[VPlan] Introduce recipe to build scalar steps.

[fhahn]

Summary

This patch adds a new VPScalarIVStepsRecipe to handle building scalar
steps.

In the first patch, it only handles the case where there is no vector
induction variable needed.

[fhahn]

rebase on top of D116123

[fhahn]

rebased after moving SCEV expansion to extra recipe D116288

[Ayal]

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

561

OldInduction moved from protected to public (in some version?) - suffice to provide a const public method to retrieve OldInduction?

2712

Can be folded with above call to createVectorIntOrFpInductionPHI(), followed by

if (NeedsScalarIV) {
  Value *ScalarIV = CreateScalarIV(Step);
  buildScalarSteps(ScalarIV, Step, EntryVal, ID, Def, State);
}

?

7920

OldInduction already set at State construction above?

Also feed Vector/TripCounts to State constructor (indep. of this patch)?

8407

Update comment above, as block in mask may no longer be the first non-phi instruction in the block?
This suggests that VPScalarIVStepsRecipes should perhaps be introduced (as first non-phi insns) by a VPlan-to-VPlan transformation after an initial VPlan is formed with ICmpULE/ActiveLaneMask (as initial first non-phi insns) and other 'initial' recipes.

8922

early-exit if !OnlyScalarStepsNeeded()

8931

Pass StepSCEV instead of II->getSteps()?

This VPExpandSCEVRecipe needs to be placed somewhere by every caller of tryToBuildScalarSteps.

8936

Place ScalarStepsForI as last parameter? Can always pass it and have the constructor check if it's equal to Phi or else should be considered a Trunc.

9078

VectorIVs and ScalarIVs seem useless?

9079

MoveAfterPhis and InductionsToMove suggest that buildVPlanWithVPRecipes() is trying to do too much too early.
Both represent recipes that could not be accurately placed when constructed, and are later (re)positioned between last phi and first non-phi: InductionsToMove are trunc-folded-phi's that must first be placed at the position of the trunc before moving to appear among phi's, and MoveAfterPhis are VPExpandSCEVRecipes (that complement VPScalarIVStepsRecipes) which have yet to be placed - after all phi's but before other non-phi's (right? Some documentation would be helpful.)

Could buildVPlanWithVPRecipes() be simplified by having

• a Trunc(-that-can-be-folded-into-phi) recipe placed according to the Trunc, with a VPlan2VPlan transformation that traverses such Truncs and replaces them with a Phi recipe
• a Phi(-that-can-be-converted-into-scalar-steps) recipe placed according to the Phi, with a VPlan2VPlan transformation that traverses such Phis and places them with pairs of VPScalarIVStepsRecipes and VPExpandSCEVRecipes
• similar to WidenMemory recipes placed according to Load/Stores, later replaced by InterleaveGroup recipes

9105

Rather than pass Instr twice, pass it once as the "Phi" and null as the potential Trunc?

9111

This attempt to build a VPScalarIVStepsRecipe given a Phi Instr should be placed inside tryToCreateWidenRecipe() along with its general handling of a Phi Instrs, rather be treated here. The fact that tryToBuildScalarSteps() generates two recipes, returning one (VPScalarIVStepsRecipes) which uses the other (VPExpandSCEVRecipe), where only a single recipe is traditionally created and returned per each Instr, can be dealt with later. In any case the actual placement of both VPScalarIVStepsRecipe
and VPExpandSCEVRecipe takes place later.

Having both "Steps" and "StepR" may be confusing.

9122

Why is this setRecipe moved?

9138

(This records both VPExpandSCEVRecipe and its VPScalarIVStepsRecipe created by tryToCreateWidenRecipe() for Trunc Instr., to be placed later.)

9733

Store per-part 0?

9735

ditto

10201

Update comment

llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h

174 ↗

(On Diff #396234)

Document?

176 ↗

(On Diff #396234)

First parameter is always a Phi; second can be a potentially null Trunc, rather than Trunc-if-exists-otherwise-Phi?

llvm/lib/Transforms/Vectorize/VPlan.cpp

645–646

assert Parent is null, as in the above insertBefore(InsertPos)?
Place next to insertBefore(InsertPos) above, so that both moveAfter and moveBefore appear next to each other?
There are in general 8 combinations: [ move | insert ][ After | Before ][ Recipe | {VPBB, I} ]

llvm/lib/Transforms/Vectorize/VPlan.h

703

(Here the insertBefores appear next to each other.)

1115

"their [vector and] scalar values"?

1128

Suffice to have a single constructor possibly with Trunc as its last optional parameter defaulted to null. Only distinction is the underlying value of VPValue which is set to (Trunc ? Trunc : IV).
Separate pointer to Trunc is redundant, can instead check if the underlying value is a Trunc.

1139

"[vectorized and]"

llvm/lib/Transforms/Vectorize/VPlanValue.h

342

Lex order

[fhahn]

Completely restructured the patch to introduce VPScalarIVStepsRecipe in VPlan2VPlan transformation, similiar to how VPWidenCanonicalIV is handled.

The code should be much simpler now. I also tried to address the existing comments, if they still applied to the updated patch.

[Ayal]

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

2701

By dropping "if (Def->needsVectorIV())" we create a vector IV even if one is (known at this point to) not be needed - all users want scalar values. This case will be cleaned up later, right?

8530

Is this change still needed or can tryToOptimizeInductionTruncate() continue to return VPWidenIntOrFpInductionRecipe?

9732

This is to run once, i.e., not per part/lane inside a replicating region - assert(!State.Instance && "..."); as above.

llvm/lib/Transforms/Vectorize/VPlan.cpp

1287

Add accessors instead of getOperand(0/1/2)?

Comment that this checks if start value is zero, aka the start of the canonical IV.

llvm/lib/Transforms/Vectorize/VPlan.h

703

Deserves a separate/additional comment?

1379

\p CanonicalIV produces a canonical scalar value per part, this recipe complements it by producing the possibly skewed scalar values per lane.

llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp

387

Have both early-exits here doing if (!IV || IV->needsVectorIV()) continue; ?

410

nit: pass IV->getPHINode() and IV->getTruncInst() instead of PhiI and TruncI? This is essentially transforming the VPWidenIntOrFpInductionRecipe IV into a VPScalarIVStepsRecipe, along with a specific Step.

415

TODO: Step should be placed in preheader?

llvm/lib/Transforms/Vectorize/VPlanTransforms.h

59

If all users of vector IV need scalar values, provide them by building scalar steps off of the canonical scalar IV, and remove the vector IV.

[fhahn]

Address latest comments, thanks!

[fhahn]

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

561

OldInduction is not needed in the latest version.

2701

Updated the comment, thanks!

2712

Folded, thanks!

7920

removed in the latest version

8407

Removed in the latest version

8530

It's a leftover from earlier iterations, it is not needed. I removed it, thanks!

8922

The whole function is gone now.

8931

The whole function is gone now.

8936

The whole function is gone now.

9078

Those changes are gone now.

9079

Those changes are gone now.

9105

Those changes are gone now.

9111

Those changes are gone now.

9122

changes gone in latest iteration

9138

changes gone in latest iteration

9732

Added assert, thanks!

llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h

176 ↗

(On Diff #396234)

This is gone in the latest version.

llvm/lib/Transforms/Vectorize/VPlan.cpp

645–646

Done!

1287

Added accessors and comment, thanks.

llvm/lib/Transforms/Vectorize/VPlan.h

1115

Removed, thanks!

1128

Simplified, thanks!

1139

Removed, thanks!

llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp

387

Moved, thanks!

410

Done, thanks!

415

Added the TODO, thanks!

llvm/lib/Transforms/Vectorize/VPlanTransforms.h

59

Much better description, updated!

llvm/lib/Transforms/Vectorize/VPlanValue.h

342

Moved, thanks!

[Ayal]

Looks good to me, thanks!

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

9236

leave this w/o change?

llvm/lib/Transforms/Vectorize/VPlan.h

703

Above comment refers to specified recipe (\p InsertPos); refer also to the \p BB and \p I interface?

1378

Define VPScalarIVStepsRecipe later to save this forward declaration of VPCanonicalIVPHIRecipe? Would also facilitate inlining of getCanonicalIV().

[fhahn]

Thanks Ayal! Latest comments should be addressed and I am planning on landing this soon.

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

public:
   /// loop.
   void addMetadata(Instruction *To, Instruction *From);
 
   /// Similar to the previous function but it adds the metadata to a
   /// vector of instructions.
   void addMetadata(ArrayRef<Value *> To, Instruction *From);
 
   // Returns the resume value (bc.merge.rdx) for a reduction as
   // generated by fixReduction.
   PHINode *getReductionResumeValue(const RecurrenceDescriptor &RdxDesc);
 
 protected:
   friend class LoopVectorizationPlanner;
 
   /// A small list of PHINodes.
   using PhiVector = SmallVector<PHINode *, 4>;
 

 void InnerLoopVectorizer::widenIntOrFpInduction(
     PHINode *IV, VPWidenIntOrFpInductionRecipe *Def, VPTransformState &State,
     Value *CanonicalIV) {
   Value *Start = Def->getStartValue()->getLiveInIRValue();
   const InductionDescriptor &ID = Def->getInductionDescriptor();
   TruncInst *Trunc = Def->getTruncInst();
   IRBuilderBase &Builder = State.Builder;
   assert(IV->getType() == ID.getStartValue()->getType() && "Types must match");
-  assert(!State.VF.isZero() && "VF must be non-zero");
+  assert(State.VF.isVector() && "must have vector VF");
 
   // The value from the original loop to which we are mapping the new induction
   // variable.
   Instruction *EntryVal = Trunc ? cast<Instruction>(Trunc) : IV;
 
   auto &DL = EntryVal->getModule()->getDataLayout();
 
   // Generate code for the induction step. Note that induction steps are
void InnerLoopVectorizer::widenIntOrFpInduction(
 
   // Fast-math-flags propagate from the original induction instruction.
   IRBuilder<>::FastMathFlagGuard FMFG(Builder);
   if (ID.getInductionBinOp() && isa<FPMathOperator>(ID.getInductionBinOp()))
     Builder.setFastMathFlags(ID.getInductionBinOp()->getFastMathFlags());
 
   // Now do the actual transformations, and start with creating the step value.
   Value *Step = CreateStepValue(ID.getStep());
-  if (State.VF.isScalar()) {
-    Value *ScalarIV = CreateScalarIV(Step);
-    Type *ScalarTy = IntegerType::get(ScalarIV->getContext(),
-                                      Step->getType()->getScalarSizeInBits());
-
-    Instruction::BinaryOps IncOp = ID.getInductionOpcode();
-    if (IncOp == Instruction::BinaryOpsEnd)
-      IncOp = Instruction::Add;
-    for (unsigned Part = 0; Part < UF; ++Part) {
-      Value *StartIdx = ConstantInt::get(ScalarTy, Part);
-      Instruction::BinaryOps MulOp = Instruction::Mul;
-      if (Step->getType()->isFloatingPointTy()) {
-        StartIdx = Builder.CreateUIToFP(StartIdx, Step->getType());
-        MulOp = Instruction::FMul;
-      }
-
-      Value *Mul = Builder.CreateBinOp(MulOp, StartIdx, Step);
-      Value *EntryPart = Builder.CreateBinOp(IncOp, ScalarIV, Mul, "induction");
-      State.set(Def, EntryPart, Part);
-      if (Trunc) {
-        assert(!Step->getType()->isFloatingPointTy() &&
-               "fp inductions shouldn't be truncated");
-        addMetadata(EntryPart, Trunc);
-      }
-    }
-    return;
-  }
 
-  // Create a new independent vector induction variable, if one is needed.
-  if (Def->needsVectorIV())
-    createVectorIntOrFpInductionPHI(ID, Step, Start, EntryVal, Def, State);
+  // Create a new independent vector induction variable. Later VPlan2VPlan
+  // optimizations will remove it, if it won't be needed, e.g. because all users
+  // of it access scalar values.
 
   if (Def->needsScalarIV()) {
     // Create scalar steps that can be used by instructions we will later
     // scalarize. Note that the addition of the scalar steps will not increase
     // the number of instructions in the loop in the common case prior to
     // InstCombine. We will be trading one vector extract for each scalar step.
     Value *ScalarIV = CreateScalarIV(Step);
     buildScalarSteps(ScalarIV, Step, EntryVal, ID, Def, State);
   }
 }
 
 void InnerLoopVectorizer::packScalarIntoVectorValue(VPValue *Def,
                                                     const VPIteration &Instance,
                                                     VPTransformState &State) {
   Value *ScalarInst = State.get(Def, Instance);
   Value *VectorValue = State.get(Def, Instance.Part);
   VectorValue = Builder.CreateInsertElement(
       VectorValue, ScalarInst,
void LoopVectorizationPlanner::executePlan(ElementCount BestVF, unsigned BestUF,
   LLVM_DEBUG(dbgs() << "Executing best plan with VF=" << BestVF << ", UF=" << BestUF
                     << '\n');
 
   // Perform the actual loop transformation.
 
