ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 05:01:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

Implement GetWorkSpaceSize from BaseOperator. (#1564)

Browse Source
This commit is contained in:
Adam Osewski
2024-10-12 08:05:11 +02:00
committed by GitHub
parent 11444e4cf2
commit 29d384d0b2
2 changed files with 20 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
include/ck/tensor_operation/gpu/device/impl/device_cgemm_4gemm_xdl_cshuffle.hpp
View File

@@ -598,10 +598,26 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
[[maybe_unused]] index_t K,
[[maybe_unused]] index_t StrideA,
[[maybe_unused]] index_t StrideB,
index_t StrideC) override
index_t StrideC) const override
{
return 2 * sizeof(CDataType) * GetCElementSpaceSize(M, N, StrideC);
}
std::size_t GetWorkSpaceSize(const BaseArgument* base_arg) const override
{
const auto* parg = dynamic_cast<const Argument*>(base_arg);
if(!parg)
{
std::ostringstream err;
err << "Provided argument pointer is not of an Argument class!"
<< " In " << __FILE__ << ":" << __LINE__ << ", in function: " << __func__;
throw std::runtime_error(err.str());
}
return GetWorkspaceSize(
parg->M, parg->N, parg->K, parg->StrideA, parg->StrideB, parg->StrideC);
}
};
} // namespace device