[llvm] [VPlan] Manage noalias/alias_scope metadata in VPlan. (NFC) (PR #136450)

Florian Hahn via llvm-commits llvm-commits at lists.llvm.org
Sat Apr 19 12:27:19 PDT 2025


https://github.com/fhahn created https://github.com/llvm/llvm-project/pull/136450

Use VPIRMetadata added in https://github.com/llvm/llvm-project/pull/135272
to also manage no-alias metadata added by versioning.

Note that this means we have to build the no-alias metadata up-front
once. If it is not used, it will be discarded automatically.

>From 38bf1af41c5425a552a53feb13c71d82873f1c18 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Wed, 9 Apr 2025 12:35:39 +0100
Subject: [PATCH 1/2] [VPlan] Manage instruction medata in VPlan.

Add a new helper to manage IR metadata that can be progated to generated
instructions for recipes.

This helps to remove a number of remaining uses of getUnderlyingInstr
during VPlan execution.

Add metadata on cloning.

!fixup address latest comments, thanks
---
 .../Transforms/Vectorize/LoopVectorize.cpp    |  34 +++--
 .../Transforms/Vectorize/VPRecipeBuilder.h    |   5 +
 llvm/lib/Transforms/Vectorize/VPlan.cpp       |  20 +--
 llvm/lib/Transforms/Vectorize/VPlan.h         | 120 +++++++++++-------
 llvm/lib/Transforms/Vectorize/VPlanHelpers.h  |   9 +-
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp |  38 ++++--
 .../Transforms/Vectorize/VPlanTransforms.cpp  |  31 +++--
 .../Transforms/Vectorize/VPlanTest.cpp        |  24 ++--
 8 files changed, 168 insertions(+), 113 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 7a5f618d09e95..398e575199b21 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -8599,11 +8599,13 @@ VPRecipeBuilder::tryToWidenMemory(Instruction *I, ArrayRef<VPValue *> Operands,
   }
   if (LoadInst *Load = dyn_cast<LoadInst>(I))
     return new VPWidenLoadRecipe(*Load, Ptr, Mask, Consecutive, Reverse,
+                                 getMetadataToPropagate(Load),
                                  I->getDebugLoc());
 
   StoreInst *Store = cast<StoreInst>(I);
   return new VPWidenStoreRecipe(*Store, Ptr, Operands[0], Mask, Consecutive,
-                                Reverse, I->getDebugLoc());
+                                Reverse, getMetadataToPropagate(Store),
+                                I->getDebugLoc());
 }
 
 /// Creates a VPWidenIntOrFpInductionRecpipe for \p Phi. If needed, it will also
@@ -8745,6 +8747,7 @@ VPSingleDefRecipe *VPRecipeBuilder::tryToWidenCall(CallInst *CI,
                 Range);
   if (ShouldUseVectorIntrinsic)
     return new VPWidenIntrinsicRecipe(*CI, ID, Ops, CI->getType(),
+                                      getMetadataToPropagate(CI),
                                       CI->getDebugLoc());
 
   Function *Variant = nullptr;
@@ -8798,7 +8801,8 @@ VPSingleDefRecipe *VPRecipeBuilder::tryToWidenCall(CallInst *CI,
     }
 
     Ops.push_back(Operands.back());
-    return new VPWidenCallRecipe(CI, Variant, Ops, CI->getDebugLoc());
+    return new VPWidenCallRecipe(CI, Variant, Ops, getMetadataToPropagate(CI),
+                                 CI->getDebugLoc());
   }
 
   return nullptr;
@@ -8836,7 +8840,8 @@ VPWidenRecipe *VPRecipeBuilder::tryToWiden(Instruction *I,
           Plan.getOrAddLiveIn(ConstantInt::get(I->getType(), 1u, false));
       auto *SafeRHS = Builder.createSelect(Mask, Ops[1], One, I->getDebugLoc());
       Ops[1] = SafeRHS;
-      return new VPWidenRecipe(*I, make_range(Ops.begin(), Ops.end()));
+      return new VPWidenRecipe(*I, make_range(Ops.begin(), Ops.end()),
+                               getMetadataToPropagate(I));
     }
     [[fallthrough]];
   }
@@ -8882,7 +8887,8 @@ VPWidenRecipe *VPRecipeBuilder::tryToWiden(Instruction *I,
       // For other binops, the legacy cost model only checks the second operand.
       NewOps[1] = GetConstantViaSCEV(NewOps[1]);
     }
-    return new VPWidenRecipe(*I, make_range(NewOps.begin(), NewOps.end()));
+    return new VPWidenRecipe(*I, make_range(NewOps.begin(), NewOps.end()),
+                             getMetadataToPropagate(I));
   }
   case Instruction::ExtractValue: {
     SmallVector<VPValue *> NewOps(Operands);
@@ -8891,7 +8897,8 @@ VPWidenRecipe *VPRecipeBuilder::tryToWiden(Instruction *I,
     assert(EVI->getNumIndices() == 1 && "Expected one extractvalue index");
     unsigned Idx = EVI->getIndices()[0];
     NewOps.push_back(Plan.getOrAddLiveIn(ConstantInt::get(I32Ty, Idx, false)));
-    return new VPWidenRecipe(*I, make_range(NewOps.begin(), NewOps.end()));
+    return new VPWidenRecipe(*I, make_range(NewOps.begin(), NewOps.end()),
+                             getMetadataToPropagate(I));
   }
   };
 }
@@ -9096,6 +9103,13 @@ bool VPRecipeBuilder::getScaledReductions(
   return false;
 }
 
+SmallVector<std::pair<unsigned, MDNode *>>
+VPRecipeBuilder::getMetadataToPropagate(Instruction *I) {
+  SmallVector<std::pair<unsigned, MDNode *>> Metadata;
+  ::getMetadataToPropagate(I, Metadata);
+  return Metadata;
+}
+
 VPRecipeBase *VPRecipeBuilder::tryToCreateWidenRecipe(
     Instruction *Instr, ArrayRef<VPValue *> Operands, VFRange &Range) {
   // First, check for specific widening recipes that deal with inductions, Phi
@@ -9168,13 +9182,14 @@ VPRecipeBase *VPRecipeBuilder::tryToCreateWidenRecipe(
                                 make_range(Operands.begin(), Operands.end()));
 
   if (auto *SI = dyn_cast<SelectInst>(Instr)) {
-    return new VPWidenSelectRecipe(
-        *SI, make_range(Operands.begin(), Operands.end()));
+    return new VPWidenSelectRecipe(*SI,
+                                   make_range(Operands.begin(), Operands.end()),
+                                   getMetadataToPropagate(SI));
   }
 
   if (auto *CI = dyn_cast<CastInst>(Instr)) {
     return new VPWidenCastRecipe(CI->getOpcode(), Operands[0], CI->getType(),
-                                 *CI);
+                                 *CI, getMetadataToPropagate(CI));
   }
 
   return tryToWiden(Instr, Operands);
@@ -9200,7 +9215,8 @@ VPRecipeBuilder::tryToCreatePartialReduction(Instruction *Reduction,
     SmallVector<VPValue *, 2> Ops;
     Ops.push_back(Plan.getOrAddLiveIn(Zero));
     Ops.push_back(BinOp);
-    BinOp = new VPWidenRecipe(*Reduction, make_range(Ops.begin(), Ops.end()));
+    BinOp = new VPWidenRecipe(*Reduction, make_range(Ops.begin(), Ops.end()),
+                              getMetadataToPropagate(Reduction));
     Builder.insert(BinOp->getDefiningRecipe());
     ReductionOpcode = Instruction::Add;
   }
diff --git a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
index fd0064a34c4c9..aab34834ae9bf 100644
--- a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
+++ b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
@@ -233,6 +233,11 @@ class VPRecipeBuilder {
     }
     return Plan.getOrAddLiveIn(V);
   }
+
+  /// Returns the metatadata that can be preserved from the original instruction
+  /// \p I, including noalias metadata guaranteed by runtime checks.
+  static SmallVector<std::pair<unsigned, MDNode *>>
+  getMetadataToPropagate(Instruction *I);
 };
 } // end namespace llvm
 
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp
index fa2d95a44609a..e7eb4b36ced07 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp
@@ -355,23 +355,13 @@ BasicBlock *VPTransformState::CFGState::getPreheaderBBFor(VPRecipeBase *R) {
   return VPBB2IRBB[LoopRegion->getPreheaderVPBB()];
 }
 
