[Mlir-commits] [mlir] [MLIR][SCF] Add support for loop pipeline peeling for dynamic loops. (PR #106436)

[llvmlistbot at llvm.org]

================
@@ -764,11 +764,44 @@ func.func @stage_0_value_escape(%A: memref<?xf32>, %result: memref<?xf32>, %ub:
 //      NOEPILOGUE:     memref.load %[[A]][%[[IV3]]] : memref<?xf32>
 //      NOEPILOGUE:   scf.yield %[[V2]], %[[L3]] : f32, f32
 
-// In case dynamic loop pipelining is off check that the transformation didn't
----------------
pawelszczerbuk wrote:

Hmm, not sure if this is worth the effort to validate that the transformation is *disabled*. I think I'm OK if you'd want to skip it.

https://github.com/llvm/llvm-project/pull/106436