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
|
/*******************************************************************************
* Copyright 2020-2025 Intel Corporation
*
* 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.
*******************************************************************************/
/// @example pooling.cpp
/// > Annotated version: @ref pooling_example_cpp
///
/// @page pooling_example_cpp_short
///
/// This C++ API example demonstrates how to create and execute a
/// [Pooling](@ref dev_guide_pooling) primitive in forward training propagation
/// mode.
///
/// @page pooling_example_cpp Pooling Primitive Example
/// @copydetails pooling_example_cpp_short
///
/// @include pooling.cpp
#include <algorithm>
#include <cmath>
#include <iostream>
#include <string>
#include <vector>
#include "example_utils.hpp"
#include "oneapi/dnnl/dnnl.hpp"
using namespace dnnl;
void pooling_example(dnnl::engine::kind engine_kind) {
// Create execution dnnl::engine.
dnnl::engine engine(engine_kind, 0);
// Create dnnl::stream.
dnnl::stream engine_stream(engine);
// Tensor dimensions.
const memory::dim N = 3, // batch size
IC = 3, // input channels
IH = 27, // input tensor height
IW = 27, // input tensor width
KH = 11, // kernel height
KW = 11, // kernel width
PH_L = 0, // height padding: left
PH_R = 0, // height padding: right
PW_L = 0, // width padding: left
PW_R = 0, // width padding: right
SH = 4, // height-wise stride
SW = 4, // width-wise stride
DH = 1, // height-wise dilation
DW = 1; // width-wise dilation
// oneDNN uses the following formula to calculate destination dimensions:
// dst = (src - ((weights - 1) * (dilation_onednn + 1) + 1)) / stride + 1
//
// PyTorch and TensorFlow use a different formula:
// dst = (src - ((weights - 1) * dilation_torch + 1)) / stride + 1
//
// As a result, the PyTorch and Tensorflow dilation parameters need to be
// adjusted by subtracting 1:
// dilation_onednn = dilation_torch - 1.
//
// Output tensor height and width.
const memory::dim OH = (IH - ((KH - 1) * DH + KH) + PH_L + PH_R) / SH + 1;
const memory::dim OW = (IW - ((KW - 1) * DW + KW) + PW_L + PW_R) / SW + 1;
// Source (src) and destination (dst) tensors dimensions.
memory::dims src_dims = {N, IC, IH, IW};
memory::dims dst_dims = {N, IC, OH, OW};
// Kernel dimensions.
memory::dims kernel_dims = {KH, KW};
// Strides, padding dimensions.
memory::dims strides_dims = {SH, SW};
memory::dims padding_dims_l = {PH_L, PW_L};
memory::dims padding_dims_r = {PH_R, PW_R};
memory::dims dilation = {DH, DW};
// Allocate buffers.
std::vector<float> src_data(product(src_dims));
std::vector<float> dst_data(product(dst_dims));
std::generate(src_data.begin(), src_data.end(), []() {
static int i = 0;
return std::cos(i++ / 10.f);
});
// Create memory descriptors and memory objects for src and dst.
auto src_md = memory::desc(
src_dims, memory::data_type::f32, memory::format_tag::nchw);
auto src_mem = memory(src_md, engine);
auto dst_md = memory::desc(
dst_dims, memory::data_type::f32, memory::format_tag::nchw);
auto dst_mem = memory(dst_md, engine);
// Write data to memory object's handle.
write_to_dnnl_memory(src_data.data(), src_mem);
// Create primitive descriptor.
auto pooling_pd = pooling_forward::primitive_desc(engine,
prop_kind::forward_training, algorithm::pooling_max, src_md, dst_md,
strides_dims, kernel_dims, dilation, padding_dims_l,
padding_dims_r);
// Create workspace memory objects using memory descriptor created by the
// primitive descriptor.
// NOTE: Here, the workspace is required to save the indices where maximum
// was found, and is used in backward pooling to perform upsampling.
auto workspace_mem = memory(pooling_pd.workspace_desc(), engine);
// Create the primitive.
auto pooling_prim = pooling_forward(pooling_pd);
// Primitive arguments. Set up in-place execution by assigning src as DST.
std::unordered_map<int, memory> pooling_args;
pooling_args.insert({DNNL_ARG_SRC, src_mem});
pooling_args.insert({DNNL_ARG_DST, dst_mem});
pooling_args.insert({DNNL_ARG_WORKSPACE, workspace_mem});
// Primitive execution: pooling.
pooling_prim.execute(engine_stream, pooling_args);
// Wait for the computation to finalize.
engine_stream.wait();
// Read data from memory object's handle.
read_from_dnnl_memory(dst_data.data(), dst_mem);
}
int main(int argc, char **argv) {
return handle_example_errors(
pooling_example, parse_engine_kind(argc, argv));
}
|
