From 6c6e77afc0059e4c5a59d97215acccdedf473a7f Mon Sep 17 00:00:00 2001 From: Pablo Tello Date: Tue, 23 Jan 2018 10:03:27 +0000 Subject: COMPMID-784: Doxygen fixes Change-Id: I35f429fbf08dece7c759242c37e0a68b0851ce49 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/117231 Reviewed-by: Anthony Barbier Tested-by: Jenkins --- .../core/NEON/kernels/NEWinogradLayerKernel.h | 92 ++++++++++++++-------- arm_compute/core/NEON/kernels/winograd/perf.h | 23 ++++++ .../core/NEON/kernels/winograd/winograd_layer.hpp | 88 ++++++++++++++------- docs/Doxyfile | 3 +- src/core/CL/kernels/CLPermuteKernel.cpp | 3 +- src/core/CPP/kernels/CPPPermuteKernel.cpp | 12 +-- src/core/NEON/kernels/NEWinogradLayerKernel.cpp | 2 +- 7 files changed, 153 insertions(+), 70 deletions(-) diff --git a/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h b/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h index 95261929ca..78ac56418f 100644 --- a/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h +++ b/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h @@ -36,21 +36,35 @@ class NEWinogradLayerKernel; class Winograd3x3F32 final { public: + /** Create a new Winograd convolution layer. + * + * @param[in] n_batches Number of batches in the input and output tensors. + * @param[in] n_input_channels Number of feature maps in a batch of the input tensor. + * @param[in] n_input_rows Number of rows in a feature map of the input tensor. + * @param[in] n_input_cols Number of columns in a feature map of the input tensor. + * @param[in] n_output_channels Number of feature maps in the output tensor. + * @param[in] same_padding Use "SAME" padding, otherwise use "VALID". + * @param[in] weights Pointer to weight tensor in spatial domain. Must be ordered as "Height x Rows x Input Feature Maps x Output Feature Maps. + * @param[out] weights_storage Pointer to storage for weight tensor in the Winograd domain. Must be at least the size returned by `get_weight_storage_size + * @param[in] input Pointer to NHWC ordered input tensor, in the spatial domain. + * @param[out] winograd_input Pointer to working space for the input tensor in the Winograd domain. Must be at least the size returned by `get_input_storage_size`. + * @param[out] output Pointer to NHWC ordered output tensor, in the spatial domain. + * @param[out] winograd_output Pointer to working space for the output tensor in the Winograd domain. Must be at least the size returned by `get_output_storage_size`. + */ friend class NEWinogradLayerKernel; Winograd3x3F32( - const int n_batches, /** Number of batches in the input and output tensors. */ - const int n_input_channels, /** Number of feature maps in a batch of the input tensor. */ - const int n_input_rows, /** Number of rows in a feature map of the input tensor. */ - const int n_input_cols, /** Number of columns in a feature map of the input tensor. */ - const int n_output_channels, /** Number of feature maps in the output tensor. */ - const bool same_padding, /** Use "SAME" padding, otherwise use "VALID". */ - const float *const weights, /** Pointer to weight tensor in spatial domain. Must be ordered as "Height x Rows x Input Feature Maps x Output Feature Maps. */ - float *const weights_storage, /** Pointer to storage for weight tensor in the Winograd domain. Must be at least the size returned by `get_weight_storage_size`. */ - const float *const input, /** Pointer to NHWC ordered input tensor, in the spatial domain. */ - float *const winograd_input, /** Pointer to working space for the input tensor in the Winograd domain. Must be at least the size returned by `get_input_storage_size`. */ - float *const output, /** Pointer to NHWC ordered output tensor, in the spatial domain. */ - float *const winograd_output /** Pointer to working space for the output tensor in the Winograd domain. Must be at least the size returned by `get_output_storage_size`. */ - ); + const int n_batches, + const int n_input_channels, + const int n_input_rows, + const int n_input_cols, + const int n_output_channels, + const bool same_padding, + const float *const weights, + float *const weights_storage, + const float *const input, + float *const winograd_input, + float *const output, + float *const