Spaces:
Sleeping
Sleeping
File size: 4,095 Bytes
181d94d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 |
int gpu_index = 0;
#ifdef GPU
#include "cuda.h"
#include "utils.h"
#include "blas.h"
#include <assert.h>
#include <stdlib.h>
#include <time.h>
void cuda_set_device(int n)
{
gpu_index = n;
cudaError_t status = cudaSetDevice(n);
check_error(status);
}
int cuda_get_device()
{
int n = 0;
cudaError_t status = cudaGetDevice(&n);
check_error(status);
return n;
}
void check_error(cudaError_t status)
{
//cudaDeviceSynchronize();
cudaError_t status2 = cudaGetLastError();
if (status != cudaSuccess)
{
const char *s = cudaGetErrorString(status);
char buffer[256];
printf("CUDA Error: %s\n", s);
assert(0);
snprintf(buffer, 256, "CUDA Error: %s", s);
error(buffer);
}
if (status2 != cudaSuccess)
{
const char *s = cudaGetErrorString(status);
char buffer[256];
printf("CUDA Error Prev: %s\n", s);
assert(0);
snprintf(buffer, 256, "CUDA Error Prev: %s", s);
error(buffer);
}
}
dim3 cuda_gridsize(size_t n){
size_t k = (n-1) / BLOCK + 1;
size_t x = k;
size_t y = 1;
if(x > 65535){
x = ceil(sqrt(k));
y = (n-1)/(x*BLOCK) + 1;
}
dim3 d = {x, y, 1};
//printf("%ld %ld %ld %ld\n", n, x, y, x*y*BLOCK);
return d;
}
#ifdef CUDNN
cudnnHandle_t cudnn_handle()
{
static int init[16] = {0};
static cudnnHandle_t handle[16];
int i = cuda_get_device();
if(!init[i]) {
cudnnCreate(&handle[i]);
init[i] = 1;
}
return handle[i];
}
#endif
cublasHandle_t blas_handle()
{
static int init[16] = {0};
static cublasHandle_t handle[16];
int i = cuda_get_device();
if(!init[i]) {
cublasCreate(&handle[i]);
init[i] = 1;
}
return handle[i];
}
float *cuda_make_array(float *x, size_t n)
{
float *x_gpu;
size_t size = sizeof(float)*n;
cudaError_t status = cudaMalloc((void **)&x_gpu, size);
check_error(status);
if(x){
status = cudaMemcpy(x_gpu, x, size, cudaMemcpyHostToDevice);
check_error(status);
} else {
fill_gpu(n, 0, x_gpu, 1);
}
if(!x_gpu) error("Cuda malloc failed\n");
return x_gpu;
}
void cuda_random(float *x_gpu, size_t n)
{
static curandGenerator_t gen[16];
static int init[16] = {0};
int i = cuda_get_device();
if(!init[i]){
curandCreateGenerator(&gen[i], CURAND_RNG_PSEUDO_DEFAULT);
curandSetPseudoRandomGeneratorSeed(gen[i], time(0));
init[i] = 1;
}
curandGenerateUniform(gen[i], x_gpu, n);
check_error(cudaPeekAtLastError());
}
float cuda_compare(float *x_gpu, float *x, size_t n, char *s)
{
float *tmp = calloc(n, sizeof(float));
cuda_pull_array(x_gpu, tmp, n);
//int i;
//for(i = 0; i < n; ++i) printf("%f %f\n", tmp[i], x[i]);
axpy_cpu(n, -1, x, 1, tmp, 1);
float err = dot_cpu(n, tmp, 1, tmp, 1);
printf("Error %s: %f\n", s, sqrt(err/n));
free(tmp);
return err;
}
int *cuda_make_int_array(int *x, size_t n)
{
int *x_gpu;
size_t size = sizeof(int)*n;
cudaError_t status = cudaMalloc((void **)&x_gpu, size);
check_error(status);
if(x){
status = cudaMemcpy(x_gpu, x, size, cudaMemcpyHostToDevice);
check_error(status);
}
if(!x_gpu) error("Cuda malloc failed\n");
return x_gpu;
}
void cuda_free(float *x_gpu)
{
cudaError_t status = cudaFree(x_gpu);
check_error(status);
}
void cuda_push_array(float *x_gpu, float *x, size_t n)
{
size_t size = sizeof(float)*n;
cudaError_t status = cudaMemcpy(x_gpu, x, size, cudaMemcpyHostToDevice);
check_error(status);
}
void cuda_pull_array(float *x_gpu, float *x, size_t n)
{
size_t size = sizeof(float)*n;
cudaError_t status = cudaMemcpy(x, x_gpu, size, cudaMemcpyDeviceToHost);
check_error(status);
}
float cuda_mag_array(float *x_gpu, size_t n)
{
float *temp = calloc(n, sizeof(float));
cuda_pull_array(x_gpu, temp, n);
float m = mag_array(temp, n);
free(temp);
return m;
}
#else
void cuda_set_device(int n){}
#endif
|