-void VPTransformState::addNewMetadata(Instruction *To,
-                                      const Instruction *Orig) {
+void VPTransformState::addNewMetadata(Value *To, const Instruction *Orig) {
+
   // If the loop was versioned with memchecks, add the corresponding no-alias
   // metadata.
-  if (LVer && isa<LoadInst, StoreInst>(Orig))
-    LVer->annotateInstWithNoAlias(To, Orig);
-}
-
-void VPTransformState::addMetadata(Value *To, Instruction *From) {
-  // No source instruction to transfer metadata from?
-  if (!From)
-    return;
-
-  if (Instruction *ToI = dyn_cast<Instruction>(To)) {
-    propagateMetadata(ToI, From);
-    addNewMetadata(ToI, From);
-  }
+  Instruction *ToI = dyn_cast<Instruction>(To);
+  if (ToI && LVer && isa<LoadInst, StoreInst>(Orig))
+    LVer->annotateInstWithNoAlias(ToI, Orig);
 }
 
 void VPTransformState::setDebugLocFrom(DebugLoc DL) {
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 7084676af6d5b..0692ad4e3bbfb 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1190,28 +1190,49 @@ struct VPIRPhi : public VPIRInstruction {
 #endif
 };
 
+using MDArrayRef = ArrayRef<std::pair<unsigned, MDNode *>>;
+
+/// Helper to manage IR metadata for recipes. It filters out metadata that
+/// cannot be proagated.
+class VPIRMetadata {
+  SmallVector<std::pair<unsigned, MDNode *>> Metadata;
+
+protected:
+  VPIRMetadata(MDArrayRef Metadata) : Metadata(Metadata) {}
+
+public:
+  /// Add all metadata to \p V if it is an instruction.
+  void applyMetadata(Value *V) const;
+
+  /// Return the IR metadata.
+  MDArrayRef getMetadata() const { return Metadata; }
+};
+
 /// VPWidenRecipe is a recipe for producing a widened instruction using the
 /// opcode and operands of the recipe. This recipe covers most of the
 /// traditional vectorization cases where each recipe transforms into a
 /// vectorized version of itself.
-class VPWidenRecipe : public VPRecipeWithIRFlags {
+class VPWidenRecipe : public VPRecipeWithIRFlags, public VPIRMetadata {
   unsigned Opcode;
 
 protected:
   template <typename IterT>
   VPWidenRecipe(unsigned VPDefOpcode, Instruction &I,
-                iterator_range<IterT> Operands)
-      : VPRecipeWithIRFlags(VPDefOpcode, Operands, I), Opcode(I.getOpcode()) {}
+                iterator_range<IterT> Operands, MDArrayRef Metadata)
+      : VPRecipeWithIRFlags(VPDefOpcode, Operands, I), VPIRMetadata(Metadata),
+        Opcode(I.getOpcode()) {}
 
 public:
   template <typename IterT>
-  VPWidenRecipe(Instruction &I, iterator_range<IterT> Operands)
-      : VPWidenRecipe(VPDef::VPWidenSC, I, Operands) {}
+  VPWidenRecipe(Instruction &I, iterator_range<IterT> Operands,
+                MDArrayRef Metadata)
+      : VPWidenRecipe(VPDef::VPWidenSC, I, Operands, Metadata) {}
 
   ~VPWidenRecipe() override = default;
 
   VPWidenRecipe *clone() override {
-    auto *R = new VPWidenRecipe(*getUnderlyingInstr(), operands());
+    auto *R =
+        new VPWidenRecipe(*getUnderlyingInstr(), operands(), getMetadata());
     R->transferFlags(*this);
     return R;
   }
@@ -1236,7 +1257,7 @@ class VPWidenRecipe : public VPRecipeWithIRFlags {
 };
 
 /// VPWidenCastRecipe is a recipe to create vector cast instructions.
-class VPWidenCastRecipe : public VPRecipeWithIRFlags {
+class VPWidenCastRecipe : public VPRecipeWithIRFlags, public VPIRMetadata {
   /// Cast instruction opcode.
   Instruction::CastOps Opcode;
 
@@ -1245,23 +1266,23 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
 
 public:
   VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy,
-                    CastInst &UI)
-      : VPRecipeWithIRFlags(VPDef::VPWidenCastSC, Op, UI), Opcode(Opcode),
-        ResultTy(ResultTy) {
+                    CastInst &UI, MDArrayRef Metadata)
+      : VPRecipeWithIRFlags(VPDef::VPWidenCastSC, Op, UI),
+        VPIRMetadata(Metadata), Opcode(Opcode), ResultTy(ResultTy) {
     assert(UI.getOpcode() == Opcode &&
            "opcode of underlying cast doesn't match");
   }
 
   VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy)
-      : VPRecipeWithIRFlags(VPDef::VPWidenCastSC, Op), Opcode(Opcode),
-        ResultTy(ResultTy) {}
+      : VPRecipeWithIRFlags(VPDef::VPWidenCastSC, Op), VPIRMetadata({}),
+        Opcode(Opcode), ResultTy(ResultTy) {}
 
   ~VPWidenCastRecipe() override = default;
 
   VPWidenCastRecipe *clone() override {
     if (auto *UV = getUnderlyingValue())
       return new VPWidenCastRecipe(Opcode, getOperand(0), ResultTy,
-                                   *cast<CastInst>(UV));
+                                   *cast<CastInst>(UV), getMetadata());
 
     return new VPWidenCastRecipe(Opcode, getOperand(0), ResultTy);
   }
@@ -1288,7 +1309,7 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
 };
 
 /// A recipe for widening vector intrinsics.
-class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
+class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags, public VPIRMetadata {
   /// ID of the vector intrinsic to widen.
   Intrinsic::ID VectorIntrinsicID;
 
@@ -1307,18 +1328,19 @@ class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
 public:
   VPWidenIntrinsicRecipe(CallInst &CI, Intrinsic::ID VectorIntrinsicID,
                          ArrayRef<VPValue *> CallArguments, Type *Ty,
-                         DebugLoc DL = {})
+                         MDArrayRef Metadata, DebugLoc DL = {})
       : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, CI),
-        VectorIntrinsicID(VectorIntrinsicID), ResultTy(Ty),
-        MayReadFromMemory(CI.mayReadFromMemory()),
+        VPIRMetadata(Metadata), VectorIntrinsicID(VectorIntrinsicID),
+        ResultTy(Ty), MayReadFromMemory(CI.mayReadFromMemory()),
         MayWriteToMemory(CI.mayWriteToMemory()),
         MayHaveSideEffects(CI.mayHaveSideEffects()) {}
 
   VPWidenIntrinsicRecipe(Intrinsic::ID VectorIntrinsicID,
                          ArrayRef<VPValue *> CallArguments, Type *Ty,
-                         DebugLoc DL = {})
+                         MDArrayRef Metadata, DebugLoc DL = {})
       : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, DL),
-        VectorIntrinsicID(VectorIntrinsicID), ResultTy(Ty) {
+        VPIRMetadata(Metadata), VectorIntrinsicID(VectorIntrinsicID),
+        ResultTy(Ty) {
     LLVMContext &Ctx = Ty->getContext();
     AttributeSet Attrs = Intrinsic::getFnAttributes(Ctx, VectorIntrinsicID);
     MemoryEffects ME = Attrs.getMemoryEffects();
@@ -1334,7 +1356,7 @@ class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
   VPWidenIntrinsicRecipe *clone() override {
     return new VPWidenIntrinsicRecipe(*cast<CallInst>(getUnderlyingValue()),
                                       VectorIntrinsicID, {op_begin(), op_end()},
-                                      ResultTy, getDebugLoc());
+                                      ResultTy, getMetadata(), getDebugLoc());
   }
 
   VP_CLASSOF_IMPL(VPDef::VPWidenIntrinsicSC)
@@ -1374,7 +1396,7 @@ class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
 };
 
 /// A recipe for widening Call instructions using library calls.
-class VPWidenCallRecipe : public VPRecipeWithIRFlags {
+class VPWidenCallRecipe : public VPRecipeWithIRFlags, public VPIRMetadata {
   /// Variant stores a pointer to the chosen function. There is a 1:1 mapping
   /// between a given VF and the chosen vectorized variant, so there will be a
   /// different VPlan for each VF with a valid variant.
@@ -1382,10 +1404,11 @@ class VPWidenCallRecipe : public VPRecipeWithIRFlags {
 
 public:
   VPWidenCallRecipe(Value *UV, Function *Variant,
-                    ArrayRef<VPValue *> CallArguments, DebugLoc DL = {})
+                    ArrayRef<VPValue *> CallArguments, MDArrayRef Metadata,
+                    DebugLoc DL = {})
       : VPRecipeWithIRFlags(VPDef::VPWidenCallSC, CallArguments,
                             *cast<Instruction>(UV)),
-        Variant(Variant) {
+        VPIRMetadata(Metadata), Variant(Variant) {
     assert(
         isa<Function>(getOperand(getNumOperands() - 1)->getLiveInIRValue()) &&
         "last operand must be the called function");
@@ -1395,7 +1418,8 @@ class VPWidenCallRecipe : public VPRecipeWithIRFlags {
 
   VPWidenCallRecipe *clone() override {
     return new VPWidenCallRecipe(getUnderlyingValue(), Variant,
-                                 {op_begin(), op_end()}, getDebugLoc());
+                                 {op_begin(), op_end()}, getMetadata(),
+                                 getDebugLoc());
   }
 
   VP_CLASSOF_IMPL(VPDef::VPWidenCallSC)
@@ -1471,16 +1495,18 @@ class VPHistogramRecipe : public VPRecipeBase {
 };
 