winograd_output); ~Winograd3x3F32(); void transform_weights(); @@ -88,31 +102,47 @@ public: // Inherited methods overridden: void run(const Window &window, const ThreadInfo &info) override; - /* Get the memory required to instantiate a new Winograd operator. - */ - static size_t get_weight_storage_size( - const int n_output_channels, /** Number of output feature maps. */ - const int n_input_channels /** Number of input feature maps. */ - ); + /** Determine how much memory (in units of TIn) to allocate for the + * transformed weights. + * + * @param[in] n_output_channels Number of output feature maps. + * @param[in] n_input_channels Number of input feature maps. + */ + static unsigned int get_weight_storage_size( + const int n_output_channels, + const int n_input_channels); + /** Determine how much memory (in units of TIn) to allocate for the + * transformed input. + * + * @param[in] n_batches Number of batches in the input tensor. + * @param[in] n_channels Number of feature maps in the input tensor. + * @param[in] n_rows Number of rows in each feature map. + * @param[in] n_cols Number of columns in each feature map. + * @param[in] same_padding Use "SAME" padding, otherwise use "VALID". + */ static unsigned int get_input_storage_size( - const int n_batches, /** Number of batches in the input tensor. */ - const int n_channels, /** Number of feature maps in the input tensor. */ - const int n_rows, /** Number of rows in each feature map. */ - const int n_cols, /** Number of columns in each feature map. */ - const bool same_padding /** Use "SAME" padding, otherwise use "VALID". */ - ); + const int n_batches, + const int n_channels, + const int n_rows, + const int n_cols, + const bool same_padding); /** Determine how much memory (in units of TOut) to allocate for the * (Winograd domain) output. + * + * @param[in] n_batches Number of batches in the output tensor. + * @param[in] n_rows Number of rows in each feature map of the input tensor. + * @param[in] n_cols Number of columns in each feature map of the input tensor. + * @param[in] n_output_channels Number of feature maps in the output tensor. + * @param[in] same_padding Use "SAME" padding, otherwise use "VALID". */ static unsigned int get_output_storage_size( - const int n_batches, /** Number of batches in the output tensor. */ - const int n_rows, /** Number of rows in each feature map of the input tensor. */ - const int n_cols, /** Number of columns in each feature map of the input tensor. */ - const int n_output_channels, /** Number of feature maps in the output tensor. */ - const bool same_padding /** Use "SAME" padding, otherwise use "VALID". */ - ); + const int n_batches, + const int n_rows, + const int n_cols, + const int n_output_channels, + const bool same_padding); protected: Winograd3x3F32 *_convolver; diff --git a/arm_compute/core/NEON/kernels/winograd/perf.h b/arm_compute/core/NEON/kernels/winograd/perf.h index 0cdf742a25..3c0d36646d 100644 --- a/arm_compute/core/NEON/kernels/winograd/perf.h +++ b/arm_compute/core/NEON/kernels/winograd/perf.h @@ -1,3 +1,26 @@ +/* + * Copyright (c) 2018 ARM Limited. + * + * SPDX-License-Identifier: MIT + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ #pragma once /* Prototypes from perf.c */ diff --git a/arm_compute/core/NEON/kernels/winograd/winograd_layer.hpp b/arm_compute/core/NEON/kernels/winograd/winograd_layer.hpp index a3b3db42dd..4559312df4 100644 --- a/arm_compute/core/NEON/kernels/winograd/winograd_layer.hpp +++ b/arm_compute/core/NEON/kernels/winograd/winograd_layer.hpp @@ -74,55 +74,87 @@ class WinogradConvolutionLayer /** Determine how much memory (in units of TIn) to allocate for the * transformed weights. + * + * @param[in] n_output_channels Number of output feature maps. + * @param[in] n_input_channels Number of input feature maps. */ static unsigned int get_weight_storage_size( - const int n_output_channels, /** Number of output feature maps. */ - const int n_input_channels /** Number of input feature maps. */ + const int n_output_channels, + const int n_input_channels ); /** Determine how much memory (in units of TIn) to allocate for the * transformed input. + * + * @param[in] n_batches Number of batches in the input