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
|
/*
* Copyright (c) 2019-2021 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 "src/core/CL/kernels/CLDepthwiseConvolutionLayerReshapeWeightsKernel.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/TensorInfo.h"
#include "arm_compute/core/Utils.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "src/core/AccessWindowStatic.h"
#include "src/core/CL/CLValidate.h"
#include "src/core/CL/ICLKernel.h"
#include "src/core/helpers/AutoConfiguration.h"
#include "src/core/helpers/WindowHelpers.h"
#include "support/StringSupport.h"
namespace arm_compute
{
namespace
{
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const DepthwiseConvolutionReshapeInfo &info)
{
ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
const size_t idx_w = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::WIDTH);
const size_t idx_h = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::HEIGHT);
ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
ARM_COMPUTE_RETURN_ERROR_ON_DATA_LAYOUT_NOT_IN(input, DataLayout::NHWC);
ARM_COMPUTE_RETURN_ERROR_ON(info.c0 != 4);
ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(idx_h) != 3);
ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(idx_w) != 3);
ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
if(output->total_size() != 0)
{
auto reshaped_weights_shape = arm_compute::misc::shape_calculator::compute_reshaped_depthwise_weights_shape(*input, info);
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), reshaped_weights_shape);
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output);
}
return Status{};
}
std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, const DepthwiseConvolutionReshapeInfo &info)
{
auto reshaped_input_shape = arm_compute::misc::shape_calculator::compute_reshaped_depthwise_weights_shape(*input, info);
auto_init_if_empty(*output, reshaped_input_shape, 1, input->data_type(), input->quantization_info());
Window win = calculate_max_window(*input, Steps(info.c0));
AccessWindowHorizontal weights_access(input, 0, info.c0);
const bool window_changed = update_window_and_padding(win, weights_access);
Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
return std::make_pair(err, win);
}
} // namespace
CLDepthwiseConvolutionLayerReshapeWeightsKernel::CLDepthwiseConvolutionLayerReshapeWeightsKernel()
: _input(nullptr), _output(nullptr)
{
}
void CLDepthwiseConvolutionLayerReshapeWeightsKernel::configure(const ICLTensor *input, ICLTensor *output, const DepthwiseConvolutionReshapeInfo &info)
{
configure(CLKernelLibrary::get().get_compile_context(), input, output, info);
}
void CLDepthwiseConvolutionLayerReshapeWeightsKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const DepthwiseConvolutionReshapeInfo &info)
{
ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), info));
auto win_config = validate_and_configure_window(input->info(), output->info(), info);
ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
ICLKernel::configure_internal(win_config.second);
_input = input;
_output = output;
// Build the kernel
CLBuildOptions build_opts;
build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(info.c0));
build_opts.add_option("-DDST_WIDTH=" + support::cpp11::to_string(_output->info()->dimension(0)));
build_opts.add_option_if(info.transpose, "-DTRANSPOSE");
build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(input->info()->element_size()));
_kernel = create_kernel(compile_context, "depthwise_convolution_reshape_weights", build_opts.options());
}
Status CLDepthwiseConvolutionLayerReshapeWeightsKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const DepthwiseConvolutionReshapeInfo &info)
{
ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, info));
ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), info).first);
return Status{};
}
void CLDepthwiseConvolutionLayerReshapeWeightsKernel::run(const Window &window, cl::CommandQueue &queue)
{
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
unsigned int idx = 0;
add_3D_tensor_argument(idx, _input, window);
add_2D_tensor_argument(idx, _output, window);
enqueue(queue, *this, window, lws_hint());
}
} // namespace arm_compute
|
