aboutsummaryrefslogtreecommitdiff
path: root/src/core/CL/kernels
diff options
context:
space:
mode:
authorManuel Bottini <manuel.bottini@arm.com>2021-07-14 17:07:23 +0100
committerManuel Bottini <manuel.bottini@arm.com>2021-07-15 09:59:56 +0000
commit7b23732bc8815c7084d4b5f453340fcd740a00fe (patch)
treec7256da469db26c446f4f2d4daa82b223b6ebb25 /src/core/CL/kernels
parente5d76e1574103de405df625e48e5294ea106060c (diff)
downloadComputeLibrary-7b23732bc8815c7084d4b5f453340fcd740a00fe.tar.gz
Port CLCol2ImKernel to ClCol2ImKernel
Resolves: COMPMID-4517 Change-Id: I50cb02116a1ab86fc29200371944c4774e830746 Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5949 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/CL/kernels')
-rw-r--r--src/core/CL/kernels/CLCol2ImKernel.cpp173
-rw-r--r--src/core/CL/kernels/CLCol2ImKernel.h106
2 files changed, 0 insertions, 279 deletions
diff --git a/src/core/CL/kernels/CLCol2ImKernel.cpp b/src/core/CL/kernels/CLCol2ImKernel.cpp
deleted file mode 100644
index 6d7b83471f..0000000000
--- a/src/core/CL/kernels/CLCol2ImKernel.cpp
+++ /dev/null
@@ -1,173 +0,0 @@
-/*
- * 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/CL/kernels/CLCol2ImKernel.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/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.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/StringSupport.h"
-
-#include <cmath>
-
-using namespace arm_compute::misc::shape_calculator;
-
-namespace arm_compute
-{
-namespace
-{
-Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const Size2D &convolved_dims, unsigned int num_groups)
-{
- ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
- ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
- ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::F16, DataType::F32);
-
- // Checks performed when output is configured
- if(output->total_size() != 0)
- {
- ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), compute_col2im_shape(*input, convolved_dims, true, num_groups));
- ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
- ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output);
- ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->data_layout() != DataLayout::NCHW, "Col2Im output's data layout must always be NCHW");
- }
-
- return Status{};
-}
-
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, const Size2D &convolved_dims, unsigned int num_groups)
-{
- ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
- // Output auto inizialitation if not yet initialized
- auto_init_if_empty(*output, input->clone()->set_tensor_shape(compute_col2im_shape(*input, convolved_dims, true, num_groups)).set_data_layout(DataLayout::NCHW));
-
- constexpr unsigned int num_elems_read_per_iteration = 8;
-
- // Configure window
- Window win = calculate_max_window(*input, Steps(num_elems_read_per_iteration));
-
- // Update window and padding just for the input tensor as we cannot access out-of-bounds elements in the output one
- AccessWindowHorizontal input_access(input, 0, num_elems_read_per_iteration);
- bool window_changed = update_window_and_padding(win, input_access);
-
- Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
- return std::make_pair(err, win);
-}
-} // namespace
-
-CLCol2ImKernel::CLCol2ImKernel()
- : _input(nullptr), _output(nullptr), _convolved_dims()
-{
- _type = CLKernelType::ELEMENTWISE;
-}
-
-void CLCol2ImKernel::configure(const ICLTensor *input, ICLTensor *output, const Size2D &convolved_dims, unsigned int num_groups)
-{
- configure(CLKernelLibrary::get().get_compile_context(), input, output, convolved_dims, num_groups);
-}
-
-void CLCol2ImKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const Size2D &convolved_dims, unsigned int num_groups)
-{
- ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-
- // Perform validation step
- ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), convolved_dims, num_groups));
-
- _input = input;
- _output = output;
- _convolved_dims = convolved_dims;
-
- const DataType data_type = input->info()->data_type();
-
- // Create kernel
- CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
- build_opts.add_option("-DELEMENT_SIZE=" + support::cpp11::to_string(input->info()->element_size()));
- build_opts.add_option("-DWIDTH_INPUT=" + support::cpp11::to_string(input->info()->dimension(0)));
- build_opts.add_option("-DWIDTH_OUTPUT=" + support::cpp11::to_string(_convolved_dims.width));
- build_opts.add_option("-DNUM_GROUPS=" + support::cpp11::to_string(num_groups));
-
- _kernel = create_kernel(compile_context, "col2im", build_opts.options());
-
- // Configure kernel window
- auto win_config = validate_and_configure_window(input->info(), output->info(), _convolved_dims, num_groups);
- ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
- ICLKernel::configure_internal(win_config.second);
-
- // Set config_id for enabling LWS tuning
- _config_id = "col2im_";
- _config_id += lower_string(string_from_data_type(input->info()->data_type()));
- _config_id += "_";
- _config_id += support::cpp11::to_string(num_groups);
- _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));
-}
-
-Status CLCol2ImKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const Size2D &convolved_dims, unsigned int num_groups)
