = alloca [2 x i8], align 2\r\n store float 1.000000e+00, ptr %_3, align 2\r\n br i1 false, label %bb1, label %bb2\r\n```\r\n\r\nEarlyCSE seems to be doing an incorrect transformation: the result of `powi.f16(1.0, 1)` should be `half` 1.0 (0x3c00), but it is returning `float` 1.0 (0x3f800000). This is incorrect and an OOB write.\r\n\r\nThis comes from the following rust code, which asserts only when optimizations are enabled:\r\n\r\n```rust\r\n#![feature(f16)]\r\n#![allow(unused)]\r\n\r\n#[inline(never)]\r\npub fn check_pow(a: f16) {\r\n assert_eq!(1.0f16.powi(1), 1.0);\r\n}\r\n\r\npub fn main() {\r\n check_pow(1.0);\r\n println!(\"finished\");\r\n}\r\n```\r\n\r\nLink to compiler explorer: https://rust.godbolt.org/z/zsbzzxGvj\r\n\r\nI'm not sure how to reduce to a llc example since the passes appear different. I have been testing on aarch64 since x86 has other f16 ABI bugs, but I don't think this is limited to aarch64.", "author": "tgross35", "labels": ["miscompilation", "llvm:analysis"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://godbolt.org/z/jhM834G3z\r\n```\r\ndefine half @test() {\r\n %x = call half @llvm.powi.f16.i32(half 0xH3C00, i32 1)\r\n ret half %x\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/lib/IR/Value.cpp:507: void llvm::Value::doRAUW(llvm::Value*, llvm::Value::ReplaceMetadataUses): Assertion `New->getType() == getType() && \"replaceAllUses of value with new value of different type!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=early-cse \r\n1.\tRunning pass \"function(early-cse<>)\" on module \"\"\r\n2.\tRunning pass \"early-cse<>\" on function \"test\"\r\n #0 0x0000000004e9b948 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e9b948)\r\n #1 0x0000000004e990bc SignalHandler(int) Signals.cpp:0:0\r\n #2 0x000078ad4b442520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x000078ad4b4969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x000078ad4b442476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x000078ad4b4287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x000078ad4b42871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x000078ad4b439e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x0000000004cce094 llvm::Value::doRAUW(llvm::Value*, llvm::Value::ReplaceMetadataUses) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4cce094)\r\n #9 0x0000000003ae3b58 (anonymous namespace)::EarlyCSE::processNode(llvm::DomTreeNodeBase*) EarlyCSE.cpp:0:0\r\n#10 0x0000000003ae5a85 (anonymous namespace)::EarlyCSE::run() EarlyCSE.cpp:0:0\r\n#11 0x0000000003ae81f8 llvm::EarlyCSEPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3ae81f8)\r\n#12 0x0000000002ac11be llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2ac11be)\r\n#13 0x0000000004c9f2d8 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c9f2d8)\r\n#14 0x0000000000dc9bee llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdc9bee)\r\n#15 0x0000000004c9dd46 llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c9dd46)\r\n#16 0x0000000000dca1fe llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdca1fe)\r\n#17 0x0000000004c9bf10 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4c9bf10)\r\n#18 0x00000000008db1d2 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8db1d2)\r\n#19 0x00000000008ce0fc optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8ce0fc)\r\n#20 0x000078ad4b429d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#21 0x000078ad4b429e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#22 0x00000000008c556e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8c556e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\n\r\nThere is a bug in constant folding.\r\n\r\n"}]}, "verified": true} {"bug_id": "112476", "issue_url": "https://github.com/llvm/llvm-project/issues/112476", "bug_type": "miscompilation", "base_commit": "3ef630ac339f31686290f9460a40eb2a9c9f5bd0", "knowledge_cutoff": "2024-10-16T04:35:15Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "09361953116770b646decf5820a9455ada2ba4fc", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": [[1738, 1744], [1844, 1850]], "llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[1448, 1453]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp": ["InstCombinerImpl::foldICmpAndConstConst", "InstCombinerImpl::foldICmpAndShift"], "llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["InstCombinerImpl::foldSelectEqualityTest"]}}, "patch": "commit 09361953116770b646decf5820a9455ada2ba4fc\nAuthor: Yingwei Zheng \nDate: Wed Oct 16 19:13:52 2024 +0800\n\n [InstCombine] Drop `samesign` in InstCombine (#112480)\n \n Closes https://github.com/llvm/llvm-project/issues/112476.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\nindex 7129499e0f8f..18a6fdcec172 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp\n@@ -1738,7 +1738,7 @@ Instruction *InstCombinerImpl::foldICmpAndShift(ICmpInst &Cmp,\n \n // Compute X & (C2 << Y).\n Value *NewAnd = Builder.CreateAnd(Shift->getOperand(0), NewShift);\n- return replaceOperand(Cmp, 0, NewAnd);\n+ return new ICmpInst(Cmp.getPredicate(), NewAnd, Cmp.getOperand(1));\n }\n \n return nullptr;\n@@ -1844,7 +1844,7 @@ Instruction *InstCombinerImpl::foldICmpAndConstConst(ICmpInst &Cmp,\n /*HasNUW=*/true),\n One, Or->getName());\n Value *NewAnd = Builder.CreateAnd(A, NewOr, And->getName());\n- return replaceOperand(Cmp, 0, NewAnd);\n+ return new ICmpInst(Cmp.getPredicate(), NewAnd, Cmp.getOperand(1));\n }\n }\n }\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex 8be2eeed84ad..623694663aa1 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -1448,6 +1448,7 @@ Instruction *InstCombinerImpl::foldSelectEqualityTest(SelectInst &Sel) {\n m_c_SpecificICmp(ICmpInst::ICMP_EQ, m_Specific(X), m_Specific(Y))))\n return nullptr;\n \n+ cast(XeqY)->setSameSign(false);\n return replaceInstUsesWith(Sel, XeqY);\n }\n \n", "tests": [{"file": "llvm/test/Transforms/InstCombine/icmp-and-shift.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "test_shr_and_1_ne_0", "test_body": "define i1 @test_shr_and_1_ne_0(i32 %a, i32 %b) {\n %shr = lshr i32 %a, %b\n %and = and i32 %shr, 1\n %cmp = icmp ne i32 %and, 0\n ret i1 %cmp\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/icmp-equality-test.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "icmp_equality_test_swift_optional_pointers", "test_body": "define i1 @icmp_equality_test_swift_optional_pointers(i64 %X, i64 %Y) {\nentry:\n %XeqC = icmp eq i64 %X, 0\n %YeqC = icmp eq i64 %Y, 0\n %either = select i1 %XeqC, i1 true, i1 %YeqC\n %both = select i1 %XeqC, i1 %YeqC, i1 false\n %XeqY = icmp eq i64 %X, %Y\n %equal = select i1 %either, i1 %both, i1 %XeqY\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_wrong_equal", "test_body": "define i1 @icmp_equality_test_wrong_equal(i64 %X, i64 %Y, i64 %Z) {\nentry:\n %XeqZ = icmp eq i64 %X, %Z\n %YeqZ = icmp eq i64 %Y, %Z\n %XeqY = icmp eq i64 %X, %Y\n %not.YeqZ = xor i1 %YeqZ, true\n %and = select i1 %not.YeqZ, i1 %XeqY, i1 false\n %equal = select i1 %XeqZ, i1 %and, i1 %YeqZ\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_commute_icmp1", "test_body": "define i1 @icmp_equality_test_commute_icmp1(i64 %X, i64 %Y, i64 %Z) {\nentry:\n %XeqZ = icmp eq i64 %Z, %X\n %YeqZ = icmp eq i64 %Z, %Y\n %XeqY = icmp eq i64 %Y, %X\n %not.YeqZ = xor i1 %YeqZ, true\n %and = select i1 %not.YeqZ, i1 %XeqY, i1 false\n %equal = select i1 %XeqZ, i1 %YeqZ, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_wrong_cmp", "test_body": "define i1 @icmp_equality_test_wrong_cmp(i64 %X, i64 %Y, i64 %Z) {\nentry:\n %XeqZ = icmp eq i64 %X, %Z\n %YeqZ = icmp eq i64 %Y, %Z\n %XeqY = icmp eq i64 %X, 999\n %not.YeqZ = xor i1 %YeqZ, true\n %and = select i1 %not.YeqZ, i1 %XeqY, i1 false\n %equal = select i1 %XeqZ, i1 %YeqZ, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_wrong_constant", "test_body": "define i1 @icmp_equality_test_wrong_constant(i64 %X, i64 %Y) {\nentry:\n %XeqC = icmp eq i64 %X, 0\n %YeqC = icmp eq i64 %Y, 999\n %XeqY = icmp eq i64 %X, %Y\n %not.YeqC = xor i1 %YeqC, true\n %and = select i1 %not.YeqC, i1 %XeqY, i1 false\n %equal = select i1 %XeqC, i1 %YeqC, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_vector", "test_body": "define <2 x i1> @icmp_equality_test_vector(<2 x i64> %X, <2 x i64> %Y) {\nentry:\n %XeqC = icmp eq <2 x i64> %X, \n %YeqC = icmp eq <2 x i64> %Y, \n %XeqY = icmp eq <2 x i64> %X, %Y\n %not.YeqC = xor <2 x i1> %YeqC, splat (i1 true)\n %and = select <2 x i1> %not.YeqC, <2 x i1> %XeqY, <2 x i1> zeroinitializer\n %equal = select <2 x i1> %XeqC, <2 x i1> %YeqC, <2 x i1> %and\n ret <2 x i1> %equal\n}\n"}, {"test_name": "icmp_equality_test_commute_select2", "test_body": "define i1 @icmp_equality_test_commute_select2(i64 %X, i64 %Y, i64 %Z) {\nentry:\n %XeqZ = icmp eq i64 %X, %Z\n %YeqZ = icmp eq i64 %Y, %Z\n %XeqY = icmp eq i64 %X, %Y\n %not.XeqZ = xor i1 %XeqZ, true\n %and = select i1 %YeqZ, i1 false, i1 %XeqY\n %equal = select i1 %not.XeqZ, i1 %and, i1 %YeqZ\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_constant", "test_body": "define i1 @icmp_equality_test_constant(i42 %X, i42 %Y) {\nentry:\n %XeqC = icmp eq i42 %X, -42\n %YeqC = icmp eq i42 %Y, -42\n %XeqY = icmp eq i42 %X, %Y\n %not.YeqC = xor i1 %YeqC, true\n %and = select i1 %not.YeqC, i1 %XeqY, i1 false\n %equal = select i1 %XeqC, i1 %YeqC, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_wrong_and", "test_body": "define i1 @icmp_equality_test_wrong_and(i64 %X, i64 %Y, i64 %Z) {\nentry:\n %XeqZ = icmp eq i64 %X, %Z\n %YeqZ = icmp eq i64 %Y, %Z\n %XeqY = icmp eq i64 %X, %Y\n %not.YeqZ = xor i1 %YeqZ, true\n %and = select i1 %not.YeqZ, i1 false, i1 %XeqY\n %equal = select i1 %XeqZ, i1 %YeqZ, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test", "test_body": "define i1 @icmp_equality_test(i64 %X, i64 %Y, i64 %Z) {\nentry:\n %XeqZ = icmp eq i64 %X, %Z\n %YeqZ = icmp eq i64 %Y, %Z\n %XeqY = icmp eq i64 %X, %Y\n %not.YeqZ = xor i1 %YeqZ, true\n %and = select i1 %not.YeqZ, i1 %XeqY, i1 false\n %equal = select i1 %XeqZ, i1 %YeqZ, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_missing_not", "test_body": "define i1 @icmp_equality_test_missing_not(i64 %X, i64 %Y, i64 %Z) {\n %XeqZ = icmp eq i64 %X, %Z\n %YeqZ = icmp eq i64 %Y, %Z\n %XeqY = icmp eq i64 %X, %Y\n %and = select i1 %YeqZ, i1 %XeqY, i1 false\n %equal = select i1 %XeqZ, i1 %YeqZ, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_constant_samesign", "test_body": "define i1 @icmp_equality_test_constant_samesign(i42 %X, i42 %Y) {\nentry:\n %XeqC = icmp eq i42 %X, -42\n %YeqC = icmp eq i42 %Y, -42\n %XeqY = icmp samesign eq i42 %X, %Y\n %not.YeqC = xor i1 %YeqC, true\n %and = select i1 %not.YeqC, i1 %XeqY, i1 false\n %equal = select i1 %XeqC, i1 %YeqC, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_commute_icmp2", "test_body": "define i1 @icmp_equality_test_commute_icmp2(i64 %X, i64 %Y, i64 %Z) {\n %XeqZ = icmp eq i64 %Z, %X\n %YeqZ = icmp eq i64 %Y, %Z\n %XeqY = icmp eq i64 %Y, %X\n %not.YeqZ = xor i1 %YeqZ, true\n %and = select i1 %not.YeqZ, i1 %XeqY, i1 false\n %equal = select i1 %XeqZ, i1 %YeqZ, i1 %and\n ret i1 %equal\n}\n"}, {"test_name": "icmp_equality_test_commute_select1", "test_body": "define i1 @icmp_equality_test_commute_select1(i64 %X, i64 %Y, i64 %Z) {\nentry:\n %XeqZ = icmp eq i64 %X, %Z\n %YeqZ = icmp eq i64 %Y, %Z\n %XeqY = icmp eq i64 %X, %Y\n %and = select i1 %YeqZ, i1 false, i1 %XeqY\n %equal = select i1 %XeqZ, i1 %YeqZ, i1 %and\n ret i1 %equal\n}\n"}]}, {"file": "llvm/test/Transforms/InstCombine/icmp.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "icmp_and_or_lshr", "test_body": "target datalayout = \"e-p:64:64:64-p1:16:16:16-p2:32:32:32-p3:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64\"\n\ndefine i1 @icmp_and_or_lshr(i32 %x, i32 %y) {\n %shf = lshr i32 %x, %y\n %or = or i32 %shf, %x\n %and = and i32 %or, 1\n %ret = icmp ne i32 %and, 0\n ret i1 %ret\n}\n"}]}], "issue": {"title": "[InstCombine] `samesign` flag should be dropped in `InstCombinerImpl::foldICmpAndConstConst`", "body": "Reproducer:\r\n```\r\ndefine i1 @src(i32 %x, i32 %y) {\r\n %shf = lshr i32 %x, %y\r\n %or = or i32 %shf, %x\r\n %and = and i32 %or, 1\r\n %ret = icmp samesign ne i32 %and, 0\r\n ret i1 %ret\r\n}\r\n\r\ndefine i1 @tgt(i32 %x, i32 %y) {\r\n %shf1 = shl nuw i32 1, %y\r\n %or2 = or i32 %shf1, 1\r\n %and3 = and i32 %x, %or2\r\n %ret = icmp samesign ne i32 %and3, 0\r\n ret i1 %ret\r\n}\r\n```\r\n```\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\ni32 %x = #x80000000 (2147483648, -2147483648)\r\ni32 %y = #x0000001f (31)\r\n\r\nSource:\r\ni32 %shf = #x00000001 (1)\r\ni32 %or = #x80000001 (2147483649, -2147483647)\r\ni32 %and = #x00000001 (1)\r\ni1 %ret = #x1 (1)\r\n\r\nTarget:\r\ni32 %shf1 = #x80000000 (2147483648, -2147483648)\r\ni32 %or2 = #x80000001 (2147483649, -2147483647)\r\ni32 %and3 = #x80000000 (2147483648, -2147483648)\r\ni1 %ret = poison\r\nSource value: #x1 (1)\r\nTarget value: poison\r\n```\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "102934", "issue_url": "https://github.com/llvm/llvm-project/issues/102934", "bug_type": "crash", "base_commit": "2913e71865dfc063a47ddfaf1e2ce07763f69614", "knowledge_cutoff": "2024-08-12T16:46:46Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "2ab910c08c4ed43d6fd9eb2ef9b3ff52e0293cf9", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[3705, 3711]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::collectLoopUniforms"]}}, "patch": "commit 2ab910c08c4ed43d6fd9eb2ef9b3ff52e0293cf9\nAuthor: Florian Hahn \nDate: Tue Aug 13 09:23:44 2024 +0100\n\n [LV] Check pointer user are in loop when checking for uniform pointers.\n \n Widening decisions are not set for users outside the loop. Avoid\n crashing by only calling isVectorizedMemAccessUse for users in the loop.\n \n Fixes https://github.com/llvm/llvm-project/issues/102934.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 5547116133e8..0d1262fa1872 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -3705,7 +3705,8 @@ void LoopVectorizationCostModel::collectLoopUniforms(ElementCount VF) {\n auto *I = cast(V);\n auto UsersAreMemAccesses =\n llvm::all_of(I->users(), [&](User *U) -> bool {\n- return isVectorizedMemAccessUse(cast(U), V);\n+ auto *UI = cast(U);\n+ return TheLoop->contains(UI) && isVectorizedMemAccessUse(UI, V);\n });\n if (UsersAreMemAccesses)\n addToWorklistIfAllowed(I);\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/gep-use-outside-loop.ll", "commands": ["opt -p loop-vectorize -mtriple=x86_64-apple-macosx -mcpu=skylake-avx512 -force-vector-width=4 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "gep_use_outside_loop", "test_body": "define void @gep_use_outside_loop(ptr noalias %dst, ptr %src) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %gep.dst = getelementptr i16, ptr %dst, i64 %iv\n %gep.src = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep.src, align 2\n %c = icmp eq i16 %l, 10\n br i1 %c, label %loop.latch, label %then\n\nthen: ; preds = %loop.header\n store i16 0, ptr %gep.dst, align 2\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 99\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n store i16 0, ptr %gep.dst, align 2\n ret void\n}\n"}, {"test_name": "gep_use_in_dead_block", "test_body": "define void @gep_use_in_dead_block(ptr noalias %dst, ptr %src) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %gep.src = getelementptr i16, ptr %src, i64 %iv\n %l = load i16, ptr %gep.src, align 2\n %c = icmp eq i16 %l, 10\n br i1 %c, label %loop.latch, label %then\n\nthen: ; preds = %loop.header\n %gep.dst = getelementptr i16, ptr %dst, i64 %iv\n store i16 0, ptr %gep.dst, align 2\n br label %loop.latch\n\ndead: ; preds = %dead\n store i16 0, ptr %gep.dst, align 2\n br label %dead\n\nloop.latch: ; preds = %then, %loop.header\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 99\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}]}], "issue": {"title": "[LoopVectorize] Assertion `WideningDecision != CM_Unknown && \"Widening decision should be ready at this moment\"' failed.", "body": "Reduced LLVM IR:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main(i64 %indvars.iv.i) #0 {\r\nentry:\r\n br label %for.body6.i\r\n\r\nfor.body6.i: ; preds = %for.inc34.i, %entry\r\n %indvars.iv.i1 = phi i64 [ 0, %entry ], [ %indvars.iv.next.i, %for.inc34.i ]\r\n br i1 false, label %for.body6.i.for.inc34.i_crit_edge, label %for.cond7.preheader.i\r\n\r\nfor.body6.i.for.inc34.i_crit_edge: ; preds = %for.body6.i\r\n br label %for.inc34.i\r\n\r\nfor.cond7.preheader.i: ; preds = %for.body6.i\r\n %arrayidx30.i = getelementptr [0 x i16], ptr null, i64 0, i64 %indvars.iv.i\r\n store i16 0, ptr %arrayidx30.i, align 2\r\n br label %for.inc34.i\r\n\r\nfor.body19.1.i: ; preds = %for.body19.1.i\r\n store i16 0, ptr %arrayidx30.i, align 2\r\n br label %for.body19.1.i\r\n\r\nfor.inc34.i: ; preds = %for.cond7.preheader.i, %for.body6.i.for.inc34.i_crit_edge\r\n %indvars.iv.next.i = add i64 %indvars.iv.i1, 1\r\n %exitcond.not.i = icmp eq i64 %indvars.iv.next.i, 0\r\n br i1 %exitcond.not.i, label %q.exit, label %for.body6.i\r\n\r\nq.exit: ; preds = %for.inc34.i\r\n ret i32 0\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl256b,+zvl32b,+zvl64b,-b,-e,-experimental-smmpm,-experimental-smnpm,-experimental-ssnpm,-experimental-sspm,-experimental-ssqosid,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nBacktrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=\"loop-vectorize\" -S reduced.ll\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:3631: llvm::LoopVectorizationCostModel::collectLoopUniforms(llvm::ElementCount)::: Assertion `WideningDecision != CM_Unknown && \"Widening decision should be ready at this moment\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt -passes=loop-vectorize -S reduced.ll\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"main\"\r\n #0 0x000059f80448a570 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f4f570)\r\n #1 0x000059f80448798f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2f4c98f)\r\n #2 0x000059f804487ae5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x0000795355842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007953558969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007953558969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007953558969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x0000795355842476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007953558287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x000079535582871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x0000795355839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x000059f80347d610 llvm::LoopVectorizationCostModel::collectLoopUniforms(llvm::ElementCount) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f42610)\r\n#12 0x000059f80348fa77 llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f54a77)\r\n#13 0x000059f803491b83 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f56b83)\r\n#14 0x000059f803494de1 llvm::LoopVectorizePass::runImpl(llvm::Function&, llvm::ScalarEvolution&, llvm::LoopInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::BlockFrequencyInfo*, llvm::TargetLibraryInfo*, llvm::DemandedBits&, llvm::AssumptionCache&, llvm::LoopAccessInfoManager&, llvm::OptimizationRemarkEmitter&, llvm::ProfileSummaryInfo*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f59de1)\r\n#15 0x000059f803495497 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x1f5a497)\r\n#16 0x000059f8022ff336 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdc4336)\r\n#17 0x000059f8042a17ff llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d667ff)\r\n#18 0x000059f8023024c6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdc74c6)\r\n#19 0x000059f8042a057b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d6557b)\r\n#20 0x000059f8022fe7c6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xdc37c6)\r\n#21 0x000059f80429e5dd llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2d635dd)\r\n#22 0x000059f801b38c36 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5fdc36)\r\n#23 0x000059f801b2a881 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5ef881)\r\n#24 0x0000795355829d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#25 0x0000795355829e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#26 0x0000795355829e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#27 0x000059f801b206c5 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x5e56c5)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\n\r\nFound via fuzzer.", "author": "patrick-rivos", "labels": ["vectorizers", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "117537", "issue_url": "https://github.com/llvm/llvm-project/issues/117537", "bug_type": "crash", "base_commit": "c979ec05642f292737d250c6682d85ed49bc7b6e", "knowledge_cutoff": "2024-11-25T10:47:02Z", "lit_test_dir": ["llvm/test/Transforms/SimpleLoopUnswitch"], "hints": {"fix_commit": "fc5c89900f2a4b50e0f3a88ef7c89115d93684f4", "components": ["SimpleLoopUnswitch"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp": [[1248, 1255]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp": ["buildClonedLoopBlocks"]}}, "patch": "commit fc5c89900f2a4b50e0f3a88ef7c89115d93684f4\nAuthor: Nikita Popov \nDate: Wed Nov 27 11:47:22 2024 +0100\n\n [SimpleLoopUnswitch] Fix LCSSA phi node invalidation\n \n Fixes https://github.com/llvm/llvm-project/issues/117537.\n\ndiff --git a/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp b/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp\nindex aa3cbc5e4bdd..d8ef450eeb9a 100644\n--- a/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp\n+++ b/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp\n@@ -1248,8 +1248,9 @@ static BasicBlock *buildClonedLoopBlocks(\n assert(VMap.lookup(&I) == &ClonedI && \"Mismatch in the value map!\");\n \n // Forget SCEVs based on exit phis in case SCEV looked through the phi.\n- if (SE && isa(I))\n- SE->forgetValue(&I);\n+ if (SE)\n+ if (auto *PN = dyn_cast(&I))\n+ SE->forgetLcssaPhiWithNewPredecessor(&L, PN);\n \n BasicBlock::iterator InsertPt = MergeBB->getFirstInsertionPt();\n \n", "tests": [{"file": "llvm/test/Transforms/SimpleLoopUnswitch/pr117537.ll", "commands": ["opt -S -passes='print,simple-loop-unswitch,print' -verify-scev < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "define void @test(ptr %p) {\nentry:\n %check = icmp eq ptr %p, null\n br label %bb0\n\nbb0: ; preds = %loop0, %entry\n br i1 %check, label %loop0, label %latch\n\nlatch: ; preds = %bb0\n br i1 %check, label %exit0, label %loop0\n\nexit0: ; preds = %latch\n ret void\n\nloop0: ; preds = %latch, %bb0\n %v = load atomic i32, ptr %p unordered, align 8\n %add = add i32 %v, 3\n br i1 true, label %preheader, label %bb0\n\npreheader: ; preds = %loop0\n br label %loop1\n\nloop1: ; preds = %backedge, %preheader\n %iv1 = phi i32 [ %add, %preheader ], [ %iv1.next, %backedge ]\n %iv1.next = add i32 %iv1, -33\n br label %loop2\n\nbackedge: ; preds = %loop2\n br i1 true, label %exit1, label %loop1\n\nloop2: ; preds = %loop2, %loop1\n %iv0 = phi i32 [ %iv1, %loop1 ], [ %iv0.next, %loop2 ]\n %iv0.next = add nsw i32 %iv0, 1\n %cmp = icmp sgt i32 %iv0.next, 0\n br i1 %cmp, label %backedge, label %loop2\n\nexit1: ; preds = %backedge\n ret void\n}\n"}]}], "issue": {"title": "[SCEV] Assertion `isAvailableAtLoopEntry(Op, L) && \"SCEVAddRecExpr operand is not available at loop entry!\"' failed.", "body": "Faulty IR:\n```\n; ModuleID = 'test.ll'\nsource_filename = \"/nfs/build-cache/zvm-dev-4598/jdk11/x86_64/fastdebug/lib/server/boilerplate.ll\"\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @test(ptr addrspace(1) %p) {\nentry:\n %check = icmp eq ptr addrspace(1) %p, null\n br label %bb0\n\nbb0: ; preds = %loop0, %entry\n br i1 %check, label %loop0, label %latch\n\nlatch: ; preds = %bb0\n br i1 %check, label %exit0, label %loop0\n\nexit0: ; preds = %latch\n ret void\n\nloop0: ; preds = %latch, %bb0\n %0 = load atomic i32, ptr addrspace(1) %p unordered, align 8\n %add = add i32 %0, 3\n br i1 true, label %preheader, label %bb0\n\npreheader: ; preds = %loop0\n br label %loop1\n\nloop1: ; preds = %backedge, %preheader\n %iv1 = phi i32 [ %add, %preheader ], [ %1, %backedge ]\n %1 = add i32 %iv1, -33\n br label %loop2\n\nbackedge: ; preds = %loop2\n br i1 true, label %exit1, label %loop1\n\nloop2: ; preds = %loop2, %loop1\n %iv0 = phi i32 [ %iv1, %loop1 ], [ %2, %loop2 ]\n %2 = add nsw i32 %iv0, 1\n %3 = icmp sgt i32 %2, 0\n br i1 %3, label %backedge, label %loop2\n\nexit1: ; preds = %backedge\n ret void\n}\n```\nfails when run with `opt -passes='print,simple-loop-unswitch,print' `\nDemo of failure: https://godbolt.org/z/qhboK5r3j\n", "author": "Nirhar", "labels": ["crash", "llvm:transforms"], "comments": [{"author": "Nirhar", "body": "Here is the IR after simple-loop-unswitch pass:\n![res](https://github.com/user-attachments/assets/8a35df98-54f4-45f6-9e96-626fd6814e7f)\n\n\nAt the point of crash, the SCEV for Exit Value of `%iv0`(which I assume, as we crashed while printing it) is `(-4 + (-1 * %1) + %.us-phi)`, which is incorrect. The crash happens because `%1` is in a block that does not properly dominate `%iv0`. Looks like, simple-loop-unswitch is either incorrectly updating the SCEV in the pass, or is incorrectly reporting SCEV to be preserved. As of now, I'm not sure which is the case. "}, {"author": "Nirhar", "body": "@nikic @dtcxzyw @EugeneZelenko any thoughts? I'm inclined to having a patch that removes SCEV from simple-loop-unswitch 's list of preserved analyses."}, {"author": "dtcxzyw", "body": "> I'm inclined to having a patch that removes SCEV from simple-loop-unswitch 's list of preserved analyses.\n\nSome SCEVs should be invalidated in `unswitchNontrivialInvariants`. Removing SCEV from the list of preserved analyses doesn't fix the issue.\n"}]}, "verified": true} {"bug_id": "109528", "issue_url": "https://github.com/llvm/llvm-project/issues/109528", "bug_type": "crash", "base_commit": "586736226e3a25b1fd647d3257270c1388f907bf", "knowledge_cutoff": "2024-09-21T14:20:16Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "6d6eea92e36c301e34a7ec11b2a40e3080f79f53", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[8420, 8440]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["VPRecipeBuilder::tryToWidenHistogram"]}}, "patch": "commit 6d6eea92e36c301e34a7ec11b2a40e3080f79f53\nAuthor: Florian Hahn \nDate: Wed Oct 2 13:45:48 2024 +0100\n\n [LV] Use SCEV to simplify wide binop operand to constant.\n \n The legacy cost model uses SCEV to determine if the second operand of a\n binary op is a constant. Update the VPlan construction logic to mirror\n the current legacy behavior, to fix a difference in the cost models.\n \n Fixes https://github.com/llvm/llvm-project/issues/109528.\n Fixes https://github.com/llvm/llvm-project/issues/110440.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex e1f9f29ac6d9..de3b981a4fe3 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -8420,21 +8420,28 @@ VPWidenRecipe *VPRecipeBuilder::tryToWiden(Instruction *I,\n case Instruction::Sub:\n case Instruction::Xor:\n case Instruction::Freeze:\n- if (I->getOpcode() == Instruction::Mul) {\n- // Simplify operands of multiplications using SCEV. This is needed at the\n- // moment to match the behavior of the legacy cost-model.\n- // TODO: Generalize to any opcode and move to VPlan transformation.\n- SmallVector NewOps(Operands);\n+ SmallVector NewOps(Operands);\n+ if (Instruction::isBinaryOp(I->getOpcode())) {\n+ // The legacy cost model uses SCEV to check if some of the operands are\n+ // constants. To match the legacy cost model's behavior, use SCEV to try\n+ // to replace operands with constants.\n ScalarEvolution &SE = *PSE.getSE();\n- for (unsigned I = 0; I < Operands.size(); ++I) {\n- Value *V = NewOps[I]->getUnderlyingValue();\n- if (!isa(V) && SE.isSCEVable(V->getType()))\n- if (auto *C = dyn_cast(PSE.getSE()->getSCEV(V)))\n- NewOps[I] = Plan.getOrAddLiveIn(C->getValue());\n- }\n- return new VPWidenRecipe(*I, make_range(NewOps.begin(), NewOps.end()));\n+ auto GetConstantViaSCEV = [this, &SE](VPValue *Op) {\n+ Value *V = Op->getUnderlyingValue();\n+ if (isa(V) || !SE.isSCEVable(V->getType()))\n+ return Op;\n+ auto *C = dyn_cast(SE.getSCEV(V));\n+ if (!C)\n+ return Op;\n+ return Plan.getOrAddLiveIn(C->getValue());\n+ };\n+ // For Mul, the legacy cost model checks both operands.\n+ if (I->getOpcode() == Instruction::Mul)\n+ NewOps[0] = GetConstantViaSCEV(NewOps[0]);\n+ // For other binops, the legacy cost model only checks the second operand.\n+ NewOps[1] = GetConstantViaSCEV(NewOps[1]);\n }\n- return new VPWidenRecipe(*I, make_range(Operands.begin(), Operands.end()));\n+ return new VPWidenRecipe(*I, make_range(NewOps.begin(), NewOps.end()));\n };\n }\n \n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/cost-constant-known-via-scev.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "test_foldable_live_in_via_scev", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i64 @test_foldable_live_in_via_scev() {\nentry:\n %conv = zext i16 -6 to i64\n %add = add nsw i64 %conv, -65528\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 1, %entry ], [ %iv.next, %loop ]\n %red = phi i64 [ 1, %entry ], [ %mul, %loop ]\n %mul = mul nsw i64 %red, %add\n %iv.next = add nsw i32 %iv, 1\n %ec = icmp eq i32 %iv.next, 100\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n %ret = phi i64 [ %mul, %loop ]\n ret i64 %ret\n}\n"}, {"test_name": "second_lshr_operand_zero_via_scev", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i64 @second_lshr_operand_zero_via_scev() {\nentry:\n %ext.0 = sext i8 0 to i32\n br label %loops\n\nloops: ; preds = %loops, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loops ]\n %red = phi i64 [ 0, %entry ], [ %red.next, %loops ]\n %c = icmp eq i64 %iv, 0\n %and = and i64 %iv, 0\n %0 = trunc i64 %iv to i32\n %shr = lshr i32 %0, %ext.0\n %conv.1 = zext i32 %shr to i64\n %red.next.v = select i1 %c, i64 %and, i64 %conv.1\n %red.next = or i64 %red.next.v, %red\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, 0\n br i1 %ec, label %exit, label %loops\n\nexit: ; preds = %loops\n %res = phi i64 [ %red.next, %loops ]\n ret i64 %res\n}\n"}]}], "issue": {"title": "[clang] Assertion failed in Vectorize", "body": "When I compiled this code with O2 flag, it crashed:\r\n\r\n```c\r\n#include \r\nstatic uint64_t\r\nsafe_sub_func_uint64_t_u_u(uint64_t ui1, uint64_t ui2)\r\n{\r\n return ui1 - ui2;\r\n}\r\nlong a;\r\nstatic int8_t c;\r\nvoid b(long d, int f) {\r\n int e = 0;\r\n for (; e < 64; e++)\r\n if (e - f)\r\n a |= e >> f;\r\n else\r\n a |= d & e;\r\n}\r\nuint32_t g() {\r\n b(6, c);\r\n for (;; c = safe_sub_func_uint64_t_u_u(c, 4))\r\n ;\r\n}\r\n```\r\n\r\nThe crash is:\r\n```\r\nclang: /root/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7383: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\n```\r\n\r\nDetails can be found here: https://godbolt.org/z/7733x1hvP", "author": "cardigan1008", "labels": ["vectorizers", "confirmed", "crash"], "comments": [{"author": "shafik", "body": "This looks like a regression in trunk: https://godbolt.org/z/eo7WG1e66"}, {"author": "hiraditya", "body": "Probably a recent change caused this? cc: @ayalz @fhahn "}]}, "verified": true} {"bug_id": "104761", "issue_url": "https://github.com/llvm/llvm-project/issues/104761", "bug_type": "crash", "base_commit": "5cfa8baef33636827e5aa8dd76888c724433b53e", "knowledge_cutoff": "2024-08-19T11:00:31Z", "lit_test_dir": ["llvm/test/Analysis/LoopCacheAnalysis"], "hints": {"fix_commit": "2e6deb1dd3a4422807633ba08773e8d786e43d4c", "components": ["LoopCacheAnalysis"], "bug_location_lineno": {"llvm/include/llvm/Analysis/LoopCacheAnalysis.h": [[16, 21], [31, 37], [192, 199]], "llvm/lib/Analysis/LoopCacheAnalysis.cpp": [[328, 333], [338, 351], [696, 702]]}, "bug_location_funcname": {"llvm/lib/Analysis/LoopCacheAnalysis.cpp": ["CacheCost::computeLoopCacheCost", "IndexedReference::computeRefCost", "IndexedReference::tryDelinearizeFixedSize"]}}, "patch": "commit 2e6deb1dd3a4422807633ba08773e8d786e43d4c\nAuthor: Sjoerd Meijer \nDate: Thu Nov 14 08:40:45 2024 +0000\n\n [LoopInterchange] Fix overflow in cost calculation (#111807)\n \n If the iteration count is really large, e.g. UINT_MAX, then the cost\n calculation can overflows and trigger an assert. So saturate the cost to\n INT_MAX if this is the case by using InstructionCost as a type which\n already supports this kind of overflow handling.\n \n This fixes #104761\n\ndiff --git a/llvm/include/llvm/Analysis/LoopCacheAnalysis.h b/llvm/include/llvm/Analysis/LoopCacheAnalysis.h\nindex 4fd2485e39d6..3e22487e5e34 100644\n--- a/llvm/include/llvm/Analysis/LoopCacheAnalysis.h\n+++ b/llvm/include/llvm/Analysis/LoopCacheAnalysis.h\n@@ -16,6 +16,7 @@\n \n #include \"llvm/Analysis/LoopAnalysisManager.h\"\n #include \"llvm/IR/PassManager.h\"\n+#include \"llvm/Support/InstructionCost.h\"\n #include \n \n namespace llvm {\n@@ -31,7 +32,7 @@ class ScalarEvolution;\n class SCEV;\n class TargetTransformInfo;\n \n-using CacheCostTy = int64_t;\n+using CacheCostTy = InstructionCost;\n using LoopVectorTy = SmallVector;\n \n /// Represents a memory reference as a base pointer and a set of indexing\n@@ -192,8 +193,6 @@ class CacheCost {\n using LoopCacheCostTy = std::pair;\n \n public:\n- static CacheCostTy constexpr InvalidCost = -1;\n-\n /// Construct a CacheCost object for the loop nest described by \\p Loops.\n /// The optional parameter \\p TRT can be used to specify the max. distance\n /// between array elements accessed in a loop so that the elements are\ndiff --git a/llvm/lib/Analysis/LoopCacheAnalysis.cpp b/llvm/lib/Analysis/LoopCacheAnalysis.cpp\nindex 7ca9f15ad5fc..2897b922f61e 100644\n--- a/llvm/lib/Analysis/LoopCacheAnalysis.cpp\n+++ b/llvm/lib/Analysis/LoopCacheAnalysis.cpp\n@@ -328,6 +328,8 @@ CacheCostTy IndexedReference::computeRefCost(const Loop &L,\n const SCEV *TripCount =\n computeTripCount(*AR->getLoop(), *Sizes.back(), SE);\n Type *WiderType = SE.getWiderType(RefCost->getType(), TripCount->getType());\n+ // For the multiplication result to fit, request a type twice as wide.\n+ WiderType = WiderType->getExtendedType();\n RefCost = SE.getMulExpr(SE.getNoopOrZeroExtend(RefCost, WiderType),\n SE.getNoopOrZeroExtend(TripCount, WiderType));\n }\n@@ -338,14 +340,18 @@ CacheCostTy IndexedReference::computeRefCost(const Loop &L,\n assert(RefCost && \"Expecting a valid RefCost\");\n \n // Attempt to fold RefCost into a constant.\n+ // CacheCostTy is a signed integer, but the tripcount value can be large\n+ // and may not fit, so saturate/limit the value to the maximum signed\n+ // integer value.\n if (auto ConstantCost = dyn_cast(RefCost))\n- return ConstantCost->getValue()->getZExtValue();\n+ return ConstantCost->getValue()->getLimitedValue(\n+ std::numeric_limits::max());\n \n LLVM_DEBUG(dbgs().indent(4)\n << \"RefCost is not a constant! Setting to RefCost=InvalidCost \"\n \"(invalid value).\\n\");\n \n- return CacheCost::InvalidCost;\n+ return CacheCostTy::getInvalid();\n }\n \n bool IndexedReference::tryDelinearizeFixedSize(\n@@ -696,7 +702,7 @@ CacheCostTy\n CacheCost::computeLoopCacheCost(const Loop &L,\n const ReferenceGroupsTy &RefGroups) const {\n if (!L.isLoopSimplifyForm())\n- return InvalidCost;\n+ return CacheCostTy::getInvalid();\n \n LLVM_DEBUG(dbgs() << \"Considering loop '\" << L.getName()\n << \"' as innermost loop.\\n\");\n", "tests": [{"file": "llvm/test/Analysis/LoopCacheAnalysis/interchange-refcost-overflow.ll", "commands": ["opt < %s -passes='print' -disable-output 2>&1"], "tests": [{"test_name": "foo", "test_body": "@A = external local_unnamed_addr global [11 x [11 x [11 x i32]]], align 16\n\ndefine void @foo(i32 noundef %b) {\nentry:\n %0 = sext i32 %b to i64\n br label %outer.loop\n\nouter.loop: ; preds = %outer.loop.cleanup, %entry\n %indvars.iv = phi i64 [ %indvars.iv.next, %outer.loop.cleanup ], [ 0, %entry ]\n br label %inner.loop\n\nouter.loop.cleanup: ; preds = %inner.loop\n %indvars.iv.next = add nsw i64 %indvars.iv, %0\n br label %outer.loop\n\ninner.loop: ; preds = %inner.loop, %outer.loop\n %inner.iv = phi i64 [ 0, %outer.loop ], [ %add, %inner.loop ]\n %arrayidx3 = getelementptr inbounds [11 x [11 x [11 x i32]]], ptr @A, i64 0, i64 %indvars.iv, i64 %inner.iv, i64 %inner.iv\n store i32 0, ptr %arrayidx3, align 4\n %add = add nuw i64 %inner.iv, 2\n %cmp = icmp ult i64 %inner.iv, -5\n br i1 %cmp, label %inner.loop, label %outer.loop.cleanup\n}\n"}]}], "issue": {"title": "[LoopInterchange] Assertion `getActiveBits() <= 64 && \"Too many bits for uint64_t\"", "body": "With options \"-O3 --target=aarch64 func.cpp -mllvm -enable-loopinterchange\" and this input:\r\n```\r\nunsigned int A[11][11][11] __attribute__((aligned(16)));\r\nvoid a(int b) {\r\n for (int c;; c += b)\r\n for (long d = 0; d < -3ULL; d += 2ULL)\r\n A[c][d][d] = 0;\r\n}\r\n```\r\n\r\nLoop-interchange runs in an assert:\r\n\r\n```\r\nllvm-project/llvm/include/llvm/ADT/APInt.h:1501: uint64_t llvm::APInt::getZExtValue() const: Assertion `getActiveBits() <= 64 && \"Too many bits for uint64_t\"' failed.\r\n```\r\n\r\nIR reproducer, compile this with \"opt -passes=loop-interchange -S\":\r\n\r\n```\r\ntarget datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-Fn32\"\r\ntarget triple = \"aarch64-unknown-linux-gnu\"\r\n\r\n@A = local_unnamed_addr global [11 x [11 x [11 x i32]]] zeroinitializer, align 16\r\n\r\ndefine void @_Z1ai(i32 noundef %b) local_unnamed_addr #0 {\r\nentry:\r\n %0 = sext i32 %b to i64\r\n br label %for.cond\r\n\r\n; Loop:\r\nfor.cond: ; preds = %for.cond.cleanup, %entry\r\n %indvars.iv = phi i64 [ %indvars.iv.next, %for.cond.cleanup ], [ 0, %entry ]\r\n br label %for.body\r\n\r\nfor.cond.cleanup: ; preds = %for.body\r\n %indvars.iv.next = add nsw i64 %indvars.iv, %0\r\n br label %for.cond\r\n\r\nfor.body: ; preds = %for.cond, %for.body\r\n %d.010 = phi i64 [ 0, %for.cond ], [ %add, %for.body ]\r\n %arrayidx3 = getelementptr inbounds [11 x [11 x [11 x i32]]], ptr @A, i64 0, i64 %indvars.iv, i64 %d.010, i64 %d.010\r\n store i32 0, ptr %arrayidx3, align 4\r\n %add = add nuw i64 %d.010, 2\r\n %cmp = icmp ult i64 %d.010, -5\r\n br i1 %cmp, label %for.body, label %for.cond.cleanup\r\n}\r\n```", "author": "sjoerdmeijer", "labels": ["crash-on-valid", "llvm:analysis"], "comments": [{"author": "RKSimon", "body": "```\r\nopt: /root/llvm-project/llvm/include/llvm/ADT/APInt.h:1501: uint64_t llvm::APInt::getZExtValue() const: Assertion `getActiveBits() <= 64 && \"Too many bits for uint64_t\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=loop-interchange -debug \r\n1.\tRunning pass \"function(loop(loop-interchange))\" on module \"\"\r\n2.\tRunning pass \"loop(loop-interchange)\" on function \"_Z1ai\"\r\n #0 0x0000000004f7f528 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4f7f528)\r\n #1 0x0000000004f7cc9c SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f497b842520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x00007f497b8969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x00007f497b842476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x00007f497b8287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x00007f497b82871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x00007f497b839e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x00000000045677cc llvm::IndexedReference::computeRefCost(llvm::Loop const&, unsigned int) const (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x45677cc)\r\n #9 0x000000000456995f llvm::CacheCost::computeLoopCacheCost(llvm::Loop const&, llvm::SmallVector>, 8u>, 8u> const&) const (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x456995f)\r\n#10 0x000000000456c137 llvm::CacheCost::calculateCacheFootprint() (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x456c137)\r\n#11 0x0000000004573d5c llvm::CacheCost::getCacheCost(llvm::Loop&, llvm::LoopStandardAnalysisResults&, llvm::DependenceInfo&, std::optional) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4573d5c)\r\n#12 0x0000000003c71d66 llvm::LoopInterchangePass::run(llvm::LoopNest&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3c71d66)\r\n#13 0x0000000002eb0a6e llvm::detail::PassModel, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::run(llvm::LoopNest&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2eb0a6e)\r\n#14 0x0000000003c85255 llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::runWithLoopNestPasses(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3c85255)\r\n#15 0x0000000003c85ce3 llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3c85ce3)\r\n#16 0x0000000003c868b8 llvm::FunctionToLoopPassAdaptor::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x3c868b8)\r\n#17 0x0000000000df128e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdf128e)\r\n#18 0x0000000004d7c368 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d7c368)\r\n#19 0x0000000000df14be llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdf14be)\r\n#20 0x0000000004d7ad5e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d7ad5e)\r\n#21 0x0000000000df0f8e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xdf0f8e)\r\n#22 0x0000000004d78ec0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4d78ec0)\r\n#23 0x00000000008e7582 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e7582)\r\n#24 0x00000000008da45c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8da45c)\r\n#25 0x00007f497b829d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#26 0x00007f497b829e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#27 0x00000000008d1eee _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8d1eee)\r\n```\r\nWhich seems to be due to the getZExtValue() call inside IndexedReference::computeRefCost\r\nhttps://github.com/llvm/llvm-project/blob/b05c55472bf7cadcd0e4cb1a669b3474695b0524/llvm/lib/Analysis/LoopCacheAnalysis.cpp#L340:L343"}, {"author": "svs-quic", "body": "cc: @caojoshua "}, {"author": "sjoerdmeijer", "body": "I have a fix, will put it up for review tomorrow. "}]}, "verified": true} {"bug_id": "114947", "issue_url": "https://github.com/llvm/llvm-project/issues/114947", "bug_type": "miscompilation", "base_commit": "1e5bfac933ea90ec4361446398551dd6b967c67f", "knowledge_cutoff": "2024-11-05T07:42:44Z", "lit_test_dir": ["llvm/test/Transforms/SCCP"], "hints": {"fix_commit": "cd88bfcb5906049e1387b856fc7256e5fae22e5f", "components": ["ConstantFold"], "bug_location_lineno": {"llvm/lib/Analysis/ConstantFolding.cpp": [[1266, 1279], [1298, 1344]]}, "bug_location_funcname": {"llvm/lib/Analysis/ConstantFolding.cpp": ["llvm::ConstantFoldBinaryOpOperands", "llvm::ConstantFoldCompareInstOperands", "llvm::ConstantFoldFPInstOperands", "llvm::FlushFPConstant"]}}, "patch": "commit cd88bfcb5906049e1387b856fc7256e5fae22e5f\nAuthor: Matt Arsenault \nDate: Wed Nov 13 07:43:50 2024 -0800\n\n ConstantFolding: Do not fold fcmp of denormal without known mode (#115407)\n \n Fixes #114947\n\ndiff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp\nindex 88db315ffd0b..1971c28fc4c4 100644\n--- a/llvm/lib/Analysis/ConstantFolding.cpp\n+++ b/llvm/lib/Analysis/ConstantFolding.cpp\n@@ -1266,14 +1266,16 @@ Constant *llvm::ConstantFoldCompareInstOperands(\n return ConstantFoldCompareInstOperands(Predicate, Ops1, Ops0, DL, TLI);\n }\n \n- // Flush any denormal constant float input according to denormal handling\n- // mode.\n- Ops0 = FlushFPConstant(Ops0, I, /* IsOutput */ false);\n- if (!Ops0)\n- return nullptr;\n- Ops1 = FlushFPConstant(Ops1, I, /* IsOutput */ false);\n- if (!Ops1)\n- return nullptr;\n+ if (CmpInst::isFPPredicate(Predicate)) {\n+ // Flush any denormal constant float input according to denormal handling\n+ // mode.\n+ Ops0 = FlushFPConstant(Ops0, I, /*IsOutput=*/false);\n+ if (!Ops0)\n+ return nullptr;\n+ Ops1 = FlushFPConstant(Ops1, I, /*IsOutput=*/false);\n+ if (!Ops1)\n+ return nullptr;\n+ }\n \n return ConstantFoldCompareInstruction(Predicate, Ops0, Ops1);\n }\n@@ -1298,47 +1300,110 @@ Constant *llvm::ConstantFoldBinaryOpOperands(unsigned Opcode, Constant *LHS,\n return ConstantFoldBinaryInstruction(Opcode, LHS, RHS);\n }\n \n-Constant *llvm::FlushFPConstant(Constant *Operand, const Instruction *I,\n- bool IsOutput) {\n- if (!I || !I->getParent() || !I->getFunction())\n- return Operand;\n+static ConstantFP *flushDenormalConstant(Type *Ty, const APFloat &APF,\n+ DenormalMode::DenormalModeKind Mode) {\n+ switch (Mode) {\n+ case DenormalMode::Dynamic:\n+ return nullptr;\n+ case DenormalMode::IEEE:\n+ return ConstantFP::get(Ty->getContext(), APF);\n+ case DenormalMode::PreserveSign:\n+ return ConstantFP::get(\n+ Ty->getContext(),\n+ APFloat::getZero(APF.getSemantics(), APF.isNegative()));\n+ case DenormalMode::PositiveZero:\n+ return ConstantFP::get(Ty->getContext(),\n+ APFloat::getZero(APF.getSemantics(), false));\n+ default:\n+ break;\n+ }\n \n- ConstantFP *CFP = dyn_cast(Operand);\n- if (!CFP)\n- return Operand;\n+ llvm_unreachable(\"unknown denormal mode\");\n+}\n+\n+/// Return the denormal mode that can be assumed when executing a floating point\n+/// operation at \\p CtxI.\n+static DenormalMode getInstrDenormalMode(const Instruction *CtxI, Type *Ty) {\n+ if (!CtxI || !CtxI->getParent() || !CtxI->getFunction())\n+ return DenormalMode::getDynamic();\n+ return CtxI->getFunction()->getDenormalMode(Ty->getFltSemantics());\n+}\n \n+static ConstantFP *flushDenormalConstantFP(ConstantFP *CFP,\n+ const Instruction *Inst,\n+ bool IsOutput) {\n const APFloat &APF = CFP->getValueAPF();\n- // TODO: Should this canonicalize nans?\n if (!APF.isDenormal())\n- return Operand;\n+ return CFP;\n \n- Type *Ty = CFP->getType();\n- DenormalMode DenormMode =\n- I->getFunction()->getDenormalMode(Ty->getFltSemantics());\n- DenormalMode::DenormalModeKind Mode =\n- IsOutput ? DenormMode.Output : DenormMode.Input;\n- switch (Mode) {\n- default:\n- llvm_unreachable(\"unknown denormal mode\");\n- case DenormalMode::Dynamic:\n- return nullptr;\n- case DenormalMode::IEEE:\n+ DenormalMode Mode = getInstrDenormalMode(Inst, CFP->getType());\n+ return flushDenormalConstant(CFP->getType(), APF,\n+ IsOutput ? Mode.Output : Mode.Input);\n+}\n+\n+Constant *llvm::FlushFPConstant(Constant *Operand, const Instruction *Inst,\n+ bool IsOutput) {\n+ if (ConstantFP *CFP = dyn_cast(Operand))\n+ return flushDenormalConstantFP(CFP, Inst, IsOutput);\n+\n+ if (isa(Operand))\n return Operand;\n- case DenormalMode::PreserveSign:\n- if (APF.isDenormal()) {\n- return ConstantFP::get(\n- Ty->getContext(),\n- APFloat::getZero(Ty->getFltSemantics(), APF.isNegative()));\n+\n+ Type *Ty = Operand->getType();\n+ VectorType *VecTy = dyn_cast(Ty);\n+ if (VecTy) {\n+ if (auto *Splat = dyn_cast_or_null(Operand->getSplatValue())) {\n+ ConstantFP *Folded = flushDenormalConstantFP(Splat, Inst, IsOutput);\n+ if (!Folded)\n+ return nullptr;\n+ return ConstantVector::getSplat(VecTy->getElementCount(), Folded);\n }\n- return Operand;\n- case DenormalMode::PositiveZero:\n- if (APF.isDenormal()) {\n- return ConstantFP::get(Ty->getContext(),\n- APFloat::getZero(Ty->getFltSemantics(), false));\n+\n+ Ty = VecTy->getElementType();\n+ }\n+\n+ if (const auto *CV = dyn_cast(Operand)) {\n+ SmallVector NewElts;\n+ for (unsigned i = 0, e = CV->getNumOperands(); i != e; ++i) {\n+ Constant *Element = CV->getAggregateElement(i);\n+ if (isa(Element)) {\n+ NewElts.push_back(Element);\n+ continue;\n+ }\n+\n+ ConstantFP *CFP = dyn_cast(Element);\n+ if (!CFP)\n+ return nullptr;\n+\n+ ConstantFP *Folded = flushDenormalConstantFP(CFP, Inst, IsOutput);\n+ if (!Folded)\n+ return nullptr;\n+ NewElts.push_back(Folded);\n }\n- return Operand;\n+\n+ return ConstantVector::get(NewElts);\n+ }\n+\n+ if (const auto *CDV = dyn_cast(Operand)) {\n+ SmallVector NewElts;\n+ for (unsigned I = 0, E = CDV->getNumElements(); I < E; ++I) {\n+ const APFloat &Elt = CDV->getElementAsAPFloat(I);\n+ if (!Elt.isDenormal()) {\n+ NewElts.push_back(ConstantFP::get(Ty, Elt));\n+ } else {\n+ DenormalMode Mode = getInstrDenormalMode(Inst, Ty);\n+ ConstantFP *Folded =\n+ flushDenormalConstant(Ty, Elt, IsOutput ? Mode.Output : Mode.Input);\n+ if (!Folded)\n+ return nullptr;\n+ NewElts.push_back(Folded);\n+ }\n+ }\n+\n+ return ConstantVector::get(NewElts);\n }\n- return Operand;\n+\n+ return nullptr;\n }\n \n Constant *llvm::ConstantFoldFPInstOperands(unsigned Opcode, Constant *LHS,\n", "tests": [{"file": "llvm/test/Transforms/SCCP/no-fold-fcmp-dynamic-denormal-mode-issue114947.ll", "commands": ["opt -S -passes=ipsccp < %s"], "tests": [{"test_name": "no_fold_fcmp_denormal_double_ieee_dynamic_vector_nonsplat", "test_body": "define <2 x i1> @no_fold_fcmp_denormal_double_ieee_dynamic_vector_nonsplat() #0 {\n %cmp = fcmp une <2 x double> , zeroinitializer\n ret <2 x i1> %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "fold_fcmp_nondenormal_double_ieee_dynamic_vector_nonsplat_undef", "test_body": "define <3 x i1> @fold_fcmp_nondenormal_double_ieee_dynamic_vector_nonsplat_undef() #0 {\n %cmp = fcmp une <3 x double> , \n ret <3 x i1> %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "no_fold_fcmp_denormal_double_ieee_dynamic_denormal_undef", "test_body": "define i1 @no_fold_fcmp_denormal_double_ieee_dynamic_denormal_undef() #0 {\n %cmp = fcmp une double 0x8000000000000, undef\n ret i1 %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "fold_fcmp_nondenormal_double_ieee_dynamic_vector_nonsplat", "test_body": "define <2 x i1> @fold_fcmp_nondenormal_double_ieee_dynamic_vector_nonsplat() #0 {\n %cmp = fcmp une <2 x double> , \n ret <2 x i1> %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "fold_fcmp_nondenormal_double_ieee_dynamic_vector_splat", "test_body": "define <2 x i1> @fold_fcmp_nondenormal_double_ieee_dynamic_vector_splat() #0 {\n %cmp = fcmp une <2 x double> splat (double 2.000000e+00), zeroinitializer\n ret <2 x i1> %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "fold_fcmp_nondenormal_double_ieee_dynamic", "test_body": "define i1 @fold_fcmp_nondenormal_double_ieee_dynamic() #0 {\n %cmp = fcmp une double 2.000000e+00, 0.000000e+00\n ret i1 %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "fold_fcmp_nondenormal_double_ieee_dynamic_zero", "test_body": "define <2 x i1> @fold_fcmp_nondenormal_double_ieee_dynamic_zero() #0 {\n %cmp = fcmp une <2 x double> zeroinitializer, zeroinitializer\n ret <2 x i1> %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "no_fold_fcmp_denormal_double_ieee_dynamic_denormal_poison", "test_body": "define i1 @no_fold_fcmp_denormal_double_ieee_dynamic_denormal_poison() #0 {\n %cmp = fcmp une double 0x8000000000000, poison\n ret i1 %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "no_fold_fcmp_denormal_double_ieee_dynamic_vector_splat", "test_body": "define <2 x i1> @no_fold_fcmp_denormal_double_ieee_dynamic_vector_splat() #0 {\n %cmp = fcmp une <2 x double> splat (double 0x8000000000000), zeroinitializer\n ret <2 x i1> %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}, {"test_name": "no_fold_fcmp_denormal_double_ieee_dynamic", "test_body": "define i1 @no_fold_fcmp_denormal_double_ieee_dynamic() #0 {\n %cmp = fcmp une double 0x8000000000000, 0.000000e+00\n ret i1 %cmp\n}\n\nattributes #0 = { \"denormal-fp-math\"=\"ieee,dynamic\" }\n"}]}], "issue": {"title": "`fcmp` miscompiled to a constant", "body": "Alive2 report: https://alive2.llvm.org/ce/z/324E7x\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine i1 @fcmp_double_dynamic_dynamic.2() denormal-fp-math=dynamic,dynamic {\r\n#0:\r\n %cmp = fcmp une double 0x0008000000000000, -0.000000\r\n ret i1 %cmp\r\n}\r\n=>\r\ndefine i1 @fcmp_double_dynamic_dynamic.2() nofree noundef willreturn denormal-fp-math=dynamic,dynamic memory(none) {\r\n#0:\r\n ret i1 1\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Value mismatch\r\n\r\nExample:\r\n\r\nSource:\r\ni1 %cmp = #x0 (0)\r\n\r\nTarget:\r\nSource value: #x0 (0)\r\nTarget value: #x1 (1)\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```\r\n", "author": "bongjunj", "labels": ["miscompilation", "floating-point", "llvm:analysis"], "comments": [{"author": "arsenm", "body": "I don't see the issue? Those values are in fact not equal, so this is true. Alive bug? "}, {"author": "jcranmer-intel", "body": "it's comparing a denormal to 0, and the `denormal-fp-mode` is set to `dynamic`, so `fcmp` may or may not be flushing denormals to 0."