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_NESCALEKERNEL_H 25 #define ARM_COMPUTE_NESCALEKERNEL_H 26 27 #include "arm_compute/core/KernelDescriptors.h" 28 #include "src/core/NEON/INEKernel.h" 29 30 namespace arm_compute 31 { 32 class ITensor; 33 34 /** NEON kernel to perform scaling on a tensor */ 35 class NEScaleKernel : public INEKernel 36 { 37 public: name()38 const char *name() const override 39 { 40 return "NEScaleKernel"; 41 } 42 /** Default constructor */ 43 NEScaleKernel(); 44 /** Prevent instances of this class from being copied (As this class contains pointers) */ 45 NEScaleKernel(const NEScaleKernel &) = delete; 46 /** Prevent instances of this class from being copied (As this class contains pointers) */ 47 NEScaleKernel &operator=(const NEScaleKernel &) = delete; 48 /** Allow instances of this class to be moved */ 49 NEScaleKernel(NEScaleKernel &&) = default; 50 /** Allow instances of this class to be moved */ 51 NEScaleKernel &operator=(NEScaleKernel &&) = default; 52 /** Default destructor */ 53 ~NEScaleKernel() = default; 54 55 /** Initialise the kernel's inputs, output and interpolation policy 56 * 57 * @note dx, dy and offsets have the same dimensions (width and height) of the output tensor 58 * @note Using @p policy Area only supports data layout NCHW and input data type U8. 59 * 60 * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/U8/S16/F16/F32. 61 * @param[in] dx Pixel's distance between the X real coordinate and the smallest X following integer. Data type supported: F32 62 * @param[in] dy Pixel's distance between the Y real coordinate and the smallest Y following integer. Data type supported: F32 63 * @param[in] offsets Offset to access the pixel with NEAREST interpolation or the top-left pixel with BILINEAR interpolation in the input tensor. Data type supported: S32. 64 * @param[out] output Destination tensor. Data types supported: Same as @p input. All but the lowest two dimensions must be the same size as in the input tensor, i.e. scaling is only performed within the XY-plane. 65 * @param[in] info @ref ScaleKernelInfo to use for configuration 66 */ 67 void configure(const ITensor *input, const ITensor *dx, const ITensor *dy, const ITensor *offsets, ITensor *output, 68 const ScaleKernelInfo &info); 69 /** Static function to check if given info will lead to a valid configuration of @ref NEScaleKernel 70 * 71 * @note dx, dy and offsets have the same dimensions (width and height) of the output tensor 72 * @note Using @p policy Area only supports data layout NCHW and input data type U8. 73 * 74 * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/U8/S16/F16/F32. 75 * @param[in] dx Pixel's distance between the X real coordinate and the smallest X following integer. Data type supported: F32 76 * @param[in] dy Pixel's distance between the Y real coordinate and the smallest Y following integer. Data type supported: F32 77 * @param[in] offsets Offset to access the pixel with NEAREST interpolation or the top-left pixel with BILINEAR interpolation in the input tensor. Data type supported: S32. 78 * @param[in] output Destination tensor. Data types supported: Same as @p input. All but the lowest two dimensions must be the same size as in the input tensor, i.e. scaling is only performed within the XY-plane. 79 * @param[in] info @ref ScaleKernelInfo to use for validation 80 */ 81 static Status validate(const ITensorInfo *input, const ITensorInfo *dx, const ITensorInfo *dy, const ITensorInfo *offsets, ITensorInfo *output, 82 const ScaleKernelInfo &info); 83 84 // Inherited methods overridden: 85 void run(const Window &window, const ThreadInfo &info) override; 86 87 private: 88 /** function to perform scale using area interpolation on the given window 89 * 90 * @note Used only in case down-sampling. 91 */ 92 void scale_area_nchw_u8(const Window &window); 93 94 /** function to perform scale using bilinear interpolation on the given window */ 95 template <typename T> 96 void scale_bilinear_nchw(const Window &window); 97 /** function to perform scale using bilinear interpolation on the given window */ 98 template <typename T> 99 void scale_bilinear_nhwc(const Window &window); 100 /** function to perform scale using bilinear interpolation on the given window */ 101 template <typename T> 102 void scale_bilinear_qasymm(const Window &window); 103 104 /** function to perform scale using nearest neighbour on the given window */ 105 template <typename T> 106 void scale_nearest_nchw(const Window &window); 107 /** function to perform scale using nearest neighbour on the given window */ 108 template <typename T> 109 void scale_nearest_nhwc(const Window &window); 110 111 /** Scale function to use for the particular function to use */ 112 using ScaleFunctionPtr = void (NEScaleKernel::*)(const Window &window); 113 114 ScaleFunctionPtr _func; 115 const ITensor *_offsets; 116 const ITensor *_dx; 117 const ITensor *_dy; 118 const ITensor *_input; 119 ITensor *_output; 120 InterpolationPolicy _policy; 121 BorderMode _border_mode; 122 PixelValue _constant_border_value; 123 float _sampling_offset; 124 bool _align_corners; 125 }; 126 } // namespace arm_compute 127 #endif /*ARM_COMPUTE_NESCALEKERNEL_H */ 128