NVIDIA/cutlass
1
0
Fork 0
mirror of https://github.com/NVIDIA/cutlass.git synced 2026-05-11 17:00:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix alignmentC=8 for imma N=128 (#822)

Browse Source 
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
This commit is contained in:
Haicheng Wu
2023-02-15 12:06:00 -05:00
committed by GitHub
parent 8f5c242426
commit 9fb38ac048
1 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
include/cutlass/epilogue/threadblock/default_epilogue_tensor_op.h
View File

@@ -270,7 +270,7 @@ struct DefaultIteratorsTensorOp<
>;
using WarpTileIterator = typename platform::conditional<
(ThreadblockShape::kN == 256),
(ThreadblockShape::kN == 256) || (ThreadblockShape::kN == 128 && ElementsPerAccess == 8),
WarpTileIteratorNotMixed,
WarpTileIteratorMixed>::type;
@@ -289,7 +289,7 @@ struct DefaultIteratorsTensorOp<
>;
using SharedLoadIterator = typename platform::conditional<
(ThreadblockShape::kN == 256),
(ThreadblockShape::kN == 256) || (ThreadblockShape::kN == 128 && ElementsPerAccess == 8),
SharedLoadIteratorNotMixed,
SharedLoadIteratorMixed>::type;
@@ -337,7 +337,7 @@ struct DefaultIteratorsTensorOp<
>;
using WarpTileIterator = typename platform::conditional<
(ThreadblockShape::kN == 256),
(ThreadblockShape::kN == 256) || (ThreadblockShape::kN == 128 && ElementsPerAccess == 8),
WarpTileIteratorNotMixed,
WarpTileIteratorMixed>::type;
@@ -356,7 +356,7 @@ struct DefaultIteratorsTensorOp<
>;
using SharedLoadIterator = typename platform::conditional<
(ThreadblockShape::kN == 256),
(ThreadblockShape::kN == 256) || (ThreadblockShape::kN == 128 && ElementsPerAccess == 8),
SharedLoadIteratorNotMixed,
SharedLoadIteratorMixed>::type;
@@ -404,7 +404,7 @@ struct DefaultIteratorsTensorOp<
>;
using WarpTileIterator = typename platform::conditional<
(ThreadblockShape::kN == 256),
(ThreadblockShape::kN == 256) || (ThreadblockShape::kN == 128 && ElementsPerAccess == 8),
WarpTileIteratorNotMixed,
WarpTileIteratorMixed>::type;
@@ -423,7 +423,7 @@ struct DefaultIteratorsTensorOp<
>;
using SharedLoadIterator = typename platform::conditional<
(ThreadblockShape::kN == 256),
(ThreadblockShape::kN == 256) || (ThreadblockShape::kN == 128 && ElementsPerAccess == 8),
SharedLoadIteratorNotMixed,
SharedLoadIteratorMixed>::type;