}, {"author": "arsenm", "body": "InstCombine doesn't fold this, IPSCCP does "}, {"author": "nikic", "body": "I expect this is because we got through ValueLattice comparison, which does not have access to the denormal mode: https://github.com/llvm/llvm-project/blob/9501af5f92bb31fa9527d12a4b17b88d9fd651c4/llvm/lib/Analysis/ValueLattice.cpp#L28-L29\r\n\r\nReally, the constant folding implementation should be defaulting to the conservatively correct behavior if the denormal mode is unknown."}]}, "verified": true} {"bug_id": "121584", "issue_url": "https://github.com/llvm/llvm-project/issues/121584", "bug_type": "miscompilation", "base_commit": "4a7c0b8afe6bf616cd6bb4f13b5b706a43c10e74", "knowledge_cutoff": "2025-01-03T17:05:45Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "24c2ba07ce65a5bf7d1113e05c517169d950b663", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp": [[2066, 2079], [2095, 2101], [2108, 2114], [2131, 2137]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp": ["simplifyIRemMulShl"]}}, "patch": "commit 24c2ba07ce65a5bf7d1113e05c517169d950b663\nAuthor: Yingwei Zheng \nDate: Sun Jan 5 01:20:54 2025 +0800\n\n [InstCombine] Drop NSW when converting `shl X, BW - 1` back into mul (#121633)\n \n `X < bool {\n+ auto MatchShiftOrMulXC = [](Value *Op, Value *&V, APInt &C,\n+ bool &PreserveNSW) -> bool {\n const APInt *Tmp = nullptr;\n if ((!V && match(Op, m_Mul(m_Value(V), m_APInt(Tmp)))) ||\n (V && match(Op, m_Mul(m_Specific(V), m_APInt(Tmp)))))\n C = *Tmp;\n else if ((!V && match(Op, m_Shl(m_Value(V), m_APInt(Tmp)))) ||\n- (V && match(Op, m_Shl(m_Specific(V), m_APInt(Tmp)))))\n+ (V && match(Op, m_Shl(m_Specific(V), m_APInt(Tmp))))) {\n C = APInt(Tmp->getBitWidth(), 1) << *Tmp;\n+ // We cannot preserve NSW when shifting by BW - 1.\n+ PreserveNSW = Tmp->ult(Tmp->getBitWidth() - 1);\n+ }\n if (Tmp != nullptr)\n return true;\n \n@@ -2095,7 +2099,9 @@ static Instruction *simplifyIRemMulShl(BinaryOperator &I,\n return false;\n };\n \n- if (MatchShiftOrMulXC(Op0, X, Y) && MatchShiftOrMulXC(Op1, X, Z)) {\n+ bool Op0PreserveNSW = true, Op1PreserveNSW = true;\n+ if (MatchShiftOrMulXC(Op0, X, Y, Op0PreserveNSW) &&\n+ MatchShiftOrMulXC(Op1, X, Z, Op1PreserveNSW)) {\n // pass\n } else if (MatchShiftCX(Op0, Y, X) && MatchShiftCX(Op1, Z, X)) {\n ShiftByX = true;\n@@ -2108,7 +2114,7 @@ static Instruction *simplifyIRemMulShl(BinaryOperator &I,\n OverflowingBinaryOperator *BO0 = cast(Op0);\n // TODO: We may be able to deduce more about nsw/nuw of BO0/BO1 based on Y >=\n // Z or Z >= Y.\n- bool BO0HasNSW = BO0->hasNoSignedWrap();\n+ bool BO0HasNSW = Op0PreserveNSW && BO0->hasNoSignedWrap();\n bool BO0HasNUW = BO0->hasNoUnsignedWrap();\n bool BO0NoWrap = IsSRem ? BO0HasNSW : BO0HasNUW;\n \n@@ -2131,7 +2137,7 @@ static Instruction *simplifyIRemMulShl(BinaryOperator &I,\n };\n \n OverflowingBinaryOperator *BO1 = cast(Op1);\n- bool BO1HasNSW = BO1->hasNoSignedWrap();\n+ bool BO1HasNSW = Op1PreserveNSW && BO1->hasNoSignedWrap();\n bool BO1HasNUW = BO1->hasNoUnsignedWrap();\n bool BO1NoWrap = IsSRem ? BO1HasNSW : BO1HasNUW;\n // (rem (mul X, Y), (mul nuw/nsw X, Z))\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/rem-mul-shl.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "srem_XY_XZ_with_CY_gt_CZ_no_nuw_out", "test_body": "define <2 x i8> @srem_XY_XZ_with_CY_gt_CZ_no_nuw_out(<2 x i8> %X) {\n %BO0 = mul nsw <2 x i8> %X, splat (i8 10)\n %BO1 = shl nuw nsw <2 x i8> %X, splat (i8 3)\n %r = srem <2 x i8> %BO0, %BO1\n ret <2 x i8> %r\n}\n"}, {"test_name": "srem_XY_XZ_with_CY_gt_CZ_drop_nsw_commuted", "test_body": "define i8 @srem_XY_XZ_with_CY_gt_CZ_drop_nsw_commuted(i8 noundef %X) {\n %BO0 = mul nsw i8 %X, 127\n %BO1 = shl nsw i8 %X, 7\n %r = srem i8 %BO0, %BO1\n ret i8 %r\n}\n"}, {"test_name": "srem_XY_XZ_with_CY_gt_CZ_drop_nsw", "test_body": "define i8 @srem_XY_XZ_with_CY_gt_CZ_drop_nsw(i8 noundef %X) {\n %BO0 = mul nsw i8 %X, 127\n %BO1 = shl nsw i8 %X, 7\n %r = srem i8 %BO1, %BO0\n ret i8 %r\n}\n"}, {"test_name": "srem_XY_XZ_with_CY_gt_CZ_fail_missing_flag1", "test_body": "define i8 @srem_XY_XZ_with_CY_gt_CZ_fail_missing_flag1(i8 %X) {\n %BO0 = mul nuw nsw i8 %X, 10\n %BO1 = mul nuw i8 %X, 6\n %r = srem i8 %BO0, %BO1\n ret i8 %r\n}\n"}]}], "issue": {"title": "[InstCombine] Miscompilation in `simplifyIRemMulShl`", "body": "Alive2: https://alive2.llvm.org/ce/z/rdeehL\n```\n\n----------------------------------------\ndefine i8 @srem_XY_XZ_with_CY_gt_CZ_no_nuw_out1(i8 noundef %X) {\n#0:\n %BO0 = mul nsw i8 noundef %X, 127\n %BO1 = shl nsw i8 noundef %X, 7\n %r = srem i8 %BO1, %BO0\n ret i8 %r\n}\n=>\ndefine i8 @srem_XY_XZ_with_CY_gt_CZ_no_nuw_out1(i8 noundef %X) {\n#0:\n %r = sub nsw i8 0, noundef %X\n ret i8 %r\n}\nTransformation doesn't verify!\n\nERROR: Value mismatch\n\nExample:\ni8 noundef %X = #xff (255, -1)\n\nSource:\ni8 %BO0 = #x81 (129, -127)\ni8 %BO1 = #x80 (128, -128)\ni8 %r = #xff (255, -1)\n\nTarget:\ni8 %r = #x01 (1)\nSource value: #xff (255, -1)\nTarget value: #x01 (1)\n\nSummary:\n 0 correct transformations\n 1 incorrect transformations\n 0 failed-to-prove transformations\n 0 Alive2 errors\n```", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "119173", "issue_url": "https://github.com/llvm/llvm-project/issues/119173", "bug_type": "miscompilation", "base_commit": "ab77db03ce28e86a61010e51ea13796ea09efc46", "knowledge_cutoff": "2024-12-09T07:19:06Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "30f3752e54fa7cd595a434a985efbe9a7abe9b65", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[6637, 6644], [8588, 8593]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["LoopVectorizationCostModel::getInstructionCost", "VPRecipeBuilder::tryToWiden"]}}, "patch": "commit 30f3752e54fa7cd595a434a985efbe9a7abe9b65\nAuthor: Florian Hahn \nDate: Mon Feb 3 17:01:02 2025 +0000\n\n [VPlan] Only use SCEV for live-ins in tryToWiden. (#125436)\n \n Replacing a recipe with a live-in may not be correct in all cases,\n e.g. when replacing recipes involving header-phi recipes, like\n reductions.\n \n For now, only use SCEV to simplify live-ins.\n \n More powerful input simplification can be built in top of\n https://github.com/llvm/llvm-project/pull/124432 in the future.\n \n \n Fixes https://github.com/llvm/llvm-project/issues/119173.\n Fixes https://github.com/llvm/llvm-project/issues/125374.\n \n PR: https://github.com/llvm/llvm-project/pull/125436\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex cbb9960959f2..ce66350669d5 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -6637,8 +6637,10 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I,\n // fold away. We can generalize this for all operations using the notion\n // of neutral elements. (TODO)\n if (I->getOpcode() == Instruction::Mul &&\n- (PSE.getSCEV(I->getOperand(0))->isOne() ||\n- PSE.getSCEV(I->getOperand(1))->isOne()))\n+ ((TheLoop->isLoopInvariant(I->getOperand(0)) &&\n+ PSE.getSCEV(I->getOperand(0))->isOne()) ||\n+ (TheLoop->isLoopInvariant(I->getOperand(1)) &&\n+ PSE.getSCEV(I->getOperand(1))->isOne())))\n return 0;\n \n // Detect reduction patterns\n@@ -8588,6 +8590,8 @@ VPWidenRecipe *VPRecipeBuilder::tryToWiden(Instruction *I,\n // to replace operands with constants.\n ScalarEvolution &SE = *PSE.getSE();\n auto GetConstantViaSCEV = [this, &SE](VPValue *Op) {\n+ if (!Op->isLiveIn())\n+ return Op;\n Value *V = Op->getUnderlyingValue();\n if (isa(V) || !SE.isSCEVable(V->getType()))\n return Op;\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/AArch64/mul-simplification.ll", "commands": ["opt -p loop-vectorize -S %s"], "tests": [{"test_name": "mul_select_operand_known_1_via_scev", "test_body": "target triple = \"arm64-apple-macosx\"\n\ndefine i64 @mul_select_operand_known_1_via_scev() {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %red = phi i64 [ 12, %entry ], [ %red.next, %loop ]\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %0 = and i32 %iv, 1\n %cmp1.i = icmp eq i32 %0, 0\n %narrow.i = select i1 %cmp1.i, i32 1, i32 %iv\n %mul = zext nneg i32 %narrow.i to i64\n %red.next = mul nsw i64 %red, %mul\n %iv.next = add nuw nsw i32 %iv, 1\n %ec = icmp eq i32 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n %res = phi i64 [ %red.next, %loop ]\n ret i64 %res\n}\n", "additional_args": "-src-unroll=4 -tgt-unroll=4"}, {"test_name": "pr125374", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i32 @main() {\nentry:\n br label %for.body.i.i\n\nfor.body.i.i: ; preds = %for.body.i.i, %entry\n %indvars.iv.i.i = phi i64 [ 0, %entry ], [ %indvars.iv.next.i.i, %for.body.i.i ]\n %conv678.i.i = phi i8 [ 1, %entry ], [ %add.i.i.i, %for.body.i.i ]\n %add.i.i.i = add i8 0, %conv678.i.i\n %indvars.iv.next.i.i = add i64 %indvars.iv.i.i, 1\n %exitcond.not.i.i = icmp eq i64 %indvars.iv.i.i, 1\n br i1 %exitcond.not.i.i, label %o.exit, label %for.body.i.i\n\no.exit: ; preds = %for.body.i.i\n %conv6.i.i = zext i8 %add.i.i.i to i32\n ret i32 %conv6.i.i\n}\n", "additional_args": "-src-unroll=2 -tgt-unroll=2"}]}], "issue": {"title": "[clang] Miscompilation at -O3", "body": "This code prints 15 at -O3, but prints 5 at -O0/1/2:\n```c\nint printf(const char *, ...);\nstatic int a[] = {4294967295, 5};\nint b, c;\nint main() {\n a[1] = b = 5;\n unsigned d = -13;\n for (; d >= 8; d = a[0] + d + 6) {\n int *e = &b;\n *e = a[0] - -1 + b;\n }\n a[c];\n printf(\"%d\\n\", b);\n}\n```\n\nCompiler Explorer: https://godbolt.org/z/4xdPqcjPb\n\nBisected to https://github.com/llvm/llvm-project/commit/6d6eea92e36c301e34a7ec11b2a40e3080f79f53, which was committed by @fhahn ", "author": "cardigan1008", "labels": ["miscompilation", "vectorizers"], "comments": [{"author": "hstk30-hw", "body": "Don't ignore the warning.\n\n```\n:2:19: warning: implicit conversion from 'long' to 'int' changes value from 4294967294 to -2 [-Wconstant-conversion]\n 2 | static int a[] = {4294967294, 5};\n | ~^~~~~~~~~~\n```"}, {"author": "antoniofrighetto", "body": "Please leave the issue opened, as it's implementation-defined behaviour, not undefined behaviour."}, {"author": "antoniofrighetto", "body": "Reduced to:\n```llvm\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine noundef i32 @src() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %entry, %for.body\n %d.06 = phi i32 [ -13, %entry ], [ %add2.reass, %for.body ]\n %add45 = phi i32 [ 5, %entry ], [ %add, %for.body ]\n %add = add i32 0, %add45\n %add2.reass = add i32 %d.06, 5\n %cmp = icmp ugt i32 %add2.reass, 7\n br i1 %cmp, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret i32 %add\n}\n```\n\n@fhahn I think the issue fundamentally here lies in the fact that we are incorrectly widening the constant 5 (retrieved from SCEV) to `<5, 5, 5, 5>`, when in fact it should be `<5, 0, 0, 0>` (VPlan value: `WIDEN-REDUCTION-PHI ir<%add45> = phi ir<5>`). The VPBasicBlock for the miscompiled basic block is the following one:\n```\nvector.body:\n EMIT vp<%4> = CANONICAL-INDUCTION ir<0>, vp<%7>\n WIDEN-INDUCTION %d.06 = phi -13, %add2.reass, ir<5>, vp<%0>\n WIDEN-REDUCTION-PHI ir<%add45> = phi ir<5>\n EMIT vp<%5> = WIDEN-CANONICAL-INDUCTION vp<%4>\n EMIT vp<%6> = icmp ule vp<%5>, vp<%3>\n```\nI'm not that familiar with VPlan, though I feel like that either we are not taking into account the active lane, or we should have not emitted a widen instruction in the first place.\n\n(Godbolt: https://llvm.godbolt.org/z/4Wz46ceh1)"}, {"author": "antoniofrighetto", "body": "@fhahn Would you be willing to have a look at it? Happy to help with any additional triage, if needed."}, {"author": "fhahn", "body": "Yep let me take a look!"}]}, "verified": true} {"bug_id": "109376", "issue_url": "https://github.com/llvm/llvm-project/issues/109376", "bug_type": "crash", "base_commit": "1c47fa9b620d0abb280647b4f361ada43784d00e", "knowledge_cutoff": "2024-09-20T05:30:57Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "3db0f8c895d4e814a18b754f9afbb1e03bd839a5", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[18668, 18673]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["tryToReduce"]}}, "patch": "commit 3db0f8c895d4e814a18b754f9afbb1e03bd839a5\nAuthor: Alexey Bataev \nDate: Mon Sep 23 10:49:54 2024 -0700\n\n [SLP]Update TrackedToOrig mappings after reduction vectorization\n \n Need to update mappings in TrackedToOrig to correctly provide mapping\n between updated reduced value after vectorization and its original\n value, otherwise the compiler might miss this update and it may cause\n compiler crash later, when it tries to find the original instruction\n mapping for the updated value.\n \n Fixes https://github.com/llvm/llvm-project/issues/109376\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex a88702b81096..04b8fc09a724 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -18668,6 +18668,14 @@ public:\n // Vectorize a tree.\n Value *VectorizedRoot =\n V.vectorizeTree(LocalExternallyUsedValues, InsertPt);\n+ // Update TrackedToOrig mapping, since the tracked values might be\n+ // updated.\n+ for (Value *RdxVal : Candidates) {\n+ Value *OrigVal = TrackedToOrig.at(RdxVal);\n+ Value *TransformedRdxVal = TrackedVals.at(OrigVal);\n+ if (TransformedRdxVal != RdxVal)\n+ TrackedToOrig.try_emplace(TransformedRdxVal, OrigVal);\n+ }\n \n Builder.SetInsertPoint(InsertPt);\n \n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/reduced-value-vectorized-later.ll", "commands": ["opt -S -passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define i16 @test() {\nentry:\n %subi = add i16 0, 0\n %sub40.i = add i16 %subi, 0\n %sub41.i = add i16 %subi, 0\n %sub42.i = add i16 %subi, 0\n %sub43.i = add i16 %subi, 0\n %sub44.i = add i16 %subi, 0\n %sub45.i = add i16 %subi, 0\n %sub46.i = add i16 0, 0\n %sub47.i = add i16 0, 0\n %sub48.i = add i16 0, 0\n %sub49.i = add i16 0, 0\n %or40.i = or i16 %sub40.i, %sub41.i\n %or41.i = or i16 %or40.i, %sub42.i\n %or42.i = or i16 %or41.i, %sub43.i\n %or43.i = or i16 %or42.i, %sub44.i\n %or44.i = or i16 %or43.i, %sub45.i\n %or45.i = or i16 %or44.i, %sub46.i\n %or46.i = or i16 %or45.i, %sub47.i\n %or47.i = or i16 %or46.i, %sub48.i\n %or48.i = or i16 %or47.i, %sub49.i\n %or50.i = or i16 %or48.i, %subi\n %subii = add i16 0, 0\n %subi16.i = add i16 %subii, 0\n %subi17.i = add i16 %subii, 0\n %0 = or i16 %subi16.i, %subi17.i\n %1 = or i16 %0, %or50.i\n ret i16 %1\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Assertion `Iter != this->end() && \"DenseMap::at failed due to a missing key\"' failed.", "body": "Reproducer: https://godbolt.org/z/v5bqhPWqb\r\n```\r\n; bin/opt -passes=slp-vectorizer reduced.ll -S\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i16 @test() {\r\nentry:\r\n %sub.i487.i.reass.i = add i16 0, 0\r\n %sub.i487.i.reass.40.i = add i16 %sub.i487.i.reass.i, 0\r\n %sub.i487.i.reass.41.i = add i16 %sub.i487.i.reass.i, 0\r\n %sub.i487.i.reass.42.i = add i16 %sub.i487.i.reass.i, 0\r\n %sub.i487.i.reass.43.i = add i16 %sub.i487.i.reass.i, 0\r\n %sub.i487.i.reass.44.i = add i16 %sub.i487.i.reass.i, 0\r\n %sub.i487.i.reass.45.i = add i16 %sub.i487.i.reass.i, 0\r\n %sub.i487.i.reass.46.i = add i16 0, 0\r\n %sub.i487.i.reass.47.i = add i16 0, 0\r\n %sub.i487.i.reass.48.i = add i16 0, 0\r\n %sub.i487.i.reass.49.i = add i16 0, 0\r\n %or374461.i.40.i = or i16 %sub.i487.i.reass.40.i, %sub.i487.i.reass.41.i\r\n %or374461.i.41.i = or i16 %or374461.i.40.i, %sub.i487.i.reass.42.i\r\n %or374461.i.42.i = or i16 %or374461.i.41.i, %sub.i487.i.reass.43.i\r\n %or374461.i.43.i = or i16 %or374461.i.42.i, %sub.i487.i.reass.44.i\r\n %or374461.i.44.i = or i16 %or374461.i.43.i, %sub.i487.i.reass.45.i\r\n %or374461.i.45.i = or i16 %or374461.i.44.i, %sub.i487.i.reass.46.i\r\n %or374461.i.46.i = or i16 %or374461.i.45.i, %sub.i487.i.reass.47.i\r\n %or374461.i.47.i = or i16 %or374461.i.46.i, %sub.i487.i.reass.48.i\r\n %or374461.i.48.i = or i16 %or374461.i.47.i, %sub.i487.i.reass.49.i\r\n %or374461.i.50.i = or i16 %or374461.i.48.i, %sub.i487.i.reass.i\r\n %sub.i487.i96.reass.i = add i16 0, 0\r\n %sub.i487.i96.reass.16.i = add i16 %sub.i487.i96.reass.i, 0\r\n %sub.i487.i96.reass.17.i = add i16 %sub.i487.i96.reass.i, 0\r\n %0 = or i16 %sub.i487.i96.reass.16.i, %sub.i487.i96.reass.17.i\r\n %1 = or i16 %0, %or374461.i.50.i\r\n ret i16 %1\r\n}\r\n```\r\n```\r\nopt: /root/llvm-project/llvm/include/llvm/ADT/DenseMap.h:204: const ValueT& llvm::DenseMapBase::at(llvm::DenseMapBase::const_arg_type_t) const [with DerivedT = llvm::DenseMap; KeyT = llvm::Value*; ValueT = llvm::Value*; KeyInfoT = llvm::DenseMapInfo; BucketT = llvm::detail::DenseMapPair; llvm::DenseMapBase::const_arg_type_t = const llvm::Value*]: Assertion `Iter != this->end() && \"DenseMap::at failed due to a missing key\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \r\n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\r\n2.\tRunning pass \"slp-vectorizer\" on function \"test\"\r\n #0 0x0000000005035d68 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x5035d68)\r\n #1 0x00000000050334ac SignalHandler(int) Signals.cpp:0:0\r\n #2 0x0000707bdcc42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #3 0x0000707bdcc969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\r\n #4 0x0000707bdcc42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #5 0x0000707bdcc287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #6 0x0000707bdcc2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #7 0x0000707bdcc39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #8 0x000000000409326e llvm::DenseMapBase, llvm::detail::DenseMapPair>, llvm::Value*, llvm::Value*, llvm::DenseMapInfo, llvm::detail::DenseMapPair>::at(llvm::Value const*) const (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x409326e)\r\n #9 0x000000000413473f (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#10 0x00000000041370d1 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41370d1)\r\n#11 0x000000000413b996 llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*) (.constprop.0) SLPVectorizer.cpp:0:0\r\n#12 0x000000000413ff36 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x413ff36)\r\n#13 0x0000000004146401 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#14 0x000000000414708b llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x414708b)\r\n#15 0x0000000002f1a1ae llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2f1a1ae)\r\n#16 0x0000000004e38158 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e38158)\r\n#17 0x0000000000e0b02e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe0b02e)\r\n#18 0x0000000004e36b7e llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e36b7e)\r\n#19 0x0000000000e0a83e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe0a83e)\r\n#20 0x0000000004e365b0 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4e365b0)\r\n#21 0x00000000008feef2 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8feef2)\r\n#22 0x00000000008f1d7c optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8f1d7c)\r\n#23 0x0000707bdcc29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#24 0x0000707bdcc29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#25 0x00000000008e984e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x8e984e)\r\nProgram terminated with signal: SIGSEGV\r\nCompiler returned: 139\r\n```\r\nllvm version: e8a739062490ce1bb30974f216ba187cd7b5a2a4", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "104637", "issue_url": "https://github.com/llvm/llvm-project/issues/104637", "bug_type": "crash", "base_commit": "13779ec29ed724666407bd60f8c4cc8228107ea5", "knowledge_cutoff": "2024-08-16T19:54:48Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "4a0bbbcbcf6ebc87e794e7b86b9f4651bffcd806", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[7144, 7149]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::buildTree_rec"]}}, "patch": "commit 4a0bbbcbcf6ebc87e794e7b86b9f4651bffcd806\nAuthor: Alexey Bataev \nDate: Fri Aug 16 13:24:03 2024 -0700\n\n [SLP]Fix PR104637: do not create new nodes for fully overlapped non-schedulable nodes\n \n If the scalars do not require scheduling and were already vectorized,\n but in the different order, compiler still tries to create the new node.\n It may cause the compiler crash for the gathered operands. Instead need\n to consider such nodes as full overlap and just reshuffle vectorized\n node.\n \n Fixes https://github.com/llvm/llvm-project/issues/104637\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 667d0df781aa..9ecd8160a978 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -7144,6 +7144,21 @@ void BoUpSLP::buildTree_rec(ArrayRef VL, unsigned Depth,\n ReuseShuffleIndices);\n return;\n }\n+ SmallPtrSet Nodes;\n+ Nodes.insert(getTreeEntry(S.OpValue));\n+ for (const TreeEntry *E : MultiNodeScalars.lookup(S.OpValue))\n+ Nodes.insert(E);\n+ SmallPtrSet Values(VL.begin(), VL.end());\n+ if (any_of(Nodes, [&](const TreeEntry *E) {\n+ return all_of(E->Scalars,\n+ [&](Value *V) { return Values.contains(V); });\n+ })) {\n+ LLVM_DEBUG(dbgs() << \"SLP: Gathering due to full overlap.\\n\");\n+ if (TryToFindDuplicates(S))\n+ newTreeEntry(VL, std::nullopt /*not vectorized*/, S, UserTreeIdx,\n+ ReuseShuffleIndices);\n+ return;\n+ }\n } else {\n // Record the reuse of the tree node. FIXME, currently this is only used\n // to properly draw the graph rather than for the actual vectorization.\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/full-non-schedulable-overlap.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define void @test(double %v) {\nentry:\n %mul.3 = fmul double 0.000000e+00, %v\n %mul.4 = fmul double 0.000000e+00, 0.000000e+00\n br label %loop\n\nloop: ; preds = %loop, %entry\n %t48.0 = phi double [ 0.000000e+00, %entry ], [ %5, %loop ]\n %t50.02 = phi double [ 0.000000e+00, %entry ], [ %3, %loop ]\n %t52.0 = phi double [ 0.000000e+00, %entry ], [ %7, %loop ]\n %0 = fmul double %t52.0, %mul.3\n %1 = fmul double %t48.0, %mul.4\n %2 = fadd double %1, %0\n %3 = fadd double %2, %v\n %4 = fmul double 0.000000e+00, %mul.3\n %5 = fadd double 0.000000e+00, %4\n %6 = fmul double 0.000000e+00, %mul.4\n %7 = fadd double 0.000000e+00, %6\n br label %loop\n}\n"}]}], "issue": {"title": "SLP Vectorizer crash due to commit ca654acc16c43191228eadfec8f7241dca10b0c3", "body": "ca654acc16c43191228eadfec8f7241dca10b0c3 change in SLP Vectorizer causes a crash for this test case:\r\n\r\n```\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @a_(double %c__fetch.3) {\r\nalloca_0:\r\n %mul.3 = fmul double 0.000000e+00, %c__fetch.3\r\n %mul.4 = fmul double 0.000000e+00, 0.000000e+00\r\n br label %loop.69\r\n\r\nloop.69: ; preds = %loop.69, %alloca_0\r\n %t48.0 = phi double [ 0.000000e+00, %alloca_0 ], [ %5, %loop.69 ]\r\n %t50.0 = phi double [ 0.000000e+00, %alloca_0 ], [ %3, %loop.69 ]\r\n %t52.0 = phi double [ 0.000000e+00, %alloca_0 ], [ %7, %loop.69 ]\r\n %0 = fmul double %t52.0, %mul.3\r\n %1 = fmul double %t48.0, %mul.4\r\n %2 = fadd double %1, %0\r\n %3 = fadd double %2, 0.000000e+00\r\n %4 = fmul double 0.000000e+00, %mul.3\r\n %5 = fadd double 0.000000e+00, %4\r\n %6 = fmul double 0.000000e+00, %mul.4\r\n %7 = fadd double 0.000000e+00, %6\r\n br label %loop.69\r\n}\r\n```\r\n\r\nAfter slp-vectorizer pass IR is being transformed to as incorrect one:\r\n```\r\ndefine void @a_(double %c__fetch.3) {\r\nalloca_0:\r\n %0 = shufflevector <2 x double> %2, <2 x double> , <2 x i32> \r\n %1 = fmul <2 x double> zeroinitializer, %0\r\n %2 = insertelement <2 x double> , double %c__fetch.3, i32 0\r\n %3 = fmul <2 x double> zeroinitializer, %2\r\n br label %loop.69\r\n\r\nloop.69: ; preds = %loop.69, %alloca_0\r\n %t50.0 = phi double [ 0.000000e+00, %alloca_0 ], [ %9, %loop.69 ]\r\n %4 = phi <2 x double> [ zeroinitializer, %alloca_0 ], [ %11, %loop.69 ]\r\n %5 = fmul <2 x double> %4, %1\r\n %6 = extractelement <2 x double> %5, i32 0\r\n %7 = extractelement <2 x double> %5, i32 1\r\n %8 = fadd double %6, %7\r\n %9 = fadd double %8, 0.000000e+00\r\n %10 = fmul <2 x double> zeroinitializer, %3\r\n %11 = fadd <2 x double> zeroinitializer, %10\r\n br label %loop.69\r\n}\r\n```\r\n\r\nError message:\r\n\r\n```\r\nInstruction does not dominate all uses!\r\n %2 = insertelement <2 x double> , double %c__fetch.3, i32 0\r\n %0 = shufflevector <2 x double> %2, <2 x double> , <2 x i32> \r\nLLVM ERROR: Broken module found, compilation aborted!\r\n```\r\n\r\nReverting the mentioned commit makes this test pass. @alexey-bataev could you please take a look?\r\n", "author": "ivankelarev", "labels": ["llvm:SLPVectorizer", "crash"], "comments": [{"author": "ivankelarev", "body": "Update: It appears that the mentioned commit merely obscures a more fundamental issue. The following test continues to fail even after the commit has been reverted:\r\n\r\n```\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @a_(double %c__fetch.3) {\r\nalloca_0:\r\n %mul.3 = fmul double 0.000000e+00, %c__fetch.3\r\n %mul.4 = fmul double 0.000000e+00, 0.000000e+00\r\n br label %loop.69\r\n\r\nloop.69: ; preds = %loop.69, %alloca_0\r\n %t48.0 = phi double [ 0.000000e+00, %alloca_0 ], [ %5, %loop.69 ]\r\n %t50.02 = phi double [ 0.000000e+00, %alloca_0 ], [ %3, %loop.69 ]\r\n %t52.0 = phi double [ 0.000000e+00, %alloca_0 ], [ %7, %loop.69 ]\r\n %0 = fmul double %t52.0, %mul.3\r\n %1 = fmul double %t48.0, %mul.4\r\n %2 = fadd double %1, %0\r\n %3 = fadd double %2, %c__fetch.3\r\n %4 = fmul double 0.000000e+00, %mul.3\r\n %5 = fadd double 0.000000e+00, %4\r\n %6 = fmul double 0.000000e+00, %mul.4\r\n %7 = fadd double 0.000000e+00, %6\r\n br label %loop.69\r\n}\r\n```"}, {"author": "ivankelarev", "body": "It appears that the actual problematic commit is the following:\r\n\r\n```\r\ncommit 66775f8ccdcc8264ef349518e1c59d96d4227823\r\nAuthor: Alexey Bataev \r\nDate: Mon Oct 16 12:50:29 2023 -0700\r\n\r\n [SLP]Fix PR69196: Instruction does not dominate all uses\r\n \r\n During emission of the postponed gathers, need to insert them before\r\n user instruction to avoid use before definition crash.\r\n```\r\nFor compiler builds prior to this change, the test passes, and the correct code is generated as follows:\r\n```\r\ndefine void @a_(double %c__fetch.3) {\r\nalloca_0:\r\n %0 = insertelement <2 x double> , double %c__fetch.3, i32 0\r\n %1 = fmul <2 x double> zeroinitializer, %0\r\n %2 = shufflevector <2 x double> %0, <2 x double> , <2 x i32> \r\n %3 = fmul <2 x double> zeroinitializer, %2\r\n br label %loop.69\r\n\r\nloop.69: ; preds = %loop.69, %alloca_0\r\n %t50.02 = phi double [ 0.000000e+00, %alloca_0 ], [ %9, %loop.69 ]\r\n %4 = phi <2 x double> [ zeroinitializer, %alloca_0 ], [ %11, %loop.69 ]\r\n %5 = fmul <2 x double> %4, %3\r\n %6 = extractelement <2 x double> %5, i32 0\r\n %7 = extractelement <2 x double> %5, i32 1\r\n %8 = fadd double %6, %7\r\n %9 = fadd double %8, %c__fetch.3\r\n %10 = fmul <2 x double> zeroinitializer, %1\r\n %11 = fadd <2 x double> zeroinitializer, %10\r\n br label %loop.69\r\n}\r\n```\r\n\r\nAlthough the commit may seem quite old, there is a test case for which it causes incorrect code generation. Alexey, could you please revisit this change and address the code generation issue for the provided test case?"}, {"author": "alexey-bataev", "body": "> It appears that the actual problematic commit is the following:\r\n> \r\n> ```\r\n> commit 66775f8ccdcc8264ef349518e1c59d96d4227823\r\n> Author: Alexey Bataev \r\n> Date: Mon Oct 16 12:50:29 2023 -0700\r\n> \r\n> [SLP]Fix PR69196: Instruction does not dominate all uses\r\n> \r\n> During emission of the postponed gathers, need to insert them before\r\n> user instruction to avoid use before definition crash.\r\n> ```\r\n> \r\n> For compiler builds prior to this change, the test passes, and the correct code is generated as follows:\r\n> \r\n> ```\r\n> define void @a_(double %c__fetch.3) {\r\n> alloca_0:\r\n> %0 = insertelement <2 x double> , double %c__fetch.3, i32 0\r\n> %1 = fmul <2 x double> zeroinitializer, %0\r\n> %2 = shufflevector <2 x double> %0, <2 x double> , <2 x i32> \r\n> %3 = fmul <2 x double> zeroinitializer, %2\r\n> br label %loop.69\r\n> \r\n> loop.69: ; preds = %loop.69, %alloca_0\r\n> %t50.02 = phi double [ 0.000000e+00, %alloca_0 ], [ %9, %loop.69 ]\r\n> %4 = phi <2 x double> [ zeroinitializer, %alloca_0 ], [ %11, %loop.69 ]\r\n> %5 = fmul <2 x double> %4, %3\r\n> %6 = extractelement <2 x double> %5, i32 0\r\n> %7 = extractelement <2 x double> %5, i32 1\r\n> %8 = fadd double %6, %7\r\n> %9 = fadd double %8, %c__fetch.3\r\n> %10 = fmul <2 x double> zeroinitializer, %1\r\n> %11 = fadd <2 x double> zeroinitializer, %10\r\n> br label %loop.69\r\n> }\r\n> ```\r\n> \r\n> Although the commit may seem quite old, there is a test case for which it causes incorrect code generation. Alexey, could you please revisit this change and address the code generation issue for the provided test case?\r\n\r\nThe actual problem is not in this commit, the fix is almost ready"}]}, "verified": true} {"bug_id": "124267", "issue_url": "https://github.com/llvm/llvm-project/issues/124267", "bug_type": "miscompilation", "base_commit": "9cf52fe1f94fdcd8e27c76f7d33a80eeb2075833", "knowledge_cutoff": "2025-01-24T13:05:38Z", "lit_test_dir": ["llvm/test/Analysis/ValueTracking"], "hints": {"fix_commit": "c546b5317c518987a5f45dd4c4d25321a955c758", "components": ["ValueTracking"], "bug_location_lineno": {"llvm/lib/Analysis/ValueTracking.cpp": [[9494, 9500], [9504, 9510]]}, "bug_location_funcname": {"llvm/lib/Analysis/ValueTracking.cpp": ["isImpliedCondICmps"]}}, "patch": "commit c546b5317c518987a5f45dd4c4d25321a955c758\nAuthor: DianQK \nDate: Fri Jan 24 23:02:50 2025 +0800\n\n [ValueTracking] Pass changed predicate `SignedLPred` to `isImpliedByMatchingCmp` (#124271)\n \n Fixes #124267.\n \n Since we are using the new predicate, we should also update the\n parameters of `isImpliedByMatchingCmp`.\n\ndiff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp\nindex 264fedd6b66b..eba728c7c8c3 100644\n--- a/llvm/lib/Analysis/ValueTracking.cpp\n+++ b/llvm/lib/Analysis/ValueTracking.cpp\n@@ -9494,7 +9494,7 @@ isImpliedCondICmps(const ICmpInst *LHS, CmpPredicate RPred, const Value *R0,\n SignedLPred == ICmpInst::ICMP_SGE) &&\n match(R0, m_NSWSub(m_Specific(L0), m_Specific(L1)))) {\n if (match(R1, m_NonPositive()) &&\n- ICmpInst::isImpliedByMatchingCmp(LPred, RPred) == false)\n+ ICmpInst::isImpliedByMatchingCmp(SignedLPred, RPred) == false)\n return false;\n }\n \n@@ -9504,7 +9504,7 @@ isImpliedCondICmps(const ICmpInst *LHS, CmpPredicate RPred, const Value *R0,\n SignedLPred == ICmpInst::ICMP_SLE) &&\n match(R0, m_NSWSub(m_Specific(L0), m_Specific(L1)))) {\n if (match(R1, m_NonNegative()) &&\n- ICmpInst::isImpliedByMatchingCmp(LPred, RPred) == true)\n+ ICmpInst::isImpliedByMatchingCmp(SignedLPred, RPred) == true)\n return true;\n }\n \n", "tests": [{"file": "llvm/test/Analysis/ValueTracking/implied-condition-samesign.ll", "commands": ["opt -passes=instsimplify -S %s"], "tests": [{"test_name": "gt_sub_nsw_ult", "test_body": "define i1 @gt_sub_nsw_ult(i8 %L0, i8 %L1, i1 %V) {\n %LHS = icmp samesign ugt i8 %L0, %L1\n br i1 %LHS, label %LHS_true, label %LHS_false\n\nLHS_true: ; preds = %0\n %R0 = sub nsw i8 %L0, %L1\n %RHS = icmp ult i8 %R0, -1\n ret i1 %RHS\n\nLHS_false: ; preds = %0\n ret i1 %V\n}\n"}, {"test_name": "lt_sub_nsw_ult", "test_body": "define i1 @lt_sub_nsw_ult(i8 %L0, i8 %L1, i1 %V) {\n %LHS = icmp samesign ult i8 %L0, %L1\n br i1 %LHS, label %LHS_true, label %LHS_false\n\nLHS_true: ; preds = %0\n %R0 = sub nsw i8 %L0, %L1\n %RHS = icmp ult i8 %R0, 1\n ret i1 %RHS\n\nLHS_false: ; preds = %0\n ret i1 %V\n}\n"}]}], "issue": {"title": "[ValueTracking] miscompile in samesign compare", "body": "#122590 and #122700 miscompile the following code:\n\n```llvm\ndefine i1 @src(i8 %L0, i8 %L1, i1 %V) {\n %LHS = icmp samesign ugt i8 %L0, %L1\n %R0 = sub nsw i8 %L0, %L1\n %RHS = icmp ult i8 %R0, -1\n br i1 %LHS, label %LHS_true, label %LHS_false\n\nLHS_true:\n br label %EXIT\n\nLHS_false:\n br label %EXIT\n\nEXIT:\n %RET = phi i1 [ %RHS, %LHS_true ], [ %V, %LHS_false ]\n ret i1 %RET\n}\n```\n\n`%RHS` should be 1: https://alive2.llvm.org/ce/z/RscAzy.", "author": "DianQK", "labels": ["miscompilation", "llvm:analysis"], "comments": []}, "verified": true} {"bug_id": "121428", "issue_url": "https://github.com/llvm/llvm-project/issues/121428", "bug_type": "miscompilation", "base_commit": "b5f21671ef04984bc00770263234dfb94833a274", "knowledge_cutoff": "2025-01-01T09:27:44Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "a37dbc1f51c70d92fd209c2e52a9d794eb15e4e7", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": [[1225, 1232]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp": ["foldSelectCttzCtlz"]}}, "patch": "commit a37dbc1f51c70d92fd209c2e52a9d794eb15e4e7\nAuthor: Yingwei Zheng \nDate: Mon Jan 6 00:04:28 2025 +0800\n\n [InstCombine] Drop noundef in `foldSelectCttzCtlz` (#121692)\n \n Close https://github.com/llvm/llvm-project/issues/121428\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\nindex e7a8e947705f..a18b927678ef 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp\n@@ -1225,8 +1225,12 @@ static Value *foldSelectCttzCtlz(ICmpInst *ICI, Value *TrueVal, Value *FalseVal,\n // zext/trunc) have one use (ending at the select), the cttz/ctlz result will\n // not be used if the input is zero. Relax to 'zero is poison' for that case.\n if (II->hasOneUse() && SelectArg->hasOneUse() &&\n- !match(II->getArgOperand(1), m_One()))\n+ !match(II->getArgOperand(1), m_One())) {\n II->setArgOperand(1, ConstantInt::getTrue(II->getContext()));\n+ // noundef attribute on the intrinsic may no longer be valid.\n+ II->dropUBImplyingAttrsAndMetadata();\n+ IC.addToWorklist(II);\n+ }\n \n return nullptr;\n }\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll", "commands": ["opt -passes=instcombine -S < %s"], "tests": [{"test_name": "test_cttz_not_bw_multiuse", "test_body": "define i32 @test_cttz_not_bw_multiuse(i32 %x) {\n %ct = tail call i32 @llvm.cttz.i32(i32 %x, i1 false)\n %cmp = icmp ne i32 %x, 0\n %sel = select i1 %cmp, i32 %ct, i32 123\n %res = or i32 %sel, %ct\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "test_cttz_not_bw", "test_body": "define i32 @test_cttz_not_bw(i32 %x) {\n %ct = tail call i32 @llvm.cttz.i32(i32 %x, i1 false)\n %cmp = icmp ne i32 %x, 0\n %res = select i1 %cmp, i32 %ct, i32 123\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "test_cttz_not_bw_noundef", "test_body": "define i32 @test_cttz_not_bw_noundef(i32 %x) {\n %ct = tail call noundef i32 @llvm.cttz.i32(i32 %x, i1 false)\n %cmp = icmp ne i32 %x, 0\n %res = select i1 %cmp, i32 %ct, i32 123\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[InstCombine] `noundef` attribute should be dropped", "body": "Alive2: https://alive2.llvm.org/ce/z/wKsUWK\n```\ndefine i32 @src(i32 noundef %x) {\n %ct = call noundef i32 @llvm.cttz.i32(i32 %x, i1 false)\n %cmp = icmp ne i32 %x, 0\n %res = select i1 %cmp, i32 %ct, i32 123\n ret i32 %res\n}\ndefine i32 @tgt(i32 noundef %x) {\n %ct = call noundef range(i32 0, 33) i32 @llvm.cttz.i32(i32 %x, i1 true)\n %cmp.not = icmp eq i32 %x, 0\n %res = select i1 %cmp.not, i32 123, i32 %ct\n ret i32 %res\n}\n```\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "119665", "issue_url": "https://github.com/llvm/llvm-project/issues/119665", "bug_type": "crash", "base_commit": "7ae78a6cdb6ce9ad1534ed10519649fb3d47aca9", "knowledge_cutoff": "2024-12-12T07:13:14Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "3706dfef660097f24fb5efbac0d7f14b424492ed", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[2937, 2946], [2950, 2955]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["InnerLoopVectorizer::fixVectorizedLoop"]}}, "patch": "commit 3706dfef660097f24fb5efbac0d7f14b424492ed\nAuthor: Florian Hahn \nDate: Mon Feb 10 16:29:42 2025 +0000\n\n [LV] Forget LCSSA phi with new pred before other SCEV invalidation. (#119897)\n \n `forgetLcssaPhiWithNewPredecessor` performs additional invalidation if\n there is an existing SCEV for the phi, but earlier\n `forgetBlockAndLoopDispositions` or `forgetLoop` may already invalidate\n the SCEV for the phi.\n \n Change the order to first call `forgetLcssaPhiWithNewPredecessor` to\n ensure it runs before its SCEV gets invalidated too eagerly.\n \n Fixes https://github.com/llvm/llvm-project/issues/119665.\n \n PR: https://github.com/llvm/llvm-project/pull/119897\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 610e4904a80a..f2241be60ce0 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -2937,10 +2937,6 @@ void InnerLoopVectorizer::fixVectorizedLoop(VPTransformState &State) {\n if (EnableVPlanNativePath)\n fixNonInductionPHIs(State);\n \n- // Forget the original basic block.\n- PSE.getSE()->forgetLoop(OrigLoop);\n- PSE.getSE()->forgetBlockAndLoopDispositions();\n-\n // After vectorization, the exit blocks of the original loop will have\n // additional predecessors. Invalidate SCEVs for the exit phis in case SE\n // looked through single-entry phis.\n@@ -2950,6 +2946,10 @@ void InnerLoopVectorizer::fixVectorizedLoop(VPTransformState &State) {\n for (PHINode &PN : Exit->phis())\n PSE.getSE()->forgetLcssaPhiWithNewPredecessor(OrigLoop, &PN);\n \n+ // Forget the original basic block.\n+ PSE.getSE()->forgetLoop(OrigLoop);\n+ PSE.getSE()->forgetBlockAndLoopDispositions();\n+\n // Don't apply optimizations below when no vector region remains, as they all\n // require a vector loop at the moment.\n if (!State.Plan->getVectorLoopRegion())\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/invalidate-scev-at-scope-after-vectorization.ll", "commands": ["opt -passes='print,loop-vectorize' -force-vector-width=4 -scalar-evolution-classify-expressions=false -S %s"], "tests": [{"test_name": "", "test_body": "\n; Test case for https://github.com/llvm/llvm-project/issues/119665.\n\n; %loop.2's backedge-taken-count depends on %add.1 from %loop.1 via its\n; corresponding SCEV at the scope of %loop.2. After vectorizing %loop.1, %add.1\n; isn't available at the entry of %loop.2 anymore and %add.1 at %loop.2's scope\n; must be invalidated, as well as %loop.2's backedge-taken count.\ndefine void @test_invalidate_scevs_at_scope(ptr %p) {\n;\nentry:\n br label %loop.1\n\nloop.1:\n %iv.1 = phi i32 [ 0, %entry ], [ %iv.1.next, %loop.1 ]\n %1 = load i32, ptr %p, align 4\n %add.1 = add i32 %1, %iv.1\n %iv.1.next = add i32 %iv.1, 1\n %c.1 = icmp eq i32 %iv.1, 100\n br i1 %c.1, label %exit.1, label %loop.1\n\nexit.1:\n %add.lcssa = phi i32 [ %add.1, %loop.1 ]\n br label %loop.2\n\nloop.2:\n %iv.2 = phi i64 [ 0, %exit.1 ], [ %iv.2.next, %loop.2 ]\n %iv.2.trunc = trunc i64 %iv.2 to i32\n %iv.2.next = add i64 %iv.2, 1\n %gep = getelementptr inbounds i64, ptr %p, i64 %iv.2\n store i64 %iv.2, ptr %gep\n %add.2 = add i32 %add.lcssa, %iv.2.trunc\n %c.2 = icmp slt i32 %add.2, 100\n br i1 %c.2, label %loop.2, label %exit.2\n\nexit.2:\n ret void\n}\n\n;.\n;."}]}], "issue": {"title": "[clang] Crash at -O2: Assertion `isAvailableAtLoopEntry(LHS, L) && \"LHS is not available at Loop Entry\"' failed.", "body": "This code crashes at `-O2`: \n```c\nint a, e;\nint *b;\nshort c[];\nshort d;\nvoid f() {\n int g;\n for (;;)\n for (; a - 3 + g < 6; g = g + 1)\n c[g] = d;\n}\nvoid h();\nvoid i() {\n h();\n f();\n}\nvoid h() {\n e = 0;\n for (; e <= 50; ++e) {\n int *j = &a;\n *j = *b + e;\n }\n}\n```\n\nCompiler Explorer: https://godbolt.org/z/jfo65eK9x\n\nCrash:\n```console\nclang: /root/llvm-project/llvm/lib/Analysis/ScalarEvolution.cpp:11744: bool llvm::ScalarEvolution::isLoopEntryGuardedByCond(const llvm::Loop*, llvm::CmpInst::Predicate, const llvm::SCEV*, const llvm::SCEV*): Assertion `isAvailableAtLoopEntry(LHS, L) && \"LHS is not available at Loop Entry\"' failed.\n```\nBacktrace:\n```console\nStack dump:\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/clang -gdwarf-4 -g -o /app/output.s -mllvm --x86-asm-syntax=intel -fno-verbose-asm -S --gcc-toolchain=/opt/compiler-explorer/gcc-snapshot -fcolor-diagnostics -fno-crash-diagnostics -O2 -Wall -Wextra -fno-strict-aliasing \n1.\t parser at end of file\n2.\tOptimizer\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"\"\n4.\tRunning pass \"loop-vectorize\" on function \"i\"\n #0 0x0000000003c19758 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3c19758)\n #1 0x0000000003c17464 llvm::sys::CleanupOnSignal(unsigned long) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3c17464)\n #2 0x0000000003b649c8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\n #3 0x000079aa17242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #4 0x000079aa172969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #5 0x000079aa17242476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #6 0x000079aa172287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #7 0x000079aa1722871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #8 0x000079aa17239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #9 0x0000000002c90a7b llvm::ScalarEvolution::isLoopEntryGuardedByCond(llvm::Loop const*, llvm::CmpInst::Predicate, llvm::SCEV const*, llvm::SCEV const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2c90a7b)\n#10 0x0000000002ca6d97 llvm::ScalarEvolution::getTripCountFromExitCount(llvm::SCEV const*, llvm::Type*, llvm::Loop const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x2ca6d97)\n#11 0x00000000058449e4 llvm::VPlan::createInitialVPlan(llvm::Type*, llvm::PredicatedScalarEvolution&, bool, bool, llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x58449e4)\n#12 0x000000000572a8ce llvm::LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(llvm::VFRange&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x572a8ce)\n#13 0x000000000572dd5a llvm::LoopVectorizationPlanner::buildVPlansWithVPRecipes(llvm::ElementCount, llvm::ElementCount) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x572dd5a)\n#14 0x000000000572e24f llvm::LoopVectorizationPlanner::plan(llvm::ElementCount, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x572e24f)\n#15 0x000000000573044f llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x573044f)\n#16 0x0000000005732a89 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x5732a89)\n#17 0x00000000057330c3 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x57330c3)\n#18 0x0000000005283abe llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x5283abe)\n#19 0x00000000035c8550 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35c8550)\n#20 0x00000000010fa0ee llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x10fa0ee)\n#21 0x00000000035c6e6b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35c6e6b)\n#22 0x00000000010f977e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x10f977e)\n#23 0x00000000035c6870 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x35c6870)\n#24 0x0000000003eca382 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\n#25 0x0000000003ecdbfd clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3ecdbfd)\n#26 0x000000000459e07e clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x459e07e)\n#27 0x000000000655704c clang::ParseAST(clang::Sema&, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x655704c)\n#28 0x000000000459e458 clang::CodeGenAction::ExecuteAction() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x459e458)\n#29 0x0000000004856fc9 clang::FrontendAction::Execute() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4856fc9)\n#30 0x00000000047d989e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x47d989e)\n#31 0x000000000494267e clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x494267e)\n#32 0x0000000000cb92df cc1_main(llvm::ArrayRef, char const*, void*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xcb92df)\n#33 0x0000000000cb0d7a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\n#34 0x00000000045e2229 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\n#35 0x0000000003b64e74 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3b64e74)\n#36 0x00000000045e281f clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\n#37 0x00000000045a7f3d clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x45a7f3d)\n#38 0x00000000045a902d clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x45a902d)\n#39 0x00000000045b0565 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x45b0565)\n#40 0x0000000000cb6123 clang_main(int, char**, llvm::ToolContext const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xcb6123)\n#41 0x0000000000b84564 main (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xb84564)\n#42 0x000079aa17229d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#43 0x000079aa17229e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#44 0x0000000000cb0825 _start (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xcb0825)\n```\n", "author": "cardigan1008", "labels": ["release:backport", "vectorizers", "crash", "regression:18"], "comments": [{"author": "shafik", "body": "Looks like a clang-18 regression: https://godbolt.org/z/K7MMKxjzb"}, {"author": "dtcxzyw", "body": "Reduced reproducer: https://godbolt.org/z/s4G9Yb1vG\n```\n; bin/opt -passes=\"print,loop-vectorize\" test.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @i(ptr %p) {\nentry:\n br label %for.body.i\n\nfor.body.i: ; preds = %for.body.i, %entry\n %0 = phi i32 [ 0, %entry ], [ %inc.i, %for.body.i ]\n %1 = load i32, ptr %p, align 4\n %add.i = add i32 %1, %0\n %inc.i = add i32 %0, 1\n %exitcond.not.i = icmp eq i32 %0, 1\n br i1 %exitcond.not.i, label %for.cond.i.preheader, label %for.body.i\n\nfor.cond.i.preheader: ; preds = %for.body.i\n br label %for.cond.i\n\nfor.cond.i.loopexit: ; preds = %for.body.i1\n br label %for.cond.i\n\nfor.cond.i: ; preds = %for.cond.i.loopexit, %for.cond.i.preheader\n br label %for.body.i1\n\nfor.body.i1: ; preds = %for.body.i1, %for.cond.i\n %indvars.iv.i = phi i64 [ %indvars.iv.next.i, %for.body.i1 ], [ 0, %for.cond.i ]\n %indvars4.i = trunc i64 %indvars.iv.i to i32\n %indvars.iv.next.i = add i64 %indvars.iv.i, 1\n %add.reass.i = add i32 %add.i, %indvars4.i\n %cmp.i = icmp slt i32 %add.reass.i, 0\n br i1 %cmp.i, label %for.body.i1, label %for.cond.i.loopexit\n}\n\n```"}, {"author": "fhahn", "body": "Thanks, let me take a look"}]}, "verified": true} {"bug_id": "113123", "issue_url": "https://github.com/llvm/llvm-project/issues/113123", "bug_type": "miscompilation", "base_commit": "67ff5ba9af9754261abe11d762af11532a816126", "knowledge_cutoff": "2024-10-21T02:18:29Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "a2ba438f3e5635e368333213914c7452a6a6a2da", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": [[1531, 1538]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": ["InstCombinerImpl::foldLogicOfFCmps"]}}, "patch": "commit a2ba438f3e5635e368333213914c7452a6a6a2da\nAuthor: XChy \nDate: Mon Oct 21 22:30:31 2024 +0800\n\n [InstCombine] Preserve the flag from RHS only if the `and` is bitwise (#113164)\n \n Fixes #113123\n Alive proof: https://alive2.llvm.org/ce/z/hnqeLC\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\nindex d72013ba223d..835eae9f5725 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n@@ -1531,8 +1531,10 @@ Value *InstCombinerImpl::foldLogicOfFCmps(FCmpInst *LHS, FCmpInst *RHS,\n }\n if (IsLessThanOrLessEqual(IsAnd ? PredL : PredR)) {\n BuilderTy::FastMathFlagGuard Guard(Builder);\n- Builder.setFastMathFlags(LHS->getFastMathFlags() |\n- RHS->getFastMathFlags());\n+ FastMathFlags NewFlag = LHS->getFastMathFlags();\n+ if (!IsLogicalSelect)\n+ NewFlag |= RHS->getFastMathFlags();\n+ Builder.setFastMathFlags(NewFlag);\n \n Value *FAbs = Builder.CreateUnaryIntrinsic(Intrinsic::fabs, LHS0);\n return Builder.CreateFCmp(PredL, FAbs,\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/fcmp-range-check-idiom.ll", "commands": ["opt -S -passes=instcombine %s"], "tests": [{"test_name": "test_and_olt_fmf_propagation_union_logical_rhs_poison", "test_body": "define i1 @test_and_olt_fmf_propagation_union_logical_rhs_poison(float %x) {\n %cmp1 = fcmp ninf olt float %x, 0x3C00000000000000\n %cmp2 = fcmp ogt float %x, 0xBC00000000000000\n %cond = select i1 %cmp2, i1 %cmp1, i1 false\n ret i1 %cond\n}\n"}, {"test_name": "test_and_olt_fmf_propagation_union_logical_both_poison", "test_body": "define i1 @test_and_olt_fmf_propagation_union_logical_both_poison(float %x) {\n %cmp1 = fcmp ninf olt float %x, 0x3C00000000000000\n %cmp2 = fcmp ninf ogt float %x, 0xBC00000000000000\n %cond = select i1 %cmp2, i1 %cmp1, i1 false\n ret i1 %cond\n}\n"}, {"test_name": "test_and_olt_fmf_propagation_union", "test_body": "define i1 @test_and_olt_fmf_propagation_union(float %x) {\n %cmp1 = fcmp nnan ninf olt float %x, 0x3C00000000000000\n %cmp2 = fcmp nnan nsz ogt float %x, 0xBC00000000000000\n %cond = and i1 %cmp1, %cmp2\n ret i1 %cond\n}\n"}, {"test_name": "test_and_olt_fmf_propagation_union_logical_lhs_poison", "test_body": "define i1 @test_and_olt_fmf_propagation_union_logical_lhs_poison(float %x) {\n %cmp1 = fcmp olt float %x, 0x3C00000000000000\n %cmp2 = fcmp ninf ogt float %x, 0xBC00000000000000\n %cond = select i1 %cmp2, i1 %cmp1, i1 false\n ret i1 %cond\n}\n"}]}], "issue": {"title": "Wrong canonicalization of a range check idiom", "body": "https://github.com/llvm/llvm-project/blob/11dad2fa5138a50d60a5a34a2c7e074b976820e2/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp#L1511-L1527\r\n\r\n\r\nAlive2 report: https://alive2.llvm.org/ce/z/qBHAmS\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine i1 @test_and_olt_fmf_propagation_union.2(float %x) {\r\n#0:\r\n %cmp1 = fcmp nnan ninf olt float %x, 0.000000\r\n %cmp2 = fcmp nnan nsz ogt float %x, -0.000000\r\n %#1 = select i1 %cmp2, i1 %cmp1, i1 %cmp2\r\n %cond = and i1 %#1, %cmp2\r\n ret i1 %cond\r\n}\r\n=>\r\ndefine i1 @test_and_olt_fmf_propagation_union.2(float %x) {\r\n#0:\r\n %#1 = fabs nnan ninf nsz float %x\r\n %#2 = fcmp nnan ninf nsz olt float %#1, 0.000000\r\n ret i1 %#2\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Target is more poisonous than source\r\n\r\nExample:\r\nfloat %x = #xff800000 (-oo)\r\n\r\nSource:\r\ni1 %cmp1 = poison\r\ni1 %cmp2 = #x0 (0)\r\ni1 %#1 = #x0 (0)\r\ni1 %cond = #x0 (0)\r\n\r\nTarget:\r\nfloat %#1 = poison\r\ni1 %#2 = poison\r\nSource value: #x0 (0)\r\nTarget value: poison\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:instcombine", "floating-point"], "comments": [{"author": "XChy", "body": "Thanks for the report.