[llvm] [LV][EVL] Introduce MergeUntilPivot VPInstruction to enable out-loop reduction in EVL vectorization. (PR #101641)

[Mel Chen via llvm-commits]

================
@@ -314,67 +348,117 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; INLOOP-NEXT:    [[SMIN_LCSSA:%.*]] = phi i32 [ [[SMIN]], [[FOR_BODY]] ], [ [[RDX_MINMAX]], [[MIDDLE_BLOCK]] ]
 ; INLOOP-NEXT:    ret i32 [[SMIN_LCSSA]]
 ;
-; IF-EVL-LABEL: @smin(
----------------
Mel-Chen wrote:

Pre-commit NFC #102907 to updated the test case.

https://github.com/llvm/llvm-project/pull/101641