1 /* 2 * Copyright (c) 2018-2021 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_GEMM_MATRIXMULTIPLY_RESHAPED_KERNEL_H 25 #define ARM_COMPUTE_CL_GEMM_MATRIXMULTIPLY_RESHAPED_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 #include "arm_compute/core/KernelDescriptors.h" 32 33 namespace arm_compute 34 { 35 namespace opencl 36 { 37 namespace kernels 38 { 39 /** OpenCL kernel to multiply matrices when both the input matrices LHS (src0) and RHS (src1) have been reshaped 40 * 41 * @note The input matrices @p src0 and @p src1 must be reshaped through: 42 * - @ref ClGemmReshapeLhsMatrixKernel 43 * - @ref ClGemmReshapeRhsMatrixKernel 44 */ 45 class ClGemmMatrixMultiplyReshapedKernel : public IClKernel 46 { 47 public: 48 ClGemmMatrixMultiplyReshapedKernel(); 49 ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClGemmMatrixMultiplyReshapedKernel); 50 /** Initialise the kernel's input and output. 51 * 52 * @note The F16 computation also supports mixed precision through the gemm_info.fp_mixed_precision flag. 53 * Mixed precision combines different floating precisions during the computation, in particular, F32 for the accumulations and F16 for the 54 * multiplications. i.e. float c = (half)a * (half)b 55 * 56 * @note If rhs_info.export_to_cl_image = true, this OpenCL kernel will fetch the RHS data using the OpenCL read_image built-in function. 57 * Reading from the OpenCL image object can increase the performance. However, since the OpenCL image object is created importing the OpenCL buffer, 58 * the following conditions are required: 59 * -# rhs_info.n0 can only be 4, 8 and 16 60 * -# rhs_info.k0 can only be 4, 8 and 16 61 * -# Data type can only be F32 62 * -# The platform should support the OpenCL cl_khr_image2d_from_buffer extension 63 * -# The stride Y for the src1 should satisfy the OpenCL pitch alignment requirement 64 * -# src1 width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4) 65 * -# src1 (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT 66 * 67 * @param[in] compile_context The compile context to be used. 68 * @param[in] src0 Input tensor containing the LHS reshaped matrix. Data type supported: F16/F32 (only F32 if rhs_info.export_to_cl_image = true). The number of dimensions for the LHS matrix must be less or equal than 4 69 * @param[in] src1 Input tensor containing the RHS reshaped matrix. Data type supported: same as @p src0. The number of dimensions for the RHS matrix must be less or equal than 3 70 * @param[in] src2 Input tensor containing the bias matrix. Data type supported: same as @p src0. 71 * @param[out] dst dst tensor to store the result of matrix multiplication. Data type supported: same as @p src0 72 * @param[in] alpha Weight of the matrix product 73 * @param[in] beta Weight of the matrix bias 74 * @param[in] lhs_info LHS matrix information used for reshaping the src0 tensor. Only the following values are supported: 75 * lhs_info.m0: 2,3,4,5,6,7,8 76 * lhs_info.k0: 2,3,4,8,16 77 * lhs_info.transpose: false 78 * @param[in] rhs_info RHS matrix information used for reshaping the src1 tensor. Only the following values are supported: 79 * rhs_info.n0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image = true) 80 * rhs_info.k0: 2,3,4,8,16 (only 4, 8 and 16 if rhs_info.export_to_cl_image = true) 81 * rhs_info.transpose: true 82 * @param[in] gemm_info GEMM information used to retrieve the original dimensions of the input matrices 83 * 84 * @note lhs_info.k0 must be equal to rhs_info.k0 85 */ 86 void configure(const ClCompileContext &compile_context, 87 const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *src2, ITensorInfo *dst, float alpha, float beta, 88 const GEMMLHSMatrixInfo &lhs_info, const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info); 89 /** Static function to check if given info will lead to a valid configuration 90 * 91 * Similar to @ref ClGemmMatrixMultiplyReshapedKernel::configure() 92 * 93 * @return a status 94 */ 95 static Status validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info, 96 const GEMMRHSMatrixInfo &rhs_info, 97 const GEMMKernelInfo &gemm_info); 98 99 // Inherited methods overridden: 100 void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override; 101 102 private: 103 bool _slide_matrix_b{ true }; 104 bool _reinterpret_output_as_3d{ false }; 105 bool _use_dummy_work_items{ false }; 106 bool _add_bias{ false }; 107 bool _export_to_cl_image{ false }; 108 signed int _m{ 1 }; 109 signed int _n{ 1 }; 110 signed int _k{ 1 }; 111 unsigned int _num_post_op_args{ 0 }; // (EXPERIMENTAL_POST_OPS) total number of post op arguments 112 }; 113 } // namespace kernels 114 } // namespace opencl 115 } // namespace arm_compute 116 #endif /* ARM_COMPUTE_CL_GEMM_MATRIXMULTIPLY_RESHAPED_KERNEL_H */