\r\n\r\nWe preserve `ninf` flag from `cmp1` when folding logical and `select cmp2, cmp1, false` into `fcmp ninf (fabs(x))`. But if `cmp2` equals to `false`, the select won't be poison whatever `cmp1` is. Thus, such fold makes it more poisonous. See https://alive2.llvm.org/ce/z/hnqeLC"}, {"author": "XChy", "body": "And it looks like you are mutating the unit tests to fuzz LLVM. Looks interesting to me. If OK, could you provide some information about it?"}, {"author": "bongjunj", "body": "> And it looks like you are mutating the unit tests to fuzz LLVM. Looks interesting to me. If OK, could you provide some information about it?\r\n\r\n@XChy Yes we are designing a fuzzer on top of LLVM to test the compiler. We can't disclose the full detail at this time. Thank you."}]}, "verified": true} {"bug_id": "112068", "issue_url": "https://github.com/llvm/llvm-project/issues/112068", "bug_type": "miscompilation", "base_commit": "b24acc06e1d465b3e3e4e28515dd437f6a7454f2", "knowledge_cutoff": "2024-10-12T02:41:57Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "a59976bea8ad76f18119a11391dc8ba3e6ba07d5", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp": [[505, 512]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp": ["foldCttzCtlz"]}}, "patch": "commit a59976bea8ad76f18119a11391dc8ba3e6ba07d5\nAuthor: Yingwei Zheng \nDate: Tue Nov 19 20:06:34 2024 +0800\n\n [InstCombine] Drop noundef attributes in `foldCttzCtlz` (#116718)\n \n Closes https://github.com/llvm/llvm-project/issues/112068.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\nindex 6cff3c7af91e..42c0acd1e45e 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp\n@@ -505,8 +505,10 @@ static Instruction *foldCttzCtlz(IntrinsicInst &II, InstCombinerImpl &IC) {\n \n // If ctlz/cttz is only used as a shift amount, set is_zero_poison to true.\n if (II.hasOneUse() && match(Op1, m_Zero()) &&\n- match(II.user_back(), m_Shift(m_Value(), m_Specific(&II))))\n+ match(II.user_back(), m_Shift(m_Value(), m_Specific(&II)))) {\n+ II.dropUBImplyingAttrsAndMetadata();\n return IC.replaceOperand(II, 1, IC.Builder.getTrue());\n+ }\n \n Constant *C;\n \n", "tests": [{"file": "llvm/test/Transforms/InstCombine/shift-cttz-ctlz.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "shl_cttz_false_multiuse", "test_body": "define i32 @shl_cttz_false_multiuse(i32 %x, i32 %y) {\nentry:\n %cttz = call i32 @llvm.cttz.i32(i32 %y, i1 false)\n call void @use(i32 %cttz)\n %res = shl i32 %x, %cttz\n ret i32 %res\n}\n\ndeclare void @use(i32)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "shl_cttz_false_noundef", "test_body": "define i32 @shl_cttz_false_noundef(i32 %x, i32 %y) {\nentry:\n %cttz = call noundef i32 @llvm.cttz.i32(i32 %y, i1 false)\n %res = shl i32 %x, %cttz\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "ashr_cttz_false", "test_body": "define i32 @ashr_cttz_false(i32 %x, i32 %y) {\nentry:\n %cttz = call i32 @llvm.cttz.i32(i32 %y, i1 false)\n %res = ashr i32 %x, %cttz\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "shl_cttz_false", "test_body": "define i32 @shl_cttz_false(i32 %x, i32 %y) {\nentry:\n %cttz = call i32 @llvm.cttz.i32(i32 %y, i1 false)\n %res = shl i32 %x, %cttz\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "shl_cttz_as_lhs", "test_body": "define i32 @shl_cttz_as_lhs(i32 %x, i32 %y) {\nentry:\n %cttz = call i32 @llvm.cttz.i32(i32 %y, i1 false)\n %res = shl i32 %cttz, %x\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "shl_ctlz_false", "test_body": "define i32 @shl_ctlz_false(i32 %x, i32 %y) {\nentry:\n %cttz = call i32 @llvm.ctlz.i32(i32 %y, i1 false)\n %res = shl i32 %x, %cttz\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctlz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "lshr_cttz_false", "test_body": "define i32 @lshr_cttz_false(i32 %x, i32 %y) {\nentry:\n %cttz = call i32 @llvm.cttz.i32(i32 %y, i1 false)\n %res = lshr i32 %x, %cttz\n ret i32 %res\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.cttz.i32(i32, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[InstCombine] `noundef` attribute should be dropped after setting `is_zero_poison` to true", "body": "Reproducer: https://alive2.llvm.org/ce/z/7-TgAf\r\n```\r\ndefine i32 @src(i32 %x, i32 %y) {\r\nentry:\r\n %cttz = call noundef i32 @llvm.cttz.i32(i32 %y, i1 false)\r\n %res = lshr i32 %x, %cttz\r\n ret i32 %res\r\n}\r\n\r\ndefine i32 @tgt(i32 %x, i32 %y) {\r\nentry:\r\n %cttz = call noundef i32 @llvm.cttz.i32(i32 %y, i1 true)\r\n %res = lshr i32 %x, %cttz\r\n ret i32 %res\r\n}\r\n\r\ndefine i32 @src2(i32 %x) {\r\n %ct = call noundef i32 @llvm.ctlz.i32(i32 %x, i1 false)\r\n %cmp = icmp ne i32 %x, 0\r\n %res = select i1 %cmp, i32 %ct, i32 123\r\n ret i32 %res\r\n}\r\n\r\ndefine i32 @tgt2(i32 %x) {\r\n %ct = call noundef i32 @llvm.ctlz.i32(i32 %x, i1 true)\r\n %cmp = icmp ne i32 %x, 0\r\n %res = select i1 %cmp, i32 %ct, i32 123\r\n ret i32 %res\r\n}\r\n```\r\n`noundef` attribute should be dropped. Otherwise we will convert a poison value to immediate UB.\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "97674", "issue_url": "https://github.com/llvm/llvm-project/issues/97674", "bug_type": "crash", "base_commit": "c67653fbc3dd4ea9ab2a32c481a9ac4411020c7b", "knowledge_cutoff": "2024-07-04T04:06:51Z", "lit_test_dir": ["llvm/test/Transforms/CorrelatedValuePropagation"], "hints": {"fix_commit": "899fe2cf318f6e3c7a183125ff17ef53944b2989", "components": ["LazyValueInfo"], "bug_location_lineno": {"llvm/lib/Analysis/LazyValueInfo.cpp": [[1723, 1729], [1758, 1764]]}, "bug_location_funcname": {"llvm/lib/Analysis/LazyValueInfo.cpp": ["LazyValueInfo::getConstantOnEdge"]}}, "patch": "commit 899fe2cf318f6e3c7a183125ff17ef53944b2989\nAuthor: goldsteinn <35538541+goldsteinn@users.noreply.github.com>\nDate: Thu Jul 4 15:14:39 2024 +0800\n\n [CVP][LVI] Fix incorrect scalar type when getting constant folded vec (#97682)\n \n Fixes #97674\n \n After #97428 added support for vectors, our constant ranges can now be\n from splat vectors so when they reduce to a singe constant value, we\n need to return the original type as opposed to just an int.\n\ndiff --git a/llvm/lib/Analysis/LazyValueInfo.cpp b/llvm/lib/Analysis/LazyValueInfo.cpp\nindex b30e6a6a367c..4209ee4a7795 100644\n--- a/llvm/lib/Analysis/LazyValueInfo.cpp\n+++ b/llvm/lib/Analysis/LazyValueInfo.cpp\n@@ -1370,6 +1370,7 @@ LazyValueInfoImpl::getEdgeValueLocal(Value *Val, BasicBlock *BBFrom,\n \n // If V is the condition of the branch itself, then we know exactly what\n // it is.\n+ // NB: The condition on a `br` can't be a vector type.\n if (Condition == Val)\n return ValueLatticeElement::get(ConstantInt::get(\n Type::getInt1Ty(Val->getContext()), isTrueDest));\n@@ -1723,7 +1724,7 @@ Constant *LazyValueInfo::getConstant(Value *V, Instruction *CxtI) {\n if (Result.isConstantRange()) {\n const ConstantRange &CR = Result.getConstantRange();\n if (const APInt *SingleVal = CR.getSingleElement())\n- return ConstantInt::get(V->getContext(), *SingleVal);\n+ return ConstantInt::get(V->getType(), *SingleVal);\n }\n return nullptr;\n }\n@@ -1758,7 +1759,7 @@ Constant *LazyValueInfo::getConstantOnEdge(Value *V, BasicBlock *FromBB,\n if (Result.isConstantRange()) {\n const ConstantRange &CR = Result.getConstantRange();\n if (const APInt *SingleVal = CR.getSingleElement())\n- return ConstantInt::get(V->getContext(), *SingleVal);\n+ return ConstantInt::get(V->getType(), *SingleVal);\n }\n return nullptr;\n }\n", "tests": [{"file": "llvm/test/Transforms/CorrelatedValuePropagation/vectors.ll", "commands": ["opt -S -passes=correlated-propagation < %s"], "tests": [{"test_name": "and_with_poison", "test_body": "define <2 x i16> @and_with_poison(<2 x i8> %a) {\n %zext = zext <2 x i8> %a to <2 x i16>\n %res = and <2 x i16> %zext, \n ret <2 x i16> %res\n}\n"}, {"test_name": "issue_97674_getConstant", "test_body": "define <4 x i64> @issue_97674_getConstant() {\nentry:\n %folds = add <4 x i64> zeroinitializer, zeroinitializer\n ret <4 x i64> %folds\n}\n"}, {"test_name": "issue_97674_getConstantOnEdge", "test_body": "define <4 x i64> @issue_97674_getConstantOnEdge(i1 %cond) {\nentry:\n br i1 %cond, label %if.then, label %if.end\n\nif.then: ; preds = %entry\n %folds = add <4 x i64> zeroinitializer, splat (i64 1)\n br label %if.end\n\nif.end: ; preds = %if.then, %entry\n %r = phi <4 x i64> [ %folds, %if.then ], [ zeroinitializer, %entry ]\n ret <4 x i64> %r\n}\n"}]}], "issue": {"title": "A/F: `getType() == V->getType() && \"All operands to PHI node must be the same type as the PHI node !\"' after 1eec81a", "body": "We have an internal test which started to hit an assertion failure when compiled which I bisected back to 1eec81a.\r\n\r\nConsider the following code:\r\n```c++\r\ntemplate using c = int;\r\ntypedef long cb __attribute__((__vector_size__(16)));\r\ncb d(cb);\r\ntypedef long e __attribute__((__vector_size__(32)));\r\nenum { f };\r\nstruct g {\r\n enum { h };\r\n};\r\ntemplate struct i;\r\ntemplate struct aa;\r\ntemplate struct j;\r\ntemplate class k;\r\ntemplate class n;\r\ntemplate class o;\r\ntemplate class p;\r\ntemplate struct q;\r\ntemplate struct r;\r\ntemplate struct t { typedef q::ab ab; };\r\ntemplate struct ad { typedef k ab; };\r\ntemplate struct u {\r\n operator s() { return ae; }\r\n u(s v) : ae(v) {}\r\n s ae;\r\n};\r\ntemplate w x(const typename r::ab &);\r\ntemplate <> struct q { typedef u ab; };\r\ntemplate <> struct r> {\r\n typedef int ab;\r\n enum { af = 8 };\r\n};\r\ntemplate <> u x(const int &) {\r\n e y{};\r\n return y;\r\n}\r\ne z;\r\ne ag(e a, u b) {\r\n __attribute__((__vector_size__(4 * sizeof(long)))) long ah = b;\r\n z = a + ah;\r\n return z;\r\n}\r\nstruct ai {\r\n u<__attribute__((__vector_size__(4 * sizeof(long)))) long>\r\n aj(u<__attribute__((__vector_size__(4 * sizeof(long)))) long> a,\r\n const u<__attribute__((__vector_size__(4 * sizeof(long)))) long> &b) {\r\n __attribute__((__vector_size__(4 * sizeof(long)))) long ak = a;\r\n return ag(ak, b);\r\n }\r\n};\r\nstruct al {\r\n template am aj() { return x(an); }\r\n int an;\r\n};\r\ntemplate long ao(ac);\r\ntemplate class k : j {\r\npublic:\r\n typedef i::ap ap;\r\n using j::aq;\r\n static o::ar, i::as>>> at(long, long);\r\n ap au() const;\r\n template ap aw(const av &) const;\r\n};\r\ntemplate struct j { ac aq() const; };\r\nstruct ax {\r\n template s aj(al az, ay) { return az.aj(); }\r\n};\r\ntemplate struct aa> {\r\n aa(o);\r\n template am bc(ay) {\r\n return be.aj(bf, d);\r\n }\r\n ba bf;\r\n ax be;\r\n};\r\ntemplate class bg : public ad::ab {};\r\ntemplate \r\nstruct i> {\r\n typedef bh ap;\r\n enum { ar, as };\r\n};\r\ntemplate \r\nstruct i> : i> {};\r\ntemplate \r\nclass p : public bg> {};\r\ntemplate struct i> : i {};\r\ntemplate \r\nclass o : public ad>>::ab {};\r\ntemplate struct bm {\r\n typedef t::ab am;\r\n enum { bn = r::af };\r\n enum { bo };\r\n};\r\ntemplate ::bo> struct M;\r\ntemplate struct M {\r\n typedef bm::am bq;\r\n template static bl bs(bl bt, bp bu, br bv) {\r\n enum { alignment };\r\n long bw = ao(bv);\r\n bq bx = bt.template bc(bw);\r\n if (bw > bm::bn) {\r\n bq by = bt.template bc(bm::bn);\r\n for (long bd; bd; bd += bm::bn) {\r\n bx = bu.aj(bx, bt.template bc(d));\r\n u bz = bt.template bc(bm::bn);\r\n by = bu.aj(by, bz);\r\n }\r\n bx = bu.aj(bx, by);\r\n }\r\n __attribute__((__vector_size__(4 * sizeof(int)))) int ca =\r\n __builtin_ia32_vextractf128_si256(bx, 0);\r\n __attribute__((__vector_size__(2 * sizeof(long)))) long b = d(ca);\r\n }\r\n};\r\nclass cc : public aa>> {\r\npublic:\r\n typedef int ap;\r\n};\r\ntemplate \r\ntemplate \r\ni::ap k::aw(const bp &bu) const {\r\n cc cd(aq());\r\n o ce;\r\n M::bs(cd, bu, ce);\r\n}\r\ntemplate i::ap k::au() const { aw(ai()); }\r\nvoid cf() { p::at(0, 0).au(); }\r\n```\r\n\r\nWhen compiled with optimizations and AVX support with a compiler before 1eec81a, the compilation succeeds:\r\n```\r\n$ ~/src/upstream/b76dd4edbfbba5593b691ef92b755f25cf63f445-linux/bin/clang -c -O2 -mavx -Wno-return-type -std=c++20 repro.cpp\r\n$\r\n```\r\n\r\nWhen compiled with the same arguments but with a compiler built from 1eec81a, the compiler hits an assertion failure:\r\n```\r\n$ ~/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang -c -O2 -mavx -Wno-return-type -std=c++20 repro.cpp\r\nclang: /home/dyung/src/upstream/llvm_clean_git/llvm/include/llvm/IR/Instructions.h:2582: void llvm::PHINode::setIncomingValue(unsigned int, llvm::Value*): Assertion `getType() == V->getType() && \"All operands to PHI node must be the same type as the PHI node\r\n!\"' failed. \r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script. \r\nStack dump:\r\n0. Program arguments: /home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang -c -O2 -mavx -Wno-return-type -std=c++20 repro.cpp\r\n1. parser at end of file\r\n2. Optimizer\r\n3. Running pass \"require,function(invalidate),require,cgscc(devirt<4>(inline,function-attrs,openmp-opt-cgscc,function(sroa,early-cse,speculative-\r\nexecution,jump-threading,correlated-propagation,simplifycfg,instcombine,aggressive-instcombine,libcalls-shrinkwrap,tailcallelim,simplifycfg,reassociate,constraint-elimination,loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-un\r\nswitch),simplifycfg,instcombine,loop(loop-idiom,indvars,simple-loop-unswitch,loop-deletion,loop-unroll-full),sroa,vector-combine,mldst-motion,gvn<>,sccp,bdce,instcombine,jump-threading,correlated-propagation,adce,memcpyopt,dse,move-auto-init,loop-mssa(licm),coro-elide,simplifycfg,instcombine),function-attrs,function(require),coro-split)),function(invalidate),cgscc(devirt<4>())\" on module \"repro.cpp\"\r\n4. Running pass \"cgscc(devirt<4>(inline,function-attrs,openmp-opt-cgscc,function(sroa,early-cse,speculative-execution,jump-threading,correlated-propagat\r\nion,simplifycfg,instcombine,aggressive-instcombine,libcalls-shrinkwrap,tailcallelim,simplifycfg,reasso\r\nciate,constraint-elimination,loop-mssa(loop-instsimplify,loop-simplifycfg,licm,loop-rotate,licm,simple-loop-unswitch),simplifycfg,instcombine,loop(loop-idiom,indvars,simple-loop-uns\r\nwitch,loop-deletion,loop-unroll-full),sroa,vector-combine,mldst-motion,gvn<>,sccp,bdce,instcombine,jump-threading,correlated-propagation,adce,memcpyo\r\npt,dse,move-auto-init,loop-mssa(licm),coro-elide,simplifycfg,instc\r\nombine),function-attrs,function(require),coro-split))\" on module \"repro.cpp\"\r\n5. Running pass \"correlated-propagation\" on function \"_ZN1MI2ai2ccLi0EE2bsI1oIiiEEES1_S1_S0_T_\"\r\n #0 0x000056411be676af llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x41856af)\r\n #1 0x000056411be6520c llvm::sys::CleanupOnSignal(unsigned long) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x418320c)\r\n #2 0x000056411bdaa728 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\r\n #3 0x00007f29a8886420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #4 0x00007f29a835300b raise /build/glibc-LcI20x/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #5 0x00007f29a8332859 abort /build/glibc-LcI20x/glibc-2.31/stdlib/abort.c:81:7\r\n #6 0x00007f29a8332729 get_sysdep_segment_value /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #7 0x00007f29a8332729 _nl_load_domain /build/glibc-LcI20x/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #8 0x00007f29a8343fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #9 0x00005641200199a5 processPHI(llvm::PHINode*, llvm::LazyValueInfo*, llvm::DominatorTree*, llvm::SimplifyQuery const&) CorrelatedValuePropagation.cpp:0:0\r\n#10 0x000056412002131e runImpl(llvm::Function&, llvm::LazyValueInfo*, llvm::DominatorTree*, llvm::SimplifyQuery const&) CorrelatedValuePropagation.cpp:0:0\r\n#11 0x0000564120022b4b llvm::CorrelatedValuePropagationPass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x8340b4b)\r\n#12 0x000056411d2b9866 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x55d7866)\r\n#13 0x000056411b80daf1 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x3b2baf1)\r\n#14 0x00005641193e3d86 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x1701d86)\r\n#15 0x000056411acf6e91 llvm::CGSCCToFunctionPassAdaptor::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x3014e91)\r\n#16 0x00005641193e5326 llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x1703326)\r\n#17 0x000056411aceef7f llvm::PassManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x300cf7f)\r\n#18 0x000056411d2b71b6 llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>, llvm::AnalysisManager, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x55d51b6)\r\n#19 0x000056411acf370d llvm::DevirtSCCRepeatedPass::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x301170d)\r\n#20 0x000056411d2b7206 llvm::detail::PassModel, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&>::run(llvm::LazyCallGraph::SCC&, llvm::AnalysisManager&, llvm::LazyCallGraph&, llvm::CGSCCUpdateResult&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x55d5206)\r\n#21 0x000056411acf150c llvm::ModuleToPostOrderCGSCCPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x300f50c)\r\n#22 0x000056411d2b7166 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x55d5166)\r\n#23 0x000056411b80a601 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x3b28601)\r\n#24 0x000056411d4afdb5 llvm::ModuleInlinerWrapperPass::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x57cddb5)\r\n#25 0x000056411d2b68f6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x55d48f6)\r\n#26 0x000056411b80a601 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x3b28601)\r\n#27 0x000056411c11c4d8 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#28 0x000056411c11f855 (anonymous namespace)::EmitAssemblyHelper::EmitAssembly(clang::BackendAction, std::unique_ptr>, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#29 0x000056411c11fef6 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x443def6)\r\n#30 0x000056411c792bea clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x4ab0bea)\r\n#31 0x000056411e56cb8c clang::ParseAST(clang::Sema&, bool, bool) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x688ab8c)\r\n#32 0x000056411c793078 clang::CodeGenAction::ExecuteAction() (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x4ab1078)\r\n#33 0x000056411ca26e79 clang::FrontendAction::Execute() (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x4d44e79)\r\n#34 0x000056411c9aa76e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x4cc876e)\r\n#35 0x000056411cb12366 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x4e30366)\r\n#36 0x0000564118fcc89f cc1_main(llvm::ArrayRef, char const*, void*) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x12ea89f)\r\n#37 0x0000564118fc594a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#38 0x000056411c7d833d void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\r\n#39 0x000056411bdaac30 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x40c8c30)\r\n#40 0x000056411c7d895f clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\r\n#41 0x000056411c79cdbc clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x4abadbc)\r\n#42 0x000056411c79d84d clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x4abb84d)\r\n#43 0x000056411c7a5925 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x4ac3925)\r\n#44 0x0000564118fc9bab clang_main(int, char**, llvm::ToolContext const&) (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x12e7bab)\r\n#45 0x0000564118ebcc0b main (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x11dac0b)\r\n#46 0x00007f29a8334083 __libc_start_main /build/glibc-LcI20x/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#47 0x0000564118fc53de _start (/home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin/clang+0x12e33de)\r\nclang: error: clang frontend command failed with exit code 134 (use -v to see invocation)\r\nclang version 19.0.0 (https://github.com/llvm/llvm-project.git 1eec81a831dcf925c0c86c153e8dce019a5b5436)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /home/dyung/src/upstream/1eec81a831dcf925c0c86c153e8dce019a5b5436-linux/bin\r\n```", "author": "dyung", "labels": ["crash", "llvm:analysis"], "comments": [{"author": "goldsteinn", "body": "Seems like when `CVP` finds a constant, its returning the scalar, not the VEC type.\r\n\r\nReducer of exact error:\r\n```\r\ndefine <4 x i64> @reduce(i1 %cond) {\r\nentry:\r\n br i1 %cond, label %if.then, label %if.end\r\n\r\nif.then:\r\n %folds = add <4 x i64> zeroinitializer, \r\n br label %if.end\r\n\r\nif.end:\r\n %r = phi <4 x i64> [ %folds, %if.then ], [ zeroinitializer, %entry ]\r\n ret <4 x i64> %r\r\n}\r\n```\r\nRepro with: `opt -passes=correlated-propagation`\r\n\r\nOr an even simpler version (different assert location):\r\n```\r\ndefine <4 x i64> @reduce() {\r\nentry:\r\n %folds = add <4 x i64> zeroinitializer, zeroinitializer\r\n ret <4 x i64> %folds\r\n}\r\n \r\n```\r\n\r\nSame command as above.\r\n"}, {"author": "goldsteinn", "body": "Ill have a fix up shortly."}, {"author": "goldsteinn", "body": "Fix at: https://github.com/llvm/llvm-project/pull/97682"}]}, "verified": true} {"bug_id": "122584", "issue_url": "https://github.com/llvm/llvm-project/issues/122584", "bug_type": "crash", "base_commit": "34ba84fe90b3e369c03e695a395ec632ef60e23d", "knowledge_cutoff": "2025-01-11T08:32:25Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "af524de1fa94e4b4cee8b745d1b68f4ea0090759", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[14935, 14942]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::processBuildVector"]}}, "patch": "commit af524de1fa94e4b4cee8b745d1b68f4ea0090759\nAuthor: Alexey Bataev \nDate: Mon Jan 13 06:48:56 2025 -0800\n\n [SLP]Do not include subvectors for fully matched buildvectors\n \n If the buildvector node fully matched another node, need to exclude\n subvectors, when building final shuffle, just a shuffle of the original\n node must be emitted.\n \n Fixes #122584\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex e3487b501534..df46c69ff3ab 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -14935,8 +14935,8 @@ ResTy BoUpSLP::processBuildVector(const TreeEntry *E, Type *ScalarTy,\n }\n }\n ShuffleBuilder.add(*FrontTE, Mask);\n- Res = ShuffleBuilder.finalize(E->getCommonMask(), SubVectors,\n- SubVectorsMask);\n+ // Full matched entry found, no need to insert subvectors.\n+ Res = ShuffleBuilder.finalize(E->getCommonMask(), {}, {});\n return Res;\n }\n if (!Resized) {\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/AArch64/vectorizable-selects-uniform-cmps.ll", "commands": ["opt -passes=slp-vectorizer -S %s", "opt -aa-pipeline=basic-aa -passes='slp-vectorizer' -S %s"], "tests": [{"test_name": "select_uniform_ugt_16xi8", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"arm64-apple-ios5.0.0\"\n\ndefine void @select_uniform_ugt_16xi8(ptr %ptr, i8 %x) {\nentry:\n %l.0 = load i8, ptr %ptr, align 1\n %cmp.0 = icmp ugt i8 %l.0, -1\n %s.0 = select i1 %cmp.0, i8 %l.0, i8 %x\n store i8 %s.0, ptr %ptr, align 2\n %gep.1 = getelementptr inbounds i8, ptr %ptr, i8 1\n %l.1 = load i8, ptr %gep.1, align 1\n %cmp.1 = icmp ugt i8 %l.1, -1\n %s.1 = select i1 %cmp.1, i8 %l.1, i8 %x\n store i8 %s.1, ptr %gep.1, align 2\n %gep.2 = getelementptr inbounds i8, ptr %ptr, i8 2\n %l.2 = load i8, ptr %gep.2, align 1\n %cmp.2 = icmp ugt i8 %l.2, -1\n %s.2 = select i1 %cmp.2, i8 %l.2, i8 %x\n store i8 %s.2, ptr %gep.2, align 2\n %gep.3 = getelementptr inbounds i8, ptr %ptr, i8 3\n %l.3 = load i8, ptr %gep.3, align 1\n %cmp.3 = icmp ugt i8 %l.3, -1\n %s.3 = select i1 %cmp.3, i8 %l.3, i8 %x\n store i8 %s.3, ptr %gep.3, align 2\n %gep.4 = getelementptr inbounds i8, ptr %ptr, i8 4\n %l.4 = load i8, ptr %gep.4, align 1\n %cmp.4 = icmp ugt i8 %l.4, -1\n %s.4 = select i1 %cmp.4, i8 %l.4, i8 %x\n store i8 %s.4, ptr %gep.4, align 2\n %gep.5 = getelementptr inbounds i8, ptr %ptr, i8 5\n %l.5 = load i8, ptr %gep.5, align 1\n %cmp.5 = icmp ugt i8 %l.5, -1\n %s.5 = select i1 %cmp.5, i8 %l.5, i8 %x\n store i8 %s.5, ptr %gep.5, align 2\n %gep.6 = getelementptr inbounds i8, ptr %ptr, i8 6\n %l.6 = load i8, ptr %gep.6, align 1\n %cmp.6 = icmp ugt i8 %l.6, -1\n %s.6 = select i1 %cmp.6, i8 %l.6, i8 %x\n store i8 %s.6, ptr %gep.6, align 2\n %gep.7 = getelementptr inbounds i8, ptr %ptr, i8 7\n %l.7 = load i8, ptr %gep.7, align 1\n %cmp.7 = icmp ugt i8 %l.7, -1\n %s.7 = select i1 %cmp.7, i8 %l.7, i8 %x\n store i8 %s.7, ptr %gep.7, align 2\n %gep.8 = getelementptr inbounds i8, ptr %ptr, i8 8\n %l.8 = load i8, ptr %gep.8, align 1\n %cmp.8 = icmp ugt i8 %l.8, -1\n %s.8 = select i1 %cmp.8, i8 %l.0, i8 %x\n store i8 %s.0, ptr %gep.8, align 2\n %gep.9 = getelementptr inbounds i8, ptr %ptr, i8 9\n %l.9 = load i8, ptr %gep.9, align 1\n %cmp.9 = icmp ugt i8 %l.9, -1\n %s.9 = select i1 %cmp.9, i8 %l.9, i8 %x\n store i8 %s.9, ptr %gep.9, align 2\n %gep.10 = getelementptr inbounds i8, ptr %ptr, i8 10\n %l.10 = load i8, ptr %gep.10, align 1\n %cmp.10 = icmp ugt i8 %l.10, -1\n %s.10 = select i1 %cmp.10, i8 %l.10, i8 %x\n store i8 %s.10, ptr %gep.10, align 2\n %gep.11 = getelementptr inbounds i8, ptr %ptr, i8 11\n %l.11 = load i8, ptr %gep.11, align 1\n %cmp.11 = icmp ugt i8 %l.11, -1\n %s.11 = select i1 %cmp.11, i8 %l.11, i8 %x\n store i8 %s.11, ptr %gep.11, align 2\n %gep.12 = getelementptr inbounds i8, ptr %ptr, i8 12\n %l.12 = load i8, ptr %gep.12, align 1\n %cmp.12 = icmp ugt i8 %l.12, -1\n %s.12 = select i1 %cmp.12, i8 %l.12, i8 %x\n store i8 %s.12, ptr %gep.12, align 2\n %gep.13 = getelementptr inbounds i8, ptr %ptr, i8 13\n %l.13 = load i8, ptr %gep.13, align 1\n %cmp.13 = icmp ugt i8 %l.13, -1\n %s.13 = select i1 %cmp.13, i8 %l.13, i8 %x\n store i8 %s.13, ptr %gep.13, align 2\n %gep.14 = getelementptr inbounds i8, ptr %ptr, i8 14\n %l.14 = load i8, ptr %gep.14, align 1\n %cmp.14 = icmp ugt i8 %l.14, -1\n %s.14 = select i1 %cmp.14, i8 %l.14, i8 %x\n store i8 %s.14, ptr %gep.14, align 2\n %gep.15 = getelementptr inbounds i8, ptr %ptr, i8 15\n %l.15 = load i8, ptr %gep.15, align 1\n %cmp.15 = icmp ugt i8 %l.15, -1\n %s.15 = select i1 %cmp.15, i8 %l.15, i8 %x\n store i8 %s.15, ptr %gep.15, align 2\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/SLPVectorizer/X86/full-matched-bv-with-subvectors.ll", "commands": ["opt -S --passes=slp-vectorizer -slp-threshold=-9999 -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test(i64 %l.549) {\nentry:\n %conv3 = sext i32 0 to i64\n br label %if.then19\n\np: ; preds = %if.end29, %if.end25\n %l.0 = phi i64 [ %xor, %if.end29 ], [ %l.5493, %if.end25 ]\n %m.0 = phi i64 [ %not21, %if.end29 ], [ %m.550, %if.end25 ]\n br i1 false, label %s, label %q\n\nq: ; preds = %land.lhs.true, %p\n %xor39 = phi i64 [ 0, %p ], [ 0, %land.lhs.true ]\n %l.1 = phi i64 [ 0, %p ], [ 0, %land.lhs.true ]\n %m.1 = phi i64 [ 0, %p ], [ 0, %land.lhs.true ]\n br i1 false, label %lor.lhs.false, label %r\n\nlor.lhs.false: ; preds = %q\n br i1 false, label %land.lhs.true, label %s\n\nr: ; preds = %if.then19, %q\n %xor38 = phi i64 [ %xor39, %q ], [ %xor, %if.then19 ]\n %j.0 = phi i64 [ %conv3, %q ], [ %not21, %if.then19 ]\n %l.2 = phi i64 [ %l.1, %q ], [ %l.549, %if.then19 ]\n %m.2 = phi i64 [ %m.1, %q ], [ %m.550, %if.then19 ]\n br i1 false, label %s, label %land.lhs.true\n\nland.lhs.true: ; preds = %r, %lor.lhs.false\n %xor37 = phi i64 [ %xor38, %r ], [ 0, %lor.lhs.false ]\n %j.1 = phi i64 [ %j.0, %r ], [ 0, %lor.lhs.false ]\n %l.3 = phi i64 [ %l.2, %r ], [ 0, %lor.lhs.false ]\n %m.3 = phi i64 [ %m.2, %r ], [ 0, %lor.lhs.false ]\n br i1 false, label %q, label %s\n\ns: ; preds = %land.lhs.true, %r, %lor.lhs.false, %p\n %xor36 = phi i64 [ %xor37, %land.lhs.true ], [ %xor38, %r ], [ %xor39, %lor.lhs.false ], [ %l.0, %p ]\n %j.2 = phi i64 [ %j.1, %land.lhs.true ], [ %j.0, %r ], [ %conv3, %lor.lhs.false ], [ %m.0, %p ]\n %l.4 = phi i64 [ %l.3, %land.lhs.true ], [ %l.2, %r ], [ %l.1, %lor.lhs.false ], [ %l.0, %p ]\n %m.4 = phi i64 [ %m.3, %land.lhs.true ], [ %m.2, %r ], [ %m.1, %lor.lhs.false ], [ %m.0, %p ]\n br label %if.then19\n\nif.then19: ; preds = %s, %entry\n %m.550 = phi i64 [ 0, %entry ], [ %m.4, %s ]\n %l.5493 = phi i64 [ 0, %entry ], [ %l.4, %s ]\n %xor = xor i64 0, 0\n %not21 = xor i64 0, 0\n br i1 false, label %r, label %if.end25\n\nif.end25: ; preds = %if.then19\n br i1 false, label %if.end29, label %p\n\nif.end29: ; preds = %if.end25\n br label %p\n}\n"}]}], "issue": {"title": "clang crashes on valid code at -O{s,2,3} on x86_64-linux-gnu: Assertion `I1 == PoisonMaskElem && \"Expected unused subvectors mask\"' failed", "body": "It appears to be a recent regression as it doesn't reproduce with 19.1.0 and earlier.\n\nCompiler Explorer: https://godbolt.org/z/vWPhsMfoz\n\n```\n[510] % clangtk -v\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 24bd9bc0b59d51c82e9a4d84c21d86d58d0ef6ce)\nTarget: x86_64-unknown-linux-gnu\nThread model: posix\nInstalledDir: /local/home/suz/suz-local/software/local/clang-trunk/bin\nBuild config: +assertions\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\nCandidate multilib: .;@m64\nSelected multilib: .;@m64\nFound CUDA installation: /usr/local/cuda, version 12.1\n[511] % \n[511] % clangtk -O2 small.c\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:10936: llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::finalize(llvm::ArrayRef, llvm::ArrayRef >, llvm::ArrayRef, unsigned int, llvm::function_ref&)>): Assertion `I1 == PoisonMaskElem && \"Expected unused subvectors mask\"' failed.\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\nStack dump:\n0.\tProgram arguments: /local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20250111-clangtk-m64-Os-build-025751/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20250111-clangtk-m64-Os-build-025751/delta -resource-dir /local/home/suz/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/home/suz/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/12/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-2771d7.o -x c small.c\n1.\t parser at end of file\n2.\tOptimizer\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\n4.\tRunning pass \"slp-vectorizer\" on function \"main\"\n #0 0x0000562bd646d500 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x4576500)\n #1 0x0000562bd646a91f llvm::sys::RunSignalHandlers() (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x457391f)\n #2 0x0000562bd646aa75 SignalHandler(int) Signals.cpp:0:0\n #3 0x00007f13d1771520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #4 0x00007f13d17c5a7c pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x96a7c)\n #5 0x00007f13d1771476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #6 0x00007f13d17577f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #7 0x00007f13d175771b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #8 0x00007f13d1768e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #9 0x0000562bd7f25a4f llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::finalize(llvm::ArrayRef, llvm::ArrayRef>, llvm::ArrayRef, unsigned int, llvm::function_ref&)>) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x602ea4f)\n#10 0x0000562bd7f9164b llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::processBuildVector, llvm::slpvectorizer::BoUpSLP, llvm::SmallPtrSetImpl>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::TargetTransformInfo&, llvm::ArrayRef&, llvm::slpvectorizer::BoUpSLP&, llvm::SmallPtrSetImpl&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x609a64b)\n#11 0x0000562bd7f92881 llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x609b881)\n#12 0x0000562bd7faaa12 llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x60b3a12)\n#13 0x0000562bd7fb43e8 llvm::SLPVectorizerPass::tryToVectorizeList(llvm::ArrayRef, llvm::slpvectorizer::BoUpSLP&, bool) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x60bd3e8)\n#14 0x0000562bd7fb92b0 bool tryToVectorizeSequence(llvm::SmallVectorImpl&, llvm::function_ref, llvm::function_ref, llvm::function_ref, bool)>, bool, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\n#15 0x0000562bd7fbabe9 llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x60c3be9)\n#16 0x0000562bd7fc268e llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\n#17 0x0000562bd7fc31d1 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x60cc1d1)\n#18 0x0000562bd79d6826 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5adf826)\n#19 0x0000562bd5de82ef llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3ef12ef)\n#20 0x0000562bd37a5226 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x18ae226)\n#21 0x0000562bd5de664b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3eef64b)\n#22 0x0000562bd37a3b36 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x18acb36)\n#23 0x0000562bd5de724d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3ef024d)\n#24 0x0000562bd67194c7 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\n#25 0x0000562bd671d213 clang::emitBackendOutput(clang::CompilerInstance&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x4826213)\n#26 0x0000562bd6df1805 clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x4efa805)\n#27 0x0000562bd8a8c2cc clang::ParseAST(clang::Sema&, bool, bool) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x6b952cc)\n#28 0x0000562bd70d6ca9 clang::FrontendAction::Execute() (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x51dfca9)\n#29 0x0000562bd7054ad5 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x515dad5)\n#30 0x0000562bd71bbf73 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x52c4f73)\n#31 0x0000562bd334425f cc1_main(llvm::ArrayRef, char const*, void*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x144d25f)\n#32 0x0000562bd333ab2a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\n#33 0x0000562bd333efe7 clang_main(int, char**, llvm::ToolContext const&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x1447fe7)\n#34 0x0000562bd322eb3b main (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x1337b3b)\n#35 0x00007f13d1758d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#36 0x00007f13d1758e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#37 0x0000562bd333a255 _start (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x1443255)\nclangtk: error: unable to execute command: Aborted\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git 24bd9bc0b59d51c82e9a4d84c21d86d58d0ef6ce)\nTarget: x86_64-unknown-linux-gnu\nThread model: posix\nInstalledDir: /local/home/suz/suz-local/software/local/clang-trunk/bin\nBuild config: +assertions\nclangtk: note: diagnostic msg: \n********************\n\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\nPreprocessed source(s) and associated run script(s) are located at:\nclangtk: note: diagnostic msg: /tmp/small-929da7.c\nclangtk: note: diagnostic msg: /tmp/small-929da7.sh\nclangtk: note: diagnostic msg: \n\n********************\n[512] % \n[512] % cat small.c\nint a, b, d, e, f, i, n;\nlong c, g, h;\nunsigned o;\nint main() {\n unsigned long j;\n long l, m;\n if (0) {\n k:\n h = ~g;\n l = g;\n m = j;\n i = d;\n p:\n b = g = i;\n if (d) {\n q:\n j = a;\n if (i || i >= a)\n r:\n if (f && h > a)\n goto q;\n goto s;\n }\n g = l;\n j = m;\n s:;\n }\n if (c) {\n n = j % -1;\n o = b;\n if (n) {\n g = d ^ j;\n j = ~(1 / o);\n if (!(n - 1))\n goto r;\n if (e)\n goto p;\n }\n goto k;\n }\n}\n```", "author": "zhendongsu", "labels": ["llvm:SLPVectorizer", "crash-on-valid", "generated by fuzzer"], "comments": [{"author": "dtcxzyw", "body": "Reduced reproducer: https://godbolt.org/z/MzcesqWz1\n```\n; bin/opt -passes=slp-vectorizer reduced.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i32 @main(i64 %l.549) {\nentry:\n %conv3 = sext i32 0 to i64\n br label %if.then19\n\np: ; preds = %if.end29, %if.end25\n %l.0 = phi i64 [ %xor, %if.end29 ], [ %l.5493, %if.end25 ]\n %m.0 = phi i64 [ %not21, %if.end29 ], [ %m.550, %if.end25 ]\n br i1 false, label %s, label %q\n\nq: ; preds = %land.lhs.true, %p\n %xor39 = phi i64 [ 0, %p ], [ 0, %land.lhs.true ]\n %l.1 = phi i64 [ 0, %p ], [ 0, %land.lhs.true ]\n %m.1 = phi i64 [ 0, %p ], [ 0, %land.lhs.true ]\n br i1 false, label %lor.lhs.false, label %r\n\nlor.lhs.false: ; preds = %q\n br i1 false, label %land.lhs.true, label %s\n\nr: ; preds = %if.then19, %q\n %xor38 = phi i64 [ %xor39, %q ], [ %xor, %if.then19 ]\n %j.0 = phi i64 [ %conv3, %q ], [ %not21, %if.then19 ]\n %l.2 = phi i64 [ %l.1, %q ], [ %l.549, %if.then19 ]\n %m.2 = phi i64 [ %m.1, %q ], [ %m.550, %if.then19 ]\n br i1 false, label %s, label %land.lhs.true\n\nland.lhs.true: ; preds = %r, %lor.lhs.false\n %xor37 = phi i64 [ %xor38, %r ], [ 0, %lor.lhs.false ]\n %j.1 = phi i64 [ %j.0, %r ], [ 0, %lor.lhs.false ]\n %l.3 = phi i64 [ %l.2, %r ], [ 0, %lor.lhs.false ]\n %m.3 = phi i64 [ %m.2, %r ], [ 0, %lor.lhs.false ]\n br i1 false, label %q, label %s\n\ns: ; preds = %land.lhs.true, %r, %lor.lhs.false, %p\n %xor36 = phi i64 [ %xor37, %land.lhs.true ], [ %xor38, %r ], [ %xor39, %lor.lhs.false ], [ %l.0, %p ]\n %j.2 = phi i64 [ %j.1, %land.lhs.true ], [ %j.0, %r ], [ %conv3, %lor.lhs.false ], [ %m.0, %p ]\n %l.4 = phi i64 [ %l.3, %land.lhs.true ], [ %l.2, %r ], [ %l.1, %lor.lhs.false ], [ %l.0, %p ]\n %m.4 = phi i64 [ %m.3, %land.lhs.true ], [ %m.2, %r ], [ %m.1, %lor.lhs.false ], [ %m.0, %p ]\n br label %if.then19\n\nif.then19: ; preds = %s, %entry\n %m.550 = phi i64 [ 0, %entry ], [ %m.4, %s ]\n %l.5493 = phi i64 [ 0, %entry ], [ %l.4, %s ]\n %xor = xor i64 0, 0\n %not21 = xor i64 0, 0\n br i1 false, label %r, label %if.end25\n\nif.end25: ; preds = %if.then19\n br i1 false, label %if.end29, label %p\n\nif.end29: ; preds = %if.end25\n br label %p\n}\n\n```"}]}, "verified": true} {"bug_id": "98838", "issue_url": "https://github.com/llvm/llvm-project/issues/98838", "bug_type": "miscompilation", "base_commit": "9ba9e480fa4ef1f37a2b6b51603ac8e446cca5ff", "knowledge_cutoff": "2024-07-14T21:26:59Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "beccecaacde405a3b50891c67594eccbcd1c8b08", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[14201, 14209]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::vectorizeTree"]}}, "patch": "commit beccecaacde405a3b50891c67594eccbcd1c8b08\nAuthor: Alexey Bataev \nDate: Mon Jul 15 07:03:22 2024 -0700\n\n [SLP]Fix PR98838: do no replace condition of select-based logical op by poison.\n \n If the reduction operation is a select-based logical op, the condition\n should be replaced by the poison, better to replace by the non-poisoning\n constant to prevent poison propagation in the vector code.\n \n Fixes https://github.com/llvm/llvm-project/issues/98838\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 5c2fc0b9320e..74a16d3fbcad 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -14201,9 +14201,23 @@ Value *BoUpSLP::vectorizeTree(\n for (Instruction *I : RemovedInsts) {\n if (getTreeEntry(I)->Idx != 0)\n continue;\n+ SmallVector LogicalOpSelects;\n I->replaceUsesWithIf(PoisonValue::get(I->getType()), [&](Use &U) {\n+ // Do not replace condition of the logical op in form select .\n+ bool IsPoisoningLogicalOp = isa(U.getUser()) &&\n+ (match(U.getUser(), m_LogicalAnd()) ||\n+ match(U.getUser(), m_LogicalOr())) &&\n+ U.getOperandNo() == 0;\n+ if (IsPoisoningLogicalOp) {\n+ LogicalOpSelects.push_back(cast(U.getUser()));\n+ return false;\n+ }\n return UserIgnoreList->contains(U.getUser());\n });\n+ // Replace conditions of the poisoning logical ops with the non-poison\n+ // constant value.\n+ for (SelectInst *SI : LogicalOpSelects)\n+ SI->setCondition(Constant::getNullValue(SI->getCondition()->getType()));\n }\n }\n // Retain to-be-deleted instructions for some debug-info bookkeeping and alias\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/select-reduction-op.ll", "commands": ["opt -S --passes=slp-vectorizer < %s"], "tests": [{"test_name": "src", "test_body": "define i1 @src(i1 %cmp4.118.i) {\n %cmp4.118.i.not = xor i1 %cmp4.118.i, true\n %brmerge = select i1 %cmp4.118.i.not, i1 true, i1 poison\n %.not = xor i1 poison, true\n %brmerge2 = select i1 %brmerge, i1 true, i1 %.not\n %.not3 = xor i1 poison, true\n %brmerge4 = select i1 %brmerge2, i1 true, i1 %.not3\n %.not5 = xor i1 poison, true\n %brmerge6 = select i1 %brmerge4, i1 true, i1 %.not5\n %.not7 = xor i1 poison, true\n %brmerge8 = select i1 %brmerge6, i1 true, i1 %.not7\n ret i1 %brmerge8\n}\n"}]}], "issue": {"title": "wrong code at -O3 on x86_64-linux-gnu", "body": "It appears to be a recent regression as it doesn't reproduce with 18.1.0 and earlier. \r\n\r\nCompiler Explorer: https://godbolt.org/z/fzsqjEPWW\r\n\r\n```\r\n[562] % clangtk -v\r\nclang version 19.0.0git (https://github.com/llvm/llvm-project.git 52139d8f9a4e3f595ca552393d62ba06b0bc082c)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/10\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/9\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\n[563] % \r\n[563] % clangtk -O2 -w small.c\r\n[564] % ./a.out\r\n0\r\n[565] % clangtk -O3 -w small.c\r\n[566] % ./a.out\r\n[567] % \r\n[567] % cat small.c\r\nint printf(const char *, ...);\r\nint a, b, e, g;\r\nvolatile int c;\r\nstatic int d[1] = {1}, f;\r\nint h() {\r\n for (b = 0; b < 4; b++) {\r\n for (e = 0; e < 4; e++)\r\n while (b) {\r\n if (c > 0 && d[1])\r\n break;\r\n return f;\r\n }\r\n }\r\n return 0;\r\n}\r\nvoid i() {\r\n int *j[1];\r\n j[g] = &d[0];\r\n}\r\nint main() {\r\n h();\r\n printf(\"%d\\n\", a);\r\n return 0;\r\n}\r\n```\r\n\r\n", "author": "zhendongsu", "labels": ["miscompilation", "llvm:SLPVectorizer", "confirmed"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/Whfr_a\r\n```\r\n; bin/opt -passes=slp-vectorizer -S test.ll\r\ndefine i1 @src(i1 %cmp4.118.i) {\r\n %cmp4.118.i.not = xor i1 %cmp4.118.i, true\r\n %brmerge = select i1 %cmp4.118.i.not, i1 true, i1 poison\r\n %.not = xor i1 poison, true\r\n %brmerge2 = select i1 %brmerge, i1 true, i1 %.not\r\n %.not3 = xor i1 poison, true\r\n %brmerge4 = select i1 %brmerge2, i1 true, i1 %.not3\r\n %.not5 = xor i1 poison, true\r\n %brmerge6 = select i1 %brmerge4, i1 true, i1 %.not5\r\n %.not7 = xor i1 poison, true\r\n %brmerge8 = select i1 %brmerge6, i1 true, i1 %.not7\r\n ret i1 %brmerge8\r\n}\r\n```\r\n```\r\ndefine i1 @tgt(i1 %cmp4.118.i) {\r\n %1 = insertelement <4 x i1> , i1 %cmp4.118.i, i32 0\r\n %2 = xor <4 x i1> %1, \r\n %.not7 = xor i1 poison, true\r\n %3 = freeze <4 x i1> %2\r\n %4 = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> %3)\r\n %op.rdx = select i1 %4, i1 true, i1 %.not7\r\n %op.rdx1 = select i1 poison, i1 true, i1 %op.rdx\r\n ret i1 %op.rdx1\r\n}\r\n```\r\ncc @alexey-bataev "}]}, "verified": true} {"bug_id": "105785", "issue_url": "https://github.com/llvm/llvm-project/issues/105785", "bug_type": "miscompilation", "base_commit": "1241c762c165972690c4edfb82ec7421c1e64658", "knowledge_cutoff": "2024-08-23T05:45:52Z", "lit_test_dir": ["llvm/test/Transforms/ConstraintElimination"], "hints": {"fix_commit": "85b6aac7c25f9d2a976a76045ace1e7afebb5965", "components": ["ConstraintElimination"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/ConstraintElimination.cpp": [[1464, 1470]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/ConstraintElimination.cpp": ["checkAndReplaceCmp"]}}, "patch": "commit 3ff9d92aae0945daa85ec6f85f05a3aeaaa9f962\nAuthor: Yingwei Zheng \nDate: Fri Aug 23 16:06:00 2024 +0800\n\n [ConstraintElim] Fix miscompilation caused by PR97974 (#105790)\n \n Fixes https://github.com/llvm/llvm-project/issues/105785.\n \n (cherry picked from commit 85b6aac7c25f9d2a976a76045ace1e7afebb5965)\n\ndiff --git a/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp b/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\nindex c31173879af1..37022104d0a9 100644\n--- a/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\n+++ b/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\n@@ -1464,7 +1464,7 @@ static bool checkAndReplaceCmp(CmpIntrinsic *I, ConstraintInfo &Info,\n ToRemove.push_back(I);\n return true;\n }\n- if (checkCondition(ICmpInst::ICMP_EQ, LHS, RHS, I, Info)) {\n+ if (checkCondition(ICmpInst::ICMP_EQ, LHS, RHS, I, Info).value_or(false)) {\n I->replaceAllUsesWith(ConstantInt::get(I->getType(), 0));\n ToRemove.push_back(I);\n return true;\n", "tests": [{"file": "llvm/test/Transforms/ConstraintElimination/pr105785.ll", "commands": ["opt -passes=constraint-elimination -S %s"], "tests": [{"test_name": "pr105785", "test_body": "define void @pr105785(ptr %p) {\nentry:\n br label %for.cond\n\nfor.cond: ; preds = %for.cond1, %entry\n %for.ind = phi i32 [ 0, %entry ], [ 1, %for.cond1 ]\n %cmp = icmp eq i32 %for.ind, 0\n br i1 %cmp, label %for.cond1, label %for.end6\n\nfor.cond1: ; preds = %for.body3, %for.cond\n %for.ind2 = phi i32 [ %inc, %for.body3 ], [ 0, %for.cond ]\n %cmp2 = icmp ult i32 %for.ind2, 3\n br i1 %cmp2, label %for.body3, label %for.cond\n\nfor.body3: ; preds = %for.cond1\n %scmp = call i32 @llvm.scmp.i32.i32(i32 %for.ind, i32 1)\n store i32 %scmp, ptr %p, align 4\n %inc = add nuw nsw i32 %for.ind2, 1\n br label %for.cond1\n\nfor.end6: ; preds = %for.cond\n ret void\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.scmp.i32.i32(i32, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[ConstraintElim] Miscompilation with cmp intrinsic", "body": "Reproducer: https://alive2.llvm.org/ce/z/Tvz2NA\r\n```\r\n; bin/opt -passes=constraint-elimination test.ll -S\r\ndefine void @h(ptr %p) {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %for.cond1, %entry\r\n %storemerge = phi i32 [ 0, %entry ], [ 1, %for.cond1 ]\r\n %cmp = icmp eq i32 %storemerge, 0\r\n br i1 %cmp, label %for.cond1, label %for.end6\r\n\r\nfor.cond1: ; preds = %for.cond, %for.body3\r\n %i.0 = phi i32 [ %inc, %for.body3 ], [ 0, %for.cond ]\r\n %cmp2 = icmp ult i32 %i.0, 3\r\n br i1 %cmp2, label %for.body3, label %for.cond\r\n\r\nfor.body3: ; preds = %for.cond1\r\n %sub.i = tail call range(i32 -1, 2) i32 @llvm.scmp.i32.i32(i32 1, i32 %storemerge)\r\n store i32 %sub.i, ptr %p, align 4\r\n %inc = add nuw nsw i32 %i.0, 1\r\n br label %for.cond1\r\n\r\nfor.end6:\r\n ret void\r\n}\r\n```\r\n\r\n```\r\ndefine void @h(ptr %p) {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %for.cond1, %entry\r\n %storemerge = phi i32 [ 0, %entry ], [ 1, %for.cond1 ]\r\n %cmp = icmp eq i32 %storemerge, 0\r\n br i1 %cmp, label %for.cond1, label %for.end6\r\n\r\nfor.cond1: ; preds = %for.body3, %for.cond\r\n %i.0 = phi i32 [ %inc, %for.body3 ], [ 0, %for.cond ]\r\n %cmp2 = icmp ult i32 %i.0, 3\r\n br i1 %cmp2, label %for.body3, label %for.cond\r\n\r\nfor.body3: ; preds = %for.cond1\r\n store i32 0, ptr %p, align 4\r\n %inc = add nuw nsw i32 %i.0, 1\r\n br label %for.cond1\r\n\r\nfor.end6: ; preds = %for.cond\r\n ret void\r\n}\r\n```\r\n\r\nThe result of scmp should be 1 instead of 0.\r\n\r\nReduced C reproducer:\r\n```\r\n#include \r\n#include \r\n\r\nint builtin_scmp(int d, int e) { return (d > e) - (d < e); }\r\nint32_t f = 0;\r\nint64_t g = 0;\r\nvoid h() {\r\n for (f = 0; f <= 0; f++) {\r\n int i;\r\n for (i = 0; i < 3; i++)\r\n g = builtin_scmp(1, f);\r\n }\r\n}\r\nint main() {\r\n h();\r\n printf(\"%d\\n\", (int)g);\r\n return 0;\r\n}\r\n```\r\n\r\nI will post a fix later.\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:transforms"], "comments": []}, "verified": true} {"bug_id": "80597", "issue_url": "https://github.com/llvm/llvm-project/issues/80597", "bug_type": "crash", "base_commit": "8f4d8945536e9fc45db0e349b91c2f4b3a9cae29", "knowledge_cutoff": "2024-02-04T14:06:35Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "cb8d83a77c25e529f58eba17bb1ec76069a04e90", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": [[802, 807]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp": ["InstCombinerImpl::SimplifyDemandedUseBits"]}}, "patch": "commit cb8d83a77c25e529f58eba17bb1ec76069a04e90\nAuthor: Yingwei Zheng \nDate: Tue Feb 6 01:29:38 2024 +0800\n\n [InstCombine] Fix assertion failure in issue80597 (#80614)\n \n The assertion in #80597 failed when we were trying to compute known bits\n of a value in an unreachable BB.\n \n https://github.com/llvm/llvm-project/blob/859b09da08c2a47026ba0a7d2f21b7dca705864d/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp#L749-L810\n \n In this case, `SignBits` is 30 (deduced from instr info), but `Known` is\n `10000101010111010011110101000?0?00000000000000000000000000000000`\n (deduced from dom cond). Setting high bits of `lshr Known, 1` will lead\n to conflict.\n \n This patch masks out high bits of `Known.Zero` to address this problem.\n \n Fixes #80597.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\nindex a8a5f9831e15..79873a9b4cbb 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp\n@@ -802,6 +802,9 @@ Value *InstCombinerImpl::SimplifyDemandedUseBits(Value *V, APInt DemandedMask,\n return InsertNewInstWith(LShr, I->getIterator());\n } else if (Known.One[BitWidth-ShiftAmt-1]) { // New bits are known one.\n Known.One |= HighBits;\n+ // SignBits may be out-of-sync with Known.countMinSignBits(). Mask out\n+ // high bits of Known.Zero to avoid conflicts.\n+ Known.Zero &= ~HighBits;\n }\n } else {\n computeKnownBits(I, Known, Depth, CxtI);\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/pr80597.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "pr80597", "test_body": "define i64 @pr80597(i1 %cond) {\nentry:\n %add = select i1 %cond, i64 0, i64 4294967293\n %add8 = shl i64 %add, 32\n %sext1 = add i64 %add8, 8836839514384105472\n %cmp = icmp ult i64 %sext1, -34359738368\n br i1 %cmp, label %if.then, label %if.else\n\nif.else: ; preds = %entry\n %sext2 = or i64 %add8, 8836839522974040064\n %ashr = ashr i64 %sext2, 1\n ret i64 %ashr\n\nif.then: ; preds = %entry\n ret i64 0\n}\n"}]}], "issue": {"title": "llvm crash: Assertion `!LHSKnown.hasConflict() && \"Bits known to be one AND zero?\"' failed.", "body": "llvm crashes in `instcombine`\r\n\r\nBisected to d77067d08a3f56dc2d0e6c95bd2852c943df743a, which was committed by @nikic \r\n\r\nCompiler explorer: https://godbolt.org/z/e4jrvf5Wr\r\n\r\n```console\r\n% cat reduced.ll\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine i32 @main(i1 %.b.i) {\r\nentry:\r\n %add.i = select i1 %.b.i, i64 0, i64 4294967293\r\n %add8.i = shl i64 %add.i, 32\r\n %sext.2.i = add i64 %add8.i, 8836839514384105472\r\n %cmp11.2.i = icmp ult i64 %sext.2.i, -34359738368\r\n br i1 %cmp11.2.i, label %return.loopexit.i, label %for.inc.2.i\r\n\r\nfor.inc.2.i: ; preds = %entry\r\n %sext.4.i = or i64 %add8.i, 8836839522974040064\r\n %conv9.4.i = ashr i64 %sext.4.i, 1\r\n %cmp11.4.i = icmp ult i64 %conv9.4.i, 0\r\n br i1 %cmp11.4.i, label %return.loopexit.i, label %return.loopexit.i\r\n\r\nreturn.loopexit.i: ; preds = %for.inc.2.i, %for.inc.2.i, %entry\r\n ret i32 0\r\n}\r\n%\r\n% opt -disable-output -passes=instcombine reduced.ll\r\nopt: /tmp/tmpzoi34eaf/tmporhhh5ti/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp:246: llvm::Value* llvm::InstCombinerImpl::SimplifyDemandedUseBits(llvm::Value*, llvm::APInt, llvm::KnownBits&, unsigned int, llvm::Instruction*): Assertion `!LHSKnown.hasConflict() && \"Bits known to be one AND zero?