1 /* 2 * Copyright (c) 2016-2022 Arm Limited. 3 * 4 * SPDX-License-Identifier: MIT 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to 8 * deal in the Software without restriction, including without limitation the 9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or 10 * sell copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in all 14 * copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 22 * SOFTWARE. 23 */ 24 #ifndef ARM_COMPUTE_CPU_GEMM_MATRIX_ADDITION_KERNEL_H 25 #define ARM_COMPUTE_CPU_GEMM_MATRIX_ADDITION_KERNEL_H 26 27 #include "src/core/common/Macros.h" 28 #include "src/cpu/ICpuKernel.h" 29 30 namespace arm_compute 31 { 32 namespace cpu 33 { 34 namespace kernels 35 { 36 /** 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: 37 * 38 * @note [ MTX_OUT = MTX_0 + beta * MTX_1 ] with MTX_0 and MTX_1 of the same size 39 * 40 * @note This stage is used to finalize the GEMM result and it is computed if and only if beta != 0.0. In case this kernel is used for finalizing GEMM result, we have: 41 * - MTX_0 = A * B * alpha, where MTX_0 is the output of @ref CpuGemmMatrixMultiplyKernel 42 * - MTX_1 = C 43 */ 44 class CpuGemmMatrixAdditionKernel : public ICpuKernel<CpuGemmMatrixAdditionKernel> 45 { 46 private: 47 using GemmMatrixAddKernelPtr = std::add_pointer<void(const ITensor *, ITensor *, const Window &, float)>::type; 48 49 public: 50 struct GemmMatrixAddKernel 51 { 52 const char *name; 53 const DataTypeISASelectorPtr is_selected; 54 GemmMatrixAddKernelPtr ukernel; 55 }; 56 CpuGemmMatrixAdditionKernel() = default; 57 ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuGemmMatrixAdditionKernel); 58 /** Initialise the kernel's input and output. 59 * 60 * @note The input and output tensor must have the same dimensions 61 * 62 * @param[in] src Input tensor info (Matrix C). Data types supported: F16/F32 63 * @param[in, out] dst Output tensor info. If this kernel is used to finalize the GEMM result, output contains the result obtained by the kernel @ref CpuGemmMatrixMultiplyKernel. Data type supported: the same as @p src. 64 * @param[in] beta Weight of matrix C 65 */ 66 void configure(const ITensorInfo *src, ITensorInfo *dst, float beta); 67 /** Static function to check if given info will lead to a valid configuration of @ref CpuGemmMatrixAdditionKernel. 68 * 69 * @note The input and output tensor must have the same dimensions 70 * 71 * Similar to @ref CpuGemmMatrixAdditionKernel::configure() 72 * 73 * @return a status 74 */ 75 static Status validate(const ITensorInfo *src, const ITensorInfo *dst, float beta); 76 77 // Inherited methods overridden: 78 void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override; 79 const char *name() const override; 80 81 static const std::vector<GemmMatrixAddKernel> &get_available_kernels(); 82 83 private: 84 /** Common signature for all the matrix addition functions 85 * 86 * @param[in] src An input tensor. Data types supported: F16/F32 87 * @param[out] dst The output tensor. Data type supported: same as @p src 88 * @param[in] window Region on which to execute the kernel. 89 * @param[in] beta Weight of matrix C 90 */ 91 /** Matrix addition function to use for the particular tensor types passed to configure() */ 92 GemmMatrixAddKernelPtr _func{ nullptr }; 93 float _beta{ 0.f }; 94 }; 95 } // namespace kernels 96 } // namespace cpu 97 } // namespace arm_compute 98 #endif /* ARM_COMPUTE_CPU_GEMM_MATRIX_ADDITION_KERNEL_H */ 99