diff options
| author | Giorgio Arena <giorgio.arena@arm.com> | 2018-04-16 11:20:11 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:49:37 +0000 |
| commit | dd03870b63784abe499761da2b26b209b33f2db2 (patch) | |
| tree | 6ecc9650493c62f70f45cccb859e77090f895d6a /src/core/CL/kernels/CLWinogradOutputTransformKernel.cpp | |
| parent | bb54e4e40b7b08c509e234cd91ebd3087af66c23 (diff) | |
| download | ComputeLibrary-dd03870b63784abe499761da2b26b209b33f2db2.tar.gz | |
COMPMID-1037 Add support for F(4x4, 5x5) in CLWinogradOutputTransformKernel
Change-Id: I0b126f03028f08687497b0d79d2e2764f7ed07c8
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/128001
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Diffstat (limited to 'src/core/CL/kernels/CLWinogradOutputTransformKernel.cpp')
| -rw-r--r-- | src/core/CL/kernels/CLWinogradOutputTransformKernel.cpp | 5 |
1 files changed, 3 insertions, 2 deletions
diff --git a/src/core/CL/kernels/CLWinogradOutputTransformKernel.cpp b/src/core/CL/kernels/CLWinogradOutputTransformKernel.cpp index b59bc79327..8ee1a82209 100644 --- a/src/core/CL/kernels/CLWinogradOutputTransformKernel.cpp +++ b/src/core/CL/kernels/CLWinogradOutputTransformKernel.cpp @@ -56,8 +56,9 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, con const Size2D kernel_size = winograd_info.kernel_size; const Size2D input_dimensions = winograd_info.input_dimensions; - ARM_COMPUTE_RETURN_ERROR_ON_MSG(kernel_size != Size2D(3U, 3U), "Only 3x3 kernels are supported"); - ARM_COMPUTE_RETURN_ERROR_ON_MSG(input->dimension(2) != 16, "Only 2x2 output tile is supported"); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(kernel_size != Size2D(3U, 3U) && kernel_size != Size2D(5U, 5U), "Only 3x3 and 5x5 kernels are supported"); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(kernel_size == Size2D(3U, 3U) && output_tile_size == Size2D(2U, 2U) && input->dimension(2) != 16, "Wrong number of batches"); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(kernel_size == Size2D(5U, 5U) && output_tile_size == Size2D(4U, 4U) && input->dimension(2) != 64, "Wrong number of batches"); // Compute number of elements to process in the X and Y direction const int num_elements_x = input_dimensions.width - (kernel_size.width - 1) + conv_info.pad_left() + conv_info.pad_right(); |