\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/opt -disable-output -passes=instcombine reduced.ll\r\n #0 0x00007f7698f4558f llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x93858f)\r\n #1 0x00007f7698f42ab4 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007f7698222090 (/lib/x86_64-linux-gnu/libc.so.6+0x43090)\r\n #3 0x00007f769822200b raise /build/glibc-wuryBv/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007f7698201859 abort /build/glibc-wuryBv/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007f7698201729 get_sysdep_segment_value /build/glibc-wuryBv/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007f7698201729 _nl_load_domain /build/glibc-wuryBv/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007f7698212fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x00007f769a44445d llvm::InstCombinerImpl::SimplifyDemandedUseBits(llvm::Value*, llvm::APInt, llvm::KnownBits&, unsigned int, llvm::Instruction*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x1e3745d)\r\n #9 0x00007f769a444d25 llvm::InstCombinerImpl::SimplifyDemandedInstructionBits(llvm::Instruction&, llvm::KnownBits&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x1e37d25)\r\n#10 0x00007f769a4459c5 llvm::InstCombinerImpl::SimplifyDemandedInstructionBits(llvm::Instruction&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x1e389c5)\r\n#11 0x00007f769a38cdc3 llvm::InstCombinerImpl::visitOr(llvm::BinaryOperator&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x1d7fdc3)\r\n#12 0x00007f769a3537b1 llvm::InstCombinerImpl::run() (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x1d467b1)\r\n#13 0x00007f769a355002 combineInstructionsOverFunction(llvm::Function&, llvm::InstructionWorklist&, llvm::AAResults*, llvm::AssumptionCache&, llvm::TargetLibraryInfo&, llvm::TargetTransformInfo&, llvm::DominatorTree&, llvm::OptimizationRemarkEmitter&, llvm::BlockFrequencyInfo*, llvm::ProfileSummaryInfo*, llvm::LoopInfo*, llvm::InstCombineOptions const&) (.isra.0) InstructionCombining.cpp:0:0\r\n#14 0x00007f769a355fbb llvm::InstCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x1d48fbb)\r\n#15 0x00007f769bea5916 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x3898916)\r\n#16 0x00007f769916239d llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0xb5539d)\r\n#17 0x00007f769be9e506 llvm::detail::PassModel>, llvm::PreservedAnalyses, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x3891506)\r\n#18 0x00007f7699160c2d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0xb53c2d)\r\n#19 0x00007f769bea5556 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0x3898556)\r\n#20 0x00007f769915eb5d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/../lib/libLLVM-18git.so+0xb51b5d)\r\n#21 0x000055c7b779e395 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/opt+0x26395)\r\n#22 0x000055c7b7791e8a main (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/opt+0x19e8a)\r\n#23 0x00007f7698203083 __libc_start_main /build/glibc-wuryBv/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#24 0x000055c7b7792a5e _start (/zdata/shaoli/compilers/ccbuilder-compilers/clang-d77067d08a3f56dc2d0e6c95bd2852c943df743a/bin/opt+0x1aa5e)\r\n[1] 880216 abort -disable-output -passes=instcombine reduced.ll\r\n%\r\n```\r\n\r\nThe above test case was reduced from\r\n```C\r\nlong a, i;\r\nstatic long b = -4;\r\nchar c;\r\nint d;\r\nstatic int *e = &d, *h = &d;\r\nstatic long *f = &b, *g = &b;\r\nint j() {\r\n for (; a;) {\r\n c = 0;\r\n for (; c <= 6; c++) {\r\n unsigned long k = 1;\r\n i = -8;\r\n *h = i >= (int)k + (int)*f + c + 2057487030 + k;\r\n if (*e)\r\n break;\r\n }\r\n return 0;\r\n }\r\n *g = 0;\r\n return 0;\r\n}\r\nint main() { j(); }\r\n```", "author": "shao-hua-li", "labels": ["release:backport", "llvm:instcombine", "crash-on-valid"], "comments": [{"author": "dtcxzyw", "body": "Minimal reproduction: https://godbolt.org/z/5bzaqP65P"}, {"author": "dtcxzyw", "body": "Reopen this issue to backport the fix to the 18.X release.\r\n"}]}, "verified": true} {"bug_id": "105904", "issue_url": "https://github.com/llvm/llvm-project/issues/105904", "bug_type": "crash", "base_commit": "499e13514aaf2efdcd85520ade791ed635502adb", "knowledge_cutoff": "2024-08-23T22:57:57Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "b9d3da8c8d277a7fc2223c659122bb377a0e54e0", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[3011, 3017]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["isOperandGatherNode"]}}, "patch": "commit b9d3da8c8d277a7fc2223c659122bb377a0e54e0\nAuthor: Alexey Bataev \nDate: Mon Aug 26 04:31:44 2024 -0700\n\n [SLP]Fix PR105904: the root node might be a gather node without user for reductions.\n \n Before checking the user components of the gather/buildvector nodes,\n need to check if the node has users at all. Root nodes might not have\n users, if it is a node for the reduction.\n \n Fixes https://github.com/llvm/llvm-project/issues/105904\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 949579772b94..def73e8d0c0d 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -3011,7 +3011,8 @@ private:\n }\n \n bool isOperandGatherNode(const EdgeInfo &UserEI) const {\n- return isGather() && UserTreeIndices.front().EdgeIdx == UserEI.EdgeIdx &&\n+ return isGather() && (Idx > 0 || !UserTreeIndices.empty()) &&\n+ UserTreeIndices.front().EdgeIdx == UserEI.EdgeIdx &&\n UserTreeIndices.front().UserTE == UserEI.UserTE;\n }\n \n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/RISCV/gather-node-with-no-users.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=riscv64-unknown-linux-gnu -mattr=+v,+zvl512b < %s"], "tests": [{"test_name": "test", "test_body": "define void @test(ptr %c) {\nentry:\n %arrayidx8.5.3 = getelementptr i8, ptr %c, i64 222\n %0 = load i8, ptr %arrayidx8.5.3, align 1\n %arrayidx8.7.3 = getelementptr i8, ptr %c, i64 228\n %1 = load i8, ptr %arrayidx8.7.3, align 1\n %arrayidx8.434 = getelementptr i8, ptr %c, i64 276\n %2 = load i8, ptr %arrayidx8.434, align 1\n %arrayidx8.1.4 = getelementptr i8, ptr %c, i64 279\n %3 = load i8, ptr %arrayidx8.1.4, align 1\n %arrayidx8.2.4 = getelementptr i8, ptr %c, i64 282\n %4 = load i8, ptr %arrayidx8.2.4, align 1\n %arrayidx8.3.4 = getelementptr i8, ptr %c, i64 285\n %5 = load i8, ptr %arrayidx8.3.4, align 1\n %arrayidx8.4.4 = getelementptr i8, ptr %c, i64 288\n %6 = load i8, ptr %arrayidx8.4.4, align 1\n %7 = load i8, ptr %c, align 1\n %8 = load i8, ptr %c, align 1\n %arrayidx8.536 = getelementptr i8, ptr %c, i64 345\n %9 = load i8, ptr %arrayidx8.536, align 1\n %arrayidx8.1.5 = getelementptr i8, ptr %c, i64 348\n %10 = load i8, ptr %arrayidx8.1.5, align 1\n %arrayidx8.2.5 = getelementptr i8, ptr %c, i64 351\n %11 = load i8, ptr %arrayidx8.2.5, align 1\n %arrayidx8.3.5 = getelementptr i8, ptr %c, i64 354\n %12 = load i8, ptr %arrayidx8.3.5, align 1\n %arrayidx8.4.5 = getelementptr i8, ptr %c, i64 357\n %13 = load i8, ptr %arrayidx8.4.5, align 1\n %arrayidx8.5.5 = getelementptr i8, ptr %c, i64 360\n %14 = load i8, ptr %arrayidx8.5.5, align 1\n %arrayidx8.6.5 = getelementptr i8, ptr %c, i64 363\n %15 = load i8, ptr %arrayidx8.6.5, align 1\n br label %for.cond\n\nfor.cond: ; preds = %for.cond, %entry\n %a.promoted2226 = phi i8 [ 0, %entry ], [ %or18.6.5, %for.cond ]\n %or18.7.3 = or i8 %0, %1\n %or18.435 = or i8 %or18.7.3, %2\n %or18.1.4 = or i8 %or18.435, %3\n %or18.2.4 = or i8 %or18.1.4, %4\n %or18.3.4 = or i8 %or18.2.4, %5\n %or18.4.4 = or i8 %or18.3.4, %6\n %or18.5.4 = or i8 %or18.4.4, %7\n %or18.6.4 = or i8 %or18.5.4, %8\n %or18.537 = or i8 %or18.6.4, %9\n %or18.1.5 = or i8 %or18.537, %10\n %or18.2.5 = or i8 %or18.1.5, %11\n %or18.3.5 = or i8 %or18.2.5, %12\n %or18.4.5 = or i8 %or18.3.5, %13\n %or18.5.5 = or i8 %or18.4.5, %14\n %or18.6.5 = or i8 %or18.5.5, %15\n br label %for.cond\n}\n"}]}], "issue": {"title": "[SLP Vectorizer] Assertion `!empty()' failed.", "body": "Testcase:\r\n```llvm ir\r\n ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @b(ptr %c) #0 {\r\nentry:\r\n %arrayidx8.5.3 = getelementptr i8, ptr %c, i64 222\r\n %0 = load i8, ptr %arrayidx8.5.3, align 1\r\n %arrayidx8.7.3 = getelementptr i8, ptr %c, i64 228\r\n %1 = load i8, ptr %arrayidx8.7.3, align 1\r\n %arrayidx8.434 = getelementptr i8, ptr %c, i64 276\r\n %2 = load i8, ptr %arrayidx8.434, align 1\r\n %arrayidx8.1.4 = getelementptr i8, ptr %c, i64 279\r\n %3 = load i8, ptr %arrayidx8.1.4, align 1\r\n %arrayidx8.2.4 = getelementptr i8, ptr %c, i64 282\r\n %4 = load i8, ptr %arrayidx8.2.4, align 1\r\n %arrayidx8.3.4 = getelementptr i8, ptr %c, i64 285\r\n %5 = load i8, ptr %arrayidx8.3.4, align 1\r\n %arrayidx8.4.4 = getelementptr i8, ptr %c, i64 288\r\n %6 = load i8, ptr %arrayidx8.4.4, align 1\r\n %7 = load i8, ptr %c, align 1\r\n %8 = load i8, ptr %c, align 1\r\n %arrayidx8.536 = getelementptr i8, ptr %c, i64 345\r\n %9 = load i8, ptr %arrayidx8.536, align 1\r\n %arrayidx8.1.5 = getelementptr i8, ptr %c, i64 348\r\n %10 = load i8, ptr %arrayidx8.1.5, align 1\r\n %arrayidx8.2.5 = getelementptr i8, ptr %c, i64 351\r\n %11 = load i8, ptr %arrayidx8.2.5, align 1\r\n %arrayidx8.3.5 = getelementptr i8, ptr %c, i64 354\r\n %12 = load i8, ptr %arrayidx8.3.5, align 1\r\n %arrayidx8.4.5 = getelementptr i8, ptr %c, i64 357\r\n %13 = load i8, ptr %arrayidx8.4.5, align 1\r\n %arrayidx8.5.5 = getelementptr i8, ptr %c, i64 360\r\n %14 = load i8, ptr %arrayidx8.5.5, align 1\r\n %arrayidx8.6.5 = getelementptr i8, ptr %c, i64 363\r\n %15 = load i8, ptr %arrayidx8.6.5, align 1\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %for.cond, %entry\r\n %a.promoted2226 = phi i8 [ 0, %entry ], [ %or18.6.5, %for.cond ]\r\n %or18.7.3 = or i8 %0, %1\r\n %or18.435 = or i8 %or18.7.3, %2\r\n %or18.1.4 = or i8 %or18.435, %3\r\n %or18.2.4 = or i8 %or18.1.4, %4\r\n %or18.3.4 = or i8 %or18.2.4, %5\r\n %or18.4.4 = or i8 %or18.3.4, %6\r\n %or18.5.4 = or i8 %or18.4.4, %7\r\n %or18.6.4 = or i8 %or18.5.4, %8\r\n %or18.537 = or i8 %or18.6.4, %9\r\n %or18.1.5 = or i8 %or18.537, %10\r\n %or18.2.5 = or i8 %or18.1.5, %11\r\n %or18.3.5 = or i8 %or18.2.5, %12\r\n %or18.4.5 = or i8 %or18.3.5, %13\r\n %or18.5.5 = or i8 %or18.4.5, %14\r\n %or18.6.5 = or i8 %or18.5.5, %15\r\n br label %for.cond\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+relax,+xsifivecflushdlone,+zicsr,+zve32x,+zve64x,+zvl1024b,+zvl128b,+zvl256b,+zvl32b,+zvl512b,+zvl64b,-a,-b,-c,-d,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-f,-h,-m,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-v,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zifencei,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-zmmul,-ztso,-zvbb,-zvbc,-zve32f,-zve64d,-zve64f,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl16384b,-zvl2048b,-zvl32768b,-zvl4096b,-zvl65536b,-zvl8192b\" }\r\n```\r\nGodbolt: https://godbolt.org/z/Thhajqn9q\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt --passes=slp-vectorizer -S reduced.ll\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/include/llvm/ADT/SmallVector.h:317: const T& llvm::SmallVectorTemplateCommon >::front() const [with T = llvm::slpvectorizer::BoUpSLP::EdgeInfo; = void; llvm::SmallVectorTemplateCommon >::const_reference = const llvm::slpvectorizer::BoUpSLP::EdgeInfo&]: Assertion `!empty()' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt --passes=slp-vectorizer -S reduced.ll\r\n1. Running pass \"function(slp-vectorizer)\" on module \"reduced.ll\"\r\n2. Running pass \"slp-vectorizer\" on function \"b\"\r\n #0 0x00005da37ad02b80 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x300eb80)\r\n #1 0x00005da37acfff9f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x300bf9f)\r\n #2 0x00005da37ad000f5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x0000733cc0242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x0000733cc02969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x0000733cc02969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x0000733cc02969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x0000733cc0242476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x0000733cc02287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x0000733cc022871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x0000733cc0239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00005da379da2beb llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20aebeb)\r\n#12 0x00005da379d9e086 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20aa086)\r\n#13 0x00005da379da2372 llvm::slpvectorizer::BoUpSLP::vectorizeOperand(llvm::slpvectorizer::BoUpSLP::TreeEntry*, unsigned int, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20ae372)\r\n#14 0x00005da379d9db47 llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20a9b47)\r\n#15 0x00005da379d9c510 llvm::slpvectorizer::BoUpSLP::createBuildVector(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20a8510)\r\n#16 0x00005da379d9cf7b llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20a8f7b)\r\n#17 0x00005da379da62cf llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::MapVector, llvm::DenseMap, llvm::detail::DenseMapPair>, llvm::SmallVector>, 0u>> const&, llvm::SmallVectorImpl>&, llvm::Instruction*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20b22cf)\r\n#18 0x00005da379db2bd5 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\r\n#19 0x00005da379db5599 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*, llvm::SmallVectorImpl&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20c1599)\r\n#20 0x00005da379db9a68 llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::TargetTransformInfo*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20c5a68)\r\n#21 0x00005da379dbd75c llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20c975c)\r\n#22 0x00005da379dc41b0 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\r\n#23 0x00005da379dc4da2 llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20d0da2)\r\n#24 0x00005da378b51cf6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5dcf6)\r\n#25 0x00005da37ab0e94f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1a94f)\r\n#26 0x00005da378b544f6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe604f6)\r\n#27 0x00005da37ab0d6cb llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e196cb)\r\n#28 0x00005da378b50796 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe5c796)\r\n#29 0x00005da37ab0b70d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e1770d)\r\n#30 0x00005da37837ae56 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x686e56)\r\n#31 0x00005da37836cbe1 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x678be1)\r\n#32 0x0000733cc0229d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#33 0x0000733cc0229e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#34 0x0000733cc0229e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#35 0x00005da378362f95 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x66ef95)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\n\r\nFound via fuzzer", "author": "patrick-rivos", "labels": ["llvm:SLPVectorizer", "crash"], "comments": []}, "verified": true} {"bug_id": "108098", "issue_url": "https://github.com/llvm/llvm-project/issues/108098", "bug_type": "crash", "base_commit": "b8239e1201f5871bed5b633b76fa9536672f287f", "knowledge_cutoff": "2024-09-10T21:48:37Z", "lit_test_dir": ["llvm/test/Analysis/CostModel", "llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "ea83e1c05a5adee5d8f9e680356ee57556ba64a1", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": [[1182, 1194]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/LoopVectorize.cpp": ["setWideningDecision"]}}, "patch": "commit ea83e1c05a5adee5d8f9e680356ee57556ba64a1\nAuthor: Florian Hahn \nDate: Wed Sep 11 21:04:33 2024 +0100\n\n [LV] Assign cost to all interleave members when not interleaving.\n \n At the moment, the full cost of all interleave group members is assigned\n to the instruction at the group's insert position, even if the decision\n was to not form an interleave group.\n \n This can lead to inaccurate cost estimates, e.g. if the instruction at\n the insert position is dead. If the decision is to not vectorize but\n scalarize or scather/gather, then the cost will be to total cost for all\n members. In those cases, assign individual the cost per member, to more\n closely reflect to choice per instruction.\n \n This fixes a divergence between legacy and VPlan-based cost model.\n \n Fixes https://github.com/llvm/llvm-project/issues/108098.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\nindex 640a7bf3d672..3b6b154b9660 100644\n--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp\n@@ -1182,13 +1182,23 @@ public:\n InstructionCost Cost) {\n assert(VF.isVector() && \"Expected VF >=2\");\n /// Broadcast this decicion to all instructions inside the group.\n- /// But the cost will be assigned to one instruction only.\n+ /// When interleaving, the cost will only be assigned one instruction, the\n+ /// insert position. For other cases, add the appropriate fraction of the\n+ /// total cost to each instruction. This ensures accurate costs are used,\n+ /// even if the insert position instruction is not used.\n+ InstructionCost InsertPosCost = Cost;\n+ InstructionCost OtherMemberCost = 0;\n+ if (W != CM_Interleave)\n+ OtherMemberCost = InsertPosCost = Cost / Grp->getNumMembers();\n+ ;\n for (unsigned Idx = 0; Idx < Grp->getFactor(); ++Idx) {\n if (auto *I = Grp->getMember(Idx)) {\n if (Grp->getInsertPos() == I)\n- WideningDecisions[std::make_pair(I, VF)] = std::make_pair(W, Cost);\n+ WideningDecisions[std::make_pair(I, VF)] =\n+ std::make_pair(W, InsertPosCost);\n else\n- WideningDecisions[std::make_pair(I, VF)] = std::make_pair(W, 0);\n+ WideningDecisions[std::make_pair(I, VF)] =\n+ std::make_pair(W, OtherMemberCost);\n }\n }\n }\n", "tests": [{"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f32-stride-5.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x float], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %in0 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.4\n %v0 = load float, ptr %in0, align 4\n %v1 = load float, ptr %in1, align 4\n %v2 = load float, ptr %in2, align 4\n %v3 = load float, ptr %in3, align 4\n %v4 = load float, ptr %in4, align 4\n %reduce.add.0 = fadd float %v0, %v1\n %reduce.add.1 = fadd float %reduce.add.0, %v2\n %reduce.add.2 = fadd float %reduce.add.1, %v3\n %reduce.add.3 = fadd float %reduce.add.2, %v4\n %reduce.add.3.narrow = fptoui float %reduce.add.3 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.3.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 5\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f32-stride-7.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x float], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %in0 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.6\n %v0 = load float, ptr %in0, align 4\n %v1 = load float, ptr %in1, align 4\n %v2 = load float, ptr %in2, align 4\n %v3 = load float, ptr %in3, align 4\n %v4 = load float, ptr %in4, align 4\n %v5 = load float, ptr %in5, align 4\n %v6 = load float, ptr %in6, align 4\n %reduce.add.0 = fadd float %v0, %v1\n %reduce.add.1 = fadd float %reduce.add.0, %v2\n %reduce.add.2 = fadd float %reduce.add.1, %v3\n %reduce.add.3 = fadd float %reduce.add.2, %v4\n %reduce.add.4 = fadd float %reduce.add.3, %v5\n %reduce.add.5 = fadd float %reduce.add.4, %v6\n %reduce.add.5.narrow = fptoui float %reduce.add.5 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.5.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 7\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f32-stride-8.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x float], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %iv.7 = add nuw nsw i64 %iv, 7\n %in0 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.6\n %in7 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.7\n %v0 = load float, ptr %in0, align 4\n %v1 = load float, ptr %in1, align 4\n %v2 = load float, ptr %in2, align 4\n %v3 = load float, ptr %in3, align 4\n %v4 = load float, ptr %in4, align 4\n %v5 = load float, ptr %in5, align 4\n %v6 = load float, ptr %in6, align 4\n %v7 = load float, ptr %in7, align 4\n %reduce.add.0 = fadd float %v0, %v1\n %reduce.add.1 = fadd float %reduce.add.0, %v2\n %reduce.add.2 = fadd float %reduce.add.1, %v3\n %reduce.add.3 = fadd float %reduce.add.2, %v4\n %reduce.add.4 = fadd float %reduce.add.3, %v5\n %reduce.add.5 = fadd float %reduce.add.4, %v6\n %reduce.add.6 = fadd float %reduce.add.5, %v7\n %reduce.add.6.narrow = fptoui float %reduce.add.6 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.6.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 8\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f64-stride-2.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x double], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %in0 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.1\n %v0 = load double, ptr %in0, align 8\n %v1 = load double, ptr %in1, align 8\n %reduce.add.0 = fadd double %v0, %v1\n %reduce.add.0.narrow = fptoui double %reduce.add.0 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.0.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 2\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f64-stride-3.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x double], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %in0 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.2\n %v0 = load double, ptr %in0, align 8\n %v1 = load double, ptr %in1, align 8\n %v2 = load double, ptr %in2, align 8\n %reduce.add.0 = fadd double %v0, %v1\n %reduce.add.1 = fadd double %reduce.add.0, %v2\n %reduce.add.1.narrow = fptoui double %reduce.add.1 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.1.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 3\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f64-stride-4.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x double], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %in0 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.3\n %v0 = load double, ptr %in0, align 8\n %v1 = load double, ptr %in1, align 8\n %v2 = load double, ptr %in2, align 8\n %v3 = load double, ptr %in3, align 8\n %reduce.add.0 = fadd double %v0, %v1\n %reduce.add.1 = fadd double %reduce.add.0, %v2\n %reduce.add.2 = fadd double %reduce.add.1, %v3\n %reduce.add.2.narrow = fptoui double %reduce.add.2 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.2.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 4\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f64-stride-5.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x double], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %in0 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.4\n %v0 = load double, ptr %in0, align 8\n %v1 = load double, ptr %in1, align 8\n %v2 = load double, ptr %in2, align 8\n %v3 = load double, ptr %in3, align 8\n %v4 = load double, ptr %in4, align 8\n %reduce.add.0 = fadd double %v0, %v1\n %reduce.add.1 = fadd double %reduce.add.0, %v2\n %reduce.add.2 = fadd double %reduce.add.1, %v3\n %reduce.add.3 = fadd double %reduce.add.2, %v4\n %reduce.add.3.narrow = fptoui double %reduce.add.3 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.3.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 5\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f64-stride-6.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x double], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %in0 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.5\n %v0 = load double, ptr %in0, align 8\n %v1 = load double, ptr %in1, align 8\n %v2 = load double, ptr %in2, align 8\n %v3 = load double, ptr %in3, align 8\n %v4 = load double, ptr %in4, align 8\n %v5 = load double, ptr %in5, align 8\n %reduce.add.0 = fadd double %v0, %v1\n %reduce.add.1 = fadd double %reduce.add.0, %v2\n %reduce.add.2 = fadd double %reduce.add.1, %v3\n %reduce.add.3 = fadd double %reduce.add.2, %v4\n %reduce.add.4 = fadd double %reduce.add.3, %v5\n %reduce.add.4.narrow = fptoui double %reduce.add.4 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.4.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 6\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f64-stride-7.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x double], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %in0 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.6\n %v0 = load double, ptr %in0, align 8\n %v1 = load double, ptr %in1, align 8\n %v2 = load double, ptr %in2, align 8\n %v3 = load double, ptr %in3, align 8\n %v4 = load double, ptr %in4, align 8\n %v5 = load double, ptr %in5, align 8\n %v6 = load double, ptr %in6, align 8\n %reduce.add.0 = fadd double %v0, %v1\n %reduce.add.1 = fadd double %reduce.add.0, %v2\n %reduce.add.2 = fadd double %reduce.add.1, %v3\n %reduce.add.3 = fadd double %reduce.add.2, %v4\n %reduce.add.4 = fadd double %reduce.add.3, %v5\n %reduce.add.5 = fadd double %reduce.add.4, %v6\n %reduce.add.5.narrow = fptoui double %reduce.add.5 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.5.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 7\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-f64-stride-8.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x double], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %iv.7 = add nuw nsw i64 %iv, 7\n %in0 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.6\n %in7 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.7\n %v0 = load double, ptr %in0, align 8\n %v1 = load double, ptr %in1, align 8\n %v2 = load double, ptr %in2, align 8\n %v3 = load double, ptr %in3, align 8\n %v4 = load double, ptr %in4, align 8\n %v5 = load double, ptr %in5, align 8\n %v6 = load double, ptr %in6, align 8\n %v7 = load double, ptr %in7, align 8\n %reduce.add.0 = fadd double %v0, %v1\n %reduce.add.1 = fadd double %reduce.add.0, %v2\n %reduce.add.2 = fadd double %reduce.add.1, %v3\n %reduce.add.3 = fadd double %reduce.add.2, %v4\n %reduce.add.4 = fadd double %reduce.add.3, %v5\n %reduce.add.5 = fadd double %reduce.add.4, %v6\n %reduce.add.6 = fadd double %reduce.add.5, %v7\n %reduce.add.6.narrow = fptoui double %reduce.add.6 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.6.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 8\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i16-stride-5.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl,+avx512bw --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i16], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %in0 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.4\n %v0 = load i16, ptr %in0, align 2\n %v1 = load i16, ptr %in1, align 2\n %v2 = load i16, ptr %in2, align 2\n %v3 = load i16, ptr %in3, align 2\n %v4 = load i16, ptr %in4, align 2\n %reduce.add.0 = add i16 %v0, %v1\n %reduce.add.1 = add i16 %reduce.add.0, %v2\n %reduce.add.2 = add i16 %reduce.add.1, %v3\n %reduce.add.3 = add i16 %reduce.add.2, %v4\n %reduce.add.3.narrow = trunc i16 %reduce.add.3 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.3.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 5\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i16-stride-7.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl,+avx512bw --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i16], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %in0 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.6\n %v0 = load i16, ptr %in0, align 2\n %v1 = load i16, ptr %in1, align 2\n %v2 = load i16, ptr %in2, align 2\n %v3 = load i16, ptr %in3, align 2\n %v4 = load i16, ptr %in4, align 2\n %v5 = load i16, ptr %in5, align 2\n %v6 = load i16, ptr %in6, align 2\n %reduce.add.0 = add i16 %v0, %v1\n %reduce.add.1 = add i16 %reduce.add.0, %v2\n %reduce.add.2 = add i16 %reduce.add.1, %v3\n %reduce.add.3 = add i16 %reduce.add.2, %v4\n %reduce.add.4 = add i16 %reduce.add.3, %v5\n %reduce.add.5 = add i16 %reduce.add.4, %v6\n %reduce.add.5.narrow = trunc i16 %reduce.add.5 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.5.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 7\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i16-stride-8.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl,+avx512bw --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i16], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %iv.7 = add nuw nsw i64 %iv, 7\n %in0 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.6\n %in7 = getelementptr inbounds [1024 x i16], ptr @A, i64 0, i64 %iv.7\n %v0 = load i16, ptr %in0, align 2\n %v1 = load i16, ptr %in1, align 2\n %v2 = load i16, ptr %in2, align 2\n %v3 = load i16, ptr %in3, align 2\n %v4 = load i16, ptr %in4, align 2\n %v5 = load i16, ptr %in5, align 2\n %v6 = load i16, ptr %in6, align 2\n %v7 = load i16, ptr %in7, align 2\n %reduce.add.0 = add i16 %v0, %v1\n %reduce.add.1 = add i16 %reduce.add.0, %v2\n %reduce.add.2 = add i16 %reduce.add.1, %v3\n %reduce.add.3 = add i16 %reduce.add.2, %v4\n %reduce.add.4 = add i16 %reduce.add.3, %v5\n %reduce.add.5 = add i16 %reduce.add.4, %v6\n %reduce.add.6 = add i16 %reduce.add.5, %v7\n %reduce.add.6.narrow = trunc i16 %reduce.add.6 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.6.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 8\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i32-stride-4-indices-01uu.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i32], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %in0 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.1\n %v0 = load i32, ptr %in0, align 4\n %v1 = load i32, ptr %in1, align 4\n %reduce.add.0 = add i32 %v0, %v1\n %reduce.add.1 = add i32 %reduce.add.0, 0\n %reduce.add.2 = add i32 %reduce.add.1, 0\n %reduce.add.2.narrow = trunc i32 %reduce.add.2 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.2.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv, 4\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i32-stride-5.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i32], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %in0 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.4\n %v0 = load i32, ptr %in0, align 4\n %v1 = load i32, ptr %in1, align 4\n %v2 = load i32, ptr %in2, align 4\n %v3 = load i32, ptr %in3, align 4\n %v4 = load i32, ptr %in4, align 4\n %reduce.add.0 = add i32 %v0, %v1\n %reduce.add.1 = add i32 %reduce.add.0, %v2\n %reduce.add.2 = add i32 %reduce.add.1, %v3\n %reduce.add.3 = add i32 %reduce.add.2, %v4\n %reduce.add.3.narrow = trunc i32 %reduce.add.3 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.3.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 5\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i32-stride-7.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i32], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %in0 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.6\n %v0 = load i32, ptr %in0, align 4\n %v1 = load i32, ptr %in1, align 4\n %v2 = load i32, ptr %in2, align 4\n %v3 = load i32, ptr %in3, align 4\n %v4 = load i32, ptr %in4, align 4\n %v5 = load i32, ptr %in5, align 4\n %v6 = load i32, ptr %in6, align 4\n %reduce.add.0 = add i32 %v0, %v1\n %reduce.add.1 = add i32 %reduce.add.0, %v2\n %reduce.add.2 = add i32 %reduce.add.1, %v3\n %reduce.add.3 = add i32 %reduce.add.2, %v4\n %reduce.add.4 = add i32 %reduce.add.3, %v5\n %reduce.add.5 = add i32 %reduce.add.4, %v6\n %reduce.add.5.narrow = trunc i32 %reduce.add.5 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.5.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 7\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i32-stride-8.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i32], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %iv.7 = add nuw nsw i64 %iv, 7\n %in0 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.6\n %in7 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.7\n %v0 = load i32, ptr %in0, align 4\n %v1 = load i32, ptr %in1, align 4\n %v2 = load i32, ptr %in2, align 4\n %v3 = load i32, ptr %in3, align 4\n %v4 = load i32, ptr %in4, align 4\n %v5 = load i32, ptr %in5, align 4\n %v6 = load i32, ptr %in6, align 4\n %v7 = load i32, ptr %in7, align 4\n %reduce.add.0 = add i32 %v0, %v1\n %reduce.add.1 = add i32 %reduce.add.0, %v2\n %reduce.add.2 = add i32 %reduce.add.1, %v3\n %reduce.add.3 = add i32 %reduce.add.2, %v4\n %reduce.add.4 = add i32 %reduce.add.3, %v5\n %reduce.add.5 = add i32 %reduce.add.4, %v6\n %reduce.add.6 = add i32 %reduce.add.5, %v7\n %reduce.add.6.narrow = trunc i32 %reduce.add.6 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.6.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 8\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i64-stride-2.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i64], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %in0 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.1\n %v0 = load i64, ptr %in0, align 8\n %v1 = load i64, ptr %in1, align 8\n %reduce.add.0 = add i64 %v0, %v1\n %reduce.add.0.narrow = trunc i64 %reduce.add.0 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.0.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 2\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i64-stride-3.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i64], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %in0 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.2\n %v0 = load i64, ptr %in0, align 8\n %v1 = load i64, ptr %in1, align 8\n %v2 = load i64, ptr %in2, align 8\n %reduce.add.0 = add i64 %v0, %v1\n %reduce.add.1 = add i64 %reduce.add.0, %v2\n %reduce.add.1.narrow = trunc i64 %reduce.add.1 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.1.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 3\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i64-stride-4.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i64], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %in0 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.3\n %v0 = load i64, ptr %in0, align 8\n %v1 = load i64, ptr %in1, align 8\n %v2 = load i64, ptr %in2, align 8\n %v3 = load i64, ptr %in3, align 8\n %reduce.add.0 = add i64 %v0, %v1\n %reduce.add.1 = add i64 %reduce.add.0, %v2\n %reduce.add.2 = add i64 %reduce.add.1, %v3\n %reduce.add.2.narrow = trunc i64 %reduce.add.2 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.2.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 4\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i64-stride-5.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i64], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %in0 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.4\n %v0 = load i64, ptr %in0, align 8\n %v1 = load i64, ptr %in1, align 8\n %v2 = load i64, ptr %in2, align 8\n %v3 = load i64, ptr %in3, align 8\n %v4 = load i64, ptr %in4, align 8\n %reduce.add.0 = add i64 %v0, %v1\n %reduce.add.1 = add i64 %reduce.add.0, %v2\n %reduce.add.2 = add i64 %reduce.add.1, %v3\n %reduce.add.3 = add i64 %reduce.add.2, %v4\n %reduce.add.3.narrow = trunc i64 %reduce.add.3 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.3.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 5\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i64-stride-6.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i64], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %in0 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.5\n %v0 = load i64, ptr %in0, align 8\n %v1 = load i64, ptr %in1, align 8\n %v2 = load i64, ptr %in2, align 8\n %v3 = load i64, ptr %in3, align 8\n %v4 = load i64, ptr %in4, align 8\n %v5 = load i64, ptr %in5, align 8\n %reduce.add.0 = add i64 %v0, %v1\n %reduce.add.1 = add i64 %reduce.add.0, %v2\n %reduce.add.2 = add i64 %reduce.add.1, %v3\n %reduce.add.3 = add i64 %reduce.add.2, %v4\n %reduce.add.4 = add i64 %reduce.add.3, %v5\n %reduce.add.4.narrow = trunc i64 %reduce.add.4 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.4.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 6\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i64-stride-7.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i64], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %in0 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.6\n %v0 = load i64, ptr %in0, align 8\n %v1 = load i64, ptr %in1, align 8\n %v2 = load i64, ptr %in2, align 8\n %v3 = load i64, ptr %in3, align 8\n %v4 = load i64, ptr %in4, align 8\n %v5 = load i64, ptr %in5, align 8\n %v6 = load i64, ptr %in6, align 8\n %reduce.add.0 = add i64 %v0, %v1\n %reduce.add.1 = add i64 %reduce.add.0, %v2\n %reduce.add.2 = add i64 %reduce.add.1, %v3\n %reduce.add.3 = add i64 %reduce.add.2, %v4\n %reduce.add.4 = add i64 %reduce.add.3, %v5\n %reduce.add.5 = add i64 %reduce.add.4, %v6\n %reduce.add.5.narrow = trunc i64 %reduce.add.5 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.5.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 7\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i64-stride-8.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i64], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %iv.7 = add nuw nsw i64 %iv, 7\n %in0 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.6\n %in7 = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.7\n %v0 = load i64, ptr %in0, align 8\n %v1 = load i64, ptr %in1, align 8\n %v2 = load i64, ptr %in2, align 8\n %v3 = load i64, ptr %in3, align 8\n %v4 = load i64, ptr %in4, align 8\n %v5 = load i64, ptr %in5, align 8\n %v6 = load i64, ptr %in6, align 8\n %v7 = load i64, ptr %in7, align 8\n %reduce.add.0 = add i64 %v0, %v1\n %reduce.add.1 = add i64 %reduce.add.0, %v2\n %reduce.add.2 = add i64 %reduce.add.1, %v3\n %reduce.add.3 = add i64 %reduce.add.2, %v4\n %reduce.add.4 = add i64 %reduce.add.3, %v5\n %reduce.add.5 = add i64 %reduce.add.4, %v6\n %reduce.add.6 = add i64 %reduce.add.5, %v7\n %reduce.add.6.narrow = trunc i64 %reduce.add.6 to i8\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.6.narrow, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 8\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i8-stride-5.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl,+avx512bw --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i8], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %in0 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.4\n %v0 = load i8, ptr %in0, align 1\n %v1 = load i8, ptr %in1, align 1\n %v2 = load i8, ptr %in2, align 1\n %v3 = load i8, ptr %in3, align 1\n %v4 = load i8, ptr %in4, align 1\n %reduce.add.0 = add i8 %v0, %v1\n %reduce.add.1 = add i8 %reduce.add.0, %v2\n %reduce.add.2 = add i8 %reduce.add.1, %v3\n %reduce.add.3 = add i8 %reduce.add.2, %v4\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.3, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 5\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i8-stride-7.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl,+avx512bw --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i8], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %in0 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.6\n %v0 = load i8, ptr %in0, align 1\n %v1 = load i8, ptr %in1, align 1\n %v2 = load i8, ptr %in2, align 1\n %v3 = load i8, ptr %in3, align 1\n %v4 = load i8, ptr %in4, align 1\n %v5 = load i8, ptr %in5, align 1\n %v6 = load i8, ptr %in6, align 1\n %reduce.add.0 = add i8 %v0, %v1\n %reduce.add.1 = add i8 %reduce.add.0, %v2\n %reduce.add.2 = add i8 %reduce.add.1, %v3\n %reduce.add.3 = add i8 %reduce.add.2, %v4\n %reduce.add.4 = add i8 %reduce.add.3, %v5\n %reduce.add.5 = add i8 %reduce.add.4, %v6\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.5, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 7\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-load-i8-stride-8.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl,+avx512bw --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i8], align 128\n@B = external global [1024 x i8], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %iv.7 = add nuw nsw i64 %iv, 7\n %in0 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.0\n %in1 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.1\n %in2 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.2\n %in3 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.3\n %in4 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.4\n %in5 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.5\n %in6 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.6\n %in7 = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.7\n %v0 = load i8, ptr %in0, align 1\n %v1 = load i8, ptr %in1, align 1\n %v2 = load i8, ptr %in2, align 1\n %v3 = load i8, ptr %in3, align 1\n %v4 = load i8, ptr %in4, align 1\n %v5 = load i8, ptr %in5, align 1\n %v6 = load i8, ptr %in6, align 1\n %v7 = load i8, ptr %in7, align 1\n %reduce.add.0 = add i8 %v0, %v1\n %reduce.add.1 = add i8 %reduce.add.0, %v2\n %reduce.add.2 = add i8 %reduce.add.1, %v3\n %reduce.add.3 = add i8 %reduce.add.2, %v4\n %reduce.add.4 = add i8 %reduce.add.3, %v5\n %reduce.add.5 = add i8 %reduce.add.4, %v6\n %reduce.add.6 = add i8 %reduce.add.5, %v7\n %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0\n store i8 %reduce.add.6, ptr %out, align 1\n %iv.next = add nuw nsw i64 %iv.0, 8\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-store-f64-stride-8.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i8], align 128\n@B = external global [1024 x double], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %iv.7 = add nuw nsw i64 %iv, 7\n %in = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.0\n %v.narrow = load i8, ptr %in, align 1\n %v = uitofp i8 %v.narrow to double\n %v0 = fadd double %v, 0.000000e+00\n %v1 = fadd double %v, 1.000000e+00\n %v2 = fadd double %v, 2.000000e+00\n %v3 = fadd double %v, 3.000000e+00\n %v4 = fadd double %v, 4.000000e+00\n %v5 = fadd double %v, 5.000000e+00\n %v6 = fadd double %v, 6.000000e+00\n %v7 = fadd double %v, 7.000000e+00\n %out0 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.0\n %out1 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.1\n %out2 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.2\n %out3 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.3\n %out4 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.4\n %out5 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.5\n %out6 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.6\n %out7 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.7\n store double %v0, ptr %out0, align 8\n store double %v1, ptr %out1, align 8\n store double %v2, ptr %out2, align 8\n store double %v3, ptr %out3, align 8\n store double %v4, ptr %out4, align 8\n store double %v5, ptr %out5, align 8\n store double %v6, ptr %out6, align 8\n store double %v7, ptr %out7, align 8\n %iv.next = add nuw nsw i64 %iv.0, 8\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/interleaved-store-i64-stride-8.ll", "commands": ["opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\n@A = external global [1024 x i8], align 128\n@B = external global [1024 x i64], align 128\n\ndefine void @test() {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]\n %iv.0 = add nuw nsw i64 %iv, 0\n %iv.1 = add nuw nsw i64 %iv, 1\n %iv.2 = add nuw nsw i64 %iv, 2\n %iv.3 = add nuw nsw i64 %iv, 3\n %iv.4 = add nuw nsw i64 %iv, 4\n %iv.5 = add nuw nsw i64 %iv, 5\n %iv.6 = add nuw nsw i64 %iv, 6\n %iv.7 = add nuw nsw i64 %iv, 7\n %in = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.0\n %v.narrow = load i8, ptr %in, align 1\n %v = zext i8 %v.narrow to i64\n %v0 = add i64 %v, 0\n %v1 = add i64 %v, 1\n %v2 = add i64 %v, 2\n %v3 = add i64 %v, 3\n %v4 = add i64 %v, 4\n %v5 = add i64 %v, 5\n %v6 = add i64 %v, 6\n %v7 = add i64 %v, 7\n %out0 = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.0\n %out1 = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.1\n %out2 = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.2\n %out3 = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.3\n %out4 = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.4\n %out5 = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.5\n %out6 = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.6\n %out7 = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.7\n store i64 %v0, ptr %out0, align 8\n store i64 %v1, ptr %out1, align 8\n store i64 %v2, ptr %out2, align 8\n store i64 %v3, ptr %out3, align 8\n store i64 %v4, ptr %out4, align 8\n store i64 %v5, ptr %out5, align 8\n store i64 %v6, ptr %out6, align 8\n store i64 %v7, ptr %out7, align 8\n %iv.next = add nuw nsw i64 %iv.0, 8\n %cmp = icmp ult i64 %iv.next, 1024\n br i1 %cmp, label %for.body, label %for.cond.cleanup\n\nfor.cond.cleanup: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Analysis/CostModel/X86/masked-interleaved-store-i16.ll", "commands": ["opt -passes=loop-vectorize -enable-interleaved-mem-accesses -prefer-predicate-over-epilogue=predicate-dont-vectorize -S -mcpu=skx --debug-only=loop-vectorize < %s 2>&1", "opt -passes=loop-vectorize -enable-interleaved-mem-accesses -enable-masked-interleaved-mem-accesses -prefer-predicate-over-epilogue=predicate-dont-vectorize -S -mcpu=skx --debug-only=loop-vectorize < %s 2>&1"], "tests": [{"test_name": "test2", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @test2(ptr noalias nocapture %points, i32 %numPoints, ptr noalias nocapture readonly %x, ptr noalias nocapture readonly %y) {\nentry:\n %cmp15 = icmp sgt i32 %numPoints, 0\n br i1 %cmp15, label %for.body.preheader, label %for.end\n\nfor.body.preheader: ; preds = %entry\n %wide.trip.count = zext i32 %numPoints to i64\n br label %for.body\n\nfor.body: ; preds = %for.body, %for.body.preheader\n %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]\n %arrayidx = getelementptr inbounds i16, ptr %x, i64 %indvars.iv\n %0 = load i16, ptr %arrayidx, align 2\n %1 = shl nsw i64 %indvars.iv, 2\n %arrayidx2 = getelementptr inbounds i16, ptr %points, i64 %1\n store i16 %0, ptr %arrayidx2, align 2\n %arrayidx4 = getelementptr inbounds i16, ptr %y, i64 %indvars.iv\n %2 = load i16, ptr %arrayidx4, align 2\n %3 = or disjoint i64 %1, 1\n %arrayidx7 = getelementptr inbounds i16, ptr %points, i64 %3\n store i16 %2, ptr %arrayidx7, align 2\n %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\n %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count\n br i1 %exitcond.not, label %for.end.loopexit, label %for.body\n\nfor.end.loopexit: ; preds = %for.body\n br label %for.end\n\nfor.end: ; preds = %for.end.loopexit, %entry\n ret void\n}\n"}, {"test_name": "test1", "test_body": "target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @test1(ptr noalias nocapture %points, ptr noalias nocapture readonly %x, ptr noalias nocapture readonly %y) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]\n %arrayidx = getelementptr inbounds i16, ptr %x, i64 %indvars.iv\n %0 = load i16, ptr %arrayidx, align 2\n %1 = shl nuw nsw i64 %indvars.iv, 2\n %arrayidx2 = getelementptr inbounds i16, ptr %points, i64 %1\n store i16 %0, ptr %arrayidx2, align 2\n %arrayidx4 = getelementptr inbounds i16, ptr %y, i64 %indvars.iv\n %2 = load i16, ptr %arrayidx4, align 2\n %3 = or disjoint i64 %1, 1\n %arrayidx7 = getelementptr inbounds i16, ptr %points, i64 %3\n store i16 %2, ptr %arrayidx7, align 2\n %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1\n %exitcond.not = icmp eq i64 %indvars.iv.next, 1024\n br i1 %exitcond.not, label %for.end, label %for.body\n\nfor.end: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/AArch64/interleaved-vs-scalar.ll", "commands": ["opt < %s -force-vector-width=2 -force-vector-interleave=1 -passes=loop-vectorize -S --debug-only=loop-vectorize 2>&1"], "tests": [{"test_name": "test", "test_body": "target datalayout = \"e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64--linux-gnu\"\n\n%pair = type { i8, i8 }\n\ndefine void @test(ptr %p, ptr %q, i64 %n) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr %pair, ptr %p, i64 %i, i32 0\n %tmp1 = load i8, ptr %tmp0, align 1\n %tmp2 = getelementptr %pair, ptr %p, i64 %i, i32 1\n %tmp3 = load i8, ptr %tmp2, align 1\n %add = add i8 %tmp1, %tmp3\n %qi = getelementptr i8, ptr %q, i64 %i\n store i8 %add, ptr %qi, align 1\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp eq i64 %i.next, %n\n br i1 %cond, label %for.end, label %for.body\n\nfor.end: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/AArch64/interleaved_cost.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=2 -debug-only=loop-vectorize -disable-output < %s 2>&1", "opt -passes=loop-vectorize -force-vector-width=4 -debug-only=loop-vectorize -disable-output < %s 2>&1", "opt -passes=loop-vectorize -force-vector-width=8 -debug-only=loop-vectorize -disable-output < %s 2>&1", "opt -passes=loop-vectorize -force-vector-width=16 -debug-only=loop-vectorize -disable-output < %s 2>&1"], "tests": [{"test_name": "i8_factor_2", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64\"\n\n%i8.2 = type { i8, i8 }\n\ndefine void @i8_factor_2(ptr %data, i64 %n) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i8.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i8.2, ptr %data, i64 %i, i32 1\n %tmp2 = load i8, ptr %tmp0, align 1\n %tmp3 = load i8, ptr %tmp1, align 1\n store i8 %tmp2, ptr %tmp0, align 1\n store i8 %tmp3, ptr %tmp1, align 1\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n"}, {"test_name": "i16_factor_2", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64\"\n\n%i16.2 = type { i16, i16 }\n\ndefine void @i16_factor_2(ptr %data, i64 %n) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i16.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i16.2, ptr %data, i64 %i, i32 1\n %tmp2 = load i16, ptr %tmp0, align 2\n %tmp3 = load i16, ptr %tmp1, align 2\n store i16 %tmp2, ptr %tmp0, align 2\n store i16 %tmp3, ptr %tmp1, align 2\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n"}, {"test_name": "i32_factor_2", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64\"\n\n%i32.2 = type { i32, i32 }\n\ndefine void @i32_factor_2(ptr %data, i64 %n) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i32.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i32.2, ptr %data, i64 %i, i32 1\n %tmp2 = load i32, ptr %tmp0, align 4\n %tmp3 = load i32, ptr %tmp1, align 4\n store i32 %tmp2, ptr %tmp0, align 4\n store i32 %tmp3, ptr %tmp1, align 4\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n"}, {"test_name": "i64_factor_2", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64\"\n\n%i64.2 = type { i64, i64 }\n\ndefine void @i64_factor_2(ptr %data, i64 %n) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i64.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i64.2, ptr %data, i64 %i, i32 1\n %tmp2 = load i64, ptr %tmp0, align 8\n %tmp3 = load i64, ptr %tmp1, align 8\n store i64 %tmp2, ptr %tmp0, align 8\n store i64 %tmp3, ptr %tmp1, align 8\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n"}, {"test_name": "i64_factor_8", "test_body": "target datalayout = \"e-m:e-i64:64-i128:128-n32:64-S128\"\ntarget triple = \"aarch64\"\n\n%i64.8 = type { i64, i64, i64, i64, i64, i64, i64, i64 }\n\ndefine void @i64_factor_8(ptr %data, i64 %n) {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i64.8, ptr %data, i64 %i, i32 2\n %tmp1 = getelementptr inbounds %i64.8, ptr %data, i64 %i, i32 6\n %tmp2 = load i64, ptr %tmp0, align 8\n %tmp3 = load i64, ptr %tmp1, align 8\n store i64 %tmp2, ptr %tmp0, align 8\n store i64 %tmp3, ptr %tmp1, align 8\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/ARM/mve-interleaved-cost.ll", "commands": ["opt -passes=loop-vectorize -force-vector-width=2 -debug-only=loop-vectorize -disable-output < %s 2>&1", "opt -passes=loop-vectorize -force-vector-width=4 -debug-only=loop-vectorize -disable-output < %s 2>&1", "opt -passes=loop-vectorize -force-vector-width=8 -debug-only=loop-vectorize -disable-output < %s 2>&1", "opt -passes=loop-vectorize -force-vector-width=16 -debug-only=loop-vectorize -disable-output < %s 2>&1"], "tests": [{"test_name": "i16_factor_3", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i16.3 = type { i16, i16, i16 }\n\ndefine void @i16_factor_3(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i16.3, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i16.3, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %i16.3, ptr %data, i64 %i, i32 2\n %tmp3 = load i16, ptr %tmp0, align 2\n %tmp4 = load i16, ptr %tmp1, align 2\n %tmp5 = load i16, ptr %tmp2, align 2\n store i16 %tmp3, ptr %tmp0, align 2\n store i16 %tmp4, ptr %tmp1, align 2\n store i16 %tmp5, ptr %tmp2, align 2\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f16_factor_2", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f16.2 = type { half, half }\n\ndefine void @f16_factor_2(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f16.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f16.2, ptr %data, i64 %i, i32 1\n %tmp2 = load half, ptr %tmp0, align 2\n %tmp3 = load half, ptr %tmp1, align 2\n store half %tmp2, ptr %tmp0, align 2\n store half %tmp3, ptr %tmp1, align 2\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i32_factor_2", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i32.2 = type { i32, i32 }\n\ndefine void @i32_factor_2(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i32.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i32.2, ptr %data, i64 %i, i32 1\n %tmp2 = load i32, ptr %tmp0, align 4\n %tmp3 = load i32, ptr %tmp1, align 4\n store i32 %tmp2, ptr %tmp0, align 4\n store i32 %tmp3, ptr %tmp1, align 4\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i64_factor_4", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i64.4 = type { i64, i64, i64, i64 }\n\ndefine void @i64_factor_4(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i64.4, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i64.4, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %i64.4, ptr %data, i64 %i, i32 2\n %tmp3 = getelementptr inbounds %i64.4, ptr %data, i64 %i, i32 3\n %tmp4 = load i64, ptr %tmp0, align 8\n %tmp5 = load i64, ptr %tmp1, align 8\n %tmp6 = load i64, ptr %tmp2, align 8\n %tmp7 = load i64, ptr %tmp3, align 8\n store i64 %tmp4, ptr %tmp0, align 8\n store i64 %tmp5, ptr %tmp1, align 8\n store i64 %tmp6, ptr %tmp2, align 8\n store i64 %tmp7, ptr %tmp3, align 8\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f32_factor_3", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f32.3 = type { float, float, float }\n\ndefine void @f32_factor_3(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f32.3, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f32.3, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %f32.3, ptr %data, i64 %i, i32 2\n %tmp3 = load float, ptr %tmp0, align 4\n %tmp4 = load float, ptr %tmp1, align 4\n %tmp5 = load float, ptr %tmp2, align 4\n store float %tmp3, ptr %tmp0, align 4\n store float %tmp4, ptr %tmp1, align 4\n store float %tmp5, ptr %tmp2, align 4\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i64_factor_3", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i64.3 = type { i64, i64, i64 }\n\ndefine void @i64_factor_3(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i64.3, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i64.3, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %i64.3, ptr %data, i64 %i, i32 2\n %tmp3 = load i64, ptr %tmp0, align 8\n %tmp4 = load i64, ptr %tmp1, align 8\n %tmp5 = load i64, ptr %tmp2, align 8\n store i64 %tmp3, ptr %tmp0, align 8\n store i64 %tmp4, ptr %tmp1, align 8\n store i64 %tmp5, ptr %tmp2, align 8\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f16_factor_3", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f16.3 = type { half, half, half }\n\ndefine void @f16_factor_3(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f16.3, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f16.3, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %f16.3, ptr %data, i64 %i, i32 2\n %tmp3 = load half, ptr %tmp0, align 2\n %tmp4 = load half, ptr %tmp1, align 2\n %tmp5 = load half, ptr %tmp2, align 2\n store half %tmp3, ptr %tmp0, align 2\n store half %tmp4, ptr %tmp1, align 2\n store half %tmp5, ptr %tmp2, align 2\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i8_factor_3", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i8.3 = type { i8, i8, i8 }\n\ndefine void @i8_factor_3(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i8.3, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i8.3, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %i8.3, ptr %data, i64 %i, i32 2\n %tmp3 = load i8, ptr %tmp0, align 1\n %tmp4 = load i8, ptr %tmp1, align 1\n %tmp5 = load i8, ptr %tmp2, align 1\n store i8 %tmp3, ptr %tmp0, align 1\n store i8 %tmp4, ptr %tmp1, align 1\n store i8 %tmp5, ptr %tmp2, align 1\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f32_factor_2", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f32.2 = type { float, float }\n\ndefine void @f32_factor_2(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f32.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f32.2, ptr %data, i64 %i, i32 1\n %tmp2 = load float, ptr %tmp0, align 4\n %tmp3 = load float, ptr %tmp1, align 4\n store float %tmp2, ptr %tmp0, align 4\n store float %tmp3, ptr %tmp1, align 4\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i8_factor_4", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i8.4 = type { i8, i8, i8, i8 }\n\ndefine void @i8_factor_4(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i8.4, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i8.4, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %i8.4, ptr %data, i64 %i, i32 2\n %tmp3 = getelementptr inbounds %i8.4, ptr %data, i64 %i, i32 3\n %tmp4 = load i8, ptr %tmp0, align 1\n %tmp5 = load i8, ptr %tmp1, align 1\n %tmp6 = load i8, ptr %tmp2, align 1\n %tmp7 = load i8, ptr %tmp3, align 1\n store i8 %tmp4, ptr %tmp0, align 1\n store i8 %tmp5, ptr %tmp1, align 1\n store i8 %tmp6, ptr %tmp2, align 1\n store i8 %tmp7, ptr %tmp3, align 1\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i16_factor_2", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i16.2 = type { i16, i16 }\n\ndefine void @i16_factor_2(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i16.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i16.2, ptr %data, i64 %i, i32 1\n %tmp2 = load i16, ptr %tmp0, align 2\n %tmp3 = load i16, ptr %tmp1, align 2\n store i16 %tmp2, ptr %tmp0, align 2\n store i16 %tmp3, ptr %tmp1, align 2\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i16_factor_4", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i16.4 = type { i16, i16, i16, i16 }\n\ndefine void @i16_factor_4(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i16.4, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i16.4, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %i16.4, ptr %data, i64 %i, i32 2\n %tmp3 = getelementptr inbounds %i16.4, ptr %data, i64 %i, i32 3\n %tmp4 = load i16, ptr %tmp0, align 2\n %tmp5 = load i16, ptr %tmp1, align 2\n %tmp6 = load i16, ptr %tmp2, align 2\n %tmp7 = load i16, ptr %tmp3, align 2\n store i16 %tmp4, ptr %tmp0, align 2\n store i16 %tmp5, ptr %tmp1, align 2\n store i16 %tmp6, ptr %tmp2, align 2\n store i16 %tmp7, ptr %tmp3, align 2\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f64_factor_3", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f64.3 = type { double, double, double }\n\ndefine void @f64_factor_3(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f64.3, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f64.3, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %f64.3, ptr %data, i64 %i, i32 2\n %tmp3 = load double, ptr %tmp0, align 8\n %tmp4 = load double, ptr %tmp1, align 8\n %tmp5 = load double, ptr %tmp2, align 8\n store double %tmp3, ptr %tmp0, align 8\n store double %tmp4, ptr %tmp1, align 8\n store double %tmp5, ptr %tmp2, align 8\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i8_factor_2", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i8.2 = type { i8, i8 }\n\ndefine void @i8_factor_2(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i8.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i8.2, ptr %data, i64 %i, i32 1\n %tmp2 = load i8, ptr %tmp0, align 1\n %tmp3 = load i8, ptr %tmp1, align 1\n store i8 %tmp2, ptr %tmp0, align 1\n store i8 %tmp3, ptr %tmp1, align 1\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i32_factor_4", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i32.4 = type { i32, i32, i32, i32 }\n\ndefine void @i32_factor_4(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i32.4, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i32.4, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %i32.4, ptr %data, i64 %i, i32 2\n %tmp3 = getelementptr inbounds %i32.4, ptr %data, i64 %i, i32 3\n %tmp4 = load i32, ptr %tmp0, align 4\n %tmp5 = load i32, ptr %tmp1, align 4\n %tmp6 = load i32, ptr %tmp2, align 4\n %tmp7 = load i32, ptr %tmp3, align 4\n store i32 %tmp4, ptr %tmp0, align 4\n store i32 %tmp5, ptr %tmp1, align 4\n store i32 %tmp6, ptr %tmp2, align 4\n store i32 %tmp7, ptr %tmp3, align 4\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i32_factor_3", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i32.3 = type { i32, i32, i32 }\n\ndefine void @i32_factor_3(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i32.3, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i32.3, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %i32.3, ptr %data, i64 %i, i32 2\n %tmp3 = load i32, ptr %tmp0, align 4\n %tmp4 = load i32, ptr %tmp1, align 4\n %tmp5 = load i32, ptr %tmp2, align 4\n store i32 %tmp3, ptr %tmp0, align 4\n store i32 %tmp4, ptr %tmp1, align 4\n store i32 %tmp5, ptr %tmp2, align 4\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f64_factor_2", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f64.2 = type { double, double }\n\ndefine void @f64_factor_2(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f64.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f64.2, ptr %data, i64 %i, i32 1\n %tmp2 = load double, ptr %tmp0, align 8\n %tmp3 = load double, ptr %tmp1, align 8\n store double %tmp2, ptr %tmp0, align 8\n store double %tmp3, ptr %tmp1, align 8\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f16_factor_4", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f16.4 = type { half, half, half, half }\n\ndefine void @f16_factor_4(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f16.4, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f16.4, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %f16.4, ptr %data, i64 %i, i32 2\n %tmp3 = getelementptr inbounds %f16.4, ptr %data, i64 %i, i32 3\n %tmp4 = load half, ptr %tmp0, align 2\n %tmp5 = load half, ptr %tmp1, align 2\n %tmp6 = load half, ptr %tmp2, align 2\n %tmp7 = load half, ptr %tmp3, align 2\n store half %tmp4, ptr %tmp0, align 2\n store half %tmp5, ptr %tmp1, align 2\n store half %tmp6, ptr %tmp2, align 2\n store half %tmp7, ptr %tmp3, align 2\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "i64_factor_2", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%i64.2 = type { i64, i64 }\n\ndefine void @i64_factor_2(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %i64.2, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %i64.2, ptr %data, i64 %i, i32 1\n %tmp2 = load i64, ptr %tmp0, align 8\n %tmp3 = load i64, ptr %tmp1, align 8\n store i64 %tmp2, ptr %tmp0, align 8\n store i64 %tmp3, ptr %tmp1, align 8\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f64_factor_4", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f64.4 = type { double, double, double, double }\n\ndefine void @f64_factor_4(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f64.4, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f64.4, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %f64.4, ptr %data, i64 %i, i32 2\n %tmp3 = getelementptr inbounds %f64.4, ptr %data, i64 %i, i32 3\n %tmp4 = load double, ptr %tmp0, align 8\n %tmp5 = load double, ptr %tmp1, align 8\n %tmp6 = load double, ptr %tmp2, align 8\n %tmp7 = load double, ptr %tmp3, align 8\n store double %tmp4, ptr %tmp0, align 8\n store double %tmp5, ptr %tmp1, align 8\n store double %tmp6, ptr %tmp2, align 8\n store double %tmp7, ptr %tmp3, align 8\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}, {"test_name": "f32_factor_4", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\ntarget triple = \"thumbv8.1m.main-none-eabi\"\n\n%f32.4 = type { float, float, float, float }\n\ndefine void @f32_factor_4(ptr %data, i64 %n) #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]\n %tmp0 = getelementptr inbounds %f32.4, ptr %data, i64 %i, i32 0\n %tmp1 = getelementptr inbounds %f32.4, ptr %data, i64 %i, i32 1\n %tmp2 = getelementptr inbounds %f32.4, ptr %data, i64 %i, i32 2\n %tmp3 = getelementptr inbounds %f32.4, ptr %data, i64 %i, i32 3\n %tmp4 = load float, ptr %tmp0, align 4\n %tmp5 = load float, ptr %tmp1, align 4\n %tmp6 = load float, ptr %tmp2, align 4\n %tmp7 = load float, ptr %tmp3, align 4\n store float %tmp4, ptr %tmp0, align 4\n store float %tmp5, ptr %tmp1, align 4\n store float %tmp6, ptr %tmp2, align 4\n store float %tmp7, ptr %tmp3, align 4\n %i.next = add nuw nsw i64 %i, 1\n %cond = icmp slt i64 %i.next, %n\n br i1 %cond, label %for.body, label %for.end\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+mve.fp\" }\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll", "commands": ["opt -p loop-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S %s"], "tests": [{"test_name": "dead_load", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @dead_load(ptr %p, i16 %start) {\nentry:\n %start.ext = sext i16 %start to i64\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %start.ext, %entry ], [ %iv.next, %loop ]\n %gep = getelementptr i16, ptr %p, i64 %iv\n store i16 0, ptr %gep, align 2\n %l = load i16, ptr %gep, align 2\n %iv.next = add i64 %iv, 3\n %cmp = icmp slt i64 %iv, 111\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "cost_of_exit_branch_and_cond_insts", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i16 @llvm.umax.i16(i16, i16) #0\n\ndefine i32 @cost_of_exit_branch_and_cond_insts(ptr %a, ptr %b, i1 %c, i16 %x) #1 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 %c, label %then, label %loop.exiting\n\nthen: ; preds = %loop.header\n %gep = getelementptr inbounds i32, ptr %b, i32 %iv\n store i1 false, ptr %a, align 1\n store i32 0, ptr %gep, align 4\n br label %loop.exiting\n\nloop.exiting: ; preds = %then, %loop.header\n %iv.next = add i32 %iv, 1\n %umax = tail call i16 @llvm.umax.i16(i16 %x, i16 111)\n %umax.ext = zext i16 %umax to i32\n %sub = sub i32 770, %umax.ext\n %ec = icmp slt i32 %iv, %sub\n br i1 %ec, label %loop.latch, label %exit\n\nloop.latch: ; preds = %loop.exiting\n br label %loop.header\n\nexit: ; preds = %loop.exiting\n br label %return\n\nreturn: ; preds = %exit\n ret i32 0\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\nattributes #1 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "gather_interleave_group_with_dead_insert_pos", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @gather_interleave_group_with_dead_insert_pos(i64 %N, ptr noalias %src, ptr noalias %dst) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]\n %gep.src.0 = getelementptr i8, ptr %src, i64 %iv\n %l.dead = load i8, ptr %gep.src.0, align 1\n %iv.1 = add i64 %iv, 1\n %gep.src.1 = getelementptr i8, ptr %src, i64 %iv.1\n %l.1 = load i8, ptr %gep.src.1, align 1\n %ext = zext i8 %l.1 to i32\n %gep.dst = getelementptr i32, ptr %dst, i64 %iv\n store i32 %ext, ptr %gep.dst, align 4\n %iv.next = add nsw i64 %iv, 4\n %ec = icmp slt i64 %iv, %N\n br i1 %ec, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-features\"=\"+64bit,+v\" }\n"}, {"test_name": "test_phi_in_latch_redundant", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine void @test_phi_in_latch_redundant(ptr %dst, i32 %a) {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n br i1 false, label %loop.latch, label %then\n\nthen: ; preds = %loop.header\n %not.a = xor i32 %a, -1\n br label %loop.latch\n\nloop.latch: ; preds = %then, %loop.header\n %p = phi i32 [ %not.a, %then ], [ 0, %loop.header ]\n %gep = getelementptr i32, ptr %dst, i64 %iv\n store i32 %p, ptr %gep, align 4\n %iv.next = add i64 %iv, 9\n %ec = icmp slt i64 %iv, 322\n br i1 %ec, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n"}, {"test_name": "dead_live_out_due_to_scalar_epilogue_required", "test_body": "target datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\ndefine i8 @dead_live_out_due_to_scalar_epilogue_required(ptr %src, ptr %dst) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %idxprom = sext i32 %iv to i64\n %gep.src = getelementptr i8, ptr %src, i64 %idxprom\n %l = load i8, ptr %gep.src, align 1\n %gep.dst = getelementptr i8, ptr %dst, i64 %idxprom\n store i8 0, ptr %gep.dst, align 1\n %iv.next = add i32 %iv, 4\n %cmp = icmp ult i32 %iv, 1001\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop\n %r = phi i8 [ %l, %loop ]\n ret i8 %r\n}\n"}]}, {"file": "llvm/test/Transforms/LoopVectorize/SystemZ/mem-interleaving-costs-03.ll", "commands": ["opt -mtriple=s390x-unknown-linux -mcpu=z16 -passes=loop-vectorize -debug-only=loop-vectorize -force-vector-width=4 -disable-output < %s 2>&1"], "tests": [{"test_name": "fun", "test_body": "define noundef i32 @fun(i32 %argc, ptr nocapture readnone %argv) {\nentry:\n %l_4774.i = alloca [4 x [2 x i128]], align 8\n call void @llvm.lifetime.start.p0(i64 128, ptr nonnull %l_4774.i)\n br label %for.cond4.preheader.i\n\nfor.cond4.preheader.i: ; preds = %for.cond4.preheader.i, %entry\n %indvars.iv8.i = phi i64 [ 0, %entry ], [ %indvars.iv.next9.i, %for.cond4.preheader.i ]\n %arrayidx10.i = getelementptr inbounds [4 x [2 x i128]], ptr %l_4774.i, i64 0, i64 %indvars.iv8.i, i64 0\n store i128 8721036757475490113, ptr %arrayidx10.i, align 8\n %arrayidx10.i.c = getelementptr inbounds [4 x [2 x i128]], ptr %l_4774.i, i64 0, i64 %indvars.iv8.i, i64 1\n store i128 8721036757475490113, ptr %arrayidx10.i.c, align 8\n %indvars.iv.next9.i = add nuw nsw i64 %indvars.iv8.i, 1\n %exitcond.not.i = icmp eq i64 %indvars.iv.next9.i, 4\n br i1 %exitcond.not.i, label %func_1.exit, label %for.cond4.preheader.i\n\nfunc_1.exit: ; preds = %for.cond4.preheader.i\n %arrayidx195.i = getelementptr inbounds [4 x [2 x i128]], ptr %l_4774.i, i64 0, i64 1\n %0 = load i128, ptr %arrayidx195.i, align 8\n %cmp200.i = icmp ne i128 %0, 0\n %conv202.i = zext i1 %cmp200.i to i64\n %call203.i = tail call i64 @safe_sub_func_int64_t_s_s(i64 noundef %conv202.i, i64 noundef 9139899272418802852)\n call void @llvm.lifetime.end.p0(i64 128, ptr nonnull %l_4774.i)\n br label %for.cond\n\nfor.cond: ; preds = %for.cond, %func_1.exit\n br label %for.cond\n}\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #0\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #0\n\ndeclare dso_local i64 @safe_sub_func_int64_t_s_s(i64, i64)\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }\n"}]}], "issue": {"title": "[VPlan] Assertion \" VPlan cost model and legacy cost model disagreed\"' failed.", "body": "Tested using: 15106c26662a WITH commits from #107894 cherry-picked on top.\r\n```\r\n> git log --oneline\r\n2bfba9bc7d37 (HEAD -> main) !fixup add TODO\r\n28acd6af75ff Add initial loop-invariant code motion transform.\r\n15106c26662a (origin/main, origin/HEAD) [flang][runtime] Fix odd \"invalid descriptor\" runtime crash\r\n```\r\nTestcase:\r\n```c\r\nchar a;\r\nextern int b[];\r\nshort c;\r\nchar d[1];\r\nvoid checksum() {\r\n for (long e = 0; e < 3;)\r\n for (long f = 0; e < 3; e = 709692)\r\n for (; f < c; f += 4) {\r\n a = d[f];\r\n b[f] = d[f - 1];\r\n }\r\n}\r\n```\r\n`clang -march=rv64gcv -O3 red.c`\r\n\r\nTestcase:\r\n```llvm ir\r\n; ModuleID = 'reduced.bc'\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine void @g(i64 %conv) #0 {\r\nentry:\r\n br label %for.body8.us23\r\n\r\nfor.body8.us23: ; preds = %for.body8.us23, %entry\r\n %f.121.us24 = phi i64 [ %add.us30, %for.body8.us23 ], [ 0, %entry ]\r\n %arrayidx.us25 = getelementptr [1 x i8], ptr null, i64 0, i64 %f.121.us24\r\n %0 = load i8, ptr %arrayidx.us25, align 1\r\n %sub.us26 = add i64 %f.121.us24, 1\r\n %arrayidx9.us27 = getelementptr [1 x i8], ptr null, i64 0, i64 %sub.us26\r\n %1 = load i8, ptr %arrayidx9.us27, align 1\r\n %conv10.us28 = zext i8 %1 to i32\r\n %arrayidx11.us29 = getelementptr [0 x i32], ptr null, i64 0, i64 %f.121.us24\r\n store i32 %conv10.us28, ptr null, align 4\r\n %add.us30 = add nsw i64 %f.121.us24, 4\r\n %cmp6.us31 = icmp slt i64 %f.121.us24, %conv\r\n br i1 %cmp6.us31, label %for.body8.us23, label %for.cond.cleanup.split.us\r\n\r\nfor.cond.cleanup.split.us: ; preds = %for.body8.us23\r\n ret void\r\n}\r\n\r\nattributes #0 = { \"target-features\"=\"+64bit,+a,+c,+d,+f,+m,+relax,+v,+zicsr,+zifencei,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b,-b,-e,-experimental-smctr,-experimental-smmpm,-experimental-smnpm,-experimental-ssctr,-experimental-ssnpm,-experimental-sspm,-experimental-supm,-experimental-zacas,-experimental-zalasr,-experimental-zicfilp,-experimental-zicfiss,-experimental-zvbc32e,-experimental-zvkgs,-h,-shcounterenw,-shgatpa,-shtvala,-shvsatpa,-shvstvala,-shvstvecd,-smaia,-smcdeleg,-smcsrind,-smepmp,-smstateen,-ssaia,-ssccfg,-ssccptr,-sscofpmf,-sscounterenw,-sscsrind,-ssqosid,-ssstateen,-ssstrict,-sstc,-sstvala,-sstvecd,-ssu64xl,-svade,-svadu,-svbare,-svinval,-svnapot,-svpbmt,-xcvalu,-xcvbi,-xcvbitmanip,-xcvelw,-xcvmac,-xcvmem,-xcvsimd,-xsfcease,-xsfvcp,-xsfvfnrclipxfqf,-xsfvfwmaccqqq,-xsfvqmaccdod,-xsfvqmaccqoq,-xsifivecdiscarddlone,-xsifivecflushdlone,-xtheadba,-xtheadbb,-xtheadbs,-xtheadcmo,-xtheadcondmov,-xtheadfmemidx,-xtheadmac,-xtheadmemidx,-xtheadmempair,-xtheadsync,-xtheadvdot,-xventanacondops,-xwchc,-za128rs,-za64rs,-zaamo,-zabha,-zalrsc,-zama16b,-zawrs,-zba,-zbb,-zbc,-zbkb,-zbkc,-zbkx,-zbs,-zca,-zcb,-zcd,-zce,-zcf,-zcmop,-zcmp,-zcmt,-zdinx,-zfa,-zfbfmin,-zfh,-zfhmin,-zfinx,-zhinx,-zhinxmin,-zic64b,-zicbom,-zicbop,-zicboz,-ziccamoa,-ziccif,-zicclsm,-ziccrse,-zicntr,-zicond,-zihintntl,-zihintpause,-zihpm,-zimop,-zk,-zkn,-zknd,-zkne,-zknh,-zkr,-zks,-zksed,-zksh,-zkt,-ztso,-zvbb,-zvbc,-zvfbfmin,-zvfbfwma,-zvfh,-zvfhmin,-zvkb,-zvkg,-zvkn,-zvknc,-zvkned,-zvkng,-zvknha,-zvknhb,-zvks,-zvksc,-zvksed,-zvksg,-zvksh,-zvkt,-zvl1024b,-zvl16384b,-zvl2048b,-zvl256b,-zvl32768b,-zvl4096b,-zvl512b,-zvl65536b,-zvl8192b\" }\r\n```\r\n\r\nCommand/backtrace:\r\n```\r\n> /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt --passes loop-vectorize reduced.ll -S\r\nopt: /scratch/tc-testing/tc-compiler-fuzz-trunk/llvm/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7394: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0. Program arguments: /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt --passes loop-vectorize reduced.ll -S\r\n1. Running pass \"function(loop-vectorize)\" on module \"reduced.ll\"\r\n2. Running pass \"loop-vectorize\" on function \"g\"\r\n #0 0x00005fabff5a9d80 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x3072d80)\r\n #1 0x00005fabff5a718f llvm::sys::RunSignalHandlers() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x307018f)\r\n #2 0x00005fabff5a72e5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007fe3d2042520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007fe3d20969fc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76\r\n #5 0x00007fe3d20969fc __pthread_kill_internal ./nptl/pthread_kill.c:78:10\r\n #6 0x00007fe3d20969fc pthread_kill ./nptl/pthread_kill.c:89:10\r\n #7 0x00007fe3d2042476 gsignal ./signal/../sysdeps/posix/raise.c:27:6\r\n #8 0x00007fe3d20287f3 abort ./stdlib/abort.c:81:7\r\n #9 0x00007fe3d202871b _nl_load_domain ./intl/loadmsgcat.c:1177:9\r\n#10 0x00007fe3d2039e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n#11 0x00005fabfe550a26 llvm::LoopVectorizationPlanner::computeBestVF() (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2019a26)\r\n#12 0x00005fabfe5678fb llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x20308fb)\r\n#13 0x00005fabfe56a911 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2033911)\r\n#14 0x00005fabfe56af87 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2033f87)\r\n#15 0x00005fabfd3adf96 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe76f96)\r\n#16 0x00005fabff3b51ef llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e7e1ef)\r\n#17 0x00005fabfd3b11c6 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe7a1c6)\r\n#18 0x00005fabff3b357b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e7c57b)\r\n#19 0x00005fabfd3ad376 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0xe76376)\r\n#20 0x00005fabff3b40dd llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x2e7d0dd)\r\n#21 0x00005fabfcbd49db llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x69d9db)\r\n#22 0x00005fabfcbc6701 optMain (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x68f701)\r\n#23 0x00007fe3d2029d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16\r\n#24 0x00007fe3d2029e40 call_init ./csu/../csu/libc-start.c:128:20\r\n#25 0x00007fe3d2029e40 __libc_start_main ./csu/../csu/libc-start.c:379:5\r\n#26 0x00005fabfcbbcc65 _start (/scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt+0x685c65)\r\nzsh: IOT instruction (core dumped) /scratch/tc-testing/tc-compiler-fuzz-trunk/build-gcv/build-llvm-linux/bin/opt\r\n```\r\nhttps://godbolt.org/z/TTchhsf9h\r\n\r\nFound via fuzzer\r\n\r\nRelated/prior issue: https://github.com/llvm/llvm-project/issues/107473\r\n\r\ncc @fhahn", "author": "patrick-rivos", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "121897", "issue_url": "https://github.com/llvm/llvm-project/issues/121897", "bug_type": "crash", "base_commit": "3b19e787fc5da27dfcc9ac6552b06a763f12ea03", "knowledge_cutoff": "2025-01-07T08:06:43Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "0eaa69eb234798774a08743b64a9aa0cf71c5356", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanUtils.h": [[49, 55]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanUtils.h": ["isUniformAfterVectorization"]}}, "patch": "commit 0eaa69eb234798774a08743b64a9aa0cf71c5356\nAuthor: Florian Hahn \nDate: Tue Jan 7 21:26:40 2025 +0000\n\n [VPlan] Handle VPExpandSCEVRecipe in isUniformAfterVectorization.\n \n VPExpandSCEVRecipes must be placed in the entry and are alway uniform.\n This fixes a crash by always identifying them as uniform, even if the\n main vector loop region has been removed.\n \n Fixes https://github.com/llvm/llvm-project/issues/121897.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanUtils.h b/llvm/lib/Transforms/Vectorize/VPlanUtils.h\nindex 965777002052..777944264f45 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanUtils.h\n+++ b/llvm/lib/Transforms/Vectorize/VPlanUtils.h\n@@ -49,7 +49,8 @@ inline bool isUniformAfterVectorization(const VPValue *VPV) {\n return all_of(GEP->operands(), isUniformAfterVectorization);\n if (auto *VPI = dyn_cast(Def))\n return VPI->isSingleScalar() || VPI->isVectorToScalar();\n- return false;\n+ // VPExpandSCEVRecipes must be placed in the entry and are alway uniform.\n+ return isa(Def);\n }\n \n /// Return true if \\p V is a header mask in \\p Plan.\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination.ll", "commands": ["opt -passes='loop-vectorize,verify' -force-vector-width=8 -force-vector-interleave=1 -S %s", "opt -passes='loop-vectorize,verify' -force-vector-width=8 -force-vector-interleave=2 -S %s", "opt -passes='loop-vectorize,verify' -force-vector-width=16 -force-vector-interleave=1 -S %s"], "tests": [{"test_name": "scev_expand_step", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\ndefine void @scev_expand_step(i64 %x, ptr %dst) {\nentry:\n %c = icmp eq i64 %x, 65536\n call void @llvm.assume(i1 %c)\n %fr = freeze i64 %x\n %step = add i64 %fr, -65534\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %iv.next = add i64 %iv, %step\n %gep.dst = getelementptr i8, ptr %dst, i64 %iv.next\n store i8 0, ptr %gep.dst, align 1\n %ec = icmp slt i64 %iv.next, 16\n br i1 %ec, label %loop, label %exit\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "remove_loop_region_outer_loop", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\n\ndeclare i1 @cond()\n\ndefine void @remove_loop_region_outer_loop(i64 range(i64 8, 17) %N, ptr noalias %src, ptr %dst) {\nentry:\n br label %outer.header\n\nouter.header: ; preds = %outer.latch, %entry\n %outer.iv = phi ptr [ %src, %entry ], [ %outer.iv.next, %outer.latch ]\n br label %inner\n\ninner: ; preds = %inner, %outer.header\n %inner.iv = phi i64 [ 0, %outer.header ], [ %iv.next, %inner ]\n %gep.src = getelementptr i8, ptr %outer.iv, i64 %inner.iv\n %l = load i8, ptr %gep.src, align 1\n %gep.dst = getelementptr i8, ptr %dst, i64 %inner.iv\n store i8 %l, ptr %gep.dst, align 1\n %iv.next = add i64 %inner.iv, 1\n %c.1 = icmp eq i64 %iv.next, %N\n br i1 %c.1, label %outer.latch, label %inner\n\nouter.latch: ; preds = %inner\n %outer.iv.next = getelementptr i8, ptr %outer.iv, i64 1\n %c.2 = call i1 @cond()\n br i1 %c.2, label %outer.header, label %exit\n\nexit: ; preds = %outer.latch\n ret void\n}\n"}, {"test_name": "remove_loop_region_with_replicate_recipe", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\n\ndefine void @remove_loop_region_with_replicate_recipe(ptr %dst, i64 range(i64 5, 10) %N) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 2, %entry ], [ %iv.next, %loop ]\n %gep.dst = getelementptr i16, ptr %dst, i64 %iv\n store i16 0, ptr %gep.dst, align 2\n %iv.next = add i64 %iv, 1\n %ec = icmp eq i64 %iv.next, %N\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}, {"test_name": "test_tc_less_than_16", "test_body": "target datalayout = \"e-m:o-i64:64-i128:128-n32:64-S128\"\n\ndefine void @test_tc_less_than_16(ptr %A, i64 %N) {\nentry:\n %and = and i64 %N, 15\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ %and, %entry ], [ %iv.next, %loop ]\n %p.src = phi ptr [ %A, %entry ], [ %p.src.next, %loop ]\n %p.src.next = getelementptr inbounds i8, ptr %p.src, i64 1\n %l = load i8, ptr %p.src, align 1\n %add = add nsw i8 %l, 10\n store i8 %add, ptr %p.src, align 1\n %iv.next = add nsw i64 %iv, -1\n %cmp = icmp eq i64 %iv.next, 0\n br i1 %cmp, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n"}]}], "issue": {"title": "[clang] Crash at -O3: Assertion `hasVectorValue(Def)' failed.", "body": "This code crashes at `-O3`:\n\n```c\nint *a;\nint b, d;\nstatic int f = 65535;\nchar c[];\nint h(int i) {\n int g = 1;\n for (; b;)\n if (a[0]) {\n g = 0;\n break;\n }\n if (g)\n a;\n else if (i)\n return 1;\n return 0;\n}\nvoid k() {\n int *e[10] = {&f};\n for (;;) {\n d = 0;\n for (; d < 8; d = f - 65535 + d + 1) {\n int j = h(f);\n c[j + f - 65534 + d] = 0;\n }\n }\n}\nvoid main() {}\n```\n\nCompiler Explorer: https://godbolt.org/z/rcGqsE196 \n\nBisected to https://github.com/llvm/llvm-project/commit/f48884ded884d982a7fd13394b0e93e6588f4143, which was comitted by @fhahn \n\nCrash:\n\n```console\nclang: /root/llvm-project/llvm/lib/Transforms/Vectorize/VPlan.cpp:236: llvm::Value* llvm::VPTransformState::get(llvm::VPValue*, const llvm::VPLane&): Assertion `hasVectorValue(Def)' failed.\n```\n\nBacktrace:\n\n```console\nStack dump:\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/clang -gdwarf-4 -g -o /app/output.s -mllvm --x86-asm-syntax=intel -fno-verbose-asm -S --gcc-toolchain=/opt/compiler-explorer/gcc-snapshot -fcolor-diagnostics -fno-crash-diagnostics -O3 \n1.\t parser at end of file\n2.\tOptimizer\n3.\tRunning pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"\"\n4.\tRunning pass \"loop-vectorize\" on function \"k\"\n #0 0x0000000003c71738 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3c71738)\n #1 0x0000000003c6f444 llvm::sys::CleanupOnSignal(unsigned long) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3c6f444)\n #2 0x0000000003bbc988 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\n #3 0x000077fbea242520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #4 0x000077fbea2969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #5 0x000077fbea242476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #6 0x000077fbea2287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #7 0x000077fbea22871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #8 0x000077fbea239e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #9 0x00000000058fd2be llvm::VPTransformState::get(llvm::VPValue*, llvm::VPLane const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x58fd2be)\n#10 0x0000000005790d0d llvm::InnerLoopVectorizer::scalarizeInstruction(llvm::Instruction const*, llvm::VPReplicateRecipe*, llvm::VPLane const&, llvm::VPTransformState&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x5790d0d)\n#11 0x00000000057916d0 llvm::VPReplicateRecipe::execute(llvm::VPTransformState&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x57916d0)\n#12 0x00000000058e3ce7 llvm::VPBasicBlock::executeRecipes(llvm::VPTransformState*, llvm::BasicBlock*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x58e3ce7)\n#13 0x00000000058ef282 llvm::VPBasicBlock::execute(llvm::VPTransformState*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x58ef282)\n#14 0x00000000058f0ef1 llvm::VPRegionBlock::execute(llvm::VPTransformState*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x58f0ef1)\n#15 0x00000000058fddb1 llvm::VPlan::execute(llvm::VPTransformState*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x58fddb1)\n#16 0x00000000057be1eb llvm::LoopVectorizationPlanner::executePlan(llvm::ElementCount, unsigned int, llvm::VPlan&, llvm::InnerLoopVectorizer&, llvm::DominatorTree*, bool, llvm::DenseMap, llvm::detail::DenseMapPair> const*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x57be1eb)\n#17 0x00000000057d24c3 llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x57d24c3)\n#18 0x00000000057d3e19 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x57d3e19)\n#19 0x00000000057d4453 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x57d4453)\n#20 0x00000000052f765e llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x52f765e)\n#21 0x000000000361c6e0 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x361c6e0)\n#22 0x00000000011305ce llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x11305ce)\n#23 0x000000000361affb llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x361affb)\n#24 0x000000000112fc5e llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x112fc5e)\n#25 0x000000000361aa00 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x361aa00)\n#26 0x0000000003f247bf (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\n#27 0x0000000003f28335 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3f28335)\n#28 0x000000000460209e clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x460209e)\n#29 0x000000000660f9cc clang::ParseAST(clang::Sema&, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x660f9cc)\n#30 0x0000000004602478 clang::CodeGenAction::ExecuteAction() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4602478)\n#31 0x00000000048be799 clang::FrontendAction::Execute() (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x48be799)\n#32 0x0000000004840e2e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4840e2e)\n#33 0x00000000049abb8e clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x49abb8e)\n#34 0x0000000000ce93ef cc1_main(llvm::ArrayRef, char const*, void*) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xce93ef)\n#35 0x0000000000ce0e8a ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\n#36 0x0000000004647ae9 void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\n#37 0x0000000003bbce34 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x3bbce34)\n#38 0x00000000046480df clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\n#39 0x000000000460c1ed clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x460c1ed)\n#40 0x000000000460d26e clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x460d26e)\n#41 0x0000000004614d95 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0x4614d95)\n#42 0x0000000000ce6233 clang_main(int, char**, llvm::ToolContext const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xce6233)\n#43 0x0000000000bb2524 main (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xbb2524)\n#44 0x000077fbea229d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#45 0x000077fbea229e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#46 0x0000000000ce0935 _start (/opt/compiler-explorer/clang-assertions-trunk/bin/clang+0xce0935)\n```", "author": "cardigan1008", "labels": ["vectorizers", "crash-on-valid"], "comments": [{"author": "fhahn", "body": "Reduced repro:\n\n```\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine void @k() {\nentry:\n %.fr = freeze i32 65535\n %invariant.op = add i32 %.fr, -65534\n br label %for.cond.us\n\nfor.cond.us: ; preds = %for.body.us.us, %entry\n br label %for.body.us.us\n\nfor.body.us.us: ; preds = %for.body.us.us, %for.cond.us\n %storemerge7.us.us = phi i32 [ 0, %for.cond.us ], [ %sub.reass.us.us, %for.body.us.us ]\n %sub.reass.us.us = add i32 %storemerge7.us.us, %invariant.op\n %idxprom.us.us = sext i32 %sub.reass.us.us to i64\n %arrayidx.us.us = getelementptr [0 x i8], ptr null, i64 0, i64 %idxprom.us.us\n store i8 0, ptr %arrayidx.us.us, align 1\n %cmp.us.us = icmp slt i32 %sub.reass.us.us, 8\n br i1 %cmp.us.us, label %for.body.us.us, label %for.cond.us\n}\n```"}]}, "verified": true} {"bug_id": "102597", "issue_url": "https://github.com/llvm/llvm-project/issues/102597", "bug_type": "miscompilation", "base_commit": "513c3726ebc0a324f7e5a11d25617bb9557324d6", "knowledge_cutoff": "2024-08-09T11:02:59Z", "lit_test_dir": ["llvm/test/Transforms/IndVarSimplify"], "hints": {"fix_commit": "3512bcc2e9ab06b0ae2cab78744550b515e54184", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[11961, 11969]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::computeConstantDifference"]}}, "patch": "commit 3512bcc2e9ab06b0ae2cab78744550b515e54184\nAuthor: Nikita Popov \nDate: Mon Aug 12 15:18:32 2024 +0200\n\n [SCEV] Fix incorrect extension in computeConstantDifference()\n \n The Mul factor was zero-extended here, resulting in incorrect\n results for integers larger than 64-bit.\n \n As we currently only multiply by 1 or -1, just split this into\n two cases -- there's no need for a full multiplication here.\n \n Fixes https://github.com/llvm/llvm-project/issues/102597.\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex 9ecdcbe2c5de..318a9d773dc2 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -11961,9 +11961,14 @@ ScalarEvolution::computeConstantDifference(const SCEV *More, const SCEV *Less) {\n SmallDenseMap Multiplicity;\n APInt Diff(BW, 0);\n auto Add = [&](const SCEV *S, int Mul) {\n- if (auto *C = dyn_cast(S))\n- Diff += C->getAPInt() * Mul;\n- else\n+ if (auto *C = dyn_cast(S)) {\n+ if (Mul == 1) {\n+ Diff += C->getAPInt();\n+ } else {\n+ assert(Mul == -1);\n+ Diff -= C->getAPInt();\n+ }\n+ } else\n Multiplicity[S] += Mul;\n };\n auto Decompose = [&](const SCEV *S, int Mul) {\n", "tests": [{"file": "llvm/test/Transforms/IndVarSimplify/pr102597.ll", "commands": ["opt -S -passes=indvars < %s"], "tests": [{"test_name": "test", "test_body": "define void @test() {\nentry:\n br label %loop\n\nloop: ; preds = %loop.latch, %entry\n %iv = phi i128 [ 3, %entry ], [ %iv.dec, %loop.latch ]\n %tobool = icmp ne i128 %iv, 0\n br i1 %tobool, label %loop.latch, label %if\n\nif: ; preds = %loop\n call void @foo()\n br label %loop.latch\n\nloop.latch: ; preds = %if, %loop\n %iv.dec = add nsw i128 %iv, -1\n %cmp = icmp sgt i128 %iv, 0\n br i1 %cmp, label %loop, label %exit\n\nexit: ; preds = %loop.latch\n ret void\n}\n\ndeclare void @foo()\n", "additional_args": "-src-unroll=4 -tgt-unroll=4"}]}], "issue": {"title": "[SCEV] WRONG code ", "body": "```\r\nint printf(const char *, ...);\r\n__int128 a = 3, c;\r\nchar b;\r\nint main() {\r\n c = 3;\r\n for (; c >= 0; c--) {\r\n b = 0;\r\n for (; b <= 3; b++) {\r\n if (c)\r\n break;\r\n a = 0;\r\n }\r\n }\r\n printf(\"%d\\n\", (int)a);\r\n}\r\n\r\n```\r\n\r\nclang -march=z15 -O1 wrong0.i -o a.out\r\n\r\nThis program should print '0' as 'a' is set to zero inside the loop. It now prints '3' instead. A bisect leads to a recent commit: 79af689 \" [SCEV] Handle more adds in computeConstantDifference() (#101339)\".\r\n\r\nSeems like IndVarSimplify has gone wrong with an i128:\r\nGOOD <> BROKEN\r\n```\r\n; *** IR Dump After IndVarSimplifyPass on loo ; *** IR Dump After IndVarSimplifyPass on loo\r\n\r\n; Preheader: ; Preheader:\r\nentry: entry:\r\n br label %for.cond1.preheader br label %for.cond1.preheader\r\n\r\n; Loop: ; Loop:\r\nfor.cond1.preheader: for.cond1.preheader: \r\n %storemerge11 = phi i128 [ 3, %entry ], [ % %storemerge11 = phi i128 [ 3, %entry ], [ %\r\n %tobool = icmp ne i128 %storemerge11, 0 | br i1 true, label %for.inc5, label %if.end.\r\n br i1 %tobool, label %for.inc5, label %if.e <\r\n\r\nif.end.lr.ph: if.end.lr.ph: \r\n store i128 0, ptr @a, align 8, !tbaa !4 store i128 0, ptr @a, align 8, !tbaa !4\r\n br label %if.end br label %if.end\r\n\r\nif.end: if.end: \r\n %storemerge89 = phi i8 [ 0, %if.end.lr.ph ] %storemerge89 = phi i8 [ 0, %if.end.lr.ph ]\r\n %inc = add nuw nsw i8 %storemerge89, 1 %inc = add nuw nsw i8 %storemerge89, 1\r\n %cmp2 = icmp ugt i8 %storemerge89, 2 %cmp2 = icmp ugt i8 %storemerge89, 2\r\n %or.cond = or i1 %tobool, %cmp2 | %or.cond = or i1 true, %cmp2\r\n br i1 %or.cond, label %for.inc5.loopexit, l br i1 %or.cond, label %for.inc5.loopexit, l\r\n\r\nfor.inc5: for.inc5: \r\n %storemerge8.lcssa = phi i8 [ 0, %for.cond1 %storemerge8.lcssa = phi i8 [ 0, %for.cond1\r\n %dec = add nsw i128 %storemerge11, -1 %dec = add nsw i128 %storemerge11, -1\r\n %cmp = icmp ugt i128 %storemerge11, 0 %cmp = icmp ugt i128 %storemerge11, 0\r\n br i1 %cmp, label %for.cond1.preheader, lab br i1 %cmp, label %for.cond1.preheader, lab\r\n\r\nfor.inc5.loopexit: for.inc5.loopexit: \r\n %inc.lcssa = phi i8 [ %inc, %if.end ] %inc.lcssa = phi i8 [ %inc, %if.end ]\r\n br label %for.inc5 br label %for.inc5\r\n\r\n; Exit blocks ; Exit blocks\r\nfor.end6: for.end6: \r\n %storemerge8.lcssa.lcssa = phi i8 [ %storem %storemerge8.lcssa.lcssa = phi i8 [ %storem\r\n store i128 -1, ptr @c, align 8, !tbaa !4 store i128 -1, ptr @c, align 8, !tbaa !4\r\n store i8 %storemerge8.lcssa.lcssa, ptr @b, store i8 %storemerge8.lcssa.lcssa, ptr @b, \r\n %0 = load i128, ptr @a, align 8, !tbaa !4 %0 = load i128, ptr @a, align 8, !tbaa !4\r\n %conv7 = trunc i128 %0 to i32 %conv7 = trunc i128 %0 to i32\r\n %call = call signext i32 (ptr, ...) @printf %call = call signext i32 (ptr, ...) @printf\r\n ret i32 0 ret i32 0\r\n\r\n```", "author": "JonPsson1", "labels": ["miscompilation", "llvm:SCEV"], "comments": [{"author": "nikic", "body": "Classic mistake: https://github.com/llvm/llvm-project/blob/2f8f58dd17a11934e8c8ec212b6474f76fb18e61/llvm/lib/Analysis/ScalarEvolution.cpp#L11967 This `* Mul` doesn't do what I wanted it to do :)\r\n\r\nI should really pick up https://github.com/llvm/llvm-project/pull/80309 again to prevent this kind of mistake in the first place."}, {"author": "JonPsson1", "body": "Is this commit self-contained in the sense that I could revert it on my testing machines? Or will you revert / fix soon?"}]}, "verified": true} {"bug_id": "79175", "issue_url": "https://github.com/llvm/llvm-project/issues/79175", "bug_type": "miscompilation", "base_commit": "e538486e90539096e7851d0deba4ea9ed94fced2", "knowledge_cutoff": "2024-01-23T17:34:28Z", "lit_test_dir": ["llvm/test/Transforms/JumpThreading"], "hints": {"fix_commit": "4f32f5d5720fbef06672714a62376f236a36aef5", "components": ["JumpThreading", "AliasAnalysis"], "bug_location_lineno": {"llvm/include/llvm/Analysis/AliasAnalysis.h": [[287, 292], [668, 673]], "llvm/include/llvm/Analysis/BasicAliasAnalysis.h": [[43, 62]], "llvm/lib/Analysis/BasicAliasAnalysis.cpp": [[89, 95], [1063, 1068], [1556, 1561], [1735, 1741]], "llvm/lib/Transforms/Scalar/JumpThreading.cpp": [[1261, 1266]]}, "bug_location_funcname": {"llvm/include/llvm/Analysis/AliasAnalysis.h": ["AAQueryInfo", "enableCrossIterationMode"], "llvm/include/llvm/Analysis/BasicAliasAnalysis.h": ["BasicAAResult"], "llvm/lib/Analysis/BasicAliasAnalysis.cpp": ["BasicAAResult::aliasCheck", "BasicAAResult::aliasGEP", "BasicAAResult::invalidate", "BasicAAResult::isValueEqualInPotentialCycles", "BasicAAResult::subtractDecomposedGEPs"], "llvm/lib/Transforms/Scalar/JumpThreading.cpp": ["JumpThreadingPass::simplifyPartiallyRedundantLoad"]}}, "patch": "commit 4f32f5d5720fbef06672714a62376f236a36aef5\nAuthor: Nikita Popov \nDate: Wed Jan 31 15:23:53 2024 +0100\n\n [AA][JumpThreading] Don't use DomTree for AA in JumpThreading (#79294)\n \n JumpThreading may perform AA queries while the dominator tree is not up\n to date, which may result in miscompilations.\n \n Fix this by adding a new AAQI option to disable the use of the dominator\n tree in BasicAA.\n \n Fixes https://github.com/llvm/llvm-project/issues/79175.\n\ndiff --git a/llvm/include/llvm/Analysis/AliasAnalysis.h b/llvm/include/llvm/Analysis/AliasAnalysis.h\nindex d6f732d35fd4..e8e4f491be5a 100644\n--- a/llvm/include/llvm/Analysis/AliasAnalysis.h\n+++ b/llvm/include/llvm/Analysis/AliasAnalysis.h\n@@ -287,6 +287,10 @@ public:\n /// store %l, ...\n bool MayBeCrossIteration = false;\n \n+ /// Whether alias analysis is allowed to use the dominator tree, for use by\n+ /// passes that lazily update the DT while performing AA queries.\n+ bool UseDominatorTree = true;\n+\n AAQueryInfo(AAResults &AAR, CaptureInfo *CI) : AAR(AAR), CI(CI) {}\n };\n \n@@ -668,6 +672,9 @@ public:\n void enableCrossIterationMode() {\n AAQI.MayBeCrossIteration = true;\n }\n+\n+ /// Disable the use of the dominator tree during alias analysis queries.\n+ void disableDominatorTree() { AAQI.UseDominatorTree = false; }\n };\n \n /// Temporary typedef for legacy code that uses a generic \\c AliasAnalysis\ndiff --git a/llvm/include/llvm/Analysis/BasicAliasAnalysis.h b/llvm/include/llvm/Analysis/BasicAliasAnalysis.h\nindex afc1811239f2..7eca82729430 100644\n--- a/llvm/include/llvm/Analysis/BasicAliasAnalysis.h\n+++ b/llvm/include/llvm/Analysis/BasicAliasAnalysis.h\n@@ -43,20 +43,26 @@ class BasicAAResult : public AAResultBase {\n const Function &F;\n const TargetLibraryInfo &TLI;\n AssumptionCache &AC;\n- DominatorTree *DT;\n+ /// Use getDT() instead of accessing this member directly, in order to\n+ /// respect the AAQI.UseDominatorTree option.\n+ DominatorTree *DT_;\n+\n+ DominatorTree *getDT(const AAQueryInfo &AAQI) const {\n+ return AAQI.UseDominatorTree ? DT_ : nullptr;\n+ }\n \n public:\n BasicAAResult(const DataLayout &DL, const Function &F,\n const TargetLibraryInfo &TLI, AssumptionCache &AC,\n DominatorTree *DT = nullptr)\n- : DL(DL), F(F), TLI(TLI), AC(AC), DT(DT) {}\n+ : DL(DL), F(F), TLI(TLI), AC(AC), DT_(DT) {}\n \n BasicAAResult(const BasicAAResult &Arg)\n : AAResultBase(Arg), DL(Arg.DL), F(Arg.F), TLI(Arg.TLI), AC(Arg.AC),\n- DT(Arg.DT) {}\n+ DT_(Arg.DT_) {}\n BasicAAResult(BasicAAResult &&Arg)\n : AAResultBase(std::move(Arg)), DL(Arg.DL), F(Arg.F), TLI(Arg.TLI),\n- AC(Arg.AC), DT(Arg.DT) {}\n+ AC(Arg.AC), DT_(Arg.DT_) {}\n \n /// Handle invalidation events in the new pass manager.\n bool invalidate(Function &Fn, const PreservedAnalyses &PA,\ndiff --git a/llvm/lib/Analysis/BasicAliasAnalysis.cpp b/llvm/lib/Analysis/BasicAliasAnalysis.cpp\nindex 3178e2d27816..1028b52a7912 100644\n--- a/llvm/lib/Analysis/BasicAliasAnalysis.cpp\n+++ b/llvm/lib/Analysis/BasicAliasAnalysis.cpp\n@@ -89,7 +89,7 @@ bool BasicAAResult::invalidate(Function &Fn, const PreservedAnalyses &PA,\n // may be created without handles to some analyses and in that case don't\n // depend on them.\n if (Inv.invalidate(Fn, PA) ||\n- (DT && Inv.invalidate(Fn, PA)))\n+ (DT_ && Inv.invalidate(Fn, PA)))\n return true;\n \n // Otherwise this analysis result remains valid.\n@@ -1063,6 +1063,7 @@ AliasResult BasicAAResult::aliasGEP(\n : AliasResult::MayAlias;\n }\n \n+ DominatorTree *DT = getDT(AAQI);\n DecomposedGEP DecompGEP1 = DecomposeGEPExpression(GEP1, DL, &AC, DT);\n DecomposedGEP DecompGEP2 = DecomposeGEPExpression(V2, DL, &AC, DT);\n \n@@ -1556,6 +1557,7 @@ AliasResult BasicAAResult::aliasCheck(const Value *V1, LocationSize V1Size,\n const Value *HintO1 = getUnderlyingObject(Hint1);\n const Value *HintO2 = getUnderlyingObject(Hint2);\n \n+ DominatorTree *DT = getDT(AAQI);\n auto ValidAssumeForPtrContext = [&](const Value *Ptr) {\n if (const Instruction *PtrI = dyn_cast(Ptr)) {\n return isValidAssumeForContext(Assume, PtrI, DT,\n@@ -1735,7 +1737,7 @@ bool BasicAAResult::isValueEqualInPotentialCycles(const Value *V,\n if (!Inst || Inst->getParent()->isEntryBlock())\n return true;\n \n- return isNotInCycle(Inst, DT, /*LI*/ nullptr);\n+ return isNotInCycle(Inst, getDT(AAQI), /*LI*/ nullptr);\n }\n \n /// Computes the symbolic difference between two de-composed GEPs.\ndiff --git a/llvm/lib/Transforms/Scalar/JumpThreading.cpp b/llvm/lib/Transforms/Scalar/JumpThreading.cpp\nindex b7cf02489631..bb33a5da288c 100644\n--- a/llvm/lib/Transforms/Scalar/JumpThreading.cpp\n+++ b/llvm/lib/Transforms/Scalar/JumpThreading.cpp\n@@ -1261,6 +1261,8 @@ bool JumpThreadingPass::simplifyPartiallyRedundantLoad(LoadInst *LoadI) {\n BasicBlock::iterator BBIt(LoadI);\n bool IsLoadCSE;\n BatchAAResults BatchAA(*AA);\n+ // The dominator tree is updated lazily and may not be valid at this point.