[Mlir-commits] [mlir] [mlir][linalg] Vectorize unpack op without masking (PR #89067)

[Andrzej Warzyński]

================
@@ -1597,6 +1597,16 @@ vectorizeAsTensorUnpackOp(RewriterBase &rewriter, tensor::UnPackOp unpackOp,
 
   RankedTensorType unpackTensorType = unpackOp.getSourceType();
 
+  // If the input vector sizes are not provided, then the vector sizes are
+  // determined by the result tensor shape. In case the vector sizes aren't
+  // provided, we update the inBounds attribute instead of masking.
+  bool doMasking = true;
+  if (inputVectorSizes.empty()) {
+    ArrayRef<int64_t> resultTensorShape = unpackOp.getDestType().getShape();
+    inputVectorSizes = resultTensorShape.take_front(unpackOp.getSourceRank());
+    doMasking = false;
+  }
----------------
banach-space wrote:

Similar comment as for https://github.com/llvm/llvm-project/pull/89119 ...

IMHO it would make more sense to introduce `useInBoundsInsteadOfMasking` in place of `doMasking`. 

The longer name better captures the underlying logic ("replace masks with in_bounds attr"). Indeed, the actual meaning of the bool is much narrower than `doMasking`.

Naming is hard and I am being a bit pedantic. And also a bit mindful of our future selves when reading this in 6 months :)

https://github.com/llvm/llvm-project/pull/89067