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

Added benchmarks for contraction on CPU.

Browse Source
This commit is contained in:
Benoit Steiner 2016-05-13 14:32:17 -07:00
parent c4fc8b70ec
commit 069a0b04d7
2 changed files with 45 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
bench/tensors/README
View File

@ -1,4 +1,6 @@
Each benchmark comes in 2 flavors: one that runs on CPU, and one that runs on GPU.
The tensor benchmark suite is made of several parts.
The first part is a generic suite, in which each benchmark comes in 2 flavors: one that runs on CPU, and one that runs on GPU.
To compile the floating point CPU benchmarks, simply call:
g++ tensor_benchmarks_cpu.cc benchmark_main.cc -I ../../ -std=c++11 -O3 -DNDEBUG -pthread -mavx -o benchmarks_cpu
@ -6,7 +8,8 @@ g++ tensor_benchmarks_cpu.cc benchmark_main.cc -I ../../ -std=c++11 -O3 -DNDEBUG
To compile the floating point GPU benchmarks, simply call:
nvcc tensor_benchmarks_gpu.cu benchmark_main.cc -I ../../ -std=c++11 -O2 -DNDEBUG -arch compute_35 -o benchmarks_gpu
To compile the half float GPU benchmarks, simply call the command line below. You'll need a recent GPU that supports compute capability 5.3 or higher to run them and nvcc 7.5 or higher to compile the code.
We also provide a version of the generic GPU tensor benchmarks that uses half floats (aka fp16) instead of regular floats. To compile these benchmarks, simply call the command line below. You'll need a recent GPU that supports compute capability 5.3 or higher to run them and nvcc 7.5 or higher to compile the code.
nvcc tensor_benchmarks_fp16_gpu.cu benchmark_main.cc -I ../../ -std=c++11 -O2 -DNDEBUG -arch compute_53 -o benchmarks_fp16_gpu
last but not least, we also provide a suite of benchmarks to measure the scalability of the contraction code on CPU. To compile these benchmarks, call
g++ contraction_benchmarks_cpu.cc benchmark_main.cc -I ../../ -std=c++11 -O3 -DNDEBUG -pthread -mavx -o benchmarks_cpu

39
bench/tensors/contraction_benchmarks_cpu.cc Normal file
View File

@ -0,0 +1,39 @@
#define EIGEN_USE_THREADS
#include <string>
#include "tensor_benchmarks.h"
#define CREATE_THREAD_POOL(threads) \
Eigen::ThreadPool pool(threads); \
Eigen::ThreadPoolDevice device(&pool, threads);
// Contractions for number of threads ranging from 1 to 32
// Dimensions are Rows, Cols, Depth
#define BM_ContractionCPU(D1, D2, D3) \
static void BM_##Contraction##_##D1##x##D2##x##D3(int iters, int Threads) { \
StopBenchmarkTiming(); \
CREATE_THREAD_POOL(Threads); \
BenchmarkSuite<Eigen::ThreadPoolDevice, float> suite(device, D1, D2, D3); \
suite.contraction(iters); \
} \
BENCHMARK_RANGE(BM_##Contraction##_##D1##x##D2##x##D3, 1, 32);
// Vector Matrix and Matrix Vector products
BM_ContractionCPU(1, 2000, 500);
BM_ContractionCPU(2000, 1, 500);
// Various skinny matrices
BM_ContractionCPU(250, 3, 512);
BM_ContractionCPU(1500, 3, 512);
BM_ContractionCPU(512, 800, 4);
BM_ContractionCPU(512, 80, 800);
BM_ContractionCPU(512, 80, 13522);
BM_ContractionCPU(1, 80, 13522);
BM_ContractionCPU(3200, 512, 4);
BM_ContractionCPU(3200, 512, 80);
BM_ContractionCPU(3200, 80, 512);