\n+ BatchAA.disableDominatorTree();\n if (Value *AvailableVal = FindAvailableLoadedValue(\n LoadI, LoadBB, BBIt, DefMaxInstsToScan, &BatchAA, &IsLoadCSE)) {\n // If the value of the load is locally available within the block, just use\n", "tests": [{"file": "llvm/test/Transforms/JumpThreading/pr79175.ll", "commands": ["opt -S -passes=jump-threading < %s"], "tests": [{"test_name": "test", "test_body": "@f = external global i32\n\ndefine i32 @test(i64 %idx, i32 %val) {\nentry:\n %cmp = icmp slt i64 %idx, 1\n br i1 %cmp, label %for.body, label %return\n\nfor.body: ; preds = %entry\n %f = load i32, ptr @f, align 4\n %cmp1 = icmp eq i32 %f, 0\n br i1 %cmp1, label %cond.end, label %cond.false\n\ncond.false: ; preds = %for.body\n br label %cond.end\n\ncond.end: ; preds = %cond.false, %for.body\n %phi = phi i32 [ %val, %cond.false ], [ 1, %for.body ]\n %cmp.i = icmp sgt i32 %phi, 0\n %sel = select i1 %cmp.i, i32 0, i32 %phi\n %f.idx = getelementptr inbounds i32, ptr @f, i64 %idx\n store i32 %sel, ptr %f.idx, align 4\n %f.reload = load i32, ptr @f, align 4\n %cmp3 = icmp slt i32 %f.reload, 1\n br i1 %cmp3, label %return2, label %return\n\nreturn: ; preds = %cond.end, %entry\n ret i32 0\n\nreturn2: ; preds = %cond.end\n ret i32 1\n}\n"}]}], "issue": {"title": "WRONG code, likely JumpThreadingPass", "body": "[wrong0.tar.gz](https://github.com/llvm/llvm-project/files/14027750/wrong0.tar.gz)\r\n(reduced c test case, should print 0)\r\n\r\nrunline:\r\nclang -O3 -march=arch13 wrong0.i -o a.out -w -mllvm -available-load-scan-limit=12\r\n\r\nThe function n() below is called two times in the reduced test case. The first time f[0] has a value of 0 at the start of the function, and a value of 1 at the end. The second time n() is called, f[0] has a value of 1 at the top, while it is then set to 0 by k().\r\n```\r\nlong k(long l, int m); // returns 0 \r\n\r\nconst int *n() {\r\n for (; C <= 0;) {\r\n A = (f[0] == 0 ? 1 : A % f[0]);\r\n f[C] = k(A, 0);\r\n g = &f[0];\r\n f[C] = 1 > *g;\r\n if (f[C])\r\n return &e;\r\n break;\r\n }\r\n return 0;\r\n}\r\n```\r\nThis is the transformation of jump threading:\r\n\r\n```\r\n > *** IR Dump After JumpThreadingPass on n ***\r\n > ; Function Attrs: nounwind\r\ndefine dso_local ptr @n() local_unnamed_addr #1 { define dso_local ptr @n() local_unnamed_addr #1 {\r\nentry: entry:\r\n %0 = load i64, ptr @C, align 8, !tbaa !4 %0 = load i64, ptr @C, align 8, !tbaa !4\r\n %cmp = icmp slt i64 %0, 1 %cmp = icmp slt i64 %0, 1\r\n br i1 %cmp, label %for.body, label %for.end br i1 %cmp, label %for.body, label %for.end\r\n\r\nfor.body: ; preds = %entry for.body: ; preds = %entry\r\n %1 = load i32, ptr @f, align 4, !tbaa !8 %1 = load i32, ptr @f, align 4, !tbaa !8\r\n %cmp1 = icmp eq i32 %1, 0 %cmp1 = icmp eq i32 %1, 0\r\n br i1 %cmp1, label %cond.end, label %cond.false | br i1 %cmp1, label %cond.end.thread, label %cond.end\r\n\r\ncond.false: ; preds = %for.body | cond.end.thread: ; preds = %for.body\r\n > store i64 1, ptr @A, align 8, !tbaa !4\r\n > br label %3\r\n >\r\n > cond.end: ; preds = %for.body\r\n %2 = load i64, ptr @A, align 8, !tbaa !4 %2 = load i64, ptr @A, align 8, !tbaa !4\r\n %conv = sext i32 %1 to i64 %conv = sext i32 %1 to i64\r\n %rem = srem i64 %2, %conv %rem = srem i64 %2, %conv\r\n br label %cond.end | store i64 %rem, ptr @A, align 8, !tbaa !4\r\n > %cmp.i = icmp sgt i64 %rem, 0\r\n > %cond.fr = freeze i1 %cmp.i\r\n > br i1 %cond.fr, label %3, label %4\r\n >\r\n > 3: ; preds = %cond.end.\r\n > %.pr = load i32, ptr @f, align 4, !tbaa !8\r\n > br label %4\r\n\r\ncond.end: ; preds = %for.body, | 4: ; preds = %cond.end,\r\n %cond = phi i64 [ %rem, %cond.false ], [ 1, %for.body ] | %5 = phi i32 [ %1, %cond.end ], [ %.pr, %3 ]\r\n store i64 %cond, ptr @A, align 8, !tbaa !4 | %6 = phi i64 [ 0, %3 ], [ %rem, %cond.end ]\r\n %cmp.i = icmp sgt i64 %cond, 0 | %conv2 = trunc i64 %6 to i32\r\n %cond.i = select i1 %cmp.i, i64 0, i64 %cond <\r\n %conv2 = trunc i64 %cond.i to i32 <\r\n %arrayidx = getelementptr inbounds [1 x i32], ptr @f, i64 0, i64 %0 %arrayidx = getelementptr inbounds [1 x i32], ptr @f, i64 0, i64 %0\r\n store i32 %conv2, ptr %arrayidx, align 4, !tbaa !8 store i32 %conv2, ptr %arrayidx, align 4, !tbaa !8\r\n store ptr @f, ptr @g, align 8, !tbaa !10 store ptr @f, ptr @g, align 8, !tbaa !10\r\n %3 = load i32, ptr @f, align 4, !tbaa !8 | %cmp3 = icmp slt i32 %5, 1\r\n %cmp3 = icmp slt i32 %3, 1 <\r\n %conv4 = zext i1 %cmp3 to i32 %conv4 = zext i1 %cmp3 to i32\r\n store i32 %conv4, ptr %arrayidx, align 4, !tbaa !8 store i32 %conv4, ptr %arrayidx, align 4, !tbaa !8\r\n br i1 %cmp3, label %return, label %for.end br i1 %cmp3, label %return, label %for.end\r\n\r\nfor.end: ; preds = %cond.end, | for.end: ; preds = %4, %entry\r\n br label %return br label %return\r\n\r\nreturn: ; preds = %cond.end, | return: ; preds = %4, %for.e\r\n %retval.0 = phi ptr [ null, %for.end ], [ @e, %cond.end ] | %retval.0 = phi ptr [ null, %for.end ], [ @e, %4 ]\r\n ret ptr %retval.0 ret ptr %retval.0\r\n} }\r\n~\r\n\r\n```\r\n\r\nBefore (left), `@f `is reloaded for the final icmp in cond.end (%3), which is necessary as the store to %conv2 just above goes to the same address. However, to the right JumpThreading has changed the final icmp to use the %5 value, which does not reflect the stored value of %conv2. This seems wrong and maybe JT has missed the fact that %arrayidx aliases `@f`?\r\n\r\n@jmorse @MatzeB @nikic \r\n", "author": "JonPsson1", "labels": ["miscompilation", "llvm:analysis", "llvm:transforms"], "comments": [{"author": "jmorse", "body": "Here's a godbolt link seemingly showing the same thing: https://godbolt.org/z/dMPaP7161 . I can't replicate this at 6a7abea47 with just `opt --passes=jump-threading`, but can if it's part of clang and inspected with `print-after-all`."}, {"author": "jmorse", "body": "Ah -- because it's sensitive to the `-available-load-scan-limit=12`, which defaults to six and needs to be at least ten for this problem to appear. Stepping through bits of opt, the AA trace believes something that's wrong:\r\n\r\n```\r\nStart %arrayidx = getelementptr inbounds [1 x i32], ptr @f, i64 0, i64 %0 @ LocationSize::precise(4), @f = dso_local global [1 x i32] zeroinitializer, align 4 @ LocationSize::precise(4)\r\n Start @f = dso_local global [1 x i32] zeroinitializer, align 4 @ LocationSize::beforeOrAfterPointer, @f = dso_local global [1 x i32] zeroinitializer, align 4 @ LocationSize::beforeOrAfterPointer\r\n End @f = dso_local global [1 x i32] zeroinitializer, align 4 @ LocationSize::beforeOrAfterPointer, @f = dso_local global [1 x i32] zeroinitializer, align 4 @ LocationSize::beforeOrAfterPointer = MustAlias\r\nEnd %arrayidx = getelementptr inbounds [1 x i32], ptr @f, i64 0, i64 %0 @ LocationSize::precise(4), @f = dso_local global [1 x i32] zeroinitializer, align 4 @ LocationSize::precise(4) = NoAlias\r\n```\r\n\r\nWhich I think is saying that the GEP is believed to not-alias the \"f\" variable, wheras it might alias depending on the value of %0. Stepping through other things during the alias query, I see `isKnownNonNullFromDominatingCondition` returning true for %0, i.e. the load from \"C\". I would imagine this means there's something wrong with the dominator tree that makes this code think it's only executing on a path where %0 is always-zero. (EDIT: actually it means always-non-zero).\r\n\r\n(However this isn't my normal wheelhouse, so that might not be true!)."}, {"author": "nikic", "body": "That sounds about right to me. JumpThreading performs lazy DT updates, so it's not legal to use DT during the transform.\r\n\r\nBasicAA in principle already supports working without DT, but it may be a bit tricky to avoid the DT use just in JumpThreading, given how this is all wired up in the pass manager."}, {"author": "nikic", "body": "Somewhat reduced test case:\r\n```llvm\r\n; RUN: opt -S -passes=jump-threading < %s\r\n@f = external global i32\r\n\r\ndefine void @test(i64 %idx, i32 %val) {\r\nentry:\r\n %cmp = icmp slt i64 %idx, 1\r\n br i1 %cmp, label %for.body, label %return\r\n\r\nfor.body:\r\n %f = load i32, ptr @f, align 4\r\n %cmp1 = icmp eq i32 %f, 0\r\n br i1 %cmp1, label %cond.end, label %cond.false\r\n\r\ncond.false:\r\n br label %cond.end\r\n\r\ncond.end:\r\n %phi = phi i32 [ %val, %cond.false ], [ 1, %for.body ]\r\n %cmp.i = icmp sgt i32 %phi, 0\r\n %sel = select i1 %cmp.i, i32 0, i32 %phi\r\n %f.idx = getelementptr inbounds i32, ptr @f, i64 %idx\r\n store i32 %sel, ptr %f.idx, align 4\r\n %f.reload = load i32, ptr @f, align 4\r\n %cmp3 = icmp slt i32 %f.reload, 1\r\n br i1 %cmp3, label %return, label %return\r\n\r\nreturn:\r\n ret void\r\n}\r\n```"}, {"author": "EugeneZelenko", "body": "Not merged yet."}, {"author": "tstellar", "body": "PR has been created, we will track the status there."}]}, "verified": true} {"bug_id": "80954", "issue_url": "https://github.com/llvm/llvm-project/issues/80954", "bug_type": "crash", "base_commit": "e60c4b61f8bab25a137a481e4d2d3dbfa656807b", "knowledge_cutoff": "2024-02-07T08:19:15Z", "lit_test_dir": ["llvm/test/Transforms/LoopIdiom"], "hints": {"fix_commit": "bec7181d5b9d1828129d78d440fd9e02d5cb63e8", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp": [[795, 801]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp": ["SCEVExpander::expandIVInc"]}}, "patch": "commit bec7181d5b9d1828129d78d440fd9e02d5cb63e8\nAuthor: Nikita Popov \nDate: Wed Feb 7 11:06:51 2024 +0100\n\n [SCEVExpander] Don't use recursive expansion for ptr IV inc\n \n Similar to the non-ptr case, directly create the getelementptr\n instruction. Going through expandAddToGEP() no longer makes sense\n with opaque pointers, where generating the necessary instruction\n is trivial.\n \n This avoids recursive expansion of (the SCEV of) StepV while the\n IR is in an inconsistent state, in particular with an incomplete\n IV phi node, which utilities may not be prepared to deal with.\n \n Fixes https://github.com/llvm/llvm-project/issues/80954.\n\ndiff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\nindex 58dbac0e85fb..3a28909473d9 100644\n--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\n+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\n@@ -795,7 +795,8 @@ Value *SCEVExpander::expandIVInc(PHINode *PN, Value *StepV, const Loop *L,\n Value *IncV;\n // If the PHI is a pointer, use a GEP, otherwise use an add or sub.\n if (PN->getType()->isPointerTy()) {\n- IncV = expandAddToGEP(SE.getSCEV(StepV), PN);\n+ // TODO: Change name to IVName.iv.next.\n+ IncV = Builder.CreatePtrAdd(PN, StepV, \"scevgep\");\n } else {\n IncV = useSubtract ?\n Builder.CreateSub(PN, StepV, Twine(IVName) + \".iv.next\") :\n", "tests": [{"file": "llvm/test/Transforms/LoopIdiom/pr80954.ll", "commands": ["opt -S -passes=loop-idiom < %s"], "tests": [{"test_name": "test", "test_body": "define void @test(ptr %p, i8 %arg, i64 %arg1, i32 %arg2) {\nentry:\n %sext = sext i8 %arg to i64\n %add = add i64 %arg1, -1\n br label %loop\n\nloop: ; preds = %latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %latch ]\n %phi = phi i32 [ %arg2, %entry ], [ %add9, %latch ]\n br label %loop2\n\nloop2: ; preds = %loop2, %loop\n %iv2 = phi i64 [ %iv2.next, %loop2 ], [ 0, %loop ]\n %gep = getelementptr [4 x i8], ptr %p, i64 %iv, i64 %iv2\n store i8 0, ptr %gep, align 1\n %iv2.next = add i64 %iv2, 1\n %icmp = icmp eq i64 %iv2.next, 0\n br i1 %icmp, label %latch, label %loop2\n\nlatch: ; preds = %loop2\n %add9 = add nsw i32 %phi, 1\n %sext10 = sext i32 %phi to i64\n %add11 = add i64 %add, %sext10\n %add12 = add i64 %add11, %sext\n %iv.next = add i64 %add12, %iv\n br label %loop\n}\n"}]}], "issue": {"title": "llvm crash: Assertion `Idx >= 0 && \"Invalid basic block argument!\"' failed.", "body": "llvm crashes in `loop-idiom` pass.\r\n\r\nBisected to 1c6e6432ca0b6832e06b93a4bcf22ead1899c14d, which was committed by @nikic \r\n\r\nCompiler explorer: https://godbolt.org/z/nxhnoE1Mb\r\n\r\n**% cat reduced.ll**\r\n```llvm\r\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"\r\ntarget triple = \"x86_64-unknown-linux-gnu\"\r\n\r\ndefine void @m(i8 %0, i64 %1, i32 %a.promoted) {\r\nentry:\r\n %conv13 = sext i8 %0 to i64\r\n %add14 = add i64 %1, -1\r\n br label %for.cond5thread-pre-split.lr.ph.split.us.us\r\n\r\nfor.cond5thread-pre-split.lr.ph.split.us.us: ; preds = %for.cond2.loopexit.us.us, %entry\r\n %inc.i1417.us = phi i32 [ %a.promoted, %entry ], [ %inc.i.us20, %for.cond2.loopexit.us.us ]\r\n %add161516.us = phi i64 [ 0, %entry ], [ %add16.us24, %for.cond2.loopexit.us.us ]\r\n br label %for.body7.us.us\r\n\r\nfor.body7.us.us: ; preds = %for.body7.us.us, %for.cond5thread-pre-split.lr.ph.split.us.us\r\n %2 = phi i64 [ %add9.us.us, %for.body7.us.us ], [ 0, %for.cond5thread-pre-split.lr.ph.split.us.us ]\r\n %arrayidx8.us.us = getelementptr [1 x [5 x i8]], ptr null, i64 0, i64 %add161516.us, i64 %2\r\n store i8 0, ptr %arrayidx8.us.us, align 1\r\n %add9.us.us = add i64 %2, 1\r\n %tobool6.not.us.us = icmp eq i64 %add9.us.us, 0\r\n br i1 %tobool6.not.us.us, label %for.cond2.loopexit.us.us, label %for.body7.us.us\r\n\r\nfor.cond2.loopexit.us.us: ; preds = %for.body7.us.us\r\n %inc.i.us20 = add nsw i32 %inc.i1417.us, 1\r\n %conv.i.us21 = sext i32 %inc.i1417.us to i64\r\n %add15.us22 = add i64 %add14, %conv.i.us21\r\n %sub.us23 = add i64 %add15.us22, %conv13\r\n %add16.us24 = add i64 %sub.us23, %add161516.us\r\n br label %for.cond5thread-pre-split.lr.ph.split.us.us\r\n}\r\n```\r\n\r\n**% opt -disable-output -passes=loop-idiom reduced.ll**\r\n```console\r\nopt: /tmp/tmpxmdtc3mt/tmpn8inmjh2/llvm/include/llvm/IR/Instructions.h:2910: llvm::Value* llvm::PHINode::getIncomingValueForBlock(const llvm::BasicBlock*) const: Assertion `Idx >= 0 && \"Invalid basic block argument!\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\r\nStack dump:\r\n0.\tProgram arguments: /zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/opt -disable-output -passes=loop-idiom reduced.ll\r\n #0 0x00007fc24b876eaf llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x914eaf)\r\n #1 0x00007fc24b874414 SignalHandler(int) Signals.cpp:0:0\r\n #2 0x00007fc24ab77090 (/lib/x86_64-linux-gnu/libc.so.6+0x43090)\r\n #3 0x00007fc24ab7700b raise /build/glibc-wuryBv/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #4 0x00007fc24ab56859 abort /build/glibc-wuryBv/glibc-2.31/stdlib/abort.c:81:7\r\n #5 0x00007fc24ab56729 get_sysdep_segment_value /build/glibc-wuryBv/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #6 0x00007fc24ab56729 _nl_load_domain /build/glibc-wuryBv/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #7 0x00007fc24ab67fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #8 0x00007fc24d68155f llvm::Loop::getCanonicalInductionVariable() const (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x271f55f)\r\n #9 0x00007fc24ca39df5 llvm::SCEVExpander::visitAddRecExpr(llvm::SCEVAddRecExpr const*) (.part.0) ScalarEvolutionExpander.cpp:0:0\r\n#10 0x00007fc24ca36745 llvm::SCEVVisitor::visit(llvm::SCEV const*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad4745)\r\n#11 0x00007fc24ca3711b llvm::SCEVExpander::expand(llvm::SCEV const*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad511b)\r\n#12 0x00007fc24ca37380 llvm::SCEVExpander::expandCodeForImpl(llvm::SCEV const*, llvm::Type*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad5380)\r\n#13 0x00007fc24ca399ec llvm::SCEVExpander::expandAddToGEP(llvm::SCEV const*, llvm::Type*, llvm::Value*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad79ec)\r\n#14 0x00007fc24ca3ab97 llvm::SCEVExpander::expandIVInc(llvm::PHINode*, llvm::Value*, llvm::Loop const*, llvm::Type*, llvm::Type*, bool) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad8b97)\r\n#15 0x00007fc24ca3bb6e llvm::SCEVExpander::getAddRecExprPHILiterally(llvm::SCEVAddRecExpr const*, llvm::Loop const*, llvm::Type*, llvm::Type*, llvm::Type*&, bool&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad9b6e)\r\n#16 0x00007fc24ca3c5d4 llvm::SCEVExpander::expandAddRecExprLiterally(llvm::SCEVAddRecExpr const*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ada5d4)\r\n#17 0x00007fc24ca36745 llvm::SCEVVisitor::visit(llvm::SCEV const*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad4745)\r\n#18 0x00007fc24ca3711b llvm::SCEVExpander::expand(llvm::SCEV const*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad511b)\r\n#19 0x00007fc24ca37380 llvm::SCEVExpander::expandCodeForImpl(llvm::SCEV const*, llvm::Type*) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ad5380)\r\n#20 0x00007fc24ce4bc8c (anonymous namespace)::LoopIdiomRecognize::processLoopStridedStore(llvm::Value*, llvm::SCEV const*, llvm::MaybeAlign, llvm::Value*, llvm::Instruction*, llvm::SmallPtrSetImpl&, llvm::SCEVAddRecExpr const*, llvm::SCEV const*, bool, bool) LoopIdiomRecognize.cpp:0:0\r\n#21 0x00007fc24ce539b7 (anonymous namespace)::LoopIdiomRecognize::processLoopStores(llvm::SmallVectorImpl&, llvm::SCEV const*, (anonymous namespace)::LoopIdiomRecognize::ForMemset) LoopIdiomRecognize.cpp:0:0\r\n#22 0x00007fc24ce5494c (anonymous namespace)::LoopIdiomRecognize::runOnCountableLoop() LoopIdiomRecognize.cpp:0:0\r\n#23 0x00007fc24ce561fd llvm::LoopIdiomRecognizePass::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1ef41fd)\r\n#24 0x00007fc24e6ee6c6 llvm::detail::PassModel, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x378c6c6)\r\n#25 0x00007fc24ce7544f std::optional llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::runSinglePass, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>, std::default_delete, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>>>>(llvm::Loop&, std::unique_ptr, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>, std::default_delete, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>>>&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&, llvm::PassInstrumentation&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1f1344f)\r\n#26 0x00007fc24ce756f6 llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::runWithoutLoopNestPasses(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1f136f6)\r\n#27 0x00007fc24ce76b0d llvm::PassManager, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&>::run(llvm::Loop&, llvm::AnalysisManager&, llvm::LoopStandardAnalysisResults&, llvm::LPMUpdater&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1f14b0d)\r\n#28 0x00007fc24ce77d62 llvm::FunctionToLoopPassAdaptor::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x1f15d62)\r\n#29 0x00007fc24e6eb0c6 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x37890c6)\r\n#30 0x00007fc24ba8a40d llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0xb2840d)\r\n#31 0x00007fc24e6e41a6 llvm::detail::PassModel>, llvm::PreservedAnalyses, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x37821a6)\r\n#32 0x00007fc24ba88c9d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0xb26c9d)\r\n#33 0x00007fc24e6eb026 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0x3789026)\r\n#34 0x00007fc24ba86c1d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/../lib/libLLVM-18git.so+0xb24c1d)\r\n#35 0x000055608ac0cc95 llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/opt+0x25c95)\r\n#36 0x000055608ac00876 main (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/opt+0x19876)\r\n#37 0x00007fc24ab58083 __libc_start_main /build/glibc-wuryBv/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#38 0x000055608ac0151e _start (/zdata/shaoli/compilers/ccbuilder-compilers/clang-1c6e6432ca0b6832e06b93a4bcf22ead1899c14d/bin/opt+0x1a51e)\r\n[1] 1629345 abort -disable-output -passes=loop-idiom reduced.ll\r\n```\r\n\r\nThe IR is reduced from the following C code:\r\n```C\r\nint a, b;\r\nlong d, h, j;\r\nshort e;\r\nchar f;\r\nstatic char *g = &f;\r\nstatic long *i = &d;\r\nchar k[1][5];\r\nlong l() {\r\n long c = a++;\r\n return c;\r\n}\r\nvoid m() {\r\n for (; b;)\r\n for (;; j = l() + *g + *i - 1 + j) {\r\n e = 0;\r\n for (; *i + e; e = 1)\r\n for (; h; h = h + 1)\r\n k[j][h] = 4;\r\n }\r\n}\r\nint main() {}\r\n```", "author": "shao-hua-li", "labels": ["llvm:crash", "llvm:SCEV"], "comments": []}, "verified": true} {"bug_id": "89958", "issue_url": "https://github.com/llvm/llvm-project/issues/89958", "bug_type": "miscompilation", "base_commit": "e913a33fcfbd667e4e3a35919b6bd9c5876a90a3", "knowledge_cutoff": "2024-04-24T17:36:34Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "b060661da8b3b53db55644e5e358bb2dca8b56d7", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h": [[126, 131]], "llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp": [[681, 687], [1376, 1386], [1395, 1400]]}, "bug_location_funcname": {"llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp": ["SCEVExpander::expandMinMaxExpr", "SCEVExpander::visitSMaxExpr", "SCEVExpander::visitUDivExpr"]}}, "patch": "commit b060661da8b3b53db55644e5e358bb2dca8b56d7\nAuthor: Florian Hahn \nDate: Thu Oct 17 13:55:20 2024 -0700\n\n [SCEVExpander] Expand UDiv avoiding UB when in seq_min/max. (#92177)\n \n Update SCEVExpander to introduce an SafeUDivMode, which is set\n when expanding operands of SCEVSequentialMinMaxExpr. In this mode,\n the expander will make sure that the divisor of the expanded UDiv is\n neither 0 nor poison.\n \n Fixes https://github.com/llvm/llvm-project/issues/89958.\n \n \n PR https://github.com/llvm/llvm-project/pull/92177\n\ndiff --git a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h\nindex 5697d983c9ad..7dd754a2bc0d 100644\n--- a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h\n+++ b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h\n@@ -126,6 +126,11 @@ class SCEVExpander : public SCEVVisitor {\n /// \"expanded\" form.\n bool LSRMode;\n \n+ /// When true, rewrite any divisors of UDiv expressions that may be 0 to\n+ /// umax(Divisor, 1) to avoid introducing UB. If the divisor may be poison,\n+ /// freeze it first.\n+ bool SafeUDivMode = false;\n+\n typedef IRBuilder BuilderType;\n BuilderType Builder;\n \ndiff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\nindex c412d0398b95..39da38e49181 100644\n--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\n+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp\n@@ -681,7 +681,21 @@ Value *SCEVExpander::visitUDivExpr(const SCEVUDivExpr *S) {\n SCEV::FlagAnyWrap, /*IsSafeToHoist*/ true);\n }\n \n- Value *RHS = expand(S->getRHS());\n+ const SCEV *RHSExpr = S->getRHS();\n+ Value *RHS = expand(RHSExpr);\n+ if (SafeUDivMode) {\n+ bool GuaranteedNotPoison =\n+ ScalarEvolution::isGuaranteedNotToBePoison(RHSExpr);\n+ if (!GuaranteedNotPoison)\n+ RHS = Builder.CreateFreeze(RHS);\n+\n+ // We need an umax if either RHSExpr is not known to be zero, or if it is\n+ // not guaranteed to be non-poison. In the later case, the frozen poison may\n+ // be 0.\n+ if (!SE.isKnownNonZero(RHSExpr) || !GuaranteedNotPoison)\n+ RHS = Builder.CreateIntrinsic(RHS->getType(), Intrinsic::umax,\n+ {RHS, ConstantInt::get(RHS->getType(), 1)});\n+ }\n return InsertBinop(Instruction::UDiv, LHS, RHS, SCEV::FlagAnyWrap,\n /*IsSafeToHoist*/ SE.isKnownNonZero(S->getRHS()));\n }\n@@ -1376,11 +1390,14 @@ Value *SCEVExpander::visitSignExtendExpr(const SCEVSignExtendExpr *S) {\n Value *SCEVExpander::expandMinMaxExpr(const SCEVNAryExpr *S,\n Intrinsic::ID IntrinID, Twine Name,\n bool IsSequential) {\n+ bool PrevSafeMode = SafeUDivMode;\n+ SafeUDivMode |= IsSequential;\n Value *LHS = expand(S->getOperand(S->getNumOperands() - 1));\n Type *Ty = LHS->getType();\n if (IsSequential)\n LHS = Builder.CreateFreeze(LHS);\n for (int i = S->getNumOperands() - 2; i >= 0; --i) {\n+ SafeUDivMode = (IsSequential && i != 0) || PrevSafeMode;\n Value *RHS = expand(S->getOperand(i));\n if (IsSequential && i != 0)\n RHS = Builder.CreateFreeze(RHS);\n@@ -1395,6 +1412,7 @@ Value *SCEVExpander::expandMinMaxExpr(const SCEVNAryExpr *S,\n }\n LHS = Sel;\n }\n+ SafeUDivMode = PrevSafeMode;\n return LHS;\n }\n \n", "tests": [{"file": "", "commands": ["opt -p loop-vectorize -force-vector-width=4 -S %s"], "tests": [{"test_name": "f", "test_body": "define i64 @f(i64 %0) {\n br label %2\n\n2: ; preds = %5, %1\n %3 = phi i64 [ 0, %1 ], [ %6, %5 ]\n %4 = icmp slt i64 %3, %0\n br i1 %4, label %5, label %9\n\n5: ; preds = %2\n %6 = add i64 %3, 1\n %7 = udiv i64 42, %0\n %8 = icmp slt i64 %3, %7\n br i1 %8, label %2, label %9\n\n9: ; preds = %5, %2\n %10 = phi i64 [ 1, %2 ], [ 0, %5 ]\n ret i64 %10\n}"}]}], "issue": {"title": "miscompilation due to LoopVectorize making a function vulnerable to integer divide-by-zero", "body": "https://alive2.llvm.org/ce/z/Kx2anL\r\n\r\nthis function:\r\n```llvm\r\ndefine i64 @f(i64 %0) {\r\n br label %2\r\n\r\n2: ; preds = %5, %1\r\n %3 = phi i64 [ 0, %1 ], [ %6, %5 ]\r\n %4 = icmp slt i64 %3, %0\r\n br i1 %4, label %5, label %9\r\n\r\n5: ; preds = %2\r\n %6 = add i64 %3, 1\r\n %7 = udiv i64 42, %0\r\n %8 = icmp slt i64 %3, %7\r\n br i1 %8, label %2, label %9\r\n\r\n9: ; preds = %5, %2\r\n %10 = phi i64 [ 1, %2 ], [ 0, %5 ]\r\n ret i64 %10\r\n}\r\n```\r\nis getting optimized to:\r\n```lllvm\r\ndefine noundef i64 @f(i64 %0) local_unnamed_addr #0 {\r\n %smax = tail call i64 @llvm.smax.i64(i64 %0, i64 0)\r\n %2 = udiv i64 42, %0\r\n %umin = tail call i64 @llvm.umin.i64(i64 %smax, i64 %2)\r\n %min.iters.check = icmp ult i64 %umin, 4\r\n br i1 %min.iters.check, label %scalar.ph.preheader, label %vector.ph\r\n\r\nvector.ph: ; preds = %1\r\n %3 = add nuw nsw i64 %umin, 1\r\n %n.mod.vf = and i64 %3, 3\r\n %4 = icmp eq i64 %n.mod.vf, 0\r\n %5 = select i1 %4, i64 4, i64 %n.mod.vf\r\n %n.vec = sub nsw i64 %3, %5\r\n br label %vector.body\r\n\r\nvector.body: ; preds = %vector.body, %vector.ph\r\n %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]\r\n %index.next = add nuw i64 %index, 4\r\n %6 = icmp eq i64 %index.next, %n.vec\r\n br i1 %6, label %scalar.ph.preheader, label %vector.body, !llvm.loop !0\r\n\r\nscalar.ph.preheader: ; preds = %vector.body, %1\r\n %.ph = phi i64 [ 0, %1 ], [ %n.vec, %vector.body ]\r\n br label %scalar.ph\r\n\r\nscalar.ph: ; preds = %scalar.ph.preheader, %9\r\n %7 = phi i64 [ %10, %9 ], [ %.ph, %scalar.ph.preheader ]\r\n %8 = icmp slt i64 %7, %0\r\n br i1 %8, label %9, label %13\r\n\r\n9: ; preds = %scalar.ph\r\n %10 = add nuw nsw i64 %7, 1\r\n %11 = udiv i64 42, %0\r\n %12 = icmp ult i64 %7, %11\r\n br i1 %12, label %scalar.ph, label %13, !llvm.loop !3\r\n\r\n13: ; preds = %9, %scalar.ph\r\n %14 = phi i64 [ 1, %scalar.ph ], [ 0, %9 ]\r\n ret i64 %14\r\n}\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i64 @llvm.smax.i64(i64, i64) #1\r\n\r\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\r\ndeclare i64 @llvm.umin.i64(i64, i64) #1\r\n\r\nattributes #0 = { nofree norecurse nosync nounwind memory(none) }\r\nattributes #1 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\r\n\r\n!0 = distinct !{!0, !1, !2}\r\n!1 = !{!\"llvm.loop.isvectorized\", i32 1}\r\n!2 = !{!\"llvm.loop.unroll.runtime.disable\"}\r\n!3 = distinct !{!3, !2, !1}\r\n```\r\n\r\nthe problem is that the optimized code can divide by zero even when the original code doesn't. to see this, pass 0 as an argument to `f`. I also independently verified on an x64-64 that the optimized code traps out with an FPE while the original code does not. it's LoopVectorize that's the culprit.\r\n\r\ncc @nunoplopes @hatsunespica\r\n", "author": "regehr", "labels": ["miscompilation", "llvm:SCEV"], "comments": [{"author": "dtcxzyw", "body": "cc @fhahn "}]}, "verified": true} {"bug_id": "99899", "issue_url": "https://github.com/llvm/llvm-project/issues/99899", "bug_type": "crash", "base_commit": "9a8b0407fc16af4ca6f79a2583297318a645d88a", "knowledge_cutoff": "2024-07-22T17:02:03Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "3cb82f49dc990dc20a765856c0e126193992fe44", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[9699, 9705]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::getEntryCost"]}}, "patch": "commit 3cb82f49dc990dc20a765856c0e126193992fe44\nAuthor: Alexey Bataev \nDate: Mon Jul 22 12:45:28 2024 -0700\n\n [SLP]Fix PR99899: Use canonical type instead of original vector of ptr.\n \n Use adjusted canonical integer type instead of the original ptr type to\n fix the crash in the TTI.\n Fixes https://github.com/llvm/llvm-project/issues/99899\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex 667c4eb311c2..cca9eeebaa53 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -9699,7 +9699,8 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef VectorizedVals,\n CanonicalType = CanonicalType->getWithNewType(IntegerType::get(\n CanonicalType->getContext(),\n DL->getTypeSizeInBits(CanonicalType->getScalarType())));\n- IntrinsicCostAttributes CostAttrs(MinMaxID, VecTy, {VecTy, VecTy});\n+ IntrinsicCostAttributes CostAttrs(MinMaxID, CanonicalType,\n+ {CanonicalType, CanonicalType});\n InstructionCost IntrinsicCost =\n TTI->getIntrinsicInstrCost(CostAttrs, CostKind);\n // If the selects are the only uses of the compares, they will be\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/SystemZ/cmp-ptr-minmax.ll", "commands": ["opt -S --passes=slp-vectorizer -mtriple=s390x-unknown-linux-gnu -mcpu=z16 -slp-threshold=-10 < %s"], "tests": [{"test_name": "", "test_body": "\ndefine i1 @test(i64 %0, i64 %1, ptr %2) {\n;\nentry:\n %gep44 = getelementptr i8, ptr null, i64 %0\n %gep45 = getelementptr i8, ptr null, i64 %1\n %4 = icmp ult ptr %gep44, %gep45\n %umin = select i1 %4, ptr %gep44, ptr %gep45\n %gep48 = getelementptr i8, ptr null, i64 %0\n %gep49 = getelementptr i8, ptr null, i64 %1\n %5 = icmp ult ptr %gep48, %gep49\n %umin50 = select i1 %5, ptr %gep48, ptr %gep49\n %b095 = icmp ult ptr %umin, %2\n %b196 = icmp ult ptr %umin50, %2\n %res = and i1 %b095, %b196\n ret i1 %res\n}"}]}], "issue": {"title": "[SLP] crash after 8ff233f", "body": "8ff233f \"[SLP]Correctly detect minnum/maxnum patterns for select/cmp operations on floats.\" seems to have introduced a problem when building SPEC on SystemZ.\r\n\r\nopt -mtriple=s390x-linux-gnu -mcpu=z16 -O3 ./tc_slp.ll -o /dev/null\r\nopt: /home/ijonpan/llvm-project/llvm/include/llvm/IR/DerivedTypes.h:704: llvm::Type* llvm::Type::getWithNewBitWidth(unsigned int) const: Assertion `isIntOrIntVectorTy() && \"Original type expected to be a vector of integers or a scalar integer.\"' failed.\r\n...\r\n#15 0x000000000519c8f8 llvm::slpvectorizer::BoUpSLP::getEntryCost\r\n\r\n[tc_slp.ll.tar.gz](https://github.com/user-attachments/files/16337301/tc_slp.ll.tar.gz)\r\n", "author": "JonPsson1", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "91078", "issue_url": "https://github.com/llvm/llvm-project/issues/91078", "bug_type": "crash", "base_commit": "d5ca2e46ca97758de6edb9f071f2114205380d87", "knowledge_cutoff": "2024-05-04T18:39:13Z", "lit_test_dir": ["llvm/test/Transforms/VectorCombine"], "hints": {"fix_commit": "d145f40963f15e94c4aefcc406a365fbe779e934", "components": ["VectorCombine"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": [[1755, 1764]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VectorCombine.cpp": ["VectorCombine::foldShuffleToIdentity"]}}, "patch": "commit d145f40963f15e94c4aefcc406a365fbe779e934\nAuthor: David Green \nDate: Sun May 5 10:47:11 2024 +0100\n\n [VectorCombine] shuffleToIdentity - guard against call instructions.\n \n The shuffleToIdentity fold needs to be a bit more careful about the difference\n between call instructions and intrinsics. The second can be handled, but the\n first should result in bailing out. This patch also adds some extra intrinsic\n tests from #91000.\n \n Fixes #91078\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\nindex 04d2ff594bf5..bbb70134870a 100644\n--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp\n@@ -1755,10 +1755,13 @@ bool VectorCombine::foldShuffleToIdentity(Instruction &I) {\n return false;\n if (IL.first->getValueID() != Item[0].first->getValueID())\n return false;\n+ if (isa(IL.first) && !isa(IL.first))\n+ return false;\n auto *II = dyn_cast(IL.first);\n return !II ||\n- II->getIntrinsicID() ==\n- cast(Item[0].first)->getIntrinsicID();\n+ (isa(Item[0].first) &&\n+ II->getIntrinsicID() ==\n+ cast(Item[0].first)->getIntrinsicID());\n }))\n return false;\n \n", "tests": [{"file": "llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll", "commands": ["opt -passes=vector-combine -S %s"], "tests": [{"test_name": "callinst", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @callinst(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = call <4 x half> @othercall(<4 x half> %at)\n %abb = call <4 x half> @llvm.fabs.v4f16(<4 x half> %ab)\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n\ndeclare <4 x half> @othercall(<4 x half>)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x half> @llvm.fabs.v4f16(<4 x half>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "callinst2", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @callinst2(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = call <4 x half> @llvm.fabs.v4f16(<4 x half> %at)\n %abb = call <4 x half> @othercall(<4 x half> %ab)\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n\ndeclare <4 x half> @othercall(<4 x half>)\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x half> @llvm.fabs.v4f16(<4 x half>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "lrint", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i32> @lrint(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = call <4 x i32> @llvm.lrint.v4i32.v4f16(<4 x half> %at)\n %abb = call <4 x i32> @llvm.lrint.v4i32.v4f16(<4 x half> %ab)\n %r = shufflevector <4 x i32> %abt, <4 x i32> %abb, <8 x i32> \n ret <8 x i32> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i32> @llvm.lrint.v4i32.v4f16(<4 x half>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "abs_different", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @abs_different(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %at, i1 true)\n %abb = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %ab, i1 false)\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.abs.v4i8(<4 x i8>, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "powi", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x half> @powi(<8 x half> %a) {\n %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> \n %abt = call <4 x half> @llvm.powi.v4f16.i32(<4 x half> %at, i32 10)\n %abb = call <4 x half> @llvm.powi.v4f16.i32(<4 x half> %ab, i32 10)\n %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> \n ret <8 x half> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x half> @llvm.powi.v4f16.i32(<4 x half>, i32) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "abs", "test_body": "target triple = \"aarch64\"\n\ndefine <8 x i8> @abs(<8 x i8> %a) {\n %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> \n %abt = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %at, i1 false)\n %abb = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %ab, i1 false)\n %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> \n ret <8 x i8> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <4 x i8> @llvm.abs.v4i8(<4 x i8>, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "A/F: isa(Val) && \"cast() argument of incompatible type!\" after commit a4d1026", "body": "A few of our internal tests started hitting an assertion failure which I bisected back to commit a4d10266d20bfe5930dfed77e17832af341ed66e. \r\n\r\nConsider the following code:\r\n```c++\r\ntemplate a b(a);\r\ntypedef float c __attribute__((__vector_size__(32)));\r\ntypedef float __attribute__((ext_vector_type(8))) d;\r\ntypedef float __attribute__((ext_vector_type(4))) e;\r\nc f;\r\n#define g(h) (b(h))\r\nd i;\r\nvoid j() {\r\n c k = __builtin_ia32_rsqrtps256(f);\r\n d l(k);\r\n d m g(i);\r\n volatile e n = __builtin_shufflevector(l, m, 15, 2, 0, 1);\r\n}\r\n```\r\nWhen compiled with optimizations and `-mavx` it hits an assertion failure:\r\n```\r\n$ ~/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang -c -O2 -mavx repro2.cpp \r\nclang: /home/dyung/src/upstream/llvm_clean_git/llvm/include/llvm/Support/Casting.h:578: decltype(auto) llvm::cast(From*) [with To = llvm::IntrinsicInst; From = llvm::Value]: Assertion `isa(Val) && \"cast() argument of incompatible type!\"' failed. \r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0. Program arguments: /home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang -c -O2 -mavx repro2.cpp\r\n1. parser at end of file\r\n2. Optimizer\r\n #0 0x0000557d533cfe2f llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x4055e2f)\r\n #1 0x0000557d533cd96c llvm::sys::CleanupOnSignal(unsigned long) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x405396c)\r\n #2 0x0000557d5331a4e8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0\r\n #3 0x00007f0ccbb8b420 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x14420)\r\n #4 0x00007f0ccb65800b raise /build/glibc-e2p3jK/glibc-2.31/signal/../sysdeps/unix/sysv/linux/raise.c:51:1\r\n #5 0x00007f0ccb637859 abort /build/glibc-e2p3jK/glibc-2.31/stdlib/abort.c:81:7\r\n #6 0x00007f0ccb637729 get_sysdep_segment_value /build/glibc-e2p3jK/glibc-2.31/intl/loadmsgcat.c:509:8\r\n #7 0x00007f0ccb637729 _nl_load_domain /build/glibc-e2p3jK/glibc-2.31/intl/loadmsgcat.c:970:34\r\n #8 0x00007f0ccb648fd6 (/lib/x86_64-linux-gnu/libc.so.6+0x33fd6)\r\n #9 0x0000557d54cd0f88 (anonymous namespace)::VectorCombine::foldShuffleToIdentity(llvm::Instruction&)::'lambda3'(std::pair)::operator()(std::pair) const (.isra.0) VectorCombine.cpp:0:0\r\n#10 0x0000557d54cdac11 (anonymous namespace)::VectorCombine::foldShuffleToIdentity(llvm::Instruction&) VectorCombine.cpp:0:0\r\n#11 0x0000557d54ce56f1 (anonymous namespace)::VectorCombine::run()::'lambda'(llvm::Instruction&)::operator()(llvm::Instruction&) const (.isra.0) VectorCombine.cpp:0:0\r\n#12 0x0000557d54ce695d llvm::VectorCombinePass::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x596c95d)\r\n#13 0x0000557d547ca386 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x5450386)\r\n#14 0x0000557d50a03b4c llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x1689b4c)\r\n#15 0x0000557d52d96d0d llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x3a1cd0d)\r\n#16 0x0000557d509e66d6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x166c6d6)\r\n#17 0x0000557d52d94b9d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x3a1ab9d)\r\n#18 0x0000557d5366aa28 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#19 0x0000557d5366dd85 (anonymous namespace)::EmitAssemblyHelper::EmitAssembly(clang::BackendAction, std::unique_ptr>, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#20 0x0000557d5366e449 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x42f4449)\r\n#21 0x0000557d53cc97ca clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x494f7ca)\r\n#22 0x0000557d55a3f65c clang::ParseAST(clang::Sema&, bool, bool) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x66c565c)\r\n#23 0x0000557d53cc8278 clang::CodeGenAction::ExecuteAction() (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x494e278)\r\n#24 0x0000557d53f56729 clang::FrontendAction::Execute() (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x4bdc729)\r\n#25 0x0000557d53edc48e clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x4b6248e)\r\n#26 0x0000557d54040e86 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x4cc6e86)\r\n#27 0x0000557d505e1bbc cc1_main(llvm::ArrayRef, char const*, void*) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x1267bbc)\r\n#28 0x0000557d505dacca ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#29 0x0000557d53d0d8ed void llvm::function_ref::callback_fn>, std::__cxx11::basic_string, std::allocator>*, bool*) const::'lambda'()>(long) Job.cpp:0:0\r\n#30 0x0000557d5331a9f0 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x3fa09f0)\r\n#31 0x0000557d53d0df0f clang::driver::CC1Command::Execute(llvm::ArrayRef>, std::__cxx11::basic_string, std::allocator>*, bool*) const (.part.0) Job.cpp:0:0\r\n#32 0x0000557d53cd2a4c clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x4958a4c)\r\n#33 0x0000557d53cd34dd clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl>&, bool) const (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x49594dd)\r\n#34 0x0000557d53cdb8f5 clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl>&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x49618f5)\r\n#35 0x0000557d505defab clang_main(int, char**, llvm::ToolContext const&) (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x1264fab)\r\n#36 0x0000557d504db96b main (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x116196b)\r\n#37 0x00007f0ccb639083 __libc_start_main /build/glibc-e2p3jK/glibc-2.31/csu/../csu/libc-start.c:342:3\r\n#38 0x0000557d505da79e _start (/home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin/clang+0x126079e)\r\nclang: error: clang frontend command failed with exit code 134 (use -v to see invocation)\r\nclang version 19.0.0 (https://github.com/llvm/llvm-project.git a4d10266d20bfe5930dfed77e17832af341ed66e)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /home/dyung/src/upstream/a4d10266d20bfe5930dfed77e17832af341ed66e-linux/bin\r\nBuild config: +assertions\r\n```", "author": "dyung", "labels": ["vectorizers", "crash"], "comments": []}, "verified": true} {"bug_id": "84051", "issue_url": "https://github.com/llvm/llvm-project/issues/84051", "bug_type": "crash", "base_commit": "b64482e23eefaef7738fde35d0b7c4174aaa6597", "knowledge_cutoff": "2024-03-05T17:59:45Z", "lit_test_dir": ["llvm/test/Transforms/DeadStoreElimination"], "hints": {"fix_commit": "eb8f379567e8d014194faefe02ce92813e237afc", "components": ["DeadStoreElimination"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp": [[1908, 1922]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp": ["tryFoldIntoCalloc"]}}, "patch": "commit eb8f379567e8d014194faefe02ce92813e237afc\nAuthor: Florian Hahn \nDate: Wed Mar 6 20:08:00 2024 +0000\n\n [DSE] Remove malloc from EarliestEscapeInfo before removing. (#84157)\n \n Not removing the malloc from earliest escape info leaves stale entries\n in the cache.\n \n Fixes https://github.com/llvm/llvm-project/issues/84051.\n \n PR: https://github.com/llvm/llvm-project/pull/84157\n\ndiff --git a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\nindex 3d1dac5ea17e..392e6ad5a66b 100644\n--- a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\n+++ b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp\n@@ -1908,15 +1908,15 @@ struct DSEState {\n Malloc->getArgOperand(0), IRB, TLI);\n if (!Calloc)\n return false;\n+\n MemorySSAUpdater Updater(&MSSA);\n auto *NewAccess =\n Updater.createMemoryAccessAfter(cast(Calloc), nullptr,\n MallocDef);\n auto *NewAccessMD = cast(NewAccess);\n Updater.insertDef(NewAccessMD, /*RenameUses=*/true);\n- Updater.removeMemoryAccess(Malloc);\n Malloc->replaceAllUsesWith(Calloc);\n- Malloc->eraseFromParent();\n+ deleteDeadInstruction(Malloc);\n return true;\n }\n \n", "tests": [{"file": "llvm/test/Transforms/DeadStoreElimination/malloc-earliest-escape-info-invalidation.ll", "commands": ["opt -p dse -S %s"], "tests": [{"test_name": "widget", "test_body": "target datalayout = \"E-m:e-i1:8:16-i8:8:16-i64:64-f128:64-v128:64-a:8:16-n32:64\"\n\ndefine void @widget(ptr %a) {\nbb:\n %call = tail call ptr @malloc(i64 1)\n tail call void @llvm.memset.p0.i64(ptr %call, i8 0, i64 1, i1 false)\n %call1 = tail call noalias ptr @malloc(i64 0)\n store ptr %call1, ptr %a, align 8\n %load = load ptr, ptr %a, align 8\n %load2 = load i32, ptr %load, align 8\n %getelementptr = getelementptr i8, ptr %call1, i64 0\n %getelementptr3 = getelementptr i8, ptr %getelementptr, i64 1\n store i8 0, ptr %call1, align 1\n %getelementptr4 = getelementptr i8, ptr %getelementptr, i64 8\n store i16 0, ptr %getelementptr4, align 4\n %getelementptr5 = getelementptr i8, ptr %getelementptr, i64 12\n store i32 0, ptr %call1, align 4\n %load6 = load i32, ptr inttoptr (i64 4 to ptr), align 4\n br label %bb48\n\nbb7: ; No predecessors!\n br label %bb9\n\nbb8: ; No predecessors!\n br label %bb53\n\nbb9: ; preds = %bb43, %bb7\n %phi = phi ptr [ %call1, %bb7 ], [ %a, %bb43 ]\n %getelementptr10 = getelementptr i8, ptr %phi, i64 0\n %getelementptr11 = getelementptr i8, ptr %phi, i64 0\n %getelementptr12 = getelementptr i8, ptr %phi, i64 0\n %getelementptr13 = getelementptr i8, ptr %getelementptr12, i64 1\n store i8 0, ptr %call1, align 1\n br label %bb29\n\nbb14: ; No predecessors!\n %getelementptr15 = getelementptr i8, ptr %getelementptr10, i64 8\n %load16 = load i16, ptr %call1, align 4\n br i1 false, label %bb22, label %bb17\n\nbb17: ; preds = %bb14\n %getelementptr18 = getelementptr i8, ptr %getelementptr11, i64 8\n %load19 = load i16, ptr %call1, align 4\n %getelementptr20 = getelementptr i8, ptr %getelementptr12, i64 8\n store i16 0, ptr %call1, align 4\n %getelementptr21 = getelementptr i8, ptr %phi, i64 0\n br label %bb25\n\nbb22: ; preds = %bb14\n %getelementptr23 = getelementptr i8, ptr %phi, i64 0\n %getelementptr24 = getelementptr i8, ptr %getelementptr23, i64 12\n br label %bb25\n\nbb25: ; preds = %bb22, %bb17\n %phi26 = phi ptr [ %a, %bb17 ], [ %call1, %bb22 ]\n %phi27 = phi ptr [ %call1, %bb17 ], [ %call1, %bb22 ]\n %phi28 = phi ptr [ %call1, %bb17 ], [ %call1, %bb22 ]\n store i32 0, ptr %call1, align 4\n br label %bb29\n\nbb29: ; preds = %bb25, %bb9\n %phi30 = phi ptr [ %call1, %bb9 ], [ %call1, %bb25 ]\n %phi31 = phi ptr [ %call1, %bb9 ], [ %call1, %bb25 ]\n %load32 = load i8, ptr %call1, align 4\n %load33 = load i8, ptr %call1, align 4\n %getelementptr34 = getelementptr i8, ptr %phi31, i64 12\n %getelementptr35 = getelementptr i8, ptr %getelementptr12, i64 12\n br label %bb86\n\nbb36: ; No predecessors!\n %getelementptr37 = getelementptr i8, ptr %phi30, i64 12\n br label %bb38\n\nbb38: ; preds = %bb38, %bb36\n %getelementptr39 = getelementptr [0 x i32], ptr %getelementptr34, i64 0, i64 0\n %load40 = load i32, ptr %call1, align 4\n %getelementptr41 = getelementptr [0 x i32], ptr %getelementptr37, i64 0, i64 0\n %load42 = load i32, ptr %call1, align 4\n br label %bb38\n\nbb43: ; No predecessors!\n %getelementptr44 = getelementptr i8, ptr %getelementptr12, i64 8\n %load45 = load i16, ptr %call1, align 4\n store i16 0, ptr %call1, align 4\n store i8 0, ptr %call1, align 4\n %getelementptr46 = getelementptr i8, ptr %getelementptr12, i64 12\n store i32 0, ptr %call1, align 4\n %getelementptr47 = getelementptr i8, ptr %getelementptr12, i64 16\n store i32 0, ptr %call1, align 4\n br label %bb9\n\nbb48: ; preds = %bb48, %bb\n %getelementptr49 = getelementptr i8, ptr %call1, i64 0\n %getelementptr50 = getelementptr i8, ptr %getelementptr49, i64 1\n store i8 0, ptr %call1, align 1\n %getelementptr51 = getelementptr i8, ptr %getelementptr49, i64 8\n store i16 0, ptr %call1, align 4\n %getelementptr52 = getelementptr i8, ptr %getelementptr49, i64 12\n store i32 0, ptr %call1, align 4\n br label %bb48\n\nbb53: ; preds = %bb71, %bb8\n %phi54 = phi ptr [ %call1, %bb8 ], [ %a, %bb71 ]\n %getelementptr55 = getelementptr i8, ptr %phi54, i64 0\n %getelementptr56 = getelementptr i8, ptr %phi54, i64 0\n %getelementptr57 = getelementptr i8, ptr %phi54, i64 0\n %getelementptr58 = getelementptr i8, ptr %getelementptr57, i64 1\n br label %bb71\n\nbb59: ; No predecessors!\n %getelementptr60 = getelementptr i8, ptr %phi54, i64 0\n %getelementptr61 = getelementptr i8, ptr %getelementptr60, i64 12\n br label %bb67\n\nbb62: ; No predecessors!\n %getelementptr63 = getelementptr i8, ptr %getelementptr56, i64 8\n %load64 = load i16, ptr %call1, align 4\n %getelementptr65 = getelementptr i8, ptr %getelementptr57, i64 8\n store i16 0, ptr %call1, align 4\n %getelementptr66 = getelementptr i8, ptr %phi54, i64 0\n br label %bb67\n\nbb67: ; preds = %bb62, %bb59\n %phi68 = phi ptr [ %a, %bb62 ], [ %call1, %bb59 ]\n %phi69 = phi ptr [ %call1, %bb62 ], [ %call1, %bb59 ]\n %phi70 = phi ptr [ %call1, %bb62 ], [ %call1, %bb59 ]\n store i32 0, ptr %call1, align 4\n br label %bb71\n\nbb71: ; preds = %bb67, %bb53\n %phi72 = phi ptr [ %call1, %bb53 ], [ %call1, %bb67 ]\n %phi73 = phi ptr [ %call1, %bb53 ], [ %call1, %bb67 ]\n %load74 = load i8, ptr %call1, align 4\n %load75 = load i8, ptr %call1, align 4\n %getelementptr76 = getelementptr i8, ptr %phi72, i64 12\n %getelementptr77 = getelementptr i8, ptr %getelementptr57, i64 12\n %getelementptr78 = getelementptr [0 x i32], ptr %getelementptr76, i64 0, i64 0\n %load79 = load i32, ptr %call1, align 4\n %getelementptr80 = getelementptr [0 x i32], ptr %getelementptr77, i64 0, i64 0\n store i32 0, ptr %call1, align 4\n %load81 = load i8, ptr %call1, align 4\n %getelementptr82 = getelementptr i8, ptr %getelementptr57, i64 8\n %load83 = load i16, ptr %call1, align 4\n store i16 0, ptr %call1, align 4\n store i8 0, ptr %call1, align 4\n %getelementptr84 = getelementptr i8, ptr %getelementptr57, i64 12\n store i32 0, ptr %call1, align 4\n %getelementptr85 = getelementptr i8, ptr %getelementptr57, i64 16\n store i32 0, ptr %call1, align 4\n br label %bb53\n\nbb86: ; preds = %bb86, %bb29\n %getelementptr87 = getelementptr [0 x i32], ptr %getelementptr34, i64 0, i64 0\n %load88 = load i32, ptr %call1, align 4\n %getelementptr89 = getelementptr [0 x i32], ptr %getelementptr35, i64 0, i64 0\n store i32 0, ptr %call1, align 4\n br label %bb86\n}\n\ndeclare ptr @malloc(i64)\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\n"}]}], "issue": {"title": "[DSE] crash", "body": "Reduced test case:\r\n[tc_dse.tar.gz](https://github.com/llvm/llvm-project/files/14499403/tc_dse.tar.gz)\r\n\r\nopt -mtriple=s390x-unknown-linux -mcpu=z16 -O3 -S ./tc_dse.ll\r\n\r\n#6 0x0000000005287456 llvm::BasicAAResult::aliasCheck\r\n#11 0x00000000048f0646 (anonymous namespace)::DSEState::isReadClobber\r\n14 0x00000000049004d8 llvm::DSEPass::run\r\nSegmentation fault (core dumped)\r\n\r\nThis is reported by user on F37 with clang-15, and is still failing on trunk.\r\n\r\n@fhahn @BK1603 @nikic ", "author": "JonPsson1", "labels": ["release:backport", "crash-on-valid", "llvm:transforms"], "comments": [{"author": "nikic", "body": "llvm-reduce result for `-passes=dse`: https://gist.github.com/nikic/d52c888425c5d4ff3fbd37b4f84778b6\r\n\r\nvalgrind reports a use-after-free. Probably not invalidating EII correctly somewhere."}, {"author": "fhahn", "body": "Also reproduces with ASAn, looking now"}]}, "verified": true} {"bug_id": "113526", "issue_url": "https://github.com/llvm/llvm-project/issues/113526", "bug_type": "crash", "base_commit": "8c4bc1e75de27adfbaead34b895b0efbaf17bd02", "knowledge_cutoff": "2024-10-24T06:14:03Z", "lit_test_dir": ["llvm/test/Transforms/LoopVectorize"], "hints": {"fix_commit": "e724226da753f10fd36fbb0ea392f04ab0fdbdab", "components": ["LoopVectorize"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": [[1524, 1529]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp": ["VPWidenCastRecipe::computeCost"]}}, "patch": "commit e724226da753f10fd36fbb0ea392f04ab0fdbdab\nAuthor: Florian Hahn \nDate: Fri Oct 25 12:35:33 2024 +0100\n\n [VPlan] Return cost of 0 for VPWidenCastRecipe without underlying value.\n \n In some cases, VPWidenCastRecipes are created but not considered in the\n legacy cost model, including truncates/extends when evaluating a reduction\n in a smaller type. Return 0 for such casts for now, to avoid divergences\n between VPlan and legacy cost models.\n \n Fixes https://github.com/llvm/llvm-project/issues/113526.