[Mlir-commits] [mlir] Fixing peelFirstIteration. Was incorrectly replacing all ub inside peeled iteration region. (PR #163720)

[llvmlistbot at llvm.org]

================
@@ -119,7 +161,7 @@ func.func @fully_dynamic_bounds(%lb : index, %ub: index, %step: index) -> i32 {
 //  CHECK-DAG:   %[[C6:.*]] = arith.constant 6 : index
 //      CHECK:   %[[FIRST:.*]] = scf.for %[[IV:.*]] = %[[C2]] to %[[C6]]
 // CHECK-SAME:       step %[[C4]] iter_args(%[[ACC:.*]] = %[[C0_I32]]) -> (i32) {
-//      CHECK:     %[[MIN:.*]] = affine.min #[[MAP]](%[[C6]], %[[IV]])[%[[C4]]]
+//      CHECK:     %[[MIN:.*]] = affine.min #[[MAP]](%[[C8]], %[[IV]])[%[[C4]]]
----------------
ddubov100 wrote:

The bottom line is the same in both: MIN == 4

https://github.com/llvm/llvm-project/pull/163720