1 /* 2 * Copyright (c) 2017-2020 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_NEGEMMLOWPOFFSETCONTRIBUTIONKERNEL_H 25 #define ARM_COMPUTE_NEGEMMLOWPOFFSETCONTRIBUTIONKERNEL_H 26 27 #include "src/core/NEON/INEKernel.h" 28 29 namespace arm_compute 30 { 31 class ITensor; 32 33 /** NEON kernel used to add the offset contribution after @ref NEGEMMLowpMatrixMultiplyKernel. The computation is performed in-place 34 * 35 * This kernel takes a final int32 accumulator value (the output of @ref NEGEMMLowpMatrixMultiplyKernel), 36 * and adds to it the offset contribution of matrix A and matrix B in-place. 37 * 38 * The final result is: 39 * 40 * mm_result[i][k] = mm_result[i][k] + 41 * (vector_sum_col[k] * a_offset) + 42 * (vector_sum_row[i] * b_offset) + 43 * (a_offset * b_offset * k) 44 * 45 */ 46 class NEGEMMLowpOffsetContributionKernel : public INEKernel 47 { 48 public: name()49 const char *name() const override 50 { 51 return "NEGEMMLowpOffsetContributionKernel"; 52 } 53 /** Constructor */ 54 NEGEMMLowpOffsetContributionKernel(); 55 /** Prevent instances of this class from being copied (As this class contains pointers)*/ 56 NEGEMMLowpOffsetContributionKernel(const NEGEMMLowpOffsetContributionKernel &) = delete; 57 /** Prevent instances of this class from being copied (As this class contains pointers)*/ 58 NEGEMMLowpOffsetContributionKernel &operator=(const NEGEMMLowpOffsetContributionKernel &) = delete; 59 /** Allow instances of this class to be moved */ 60 NEGEMMLowpOffsetContributionKernel(NEGEMMLowpOffsetContributionKernel &&) = default; 61 /** Allow instances of this class to be moved */ 62 NEGEMMLowpOffsetContributionKernel &operator=(NEGEMMLowpOffsetContributionKernel &&) = default; 63 /** Default destructor */ 64 ~NEGEMMLowpOffsetContributionKernel() = default; 65 /** Initialise the kernel's input and output. 66 * 67 * @param[in, out] mm_result Input tensor containing the result of @ref NEGEMMLowpMatrixMultiplyKernel. Data type supported: S32 68 * @param[in] vector_sum_col Input row-vector of sums of all the entries in each column of matrix B. 69 * Note: vector_sum_col can be a nullptr in case a_offset = 0. Data type supported: same as @p mm_result 70 * @param[in] vector_sum_row Input row-vector of sums of all the entries in each row of matrix A. 71 * Note: vector_sum_row can be a nullptr in case b_offset = 0. Data type supported: same as @p mm_result 72 * @param[in] k Number of matrix A columns or Matrix B rows 73 * @param[in] a_offset Offset to be added to each element of the matrix A. 74 * @param[in] b_offset Offset to be added to each element of the matrix B. 75 */ 76 void configure(ITensor *mm_result, const ITensor *vector_sum_col, const ITensor *vector_sum_row, int32_t k, int32_t a_offset, int32_t b_offset); 77 /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMLowpOffsetContributionKernel 78 * 79 * @param[in] mm_result Input tensor containing the result of @ref NEGEMMLowpMatrixMultiplyKernel. Data type supported: S32 80 * @param[in] vector_sum_col Input row-vector of sums of all the entries in each column of matrix B. 81 * Note: vector_sum_col can be a nullptr in case a_offset = 0. Data type supported: same as @p mm_result 82 * @param[in] vector_sum_row Input row-vector of sums of all the entries in each row of matrix A. 83 * Note: vector_sum_row can be a nullptr in case b_offset = 0. Data type supported: same as @p mm_result 84 * @param[in] a_offset Offset to be added to each element of the matrix A. 85 * @param[in] b_offset Offset to be added to each element of the matrix B. 86 * 87 * @return a status 88 */ 89 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); 90 91 // Inherited methods overridden: 92 void run(const Window &window, const ThreadInfo &info) override; 93 94 private: 95 const ITensor *_vector_sum_col; 96 const ITensor *_vector_sum_row; 97 ITensor *_mm_result; 98 int32_t _a_offset; 99 int32_t _b_offset; 100 int32_t _k_offset; 101 bool _slide_vector_sum_col; 102 }; 103 } // namespace arm_compute 104 105 #endif /* ARM_COMPUTE_NEGEMMLOWPOFFSETCONTRIBUTIONKERNEL_H */ 106