   // 1. Create a new empty loop. Unlink the old loop and connect the new one.
   VPTransformState State{BestVF, BestUF, LI, DT, ILV.Builder, &ILV, &BestVPlan};
   Value *CanonicalIVStartValue;
   std::tie(State.CFG.PrevBB, CanonicalIVStartValue) =
       ILV.createVectorizedLoopSkeleton();
   ILV.collectPoisonGeneratingRecipes(State);
 
   ILV.printDebugTracesAtStart();
 
   //===------------------------------------------------===//
   //
   // Notice: any optimization or new instruction that go
if (OrigLoop->getHeader() == BB) {
     auto NewInsertionPoint = HeaderVPBB->getFirstNonPhi();
     auto *IV = new VPWidenCanonicalIVRecipe(Plan->getCanonicalIV());
     HeaderVPBB->insert(IV, HeaderVPBB->getFirstNonPhi());
 
     VPBuilder::InsertPointGuard Guard(Builder);
     Builder.setInsertPoint(HeaderVPBB, NewInsertionPoint);
     if (CM.TTI.emitGetActiveLaneMask()) {
       VPValue *TC = Plan->getOrCreateTripCount();
       BlockMask = Builder.createNaryOp(VPInstruction::ActiveLaneMask, {IV, TC});
     } else {
       VPValue *BTC = Plan->getOrCreateBackedgeTakenCount();
       BlockMask = Builder.createNaryOp(VPInstruction::ICmpULE, {IV, BTC});
     }
     return BlockMaskCache[BB] = BlockMask;
   }
 
   // This is the block mask. We OR all incoming edges.
VPWidenIntOrFpInductionRecipe *VPRecipeBuilder::tryToOptimizeInductionPHI(
   // produces its scalar and vector values.
   if (auto *II = Legal->getIntOrFpInductionDescriptor(Phi))
     return createWidenInductionRecipe(Phi, Phi, Operands[0], *II, CM, *OrigLoop,
                                       Range);
 
   return nullptr;
 }
 
 VPWidenIntOrFpInductionRecipe *VPRecipeBuilder::tryToOptimizeInductionTruncate(
     TruncInst *I, ArrayRef<VPValue *> Operands, VFRange &Range,
     VPlan &Plan) const {
   // Optimize the special case where the source is a constant integer
   // induction variable. Notice that we can only optimize the 'trunc' case
   // because (a) FP conversions lose precision, (b) sext/zext may wrap, and
   // (c) other casts depend on pointer size.
 
   // Determine whether \p K is a truncation based on an induction variable that
void LoopVectorizationPlanner::buildVPlansWithVPRecipes(ElementCount MinVF,
 
   // Collect instructions from the original loop that will become trivially dead
   // in the vectorized loop. We don't need to vectorize these instructions. For
   // example, original induction update instructions can become dead because we
   // separately emit induction "steps" when generating code for the new loop.
   // Similarly, we create a new latch condition when setting up the structure
   // of the new loop, so the old one can become dead.
   SmallPtrSet<Instruction *, 4> DeadInstructions;
   collectTriviallyDeadInstructions(DeadInstructions);
 
   // Add assume instructions we need to drop to DeadInstructions, to prevent
   // them from being added to the VPlan.
   // TODO: We only need to drop assumes in blocks that get flattend. If the
   // control flow is preserved, we should keep them.
   auto &ConditionalAssumes = Legal->getConditionalAssumes();
   DeadInstructions.insert(ConditionalAssumes.begin(), ConditionalAssumes.end());
 
   MapVector<Instruction *, Instruction *> &SinkAfter = Legal->getSinkAfter();
   // Dead instructions do not need sinking. Remove them from SinkAfter.
   for (Instruction *I : DeadInstructions)
     SinkAfter.erase(I);
 
   // Cannot sink instructions after dead instructions (there won't be any
   // recipes for them). Instead, find the first non-dead previous instruction.
   for (auto &P : Legal->getSinkAfter()) {
     Instruction *SinkTarget = P.second;
     Instruction *FirstInst = &*SinkTarget->getParent()->begin();
     (void)FirstInst;
     while (DeadInstructions.contains(SinkTarget)) {
       assert(
           SinkTarget != FirstInst &&
VPlanPtr LoopVectorizationPlanner::buildVPlanWithVPRecipes(
   // Scan the body of the loop in a topological order to visit each basic block
   // after having visited its predecessor basic blocks.
   LoopBlocksDFS DFS(OrigLoop);
   DFS.perform(LI);
 
   VPBasicBlock *VPBB = HeaderVPBB;
   SmallVector<VPWidenIntOrFpInductionRecipe *> InductionsToMove;
   for (BasicBlock *BB : make_range(DFS.beginRPO(), DFS.endRPO())) {
     // Relevant instructions from basic block BB will be grouped into VPRecipe
     // ingredients and fill a new VPBasicBlock.
     unsigned VPBBsForBB = 0;
     VPBB->setName(BB->getName());
     Builder.setInsertPoint(VPBB);
 
     // Introduce each ingredient into VPlan.
     // TODO: Model and preserve debug instrinsics in VPlan.
     for (Instruction &I : BB->instructionsWithoutDebug()) {
       Instruction *Instr = &I;
for (Instruction &I : BB->instructionsWithoutDebug()) {
         Operands.push_back(Plan->getOrAddVPValue(
             Phi->getIncomingValueForBlock(OrigLoop->getLoopPreheader())));
       } else {
         auto OpRange = Plan->mapToVPValues(Instr->operands());
         Operands = {OpRange.begin(), OpRange.end()};
       }
       if (auto RecipeOrValue = RecipeBuilder.tryToCreateWidenRecipe(
               Instr, Operands, Range, Plan)) {
         // If Instr can be simplified to an existing VPValue, use it.
         if (RecipeOrValue.is<VPValue *>()) {
           auto *VPV = RecipeOrValue.get<VPValue *>();
           Plan->addVPValue(Instr, VPV);
           // If the re-used value is a recipe, register the recipe for the
           // instruction, in case the recipe for Instr needs to be recorded.
           if (auto *R = dyn_cast_or_null<VPRecipeBase>(VPV->getDef()))
             RecipeBuilder.setRecipe(Instr, R);
           continue;
         }
         // Otherwise, add the new recipe.
         VPRecipeBase *Recipe = RecipeOrValue.get<VPRecipeBase *>();
         for (auto *Def : Recipe->definedValues()) {
           auto *UV = Def->getUnderlyingValue();
           Plan->addVPValue(UV, Def);
         }
 
         if (isa<VPWidenIntOrFpInductionRecipe>(Recipe) &&
             HeaderVPBB->getFirstNonPhi() != VPBB->end()) {
           // Keep track of VPWidenIntOrFpInductionRecipes not in the phi section
           // of the header block. That can happen for truncates of induction
           // variables. Those recipes are moved to the phi section of the header
           // block after applying SinkAfter, which relies on the original
           // position of the trunc.
           assert(isa<TruncInst>(Instr));
           InductionsToMove.push_back(
               cast<VPWidenIntOrFpInductionRecipe>(Recipe));
         }
         RecipeBuilder.setRecipe(Instr, Recipe);
         VPBB->appendRecipe(Recipe);
         continue;
       }
 
       // Otherwise, if all widening options failed, Instruction is to be
       // replicated. This may create a successor for VPBB.
       VPBasicBlock *NextVPBB =
           RecipeBuilder.handleReplication(Instr, Range, VPBB, Plan);
       if (NextVPBB != VPBB) {
         VPBB = NextVPBB;
         VPBB->setName(BB->hasName() ? BB->getName() + "." + Twine(VPBBsForBB++)
                                     : "");
       }
for (auto &Entry : SinkAfter) {
     }
   }
 
   VPlanTransforms::removeRedundantCanonicalIVs(*Plan);
   VPlanTransforms::removeRedundantInductionCasts(*Plan);
 
   // Now that sink-after is done, move induction recipes for optimized truncates
   // to the phi section of the header block.
   for (VPWidenIntOrFpInductionRecipe *Ind : InductionsToMove)
     Ind->moveBefore(*HeaderVPBB, HeaderVPBB->getFirstNonPhi());
 
   // Adjust the recipes for any inloop reductions.
   adjustRecipesForReductions(cast<VPBasicBlock>(TopRegion->getExit()), Plan,
                              RecipeBuilder, Range.Start);
 
   // Introduce a recipe to combine the incoming and previous values of a
   // first-order recurrence.
for (unsigned i = 0; i < IG->getFactor(); ++i)
         RecipeBuilder.getRecipe(Member)->eraseFromParent();
       }
   }
 
   // From this point onwards, VPlan-to-VPlan transformations may change the plan
   // in ways that accessing values using original IR values is incorrect.
   Plan->disableValue2VPValue();
 
+  VPlanTransforms::optimizeInductions(*Plan, *PSE.getSE());
   VPlanTransforms::sinkScalarOperands(*Plan);
   VPlanTransforms::mergeReplicateRegions(*Plan);
   VPlanTransforms::removeDeadRecipes(*Plan, *OrigLoop);
 
   std::string PlanName;
   raw_string_ostream RSO(PlanName);
   ElementCount VF = Range.Start;
   Plan->addVF(VF);

 }
 
 void VPWidenIntOrFpInductionRecipe::execute(VPTransformState &State) {
   assert(!State.Instance && "Int or FP induction being replicated.");
   auto *CanonicalIV = State.get(getParent()->getPlan()->getCanonicalIV(), 0);
   State.ILV->widenIntOrFpInduction(IV, this, State, CanonicalIV);
 }
 
+void VPScalarIVStepsRecipe::execute(VPTransformState &State) {
+  assert(!State.Instance && "VPScalarIVStepsRecipe being replicated.");
+
+  // Fast-math-flags propagate from the original induction instruction.
+  IRBuilder<>::FastMathFlagGuard FMFG(State.Builder);
+  if (IndDesc.getInductionBinOp() &&
+      isa<FPMathOperator>(IndDesc.getInductionBinOp()))
+    State.Builder.setFastMathFlags(
+        IndDesc.getInductionBinOp()->getFastMathFlags());
+
+  Value *Step = State.get(getStepValue(), VPIteration(0, 0));
+  auto *Trunc = dyn_cast<TruncInst>(getUnderlyingValue());
+  auto CreateScalarIV = [&](Value *&Step) -> Value * {
+    Value *ScalarIV = State.get(getCanonicalIV(), VPIteration(0, 0));
+    auto *CanonicalIV = State.get(getParent()->getPlan()->getCanonicalIV(), 0);
+    if (!isCanonical() || CanonicalIV->getType() != IV->getType()) {
+      ScalarIV = IV->getType()->isIntegerTy()
+                     ? State.Builder.CreateSExtOrTrunc(ScalarIV, IV->getType())
+                     : State.Builder.CreateCast(Instruction::SIToFP, ScalarIV,
+                                                IV->getType());
+      ScalarIV = emitTransformedIndex(State.Builder, ScalarIV,
+                                      getStartValue()->getLiveInIRValue(), Step,
+                                      IndDesc);
+      ScalarIV->setName("offset.idx");
+    }
+    if (Trunc) {
+      auto *TruncType = cast<IntegerType>(Trunc->getType());
+      assert(Step->getType()->isIntegerTy() &&
+             "Truncation requires an integer step");
+      ScalarIV = State.Builder.CreateTrunc(ScalarIV, TruncType);
+      Step = State.Builder.CreateTrunc(Step, TruncType);
+    }
+    return ScalarIV;
+  };
+
+  Value *ScalarIV = CreateScalarIV(Step);
+  if (State.VF.isVector()) {
+    buildScalarSteps(ScalarIV, Step, IV, IndDesc, this, State);
+    return;
+  }
+
+  for (unsigned Part = 0; Part < State.UF; ++Part) {
+    assert(!State.VF.isScalable() && "scalable vectors not yet supported.");
+    Value *EntryPart;
+    if (Step->getType()->isFloatingPointTy()) {
+      Value *StartIdx =
+          getRuntimeVFAsFloat(State.Builder, Step->getType(), State.VF * Part);
+      // Floating-point operations inherit FMF via the builder's flags.
+      Value *MulOp = State.Builder.CreateFMul(StartIdx, Step);
+      EntryPart = State.Builder.CreateBinOp(IndDesc.getInductionOpcode(),
+                                            ScalarIV, MulOp);
+    } else {
+      Value *StartIdx =
+          getRuntimeVF(State.Builder, Step->getType(), State.VF * Part);
+      EntryPart = State.Builder.CreateAdd(
+          ScalarIV, State.Builder.CreateMul(StartIdx, Step), "induction");
+    }
+    State.set(this, EntryPart, Part);
+    if (Trunc)
+      State.ILV->addMetadata(EntryPart, Trunc);
+  }
+}
+
 void VPWidenPHIRecipe::execute(VPTransformState &State) {
   State.ILV->widenPHIInstruction(cast<PHINode>(getUnderlyingValue()), this,
                                  State);
 }
 
 void VPBlendRecipe::execute(VPTransformState &State) {
   State.ILV->setDebugLocFromInst(Phi, &State.Builder);
   // We know that all PHIs in non-header blocks are converted into
Value *VPTransformState::get(VPValue *Def, unsigned Part) {
   }
 
   auto *RepR = dyn_cast<VPReplicateRecipe>(Def);
   bool IsUniform = RepR && RepR->isUniform();
 
   unsigned LastLane = IsUniform ? 0 : VF.getKnownMinValue() - 1;
   // Check if there is a scalar value for the selected lane.
   if (!hasScalarValue(Def, {Part, LastLane})) {
     // At the moment, VPWidenIntOrFpInductionRecipes can also be uniform.
-    assert(isa<VPWidenIntOrFpInductionRecipe>(Def->getDef()) &&
+    assert((isa<VPWidenIntOrFpInductionRecipe>(Def->getDef()) ||
+            isa<VPScalarIVStepsRecipe>(Def->getDef())) &&
            "unexpected recipe found to be invariant");
     IsUniform = true;
     LastLane = 0;
   }
 
   auto *LastInst = cast<Instruction>(get(Def, {Part, LastLane}));
   // Set the insert point after the last scalarized instruction or after the
   // last PHI, if LastInst is a PHI. This ensures the insertelement sequence

llvm/lib/Transforms/Vectorize/VPlan.h

public:
 
