Georgios Pinitas | ae54e02 | 2018-07-16 15:41:27 +0100 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2018 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/NEWidthConcatenateLayer.h" |
| 25 | |
| 26 | #include "arm_compute/core/Error.h" |
| 27 | #include "arm_compute/core/Helpers.h" |
| 28 | #include "arm_compute/core/TensorInfo.h" |
| 29 | #include "arm_compute/core/Types.h" |
| 30 | #include "arm_compute/core/utils/misc/ShapeCalculator.h" |
| 31 | #include "arm_compute/runtime/NEON/NEScheduler.h" |
| 32 | #include "arm_compute/runtime/Tensor.h" |
| 33 | #include "support/ToolchainSupport.h" |
| 34 | |
| 35 | using namespace arm_compute; |
| 36 | |
| 37 | NEWidthConcatenateLayer::NEWidthConcatenateLayer() |
| 38 | : _concat_kernels_vector(), |
| 39 | _num_inputs(0) |
| 40 | { |
| 41 | } |
| 42 | |
| 43 | Status NEWidthConcatenateLayer::validate(const std::vector<ITensorInfo *> &inputs_vector, const ITensorInfo *output) |
| 44 | { |
| 45 | ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output); |
| 46 | ARM_COMPUTE_RETURN_ERROR_ON(inputs_vector.size() < 2); |
| 47 | |
| 48 | // Output auto inizialitation if not yet initialized |
| 49 | TensorInfo tmp_output_info = *output->clone(); |
| 50 | TensorShape output_shape = arm_compute::misc::shape_calculator::calculate_width_concatenate_shape(inputs_vector); |
| 51 | auto_init_if_empty(tmp_output_info, output_shape, 1, inputs_vector[0]->data_type()); |
| 52 | |
| 53 | unsigned int width_offset = 0; |
| 54 | for(const auto &input : inputs_vector) |
| 55 | { |
| 56 | ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input); |
| 57 | ARM_COMPUTE_RETURN_ON_ERROR(NEWidthConcatenateLayerKernel::validate(input, width_offset, &tmp_output_info)); |
| 58 | width_offset += input->dimension(0); |
| 59 | } |
| 60 | |
| 61 | return Status{}; |
| 62 | } |
| 63 | |
| 64 | void NEWidthConcatenateLayer::configure(std::vector<ITensor *> inputs_vector, ITensor *output) |
| 65 | { |
| 66 | _num_inputs = inputs_vector.size(); |
| 67 | |
| 68 | std::vector<ITensorInfo *> inputs_vector_info; |
| 69 | for(unsigned int i = 0; i < _num_inputs; i++) |
| 70 | { |
| 71 | inputs_vector_info.emplace_back(inputs_vector.at(i)->info()); |
| 72 | } |
| 73 | TensorShape output_shape = arm_compute::misc::shape_calculator::calculate_width_concatenate_shape(inputs_vector); |
| 74 | |
| 75 | // Output auto inizialitation if not yet initialized |
| 76 | auto_init_if_empty(*output->info(), output_shape, 1, inputs_vector[0]->info()->data_type()); |
| 77 | ARM_COMPUTE_ERROR_THROW_ON(NEWidthConcatenateLayer::validate(inputs_vector_info, output->info())); |
| 78 | |
| 79 | unsigned int width_offset = 0; |
| 80 | |
| 81 | _concat_kernels_vector = arm_compute::support::cpp14::make_unique<NEWidthConcatenateLayerKernel[]>(_num_inputs); |
| 82 | |
| 83 | for(unsigned int i = 0; i < _num_inputs; i++) |
| 84 | { |
| 85 | _concat_kernels_vector[i].configure(inputs_vector.at(i), width_offset, output); |
| 86 | width_offset += inputs_vector.at(i)->info()->dimension(0); |
| 87 | } |
| 88 | } |
| 89 | |
| 90 | void NEWidthConcatenateLayer::run() |
| 91 | { |
| 92 | for(unsigned i = 0; i < _num_inputs; i++) |
| 93 | { |
| 94 | NEScheduler::get().schedule(_concat_kernels_vector.get() + i, Window::DimY); |
| 95 | } |
| 96 | } |