\n\ndiff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\nindex 0eb4f7c7c88c..2080b77157b6 100644\n--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp\n@@ -1524,6 +1524,11 @@ void VPWidenCastRecipe::execute(VPTransformState &State) {\n \n InstructionCost VPWidenCastRecipe::computeCost(ElementCount VF,\n VPCostContext &Ctx) const {\n+ // TODO: In some cases, VPWidenCastRecipes are created but not considered in\n+ // the legacy cost model, including truncates/extends when evaluating a\n+ // reduction in a smaller type.\n+ if (!getUnderlyingValue())\n+ return 0;\n // Computes the CastContextHint from a recipes that may access memory.\n auto ComputeCCH = [&](const VPRecipeBase *R) -> TTI::CastContextHint {\n if (VF.isScalar())\n", "tests": [{"file": "llvm/test/Transforms/LoopVectorize/X86/cost-model.ll", "commands": ["opt < %s -passes=loop-vectorize -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -S"], "tests": [{"test_name": "narrowed_reduction", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i32 @narrowed_reduction(ptr %a, i1 %cmp) #0 {\nentry:\n %conv = zext i1 %cmp to i32\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i32 [ 1, %entry ], [ %inc, %loop ]\n %or13 = phi i32 [ 0, %entry ], [ %or, %loop ]\n %and = and i32 %or13, 1\n %or = or i32 %and, %conv\n %inc = add i32 %iv, 1\n %ec = icmp eq i32 %iv, 0\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i32 %or\n}\n\nattributes #0 = { \"target-cpu\"=\"penryn\" }\n"}, {"test_name": "cost_assume", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i64 @cost_assume(ptr %end, i64 %N) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]\n %0 = phi i64 [ 0, %entry ], [ %1, %loop ]\n %1 = add i64 %0, 1\n %iv.next = add nsw i64 %iv, 1\n %c = icmp ne i64 %N, 0\n tail call void @llvm.assume(i1 %c)\n %gep = getelementptr nusw [9 x i8], ptr null, i64 %iv.next\n %ec = icmp eq ptr %gep, %end\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i64 %1\n}\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #0\n\nattributes #0 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "PR27826", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine float @PR27826(ptr nocapture readonly %a, ptr nocapture readonly %b, i32 %n) {\nentry:\n %cmp = icmp sgt i32 %n, 0\n br i1 %cmp, label %preheader, label %for.end\n\npreheader: ; preds = %entry\n %t0 = sext i32 %n to i64\n br label %for\n\nfor: ; preds = %for, %preheader\n %indvars.iv = phi i64 [ 0, %preheader ], [ %indvars.iv.next, %for ]\n %s.02 = phi float [ 0.000000e+00, %preheader ], [ %add4, %for ]\n %arrayidx = getelementptr inbounds float, ptr %a, i64 %indvars.iv\n %t1 = load float, ptr %arrayidx, align 4\n %arrayidx3 = getelementptr inbounds float, ptr %b, i64 %indvars.iv\n %t2 = load float, ptr %arrayidx3, align 4\n %add = fadd fast float %t1, %s.02\n %add4 = fadd fast float %add, %t2\n %indvars.iv.next = add nuw nsw i64 %indvars.iv, 32\n %cmp1 = icmp slt i64 %indvars.iv.next, %t0\n br i1 %cmp1, label %for, label %loopexit\n\nloopexit: ; preds = %for\n %add4.lcssa = phi float [ %add4, %for ]\n br label %for.end\n\nfor.end: ; preds = %loopexit, %entry\n %s.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %add4.lcssa, %loopexit ]\n ret float %s.0.lcssa\n}\n"}, {"test_name": "reduction_store", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine void @reduction_store(ptr noalias %src, ptr %dst, i1 %x) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %red = phi i32 [ 0, %entry ], [ %red.next, %loop ]\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]\n %gep.src = getelementptr inbounds i32, ptr %src, i32 %iv\n %l = load i32, ptr %gep.src, align 4\n %l.and = and i32 %l, 3\n store i32 %l.and, ptr %dst, align 4\n %x.ext = zext i1 %x to i64\n %lshr = lshr i64 %x.ext, 12\n %t = trunc i64 %lshr to i32\n %red.next = and i32 %red, %t\n store i32 %red.next, ptr %dst, align 4\n %iv.next = add i32 %iv, 1\n %ec = icmp eq i32 %iv, 29\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"znver3\" }\n"}, {"test_name": "cost_loop_invariant_recipes", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i64 @cost_loop_invariant_recipes(i1 %x, i64 %y) {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next.i.i.i, %loop ]\n %red = phi i64 [ 1, %entry ], [ %red.mul, %loop ]\n %not.x = xor i1 %x, true\n %ext = zext i1 %not.x to i64\n %shl = shl i64 %y, %ext\n %red.mul = mul i64 %shl, %red\n %iv.next.i.i.i = add i64 %iv, 1\n %ec = icmp eq i64 %iv, 1\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i64 %red.mul\n}\n"}, {"test_name": "multi_exit", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine void @multi_exit(ptr %dst, ptr %src.1, ptr %src.2, i64 %A, i64 %B) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop.latch, %entry\n %iv.1.wide = phi i64 [ 0, %entry ], [ %iv.1.next.wide, %loop.latch ]\n %iv.1 = phi i32 [ 0, %entry ], [ %iv.1.next, %loop.latch ]\n %ec.1 = icmp ult i64 %iv.1.wide, %A\n br i1 %ec.1, label %loop.latch, label %exit\n\nloop.latch: ; preds = %loop\n %l.1 = load i64, ptr %src.1, align 8\n %l.2 = load i64, ptr %src.2, align 8\n %cmp55.us = icmp eq i64 %l.1, 0\n %cmp.i.us = icmp ne i64 %l.2, 0\n %and = and i1 %cmp.i.us, %cmp55.us\n %ext = zext i1 %and to i8\n store i8 %ext, ptr %dst, align 1\n %iv.1.next = add i32 %iv.1, 1\n %iv.1.next.wide = zext i32 %iv.1.next to i64\n %ec.2 = icmp ult i64 %iv.1.next.wide, %B\n br i1 %ec.2, label %loop, label %exit\n\nexit: ; preds = %loop.latch, %loop\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"penryn\" }\n"}, {"test_name": "cost_model_1", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\n@c = external global [2048 x i32], align 16\n@b = external global [2048 x i32], align 16\n@d = external global [2048 x i32], align 16\n@a = external global [2048 x i32], align 16\n\n; Function Attrs: noinline nounwind ssp uwtable\ndefine void @cost_model_1() #0 {\nentry:\n br label %for.body\n\nfor.body: ; preds = %for.body, %entry\n %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]\n %0 = shl nsw i64 %indvars.iv, 1\n %arrayidx = getelementptr inbounds [2048 x i32], ptr @c, i64 0, i64 %0\n %1 = load i32, ptr %arrayidx, align 8\n %idxprom1 = sext i32 %1 to i64\n %arrayidx2 = getelementptr inbounds [2048 x i32], ptr @b, i64 0, i64 %idxprom1\n %2 = load i32, ptr %arrayidx2, align 4\n %arrayidx4 = getelementptr inbounds [2048 x i32], ptr @d, i64 0, i64 %indvars.iv\n %3 = load i32, ptr %arrayidx4, align 4\n %idxprom5 = sext i32 %3 to i64\n %arrayidx6 = getelementptr inbounds [2048 x i32], ptr @a, i64 0, i64 %idxprom5\n store i32 %2, ptr %arrayidx6, align 4\n %indvars.iv.next = add i64 %indvars.iv, 1\n %lftr.wideiv = trunc i64 %indvars.iv.next to i32\n %exitcond = icmp eq i32 %lftr.wideiv, 256\n br i1 %exitcond, label %for.end, label %for.body\n\nfor.end: ; preds = %for.body\n ret void\n}\n\nattributes #0 = { noinline nounwind ssp uwtable }\n"}, {"test_name": "any_of_cost", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i1 @any_of_cost(ptr %start, ptr %end) #0 {\nentry:\n br label %loop\n\nloop: ; preds = %loop, %entry\n %any.of = phi i1 [ false, %entry ], [ %any.of.next, %loop ]\n %ptr.iv = phi ptr [ %start, %entry ], [ %ptr.iv.next, %loop ]\n %gep = getelementptr i8, ptr %ptr.iv, i64 8\n %l = load ptr, ptr %gep, align 8\n %cmp13.not.not = icmp eq ptr %l, null\n %any.of.next = select i1 %cmp13.not.not, i1 %any.of, i1 false\n %ptr.iv.next = getelementptr inbounds i8, ptr %ptr.iv, i64 40\n %cmp.not = icmp eq ptr %ptr.iv, %end\n br i1 %cmp.not, label %exit, label %loop\n\nexit: ; preds = %loop\n ret i1 %any.of.next\n}\n\nattributes #0 = { \"target-cpu\"=\"penryn\" }\n"}, {"test_name": "cost_duplicate_recipe_for_sinking", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine void @cost_duplicate_recipe_for_sinking(ptr %A, i64 %N) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %iv.shl = shl nsw i64 %iv, 2\n %gep.0 = getelementptr nusw double, ptr %A, i64 %iv.shl\n %l = load double, ptr %gep.0, align 8\n %c = fcmp oeq double %l, 0.000000e+00\n br i1 %c, label %if.then, label %loop.latch\n\nif.then: ; preds = %loop.header\n %gep.1 = getelementptr double, ptr %A, i64 %iv.shl\n store double 0.000000e+00, ptr %gep.1, align 8\n br label %loop.latch\n\nloop.latch: ; preds = %if.then, %loop.header\n %iv.next = add nsw i64 %iv, 1\n %ec = icmp eq i64 %iv, %N\n br i1 %ec, label %exit, label %loop.header\n\nexit: ; preds = %loop.latch\n ret void\n}\n\nattributes #0 = { \"target-cpu\"=\"znver3\" }\n"}, {"test_name": "avx512_cond_load_cost", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i64 @avx512_cond_load_cost(ptr %src, i32 %a, i64 %b, i32 %c, i32 %d) #0 {\nentry:\n br label %loop.header\n\nloop.header: ; preds = %loop.latch, %entry\n %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]\n %c.1 = icmp slt i32 %iv, 0\n br i1 %c.1, label %if.then, label %loop.latch\n\nif.then: ; preds = %loop.header\n %0 = urem i32 %a, %c\n %mul = sub i32 0, %0\n %div = udiv i32 %c, %d\n %or = or i32 %div, %mul\n %ext = sext i32 %or to i64\n %gep = getelementptr { i64, i64, i64 }, ptr %src, i64 %ext, i32 2\n %l = load i64, ptr %gep, align 8\n %or.2 = or i64 %l, %b\n br label %loop.latch\n\nloop.latch: ; preds = %if.then, %loop.header\n %res = phi i64 [ 0, %loop.header ], [ %or.2, %if.then ]\n %iv.next = add i32 %iv, 1\n %ec = icmp ult i32 %iv, %c\n br i1 %ec, label %loop.header, label %exit\n\nexit: ; preds = %loop.latch\n ret i64 %res\n}\n\nattributes #0 = { \"target-features\"=\"+avx512bw,+avx512cd,+avx512dq,+avx512f,+avx512vl\" }\n"}, {"test_name": "live_in_known_1_via_scev", "test_body": "target datalayout = \"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-apple-macosx10.8.0\"\n\ndefine i64 @live_in_known_1_via_scev() {\nentry:\n %sel = select i1 false, i32 3, i32 0\n br label %ph\n\nph: ; preds = %entry\n %p = phi i32 [ 1, %entry ]\n %N = add nuw nsw i32 %sel, 6\n %p.ext = zext nneg i32 %p to i64\n br label %loop\n\nloop: ; preds = %loop, %ph\n %iv = phi i32 [ 0, %ph ], [ %iv.next, %loop ]\n %red = phi i64 [ 3, %ph ], [ %red.mul, %loop ]\n %red.mul = mul nsw i64 %red, %p.ext\n %iv.next = add nuw nsw i32 %iv, 1\n %ec = icmp eq i32 %iv.next, %N\n br i1 %ec, label %exit, label %loop\n\nexit: ; preds = %loop\n %res = phi i64 [ %red.mul, %loop ]\n ret i64 %res\n}\n"}]}], "issue": {"title": "clang crashes at -O{s,2,3} on x86_64-linux-gnu: Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed ", "body": "It appears to be a recent regression as it doesn't reproduce with 19.1.0 and earlier.\r\n\r\nCompiler Explorer: https://godbolt.org/z/jPaW9fnKj\r\n\r\n```\r\n[531] % clangtk -v\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git a9050525954cbe11d45b415e2248d9e25e004bfe)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/home/suz/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/11\r\nFound candidate GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\r\nSelected GCC installation: /usr/lib/gcc/x86_64-linux-gnu/12\r\nCandidate multilib: .;@m64\r\nSelected multilib: .;@m64\r\nFound CUDA installation: /usr/local/cuda, version 11.8\r\n[532] %\r\n[532] % clangtk -O3 small.c\r\nclang-20: /local/suz-local/software/clangbuild/llvm-project/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7519: llvm::VectorizationFactor llvm::LoopVectorizationPlanner::computeBestVF(): Assertion `(BestFactor.Width == LegacyVF.Width || planContainsAdditionalSimplifications(getPlanFor(BestFactor.Width), CostCtx, OrigLoop)) && \" VPlan cost model and legacy cost model disagreed\"' failed.\r\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.\r\nStack dump:\r\n0. Program arguments: /local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20 -cc1 -triple x86_64-unknown-linux-gnu -emit-obj -dumpdir a- -disable-free -clear-ast-before-backend -main-file-name small.c -mrelocation-model pic -pic-level 2 -pic-is-pie -mframe-pointer=none -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fdebug-compilation-dir=/local/suz-local/software/emitesting/bugs/20241024-clangtk-m64-Os-build-062114/delta -fcoverage-compilation-dir=/local/suz-local/software/emitesting/bugs/20241024-clangtk-m64-Os-build-062114/delta -resource-dir /local/home/suz/suz-local/software/local/clang-trunk/lib/clang/20 -I /usr/local/include -I /local/suz-local/software/local/include -internal-isystem /local/home/suz/suz-local/software/local/clang-trunk/lib/clang/20/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/12/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O3 -ferror-limit 19 -fgnuc-version=4.2.1 -fskip-odr-check-in-gmf -fcolor-diagnostics -vectorize-loops -vectorize-slp -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/small-22743c.o -x c small.c\r\n1. parser at end of file\r\n2. Optimizer\r\n3. Running pass \"function(float2int,lower-constant-intrinsics,chr,loop(loop-rotate,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize,infer-alignment,loop-load-elim,instcombine,simplifycfg,slp-vectorizer,vector-combine,instcombine,loop-unroll,transform-warning,sroa,infer-alignment,instcombine,loop-mssa(licm),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg)\" on module \"small.c\"\r\n4. Running pass \"loop-vectorize\" on function \"main\"\r\n #0 0x000056385e381f80 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x43e3f80)\r\n #1 0x000056385e37f38f llvm::sys::RunSignalHandlers() (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x43e138f)\r\n #2 0x000056385e37f4e5 SignalHandler(int) Signals.cpp:0:0\r\n #3 0x00007fe699585520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\r\n #4 0x00007fe6995d9a7c pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x96a7c)\r\n #5 0x00007fe699585476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\r\n #6 0x00007fe69956b7f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\r\n #7 0x00007fe69956b71b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\r\n #8 0x00007fe69957ce96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\r\n #9 0x000056385fd39986 llvm::LoopVectorizationPlanner::computeBestVF() (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5d9b986)\r\n#10 0x000056385fd50d8c llvm::LoopVectorizePass::processLoop(llvm::Loop*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5db2d8c)\r\n#11 0x000056385fd53ea1 llvm::LoopVectorizePass::runImpl(llvm::Function&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5db5ea1)\r\n#12 0x000056385fd54517 llvm::LoopVectorizePass::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5db6517)\r\n#13 0x000056385f8a7d86 llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x5909d86)\r\n#14 0x000056385dd0d52f llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3d6f52f)\r\n#15 0x000056385b779146 llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x17db146)\r\n#16 0x000056385dd0b8bb llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3d6d8bb)\r\n#17 0x000056385b777be6 llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x17d9be6)\r\n#18 0x000056385dd0c41d llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x3d6e41d)\r\n#19 0x000056385e6253b6 (anonymous namespace)::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction, std::unique_ptr>&, std::unique_ptr>&, clang::BackendConsumer*) BackendUtil.cpp:0:0\r\n#20 0x000056385e628d53 clang::EmitBackendOutput(clang::DiagnosticsEngine&, clang::HeaderSearchOptions const&, clang::CodeGenOptions const&, clang::TargetOptions const&, clang::LangOptions const&, llvm::StringRef, llvm::Module*, clang::BackendAction, llvm::IntrusiveRefCntPtr, std::unique_ptr>, clang::BackendConsumer*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x468ad53)\r\n#21 0x000056385eceb3f5 clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x4d4d3f5)\r\n#22 0x0000563860a56ffc clang::ParseAST(clang::Sema&, bool, bool) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x6ab8ffc)\r\n#23 0x000056385efcbdd9 clang::FrontendAction::Execute() (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x502ddd9)\r\n#24 0x000056385ef45205 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x4fa7205)\r\n#25 0x000056385f0aa403 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x510c403)\r\n#26 0x000056385b330b57 cc1_main(llvm::ArrayRef, char const*, void*) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x1392b57)\r\n#27 0x000056385b3275ea ExecuteCC1Tool(llvm::SmallVectorImpl&, llvm::ToolContext const&) driver.cpp:0:0\r\n#28 0x000056385b32b96f clang_main(int, char**, llvm::ToolContext const&) (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x138d96f)\r\n#29 0x000056385b21ef2b main (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x1280f2b)\r\n#30 0x00007fe69956cd90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\r\n#31 0x00007fe69956ce40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\r\n#32 0x000056385b326d15 _start (/local/home/suz/suz-local/software/local/clang-trunk/bin/clang-20+0x1388d15)\r\nclangtk: error: unable to execute command: Aborted\r\nclangtk: error: clang frontend command failed due to signal (use -v to see invocation)\r\nclang version 20.0.0git (https://github.com/llvm/llvm-project.git a9050525954cbe11d45b415e2248d9e25e004bfe)\r\nTarget: x86_64-unknown-linux-gnu\r\nThread model: posix\r\nInstalledDir: /local/home/suz/suz-local/software/local/clang-trunk/bin\r\nBuild config: +assertions\r\nclangtk: note: diagnostic msg:\r\n********************\r\n\r\nPLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:\r\nPreprocessed source(s) and associated run script(s) are located at:\r\nclangtk: note: diagnostic msg: /tmp/small-ed84bf.c\r\nclangtk: note: diagnostic msg: /tmp/small-ed84bf.sh\r\nclangtk: note: diagnostic msg:\r\n\r\n********************\r\n[533] %\r\n[533] % cat small.c\r\nint a, b, c;\r\nint main() {\r\n for (; a; a++) {\r\n c &= 1;\r\n c |= b == 1;\r\n }\r\n return 0;\r\n}\r\n```", "author": "zhendongsu", "labels": ["vectorizers", "crash"], "comments": [{"author": "zhendongsu", "body": "@fhahn "}]}, "verified": true} {"bug_id": "85333", "issue_url": "https://github.com/llvm/llvm-project/issues/85333", "bug_type": "miscompilation", "base_commit": "fa67986d5b309ddd4c2ea74e8a5eeb0559aa8022", "knowledge_cutoff": "2024-03-14T23:23:02Z", "lit_test_dir": ["llvm/test/Transforms/GVNSink"], "hints": {"fix_commit": "1c979ab7e59feb03c58cac631a93143b0c776c20", "components": ["GVNSink"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/GVNSink.cpp": [[719, 730]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/GVNSink.cpp": ["GVNSink::analyzeInstructionForSinking"]}}, "patch": "commit 1c979ab7e59feb03c58cac631a93143b0c776c20\nAuthor: AdityaK \nDate: Tue Apr 30 11:31:45 2024 -0700\n\n [GVNSink] Fix incorrect codegen with respect to GEPs #85333 (#88440)\n \n As mentioned in #68882 and\n https://discourse.llvm.org/t/rfc-replacing-getelementptr-with-ptradd/68699\n \n Gep arithmetic isn't consistent with different types. GVNSink didn't\n realize this and sank all geps\n as long as their operands can be wired via PHIs\n in a post-dominator.\n \n Fixes: #85333\n\ndiff --git a/llvm/lib/Transforms/Scalar/GVNSink.cpp b/llvm/lib/Transforms/Scalar/GVNSink.cpp\nindex d4907326eb0a..1d2577e1da33 100644\n--- a/llvm/lib/Transforms/Scalar/GVNSink.cpp\n+++ b/llvm/lib/Transforms/Scalar/GVNSink.cpp\n@@ -719,12 +719,11 @@ GVNSink::analyzeInstructionForSinking(LockstepReverseIterator &LRI,\n // try and continue making progress.\n Instruction *I0 = NewInsts[0];\n \n- // If all instructions that are going to participate don't have the same\n- // number of operands, we can't do any useful PHI analysis for all operands.\n- auto hasDifferentNumOperands = [&I0](Instruction *I) {\n- return I->getNumOperands() != I0->getNumOperands();\n+ auto isNotSameOperation = [&I0](Instruction *I) {\n+ return !I0->isSameOperationAs(I);\n };\n- if (any_of(NewInsts, hasDifferentNumOperands))\n+\n+ if (any_of(NewInsts, isNotSameOperation))\n return std::nullopt;\n \n for (unsigned OpNum = 0, E = I0->getNumOperands(); OpNum != E; ++OpNum) {\n", "tests": [{"file": "llvm/test/Transforms/GVNSink/different-gep-types.ll", "commands": ["opt -passes=gvn-sink -S %s"], "tests": [{"test_name": "bar", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\n\n%\"struct.std::pair\" = type <{ i32, %struct.substruct, [2 x i8] }>\n%struct.substruct = type { i8, i8 }\n\ndefine void @bar(ptr noundef nonnull align 4 dereferenceable(4) %__i, i32 noundef %__n) {\nentry:\n %cmp = icmp eq i32 %__n, 1\n br i1 %cmp, label %if.then, label %if.else\n\nif.then: ; preds = %entry\n %0 = load ptr, ptr %__i, align 4\n %incdec.ptr4 = getelementptr inbounds i8, ptr %0, i32 -8\n br label %if.end6\n\nif.else: ; preds = %entry\n %1 = load ptr, ptr %__i, align 4\n %add.ptr = getelementptr inbounds %\"struct.std::pair\", ptr %1, i32 %__n\n br label %if.end6\n\nif.end6: ; preds = %if.else, %if.then\n %incdec.ptr.sink = phi ptr [ %incdec.ptr4, %if.then ], [ %add.ptr, %if.else ]\n store ptr %incdec.ptr.sink, ptr %__i, align 4\n ret void\n}\n"}, {"test_name": "foo", "test_body": "target datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\n\ndefine void @foo(ptr noundef nonnull align 4 dereferenceable(4) %__i, i32 noundef %__n) {\nentry:\n %cmp = icmp eq i32 %__n, 1\n br i1 %cmp, label %if.then, label %if.else\n\nif.then: ; preds = %entry\n %0 = load ptr, ptr %__i, align 4\n %incdec.ptr = getelementptr inbounds i8, ptr %0, i32 8\n store ptr %incdec.ptr, ptr %__i, align 4\n br label %if.end6\n\nif.else: ; preds = %entry\n %cmp2 = icmp eq i32 %__n, -1\n br i1 %cmp2, label %if.then3, label %if.else5\n\nif.then3: ; preds = %if.else\n %1 = load ptr, ptr %__i, align 4\n %incdec.ptr4 = getelementptr inbounds i8, ptr %1, i32 -8\n store ptr %incdec.ptr4, ptr %__i, align 4\n br label %if.end6\n\nif.else5: ; preds = %if.else\n %2 = load ptr, ptr %__i, align 4\n %add.ptr = getelementptr inbounds i8, ptr %2, i32 -4\n store ptr %add.ptr, ptr %__i, align 4\n br label %if.end6\n\nif.end6: ; preds = %if.else5, %if.then3, %if.then\n ret void\n}\n"}]}], "issue": {"title": "GVNSink produces incorrect codegen with respect to GEPs", "body": "```llvm\r\ntarget datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\r\ntarget triple = \"thumbv7em-none-unknown-eabi\"\r\n\r\n%\"struct.std::pair\" = type <{ i32, %struct.substruct, [2 x i8] }>\r\n%struct.substruct = type { i8, i8 }\r\n%\"struct.std::random_access_iterator_tag\" = type { i8 }\r\n\r\ndefine linkonce_odr dso_local void @__advance(ptr noundef nonnull align 4 dereferenceable(4) %__i, i32 noundef %__n) local_unnamed_addr {\r\nentry:\r\n %0 = call i1 @llvm.is.constant.i32(i32 %__n)\r\n %cmp = icmp eq i32 %__n, 1\r\n %or.cond = and i1 %0, %cmp\r\n br i1 %or.cond, label %if.then, label %if.else\r\n\r\nif.then: ; preds = %entry\r\n %1 = load ptr, ptr %__i, align 4\r\n %incdec.ptr = getelementptr inbounds i8, ptr %1, i32 8\r\n store ptr %incdec.ptr, ptr %__i, align 4\r\n br label %if.end6\r\n\r\nif.else: ; preds = %entry\r\n %2 = call i1 @llvm.is.constant.i32(i32 %__n)\r\n %cmp2 = icmp eq i32 %__n, -1\r\n %or.cond7 = and i1 %2, %cmp2\r\n br i1 %or.cond7, label %if.then3, label %if.else5\r\n\r\nif.then3: ; preds = %if.else\r\n %3 = load ptr, ptr %__i, align 4\r\n %incdec.ptr4 = getelementptr inbounds i8, ptr %3, i32 -8\r\n store ptr %incdec.ptr4, ptr %__i, align 4\r\n br label %if.end6\r\n\r\nif.else5: ; preds = %if.else\r\n %4 = load ptr, ptr %__i, align 4\r\n %add.ptr = getelementptr inbounds %\"struct.std::pair\", ptr %4, i32 %__n\r\n store ptr %add.ptr, ptr %__i, align 4\r\n br label %if.end6\r\n\r\nif.end6: ; preds = %if.then3, %if.else5, %if.then\r\n ret void\r\n}\r\n```\r\n\r\nThe above IR when piped through gvn-sink via `./bin/opt < /tmp/repro.ll -passes=gvn-sink -S` produces:\r\n\r\n```llvm\r\n; ModuleID = ''\r\nsource_filename = \"\"\r\ntarget datalayout = \"e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64\"\r\ntarget triple = \"thumbv7em-none-unknown-eabi\"\r\n\r\ndefine linkonce_odr dso_local void @__advance(ptr noundef nonnull align 4 dereferenceable(4) %__i, i32 noundef %__n) local_unnamed_addr {\r\nentry:\r\n %0 = call i1 @llvm.is.constant.i32(i32 %__n)\r\n %cmp = icmp eq i32 %__n, 1\r\n %or.cond = and i1 %0, %cmp\r\n br i1 %or.cond, label %if.then, label %if.else\r\n\r\nif.then: ; preds = %entry\r\n br label %if.end6\r\n\r\nif.else: ; preds = %entry\r\n %1 = call i1 @llvm.is.constant.i32(i32 %__n)\r\n %cmp2 = icmp eq i32 %__n, -1\r\n %or.cond7 = and i1 %1, %cmp2\r\n br i1 %or.cond7, label %if.then3, label %if.else5\r\n\r\nif.then3: ; preds = %if.else\r\n br label %if.end6\r\n\r\nif.else5: ; preds = %if.else\r\n br label %if.end6\r\n\r\nif.end6: ; preds = %if.else5, %if.then3, %if.then\r\n %.sink1 = phi i32 [ 8, %if.then ], [ -8, %if.then3 ], [ %__n, %if.else5 ]\r\n %2 = load ptr, ptr %__i, align 4\r\n %incdec.ptr = getelementptr inbounds i8, ptr %2, i32 %.sink1\r\n store ptr %incdec.ptr, ptr %__i, align 4\r\n ret void\r\n}\r\n\r\n; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none)\r\ndeclare i1 @llvm.is.constant.i32(i32) #0\r\n\r\nattributes #0 = { convergent nocallback nofree nosync nounwind willreturn memory(none) }\r\n```\r\n\r\nThe GEP in the result is incorrect. Prior, the GEPs were\r\n```llvm\r\n %incdec.ptr = getelementptr inbounds i8, ptr %1, i32 8\r\n %incdec.ptr4 = getelementptr inbounds i8, ptr %3, i32 -8\r\n %add.ptr = getelementptr inbounds %\"struct.std::pair\", ptr %4, i32 %__n\r\n```\r\nbut now it's\r\n```llvm\r\n%incdec.ptr = getelementptr inbounds i8, ptr %2, i32 %.sink1\r\n```\r\nThis results in an incorrect offset for `%add.ptr = getelementptr inbounds %\"struct.std::pair\", ptr %4, i32 %__n` since we'd now just add `%.sink1` bytes (where `%.sink1 = %__n` which is the second argument), whereas before we'd add `%__n * 8` bytes (where 8 is the sizeof `%\"struct.std::pair\"`).\r\n\r\nThe before ASM is:\r\n```asm\r\n__advance:\r\n ldr r2, [r0]\r\n add.w r1, r2, r1, lsl #3\r\n str r1, [r0]\r\n bx lr\r\n```\r\nbut the after asm is:\r\n```asm\r\n__advance:\r\n ldr r2, [r0]\r\n add r1, r2\r\n str r1, [r0]\r\n bx lr\r\n```\r\n\r\nThe GEP should not be a candidate for sinking.", "author": "PiJoules", "labels": ["bug", "miscompilation", "llvm:GVN"], "comments": []}, "verified": true} {"bug_id": "96012", "issue_url": "https://github.com/llvm/llvm-project/issues/96012", "bug_type": "hang", "base_commit": "aad27bf534b59645f47a92f072af798687b1dd0d", "knowledge_cutoff": "2024-06-19T02:31:53Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "48ae61470104e9d7a8be5beb8739c24f52cc33c0", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineInternal.h": [[780, 790]], "llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp": [[90, 102], [309, 314], [537, 543]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp": ["Negator::Negate", "Negator::Negator", "Negator::visitImpl", "Negator::~Negator"]}}, "patch": "commit 48ae61470104e9d7a8be5beb8739c24f52cc33c0\nAuthor: Yingwei Zheng \nDate: Sat Aug 17 16:48:29 2024 +0800\n\n [InstCombine] Avoid infinite loop when negating phi nodes (#104581)\n \n Closes https://github.com/llvm/llvm-project/issues/96012\n \n ---------\n \n Co-authored-by: Nikita Popov \n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\nindex 16f1c3ba15eb..a0f2399972b5 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h\n@@ -780,11 +780,14 @@ class Negator final {\n using BuilderTy = IRBuilder;\n BuilderTy Builder;\n \n+ const DominatorTree &DT;\n+\n const bool IsTrulyNegation;\n \n SmallDenseMap NegationsCache;\n \n- Negator(LLVMContext &C, const DataLayout &DL, bool IsTrulyNegation);\n+ Negator(LLVMContext &C, const DataLayout &DL, const DominatorTree &DT,\n+ bool IsTrulyNegation);\n \n #if LLVM_ENABLE_STATS\n unsigned NumValuesVisitedInThisNegator = 0;\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp b/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\nindex 2dd1db6a4a75..92293ef40146 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp\n@@ -90,13 +90,14 @@ static cl::opt\n cl::desc(\"What is the maximal lookup depth when trying to \"\n \"check for viability of negation sinking.\"));\n \n-Negator::Negator(LLVMContext &C, const DataLayout &DL, bool IsTrulyNegation_)\n+Negator::Negator(LLVMContext &C, const DataLayout &DL, const DominatorTree &DT_,\n+ bool IsTrulyNegation_)\n : Builder(C, TargetFolder(DL),\n IRBuilderCallbackInserter([&](Instruction *I) {\n ++NegatorNumInstructionsCreatedTotal;\n NewInstructions.push_back(I);\n })),\n- IsTrulyNegation(IsTrulyNegation_) {}\n+ DT(DT_), IsTrulyNegation(IsTrulyNegation_) {}\n \n #if LLVM_ENABLE_STATS\n Negator::~Negator() {\n@@ -309,6 +310,9 @@ std::array Negator::getSortedOperandsOfBinOp(Instruction *I) {\n auto *PHI = cast(I);\n SmallVector NegatedIncomingValues(PHI->getNumOperands());\n for (auto I : zip(PHI->incoming_values(), NegatedIncomingValues)) {\n+ // Don't negate indvars to avoid infinite loops.\n+ if (DT.dominates(PHI->getParent(), std::get<0>(I)))\n+ return nullptr;\n if (!(std::get<1>(I) =\n negate(std::get<0>(I), IsNSW, Depth + 1))) // Early return.\n return nullptr;\n@@ -537,7 +541,8 @@ std::array Negator::getSortedOperandsOfBinOp(Instruction *I) {\n if (!NegatorEnabled || !DebugCounter::shouldExecute(NegatorCounter))\n return nullptr;\n \n- Negator N(Root->getContext(), IC.getDataLayout(), LHSIsZero);\n+ Negator N(Root->getContext(), IC.getDataLayout(), IC.getDominatorTree(),\n+ LHSIsZero);\n std::optional Res = N.run(Root, IsNSW);\n if (!Res) { // Negation failed.\n LLVM_DEBUG(dbgs() << \"Negator: failed to sink negation into \" << *Root\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/pr96012.ll", "commands": ["opt -S -passes=instcombine < %s"], "tests": [{"test_name": "", "test_body": "\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\n\n; Make sure we don't hang here.\ndefine i32 @pr96012() {\n;\nentry:\n br label %for.cond\n\nfor.cond:\n %indvar = phi i32 [ 1, %entry ], [ %shl, %for.cond ]\n %conv1 = trunc i32 %indvar to i8\n %neg = sub i8 0, %conv1\n %conv2 = zext i8 %neg to i32\n %shl = shl nuw i32 %conv2, 24\n br label %for.cond\n}"}]}], "issue": {"title": "InstCombine gets stuck when sinking negation", "body": "Minimal reproducer (It looks over-reduced): https://godbolt.org/z/9zj5K8seP\r\n```\r\n; opt -passes=instcombine test.ll -S -debug\r\ntarget datalayout = \"e-m:e-p:64:64-i64:64-i128:128-n32:64-S128\"\r\ntarget triple = \"riscv64-unknown-linux-gnu\"\r\n\r\ndefine i32 @func_112() {\r\nentry:\r\n br label %for.cond\r\n\r\nfor.cond: ; preds = %for.cond, %entry\r\n %l_139.0 = phi i32 [ 1, %entry ], [ %0, %for.cond ]\r\n %conv5 = trunc i32 %l_139.0 to i8\r\n %call6 = sub i8 0, %conv5\r\n %conv7 = zext i8 %call6 to i32\r\n %0 = shl nuw i32 %conv7, 24\r\n br label %for.cond\r\n}\r\n```\r\n\r\n```\r\nINSTCOMBINE ITERATION #1 on func_112\r\nADD: br label %for.cond\r\nADD: %0 = shl nuw i32 %conv7, 24\r\nADD: %conv7 = zext i8 %call6 to i32\r\nADD: %call6 = sub i8 0, %conv5\r\nADD: %conv5 = trunc i32 %l_139.0 to i8\r\nADD: %l_139.0 = phi i32 [ 1, %entry ], [ %0, %for.cond ]\r\nADD: br label %for.cond\r\nIC: Visiting: br label %for.cond\r\nIC: Visiting: %l_139.0 = phi i32 [ 1, %entry ], [ %0, %for.cond ]\r\nIC: Visiting: %conv5 = trunc i32 %l_139.0 to i8\r\nIC: Visiting: %call6 = sub i8 0, %conv5\r\nNegator: attempting to sink negation into %conv5 = trunc i32 %l_139.0 to i8\r\nNegator: successfully sunk negation into %conv5 = trunc i32 %l_139.0 to i8\r\n NEW: %conv5.neg = trunc i32 %l_139.0.neg to i8\r\nNegator: Propagating 3 instrs to InstCombine\r\nADD DEFERRED: %.neg = mul i32 %conv7, -16777216\r\nADD DEFERRED: %l_139.0.neg = phi i32 [ -1, %entry ], [ %.neg, %for.cond ]\r\nADD DEFERRED: %conv5.neg = trunc i32 %l_139.0.neg to i8\r\nIC: Old = %call6 = sub i8 0, %conv5\r\n New = = add i8 %conv5.neg, 0\r\nADD: %call6 = add i8 %conv5.neg, 0\r\nIC: ERASE %0 = sub i8 0, %conv5\r\nADD DEFERRED: %conv5 = trunc i32 %l_139.0 to i8\r\nIC: ERASE %conv5 = trunc i32 %l_139.0 to i8\r\nADD DEFERRED: %l_139.0 = phi i32 [ 1, %entry ], [ %0, %for.cond ]\r\nIC: ERASE %l_139.0 = phi i32 [ 1, %entry ], [ %0, %for.cond ]\r\nADD DEFERRED: %0 = shl nuw i32 %conv7, 24\r\nIC: ERASE %0 = shl nuw i32 %conv7, 24\r\nADD DEFERRED: %conv7 = zext i8 %call6 to i32\r\nADD: %conv5.neg = trunc i32 %l_139.0.neg to i8\r\nADD: %l_139.0.neg = phi i32 [ -1, %entry ], [ %.neg, %for.cond ]\r\nADD: %.neg = mul i32 %conv7, -16777216\r\nIC: Visiting: %.neg = mul i32 %conv7, -16777216\r\nNegator: attempting to sink negation into %conv7 = zext i8 %call6 to i32\r\nNegator: failed to sink negation into %conv7 = zext i8 %call6 to i32\r\nADD DEFERRED: %call6.neg = sub i8 0, %call6\r\nADD DEFERRED: %call6.neg.z = zext i8 %call6.neg to i32\r\nIC: Old = %.neg = mul i32 %conv7, -16777216\r\n New = = shl i32 %call6.neg.z, 24\r\nADD: %.neg = shl i32 %call6.neg.z, 24\r\nIC: ERASE %0 = mul i32 %conv7, -16777216\r\nADD DEFERRED: %conv7 = zext i8 %call6 to i32\r\nIC: ERASE %conv7 = zext i8 %call6 to i32\r\nADD DEFERRED: %call6 = add i8 %conv5.neg, 0\r\nADD: %call6.neg.z = zext i8 %call6.neg to i32\r\nADD: %call6.neg = sub i8 0, %call6\r\nIC: Visiting: %call6.neg = sub i8 0, %call6\r\nIC: Old = %call6.neg = sub i8 0, %call6\r\n New = = sub i8 0, %conv5.neg\r\nADD: %call6.neg = sub i8 0, %conv5.neg\r\nIC: ERASE %0 = sub i8 0, %call6\r\nADD DEFERRED: %call6 = add i8 %conv5.neg, 0\r\nIC: ERASE %call6 = add i8 %conv5.neg, 0\r\nADD DEFERRED: %conv5.neg = trunc i32 %l_139.0.neg to i8\r\nADD DEFERRED: %call6.neg = sub i8 0, %conv5.neg\r\nIC: Visiting: %call6.neg = sub i8 0, %conv5.neg\r\nNegator: attempting to sink negation into %conv5.neg = trunc i32 %l_139.0.neg to i8\r\nNegator: successfully sunk negation into %conv5.neg = trunc i32 %l_139.0.neg to i8\r\n NEW: %conv5.neg.neg = trunc i32 %l_139.0.neg.neg to i8\r\nNegator: Propagating 3 instrs to InstCombine\r\nADD DEFERRED: %.neg.neg = mul i32 %call6.neg.z, -16777216\r\nADD DEFERRED: %l_139.0.neg.neg = phi i32 [ 1, %entry ], [ %.neg.neg, %for.cond ]\r\nADD DEFERRED: %conv5.neg.neg = trunc i32 %l_139.0.neg.neg to i8\r\nIC: Old = %call6.neg = sub i8 0, %conv5.neg\r\n New = = add i8 %conv5.neg.neg, 0\r\nADD: %call6.neg = add i8 %conv5.neg.neg, 0\r\nIC: ERASE %0 = sub i8 0, %conv5.neg\r\nADD DEFERRED: %conv5.neg = trunc i32 %l_139.0.neg to i8\r\nIC: ERASE %conv5.neg = trunc i32 %l_139.0.neg to i8\r\nADD DEFERRED: %l_139.0.neg = phi i32 [ -1, %entry ], [ %.neg, %for.cond ]\r\nIC: ERASE %l_139.0.neg = phi i32 [ -1, %entry ], [ %.neg, %for.cond ]\r\nADD DEFERRED: %.neg = shl i32 %call6.neg.z, 24\r\nIC: ERASE %.neg = shl i32 %call6.neg.z, 24\r\nADD DEFERRED: %call6.neg.z = zext i8 %call6.neg to i32\r\nADD: %conv5.neg.neg = trunc i32 %l_139.0.neg.neg to i8\r\nADD: %l_139.0.neg.neg = phi i32 [ 1, %entry ], [ %.neg.neg, %for.cond ]\r\nADD: %.neg.neg = mul i32 %call6.neg.z, -16777216\r\nIC: Visiting: %.neg.neg = mul i32 %call6.neg.z, -16777216\r\nNegator: attempting to sink negation into %call6.neg.z = zext i8 %call6.neg to i32\r\nNegator: failed to sink negation into %call6.neg.z = zext i8 %call6.neg to i32\r\nADD DEFERRED: %call6.neg.neg = sub i8 0, %call6.neg\r\nADD DEFERRED: %call6.neg.neg.z = zext i8 %call6.neg.neg to i32\r\nIC: Old = %.neg.neg = mul i32 %call6.neg.z, -16777216\r\n New = = shl i32 %call6.neg.neg.z, 24\r\nADD: %.neg.neg = shl i32 %call6.neg.neg.z, 24\r\nIC: ERASE %0 = mul i32 %call6.neg.z, -16777216\r\nADD DEFERRED: %call6.neg.z = zext i8 %call6.neg to i32\r\nIC: ERASE %call6.neg.z = zext i8 %call6.neg to i32\r\nADD DEFERRED: %call6.neg = add i8 %conv5.neg.neg, 0\r\nADD: %call6.neg.neg.z = zext i8 %call6.neg.neg to i32\r\nADD: %call6.neg.neg = sub i8 0, %call6.neg\r\nIC: Visiting: %call6.neg.neg = sub i8 0, %call6.neg\r\nIC: Old = %call6.neg.neg = sub i8 0, %call6.neg\r\n New = = sub i8 0, %conv5.neg.neg\r\nADD: %call6.neg.neg = sub i8 0, %conv5.neg.neg\r\nIC: ERASE %0 = sub i8 0, %call6.neg\r\nADD DEFERRED: %call6.neg = add i8 %conv5.neg.neg, 0\r\nIC: ERASE %call6.neg = add i8 %conv5.neg.neg, 0\r\nADD DEFERRED: %conv5.neg.neg = trunc i32 %l_139.0.neg.neg to i8\r\nADD DEFERRED: %call6.neg.neg = sub i8 0, %conv5.neg.neg\r\nIC: Visiting: %call6.neg.neg = sub i8 0, %conv5.neg.neg\r\nNegator: attempting to sink negation into %conv5.neg.neg = trunc i32 %l_139.0.neg.neg to i8\r\nNegator: successfully sunk negation into %conv5.neg.neg = trunc i32 %l_139.0.neg.neg to i8\r\n NEW: %conv5.neg.neg.neg = trunc i32 %l_139.0.neg.neg.neg to i8\r\nNegator: Propagating 3 instrs to InstCombine\r\nADD DEFERRED: %.neg.neg.neg = mul i32 %call6.neg.neg.z, -16777216\r\nADD DEFERRED: %l_139.0.neg.neg.neg = phi i32 [ -1, %entry ], [ %.neg.neg.neg, %for.cond ]\r\nADD DEFERRED: %conv5.neg.neg.neg = trunc i32 %l_139.0.neg.neg.neg to i8\r\nIC: Old = %call6.neg.neg = sub i8 0, %conv5.neg.neg\r\n New = = add i8 %conv5.neg.neg.neg, 0\r\nADD: %call6.neg.neg = add i8 %conv5.neg.neg.neg, 0\r\nIC: ERASE %0 = sub i8 0, %conv5.neg.neg\r\nADD DEFERRED: %conv5.neg.neg = trunc i32 %l_139.0.neg.neg to i8\r\nIC: ERASE %conv5.neg.neg = trunc i32 %l_139.0.neg.neg to i8\r\nADD DEFERRED: %l_139.0.neg.neg = phi i32 [ 1, %entry ], [ %.neg.neg, %for.cond ]\r\nIC: ERASE %l_139.0.neg.neg = phi i32 [ 1, %entry ], [ %.neg.neg, %for.cond ]\r\nADD DEFERRED: %.neg.neg = shl i32 %call6.neg.neg.z, 24\r\nIC: ERASE %.neg.neg = shl i32 %call6.neg.neg.z, 24\r\nADD DEFERRED: %call6.neg.neg.z = zext i8 %call6.neg.neg to i32\r\nADD: %conv5.neg.neg.neg = trunc i32 %l_139.0.neg.neg.neg to i8\r\nADD: %l_139.0.neg.neg.neg = phi i32 [ -1, %entry ], [ %.neg.neg.neg, %for.cond ]\r\nADD: %.neg.neg.neg = mul i32 %call6.neg.neg.z, -16777216\r\nIC: Visiting: %.neg.neg.neg = mul i32 %call6.neg.neg.z, -16777216\r\nNegator: attempting to sink negation into %call6.neg.neg.z = zext i8 %call6.neg.neg to i32\r\nNegator: failed to sink negation into %call6.neg.neg.z = zext i8 %call6.neg.neg to i32\r\nADD DEFERRED: %call6.neg.neg.neg = sub i8 0, %call6.neg.neg\r\nADD DEFERRED: %call6.neg.neg.neg.z = zext i8 %call6.neg.neg.neg to i32\r\nIC: Old = %.neg.neg.neg = mul i32 %call6.neg.neg.z, -16777216\r\n New = = shl i32 %call6.neg.neg.neg.z, 24\r\nADD: %.neg.neg.neg = shl i32 %call6.neg.neg.neg.z, 24\r\nIC: ERASE %0 = mul i32 %call6.neg.neg.z, -16777216\r\nADD DEFERRED: %call6.neg.neg.z = zext i8 %call6.neg.neg to i32\r\nIC: ERASE %call6.neg.neg.z = zext i8 %call6.neg.neg to i32\r\nADD DEFERRED: %call6.neg.neg = add i8 %conv5.neg.neg.neg, 0\r\nADD: %call6.neg.neg.neg.z = zext i8 %call6.neg.neg.neg to i32\r\nADD: %call6.neg.neg.neg = sub i8 0, %call6.neg.neg\r\nIC: Visiting: %call6.neg.neg.neg = sub i8 0, %call6.neg.neg\r\nIC: Old = %call6.neg.neg.neg = sub i8 0, %call6.neg.neg\r\n New = = sub i8 0, %conv5.neg.neg.neg\r\nADD: %call6.neg.neg.neg = sub i8 0, %conv5.neg.neg.neg\r\nIC: ERASE %0 = sub i8 0, %call6.neg.neg\r\nADD DEFERRED: %call6.neg.neg = add i8 %conv5.neg.neg.neg, 0\r\nIC: ERASE %call6.neg.neg = add i8 %conv5.neg.neg.neg, 0\r\nADD DEFERRED: %conv5.neg.neg.neg = trunc i32 %l_139.0.neg.neg.neg to i8\r\nADD DEFERRED: %call6.neg.neg.neg = sub i8 0, %conv5.neg.neg.neg\r\nIC: Visiting: %call6.neg.neg.neg = sub i8 0, %conv5.neg.neg.neg\r\nNegator: attempting to sink negation into %conv5.neg.neg.neg = trunc i32 %l_139.0.neg.neg.neg to i8\r\nNegator: successfully sunk negation into %conv5.neg.neg.neg = trunc i32 %l_139.0.neg.neg.neg to i8\r\n NEW: %conv5.neg.neg.neg.neg = trunc i32 %l_139.0.neg.neg.neg.neg to i8\r\n...\r\n```", "author": "dtcxzyw", "labels": ["llvm:instcombine", "llvm:hang"], "comments": []}, "verified": true} {"bug_id": "98610", "issue_url": "https://github.com/llvm/llvm-project/issues/98610", "bug_type": "hang", "base_commit": "9ad72df55cb74b29193270c28f6974d2af8e0b71", "knowledge_cutoff": "2024-07-12T10:31:22Z", "lit_test_dir": ["llvm/test/Transforms/MemCpyOpt"], "hints": {"fix_commit": "71051deff27928cff908ea794e09806eee662801", "components": ["MemCpyOptimizer"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp": [[1296, 1301], [1312, 1318], [1726, 1733]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp": ["MemCpyOptPass::processMemCpy", "MemCpyOptPass::processMemSetMemCpyDependence"]}}, "patch": "commit 71051deff27928cff908ea794e09806eee662801\nAuthor: Nikita Popov \nDate: Mon Jul 15 09:41:11 2024 +0200\n\n [MemCpyOpt] Fix infinite loop in memset+memcpy fold (#98638)\n \n For the case where the memcpy size is zero, this transform is a complex\n no-op. This can lead to an infinite loop when the size is zero in a way\n that BasicAA understands, because it can still understand that dst and\n dst + src_size are MustAlias.\n \n I've tried to mitigate this before using the isZeroSize() check, but we\n can hit cases where InstSimplify doesn't understand that the size is\n zero, but BasicAA does.\n \n As such, this bites the bullet and adds an explicit isKnownNonZero()\n check to guard against no-op transforms.\n \n Fixes https://github.com/llvm/llvm-project/issues/98610.\n\ndiff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\nindex 4063762c88a2..cee34f0a6da1 100644\n--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\n+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp\n@@ -1296,6 +1296,15 @@ bool MemCpyOptPass::processMemSetMemCpyDependence(MemCpyInst *MemCpy,\n if (!BAA.isMustAlias(MemSet->getDest(), MemCpy->getDest()))\n return false;\n \n+ // Don't perform the transform if src_size may be zero. In that case, the\n+ // transform is essentially a complex no-op and may lead to an infinite\n+ // loop if BasicAA is smart enough to understand that dst and dst + src_size\n+ // are still MustAlias after the transform.\n+ Value *SrcSize = MemCpy->getLength();\n+ if (!isKnownNonZero(SrcSize,\n+ SimplifyQuery(MemCpy->getDataLayout(), DT, AC, MemCpy)))\n+ return false;\n+\n // Check that src and dst of the memcpy aren't the same. While memcpy\n // operands cannot partially overlap, exact equality is allowed.\n if (isModSet(BAA.getModRefInfo(MemCpy, MemoryLocation::getForSource(MemCpy))))\n@@ -1312,7 +1321,6 @@ bool MemCpyOptPass::processMemSetMemCpyDependence(MemCpyInst *MemCpy,\n // Use the same i8* dest as the memcpy, killing the memset dest if different.\n Value *Dest = MemCpy->getRawDest();\n Value *DestSize = MemSet->getLength();\n- Value *SrcSize = MemCpy->getLength();\n \n if (mayBeVisibleThroughUnwinding(Dest, MemSet, MemCpy))\n return false;\n@@ -1726,8 +1734,7 @@ bool MemCpyOptPass::processMemCpy(MemCpyInst *M, BasicBlock::iterator &BBI) {\n return true;\n }\n \n- // If the size is zero, remove the memcpy. This also prevents infinite loops\n- // in processMemSetMemCpyDependence, which is a no-op for zero-length memcpys.\n+ // If the size is zero, remove the memcpy.\n if (isZeroSize(M->getLength())) {\n ++BBI;\n eraseInstruction(M);\n", "tests": [{"file": "llvm/test/Transforms/MemCpyOpt/memcpy-zero-size.ll", "commands": ["opt -S -passes=memcpyopt < %s"], "tests": [{"test_name": "pr64886", "test_body": "; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\ndefine void @pr64886(i64 %len, ptr noalias %p) {\n call void @llvm.memset.p0.i64(ptr inttoptr (i64 -1 to ptr), i8 0, i64 %len, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr inttoptr (i64 -1 to ptr), ptr %p, i64 poison, i1 false)\n ret void\n}\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}, {"test_name": "pr98610", "test_body": "; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\ndefine void @pr98610(ptr %p, ptr noalias %p2) {\n call void @llvm.memset.p0.i64(ptr %p, i8 0, i64 1, i1 false)\n %zero.ext = zext i32 0 to i64\n %mul = mul i64 %zero.ext, 1\n call void @llvm.memcpy.p0.p0.i64(ptr %p, ptr %p2, i64 %mul, i1 false)\n ret void\n}\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}]}, {"file": "llvm/test/Transforms/MemCpyOpt/memset-memcpy-dbgloc.ll", "commands": ["opt -passes=memcpyopt -S %s -verify-memoryssa"], "tests": [{"test_name": "test_constant", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\n@C = external constant [0 x i8]\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\ndefine void @test_constant(i64 %src_size, ptr %dst, i64 %dst_size, i8 %c) !dbg !5 {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false), !dbg !11\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr @C, i64 %src_size, i1 false), !dbg !12\n ret void, !dbg !13\n}\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n\n!llvm.dbg.cu = !{!0}\n!llvm.debugify = !{!2, !3}\n!llvm.module.flags = !{!4}\n\n!0 = distinct !DICompileUnit(language: DW_LANG_C, file: !1, producer: \"debugify\", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug)\n!1 = !DIFile(filename: \"memset-memcpy-dbgloc.ll\", directory: \"/\")\n!2 = !{i32 3}\n!3 = !{i32 1}\n!4 = !{i32 2, !\"Debug Info Version\", i32 3}\n!5 = distinct !DISubprogram(name: \"test_constant\", linkageName: \"test_constant\", scope: null, file: !1, line: 1, type: !6, scopeLine: 1, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !8)\n!6 = !DISubroutineType(types: !7)\n!7 = !{}\n!8 = !{!9}\n!9 = !DILocalVariable(name: \"1\", scope: !5, file: !1, line: 3, type: !10)\n!10 = !DIBasicType(name: \"ty32\", size: 32, encoding: DW_ATE_unsigned)\n!11 = !DILocation(line: 1, column: 1, scope: !5)\n!12 = !DILocation(line: 2, column: 1, scope: !5)\n!13 = !DILocation(line: 3, column: 1, scope: !5)\n"}]}, {"file": "llvm/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll", "commands": ["opt -passes=memcpyopt -S %s -verify-memoryssa"], "tests": [{"test_name": "test_different_types_i32_i128", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_different_types_i32_i128(ptr noalias %dst, ptr %src, i32 %dst_size, i128 %src_size, i8 %c) {\n %non.zero = icmp ne i128 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i32(ptr %dst, i8 %c, i32 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i128(ptr %dst, ptr %src, i128 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i32(ptr nocapture writeonly, i8, i32, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i128(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i128, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_missing_noalias", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_missing_noalias(ptr %src, i64 %src_size, ptr %dst, i64 %dst_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_different_types_i128_i32", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_different_types_i128_i32(ptr noalias %dst, ptr %src, i128 %dst_size, i32 %src_size, i8 %c) {\n %non.zero = icmp ne i32 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i128(ptr %dst, i8 %c, i128 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i32(ptr %dst, ptr %src, i32 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i32(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i32, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i128(ptr nocapture writeonly, i8, i128, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_align_memcpy", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_align_memcpy(ptr %src, ptr noalias %dst, i64 %dst_size) {\n call void @llvm.memset.p0.i64(ptr %dst, i8 0, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr align 8 %dst, ptr align 8 %src, i64 80, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}, {"test_name": "test_different_dst", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_different_dst(ptr noalias %dst2, ptr %src, i64 %src_size, ptr noalias %dst, i64 %dst_size) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 0, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst2, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_non_i8_dst_type", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_non_i8_dst_type(ptr %src, i64 %src_size, ptr noalias %dst_pi64, i64 %dst_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst_pi64, i8 %c, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst_pi64, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_different_types_i64_i32", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_different_types_i64_i32(ptr noalias %dst, ptr %src, i64 %dst_size, i32 %src_size, i8 %c) {\n %non.zero = icmp ne i32 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i32(ptr %dst, ptr %src, i32 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i32(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i32, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_intermediate_write", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\n%struct = type { [8 x i8], [8 x i8] }\n\ndefine void @test_intermediate_write(ptr %b) {\n %a = alloca %struct, align 8\n %a1 = getelementptr %struct, ptr %a, i32 0, i32 1, i32 0\n call void @llvm.memset.p0.i64(ptr %a, i8 0, i64 16, i1 false)\n store i8 1, ptr %a1, align 1\n call void @llvm.memcpy.p0.p0.i64(ptr %a, ptr %b, i64 8, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}, {"test_name": "test", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test(ptr %src, i64 %src_size, ptr noalias %dst, i64 %dst_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_same_dynamic_size", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_same_dynamic_size(ptr noalias %src, ptr noalias %dst, i64 %size, i8 %c) {\n %non.zero = icmp ne i64 %size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 %size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_weird_element_type", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_weird_element_type(ptr %src, i64 %src_size, ptr noalias %dst, i64 %dst_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_throwing_call", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_throwing_call(ptr %src, i64 %src_size, ptr noalias %dst, i64 %dst_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false)\n call void @call() #3\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\ndeclare void @call()\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\nattributes #3 = { memory(none) }\n"}, {"test_name": "test_same_const_size", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_same_const_size(ptr noalias %src, ptr noalias %dst, i8 %c) {\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 16, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 16, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}, {"test_name": "test_must_alias_different_size", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_must_alias_different_size(ptr noalias %src, i64 %src_size, ptr noalias %dst, i64 %dst_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n %gep1 = getelementptr i8, ptr %dst, i64 16\n call void @llvm.memset.p0.i64(ptr %gep1, i8 %c, i64 %dst_size, i1 false)\n %gep2 = getelementptr i8, ptr %dst, i64 16\n call void @llvm.memcpy.p0.p0.i64(ptr %gep2, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_throwing_call_alloca", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_throwing_call_alloca(ptr %src, i64 %src_size, i64 %dst_size, i8 %c) {\n %dst = alloca i8, align 1\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false)\n call void @call() #3\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\ndeclare void @call()\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\nattributes #3 = { memory(none) }\n"}, {"test_name": "test_addrspace", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_addrspace(ptr addrspace(1) %src, i64 %src_size, ptr addrspace(1) noalias %dst, i64 %dst_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p1.i64(ptr addrspace(1) %dst, i8 %c, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) %dst, ptr addrspace(1) %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p1.i64(ptr addrspace(1) nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) noalias nocapture writeonly, ptr addrspace(1) noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_constant", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\n@C = external constant [0 x i8]\n\ndefine void @test_constant(i64 %src_size, ptr %dst, i64 %dst_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr @C, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_different_types_i32_i64", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_different_types_i32_i64(ptr noalias %dst, ptr %src, i32 %dst_size, i64 %src_size, i8 %c) {\n %non.zero = icmp ne i64 %src_size, 0\n call void @llvm.assume(i1 %non.zero)\n call void @llvm.memset.p0.i32(ptr %dst, i8 %c, i32 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i32(ptr nocapture writeonly, i8, i32, i1 immarg) #1\n\n; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write)\ndeclare void @llvm.assume(i1 noundef) #2\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #2 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: write) }\n"}, {"test_name": "test_must_alias_same_size", "test_body": "target datalayout = \"e-m:o-i64:64-f80:128-n8:16:32:64-S128\"\n\ndefine void @test_must_alias_same_size(ptr noalias %src, ptr noalias %dst, i8 %c) {\n %gep1 = getelementptr i8, ptr %dst, i64 16\n call void @llvm.memset.p0.i64(ptr %gep1, i8 %c, i64 16, i1 false)\n %gep2 = getelementptr i8, ptr %dst, i64 16\n call void @llvm.memcpy.p0.p0.i64(ptr %gep2, ptr %src, i64 16, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}]}, {"file": "llvm/test/Transforms/MemCpyOpt/opaque-ptr.ll", "commands": ["opt -passes=memcpyopt -S %s -verify-memoryssa"], "tests": [{"test_name": "test_memset_memcpy", "test_body": "define void @test_memset_memcpy(ptr %src, i64 range(i64 1, 42) %src_size, ptr noalias %dst, i64 %dst_size, i8 %c) {\n call void @llvm.memset.p0.i64(ptr %dst, i8 %c, i64 %dst_size, i1 false)\n call void @llvm.memcpy.p0.p0.i64(ptr %dst, ptr %src, i64 %src_size, i1 false)\n ret void\n}\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)\ndeclare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i1 immarg) #0\n\n; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)\ndeclare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #1\n\nattributes #0 = { nocallback nofree nounwind willreturn memory(argmem: write) }\nattributes #1 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }\n"}]}], "issue": {"title": "Clang hangs forever building under -O1", "body": "ToT Clang hangs when building under -O1.