   /// The method which generates the output IR instructions that correspond to
   /// this VPRecipe, thereby "executing" the VPlan.
   virtual void execute(struct VPTransformState &State) = 0;
 
   /// Insert an unlinked recipe into a basic block immediately before
   /// the specified recipe.
   void insertBefore(VPRecipeBase *InsertPos);
+  /// Insert an unlinked recipe into \p BB immediately before the insertion
+  /// point \p IP;
+  void insertBefore(VPBasicBlock &BB, iplist<VPRecipeBase>::iterator IP);
 
   /// Insert an unlinked Recipe into a basic block immediately after
   /// the specified Recipe.
   void insertAfter(VPRecipeBase *InsertPos);
 
   /// Unlink this recipe from its current VPBasicBlock and insert it into
   /// the VPBasicBlock that MovePos lives in, right after MovePos.
   void moveAfter(VPRecipeBase *MovePos);
#endif
   /// otherwise.
   TruncInst *getTruncInst() {
     return dyn_cast_or_null<TruncInst>(getVPValue(0)->getUnderlyingValue());
   }
   const TruncInst *getTruncInst() const {
     return dyn_cast_or_null<TruncInst>(getVPValue(0)->getUnderlyingValue());
   }
 
+  PHINode *getPHINode() { return IV; }
+
   /// Returns the induction descriptor for the recipe.
   const InductionDescriptor &getInductionDescriptor() const { return IndDesc; }
 
   /// Returns true if the induction is canonical, i.e. starting at 0 and
   /// incremented by UF * VF (= the original IV is incremented by 1).
   bool isCanonical() const;
 
   /// Returns the scalar type of the induction.
   const Type *getScalarType() const {
     const TruncInst *TruncI = getTruncInst();
     return TruncI ? TruncI->getType() : IV->getType();
   }
 
   /// Returns true if a scalar phi needs to be created for the induction.
   bool needsScalarIV() const { return NeedsScalarIV; }
 
   /// Returns true if a vector phi needs to be created for the induction.
   bool needsVectorIV() const { return NeedsVectorIV; }
 };
 
 /// A pure virtual base class for all recipes modeling header phis, including
 /// phis for first order recurrences, pointer inductions and reductions. The
 /// start value is the first operand of the recipe and the incoming value from
 /// the backedge is the second operand.
 class VPHeaderPHIRecipe : public VPRecipeBase, public VPValue {
 protected:
   VPHeaderPHIRecipe(unsigned char VPVID, unsigned char VPDefID, PHINode *Phi,
                     VPValue *Start = nullptr)
       : VPRecipeBase(VPDefID, {}), VPValue(VPVID, Phi, this) {
     if (Start)
       addOperand(Start);
   }
 
 public:
   ~VPHeaderPHIRecipe() override = default;
 
   /// Method to support type inquiry through isa, cast, and dyn_cast.
bool onlyFirstLaneUsed(const VPValue *Op) const override {
     // Recursing through Blend recipes only, must terminate at header phi's the
     // latest.
     return all_of(users(), [this](VPUser *U) {
       return cast<VPRecipeBase>(U)->onlyFirstLaneUsed(this);
     });
   }
 };
 
 /// VPInterleaveRecipe is a recipe for transforming an interleave group of load
 /// or stores into one wide load/store and shuffles. The first operand of a
 /// VPInterleave recipe is the address, followed by the stored values, followed
 /// by an optional mask.
 class VPInterleaveRecipe : public VPRecipeBase {
   const InterleaveGroup<Instruction> *IG;
 
   bool HasMask = false;
 
 public:
VPCanonicalIVPHIRecipe(VPValue *StartV, DebugLoc DL)
         DL(DL) {}
 
   ~VPCanonicalIVPHIRecipe() override = default;
 
   /// Method to support type inquiry through isa, cast, and dyn_cast.
   static inline bool classof(const VPDef *D) {
     return D->getVPDefID() == VPCanonicalIVPHISC;
   }
+  static inline bool classof(const VPHeaderPHIRecipe *D) {
+    return D->getVPDefID() == VPCanonicalIVPHISC;
+  }
+  static inline bool classof(const VPValue *V) {
+    return V->getVPValueID() == VPValue::VPVCanonicalIVPHISC;
+  }
 
   /// Generate the canonical scalar induction phi of the vector loop.
   void execute(VPTransformState &State) override;
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
   /// Print the recipe.
   void print(raw_ostream &O, const Twine &Indent,
              VPSlotTracker &SlotTracker) const override;
#endif
 
   /// Returns the scalar type of the induction.
   const Type *getScalarType() const {
     return cast<VPCanonicalIVPHIRecipe>(getOperand(0)->getDef())
         ->getScalarType();
   }
 };
 
+/// A recipe for handling phi nodes of integer and floating-point inductions,
+/// producing their scalar values.
+class VPScalarIVStepsRecipe : public VPRecipeBase, public VPValue {
+  PHINode *IV;
+  const InductionDescriptor &IndDesc;
+
+public:
+  VPScalarIVStepsRecipe(PHINode *IV, const InductionDescriptor &IndDesc,
+                        VPValue *CanonicalIV, VPValue *Start, VPValue *Step,
+                        Instruction *Trunc)
+      : VPRecipeBase(VPScalarIVStepsSC, {CanonicalIV, Start, Step}),
+        VPValue(Trunc ? Trunc : IV, this), IV(IV), IndDesc(IndDesc) {}
+
+  ~VPScalarIVStepsRecipe() override = default;
+
+  /// Method to support type inquiry through isa, cast, and dyn_cast.
+  static inline bool classof(const VPDef *D) {
+    return D->getVPDefID() == VPRecipeBase::VPScalarIVStepsSC;
+  }
+  /// Extra classof implementations to allow directly casting from VPUser ->
+  /// VPScalarIVStepsRecipe.
+  static inline bool classof(const VPUser *U) {
+    auto *R = dyn_cast<VPRecipeBase>(U);
+    return R && R->getVPDefID() == VPRecipeBase::VPScalarIVStepsSC;
+  }
+  static inline bool classof(const VPRecipeBase *R) {
+    return R->getVPDefID() == VPRecipeBase::VPScalarIVStepsSC;
+  }
+
+  /// Generate the scalarized versions of the phi node as needed by their users.
+  void execute(VPTransformState &State) override;
+
+#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
+  /// Print the recipe.
+  void print(raw_ostream &O, const Twine &Indent,
+             VPSlotTracker &SlotTracker) const override;
+#endif
+
+  /// Returns true if the induction is canonical, i.e. starting at 0 and
+  /// incremented by UF * VF (= the original IV is incremented by 1).
+  bool isCanonical() const;
+
+  VPCanonicalIVPHIRecipe *getCanonicalIV() const;
+  VPValue *getStartValue() const { return getOperand(1); }
+  VPValue *getStepValue() const { return getOperand(2); }
+};
+
 /// VPBasicBlock serves as the leaf of the Hierarchical Control-Flow Graph. It
 /// holds a sequence of zero or more VPRecipe's each representing a sequence of
 /// output IR instructions. All PHI-like recipes must come before any non-PHI recipes.
 class VPBasicBlock : public VPBlockBase {
 public:
   using RecipeListTy = iplist<VPRecipeBase>;
 
 private:

llvm/lib/Transforms/Vectorize/VPlan.cpp

case VPBranchOnMaskSC:
     return false;
   case VPWidenIntOrFpInductionSC:
   case VPWidenCanonicalIVSC:
   case VPWidenPHISC:
   case VPBlendSC:
   case VPWidenSC:
   case VPWidenGEPSC:
   case VPReductionSC:
-  case VPWidenSelectSC: {
+  case VPWidenSelectSC:
+  case VPScalarIVStepsSC: {
     const Instruction *I =
         dyn_cast_or_null<Instruction>(getVPSingleValue()->getUnderlyingValue());
     (void)I;
     assert((!I || !I->mayHaveSideEffects()) &&
            "underlying instruction has side-effects");
     return false;
   }
   case VPReplicateSC: {
     auto *R = cast<VPReplicateRecipe>(this);
     return R->getUnderlyingInstr()->mayHaveSideEffects();
   }
   default:
     return true;
   }
 }
 
 void VPRecipeBase::insertBefore(VPRecipeBase *InsertPos) {
   assert(!Parent && "Recipe already in some VPBasicBlock");
   assert(InsertPos->getParent() &&
          "Insertion position not in any VPBasicBlock");
   Parent = InsertPos->getParent();
   Parent->getRecipeList().insert(InsertPos->getIterator(), this);
 }
 
+void VPRecipeBase::insertBefore(VPBasicBlock &BB,
+                                iplist<VPRecipeBase>::iterator I) {
+  assert(!Parent && "Recipe already in some VPBasicBlock");
+  assert(I == BB.end() || I->getParent() == &BB);
+  Parent = &BB;
+  BB.getRecipeList().insert(I, this);
+}
+
 void VPRecipeBase::insertAfter(VPRecipeBase *InsertPos) {
   assert(!Parent && "Recipe already in some VPBasicBlock");
   assert(InsertPos->getParent() &&
          "Insertion position not in any VPBasicBlock");
   Parent = InsertPos->getParent();
   Parent->getRecipeList().insertAfter(InsertPos->getIterator(), this);
 }
 

 }
 
 void VPRecipeBase::moveAfter(VPRecipeBase *InsertPos) {
   removeFromParent();
   insertAfter(InsertPos);
 }
 
 void VPRecipeBase::moveBefore(VPBasicBlock &BB,
                               iplist<VPRecipeBase>::iterator I) {
-  assert(I == BB.end() || I->getParent() == &BB);
   removeFromParent();
-  Parent = &BB;
-  BB.getRecipeList().insert(I, this);
+  insertBefore(BB, I);
 }
 
 void VPInstruction::generateInstruction(VPTransformState &State,
                                         unsigned Part) {
   IRBuilderBase &Builder = State.Builder;
   Builder.SetCurrentDebugLocation(DL);
 
   if (Instruction::isBinaryOp(getOpcode())) {
void VPlan::prepareToExecute(Value *TripCountV, Value *VectorTripCountV,
   // When vectorizing the epilogue loop, the canonical induction start value
   // needs to be changed from zero to the value after the main vector loop.
   if (CanonicalIVStartValue) {
     VPValue *VPV = new VPValue(CanonicalIVStartValue);
     addExternalDef(VPV);
     auto *IV = getCanonicalIV();
     assert(all_of(IV->users(),
                   [](const VPUser *U) {
+                    if (isa<VPScalarIVStepsRecipe>(U))
+                      return true;
                     auto *VPI = cast<VPInstruction>(U);
                     return VPI->getOpcode() ==
                                VPInstruction::CanonicalIVIncrement ||
                            VPI->getOpcode() ==
                                VPInstruction::CanonicalIVIncrementNUW;
                   }) &&
-           "the canonical IV should only be used by its increments when "
+           "the canonical IV should only be used by its increments or "
+           "ScalarIVSteps when "
            "resetting the start value");
     IV->setOperand(0, VPV);
   }
 }
 
 /// Generate the code inside the body of the vectorized loop. Assumes a single
 /// LoopVectorBody basic-block was created for this. Introduce additional
 /// basic-blocks as needed, and fill them all.

