[Mlir-commits] [mlir] [MLIR][Vector] Step op warp distribution (PR #155425)

[Adam Siemieniuk]

================
@@ -705,6 +705,45 @@ struct WarpOpConstant : public WarpDistributionPattern {
   }
 };
 
+/// Sink out step op feeding into a warp op yield.
+/// Vector step op is treated similar to arith.constant, apart from
+/// the result that represents a sequence [0, vec_size).
+/// The sequence is semantically equivalent to warp's threads/lanes indices.
----------------
adam-smnk wrote:

It looks to me that this whole rewrite assumes that `vec_size == warp_size` while `vector.step` can have an arbitrary size.

https://github.com/llvm/llvm-project/pull/155425