diff options
author | SiCongLi <sicong.li@arm.com> | 2021-10-29 15:05:49 +0100 |
---|---|---|
committer | SiCong Li <sicong.li@arm.com> | 2021-11-01 14:29:51 +0000 |
commit | eb8bd81a625f0f87080dbde55b434362ad57324a (patch) | |
tree | fda1de0843be17266388d0d137908f392a7f694e /src/gpu/cl | |
parent | 1af5416917268692fcd4b34b1d7ffebd3a2aea8a (diff) | |
download | ComputeLibrary-eb8bd81a625f0f87080dbde55b434362ad57324a.tar.gz |
Fix dst "widening" validation
* Auto-initialize the dst tensor before checking for PostOp shape
compliance so that we catch the invalid case of "widening" dst tensor
shape
* Rework post op validate test cases to be more readable
Partially resolves: COMPMID-4435
Change-Id: I79943994182942f962e4d59a7fa0d6f017ae9ac7
Signed-off-by: SiCongLi <sicong.li@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6548
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/gpu/cl')
-rw-r--r-- | src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.cpp | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.cpp b/src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.cpp index 4b28e2badc..8ee72d3f03 100644 --- a/src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.cpp +++ b/src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.cpp @@ -182,11 +182,11 @@ void ClGemmMatrixMultiplyReshapedKernel::configure(const CLCompileContext &compi { ARM_COMPUTE_ERROR_ON_NULLPTR(src0, src1, dst); - ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src0, src1, src2, dst, alpha, beta, lhs_info, rhs_info, gemm_info)); - // dst tensor auto initialization if not yet initialized auto_init_if_empty(*dst, src0->clone()->set_tensor_shape(misc::shape_calculator::compute_mm_shape(*src0, *src1, gemm_info))); + ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src0, src1, src2, dst, alpha, beta, lhs_info, rhs_info, gemm_info)); + auto padding_info = get_padding_info({ src0, src1, src2, dst }); _reinterpret_output_as_3d = gemm_info.depth_output_gemm3d != 0; _use_dummy_work_items = preferred_dummy_work_items_support(CLKernelLibrary::get().get_device()); |