tensor. + * @param[in] n_channels Number of feature maps in the input tensor. + * @param[in] n_rows Number of rows in each feature map. + * @param[in] n_cols Number of columns in each feature map. + * @param[in] same_padding Use "SAME" padding, otherwise use "VALID". */ static unsigned int get_input_storage_size( - const int n_batches, /** Number of batches in the input tensor. */ - const int n_channels, /** Number of feature maps in the input tensor. */ - const int n_rows, /** Number of rows in each feature map. */ - const int n_cols, /** Number of columns in each feature map. */ - const bool same_padding /** Use "SAME" padding, otherwise use "VALID". */ + const int n_batches, + const int n_channels, + const int n_rows, + const int n_cols, + const bool same_padding ); /** Determine how much memory (in units of TOut) to allocate for the * (Winograd domain) output. + * + * @param[in] n_batches Number of batches in the output tensor. + * @param[in] n_rows Number of rows in each feature map of the input tensor. + * @param[in] n_cols Number of columns in each feature map of the input tensor. + * @param[in] n_output_channels Number of feature maps in the output tensor. + * @param[in] same_padding Use "SAME" padding, otherwise use "VALID". */ static unsigned int get_output_storage_size( - const int n_batches, /** Number of batches in the output tensor. */ - const int n_rows, /** Number of rows in each feature map of the input tensor. */ - const int n_cols, /** Number of columns in each feature map of the input tensor. */ - const int n_output_channels, /** Number of feature maps in the output tensor. */ - const bool same_padding /** Use "SAME" padding, otherwise use "VALID". */ + const int n_batches, + const int n_rows, + const int n_cols, + const int n_output_channels, + const bool same_padding ); - /** Get the shape (rows, cols) of a feature map of the output tensor. */ + /** Get the shape (rows, cols) of a feature map of the output tensor. + * + * @param[in] n_input_rows Number of rows in the input feature map. + * @param[in] n_input_cols Number of columns in the input feature map. + * @param[in] same_padding Use "SAME" padding, otherwise use "VALID". + */ static std::pair get_output_feature_map_shape( - const int n_input_rows, /** Number of rows in the input feature map. */ - const int n_input_cols, /** Number of columns in the input feature map. */ - const bool same_padding /** Use "SAME" padding, otherwise use "VALID". */ + const int n_input_rows, + const int n_input_cols, + const bool same_padding ); /** Create a new Winograd convolution layer. + * @param[in] n_batches Number of batches in the input and output tensors. + * @param[in] n_input_channels Number of feature maps in a batch of the input tensor. + * @param[in] n_input_rows Number of rows in a feature map of the input tensor. + * @param[in] n_input_cols Number of columns in a feature map of the input tensor. + * @param[in] n_output_channels Number of feature maps in the output tensor. + * @param[in] same_padding Use "SAME" padding, otherwise use "VALID". + * @param[in] weights Pointer to weight tensor in spatial domain. Must be ordered as "Height x Rows x Input Feature Maps x Output Feature Maps. + * @param[out] weights_storage Pointer to storage for weight tensor in the Winograd domain. Must be at least the size returned by `get_weight_storage_size + * @param[in] input Pointer to NHWC ordered input tensor, in the spatial domain. + * @param[out] winograd_input Pointer to working space for the input tensor in the Winograd domain. Must be at least the size returned by `get_input_storage_size`. + * @param[out] output Pointer to NHWC ordered output tensor, in the spatial domain. + * @param[out] winograd_output Pointer to working space for the output tensor in the Winograd domain. Must be at least the size returned by `get_output_storage_size`. */ WinogradConvolutionLayer( - const int n_batches, /** Number of batches in the input and output tensors. */ - const int n_input_channels, /** Number of feature maps in a batch of the input tensor. */ - const int n_input_rows, /** Number of rows in a feature map of the input tensor. */ - const int n_input_cols, /** Number of columns in a