 /// A recipe for widening select instructions.
-struct VPWidenSelectRecipe : public VPRecipeWithIRFlags {
+struct VPWidenSelectRecipe : public VPRecipeWithIRFlags, public VPIRMetadata {
   template <typename IterT>
-  VPWidenSelectRecipe(SelectInst &I, iterator_range<IterT> Operands)
-      : VPRecipeWithIRFlags(VPDef::VPWidenSelectSC, Operands, I) {}
+  VPWidenSelectRecipe(SelectInst &I, iterator_range<IterT> Operands,
+                      MDArrayRef Metadata)
+      : VPRecipeWithIRFlags(VPDef::VPWidenSelectSC, Operands, I),
+        VPIRMetadata(Metadata) {}
 
   ~VPWidenSelectRecipe() override = default;
 
   VPWidenSelectRecipe *clone() override {
     return new VPWidenSelectRecipe(*cast<SelectInst>(getUnderlyingInstr()),
-                                   operands());
+                                   operands(), getMetadata());
   }
 
   VP_CLASSOF_IMPL(VPDef::VPWidenSelectSC)
@@ -2602,7 +2628,7 @@ class VPPredInstPHIRecipe : public VPSingleDefRecipe {
 
 /// A common base class for widening memory operations. An optional mask can be
 /// provided as the last operand.
-class VPWidenMemoryRecipe : public VPRecipeBase {
+class VPWidenMemoryRecipe : public VPRecipeBase, public VPIRMetadata {
 protected:
   Instruction &Ingredient;
 
@@ -2625,9 +2651,10 @@ class VPWidenMemoryRecipe : public VPRecipeBase {
 
   VPWidenMemoryRecipe(const char unsigned SC, Instruction &I,
                       std::initializer_list<VPValue *> Operands,
-                      bool Consecutive, bool Reverse, DebugLoc DL)
-      : VPRecipeBase(SC, Operands, DL), Ingredient(I), Consecutive(Consecutive),
-        Reverse(Reverse) {
+                      bool Consecutive, bool Reverse, MDArrayRef Metadata,
+                      DebugLoc DL)
+      : VPRecipeBase(SC, Operands, DL), VPIRMetadata(Metadata), Ingredient(I),
+        Consecutive(Consecutive), Reverse(Reverse) {
     assert((Consecutive || !Reverse) && "Reverse implies consecutive");
   }
 
@@ -2684,17 +2711,19 @@ class VPWidenMemoryRecipe : public VPRecipeBase {
 /// optional mask.
 struct VPWidenLoadRecipe final : public VPWidenMemoryRecipe, public VPValue {
   VPWidenLoadRecipe(LoadInst &Load, VPValue *Addr, VPValue *Mask,
-                    bool Consecutive, bool Reverse, DebugLoc DL)
+                    bool Consecutive, bool Reverse, MDArrayRef Metadata,
+                    DebugLoc DL)
       : VPWidenMemoryRecipe(VPDef::VPWidenLoadSC, Load, {Addr}, Consecutive,
-                            Reverse, DL),
+                            Reverse, Metadata, DL),
         VPValue(this, &Load) {
     setMask(Mask);
   }
 
   VPWidenLoadRecipe *clone() override {
-    return new VPWidenLoadRecipe(cast<LoadInst>(Ingredient), getAddr(),
-                                 getMask(), Consecutive, Reverse,
-                                 getDebugLoc());
+    auto *Copy = new VPWidenLoadRecipe(cast<LoadInst>(Ingredient), getAddr(),
+                                       getMask(), Consecutive, Reverse,
+                                       getMetadata(), getDebugLoc());
+    return Copy;
   }
 
   VP_CLASSOF_IMPL(VPDef::VPWidenLoadSC);
@@ -2725,7 +2754,7 @@ struct VPWidenLoadEVLRecipe final : public VPWidenMemoryRecipe, public VPValue {
   VPWidenLoadEVLRecipe(VPWidenLoadRecipe &L, VPValue &EVL, VPValue *Mask)
       : VPWidenMemoryRecipe(VPDef::VPWidenLoadEVLSC, L.getIngredient(),
                             {L.getAddr(), &EVL}, L.isConsecutive(),
-                            L.isReverse(), L.getDebugLoc()),
+                            L.isReverse(), L.getMetadata(), L.getDebugLoc()),
         VPValue(this, &getIngredient()) {
     setMask(Mask);
   }
@@ -2762,16 +2791,18 @@ struct VPWidenLoadEVLRecipe final : public VPWidenMemoryRecipe, public VPValue {
 /// to store to and an optional mask.
 struct VPWidenStoreRecipe final : public VPWidenMemoryRecipe {
   VPWidenStoreRecipe(StoreInst &Store, VPValue *Addr, VPValue *StoredVal,
-                     VPValue *Mask, bool Consecutive, bool Reverse, DebugLoc DL)
+                     VPValue *Mask, bool Consecutive, bool Reverse,
+                     MDArrayRef Metadata, DebugLoc DL)
       : VPWidenMemoryRecipe(VPDef::VPWidenStoreSC, Store, {Addr, StoredVal},
-                            Consecutive, Reverse, DL) {
+                            Consecutive, Reverse, Metadata, DL) {
     setMask(Mask);
   }
 
   VPWidenStoreRecipe *clone() override {
-    return new VPWidenStoreRecipe(cast<StoreInst>(Ingredient), getAddr(),
-                                  getStoredValue(), getMask(), Consecutive,
-                                  Reverse, getDebugLoc());
+    auto *Copy = new VPWidenStoreRecipe(
+        cast<StoreInst>(Ingredient), getAddr(), getStoredValue(), getMask(),
+        Consecutive, Reverse, getMetadata(), getDebugLoc());
+    return Copy;
   }
 
   VP_CLASSOF_IMPL(VPDef::VPWidenStoreSC);
@@ -2805,7 +2836,8 @@ struct VPWidenStoreEVLRecipe final : public VPWidenMemoryRecipe {
   VPWidenStoreEVLRecipe(VPWidenStoreRecipe &S, VPValue &EVL, VPValue *Mask)
       : VPWidenMemoryRecipe(VPDef::VPWidenStoreEVLSC, S.getIngredient(),
                             {S.getAddr(), S.getStoredValue(), &EVL},
-                            S.isConsecutive(), S.isReverse(), S.getDebugLoc()) {
+                            S.isConsecutive(), S.isReverse(), S.getMetadata(),
+                            S.getDebugLoc()) {
     setMask(Mask);
   }
 
diff --git a/llvm/lib/Transforms/Vectorize/VPlanHelpers.h b/llvm/lib/Transforms/Vectorize/VPlanHelpers.h
index bebea1915690f..4826dfd562c2d 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanHelpers.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanHelpers.h
@@ -288,14 +288,7 @@ struct VPTransformState {
   /// Currently this is used to add the noalias annotations based on the
   /// inserted memchecks.  Use this for instructions that are *cloned* into the
   /// vector loop.
-  void addNewMetadata(Instruction *To, const Instruction *Orig);
-
-  /// Add metadata from one instruction to another.
-  ///
-  /// This includes both the original MDs from \p From and additional ones (\see
-  /// addNewMetadata).  Use this for *newly created* instructions in the vector
-  /// loop.
-  void addMetadata(Value *To, Instruction *From);
+  void addNewMetadata(Value *To, const Instruction *Orig);
 
   /// Set the debug location in the builder using the debug location \p DL.
   void setDebugLocFrom(DebugLoc DL);
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 2cc558f49ccce..b5a5d8b7ace7d 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -1204,6 +1204,14 @@ void VPIRPhi::print(raw_ostream &O, const Twine &Indent,
 }
 #endif
 
+void VPIRMetadata::applyMetadata(Value *V) const {
+  auto *I = dyn_cast<Instruction>(V);
+  if (!I)
+    return;
+  for (const auto &[Kind, Node] : Metadata)
+    I->setMetadata(Kind, Node);
+}
+
 void VPWidenCallRecipe::execute(VPTransformState &State) {
   assert(State.VF.isVector() && "not widening");
 
@@ -1231,10 +1239,10 @@ void VPWidenCallRecipe::execute(VPTransformState &State) {
 
   CallInst *V = State.Builder.CreateCall(Variant, Args, OpBundles);
   setFlags(V);
+  applyMetadata(V);
 
   if (!V->getType()->isVoidTy())
     State.set(this, V);
-  State.addMetadata(V, CI);
 }
 
 InstructionCost VPWidenCallRecipe::computeCost(ElementCount VF,
@@ -1310,10 +1318,10 @@ void VPWidenIntrinsicRecipe::execute(VPTransformState &State) {
   CallInst *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
 
   setFlags(V);
+  applyMetadata(V);
 
   if (!V->getType()->isVoidTy())
     State.set(this, V);
-  State.addMetadata(V, CI);
 }
 
