diff options
Diffstat (limited to 'src/core/CL/kernels/CLWidthConcatenateLayerKernel.cpp')
-rw-r--r-- | src/core/CL/kernels/CLWidthConcatenateLayerKernel.cpp | 147 |
1 files changed, 0 insertions, 147 deletions
diff --git a/src/core/CL/kernels/CLWidthConcatenateLayerKernel.cpp b/src/core/CL/kernels/CLWidthConcatenateLayerKernel.cpp deleted file mode 100644 index 8eba293487..0000000000 --- a/src/core/CL/kernels/CLWidthConcatenateLayerKernel.cpp +++ /dev/null @@ -1,147 +0,0 @@ -/* - * Copyright (c) 2018-2020 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/CLWidthConcatenateLayerKernel.h" - -#include "arm_compute/core/CL/CLHelpers.h" -#include "arm_compute/core/CL/CLKernelLibrary.h" -#include "arm_compute/core/CL/CLValidate.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/TensorInfo.h" -#include "arm_compute/core/Utils.h" -#include "arm_compute/core/Window.h" -#include "arm_compute/core/utils/misc/ShapeCalculator.h" - -#include "support/StringSupport.h" - -#include <map> - -namespace arm_compute -{ -namespace -{ -constexpr unsigned int num_elems_processed_per_iteration = 16; - -std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, unsigned int width_offset, ITensorInfo *output) -{ - // The window needs to be based on input as we copy all the widths of input - Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration)); - AccessWindowHorizontal input_access(input, 0, num_elems_processed_per_iteration); - AccessWindowHorizontal output_access(output, width_offset, num_elems_processed_per_iteration); - bool window_changed = update_window_and_padding(win, input_access, output_access); - - Window win_collapsed = win.collapse(win, Window::DimZ); - - Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{}; - return std::make_pair(err, win_collapsed); -} -Status validate_arguments(const ITensorInfo *input, unsigned int width_offset, const ITensorInfo *output) -{ - ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output); - ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input); - ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN); - - ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); - ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(0) + width_offset > output->dimension(0)); - - for(size_t i = 1; i < Coordinates::num_max_dimensions; ++i) - { - ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(i) != output->dimension(i)); - } - ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4); - - return Status{}; -} -} // namespace - -CLWidthConcatenateLayerKernel::CLWidthConcatenateLayerKernel() - : _input(nullptr), _output(nullptr), _width_offset(0) -{ -} - -Status CLWidthConcatenateLayerKernel::validate(const ITensorInfo *input, unsigned int width_offset, const ITensorInfo *output) -{ - ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, width_offset, output)); - ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), width_offset, output->clone().get()).first); - return Status{}; -} - -void CLWidthConcatenateLayerKernel::configure(const ICLTensor *input, unsigned int width_offset, ICLTensor *output) -{ - configure(CLKernelLibrary::get().get_compile_context(), input, width_offset, output); -} - -void CLWidthConcatenateLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, unsigned int width_offset, ICLTensor *output) -{ - ARM_COMPUTE_ERROR_ON_NULLPTR(input, output); - ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), width_offset, output->info())); - - _input = input; - _output = output; - _width_offset = width_offset; - - // Add build options - CLBuildOptions build_opts; - build_opts.add_option("-DDATA_TYPE=" + get_underlying_cl_type_from_data_type(input->info()->data_type())); - build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration)); - build_opts.add_option("-DWIDTH_OFFSET=" + support::cpp11::to_string(_width_offset)); - build_opts.add_option("-DDEPTH=" + support::cpp11::to_string(input->info()->dimension(2))); - - if(is_data_type_quantized_asymmetric(input->info()->data_type()) && input->info()->quantization_info() != output->info()->quantization_info()) - { - const UniformQuantizationInfo iqinfo = input->info()->quantization_info().uniform(); - const UniformQuantizationInfo oqinfo = output->info()->quantization_info().uniform(); - - build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iqinfo.offset)); - build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oqinfo.offset)); - build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iqinfo.scale)); - build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oqinfo.scale)); - } - - // Create kernel - _kernel = create_kernel(compile_context, "concatenate_width", build_opts.options()); - // Configure kernel window - auto win_config = validate_and_configure_window(input->info(), width_offset, output->info()); - ARM_COMPUTE_ERROR_THROW_ON(std::get<0>(win_config)); - - ICLKernel::configure_internal(std::get<1>(win_config)); - - // Set output valid region - output->info()->set_valid_region(ValidRegion(Coordinates(), output->info()->tensor_shape())); -} - -void CLWidthConcatenateLayerKernel::run(const Window &window, cl::CommandQueue &queue) -{ - ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); - ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window); - - unsigned int idx = 0; - add_4D_tensor_argument(idx, _input, window); - add_4D_tensor_argument(idx, _output, window); - enqueue(queue, *this, window, lws_hint()); -} -} // namespace arm_compute |