xref: /aosp_15_r20/external/ComputeLibrary/src/cpu/kernels/CpuGemmTranspose1xWKernel.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_GEMM_TRANSPOSE1xW_KERNEL_H
25 #define ARM_COMPUTE_CPU_GEMM_TRANSPOSE1xW_KERNEL_H
26 
27 #include "src/core/common/Macros.h"
28 #include "src/cpu/ICpuKernel.h"
29 
30 namespace arm_compute
31 {
32 namespace cpu
33 {
34 namespace kernels
35 {
36 /** Kernel which transposes the elements of a matrix in chunks of 1xW, where W is equal to (16 / element size of the tensor)
37  *
38  * Following an example of how the transposition1xW works when the input data is F32
39  *
40  * @f[
41  * \left( \begin{array}{cccc}
42  * a00 & a01 & a02 & a03 \\
43  * a10 & a11 & a12 & a13 \\
44  * a20 & a21 & a22 & a23 \\
45  * a30 & a31 & a32 & a33 \\
46  * \end{array} \right)
47  * \rightarrow
48  * \left( \begin{array}{ccccccccccccccccc}
49  * a00 & a01 & a02 & a03 & a10 & a11 & a12 & a13 & a20 & a21 & a22 & a23 & a30 & a31 & a32 & a33 \\
50  * \end{array} \right)
51  * @f]
52  *
53  * Following an example of how the transposition1xW works when the input data type is F16
54  *
55  * @f[
56  * \left( \begin{array}{cccccccc}
57  * a00 & a01 & a02 & a03 & a04 & a05 & a06 & a07 \\
58  * a10 & a11 & a12 & a13 & a14 & a15 & a16 & a17 \\
59  * a20 & a21 & a22 & a23 & a24 & a25 & a26 & a27 \\
60  * a30 & a31 & a32 & a33 & a34 & a35 & a36 & a37 \\
61  * \end{array} \right)
62  * \rightarrow
63  * \left( \begin{array}{cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc}
64  * 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\\
65  * \end{array} \right)
66  * @f]
67  *
68  * @note The output matrix will have the following shape: [ height * W, ceil(width / W) ], where W = (16 / element size of the tensor)
69  *
70  */
71 class CpuGemmTranspose1xWKernel : public ICpuKernel<CpuGemmTranspose1xWKernel>
72 {
73 public:
74     CpuGemmTranspose1xWKernel() = default;
75     ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuGemmTranspose1xWKernel);
76     /** Configure kernel for a given list of arguments
77      *
78      * @param[in]  src Input tensor info. Data types supported: All
79      * @param[out] dst Output tensor info. Data type supported: same as @p src.
80      */
81     void configure(const ITensorInfo *src, ITensorInfo *dst);
82     /** Static function to check if given info will lead to a valid configuration of @ref CpuGemmTranspose1xWKernel
83      *
84      * Similar to @ref CpuGemmTranspose1xWKernel::configure()
85      *
86      * @return a status
87      */
88     static Status validate(const ITensorInfo *src, const ITensorInfo *dst);
89 
90     // Inherited methods overridden:
91     void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
92     const char *name() const override;
93 };
94 } // namespace kernels
95 } // namespace cpu
96 } // namespace arm_compute
97 #endif /* ARM_COMPUTE_CPU_GEMM_TRANSPOSE1xW_KERNEL_H */
98