diff options
| author | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2022-06-14 15:13:16 +0100 |
|---|---|---|
| committer | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2022-06-15 14:47:02 +0000 |
| commit | 0c687044c3f5d7f294858debfc7c4c070228a9b4 (patch) | |
| tree | ed0794cc6e2b56e40eda7aa44e83139c46241213 /src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp | |
| parent | 894659a98e76d84bf209da27d8ecb6d9ed05b13d (diff) | |
| download | ComputeLibrary-0c687044c3f5d7f294858debfc7c4c070228a9b4.tar.gz | |
Fix performance regression in Winograd Output Transform (OpenCL)
The regression was caused by NUM_TILES_X passed at runtime.
Resolves COMPMID-5327
Change-Id: Id6ccd93784eda93af09f420c0d786050e2bbccd7
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/7727
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp')
| -rw-r--r-- | src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp b/src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp index ff57c83959..a664d1ec1d 100644 --- a/src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp +++ b/src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp @@ -187,6 +187,7 @@ void ClWinogradOutputTransformKernel::configure(const ClCompileContext &compile_ build_opts.add_option_if(total_batches > 1, "-DSRC_DEPTH=" + support::cpp11::to_string(src->dimension(2))); build_opts.add_option_if(winograd_info.kernel_size.height == 1, "-DWINOGRAD_OUTPUT_TRANSFORM_HORIZONTAL"); build_opts.add_option_if(winograd_info.kernel_size.width == 1, "-DWINOGRAD_OUTPUT_TRANSFORM_VERTICAL"); + build_opts.add_option("-DNUM_TILES_X=" + support::cpp11::to_string(_num_tiles_x)); } else { @@ -279,7 +280,6 @@ void ClWinogradOutputTransformKernel::run_op(ITensorPack &tensors, const Window _kernel.setArg<cl_int>(idx2++, _src_height); _kernel.setArg<cl_int>(idx2++, _dst_width); _kernel.setArg<cl_int>(idx2++, _dst_height); - _kernel.setArg<cl_int>(idx2++, _num_tiles_x); } do |