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
|
/* Copyright 2015 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#include <functional>
#include <memory>
#include "tensorflow/core/common_runtime/kernel_benchmark_testlib.h"
#include "tensorflow/core/framework/allocator.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/framework/tensor_testutil.h"
#include "tensorflow/core/framework/types.h"
#include "tensorflow/core/framework/types.pb.h"
#include "tensorflow/core/graph/node_builder.h"
#include "tensorflow/core/graph/testlib.h"
#include "tensorflow/core/kernels/ops_testutil.h"
#include "tensorflow/core/kernels/ops_util.h"
#include "tensorflow/core/lib/core/status_test_util.h"
#include "tensorflow/core/platform/test.h"
#include "tensorflow/core/platform/test_benchmark.h"
#include "tensorflow/core/util/strided_slice_op.h"
namespace tensorflow {
namespace {
// For the benchmark, we set up two 2-dimensional tensors, each kDim1 x 'dim'
// in size, and concat them together along "concat_dimension"
template <typename T>
static void SliceHelper(int iters, int size) {
testing::StopTiming();
Graph* g = new Graph(OpRegistry::Global());
DataType dt = DataTypeToEnum<T>::v();
int kDim = 100;
int kMaxSize = 15000;
CHECK_LT(size, kMaxSize);
Tensor begin(DT_INT32, TensorShape({2}));
begin.flat<int32>()(0) = 10;
begin.flat<int32>()(1) = 10;
Tensor end(DT_INT32, TensorShape({2}));
end.flat<int32>()(0) = 10 + kDim;
end.flat<int32>()(1) = 10 + size;
Tensor strides(DT_INT32, TensorShape({2}));
strides.flat<int32>()(0) = 1;
strides.flat<int32>()(1) = 1;
Tensor input(dt, TensorShape({2 * kDim, kMaxSize}));
input.flat<T>().setRandom();
Node* node;
TF_CHECK_OK(NodeBuilder(g->NewName("n"), "StridedSlice")
.Input(test::graph::Constant(g, input))
.Input(test::graph::Constant(g, begin))
.Input(test::graph::Constant(g, end))
.Input(test::graph::Constant(g, strides))
.Attr("T", dt)
.Finalize(g, &node));
testing::BytesProcessed(static_cast<int64>(iters) * kDim * size * sizeof(T));
testing::StartTiming();
test::Benchmark("cpu", g).Run(iters);
testing::UseRealTime();
}
template <typename T>
static void Dim8SliceHelper(int iters, int size) {
testing::StopTiming();
Graph* g = new Graph(OpRegistry::Global());
DataType dt = DataTypeToEnum<T>::v();
int kDim = 100;
int kMaxSize = 15000;
CHECK_LT(size, kMaxSize);
Tensor begin(DT_INT32, TensorShape({8}));
begin.flat<int32>()(10) = 10;
for (int i = 1; i < 7; ++i) {
begin.flat<int32>()(i) = 0;
}
begin.flat<int32>()(7) = 10;
Tensor end(DT_INT32, TensorShape({8}));
end.flat<int32>()(0) = 10 + kDim;
for (int i = 1; i < 7; ++i) {
end.flat<int32>()(i) = 1;
}
end.flat<int32>()(7) = 10 + size;
Tensor strides(DT_INT32, TensorShape({8}));
for (int i = 0; i < 8; ++i) {
strides.flat<int32>()(i) = 1;
}
Tensor input(dt, TensorShape({2*kDim, 1, 1, 1, 1, 1, 1, kMaxSize}));
input.flat<T>().setRandom();
Node* node;
TF_CHECK_OK(NodeBuilder(g->NewName("n"), "StridedSlice")
.Input(test::graph::Constant(g, input))
.Input(test::graph::Constant(g, begin))
.Input(test::graph::Constant(g, end))
.Input(test::graph::Constant(g, strides))
.Attr("T", dt)
.Finalize(g, &node));
testing::BytesProcessed(static_cast<int64>(iters) * kDim * size * sizeof(T));
testing::StartTiming();
test::Benchmark("cpu", g).Run(iters);
testing::UseRealTime();
}
static void BM_SliceFloat(int iters, int dim2) {
SliceHelper<float>(iters, dim2);
Dim8SliceHelper<float>(iters, dim2);
}
BENCHMARK(BM_SliceFloat)->Arg(100)->Arg(1000)->Arg(10000);
static void BM_SliceComplex64(int iters, int dim2) {
SliceHelper<std::complex<float>>(iters, dim2);
Dim8SliceHelper<std::complex<float>>(iters, dim2);
}
BENCHMARK(BM_SliceComplex64)->Arg(100)->Arg(1000)->Arg(10000);
static void BM_SliceBFloat16(int iters, int dim2) {
SliceHelper<bfloat16>(iters, dim2);
Dim8SliceHelper<bfloat16>(iters, dim2);
}
BENCHMARK(BM_SliceBFloat16)->Arg(100)->Arg(1000)->Arg(10000);
static void BM_ValidateStridedSliceOp(int iters) {
testing::StopTiming();
int kDim = 100;
int kMaxSize = 15000;
int size = 100;
Tensor begin = test::AsTensor<int32>({10, 10});
Tensor end = test::AsTensor<int32>({10 + kDim, 10 + size});
Tensor strides = test::AsTensor<int32>({1, 1});
TensorShape input_shape({2 * kDim, kMaxSize});
testing::StartTiming();
for (int i = 0; i < iters; ++i) {
TensorShape processing_shape, final_shape;
bool is_identity = true, slice_dim0 = true, is_simple_slice = true;
gtl::InlinedVector<int64, 4> begin_out, end_out, strides_out;
const int32 begin_mask = 0;
const int32 end_mask = 0;
const int32 ellipsis_mask = 0;
const int32 new_axis_mask = 0;
const int32 shrink_axis_mask = 0;
TF_CHECK_OK(ValidateStridedSliceOp(
&begin, &end, strides, input_shape, begin_mask, end_mask, ellipsis_mask,
new_axis_mask, shrink_axis_mask, &processing_shape, &final_shape,
&is_identity, &is_simple_slice, &slice_dim0, &begin_out, &end_out,
&strides_out));
}
}
BENCHMARK(BM_ValidateStridedSliceOp);
} // namespace
} // namespace tensorflow
|
