[SLP]Fix undef handling in gather function.

[ABataev]

Summary

SLP cannot replace undefs with poison without extra checks, plus need to
preserve the original order of scalars, if decided not to shuffle the
final vectorbuild sequence.

[ABataev]

Rebase

[vporpo]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

Doesn't an undef here mean that lane 4 can potentially be poison? Wouldn't that be incorrect?

[hvdijk]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

A mask of undef means the result is undef, not poison, even if the input contains poison elements. See https://llvm.org/docs/LangRef.html#shufflevector-instruction (which is less clear than I would have liked, it used to explicitly say undef, it now says "undefined" to be more readable, but it's not obvious that it is still meant to refer to undef rather than poison)

[nikic]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

This will be switched to return a poison element in the future, so it would be best not to rely on it. (cc @nlopes).

[nlopes]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

Right, thanks!
We need to switch that behavior to fix a bunch of optimizations.

[hvdijk]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

D93818 linked for reference. Okay, are you saying for the purpose of SLPVectorizer, you would prefer that we already treat undef masks as selecting poison? That should always result in code that is already correct today, but it may not always be optimal under the current rules.

[vporpo]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

Thanks for the explanation!

38 ↗

(On Diff #434535)

SHUFFLE is: LD LD LD Undef
SHUFFLE1 is: LD LD LD LD
Shouldn't we reuse SHUFFLE` for both operands of mul` ?

[nlopes]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

That would be awesome, yes!
It would also motivate us to get that done (@aqjune :)

[ABataev]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

I'll try to update this patch to follow this new rule

[ABataev]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

Just one question. Shall I assume that <0, 0, 0, 0> and <0, 0, 0, undef> can be safely merged to <0, 0, 0, 0> then? Or better to wait for D93818?

[hvdijk]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

I think my D127073 actually already achieved that. The concerns there of regressions there are cases where under the current rules, we can use undef masks, but under those proposed new rules of undef masks, we cannot as we need to ensure the result is not poison. However, this diff is intended to also perform some additional optimisations so it may still be worth updating this one.

[ABataev]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

Yep, my update would be pretty similar to your original patch, just some situation will be handled a bit better. Just need to clarify the question, if shuffle <ld, ld, ld, undef> can be safely replaced by shuffle <ld, ld, ld, ld>

[hvdijk]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

If we ensure that the vectorizer only ever generates a mask of <0, 0, 0, undef> when we element 3 is meant to be poison, it will be valid to merge that with <0, 0, 0, 0> even without waiting for D93818, I think: we do not have to support arbitrary existing shufflevectors here that may rely on undef masks giving undef results, we only have to worry about the new shufflevectors we generate.

[nlopes]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

Just need to clarify the question, if shuffle <ld, ld, ld, undef> can be safely replaced by shuffle <ld, ld, ld, ld>

It's correct to do that in just 2 cases:

• ld is known to be non-poison (eg via ValueTracking), or
• ld is present in the expression tree already. so x * undef can be replaced with x * x, but not with x * y.

If instead of undef, you've poison, then it can safely be replaced with ld.
We've been moving away from initializing vectors with undef, so hopefully undef should be rare in vectors these days. The only exception left is shufflevector.
It would be nice to duplicate some of the SLP tests to use poison rather than undef to ensure the code is simplifying this appropriately.

[hvdijk]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

This explanation is correct for arbitrary existing shuffles, but I wrote earlier that I do not think that is the case we are concerned with. We are concerned with new shuffle instructions that we generate. If we know that the undef mask was generated by SLPVectorizer (because it is a mask that is not even actually the operand to a shufflevector instruction yet, because it is a mask that we have built up in order to potentially generate a shufflevector instruction), and we decide that in SLPVectorizer, we only generate an undef mask if we originally wanted the result to be poison, that is also a case where we can treat an undef mask as selecting poison.

[aqjune]

llvm/test/Transforms/SLPVectorizer/X86/diamond_broadcast_extra_shuffle.ll

37 ↗

(On Diff #434535)

Let me slightly make a progress to that patch and its friends. I think I will have some time for this.

[ABataev]

Address comments

[hvdijk]

Comparing the changes to the tests in this diff to those in D127073, I am seeing a number of tests where we have more shufflevectors, and none where we have fewer. Are there improvements that are not as obvious to see?

[ABataev]

In D127119#3564461, @hvdijk wrote:

Comparing the changes to the tests in this diff to those in D127073, I am seeing a number of tests where we have more shufflevectors, and none where we have fewer. Are there improvements that are not as obvious to see?

Yes, there are. These extra shuffles caused by changes in performExtractsShuffleAction() and in IsIdenticalOrLessDefined lambda, these changes (they treat UndefMaskElem as possible poison) increase number of shuffles. Without them, there are less shuffles, these extra changes are required for correct handling of UndefMaskElem as posion.

[hvdijk]

In D127119#3564522, @ABataev wrote:

In D127119#3564461, @hvdijk wrote:

Comparing the changes to the tests in this diff to those in D127073, I am seeing a number of tests where we have more shufflevectors, and none where we have fewer. Are there improvements that are not as obvious to see?

Yes, there are. These extra shuffles caused by changes in performExtractsShuffleAction() and in IsIdenticalOrLessDefined lambda, these changes (they treat UndefMaskElem as possible poison) increase number of shuffles. Without them, there are less shuffles, these extra changes are required for correct handling of UndefMaskElem as posion.

That suggests that D127073 still results in incorrect code in some cases. I was under the impression that it was already correct, just not optimal. Can you point to specific tests where you believe D127073 results in wrong code?

