[VPlan] Support single-scalar VPReplicateRecipes when narrowing IGs.

When narrowing interleave groups, we can treat single scalar
VPReplicateRecipes as already narrowed.

Author: fhahn

llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp

@@ -3140,7 +3140,12 @@ static bool isConsecutiveInterleaveGroup(VPInterleaveRecipe *InterleaveR,
 }
 
 /// Returns true if \p VPValue is a narrow VPValue.
-static bool isAlreadyNarrow(VPValue *VPV) { return VPV->isLiveIn(); }
+static bool isAlreadyNarrow(VPValue *VPV) {
+  if (VPV->isLiveIn())
+    return true;
+  auto *RepR = dyn_cast<VPReplicateRecipe>(VPV);
+  return RepR && RepR->isSingleScalar();
+}
 
 void VPlanTransforms::narrowInterleaveGroups(VPlan &Plan, ElementCount VF,
                                              unsigned VectorRegWidth) {

llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll

@@ -463,10 +463,8 @@ define void @single_uniform_load_store_interleave_group(ptr noalias %src, ptr no
 ; VF2-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[TMP1]], i64 0
 ; VF2-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLAT]], <2 x i64> [[BROADCAST_SPLAT]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; VF2-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <4 x i64> [[TMP3]], <4 x i64> poison, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
-; VF2-NEXT:    store <4 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8
-; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VF2-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP2]], align 8
+; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 1
 ; VF2-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]: