tensor_benchmarks_cpu.cc File Reference

#include <string>
#include "tensor_benchmarks.h"

Include dependency graph for tensor_benchmarks_cpu.cc:

Go to the source code of this file.

Macros
#define	BM_FuncCPU(FUNC, THREADS)
#define	BM_FuncWithInputDimsCPU(FUNC, D1, D2, D3, THREADS)
#define	BM_FuncWithKernelDimsCPU(FUNC, DIM1, DIM2, THREADS)
#define	CREATE_THREAD_POOL(threads)
#define	EIGEN_USE_THREADS

Functions
	BM_FuncCPU (memcpy, 4)
	BM_FuncCPU (memcpy, 8)
	BM_FuncCPU (memcpy, 12)
	BM_FuncCPU (typeCasting, 4)
	BM_FuncCPU (typeCasting, 8)
	BM_FuncCPU (typeCasting, 12)
	BM_FuncCPU (random, 4)
	BM_FuncCPU (random, 8)
	BM_FuncCPU (random, 12)
	BM_FuncCPU (slicing, 4)
	BM_FuncCPU (slicing, 8)
	BM_FuncCPU (slicing, 12)
	BM_FuncCPU (rowChip, 4)
	BM_FuncCPU (rowChip, 8)
	BM_FuncCPU (rowChip, 12)
	BM_FuncCPU (colChip, 4)
	BM_FuncCPU (colChip, 8)
	BM_FuncCPU (colChip, 12)
	BM_FuncCPU (shuffling, 4)
	BM_FuncCPU (shuffling, 8)
	BM_FuncCPU (shuffling, 12)
	BM_FuncCPU (padding, 4)
	BM_FuncCPU (padding, 8)
	BM_FuncCPU (padding, 12)
	BM_FuncCPU (striding, 4)
	BM_FuncCPU (striding, 8)
	BM_FuncCPU (striding, 12)
	BM_FuncCPU (broadcasting, 4)
	BM_FuncCPU (broadcasting, 8)
	BM_FuncCPU (broadcasting, 12)
	BM_FuncCPU (coeffWiseOp, 4)
	BM_FuncCPU (coeffWiseOp, 8)
	BM_FuncCPU (coeffWiseOp, 12)
	BM_FuncCPU (algebraicFunc, 4)
	BM_FuncCPU (algebraicFunc, 8)
	BM_FuncCPU (algebraicFunc, 12)
	BM_FuncCPU (transcendentalFunc, 4)
	BM_FuncCPU (transcendentalFunc, 8)
	BM_FuncCPU (transcendentalFunc, 12)
	BM_FuncCPU (rowReduction, 4)
	BM_FuncCPU (rowReduction, 8)
	BM_FuncCPU (rowReduction, 12)
	BM_FuncCPU (colReduction, 4)
	BM_FuncCPU (colReduction, 8)
	BM_FuncCPU (colReduction, 12)
	BM_FuncWithInputDimsCPU (contraction, N, N, N, 1)
	BM_FuncWithInputDimsCPU (contraction, N, N, N, 4)
	BM_FuncWithInputDimsCPU (contraction, N, N, N, 8)
	BM_FuncWithInputDimsCPU (contraction, N, N, N, 12)
	BM_FuncWithInputDimsCPU (contraction, N, N, N, 16)
	BM_FuncWithInputDimsCPU (contraction, 64, N, N, 1)
	BM_FuncWithInputDimsCPU (contraction, 64, N, N, 4)
	BM_FuncWithInputDimsCPU (contraction, 64, N, N, 8)
	BM_FuncWithInputDimsCPU (contraction, 64, N, N, 12)
	BM_FuncWithInputDimsCPU (contraction, 64, N, N, 16)
	BM_FuncWithInputDimsCPU (contraction, N, 64, N, 1)
	BM_FuncWithInputDimsCPU (contraction, N, 64, N, 4)
	BM_FuncWithInputDimsCPU (contraction, N, 64, N, 8)
	BM_FuncWithInputDimsCPU (contraction, N, 64, N, 12)
	BM_FuncWithInputDimsCPU (contraction, N, 64, N, 16)
	BM_FuncWithInputDimsCPU (contraction, N, N, 64, 1)
	BM_FuncWithInputDimsCPU (contraction, N, N, 64, 4)
	BM_FuncWithInputDimsCPU (contraction, N, N, 64, 8)
	BM_FuncWithInputDimsCPU (contraction, N, N, 64, 12)
	BM_FuncWithInputDimsCPU (contraction, N, N, 64, 16)
	BM_FuncWithInputDimsCPU (contraction, 1, N, N, 1)
	BM_FuncWithInputDimsCPU (contraction, 1, N, N, 4)
	BM_FuncWithInputDimsCPU (contraction, 1, N, N, 8)
	BM_FuncWithInputDimsCPU (contraction, 1, N, N, 12)
	BM_FuncWithInputDimsCPU (contraction, 1, N, N, 16)
	BM_FuncWithInputDimsCPU (contraction, N, N, 1, 1)
	BM_FuncWithInputDimsCPU (contraction, N, N, 1, 4)
	BM_FuncWithInputDimsCPU (contraction, N, N, 1, 8)
	BM_FuncWithInputDimsCPU (contraction, N, N, 1, 12)
	BM_FuncWithInputDimsCPU (contraction, N, N, 1, 16)
	BM_FuncWithKernelDimsCPU (convolution, 7, 1, 4)
	BM_FuncWithKernelDimsCPU (convolution, 7, 1, 8)
	BM_FuncWithKernelDimsCPU (convolution, 7, 1, 12)
	BM_FuncWithKernelDimsCPU (convolution, 1, 7, 4)
	BM_FuncWithKernelDimsCPU (convolution, 1, 7, 8)
	BM_FuncWithKernelDimsCPU (convolution, 1, 7, 12)
	BM_FuncWithKernelDimsCPU (convolution, 7, 4, 4)
	BM_FuncWithKernelDimsCPU (convolution, 7, 4, 8)
	BM_FuncWithKernelDimsCPU (convolution, 7, 4, 12)
	BM_FuncWithKernelDimsCPU (convolution, 4, 7, 4)
	BM_FuncWithKernelDimsCPU (convolution, 4, 7, 8)
	BM_FuncWithKernelDimsCPU (convolution, 4, 7, 12)
	BM_FuncWithKernelDimsCPU (convolution, 7, 64, 4)
	BM_FuncWithKernelDimsCPU (convolution, 7, 64, 8)
	BM_FuncWithKernelDimsCPU (convolution, 7, 64, 12)
	BM_FuncWithKernelDimsCPU (convolution, 64, 7, 4)
	BM_FuncWithKernelDimsCPU (convolution, 64, 7, 8)
	BM_FuncWithKernelDimsCPU (convolution, 64, 7, 12)

Macro Definition Documentation

BM_FuncCPU

#define BM_FuncCPU	(		FUNC,
			THREADS
	)

Value:

static void BM_##FUNC##_##THREADS##T(int iters, int N) {           \
    StopBenchmarkTiming();                                           \
    CREATE_THREAD_POOL(THREADS);                                     \
    BenchmarkSuite<Eigen::ThreadPoolDevice, float> suite(device, N); \
    suite.FUNC(iters);                                               \
  }                                                                  \
  BENCHMARK_RANGE(BM_##FUNC##_##THREADS##T, 10, 5000);

Definition at line 12 of file tensor_benchmarks_cpu.cc.

BM_FuncWithInputDimsCPU

#define BM_FuncWithInputDimsCPU	(		FUNC,
			D1,
			D2,
			D3,
			THREADS
	)

Value:

static void BM_##FUNC##_##D1##x##D2##x##D3##_##THREADS##T(int iters, int N) { \
    StopBenchmarkTiming();                                                      \
    if (THREADS == 1) {                                                         \
      Eigen::DefaultDevice device;                                              \
      BenchmarkSuite<Eigen::DefaultDevice, float> suite(device, D1, D2, D3);    \
      suite.FUNC(iters);                                                        \
    } else {                                                                    \
      CREATE_THREAD_POOL(THREADS);                                              \
      BenchmarkSuite<Eigen::ThreadPoolDevice, float> suite(device, D1, D2, D3); \
      suite.FUNC(iters);                                                        \
    }                                                                           \
  }                                                                             \
  BENCHMARK_RANGE(BM_##FUNC##_##D1##x##D2##x##D3##_##THREADS##T, 10, 5000);

Definition at line 83 of file tensor_benchmarks_cpu.cc.

BM_FuncWithKernelDimsCPU

#define BM_FuncWithKernelDimsCPU	(		FUNC,
			DIM1,
			DIM2,
			THREADS
	)

Value:

static void BM_##FUNC##_##DIM1##x##DIM2##_##THREADS##T(int iters, int N) {   \
    StopBenchmarkTiming();                                                     \
    CREATE_THREAD_POOL(THREADS);                                               \
    BenchmarkSuite<Eigen::ThreadPoolDevice, float> suite(device, N);           \
    suite.FUNC(iters, DIM1, DIM2);                                             \
  }                                                                            \
  BENCHMARK_RANGE(BM_##FUNC##_##DIM1##x##DIM2##_##THREADS##T, 128, 5000);

Definition at line 137 of file tensor_benchmarks_cpu.cc.

CREATE_THREAD_POOL

#define CREATE_THREAD_POOL

(

threads

)

Value:

Eigen::ThreadPool pool(threads);                \
Eigen::ThreadPoolDevice device(&pool, threads);

Definition at line 7 of file tensor_benchmarks_cpu.cc.

EIGEN_USE_THREADS

#define EIGEN_USE_THREADS

Definition at line 1 of file tensor_benchmarks_cpu.cc.

Function Documentation

BM_FuncCPU() [1/45]

BM_FuncCPU	(	memcpy	,
		4
	)

BM_FuncCPU() [2/45]

BM_FuncCPU	(	memcpy	,
		8
	)

BM_FuncCPU() [3/45]

BM_FuncCPU	(	memcpy	,
		12
	)

BM_FuncCPU() [4/45]

BM_FuncCPU	(	typeCasting	,
		4
	)

BM_FuncCPU() [5/45]

BM_FuncCPU	(	typeCasting	,
		8
	)

BM_FuncCPU() [6/45]

BM_FuncCPU	(	typeCasting	,
		12
	)

BM_FuncCPU() [7/45]

BM_FuncCPU	(	random	,
		4
	)

BM_FuncCPU() [8/45]

BM_FuncCPU	(	random	,
		8
	)

BM_FuncCPU() [9/45]

BM_FuncCPU	(	random	,
		12
	)

BM_FuncCPU() [10/45]

BM_FuncCPU	(	slicing	,
		4
	)

BM_FuncCPU() [11/45]

BM_FuncCPU	(	slicing	,
		8
	)

BM_FuncCPU() [12/45]

BM_FuncCPU	(	slicing	,
		12
	)

BM_FuncCPU() [13/45]

BM_FuncCPU	(	rowChip	,
		4
	)

BM_FuncCPU() [14/45]

BM_FuncCPU	(	rowChip	,
		8
	)

BM_FuncCPU() [15/45]

BM_FuncCPU	(	rowChip	,
		12
	)

BM_FuncCPU() [16/45]

BM_FuncCPU	(	colChip	,
		4
	)

BM_FuncCPU() [17/45]

BM_FuncCPU	(	colChip	,
		8
	)

BM_FuncCPU() [18/45]

BM_FuncCPU	(	colChip	,
		12
	)

BM_FuncCPU() [19/45]

BM_FuncCPU	(	shuffling	,
		4
	)

BM_FuncCPU() [20/45]

BM_FuncCPU	(	shuffling	,
		8
	)

BM_FuncCPU() [21/45]

BM_FuncCPU	(	shuffling	,
		12
	)

BM_FuncCPU() [22/45]

BM_FuncCPU	(	padding	,
		4
	)

BM_FuncCPU() [23/45]

BM_FuncCPU	(	padding	,
		8
	)

BM_FuncCPU() [24/45]

BM_FuncCPU	(	padding	,
		12
	)

BM_FuncCPU() [25/45]

BM_FuncCPU	(	striding	,
		4
	)

BM_FuncCPU() [26/45]

BM_FuncCPU	(	striding	,
		8
	)

BM_FuncCPU() [27/45]

BM_FuncCPU	(	striding	,
		12
	)

BM_FuncCPU() [28/45]

BM_FuncCPU	(	broadcasting	,
		4
	)

BM_FuncCPU() [29/45]

BM_FuncCPU	(	broadcasting	,
		8
	)

BM_FuncCPU() [30/45]

BM_FuncCPU	(	broadcasting	,
		12
	)

BM_FuncCPU() [31/45]

BM_FuncCPU	(	coeffWiseOp	,
		4
	)

BM_FuncCPU() [32/45]

BM_FuncCPU	(	coeffWiseOp	,
		8
	)

BM_FuncCPU() [33/45]

BM_FuncCPU	(	coeffWiseOp	,
		12
	)

BM_FuncCPU() [34/45]

BM_FuncCPU	(	algebraicFunc	,
		4
	)

BM_FuncCPU() [35/45]

BM_FuncCPU	(	algebraicFunc	,
		8
	)

BM_FuncCPU() [36/45]

BM_FuncCPU	(	algebraicFunc	,
		12
	)

BM_FuncCPU() [37/45]

BM_FuncCPU	(	transcendentalFunc	,
		4
	)

BM_FuncCPU() [38/45]

BM_FuncCPU	(	transcendentalFunc	,
		8
	)

BM_FuncCPU() [39/45]

BM_FuncCPU	(	transcendentalFunc	,
		12
	)

BM_FuncCPU() [40/45]

BM_FuncCPU	(	rowReduction	,
		4
	)

BM_FuncCPU() [41/45]

BM_FuncCPU	(	rowReduction	,
		8
	)

BM_FuncCPU() [42/45]

BM_FuncCPU	(	rowReduction	,
		12
	)

BM_FuncCPU() [43/45]

