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
|
/*
* Copyright (c) 2018 Samsung Electronics Co., Ltd. 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 <OperationsUtils.h>
#include <arm_compute/core/TensorShape.h>
#include <arm_compute/core/TensorInfo.h>
#include "../IO_accessor.h"
#include "../shape.h"
#include "../CLUniqueTensor.h"
#include <cassert>
namespace nnfw {
namespace kernel {
namespace acl {
bool maxPoolFloat32(const float* inputData, const nnfw::rt::Shape& inputShape,
int32_t padding_left, int32_t padding_right,
int32_t padding_top, int32_t padding_bottom,
int32_t stride_width, int32_t stride_height,
int32_t filter_width, int32_t filter_height,
int32_t activation,
float* outputData, const nnfw::rt::Shape& outputShape)
{
arm_compute::TensorShape input_shape = util::fromNNShape(inputShape);
arm_compute::TensorShape output_shape = util::fromNNShape(outputShape);
std::vector<std::shared_ptr<arm_compute::IFunction>> fns;
arm_compute::PadStrideInfo pad_info = arm_compute::PadStrideInfo(stride_width, stride_height,
padding_left, padding_right,
padding_top, padding_bottom,
arm_compute::DimensionRoundingType::FLOOR);
arm_compute::PoolingLayerInfo maxpool_info = arm_compute::PoolingLayerInfo(arm_compute::PoolingType::MAX,
arm_compute::Size2D(filter_width,filter_height),
pad_info, false);
CLUniqueTensor input(arm_compute::TensorInfo(input_shape, arm_compute::Format::F32));
CLUniqueTensor output(arm_compute::TensorInfo(output_shape, arm_compute::Format::F32));
auto pool_f = std::make_shared<arm_compute::CLPoolingLayer>();
pool_f->configure(input.ptr(), output.ptr(), maxpool_info);
fns.emplace_back(pool_f);
input.allocate();
output.allocate();
util::insertFusedActivationLayer<CLUniqueTensor, arm_compute::CLActivationLayer>(output, activation, fns);
TensorAccess<InputAccessor>(input.ref(), inputData, inputShape);
for (const auto &fn : fns)
{
fn->run();
}
arm_compute::CLScheduler::get().sync();
TensorAccess<OutputAccessor>(output.ref(), outputData, outputShape);
return true;
}
bool averagePoolFloat32(const float* inputData, const nnfw::rt::Shape& inputShape,
int32_t padding_left, int32_t padding_right,
int32_t padding_top, int32_t padding_bottom,
int32_t stride_width, int32_t stride_height,
int32_t filter_width, int32_t filter_height,
int32_t activation,
float* outputData, const nnfw::rt::Shape& outputShape)
{
arm_compute::TensorShape input_shape = util::fromNNShape(inputShape);
arm_compute::TensorShape output_shape = util::fromNNShape(outputShape);
std::vector<std::shared_ptr<arm_compute::IFunction>> fns;
arm_compute::PadStrideInfo pad_info = arm_compute::PadStrideInfo(stride_width, stride_height,
padding_left, padding_right,
padding_top, padding_bottom,
arm_compute::DimensionRoundingType::FLOOR);
arm_compute::PoolingLayerInfo pool_info = arm_compute::PoolingLayerInfo(arm_compute::PoolingType::AVG,
arm_compute::Size2D(filter_width,filter_height),
pad_info, true);
CLUniqueTensor input(arm_compute::TensorInfo(input_shape, arm_compute::Format::F32));
CLUniqueTensor output(arm_compute::TensorInfo(output_shape, arm_compute::Format::F32));
auto pool_f = std::make_shared<arm_compute::CLPoolingLayer>();
pool_f->configure(input.ptr(), output.ptr(), pool_info);
fns.emplace_back(pool_f);
input.allocate();
output.allocate();
util::insertFusedActivationLayer<CLUniqueTensor, arm_compute::CLActivationLayer>(output, activation, fns);
TensorAccess<InputAccessor>(input.ref(), inputData, inputShape);
for (const auto &fn : fns)
{
fn->run();
}
arm_compute::CLScheduler::get().sync();
TensorAccess<OutputAccessor>(output.ref(), outputData, outputShape);
return true;
}
} // namespace acl
} // namespace kernel
} // namespace nnfw
|