-{
- ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
- ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, convolved_dims, num_groups));
- ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), convolved_dims, num_groups).first);
- return Status{};
-}
-
-void CLCol2ImKernel::run(const Window &window, cl::CommandQueue &queue)
-{
- ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
- ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
-
- bool is_collapsed = false;
- bool is_collapsed_out = false;
-
- Window out_window;
- out_window.use_tensor_dimensions(_output->info()->tensor_shape());
-
- Window collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimZ, &is_collapsed);
- Window collapsed_out = out_window.collapse_if_possible(out_window, 3, &is_collapsed_out);
-
- ARM_COMPUTE_ERROR_ON(is_collapsed != is_collapsed_out);
-
- Window slice = collapsed.first_slice_window_3D();
- Window slice_out = collapsed_out.first_slice_window_4D();
- do
- {
- // Set inputs
- unsigned int idx = 0;
- add_3D_tensor_argument(idx, _input, slice);
- add_4D_tensor_argument(idx, _output, slice_out);
- enqueue(queue, *this, slice, lws_hint());
- }
- while(collapsed.slide_window_slice_3D(slice) && collapsed_out.slide_window_slice_4D(slice_out));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLCol2ImKernel.h b/src/core/CL/kernels/CLCol2ImKernel.h
deleted file mode 100644
index 710e048bca..0000000000
--- a/src/core/CL/kernels/CLCol2ImKernel.h
+++ /dev/null
@@ -1,106 +0,0 @@
-/*
- * Copyright (c) 2017-2020 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.
- */
-#ifndef ARM_COMPUTE_CLCOL2IMKERNEL_H
-#define ARM_COMPUTE_CLCOL2IMKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-class ICLTensor;
-
-/** Interface for the col2im reshaping kernel.
- *
- * Rearranges each matrix column into image blocks. It's the inverse operation of @ref CLIm2ColKernel.
- *
- * For example, a vector of 9 elements can be reshaped to a block(image) of 3x3:
- *
- * @f[
- * \left( \begin{array}{ccccccccc}
- * a0 & a1 & a2 & a3 & a4 & a5 & a6 & a7 & a8 \\
- * \end{array} \right)
- * \rightarrow
- * \left( \begin{array}{ccc}
- * a0 & a1 & a2 \\
- * a3 & a4 & a5 \\
- * a6 & a7 & a8 \\
- * \end{array} \right)
- * @f]
- */
-class CLCol2ImKernel : public ICLKernel
-{
-public:
- /** Default constructor */
- CLCol2ImKernel();
- /** Prevent instances of this class from being copied (As this class contains pointers) */
- CLCol2ImKernel(const CLCol2ImKernel &) = delete;
- /** Prevent instances of this class from being copied (As this class contains pointers) */
- CLCol2ImKernel &operator=(const CLCol2ImKernel &) = delete;
- /** Allow instances of this class to be moved */
- CLCol2ImKernel(CLCol2ImKernel &&) = default;
- /** Allow instances of this class to be moved */
- CLCol2ImKernel &operator=(CLCol2ImKernel &&) = default;
- /** Default destructor */
- ~CLCol2ImKernel() = default;
- /** Set the input and output of the kernel.
- *
- * @param[in] input The input tensor to convert. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32
- * @param[out] output The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
- * while the rest represent batch of outputs. Data types supported: Same as @p input. Data layout: NCHW
- * @param[in] convolved_dims Output convolved dimensions.
- * @param[in] num_groups (Optional) Number of groups when performing a grouped convolution
- */
- void configure(const ICLTensor *input, ICLTensor *output, const Size2D &convolved_dims, unsigned int num_groups = 1);
- /** Set the input and output of the kernel.
- *
- * @param[in] compile_context The compile context to be used.
- * @param[in] input The input tensor to convert. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32
- * @param[out] output The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
- * while the rest represent batch of outputs. Data types supported: Same as @p input. Data layout: NCHW
- * @param[in] convolved_dims Output convolved dimensions.
- * @param[in] num_groups (Optional) Number of groups when performing a grouped convolution
- */
- void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, const Size2D &convolved_dims, unsigned int num_groups = 1);
- /** Static function to check if given info will lead to a valid configuration of @ref CLCol2ImKernel
- *
- * @param[in] input The input tensor to convert. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32
- * @param[in] output The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
- * while the rest represent batch of outputs. Data types supported: Same as @p input. Data layout: NCHW
- * @param[in] convolved_dims Output convolved dimensions.
- * @param[in] num_groups (Optional) Number of groups when performing a grouped convolution
- *
- * @return a status
- */
- static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Size2D &convolved_dims, unsigned int num_groups = 1);
-
- // Inherited methods overridden:
- void run(const Window &window, cl::CommandQueue &queue) override;
-
-public:
- const ICLTensor *_input;
- ICLTensor *_output;
- Size2D _convolved_dims;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLCOL2IMKERNEL_H */