xref: /aosp_15_r20/external/ComputeLibrary/src/cpu/kernels/CpuScaleKernel.h (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 2016-2022 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_CPU_SCALEKERNEL_H
25 #define ARM_COMPUTE_CPU_SCALEKERNEL_H
26 
27 #include "arm_compute/core/KernelDescriptors.h"
28 #include "src/core/common/Macros.h"
29 #include "src/cpu/ICpuKernel.h"
30 
31 namespace arm_compute
32 {
33 namespace cpu
34 {
35 namespace kernels
36 {
37 /** Arm(R) Neon(TM) kernel to perform scaling on a tensor */
38 class CpuScaleKernel : public ICpuKernel<CpuScaleKernel>
39 {
40 private:
41     /** Scale function to use for the particular function to use */
42     using ScaleFunctionPtr = void (CpuScaleKernel::*)(const ITensor *, ITensor *, const ITensor *, const ITensor *, const ITensor *, const Window &window);
43     using ScaleKernelPtr   = std::add_pointer<void(const ITensor *, ITensor *, const ITensor *, const ITensor *, const ITensor *,
44                                                    InterpolationPolicy, BorderMode, PixelValue, float, bool, const Window &)>::type;
45 
46 public:
47     CpuScaleKernel() = default;
48     ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuScaleKernel);
49     /** Initialise the kernel's inputs, output and interpolation policy
50      *
51      * @note dx, dy and offsets have the same dimensions (width and height) of the output tensor
52      * @note Using @p policy Area only supports data layout NCHW and input data type U8.
53      * @note Using S8 data type only supports NHWC, @p border_mode Replicate, and @p policy Bilinear
54      *
55      * @param[in]  src     Source tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/U8/S8/S16/F16/F32.
56      * @param[in]  dx      Distance x tensor info. Pixel's distance between the X real coordinate and the smallest X following integer. Data type supported: F32
57      * @param[in]  dy      Distance y tensor info. Pixel's distance between the Y real coordinate and the smallest Y following integer. Data type supported: F32
58      * @param[in]  offsets Offset tensor info. Offset to access the pixel with NEAREST interpolation or the top-left pixel with BILINEAR interpolation in the input tensor. Data type supported: S32.
59      * @param[out] dst     Destination tensor info. 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.
60      * @param[in]  info    @ref ScaleKernelInfo to use for configuration
61      */
62     void configure(const ITensorInfo *src, const ITensorInfo *dx, const ITensorInfo *dy, const ITensorInfo *offsets, ITensorInfo *dst,
63                    const ScaleKernelInfo &info);
64     /** Static function to check if given info will lead to a valid configuration
65      *
66      * Similar to CpuScaleKernel::configure()
67      *
68      * @return a status
69      */
70     static Status validate(const ITensorInfo *src, const ITensorInfo *dx, const ITensorInfo *dy, const ITensorInfo *offsets, ITensorInfo *dst,
71                            const ScaleKernelInfo &info);
72 
73     // Inherited methods overridden:
74     void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
75     const char *name() const override;
76 
77     struct ScaleKernel
78     {
79         const char                                 *name;
80         const ScaleKernelDataTypeISASelectorDataPtr is_selected;
81         ScaleKernelPtr                              ukernel;
82     };
83 
84     static const std::vector<ScaleKernel> &get_available_kernels();
85 
86 private:
87 #ifdef ENABLE_NCHW_KERNELS
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 ITensor *src, ITensor *dst, const ITensor *dx, const ITensor *dy, const ITensor *offsets, 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 ITensor *src, ITensor *dst, const ITensor *dx, const ITensor *dy, const ITensor *offsets, const Window &window);
97     /** function to perform scale using bilinear interpolation on the given window */
98     template <typename T>
99     void scale_bilinear_qasymm(const ITensor *src, ITensor *dst, const ITensor *dx, const ITensor *dy, const ITensor *offsets, const Window &window);
100 
101     /** function to perform scale using nearest neighbour on the given window */
102     template <typename T>
103     void scale_nearest_nchw(const ITensor *src, ITensor *dst, const ITensor *dx, const ITensor *dy, const ITensor *offsets, const Window &window);
104 #endif // ENABLE_NCHW_KERNELS
105 
106     ScaleFunctionPtr    _func{ nullptr };
107     InterpolationPolicy _policy{};
108     BorderMode          _border_mode{};
109     PixelValue          _constant_border_value{};
110     float               _sampling_offset{ 0 };
111     bool                _align_corners{ false };
112     DataLayout          _data_layout{ DataLayout::UNKNOWN };
113     ScaleKernelPtr      _run_method{ nullptr };
114     std::string         _name{};
115 };
116 } // namespace kernels
117 } // namespace cpu
118 } // namespace arm_compute
119 #endif /* ARM_COMPUTE_CPU_SCALEKERNEL_H */
120