feature map of the input tensor. */ - const int n_output_channels, /** Number of feature maps in the output tensor. */ - const bool same_padding, /** Use "SAME" padding, otherwise use "VALID". */ - const TIn* const weights, /** Pointer to weight tensor in spatial domain. Must be ordered as "Height x Rows x Input Feature Maps x Output Feature Maps. */ - TIn* const weights_storage, /** Pointer to storage for weight tensor in the Winograd domain. Must be at least the size returned by `get_weight_storage_size`. */ - const TIn* const input, /** Pointer to NHWC ordered input tensor, in the spatial domain. */ - TIn* const winograd_input, /** Pointer to working space for the input tensor in the Winograd domain. Must be at least the size returned by `get_input_storage_size`. */ - TOut* const output, /** Pointer to NHWC ordered output tensor, in the spatial domain. */ - TOut* const winograd_output /** Pointer to working space for the output tensor in the Winograd domain. Must be at least the size returned by `get_output_storage_size`. */ + const int n_batches, + const int n_input_channels, + const int n_input_rows, + const int n_input_cols, + const int n_output_channels, + const bool same_padding, + const TIn* const weights, + TIn* const weights_storage, + const TIn* const input, + TIn* const winograd_input, + TOut* const output, + TOut* const winograd_output ); }; diff --git a/docs/Doxyfile b/docs/Doxyfile index de0320f38b..59aa780fd3 100644 --- a/docs/Doxyfile +++ b/docs/Doxyfile @@ -855,7 +855,8 @@ RECURSIVE = YES # Note that relative paths are relative to the directory from which doxygen is # run. -EXCLUDE = +EXCLUDE = ./arm_compute/core/NEON/kernels/assembly/ \ + ./arm_compute/core/NEON/kernels/winograd/ # The EXCLUDE_SYMLINKS tag can be used to select whether or not files or # directories that are symbolic links (a Unix file system feature) are excluded diff --git a/src/core/CL/kernels/CLPermuteKernel.cpp b/src/core/CL/kernels/CLPermuteKernel.cpp index 7001c5565b..132de60b68 100644 --- a/src/core/CL/kernels/CLPermuteKernel.cpp +++ b/src/core/CL/kernels/CLPermuteKernel.cpp @@ -46,7 +46,8 @@ TensorShape get_output_shape(const ITensorInfo *input, const PermutationVector & permute(output_shape, perm); return output_shape; } -} +} // namespace + void CLPermuteKernel::configure(const ICLTensor *input, ICLTensor *output, const PermutationVector &perm) { ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::QASYMM8, diff --git a/src/core/CPP/kernels/CPPPermuteKernel.cpp b/src/core/CPP/kernels/CPPPermuteKernel.cpp index c7bae870d1..298c700809 100644 --- a/src/core/CPP/kernels/CPPPermuteKernel.cpp +++ b/src/core/CPP/kernels/CPPPermuteKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017, 2018 ARM Limited. + * Copyright (c) 2017-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -76,17 +76,13 @@ inline void permute_strides(Dimensions &dimensions, const PermutationVector & } // namespace - - - template void CPPPermuteKernel::run_permute(const Window &window) { - - Strides strides = _output->info()->strides_in_bytes(); + Strides strides = _output->info()->strides_in_bytes(); Strides perm_strides = strides; - permute_strides(perm_strides,_perm); - const int output_stride_w = strides[3]; + permute_strides(perm_strides, _perm); + const int output_stride_w = strides[3]; Window window_out(window); const Window::Dimension zero_window = Window::Dimension(0, 0, 0); for(size_t d = 0; d <= _perm.num_dimensions(); ++d) diff --git a/src/core/NEON/kernels/NEWinogradLayerKernel.cpp b/src/core/NEON/kernels/NEWinogradLayerKernel.cpp index 24d72eddd8..ea48e1f32b 100644 --- a/src/core/NEON/kernels/NEWinogradLayerKernel.cpp +++ b/src/core/NEON/kernels/NEWinogradLayerKernel.cpp @@ -118,7 +118,7 @@ unsigned int NEWinogradLayerKernel::get_output_storage_size( return T::get_output_storage_size(n_batches, n_rows, n_cols, n_output_channels, same_padding); } -size_t NEWinogradLayerKernel::get_weight_storage_size(const int n_output_channels, const int n_input_channels) +unsigned int NEWinogradLayerKernel::get_weight_storage_size(const int n_output_channels, const int n_input_channels) { return T::get_weight_storage_size(n_output_channels, n_input_channels); } -- cgit v1.2.1