From 3b0a2654034714c16f5930d2b24936d8be7b18a6 Mon Sep 17 00:00:00 2001 From: Gian Marco Iodice Date: Fri, 7 Dec 2018 11:18:09 +0000 Subject: COMPMID-1775: Implement CLGEMMReshapeRHSMatrixKernel to reshape the RHS matrix of GEMM/GEMMLowp Change-Id: I77f2bfcc5d170bcc2428a2f27104942c1ec877d7 Reviewed-on: https://review.mlplatform.org/375 Reviewed-by: Michele Di Giorgio Tested-by: Arm Jenkins --- arm_compute/core/CL/CLKernels.h | 1 + .../core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h | 86 ++++++++++++++++++++++ 2 files changed, 87 insertions(+) create mode 100644 arm_compute/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h (limited to 'arm_compute/core/CL') diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h index 7bfd44721f..37b92f2d6c 100644 --- a/arm_compute/core/CL/CLKernels.h +++ b/arm_compute/core/CL/CLKernels.h @@ -80,6 +80,7 @@ #include "arm_compute/core/CL/kernels/CLGEMMMatrixMultiplyKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMMatrixVectorMultiplyKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h" +#include "arm_compute/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h" #include "arm_compute/core/CL/kernels/CLGEMMTranspose1xWKernel.h" #include "arm_compute/core/CL/kernels/CLGaussian3x3Kernel.h" #include "arm_compute/core/CL/kernels/CLGaussian5x5Kernel.h" diff --git a/arm_compute/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h b/arm_compute/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h new file mode 100644 index 0000000000..611549a1cb --- /dev/null +++ b/arm_compute/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h @@ -0,0 +1,86 @@ +/* + * Copyright (c) 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_CLGEMMRESHAPERHSMATRIXKERNEL_H__ +#define __ARM_COMPUTE_CLGEMMRESHAPERHSMATRIXKERNEL_H__ + +#include "arm_compute/core/CL/ICLKernel.h" + +namespace arm_compute +{ +class ICLTensor; + +/** OpenCL kernel to reshape the RHS matrix when performing the matrix multiplication + * In particular, this kernel splits the input matrix in blocks of size K0xN0 and stores each one in + * the output matrix unrolling the values */ +class CLGEMMReshapeRHSMatrixKernel : public ICLKernel +{ +public: + /** Default constructor */ + CLGEMMReshapeRHSMatrixKernel(); + /** Prevent instances of this class from being copied (As this class contains pointers) */ + CLGEMMReshapeRHSMatrixKernel(const CLGEMMReshapeRHSMatrixKernel &) = delete; + /** Prevent instances of this class from being copied (As this class contains pointers) */ + CLGEMMReshapeRHSMatrixKernel &operator=(const CLGEMMReshapeRHSMatrixKernel &) = delete; + /** Allow instances of this class to be moved */ + CLGEMMReshapeRHSMatrixKernel(CLGEMMReshapeRHSMatrixKernel &&) = default; + /** Allow instances of this class to be moved */ + CLGEMMReshapeRHSMatrixKernel &operator=(CLGEMMReshapeRHSMatrixKernel &&) = default; + /** Initialise the kernel's input and output. + * + * @param[in] input Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32 + * @param[out] output Output tensor. Data type supported: same as @p input + * @param[in] rhs_info RHS matrix information to be used for reshaping. This object contains all the necessary + * information to reshape the input tensor. Only the following values are supported: + * rhs_info.n0: 2,4,8,16 + * rhs_info.k0: 1,2,4,8,16 (k0 = 1 and k0 = 2 only if rhs_info.transpose = false) + * rhs_info.h0: greater than 0 + * rhs_info.transpose: true, false + * rhs_info.interleave: true, false + */ + void configure(const ICLTensor *input, ICLTensor *output, const GEMMRHSMatrixInfo &rhs_info); + /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMReshapeRHSMatrixKernel + * + * @param[in] input Input tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32 + * @param[in] output Output tensor info which stores the interleaved matrix. Data type supported: same as @p input. + * @param[in] rhs_info RHS matrix information to be used for reshaping. This object contains all the necessary + * information to reshape the input tensor. Only the following values are supported: + * rhs_info.n0: 2,4,8,16 + * rhs_info.k0: 1,2,4,8,16 (k0 = 1 and k0 = 2 only if rhs_info.transpose = false) + * rhs_info.h0: greater than 0 + * rhs_info.transpose: true, false + * rhs_info.interleave: true, false + * + * @return a status + */ + static Status validate(const ITensorInfo *input, const ITensorInfo *output, const GEMMRHSMatrixInfo &rhs_info); + + // Inherited methods overridden + void run(const Window &window, cl::CommandQueue &queue) override; + +private: + const ICLTensor *_input; + ICLTensor *_output; +}; +} // namespace arm_compute +#endif /* __ARM_COMPUTE_CLGEMMRESHAPERHSMATRIXKERNEL_H__ */ \ No newline at end of file -- cgit v1.2.1