xref: /aosp_15_r20/external/ComputeLibrary/src/runtime/NEON/functions/NEStackLayer.cpp (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 2018-2021 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 #include "arm_compute/runtime/NEON/functions/NEStackLayer.h"
25 
26 #include "arm_compute/core/Error.h"
27 #include "arm_compute/core/Helpers.h"
28 #include "arm_compute/core/ITensor.h"
29 #include "arm_compute/core/TensorInfo.h"
30 #include "arm_compute/core/Types.h"
31 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
32 #include "arm_compute/runtime/NEON/NEScheduler.h"
33 #include "src/common/utils/Log.h"
34 #include "src/core/NEON/kernels/NEStackLayerKernel.h"
35 
36 namespace arm_compute
37 {
38 NEStackLayer::~NEStackLayer() = default;
39 
NEStackLayer()40 NEStackLayer::NEStackLayer() // NOLINT
41     : _input(),
42       _stack_kernels(),
43       _num_inputs(0)
44 {
45 }
46 
configure(const std::vector<ITensor * > & input,int axis,ITensor * output)47 void NEStackLayer::configure(const std::vector<ITensor *> &input, int axis, ITensor *output)
48 {
49     ARM_COMPUTE_LOG_PARAMS(input, axis, output);
50 
51     _num_inputs = input.size();
52     _stack_kernels.resize(_num_inputs);
53 
54     // Wrap around negative values
55     const unsigned int axis_u = wrap_around(axis, static_cast<int>(input[0]->info()->num_dimensions() + 1));
56 
57     for(unsigned int i = 0; i < _num_inputs; i++)
58     {
59         _stack_kernels[i] = std::make_unique<NEStackLayerKernel>();
60         _stack_kernels[i]->configure(input[i], axis_u, i, _num_inputs, output);
61     }
62 }
63 
validate(const std::vector<ITensorInfo * > & input,int axis,const ITensorInfo * output)64 Status NEStackLayer::validate(const std::vector<ITensorInfo *> &input, int axis, const ITensorInfo *output)
65 {
66     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output);
67     ARM_COMPUTE_RETURN_ERROR_ON(input.empty());
68 
69     // Wrap around negative values
70     const size_t       rank   = input[0]->num_dimensions();
71     const unsigned int axis_u = wrap_around(axis, static_cast<int>(rank + 1));
72 
73     const unsigned int num_inputs = input.size();
74 
75     for(unsigned int i = 0; i < num_inputs; i++)
76     {
77         // All the tensors must have the same rank
78         ARM_COMPUTE_RETURN_ERROR_ON(input[i]->num_dimensions() != rank);
79         // Validate Kernel
80         ARM_COMPUTE_RETURN_ON_ERROR(NEStackLayerKernel::validate(input[i], axis_u, i, num_inputs, output));
81     }
82 
83     return Status{};
84 }
85 
run()86 void NEStackLayer::run()
87 {
88     for(unsigned i = 0; i < _num_inputs; i++)
89     {
90         NEScheduler::get().schedule(_stack_kernels[i].get(), Window::DimY);
91     }
92 }
93 } // namespace arm_compute
94