Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame^] | 1 | /* |
| 2 | * Copyright (c) 2017 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/NEDirectConvolutionLayer.h" |
| 25 | |
| 26 | #include "arm_compute/core/PixelValue.h" |
| 27 | #include "arm_compute/core/Utils.h" |
| 28 | #include "arm_compute/core/Validate.h" |
| 29 | #include "arm_compute/runtime/NEON/NEScheduler.h" |
| 30 | |
| 31 | #include <cmath> |
| 32 | #include <tuple> |
| 33 | |
| 34 | using namespace arm_compute; |
| 35 | |
| 36 | NEDirectConvolutionLayer::NEDirectConvolutionLayer() |
| 37 | : _accumulate_bias_kernel(), _conv_kernel(), _input_border_handler(), _accumulator() |
| 38 | { |
| 39 | } |
| 40 | |
| 41 | void NEDirectConvolutionLayer::configure(ITensor *input, const ITensor *weights, const ITensor *bias, ITensor *output, const PadStrideInfo &conv_info) |
| 42 | { |
| 43 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::QS8, DataType::F32); |
| 44 | |
| 45 | // Free accumulator |
| 46 | if(_accumulator.buffer() != nullptr) |
| 47 | { |
| 48 | _accumulator.allocator()->free(); |
| 49 | } |
| 50 | |
| 51 | // Allocate the intermediate accumulator tensor in case of fixed point input |
| 52 | if(output->info()->data_type() == DataType::QS8) |
| 53 | { |
| 54 | _accumulator.allocator()->init(TensorInfo(output->info()->tensor_shape(), 1, DataType::QS16, output->info()->fixed_point_position())); |
| 55 | _conv_kernel.configure(input, weights, &_accumulator, conv_info); |
| 56 | _accumulate_bias_kernel.configure(&_accumulator, bias, output); |
| 57 | _accumulator.allocator()->allocate(); |
| 58 | } |
| 59 | else |
| 60 | { |
| 61 | _conv_kernel.configure(input, weights, output, conv_info); |
| 62 | _accumulate_bias_kernel.configure(output, bias); |
| 63 | } |
| 64 | |
| 65 | // Add zero padding XY |
| 66 | _input_border_handler.configure(input, _conv_kernel.border_size(), BorderMode::CONSTANT, PixelValue(0)); |
| 67 | } |
| 68 | |
| 69 | void NEDirectConvolutionLayer::run() |
| 70 | { |
| 71 | _input_border_handler.run(_input_border_handler.window()); |
| 72 | |
| 73 | NEScheduler::get().schedule(&_conv_kernel, Window::DimZ); |
| 74 | NEScheduler::get().schedule(&_accumulate_bias_kernel, Window::DimY); |
| 75 | } |