ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 13:11:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

added padding of K into gemm_v2r3 (#887)

Browse Source 
* added kpad support into v2r3

* add generic instances

* fixed comments

* fixed mnk padding

* Update device_batched_gemm_xdl.hpp

---------

Co-authored-by: Jing Zhang <jizha@amd.com>
This commit is contained in:
zjing14
2023-09-06 10:15:52 -05:00
committed by GitHub
parent a61b8b785e
commit 3786bfe1cc
4 changed files with 72 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_xdl.hpp
View File

@@ -185,7 +185,7 @@ struct DeviceBatchedGemmXdl : public DeviceBatchedGemm<ALayout,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
GemmSpecialization::MNPadding,
GemmSpecialization::MNKPadding,
MPerBlock,
NPerBlock,
K0PerBlock,
@@ -315,11 +315,6 @@ struct DeviceBatchedGemmXdl : public DeviceBatchedGemm<ALayout,
return false;
}
if(problem.K % K1 != 0)
{
return false;
}
return GridwiseGemm::CheckValidity(problem);
}
@@ -416,7 +411,12 @@ struct DeviceBatchedGemmXdl : public DeviceBatchedGemm<ALayout,
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock
<< K0PerBlock << ", "
<< K1 << ", "
<< MPerXDL << ", "
<< NPerXDL << ", "
<< MXdlPerWave << ", "
<< NXdlPerWave << ", "
<< ">"
<< " NumGemmKPrefetchStage: "
<< NumGemmKPrefetchStage << ", "