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
|
/*
* Copyright (c) 2017-2018 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include "ConvolutionLayer.h"
#include "tests/validation/FixedPoint.h"
#include "tests/validation/Helpers.h"
#include "tests/validation/reference/Convolution3d.h"
#include "tests/validation/reference/Utils.h"
#include "tests/validation/reference/UtilsQuantizedAsymm.h"
#include "tests/framework/Asserts.h"
#include "arm_compute/core/utils/quantization/AsymmHelpers.h"
namespace arm_compute
{
namespace test
{
namespace validation
{
namespace reference
{
namespace
{
} // namespace
template <typename T, typename TB>
SimpleTensor<T> convolution_layer(const SimpleTensor<T> &src, const SimpleTensor<T> &weights, const SimpleTensor<TB> &bias, const TensorShape &output_shape, const PadStrideInfo &info,
const Size2D &dilation)
{
// Create reference
SimpleTensor<T> dst{ output_shape, src.data_type(), 1, src.fixed_point_position(), src.quantization_info() };
// Compute reference
const int width_in = src.shape().x();
const int height_in = src.shape().y();
const int depth_in = src.shape().z();
const int width_out = dst.shape().x();
const int height_out = dst.shape().y();
const int depth_out = dst.shape().z();
const int width_weights = weights.shape().x();
const int height_weights = weights.shape().y();
const int depth_weights = weights.shape().z();
const int pad_left = info.pad_left();
const int pad_top = info.pad_top();
const int stride_xi = info.stride().first;
const int stride_yi = info.stride().second;
auto output_wh = scaled_dimensions(width_in, height_in, width_weights, height_weights, info, dilation);
const int start_xi = (dilation.x() * (width_weights - 1) + 1) / 2 - pad_left;
const int start_yi = (dilation.y() * (height_weights - 1) + 1) / 2 - pad_top;
const int end_xi = output_wh.first * stride_xi;
const int end_yi = output_wh.second * stride_yi;
const int num_batches = src.shape().total_size() / (width_in * height_in * depth_in);
for(int r = 0; r < num_batches; ++r)
{
for(int yi = start_yi; yi < start_yi + end_yi; yi += stride_yi)
{
for(int xi = start_xi; xi < start_xi + end_xi; xi += stride_xi)
{
for(int ofm = 0; ofm < depth_out; ++ofm)
{
// Compute input and output offsets
const int offset_in = r * width_in * height_in * depth_in;
const int xo = (xi - start_xi) / stride_xi;
const int yo = (yi - start_yi) / stride_yi;
const int offset_out = xo + yo * width_out + ofm * width_out * height_out + r * width_out * height_out * depth_out;
ARM_COMPUTE_ASSERT(xo < width_out);
ARM_COMPUTE_ASSERT(yo < height_out);
// Compute 3D convolution
convolution_3d::detail::convolution3d(src, weights, bias, dst,
offset_in, ofm * width_weights * height_weights * depth_weights, ofm, offset_out,
xi, yi,
width_in, height_in, depth_in,
width_weights, height_weights, dilation.x(), dilation.y());
}
}
}
}
return dst;
}
template SimpleTensor<float> convolution_layer(const SimpleTensor<float> &src, const SimpleTensor<float> &weights, const SimpleTensor<float> &bias, const TensorShape &output_shape,
const PadStrideInfo &info, const Size2D &dilation);
template SimpleTensor<half> convolution_layer(const SimpleTensor<half> &src, const SimpleTensor<half> &weights, const SimpleTensor<half> &bias, const TensorShape &output_shape,
const PadStrideInfo &info, const Size2D &dilation);
template SimpleTensor<qint8_t> convolution_layer(const SimpleTensor<qint8_t> &src, const SimpleTensor<qint8_t> &weights, const SimpleTensor<qint8_t> &bias, const TensorShape &output_shape,
const PadStrideInfo &info, const Size2D &dilation);
template SimpleTensor<qint16_t> convolution_layer(const SimpleTensor<qint16_t> &src, const SimpleTensor<qint16_t> &weights, const SimpleTensor<qint16_t> &bias, const TensorShape &output_shape,
const PadStrideInfo &info, const Size2D &dilation);
template SimpleTensor<uint8_t> convolution_layer(const SimpleTensor<uint8_t> &src, const SimpleTensor<uint8_t> &weights, const SimpleTensor<int32_t> &bias, const TensorShape &output_shape,
const PadStrideInfo &info, const Size2D &dilation);
} // namespace reference
} // namespace validation
} // namespace test
} // namespace arm_compute
|
