2 * Copyright (c) 2017 ARM Limited.
4 * SPDX-License-Identifier: MIT
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:
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
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
25 #include "arm_compute/runtime/GLES_COMPUTE/functions/GCNormalizationLayer.h"
27 #include "arm_compute/core/Error.h"
28 #include "arm_compute/core/PixelValue.h"
29 #include "arm_compute/core/TensorInfo.h"
30 #include "arm_compute/core/Types.h"
31 #include "arm_compute/core/Validate.h"
32 #include "arm_compute/runtime/GLES_COMPUTE/GCScheduler.h"
34 using namespace arm_compute;
36 GCNormalizationLayer::GCNormalizationLayer()
37 : _squared_input(), _norm_kernel(), _multiply_kernel(), _border_handler()
41 void GCNormalizationLayer::configure(const IGCTensor *input, IGCTensor *output, const NormalizationLayerInfo &norm_info)
43 ARM_COMPUTE_ERROR_ON(input == nullptr);
45 _squared_input.allocator()->init(TensorInfo(input->info()->tensor_shape(), 1, input->info()->data_type()));
47 _norm_kernel.configure(input, &_squared_input, output, norm_info);
48 _multiply_kernel.configure(input, input, &_squared_input, 1.0f);
49 // Fill the border by 3 elements since we need vload4 in the IN_MAP normalization kernel
50 _border_handler.configure(&_squared_input, _norm_kernel.border_size(), BorderMode::CONSTANT, PixelValue(0));
52 // Allocate intermediate buffers
53 _squared_input.allocator()->allocate();
56 void GCNormalizationLayer::run()
58 GCScheduler::get().dispatch(_multiply_kernel, false);
59 GCScheduler::get().memory_barrier();
60 GCScheduler::get().dispatch(_border_handler, false);
61 GCScheduler::get().memory_barrier();
62 GCScheduler::get().dispatch(_norm_kernel, true);