Gian Marco | 05288a2 | 2017-11-21 10:57:50 +0000 | [diff] [blame] | 1 | /* |
Manuel Bottini | 959c26d | 2019-12-02 16:22:35 +0000 | [diff] [blame] | 2 | * Copyright (c) 2017-2020 ARM Limited. |
Gian Marco | 05288a2 | 2017-11-21 10:57:50 +0000 | [diff] [blame] | 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 | */ |
Michalis Spyrou | f464337 | 2019-11-29 16:17:13 +0000 | [diff] [blame] | 24 | #ifndef ARM_COMPUTE_CLGEMMLOWREDUCTIONKERNEL_H |
| 25 | #define ARM_COMPUTE_CLGEMMLOWREDUCTIONKERNEL_H |
Gian Marco | 05288a2 | 2017-11-21 10:57:50 +0000 | [diff] [blame] | 26 | |
| 27 | #include "arm_compute/core/CL/ICLKernel.h" |
| 28 | |
| 29 | namespace arm_compute |
| 30 | { |
| 31 | class ICLTensor; |
| 32 | |
| 33 | /** Common interface for all OpenCL reduction kernels */ |
| 34 | class ICLGEMMLowpReductionKernel : public ICLKernel |
| 35 | { |
| 36 | public: |
| 37 | /** Constructor */ |
| 38 | ICLGEMMLowpReductionKernel(); |
| 39 | /** Prevent instances of this class from being copied (As this class contains pointers)*/ |
| 40 | ICLGEMMLowpReductionKernel(const ICLGEMMLowpReductionKernel &) = delete; |
| 41 | /** Prevent instances of this class from being copied (As this class contains pointers)*/ |
| 42 | ICLGEMMLowpReductionKernel &operator=(const ICLGEMMLowpReductionKernel &) = delete; |
| 43 | /** Allow instances of this class to be moved */ |
| 44 | ICLGEMMLowpReductionKernel(ICLGEMMLowpReductionKernel &&) = default; |
| 45 | /** Allow instances of this class to be moved */ |
| 46 | ICLGEMMLowpReductionKernel &operator=(ICLGEMMLowpReductionKernel &&) = default; |
| 47 | |
| 48 | /** Initialise the kernel's input and output. |
| 49 | * |
| 50 | * @param[in] input Input tensor. Data type supported: S8 |
| 51 | * @param[out] output Output row-vector of sums of all the entries in each row/col of input tensor. Data type supported: S32 |
| 52 | */ |
| 53 | virtual void configure(const ICLTensor *input, ICLTensor *output) = 0; |
| 54 | |
| 55 | protected: |
| 56 | const ICLTensor *_input; |
| 57 | ICLTensor *_output; |
| 58 | }; |
| 59 | |
| 60 | /** OpenCL kernel used to compute the row-vectors of sums of all the entries in each row of Matrix A. |
| 61 | * |
| 62 | * @note This stage is needed to handle the offset of matrix product |
| 63 | * https://github.com/google/gemmlowp/blob/master/doc/low-precision.md |
| 64 | */ |
| 65 | class CLGEMMLowpMatrixAReductionKernel : public ICLGEMMLowpReductionKernel |
| 66 | { |
| 67 | public: |
| 68 | /** Initialise the kernel's input and output. |
| 69 | * |
Manuel Bottini | 959c26d | 2019-12-02 16:22:35 +0000 | [diff] [blame] | 70 | * @param[in] mtx_a Input tensor. Data type supported: QASYMM8/QASYMM8_SIGNED |
Gian Marco | 05288a2 | 2017-11-21 10:57:50 +0000 | [diff] [blame] | 71 | * @param[out] vector_sum_row Output row-vector of sums of all the entries in each row of mtx_a. Data type supported: S32 |
| 72 | */ |
| 73 | void configure(const ICLTensor *mtx_a, ICLTensor *vector_sum_row) override; |
Georgios Pinitas | 358ca20 | 2017-12-07 16:47:52 +0000 | [diff] [blame] | 74 | /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMLowpMatrixAReductionKernel |
| 75 | * |
Manuel Bottini | 959c26d | 2019-12-02 16:22:35 +0000 | [diff] [blame] | 76 | * @param[in] mtx_a Input tensor. Data type supported: QASYMM8/QASYMM8_SIGNED |
Georgios Pinitas | 358ca20 | 2017-12-07 16:47:52 +0000 | [diff] [blame] | 77 | * @param[in] vector_sum_row Output row-vector of sums of all the entries in each row of mtx_a. Data type supported: S32 |
| 78 | * |
| 79 | * @return a status |
| 80 | */ |
| 81 | static Status validate(const ITensorInfo *mtx_a, const ITensorInfo *vector_sum_row); |
Gian Marco | 05288a2 | 2017-11-21 10:57:50 +0000 | [diff] [blame] | 82 | |
| 83 | // Inherited methods overridden: |
| 84 | void run(const Window &window, cl::CommandQueue &queue) override; |
| 85 | }; |
| 86 | |
| 87 | /** OpenCL kernel used to compute the row-vectors of sums of all the entries in each column of Matrix B. |
| 88 | * |
| 89 | * @note This stage is needed to handle the offset of matrix product |
| 90 | * https://github.com/google/gemmlowp/blob/master/doc/low-precision.md |
| 91 | */ |
| 92 | class CLGEMMLowpMatrixBReductionKernel : public ICLGEMMLowpReductionKernel |
| 93 | { |
| 94 | public: |
| 95 | /** Initialise the kernel's input and output. |
| 96 | * |
Manuel Bottini | 959c26d | 2019-12-02 16:22:35 +0000 | [diff] [blame] | 97 | * @param[in] mtx_b Input tensor. Data type supported: Data type supported: QASYMM8/QASYMM8_SIGNED |
Gian Marco | 05288a2 | 2017-11-21 10:57:50 +0000 | [diff] [blame] | 98 | * @param[out] vector_sum_col Output row-vector of sums of all the entries in each column of mtx_b. Data type supported: S32 |
| 99 | */ |
| 100 | void configure(const ICLTensor *mtx_b, ICLTensor *vector_sum_col) override; |
Georgios Pinitas | 358ca20 | 2017-12-07 16:47:52 +0000 | [diff] [blame] | 101 | /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMLowpMatrixBReductionKernel |
| 102 | * |
Manuel Bottini | 959c26d | 2019-12-02 16:22:35 +0000 | [diff] [blame] | 103 | * @param[in] mtx_b Input tensor. Data type supported: Data type supported: QASYMM8/QASYMM8_SIGNED |
Georgios Pinitas | 358ca20 | 2017-12-07 16:47:52 +0000 | [diff] [blame] | 104 | * @param[in] vector_sum_col Output row-vector of sums of all the entries in each column of mtx_b. Data type supported: S32 |
| 105 | * |
| 106 | * @return a status |
| 107 | */ |
| 108 | static Status validate(const ITensorInfo *mtx_b, const ITensorInfo *vector_sum_col); |
Gian Marco | 05288a2 | 2017-11-21 10:57:50 +0000 | [diff] [blame] | 109 | |
| 110 | // Inherited methods overridden: |
| 111 | void run(const Window &window, cl::CommandQueue &queue) override; |
| 112 | }; |
| 113 | } // namespace arm_compute |
Michalis Spyrou | f464337 | 2019-11-29 16:17:13 +0000 | [diff] [blame] | 114 | #endif /* ARM_COMPUTE_CLGEMMLOWREDUCTIONKERNEL_H */ |