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
|
/*
* Copyright (c) 2019 Samsung Electronics Co., Ltd. All Rights Reserved
* Copyright 2017 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.
*/
#ifndef __NNFW_CKER_AVERAGE_POOL_H__
#define __NNFW_CKER_AVERAGE_POOL_H__
#if defined(CKER_OPTIMIZED_EIGEN)
#include "cker/operation/optimized/AveragePool.h"
#endif // defined(CKER_OPTIMIZED_EIGEN)
#include "cker/operation/reference/AveragePool.h"
namespace nnfw
{
namespace cker
{
inline void AveragePool(const PoolParams ¶ms, const Shape &input_shape, const float *input_data,
const Shape &output_shape, float *output_data)
{
#if defined(CKER_OPTIMIZED_EIGEN)
optimized::AveragePool(params, input_shape, input_data, output_shape, output_data);
#else // defined(CKER_OPTIMIZED_EIGEN)
reference::AveragePool(params, input_shape, input_data, output_shape, output_data);
#endif // defined(CKER_OPTIMIZED_EIGEN)
}
inline void AveragePool(const PoolParams ¶ms, const Shape &input_shape,
const uint8_t *input_data, const Shape &output_shape, uint8_t *output_data)
{
assert(params.quantized_activation_min <= params.quantized_activation_max);
assert(input_shape.DimensionsCount() == 4);
assert(output_shape.DimensionsCount() == 4);
const int batches = MatchingDim(input_shape, 0, output_shape, 0);
const int depth = MatchingDim(input_shape, 3, output_shape, 3);
const int input_height = input_shape.Dims(1);
const int input_width = input_shape.Dims(2);
const int output_height = output_shape.Dims(1);
const int output_width = output_shape.Dims(2);
const int stride_height = params.stride_height;
const int stride_width = params.stride_width;
for (int batch = 0; batch < batches; ++batch)
{
for (int out_y = 0; out_y < output_height; ++out_y)
{
for (int out_x = 0; out_x < output_width; ++out_x)
{
const int in_x_origin = (out_x * stride_width) - params.padding_values.width;
const int in_y_origin = (out_y * stride_height) - params.padding_values.height;
// Compute the boundaries of the filter region clamped so as to
// ensure that the filter window fits in the input array.
const int filter_x_start = std::max(0, -in_x_origin);
const int filter_x_end = std::min(params.filter_width, input_width - in_x_origin);
const int filter_y_start = std::max(0, -in_y_origin);
const int filter_y_end = std::min(params.filter_height, input_height - in_y_origin);
int filter_count = (filter_y_end - filter_y_start) * (filter_x_end - filter_x_start);
if (filter_count <= 0)
{
continue;
}
for (int channel = 0; channel < depth; ++channel)
{
int32_t acc = 0;
for (int filter_y = filter_y_start; filter_y < filter_y_end; ++filter_y)
{
for (int filter_x = filter_x_start; filter_x < filter_x_end; ++filter_x)
{
const int in_x = in_x_origin + filter_x;
const int in_y = in_y_origin + filter_y;
acc += input_data[Offset(input_shape, batch, in_y, in_x, channel)];
}
}
acc = (acc + filter_count / 2) / filter_count;
acc = std::max(acc, params.quantized_activation_min);
acc = std::min(acc, params.quantized_activation_max);
output_data[Offset(output_shape, batch, out_y, out_x, channel)] =
static_cast<uint8_t>(acc);
}
}
}
}
}
} // namespace cker
} // namespace nnfw
#endif // __NNFW_CKER_AVERAGE_POOL_H__
|
