[llvm] [VPlan] Allow wide-stores in narrow to single scalar (PR #157869)

[Ramkumar Ramachandra via llvm-commits]

================
@@ -26,12 +26,13 @@ define void @m(ptr nocapture %p, ptr nocapture %p2, i32 %q) {
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP3]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT2]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = sub nsw <4 x i32> [[TMP2]], [[BROADCAST_SPLAT3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[ARRAYIDX9_2]], align 4, !alias.scope [[META0]]
+; CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[ARRAYIDX9_2]], align 4, !alias.scope [[META0]]
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <4 x i32> [[TMP4]], i32 0
+; CHECK-NEXT:    [[TMP5:%.*]] = sub nsw i32 [[TMP9]], [[TMP6]]
----------------
artagnon wrote:

Damn, sorry I missed this. Investigating.

https://github.com/llvm/llvm-project/pull/157869