 InstructionCost VPWidenIntrinsicRecipe::computeCost(ElementCount VF,
@@ -1510,7 +1518,7 @@ void VPWidenSelectRecipe::execute(VPTransformState &State) {
   State.set(this, Sel);
   if (isa<FPMathOperator>(Sel))
     setFlags(cast<Instruction>(Sel));
-  State.addMetadata(Sel, dyn_cast_or_null<Instruction>(getUnderlyingValue()));
+  applyMetadata(Sel);
 }
 
 InstructionCost VPWidenSelectRecipe::computeCost(ElementCount VF,
@@ -1641,12 +1649,13 @@ void VPWidenRecipe::execute(VPTransformState &State) {
 
     Value *V = Builder.CreateNAryOp(Opcode, Ops);
 
-    if (auto *VecOp = dyn_cast<Instruction>(V))
+    if (auto *VecOp = dyn_cast<Instruction>(V)) {
       setFlags(VecOp);
+      applyMetadata(V);
+    }
 
     // Use this vector value for all users of the original instruction.
     State.set(this, V);
-    State.addMetadata(V, dyn_cast_or_null<Instruction>(getUnderlyingValue()));
     break;
   }
   case Instruction::ExtractValue: {
@@ -1678,8 +1687,8 @@ void VPWidenRecipe::execute(VPTransformState &State) {
     } else {
       C = Builder.CreateICmp(getPredicate(), A, B);
     }
+    applyMetadata(C);
     State.set(this, C);
-    State.addMetadata(C, dyn_cast_or_null<Instruction>(getUnderlyingValue()));
     break;
   }
   default:
@@ -1795,9 +1804,10 @@ void VPWidenCastRecipe::execute(VPTransformState &State) {
   Value *A = State.get(Op);
   Value *Cast = Builder.CreateCast(Instruction::CastOps(Opcode), A, DestTy);
   State.set(this, Cast);
-  State.addMetadata(Cast, cast_or_null<Instruction>(getUnderlyingValue()));
-  if (auto *CastOp = dyn_cast<Instruction>(Cast))
+  if (auto *CastOp = dyn_cast<Instruction>(Cast)) {
     setFlags(CastOp);
+    applyMetadata(CastOp);
+  }
 }
 
 InstructionCost VPWidenCastRecipe::computeCost(ElementCount VF,
@@ -2749,7 +2759,8 @@ void VPWidenLoadRecipe::execute(VPTransformState &State) {
     NewLI = Builder.CreateAlignedLoad(DataTy, Addr, Alignment, "wide.load");
   }
   // Add metadata to the load, but setVectorValue to the reverse shuffle.
-  State.addMetadata(NewLI, LI);
+  State.addNewMetadata(NewLI, LI);
+  applyMetadata(NewLI);
   if (Reverse)
     NewLI = Builder.CreateVectorReverse(NewLI, "reverse");
   State.set(this, NewLI);
@@ -2809,7 +2820,8 @@ void VPWidenLoadEVLRecipe::execute(VPTransformState &State) {
   }
   NewLI->addParamAttr(
       0, Attribute::getWithAlignment(NewLI->getContext(), Alignment));
-  State.addMetadata(NewLI, LI);
+  State.addNewMetadata(NewLI, LI);
+  applyMetadata(NewLI);
   Instruction *Res = NewLI;
   if (isReverse())
     Res = createReverseEVL(Builder, Res, EVL, "vp.reverse");
@@ -2885,7 +2897,8 @@ void VPWidenStoreRecipe::execute(VPTransformState &State) {
     NewSI = Builder.CreateMaskedStore(StoredVal, Addr, Alignment, Mask);
   else
     NewSI = Builder.CreateAlignedStore(StoredVal, Addr, Alignment);
-  State.addMetadata(NewSI, SI);
+  State.addNewMetadata(NewSI, SI);
+  applyMetadata(NewSI);
 }
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
@@ -2932,7 +2945,8 @@ void VPWidenStoreEVLRecipe::execute(VPTransformState &State) {
   }
   NewSI->addParamAttr(
       1, Attribute::getWithAlignment(NewSI->getContext(), Alignment));
-  State.addMetadata(NewSI, SI);
+  State.addNewMetadata(NewSI, SI);
+  applyMetadata(NewSI);
 }
 
 InstructionCost VPWidenStoreEVLRecipe::computeCost(ElementCount VF,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index b80fe18d1bd66..445bcac5919fc 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -70,6 +70,7 @@ bool VPlanTransforms::tryToConvertVPInstructionsToVPRecipes(
         NewRecipe = new VPWidenIntOrFpInductionRecipe(
             Phi, Start, Step, &Plan->getVF(), *II, Ingredient.getDebugLoc());
       } else {
+        auto Metadata = VPRecipeBuilder::getMetadataToPropagate(Inst);
         assert(isa<VPInstruction>(&Ingredient) &&
                "only VPInstructions expected here");
         assert(!isa<PHINode>(Inst) && "phis should be handled above");
@@ -77,13 +78,13 @@ bool VPlanTransforms::tryToConvertVPInstructionsToVPRecipes(
         if (LoadInst *Load = dyn_cast<LoadInst>(Inst)) {
           NewRecipe = new VPWidenLoadRecipe(
               *Load, Ingredient.getOperand(0), nullptr /*Mask*/,
-              false /*Consecutive*/, false /*Reverse*/,
+              false /*Consecutive*/, false /*Reverse*/, Metadata,
               Ingredient.getDebugLoc());
         } else if (StoreInst *Store = dyn_cast<StoreInst>(Inst)) {
           NewRecipe = new VPWidenStoreRecipe(
               *Store, Ingredient.getOperand(1), Ingredient.getOperand(0),
               nullptr /*Mask*/, false /*Consecutive*/, false /*Reverse*/,
-              Ingredient.getDebugLoc());
+              Metadata, Ingredient.getDebugLoc());
         } else if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Inst)) {
           NewRecipe = new VPWidenGEPRecipe(GEP, Ingredient.operands());
         } else if (CallInst *CI = dyn_cast<CallInst>(Inst)) {
@@ -93,14 +94,16 @@ bool VPlanTransforms::tryToConvertVPInstructionsToVPRecipes(
           NewRecipe = new VPWidenIntrinsicRecipe(
               *CI, getVectorIntrinsicIDForCall(CI, &TLI),
               {Ingredient.op_begin(), Ingredient.op_end() - 1}, CI->getType(),
-              CI->getDebugLoc());
+              Metadata, CI->getDebugLoc());
         } else if (SelectInst *SI = dyn_cast<SelectInst>(Inst)) {
-          NewRecipe = new VPWidenSelectRecipe(*SI, Ingredient.operands());
+          NewRecipe =
+              new VPWidenSelectRecipe(*SI, Ingredient.operands(), Metadata);
         } else if (auto *CI = dyn_cast<CastInst>(Inst)) {
-          NewRecipe = new VPWidenCastRecipe(
-              CI->getOpcode(), Ingredient.getOperand(0), CI->getType(), *CI);
+          NewRecipe =
+              new VPWidenCastRecipe(CI->getOpcode(), Ingredient.getOperand(0),
+                                    CI->getType(), *CI, Metadata);
         } else {
-          NewRecipe = new VPWidenRecipe(*Inst, Ingredient.operands());
+          NewRecipe = new VPWidenRecipe(*Inst, Ingredient.operands(), Metadata);
         }
       }
 
@@ -2009,9 +2012,9 @@ static VPRecipeBase *createEVLRecipe(VPValue *HeaderMask,
       .Case<VPWidenSelectRecipe>([&](VPWidenSelectRecipe *Sel) {
         SmallVector<VPValue *> Ops(Sel->operands());
         Ops.push_back(&EVL);
-        return new VPWidenIntrinsicRecipe(Intrinsic::vp_select, Ops,
-                                          TypeInfo.inferScalarType(Sel),
-                                          Sel->getDebugLoc());
+        return new VPWidenIntrinsicRecipe(
+            Intrinsic::vp_select, Ops, TypeInfo.inferScalarType(Sel),
+            Sel->getMetadata(), Sel->getDebugLoc());
       })
       .Case<VPInstruction>([&](VPInstruction *VPI) -> VPRecipeBase * {
         if (VPI->getOpcode() == VPInstruction::FirstOrderRecurrenceSplice) {
@@ -2023,7 +2026,7 @@ static VPRecipeBase *createEVLRecipe(VPValue *HeaderMask,
           Ops.append({MinusOneVPV, &AllOneMask, PrevEVL, &EVL});
           return new VPWidenIntrinsicRecipe(Intrinsic::experimental_vp_splice,
                                             Ops, TypeInfo.inferScalarType(VPI),
-                                            VPI->getDebugLoc());
+                                            {}, VPI->getDebugLoc());
         }
 
