[mlir][vector] Propagate scalability in TransferWriteNonPermutationLowering

[cfRod]

Updates extendVectorRank so that scalability in patterns
that use it (in particular, TransferWriteNonPermutationLowering),
is correctly propagated.

Closed related previous PR #85270

[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.

[cfRod]

@banach-space @c-rhodes @MacDue

[llvmbot]

@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-vector

Author: Crefeda Rodrigues (cfRod)

Changes

…ectors

---

Full diff: https://github.com/llvm/llvm-project/pull/85632.diff

2 Files Affected:

• (modified) mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp (+10-2)
• (modified) mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir (+19)

diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
index 4a5e8fcfb6edaf..6c63928a009377 100644
--- a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
@@ -41,8 +41,16 @@ static Value extendVectorRank(OpBuilder &builder, Location loc, Value vec,
   SmallVector<int64_t> newShape(addedRank, 1);
   newShape.append(originalVecType.getShape().begin(),
                   originalVecType.getShape().end());
-  VectorType newVecType =
-      VectorType::get(newShape, originalVecType.getElementType());
+
+  ArrayRef<bool> originalScalableDims = originalVecType.getScalableDims();
+  SmallVector<bool> tempScalableDims(originalVecType.getShape().size());
+  for (const auto &pos : llvm::enumerate(originalScalableDims)) {
+    tempScalableDims[pos.index()] = originalScalableDims[pos.index()];
+  }
+  SmallVector<bool> newScalableDims(addedRank, 0);
+  newScalableDims.append(tempScalableDims);
+  VectorType newVecType = VectorType::get(
+      newShape, originalVecType.getElementType(), newScalableDims);
   return builder.create<vector::BroadcastOp>(loc, newVecType, vec);
 }
 
diff --git a/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir b/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
index 13e07f59a72a77..20afa574327d05 100644
--- a/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
+++ b/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
@@ -41,6 +41,25 @@ func.func @permutation_with_mask_scalable(%2: memref<?x?xf32>, %dim_1: index, %d
   return %1 : vector<8x[4]x2xf32>
 }
 
+// CHECK:           func.func @permutation_with_mask_transfer_write_scalable(
+// CHECK-SAME:        %[[VAL_0:.*]]: vector<4x[8]xi16>,
+// CHECK-SAME:        %[[VAL_1:.*]]: memref<1x4x?x1x1x1x1xi16>,
+// CHECK-SAME:        %[[VAL_2:.*]]: vector<4x[8]xi1>) {
+// CHECK:             %[[VAL_3:.*]] = arith.constant 0 : index
+// CHECK:             %[[VAL_4:.*]] = vector.broadcast %[[VAL_0]] : vector<4x[8]xi16> to vector<1x1x1x1x4x[8]xi16>
+// CHECK:             %[[VAL_5:.*]] = vector.broadcast %[[VAL_2]] : vector<4x[8]xi1> to vector<1x1x1x1x4x[8]xi1>
+// CHECK:             %[[VAL_6:.*]] = vector.transpose %[[VAL_5]], [4, 5, 0, 1, 2, 3] : vector<1x1x1x1x4x[8]xi1> to vector<4x[8]x1x1x1x1xi1>
+// CHECK:             %[[VAL_7:.*]] = vector.transpose %[[VAL_4]], [4, 5, 0, 1, 2, 3] : vector<1x1x1x1x4x[8]xi16> to vector<4x[8]x1x1x1x1xi16>
+// CHECK:             vector.transfer_write %[[VAL_7]], %[[VAL_1]]{{\[}}%[[VAL_3]], %[[VAL_3]], %[[VAL_3]], %[[VAL_3]], %[[VAL_3]], %[[VAL_3]], %[[VAL_3]]], %[[VAL_6]] {in_bounds = [true, true, true, true, true, true]} : vector<4x[8]x1x1x1x1xi16>, memref<1x4x?x1x1x1x1xi16>
+// CHECK:             return
+// CHECK:           }
+func.func @permutation_with_mask_transfer_write_scalable(%arg0: vector<4x[8]xi16>, %arg1: memref<1x4x?x1x1x1x1xi16>, %mask:  vector<4x[8]xi1>){
+     %c0 = arith.constant 0 : index
+      vector.transfer_write %arg0, %arg1[%c0, %c0, %c0, %c0, %c0, %c0, %c0], %mask {in_bounds = [true, true], permutation_map = affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d1, d2)>
+} : vector<4x[8]xi16>, memref<1x4x?x1x1x1x1xi16>
+
+    return
+}
 module attributes {transform.with_named_sequence} {
   transform.named_sequence @__transform_main(%module_op: !transform.any_op {transform.readonly}) {
     %f = transform.structured.match ops{["func.func"]} in %module_op

[MacDue]

Does this lower all the way to llvmir? The transposes here are to illegal types (vector<4x[8]x1x1x1x1xi16>), so they must be somehow be eliminated/cancel out.

[cfRod]

Will raise a separate PR to address this.

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[MacDue]

LGTM (just a final little nit)

P.s. I renamed the PR to make the intent a little clearer :)

[banach-space]

Nice! +1 to comments from Ben, otherwise LGTM 🙏🏻

[banach-space]

@cfRod Shall I land this for you?

[github-actions]

@cfRod 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 recieve 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!