 #endif
 
 bool VPWidenIntOrFpInductionRecipe::isCanonical() const {
   auto *StartC = dyn_cast<ConstantInt>(getStartValue()->getLiveInIRValue());
   auto *StepC = dyn_cast<SCEVConstant>(getInductionDescriptor().getStep());
   return StartC && StartC->isZero() && StepC && StepC->isOne();
 }
 
+VPCanonicalIVPHIRecipe *VPScalarIVStepsRecipe::getCanonicalIV() const {
+  return cast<VPCanonicalIVPHIRecipe>(getOperand(0));
+}
+
+bool VPScalarIVStepsRecipe::isCanonical() const {
+  auto *CanIV = getCanonicalIV();
+  // The start value of the steps-recipe must match the start value of the
+  // canonical induction and it must step by 1.
+  if (CanIV->getStartValue() != getStartValue())
+    return false;
+  auto *StepVPV = getStepValue();
+  if (StepVPV->getDef())
+    return false;
+  auto *StepC = dyn_cast_or_null<ConstantInt>(StepVPV->getLiveInIRValue());
+  return StepC && StepC->isOne();
+}
+
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
+void VPScalarIVStepsRecipe::print(raw_ostream &O, const Twine &Indent,
+                                  VPSlotTracker &SlotTracker) const {
+  O << Indent;
+  printAsOperand(O, SlotTracker);
+  O << Indent << "= SCALAR-STEPS ";
+  printOperands(O, SlotTracker);
+}
+
 void VPWidenGEPRecipe::print(raw_ostream &O, const Twine &Indent,
                              VPSlotTracker &SlotTracker) const {
   O << Indent << "WIDEN-GEP ";
   O << (IsPtrLoopInvariant ? "Inv" : "Var");
   size_t IndicesNumber = IsIndexLoopInvariant.size();
   for (size_t I = 0; I < IndicesNumber; ++I)
     O << "[" << (IsIndexLoopInvariant[I] ? "Inv" : "Var") << "]";
 

llvm/lib/Transforms/Vectorize/VPlanTransforms.h

struct VPlanTransforms {
 
   /// Try to replace VPWidenCanonicalIVRecipes with a widened canonical IV
   /// recipe, if it exists.
   static void removeRedundantCanonicalIVs(VPlan &Plan);
 
   /// Try to remove dead recipes. At the moment, only dead header recipes are
   /// removed.
   static void removeDeadRecipes(VPlan &Plan, Loop &OrigLoop);
+
+  //  If all users of a vector IV need scalar values, provide them by building
+  //  scalar steps off of the canonical scalar IV, and remove the vector IV.
+  static void optimizeInductions(VPlan &Plan, ScalarEvolution &SE);
 };
 
 } // namespace llvm
 
 #endif // LLVM_TRANSFORMS_VECTORIZE_VPLANTRANSFORMS_H

llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp

if (R.mayHaveSideEffects() ||
            // Check for live-out users currently not modeled in VPlan.
            // TODO: Remove once live-outs are modeled in VPlan.
            return !OrigLoop.contains(cast<Instruction>(U));
          })))
       continue;
     R.eraseFromParent();
   }
 }
+
+void VPlanTransforms::optimizeInductions(VPlan &Plan, ScalarEvolution &SE) {
+  SmallVector<VPRecipeBase *> ToRemove;
+  VPBasicBlock *HeaderVPBB = Plan.getVectorLoopRegion()->getEntryBasicBlock();
+  for (VPRecipeBase &Phi : HeaderVPBB->phis()) {
+    auto *IV = dyn_cast<VPWidenIntOrFpInductionRecipe>(&Phi);
+    if (!IV || IV->needsVectorIV())
+      continue;
+
+    const InductionDescriptor &ID = IV->getInductionDescriptor();
+    const SCEV *StepSCEV = ID.getStep();
+    VPValue *Step = nullptr;
+    if (auto *E = dyn_cast<SCEVConstant>(StepSCEV)) {
+      Step = new VPValue(E->getValue());
+      Plan.addExternalDef(Step);
+    } else if (auto *E = dyn_cast<SCEVUnknown>(StepSCEV)) {
+      Step = new VPValue(E->getValue());
+      Plan.addExternalDef(Step);
+    } else {
+      Step = new VPExpandSCEVRecipe(StepSCEV, SE);
+    }
+
+    VPScalarIVStepsRecipe *Steps = new VPScalarIVStepsRecipe(
+        IV->getPHINode(), ID, Plan.getCanonicalIV(), IV->getStartValue(), Step,
+        IV->getTruncInst());
+
+    HeaderVPBB->insert(Steps, HeaderVPBB->getFirstNonPhi());
+    if (Step->getDef()) {
+      // TODO: Place the step in the preheader, once it is explicitly modeled in
+      // VPlan.
+      HeaderVPBB->insert(cast<VPRecipeBase>(Step->getDef()),
+                         HeaderVPBB->getFirstNonPhi());
+    }
+    IV->replaceAllUsesWith(Steps);
+  }
+}

llvm/lib/Transforms/Vectorize/VPlanValue.h

public:
   /// type identification.
   using VPRecipeTy = enum {
     VPBranchOnMaskSC,
     VPExpandSCEVSC,
     VPInstructionSC,
     VPInterleaveSC,
     VPReductionSC,
     VPReplicateSC,
+    VPScalarIVStepsSC,
     VPWidenCallSC,
     VPWidenCanonicalIVSC,
     VPWidenGEPSC,
     VPWidenMemoryInstructionSC,
     VPWidenSC,
     VPWidenSelectSC,
 
     // Phi-like recipes. Need to be kept together.
     VPBlendSC,
     VPCanonicalIVPHISC,
     VPFirstOrderRecurrencePHISC,
     VPWidenPHISC,
     VPWidenIntOrFpInductionSC,
     VPPredInstPHISC,
     VPReductionPHISC,

llvm/test/Transforms/LoopVectorize/ARM/tail-folding-scalar-epilogue-fallback.ll

 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <16 x i8>*
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, <16 x i8>* [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to <16 x i8>*
 ; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], <16 x i8>* [[TMP6]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[SIZE]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[END:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[SIZE]], [[HEADER:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i8* [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[PTR]], [[HEADER]] ]
 ; CHECK-NEXT:    br label [[BODY:%.*]]
 ; CHECK:       body:
 ; CHECK-NEXT:    [[DEC66:%.*]] = phi i32 [ [[DEC:%.*]], [[BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[BUFF:%.*]] = phi i8* [ [[INCDEC_PTR:%.*]], [[BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds i8, i8* [[BUFF]], i32 1
 ; CHECK-NEXT:    [[DEC]] = add nsw i32 [[DEC66]], -1
 ; CHECK-NEXT:    [[TMP8:%.*]] = load i8, i8* [[INCDEC_PTR]], align 1
 ; CHECK-NEXT:    store i8 [[TMP8]], i8* [[BUFF]], align 1
 ; CHECK-NEXT:    [[TOBOOL11:%.*]] = icmp eq i32 [[DEC]], 0
-; CHECK-NEXT:    br i1 [[TOBOOL11]], label [[END]], label [[BODY]], !llvm.loop !2
+; CHECK-NEXT:    br i1 [[TOBOOL11]], label [[END]], label [[BODY]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK:       end:
 ; CHECK-NEXT:    [[INCDEC_PTR_LCSSA:%.*]] = phi i8* [ [[INCDEC_PTR]], [[BODY]] ], [ [[IND_END2]], [[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    store i8* [[INCDEC_PTR_LCSSA]], i8** [[POS]], align 4
 ; CHECK-NEXT:    ret void
 ;
 header:
   %ptr0 = load i8*, i8** %pos, align 4
   br label %body

llvm/test/Transforms/LoopVectorize/PowerPC/optimal-epilog-vectorization.ll

 ; VF-TWO-CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32
 ; VF-TWO-CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VF-TWO-CHECK:       vector.ph:
 ; VF-TWO-CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32
 ; VF-TWO-CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
 ; VF-TWO-CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VF-TWO-CHECK:       vector.body:
 ; VF-TWO-CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; VF-TWO-CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 0
-; VF-TWO-CHECK-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 4
-; VF-TWO-CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 8
-; VF-TWO-CHECK-NEXT:    [[TMP15:%.*]] = add i64 [[INDEX]], 12
-; VF-TWO-CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], 16
-; VF-TWO-CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 20
-; VF-TWO-CHECK-NEXT:    [[TMP18:%.*]] = add i64 [[INDEX]], 24
-; VF-TWO-CHECK-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], 28
 ; VF-TWO-CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
 ; VF-TWO-CHECK-NEXT:    [[TMP20:%.*]] = add i32 [[OFFSET_IDX]], 0
 ; VF-TWO-CHECK-NEXT:    [[TMP21:%.*]] = add i32 [[OFFSET_IDX]], 4
 ; VF-TWO-CHECK-NEXT:    [[TMP22:%.*]] = add i32 [[OFFSET_IDX]], 8
 ; VF-TWO-CHECK-NEXT:    [[TMP23:%.*]] = add i32 [[OFFSET_IDX]], 12
 ; VF-TWO-CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[OFFSET_IDX]], 16
 ; VF-TWO-CHECK-NEXT:    [[TMP25:%.*]] = add i32 [[OFFSET_IDX]], 20
 ; VF-TWO-CHECK-NEXT:    [[TMP26:%.*]] = add i32 [[OFFSET_IDX]], 24
 ; VF-TWO-CHECK-NEXT:    [[TMP27:%.*]] = add i32 [[OFFSET_IDX]], 28
+; VF-TWO-CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 0
+; VF-TWO-CHECK-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 4
+; VF-TWO-CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 8
+; VF-TWO-CHECK-NEXT:    [[TMP15:%.*]] = add i64 [[INDEX]], 12
+; VF-TWO-CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], 16
+; VF-TWO-CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 20
+; VF-TWO-CHECK-NEXT:    [[TMP18:%.*]] = add i64 [[INDEX]], 24
+; VF-TWO-CHECK-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], 28
 ; VF-TWO-CHECK-NEXT:    [[TMP28:%.*]] = xor i32 [[TMP20]], -1
 ; VF-TWO-CHECK-NEXT:    [[TMP29:%.*]] = xor i32 [[TMP21]], -1
 ; VF-TWO-CHECK-NEXT:    [[TMP30:%.*]] = xor i32 [[TMP22]], -1
 ; VF-TWO-CHECK-NEXT:    [[TMP31:%.*]] = xor i32 [[TMP23]], -1
 ; VF-TWO-CHECK-NEXT:    [[TMP32:%.*]] = xor i32 [[TMP24]], -1
 ; VF-TWO-CHECK-NEXT:    [[TMP33:%.*]] = xor i32 [[TMP25]], -1
 ; VF-TWO-CHECK-NEXT:    [[TMP34:%.*]] = xor i32 [[TMP26]], -1
 ; VF-TWO-CHECK-NEXT:    [[TMP35:%.*]] = xor i32 [[TMP27]], -1

 ; VF-TWO-CHECK:       vec.epilog.ph:
 ; VF-TWO-CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
 ; VF-TWO-CHECK-NEXT:    [[N_MOD_VF16:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2
 ; VF-TWO-CHECK-NEXT:    [[N_VEC17:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF16]]
 ; VF-TWO-CHECK-NEXT:    [[IND_END:%.*]] = trunc i64 [[N_VEC17]] to i32
 ; VF-TWO-CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; VF-TWO-CHECK:       vec.epilog.vector.body:
 ; VF-TWO-CHECK-NEXT:    [[INDEX18:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT19:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; VF-TWO-CHECK-NEXT:    [[TMP117:%.*]] = add i64 [[INDEX18]], 0
 ; VF-TWO-CHECK-NEXT:    [[OFFSET_IDX23:%.*]] = trunc i64 [[INDEX18]] to i32
 ; VF-TWO-CHECK-NEXT:    [[TMP118:%.*]] = add i32 [[OFFSET_IDX23]], 0
+; VF-TWO-CHECK-NEXT:    [[TMP117:%.*]] = add i64 [[INDEX18]], 0
 ; VF-TWO-CHECK-NEXT:    [[TMP119:%.*]] = xor i32 [[TMP118]], -1
 ; VF-TWO-CHECK-NEXT:    [[TMP120:%.*]] = add i32 [[TMP119]], [[N]]
 ; VF-TWO-CHECK-NEXT:    [[TMP121:%.*]] = sext i32 [[TMP120]] to i64
 ; VF-TWO-CHECK-NEXT:    [[TMP122:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP121]]
 ; VF-TWO-CHECK-NEXT:    [[TMP123:%.*]] = getelementptr inbounds float, float* [[TMP122]], i32 0
 ; VF-TWO-CHECK-NEXT:    [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP123]], i32 -1
 ; VF-TWO-CHECK-NEXT:    [[TMP125:%.*]] = bitcast float* [[TMP124]] to <2 x float>*
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD24:%.*]] = load <2 x float>, <2 x float>* [[TMP125]], align 4

