xref: /aosp_15_r20/external/ComputeLibrary/src/gpu/cl/operators/ClConcatenate.h (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 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_CLCONCATENATE_H
25 #define ARM_COMPUTE_CLCONCATENATE_H
26 
27 #include "src/gpu/cl/ClCompileContext.h"
28 #include "src/gpu/cl/IClKernel.h"
29 #include "src/gpu/cl/IClOperator.h"
30 
31 #include <vector>
32 
33 namespace arm_compute
34 {
35 namespace opencl
36 {
37 /** Basic function to execute concatenate tensors along a given axis. This function calls the following kernels:
38  *
39  * -# @ref kernels::ClWidthConcatenateKernel (if underlying concatenation axis is 0).
40  * -# @ref kernels::ClHeightConcatenateKernel (if underlying concatenation axis is 1).
41  * -# @ref kernels::ClDepthConcatenateKernel (if underlying concatenation axis is 2).
42  * -# @ref kernels::ClBatchConcatenateKernel (if underlying concatenation axis is 3).
43  */
44 class ClConcatenate : public IClOperator
45 {
46 public:
47     ClConcatenate() = default;
48     /** Initialise the kernel's inputs vector and dst.
49      *
50      * @note Input and dst tensor dimensions preconditions defer depending on the concatenation axis.
51      * @note Preconditions can be found respectively at @ref kernels::ClWidthConcatenateKernel,
52      *       @ref kernels::ClHeightConcatenateKernel and @ref kernels::ClDepthConcatenateKernel.
53      *
54      *
55      * @param[in]     compile_context The compile context to be used.
56      * @param[in,out] src_vector      The vectors containing all the tensors info to concatenate. Data types supported: All
57      * @param[out]    dst             Destination tensor info. Data types supported: same as @p src_vector.
58      * @param[in]     axis            Concatenation axis. Supported underlying concatenation axis are 0, 1, 2 and 3.
59      */
60     void configure(const ClCompileContext &compile_context, const std::vector<ITensorInfo *> &src_vector, ITensorInfo *dst, size_t axis);
61     /** Static function to check if given info will lead to a valid configuration
62      *
63      * Similar to @ref ClConcatenate::configure()
64      *
65      * @return a status
66      */
67     static Status validate(const std::vector<const ITensorInfo *> &src_vector, const ITensorInfo *dst, size_t axis);
68 
69     // Inherited methods overridden:
70     void run(ITensorPack &tensors) override;
71 
72 private:
73     std::vector<std::unique_ptr<IClKernel>> _concat_kernels{};
74     unsigned int                            _num_inputs{ 0 };
75     unsigned int                            _axis{ 0 };
76 };
77 } // namespace opencl
78 } // namespace arm_compute
79 #endif /* ARM_COMPUTE_CL_CONCATENATE_H */
80