ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-14 11:07:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

refine

Browse Source
This commit is contained in:
Jin Zhou
2025-09-08 02:46:41 +00:00
parent 66d6f152a7
commit 98c58fd94b
1 changed files with 6 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_b_preshuffle.hpp
View File

@@ -324,6 +324,7 @@ struct DeviceGemmMultiD_Xdl_CShuffle_V3_BPreshuffle
Run(kernel);
}
}
else
{
if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
{
@@ -336,6 +337,7 @@ struct DeviceGemmMultiD_Xdl_CShuffle_V3_BPreshuffle
TailNumber::Odd>;
Run(kernel);
}
else
{
const auto kernel =
kernel_gemm_xdl_cshuffle_v3_multi_d_b_preshuffle_2lds<
@@ -353,10 +355,12 @@ struct DeviceGemmMultiD_Xdl_CShuffle_V3_BPreshuffle
throw std::runtime_error("todo: only v1 v2 and v3 support now");
}
}
#if 0
else
{
if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v1)
{
#if 0
if(arg.KBatch > 1)
{
if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
@@ -403,6 +407,7 @@ struct DeviceGemmMultiD_Xdl_CShuffle_V3_BPreshuffle
Run(kernel);
}
}
#endif
if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
{
const auto kernel = kernel_gemm_xdl_cshuffle_v3_multi_d_b_preshuffle<
@@ -482,7 +487,7 @@ struct DeviceGemmMultiD_Xdl_CShuffle_V3_BPreshuffle
throw std::runtime_error("todo: only v3 support now");
}
}
#endif
return ave_time;
}