1 /* 2 * Copyright (c) 2017-2022 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_GEMMLOWP_OFFSETCONTRIBUTION_KERNEL_H 25 #define ARM_COMPUTE_CPU_GEMMLOWP_OFFSETCONTRIBUTION_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 used to add the offset contribution after @ref CpuGemmLowpMatrixMultiplyKernel. The computation is performed in-place 37 * 38 * This kernel takes a final int32 accumulator value (the output of @ref CpuGemmLowpMatrixMultiplyKernel), 39 * and adds to it the offset contribution of matrix A and matrix B in-place. 40 * 41 * The final result is: 42 * 43 * mm_result[i][k] = mm_result[i][k] + 44 * (vector_sum_col[k] * a_offset) + 45 * (vector_sum_row[i] * b_offset) + 46 * (a_offset * b_offset * k) 47 * 48 */ 49 class CpuGemmLowpOffsetContributionKernel : public ICpuKernel<CpuGemmLowpOffsetContributionKernel> 50 { 51 public: 52 /** Default constructor */ 53 CpuGemmLowpOffsetContributionKernel() = default; 54 ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuGemmLowpOffsetContributionKernel); 55 /** Initialise the kernel's input and output. 56 * 57 * @param[in, out] mm_result Input tensor containing the result of @ref CpuGemmLowpMatrixMultiplyKernel. Data type supported: S32 58 * @param[in] vector_sum_col Input row-vector of sums of all the entries in each column of matrix B. 59 * Note: vector_sum_col can be a nullptr in case a_offset = 0. Data type supported: same as @p mm_result 60 * @param[in] vector_sum_row Input row-vector of sums of all the entries in each row of matrix A. 61 * Note: vector_sum_row can be a nullptr in case b_offset = 0. Data type supported: same as @p mm_result 62 * @param[in] k Number of matrix A columns or Matrix B rows 63 * @param[in] a_offset Offset to be added to each element of the matrix A. 64 * @param[in] b_offset Offset to be added to each element of the matrix B. 65 */ 66 void configure(ITensorInfo *mm_result, ITensorInfo *vector_sum_col, ITensorInfo *vector_sum_row, int32_t k, int32_t a_offset, int32_t b_offset); 67 /** Static function to check if given info will lead to a valid configuration 68 * 69 * Similar to CpuGemmLowpOffsetContributionKernel::configure() 70 * 71 * @return a status 72 */ 73 static Status validate(const ITensorInfo *mm_result, const ITensorInfo *vector_sum_col, const ITensorInfo *vector_sum_row, int32_t a_offset, int32_t b_offset); 74 75 // Inherited methods overridden: 76 void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override; 77 const char *name() const override; 78 79 private: 80 int32_t _a_offset{ 0 }; 81 int32_t _b_offset{ 0 }; 82 int32_t _k_offset{ 0 }; 83 bool _slide_vector_sum_col{ true }; 84 }; 85 } // namespace kernels 86 } // namespace cpu 87 } // namespace arm_compute 88 #endif /* ARM_COMPUTE_CPU_GEMMLOWP_OFFSETCONTRIBUTION_KERNEL_H */ 89