Skip to content

[MLIR][AMDGPU] Add a wrapper for global LDS load intrinsics in AMDGPU #133498

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

Merged
merged 17 commits into from
Apr 8, 2025
Merged

[MLIR][AMDGPU] Add a wrapper for global LDS load intrinsics in AMDGPU #133498

Changes from 1 commit
Commits
Show all changes
17 commits
Select commit Hold shift + click to select a range
ba38466
make it compiable.
lialan Mar 25, 2025
8e46990
Adding a verifier.
lialan Mar 25, 2025
c072e66
checkpoint
lialan Mar 27, 2025
564ebc8
Make it work
lialan Mar 27, 2025
92a1ef9
update AMDGPU description.
lialan Mar 27, 2025
a17e854
Address comments.
lialan Apr 2, 2025
4ed2006
update lowering
lialan Apr 2, 2025
81bb8bc
update test files.
lialan Apr 2, 2025
73629f4
linting
lialan Apr 2, 2025
b483701
Update tests.
lialan Apr 3, 2025
1a40d6c
update again
lialan Apr 3, 2025
d68db39
Final touch.
lialan Apr 7, 2025
33fbbc3
Merge branch 'main' into users/lialan/global_load_lds
lialan Apr 7, 2025
7aad9fb
Small update
lialan Apr 7, 2025
85df43b
Rebase
lialan Apr 7, 2025
fd97908
Fix a small issue.
lialan Apr 8, 2025
cd22692
fix error
lialan Apr 8, 2025
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
13 changes: 8 additions & 5 deletions mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1031,11 +1031,14 @@ struct GatherToLDSOpLowering : public ConvertOpToLLVMPattern<GatherToLDSOp> {
// augment it to transfer multiple elements per thread by issuing multiple
// `global_load_lds` instructions.
Type transferType = op.getTransferType();
size_t loadWidth = transferType.getIntOrFloatBitWidth() / 8;
if (auto transferVectorType = dyn_cast<VectorType>(transferType)) {
loadWidth = transferVectorType.getNumElements() *
(transferVectorType.getElementTypeBitWidth() / 8);
}
size_t loadWidth = [&]() -> size_t {
if (auto transferVectorType = dyn_cast<VectorType>(transferType)) {
return transferVectorType.getNumElements() *
(transferVectorType.getElementTypeBitWidth() / 8);
} else {
return transferType.getIntOrFloatBitWidth() / 8;
}
Comment on lines +1038 to +1040
Copy link
Member

@kuhar kuhar Apr 8, 2025

Choose a reason for hiding this comment

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

no else after return: https://llvm.org/docs/CodingStandards.html#don-t-use-else-after-a-return

}();

// Currently only 1, 2, and 4 byte loads are supported.
if (loadWidth != 1 && loadWidth != 2 && loadWidth != 4)
Expand Down
Loading