gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fixed CUDA benchmark

Browse Source
This commit is contained in:
Philippe
2015-02-05 23:42:31 -05:00
parent 58fdc5d18e
commit 7fc2348924
3 changed files with 13 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
bench/blas.cpp
View File

@@ -53,20 +53,20 @@ void bench(ad::numeric_type dtype)
{\
std::vector<long> times;\
double total_time = 0;\
double time;\
float time;\
cudaEvent_t start, stop;\
cudaEventCreate(&start);\
cudaEventCreate(&stop);\
while(total_time*1e-3 < 1e-1){\
cudaEvent_t start, stop;\
cudaEventCreate(&start);\
cudaEventCreate(&stop);\
cudaEventRecord(start);\
cudaEventRecord(start,0);\
OP;\
cudaEventRecord(stop);\
cudaEventSynchronize();\
cudaEventRecord(stop,0);\
cudaEventSynchronize(stop);\
cudaEventElapsedTime(&time, start, stop);\
times.push_back(time);\
times.push_back(time*1e6);\
total_time+=time;\
}\
double t = 1e-6*median(times);\
double t = median(times);\
std::cout << " " << PERF << std::flush;\
}
@@ -98,7 +98,7 @@ void bench(ad::numeric_type dtype)
T *cux, *cuy;
cudaMalloc((void**) &cux, N * sizeof(T));
cudaMalloc((void**) &cuy, N * sizeof(T));
BENCHMARK(cublasSaxpy(N, 2, cux, 1, cuy, 1), 3*N*dtsize/t)
BENCHMARK_CUDA(cublasSaxpy(N, 2, cux, 1, cuy, 1), 3*N*dtsize/t)
cudaFree(cux);
cudaFree(cuy);
#endif