giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2019 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_CLGEMMMATRIXMULTIPLYNATIVEKERNEL_H__ |
| 25 | #define __ARM_COMPUTE_CLGEMMMATRIXMULTIPLYNATIVEKERNEL_H__ |
| 26 | |
| 27 | #include "arm_compute/core/CL/ICLKernel.h" |
| 28 | |
| 29 | namespace arm_compute |
| 30 | { |
| 31 | class ICLTensor; |
| 32 | |
| 33 | /** OpenCL kernel to multiply matrices when neither of the input matrices have been reshaped */ |
| 34 | class CLGEMMMatrixMultiplyNativeKernel : public ICLKernel |
| 35 | { |
| 36 | public: |
| 37 | /** Default Constructor */ |
| 38 | CLGEMMMatrixMultiplyNativeKernel(); |
| 39 | /** Prevent instances of this class from being copied (As this class contains pointers) */ |
| 40 | CLGEMMMatrixMultiplyNativeKernel(const CLGEMMMatrixMultiplyNativeKernel &) = delete; |
| 41 | /** Prevent instances of this class from being copied (As this class contains pointers) */ |
| 42 | CLGEMMMatrixMultiplyNativeKernel &operator=(const CLGEMMMatrixMultiplyNativeKernel &) = delete; |
| 43 | /** Allow instances of this class to be moved */ |
| 44 | CLGEMMMatrixMultiplyNativeKernel(CLGEMMMatrixMultiplyNativeKernel &&) = default; |
| 45 | /** Allow instances of this class to be moved */ |
| 46 | CLGEMMMatrixMultiplyNativeKernel &operator=(CLGEMMMatrixMultiplyNativeKernel &&) = default; |
| 47 | /** Initialise the kernel's input and output. |
| 48 | * |
| 49 | * @param[in] input0 Input tensor for the LHS matrix. Data type supported: F32/F16. The number of dimensions for the LHS matrix must be less or equal than 4. |
| 50 | * @param[in] input1 Input tensor for the RHS matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3. |
Gian Marco Iodice | 944170e | 2019-06-24 14:40:30 +0100 | [diff] [blame] | 51 | * @param[in] input2 Input tensor containing the bias matrix. Data type supported: same as @p input0. |
giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 52 | * @param[out] output Output tensor info. Data type supported: same as @p input0 |
| 53 | * @param[in] alpha Weight of the matrix product |
Gian Marco Iodice | 944170e | 2019-06-24 14:40:30 +0100 | [diff] [blame] | 54 | * @param[in] beta Weight of the matrix bias |
giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 55 | * @param[in] lhs_info LHS matrix information used to retrieve the number of rows and accumulations to be processed by each thread. Only the following values are supported: |
| 56 | * lhs_info.m0: 1,2,3,4,5,6,7,8 |
| 57 | * lhs_info.k0: 2,3,4,8,16 |
| 58 | * @param[in] rhs_info RHS matrix information used to retrieve the number of columns and accumulations to be processed by each thread. Only the following values are supported: |
| 59 | * rhs_info.n0: 2,3,4,8,16 |
| 60 | * rhs_info.k0: same of lhs_info.k0 |
| 61 | * @param[in] gemm_info GEMM information used to retrieve the original dimensions of the input matrices |
| 62 | */ |
Gian Marco Iodice | 944170e | 2019-06-24 14:40:30 +0100 | [diff] [blame] | 63 | void configure(const ICLTensor *input0, const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info, |
| 64 | const GEMMRHSMatrixInfo &rhs_info, |
giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 65 | const GEMMReshapeInfo &gemm_info); |
| 66 | /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMMatrixMultiplyNativeKernel |
| 67 | * |
| 68 | * @param[in] input0 Input tensor info for the LHS matrix. Data type supported: F32/F16. The number of dimensions for the LHS matrix must be less or equal than 4. |
| 69 | * @param[in] input1 Input tensor info for the RHS matrix. Data type supported: same as @p input0. The number of dimensions for the RHS matrix must be less or equal than 3. |
Gian Marco Iodice | 944170e | 2019-06-24 14:40:30 +0100 | [diff] [blame] | 70 | * @param[in] input2 Input tensor info containing the bias matrix. Data type supported: same as @p input0. |
giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 71 | * @param[in] output Output tensor info. Data type supported: same as @p input0 |
| 72 | * @param[in] alpha Weight of the matrix product |
Gian Marco Iodice | 944170e | 2019-06-24 14:40:30 +0100 | [diff] [blame] | 73 | * @param[in] beta Weight of the matrix bias |
giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 74 | * @param[in] lhs_info LHS matrix information used to retrieve the number of rows and accumulations to be processed by each thread. Only the following values are supported: |
| 75 | * lhs_info.m0: 1,2,3,4,5,6,7,8 |
| 76 | * lhs_info.k0: 2,3,4,8,16 |
| 77 | * @param[in] rhs_info RHS matrix information used to retrieve the number of columns and accumulations to be processed by each thread. Only the following values are supported: |
| 78 | * rhs_info.n0: 2,3,4,8,16 |
| 79 | * rhs_info.k0: same of lhs_info.k0 |
| 80 | * @param[in] gemm_info GEMM information used to retrieve the original dimensions of the input matrices |
| 81 | * |
| 82 | * @return a status |
| 83 | */ |
Gian Marco Iodice | 944170e | 2019-06-24 14:40:30 +0100 | [diff] [blame] | 84 | static Status validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info, |
| 85 | const GEMMRHSMatrixInfo &rhs_info, |
giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 86 | const GEMMReshapeInfo &gemm_info); |
| 87 | |
| 88 | // Inherited methods overridden: |
| 89 | void run(const Window &window, cl::CommandQueue &queue) override; |
| 90 | |
| 91 | private: |
| 92 | const ICLTensor *_input0; |
| 93 | const ICLTensor *_input1; |
Gian Marco Iodice | 944170e | 2019-06-24 14:40:30 +0100 | [diff] [blame] | 94 | const ICLTensor *_input2; |
giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 95 | ICLTensor *_output; |
| 96 | bool _slide_matrix_b; |
| 97 | bool _reinterpret_input_as_3d; |
| 98 | bool _reinterpret_output_as_3d; |
| 99 | bool _use_dummy_work_items; |
Gian Marco Iodice | 944170e | 2019-06-24 14:40:30 +0100 | [diff] [blame] | 100 | bool _add_bias; |
| 101 | bool _broadcast_bias; |
giuros01 | b3204e7 | 2019-04-01 13:50:22 +0100 | [diff] [blame] | 102 | }; |
| 103 | } // namespace arm_compute |
| 104 | #endif /*__ARM_COMPUTE_CLGEMMMATRIXMULTIPLYNATIVEKERNEL_H__*/ |