gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[Triton-MLIR] Minor fixes related with scf/swizzling support (#791)

Browse Source 
1, Disable static loop unrolling in the frontend by default;
2, A minor fix in axisAnalysis in order to support scf;
3, A minor fix in TritonGPUToLLVM to support swizzling.
This commit is contained in:
goostavz
2022-10-21 11:46:28 +08:00
committed by GitHub
parent dc0588a898
commit c4726333bf
4 changed files with 79 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
lib/Conversion/TritonGPUToLLVM/TritonGPUToLLVM.cpp
View File

@@ -1867,8 +1867,8 @@ LogicalResult ConvertLayoutOpConversion::lowerBlockedToShared(
unsigned linearIdxInNanoTile = i % srcAccumSizeInThreads;
auto multiDimIdxInNanoTile = getMultiDimIndex<unsigned>(
linearIdxInNanoTile, srcBlockedLayout.getSizePerThread());
multiDimIdxInNanoTile[inOrd[0]] /= minVec;
unsigned pos = multiDimIdxInNanoTile[inOrd[0]] % minVec;
multiDimIdxInNanoTile[inOrd[0]] /= minVec;
unsigned wordVecIdx =
getLinearIndex<unsigned>(multiDimIdxInNanoTile, wordsInEachRep);
wordVecs[wordVecIdx] =