         VPValue *LHS, *RHS;
@@ -2038,7 +2041,7 @@ static VPRecipeBase *createEVLRecipe(VPValue *HeaderMask,
         // limited to selects whose condition is a header mask.
         return new VPWidenIntrinsicRecipe(
             Intrinsic::vp_merge, {&AllOneMask, LHS, RHS, &EVL},
-            TypeInfo.inferScalarType(LHS), VPI->getDebugLoc());
+            TypeInfo.inferScalarType(LHS), {}, VPI->getDebugLoc());
       })
       .Default([&](VPRecipeBase *R) { return nullptr; });
 }
@@ -2755,7 +2758,7 @@ void VPlanTransforms::narrowInterleaveGroups(VPlan &Plan, ElementCount VF,
       auto *L = new VPWidenLoadRecipe(
           *cast<LoadInst>(LoadGroup->getInterleaveGroup()->getInsertPos()),
           LoadGroup->getAddr(), LoadGroup->getMask(), /*Consecutive=*/true,
-          /*Reverse=*/false, LoadGroup->getDebugLoc());
+          /*Reverse=*/false, {}, LoadGroup->getDebugLoc());
       L->insertBefore(LoadGroup);
       return L;
     }
@@ -2786,7 +2789,7 @@ void VPlanTransforms::narrowInterleaveGroups(VPlan &Plan, ElementCount VF,
     auto *S = new VPWidenStoreRecipe(
         *cast<StoreInst>(StoreGroup->getInterleaveGroup()->getInsertPos()),
         StoreGroup->getAddr(), Res, nullptr, /*Consecutive=*/true,
-        /*Reverse=*/false, StoreGroup->getDebugLoc());
+        /*Reverse=*/false, {}, StoreGroup->getDebugLoc());
     S->insertBefore(StoreGroup);
     StoreGroup->eraseFromParent();
   }
diff --git a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
index cb7545171744e..277e03cbe8588 100644
--- a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
+++ b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
@@ -929,7 +929,7 @@ TEST_F(VPRecipeTest, CastVPWidenRecipeToVPUser) {
   SmallVector<VPValue *, 2> Args;
   Args.push_back(Op1);
   Args.push_back(Op2);
-  VPWidenRecipe WidenR(*AI, make_range(Args.begin(), Args.end()));
+  VPWidenRecipe WidenR(*AI, make_range(Args.begin(), Args.end()), {});
   EXPECT_TRUE(isa<VPUser>(&WidenR));
   VPRecipeBase *WidenRBase = &WidenR;
   EXPECT_TRUE(isa<VPUser>(WidenRBase));
@@ -950,7 +950,7 @@ TEST_F(VPRecipeTest, CastVPWidenCallRecipeToVPUserAndVPDef) {
   Args.push_back(Op1);
   Args.push_back(Op2);
   Args.push_back(CalledFn);
-  VPWidenCallRecipe Recipe(Call, Fn, Args);
+  VPWidenCallRecipe Recipe(Call, Fn, Args, {});
   EXPECT_TRUE(isa<VPUser>(&Recipe));
   VPRecipeBase *BaseR = &Recipe;
   EXPECT_TRUE(isa<VPUser>(BaseR));
@@ -978,7 +978,7 @@ TEST_F(VPRecipeTest, CastVPWidenSelectRecipeToVPUserAndVPDef) {
   Args.push_back(Op2);
   Args.push_back(Op3);
   VPWidenSelectRecipe WidenSelectR(*SelectI,
-                                   make_range(Args.begin(), Args.end()));
+                                   make_range(Args.begin(), Args.end()), {});
   EXPECT_TRUE(isa<VPUser>(&WidenSelectR));
   VPRecipeBase *BaseR = &WidenSelectR;
   EXPECT_TRUE(isa<VPUser>(BaseR));
@@ -1084,7 +1084,7 @@ TEST_F(VPRecipeTest, CastVPWidenMemoryRecipeToVPUserAndVPDef) {
       new LoadInst(Int32, PoisonValue::get(Int32Ptr), "", false, Align(1));
   VPValue *Addr = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 1));
   VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 2));
-  VPWidenLoadRecipe Recipe(*Load, Addr, Mask, true, false, {});
+  VPWidenLoadRecipe Recipe(*Load, Addr, Mask, true, false, {}, {});
   EXPECT_TRUE(isa<VPUser>(&Recipe));
   VPRecipeBase *BaseR = &Recipe;
   EXPECT_TRUE(isa<VPUser>(BaseR));
@@ -1111,7 +1111,7 @@ TEST_F(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
     SmallVector<VPValue *, 2> Args;
     Args.push_back(Op1);
     Args.push_back(Op2);
-    VPWidenRecipe Recipe(*AI, make_range(Args.begin(), Args.end()));
+    VPWidenRecipe Recipe(*AI, make_range(Args.begin(), Args.end()), {});
     EXPECT_FALSE(Recipe.mayHaveSideEffects());
     EXPECT_FALSE(Recipe.mayReadFromMemory());
     EXPECT_FALSE(Recipe.mayWriteToMemory());
@@ -1130,7 +1130,8 @@ TEST_F(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
     Args.push_back(Op1);
     Args.push_back(Op2);
     Args.push_back(Op3);
-    VPWidenSelectRecipe Recipe(*SelectI, make_range(Args.begin(), Args.end()));
+    VPWidenSelectRecipe Recipe(*SelectI, make_range(Args.begin(), Args.end()),
+                               {});
     EXPECT_FALSE(Recipe.mayHaveSideEffects());
     EXPECT_FALSE(Recipe.mayReadFromMemory());
     EXPECT_FALSE(Recipe.mayWriteToMemory());
@@ -1201,7 +1202,7 @@ TEST_F(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
         new LoadInst(Int32, PoisonValue::get(Int32Ptr), "", false, Align(1));
     VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 1));
     VPValue *Addr = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 2));
-    VPWidenLoadRecipe Recipe(*Load, Addr, Mask, true, false, {});
+    VPWidenLoadRecipe Recipe(*Load, Addr, Mask, true, false, {}, {});
     EXPECT_FALSE(Recipe.mayHaveSideEffects());
     EXPECT_TRUE(Recipe.mayReadFromMemory());
     EXPECT_FALSE(Recipe.mayWriteToMemory());
@@ -1215,7 +1216,8 @@ TEST_F(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
     VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 1));
     VPValue *Addr = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 2));
     VPValue *StoredV = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 3));
-    VPWidenStoreRecipe Recipe(*Store, Addr, StoredV, Mask, false, false, {});
+    VPWidenStoreRecipe Recipe(*Store, Addr, StoredV, Mask, false, false, {},
+                              {});
     EXPECT_TRUE(Recipe.mayHaveSideEffects());
     EXPECT_FALSE(Recipe.mayReadFromMemory());
     EXPECT_TRUE(Recipe.mayWriteToMemory());
@@ -1234,7 +1236,7 @@ TEST_F(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
     Args.push_back(Op1);
     Args.push_back(Op2);
     Args.push_back(CalledFn);
-    VPWidenCallRecipe Recipe(Call, Fn, Args);
+    VPWidenCallRecipe Recipe(Call, Fn, Args, {});
     EXPECT_TRUE(Recipe.mayHaveSideEffects());
     EXPECT_TRUE(Recipe.mayReadFromMemory());
     EXPECT_TRUE(Recipe.mayWriteToMemory());
@@ -1257,7 +1259,7 @@ TEST_F(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
     Args.push_back(Op1);
     Args.push_back(Op2);
     Args.push_back(CalledFn);
-    VPWidenCallRecipe Recipe(Call, TheFn, Args);
+    VPWidenCallRecipe Recipe(Call, TheFn, Args, {});
     EXPECT_FALSE(Recipe.mayHaveSideEffects());
     EXPECT_FALSE(Recipe.mayReadFromMemory());
     EXPECT_FALSE(Recipe.mayWriteToMemory());
@@ -1314,7 +1316,7 @@ TEST_F(VPRecipeTest, dumpRecipeInPlan) {
   Args.push_back(ExtVPV1);
   Args.push_back(ExtVPV2);
   VPWidenRecipe *WidenR =
-      new VPWidenRecipe(*AI, make_range(Args.begin(), Args.end()));
+      new VPWidenRecipe(*AI, make_range(Args.begin(), Args.end()), {});
   VPBB1->appendRecipe(WidenR);
 
   {

>From 2fd7844cfdf5ec0f1c2ce0b9b3ae0763245b6922 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Mon, 14 Apr 2025 21:04:42 +0200
Subject: [PATCH 2/2] [VPlan] Manage noalias/alias_scope metadata in VPlan.

Use VPIRMetadata added in https://github.com/llvm/llvm-project/pull/135272
to also manage no-alias metadata added by versioning.

