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
198
199
200
201
202
203
204
205
206
207
208
209
210
211
|
/*
* 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/CLFastCornersKernel.h"
#include "arm_compute/core/CL/CLKernelLibrary.h"
#include "arm_compute/core/CL/ICLTensor.h"
#include "arm_compute/core/CL/OpenCL.h"
#include "arm_compute/core/Error.h"
#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/IAccessWindow.h"
#include "arm_compute/core/Validate.h"
#include "arm_compute/core/Window.h"
#include "support/StringSupport.h"
#include <set>
#include <string>
using namespace arm_compute;
CLFastCornersKernel::CLFastCornersKernel()
: ICLKernel(), _input(nullptr), _output(nullptr)
{
}
BorderSize CLFastCornersKernel::border_size() const
{
return BorderSize(3);
}
void CLFastCornersKernel::configure(const ICLImage *input, ICLImage *output, float threshold, bool non_max_suppression, BorderMode border_mode)
{
configure(CLKernelLibrary::get().get_compile_context(), input, output, threshold, non_max_suppression, border_mode);
}
void CLFastCornersKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, ICLImage *output, float threshold, bool non_max_suppression, BorderMode border_mode)
{
ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(output);
ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8);
ARM_COMPUTE_ERROR_ON_MSG(border_mode != BorderMode::UNDEFINED, "Not implemented");
_input = input;
_output = output;
// Create build options
std::set<std::string> build_opts;
if(non_max_suppression)
{
build_opts.emplace("-DUSE_MAXSUPPRESSION");
}
// Create kernel
const std::string kernel_name = std::string("fast_corners");
_kernel = create_kernel(compile_context, kernel_name, build_opts);
// Set static kernel arguments
unsigned int idx = 2 * num_arguments_per_2D_tensor(); // Skip the input and output parameters
_kernel.setArg<cl_float>(idx, static_cast<float>(threshold));
// Configure kernel window
constexpr unsigned int num_elems_processed_per_iteration = 1;
constexpr unsigned int num_elems_read_per_iteration = 7;
constexpr unsigned int num_rows_read_per_iteration = 3;
Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_mode == BorderMode::UNDEFINED, BorderSize(3));
AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
AccessWindowRectangle input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
update_window_and_padding(win, input_access, output_access);
output_access.set_valid_region(win, input->info()->valid_region(), border_mode == BorderMode::UNDEFINED, border_size());
ICLKernel::configure_internal(win);
// Set config_id for enabling LWS tuning
_config_id = kernel_name;
_config_id += "_";
_config_id += lower_string(string_from_data_type(input->info()->data_type()));
_config_id += "_";
_config_id += support::cpp11::to_string(input->info()->dimension(0));
_config_id += "_";
_config_id += support::cpp11::to_string(input->info()->dimension(1));
_config_id += "_";
_config_id += support::cpp11::to_string(output->info()->dimension(0));
_config_id += "_";
_config_id += support::cpp11::to_string(output->info()->dimension(1));
_config_id += "_";
_config_id += support::cpp11::to_string(non_max_suppression);
_config_id += "_";
_config_id += lower_string(string_from_border_mode(border_mode));
}
void CLFastCornersKernel::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
{
unsigned int idx = 0;
add_2D_tensor_argument(idx, _input, slice);
add_2D_tensor_argument(idx, _output, slice);
enqueue(queue, *this, slice, lws_hint());
}
while(window.slide_window_slice_2D(slice));
}
CLCopyToArrayKernel::CLCopyToArrayKernel()
: ICLKernel(), _input(nullptr), _corners(nullptr), _num_buffer(nullptr)
{
}
void CLCopyToArrayKernel::configure(const ICLImage *input, bool update_number, ICLKeyPointArray *corners, cl::Buffer *num_buffers)
{
configure(CLKernelLibrary::get().get_compile_context(), input, update_number, corners, num_buffers);
}
void CLCopyToArrayKernel::configure(const CLCompileContext &compile_context, const ICLImage *input, bool update_number, ICLKeyPointArray *corners, cl::Buffer *num_buffers)
{
ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
ARM_COMPUTE_ERROR_ON(corners == nullptr);
ARM_COMPUTE_ERROR_ON(num_buffers == nullptr);
_input = input;
_corners = corners;
_num_buffer = num_buffers;
std::set<std::string> build_opts;
if(update_number)
{
build_opts.emplace("-DUPDATE_NUMBER");
}
// Create kernel
const std::string kernel_name = std::string("copy_to_keypoint");
_kernel = create_kernel(compile_context, kernel_name, build_opts);
//Get how many pixels skipped in the x dimension in the previous stages
unsigned int offset = _input->info()->valid_region().anchor.x();
// Set static kernel arguments
unsigned int idx = num_arguments_per_2D_tensor(); // Skip the input and output parameters
_kernel.setArg<unsigned int>(idx++, _corners->max_num_values());
_kernel.setArg<cl_uint>(idx++, offset);
_kernel.setArg(idx++, *_num_buffer);
_kernel.setArg(idx++, _corners->cl_buffer());
// Configure kernel window
constexpr unsigned int num_elems_processed_per_iteration = 1;
Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
update_window_and_padding(win,
AccessWindowHorizontal(input->info(), 0, num_elems_processed_per_iteration));
ICLKernel::configure_internal(win);
// Set config_id for enabling LWS tuning
_config_id = kernel_name;
_config_id += "_";
_config_id += lower_string(string_from_data_type(input->info()->data_type()));
_config_id += "_";
_config_id += support::cpp11::to_string(input->info()->dimension(0));
_config_id += "_";
_config_id += support::cpp11::to_string(input->info()->dimension(1));
}
void CLCopyToArrayKernel::run(const Window &window, cl::CommandQueue &queue)
{
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
//Initialise the _num_buffer as it used as both input and output
static const unsigned int zero_init = 0;
queue.enqueueWriteBuffer(*_num_buffer, CL_FALSE, 0, sizeof(unsigned int), &zero_init);
Window slice = window.first_slice_window_2D();
do
{
unsigned int idx = 0;
add_2D_tensor_argument(idx, _input, slice);
enqueue(queue, *this, slice, lws_hint());
}
while(window.slide_window_slice_2D(slice));
}
|
