[VPlan] Implement initial vector code generation support for simple outer loops.

[sguggill]

Summary

[VPlan] Implement vector code generation support for simple outer loops.

Context: Patch Series #1 for outer loop vectorization support in LV using VPlan. (RFC: http://lists.llvm.org/pipermail/llvm-dev/2017-December/119523.html).

This patch introduces vector code generation support for simple outer loops that are currently supported in the VPlanNativePath. Changes here essentially do the following:

• force vector code generation using explicit vectorize_width
• add conservative early returns in cost model and other places for VPlanNativePath
• add code for setting up outer loop inductions
• support for widening non-induction PHIs that can result from inner loops and uniform conditional branches
• support for generating uniform inner branches

We plan to add a handful C outer loop executable tests once the initial code generation support is committed. This patch is expected to be NFC for the inner loop vectorizer path. Since we are moving in the direction of supporting outer loop vectorization in LV, it may also be time to rename classes such as InnerLoopVectorizer.

[fhahn]

Great to see this. I've left a few small comments after a first pass and will take a closer look in the next few days!

We plan to add a handful C outer loop executable tests once the initial code generation support is committed. This patch is expected to be NFC for the inner loop vectorizer path. Since we are moving in the direction of supporting outer loop vectorization in LV, it may also be time to rename classes such as InnerLoopVectorizer.

Excellent, having some test cases in the test-suite to exercise the VPlan native path will be great.

lib/Transforms/Vectorize/LoopVectorizationLegality.cpp

592

can we just use Header->phis() here? Using something like llvm::any_of/llvm::all_of might also make the code slightly simpler.

lib/Transforms/Vectorize/LoopVectorize.cpp

7807

Is it worth explicitly calling this out as FIXME/TODO? IIUC this could be done independently and I think it would be good to start preserving this sooner rather than later :)

lib/Transforms/Vectorize/VPlan.cpp

241

nit: if (!EnableVPlanNativePath) continue;?

[sguggill]

In D50820#1202465, @fhahn wrote:

Great to see this. I've left a few small comments after a first pass and will take a closer look in the next few days!

Thanks!

[sguggill]

lib/Transforms/Vectorize/LoopVectorizationLegality.cpp

592

I will make the change.

lib/Transforms/Vectorize/LoopVectorize.cpp

7807

Yes, I plan to do this independently soon afterwards. I will add a TODO.

lib/Transforms/Vectorize/VPlan.cpp

241

I hope I am not missing something here. We want to call execute for all blocks for non VPlan-native path and skip calling execute for outer loop's PH/exit blocks.

[sguggill]

Addressed initial comments from Florian.

[sguggill]

Minor formatting changes.

[sguggill]

Ping.

[fhahn]

lib/Transforms/Vectorize/LoopVectorize.cpp

2800

Should we ever hit this code path in the vplan native path at the moment? If not, I would turn it into an assert.

3857

nit: ScalarBBPredecessors be initialized in the constructor so we don't need this loop (Same for VectorBBPredecessors)? There's a constructor that takes an iterator range.

7186

Could we use unsigned VF = Range.Start; at the start of the for loop, so we do not need plan->addVF outside?

7400

as it is currently, we should only get to this point for outer loops, right? Then this could be an assert.

lib/Transforms/Vectorize/VPlan.cpp

403–405

This message needs to be updated I think. Something like 'Expected InnerLoop VPlan CFG to terminate with unreachable')

[sguggill]

Addressing latest comments from Florian.

[sguggill]

lib/Transforms/Vectorize/LoopVectorize.cpp

2800

We do hit this code in vplan native path and later hit a seg fault as getLAI returns NULL.

[sguggill]

lib/Transforms/Vectorize/LoopVectorize.cpp

7400

Yes, that is right. I have removed this code as the call to this function is guarded by a check for non-empty loops.

[fhahn]

I just tried to build your patch, and it triggers

lib/Transforms/Vectorize/LoopVectorize.cpp:1288: bool llvm::LoopVectorizationCostModel::isScalarAfterVectorization(llvm::Instruction*, unsigned int) const: Assertion `ScalarsPerVF != Scalars.end() && "Scalar values are not calculated for VF"' failed.

for the vplan tests.

lib/Transforms/Vectorize/LoopVectorize.cpp

2800

Right! I think for now it is better to exit early here then,

[fhahn]

In D50820#1223015, @fhahn wrote:

I just tried to build your patch, and it triggers

lib/Transforms/Vectorize/LoopVectorize.cpp:1288: bool llvm::LoopVectorizationCostModel::isScalarAfterVectorization(llvm::Instruction*, unsigned int) const: Assertion `ScalarsPerVF != Scalars.end() && "Scalar values are not calculated for VF"' failed.

for the vplan tests.

Nevermind, it was because the patch did not apply cleanly. Could you rebase the patch on latest trunk?

[sguggill]

Rebased my changes.

[fhahn]

LGTM with some more nits to the test cases. The EnableVPlanNativePath sprinkled around are not ideal, but I don't think there is a better alternative and we should be able to get rid of them as we go along. Please wait a few days with submitting this, to give other people a change to raise any additional concerns.

test/Transforms/LoopVectorize/outer_loop_test1.ll

18

Using CHECK-LABEL to check for blocks gives better error messages in case of a mismatch.

19

could you also add checks for the generated PHIs, increments for the induction vars, address calculation and exit condition check? Same for the second test.

26

This line and the one below can be dropped.

29

dso_local local_unnamed_addr can be dropped

33

dso_local and local_unnamed_addr #0 can be dropped, same for the comment above.

test/Transforms/LoopVectorize/outer_loop_test2.ll

38

IIUC the test does not rely on any X86 specifics, so this and the target triple can be dropped?

[sguggill]

In D50820#1224411, @fhahn wrote:

LGTM with some more nits to the test cases. The EnableVPlanNativePath sprinkled around are not ideal, but I don't think there is a better alternative and we should be able to get rid of them as we go along. Please wait a few days with submitting this, to give other people a change to raise any additional concerns.

Thanks for the approval Florian! I will take care of the test case changes and wait a few days to submit the changes.

[sguggill]

Florian, I have made the suggested test case changes and uploaded the same. Please take a look.

[rkruppe]

I think I found a typo, but otherwise LGTM too!

lib/Transforms/Vectorize/VPlan.cpp

392

Shouldn't this call getEntryBasicBlock()? If the successor is a region, the corresponding IR block should jump to the entry block of that region, not to its exit.

[sguggill]

Addressing Robin's comment.

[sguggill]

lib/Transforms/Vectorize/VPlan.cpp

392

Thank you for pointing this out. This was an oversight. The code was working as all successors were VPBasicBlocks in VPlanNativePath at this point. I have fixed this.

[sguggill]

I have requested Hideki Saito to submit this change on my behalf.

[fhahn]

lib/Transforms/Vectorize/VPlan.cpp

392

I am not sure if it is possible to construct a test case for that with the restrictions at the moment, but if it is, could you add a test case covering this?

[hsaito]

In D50820#1231293, @sguggill wrote:

I have requested Hideki Saito to submit this change on my behalf.

Status Update:
There are a few failures found in our internal testing. Waiting for @sguggill to look into it.

[hsaito]

Accepting the last revision prior to the commit. Hope arc will automatically attribute the patch to @sguggill.

[hsaito]

Fix for the buildbot failures, provided by @sguggill . There was a memory leak in the D50820 patch.

http://llvm.org/viewvc/llvm-project?view=revision&revision=342201

include/llvm/Transforms/Vectorize/LoopVectorizationLegality.h

private:
   /// its nested loops are considered legal for vectorization. These legal
   /// checks are common for inner and outer loop vectorization.
   /// Temporarily taking UseVPlanNativePath parameter. If true, take
   /// the new code path being implemented for outer loop vectorization
   /// (should be functional for inner loop vectorization) based on VPlan.
   /// If false, good old LV code.
   bool canVectorizeLoopNestCFG(Loop *Lp, bool UseVPlanNativePath);
 
+  /// Set up outer loop inductions by checking Phis in outer loop header for
+  /// supported inductions (int inductions). Return false if any of these Phis
+  /// is not a supported induction or if we fail to find an induction.
+  bool setupOuterLoopInductions();
+
   /// Return true if the pre-header, exiting and latch blocks of \p Lp
   /// (non-recursive) are considered legal for vectorization.
   /// Temporarily taking UseVPlanNativePath parameter. If true, take
   /// the new code path being implemented for outer loop vectorization
   /// (should be functional for inner loop vectorization) based on VPlan.
   /// If false, good old LV code.
   bool canVectorizeLoopCFG(Loop *Lp, bool UseVPlanNativePath);
 

lib/Transforms/Vectorize/LoopVectorizationLegality.cpp

if (!isUniformLoopNest(TheLoop /*loop nest*/,
     ORE->emit(createMissedAnalysis("CFGNotUnderstood")
               << "loop control flow is not understood by vectorizer");
     if (DoExtraAnalysis)
       Result = false;
     else
       return false;
   }
 
+  // Check whether we are able to set up outer loop induction.
+  if (!setupOuterLoopInductions()) {
+    LLVM_DEBUG(
+        dbgs() << "LV: Not vectorizing: Unsupported outer loop Phi(s).\n");
+    ORE->emit(createMissedAnalysis("UnsupportedPhi")
+              << "Unsupported outer loop Phi(s)");
+    if (DoExtraAnalysis)
+      Result = false;
+    else
+      return false;
+  }
+
   return Result;
 }
 
 void LoopVectorizationLegality::addInductionPhi(
     PHINode *Phi, const InductionDescriptor &ID,
     SmallPtrSetImpl<Value *> &AllowedExit) {
   Inductions[Phi] = ID;
 
void LoopVectorizationLegality::addInductionPhi(
   if (PSE.getUnionPredicate().isAlwaysTrue()) {
     AllowedExit.insert(Phi);
     AllowedExit.insert(Phi->getIncomingValueForBlock(TheLoop->getLoopLatch()));
   }
 
   LLVM_DEBUG(dbgs() << "LV: Found an induction variable.\n");
 }
 
+bool LoopVectorizationLegality::setupOuterLoopInductions() {
+  BasicBlock *Header = TheLoop->getHeader();
+
+  // Returns true if a given Phi is a supported induction.
+  auto isSupportedPhi = [&] (PHINode &Phi) -> bool {
+    InductionDescriptor ID;
+    if (InductionDescriptor::isInductionPHI(&Phi, TheLoop, PSE, ID) &&
+        ID.getKind() == InductionDescriptor::IK_IntInduction) {
+      addInductionPhi(&Phi, ID, AllowedExit);
+      return true;
+    } else {
+      // Bail out for any Phi in the outer loop header that is not a supported
+      // induction.
+      LLVM_DEBUG(dbgs() << "LV: Found unsupported PHI for outer loop vectorization.\n");
+      return false;
+    }
+  };
+
+  if (llvm::all_of(Header->phis(), isSupportedPhi))
+    return true;
+  else 
+    return false;
+}
+
 bool LoopVectorizationLegality::canVectorizeInstrs() {
   BasicBlock *Header = TheLoop->getHeader();
 
   // Look for the attribute signaling the absence of NaNs.
   Function &F = *Header->getParent();
   HasFunNoNaNAttr =
       F.getFnAttribute("no-nans-fp-math").getValueAsString() == "true";
 

lib/Transforms/Vectorize/LoopVectorize.cpp

 //  Vectorizing Compilers.
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Transforms/Vectorize/LoopVectorize.h"
 #include "LoopVectorizationPlanner.h"
 #include "VPRecipeBuilder.h"
 #include "VPlanHCFGBuilder.h"
+#include "VPlanHCFGTransforms.h"
 #include "llvm/ADT/APInt.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/DenseMapInfo.h"
 #include "llvm/ADT/Hashing.h"
 #include "llvm/ADT/MapVector.h"
 #include "llvm/ADT/None.h"
 #include "llvm/ADT/Optional.h"
static cl::opt<bool> EnableCondStoresVectorization(
     "enable-cond-stores-vec", cl::init(true), cl::Hidden,
     cl::desc("Enable if predication of stores during vectorization."));
 
 static cl::opt<unsigned> MaxNestedScalarReductionIC(
     "max-nested-scalar-reduction-interleave", cl::init(2), cl::Hidden,
     cl::desc("The maximum interleave count to use when interleaving a scalar "
              "reduction in a nested loop."));
 
-static cl::opt<bool> EnableVPlanNativePath(
+cl::opt<bool> EnableVPlanNativePath(
     "enable-vplan-native-path", cl::init(false), cl::Hidden,
     cl::desc("Enable VPlan-native vectorization path with "
              "support for outer loop vectorization."));
 
 // This flag enables the stress testing of the VPlan H-CFG construction in the
 // VPlan-native vectorization path. It must be used in conjuction with
 // -enable-vplan-native-path. -vplan-verify-hcfg can also be used to enable the
 // verification of the H-CFGs built.
public:
   /// operations if \p BlockInMask is non-null.
   void vectorizeMemoryInstruction(Instruction *Instr,
                                   VectorParts *BlockInMask = nullptr);
 
   /// Set the debug location in the builder using the debug location in
   /// the instruction.
   void setDebugLocFromInst(IRBuilder<> &B, const Value *Ptr);
 
+  /// Fix the non-induction PHIs in the OrigPHIsToFix vector.
+  void fixNonInductionPHIs(void);
+
 protected:
   friend class LoopVectorizationPlanner;
 
   /// A small list of PHINodes.
   using PhiVector = SmallVector<PHINode *, 4>;
 
   /// A type for scalarized values in the new loop. Each value from the
   /// original loop, when scalarized, is represented by UF x VF scalar values
protected:
   LoopVectorizationCostModel *Cost;
 
   // Record whether runtime checks are added.
   bool AddedSafetyChecks = false;
 
   // Holds the end values for each induction variable. We save the end values
   // so we can later fix-up the external users of the induction variables.
   DenseMap<PHINode *, Value *> IVEndValues;
+
+  // Vector of original scalar PHIs whose corresponding widened PHIs need to be
+  // fixed up at the end of vector code generation.
+  SmallVector<PHINode *, 8> OrigPHIsToFix;
 };
 
 class InnerLoopUnroller : public InnerLoopVectorizer {
 public:
   InnerLoopUnroller(Loop *OrigLoop, PredicatedScalarEvolution &PSE,
                     LoopInfo *LI, DominatorTree *DT,
                     const TargetLibraryInfo *TLI,
                     const TargetTransformInfo *TTI, AssumptionCache *AC,
public:
   const MapVector<Instruction *, uint64_t> &getMinimalBitwidths() const {
     return MinBWs;
   }
 
   /// \returns True if it is more profitable to scalarize instruction \p I for
   /// vectorization factor \p VF.
   bool isProfitableToScalarize(Instruction *I, unsigned VF) const {
     assert(VF > 1 && "Profitable to scalarize relevant only for VF > 1.");
+
+    // Cost model is not run in the VPlan-native path - return conservative
+    // result until this changes.
+    if (EnableVPlanNativePath)
+      return false;
+
     auto Scalars = InstsToScalarize.find(VF);
     assert(Scalars != InstsToScalarize.end() &&
            "VF not yet analyzed for scalarization profitability");
     return Scalars->second.count(I);
   }
 
   /// Returns true if \p I is known to be uniform after vectorization.
   bool isUniformAfterVectorization(Instruction *I, unsigned VF) const {
     if (VF == 1)
       return true;
+
+    // Cost model is not run in the VPlan-native path - return conservative
+    // result until this changes.
+    if (EnableVPlanNativePath)
+      return false;
+
     assert(Uniforms.count(VF) && "VF not yet analyzed for uniformity");
     auto UniformsPerVF = Uniforms.find(VF);
     return UniformsPerVF->second.count(I);
   }
 
   /// Returns true if \p I is known to be scalar after vectorization.
   bool isScalarAfterVectorization(Instruction *I, unsigned VF) const {
     if (VF == 1)
       return true;
+
+    // Cost model is not run in the VPlan-native path - return conservative
+    // result until this changes.
+    if (EnableVPlanNativePath)
+      return false;
+
     assert(Scalars.count(VF) && "Scalar values are not calculated for VF");
     auto ScalarsPerVF = Scalars.find(VF);
     return ScalarsPerVF->second.count(I);
   }
 
   /// \returns True if instruction \p I can be truncated to a smaller bitwidth
   /// for vectorization factor \p VF.
   bool canTruncateToMinimalBitwidth(Instruction *I, unsigned VF) const {
void setWideningDecision(const InterleaveGroup *Grp, unsigned VF,
     }
   }
 
   /// Return the cost model decision for the given instruction \p I and vector
   /// width \p VF. Return CM_Unknown if this instruction did not pass
   /// through the cost modeling.
   InstWidening getWideningDecision(Instruction *I, unsigned VF) {
     assert(VF >= 2 && "Expected VF >=2");
+
+    // Cost model is not run in the VPlan-native path - return conservative
+    // result until this changes.
+    if (EnableVPlanNativePath)
+      return CM_GatherScatter;
+
     std::pair<Instruction *, unsigned> InstOnVF = std::make_pair(I, VF);
     auto Itr = WideningDecisions.find(InstOnVF);
     if (Itr == WideningDecisions.end())
       return CM_Unknown;
     return Itr->second.first;
   }
 
   /// Return the vectorization cost for the given instruction \p I and vector
void getAnalysisUsage(AnalysisUsage &AU) const override {
     AU.addRequired<DominatorTreeWrapperPass>();
     AU.addRequired<LoopInfoWrapperPass>();
     AU.addRequired<ScalarEvolutionWrapperPass>();
     AU.addRequired<TargetTransformInfoWrapperPass>();
     AU.addRequired<AAResultsWrapperPass>();
     AU.addRequired<LoopAccessLegacyAnalysis>();
     AU.addRequired<DemandedBitsWrapperPass>();
     AU.addRequired<OptimizationRemarkEmitterWrapperPass>();
+
+    // We currently do not preserve loopinfo/dominator analyses with outer loop
+    // vectorization. Until this is addressed, mark these analyses as preserved
+    // only for non-VPlan-native path.
+    // TODO: Preserve Loop and Dominator analyses for VPlan-native path.
+    if (!EnableVPlanNativePath) {
       AU.addPreserved<LoopInfoWrapperPass>();
       AU.addPreserved<DominatorTreeWrapperPass>();
+    }
+
     AU.addPreserved<BasicAAWrapperPass>();
     AU.addPreserved<GlobalsAAWrapperPass>();
   }
 };
 
 } // end anonymous namespace
 
 //===----------------------------------------------------------------------===//
void InnerLoopVectorizer::buildScalarSteps(Value *ScalarIV, Value *Step,
   }
 }
 
 Value *InnerLoopVectorizer::getOrCreateVectorValue(Value *V, unsigned Part) {
   assert(V != Induction && "The new induction variable should not be used.");
   assert(!V->getType()->isVectorTy() && "Can't widen a vector");
   assert(!V->getType()->isVoidTy() && "Type does not produce a value");
 
-  // If we have a stride that is replaced by one, do it here.
-  if (Legal->hasStride(V))
+  // If we have a stride that is replaced by one, do it here. Defer this for
+  // the VPlan-native path until we start running Legal checks in that path.
+  if (!EnableVPlanNativePath && Legal->hasStride(V))
     V = ConstantInt::get(V->getType(), 1);
 
   // If we have a vector mapped to this value, return it.
   if (VectorLoopValueMap.hasVectorValue(V, Part))
     return VectorLoopValueMap.getVectorValue(V, Part);
 
   // If the value has not been vectorized, check if it has been scalarized
   // instead. If it has been scalarized, and we actually need the value in
if (L->getParentLoop())
     L->getParentLoop()->addBasicBlockToLoop(NewBB, *LI);
   ReplaceInstWithInst(BB->getTerminator(),
                       BranchInst::Create(Bypass, NewBB, SCEVCheck));
   LoopBypassBlocks.push_back(BB);
   AddedSafetyChecks = true;
 }
 
 void InnerLoopVectorizer::emitMemRuntimeChecks(Loop *L, BasicBlock *Bypass) {
+  // VPlan-native path does not do any analysis for runtime checks currently.
+  if (EnableVPlanNativePath)
+    return;
+
   BasicBlock *BB = L->getLoopPreheader();
 
   // Generate the code that checks in runtime if arrays overlap. We put the
   // checks into a separate block to make the more common case of few elements
   // faster.
   Instruction *FirstCheckInst;
   Instruction *MemRuntimeCheck;
   std::tie(FirstCheckInst, MemRuntimeCheck) =

 }
 
 void InnerLoopVectorizer::fixVectorizedLoop() {
   // Insert truncates and extends for any truncated instructions as hints to
   // InstCombine.
   if (VF > 1)
     truncateToMinimalBitwidths();
 
+  // Fix widened non-induction PHIs by setting up the PHI operands.
+  if (OrigPHIsToFix.size()) {
+    assert(EnableVPlanNativePath &&
+           "Unexpected non-induction PHIs for fixup in non VPlan-native path");
+    fixNonInductionPHIs();
+  }
+
   // At this point every instruction in the original loop is widened to a
   // vector form. Now we need to fix the recurrences in the loop. These PHI
   // nodes are currently empty because we did not want to introduce cycles.
   // This is the second stage of vectorizing recurrences.
   fixCrossIterationPHIs();
 
   // Update the dominator tree.
   //
while (!Worklist.empty()) {
 
       // The sinking may have enabled other instructions to be sunk, so we will
       // need to iterate.
       Changed = true;
     }
   } while (Changed);
 }
 
+void InnerLoopVectorizer::fixNonInductionPHIs() {
+  for (PHINode *OrigPhi : OrigPHIsToFix) {
+    PHINode *NewPhi =
+        cast<PHINode>(VectorLoopValueMap.getVectorValue(OrigPhi, 0));
+    unsigned NumIncomingValues = OrigPhi->getNumIncomingValues();
+
+    SmallVector<BasicBlock *, 2> ScalarBBPredecessors;
+    for (BasicBlock *BB : predecessors(OrigPhi->getParent()))
+      ScalarBBPredecessors.push_back(BB);
+
+    SmallVector<BasicBlock *, 2> VectorBBPredecessors;
+    for (BasicBlock *BB : predecessors(NewPhi->getParent()))
+      VectorBBPredecessors.push_back(BB);
+
+    assert(ScalarBBPredecessors.size() == VectorBBPredecessors.size() &&
+           "Scalar and Vector BB should have the same number of predecessors");
+
+    // The insertion point in Builder may be invalidated by the time we get
+    // here. Force the Builder insertion point to something valid so that we do
+    // not run into issues during insertion point restore in
+    // getOrCreateVectorValue calls below.
+    Builder.SetInsertPoint(NewPhi);
+
+    // The predecessor order is preserved and we can rely on mapping between
+    // scalar and vector block predecessors.
+    for (unsigned i = 0; i < NumIncomingValues; ++i) {
+      BasicBlock *NewPredBB = VectorBBPredecessors[i];
+
+      // When looking up the new scalar/vector values to fix up, use incoming
+      // values from original phi.
+      Value *ScIncV =
+          OrigPhi->getIncomingValueForBlock(ScalarBBPredecessors[i]);
+
+      // Scalar incoming value may need a broadcast
+      Value *NewIncV = getOrCreateVectorValue(ScIncV, 0);
+      NewPhi->addIncoming(NewIncV, NewPredBB);
+    }
+  }
+}
+
 void InnerLoopVectorizer::widenPHIInstruction(Instruction *PN, unsigned UF,
                                               unsigned VF) {
+  PHINode *P = cast<PHINode>(PN);
+  if (EnableVPlanNativePath) {
+    // Currently we enter here in the VPlan-native path for non-induction
+    // PHIs where all control flow is uniform. We simply widen these PHIs.
+    // Create a vector phi with no operands - the vector phi operands will be
+    // set at the end of vector code generation.
+    Type *VecTy =
+        (VF == 1) ? PN->getType() : VectorType::get(PN->getType(), VF);
+    Value *VecPhi = Builder.CreatePHI(VecTy, PN->getNumOperands(), "vec.phi");
+    VectorLoopValueMap.setVectorValue(P, 0, VecPhi);
+    OrigPHIsToFix.push_back(P);
+
+    return;
+  }
+
   assert(PN->getParent() == OrigLoop->getHeader() &&
          "Non-header phis should have been handled elsewhere");
 
-  PHINode *P = cast<PHINode>(PN);
   // In order to support recurrences we need to be able to vectorize Phi nodes.
   // Phi nodes have cycles, so we need to vectorize them in two stages. This is
   // stage #1: We create a new vector PHI node with no incoming edges. We'll use
   // this value when we vectorize all of the instructions that use the PHI.
   if (Legal->isReductionVariable(P) || Legal->isFirstOrderRecurrence(P)) {
     for (unsigned Part = 0; Part < UF; ++Part) {
       // This is phase one of vectorizing PHIs.
       Type *VecTy =
default:
     llvm_unreachable("Unhandled instruction!");
   } // end of switch.
 }
 
 void InnerLoopVectorizer::updateAnalysis() {
   // Forget the original basic block.
   PSE.getSE()->forgetLoop(OrigLoop);
 
+  // DT is not kept up-to-date for outer loop vectorization
+  if (EnableVPlanNativePath)
+    return;
+
   // Update the dominator tree information.
   assert(DT->properlyDominates(LoopBypassBlocks.front(), LoopExitBlock) &&
          "Entry does not dominate exit.");
 
   DT->addNewBlock(LoopMiddleBlock,
                   LI->getLoopFor(LoopVectorBody)->getLoopLatch());
   DT->addNewBlock(LoopScalarPreHeader, LoopBypassBlocks[0]);
   DT->changeImmediateDominator(LoopScalarBody, LoopScalarPreHeader);
LoopVectorizationPlanner::buildVPlan(VFRange &Range) {
 
   // Create new empty VPlan
   auto Plan = llvm::make_unique<VPlan>();
 
   // Build hierarchical CFG
   VPlanHCFGBuilder HCFGBuilder(OrigLoop, LI, *Plan);
   HCFGBuilder.buildHierarchicalCFG();
 
+  SmallPtrSet<Instruction *, 1> DeadInstructions;
+  VPlanHCFGTransforms::VPInstructionsToVPRecipes(
+      Plan, Legal->getInductionVars(), DeadInstructions);
+
+  unsigned VF = Range.Start;
+  Plan->addVF(VF);
+  for (VF *= 2; VF < Range.End; VF *= 2) {
+    Plan->addVF(VF);
+  }
   return Plan;
 }
 
 Value* LoopVectorizationPlanner::VPCallbackILV::
 getOrCreateVectorValues(Value *V, unsigned Part) {
       return ILV.getOrCreateVectorValue(V, Part);
 }
 
static bool processLoopInVPlanNativePath(
   unsigned UserVF = Hints.getWidth();
 
   // Check the function attributes to find out if this function should be
   // optimized for size.
   bool OptForSize =
       Hints.getForce() != LoopVectorizeHints::FK_Enabled && F->optForSize();
 
   // Plan how to best vectorize, return the best VF and its cost.
-  LVP.planInVPlanNativePath(OptForSize, UserVF);
+  VectorizationFactor VF = LVP.planInVPlanNativePath(OptForSize, UserVF);
 
-  // Returning false. We are currently not generating vector code in the VPlan
-  // native path.
+  // If we are stress testing VPlan builds, do not attempt to generate vector
+  // code.
+  if (VPlanBuildStressTest)
     return false;
+
+  // Limit vector code generation to supported explicit outer loops in native
+  // path. By the time we get here, we have already done checks for supported
+  // outer loops - a check for empty loop is sufficient here.
+  if (L->empty())
+    return false;
+
+  LVP.setBestPlan(VF.Width, 1);
+
+  InnerLoopVectorizer LB(L, PSE, LI, DT, TLI, TTI, AC, ORE, UserVF, 1, LVL,
+                         &CM);
+  LLVM_DEBUG(dbgs() << "Vectorizing outer loop in \""
+                    << L->getHeader()->getParent()->getName() << "\"\n");
+  LVP.executePlan(LB, DT);
+
+  // Mark the loop as already vectorized to avoid vectorizing again.
+  Hints.setAlreadyVectorized();
+
+  LLVM_DEBUG(verifyFunction(*L->getHeader()->getParent()));
+  return true;
 }
 
 bool LoopVectorizePass::processLoop(Loop *L) {
   assert((EnableVPlanNativePath || L->empty()) &&
          "VPlan-native path is not enabled. Only process inner loops.");
 
 #ifndef NDEBUG
   const std::string DebugLocStr = getDebugLocString(L);
std::function<const LoopAccessInfo &(Loop &)> GetLAA =
       LoopStandardAnalysisResults AR = {AA, AC, DT, LI, SE, TLI, TTI, nullptr};
       return LAM.getResult<LoopAccessAnalysis>(L, AR);
     };
     bool Changed =
         runImpl(F, SE, LI, TTI, DT, BFI, &TLI, DB, AA, AC, GetLAA, ORE);
     if (!Changed)
       return PreservedAnalyses::all();
     PreservedAnalyses PA;
+
+    // We currently do not preserve loopinfo/dominator analyses with outer loop
+    // vectorization. Until this is addressed, mark these analyses as preserved
+    // only for non-VPlan-native path.
+    // TODO: Preserve Loop and Dominator analyses for VPlan-native path.
+    if (!EnableVPlanNativePath) {
       PA.preserve<LoopAnalysis>();
       PA.preserve<DominatorTreeAnalysis>();
+    }
     PA.preserve<BasicAA>();
     PA.preserve<GlobalsAA>();
     return PA;
 }

lib/Transforms/Vectorize/VPlan.h

struct CFGState {
     /// The last IR BasicBlock in the output IR. Set to the new latch
     /// BasicBlock, used for placing the newly created BasicBlocks.
     BasicBlock *LastBB = nullptr;
 
     /// A mapping of each VPBasicBlock to the corresponding BasicBlock. In case
     /// of replication, maps the BasicBlock of the last replica created.
     SmallDenseMap<VPBasicBlock *, BasicBlock *> VPBB2IRBB;
 
+    /// Vector of VPBasicBlocks whose terminator instruction needs to be fixed
+    /// up at the end of vector code generation.
+    SmallVector<VPBasicBlock *, 8> VPBBsToFix;
+
     CFGState() = default;
   } CFG;
 
   /// Hold a pointer to LoopInfo to register new basic blocks in the loop.
   LoopInfo *LI;
 
   /// Hold a pointer to Dominator Tree to register new basic blocks in the loop.
   DominatorTree *DT;

lib/Transforms/Vectorize/VPlan.cpp

 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include <cassert>
 #include <iterator>
 #include <string>
 #include <vector>
 
 using namespace llvm;
+extern cl::opt<bool> EnableVPlanNativePath;
 
 #define DEBUG_TYPE "vplan"
 
 raw_ostream &llvm::operator<<(raw_ostream &OS, const VPValue &V) {
   if (const VPInstruction *Instr = dyn_cast<VPInstruction>(&V))
     Instr->print(OS);
   else
     V.printAsOperand(OS);
BasicBlock *NewBB = BasicBlock::Create(PrevBB->getContext(), getName(),
                                          PrevBB->getParent(), CFG.LastBB);
   LLVM_DEBUG(dbgs() << "LV: created " << NewBB->getName() << '\n');
 
   // Hook up the new basic block to its predecessors.
   for (VPBlockBase *PredVPBlock : getHierarchicalPredecessors()) {
     VPBasicBlock *PredVPBB = PredVPBlock->getExitBasicBlock();
     auto &PredVPSuccessors = PredVPBB->getSuccessors();
     BasicBlock *PredBB = CFG.VPBB2IRBB[PredVPBB];
+
+    // In outer loop vectorization scenario, the predecessor BBlock may not yet
+    // be visited(backedge). Mark the VPBasicBlock for fixup at the end of
+    // vectorization. We do not encounter this case in inner loop vectorization
+    // as we start out by building a loop skeleton with the vector loop header
+    // and latch blocks. As a result, we never enter this function for the
+    // header block in the non VPlan-native path.
+    if (!PredBB) {
+      assert(EnableVPlanNativePath &&
+             "Unexpected null predecessor in non VPlan-native path");
+      CFG.VPBBsToFix.push_back(PredVPBB);
+      continue;
+    }
+
     assert(PredBB && "Predecessor basic-block not found building successor.");
     auto *PredBBTerminator = PredBB->getTerminator();
     LLVM_DEBUG(dbgs() << "LV: draw edge from" << PredBB->getName() << '\n');
     if (isa<UnreachableInst>(PredBBTerminator)) {
       assert(PredVPSuccessors.size() == 1 &&
              "Predecessor ending w/o branch must have single successor.");
       PredBBTerminator->eraseFromParent();
       BranchInst::Create(NewBB, PredBB);
LLVM_DEBUG(dbgs() << "LV: vectorizing VPBB:" << getName()
                     << " in BB:" << NewBB->getName() << '\n');
 
   State->CFG.VPBB2IRBB[this] = NewBB;
   State->CFG.PrevVPBB = this;
 
   for (VPRecipeBase &Recipe : Recipes)
     Recipe.execute(*State);
 
+  VPValue *CBV;
+  if (EnableVPlanNativePath && (CBV = getCondBit())) {
+    Value *IRCBV = CBV->getUnderlyingValue();
+    assert(IRCBV && "Unexpected null underlying value for condition bit");
+
+    // Delete the condition bit at this point - it should be no longer needed.
+    delete CBV;
+    setCondBit(nullptr);
+
+    // Condition bit value in a VPBasicBlock is used as the branch selector. In
+    // the VPlan-native path case, since all branches are uniform we generate a
+    // branch instruction using the condition value from vector lane 0 and dummy
+    // successors. The successors are fixed later when the successor blocks are
+    // visited.
+    Value *NewCond = State->Callback.getOrCreateVectorValues(IRCBV, 0);
+    NewCond = State->Builder.CreateExtractElement(NewCond,
+                                                  State->Builder.getInt32(0));
+
+    // Replace the temporary unreachable terminator with the new conditional
+    // branch.
+    auto *CurrentTerminator = NewBB->getTerminator();
+    assert(isa<UnreachableInst>(CurrentTerminator) &&
+           "Expected to replace unreachable terminator with conditional "
+           "branch.");
+    auto *CondBr = BranchInst::Create(NewBB, nullptr, NewCond);
+    CondBr->setSuccessor(0, nullptr);
+    ReplaceInstWithInst(CurrentTerminator, CondBr);
+  }
+
   LLVM_DEBUG(dbgs() << "LV: filled BB:" << *NewBB);
 }
 
 void VPRegionBlock::execute(VPTransformState *State) {
   ReversePostOrderTraversal<VPBlockBase *> RPOT(Entry);
 
   if (!isReplicator()) {
     // Visit the VPBlocks connected to "this", starting from it.
     for (VPBlockBase *Block : RPOT) {
+      if (EnableVPlanNativePath) {
+        // The inner loop vectorization path does not represent loop preheader
+        // and exit blocks as part of the VPlan. In the VPlan-native path, skip
+        // vectorizing loop preheader block. In future, we may replace this
+        // check with the check for loop preheader.
+        if (Block->getNumPredecessors() == 0)
+          continue;
+
+        // Skip vectorizing loop exit block. In future, we may replace this
+        // check with the check for loop exit.
+        if (Block->getNumSuccessors() == 0)
+          continue;
+      }
+
       LLVM_DEBUG(dbgs() << "LV: VPBlock in RPO " << Block->getName() << '\n');
       Block->execute(State);
     }
     return;
   }
 
   assert(!State->Instance && "Replicating a Region with non-null instance.");
 
void VPlan::execute(VPTransformState *State) {
   // 2. Generate code in loop body.
   State->CFG.PrevVPBB = nullptr;
   State->CFG.PrevBB = VectorHeaderBB;
   State->CFG.LastBB = VectorLatchBB;
 
   for (VPBlockBase *Block : depth_first(Entry))
     Block->execute(State);
 
+  // Setup branch terminator successors for VPBBs in VPBBsToFix based on
+  // VPBB's successors.
+  for (auto VPBB : State->CFG.VPBBsToFix) {
+    assert(EnableVPlanNativePath &&
+           "Unexpected VPBBsToFix in non VPlan-native path");
+    BasicBlock *BB = State->CFG.VPBB2IRBB[VPBB];
+    assert(BB && "Unexpected null basic block for VPBB");
+
+    unsigned Idx = 0;
+    auto *BBTerminator = BB->getTerminator();
+
+    for (VPBlockBase *SuccVPBlock : VPBB->getHierarchicalSuccessors()) {
+      VPBasicBlock *SuccVPBB = SuccVPBlock->getExitBasicBlock();
+      BBTerminator->setSuccessor(Idx, State->CFG.VPBB2IRBB[SuccVPBB]);
+      ++Idx;
+    }
+  }
+
   // 3. Merge the temporary latch created with the last basic-block filled.
   BasicBlock *LastBB = State->CFG.PrevBB;
   // Connect LastBB to VectorLatchBB to facilitate their merge.
-  assert(isa<UnreachableInst>(LastBB->getTerminator()) &&
+  assert((EnableVPlanNativePath ||
+          isa<UnreachableInst>(LastBB->getTerminator())) &&
          "Expected VPlan CFG to terminate with unreachable");
+  assert((!EnableVPlanNativePath || isa<BranchInst>(LastBB->getTerminator())) &&
+         "Expected VPlan CFG to terminate with branch instruction");
   LastBB->getTerminator()->eraseFromParent();
   BranchInst::Create(VectorLatchBB, LastBB);
 
   // Merge LastBB with Latch.
   bool Merged = MergeBlockIntoPredecessor(VectorLatchBB, nullptr, State->LI);
   (void)Merged;
   assert(Merged && "Could not merge last basic block with latch.");
   VectorLatchBB = LastBB;
 
+  // We do not attempt to preserve DT for outer loop vectorization currently.
+  if (!EnableVPlanNativePath)
     updateDominatorTree(State->DT, VectorPreHeaderBB, VectorLatchBB);
 }
 
 void VPlan::updateDominatorTree(DominatorTree *DT, BasicBlock *LoopPreHeaderBB,
                                 BasicBlock *LoopLatchBB) {
   BasicBlock *LoopHeaderBB = LoopPreHeaderBB->getSingleSuccessor();
   assert(LoopHeaderBB && "Loop preheader does not have a single successor.");
   DT->addNewBlock(LoopHeaderBB, LoopPreHeaderBB);
   // The vector body may be more than a single basic-block by this point.

lib/Transforms/Vectorize/VPlanHCFGTransforms.cpp

 
 void VPlanHCFGTransforms::VPInstructionsToVPRecipes(
     VPlanPtr &Plan,
     LoopVectorizationLegality::InductionList *Inductions,
     SmallPtrSetImpl<Instruction *> &DeadInstructions) {
 
   VPRegionBlock *TopRegion = dyn_cast<VPRegionBlock>(Plan->getEntry());
   ReversePostOrderTraversal<VPBlockBase *> RPOT(TopRegion->getEntry());
+
+  // Condition bit VPValues get deleted during transformation to VPRecipes.
+  // Create new VPValues and save away as condition bits. These will be deleted
+  // after finalizing the vector IR basic blocks.
+  for (VPBlockBase *Base : RPOT) {
+    VPBasicBlock *VPBB = Base->getEntryBasicBlock();
+    if (auto *CondBit = VPBB->getCondBit()) {
+      auto *NCondBit = new VPValue(CondBit->getUnderlyingValue());
+      VPBB->setCondBit(NCondBit);
+    }
+  }
   for (VPBlockBase *Base : RPOT) {
     // Do not widen instructions in pre-header and exit blocks.
     if (Base->getNumPredecessors() == 0 || Base->getNumSuccessors() == 0)
       continue;
 
     VPBasicBlock *VPBB = Base->getEntryBasicBlock();
     VPRecipeBase *LastRecipe = nullptr;
     // Introduce each ingredient into VPlan.

lib/Transforms/Vectorize/VPlanValue.h

 class VPUser;
 
 // This is the base class of the VPlan Def/Use graph, used for modeling the data
 // flow into, within and out of the VPlan. VPValues can stand for live-ins
 // coming from the input IR, instructions which VPlan will generate if executed
 // and live-outs which the VPlan will need to fix accordingly.
 class VPValue {
   friend class VPBuilder;
+  friend class VPlanHCFGTransforms;
+  friend class VPBasicBlock;
+
 private:
   const unsigned char SubclassID; ///< Subclass identifier (for isa/dyn_cast).
 
   SmallVector<VPUser *, 1> Users;
 
 protected:
   // Hold the underlying Value, if any, attached to this VPValue.
   Value *UnderlyingVal;

test/Transforms/LoopVectorize/outer_loop_test1.ll

+; extern int arr[8][8];
+; extern int arr2[8];
+;
+; void foo(int n)
+; {
+;   int i1, i2;
+;
+; #pragma clang loop vectorize(enable) vectorize_width(4)
+;   for (i1 = 0; i1 < 8; i1++) {
+;     arr2[i1] = i1;
+;     for (i2 = 0; i2 < 8; i2++)
+;       arr[i2][i1] = i1 + n;
+;   }
+; }
+;
+; RUN: opt -S -loop-vectorize -enable-vplan-native-path < %s | FileCheck %s
+; CHECK: vector.body:
+; CHECK: call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32>
+; CHECK: br label %[[for_loop:.+]]
+; CHECK: [[for_loop]]:
+; CHECK: call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32>
+; CHECK: %[[cond0:.+]] = extractelement {{.*}}, i32 0
+; CHECK: br i1 %[[cond0]], {{.*}}label %[[for_loop]]
+; CHECK: br i1 {{.*}} label %vector.body
+; ModuleID = 't1.c'
+source_filename = "t1.c"
+
+@arr2 = external dso_local local_unnamed_addr global [8 x i32], align 16
+@arr = external dso_local local_unnamed_addr global [8 x [8 x i32]], align 16
+
+; Function Attrs: norecurse nounwind uwtable
+define dso_local void @foo(i32 %n) local_unnamed_addr #0 {
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.inc8, %entry
+  %indvars.iv21 = phi i64 [ 0, %entry ], [ %indvars.iv.next22, %for.inc8 ]
+  %arrayidx = getelementptr inbounds [8 x i32], [8 x i32]* @arr2, i64 0, i64 %indvars.iv21
+  %0 = trunc i64 %indvars.iv21 to i32
+  store i32 %0, i32* %arrayidx, align 4
+  %1 = trunc i64 %indvars.iv21 to i32
+  %add = add nsw i32 %1, %n
+  br label %for.body3
+
+for.body3:                                        ; preds = %for.body3, %for.body
+  %indvars.iv = phi i64 [ 0, %for.body ], [ %indvars.iv.next, %for.body3 ]
+  %arrayidx7 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* @arr, i64 0, i64 %indvars.iv, i64 %indvars.iv21
+  store i32 %add, i32* %arrayidx7, align 4
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, 8
+  br i1 %exitcond, label %for.inc8, label %for.body3
+
+for.inc8:                                         ; preds = %for.body3
+  %indvars.iv.next22 = add nuw nsw i64 %indvars.iv21, 1
+  %exitcond23 = icmp eq i64 %indvars.iv.next22, 8
+  br i1 %exitcond23, label %for.end10, label %for.body, !llvm.loop !1
+
+for.end10:                                        ; preds = %for.inc8
+  ret void
+}
+
+!1 = distinct !{!1, !2, !3}
+!2 = !{!"llvm.loop.vectorize.width", i32 4}
+!3 = !{!"llvm.loop.vectorize.enable", i1 true}

test/Transforms/LoopVectorize/outer_loop_test2.ll

+; int A[1024], B[1024];
+;
+; void foo(int iCount, int c, int jCount)
+; {
+;
+;   int i, j;
+;
+; #pragma clang loop vectorize(enable) vectorize_width(4)
+;   for (i = 0; i < iCount; i++) {
+;     A[i] = c;
+;     for (j = 0; j < jCount; j++) {
+;       A[i] += B[j] + i;
+;     }
+;   }
+; }
+; RUN: opt -S -loop-vectorize -enable-vplan-native-path < %s | FileCheck %s
+; CHECK: %[[ZeroTripChk:.*]] = icmp sgt i32 %jCount, 0
+; CHECK: vector.ph:
+; CHECK: %[[CVal0:.*]] = insertelement <4 x i32> undef, i32 %c, i32 0
+; CHECK-NEXT: %[[CSplat:.*]] = shufflevector <4 x i32> %[[CVal0]], <4 x i32> undef, <4 x i32> zeroinitializer
+; CHECK: %[[ZVal0:.*]] = insertelement <4 x i1> undef, i1 %[[ZeroTripChk]], i32 0
+; CHECK-NEXT: %[[ZSplat:.*]] = shufflevector <4 x i1> %[[ZVal0]], <4 x i1> undef, <4 x i32> zeroinitializer
+; CHECK: vector.body:
+; CHECK: %[[VecInd:.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %vector.ph ]
+; CHECK: %[[AAddr:.*]] = getelementptr inbounds [1024 x i32], [1024 x i32]* @A, i64 0, <4 x i64> %[[VecInd]]
+; CHECK: call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> %[[CSplat]], <4 x i32*> %[[AAddr]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
+; CHECK: %[[ZCmpExtr:.*]] = extractelement <4 x i1> %[[ZSplat]], i32 0
+; CHECK: br i1 %[[ZCmpExtr]], label %[[InnerForPh:.*]], label %[[OuterInc:.*]]
+; CHECK: [[InnerForPh]]:
+; CHECK: br label %[[InnerForBody:.*]]
+; CHECK: [[InnerForBody]]:
+; CHECK: %[[cond0:.+]] = extractelement {{.*}}, i32 0
+; CHECK: br i1 %[[cond0]], {{.*}}, label %[[InnerForBody]]
+; CHECK: br i1 {{.*}} label %vector.body
+; ModuleID = 's1.c'
+source_filename = "s1.c"
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@A = common dso_local local_unnamed_addr global [1024 x i32] zeroinitializer, align 16
+@B = common dso_local local_unnamed_addr global [1024 x i32] zeroinitializer, align 16
+
+; Function Attrs: norecurse nounwind uwtable
+define dso_local void @foo(i32 %iCount, i32 %c, i32 %jCount) local_unnamed_addr #0 {
+entry:
+  %cmp22 = icmp sgt i32 %iCount, 0
+  br i1 %cmp22, label %for.body.lr.ph, label %for.end11
+
+for.body.lr.ph:                                   ; preds = %entry
+  %cmp220 = icmp sgt i32 %jCount, 0
+  %wide.trip.count = zext i32 %jCount to i64
+  %wide.trip.count27 = zext i32 %iCount to i64
+  br label %for.body
+
+for.body:                                         ; preds = %for.inc9, %for.body.lr.ph
+  %indvars.iv25 = phi i64 [ 0, %for.body.lr.ph ], [ %indvars.iv.next26, %for.inc9 ]
+  %arrayidx = getelementptr inbounds [1024 x i32], [1024 x i32]* @A, i64 0, i64 %indvars.iv25
+  store i32 %c, i32* %arrayidx, align 4
+  br i1 %cmp220, label %for.body3.lr.ph, label %for.inc9
+
+for.body3.lr.ph:                                  ; preds = %for.body
+  %arrayidx.promoted = load i32, i32* %arrayidx, align 4
+  %0 = trunc i64 %indvars.iv25 to i32
+  br label %for.body3
+
+for.body3:                                        ; preds = %for.body3, %for.body3.lr.ph
+  %indvars.iv = phi i64 [ 0, %for.body3.lr.ph ], [ %indvars.iv.next, %for.body3 ]
+  %1 = phi i32 [ %arrayidx.promoted, %for.body3.lr.ph ], [ %add8, %for.body3 ]
+  %arrayidx5 = getelementptr inbounds [1024 x i32], [1024 x i32]* @B, i64 0, i64 %indvars.iv
+  %2 = load i32, i32* %arrayidx5, align 4
+  %add = add nsw i32 %2, %0
+  %add8 = add nsw i32 %add, %1
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, %wide.trip.count
+  br i1 %exitcond, label %for.cond1.for.inc9_crit_edge, label %for.body3
+
+for.cond1.for.inc9_crit_edge:                     ; preds = %for.body3
+  store i32 %add8, i32* %arrayidx, align 4
+  br label %for.inc9
+
+for.inc9:                                         ; preds = %for.cond1.for.inc9_crit_edge, %for.body
+  %indvars.iv.next26 = add nuw nsw i64 %indvars.iv25, 1
+  %exitcond28 = icmp eq i64 %indvars.iv.next26, %wide.trip.count27
+  br i1 %exitcond28, label %for.end11, label %for.body, !llvm.loop !1
+
+for.end11:                                        ; preds = %for.inc9, %entry
+  ret void
+}
+
+!1 = distinct !{!1, !2, !3}
+!2 = !{!"llvm.loop.vectorize.width", i32 4}
+!3 = !{!"llvm.loop.vectorize.enable", i1 true}