Maintain previous behaviour by always generating vector for WidePtrAdd

Author: lukel97

llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp

@@ -514,7 +514,6 @@ bool VPInstruction::canGenerateScalarForFirstLane() const {
   case VPInstruction::CalculateTripCountMinusVF:
   case VPInstruction::CanonicalIVIncrementForPart:
   case VPInstruction::PtrAdd:
-  case VPInstruction::WidePtrAdd:
   case VPInstruction::ExplicitVectorLength:
   case VPInstruction::AnyOf:
     return true;
@@ -858,7 +857,7 @@ Value *VPInstruction::generate(VPTransformState &State) {
   }
   case VPInstruction::WidePtrAdd: {
     Value *Ptr = State.get(getOperand(0), true);
-    Value *Addend = State.get(getOperand(1), vputils::onlyFirstLaneUsed(this));
+    Value *Addend = State.get(getOperand(1));
     return Builder.CreatePtrAdd(Ptr, Addend, Name, getGEPNoWrapFlags());
   }
   case VPInstruction::AnyOf: {
@@ -1090,8 +1089,9 @@ bool VPInstruction::onlyFirstLaneUsed(const VPValue *Op) const {
   case VPInstruction::ReductionStartVector:
     return true;
   case VPInstruction::PtrAdd:
-  case VPInstruction::WidePtrAdd:
     return Op == getOperand(0) || vputils::onlyFirstLaneUsed(this);
+  case VPInstruction::WidePtrAdd:
+    return Op == getOperand(0);
   case VPInstruction::ComputeAnyOfResult:
   case VPInstruction::ComputeFindIVResult:
     return Op == getOperand(1);

llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll

@@ -162,11 +162,10 @@ define void @pointer_induction(ptr noalias %start, i64 %N) {
 ; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP11]], i64 0
 ; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP12:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
-; CHECK-NEXT:    [[TMP20:%.*]] = extractelement <vscale x 2 x i64> [[DOTSPLAT]], i32 0
-; CHECK-NEXT:    [[TMP21:%.*]] = extractelement <vscale x 2 x i64> [[TMP12]], i32 0
-; CHECK-NEXT:    [[TMP13:%.*]] = add i64 [[TMP20]], [[TMP21]]
-; CHECK-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], 1
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], i64 [[TMP14]]
+; CHECK-NEXT:    [[TMP13:%.*]] = add <vscale x 2 x i64> [[DOTSPLAT]], [[TMP12]]
+; CHECK-NEXT:    [[TMP14:%.*]] = mul <vscale x 2 x i64> [[TMP13]], splat (i64 1)
+; CHECK-NEXT:    [[VECTOR_GEP:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <vscale x 2 x i64> [[TMP14]]
+; CHECK-NEXT:    [[TMP15:%.*]] = extractelement <vscale x 2 x ptr> [[VECTOR_GEP]], i32 0
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[TMP15]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP16]], align 1
 ; CHECK-NEXT:    [[TMP17:%.*]] = add <vscale x 2 x i8> [[WIDE_LOAD]], splat (i8 1)

llvm/test/Transforms/LoopVectorize/X86/pr48340.ll

@@ -28,7 +28,7 @@ define ptr @foo(ptr %p, ptr %p.last) unnamed_addr #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[P]], [[VECTOR_PH]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 12288, i64 13312, i64 14336, i64 15360>
-; CHECK-NEXT:    [[VECTOR_GEP4:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], i64 0
+; CHECK-NEXT:    [[VECTOR_GEP4:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 0, i64 1024, i64 2048, i64 3072>
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER6:%.*]] = call <4 x ptr> @llvm.masked.gather.v4p0.v4p0(<4 x ptr> [[TMP8]], i32 8, <4 x i1> splat (i1 true), <4 x ptr> poison)
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 16384
@@ -86,7 +86,7 @@ define ptr @bar(ptr %p, ptr %p.last) unnamed_addr #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[P]], [[VECTOR_PH]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 12288, i64 13312, i64 14336, i64 15360>
-; CHECK-NEXT:    [[VECTOR_GEP4:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], i64 0
+; CHECK-NEXT:    [[VECTOR_GEP4:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 0, i64 1024, i64 2048, i64 3072>
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER6:%.*]] = call <4 x ptr> @llvm.masked.gather.v4p0.v4p0(<4 x ptr> [[TMP8]], i32 8, <4 x i1> splat (i1 true), <4 x ptr> poison)
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 16384