\r\n\r\n```\r\nclang++ test.i -fno-exceptions -O1 -c -emit-llvm\r\n```\r\n\r\nThe reproducer, `test.i`:\r\n\r\n```c++\r\nclass MockDevice;\r\nstruct sp {\r\n MockDevice operator*();\r\n};\r\nvoid memset(void *__dest, int __ch, unsigned __len) {\r\n long __trans_tmp_1 = __builtin_object_size(__dest, 0);\r\n __builtin___memset_chk(__dest, __ch, __len, __trans_tmp_1);\r\n}\r\ntemplate \r\nclass function;\r\ntemplate \r\nclass function<_Rp(_ArgTypes...)> {\r\n public:\r\n template \r\n function operator=(_Fp);\r\n};\r\ntemplate \r\nstruct hidl_vec {\r\n hidl_vec() : mSize() { memset(mPad, 0, 0); }\r\n hidl_vec(hidl_vec &other) { *this = other; }\r\n hidl_vec &operator=(hidl_vec &other) {\r\n copyFrom(other, other.mSize);\r\n return *this;\r\n }\r\n T &operator[](unsigned index) { return mBuffer[index]; }\r\n T* mBuffer;\r\n unsigned mSize;\r\n unsigned mPad[];\r\n template \r\n void copyFrom(Array &data, unsigned size) {\r\n mSize = size;\r\n if (mSize) mBuffer = new T();\r\n for (int i = 0; i < size; ++i) mBuffer[i] = data[i];\r\n }\r\n};\r\nstruct Capabilities {\r\n struct OperandPerformance {\r\n int a, b, c;\r\n };\r\n hidl_vec operandPerformance;\r\n};\r\ntemplate \r\nclass Action;\r\ntemplate \r\nclass Action {\r\n using F = R();\r\n public:\r\n template \r\n Action(G fun) {\r\n fun_ = fun;\r\n }\r\n function fun_;\r\n};\r\nstruct OnCallSpec {\r\n OnCallSpec WillByDefault(Action);\r\n};\r\nstruct MockSpec {\r\n OnCallSpec InternalDefaultActionSetAt();\r\n};\r\nstruct MockDevice {\r\n MockSpec gmock_getCapabilities();\r\n};\r\nsp createMockDevice_mockDevice;\r\nMockDevice createMockDevice() {\r\n Capabilities getCapabilities_ret;\r\n (*createMockDevice_mockDevice)\r\n .gmock_getCapabilities()\r\n .InternalDefaultActionSetAt()\r\n .WillByDefault(getCapabilities_ret);\r\n return *createMockDevice_mockDevice;\r\n}\r\n```", "author": "kongy", "labels": ["confirmed", "llvm:hang", "llvm:transforms"], "comments": [{"author": "AaronBallman", "body": "This seems to be hanging somewhere in this vicinity:\r\n```\r\n>\tclang.exe!llvm::detail::CheckedInt::from(int FromValue) Line 136\tC++\r\n \tclang.exe!llvm::detail::CheckedInt::from(llvm::IRMemLocation FromValue) Line 143\tC++\r\n \tclang.exe!llvm::detail::SafeIntIterator::SafeIntIterator(llvm::IRMemLocation Value) Line 196\tC++\r\n \tclang.exe!llvm::iota_range::iota_range(llvm::IRMemLocation Begin, llvm::IRMemLocation End, bool Inclusive) Line 275\tC++\r\n \tclang.exe!llvm::enum_seq_inclusive(llvm::IRMemLocation Begin, llvm::IRMemLocation End, llvm::force_iteration_on_noniterable_enum_t __formal) Line 381\tC++\r\n \tclang.exe!llvm::MemoryEffectsBase::locations() Line 96\tC++\r\n \tclang.exe!llvm::MemoryEffectsBase::MemoryEffectsBase(llvm::ModRefInfo MR) Line 107\tC++\r\n \tclang.exe!llvm::MemoryEffectsBase::unknown() Line 113\tC++\r\n \tclang.exe!llvm::GlobalsAAResult::getMemoryEffects(const llvm::Function * F) Line 244\tC++\r\n \tclang.exe!llvm::AAResults::Model::getMemoryEffects(const llvm::Function * F) Line 784\tC++\r\n \tclang.exe!llvm::AAResults::getMemoryEffects(const llvm::Function * F) Line 396\tC++\r\n \tclang.exe!llvm::BasicAAResult::getMemoryEffects(const llvm::CallBase * Call, llvm::AAQueryInfo & AAQI) Line 830\tC++\r\n \tclang.exe!llvm::AAResults::Model::getMemoryEffects(const llvm::CallBase * Call, llvm::AAQueryInfo & AAQI) Line 780\tC++\r\n \tclang.exe!llvm::AAResults::getMemoryEffects(const llvm::CallBase * Call, llvm::AAQueryInfo & AAQI) Line 377\tC++\r\n \tclang.exe!llvm::AAResults::getModRefInfo(const llvm::CallBase * Call, const llvm::MemoryLocation & Loc, llvm::AAQueryInfo & AAQI) Line 230\tC++\r\n \tclang.exe!llvm::AAResults::getModRefInfo(const llvm::Instruction * I, const std::optional & OptLoc, llvm::AAQueryInfo & AAQIP) Line 628\tC++\r\n \tclang.exe!llvm::BatchAAResults::getModRefInfo(const llvm::Instruction * I, const std::optional & OptLoc) Line 646\tC++\r\n \tclang.exe!llvm::MemCpyOptPass::processMemSetMemCpyDependence(llvm::MemCpyInst * MemCpy, llvm::MemSetInst * MemSet, llvm::BatchAAResults & BAA) Line 1252\tC++\r\n \tclang.exe!llvm::MemCpyOptPass::processMemCpy(llvm::MemCpyInst * M, llvm::ilist_iterator_w_bits,0,0> & BBI) Line 1725\tC++\r\n \tclang.exe!llvm::MemCpyOptPass::iterateOnFunction(llvm::Function & F) Line 2013\tC++\r\n \tclang.exe!llvm::MemCpyOptPass::runImpl(llvm::Function & F, llvm::TargetLibraryInfo * TLI_, llvm::AAResults * AA_, llvm::AssumptionCache * AC_, llvm::DominatorTree * DT_, llvm::PostDominatorTree * PDT_, llvm::MemorySSA * MSSA_) Line 2070\tC++\r\n \tclang.exe!llvm::MemCpyOptPass::run(llvm::Function & F, llvm::AnalysisManager & AM) Line 2045\tC++\r\n \tclang.exe!llvm::detail::PassModel>::run(llvm::Function & IR, llvm::AnalysisManager & AM) Line 90\tC++\r\n \tclang.exe!llvm::PassManager>::run(llvm::Function & IR, llvm::AnalysisManager & AM) Line 85\tC++\r\n \tclang.exe!llvm::detail::PassModel>,llvm::AnalysisManager>::run(llvm::Function & IR, llvm::AnalysisManager & AM) Line 90\tC++\r\n \tclang.exe!llvm::CGSCCToFunctionPassAdaptor::run(llvm::LazyCallGraph::SCC & C, llvm::AnalysisManager & AM, llvm::LazyCallGraph & CG, llvm::CGSCCUpdateResult & UR) Line 543\tC++\r\n \tclang.exe!llvm::detail::PassModel,llvm::LazyCallGraph &,llvm::CGSCCUpdateResult &>::run(llvm::LazyCallGraph::SCC & IR, llvm::AnalysisManager & AM, llvm::LazyCallGraph & , llvm::CGSCCUpdateResult & ) Line 90\tC++\r\n \tclang.exe!llvm::PassManager,llvm::LazyCallGraph &,llvm::CGSCCUpdateResult &>::run(llvm::LazyCallGraph::SCC & InitialC, llvm::AnalysisManager & AM, llvm::LazyCallGraph & G, llvm::CGSCCUpdateResult & UR) Line 90\tC++\r\n \tclang.exe!llvm::detail::PassModel,llvm::LazyCallGraph &,llvm::CGSCCUpdateResult &>,llvm::AnalysisManager,llvm::LazyCallGraph &,llvm::CGSCCUpdateResult &>::run(llvm::LazyCallGraph::SCC & IR, llvm::AnalysisManager & AM, llvm::LazyCallGraph & , llvm::CGSCCUpdateResult & ) Line 90\tC++\r\n \tclang.exe!llvm::DevirtSCCRepeatedPass::run(llvm::LazyCallGraph::SCC & InitialC, llvm::AnalysisManager & AM, llvm::LazyCallGraph & CG, llvm::CGSCCUpdateResult & UR) Line 415\tC++\r\n \tclang.exe!llvm::detail::PassModel,llvm::LazyCallGraph &,llvm::CGSCCUpdateResult &>::run(llvm::LazyCallGraph::SCC & IR, llvm::AnalysisManager & AM, llvm::LazyCallGraph & , llvm::CGSCCUpdateResult & ) Line 90\tC++\r\n \tclang.exe!llvm::ModuleToPostOrderCGSCCPassAdaptor::run(llvm::Module & M, llvm::AnalysisManager & AM) Line 277\tC++\r\n \tclang.exe!llvm::detail::PassModel>::run(llvm::Module & IR, llvm::AnalysisManager & AM) Line 90\tC++\r\n \tclang.exe!llvm::PassManager>::run(llvm::Module & IR, llvm::AnalysisManager & AM) Line 85\tC++\r\n \tclang.exe!llvm::ModuleInlinerWrapperPass::run(llvm::Module & M, llvm::AnalysisManager & MAM) Line 631\tC++\r\n \tclang.exe!llvm::detail::PassModel>::run(llvm::Module & IR, llvm::AnalysisManager & AM) Line 90\tC++\r\n \tclang.exe!llvm::PassManager>::run(llvm::Module & IR, llvm::AnalysisManager & AM) Line 85\tC++\r\n \tclang.exe!`anonymous namespace'::EmitAssemblyHelper::RunOptimizationPipeline(clang::BackendAction Action, std::unique_ptr> & OS, std::unique_ptr> & ThinLinkOS, clang::BackendConsumer * BC) Line 1114\tC++\r\n \tclang.exe!`anonymous namespace'::EmitAssemblyHelper::EmitAssembly(clang::BackendAction Action, std::unique_ptr> OS, clang::BackendConsumer * BC) Line 1180\tC++\r\n \tclang.exe!clang::EmitBackendOutput(clang::DiagnosticsEngine & Diags, const clang::HeaderSearchOptions & HeaderOpts, const clang::CodeGenOptions & CGOpts, const clang::TargetOptions & TOpts, const clang::LangOptions & LOpts, llvm::StringRef TDesc, llvm::Module * M, clang::BackendAction Action, llvm::IntrusiveRefCntPtr VFS, std::unique_ptr> OS, clang::BackendConsumer * BC) Line 1341\tC++\r\n \tclang.exe!clang::BackendConsumer::HandleTranslationUnit(clang::ASTContext & C) Line 354\tC++\r\n \tclang.exe!clang::ParseAST(clang::Sema & S, bool PrintStats, bool SkipFunctionBodies) Line 183\tC++\r\n \tclang.exe!clang::ASTFrontendAction::ExecuteAction() Line 1192\tC++\r\n \tclang.exe!clang::CodeGenAction::ExecuteAction() Line 1144\tC++\r\n \tclang.exe!clang::FrontendAction::Execute() Line 1078\tC++\r\n \tclang.exe!clang::CompilerInstance::ExecuteAction(clang::FrontendAction & Act) Line 1061\tC++\r\n \tclang.exe!clang::ExecuteCompilerInvocation(clang::CompilerInstance * Clang) Line 280\tC++\r\n \tclang.exe!cc1_main(llvm::ArrayRef Argv, const char * Argv0, void * MainAddr) Line 283\tC++\r\n \tclang.exe!ExecuteCC1Tool(llvm::SmallVectorImpl & ArgV, const llvm::ToolContext & ToolContext) Line 215\tC++\r\n \tclang.exe!clang_main(int Argc, char * * Argv, const llvm::ToolContext & ToolContext) Line 256\tC++\r\n \tclang.exe!main(int argc, char * * argv) Line 17\tC++\r\n \tclang.exe!invoke_main() Line 79\tC++\r\n \tclang.exe!__scrt_common_main_seh() Line 288\tC++\r\n \tclang.exe!__scrt_common_main() Line 331\tC++\r\n \tclang.exe!mainCRTStartup(void * __formal) Line 17\tC++\r\n \tkernel32.dll!00007ff830e07344()\tUnknown\r\n \tntdll.dll!00007ff83149cc91()\tUnknown\r\n```"}, {"author": "nikic", "body": "Reduced:\r\n```llvm\r\n; RUN: opt -S -passes=memcpyopt < %s\r\ndefine void @test(ptr %p, ptr noalias %p2) {\r\n call void @llvm.memset.p0.i64(ptr %p, i8 0, i64 1, i1 false)\r\n %zero.ext = zext i32 0 to i64\r\n %mul = mul i64 %zero.ext, 1\r\n call void @llvm.memcpy.p0.p0.i64(ptr %p, ptr %p2, i64 %mul, i1 false)\r\n ret void\r\n}\r\n```"}, {"author": "nikic", "body": "This is an infinite loop in the processMemSetMemCpyDependence() transform. I've tried to mitigate this in the past, but it looks like the mitigation doesn't work for this case. I guess we'll have to add an isKnownNonZero() check to prevent it."}]}, "verified": true} {"bug_id": "115574", "issue_url": "https://github.com/llvm/llvm-project/issues/115574", "bug_type": "miscompilation", "base_commit": "00bdce1c373a1c5b756f4cf694a952ef702d0294", "knowledge_cutoff": "2024-11-09T01:49:48Z", "lit_test_dir": ["llvm/test/Transforms/InstSimplify"], "hints": {"fix_commit": "1af627b592dd15bbe58136f902ced46251fc344d", "components": ["InstructionSimplify"], "bug_location_lineno": {"llvm/lib/Analysis/InstructionSimplify.cpp": [[27, 32], [4731, 4742]]}, "bug_location_funcname": {"llvm/lib/Analysis/InstructionSimplify.cpp": ["simplifySelectWithICmpCond"]}}, "patch": "commit 1af627b592dd15bbe58136f902ced46251fc344d\nAuthor: Yingwei Zheng \nDate: Sun Feb 2 19:04:23 2025 +0800\n\n [InstSimplify] Add additional checks when substituting pointers (#125385)\n \n Compile-time impact:\n https://llvm-compile-time-tracker.com/compare.php?from=d09b521624f263b5f1296f8d4771836b97e600cb&to=e437ba2cb83bb965e13ef00727671896f03ff84f&stat=instructions:u\n IR diff looks acceptable.\n Closes https://github.com/llvm/llvm-project/issues/115574\n\ndiff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp\nindex 21c937530cd8..3cbc4107433e 100644\n--- a/llvm/lib/Analysis/InstructionSimplify.cpp\n+++ b/llvm/lib/Analysis/InstructionSimplify.cpp\n@@ -27,6 +27,7 @@\n #include \"llvm/Analysis/CmpInstAnalysis.h\"\n #include \"llvm/Analysis/ConstantFolding.h\"\n #include \"llvm/Analysis/InstSimplifyFolder.h\"\n+#include \"llvm/Analysis/Loads.h\"\n #include \"llvm/Analysis/LoopAnalysisManager.h\"\n #include \"llvm/Analysis/MemoryBuiltins.h\"\n #include \"llvm/Analysis/OverflowInstAnalysis.h\"\n@@ -4731,12 +4732,16 @@ static Value *simplifySelectWithICmpCond(Value *CondVal, Value *TrueVal,\n // the arms of the select. See if substituting this value into the arm and\n // simplifying the result yields the same value as the other arm.\n if (Pred == ICmpInst::ICMP_EQ) {\n- if (Value *V = simplifySelectWithEquivalence({{CmpLHS, CmpRHS}}, TrueVal,\n- FalseVal, Q, MaxRecurse))\n- return V;\n- if (Value *V = simplifySelectWithEquivalence({{CmpRHS, CmpLHS}}, TrueVal,\n- FalseVal, Q, MaxRecurse))\n- return V;\n+ if (CmpLHS->getType()->isIntOrIntVectorTy() ||\n+ canReplacePointersIfEqual(CmpLHS, CmpRHS, Q.DL))\n+ if (Value *V = simplifySelectWithEquivalence({{CmpLHS, CmpRHS}}, TrueVal,\n+ FalseVal, Q, MaxRecurse))\n+ return V;\n+ if (CmpLHS->getType()->isIntOrIntVectorTy() ||\n+ canReplacePointersIfEqual(CmpRHS, CmpLHS, Q.DL))\n+ if (Value *V = simplifySelectWithEquivalence({{CmpRHS, CmpLHS}}, TrueVal,\n+ FalseVal, Q, MaxRecurse))\n+ return V;\n \n Value *X;\n Value *Y;\n", "tests": [{"file": "llvm/test/Transforms/InstSimplify/select-icmp.ll", "commands": ["opt < %s -passes=instsimplify -S"], "tests": [{"test_name": "icmp_ptr_eq_replace_null", "test_body": "target datalayout = \"p:8:8:8\"\n\ndefine ptr @icmp_ptr_eq_replace_null(ptr %a) {\n %cmp = icmp eq ptr %a, null\n %sel = select i1 %cmp, ptr null, ptr %a\n ret ptr %sel\n}\n"}, {"test_name": "ptr_eq_replace_same_underlying_object", "test_body": "target datalayout = \"p:8:8:8\"\n\ndefine ptr @ptr_eq_replace_same_underlying_object(ptr %st, i64 %i, i64 %j) {\n %a = getelementptr inbounds i8, ptr %st, i64 %i\n %b = getelementptr inbounds i8, ptr %st, i64 %j\n %cmp = icmp eq ptr %a, %b\n %sel = select i1 %cmp, ptr %a, ptr %b\n ret ptr %sel\n}\n"}, {"test_name": "icmp_ptr_eq_replace", "test_body": "target datalayout = \"p:8:8:8\"\n\ndefine ptr @icmp_ptr_eq_replace(ptr %a, ptr %b) {\n %cmp = icmp eq ptr %a, %b\n %sel = select i1 %cmp, ptr %a, ptr %b\n ret ptr %sel\n}\n"}]}], "issue": {"title": "[InstSimplify] wrong folding of pointer comparison `select (icmp eq ptr ...)`", "body": "https://github.com/llvm/llvm-project/blob/c93e001ca695e905cb965b36d63f7a348d1dd809/llvm/lib/Analysis/InstructionSimplify.cpp#L148-L167\r\n\r\nAlive2 report: https://alive2.llvm.org/ce/z/aux2zY\r\n\r\n```llvm\r\n----------------------------------------\r\ndefine ptr @smin_test8.2(ptr %a, ptr %b, ptr %c) {\r\n#0:\r\n %cmp1 = icmp eq ptr %a, %b\r\n %umin1 = select i1 %cmp1, ptr %a, ptr %b\r\n %cmp2 = icmp slt ptr %b, %c\r\n %umin2 = select i1 %cmp2, ptr %b, ptr %c\r\n %cmp3 = icmp ult ptr %umin2, %a\r\n %umin3 = select i1 %cmp3, ptr %umin2, ptr %a\r\n %cmp4 = icmp slt ptr %c, %umin3\r\n %res = select i1 %cmp4, ptr %umin1, ptr %umin3\r\n ret ptr %res\r\n}\r\n=>\r\ndefine ptr @smin_test8.2(ptr %a, ptr %b, ptr %c) {\r\n#0:\r\n %cmp2 = icmp slt ptr %b, %c\r\n %umin2 = select i1 %cmp2, ptr %b, ptr %c\r\n %cmp3 = icmp ult ptr %umin2, %a\r\n %umin3 = select i1 %cmp3, ptr %umin2, ptr %a\r\n %cmp4 = icmp slt ptr %c, %umin3\r\n %res = select i1 %cmp4, ptr %b, ptr %umin3\r\n ret ptr %res\r\n}\r\nTransformation doesn't verify!\r\n\r\nERROR: Value mismatch\r\n\r\nExample:\r\nptr %a = pointer(non-local, block_id=0, offset=1) / Address=#x1\r\nptr %b = pointer(non-local, block_id=1, offset=0) / Address=#x1\r\nptr %c = pointer(non-local, block_id=0, offset=-8) / Address=#x8\r\n\r\nSource:\r\ni1 %cmp1 = #x1 (1)\r\nptr %umin1 = pointer(non-local, block_id=0, offset=1) / Address=#x1\r\ni1 %cmp2 = #x0 (0)\r\nptr %umin2 = pointer(non-local, block_id=0, offset=-8) / Address=#x8\r\ni1 %cmp3 = #x0 (0)\r\nptr %umin3 = pointer(non-local, block_id=0, offset=1) / Address=#x1\r\ni1 %cmp4 = #x1 (1)\r\nptr %res = pointer(non-local, block_id=0, offset=1) / Address=#x1\r\n\r\nSOURCE MEMORY STATE\r\n===================\r\nNON-LOCAL BLOCKS:\r\nBlock 0 >\tsize: 0\talign: 1\talloc type: 0\talive: false\taddress: 0\r\nBlock 1 >\tsize: 1\talign: 1\talloc type: 0\talive: true\taddress: 1\r\nBlock 2 >\tsize: 1\talign: 1\talloc type: 0\talive: true\taddress: 2\r\nBlock 3 >\tsize: 3\talign: 1\talloc type: 0\talive: true\taddress: 8\r\n\r\nTarget:\r\ni1 %cmp2 = #x0 (0)\r\nptr %umin2 = pointer(non-local, block_id=0, offset=-8) / Address=#x8\r\ni1 %cmp3 = #x0 (0)\r\nptr %umin3 = pointer(non-local, block_id=0, offset=1) / Address=#x1\r\ni1 %cmp4 = #x1 (1)\r\nptr %res = pointer(non-local, block_id=1, offset=0) / Address=#x1\r\nSource value: pointer(non-local, block_id=0, offset=1) / Address=#x1\r\nTarget value: pointer(non-local, block_id=1, offset=0) / Address=#x1\r\n\r\nSummary:\r\n 0 correct transformations\r\n 1 incorrect transformations\r\n 0 failed-to-prove transformations\r\n 0 Alive2 errors\r\n```", "author": "bongjunj", "labels": ["miscompilation", "llvm:analysis"], "comments": []}, "verified": true} {"bug_id": "116553", "issue_url": "https://github.com/llvm/llvm-project/issues/116553", "bug_type": "miscompilation", "base_commit": "c25e09e238c6f872a116d10bbefba0beff145a57", "knowledge_cutoff": "2024-11-17T18:49:16Z", "lit_test_dir": ["llvm/test/Transforms/ConstraintElimination"], "hints": {"fix_commit": "52361d0368b79841be12156bf03cf8c1851e5df7", "components": ["ConstraintElimination"], "bug_location_lineno": {"llvm/lib/Transforms/Scalar/ConstraintElimination.cpp": [[1034, 1042], [1044, 1051]]}, "bug_location_funcname": {"llvm/lib/Transforms/Scalar/ConstraintElimination.cpp": ["State::addInfoForInductions"]}}, "patch": "commit 52361d0368b79841be12156bf03cf8c1851e5df7\nAuthor: Yingwei Zheng \nDate: Mon Nov 18 23:41:04 2024 +0800\n\n [ConstraintElim] Bail out on non-dedicated exits when adding exiting conditions (#116627)\n \n This patch bails out non-dedicated exits to avoid adding exiting\n conditions to invalid context.\n Closes https://github.com/llvm/llvm-project/issues/116553.\n\ndiff --git a/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp b/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\nindex d2e532d29d07..64433cb9216c 100644\n--- a/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\n+++ b/llvm/lib/Transforms/Scalar/ConstraintElimination.cpp\n@@ -1034,9 +1034,9 @@ void State::addInfoForInductions(BasicBlock &BB) {\n DTN, CmpInst::ICMP_SLT, PN, B,\n ConditionTy(CmpInst::ICMP_SLE, StartValue, B)));\n \n- // Try to add condition from header to the exit blocks. When exiting either\n- // with EQ or NE in the header, we know that the induction value must be u<=\n- // B, as other exits may only exit earlier.\n+ // Try to add condition from header to the dedicated exit blocks. When exiting\n+ // either with EQ or NE in the header, we know that the induction value must\n+ // be u<= B, as other exits may only exit earlier.\n assert(!StepOffset.isNegative() && \"induction must be increasing\");\n assert((Pred == CmpInst::ICMP_EQ || Pred == CmpInst::ICMP_NE) &&\n \"unsupported predicate\");\n@@ -1044,8 +1044,11 @@ void State::addInfoForInductions(BasicBlock &BB) {\n SmallVector ExitBBs;\n L->getExitBlocks(ExitBBs);\n for (BasicBlock *EB : ExitBBs) {\n- WorkList.emplace_back(FactOrCheck::getConditionFact(\n- DT.getNode(EB), CmpInst::ICMP_ULE, A, B, Precond));\n+ // Bail out on non-dedicated exits.\n+ if (DT.dominates(&BB, EB)) {\n+ WorkList.emplace_back(FactOrCheck::getConditionFact(\n+ DT.getNode(EB), CmpInst::ICMP_ULE, A, B, Precond));\n+ }\n }\n }\n \n", "tests": [{"file": "llvm/test/Transforms/ConstraintElimination/induction-condition-in-loop-exit.ll", "commands": ["opt -p constraint-elimination -S %s"], "tests": [{"test_name": "test_non_dedicated_exit", "test_body": "define i1 @test_non_dedicated_exit(i16 %n) {\nentry:\n %cond = icmp slt i16 %n, 1\n br i1 %cond, label %exit, label %loop.preheader\n\nloop.preheader: ; preds = %entry\n %sub = add nsw i16 %n, -1\n %ext = zext nneg i16 %sub to i32\n br label %loop\n\nloop: ; preds = %loop.latch, %loop.preheader\n %indvar = phi i32 [ %indvar.inc, %loop.latch ], [ 0, %loop.preheader ]\n %exitcond = icmp eq i32 %indvar, %ext\n br i1 %exitcond, label %exit, label %loop.latch\n\nloop.latch: ; preds = %loop\n %indvar.inc = add nuw nsw i32 %indvar, 1\n br label %loop\n\nexit: ; preds = %loop, %entry\n %cmp = icmp sgt i16 %n, 0\n ret i1 %cmp\n}\n"}]}], "issue": {"title": "[clang] Miscompile at O3 with SIGKILL", "body": "When I compiled this code with O3, it triggered SIGKILL. With O0/1/2, it returned 0.\r\n\r\nBisected to https://github.com/llvm/llvm-project/commit/5b927130b0e15a442a6ed171f43a612e6a40bbcd, which was committed by @fhahn \r\n\r\nCompiler Explorer: https://godbolt.org/z/q9j75feYG\r\n\r\nCode:\r\n```c\r\nint printf(const char *, ...);\r\nint a, c, d;\r\nint *b;\r\nchar e;\r\nshort f;\r\nint g(int i) {\r\n if (i <= 0)\r\n return 1;\r\n int h = 0;\r\n for (; h < i - 1; h++) {\r\n for (; d - h - 1;)\r\n if (c)\r\n b[d] = e = 1;\r\n if (e)\r\n break;\r\n }\r\n return 0;\r\n}\r\nvoid j() { g(f); }\r\nint main(int i, char *k[]) {\r\n j();\r\n while (0 < f)\r\n ;\r\n printf(\"%X\\n\", a);\r\n}\r\n```", "author": "cardigan1008", "labels": ["miscompilation", "confirmed", "llvm:transforms"], "comments": [{"author": "dtcxzyw", "body": "Reproducer: https://alive2.llvm.org/ce/z/fLN5gw\r\n```\r\n; bin/opt -passes=constraint-elimination test.ll -S\r\ndefine i1 @test(i16 %n) {\r\nentry:\r\n %cond = icmp slt i16 %n, 1\r\n br i1 %cond, label %exit, label %loop.preheader\r\n\r\nloop.preheader:\r\n %sub = add nsw i16 %n, -1\r\n %ext = zext nneg i16 %sub to i32\r\n br label %loop\r\n\r\nloop:\r\n %indvar = phi i32 [ %indvar.inc, %loop.latch ], [ 0, %loop.preheader ]\r\n %12 = icmp eq i32 %indvar, %ext\r\n br i1 %12, label %exit, label %loop.latch\r\n\r\nloop.latch:\r\n %indvar.inc = add nuw nsw i32 %indvar, 1\r\n br label %loop\r\n\r\nexit:\r\n %cmp = icmp sgt i16 %n, 0\r\n ret i1 %cmp\r\n}\r\n\r\n```\r\nI will post a fix later.\r\n"}]}, "verified": true} {"bug_id": "119393", "issue_url": "https://github.com/llvm/llvm-project/issues/119393", "bug_type": "crash", "base_commit": "13539c219c49c581d6ba5270d09b19da9b1e3975", "knowledge_cutoff": "2024-12-10T15:25:03Z", "lit_test_dir": ["llvm/test/Transforms/SLPVectorizer"], "hints": {"fix_commit": "a42aa8f265372a15d31a2af10b548fc239ed4b51", "components": ["SLPVectorizer"], "bug_location_lineno": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": [[13222, 13230]]}, "bug_location_funcname": {"llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp": ["BoUpSLP::isGatherShuffledSingleRegisterEntry"]}}, "patch": "commit a42aa8f265372a15d31a2af10b548fc239ed4b51\nAuthor: Alexey Bataev \nDate: Tue Dec 10 09:19:48 2024 -0800\n\n [SLP]Fix adjusting of the mask for the fully matched nodes.\n \n When checking for the poison elements in the matches node, need to\n consider the register number, when clearing the corresponding mask\n element.\n \n Fixes #119393\n\ndiff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\nindex c729c6cc9195..d90d7cc8f29e 100644\n--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp\n@@ -13222,9 +13222,9 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(\n copy(CommonMask, Mask.begin());\n }\n // Clear undef scalars.\n- for (int I = 0, Sz = VL.size(); I < Sz; ++I)\n+ for (unsigned I : seq(VL.size()))\n if (isa(VL[I]))\n- Mask[I] = PoisonMaskElem;\n+ Mask[Part * VL.size() + I] = PoisonMaskElem;\n return TargetTransformInfo::SK_PermuteSingleSrc;\n }\n // No perfect match, just shuffle, so choose the first tree node from the\n", "tests": [{"file": "llvm/test/Transforms/SLPVectorizer/X86/full-match-with-poison-scalar.ll", "commands": ["opt -S --passes=slp-vectorizer -slp-threshold=-200 -mtriple=x86_64-unknown-linux-gnu < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @test() {\nentry:\n br label %func_135.exit.i\n\nfunc_135.exit.i: ; preds = %entry\n %g_228.promoted166.i1105.i = phi i32 [ 0, %entry ]\n %cmp55.i.i1199.i = icmp ugt i32 %g_228.promoted166.i1105.i, 0\n %cmp55.1.i.i1200.i = icmp ugt i32 %g_228.promoted166.i1105.i, 0\n %cmp55.2.i.i1201.i = icmp ugt i32 %g_228.promoted166.i1105.i, 0\n %cmp55.3.i.i1202.i = icmp ugt i32 %g_228.promoted166.i1105.i, 0\n %0 = xor i1 %cmp55.2.i.i1201.i, %cmp55.3.i.i1202.i\n %1 = xor i1 %cmp55.1.i.i1200.i, %0\n %2 = xor i1 %cmp55.i.i1199.i, %1\n %dec.i.i1204.i = add i32 %g_228.promoted166.i1105.i, 0\n %cmp55.i.i1199.1.i = icmp ugt i32 %dec.i.i1204.i, 0\n %cmp55.1.i.i1200.1.i = icmp ugt i32 %dec.i.i1204.i, 0\n %cmp55.2.i.i1201.1.i = icmp ugt i32 %dec.i.i1204.i, 0\n %cmp55.3.i.i1202.1.i = icmp ugt i32 %dec.i.i1204.i, 0\n %3 = xor i1 %cmp55.2.i.i1201.1.i, %cmp55.3.i.i1202.1.i\n %4 = xor i1 %cmp55.1.i.i1200.1.i, %3\n %5 = xor i1 %cmp55.i.i1199.1.i, %4\n %6 = xor i1 %2, %5\n %dec.i.i1204.1.i = add i32 %g_228.promoted166.i1105.i, 0\n %cmp55.i.i1199.2.i = icmp ugt i32 %dec.i.i1204.1.i, 0\n %cmp55.1.i.i1200.2.i = icmp ugt i32 %dec.i.i1204.1.i, 0\n %cmp55.2.i.i1201.2.i = icmp ugt i32 %dec.i.i1204.1.i, 0\n %cmp55.3.i.i1202.2.i = icmp ugt i32 %dec.i.i1204.1.i, 0\n %7 = add i32 0, 0\n %8 = icmp ult i32 %7, 0\n %9 = xor i1 %cmp55.3.i.i1202.2.i, %8\n %10 = xor i1 %cmp55.2.i.i1201.2.i, %9\n %11 = xor i1 %cmp55.1.i.i1200.2.i, %10\n %12 = xor i1 %cmp55.i.i1199.2.i, %11\n %13 = xor i1 %12, %6\n %dec.i.i1204.2.i = add i32 %g_228.promoted166.i1105.i, 0\n %cmp55.i.i1199.3.i = icmp ugt i32 %dec.i.i1204.2.i, 0\n %cmp55.1.i.i1200.3.i = icmp ugt i32 %dec.i.i1204.2.i, 0\n %cmp55.2.i.i1201.3.i = icmp ugt i32 %dec.i.i1204.2.i, 0\n %14 = xor i1 %cmp55.1.i.i1200.3.i, %cmp55.2.i.i1201.3.i\n %15 = xor i1 %cmp55.i.i1199.3.i, %14\n %16 = xor i1 %15, %13\n %17 = zext i1 %16 to i16\n br label %if.then.sink.split.i\n\nfor.cond7.preheader.i1132.1.i: ; No predecessors!\n br label %if.then.sink.split.i\n\nif.then.sink.split.i: ; preds = %for.cond7.preheader.i1132.1.i, %func_135.exit.i\n %xor58.5.i.i1203.3.3.sink.i = phi i16 [ 0, %for.cond7.preheader.i1132.1.i ], [ %17, %func_135.exit.i ]\n ret i32 0\n}\n"}]}], "issue": {"title": "[SLPVectorizer] Assertion `I1 == PoisonMaskElem && \"Expected unused subvectors mask\"' failed.", "body": "Reproducer: https://godbolt.org/z/3dahr6Mb4\n```\n; bin/opt -passes=slp-vectorizer reduced.ll -S\ntarget datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-unknown-linux-gnu\"\n\ndefine i32 @main() {\nentry:\n br label %func_135.exit.i\n\nfunc_135.exit.i:\n %g_228.promoted166.i1105.i = phi i32 [ 0, %entry ]\n %cmp55.i.i1199.i = icmp ugt i32 %g_228.promoted166.i1105.i, 0\n %cmp55.1.i.i1200.i = icmp ugt i32 %g_228.promoted166.i1105.i, 0\n %cmp55.2.i.i1201.i = icmp ugt i32 %g_228.promoted166.i1105.i, 0\n %cmp55.3.i.i1202.i = icmp ugt i32 %g_228.promoted166.i1105.i, 0\n %0 = xor i1 %cmp55.2.i.i1201.i, %cmp55.3.i.i1202.i\n %1 = xor i1 %cmp55.1.i.i1200.i, %0\n %2 = xor i1 %cmp55.i.i1199.i, %1\n %dec.i.i1204.i = add i32 %g_228.promoted166.i1105.i, 0\n %cmp55.i.i1199.1.i = icmp ugt i32 %dec.i.i1204.i, 0\n %cmp55.1.i.i1200.1.i = icmp ugt i32 %dec.i.i1204.i, 0\n %cmp55.2.i.i1201.1.i = icmp ugt i32 %dec.i.i1204.i, 0\n %cmp55.3.i.i1202.1.i = icmp ugt i32 %dec.i.i1204.i, 0\n %3 = xor i1 %cmp55.2.i.i1201.1.i, %cmp55.3.i.i1202.1.i\n %4 = xor i1 %cmp55.1.i.i1200.1.i, %3\n %5 = xor i1 %cmp55.i.i1199.1.i, %4\n %6 = xor i1 %2, %5\n %dec.i.i1204.1.i = add i32 %g_228.promoted166.i1105.i, 0\n %cmp55.i.i1199.2.i = icmp ugt i32 %dec.i.i1204.1.i, 0\n %cmp55.1.i.i1200.2.i = icmp ugt i32 %dec.i.i1204.1.i, 0\n %cmp55.2.i.i1201.2.i = icmp ugt i32 %dec.i.i1204.1.i, 0\n %cmp55.3.i.i1202.2.i = icmp ugt i32 %dec.i.i1204.1.i, 0\n %7 = add i32 0, 0\n %8 = icmp ult i32 %7, 0\n %9 = xor i1 %cmp55.3.i.i1202.2.i, %8\n %10 = xor i1 %cmp55.2.i.i1201.2.i, %9\n %11 = xor i1 %cmp55.1.i.i1200.2.i, %10\n %12 = xor i1 %cmp55.i.i1199.2.i, %11\n %13 = xor i1 %12, %6\n %dec.i.i1204.2.i = add i32 %g_228.promoted166.i1105.i, 0\n %cmp55.i.i1199.3.i = icmp ugt i32 %dec.i.i1204.2.i, 0\n %cmp55.1.i.i1200.3.i = icmp ugt i32 %dec.i.i1204.2.i, 0\n %cmp55.2.i.i1201.3.i = icmp ugt i32 %dec.i.i1204.2.i, 0\n %14 = xor i1 %cmp55.1.i.i1200.3.i, %cmp55.2.i.i1201.3.i\n %15 = xor i1 %cmp55.i.i1199.3.i, %14\n %16 = xor i1 %15, %13\n %17 = zext i1 %16 to i16\n br label %if.then.sink.split.i\n\nfor.cond7.preheader.i1132.1.i: ; No predecessors!\n br label %if.then.sink.split.i\n\nif.then.sink.split.i: ; preds = %for.cond7.preheader.i1132.1.i, %func_135.exit.i\n %xor58.5.i.i1203.3.3.sink.i = phi i16 [ 0, %for.cond7.preheader.i1132.1.i ], [ %17, %func_135.exit.i ]\n ret i32 0\n}\n```\n```\nopt: /root/llvm-project/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:10913: llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::finalize(llvm::ArrayRef, llvm::ArrayRef >, llvm::ArrayRef, unsigned int, llvm::function_ref&)>): Assertion `I1 == PoisonMaskElem && \"Expected unused subvectors mask\"' failed.\nPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.\nStack dump:\n0.\tProgram arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/opt -o /app/output.s -S -passes=slp-vectorizer \n1.\tRunning pass \"function(slp-vectorizer)\" on module \"\"\n2.\tRunning pass \"slp-vectorizer\" on function \"main\"\n #0 0x00000000051c9e58 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x51c9e58)\n #1 0x00000000051c785c SignalHandler(int) Signals.cpp:0:0\n #2 0x000078cea7e42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)\n #3 0x000078cea7e969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)\n #4 0x000078cea7e42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)\n #5 0x000078cea7e287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)\n #6 0x000078cea7e2871b (/lib/x86_64-linux-gnu/libc.so.6+0x2871b)\n #7 0x000078cea7e39e96 (/lib/x86_64-linux-gnu/libc.so.6+0x39e96)\n #8 0x00000000041fda54 llvm::slpvectorizer::BoUpSLP::ShuffleCostEstimator::finalize(llvm::ArrayRef, llvm::ArrayRef>, llvm::ArrayRef, unsigned int, llvm::function_ref&)>) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x41fda54)\n #9 0x00000000042390b0 llvm::InstructionCost llvm::slpvectorizer::BoUpSLP::processBuildVector, llvm::slpvectorizer::BoUpSLP, llvm::SmallPtrSetImpl>(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::Type*, llvm::TargetTransformInfo&, llvm::ArrayRef&, llvm::slpvectorizer::BoUpSLP&, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42390b0)\n#10 0x0000000004243fa7 llvm::slpvectorizer::BoUpSLP::getEntryCost(llvm::slpvectorizer::BoUpSLP::TreeEntry const*, llvm::ArrayRef, llvm::SmallPtrSetImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4243fa7)\n#11 0x0000000004277f91 llvm::slpvectorizer::BoUpSLP::getTreeCost(llvm::ArrayRef) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4277f91)\n#12 0x000000000427da60 (anonymous namespace)::HorizontalReduction::tryToReduce(llvm::slpvectorizer::BoUpSLP&, llvm::DataLayout const&, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo const&) SLPVectorizer.cpp:0:0\n#13 0x0000000004280ea9 llvm::SLPVectorizerPass::vectorizeHorReduction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&, llvm::SmallVectorImpl&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4280ea9)\n#14 0x000000000428563b llvm::SLPVectorizerPass::vectorizeRootInstruction(llvm::PHINode*, llvm::Instruction*, llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (.constprop.0) SLPVectorizer.cpp:0:0\n#15 0x0000000004289c9f llvm::SLPVectorizerPass::vectorizeChainsInBlock(llvm::BasicBlock*, llvm::slpvectorizer::BoUpSLP&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4289c9f)\n#16 0x0000000004290786 llvm::SLPVectorizerPass::runImpl(llvm::Function&, llvm::ScalarEvolution*, llvm::TargetTransformInfo*, llvm::TargetLibraryInfo*, llvm::AAResults*, llvm::LoopInfo*, llvm::DominatorTree*, llvm::AssumptionCache*, llvm::DemandedBits*, llvm::OptimizationRemarkEmitter*) (.part.0) SLPVectorizer.cpp:0:0\n#17 0x00000000042912cb llvm::SLPVectorizerPass::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x42912cb)\n#18 0x0000000002fdbfee llvm::detail::PassModel>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x2fdbfee)\n#19 0x0000000004fc8d00 llvm::PassManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fc8d00)\n#20 0x0000000000e3b51e llvm::detail::PassModel>, llvm::AnalysisManager>::run(llvm::Function&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3b51e)\n#21 0x0000000004fc761b llvm::ModuleToFunctionPassAdaptor::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fc761b)\n#22 0x0000000000e3abae llvm::detail::PassModel>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0xe3abae)\n#23 0x0000000004fc7020 llvm::PassManager>::run(llvm::Module&, llvm::AnalysisManager&) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x4fc7020)\n#24 0x00000000009177ca llvm::runPassPipeline(llvm::StringRef, llvm::Module&, llvm::TargetMachine*, llvm::TargetLibraryInfoImpl*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::ToolOutputFile*, llvm::StringRef, llvm::ArrayRef, llvm::ArrayRef>, llvm::opt_tool::OutputKind, llvm::opt_tool::VerifierKind, bool, bool, bool, bool, bool, bool, bool) (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x9177ca)\n#25 0x0000000000909e49 optMain (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x909e49)\n#26 0x000078cea7e29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)\n#27 0x000078cea7e29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)\n#28 0x0000000000901705 _start (/opt/compiler-explorer/clang-assertions-trunk/bin/opt+0x901705)\nProgram terminated with signal: SIGSEGV\nCompiler returned: 139\n```", "author": "dtcxzyw", "labels": ["llvm:SLPVectorizer", "crash-on-valid"], "comments": []}, "verified": true} {"bug_id": "111934", "issue_url": "https://github.com/llvm/llvm-project/issues/111934", "bug_type": "miscompilation", "base_commit": "f74f568b29885c3fa63c44e33f91f3bb7281138e", "knowledge_cutoff": "2024-10-11T03:29:55Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "6a65e98fa7901dc1de91172d065fafb16ce89d77", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": [[955, 963], [971, 977], [980, 986], [3375, 3381]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": ["InstCombinerImpl::foldAndOrOfICmps", "foldIsPowerOf2"]}}, "patch": "commit 6a65e98fa7901dc1de91172d065fafb16ce89d77\nAuthor: Yingwei Zheng \nDate: Fri Oct 11 18:19:21 2024 +0800\n\n [InstCombine] Drop range attributes in `foldIsPowerOf2` (#111946)\n \n Fixes https://github.com/llvm/llvm-project/issues/111934.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\nindex 688601a8ffa5..964616a4eb35 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n@@ -955,9 +955,11 @@ static Value *foldIsPowerOf2OrZero(ICmpInst *Cmp0, ICmpInst *Cmp1, bool IsAnd,\n }\n \n /// Reduce a pair of compares that check if a value has exactly 1 bit set.\n-/// Also used for logical and/or, must be poison safe.\n+/// Also used for logical and/or, must be poison safe if range attributes are\n+/// dropped.\n static Value *foldIsPowerOf2(ICmpInst *Cmp0, ICmpInst *Cmp1, bool JoinedByAnd,\n- InstCombiner::BuilderTy &Builder) {\n+ InstCombiner::BuilderTy &Builder,\n+ InstCombinerImpl &IC) {\n // Handle 'and' / 'or' commutation: make the equality check the first operand.\n if (JoinedByAnd && Cmp1->getPredicate() == ICmpInst::ICMP_NE)\n std::swap(Cmp0, Cmp1);\n@@ -971,7 +973,10 @@ static Value *foldIsPowerOf2(ICmpInst *Cmp0, ICmpInst *Cmp1, bool JoinedByAnd,\n match(Cmp1, m_SpecificICmp(ICmpInst::ICMP_ULT,\n m_Intrinsic(m_Specific(X)),\n m_SpecificInt(2)))) {\n- Value *CtPop = Cmp1->getOperand(0);\n+ auto *CtPop = cast(Cmp1->getOperand(0));\n+ // Drop range attributes and re-infer them in the next iteration.\n+ CtPop->dropPoisonGeneratingAnnotations();\n+ IC.addToWorklist(CtPop);\n return Builder.CreateICmpEQ(CtPop, ConstantInt::get(CtPop->getType(), 1));\n }\n // (X == 0) || (ctpop(X) u> 1) --> ctpop(X) != 1\n@@ -980,7 +985,10 @@ static Value *foldIsPowerOf2(ICmpInst *Cmp0, ICmpInst *Cmp1, bool JoinedByAnd,\n match(Cmp1, m_SpecificICmp(ICmpInst::ICMP_UGT,\n m_Intrinsic(m_Specific(X)),\n m_SpecificInt(1)))) {\n- Value *CtPop = Cmp1->getOperand(0);\n+ auto *CtPop = cast(Cmp1->getOperand(0));\n+ // Drop range attributes and re-infer them in the next iteration.\n+ CtPop->dropPoisonGeneratingAnnotations();\n+ IC.addToWorklist(CtPop);\n return Builder.CreateICmpNE(CtPop, ConstantInt::get(CtPop->getType(), 1));\n }\n return nullptr;\n@@ -3375,7 +3383,7 @@ Value *InstCombinerImpl::foldAndOrOfICmps(ICmpInst *LHS, ICmpInst *RHS,\n if (Value *V = foldSignedTruncationCheck(LHS, RHS, I, Builder))\n return V;\n \n- if (Value *V = foldIsPowerOf2(LHS, RHS, IsAnd, Builder))\n+ if (Value *V = foldIsPowerOf2(LHS, RHS, IsAnd, Builder, *this))\n return V;\n \n if (Value *V = foldPowerOf2AndShiftedMask(LHS, RHS, IsAnd, Builder))\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/ispow2.ll", "commands": ["opt < %s -passes=instcombine -S"], "tests": [{"test_name": "has_single_bit", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @has_single_bit(i32 %x) {\nentry:\n %cmp1 = icmp ne i32 %x, 0\n %popcnt = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)\n %cmp2 = icmp ult i32 %popcnt, 2\n %sel = select i1 %cmp1, i1 %cmp2, i1 false\n ret i1 %sel\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "has_single_bit_inv", "test_body": "; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare i32 @llvm.ctpop.i32(i32) #0\n\ndefine i1 @has_single_bit_inv(i32 %x) {\nentry:\n %cmp1 = icmp eq i32 %x, 0\n %popcnt = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)\n %cmp2 = icmp ugt i32 %popcnt, 1\n %sel = select i1 %cmp1, i1 true, i1 %cmp2\n ret i1 %sel\n}\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}, {"test_name": "not_pow2_or_z_known_bits_fail_wrong_cmp", "test_body": "define <2 x i1> @not_pow2_or_z_known_bits_fail_wrong_cmp(<2 x i32> %xin) {\n %x = or <2 x i32> %xin, splat (i32 64)\n %cnt = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %x)\n %r = icmp ugt <2 x i32> %cnt, splat (i32 2)\n ret <2 x i1> %r\n}\n\n; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)\ndeclare <2 x i32> @llvm.ctpop.v2i32(<2 x i32>) #0\n\nattributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }\n"}]}], "issue": {"title": "[InstCombine] poison-generating attributes are not dropped when folding logical and/or of icmps", "body": "Reproducer: https://alive2.llvm.org/ce/z/rSUfvo\r\n```\r\ndefine i1 @src(i32 noundef %Value) {\r\nentry:\r\n %cmp.not = icmp eq i32 %Value, 0\r\n %0 = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %Value)\r\n %cmp1 = icmp ult i32 %0, 2\r\n %1 = select i1 %cmp.not, i1 false, i1 %cmp1\r\n ret i1 %1\r\n}\r\n\r\ndefine i1 @tgt(i32 noundef %Value) {\r\nentry:\r\n %0 = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %Value)\r\n %1 = icmp eq i32 %0, 1\r\n ret i1 %1\r\n}\r\n```\r\nWhen `%Value == 0`, `@src` returns `false` but `@tgt` returns `poison`.\r\nThis miscompilation was triggered after https://github.com/llvm/llvm-project/pull/111284 was landed.\r\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "120361", "issue_url": "https://github.com/llvm/llvm-project/issues/120361", "bug_type": "miscompilation", "base_commit": "c6967efe780d6cc5d70fc8cadbd227353b6768f1", "knowledge_cutoff": "2024-12-18T05:27:30Z", "lit_test_dir": ["llvm/test/Transforms/InstCombine"], "hints": {"fix_commit": "6f68010f9123aae9f6f105d7a11af22458518ad7", "components": ["InstCombine"], "bug_location_lineno": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": [[455, 468]]}, "bug_location_funcname": {"llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp": ["foldLogOpOfMaskedICmps_NotAllZeros_BMask_Mixed"]}}, "patch": "commit 6f68010f9123aae9f6f105d7a11af22458518ad7\nAuthor: Yingwei Zheng \nDate: Wed Dec 18 20:40:33 2024 +0800\n\n [InstCombine] Drop samesign flags in `foldLogOpOfMaskedICmps_NotAllZeros_BMask_Mixed` (#120373)\n \n Counterexamples: https://alive2.llvm.org/ce/z/6Ks8Qz\n Closes https://github.com/llvm/llvm-project/issues/120361.\n\ndiff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\nindex dff9304be64d..e576eea4ca36 100644\n--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp\n@@ -455,14 +455,20 @@ static Value *foldLogOpOfMaskedICmps_NotAllZeros_BMask_Mixed(\n // RHS. For example,\n // (icmp ne (A & 255), 0) & (icmp eq (A & 15), 8) -> (icmp eq (A & 15), 8).\n // (icmp ne (A & 15), 0) & (icmp eq (A & 15), 8) -> (icmp eq (A & 15), 8).\n- if (IsSuperSetOrEqual(BCst, DCst))\n+ if (IsSuperSetOrEqual(BCst, DCst)) {\n+ // We can't guarantee that samesign hold after this fold.\n+ RHS->setSameSign(false);\n return RHS;\n+ }\n // Otherwise, B is a subset of D. If B and E have a common bit set,\n // ie. (B & E) != 0, then LHS is subsumed by RHS. For example.\n // (icmp ne (A & 12), 0) & (icmp eq (A & 15), 8) -> (icmp eq (A & 15), 8).\n assert(IsSubSetOrEqual(BCst, DCst) && \"Precondition due to above code\");\n- if ((*BCst & ECst) != 0)\n+ if ((*BCst & ECst) != 0) {\n+ // We can't guarantee that samesign hold after this fold.\n+ RHS->setSameSign(false);\n return RHS;\n+ }\n // Otherwise, LHS and RHS contradict and the whole expression becomes false\n // (or true if negated.) For example,\n // (icmp ne (A & 7), 0) & (icmp eq (A & 15), 8) -> false.\n", "tests": [{"file": "llvm/test/Transforms/InstCombine/icmp-logical.ll", "commands": ["opt -passes=instcombine -S -o - %s"], "tests": [{"test_name": "pr120361", "test_body": "define i1 @pr120361(i8 %x, i8 %y) {\n %cmp1 = icmp samesign eq i8 %x, -1\n %cmp2 = icmp ne i8 %x, 0\n %result = select i1 %cmp2, i1 %cmp1, i1 false\n ret i1 %result\n}\n"}, {"test_name": "pr120361_v2", "test_body": "define i1 @pr120361_v2(i32 %x) {\n %and1 = and i32 %x, 15\n %cmp1 = icmp ne i32 %and1, 0\n %and2 = and i32 %x, -113\n %cmp2 = icmp samesign eq i32 %and2, 15\n %and = select i1 %cmp1, i1 %cmp2, i1 false\n ret i1 %and\n}\n"}, {"test_name": "masked_icmps_bmask_notmixed_not_subset_notoptimized", "test_body": "define i1 @masked_icmps_bmask_notmixed_not_subset_notoptimized(i32 %A) {\n %mask1 = and i32 %A, 254\n %tst1 = icmp ne i32 %mask1, 252\n %mask2 = and i32 %A, 253\n %tst2 = icmp ne i32 %mask2, 252\n %res = and i1 %tst1, %tst2\n ret i1 %res\n}\n"}]}], "issue": {"title": "[InstCombine] `samesign` should be dropped when folding logical and of icmps", "body": "Reproducer: https://alive2.llvm.org/ce/z/YSW_NC\n```\ndefine i1 @src(i8 %0, i8 %1) {\nstart:\n %2 = icmp samesign eq i8 %0, -1\n %3 = icmp ne i8 %0, 0\n %4 = icmp ugt i8 %0, %0\n %result = select i1 %3, i1 %2, i1 %4\n ret i1 %result\n}\n\ndefine i1 @tgt(i8 %0, i8 %1) {\nstart:\n %2 = icmp samesign eq i8 %0, -1\n ret i1 %2\n}\n```\n```\n\n----------------------------------------\ndefine i1 @src(i8 %#0, i8 %#1) {\nstart:\n %#2 = icmp samesign eq i8 %#0, 255\n %#3 = icmp ne i8 %#0, 0\n %#4 = icmp ugt i8 %#0, %#0\n %result = select i1 %#3, i1 %#2, i1 %#4\n ret i1 %result\n}\n=>\ndefine i1 @tgt(i8 %#0, i8 %#1) {\nstart:\n %#2 = icmp samesign eq i8 %#0, 255\n ret i1 %#2\n}\nTransformation doesn't verify!\n\nERROR: Target is more poisonous than source\n\nExample:\ni8 %#0 = #x00 (0)\ni8 %#1 = poison\n\nSource:\ni1 %#2 = poison\ni1 %#3 = #x0 (0)\ni1 %#4 = #x0 (0)\ni1 %result = #x0 (0)\n\nTarget:\ni1 %#2 = poison\nSource value: #x0 (0)\nTarget value: poison\n```\nReported by https://github.com/dtcxzyw/llvm-mutation-based-fuzz-service/issues/14 while reviewing https://github.com/llvm/llvm-project/pull/120177.\n\n\n", "author": "dtcxzyw", "labels": ["miscompilation", "llvm:instcombine"], "comments": []}, "verified": true} {"bug_id": "123550", "issue_url": "https://github.com/llvm/llvm-project/issues/123550", "bug_type": "miscompilation", "base_commit": "a3aa452a21f983237873fa85c866b9f0224789bd", "knowledge_cutoff": "2025-01-20T06:37:58Z", "lit_test_dir": ["llvm/test/Analysis/ScalarEvolution"], "hints": {"fix_commit": "07efe2c18a63423943a4f9d9daeada23601f84c8", "components": ["ScalarEvolution"], "bug_location_lineno": {"llvm/lib/Analysis/ScalarEvolution.cpp": [[5917, 5936]]}, "bug_location_funcname": {"llvm/lib/Analysis/ScalarEvolution.cpp": ["ScalarEvolution::createAddRecFromPHI"]}}, "patch": "commit 07efe2c18a63423943a4f9d9daeada23601f84c8\nAuthor: Nikita Popov \nDate: Wed Jan 29 09:09:14 2025 +0100\n\n [SCEV] Check correct value for UB (#124302)\n \n This is a followup to #117152. That patch introduced a check for\n UB/poison on BEValue. However, the SCEV we're actually going to use is\n Shifted. In some cases, it's possible for Shifted to contain UB, while\n BEValue doesn't.\n \n In the test case the values are:\n \n BEValue: (-1 * (zext i8 (-83 + ((-83 /u {1,+,1}<%loop>) *\n {-1,+,-1}<%loop>)) to i32))\n Shifted: (-173 + (-1 * (zext i8 ((-83 /u {0,+,1}<%loop>) *\n {0,+,-1}<%loop>) to i32)))\n \n Fixes https://github.com/llvm/llvm-project/issues/123550.\n\ndiff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp\nindex 7d7d37b3d228..2ce40877b523 100644\n--- a/llvm/lib/Analysis/ScalarEvolution.cpp\n+++ b/llvm/lib/Analysis/ScalarEvolution.cpp\n@@ -5917,20 +5917,18 @@ const SCEV *ScalarEvolution::createAddRecFromPHI(PHINode *PN) {\n // PHI(f(0), f({1,+,1})) --> f({0,+,1})\n \n // Do not allow refinement in rewriting of BEValue.\n- if (isGuaranteedNotToCauseUB(BEValue)) {\n- const SCEV *Shifted = SCEVShiftRewriter::rewrite(BEValue, L, *this);\n- const SCEV *Start = SCEVInitRewriter::rewrite(Shifted, L, *this, false);\n- if (Shifted != getCouldNotCompute() && Start != getCouldNotCompute() &&\n- ::impliesPoison(BEValue, Start)) {\n- const SCEV *StartVal = getSCEV(StartValueV);\n- if (Start == StartVal) {\n- // Okay, for the entire analysis of this edge we assumed the PHI\n- // to be symbolic. We now need to go back and purge all of the\n- // entries for the scalars that use the symbolic expression.\n- forgetMemoizedResults(SymbolicName);\n- insertValueToMap(PN, Shifted);\n- return Shifted;\n- }\n+ const SCEV *Shifted = SCEVShiftRewriter::rewrite(BEValue, L, *this);\n+ const SCEV *Start = SCEVInitRewriter::rewrite(Shifted, L, *this, false);\n+ if (Shifted != getCouldNotCompute() && Start != getCouldNotCompute() &&\n+ isGuaranteedNotToCauseUB(Shifted) && ::impliesPoison(Shifted, Start)) {\n+ const SCEV *StartVal = getSCEV(StartValueV);\n+ if (Start == StartVal) {\n+ // Okay, for the entire analysis of this edge we assumed the PHI\n+ // to be symbolic. We now need to go back and purge all of the\n+ // entries for the scalars that use the symbolic expression.\n+ forgetMemoizedResults(SymbolicName);\n+ insertValueToMap(PN, Shifted);\n+ return Shifted;\n }\n }\n }\n", "tests": [{"file": "llvm/test/Analysis/ScalarEvolution/pr123550.ll", "commands": ["opt -S -passes=indvars < %s"], "tests": [{"test_name": "test", "test_body": "define i32 @src() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb4, %bb\n %phi = phi i32 [ -173, %bb ], [ %sub, %bb4 ]\n %phi2 = phi i32 [ 1, %bb ], [ %add, %bb4 ]\n %srem = srem i32 729259140, %phi\n %trunc = trunc i32 %phi2 to i8\n %urem = urem i8 -83, %trunc\n %zext = zext i8 %urem to i32\n %sub = sub nsw i32 0, %zext\n br label %bb4\n\nbb4: ; preds = %bb1\n %add = add nuw nsw i32 %phi2, 1\n %icmp = icmp eq i32 %sub, 0\n br i1 %icmp, label %bb5, label %bb1\n\nbb5: ; preds = %bb4\n %phi6 = phi i32 [ %srem, %bb4 ]\n ret i32 %phi6\n}\n"}]}], "issue": {"title": "[Indvars] Miscompile: SSA value incorrectly simplified", "body": "One can find the incorrect transformation here: https://godbolt.org/z/T9qMhT7rb\nAnd the proof of incorrect transformation here: https://alive2.llvm.org/ce/z/PXjb5J\n\nFor the record, I describe the problem here. Given the IR:\n```\ndefine i32 @widget() {\nbb:\n br label %bb1\n\nbb1: ; preds = %bb4, %bb\n %phi = phi i32 [ -173, %bb ], [ %sub, %bb4 ]\n %phi2 = phi i32 [ 1, %bb ], [ %add, %bb4 ]\n %srem = srem i32 729259140, %phi\n %trunc = trunc i32 %phi2 to i8\n %urem = urem i8 -83, %trunc\n %zext = zext i8 %urem to i32\n %sub = sub nsw i32 0, %zext\n br label %bb4\n\nbb4: ; preds = %bb1\n %add = add nuw nsw i32 %phi2, 1\n %icmp = icmp eq i32 %sub, 0\n br i1 %icmp, label %bb5, label %bb1\n\nbb5: ; preds = %bb4\n %phi6 = phi i32 [ %srem, %bb4 ]\n ret i32 %phi6\n}\n```\nwhich returns `srem i32 729259140, -173` which is `130`. The above IR under the `indvars` pass transforms to:\n```\ndefine i32 @widget() {\n br label %bb1\n\nbb1: ; preds = %bb4, %bb\n br label %bb4\n\nbb4: ; preds = %bb1\n br i1 true, label %bb5, label %bb1\n\nbb5: ; preds = %bb4\n ret i32 729259140\n}\n```\nwhich returns `729259140` which is the wrong answer. ", "author": "Nirhar", "labels": ["miscompilation", "release:backport", "llvm:SCEV"], "comments": [{"author": "nikic", "body": "Slightly cleaned up:\n```\ndefine i32 @test() {\nbb:\n br label %loop \n\nloop: \n %phi = phi i32 [ -173, %bb ], [ %sub, %loop ]\n %iv2 = phi i32 [ 1, %bb ], [ %iv2.inc, %loop ]\n %srem = srem i32 729259140, %phi\n %trunc = trunc i32 %iv2 to i8\n %urem = urem i8 -83, %trunc\n %zext = zext i8 %urem to i32\n %sub = sub i32 0, %zext \n %iv2.inc = add i32 %iv2, 1\n %icmp = icmp eq i32 %zext, 0\n br i1 %icmp, label %exit, label %loop\n\nexit:\n %srem.lcssa = phi i32 [ %srem, %loop ]\n ret i32 %srem.lcssa\n}\n```\n\nIncorrect exit value is computed by SCEV:\n```\n --> (729259140 + (-1 * (-173 + (-1 * (zext i8 ((-83 /u {0,+,1}<%loop>) * {0,+,-1}<%loop>) to i32))) * (729259140 /u (-173 + (-1 * (zext i8 ((-83 /u {0,+,1}<%loop>) * {0,+,-1}<%loop>) to i32)))))) U: empty-set S: empty-set\t\tExits: 729259140\t\tLoopDispositions: { %loop: Computable }\n```"}]}, "verified": true}