Taking a random example

--- a/llvm/test/Transforms/SLPVectorizer/X86/insert-shuffle.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/insert-shuffle.ll
@@ -19,9 +19,11 @@ define { <2 x float>, <2 x float> } @foo(%struct.sw* %v) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = fadd <4 x float> [[TMP7]], undef
 ; CHECK-NEXT:    [[TMP9:%.*]] = fadd <4 x float> [[TMP8]], undef
 ; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> poison, <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
-; CHECK-NEXT:    [[INS1:%.*]] = insertvalue { <2 x float>, <2 x float> } undef, <2 x float> [[TMP10]], 0
-; CHECK-NEXT:    [[INS2:%.*]] = insertvalue { <2 x float>, <2 x float> } [[INS1]], <2 x float> [[TMP11]], 1
+; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <2 x float> undef, <2 x float> [[TMP10]], <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[TMP12:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x float> undef, <2 x float> [[TMP12]], <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[INS1:%.*]] = insertvalue { <2 x float>, <2 x float> } undef, <2 x float> [[TMP11]], 0
+; CHECK-NEXT:    [[INS2:%.*]] = insertvalue { <2 x float>, <2 x float> } [[INS1]], <2 x float> [[TMP13]], 1
 ; CHECK-NEXT:    ret { <2 x float>, <2 x float> } [[INS2]]
 ;
 entry:

The extra shufflevectors here, TMP11 and TMP13, are identity shuffles that should not have been generated, they do not change the handling of undef/poison.

[ABataev]

In D127119#3564531, @hvdijk wrote:

In D127119#3564522, @ABataev wrote:

In D127119#3564461, @hvdijk wrote:

Comparing the changes to the tests in this diff to those in D127073, I am seeing a number of tests where we have more shufflevectors, and none where we have fewer. Are there improvements that are not as obvious to see?

Yes, there are. These extra shuffles caused by changes in performExtractsShuffleAction() and in IsIdenticalOrLessDefined lambda, these changes (they treat UndefMaskElem as possible poison) increase number of shuffles. Without them, there are less shuffles, these extra changes are required for correct handling of UndefMaskElem as posion.

That suggests that D127073 still results in incorrect code in some cases. I was under the impression that it was already correct, just not optimal. Can you point to specific tests where you believe D127073 results in wrong code?

Taking a random example

--- a/llvm/test/Transforms/SLPVectorizer/X86/insert-shuffle.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/insert-shuffle.ll
@@ -19,9 +19,11 @@ define { <2 x float>, <2 x float> } @foo(%struct.sw* %v) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = fadd <4 x float> [[TMP7]], undef
 ; CHECK-NEXT:    [[TMP9:%.*]] = fadd <4 x float> [[TMP8]], undef
 ; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> poison, <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
-; CHECK-NEXT:    [[INS1:%.*]] = insertvalue { <2 x float>, <2 x float> } undef, <2 x float> [[TMP10]], 0
-; CHECK-NEXT:    [[INS2:%.*]] = insertvalue { <2 x float>, <2 x float> } [[INS1]], <2 x float> [[TMP11]], 1
+; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <2 x float> undef, <2 x float> [[TMP10]], <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[TMP12:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x float> undef, <2 x float> [[TMP12]], <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[INS1:%.*]] = insertvalue { <2 x float>, <2 x float> } undef, <2 x float> [[TMP11]], 0
+; CHECK-NEXT:    [[INS2:%.*]] = insertvalue { <2 x float>, <2 x float> } [[INS1]], <2 x float> [[TMP13]], 1
 ; CHECK-NEXT:    ret { <2 x float>, <2 x float> } [[INS2]]
 ;
 entry:

The extra shufflevectors here, TMP11 and TMP13, are identity shuffles that should not have been generated, they do not change the handling of undef/poison.

Yes, I see some extra shuffles, that can be removed safely. Working on the improvements.

[ABataev]

Address comments.

[hvdijk]

Thanks, I am seeing improvements with this new version. I'll try to go over the changes in more detail later, some initial superficial comments now.

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

6992

This extra overload does not look like it adds anything: it takes a parameter U if and only if U is Value *. The previous overload takes Value * if and only if U is Value *. That is the same thing; this new overload does not appear to add anything and the file compiles successfully for me if I remove this.

6997

This extra overload goes against the documentation above ValueSelect. The documentation for ValueSelect says it takes a Value *, and if a Value * is wanted, returns that value, otherwise returns a default value. This extra overload is needed because ValueSelect::get now gets called with a const TreeEntry * in the instantiation of performExtractsShuffleAction<const TreeEntry>, but that is contrary to its documentation and it is not clear what it is supposed to mean.

7028

a && b || !a && c is simpler expressed as a ? b : c.

7037

(a ? b : c) == b is simpler expressed as a || c == b.

7043

This comment seems like it does not match how IsCompleteIdentity is set: we can get IsCompleteIdentity to be false even when we select elements only from a single vector.

[ABataev]

Address comments

[ABataev]

Ping!

[vporpo]

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

9136

These ->getAggregateElement(...) and ->getOperand(...) expressions repeat several times in the code and make this code quite verbose. Is there a way to avoid this repetition?

[ABataev]

Address comments

[vporpo]

llvm/test/Transforms/SLPVectorizer/X86/buildvector-same-lane-insert.ll

50

TMP9 seems to be redundant here, it looks like it is a copy of TMP3:
TMP8 is: TMP3[0], undef
TMP9 is: TMP3[0], TMP3[1]

I guess this was an issue even before this patch: TMP8 was a copy of TMP3, so the TMP8 shufflevector was redundant.

[ABataev]

Address comments

[vporpo]

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

9090

Please add a comment describing the lambda.

9155–9158

This also repeats in line 9054. Perhaps move it to a lambda like getOperandIndex(SI2, VecOp2) ?

[hvdijk]

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

9155–9158

Actually... That just looks very wrong, = rather than ==.

[ABataev]

Address comments.

[ABataev]

Fixes and improvements.

[ABataev]

Rebase + fixes

[ABataev]

Rebase

[ABataev]

Rebase

[ABataev]

Rebase

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

for (Instruction *Inst : OrderedScalars) {
     }
 
     PrevInst = Inst;
   }
 
   return Cost;
 }
 
+/// Checks if the buildvector last insertelement, with \p I as a part of this
+/// buildvector, dominates all other uses of \p Root.
+static bool dominatesAllUses(InsertElementInst *I, InsertElementInst *Root) {
+  if (I->getParent() != Root->getParent())
+    return false;
+  // Find top insertelement of the buildvector.
+  Value *V = I;
+  InsertElementInst *Top = I;
+  while (auto *IE = dyn_cast<InsertElementInst>(V)) {
+    if (!IE->hasOneUse())
+      break;
+    Top = IE;
+    V = IE->user_back();
+  }
+  if (Top->getParent() != Root->getParent())
+    return false;
+  BasicBlock *BB = Top->getParent();
+  unsigned Cnt = 0;
+  for (const User *U : Root->users()) {
+    auto *UInst = dyn_cast<Instruction>(U);
+    if (!UInst || UInst->getParent() != BB)
+      continue;
+    if (UInst->comesBefore(Top) || UInst == Top) {
+      ++Cnt;
+      if (Cnt > 1)
+        return false;
+    }
+  }
+  return true;
+}
+
 /// Check if two insertelement instructions are from the same buildvector.
 static bool areTwoInsertFromSameBuildVector(InsertElementInst *VU,
                                             InsertElementInst *V) {
   // Instructions must be from the same basic blocks.
   if (VU->getParent() != V->getParent())
     return false;
   // Checks if 2 insertelements are from the same buildvector.
   if (VU->getType() != V->getType())
     return false;
   // Multiple used inserts are separate nodes.
   if (!VU->hasOneUse() && !V->hasOneUse())
     return false;
+  auto &&IsAllowedToBeInBuildVector = [](InsertElementInst *V,
+                                         InsertElementInst *TopI) {
+    return V->hasOneUse() || dominatesAllUses(TopI, V);
+  };
   auto *IE1 = VU;
   auto *IE2 = V;
   unsigned Idx1 = *getInsertIndex(IE1);
   unsigned Idx2 = *getInsertIndex(IE2);
   // Go through the vector operand of insertelement instructions trying to find
   // either VU as the original vector for IE2 or V as the original vector for
   // IE1.
   do {
     if (IE2 == VU)
-      return VU->hasOneUse();
+      return IsAllowedToBeInBuildVector(VU, V);
     if (IE1 == V)
-      return V->hasOneUse();
+      return IsAllowedToBeInBuildVector(V, VU);
     if (IE1) {
       if ((IE1 != VU && !IE1->hasOneUse()) ||
           getInsertIndex(IE1).value_or(Idx2) == Idx2)
         IE1 = nullptr;
       else
         IE1 = dyn_cast<InsertElementInst>(IE1->getOperand(0));
     }
     if (IE2) {
if (I2 && ((I2 == IE2 || I2->hasOneUse())) &&
         getInsertIndex(I2).value_or(Idx1) != Idx1)
       I2 = dyn_cast<InsertElementInst>(I2->getOperand(0));
   } while ((I1 && PrevI1 != I1) || (I2 && PrevI2 != I2));
   llvm_unreachable("Two different buildvectors not expected.");
 }
 
 namespace {
 /// Returns incoming Value *, if the requested type is Value * too, or a default
-/// value, otherwise.
+/// value, otherwise. Also, if the requested type is Value * and the incoming
+/// value is not Value *, returns nullptr.
 struct ValueSelect {
   template <typename U>
   static typename std::enable_if<std::is_same<Value *, U>::value, Value *>::type
   get(Value *V) {
     return V;
   }
   template <typename U>
+  static typename std::enable_if<!std::is_same<Value *, U>::value, Value *>::type
+  get(U) {
+    return nullptr;
+  }
+  template <typename U>
   static typename std::enable_if<!std::is_same<Value *, U>::value, U>::type
   get(Value *) {
     return U();
   }
 };
 } // namespace
 
 /// Does the analysis of the provided shuffle masks and performs the requested
 /// actions on the vectors with the given shuffle masks. It tries to do it in
 /// several steps.
 /// 1. If the Base vector is not undef vector, resizing the very first mask to
 /// have common VF and perform action for 2 input vectors (including non-undef
static T *performExtractsShuffleAction(
     MutableArrayRef<std::pair<T *, SmallVector<int>>> ShuffleMask, Value *Base,
     function_ref<unsigned(T *)> GetVF,
     function_ref<std::pair<T *, bool>(T *, ArrayRef<int>)> ResizeAction,
     function_ref<T *(ArrayRef<int>, ArrayRef<T *>)> Action) {
   assert(!ShuffleMask.empty() && "Empty list of shuffles for inserts.");
   SmallVector<int> Mask(ShuffleMask.begin()->second);
   auto VMIt = std::next(ShuffleMask.begin());
   T *Prev = nullptr;
-  bool IsBaseNotUndef = !isUndefVector(Base);
+  Value *FirstV = ValueSelect::get<T *>(ShuffleMask.begin()->first);
+  bool IsBaseNotUndef =
+      !(FirstV ? (isa<PoisonValue>(Base) ||
+                  (isUndefVector(Base) && isGuaranteedNotToBePoison(FirstV)))
+               : isUndefVector(Base));
   if (IsBaseNotUndef) {
     // Base is not undef, need to combine it with the next subvectors.
     std::pair<T *, bool> Res = ResizeAction(ShuffleMask.begin()->first, Mask);
-    for (unsigned Idx = 0, VF = Mask.size(); Idx < VF; ++Idx) {
-      if (Mask[Idx] == UndefMaskElem)
+    bool IsCompleteIdentity = true;
+    for (int Idx = 0, VF = Mask.size(); Idx < VF; ++Idx) {
+      if (Mask[Idx] == UndefMaskElem) {
         Mask[Idx] = Idx;
-      else
+        IsCompleteIdentity = false;
+      } else {
+        IsCompleteIdentity &= (Res.second || Mask[Idx] == Idx);
         Mask[Idx] = (Res.second ? Idx : Mask[Idx]) + VF;
       }
+    }
+    if (IsCompleteIdentity) {
+      // Found complete identity, i.e. elements only from single vector are
+      // selected in the identity order.
+      Prev = Res.first;
+      std::iota(Mask.begin(), Mask.end(), 0);
+    } else {
       auto *V = ValueSelect::get<T *>(Base);
       (void)V;
       assert((!V || GetVF(V) == Mask.size()) &&
              "Expected base vector of VF number of elements.");
       Prev = Action(Mask, {nullptr, Res.first});
+    }
   } else if (ShuffleMask.size() == 1) {
     // Base is undef and only 1 vector is shuffled - perform the action only for
     // single vector, if the mask is not the identity mask.
     std::pair<T *, bool> Res = ResizeAction(ShuffleMask.begin()->first, Mask);
     if (Res.second)
       // Identity mask is found.
       Prev = Res.first;
     else
Value *BoUpSLP::createBuildVector(ArrayRef<Value *> VL) {
   unsigned VF = VL.size();
   // Exploit possible reuse of values across lanes.
   SmallVector<int> ReuseShuffleIndicies;
   SmallVector<Value *> UniqueValues;
   if (VL.size() > 2) {
     DenseMap<Value *, unsigned> UniquePositions;
     unsigned NumValues =
         std::distance(VL.begin(), find_if(reverse(VL), [](Value *V) {
-                                    return !isa<UndefValue>(V);
+                                    return !isa<PoisonValue>(V);
                                   }).base());
-    VF = std::max<unsigned>(VF, PowerOf2Ceil(NumValues));
-    int UniqueVals = 0;
+    VF = std::min<unsigned>(VF, PowerOf2Ceil(NumValues));
+    bool IsIdentity = true;
+    int UndefIdx = UndefMaskElem;
+    int ReplaceUndefIdx = UndefMaskElem;
     for (Value *V : VL.drop_back(VL.size() - VF)) {
-      if (isa<UndefValue>(V)) {
+      if (isa<PoisonValue>(V)) {
         ReuseShuffleIndicies.emplace_back(UndefMaskElem);
         continue;
       }
       if (isConstant(V)) {
+        if (isa<UndefValue>(V)) {
+          if (UndefIdx == UndefMaskElem) {
+            UndefIdx = UniqueValues.size();
+            UniqueValues.emplace_back(V);
+          }
+          ReuseShuffleIndicies.emplace_back(UndefIdx);
+        } else {
           ReuseShuffleIndicies.emplace_back(UniqueValues.size());
           UniqueValues.emplace_back(V);
+        }
         continue;
       }
       auto Res = UniquePositions.try_emplace(V, UniqueValues.size());
       ReuseShuffleIndicies.emplace_back(Res.first->second);
       if (Res.second) {
         UniqueValues.emplace_back(V);
-        ++UniqueVals;
-      }
-    }
-    if (UniqueVals == 1 && UniqueValues.size() == 1) {
-      // Emit pure splat vector.
-      ReuseShuffleIndicies.append(VF - ReuseShuffleIndicies.size(),
-                                  UndefMaskElem);
-    } else if (UniqueValues.size() >= VF - 1 || UniqueValues.size() <= 1) {
-      if (UniqueValues.empty()) {
-        assert(all_of(VL, UndefValue::classof) && "Expected list of undefs.");
-        NumValues = VF;
+        if (ReplaceUndefIdx == UndefMaskElem)
+          ReplaceUndefIdx = Res.first->second;
+      } else {
+        IsIdentity = false;
       }
-      ReuseShuffleIndicies.clear();
-      UniqueValues.clear();
-      UniqueValues.append(VL.begin(), std::next(VL.begin(), NumValues));
     }
-    UniqueValues.append(VF - UniqueValues.size(),
+    // Check if undef values can be safely replaced by some non-const vals.
+    if (UndefIdx != UndefMaskElem && ReplaceUndefIdx != UndefMaskElem &&
+        !IsIdentity) {
+      int MinIdx = std::min(ReplaceUndefIdx, UndefIdx);
+      for (unsigned I = 0; I < VF; ++I)
+        if (ReuseShuffleIndicies[I] == UndefIdx ||
+            ReuseShuffleIndicies[I] == ReplaceUndefIdx)
+          ReuseShuffleIndicies[I] = MinIdx;
+      if (MinIdx != ReplaceUndefIdx) {
+        std::swap(UniqueValues[ReplaceUndefIdx], UniqueValues[UndefIdx]);
+        UndefIdx = ReplaceUndefIdx;
+      }
+      UniqueValues.erase(std::next(UniqueValues.begin(), UndefIdx));
+    }
+    if (!IsIdentity) {
+      UniqueValues.append((UniqueValues.size() == 1 ? VL.size() : VF) -
+                              UniqueValues.size(),
                           PoisonValue::get(VL[0]->getType()));
+      ReuseShuffleIndicies.append(VL.size() - VF, UndefMaskElem);
       VL = UniqueValues;
+    } else {
+      ReuseShuffleIndicies.clear();
+    }
   }
 
   ShuffleInstructionBuilder ShuffleBuilder(Builder, VF, GatherShuffleSeq,
                                            CSEBlocks);
   Value *Vec = gather(VL);
   if (!ReuseShuffleIndicies.empty()) {
     ShuffleBuilder.addMask(ReuseShuffleIndicies);
     Vec = ShuffleBuilder.finalize(Vec);
case Instruction::InsertElement: {
       if (!IsIdentity || NumElts != NumScalars) {
         V = Builder.CreateShuffleVector(V, Mask);
         if (auto *I = dyn_cast<Instruction>(V)) {
           GatherShuffleSeq.insert(I);
           CSEBlocks.insert(I->getParent());
         }
       }
 
-      if ((!IsIdentity || Offset != 0 ||
-           !isUndefVector(FirstInsert->getOperand(0))) &&
+      bool IsUndefFirstOp = isUndefVector(FirstInsert->getOperand(0));
+      if ((!IsIdentity || Offset != 0 || !IsUndefFirstOp) &&
           NumElts != NumScalars) {
-        SmallVector<int> InsertMask(NumElts);
+        SmallVector<int> InsertMask(NumElts, UndefMaskElem);
+        if (!IsUndefFirstOp)
           std::iota(InsertMask.begin(), InsertMask.end(), 0);
         for (unsigned I = 0; I < NumElts; I++) {
           if (Mask[I] != UndefMaskElem)
             InsertMask[Offset + I] = NumElts + I;
         }
 
         V = Builder.CreateShuffleVector(
             FirstInsert->getOperand(0), V, InsertMask,
             cast<Instruction>(E->Scalars.back())->getName());
for (Instruction *II : reverse(Inserts)) {
       NewInst = II;
     }
     LastInsert->replaceAllUsesWith(NewInst);
     for (InsertElementInst *IE : reverse(ShuffledInserts[I].InsertElements)) {
       IE->replaceUsesOfWith(IE->getOperand(0),
                             PoisonValue::get(IE->getOperand(0)->getType()));
       IE->replaceUsesOfWith(IE->getOperand(1),
                             PoisonValue::get(IE->getOperand(1)->getType()));
+      if (IE == FirstInsert && dominatesAllUses(LastInsert, FirstInsert))
+        FirstInsert->replaceAllUsesWith(NewInst);
       eraseInstruction(IE);
     }
     CSEBlocks.insert(LastInsert->getParent());
   }
 
   // For each vectorized value:
   for (auto &TEPtr : VectorizableTree) {
     TreeEntry *Entry = TEPtr.get();
auto &&IsIdenticalOrLessDefined = [this](Instruction *I1, Instruction *I2,
     if (I1->getType() != I2->getType())
       return false;
     auto *SI1 = dyn_cast<ShuffleVectorInst>(I1);
     auto *SI2 = dyn_cast<ShuffleVectorInst>(I2);
     if (!SI1 || !SI2)
       return I1->isIdenticalTo(I2);
     if (SI1->isIdenticalTo(SI2))
       return true;
-    for (int I = 0, E = SI1->getNumOperands(); I < E; ++I)
-      if (SI1->getOperand(I) != SI2->getOperand(I))
+    for (int I = 0, E = SI1->getNumOperands(); I < E; ++I) {
+      Value *SIOp = SI1->getOperand(I);
+      if (isa<UndefValue>(SIOp) &&
+          SIOp->getType() == SI2->getOperand(I)->getType())
+        continue;
+      if (!any_of(SI2->operands(), [SIOp](Value *Op) { return Op == SIOp; }) ||
+          any_of(SI2->operands(), [SI1](Value *Op) { return Op == SI1; }))
         return false;
+    }
     // Check if the second instruction is more defined than the first one.
     NewMask.assign(SI2->getShuffleMask().begin(), SI2->getShuffleMask().end());
     ArrayRef<int> SM1 = SI1->getShuffleMask();
     // Count trailing undefs in the mask to check the final number of used
     // registers.
+    auto &&IsStrictUndef = [](Value *V) {
+      return isa_and_nonnull<UndefValue>(V) && !isa<PoisonValue>(V);
+    };
     unsigned LastUndefsCnt = 0;
+    unsigned VF = cast<VectorType>(SI1->getOperand(0)->getType())
+                      ->getElementCount()
+                      .getKnownMinValue();
+    // Tries to fetch the vector element (scalar value) for \p VecOp using given
+    // index \p Idx. \p RootOp is used to stop lookup through shuffles, if \p
+    // RootOp is a root of the \p VecOp shuffle.
+    auto &&GetVecElem = [VF](Value *&VecOp, int Idx, Value *RootOp) -> Value * {
+      unsigned EVF = VF;
+      while (auto *SVOp = dyn_cast<ShuffleVectorInst>(VecOp)) {
+        VecOp = SVOp->getOperand(Idx / EVF);
+        Idx = SVOp->getMaskValue(Idx % EVF);
+        auto *FTy = dyn_cast<FixedVectorType>(VecOp->getType());
+        if (!FTy)
+          break;
+        EVF = FTy->getElementCount().getKnownMinValue();
+        if (Idx == UndefMaskElem)
+          return PoisonValue::get(
+              cast<VectorType>(VecOp->getType())->getElementType());
+        if (VecOp == RootOp)
+          break;
+      }
+      if (isa<PoisonValue>(VecOp))
+        return PoisonValue::get(
+            cast<VectorType>(VecOp->getType())->getElementType());
+      if (isa<UndefValue>(VecOp))
+        return UndefValue::get(
+            cast<VectorType>(VecOp->getType())->getElementType());
+      if (auto *CV1 = dyn_cast<ConstantVector>(VecOp))
+        return CV1->getAggregateElement(Idx % EVF);
+      return nullptr;
+    };
     for (int I = 0, E = NewMask.size(); I < E; ++I) {
       if (SM1[I] == UndefMaskElem)
         ++LastUndefsCnt;
       else
         LastUndefsCnt = 0;
-      if (NewMask[I] != UndefMaskElem && SM1[I] != UndefMaskElem &&
-          NewMask[I] != SM1[I])
-        return false;
-      if (NewMask[I] == UndefMaskElem)
+      if (NewMask[I] == UndefMaskElem) {
         NewMask[I] = SM1[I];
+        continue;
+      }
+      Value *VecOp1 = SI2->getOperand(NewMask[I] / VF);
+      unsigned VecOp2Pos = SM1[I] / VF;
+      Value *VecOp2 =
+          SM1[I] == UndefMaskElem ? nullptr : SI1->getOperand(VecOp2Pos);
+      if (SM1[I] == UndefMaskElem || (NewMask[I] == SM1[I] && VecOp1 == VecOp2))
+        continue;
+      // Check if one mask can be safely replaced by another.
+      Value *CV1Elem = GetVecElem(VecOp1, NewMask[I], VecOp2);
+      Value *CV2Elem = GetVecElem(VecOp2, SM1[I], VecOp1);
+      if (isa_and_nonnull<PoisonValue>(CV1Elem) ||
+          (IsStrictUndef(CV1Elem) &&
+           (IsStrictUndef(CV2Elem) ||
+            (CV2Elem && isGuaranteedNotToBePoison(CV2Elem)) ||
+            (!CV2Elem && isGuaranteedNotToBePoison(VecOp2))))) {
+        NewMask[I] = SM1[I] % VF + VecOp2Pos * VF;
+        continue;
+      }
+      if ((CV1Elem && CV1Elem == CV2Elem) ||
+          isa_and_nonnull<PoisonValue>(CV2Elem) ||
+          (IsStrictUndef(CV2Elem) &&
+           (IsStrictUndef(CV1Elem) ||
+            (CV1Elem && isGuaranteedNotToBePoison(CV1Elem)) ||
+            (!CV1Elem && isGuaranteedNotToBePoison(VecOp1))))) {
+        NewMask[I] = NewMask[I] % VF + VecOp2Pos * VF;
+        continue;
+      }
+      return false;
     }
     // Check if the last undefs actually change the final number of used vector
     // registers.
     return SM1.size() - LastUndefsCnt > 1 &&
            TTI->getNumberOfParts(SI1->getType()) ==
                TTI->getNumberOfParts(
                    FixedVectorType::get(SI1->getType()->getElementType(),
                                         SM1.size() - LastUndefsCnt));
   };
   // Perform O(N^2) search over the gather/shuffle sequences and merge identical
   // instructions. TODO: We can further optimize this scan if we split the
   // instructions into different buckets based on the insert lane.
   SmallVector<Instruction *, 16> Visited;

llvm/test/Transforms/SLPVectorizer/AArch64/tsc-s116.ll

 ; contiguous. The score estimation needs to be corrected, so that these 4 loads
 ; are not selected for vectorization. Instead we should vectorize with
 ; contiguous loads, from %a plus offsets 0 to 3, or offsets 1 to 4.
 
 define void @s116_modified(float* %a) {
 ; CHECK-LABEL: @s116_modified(
 ; CHECK-NEXT:    [[GEP0:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 0
 ; CHECK-NEXT:    [[GEP1:%.*]] = getelementptr inbounds float, float* [[A]], i64 1
-; CHECK-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, float* [[A]], i64 2
-; CHECK-NEXT:    [[GEP4:%.*]] = getelementptr inbounds float, float* [[A]], i64 4
-; CHECK-NEXT:    [[LD1:%.*]] = load float, float* [[GEP1]], align 4
+; CHECK-NEXT:    [[GEP3:%.*]] = getelementptr inbounds float, float* [[A]], i64 3
 ; CHECK-NEXT:    [[LD0:%.*]] = load float, float* [[GEP0]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[GEP2]] to <2 x float>*
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[GEP1]] to <2 x float>*
 ; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, <2 x float>* [[TMP1]], align 4
-; CHECK-NEXT:    [[LD4:%.*]] = load float, float* [[GEP4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> poison, float [[LD0]], i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
-; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP4]], <4 x i32> <i32 0, i32 4, i32 5, i32 3>
-; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <4 x float> [[TMP5]], float [[LD4]], i32 3
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x float> poison, float [[LD1]], i32 0
-; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <4 x float> [[TMP7]], float [[LD1]], i32 1
-; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <4 x float> [[TMP8]], <4 x float> [[TMP4]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
-; CHECK-NEXT:    [[TMP10:%.*]] = fmul fast <4 x float> [[TMP6]], [[TMP9]]
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[GEP3]] to <2 x float>*
+; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[TMP3]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> poison, float [[LD0]], i32 0
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> [[TMP4]], <4 x i32> <i32 0, i32 1, i32 2, i32 undef>
+; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP6]], <4 x i32> <i32 0, i32 5, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <4 x float> [[TMP7]], <4 x float> [[TMP8]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x float> [[TMP6]], <4 x float> poison, <4 x i32> <i32 0, i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[TMP10:%.*]] = fmul fast <4 x float> [[TMP9]], [[SHUFFLE]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = bitcast float* [[GEP0]] to <4 x float>*
 ; CHECK-NEXT:    store <4 x float> [[TMP10]], <4 x float>* [[TMP11]], align 4
 ; CHECK-NEXT:    ret void
 ;
   %gep0 = getelementptr inbounds float, float* %a, i64 0
   %gep1 = getelementptr inbounds float, float* %a, i64 1
   %gep2 = getelementptr inbounds float, float* %a, i64 2
   %gep3 = getelementptr inbounds float, float* %a, i64 3

llvm/test/Transforms/SLPVectorizer/AArch64/vectorize-free-extracts-inserts.ll

 ; CHECK-NEXT:    [[V2_LANE_1:%.*]] = extractelement <4 x double> [[V_2]], i32 1
 ; CHECK-NEXT:    [[V2_LANE_2:%.*]] = extractelement <4 x double> [[V_2]], i32 2
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x double> poison, double [[V1_LANE_0]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x double> [[TMP0]], double [[V1_LANE_2]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x double> [[TMP1]], double [[V1_LANE_1]], i32 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x double> [[TMP2]], double [[V1_LANE_3]], i32 3
 ; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x double> poison, double [[V2_LANE_2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x double> [[TMP4]], double [[V2_LANE_1]], i32 1
-; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <4 x double> [[TMP5]], double [[V2_LANE_2]], i32 2
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x double> [[TMP6]], double [[V2_LANE_0]], i32 3
-; CHECK-NEXT:    [[TMP8:%.*]] = fmul <4 x double> [[TMP3]], [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <4 x double> [[TMP8]], <4 x double> poison, <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <4 x double> [[TMP5]], double [[V2_LANE_0]], i32 2
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x double> [[TMP6]], <4 x double> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 2>
+; CHECK-NEXT:    [[TMP7:%.*]] = fmul <4 x double> [[TMP3]], [[SHUFFLE]]
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <4 x double> [[TMP7]], <4 x double> poison, <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; CHECK-NEXT:    call void @use(double [[V1_LANE_0]])
 ; CHECK-NEXT:    call void @use(double [[V1_LANE_1]])
 ; CHECK-NEXT:    call void @use(double [[V1_LANE_2]])
 ; CHECK-NEXT:    call void @use(double [[V1_LANE_3]])
-; CHECK-NEXT:    store <9 x double> [[TMP9]], <9 x double>* [[PTR_1]], align 8
+; CHECK-NEXT:    store <9 x double> [[TMP8]], <9 x double>* [[PTR_1]], align 8
 ; CHECK-NEXT:    ret void
 ;
 bb:
   %v.1 = load <9 x double>, <9 x double>* %ptr.1, align 8
   %v1.lane.0 = extractelement <9 x double> %v.1, i32 0
   %v1.lane.1 = extractelement <9 x double> %v.1, i32 1
   %v1.lane.2 = extractelement <9 x double> %v.1, i32 2
   %v1.lane.3 = extractelement <9 x double> %v.1, i32 3

llvm/test/Transforms/SLPVectorizer/X86/PR35865-inseltpoison.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -slp-vectorizer < %s -S -o - -mtriple=x86_64-apple-macosx10.10.0 -mcpu=core2 | FileCheck %s
 
 define void @_Z10fooConvertPDv4_xS0_S0_PKS_() {
 ; CHECK-LABEL: @_Z10fooConvertPDv4_xS0_S0_PKS_(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <16 x half> undef, i32 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <16 x half> undef, i32 5
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x half> poison, half [[TMP0]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x half> [[TMP2]], half [[TMP1]], i32 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = fpext <2 x half> [[TMP3]] to <2 x float>
 ; CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x float> [[TMP4]] to <2 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i32> [[TMP5]], <2 x i32> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; CHECK-NEXT:    [[VECINS_I_5_I1:%.*]] = shufflevector <8 x i32> poison, <8 x i32> [[TMP6]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 6, i32 7>
+; CHECK-NEXT:    [[VECINS_I_5_I1:%.*]] = shufflevector <8 x i32> poison, <8 x i32> [[TMP6]], <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 8, i32 9, i32 undef, i32 undef>
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %0 = extractelement <16 x half> undef, i32 4
   %conv.i.4.i = fpext half %0 to float
   %1 = bitcast float %conv.i.4.i to i32
   %vecins.i.4.i = insertelement <8 x i32> poison, i32 %1, i32 4
   %2 = extractelement <16 x half> undef, i32 5
   %conv.i.5.i = fpext half %2 to float
   %3 = bitcast float %conv.i.5.i to i32
   %vecins.i.5.i = insertelement <8 x i32> %vecins.i.4.i, i32 %3, i32 5
   ret void
 }

llvm/test/Transforms/SLPVectorizer/X86/PR35865.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -slp-vectorizer < %s -S -o - -mtriple=x86_64-apple-macosx10.10.0 -mcpu=core2 | FileCheck %s
 