 ; VF-FOUR-CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32
 ; VF-FOUR-CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VF-FOUR-CHECK:       vector.ph:
 ; VF-FOUR-CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32
 ; VF-FOUR-CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
 ; VF-FOUR-CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VF-FOUR-CHECK:       vector.body:
 ; VF-FOUR-CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; VF-FOUR-CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 0
-; VF-FOUR-CHECK-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 4
-; VF-FOUR-CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 8
-; VF-FOUR-CHECK-NEXT:    [[TMP15:%.*]] = add i64 [[INDEX]], 12
-; VF-FOUR-CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], 16
-; VF-FOUR-CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 20
-; VF-FOUR-CHECK-NEXT:    [[TMP18:%.*]] = add i64 [[INDEX]], 24
-; VF-FOUR-CHECK-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], 28
 ; VF-FOUR-CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
 ; VF-FOUR-CHECK-NEXT:    [[TMP20:%.*]] = add i32 [[OFFSET_IDX]], 0
 ; VF-FOUR-CHECK-NEXT:    [[TMP21:%.*]] = add i32 [[OFFSET_IDX]], 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP22:%.*]] = add i32 [[OFFSET_IDX]], 8
 ; VF-FOUR-CHECK-NEXT:    [[TMP23:%.*]] = add i32 [[OFFSET_IDX]], 12
 ; VF-FOUR-CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[OFFSET_IDX]], 16
 ; VF-FOUR-CHECK-NEXT:    [[TMP25:%.*]] = add i32 [[OFFSET_IDX]], 20
 ; VF-FOUR-CHECK-NEXT:    [[TMP26:%.*]] = add i32 [[OFFSET_IDX]], 24
 ; VF-FOUR-CHECK-NEXT:    [[TMP27:%.*]] = add i32 [[OFFSET_IDX]], 28
+; VF-FOUR-CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 0
+; VF-FOUR-CHECK-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 4
+; VF-FOUR-CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 8
+; VF-FOUR-CHECK-NEXT:    [[TMP15:%.*]] = add i64 [[INDEX]], 12
+; VF-FOUR-CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], 16
+; VF-FOUR-CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 20
+; VF-FOUR-CHECK-NEXT:    [[TMP18:%.*]] = add i64 [[INDEX]], 24
+; VF-FOUR-CHECK-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], 28
 ; VF-FOUR-CHECK-NEXT:    [[TMP28:%.*]] = xor i32 [[TMP20]], -1
 ; VF-FOUR-CHECK-NEXT:    [[TMP29:%.*]] = xor i32 [[TMP21]], -1
 ; VF-FOUR-CHECK-NEXT:    [[TMP30:%.*]] = xor i32 [[TMP22]], -1
 ; VF-FOUR-CHECK-NEXT:    [[TMP31:%.*]] = xor i32 [[TMP23]], -1
 ; VF-FOUR-CHECK-NEXT:    [[TMP32:%.*]] = xor i32 [[TMP24]], -1
 ; VF-FOUR-CHECK-NEXT:    [[TMP33:%.*]] = xor i32 [[TMP25]], -1
 ; VF-FOUR-CHECK-NEXT:    [[TMP34:%.*]] = xor i32 [[TMP26]], -1
 ; VF-FOUR-CHECK-NEXT:    [[TMP35:%.*]] = xor i32 [[TMP27]], -1

 ; VF-FOUR-CHECK:       vec.epilog.ph:
 ; VF-FOUR-CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
 ; VF-FOUR-CHECK-NEXT:    [[N_MOD_VF16:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
 ; VF-FOUR-CHECK-NEXT:    [[N_VEC17:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF16]]
 ; VF-FOUR-CHECK-NEXT:    [[IND_END:%.*]] = trunc i64 [[N_VEC17]] to i32
 ; VF-FOUR-CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; VF-FOUR-CHECK:       vec.epilog.vector.body:
 ; VF-FOUR-CHECK-NEXT:    [[INDEX18:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT19:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; VF-FOUR-CHECK-NEXT:    [[TMP117:%.*]] = add i64 [[INDEX18]], 0
 ; VF-FOUR-CHECK-NEXT:    [[OFFSET_IDX23:%.*]] = trunc i64 [[INDEX18]] to i32
 ; VF-FOUR-CHECK-NEXT:    [[TMP118:%.*]] = add i32 [[OFFSET_IDX23]], 0
+; VF-FOUR-CHECK-NEXT:    [[TMP117:%.*]] = add i64 [[INDEX18]], 0
 ; VF-FOUR-CHECK-NEXT:    [[TMP119:%.*]] = xor i32 [[TMP118]], -1
 ; VF-FOUR-CHECK-NEXT:    [[TMP120:%.*]] = add i32 [[TMP119]], [[N]]
 ; VF-FOUR-CHECK-NEXT:    [[TMP121:%.*]] = sext i32 [[TMP120]] to i64
 ; VF-FOUR-CHECK-NEXT:    [[TMP122:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP121]]
 ; VF-FOUR-CHECK-NEXT:    [[TMP123:%.*]] = getelementptr inbounds float, float* [[TMP122]], i32 0
 ; VF-FOUR-CHECK-NEXT:    [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP123]], i32 -3
 ; VF-FOUR-CHECK-NEXT:    [[TMP125:%.*]] = bitcast float* [[TMP124]] to <4 x float>*
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD24:%.*]] = load <4 x float>, <4 x float>* [[TMP125]], align 4

llvm/test/Transforms/LoopVectorize/X86/pr36524.ll

 ; RUN: opt -S -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 < %s | FileCheck %s
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128-ni:1"
 
 define void @foo(i64* %ptr, i32* %ptr.2) {
 ; CHECK-LABEL: @foo(
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY:%.*]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 2, i64 3, i64 4, i64 5>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND_TRUNC:%.+]] = phi <4 x i32> [ <i32 2, i32 3, i32 4, i32 5>, %vector.ph ], [ [[VEC_IND_TRUNC_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:    = add i64 [[INDEX]], 0
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 2, [[INDEX]]
 ; CHECK-NEXT:    [[TRUNC:%.+]] = trunc i64 [[OFFSET_IDX]] to i32
 ; CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TRUNC]], 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = add i32 [[TRUNC]], 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = add i32 [[TRUNC]], 2
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TRUNC]], 3
+; CHECK-NEXT:    = add i64 [[INDEX]], 0
 ; CHECK-NEXT:    store i32 [[TMP7]], i32* %ptr.2, align 4
 ; CHECK-NEXT:    store i32 [[TMP8]], i32* %ptr.2, align 4
 ; CHECK-NEXT:    store i32 [[TMP9]], i32* %ptr.2, align 4
 ; CHECK-NEXT:    store i32 [[TMP10]], i32* %ptr.2, align 4
 ; CHECK:         store <4 x i64> [[VEC_IND]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], <i64 4, i64 4, i64 4, i64 4>
 ; CHECK-NEXT:    [[VEC_IND_TRUNC_NEXT]] = add <4 x i32> [[VEC_IND_TRUNC]], <i32 4, i32 4, i32 4, i32 4>

llvm/test/Transforms/LoopVectorize/float-induction.ll

 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
 ; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[FPINC]], [[CAST_CRD]]
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP3]]
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC1_INTERL2:       vector.body:
 ; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = sitofp i64 [[INDEX]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[FPINC]], [[TMP4]]
 ; VEC1_INTERL2-NEXT:    [[OFFSET_IDX:%.*]] = fsub fast float [[INIT]], [[TMP5]]
 ; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fsub fast float [[OFFSET_IDX]], [[FPINC]]
+; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
 ; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
 ; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX]], float* [[TMP7]], align 4
 ; VEC1_INTERL2-NEXT:    store float [[TMP6]], float* [[TMP8]], align 4
 ; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VEC1_INTERL2:       middle.block:

 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
 ; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = fmul reassoc float [[FPINC]], [[CAST_CRD]]
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fsub reassoc float [[INIT:%.*]], [[TMP3]]
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC1_INTERL2:       vector.body:
 ; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = sitofp i64 [[INDEX]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul reassoc float [[FPINC]], [[TMP4]]
 ; VEC1_INTERL2-NEXT:    [[OFFSET_IDX:%.*]] = fsub reassoc float [[INIT]], [[TMP5]]
 ; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fmul reassoc float [[FPINC]], 0.000000e+00
 ; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = fsub reassoc float [[OFFSET_IDX]], [[TMP6]]
 ; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = fsub reassoc float [[OFFSET_IDX]], [[FPINC]]
+; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
 ; VEC1_INTERL2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
 ; VEC1_INTERL2-NEXT:    store float [[TMP7]], float* [[TMP9]], align 4
 ; VEC1_INTERL2-NEXT:    store float [[TMP8]], float* [[TMP10]], align 4
 ; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC1_INTERL2-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; VEC1_INTERL2:       middle.block:

 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
 ; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC1_INTERL2:       vector.body:
 ; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = sitofp i64 [[INDEX]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP4]], 5.000000e-01
 ; VEC1_INTERL2-NEXT:    [[OFFSET_IDX:%.*]] = fadd fast float [[TMP5]], [[INIT]]
 ; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fadd fast float [[OFFSET_IDX]], 5.000000e-01
+; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
 ; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
 ; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX]], float* [[TMP7]], align 4
 ; VEC1_INTERL2-NEXT:    store float [[TMP6]], float* [[TMP8]], align 4
 ; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; VEC1_INTERL2:       middle.block:

 ; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = fmul fast float [[CAST_CRD]], -5.000000e-01
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP4]], 0x3FB99999A0000000
 ; VEC1_INTERL2-NEXT:    [[CAST_CRD3:%.*]] = sitofp i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP0]], [[CAST_CRD3]]
 ; VEC1_INTERL2-NEXT:    [[IND_END4:%.*]] = fadd fast float [[TMP5]], [[INIT:%.*]]
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC1_INTERL2:       vector.body:
 ; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; VEC1_INTERL2-NEXT:    [[INDUCTION5:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = sitofp i64 [[INDEX]] to float
-; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = fmul fast float [[TMP6]], -5.000000e-01
-; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = sitofp i64 [[INDEX]] to float
-; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = fmul fast float [[TMP0]], [[TMP8]]
-; VEC1_INTERL2-NEXT:    [[OFFSET_IDX8:%.*]] = fadd fast float [[TMP9]], [[INIT]]
-; VEC1_INTERL2-NEXT:    [[TMP10:%.*]] = fadd fast float [[OFFSET_IDX8]], [[TMP0]]
+; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = fmul fast float [[TMP0]], [[TMP6]]
+; VEC1_INTERL2-NEXT:    [[OFFSET_IDX:%.*]] = fadd fast float [[TMP7]], [[INIT]]
+; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = fadd fast float [[OFFSET_IDX]], [[TMP0]]
+; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC1_INTERL2-NEXT:    [[TMP10:%.*]] = fmul fast float [[TMP9]], -5.000000e-01
+; VEC1_INTERL2-NEXT:    [[INDUCTION6:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
-; VEC1_INTERL2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION5]]
-; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX8]], float* [[TMP11]], align 4
-; VEC1_INTERL2-NEXT:    store float [[TMP10]], float* [[TMP12]], align 4
-; VEC1_INTERL2-NEXT:    [[TMP13:%.*]] = fadd fast float [[OFFSET_IDX8]], [[TMP0]]
-; VEC1_INTERL2-NEXT:    [[TMP14:%.*]] = fadd fast float [[TMP10]], [[TMP0]]
-; VEC1_INTERL2-NEXT:    [[TMP15:%.*]] = fadd fast float [[TMP7]], 0xBFD99999A0000000
-; VEC1_INTERL2-NEXT:    [[TMP16:%.*]] = fadd fast float [[TMP7]], 0xBFECCCCCC0000000
+; VEC1_INTERL2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION6]]
+; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX]], float* [[TMP11]], align 4
+; VEC1_INTERL2-NEXT:    store float [[TMP8]], float* [[TMP12]], align 4
+; VEC1_INTERL2-NEXT:    [[TMP13:%.*]] = fadd fast float [[OFFSET_IDX]], [[TMP0]]
+; VEC1_INTERL2-NEXT:    [[TMP14:%.*]] = fadd fast float [[TMP8]], [[TMP0]]
+; VEC1_INTERL2-NEXT:    [[TMP15:%.*]] = fadd fast float [[TMP10]], 0xBFD99999A0000000
+; VEC1_INTERL2-NEXT:    [[TMP16:%.*]] = fadd fast float [[TMP10]], 0xBFECCCCCC0000000
 ; VEC1_INTERL2-NEXT:    [[TMP17:%.*]] = fadd fast float [[TMP15]], [[TMP13]]
 ; VEC1_INTERL2-NEXT:    [[TMP18:%.*]] = fadd fast float [[TMP16]], [[TMP14]]
 ; VEC1_INTERL2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[INDEX]]
-; VEC1_INTERL2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDUCTION5]]
+; VEC1_INTERL2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDUCTION6]]
 ; VEC1_INTERL2-NEXT:    store float [[TMP17]], float* [[TMP19]], align 4
 ; VEC1_INTERL2-NEXT:    store float [[TMP18]], float* [[TMP20]], align 4
 ; VEC1_INTERL2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, float* [[C:%.*]], i64 [[INDEX]]
-; VEC1_INTERL2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[INDUCTION5]]
+; VEC1_INTERL2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[INDUCTION6]]
 ; VEC1_INTERL2-NEXT:    store float [[TMP15]], float* [[TMP21]], align 4
 ; VEC1_INTERL2-NEXT:    store float [[TMP16]], float* [[TMP22]], align 4
 ; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC1_INTERL2-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; VEC1_INTERL2:       middle.block:
 ; VEC1_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]

 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
 ; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], 1.000000e+00
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC1_INTERL2:       vector.body:
 ; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = sitofp i64 [[INDEX]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP4]], 5.000000e-01
 ; VEC1_INTERL2-NEXT:    [[OFFSET_IDX:%.*]] = fadd fast float [[TMP5]], 1.000000e+00
 ; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fadd fast float [[TMP5]], 1.500000e+00
+; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
 ; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
 ; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX]], float* [[TMP7]], align 4
 ; VEC1_INTERL2-NEXT:    store float [[TMP6]], float* [[TMP8]], align 4
 ; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; VEC1_INTERL2:       middle.block:

 ; VEC4_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 8
 ; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL2:       vector.ph:
 ; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775800
 ; VEC4_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
 ; VEC4_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC4_INTERL2:       vector.body:
 ; VEC4_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE17:%.*]] ]
-; VEC4_INTERL2-NEXT:    [[TMP0:%.*]] = or i64 [[INDEX]], 4
-; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC4_INTERL2-NEXT:    [[TMP0:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = or i64 [[INDEX]], 4
 ; VEC4_INTERL2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
 ; VEC4_INTERL2-NEXT:    [[TMP3:%.*]] = bitcast float* [[TMP2]] to <4 x float>*
 ; VEC4_INTERL2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP3]], align 4
 ; VEC4_INTERL2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 4
 ; VEC4_INTERL2-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP4]] to <4 x float>*
 ; VEC4_INTERL2-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x float>, <4 x float>* [[TMP5]], align 4
 ; VEC4_INTERL2-NEXT:    [[TMP6:%.*]] = fcmp fast oeq <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; VEC4_INTERL2-NEXT:    [[TMP7:%.*]] = fcmp fast oeq <4 x float> [[WIDE_LOAD3]], zeroinitializer
 ; VEC4_INTERL2-NEXT:    [[TMP8:%.*]] = extractelement <4 x i1> [[TMP6]], i64 0
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP8]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
 ; VEC4_INTERL2:       pred.store.if:
 ; VEC4_INTERL2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDEX]]
-; VEC4_INTERL2-NEXT:    store float [[TMP1]], float* [[TMP9]], align 4
+; VEC4_INTERL2-NEXT:    store float [[TMP0]], float* [[TMP9]], align 4
 ; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE]]
 ; VEC4_INTERL2:       pred.store.continue:
 ; VEC4_INTERL2-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP6]], i64 1
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP10]], label [[PRED_STORE_IF4:%.*]], label [[PRED_STORE_CONTINUE5:%.*]]
 ; VEC4_INTERL2:       pred.store.if4:
-; VEC4_INTERL2-NEXT:    [[TMP11:%.*]] = fadd fast float [[TMP1]], 1.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP11:%.*]] = fadd fast float [[TMP0]], 1.000000e+00
 ; VEC4_INTERL2-NEXT:    [[TMP12:%.*]] = or i64 [[INDEX]], 1
 ; VEC4_INTERL2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]]
 ; VEC4_INTERL2-NEXT:    store float [[TMP11]], float* [[TMP13]], align 4
 ; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE5]]
 ; VEC4_INTERL2:       pred.store.continue5:
 ; VEC4_INTERL2-NEXT:    [[TMP14:%.*]] = extractelement <4 x i1> [[TMP6]], i64 2
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP14]], label [[PRED_STORE_IF6:%.*]], label [[PRED_STORE_CONTINUE7:%.*]]
 ; VEC4_INTERL2:       pred.store.if6:
-; VEC4_INTERL2-NEXT:    [[TMP15:%.*]] = fadd fast float [[TMP1]], 2.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP15:%.*]] = fadd fast float [[TMP0]], 2.000000e+00
 ; VEC4_INTERL2-NEXT:    [[TMP16:%.*]] = or i64 [[INDEX]], 2
 ; VEC4_INTERL2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP16]]
 ; VEC4_INTERL2-NEXT:    store float [[TMP15]], float* [[TMP17]], align 4
 ; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE7]]
 ; VEC4_INTERL2:       pred.store.continue7:
 ; VEC4_INTERL2-NEXT:    [[TMP18:%.*]] = extractelement <4 x i1> [[TMP6]], i64 3
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP18]], label [[PRED_STORE_IF8:%.*]], label [[PRED_STORE_CONTINUE9:%.*]]
 ; VEC4_INTERL2:       pred.store.if8:
-; VEC4_INTERL2-NEXT:    [[TMP19:%.*]] = fadd fast float [[TMP1]], 3.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP19:%.*]] = fadd fast float [[TMP0]], 3.000000e+00
 ; VEC4_INTERL2-NEXT:    [[TMP20:%.*]] = or i64 [[INDEX]], 3
 ; VEC4_INTERL2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP20]]
 ; VEC4_INTERL2-NEXT:    store float [[TMP19]], float* [[TMP21]], align 4
 ; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE9]]
 ; VEC4_INTERL2:       pred.store.continue9:
 ; VEC4_INTERL2-NEXT:    [[TMP22:%.*]] = extractelement <4 x i1> [[TMP7]], i64 0
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP22]], label [[PRED_STORE_IF10:%.*]], label [[PRED_STORE_CONTINUE11:%.*]]
 ; VEC4_INTERL2:       pred.store.if10:
-; VEC4_INTERL2-NEXT:    [[TMP23:%.*]] = fadd fast float [[TMP1]], 4.000000e+00
-; VEC4_INTERL2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP0]]
+; VEC4_INTERL2-NEXT:    [[TMP23:%.*]] = fadd fast float [[TMP0]], 4.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP1]]
 ; VEC4_INTERL2-NEXT:    store float [[TMP23]], float* [[TMP24]], align 4
 ; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE11]]
 ; VEC4_INTERL2:       pred.store.continue11:
 ; VEC4_INTERL2-NEXT:    [[TMP25:%.*]] = extractelement <4 x i1> [[TMP7]], i64 1
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP25]], label [[PRED_STORE_IF12:%.*]], label [[PRED_STORE_CONTINUE13:%.*]]
 ; VEC4_INTERL2:       pred.store.if12:
-; VEC4_INTERL2-NEXT:    [[TMP26:%.*]] = fadd fast float [[TMP1]], 5.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP26:%.*]] = fadd fast float [[TMP0]], 5.000000e+00
 ; VEC4_INTERL2-NEXT:    [[TMP27:%.*]] = or i64 [[INDEX]], 5
 ; VEC4_INTERL2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP27]]
 ; VEC4_INTERL2-NEXT:    store float [[TMP26]], float* [[TMP28]], align 4
 ; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE13]]
 ; VEC4_INTERL2:       pred.store.continue13:
 ; VEC4_INTERL2-NEXT:    [[TMP29:%.*]] = extractelement <4 x i1> [[TMP7]], i64 2
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP29]], label [[PRED_STORE_IF14:%.*]], label [[PRED_STORE_CONTINUE15:%.*]]
 ; VEC4_INTERL2:       pred.store.if14:
-; VEC4_INTERL2-NEXT:    [[TMP30:%.*]] = fadd fast float [[TMP1]], 6.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP30:%.*]] = fadd fast float [[TMP0]], 6.000000e+00
 ; VEC4_INTERL2-NEXT:    [[TMP31:%.*]] = or i64 [[INDEX]], 6
 ; VEC4_INTERL2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP31]]
 ; VEC4_INTERL2-NEXT:    store float [[TMP30]], float* [[TMP32]], align 4
 ; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE15]]
 ; VEC4_INTERL2:       pred.store.continue15:
 ; VEC4_INTERL2-NEXT:    [[TMP33:%.*]] = extractelement <4 x i1> [[TMP7]], i64 3
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP33]], label [[PRED_STORE_IF16:%.*]], label [[PRED_STORE_CONTINUE17]]
 ; VEC4_INTERL2:       pred.store.if16:
-; VEC4_INTERL2-NEXT:    [[TMP34:%.*]] = fadd fast float [[TMP1]], 7.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP34:%.*]] = fadd fast float [[TMP0]], 7.000000e+00
 ; VEC4_INTERL2-NEXT:    [[TMP35:%.*]] = or i64 [[INDEX]], 7
 ; VEC4_INTERL2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP35]]
 ; VEC4_INTERL2-NEXT:    store float [[TMP34]], float* [[TMP36]], align 4
 ; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE17]]
 ; VEC4_INTERL2:       pred.store.continue17:
 ; VEC4_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; VEC4_INTERL2-NEXT:    [[TMP37:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP37]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]

 ; VEC1_INTERL2-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
 ; VEC1_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
 ; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
 ; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC1_INTERL2:       vector.body:
-; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE5:%.*]] ]
-; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
+; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE4:%.*]] ]
 ; VEC1_INTERL2-NEXT:    [[TMP0:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
 ; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
 ; VEC1_INTERL2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
 ; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP1]], align 4
 ; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = load float, float* [[TMP2]], align 4
 ; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fcmp fast oeq float [[TMP3]], 0.000000e+00
 ; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fcmp fast oeq float [[TMP4]], 0.000000e+00
 ; VEC1_INTERL2-NEXT:    br i1 [[TMP5]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
 ; VEC1_INTERL2:       pred.store.if:
 ; VEC1_INTERL2-NEXT:    store float [[TMP0]], float* [[TMP1]], align 4
 ; VEC1_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE]]
 ; VEC1_INTERL2:       pred.store.continue:
-; VEC1_INTERL2-NEXT:    br i1 [[TMP6]], label [[PRED_STORE_IF4:%.*]], label [[PRED_STORE_CONTINUE5]]
-; VEC1_INTERL2:       pred.store.if6:
+; VEC1_INTERL2-NEXT:    br i1 [[TMP6]], label [[PRED_STORE_IF3:%.*]], label [[PRED_STORE_CONTINUE4]]
+; VEC1_INTERL2:       pred.store.if3:
 ; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = fadd fast float [[TMP0]], 1.000000e+00
 ; VEC1_INTERL2-NEXT:    store float [[TMP7]], float* [[TMP2]], align 4
-; VEC1_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE5]]
-; VEC1_INTERL2:       pred.store.continue7:
+; VEC1_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE4]]
+; VEC1_INTERL2:       pred.store.continue4:
 ; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; VEC1_INTERL2:       middle.block:
 ; VEC1_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; VEC1_INTERL2:       scalar.ph:
 ; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]

llvm/test/Transforms/LoopVectorize/if-pred-stores.ll

 define void @minimal_bit_widths(i1 %c) {
 ; UNROLL-LABEL: @minimal_bit_widths(
 ; UNROLL-NEXT:  entry:
 ; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL:       vector.body:
 ; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ]
 ; UNROLL-NEXT:    br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE6]]
 ; UNROLL:       pred.store.if:
-; UNROLL-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
-; UNROLL-NEXT:    [[TMP0:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION]]
+; UNROLL-NEXT:    [[INDUCTION3:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NEXT:    [[TMP0:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION3]]
 ; UNROLL-NEXT:    [[TMP1:%.*]] = load i8, i8* [[TMP0]], align 1
 ; UNROLL-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
 ; UNROLL-NEXT:    [[TMP3:%.*]] = trunc i32 [[TMP2]] to i8
 ; UNROLL-NEXT:    store i8 [[TMP3]], i8* [[TMP0]], align 1
-; UNROLL-NEXT:    [[INDUCTION2:%.*]] = add i64 [[INDEX]], 1
-; UNROLL-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION2]]
+; UNROLL-NEXT:    [[INDUCTION4:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION4]]
 ; UNROLL-NEXT:    [[TMP5:%.*]] = load i8, i8* [[TMP4]], align 1
 ; UNROLL-NEXT:    [[TMP6:%.*]] = zext i8 [[TMP5]] to i32
 ; UNROLL-NEXT:    [[TMP7:%.*]] = trunc i32 [[TMP6]] to i8
 ; UNROLL-NEXT:    store i8 [[TMP7]], i8* [[TMP4]], align 1
 ; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE6]]
 ; UNROLL:       pred.store.continue4:
 ; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; UNROLL-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef

 ; UNROLL-NOSIMPLIFY-NEXT:  entry:
 ; UNROLL-NOSIMPLIFY-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NOSIMPLIFY:       vector.ph:
 ; UNROLL-NOSIMPLIFY-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NOSIMPLIFY:       vector.body:
 ; UNROLL-NOSIMPLIFY-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ]
 ; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
 ; UNROLL-NOSIMPLIFY:       pred.store.if:
-; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
-; UNROLL-NOSIMPLIFY-NEXT:    [[TMP0:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION]]
+; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION3:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NOSIMPLIFY-NEXT:    [[TMP0:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION3]]
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP1:%.*]] = load i8, i8* [[TMP0]], align 1
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP3:%.*]] = trunc i32 [[TMP2]] to i8
 ; UNROLL-NOSIMPLIFY-NEXT:    store i8 [[TMP3]], i8* [[TMP0]], align 1
 ; UNROLL-NOSIMPLIFY-NEXT:    br label [[PRED_STORE_CONTINUE]]
 ; UNROLL-NOSIMPLIFY:       pred.store.continue:
 ; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[C]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6]]
 ; UNROLL-NOSIMPLIFY:       pred.store.if3:

llvm/test/Transforms/LoopVectorize/induction.ll

 ; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
 ; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
 ; UNROLL-NO-IC-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
 ; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = add i32 190, [[CAST_CRD]]
 ; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-IC:       vector.body:
 ; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
 ; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 2
-; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
 ; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP3]]
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
 ; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <2 x i32>*
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP8]], align 4
 ; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 2
 ; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <2 x i32>*
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP10]], align 4

 ; UNROLL-NO-IC:       vector.ph:
 ; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 4
 ; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
 ; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
 ; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-IC:       vector.body:
 ; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD4:%.*]] = add <2 x i32> [[VEC_IND3]], <i32 4, i32 4>
 ; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
 ; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 2
-; UNROLL-NO-IC-NEXT:    [[STEP_ADD4:%.*]] = add <2 x i32> [[VEC_IND3]], <i32 4, i32 4>
 ; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
 ; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP1]]
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND3]], <2 x i32>* [[TMP5]], align 4
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 2
 ; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD4]], <2 x i32>* [[TMP7]], align 4

 ; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = mul i32 [[STEP]], 2
 ; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP20]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
 ; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-IC:       vector.body:
 ; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
 ; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], 0
 ; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = add i64 [[INDEX]], 2
-; UNROLL-NO-IC-NEXT:    [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
 ; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
 ; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>
 ; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[TMP21]]
 ; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[TMP22]]
 ; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, i32* [[TMP25]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = bitcast i32* [[TMP27]] to <2 x i32>*
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP23]], <2 x i32>* [[TMP28]], align 4
 ; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, i32* [[TMP25]], i32 2

llvm/test/Transforms/LoopVectorize/induction_plus.ll

 ; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 
 @array = common global [1024 x i32] zeroinitializer, align 16
 
 define void @array_at_plus_one(i32 %n) {
 ; CHECK-LABEL: @array_at_plus_one(
 ; CHECK: %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK-NEXT: [[VEC_IV_1:%.+]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %vector.ph ], [ [[VEC_IV_1_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT: [[VEC_IV_TRUNC:%.+]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %vector.ph ], [ [[VEC_IV_TRUNC_NEXT:%.+]], %vector.body ]
+; CHECK: [[VEC_IV_TRUNC:%.+]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %vector.ph ], [ [[VEC_IV_TRUNC_NEXT:%.+]], %vector.body ]
 ; CHECK: [[T1:%.+]] = add i64 %index, 0
 ; CHECK: [[T2:%.+]] = add nsw i64 [[T1]], 12
 ; CHECK-NEXT: [[GEP:%.+]] = getelementptr inbounds [1024 x i32], [1024 x i32]* @array, i64 0, i64 [[T2]]
 ; CHECK-NEXT: [[GEP0:%.+]] = getelementptr inbounds i32, i32* [[GEP]], i32 0
 ; CHECK-NEXT: [[BC:%.+]] = bitcast i32* [[GEP0]] to <4 x i32>*
 ; CHECK-NEXT: store <4 x i32> [[VEC_IV_TRUNC]], <4 x i32>* [[BC]]
-; CHECK: [[VEC_IV_1_NEXT]] = add <4 x i64> [[VEC_IV_1]], <i64 4, i64 4, i64 4, i64 4>
 ; CHECK: [[VEC_IV_TRUNC_NEXT]] = add <4 x i32> [[VEC_IV_TRUNC]], <i32 4, i32 4, i32 4, i32 4>
 ; CHECK: ret void
 ;
 entry:
   br label %loop
 
 loop:
   %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]

llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization.ll

 ; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
 ; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 0
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
-; CHECK-NEXT:    [[TMP13:%.*]] = add i32 [[OFFSET_IDX]], 0
-; CHECK-NEXT:    [[TMP14:%.*]] = xor i32 [[TMP13]], -1
+; CHECK-NEXT:    [[TMP12:%.*]] = add i32 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP14:%.*]] = xor i32 [[TMP12]], -1
 ; CHECK-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], [[N]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = sext i32 [[TMP15]] to i64
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP16]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[TMP17]], i32 0
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, float* [[TMP18]], i32 -3
 ; CHECK-NEXT:    [[TMP20:%.*]] = bitcast float* [[TMP19]] to <4 x float>*
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP20]], align 4
 ; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x float> [[WIDE_LOAD]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP21:%.*]] = fadd fast <4 x float> [[REVERSE]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
-; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP12]]
+; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP13]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, float* [[TMP22]], i32 0
 ; CHECK-NEXT:    [[TMP24:%.*]] = bitcast float* [[TMP23]] to <4 x float>*
 ; CHECK-NEXT:    store <4 x float> [[TMP21]], <4 x float>* [[TMP24]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
 ; CHECK:       vec.epilog.iter.check:
 ; CHECK-NEXT:    [[IND_END7:%.*]] = trunc i64 [[N_VEC]] to i32
 ; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
 ; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
 ; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
 ; CHECK:       vec.epilog.ph:
 ; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
 ; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
 ; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF2]]
 ; CHECK-NEXT:    [[IND_END:%.*]] = trunc i64 [[N_VEC3]] to i32
 ; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT5:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP26:%.*]] = add i64 [[INDEX4]], 0
 ; CHECK-NEXT:    [[OFFSET_IDX9:%.*]] = trunc i64 [[INDEX4]] to i32
-; CHECK-NEXT:    [[TMP27:%.*]] = add i32 [[OFFSET_IDX9]], 0
-; CHECK-NEXT:    [[TMP28:%.*]] = xor i32 [[TMP27]], -1
+; CHECK-NEXT:    [[TMP26:%.*]] = add i32 [[OFFSET_IDX9]], 0
+; CHECK-NEXT:    [[TMP27:%.*]] = add i64 [[INDEX4]], 0
+; CHECK-NEXT:    [[TMP28:%.*]] = xor i32 [[TMP26]], -1
 ; CHECK-NEXT:    [[TMP29:%.*]] = add i32 [[TMP28]], [[N]]
 ; CHECK-NEXT:    [[TMP30:%.*]] = sext i32 [[TMP29]] to i64
 ; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP30]]
 ; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP31]], i32 0
 ; CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, float* [[TMP32]], i32 -3
 ; CHECK-NEXT:    [[TMP34:%.*]] = bitcast float* [[TMP33]] to <4 x float>*
 ; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP34]], align 4
 ; CHECK-NEXT:    [[REVERSE11:%.*]] = shufflevector <4 x float> [[WIDE_LOAD10]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP35:%.*]] = fadd fast <4 x float> [[REVERSE11]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
-; CHECK-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP26]]
+; CHECK-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP27]]
 ; CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, float* [[TMP36]], i32 0
 ; CHECK-NEXT:    [[TMP38:%.*]] = bitcast float* [[TMP37]] to <4 x float>*
 ; CHECK-NEXT:    store <4 x float> [[TMP35]], <4 x float>* [[TMP38]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT5]] = add nuw i64 [[INDEX4]], 4
 ; CHECK-NEXT:    [[TMP39:%.*]] = icmp eq i64 [[INDEX_NEXT5]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[TMP39]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
 ; CHECK-NEXT:    [[CMP_N8:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC3]]

llvm/test/Transforms/LoopVectorize/pointer-induction.ll

 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8*, i8** [[START_1]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, i8* [[POINTER_PHI]], <4 x i64> <i64 0, i64 1, i64 2, i64 3>
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, <4 x i8*> [[TMP2]], i64 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8*, i8** [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to <4 x i8*>*
 ; CHECK-NEXT:    store <4 x i8*> [[TMP3]], <4 x i8*>* [[TMP5]], align 8
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x i8*> [[TMP2]], i32 0
+; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x i8*> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, i8* [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to <4 x i8>*
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, <4 x i8>* [[TMP8]], align 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = add <4 x i8> [[WIDE_LOAD]], <i8 1, i8 1, i8 1, i8 1>
 ; CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP7]] to <4 x i8>*
 ; CHECK-NEXT:    store <4 x i8> [[TMP9]], <4 x i8>* [[TMP10]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, i8* [[POINTER_PHI]], i64 4

llvm/test/Transforms/LoopVectorize/reverse_induction.ll

for.body:
   br i1 %exitcond, label %for.body, label %loopend
 
 loopend:
   ret i32 %inc.redux
 }
 
 ; CHECK-LABEL: @reverse_induction_i128(
 ; CHECK: %index = phi i128 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK: %offset.idx = sub i128 %startval, %index
-; CHECK: %[[a0:.+]] = add i128 %offset.idx, 0
-; CHECK: %[[a4:.+]] = add i128 %offset.idx, -4
+; CHECK: [[OFFSET_IDX:%.+]] = sub i128 %startval, %index
+; CHECK: %[[a0:.+]] = add i128 [[OFFSET_IDX]], 0
+; CHECK: %[[a4:.+]] = add i128 [[OFFSET_IDX]], -4
 
 define i32 @reverse_induction_i128(i128 %startval, i32 * %ptr) {
 entry:
   br label %for.body
 
 for.body:
   %add.i7 = phi i128 [ %startval, %entry ], [ %add.i, %for.body ]
   %i.06 = phi i32 [ 0, %entry ], [ %inc4, %for.body ]

llvm/test/Transforms/LoopVectorize/runtime-check-needed-but-empty.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -loop-vectorize -force-vector-width=4 -S %s | FileCheck %s
 
 define void @test(float* %A, i32 %x) {
 ; CHECK-LABEL: @test(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
 ; CHECK:       vector.scevcheck:
 ; CHECK-NEXT:    [[IDENT_CHECK:%.*]] = icmp ne i32 [[X:%.*]], 1
 ; CHECK-NEXT:    br i1 [[IDENT_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP9:%.*]] = trunc i64 [[INDEX]] to i32
-; CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 0
-; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i64 [[TMP2]] to i32
-; CHECK-NEXT:    [[TMP4:%.*]] = mul i32 [[TMP3]], [[X]]
-; CHECK-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[TMP6]], i32 0
-; CHECK-NEXT:    [[TMP8:%.*]] = bitcast float* [[TMP7]] to <4 x float>*
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP8]], align 4
-; CHECK-NEXT:    [[TMP11:%.*]] = mul i32 [[TMP10]], [[X]]
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDEX]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[TMP1]], 0
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[TMP3]], 1
+; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[TMP4]] to i32
+; CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[TMP5]], [[X]]
+; CHECK-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 0
+; CHECK-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to <4 x float>*
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP10]], align 4
+; CHECK-NEXT:    [[TMP11:%.*]] = mul i32 [[TMP2]], [[X]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[TMP13]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = bitcast float* [[TMP14]] to <4 x float>*
 ; CHECK-NEXT:    store <4 x float> [[WIDE_LOAD]], <4 x float>* [[TMP15]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]

llvm/test/Transforms/LoopVectorize/use-scalar-epilogue-if-tp-fails.ll

 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <4 x i8>*
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, <4 x i8>* [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to <4 x i8>*
 ; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD]], <4 x i8>* [[TMP6]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[SIZE]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[END:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[SIZE]], [[HEADER:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i8* [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[PTR]], [[HEADER]] ]
 ; CHECK-NEXT:    br label [[BODY:%.*]]
 ; CHECK:       body:
 ; CHECK-NEXT:    [[DEC66:%.*]] = phi i32 [ [[DEC:%.*]], [[BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[BUFF:%.*]] = phi i8* [ [[INCDEC_PTR:%.*]], [[BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds i8, i8* [[BUFF]], i32 1
 ; CHECK-NEXT:    [[DEC]] = add nsw i32 [[DEC66]], -1
 ; CHECK-NEXT:    [[TMP8:%.*]] = load i8, i8* [[INCDEC_PTR]], align 1
 ; CHECK-NEXT:    store i8 [[TMP8]], i8* [[BUFF]], align 1
 ; CHECK-NEXT:    [[TOBOOL11:%.*]] = icmp eq i32 [[DEC]], 0
-; CHECK-NEXT:    br i1 [[TOBOOL11]], label [[END]], label [[BODY]], !llvm.loop !2
+; CHECK-NEXT:    br i1 [[TOBOOL11]], label [[END]], label [[BODY]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK:       end:
 ; CHECK-NEXT:    [[INCDEC_PTR_LCSSA:%.*]] = phi i8* [ [[INCDEC_PTR]], [[BODY]] ], [ [[IND_END2]], [[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    store i8* [[INCDEC_PTR_LCSSA]], i8** [[POS]], align 4
 ; CHECK-NEXT:    ret void
 ;
 header:
   %ptr0 = load i8*, i8** %pos, align 4
   br label %body

 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <4 x i8>*
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, <4 x i8>* [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to <4 x i8>*
 ; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD]], <4 x i8>* [[TMP6]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !4
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[SIZE]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[END:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[SIZE]], [[HEADER:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i8* [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[PTR]], [[HEADER]] ]
 ; CHECK-NEXT:    br label [[BODY:%.*]]
 ; CHECK:       body:
 ; CHECK-NEXT:    [[DEC66:%.*]] = phi i32 [ [[DEC:%.*]], [[BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[BUFF:%.*]] = phi i8* [ [[INCDEC_PTR:%.*]], [[BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds i8, i8* [[BUFF]], i32 1
 ; CHECK-NEXT:    [[DEC]] = add nsw i32 [[DEC66]], -1
 ; CHECK-NEXT:    [[TMP8:%.*]] = load i8, i8* [[INCDEC_PTR]], align 1
 ; CHECK-NEXT:    store i8 [[TMP8]], i8* [[BUFF]], align 1
 ; CHECK-NEXT:    [[TOBOOL11:%.*]] = icmp eq i32 [[DEC]], 0
-; CHECK-NEXT:    br i1 [[TOBOOL11]], label [[END]], label [[BODY]], !llvm.loop !5
+; CHECK-NEXT:    br i1 [[TOBOOL11]], label [[END]], label [[BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       end:
 ; CHECK-NEXT:    [[INCDEC_PTR_LCSSA:%.*]] = phi i8* [ [[INCDEC_PTR]], [[BODY]] ], [ [[IND_END2]], [[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    store i8* [[INCDEC_PTR_LCSSA]], i8** [[POS]], align 4
 ; CHECK-NEXT:    ret void
 ;
 header:
   %ptr0 = load i8*, i8** %pos, align 4
   br label %body

llvm/test/Transforms/LoopVectorize/vplan-dot-printing.ll

 ; REQUIRES: asserts
 
 ; RUN: opt -loop-vectorize -debug-only=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -vplan-print-in-dot-format -disable-output %s 2>&1 | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 
 ; Verify that -vplan-print-in-dot-format option works.
 
 define void @print_call_and_memory(i64 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
 ; CHECK:       subgraph cluster_N0 {
 ; CHECK-NEXT:    fontname=Courier
 ; CHECK-NEXT:    label="\<x1\> vector loop"
 ; CHECK-NEXT:    N1 [label =
 ; CHECK-NEXT:    "for.body:\l" +
 ; CHECK-NEXT:    "  EMIT vp\<[[CAN_IV:%.+]]\> = CANONICAL-INDUCTION\l" +
-; CHECK-NEXT:    "  WIDEN-INDUCTION %iv = phi %iv.next, 0\l" +
+; CHECK-NEXT:    "  ir\<%iv\> = SCALAR-STEPS vp\<[[CAN_IV]]\>, ir\<0\>, ir\<1\>\l" +
 ; CHECK-NEXT:    "  CLONE ir\<%arrayidx\> = getelementptr ir\<%y\>, ir\<%iv\>\l" +
 ; CHECK-NEXT:    "  WIDEN ir\<%lv\> = load ir\<%arrayidx\>\l" +
 ; CHECK-NEXT:    "  WIDEN-CALL ir\<%call\> = call @llvm.sqrt.f32(ir\<%lv\>)\l" +
 ; CHECK-NEXT:    "  CLONE ir\<%arrayidx2\> = getelementptr ir\<%x\>, ir\<%iv\>\l" +
 ; CHECK-NEXT:    "  WIDEN store ir\<%arrayidx2\>, ir\<%call\>\l" +
 ; CHECK-NEXT:    "  EMIT vp\<[[CAN_IV_NEXT:%.+]]\> = VF * UF +(nuw) vp\<[[CAN_IV]]\>\l" +
 ; CHECK-NEXT:    "  EMIT branch-on-count vp\<[[CAN_IV_NEXT]]\> vp\<{{.+}}\>\l" +
 ; CHECK-NEXT:    "No successors\l"

llvm/test/Transforms/LoopVectorize/vplan-printing.ll

 ; REQUIRES: asserts
 
 ; RUN: opt -loop-vectorize -debug-only=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -prefer-inloop-reductions -enable-interleaved-mem-accesses=true -enable-masked-interleaved-mem-accesses -disable-output %s 2>&1 | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 
 ; Tests for printing VPlans.
 
 define void @print_call_and_memory(i64 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
 ; CHECK-LABEL: Checking a loop in "print_call_and_memory"
 ; CHECK:      VPlan 'Initial VPlan for VF={4},UF>=1' {
 ; CHECK-NEXT: Live-in vp<[[VEC_TC:%.+]]> = vector-trip-count
 ; CHECK-EMPTY:
 ; CHECK-NEXT: <x1> vector loop: {
 ; CHECK-NEXT: for.body:
 ; CHECK-NEXT:   EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION
-; CHECK-NEXT:   WIDEN-INDUCTION %iv = phi %iv.next, 0
+; CHECK-NEXT:   SCALAR-STEPS vp<[[CAN_IV]]>, ir<0>, ir<1>
 ; CHECK-NEXT:   CLONE ir<%arrayidx> = getelementptr ir<%y>, ir<%iv>
 ; CHECK-NEXT:   WIDEN ir<%lv> = load ir<%arrayidx>
 ; CHECK-NEXT:   WIDEN-CALL ir<%call> = call @llvm.sqrt.f32(ir<%lv>)
 ; CHECK-NEXT:   CLONE ir<%arrayidx2> = getelementptr ir<%x>, ir<%iv>
 ; CHECK-NEXT:   WIDEN store ir<%arrayidx2>, ir<%call>
 ; CHECK-NEXT:   EMIT vp<[[CAN_IV_NEXT:%.+]]> = VF * UF +(nuw) vp<[[CAN_IV]]>
 ; CHECK-NEXT:   EMIT branch-on-count vp<[[CAN_IV_NEXT]]> vp<[[VEC_TC]]>
 ; CHECK-NEXT: No successors

 define float @print_reduction(i64 %n, float* noalias %y) {
 ; CHECK-LABEL: Checking a loop in "print_reduction"
 ; CHECK:      VPlan 'Initial VPlan for VF={4},UF>=1' {
 ; CHECK-NEXT: Live-in vp<[[VEC_TC:%.+]]> = vector-trip-count
 ; CHECK-EMPTY:
 ; CHECK-NEXT: <x1> vector loop: {
 ; CHECK-NEXT: for.body:
 ; CHECK-NEXT:   EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION
-; CHECK-NEXT:   WIDEN-INDUCTION %iv = phi %iv.next, 0
 ; CHECK-NEXT:   WIDEN-REDUCTION-PHI ir<%red> = phi ir<0.000000e+00>, ir<%red.next>
+; CHECK-NEXT:   SCALAR-STEPS vp<[[CAN_IV]]>, ir<0>, ir<1>
 ; CHECK-NEXT:   CLONE ir<%arrayidx> = getelementptr ir<%y>, ir<%iv>
 ; CHECK-NEXT:   WIDEN ir<%lv> = load ir<%arrayidx>
 ; CHECK-NEXT:   REDUCE ir<%red.next> = ir<%red> + fast reduce.fadd (ir<%lv>)
 ; CHECK-NEXT:   EMIT vp<[[CAN_IV_NEXT:%.+]]> = VF * UF +(nuw) vp<[[CAN_IV]]>
 ; CHECK-NEXT:   EMIT branch-on-count vp<[[CAN_IV_NEXT]]> vp<[[VEC_TC]]>
 ; CHECK-NEXT: No successors
 ; CHECK-NEXT: }
 ; CHECK-NEXT: No successors

 define void @print_interleave_groups(i32 %C, i32 %D) {
 ; CHECK-LABEL: Checking a loop in "print_interleave_groups"
 ; CHECK:       VPlan 'Initial VPlan for VF={4},UF>=1' {
 ; CHECK-NEXT: Live-in vp<[[VEC_TC:%.+]]> = vector-trip-count
 ; CHECK-EMPTY:
 ; CHECK-NEXT: <x1> vector loop: {
 ; CHECK-NEXT:  for.body:
 ; CHECK-NEXT:   EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION
-; CHECK-NEXT:   WIDEN-INDUCTION %iv = phi 0, %iv.next
+; CHECK-NEXT:   SCALAR-STEPS vp<[[CAN_IV]]>, ir<0>, ir<4>
 ; CHECK-NEXT:   CLONE ir<%gep.AB.0> = getelementptr ir<@AB>, ir<0>, ir<%iv>
 ; CHECK-NEXT:   INTERLEAVE-GROUP with factor 4 at %AB.0, ir<%gep.AB.0>
 ; CHECK-NEXT:     ir<%AB.0> = load from index 0
 ; CHECK-NEXT:     ir<%AB.1> = load from index 1
 ; CHECK-NEXT:     ir<%AB.3> = load from index 3
 ; CHECK-NEXT:   CLONE ir<%iv.plus.3> = add ir<%iv>, ir<3>
 ; CHECK-NEXT:   WIDEN ir<%add> = add ir<%AB.0>, ir<%AB.1>
 ; CHECK-NEXT:   CLONE ir<%gep.CD.3> = getelementptr ir<@CD>, ir<0>, ir<%iv.plus.3>

 define float @print_fmuladd_strict(float* %a, float* %b, i64 %n) {
 ; CHECK-LABEL: Checking a loop in "print_fmuladd_strict"
 ; CHECK:      VPlan 'Initial VPlan for VF={4},UF>=1' {
 ; CHECK-NEXT: Live-in vp<[[VEC_TC:%.+]]> = vector-trip-count
 ; CHECK-EMPTY:
 ; CHECK-NEXT: <x1> vector loop: {
 ; CHECK-NEXT: for.body:
 ; CHECK-NEXT:   EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION
-; CHECK-NEXT:   WIDEN-INDUCTION %iv = phi 0, %iv.next
 ; CHECK-NEXT:   WIDEN-REDUCTION-PHI ir<%sum.07> = phi ir<0.000000e+00>, ir<%muladd>
+; CHECK-NEXT:   SCALAR-STEPS vp<[[CAN_IV]]>, ir<0>, ir<1>
 ; CHECK-NEXT:   CLONE ir<%arrayidx> = getelementptr ir<%a>, ir<%iv>
 ; CHECK-NEXT:   WIDEN ir<%l.a> = load ir<%arrayidx>
 ; CHECK-NEXT:   CLONE ir<%arrayidx2> = getelementptr ir<%b>, ir<%iv>
 ; CHECK-NEXT:   WIDEN ir<%l.b> = load ir<%arrayidx2>
 ; CHECK-NEXT:   EMIT vp<[[FMUL:%.]]> = fmul nnan ninf nsz ir<%l.a> ir<%l.b>
 ; CHECK-NEXT:   REDUCE ir<[[MULADD:%.+]]> = ir<%sum.07> + nnan ninf nsz reduce.fadd (vp<[[FMUL]]>)
 ; CHECK-NEXT:   EMIT vp<[[CAN_IV_NEXT:%.+]]> = VF * UF +(nuw) vp<[[CAN_IV]]>
 ; CHECK-NEXT:   EMIT branch-on-count vp<[[CAN_IV_NEXT]]> vp<[[VEC_TC]]>

 define void @debug_loc_vpinstruction(i32* nocapture %asd, i32* nocapture %bsd) !dbg !5 {
 ; CHECK-LABEL: Checking a loop in "debug_loc_vpinstruction"
 ; CHECK:    VPlan 'Initial VPlan for VF={4},UF>=1' {
 ; CHECK-NEXT: Live-in vp<[[VEC_TC:%.+]]> = vector-trip-count
 ; CHECK-EMPTY:
 ; CHECK-NEXT: <x1> vector loop: {
 ; CHECK-NEXT:  loop:
 ; CHECK-NEXT:    EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION
-; CHECK-NEXT:    WIDEN-INDUCTION %iv = phi 0, %iv.next
+; CHECK-NEXT:    SCALAR-STEPS vp<[[CAN_IV]]>, ir<0>, ir<1>
 ; CHECK-NEXT:    CLONE ir<%isd> = getelementptr ir<%asd>, ir<%iv>
 ; CHECK-NEXT:    WIDEN ir<%lsd> = load ir<%isd>
 ; CHECK-NEXT:    WIDEN ir<%psd> = add ir<%lsd>, ir<23>
 ; CHECK-NEXT:    WIDEN ir<%cmp1> = icmp ir<%lsd>, ir<100>
 ; CHECK-NEXT:  Successor(s): check
 ; CHECK-EMPTY:
 ; CHECK-NEXT:  check:
 ; CHECK-NEXT:    WIDEN ir<%cmp2> = icmp ir<%lsd>, ir<200>

llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge-vf1.ll

 ; REQUIRES: asserts
 
 ; RUN: opt -loop-vectorize -force-vector-interleave=2 -force-vector-width=1 -debug -disable-output %s 2>&1 | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 
 ; Make sure recipes with side-effects are not sunk.
 define void @sink_with_sideeffects(i1 %c, i8* %ptr) {
 ; CHECK-LABEL: sink_with_sideeffects
 ; CHECK:      VPlan 'Initial VPlan for VF={1},UF>=1' {
 ; CHECK-NEXT: Live-in vp<[[VEC_TC:%.+]]> = vector-trip-count
 ; CHECK-EMPTY:
 ; CHECK-NEXT: <x1> vector loop: {
 ; CHECK-NEXT: for.body:
 ; CHECK-NEXT:   EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION
-; CHECK-NEXT:   WIDEN-INDUCTION %tmp0 = phi %tmp6, 0
+; CHECK-NEXT:   ir<%tmp0> = SCALAR-STEPS vp<[[CAN_IV]]>, ir<0>, ir<1>
 ; CHECK-NEXT:   CLONE ir<%tmp2> = getelementptr ir<%ptr>, ir<%tmp0>
 ; CHECK-NEXT:   CLONE ir<%tmp3> = load ir<%tmp2>
 ; CHECK-NEXT:   CLONE store ir<0>, ir<%tmp2>
 ; CHECK-NEXT:   CLONE ir<%tmp4> = zext ir<%tmp3>
 ; CHECK-NEXT:   CLONE ir<%tmp5> = trunc ir<%tmp4>
 ; CHECK-NEXT: Successor(s): if.then
 
 ; CHECK:      if.then: