[Mlir-commits] [mlir] [mlir] Replace dynamic sizes in insert_slice of tensor.cast canonicalization (PR #91352)

[Han-Chung Wang]

================
@@ -1890,21 +1918,6 @@ func.func @splat_dynamic_no_fold(%m: index) -> tensor<4x?xf32> {
 
 // -----
 
-// There was an issue in cast + insert_slice folding generating invalid ir.
-// https://github.com/llvm/llvm-project/issues/53099
----------------
hanhanW wrote:

It is moved to l.758. I think the revision adds the support for the case. It is generating valid IR now.

https://github.com/llvm/llvm-project/pull/91352