diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-08-16 12:38:54 +0100 |
---|---|---|
committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-08-20 16:26:16 +0000 |
commit | 19884630c37ae9de2f65a88ea2cda5630a551bad (patch) | |
tree | 0fdffe84c5d66ffe8a1d320b798a247afa1dfdae /src/runtime/gpu/cl/operators/ClGemmConvolution.h | |
parent | 73df9310f4e94e43597f283307e3cde0653d8bae (diff) | |
download | ComputeLibrary-19884630c37ae9de2f65a88ea2cda5630a551bad.tar.gz |
Rename [Cl|Cpu]GemmConvolution to [Cl|Gpu]GemmConv2d
Renaming the gemm-based convolution operators to accomodate for new
operators with higher convolution dimensonality
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: Id2f2cf11404221f0e87baa0e5d08ad5d63eaf78e
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6113
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/runtime/gpu/cl/operators/ClGemmConvolution.h')
-rw-r--r-- | src/runtime/gpu/cl/operators/ClGemmConvolution.h | 185 |
1 files changed, 0 insertions, 185 deletions
diff --git a/src/runtime/gpu/cl/operators/ClGemmConvolution.h b/src/runtime/gpu/cl/operators/ClGemmConvolution.h deleted file mode 100644 index 444516eaaa..0000000000 --- a/src/runtime/gpu/cl/operators/ClGemmConvolution.h +++ /dev/null @@ -1,185 +0,0 @@ -/* - * Copyright (c) 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. - */ -#ifndef ARM_COMPUTE_CL_GEMMCONVOLUTION_H -#define ARM_COMPUTE_CL_GEMMCONVOLUTION_H - -#include "arm_compute/core/TensorInfo.h" -#include "arm_compute/core/Types.h" -#include "arm_compute/runtime/FunctionDescriptors.h" -#include "src/core/gpu/cl/ClCompileContext.h" -#include "src/runtime/gpu/cl/IClOperator.h" - -#include <memory> - -namespace arm_compute -{ -namespace opencl -{ -class ClGemm; -class ClGemmLowpMatrixMultiplyCore; -namespace kernels -{ -class ClIm2ColKernel; -class ClCol2ImKernel; -class ClWeightsReshapeKernel; -class ClActivationKernel; -} // namespace kernels - -/** Basic function to compute the convolution layer. This function calls the following OpenCL kernels/functions: - * - * -# @ref opencl::kernels::ClIm2ColKernel - * -# @ref ClGemm (if the data type is FP32 or FP16) - * -# @ref CLGEMMLowpMatrixMultiplyCore (if the data type is QASYMM8/QASYMM8_SIGNED) - * -# @ref ClGemmLowpOutputStage with QUANTIZE_DOWN_FIXEDPOINT type of quantization (if the data type is QASYMM8/QASYMM8_SIGNED) - * -# @ref opencl::kernels::ClCol2ImKernel (if NCHW data layout) - * -# @ref opencl::kernels::ClActivationKernel - */ -class ClGemmConvolution : public IClOperator -{ -public: - /** Constructor */ - ClGemmConvolution(); - /** Prevent instances of this class from being copied (As this class contains pointers) */ - ClGemmConvolution(const ClGemmConvolution &) = delete; - /** Default move constructor */ - ClGemmConvolution(ClGemmConvolution &&) = default; - /** Prevent instances of this class from being copied (As this class contains pointers) */ - ClGemmConvolution &operator=(const ClGemmConvolution &) = delete; - /** Default move assignment operator */ - ClGemmConvolution &operator=(ClGemmConvolution &&) = default; - /**Default destructor */ - ~ClGemmConvolution(); - /** Set the input and output tensors. - * - * Valid data layouts: - * - NHWC - * - NCHW - * - * Valid data type configurations: - * |src0 |src1 |src2 |dst | - * |:--------------|:------------------|:--------|:--------------| - * |F16 |F16 |F16 |F16 | - * |F32 |F32 |F32 |F32 | - * |QASYMM8 |QASYMM8 |S32 |QASYMM8 | - * |QASYMM8 |QSYMM8_PER_CHANNEL |S32 |QASYMM8 | - * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |QASYMM8_SIGNED | - * |QASYMM8_SIGNED |QSYMM8_PER_CHANNEL |S32 |QASYMM8_SIGNED | - * - * @param[in] compile_context The compile context to be used. - * @param[in] src Source tensor info. 3 lower dimensions represent a single input [width, height, IFM], - * while every optional dimension from 4 and above represent a batch of inputs. - * Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] weights Weights tensor info. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. - * Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8 or QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8_SIGNED. - * @param[in] biases Biases tensor info. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of quantized type where biases should be of S32 type. - * @param[out] dst Destination tensor info. