1 /* 2 * Copyright (c) 2017-2021 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_CL_GEMMLOWP_OFFSET_CONTRIBUTION_KERNEL_H 25 #define ARM_COMPUTE_CL_GEMMLOWP_OFFSET_CONTRIBUTION_KERNEL_H 26 27 #include "src/core/common/Macros.h" 28 #include "src/gpu/cl/ClCompileContext.h" 29 #include "src/gpu/cl/IClKernel.h" 30 31 namespace arm_compute 32 { 33 namespace opencl 34 { 35 namespace kernels 36 { 37 /** OpenCL kernel used to add the offset contribution after the matrix multiplication. The computation is performed in-place 38 * 39 * This kernel takes a final int32 accumulator value (the output of the matrix multiplication), 40 * and adds to it the offset contribution of matrix A and matrix B in-place. 41 * 42 * The final result is: 43 * 44 * mm_result[i][k] = mm_result[i][k] + 45 * (vector_sum_col[k] * a_offset) + 46 * (vector_sum_row[i] * b_offset) + 47 * (a_offset * b_offset * k) 48 * 49 */ 50 class ClGemmLowpOffsetContributionKernel : public IClKernel 51 { 52 public: 53 ClGemmLowpOffsetContributionKernel(); 54 ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClGemmLowpOffsetContributionKernel); 55 /** Initialise the kernel's input and output. 56 * 57 * @param[in] compile_context The compile context to be used. 58 * @param[in, out] mm_result Input tensor containing the result of the matrix multiplication. Data type supported: S32 59 * @param[in] vector_sum_col Input row-vector of sums of all the entries in each column of matrix B. 60 * Note: vector_sum_col can be a nullptr in case a_offset = 0. Data type supported: same as @p mm_result 61 * @param[in] vector_sum_row Input row-vector of sums of all the entries in each row of matrix A. 62 * Note: vector_sum_row can be a nullptr in case b_offset = 0. Data type supported: same as @p mm_result 63 * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the addition of biases is not required. 64 * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. 65 * @param[in] k Number of matrix A columns or Matrix B rows 66 * @param[in] a_offset Offset to be added to each element of the matrix A. 67 * @param[in] b_offset Offset to be added to each element of the matrix B. 68 */ 69 void configure(const CLCompileContext &compile_context, 70 const ITensorInfo *mm_result, const ITensorInfo *vector_sum_col, const ITensorInfo *vector_sum_row, const ITensorInfo *bias, 71 int32_t k, int32_t a_offset, int32_t b_offset); 72 /** Static function to check if given info will lead to a valid configuration 73 * 74 * Similar to @ref ClGemmLowpOffsetContributionKernel::configure() 75 * 76 * @return a status 77 */ 78 static Status validate(const ITensorInfo *mm_result, const ITensorInfo *vector_sum_col, const ITensorInfo *vector_sum_row, const ITensorInfo *bias, int32_t a_offset, int32_t b_offset); 79 80 // Inherited methods overridden: 81 void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override; 82 }; 83 } // namespace kernels 84 } // namespace opencl 85 } // namespace arm_compute 86 #endif /* ARM_COMPUTE_CL_GEMMLOWP_OFFSET_CONTRIBUTION_KERNEL_H */ 87