Skip to content

Fix LDS vectorization logic to allow for bank conflict resolution when kPack > kPerThread #2196

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
umangyadav wants to merge 4 commits into
base: develop
Choose a base branch
from
+3 −1
Open

Fix LDS vectorization logic to allow for bank conflict resolution when kPack > kPerThread #2196

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
6b7896c
if copyKPerThread<kPack then D will not be contiguous in LDS and ther…
umangyadav Jan 5, 2026
2259b4c
Merge branch 'develop' into ldsVectorization
umangyadav Jan 6, 2026
e3306f0
formatting
umangyadav Jan 6, 2026
e82c331
Merge branch 'develop' into ldsVectorization
umangyadav Jan 6, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion mlir/lib/Dialect/Rock/Transforms/GridwiseGemmToBlockwise.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -287,12 +287,14 @@ static LDSLayoutConfigDim getLDSLayoutConfigDim(Type elementType, int64_t kpack,
LDSLayoutConfigDim cfg;
int64_t maxVlen = 128 / elementType.getIntOrFloatBitWidth();
int64_t copyDPerThread = vecDimInfo.inDPerThread;
int64_t copyKPerThread = vecDimInfo.inKPerThread;
bool isKContiguousDim = vecDimInfo.vectorDim == GemmDimension::K;
// If kpack is less than the hardware max vector length, and we are
// writing more contiguous kpack elements, there is a possibility to
// vectorize that we want to preserve (i.e., we favour vectorization over
// bank conflicts resolution)
bool isPossibleToVectorizeD = (kpack < maxVlen && copyDPerThread > 1);
bool isPossibleToVectorizeD =
(kpack < maxVlen && copyDPerThread > 1) && (copyKPerThread >= kpack);
Copy link
Contributor

@dhernandez0 dhernandez0 Jan 7, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can you add a comment to explain this?

Copy link
Contributor

@dhernandez0 dhernandez0 Jan 7, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: no need to have two parentheses.

cfg.doRotateWithK = isKContiguousDim && !isPossibleToVectorizeD;
cfg.doSwapThreadIterSubDims = !isKContiguousDim && !isPossibleToVectorizeD;
cfg.ldsLayoutDxK = false;
Expand Down