mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-03-07 18:27:40 +08:00
Code Issues Packages Projects Releases Wiki Activity

Updated the benchmarking code to use Eigen::half instead of half

Browse Source
This commit is contained in:
Benoit Steiner 2016-03-24 18:00:33 -07:00
parent 044efea965
commit 0968e925a0
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
bench/tensors/tensor_benchmarks_fp16_gpu.cu
View File

@ -12,7 +12,7 @@
StopBenchmarkTiming(); \ StopBenchmarkTiming(); \
Eigen::CudaStreamDevice stream; \ Eigen::CudaStreamDevice stream; \
Eigen::GpuDevice device(&stream); \ Eigen::GpuDevice device(&stream); \
BenchmarkSuite<Eigen::GpuDevice, half> suite(device, N); \ BenchmarkSuite<Eigen::GpuDevice, Eigen::half> suite(device, N); \
cudaDeviceSynchronize(); \ cudaDeviceSynchronize(); \
suite.FUNC(iters); \ suite.FUNC(iters); \
} \ } \
@ -41,7 +41,7 @@ BM_FuncGPU(colReduction);
StopBenchmarkTiming(); \ StopBenchmarkTiming(); \
Eigen::CudaStreamDevice stream; \ Eigen::CudaStreamDevice stream; \
Eigen::GpuDevice device(&stream); \ Eigen::GpuDevice device(&stream); \
BenchmarkSuite<Eigen::GpuDevice, half> suite(device, D1, D2, D3); \ BenchmarkSuite<Eigen::GpuDevice, Eigen::half> suite(device, D1, D2, D3); \
cudaDeviceSynchronize(); \ cudaDeviceSynchronize(); \
suite.FUNC(iters); \ suite.FUNC(iters); \
} \ } \
@ -60,7 +60,7 @@ BM_FuncWithInputDimsGPU(contraction, N, N, 64);
StopBenchmarkTiming(); \ StopBenchmarkTiming(); \
Eigen::CudaStreamDevice stream; \ Eigen::CudaStreamDevice stream; \
Eigen::GpuDevice device(&stream); \ Eigen::GpuDevice device(&stream); \
BenchmarkSuite<Eigen::GpuDevice, half> suite(device, N); \ BenchmarkSuite<Eigen::GpuDevice, Eigen::half> suite(device, N); \
cudaDeviceSynchronize(); \ cudaDeviceSynchronize(); \
suite.FUNC(iters, DIM1, DIM2); \ suite.FUNC(iters, DIM1, DIM2); \
} \ } \
@ -73,4 +73,4 @@ BM_FuncWithKernelDimsGPU(convolution, 7, 4);
BM_FuncWithKernelDimsGPU(convolution, 4, 7); BM_FuncWithKernelDimsGPU(convolution, 4, 7);
BM_FuncWithKernelDimsGPU(convolution, 7, 64); BM_FuncWithKernelDimsGPU(convolution, 7, 64);
BM_FuncWithKernelDimsGPU(convolution, 64, 7); BM_FuncWithKernelDimsGPU(convolution, 64, 7);
*/ */