BM_FuncCPU	(	colReduction	,
		4
	)

BM_FuncCPU() [44/45]

BM_FuncCPU	(	colReduction	,
		8
	)

BM_FuncCPU() [45/45]

BM_FuncCPU	(	colReduction	,
		12
	)

BM_FuncWithInputDimsCPU() [1/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		N	,
		1
	)

BM_FuncWithInputDimsCPU() [2/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		N	,
		4
	)

BM_FuncWithInputDimsCPU() [3/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		N	,
		8
	)

BM_FuncWithInputDimsCPU() [4/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		N	,
		12
	)

BM_FuncWithInputDimsCPU() [5/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		N	,
		16
	)

BM_FuncWithInputDimsCPU() [6/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		64	,
		N	,
		N	,
		1
	)

BM_FuncWithInputDimsCPU() [7/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		64	,
		N	,
		N	,
		4
	)

BM_FuncWithInputDimsCPU() [8/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		64	,
		N	,
		N	,
		8
	)

BM_FuncWithInputDimsCPU() [9/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		64	,
		N	,
		N	,
		12
	)

BM_FuncWithInputDimsCPU() [10/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		64	,
		N	,
		N	,
		16
	)

BM_FuncWithInputDimsCPU() [11/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		64	,
		N	,
		1
	)

BM_FuncWithInputDimsCPU() [12/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		64	,
		N	,
		4
	)

BM_FuncWithInputDimsCPU() [13/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		64	,
		N	,
		8
	)

BM_FuncWithInputDimsCPU() [14/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		64	,
		N	,
		12
	)

BM_FuncWithInputDimsCPU() [15/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		64	,
		N	,
		16
	)

BM_FuncWithInputDimsCPU() [16/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		64	,
		1
	)

BM_FuncWithInputDimsCPU() [17/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		64	,
		4
	)

BM_FuncWithInputDimsCPU() [18/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		64	,
		8
	)

BM_FuncWithInputDimsCPU() [19/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		64	,
		12
	)

BM_FuncWithInputDimsCPU() [20/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		64	,
		16
	)

BM_FuncWithInputDimsCPU() [21/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		1	,
		N	,
		N	,
		1
	)

BM_FuncWithInputDimsCPU() [22/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		1	,
		N	,
		N	,
		4
	)

BM_FuncWithInputDimsCPU() [23/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		1	,
		N	,
		N	,
		8
	)

BM_FuncWithInputDimsCPU() [24/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		1	,
		N	,
		N	,
		12
	)

BM_FuncWithInputDimsCPU() [25/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		1	,
		N	,
		N	,
		16
	)

BM_FuncWithInputDimsCPU() [26/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		1	,
		1
	)

BM_FuncWithInputDimsCPU() [27/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		1	,
		4
	)

BM_FuncWithInputDimsCPU() [28/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		1	,
		8
	)

BM_FuncWithInputDimsCPU() [29/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		1	,
		12
	)

BM_FuncWithInputDimsCPU() [30/30]

BM_FuncWithInputDimsCPU	(	contraction	,
		N	,
		N	,
		1	,
		16
	)

BM_FuncWithKernelDimsCPU() [1/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		1	,
		4
	)

BM_FuncWithKernelDimsCPU() [2/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		1	,
		8
	)

BM_FuncWithKernelDimsCPU() [3/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		1	,
		12
	)

BM_FuncWithKernelDimsCPU() [4/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		1	,
		7	,
		4
	)

BM_FuncWithKernelDimsCPU() [5/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		1	,
		7	,
		8
	)

BM_FuncWithKernelDimsCPU() [6/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		1	,
		7	,
		12
	)

BM_FuncWithKernelDimsCPU() [7/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		4	,
		4
	)

BM_FuncWithKernelDimsCPU() [8/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		4	,
		8
	)

BM_FuncWithKernelDimsCPU() [9/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		4	,
		12
	)

BM_FuncWithKernelDimsCPU() [10/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		4	,
		7	,
		4
	)

BM_FuncWithKernelDimsCPU() [11/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		4	,
		7	,
		8
	)

BM_FuncWithKernelDimsCPU() [12/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		4	,
		7	,
		12
	)

BM_FuncWithKernelDimsCPU() [13/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		64	,
		4
	)

BM_FuncWithKernelDimsCPU() [14/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		64	,
		8
	)

BM_FuncWithKernelDimsCPU() [15/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		7	,
		64	,
		12
	)

BM_FuncWithKernelDimsCPU() [16/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		64	,
		7	,
		4
	)

BM_FuncWithKernelDimsCPU() [17/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		64	,
		7	,
		8
	)

BM_FuncWithKernelDimsCPU() [18/18]

BM_FuncWithKernelDimsCPU	(	convolution	,
		64	,
		7	,
		12
	)