[Mlir-commits] [mlir] [mlir][Vector] Refactor VectorEmulateNarrowType.cpp (PR #123529)

[Andrzej Warzyński]

================
@@ -1091,6 +1091,38 @@ struct ConvertVectorMaskedLoad final
   }
 };
 
+/// Check whether `subByteVecTy` fits wthin a vector of `multiByteScalarTy`
+///
+/// "Fitting" means that `subByteVecTy` (a vector of sub-byte elements, e.g.
+/// vector<4xi4>), can fit within N scalar elements of type `multiByteScalarTy`
+/// (a multi-byte scalar, e.g. i16), where N is some integer.
+///
+/// Put differently, this method checks whether this would be valid:
+///
+///   vector.bitcast subByteVecTy into vector<N x multiByteScalarTy>
+///
+/// EXAMPLES:
+///   * vector<4xi4> -> i16 - yes (N = 1)
+///   * vector<4xi4> -> i8 - yes (N = 2)
+///   * vector<3xi4> -> i8 - no (N would have to be 1.5)
+///   * vector<3xi2> -> i16 - no (N would have to be 0.5)
+static bool isSubByteVecFittable(VectorType subByteVecTy,
----------------
banach-space wrote:

Naming is hard! Thanks for the suggestion. How about: `fitsInMultiByteContainerTy`:
* dropped `size` as we are not passing `size`, and 
* replaced `Scalar` with `Container` to solidify the intent (while it is a "scalar", it's a special scalar that is meant to be a container 😅 )

WDYT? Please give me your honest answer, it's important to make this clear for everyone.

https://github.com/llvm/llvm-project/pull/123529