diff options
| author | Pablo Tello <pablo.tello@arm.com> | 2018-01-09 10:30:27 +0000 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:42:33 +0000 |
| commit | 4e2c139dae6afeda11da165bfff8fba60d2e84f3 (patch) | |
| tree | f3fecbeaa19a2e8fe2ac37f66d948ea96055f683 /arm_compute | |
| parent | e671d6a4d0ba5cfa86ae1856433eab597bd996d2 (diff) | |
| download | ComputeLibrary-4e2c139dae6afeda11da165bfff8fba60d2e84f3.tar.gz | |
COMPMID-784: Winograd refactoring
Removed the code that created a subtensor and imported memory from the workspace in the function run() method.
The subtensor is no longer needed because we perform the reordering of the tensors with NEPermute. The call to the method
winograd::Winograd2x2_3x3GEMM<TOut, TIn>::reshape_output() will transform the results from the winograd domain
into the spatial domain and this will be stored in the member _output_nhwc.
Change-Id: Iae09d26c7587cd2eed98968c3ce214e20031038e
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/115483
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'arm_compute')
| -rw-r--r-- | arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h | 7 |
1 files changed, 2 insertions, 5 deletions
diff --git a/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h b/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h index 9ab2e4de11..73b7e8d2b7 100644 --- a/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h +++ b/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017 ARM Limited. + * Copyright (c) 2017, 2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -32,18 +32,15 @@ namespace arm_compute class ITensor; class NEWinogradLayerKernel; -class Winograd3x3F32 +class Winograd3x3F32 final { public: friend class NEWinogradLayerKernel; Winograd3x3F32(const KernelShape &kernel_shape, const Tensor4DShape input_shape, const PaddingType padding_type, void *kernel_storage); ~Winograd3x3F32(); - std::pair<void *, void *> get_nhwc_ptrs(const Tensor4DShape &input_shape, const PaddingType padding_type, void *working_space); void transform_weights(const void *const kernel, void *transform_working_space); void reshape_input(const Tensor4DShape &input_shape, const PaddingType padding_type, const void *const input, void *working_space); void reshape_output(const Tensor4DShape &input_shape, const PaddingType padding_type, void *const output); - void nchw2nhwc(const Tensor4DShape &input_shape, const PaddingType padding_type, void *working_space, const void *const input); - void nhwc2nchw(const Tensor4DShape &input_shape, const PaddingType padding_type, void *working_space, void *const output); private: class Private; |