From afd38f0c617d6f89b2b4532c6c44f116617e2b6f Mon Sep 17 00:00:00 2001 From: Felix Thomasmathibalan Date: Wed, 27 Sep 2023 17:46:17 +0100 Subject: Apply clang-format on repository Code is formatted as per a revised clang format configuration file(not part of this delivery). Version 14.0.6 is used. Exclusion List: - files with .cl extension - files that are not strictly C/C++ (e.g. Android.bp, Sconscript ...) And the following directories - compute_kernel_writer/validation/ - tests/ - include/ - src/core/NEON/kernels/convolution/ - src/core/NEON/kernels/arm_gemm/ - src/core/NEON/kernels/arm_conv/ - data/ There will be a follow up for formatting of .cl files and the files under tests/ and compute_kernel_writer/validation/. Signed-off-by: Felix Thomasmathibalan Change-Id: Ib7eb1fcf4e7537b9feaefcfc15098a804a3fde0a Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10391 Benchmark: Arm Jenkins Tested-by: Arm Jenkins Reviewed-by: Gunes Bayir --- .../NEON/functions/NEDeconvolutionLayer.cpp | 116 +++++++++++++-------- 1 file changed, 75 insertions(+), 41 deletions(-) (limited to 'src/runtime/NEON/functions/NEDeconvolutionLayer.cpp') diff --git a/src/runtime/NEON/functions/NEDeconvolutionLayer.cpp b/src/runtime/NEON/functions/NEDeconvolutionLayer.cpp index 439aff0840..3987370d9e 100644 --- a/src/runtime/NEON/functions/NEDeconvolutionLayer.cpp +++ b/src/runtime/NEON/functions/NEDeconvolutionLayer.cpp @@ -25,9 +25,10 @@ #include "arm_compute/core/Helpers.h" #include "arm_compute/core/Utils.h" -#include "arm_compute/core/Validate.h" #include "arm_compute/core/utils/misc/ShapeCalculator.h" +#include "arm_compute/core/Validate.h" #include "arm_compute/runtime/NEON/NEScheduler.h" + #include "src/common/utils/Log.h" #include "src/core/helpers/AutoConfiguration.h" @@ -61,7 +62,8 @@ PadStrideInfo compute_upsample_info(const PadStrideInfo &info, uint32_t deconv_p deconv_pad_top += deconv_pad_y / 2; deconv_pad_bottom += deconv_pad_y / 2; - return PadStrideInfo(stride_x, stride_y, deconv_pad_left, deconv_pad_right, deconv_pad_top, deconv_pad_bottom, DimensionRoundingType::FLOOR); + return PadStrideInfo(stride_x, stride_y, deconv_pad_left, deconv_pad_right, deconv_pad_top, deconv_pad_bottom, + DimensionRoundingType::FLOOR); } } // namespace @@ -82,17 +84,24 @@ NEDeconvolutionLayer::NEDeconvolutionLayer(std::shared_ptr memor { } -Status NEDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, const ITensorInfo *output, const PadStrideInfo &info, - bool enable_fast_math, const WeightsInfo &weights_info) +Status NEDeconvolutionLayer::validate(const ITensorInfo *input, + const ITensorInfo *weights, + const ITensorInfo *bias, + const ITensorInfo *output, + const PadStrideInfo &info, + bool enable_fast_math, + const WeightsInfo &weights_info) { ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, weights, output); - ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32, DataType::F16, DataType::QASYMM8, DataType::QASYMM8_SIGNED); - const unsigned int width_idx = get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::WIDTH); - const unsigned int height_idx = get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::HEIGHT); + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32, DataType::F16, DataType::QASYMM8, + DataType::QASYMM8_SIGNED); + const unsigned int width_idx = get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::WIDTH); + const unsigned int height_idx = + get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::HEIGHT); ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(width_idx) < 1); ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(height_idx) < 1); ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_LAYOUT(weights, input); - if(is_data_type_quantized_per_channel(weights->data_type()) && is_data_type_quantized(input->data_type())) + if (is_data_type_quantized_per_channel(weights->data_type()) && is_data_type_quantized(input->data_type())) { ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(weights, 1, DataType::QSYMM8_PER_CHANNEL); } @@ -101,11 +110,13 @@ Status NEDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInf ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, weights); } - auto out_dims = deconvolution_output_dimensions(input->dimension(width_idx), input->dimension(height_idx), weights->dimension(width_idx), weights->dimension(height_idx), info); + auto out_dims = + deconvolution_output_dimensions(input->dimension(width_idx), input->dimension(height_idx), + weights->dimension(width_idx), weights->dimension(height_idx), info); - if(bias != nullptr) + if (bias != nullptr) { - if(is_data_type_quantized_asymmetric(input->data_type())) + if (is_data_type_quantized_asymmetric(input->data_type())) { ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(bias, 1, DataType::S32); } @@ -115,15 +126,18 @@ Status NEDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInf } } - if(output->tensor_shape().total_size() > 0) + if (output->tensor_shape().total_size() > 0) { ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); const TensorShape output_shape = compute_deconvolution_output_shape(out_dims, *input, *weights); - ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(Window::DimX) != output_shape.x(), "Output's width is invalid."); - ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(Window::DimY) != output_shape.y(), "Output's height is invalid."); - ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(Window::DimZ) != output_shape.z(), "Output's depth is invalid."); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(Window::DimX) != output_shape.x(), + "Output's width is invalid."); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(Window::DimY) != output_shape.y(), + "Output's height is invalid."); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(Window::DimZ) != output_shape.z(), + "Output's depth is invalid."); } uint32_t deconv_pad_x = 0; @@ -141,44 +155,61 @@ Status NEDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInf ARM_COMPUTE_RETURN_ERROR_ON((out_x - weights->dimension(idx_w) + 1) > out_dims.first); ARM_COMPUTE_RETURN_ERROR_ON((out_y - weights->dimension(idx_h) + 1) > out_dims.second); - const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input, *weights, stride_x, stride_y, out_dims, deconv_pad_x, deconv_pad_y); - TensorInfo scale_out_info(input->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(scale_out_shape)); + const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input, *weights, stride_x, stride_y, + out_dims, deconv_pad_x, deconv_pad_y); + TensorInfo scale_out_info(input->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(scale_out_shape)); const PadStrideInfo upsample_info = compute_upsample_info(info, deconv_pad_x, deconv_pad_y); // Do not perform upsampling when the operation uses unit stride in all dimensions const bool do_upsampling = stride_x != 1 || stride_y != 1; - const unsigned int batches_idx = get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::BATCHES); - const unsigned int channel_idx = get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::CHANNEL); + const unsigned int batches_idx = + get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::BATCHES); + const unsigned int channel_idx = + get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::CHANNEL); ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(batches_idx) != scale_out_info.dimension(batches_idx)); ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(channel_idx) != scale_out_info.dimension(channel_idx)); - if(do_upsampling) + if (do_upsampling) { const PadStrideInfo conv_info(1, 1, 0, 0, 0, 0, DimensionRoundingType::CEIL); - ARM_COMPUTE_RETURN_ON_ERROR(NEConvolutionLayer::validate(&scale_out_info, weights, bias, output, conv_info, weights_info, Size2D(1U, 1U), ActivationLayerInfo(), enable_fast_math)); + ARM_COMPUTE_RETURN_ON_ERROR(NEConvolutionLayer::validate(&scale_out_info, weights, bias, output, conv_info, + weights_info, Size2D(1U, 1U), ActivationLayerInfo(), + enable_fast_math)); } else { - const PadStrideInfo conv_info(1, 1, upsample_info.pad_left(), upsample_info.pad_right(), upsample_info.pad_top(), upsample_info.pad_bottom(), DimensionRoundingType::CEIL); - ARM_COMPUTE_RETURN_ON_ERROR(NEConvolutionLayer::validate(input, weights, bias, output, conv_info, weights_info, Size2D(1U, 1U), ActivationLayerInfo(), enable_fast_math)); + const PadStrideInfo conv_info(1, 1, upsample_info.pad_left(), upsample_info.pad_right(), + upsample_info.pad_top(), upsample_info.pad_bottom(), DimensionRoundingType::CEIL); + ARM_COMPUTE_RETURN_ON_ERROR(NEConvolutionLayer::validate(input, weights, bias, output, conv_info, weights_info, + Size2D(1U, 1U), ActivationLayerInfo(), + enable_fast_math)); } return Status{}; } -void NEDeconvolutionLayer::configure(ITensor *input, const ITensor *weights, const ITensor *bias, ITensor *output, const PadStrideInfo &info, bool enable_fast_math, const WeightsInfo &weights_info) +void NEDeconvolutionLayer::configure(ITensor *input, + const ITensor *weights, + const ITensor *bias, + ITensor *output, + const PadStrideInfo &info, + bool enable_fast_math, + const WeightsInfo &weights_info) { // Perform validation step ARM_COMPUTE_ERROR_ON_NULLPTR(input, weights, output); - ARM_COMPUTE_ERROR_THROW_ON(NEDeconvolutionLayer::validate(input->info(), weights->info(), (bias == nullptr) ? nullptr : bias->info(), output->info(), info, enable_fast_math, weights_info)); + ARM_COMPUTE_ERROR_THROW_ON(NEDeconvolutionLayer::validate(input->info(), weights->info(), + (bias == nullptr) ? nullptr : bias->info(), + output->info(), info, enable_fast_math, weights_info)); ARM_COMPUTE_LOG_PARAMS(input, weights, bias, output, info, enable_fast_math, weights_info); const DataLayout data_layout = input->info()->data_layout(); const unsigned int width_idx = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH); const unsigned int height_idx = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT); - auto out_dims = deconvolution_output_dimensions(input->info()->dimension(width_idx), input->info()->dimension(height_idx), - weights->info()->dimension(width_idx), weights->info()->dimension(height_idx), info); + auto out_dims = deconvolution_output_dimensions( + input->info()->dimension(width_idx), input->info()->dimension(height_idx), + weights->info()->dimension(width_idx), weights->info()->dimension(height_idx), info); const TensorShape output_shape = compute_deconvolution_output_shape(out_dims, *input->info(), *weights->info()); @@ -191,7 +222,8 @@ void NEDeconvolutionLayer::configure(ITensor *input, const ITensor *weights, con const unsigned int stride_y = info.stride().second; // Output auto initialization if not yet initialized - auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->quantization_info()); + auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), + input->info()->quantization_info()); _flip_axis.allocator()->init(TensorInfo(TensorShape(2U), 1, DataType::U32)); @@ -199,12 +231,11 @@ void NEDeconvolutionLayer::configure(ITensor *input, const ITensor *weights, con _flip_weights.configure(weights, &_weights_flipped, &_flip_axis); // setup the function to convolve the upscaled output - uint32_t deconv_pad_x = 0; - uint32_t deconv_pad_y = 0; - const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input->info(), *weights->info(), - stride_x, stride_y, - out_dims, deconv_pad_x, deconv_pad_y); - const PadStrideInfo upsample_info = compute_upsample_info(info, deconv_pad_x, deconv_pad_y); + uint32_t deconv_pad_x = 0; + uint32_t deconv_pad_y = 0; + const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape( + *input->info(), *weights->info(), stride_x, stride_y, out_dims, deconv_pad_x, deconv_pad_y); + const PadStrideInfo upsample_info = compute_upsample_info(info, deconv_pad_x, deconv_pad_y); // Do not perform upsampling when the operation uses unit stride in all dimensions _do_upsampling = stride_x != 1 || stride_y != 1; @@ -216,12 +247,12 @@ void NEDeconvolutionLayer::configure(ITensor *input, const ITensor *weights, con axis_data[1] = static_cast(height_idx); // Setup convolution and upsampling, if needed - if(_do_upsampling) + if (_do_upsampling) { _memory_group.manage(&_scaled_output); const PadStrideInfo conv_info(1, 1, 0, 0, 0, 0, DimensionRoundingType::CEIL); - TensorInfo scale_out_info(scale_out_shape, 1, input->info()->data_type(), input->info()->quantization_info()); + TensorInfo scale_out_info(scale_out_shape, 1, input->info()->data_type(), input->info()->quantization_info()); scale_out_info.set_data_layout(data_layout); _scaled_output.allocator()->init(scale_out_info); @@ -229,14 +260,17 @@ void NEDeconvolutionLayer::configure(ITensor *input, const ITensor *weights, con // The padding amount can be given as input to the convolution layer. _upsample_f.configure(input, &_scaled_output, upsample_info); - _conv_f.configure(&_scaled_output, &_weights_flipped, bias, output, conv_info, weights_info, Size2D(1U, 1U), ActivationLayerInfo(), enable_fast_math); + _conv_f.configure(&_scaled_output, &_weights_flipped, bias, output, conv_info, weights_info, Size2D(1U, 1U), + ActivationLayerInfo(), enable_fast_math); _scaled_output.allocator()->allocate(); } else { - const PadStrideInfo conv_info(1, 1, upsample_info.pad_left(), upsample_info.pad_right(), upsample_info.pad_top(), upsample_info.pad_bottom(), DimensionRoundingType::CEIL); - _conv_f.configure(input, &_weights_flipped, bias, output, conv_info, weights_info, Size2D(1U, 1U), ActivationLayerInfo(), enable_fast_math); + const PadStrideInfo conv_info(1, 1, upsample_info.pad_left(), upsample_info.pad_right(), + upsample_info.pad_top(), upsample_info.pad_bottom(), DimensionRoundingType::CEIL); + _conv_f.configure(input, &_weights_flipped, bias, output, conv_info, weights_info, Size2D(1U, 1U), + ActivationLayerInfo(), enable_fast_math); } } @@ -246,7 +280,7 @@ void NEDeconvolutionLayer::run() MemoryGroupResourceScope scope_mg(_memory_group); - if(_do_upsampling) + if (_do_upsampling) { _upsample_f.run(); } @@ -255,7 +289,7 @@ void NEDeconvolutionLayer::run() void NEDeconvolutionLayer::prepare() { - if(!_is_prepared) + if (!_is_prepared) { ARM_COMPUTE_ERROR_ON(!_original_weights->is_used()); -- cgit v1.2.1