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs. - * Data types supported: Same as @p input. - * @param[in] conv2d_info Contains convolution 2d info described in @ref Conv2dInfo. - * @param[in] weights_info Specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel. If this is not part of the fully connected layer the weights - * tensor has also been transposed with CLGEMMReshapeRHSMatrixKernel. Data type supported: Same as @p input. - */ - void configure(const ClCompileContext &compile_context, ITensorInfo *src, ITensorInfo *weights, ITensorInfo *biases, ITensorInfo *dst, const Conv2dInfo &conv2d_info, - const WeightsInfo &weights_info = WeightsInfo()); - /** Static function to check if given info will lead to a valid configuration - * - * Similar to ClGemmConvolution::configure() - * - * @return a status - */ - static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const Conv2dInfo &conv2d_info, - const WeightsInfo &weights_info = WeightsInfo()); - - // Inherited methods overridden: - void run(ITensorPack &tensors) override; - void prepare(ITensorPack &constants) override; - experimental::MemoryRequirements workspace() const override; - -private: - /** Configures the appropriate matrix multiply routine - * - * @param[in] compile_context The compile context to be used. - * @param[in] src Input tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] weights Weights tensor info. Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8 or - * QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8_SIGNED. - * @param[in] biases Biases tensor info. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of quantized type where biases should be of S32 type. - * @param[in, out] dst Output tensor info. Data types supported: same as @p input. - * @param[in] gemmlowp_output_stage GEMMLowp output stage info - * @param[in] gemm_3d_depth Depth of GEMM 3D - * @param[in] act_info Activation to apply after the matrix multiplication - */ - void configure_mm(const CLCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *weights, ITensorInfo *biases, ITensorInfo *dst, - const GEMMLowpOutputStageInfo &gemmlowp_output_stage, - int gemm_3d_depth, const ActivationLayerInfo &act_info); - /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMConvolutionLayer matrix multiply routines - * - * @param[in] src Input tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] weights Weights tensor info. Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8 or - * QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8_SIGNED. - * @param[in] biases Biases tensor info. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of quantized type where biases should be of S32 type. - * @param[in] dst Output tensor info. Data types supported: same as @p input. - * @param[in] gemmlowp_output_stage GEMMLowp output stage info - * @param[in] gemm_3d_depth Depth of GEMM 3D - * @param[in] skip_im2col Flag which specifies if im2col has to be skipped. i.e. 1x1 convolution with NHWC data layout. - * @param[in] act_info Activation to apply after the matrix multiplication - * - * @return a status - */ - static Status validate_mm(const ITensorInfo *src, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *dst, const GEMMLowpOutputStageInfo &gemmlowp_output_stage, - int gemm_3d_depth, bool skip_im2col, const ActivationLayerInfo &act_info); - - enum AuxTensorIdx - { - // ClGemmLowpMatrixMultiplyCore has up to 7 internal tensors - Im2ColOutput = 8, - WeightsReshaped, - GemmOutput, - Count - }; - - std::unique_ptr<kernels::ClWeightsReshapeKernel> _weights_reshape_kernel; - std::unique_ptr<kernels::ClIm2ColKernel> _im2col_kernel; - std::unique_ptr<ClGemm> _mm_gemm; - std::unique_ptr<ClGemmLowpMatrixMultiplyCore> _mm_gemmlowp; - std::unique_ptr<opencl::kernels::ClCol2ImKernel> _col2im_kernel; - std::unique_ptr<kernels::ClActivationKernel> _activation_kernel; - - TensorInfo _im2col_output; - TensorInfo _weights_reshaped; - TensorInfo _gemm_output; - - bool _skip_im2col; - bool _skip_col2im; - bool _is_quantized; - bool _fuse_activation; - bool _append_bias; - bool _is_prepared; - - experimental::MemoryRequirements _aux_mem; -}; -} // namespace opencl -} // namespace arm_compute -#endif /* ARM_COMPUTE_CL_GEMMCONVOLUTION_H */ |