Note that this means we have to build the no-alias metadata up-front
once. If it is not used, it will be discarded automatically.
---
 .../Vectorize/LoopVectorizationPlanner.h      |  3 +-
 .../Transforms/Vectorize/LoopVectorize.cpp    | 67 ++++++++++---------
 .../Transforms/Vectorize/VPRecipeBuilder.h    | 13 ++--
 llvm/lib/Transforms/Vectorize/VPlan.cpp       | 13 +---
 llvm/lib/Transforms/Vectorize/VPlan.h         | 13 ++--
 llvm/lib/Transforms/Vectorize/VPlanHelpers.h  | 15 -----
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp | 12 ----
 .../Transforms/Vectorize/VPlanTransforms.cpp  | 12 ++--
 .../LoopVectorize/AArch64/store-costs-sve.ll  | 17 ++---
 9 files changed, 69 insertions(+), 96 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h b/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
index f639f0adb9c43..d46c14f53b24f 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
@@ -36,6 +36,7 @@ class LoopVectorizationLegality;
 class LoopVectorizationCostModel;
 class PredicatedScalarEvolution;
 class LoopVectorizeHints;
+class LoopVersioning;
 class OptimizationRemarkEmitter;
 class TargetTransformInfo;
 class TargetLibraryInfo;
@@ -515,7 +516,7 @@ class LoopVectorizationPlanner {
   /// returned VPlan is valid for. If no VPlan can be built for the input range,
   /// set the largest included VF to the maximum VF for which no plan could be
   /// built.
-  VPlanPtr tryToBuildVPlanWithVPRecipes(VFRange &Range);
+  VPlanPtr tryToBuildVPlanWithVPRecipes(VFRange &Range, LoopVersioning *LVer);
 
   /// Build VPlans for power-of-2 VF's between \p MinVF and \p MaxVF inclusive,
   /// according to the information gathered by Legal when it checked if it is
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 398e575199b21..f965d53f91bfb 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -2371,7 +2371,7 @@ void InnerLoopVectorizer::scalarizeInstruction(const Instruction *Instr,
       InputLane = VPLane::getFirstLane();
     Cloned->setOperand(I.index(), State.get(Operand, InputLane));
   }
-  State.addNewMetadata(Cloned, Instr);
+  RepRecipe->applyMetadata(Cloned);
 
   // Place the cloned scalar in the new loop.
   State.Builder.Insert(Cloned);
@@ -7989,24 +7989,6 @@ DenseMap<const SCEV *, Value *> LoopVectorizationPlanner::executePlan(
   if (VectorizingEpilogue)
     VPlanTransforms::removeDeadRecipes(BestVPlan);
 
-  // Only use noalias metadata when using memory checks guaranteeing no overlap
-  // across all iterations.
-  const LoopAccessInfo *LAI = ILV.Legal->getLAI();
-  std::unique_ptr<LoopVersioning> LVer = nullptr;
-  if (LAI && !LAI->getRuntimePointerChecking()->getChecks().empty() &&
-      !LAI->getRuntimePointerChecking()->getDiffChecks()) {
-
-    //  We currently don't use LoopVersioning for the actual loop cloning but we
-    //  still use it to add the noalias metadata.
-    //  TODO: Find a better way to re-use LoopVersioning functionality to add
-    //        metadata.
-    LVer = std::make_unique<LoopVersioning>(
-        *LAI, LAI->getRuntimePointerChecking()->getChecks(), OrigLoop, LI, DT,
-        PSE.getSE());
-    State.LVer = &*LVer;
-    State.LVer->prepareNoAliasMetadata();
-  }
-
   ILV.printDebugTracesAtStart();
 
   //===------------------------------------------------===//
@@ -8597,15 +8579,14 @@ VPRecipeBuilder::tryToWidenMemory(Instruction *I, ArrayRef<VPValue *> Operands,
     Builder.insert(VectorPtr);
     Ptr = VectorPtr;
   }
+  auto Metadata = getMetadataToPropagate(I);
   if (LoadInst *Load = dyn_cast<LoadInst>(I))
     return new VPWidenLoadRecipe(*Load, Ptr, Mask, Consecutive, Reverse,
-                                 getMetadataToPropagate(Load),
-                                 I->getDebugLoc());
+                                 Metadata, I->getDebugLoc());
 
   StoreInst *Store = cast<StoreInst>(I);
   return new VPWidenStoreRecipe(*Store, Ptr, Operands[0], Mask, Consecutive,
-                                Reverse, getMetadataToPropagate(Store),
-                                I->getDebugLoc());
+                                Reverse, Metadata, I->getDebugLoc());
 }
 
 /// Creates a VPWidenIntOrFpInductionRecpipe for \p Phi. If needed, it will also
@@ -8985,8 +8966,9 @@ VPRecipeBuilder::handleReplication(Instruction *I, ArrayRef<VPValue *> Operands,
   assert((Range.Start.isScalar() || !IsUniform || !IsPredicated ||
           (Range.Start.isScalable() && isa<IntrinsicInst>(I))) &&
          "Should not predicate a uniform recipe");
-  auto *Recipe = new VPReplicateRecipe(
-      I, make_range(Operands.begin(), Operands.end()), IsUniform, BlockInMask);
+  auto *Recipe =
+      new VPReplicateRecipe(I, make_range(Operands.begin(), Operands.end()),
+                            IsUniform, BlockInMask, getMetadataToPropagate(I));
   return Recipe;
 }
 
@@ -9104,9 +9086,16 @@ bool VPRecipeBuilder::getScaledReductions(
 }
 
 SmallVector<std::pair<unsigned, MDNode *>>
-VPRecipeBuilder::getMetadataToPropagate(Instruction *I) {
+VPRecipeBuilder::getMetadataToPropagate(Instruction *I) const {
   SmallVector<std::pair<unsigned, MDNode *>> Metadata;
   ::getMetadataToPropagate(I, Metadata);
+  if (LVer && isa<LoadInst, StoreInst>(I)) {
+    const auto &[AliasScopeMD, NoAliasMD] = LVer->getNoAliasMetadataFor(I);
+    if (AliasScopeMD)
+      Metadata.emplace_back(LLVMContext::MD_alias_scope, AliasScopeMD);
+    if (NoAliasMD)
+      Metadata.emplace_back(LLVMContext::MD_noalias, NoAliasMD);
+  }
   return Metadata;
 }
 
@@ -9239,10 +9228,22 @@ void LoopVectorizationPlanner::buildVPlansWithVPRecipes(ElementCount MinVF,
                                                         ElementCount MaxVF) {
   assert(OrigLoop->isInnermost() && "Inner loop expected.");
 
+  // Only use noalias metadata when using memory checks guaranteeing no overlap
+  // across all iterations.
+  const LoopAccessInfo *LAI = Legal->getLAI();
+  std::unique_ptr<LoopVersioning> LVer = nullptr;
+  if (LAI && !LAI->getRuntimePointerChecking()->getChecks().empty() &&
+      !LAI->getRuntimePointerChecking()->getDiffChecks()) {
+    LVer = std::make_unique<LoopVersioning>(
+        *LAI, LAI->getRuntimePointerChecking()->getChecks(), OrigLoop, LI, DT,
+        PSE.getSE());
+    LVer->prepareNoAliasMetadata();
+  }
+
   auto MaxVFTimes2 = MaxVF * 2;
   for (ElementCount VF = MinVF; ElementCount::isKnownLT(VF, MaxVFTimes2);) {
     VFRange SubRange = {VF, MaxVFTimes2};
-    if (auto Plan = tryToBuildVPlanWithVPRecipes(SubRange)) {
+    if (auto Plan = tryToBuildVPlanWithVPRecipes(SubRange, LVer.get())) {
       bool HasScalarVF = Plan->hasScalarVFOnly();
       // Now optimize the initial VPlan.
       if (!HasScalarVF)
@@ -9550,7 +9551,8 @@ static void addExitUsersForFirstOrderRecurrences(
 }
 
 VPlanPtr
-LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {
+LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range,
+                                                       LoopVersioning *LVer) {
 
   using namespace llvm::VPlanPatternMatch;
   SmallPtrSet<const InterleaveGroup<Instruction> *, 1> InterleaveGroups;
@@ -9596,7 +9598,7 @@ LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {
   addCanonicalIVRecipes(*Plan, Legal->getWidestInductionType(), HasNUW, DL);
 
   VPRecipeBuilder RecipeBuilder(*Plan, OrigLoop, TLI, &TTI, Legal, CM, PSE,
-                                Builder);
+                                Builder, LVer);
 
   // ---------------------------------------------------------------------------
   // Pre-construction: record ingredients whose recipes we'll need to further
@@ -9710,8 +9712,9 @@ LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {
           Legal->isInvariantAddressOfReduction(SI->getPointerOperand())) {
         // Only create recipe for the final invariant store of the reduction.
         if (Legal->isInvariantStoreOfReduction(SI)) {
-          auto *Recipe =
-              new VPReplicateRecipe(SI, R.operands(), true /* IsUniform */);
+          auto *Recipe = new VPReplicateRecipe(
+              SI, R.operands(), true /* IsUniform */, /*Mask*/ nullptr,
+              RecipeBuilder.getMetadataToPropagate(SI));
           Recipe->insertBefore(*MiddleVPBB, MBIP);
         }
         R.eraseFromParent();
@@ -9897,7 +9900,7 @@ VPlanPtr LoopVectorizationPlanner::tryToBuildVPlan(VFRange &Range) {
   // Collect mapping of IR header phis to header phi recipes, to be used in
   // addScalarResumePhis.
   VPRecipeBuilder RecipeBuilder(*Plan, OrigLoop, TLI, &TTI, Legal, CM, PSE,
-                                Builder);
+                                Builder, nullptr);
   for (auto &R : Plan->getVectorLoopRegion()->getEntryBasicBlock()->phis()) {
     if (isa<VPCanonicalIVPHIRecipe>(&R))
       continue;
diff --git a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
index aab34834ae9bf..357a68972cb43 100644
--- a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
+++ b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
@@ -90,6 +90,10 @@ class VPRecipeBuilder {
   /// A mapping of partial reduction exit instructions to their scaling factor.
   DenseMap<const Instruction *, unsigned> ScaledReductionMap;
 
+  /// Loop versioning instance for getting noalias metadata guaranteed by
+  /// runtime checks.
+  LoopVersioning *LVer;
+
   /// Check if \p I can be widened at the start of \p Range and possibly
   /// decrease the range such that the returned value holds for the entire \p
   /// Range. The function should not be called for memory instructions or calls.
@@ -155,9 +159,10 @@ class VPRecipeBuilder {
                   const TargetTransformInfo *TTI,
                   LoopVectorizationLegality *Legal,
                   LoopVectorizationCostModel &CM,
-                  PredicatedScalarEvolution &PSE, VPBuilder &Builder)
+                  PredicatedScalarEvolution &PSE, VPBuilder &Builder,
+                  LoopVersioning *LVer)
       : Plan(Plan), OrigLoop(OrigLoop), TLI(TLI), TTI(TTI), Legal(Legal),
-        CM(CM), PSE(PSE), Builder(Builder) {}
+        CM(CM), PSE(PSE), Builder(Builder), LVer(LVer) {}
 
   std::optional<unsigned> getScalingForReduction(const Instruction *ExitInst) {
     auto It = ScaledReductionMap.find(ExitInst);
@@ -236,8 +241,8 @@ class VPRecipeBuilder {
 
   /// Returns the metatadata that can be preserved from the original instruction
   /// \p I, including noalias metadata guaranteed by runtime checks.
-  static SmallVector<std::pair<unsigned, MDNode *>>
-  getMetadataToPropagate(Instruction *I);
+  SmallVector<std::pair<unsigned, MDNode *>>
+  getMetadataToPropagate(Instruction *I) const;
 };
 } // end namespace llvm
 
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp
index e7eb4b36ced07..7ca5008c607cf 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp
@@ -220,8 +220,8 @@ VPTransformState::VPTransformState(const TargetTransformInfo *TTI,
                                    InnerLoopVectorizer *ILV, VPlan *Plan,
                                    Loop *CurrentParentLoop, Type *CanonicalIVTy)
     : TTI(TTI), VF(VF), CFG(DT), LI(LI), Builder(Builder), ILV(ILV), Plan(Plan),
-      CurrentParentLoop(CurrentParentLoop), LVer(nullptr),
-      TypeAnalysis(CanonicalIVTy), VPDT(*Plan) {}
+      CurrentParentLoop(CurrentParentLoop), TypeAnalysis(CanonicalIVTy),
+      VPDT(*Plan) {}
 
 Value *VPTransformState::get(const VPValue *Def, const VPLane &Lane) {
   if (Def->isLiveIn())
@@ -355,15 +355,6 @@ BasicBlock *VPTransformState::CFGState::getPreheaderBBFor(VPRecipeBase *R) {
   return VPBB2IRBB[LoopRegion->getPreheaderVPBB()];
 }
 
-void VPTransformState::addNewMetadata(Value *To, const Instruction *Orig) {
-
-  // If the loop was versioned with memchecks, add the corresponding no-alias
-  // metadata.
-  Instruction *ToI = dyn_cast<Instruction>(To);
-  if (ToI && LVer && isa<LoadInst, StoreInst>(Orig))
-    LVer->annotateInstWithNoAlias(ToI, Orig);
-}
-
 void VPTransformState::setDebugLocFrom(DebugLoc DL) {
   const DILocation *DIL = DL;
   // When a FSDiscriminator is enabled, we don't need to add the multiply
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 0692ad4e3bbfb..585fa3cd12ef6 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -2469,7 +2469,7 @@ class VPReductionEVLRecipe : public VPReductionRecipe {
 /// copies of the original scalar type, one per lane, instead of producing a
 /// single copy of widened type for all lanes. If the instruction is known to be
 /// uniform only one copy, per lane zero, will be generated.
-class VPReplicateRecipe : public VPRecipeWithIRFlags {
+class VPReplicateRecipe : public VPRecipeWithIRFlags, public VPIRMetadata {
   /// Indicator if only a single replica per lane is needed.
   bool IsUniform;
 
@@ -2479,9 +2479,10 @@ class VPReplicateRecipe : public VPRecipeWithIRFlags {
 public:
   template <typename IterT>
   VPReplicateRecipe(Instruction *I, iterator_range<IterT> Operands,
-                    bool IsUniform, VPValue *Mask = nullptr)
+                    bool IsUniform, VPValue *Mask = nullptr,
+                    ArrayRef<std::pair<unsigned, MDNode *>> Metadata = {})
       : VPRecipeWithIRFlags(VPDef::VPReplicateSC, Operands, *I),
-        IsUniform(IsUniform), IsPredicated(Mask) {
+        VPIRMetadata(Metadata), IsUniform(IsUniform), IsPredicated(Mask) {
     if (Mask)
       addOperand(Mask);
   }
@@ -2489,9 +2490,9 @@ class VPReplicateRecipe : public VPRecipeWithIRFlags {
   ~VPReplicateRecipe() override = default;
 
   VPReplicateRecipe *clone() override {
-    auto *Copy =
-        new VPReplicateRecipe(getUnderlyingInstr(), operands(), IsUniform,
-                              isPredicated() ? getMask() : nullptr);
+    auto *Copy = new VPReplicateRecipe(
+        getUnderlyingInstr(), operands(), IsUniform,
+        isPredicated() ? getMask() : nullptr, getMetadata());
     Copy->transferFlags(*this);
     return Copy;
   }
diff --git a/llvm/lib/Transforms/Vectorize/VPlanHelpers.h b/llvm/lib/Transforms/Vectorize/VPlanHelpers.h
index 4826dfd562c2d..9bbc8af7c13eb 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanHelpers.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanHelpers.h
@@ -38,7 +38,6 @@ class VPBasicBlock;
 class VPRegionBlock;
 class VPlan;
 class Value;
-class LoopVersioning;
 
 /// Returns a calculation for the total number of elements for a given \p VF.
 /// For fixed width vectors this value is a constant, whereas for scalable
@@ -283,13 +282,6 @@ struct VPTransformState {
     Iter->second[CacheIdx] = V;
   }
 
-  /// Add additional metadata to \p To that was not present on \p Orig.
-  ///
-  /// Currently this is used to add the noalias annotations based on the
-  /// inserted memchecks.  Use this for instructions that are *cloned* into the
-  /// vector loop.
-  void addNewMetadata(Value *To, const Instruction *Orig);
-
   /// Set the debug location in the builder using the debug location \p DL.
   void setDebugLocFrom(DebugLoc DL);
 
@@ -341,13 +333,6 @@ struct VPTransformState {
   /// The parent loop object for the current scope, or nullptr.
   Loop *CurrentParentLoop = nullptr;
 
-  /// LoopVersioning.  It's only set up (non-null) if memchecks were
-  /// used.
-  ///
-  /// This is currently only used to add no-alias metadata based on the
-  /// memchecks.  The actually versioning is performed manually.
-  LoopVersioning *LVer = nullptr;
-
   /// VPlan-based type analysis.
   VPTypeAnalysis TypeAnalysis;
 
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index b5a5d8b7ace7d..0645642dfca3f 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -2729,8 +2729,6 @@ InstructionCost VPWidenMemoryRecipe::computeCost(ElementCount VF,
 }
 
 void VPWidenLoadRecipe::execute(VPTransformState &State) {
-  auto *LI = cast<LoadInst>(&Ingredient);
-
   Type *ScalarDataTy = getLoadStoreType(&Ingredient);
   auto *DataTy = VectorType::get(ScalarDataTy, State.VF);
   const Align Alignment = getLoadStoreAlignment(&Ingredient);
@@ -2759,7 +2757,6 @@ void VPWidenLoadRecipe::execute(VPTransformState &State) {
     NewLI = Builder.CreateAlignedLoad(DataTy, Addr, Alignment, "wide.load");
   }
   // Add metadata to the load, but setVectorValue to the reverse shuffle.
-  State.addNewMetadata(NewLI, LI);
   applyMetadata(NewLI);
   if (Reverse)
     NewLI = Builder.CreateVectorReverse(NewLI, "reverse");
@@ -2788,8 +2785,6 @@ static Instruction *createReverseEVL(IRBuilderBase &Builder, Value *Operand,
 }
 
 void VPWidenLoadEVLRecipe::execute(VPTransformState &State) {
-  auto *LI = cast<LoadInst>(&Ingredient);
-
   Type *ScalarDataTy = getLoadStoreType(&Ingredient);
   auto *DataTy = VectorType::get(ScalarDataTy, State.VF);
   const Align Alignment = getLoadStoreAlignment(&Ingredient);
@@ -2820,7 +2815,6 @@ void VPWidenLoadEVLRecipe::execute(VPTransformState &State) {
   }
   NewLI->addParamAttr(
       0, Attribute::getWithAlignment(NewLI->getContext(), Alignment));
-  State.addNewMetadata(NewLI, LI);
   applyMetadata(NewLI);
   Instruction *Res = NewLI;
   if (isReverse())
@@ -2864,8 +2858,6 @@ void VPWidenLoadEVLRecipe::print(raw_ostream &O, const Twine &Indent,
 #endif
 
 void VPWidenStoreRecipe::execute(VPTransformState &State) {
-  auto *SI = cast<StoreInst>(&Ingredient);
-
   VPValue *StoredVPValue = getStoredValue();
   bool CreateScatter = !isConsecutive();
   const Align Alignment = getLoadStoreAlignment(&Ingredient);
@@ -2897,7 +2889,6 @@ void VPWidenStoreRecipe::execute(VPTransformState &State) {
     NewSI = Builder.CreateMaskedStore(StoredVal, Addr, Alignment, Mask);
   else
     NewSI = Builder.CreateAlignedStore(StoredVal, Addr, Alignment);
-  State.addNewMetadata(NewSI, SI);
   applyMetadata(NewSI);
 }
 
@@ -2910,8 +2901,6 @@ void VPWidenStoreRecipe::print(raw_ostream &O, const Twine &Indent,
 #endif
 
 void VPWidenStoreEVLRecipe::execute(VPTransformState &State) {
-  auto *SI = cast<StoreInst>(&Ingredient);
-
   VPValue *StoredValue = getStoredValue();
   bool CreateScatter = !isConsecutive();
   const Align Alignment = getLoadStoreAlignment(&Ingredient);
@@ -2945,7 +2934,6 @@ void VPWidenStoreEVLRecipe::execute(VPTransformState &State) {
   }
   NewSI->addParamAttr(
       1, Attribute::getWithAlignment(NewSI->getContext(), Alignment));
-  State.addNewMetadata(NewSI, SI);
   applyMetadata(NewSI);
 }
 
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 445bcac5919fc..9b6d4f24f735d 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -70,7 +70,8 @@ bool VPlanTransforms::tryToConvertVPInstructionsToVPRecipes(
         NewRecipe = new VPWidenIntOrFpInductionRecipe(
             Phi, Start, Step, &Plan->getVF(), *II, Ingredient.getDebugLoc());
       } else {
-        auto Metadata = VPRecipeBuilder::getMetadataToPropagate(Inst);
+        SmallVector<std::pair<unsigned, MDNode *>> Metadata;
+        ::getMetadataToPropagate(Inst, Metadata);
         assert(isa<VPInstruction>(&Ingredient) &&
                "only VPInstructions expected here");
         assert(!isa<PHINode>(Inst) && "phis should be handled above");
@@ -177,7 +178,9 @@ static bool sinkScalarOperands(VPlan &Plan) {
       if (ScalarVFOnly)
         continue;
       Instruction *I = SinkCandidate->getUnderlyingInstr();
-      auto *Clone = new VPReplicateRecipe(I, SinkCandidate->operands(), true);
+      auto *Clone = new VPReplicateRecipe(
+          I, SinkCandidate->operands(), true, /*Mask*/ nullptr,
+          cast<VPReplicateRecipe>(SinkCandidate)->getMetadata());
       // TODO: add ".cloned" suffix to name of Clone's VPValue.
 
       Clone->insertBefore(SinkCandidate);
@@ -336,7 +339,7 @@ static VPRegionBlock *createReplicateRegion(VPReplicateRecipe *PredRecipe,
   auto *RecipeWithoutMask = new VPReplicateRecipe(
       PredRecipe->getUnderlyingInstr(),
       make_range(PredRecipe->op_begin(), std::prev(PredRecipe->op_end())),
-      PredRecipe->isUniform());
+      PredRecipe->isUniform(), /*Mask*/ nullptr, PredRecipe->getMetadata());
   auto *Pred =
       Plan.createVPBasicBlock(Twine(RegionName) + ".if", RecipeWithoutMask);
 
@@ -2768,7 +2771,8 @@ void VPlanTransforms::narrowInterleaveGroups(VPlan &Plan, ElementCount VF,
     // Narrow wide load to uniform scalar load, as transformed VPlan will only
     // process one original iteration.
     auto *N = new VPReplicateRecipe(&WideLoad->getIngredient(),
-                                    WideLoad->operands(), /*IsUniform*/ true);
+                                    WideLoad->operands(), /*IsUniform*/ true,
+                                    /*Mask*/ nullptr, WideLoad->getMetadata());
     N->insertBefore(WideLoad);
     return N;
   };
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll b/llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll
index 9b6a1686eee6e..15c5258b57cc9 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll
@@ -188,17 +188,17 @@ define void @trunc_store(ptr %dst, ptr %src, i16 %x) #1 {
 ; DEFAULT-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; DEFAULT:       vec.epilog.vector.body:
 ; DEFAULT-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; DEFAULT-NEXT:    [[TMP16:%.*]] = load i64, ptr [[SRC]], align 8, !alias.scope [[META11:![0-9]+]]
+; DEFAULT-NEXT:    [[TMP16:%.*]] = load i64, ptr [[SRC]], align 8, !alias.scope [[META5]]
 ; DEFAULT-NEXT:    [[BROADCAST_SPLATINSERT7:%.*]] = insertelement <8 x i64> poison, i64 [[TMP16]], i64 0
 ; DEFAULT-NEXT:    [[BROADCAST_SPLAT8:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT7]], <8 x i64> poison, <8 x i32> zeroinitializer
 ; DEFAULT-NEXT:    [[TMP18:%.*]] = trunc <8 x i64> [[BROADCAST_SPLAT8]] to <8 x i8>
 ; DEFAULT-NEXT:    [[TMP14:%.*]] = and <8 x i8> [[TMP18]], [[TMP15]]
 ; DEFAULT-NEXT:    [[TMP26:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX5]]
 ; DEFAULT-NEXT:    [[TMP27:%.*]] = getelementptr i8, ptr [[TMP26]], i32 0
-; DEFAULT-NEXT:    store <8 x i8> [[TMP14]], ptr [[TMP27]], align 1, !alias.scope [[META14:![0-9]+]], !noalias [[META11]]
+; DEFAULT-NEXT:    store <8 x i8> [[TMP14]], ptr [[TMP27]], align 1, !alias.scope [[META8]], !noalias [[META5]]
 ; DEFAULT-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX5]], 8
 ; DEFAULT-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT8]], 1000
-; DEFAULT-NEXT:    br i1 [[TMP17]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; DEFAULT-NEXT:    br i1 [[TMP17]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; DEFAULT:       vec.epilog.middle.block:
 ; DEFAULT-NEXT:    br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
 ; DEFAULT:       vec.epilog.scalar.ph:
@@ -214,7 +214,7 @@ define void @trunc_store(ptr %dst, ptr %src, i16 %x) #1 {
 ; DEFAULT-NEXT:    store i8 [[TRUNC]], ptr [[GEP]], align 1
 ; DEFAULT-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; DEFAULT-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
-; DEFAULT-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP17:![0-9]+]]
+; DEFAULT-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP12:![0-9]+]]
 ; DEFAULT:       exit:
 ; DEFAULT-NEXT:    ret void
 ;
@@ -313,13 +313,8 @@ attributes #1 = { vscale_range(1,16) "target-features"="+sve" }
 ; DEFAULT: [[META8]] = !{[[META9:![0-9]+]]}
 ; DEFAULT: [[META9]] = distinct !{[[META9]], [[META7]]}
 ; DEFAULT: [[LOOP10]] = distinct !{[[LOOP10]], [[META1]], [[META2]]}
-; DEFAULT: [[META11]] = !{[[META12:![0-9]+]]}
-; DEFAULT: [[META12]] = distinct !{[[META12]], [[META13:![0-9]+]]}
-; DEFAULT: [[META13]] = distinct !{[[META13]], !"LVerDomain"}
-; DEFAULT: [[META14]] = !{[[META15:![0-9]+]]}
-; DEFAULT: [[META15]] = distinct !{[[META15]], [[META13]]}
-; DEFAULT: [[LOOP16]] = distinct !{[[LOOP16]], [[META1]], [[META2]]}
-; DEFAULT: [[LOOP17]] = distinct !{[[LOOP17]], [[META1]]}
+; DEFAULT: [[LOOP11]] = distinct !{[[LOOP11]], [[META1]], [[META2]]}
+; DEFAULT: [[LOOP12]] = distinct !{[[LOOP12]], [[META1]]}
 ;.
 ; PRED: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
 ; PRED: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}



More information about the llvm-commits mailing list