1 /* 2 * Copyright (c) 2016-2020 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_NEGEMMTRANSPOSE1xWKERNEL_H 25 #define ARM_COMPUTE_NEGEMMTRANSPOSE1xWKERNEL_H 26 27 #include "src/core/NEON/INESimpleKernel.h" 28 29 namespace arm_compute 30 { 31 // Forward declarations 32 class ITensor; 33 34 /** NEON kernel which transposes the elements of a matrix in chunks of 1xW, where W is equal to (16 / element size of the tensor) 35 * 36 * Following an example of how the transposition1xW works when the input data is F32 37 * 38 * @f[ 39 * \left( \begin{array}{cccc} 40 * a00 & a01 & a02 & a03 \\ 41 * a10 & a11 & a12 & a13 \\ 42 * a20 & a21 & a22 & a23 \\ 43 * a30 & a31 & a32 & a33 \\ 44 * \end{array} \right) 45 * \rightarrow 46 * \left( \begin{array}{ccccccccccccccccc} 47 * a00 & a01 & a02 & a03 & a10 & a11 & a12 & a13 & a20 & a21 & a22 & a23 & a30 & a31 & a32 & a33 \\ 48 * \end{array} \right) 49 * @f] 50 * 51 * Following an example of how the transposition1xW works when the input data type is F16 52 * 53 * @f[ 54 * \left( \begin{array}{cccccccc} 55 * a00 & a01 & a02 & a03 & a04 & a05 & a06 & a07 \\ 56 * a10 & a11 & a12 & a13 & a14 & a15 & a16 & a17 \\ 57 * a20 & a21 & a22 & a23 & a24 & a25 & a26 & a27 \\ 58 * a30 & a31 & a32 & a33 & a34 & a35 & a36 & a37 \\ 59 * \end{array} \right) 60 * \rightarrow 61 * \left( \begin{array}{cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc} 62 * a00 & a01 & a02 & a03 & a04 & a05 & a06 & a07 & a10 & a11 & a12 & a13 & a14 & a15 & a16 & a17 & a20 & a21 & a22 & a23 & a24 & a25 & a26 & a27 & a30 & a31 & a32 & a33 & a34 & a35 & a36 & a37\\ 63 * \end{array} \right) 64 * @f] 65 * 66 * @note The output matrix will have the following shape: [ height * W, ceil(width / W) ], where W = (16 / element size of the tensor) 67 * 68 */ 69 class NEGEMMTranspose1xWKernel : public INESimpleKernel 70 { 71 public: name()72 const char *name() const override 73 { 74 return "NEGEMMTranspose1xWKernel"; 75 } 76 /** Constructor */ 77 NEGEMMTranspose1xWKernel() = default; 78 /** Prevent instances of this class from being copied (As this class contains pointers) */ 79 NEGEMMTranspose1xWKernel(const NEGEMMTranspose1xWKernel &) = delete; 80 /** Prevent instances of this class from being copied (As this class contains pointers) */ 81 NEGEMMTranspose1xWKernel &operator=(const NEGEMMTranspose1xWKernel &) = delete; 82 /** Allow instances of this class to be moved */ 83 NEGEMMTranspose1xWKernel(NEGEMMTranspose1xWKernel &&) = default; 84 /** Allow instances of this class to be moved */ 85 NEGEMMTranspose1xWKernel &operator=(NEGEMMTranspose1xWKernel &&) = default; 86 /** Default destructor */ 87 ~NEGEMMTranspose1xWKernel() = default; 88 /** Initialise the kernel's input and output. 89 * 90 * @param[in] input Input tensor. Data types supported: All 91 * @param[out] output Output tensor. Data type supported: same as @p input. 92 */ 93 void configure(const ITensor *input, ITensor *output); 94 /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMTranspose1xWKernel 95 * 96 * @param[in] input Input tensor info. Data types supported: All 97 * @param[in] output Output tensor info. Data type supported: same as @p input. 98 * 99 * @return a status 100 */ 101 static Status validate(const ITensorInfo *input, const ITensorInfo *output); 102 103 // Inherited methods overridden: 104 void run(const Window &window, const ThreadInfo &info) override; 105 }; 106 } // namespace arm_compute 107 #endif /*ARM_COMPUTE_NEGEMMTRANSPOSE1xWKERNEL_H */ 108