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
|
/*
* 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 "arm_compute/core/CL/kernels/CLMinMaxLayerKernel.h"
#include "arm_compute/core/AccessWindowStatic.h"
#include "arm_compute/core/CL/CLHelpers.h"
#include "arm_compute/core/CL/CLKernelLibrary.h"
#include "arm_compute/core/CL/ICLTensor.h"
#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/Validate.h"
#include "arm_compute/core/Window.h"
#include <climits>
using namespace arm_compute;
CLMinMaxLayerKernel::CLMinMaxLayerKernel()
: _input(nullptr), _output(nullptr)
{
}
void CLMinMaxLayerKernel::configure(const ICLTensor *input, ICLTensor *output)
{
ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32);
ARM_COMPUTE_ERROR_ON(input->info()->num_dimensions() < 3);
ARM_COMPUTE_ERROR_ON_NULLPTR(output);
TensorShape output_shape{ input->info()->tensor_shape() };
output_shape.set(Window::DimX, 2);
output_shape.remove_dimension(1);
output_shape.remove_dimension(1);
// Output auto initialization if not yet initialized
auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position());
ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
_input = input;
_output = output;
const unsigned int num_elems_processed_per_iteration = 1;
std::set<std::string> build_opts;
build_opts.emplace("-DWIDTH=" + support::cpp11::to_string(input->info()->dimension(0)));
build_opts.emplace("-DHEIGHT=" + support::cpp11::to_string(input->info()->dimension(1)));
build_opts.emplace("-DDEPTH=" + support::cpp11::to_string(input->info()->dimension(2)));
// Create kernel
_kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("minmax_layer", build_opts));
// Configure kernel 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);
AccessWindowStatic output_access(output->info(), 0, 0, 2, output->info()->dimension(1));
update_window_and_padding(win, input_access, output_access);
output_access.set_valid_region(win, ValidRegion(Coordinates(), output->info()->tensor_shape()));
ICLKernel::configure(win);
}
void CLMinMaxLayerKernel::reset(cl::CommandQueue &queue)
{
_output->map(queue, true);
Window window_output;
window_output.use_tensor_dimensions(_output->info()->tensor_shape());
window_output.set(Window::DimX, Window::Dimension(0, 1, 1));
Iterator output(_output, window_output);
// Reset output
execute_window_loop(window_output, [&](const Coordinates & id)
{
auto *ptr = reinterpret_cast<float *>(output.ptr());
ptr[0] = std::numeric_limits<float>::max();
ptr[1] = std::numeric_limits<float>::min();
},
output);
_output->unmap(queue);
}
void CLMinMaxLayerKernel::run(const Window &window, cl::CommandQueue &queue)
{
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
Window window_collapsed = window.collapse_if_possible(ICLKernel::window(), 3);
Window slice = window_collapsed.first_slice_window_3D();
slice.set(Window::DimX, Window::Dimension(0, 1, 1));
slice.set(Window::DimY, Window::Dimension(0, 1, 1));
slice.set(Window::DimZ, Window::Dimension(0, 1, 1));
do
{
Window output_slice = slice.shift_dimensions(2);
unsigned int idx = 0;
// Set inputs
add_3D_tensor_argument(idx, _input, slice);
add_1D_tensor_argument(idx, _output, output_slice);
enqueue(queue, *this, slice);
}
while(window_collapsed.slide_window_slice_3D(slice));
}
|
