Skip to content

[OptRed] Do not assert on reductions after 3D dot ops #2590

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 1 commit into from
Oct 30, 2024
Merged

[OptRed] Do not assert on reductions after 3D dot ops #2590

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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 third_party/intel/lib/TritonIntelGPUTransforms/OptimizeReductionLocality.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,13 +153,15 @@ struct DpasOperandPattern final : OpRewritePattern<ReduceOp> {
// Check this is has `triton_intel_gpu.dpas` encoding.
Value operand = operands.front();
auto type = cast<RankedTensorType>(operand.getType());
// Only support reduction after 2D-dot for now.
if (type.getRank() != 2)
return failure();
auto encoding =
llvm::dyn_cast_or_null<DpasEncodingAttr>(type.getEncoding());
if (!encoding)
return failure();

// Axis 1 will lead to within-warp reduction.
assert(type.getRank() == 2 && "Expecting 2D tensor");
if (op.getAxis() != preferredReductionAxis)
return failure();

Expand Down