xref: /aosp_15_r20/external/ComputeLibrary/src/gpu/cl/kernels/ClCol2ImKernel.h (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 2017-2021 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_COL2IM_KERNEL_H
25 #define ARM_COMPUTE_CL_COL2IM_KERNEL_H
26 
27 #include "arm_compute/core/Size2D.h"
28 #include "src/core/common/Macros.h"
29 #include "src/gpu/cl/ClCompileContext.h"
30 #include "src/gpu/cl/IClKernel.h"
31 
32 namespace arm_compute
33 {
34 namespace opencl
35 {
36 namespace kernels
37 {
38 /** Interface for the col2im reshaping kernel.
39  *
40  * Rearranges each matrix column into image blocks. It's the inverse operation of @ref opencl::kernels::ClIm2ColKernel.
41  *
42  * For example, a vector of 9 elements can be reshaped to a block(image) of 3x3:
43  *
44  * @f[
45  * \left( \begin{array}{ccccccccc}
46  * a0 & a1 & a2 & a3 & a4 & a5 & a6 & a7 & a8 \\
47  * \end{array} \right)
48  * \rightarrow
49  * \left( \begin{array}{ccc}
50  * a0 & a1 & a2 \\
51  * a3 & a4 & a5 \\
52  * a6 & a7 & a8 \\
53  * \end{array} \right)
54  * @f]
55  */
56 class ClCol2ImKernel : public IClKernel
57 {
58 public:
59     /** Default constructor */
60     ClCol2ImKernel();
61     ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClCol2ImKernel);
62     /** Set the input and output of the kernel.
63      *
64      * @param[in]  compile_context The compile context to be used.
65      * @param[in]  src             The input tensor info to convert. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32
66      * @param[out] dst             The output tensor info. 3 lower dimensions represent a single output [width, height, OFM],
67      *                             while the rest represent batch of outputs. Data types supported: Same as @p input. Data layout: NCHW
68      * @param[in]  convolved_dims  Output convolved dimensions.
69      * @param[in]  num_groups      (Optional) Number of groups when performing a grouped convolution
70      */
71     void configure(const CLCompileContext &compile_context, ITensorInfo *src, ITensorInfo *dst, const Size2D &convolved_dims, unsigned int num_groups = 1);
72     /** Static function to check if given info will lead to a valid configuration
73      *
74      * Similar to ClCol2ImKernel::configure()
75      *
76      * @return a status
77      */
78     static Status validate(const ITensorInfo *src, const ITensorInfo *dst, const Size2D &convolved_dims, unsigned int num_groups = 1);
79 
80     // Inherited methods overridden:
81     void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
82 
83 public:
84     Size2D _convolved_dims;
85 };
86 } // namespace kernels
87 } // namespace opencl
88 } // namespace arm_compute
89 #endif /*ARM_COMPUTE_CL_COL2IM_KERNEL_H */
90