[mlir][linalg] Fix empty outer dim case for packing reshape op

[yifeizh2]

This PR fixes the issue reported in comment.

[github-actions]

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be
notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write
permissions for the repository. In which case you can instead tag reviewers by
name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review
by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate
is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

[llvmbot]

@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-linalg

Author: None (yifeizh2)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/96732.diff

1 Files Affected:

• (modified) mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp (+14-3)

diff --git a/mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp b/mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp
index e51ae2264a36a..699bf56f96581 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp
@@ -641,7 +641,14 @@ bubbleUpPackOpThroughCollapseShape(tensor::CollapseShapeOp collapseOp,
                                    PatternRewriter &rewriter) {
   SmallVector<int64_t> innerTileSizes = packOp.getStaticTiles();
   ArrayRef<int64_t> innerDimsPos = packOp.getInnerDimsPos();
-  ArrayRef<int64_t> outerDimsPerm = packOp.getOuterDimsPerm();
+  auto numOuterDims =
+      dyn_cast<RankedTensorType>(packOp.getDpsInputs()[0].getType())
+          .getShape()
+          .size();
+  SmallVector<int64_t> outerDimsPerm =
+      packOp.getOuterDimsPerm().empty()
+          ? llvm::to_vector(llvm::seq<int64_t>(0, numOuterDims))
+          : SmallVector<int64_t>(packOp.getOuterDimsPerm());
 
   ArrayRef<int64_t> srcShape = collapseOp.getSrcType().getShape();
   SmallVector<ReassociationIndices> reassocIndices =
@@ -885,8 +892,12 @@ pushDownUnPackOpThroughExpandShape(tensor::UnPackOp unPackOp,
                                    PatternRewriter &rewriter) {
   SmallVector<int64_t> innerTileSizes = unPackOp.getStaticTiles();
   ArrayRef<int64_t> innerDimsPos = unPackOp.getInnerDimsPos();
-  ArrayRef<int64_t> outerDimsPerm = unPackOp.getOuterDimsPerm();
-
+  auto numOuterDims =
+      dyn_cast<RankedTensorType>(unPackOp.getType()).getShape().size();
+  SmallVector<int64_t> outerDimsPerm =
+      unPackOp.getOuterDimsPerm().empty()
+          ? llvm::to_vector(llvm::seq<int64_t>(0, numOuterDims))
+          : SmallVector<int64_t>(unPackOp.getOuterDimsPerm());
   auto expandTy = dyn_cast<RankedTensorType>(expandOp.getType());
   if (!expandTy)
     return failure();

[yifeizh2]

CC @adam-smnk . Not sure whether the fix here is acceptable.

[adam-smnk]

Thanks a lot for the fix.
Current approach works but I wonder if it could be a bit simpler to address it at the helper applyPermutationAndReindexReassoc.
It might enough to skip applyPermutationToVector when permutation is empty.

[yifeizh2]

Thanks a lot for the fix. Current approach works but I wonder if it could be a bit simpler to address it at the helper applyPermutationAndReindexReassoc. It might enough to skip applyPermutationToVector when permutation is empty.

Thanks for the suggestion! I changed the logic to be skipping applyPermutationToVector when permutation is empty.

[adam-smnk]

Looks good, thanks

[ScottTodd]

Thanks! Tested downstream and the issue I saw goes away with this fix.

[MaheshRavishankar]

Thanks!

[AmosLewis]

Thanks! the mit-b0 model issue go away with this fix.

[github-actions]

@yifeizh2 Congratulations on having your first Pull Request (PR) merged into the LLVM Project!

Your changes will be combined with recent changes from other authors, then tested
by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR.

Please check whether problems have been caused by your change specifically, as
the builds can include changes from many authors. It is not uncommon for your
change to be included in a build that fails due to someone else's changes, or
infrastructure issues.

How to do this, and the rest of the post-merge process, is covered in detail here.

If your change does cause a problem, it may be reverted, or you can revert it yourself.
This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again.

If you don't get any reports, no action is required from you. Your changes are working as expected, well done!