From 57a896172ff77c13655b1dc5acc9cfb2930e0570 Mon Sep 17 00:00:00 2001 From: Gian Marco Iodice Date: Thu, 22 Aug 2019 14:10:27 +0100 Subject: COMPMID-2631: Deprecate CLGEMMMatrixAdditionKernel Change-Id: Ia6c2399ea839a6925493d92ae074e513d7b1b0fc Signed-off-by: Gian Marco Iodice Reviewed-on: https://review.mlplatform.org/c/1792 Reviewed-by: Pablo Marquez Comments-Addressed: Arm Jenkins Tested-by: Arm Jenkins --- arm_compute/core/CL/CLKernels.h | 1 - .../core/CL/kernels/CLGEMMMatrixAdditionKernel.h | 78 ------------- arm_compute/runtime/CL/functions/CLGEMM.h | 2 - docs/00_introduction.dox | 2 +- src/core/CL/kernels/CLGEMMMatrixAdditionKernel.cpp | 125 --------------------- 5 files changed, 1 insertion(+), 207 deletions(-) delete mode 100644 arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h delete mode 100644 src/core/CL/kernels/CLGEMMMatrixAdditionKernel.cpp diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h index e002de7b89..2bb670c275 100644 --- a/arm_compute/core/CL/CLKernels.h +++ b/arm_compute/core/CL/CLKernels.h @@ -87,7 +87,6 @@ #include "arm_compute/core/CL/kernels/CLGEMMLowpQuantizeDownInt32ToUint8ScaleKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMLowpReductionKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.h" -#include "arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.h" diff --git a/arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h b/arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h deleted file mode 100644 index bf8e1d4b17..0000000000 --- a/arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h +++ /dev/null @@ -1,78 +0,0 @@ -/* - * 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. - */ -#ifndef __ARM_COMPUTE_CLGEMMMATRIXADDITIONKERNEL_H__ -#define __ARM_COMPUTE_CLGEMMMATRIXADDITIONKERNEL_H__ - -#include "arm_compute/core/CL/ICLKernel.h" - -namespace arm_compute -{ -class ICLTensor; - -/** OpenCL kernel to perform the in-place matrix addition between 2 matrices, taking into account that the second matrix might be weighted by a scalar value beta. - * The matrices must have the same dimensions - * - * @note This kernel is computed if and only if beta != 0.0. - */ -class CLGEMMMatrixAdditionKernel : public ICLKernel -{ -public: - /** Default constructor */ - CLGEMMMatrixAdditionKernel(); - /** Prevent instances of this class from being copied (As this class contains pointers) */ - CLGEMMMatrixAdditionKernel(const CLGEMMMatrixAdditionKernel &) = delete; - /** Prevent instances of this class from being copied (As this class contains pointers) */ - CLGEMMMatrixAdditionKernel &operator=(const CLGEMMMatrixAdditionKernel &) = delete; - /** Allow instances of this class to be moved */ - CLGEMMMatrixAdditionKernel(CLGEMMMatrixAdditionKernel &&) = default; - /** Allow instances of this class to be moved */ - CLGEMMMatrixAdditionKernel &operator=(CLGEMMMatrixAdditionKernel &&) = default; - /** Initialise the kernel's input, output and beta value - * - * @note The input and output tensors must have the same dimensions - * - * @param[in] input Input tensor (Matrix C). Data types supported: F16/F32 - * @param[in, out] output Output tensor. If this kernel is used to finalize the GEMM result (alpha * AB + beta * C), output must contain the result obtained by @ref CLGEMMMatrixMultiplyKernel. Data type supported: same as @p input - * @param[in] beta Weight of matrix C - */ - void configure(const ICLTensor *input, ICLTensor *output, float beta); - /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMMatrixAdditionKernel. - * - * @param[in] input Input tensor (Matrix C). Data types supported: F16/F32 - * @param[in] output Output tensor. If this kernel is used to finalize the GEMM result (alpha * AB + beta * C), output must contain the result obtained by @ref CLGEMMMatrixMultiplyKernel. Data type supported: same as @p input - * @param[in] beta Weight of matrix C - * - * @return a status - */ - static Status validate(const ITensorInfo *input, const ITensorInfo *output, float beta); - - // Inherited methods overridden: - void run(const Window &window, cl::CommandQueue &queue) override; - -private: - const ICLTensor *_input; - ICLTensor *_output; -}; -} // namespace arm_compute -#endif /* __ARM_COMPUTE_CLGEMMMATRIXADDITIONKERNEL_H__ */ diff --git a/arm_compute/runtime/CL/functions/CLGEMM.h b/arm_compute/runtime/CL/functions/CLGEMM.h index e2a92a8a37..7efeb129db 100644 --- a/arm_compute/runtime/CL/functions/CLGEMM.h +++ b/arm_compute/runtime/CL/functions/CLGEMM.h @@ -24,7 +24,6 @@ #ifndef __ARM_COMPUTE_CLGEMM_H__ #define __ARM_COMPUTE_CLGEMM_H__ -#include "arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMMatrixMultiplyReshapedOnlyRHSKernel.h" @@ -46,7 +45,6 @@ class ICLTensor; * -# @ref CLGEMMMatrixMultiplyKernel (only if either the NATIVE or RESHAPED_V1 is selected by the select_gemm_type method()) * -# @ref CLGEMMMatrixMultiplyReshapedKernel (only if RESHAPED_V1 is selected by the select_gemm_type method()) * -# @ref CLGEMMMatrixMultiplyReshapedOnlyRHSKernel (only if RESHAPED_ONLY_RHS is selected by the select_gemm_type method()) - * -# @ref CLGEMMMatrixAdditionKernel (if c != nullptr and beta != 0.0) * */ class CLGEMM : public IFunction diff --git a/docs/00_introduction.dox b/docs/00_introduction.dox index 2013185e1c..92a4b9c954 100644 --- a/docs/00_introduction.dox +++ b/docs/00_introduction.dox @@ -815,7 +815,7 @@ v17.03.1 First Major public release of the sources v17.03 Sources preview - New OpenCL kernels / functions: - @ref CLGradientKernel, @ref CLEdgeNonMaxSuppressionKernel, @ref CLEdgeTraceKernel / @ref CLCannyEdge - - GEMM refactoring + FP16 support: CLGEMMInterleave4x4Kernel, CLGEMMTranspose1xWKernel, @ref CLGEMMMatrixMultiplyKernel, @ref CLGEMMMatrixAdditionKernel / @ref CLGEMM + - GEMM refactoring + FP16 support: CLGEMMInterleave4x4Kernel, CLGEMMTranspose1xWKernel, @ref CLGEMMMatrixMultiplyKernel, CLGEMMMatrixAdditionKernel / @ref CLGEMM - @ref CLGEMMMatrixAccumulateBiasesKernel / @ref CLFullyConnectedLayer - @ref CLTransposeKernel / @ref CLTranspose - @ref CLLKTrackerInitKernel, @ref CLLKTrackerStage0Kernel, @ref CLLKTrackerStage1Kernel, @ref CLLKTrackerFinalizeKernel / @ref CLOpticalFlow diff --git a/src/core/CL/kernels/CLGEMMMatrixAdditionKernel.cpp b/src/core/CL/kernels/CLGEMMMatrixAdditionKernel.cpp deleted file mode 100644 index 803ed30d84..0000000000 --- a/src/core/CL/kernels/CLGEMMMatrixAdditionKernel.cpp +++ /dev/null @@ -1,125 +0,0 @@ -/* - * Copyright (c) 2017-2019 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/CLGEMMMatrixAdditionKernel.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/Types.h" -#include "arm_compute/core/Window.h" - -using namespace arm_compute; - -namespace -{ -std::pair validate_and_configure_window(ITensorInfo *input, ITensorInfo *output) -{ - const unsigned int num_elems_processed_per_iteration = max_cl_vector_width / data_size_from_type(input->data_type()); - // Configure kernel window - 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, 0, num_elems_processed_per_iteration); - - bool window_changed = update_window_and_padding(win, input_access, output_access); - - output_access.set_valid_region(win, input->valid_region()); - - Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{}; - return std::make_pair(err, win); -} -} // namespace - -namespace -{ -Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, float beta) -{ - ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output); - ARM_COMPUTE_UNUSED(beta); - - ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input); - ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32); - ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); - ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output); - - return Status{}; -} -} // namespace - -CLGEMMMatrixAdditionKernel::CLGEMMMatrixAdditionKernel() - : _input(nullptr), _output(nullptr) -{ -} - -void CLGEMMMatrixAdditionKernel::configure(const ICLTensor *input, ICLTensor *output, float beta) -{ - ARM_COMPUTE_ERROR_ON_NULLPTR(input, output); - - // Perform validation step - ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), beta)); - - _input = input; - _output = output; - - // Create build options - CLBuildOptions build_opts; - build_opts.add_option("-DBETA=" + float_to_string_with_full_precision(beta)); - - // Create kernel - std::string data_type_name = lower_string(string_from_data_type(input->info()->data_type())); - _kernel = static_cast(CLKernelLibrary::get().create_kernel(("gemm_ma_" + data_type_name), build_opts.options())); - - // Configure kernel window - auto win_config = validate_and_configure_window(input->info(), output->info()); - ARM_COMPUTE_ERROR_THROW_ON(win_config.first); - ICLKernel::configure_internal(win_config.second); -} - -Status CLGEMMMatrixAdditionKernel::validate(const ITensorInfo *input, const ITensorInfo *output, float beta) -{ - ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, beta)); - ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get()).first); - return Status{}; -} - -void CLGEMMMatrixAdditionKernel::run(const Window &window, cl::CommandQueue &queue) -{ - ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); - ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window); - - Window slice = window.first_slice_window_3D(); - - do - { - unsigned int idx = 0; - add_3D_tensor_argument(idx, _input, slice); - add_3D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); - } - while(window.slide_window_slice_3D(slice)); -} -- cgit v1.2.1