diff options
| author | Michalis Spyrou <michalis.spyrou@arm.com> | 2020-10-21 00:04:14 +0100 |
|---|---|---|
| committer | Michalis Spyrou <michalis.spyrou@arm.com> | 2020-11-03 15:10:47 +0000 |
| commit | ebcebf1dee7f8314976b1e0cabd62b4cf893d765 (patch) | |
| tree | 95d3e691a0e88a3e213a1d30446a9224497f2055 /src/runtime/NEON/functions/NESpaceToBatchLayer.cpp | |
| parent | da4b1b2055d96aaf73704eb9b0b82d74dc2d699c (diff) | |
| download | ComputeLibrary-ebcebf1dee7f8314976b1e0cabd62b4cf893d765.tar.gz | |
COMPMID-3638: Move NEON kernels
Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com>
Change-Id: Ieed3e4bc8be7fef80c90c5094599b477a56fc473
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4285
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/runtime/NEON/functions/NESpaceToBatchLayer.cpp')
| -rw-r--r-- | src/runtime/NEON/functions/NESpaceToBatchLayer.cpp | 27 |
1 files changed, 18 insertions, 9 deletions
diff --git a/src/runtime/NEON/functions/NESpaceToBatchLayer.cpp b/src/runtime/NEON/functions/NESpaceToBatchLayer.cpp index 97e793f6fb..516e8d604c 100644 --- a/src/runtime/NEON/functions/NESpaceToBatchLayer.cpp +++ b/src/runtime/NEON/functions/NESpaceToBatchLayer.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019 Arm Limited. + * Copyright (c) 2019-2020 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -29,9 +29,14 @@ #include "arm_compute/core/Types.h" #include "arm_compute/core/Validate.h" #include "arm_compute/runtime/NEON/NEScheduler.h" +#include "src/core/NEON/kernels/NEMemsetKernel.h" +#include "src/core/NEON/kernels/NESpaceToBatchLayerKernel.h" +#include "support/MemorySupport.h" namespace arm_compute { +NESpaceToBatchLayer::~NESpaceToBatchLayer() = default; + NESpaceToBatchLayer::NESpaceToBatchLayer() : _space_to_batch_kernel(), _memset_kernel(), _has_padding(false) { @@ -43,10 +48,12 @@ void NESpaceToBatchLayer::configure(const ITensor *input, const ITensor *block_s if(input->info()->tensor_shape().total_size() != output->info()->tensor_shape().total_size()) { - _has_padding = true; - _memset_kernel.configure(output, PixelValue(0, input->info()->data_type(), input->info()->quantization_info())); + _has_padding = true; + _memset_kernel = arm_compute::support::cpp14::make_unique<NEMemsetKernel>(); + _memset_kernel->configure(output, PixelValue(0, input->info()->data_type(), input->info()->quantization_info())); } - _space_to_batch_kernel.configure(input, block_shape, paddings, output); + _space_to_batch_kernel = arm_compute::support::cpp14::make_unique<NESpaceToBatchLayerKernel>(); + _space_to_batch_kernel->configure(input, block_shape, paddings, output); } void NESpaceToBatchLayer::configure(const ITensor *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right, ITensor *output) @@ -55,10 +62,12 @@ void NESpaceToBatchLayer::configure(const ITensor *input, const int block_shape_ if(input->info()->tensor_shape().total_size() != output->info()->tensor_shape().total_size()) { - _has_padding = true; - _memset_kernel.configure(output, PixelValue(0, input->info()->data_type(), input->info()->quantization_info())); + _has_padding = true; + _memset_kernel = arm_compute::support::cpp14::make_unique<NEMemsetKernel>(); + _memset_kernel->configure(output, PixelValue(0, input->info()->data_type(), input->info()->quantization_info())); } - _space_to_batch_kernel.configure(input, block_shape_x, block_shape_y, padding_left, padding_right, output); + _space_to_batch_kernel = arm_compute::support::cpp14::make_unique<NESpaceToBatchLayerKernel>(); + _space_to_batch_kernel->configure(input, block_shape_x, block_shape_y, padding_left, padding_right, output); } Status NESpaceToBatchLayer::validate(const ITensorInfo *input, const ITensorInfo *block_shape, const ITensorInfo *paddings, const ITensorInfo *output) @@ -81,8 +90,8 @@ void NESpaceToBatchLayer::run() // Zero out output only if we have paddings if(_has_padding) { - NEScheduler::get().schedule(&_memset_kernel, Window::DimY); + NEScheduler::get().schedule(_memset_kernel.get(), Window::DimY); } - NEScheduler::get().schedule(&_space_to_batch_kernel, Window::DimY); + NEScheduler::get().schedule(_space_to_batch_kernel.get(), Window::DimY); } } // namespace arm_compute |