[llvm] [VPlan] Strip first-lane handling in legalizeOptzInd (PR #194795)

[Florian Hahn via llvm-commits]

================
@@ -120,9 +122,9 @@ define void @sdiv_feeding_gep_predicated(ptr %dst, i32 %x, i64 %M, i64 %conv6, i
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[TMP15]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP22:%.*]] = icmp ule <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP22]], <vscale x 2 x i1> zeroinitializer
-; CHECK-NEXT:    [[TMP24:%.*]] = extractelement <vscale x 2 x i1> [[TMP23]], i64 0
-; CHECK-NEXT:    [[TMP25:%.*]] = select i1 [[TMP24]], i64 [[CONV6]], i64 1
-; CHECK-NEXT:    [[TMP26:%.*]] = sdiv i64 [[M]], [[TMP25]]
+; CHECK-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[TMP23]], <vscale x 2 x i64> [[BROADCAST_SPLAT2]], <vscale x 2 x i64> splat (i64 1)
+; CHECK-NEXT:    [[TMP13:%.*]] = sdiv <vscale x 2 x i64> [[BROADCAST_SPLAT]], [[TMP12]]
+; CHECK-NEXT:    [[TMP26:%.*]] = extractelement <vscale x 2 x i64> [[TMP13]], i64 0
----------------
fhahn wrote:

Hmm, is this actually desirable? Looks like a missed scalarization opportunity?

https://github.com/llvm/llvm-project/pull/194795