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
|
/*
* Copyright (c) 2017-2021 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/gpu/cl/kernels/ClTransposeKernel.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/Validate.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "src/core/CL/CLValidate.h"
#include "src/core/helpers/AutoConfiguration.h"
#include "src/core/helpers/WindowHelpers.h"
#include "support/Cast.h"
#include "support/StringSupport.h"
namespace arm_compute
{
namespace opencl
{
namespace kernels
{
void ClTransposeKernel::configure(const CLCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst)
{
ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);
// Output auto initialization if not yet initialized
const TensorShape dst_shape = misc::shape_calculator::compute_transposed_shape(*src);
auto_init_if_empty(*dst, src->clone()->set_tensor_shape(dst_shape));
ARM_COMPUTE_ERROR_THROW_ON(ClTransposeKernel::validate(src, dst));
auto padding_info = get_padding_info({ src, dst });
// Create kernel
const unsigned int vec_size_x = adjust_vec_size(max_cl_vector_width / src->element_size(), src->dimension(0));
const int vec_size_x_leftovers = src->dimension(0) % vec_size_x;
const unsigned int vec_size_y = adjust_vec_size(max_cl_vector_width / src->element_size(), src->dimension(1));
const int vec_size_y_leftovers = src->dimension(1) % vec_size_y;
CLBuildOptions build_opts;
build_opts.add_option("-DDATA_TYPE_IN_BYTES=" + support::cpp11::to_string(src->element_size()));
build_opts.add_option("-DVEC_SIZE_X=" + support::cpp11::to_string(vec_size_x));
build_opts.add_option("-DVEC_SIZE_LEFTOVER_X=" + support::cpp11::to_string(vec_size_x_leftovers));
build_opts.add_option("-DVEC_SIZE_Y=" + support::cpp11::to_string(vec_size_y));
build_opts.add_option("-DVEC_SIZE_LEFTOVER_Y=" + support::cpp11::to_string(vec_size_y_leftovers));
_kernel = create_kernel(compile_context, "transpose", build_opts.options());
// Configure kernel window
Window win = calculate_max_window(*src, Steps(vec_size_x, vec_size_y));
ICLKernel::configure_internal(win, cl::NDRange(2, 8));
ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
}
Status ClTransposeKernel::validate(const ITensorInfo *src, const ITensorInfo *dst)
{
ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, dst);
ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src);
ARM_COMPUTE_RETURN_ERROR_ON(src->data_type() == DataType::UNKNOWN);
ARM_COMPUTE_RETURN_ERROR_ON_MSG(src->num_dimensions() > 2, "Transpose up to 2-D src tensor is supported");
// Validate configured dst
if(dst->total_size() != 0)
{
const TensorInfo dst_info = src->clone()->set_tensor_shape(misc::shape_calculator::compute_transposed_shape(*src));
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(dst, &dst_info);
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(src, dst);
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst);
}
return Status{};
}
void ClTransposeKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
{
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
auto dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
Window slice = window.first_slice_window_2D();
do
{
unsigned int idx = 0;
add_2D_tensor_argument(idx, src, slice);
add_2D_tensor_argument(idx, dst, slice);
enqueue(queue, *this, slice, lws_hint());
}
while(window.slide_window_slice_2D(slice));
}
} // namespace kernels
} // namespace opencl
} // namespace arm_compute
|
