[Mlir-commits] [mlir] [mlir][scf]: Avoid inserting affine.min when tiling dynamic operation sizes if possible (PR #113819)

[Matthias Springer]

================
@@ -186,11 +186,35 @@ static void checkSafeToTileToForall(TilingInterface op,
   }
 }
 
+/// Returns true if `size` is dynamic multiplication of `stride`.
----------------
matthias-springer wrote:

I haven't looked into this pass for a long time. What I vaguely remember is that `-scf-for-loop-canonicalization` has always been a bit fragile due to limitations around semi-affine expressions in `FlatAffineValueConstraints`.

https://github.com/llvm/llvm-project/pull/113819