[mlir][vector] Restrict vector.shape_cast (scalable vectors) #100331
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Updates the verifier for `vector.shape_cast` so that the following incorrect cases are immediately rejected: ```mlir vector.shape_cast %vec : vector<1x1x[4]xindex> to vector<4xindex> ``` Seperately, here's a fix for the Linalg vectorizer to prevent the vectorizer from generating such shape casts (*): * llvm#100325 (*) Note, that's just one specific case that I've identified so far.
|
@llvm/pr-subscribers-mlir-ods @llvm/pr-subscribers-mlir-vector Author: Andrzej Warzyński (banach-space) ChangesUpdates the verifier for vector.shape_cast %vec : vector<1x1x[4]xindex> to vector<4xindex>Seperately, here's a fix for the Linalg vectorizer to prevent the (*) Note, that's just one specific case that I've identified so far. Full diff: https://github.com/llvm/llvm-project/pull/100331.diff 2 Files Affected:
diff --git a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
index df3a59ed80ad4..f145dc9f8817d 100644
--- a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
+++ b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
@@ -5238,6 +5238,11 @@ static LogicalResult verifyVectorShapeCast(Operation *op,
if (!isValidShapeCast(resultShape, sourceShape))
return op->emitOpError("invalid shape cast");
}
+
+ // Check that (non-)scalability is preserved
+ if (sourceVectorType.isScalable() != resultVectorType.isScalable())
+ return op->emitOpError("non-matching scalability flags");
+
return success();
}
diff --git a/mlir/test/Dialect/Vector/invalid.mlir b/mlir/test/Dialect/Vector/invalid.mlir
index 208982a3e0e7b..3fad61198b474 100644
--- a/mlir/test/Dialect/Vector/invalid.mlir
+++ b/mlir/test/Dialect/Vector/invalid.mlir
@@ -1182,6 +1182,13 @@ func.func @shape_cast_invalid_rank_expansion(%arg0 : vector<15x2xf32>) {
// -----
+func.func @shape_cast_scalability_flag_is_dropped(%arg0 : vector<15x[2]xf32>) {
+ // expected-error@+1 {{non-matching scalability flags}}
+ %0 = vector.shape_cast %arg0 : vector<15x[2]xf32> to vector<30xf32>
+}
+
+// -----
+
func.func @bitcast_not_vector(%arg0 : vector<5x1x3x2xf32>) {
// expected-error@+1 {{'vector.bitcast' invalid kind of type specified}}
%0 = vector.bitcast %arg0 : vector<5x1x3x2xf32> to f32
|
MacDue
reviewed
Jul 24, 2024
Make the verifier even stricter
MacDue
reviewed
Jul 24, 2024
kuhar
reviewed
Jul 24, 2024
…ble vectors) Improve diagnostic
MacDue
approved these changes
Jul 24, 2024
MacDue
reviewed
Jul 24, 2024
… (scalable vectors) Spell out auto
…pe_cast (scalable vectors) Remove blank line
yuxuanchen1997
pushed a commit
that referenced
this pull request
Jul 25, 2024
Summary: Updates the verifier for `vector.shape_cast` so that incorrect cases where "scalability" is dropped are immediately rejected. For example: ```mlir vector.shape_cast %vec : vector<1x1x[4]xindex> to vector<4xindex> ``` Also, as a separate PR, I've prepared a fix for the Linalg vectorizer to avoid generating such shape casts (*): * #100325 (*) Note, that's just one specific case that I've identified so far. Test Plan: Reviewers: Subscribers: Tasks: Tags: Differential Revision: https://phabricator.intern.facebook.com/D60250589
banach-space
added a commit
to banach-space/llvm-project
that referenced
this pull request
Jul 26, 2024
Updates the codebase to use the recently introduced VectorType helper: * `getNumScalableDims` (introduced in llvm#100331).
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Updates the verifier for
vector.shape_castso that the followingincorrect cases are immediately rejected:
Seperately, here's a fix for the Linalg vectorizer to prevent the
vectorizer from generating such shape casts (*):
(*) Note, that's just one specific case that I've identified so far.