Skip to content

[MLIR] Fix mlir-opt crash in ReshapeOpsUtils.cpp when collapse_shape index is invalid #173791

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
akparmar004 wants to merge 1 commit into
base: main
Choose a base branch
from
+21 −1
Open

[MLIR] Fix mlir-opt crash in ReshapeOpsUtils.cpp when collapse_shape index is invalid #173791

Show file tree
Hide file tree
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
5 changes: 5 additions & 0 deletions mlir/lib/Dialect/Tensor/IR/TensorOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2072,6 +2072,11 @@ LogicalResult ExpandShapeOp::verify() {
}

LogicalResult CollapseShapeOp::verify() {
CollapseShapeOp op = *this;
if (llvm::any_of(op.getReassociationIndices(),
[](auto &group) { return group.empty(); })) {
Copy link
Member

@ftynse ftynse Dec 29, 2025

Choose a reason for hiding this comment

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

Nit: please expand auto unless the type is obvious from the statement-level context or unavoidable. Here I suspect it should be something like ValueRange, at which point you don't need a reference.

return op.emitOpError("reassociation indices must not be empty");
}
return verifyTensorReshapeOp(*this, getSrcType(), getResultType());
}

Expand Down
4 changes: 3 additions & 1 deletion mlir/lib/Dialect/Utils/ReshapeOpsUtils.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -451,7 +451,9 @@ mlir::getSymbolLessAffineMaps(ArrayRef<ReassociationExprs> reassociation) {
SmallVector<AffineMap, 4> maps;
maps.reserve(reassociation.size());
for (const auto &exprs : reassociation) {
assert(!exprs.empty());
if (exprs.empty()) {
return {};
}
Comment on lines +454 to +456
Copy link
Member

@ftynse ftynse Dec 29, 2025

Choose a reason for hiding this comment

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

Generally, compiler code should assume/assert that the IR has passed the verifier, so this should stay as an assertion unless it is somehow called on unverified code.

maps.push_back(AffineMap::get(maxDim + 1, 0, exprs, exprs[0].getContext()));
}
return maps;
Expand Down
13 changes: 13 additions & 0 deletions mlir/test/Dialect/Tensor/invalid-collapse-shape.mlir
View file
Open in desktop
Copy link
Member

@ftynse ftynse Dec 29, 2025

Choose a reason for hiding this comment

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

This should go to

llvm-project/mlir/test/Dialect/Tensor/invalid.mlir

Line 1 in 367e388

// RUN: mlir-opt %s -split-input-file -verify-diagnostics
, we don't want one file per a tiny test.

Original file line number Diff line number Diff line change
@@ -0,0 +1,13 @@
// RUN: mlir-opt %s -split-input-file -verify-diagnostics

// This test checks that an empty reassociation group in `tensor.collapse_shape`
// results in a proper error instead of an assert/crash.

// -----

func.func @test_empty_reassociation(%arg0: tensor<1x?xf32>) -> tensor<?x10xf32> {
// expected-error@+1 {{'tensor.collapse_shape' op reassociation indices must not be empty}}
Copy link
Member

@ftynse ftynse Dec 29, 2025

Choose a reason for hiding this comment

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

Suggested change
// expected-error@+1 {{'tensor.collapse_shape' op reassociation indices must not be empty}}
// expected-error@below {{'tensor.collapse_shape' op reassociation indices must not be empty}}

%0 = tensor.collapse_shape %arg0 [[0, 1], []] : tensor<1x?xf32> into tensor<?x10xf32>
return %0 : tensor<?x10xf32>
}