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
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
|
/*
* Copyright (c) 2016-2020 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 "arm_compute/core/CL/kernels/CLChannelExtractKernel.h"
#include "arm_compute/core/CL/CLKernelLibrary.h"
#include "arm_compute/core/CL/ICLMultiImage.h"
#include "arm_compute/core/CL/ICLTensor.h"
#include "arm_compute/core/CL/OpenCL.h"
#include "arm_compute/core/Coordinates.h"
#include "arm_compute/core/Error.h"
#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/MultiImageInfo.h"
#include "arm_compute/core/TensorInfo.h"
#include "arm_compute/core/Types.h"
#include "arm_compute/core/Utils.h"
#include "arm_compute/core/Validate.h"
#include "arm_compute/core/Window.h"
#include <set>
#include <string>
using namespace arm_compute;
CLChannelExtractKernel::CLChannelExtractKernel()
: _input(nullptr), _output(nullptr), _num_elems_processed_per_iteration(8), _subsampling(1)
{
}
void CLChannelExtractKernel::configure(const ICLTensor *input, Channel channel, ICLTensor *output)
{
configure(CLKernelLibrary::get().get_compile_context(), input, channel, output);
}
void CLChannelExtractKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, Channel channel, ICLTensor *output)
{
ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
ARM_COMPUTE_ERROR_ON(input == output);
set_format_if_unknown(*output->info(), Format::U8);
// Check if input tensor has a valid format
ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(input, Format::RGB888, Format::RGBA8888, Format::YUYV422, Format::UYVY422);
ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(output, Format::U8);
ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
// Check if channel is valid for given format
const Format format = input->info()->format();
ARM_COMPUTE_ERROR_ON_CHANNEL_NOT_IN_KNOWN_FORMAT(format, channel);
// Half the processed elements for U,V channels due to sub-sampling of 2
_subsampling = 1;
if(format == Format::YUYV422 || format == Format::UYVY422)
{
// Check if the width of the tensor shape is even for formats with subsampled channels (UYVY422 and YUYV422)
ARM_COMPUTE_ERROR_ON_TENSORS_NOT_EVEN(format, input);
if(channel != Channel::Y)
{
_subsampling = 2;
}
}
// Calculate output tensor shape using subsampling
TensorShape output_shape = calculate_subsampled_shape(input->info()->tensor_shape(), format, channel);
set_shape_if_empty(*output->info(), output_shape);
ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
_input = input;
_output = output;
// Create kernel
std::string kernel_name = "channel_extract_" + string_from_format(format);
std::set<std::string> build_opts = { ("-DCHANNEL_" + string_from_channel(channel)) };
_kernel = create_kernel(compile_context, kernel_name, build_opts);
// Configure window
Window win = calculate_max_window(*input->info(), Steps(_num_elems_processed_per_iteration));
AccessWindowHorizontal input_access(input->info(), 0, _num_elems_processed_per_iteration);
AccessWindowRectangle output_access(output->info(), 0, 0, _num_elems_processed_per_iteration, 1, 1.f / _subsampling, 1.f / _subsampling);
update_window_and_padding(win, input_access, output_access);
ValidRegion input_valid_region = input->info()->valid_region();
output_access.set_valid_region(win, ValidRegion(input_valid_region.anchor, output->info()->tensor_shape()));
ICLKernel::configure_internal(win);
}
void CLChannelExtractKernel::configure(const ICLMultiImage *input, Channel channel, ICLImage *output)
{
configure(CLKernelLibrary::get().get_compile_context(), input, channel, output);
}
void CLChannelExtractKernel::configure(const CLCompileContext &compile_context, const ICLMultiImage *input, Channel channel, ICLImage *output)
{
ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
set_format_if_unknown(*output->info(), Format::U8);
// Check if channel is valid for given format
const Format format = input->info()->format();
ARM_COMPUTE_ERROR_ON_CHANNEL_NOT_IN_KNOWN_FORMAT(format, channel);
// Get input plane from the given channel
const ICLImage *input_plane = input->cl_plane(plane_idx_from_channel(format, channel));
ARM_COMPUTE_ERROR_ON_NULLPTR(input_plane);
if(Channel::Y == channel && format != Format::YUV444)
{
// Check if the width of the tensor shape is even for formats with subsampled channels (UYVY422 and YUYV422)
ARM_COMPUTE_ERROR_ON_TENSORS_NOT_EVEN(format, input_plane);
}
// Calculate 2x2 subsampled tensor shape
TensorShape output_shape = calculate_subsampled_shape(input->cl_plane(0)->info()->tensor_shape(), format, channel);
set_shape_if_empty(*output->info(), output_shape);
ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output_shape, output->info()->tensor_shape());
// Check if input tensor has a valid format
ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(input, Format::NV12, Format::NV21, Format::IYUV, Format::YUV444);
ARM_COMPUTE_ERROR_ON_FORMAT_NOT_IN(output, Format::U8);
_output = output;
_input = input_plane;
_subsampling = 1;
// Create kernel
std::string kernel_name;
std::set<std::string> build_opts;
if(Channel::Y == channel || Format::IYUV == format || Format::YUV444 == format)
{
kernel_name = "copy_plane";
}
else
{
kernel_name = "channel_extract_" + string_from_format(format);
build_opts.insert(("-DCHANNEL_" + string_from_channel(channel)));
}
_kernel = create_kernel(compile_context, kernel_name, build_opts);
// Configure window
Window win = calculate_max_window(*input_plane->info(), Steps(_num_elems_processed_per_iteration));
AccessWindowHorizontal input_access(input_plane->info(), 0, _num_elems_processed_per_iteration);
AccessWindowHorizontal output_access(output->info(), 0, _num_elems_processed_per_iteration);
update_window_and_padding(win, input_access, output_access);
output_access.set_valid_region(win, input_plane->info()->valid_region());
ICLKernel::configure_internal(win);
}
void CLChannelExtractKernel::run(const Window &window, cl::CommandQueue &queue)
{
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
Window slice = window.first_slice_window_2D();
do
{
Window win_sub(slice);
win_sub.set(Window::DimX, Window::Dimension(win_sub.x().start() / _subsampling, win_sub.x().end() / _subsampling, win_sub.x().step() / _subsampling));
win_sub.set(Window::DimY, Window::Dimension(win_sub.y().start() / _subsampling, win_sub.y().end() / _subsampling, 1));
unsigned int idx = 0;
add_2D_tensor_argument(idx, _input, slice);
add_2D_tensor_argument(idx, _output, win_sub);
enqueue(queue, *this, slice, lws_hint());
}
while(window.slide_window_slice_2D(slice));
}
|
