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_NEGEMMLOWPMATRIXMULTIPLYKERNEL_H 25 #define ARM_COMPUTE_NEGEMMLOWPMATRIXMULTIPLYKERNEL_H 26 27 #include "src/core/NEON/INEKernel.h" 28 29 namespace arm_compute 30 { 31 class ITensor; 32 33 /** NEON kernel to multiply matrices 34 * 35 * @note @ref NEGEMMLowpMatrixMultiplyKernel low precision matrix product kernel 36 * This kernel performs the following computation: 37 * 38 * -# Convert a values from int8 to int32 39 * -# Convert b values from int8 to int32 40 * -# Compute the int32 matrix product of the resulting a * b and store the result as int32 41 * 42 */ 43 class NEGEMMLowpMatrixMultiplyKernel : public INEKernel 44 { 45 public: name()46 const char *name() const override 47 { 48 return "NEGEMMLowpMatrixMultiplyKernel"; 49 } 50 /** Constructor */ 51 NEGEMMLowpMatrixMultiplyKernel(); 52 /** Prevent instances of this class from being copied (As this class contains pointers)*/ 53 NEGEMMLowpMatrixMultiplyKernel(const NEGEMMLowpMatrixMultiplyKernel &) = delete; 54 /** Prevent instances of this class from being copied (As this class contains pointers)*/ 55 NEGEMMLowpMatrixMultiplyKernel &operator=(const NEGEMMLowpMatrixMultiplyKernel &) = delete; 56 /** Allow instances of this class to be moved */ 57 NEGEMMLowpMatrixMultiplyKernel(NEGEMMLowpMatrixMultiplyKernel &&) = default; 58 /** Allow instances of this class to be moved */ 59 NEGEMMLowpMatrixMultiplyKernel &operator=(NEGEMMLowpMatrixMultiplyKernel &&) = default; 60 /** Default destructor */ 61 ~NEGEMMLowpMatrixMultiplyKernel() = default; 62 /** Initialise the kernel's input and output. 63 * 64 * The input matrices @p input0 and @p input1 must be the output of the kernels: @ref NEGEMMInterleave4x4Kernel and @ref NEGEMMTranspose1xWKernel. These two 65 * kernels change the layout of the original matrices to be more cache-friendly. 66 * 67 * @param[in] input0 Input tensor containing the interleaved Matrix A. Data type supported: U8/QASYMM8/S8/QASYMM8_SIGNED 68 * @param[in] input1 Input tensor containing the transposed1xW Matrix B. Data type supported: U8/QASYMM8/S8/QASYMM8_SIGNED/QSYMM8/QSYMM8_PER_CHANNEL 69 * @param[out] output Output tensor to store the result of matrix multiplication. Data type supported: S32 70 */ 71 void configure(const ITensor *input0, const ITensor *input1, ITensor *output); 72 /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMLowpMatrixMultiplyKernel 73 * 74 * @param[in] input0 Input tensor info containing the interleaved Matrix A. Data type supported: U8/QASYMM8/S8/QASYMM8_SIGNED 75 * @param[in] input1 Input tensor info containing the transposed Matrix B. Data type supported: U8/QASYMM8/S8/QASYMM8_SIGNED/QSYMM8/QSYMM8_PER_CHANNEL 76 * @param[in] output Output tensor info to store the result of matrix multiplication. Data type supported: S32 77 * 78 * @return a status 79 */ 80 static Status validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *output); 81 82 // Inherited methods overridden: 83 void run(const Window &window, const ThreadInfo &info) override; 84 85 private: 86 const ITensor *_input0; 87 const ITensor *_input1; 88 ITensor *_output; 89 bool _slide_matrix_b; 90 }; 91 } // namespace arm_compute 92 #endif /*ARM_COMPUTE_NEGEMMLOWPMATRIXMULTIPLYKERNEL_H*/ 93