diff options
author | Giorgio Arena <giorgio.arena@arm.com> | 2021-06-28 11:00:27 +0100 |
---|---|---|
committer | Giorgio Arena <giorgio.arena@arm.com> | 2021-06-29 11:06:10 +0000 |
commit | 4a95bba6ca61ce99995ece6fd237b5498c9f322c (patch) | |
tree | 04ef2dee1cdc015ce14ce6b664d9a956a96c3a05 /src/core/gpu/cl/kernels/ClSoftmaxKernel.cpp | |
parent | 731fe667e3009bfbfee8b0eb74ecb68b291e4311 (diff) | |
download | ComputeLibrary-4a95bba6ca61ce99995ece6fd237b5498c9f322c.tar.gz |
Set up the framework to choose the default LWS
Resolve COMPMID-4486
Signed-off-by: Giorgio Arena <giorgio.arena@arm.com>
Change-Id: Ib38b7943bd776a6d75d1da163908724c49eae73d
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5864
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/gpu/cl/kernels/ClSoftmaxKernel.cpp')
-rw-r--r-- | src/core/gpu/cl/kernels/ClSoftmaxKernel.cpp | 10 |
1 files changed, 10 insertions, 0 deletions
diff --git a/src/core/gpu/cl/kernels/ClSoftmaxKernel.cpp b/src/core/gpu/cl/kernels/ClSoftmaxKernel.cpp index 000c9ad04d..1dd905d66e 100644 --- a/src/core/gpu/cl/kernels/ClSoftmaxKernel.cpp +++ b/src/core/gpu/cl/kernels/ClSoftmaxKernel.cpp @@ -154,6 +154,11 @@ const unsigned int ClLogits1DMaxShiftExpSumKernel::_serial_vector_size = 8; /**< Vector size in the parallel case (obtained through auto-tuning, enables the best memory access pattern for Bifrost) .*/ const unsigned int ClLogits1DMaxShiftExpSumKernel::_parallel_vector_size = 4; +ClLogits1DMaxShiftExpSumKernel::ClLogits1DMaxShiftExpSumKernel() +{ + _type = CLKernelType::ELEMENTWISE; +} + void ClLogits1DMaxShiftExpSumKernel::configure(const CLCompileContext &compile_context, const ITensorInfo &src, ITensorInfo &max, ITensorInfo &dst, ITensorInfo &sum, const SoftmaxKernelInfo &info) { auto padding_info = get_padding_info({ &src, &max, &dst, &sum }); @@ -273,6 +278,11 @@ void ClLogits1DMaxShiftExpSumKernel::run_op(ITensorPack &tensors, const Window & while(window_collapsed.slide_window_slice_3D(slice)); } +ClLogits1DNormKernel::ClLogits1DNormKernel() +{ + _type = CLKernelType::ELEMENTWISE; +} + void ClLogits1DNormKernel::configure(const CLCompileContext &compile_context, const ITensorInfo &src, const ITensorInfo &sum, ITensorInfo &dst, const SoftmaxKernelInfo &info) { auto padding_info = get_padding_info({ &src, &dst, &sum }); |