 define void @_Z10fooConvertPDv4_xS0_S0_PKS_() {
 ; CHECK-LABEL: @_Z10fooConvertPDv4_xS0_S0_PKS_(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <16 x half> undef, i32 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <16 x half> undef, i32 5
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x half> poison, half [[TMP0]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x half> [[TMP2]], half [[TMP1]], i32 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = fpext <2 x half> [[TMP3]] to <2 x float>
 ; CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x float> [[TMP4]] to <2 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i32> [[TMP5]], <2 x i32> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; CHECK-NEXT:    [[VECINS_I_5_I1:%.*]] = shufflevector <8 x i32> undef, <8 x i32> [[TMP6]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 6, i32 7>
+; CHECK-NEXT:    [[VECINS_I_5_I1:%.*]] = shufflevector <8 x i32> undef, <8 x i32> [[TMP6]], <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 8, i32 9, i32 undef, i32 undef>
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %0 = extractelement <16 x half> undef, i32 4
   %conv.i.4.i = fpext half %0 to float
   %1 = bitcast float %conv.i.4.i to i32
   %vecins.i.4.i = insertelement <8 x i32> undef, i32 %1, i32 4
   %2 = extractelement <16 x half> undef, i32 5
   %conv.i.5.i = fpext half %2 to float
   %3 = bitcast float %conv.i.5.i to i32
   %vecins.i.5.i = insertelement <8 x i32> %vecins.i.4.i, i32 %3, i32 5
   ret void
 }

llvm/test/Transforms/SLPVectorizer/X86/alternate-cmp-swapped-pred.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -mtriple=x86_64-unknown -slp-vectorizer -S | FileCheck %s
 
 define i16 @test(i16 %call37) {
 ; CHECK-LABEL: @test(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[CALL:%.*]] = load i16, i16* undef, align 2
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i16> <i16 poison, i16 0, i16 0, i16 0, i16 poison, i16 0, i16 0, i16 0>, i16 [[CALL37:%.*]], i32 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i16> [[TMP0]], i16 [[CALL]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <8 x i16> <i16 0, i16 0, i16 0, i16 poison, i16 0, i16 0, i16 poison, i16 0>, i16 [[CALL37]], i32 3
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <8 x i16> [[TMP2]], i16 [[CALL37]], i32 6
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp slt <8 x i16> [[TMP1]], [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = icmp sgt <8 x i16> [[TMP1]], [[TMP3]]
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i1> [[TMP4]], <8 x i1> [[TMP5]], <8 x i32> <i32 0, i32 1, i32 2, i32 11, i32 12, i32 5, i32 14, i32 7>
-; CHECK-NEXT:    [[TMP7:%.*]] = zext <8 x i1> [[TMP6]] to <8 x i16>
-; CHECK-NEXT:    [[TMP8:%.*]] = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> [[TMP7]])
-; CHECK-NEXT:    [[OP_EXTRA:%.*]] = add i16 [[TMP8]], 0
-; CHECK-NEXT:    ret i16 [[OP_EXTRA]]
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <8 x i16> <i16 0, i16 0, i16 0, i16 poison, i16 0, i16 0, i16 0, i16 poison>, i16 [[CALL37]], i32 3
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i16> [[TMP2]], <8 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 3, i32 6>
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp slt <8 x i16> [[TMP1]], [[SHUFFLE]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp sgt <8 x i16> [[TMP1]], [[SHUFFLE]]
+; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <8 x i1> [[TMP3]], <8 x i1> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 11, i32 12, i32 5, i32 14, i32 7>
+; CHECK-NEXT:    [[TMP6:%.*]] = zext <8 x i1> [[TMP5]] to <8 x i16>
+; CHECK-NEXT:    [[TMP7:%.*]] = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> [[TMP6]])
+; CHECK-NEXT:    [[OP_RDX:%.*]] = add i16 [[TMP7]], 0
+; CHECK-NEXT:    ret i16 [[OP_RDX]]
 ;
 entry:
   %call = load i16, i16* undef, align 2
   %0 = icmp slt i16 %call, 0
   %cond = zext i1 %0 to i16
   %1 = add i16 %cond, 0
   %2 = icmp slt i16 0, 0
   %cond32 = zext i1 %2 to i16

llvm/test/Transforms/SLPVectorizer/X86/broadcast_long.ll

 ; YAML-NEXT:  - TreeSize:        '2'
 
 define void @bcast_long(i32 *%A, i32 *%S) {
 ; CHECK-LABEL: @bcast_long(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[A0:%.*]] = load i32, i32* [[A:%.*]], align 8
 ; CHECK-NEXT:    [[IDXS0:%.*]] = getelementptr inbounds i32, i32* [[S:%.*]], i64 0
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i32> poison, i32 [[A0]], i32 0
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i32> [[TMP0]], <8 x i32> poison, <8 x i32> <i32 0, i32 0, i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0>
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i32> [[TMP0]], <8 x i32> poison, <8 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[IDXS0]] to <8 x i32>*
 ; CHECK-NEXT:    store <8 x i32> [[SHUFFLE]], <8 x i32>* [[TMP1]], align 8
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %A0 = load i32, i32 *%A, align 8
 
   %idxS0 = getelementptr inbounds i32, i32* %S, i64 0

llvm/test/Transforms/SLPVectorizer/X86/buildvector-same-lane-insert.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ;RUN: opt -S -slp-vectorizer -mtriple=x86_64-unknown-linux-android23 < %s | FileCheck %s
 
 define void @test() {
 ; CHECK-LABEL: @test(
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr undef, i32 2
 ; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, ptr undef, align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = fsub <2 x float> [[TMP2]], [[TMP3]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x float> [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x float> [[TMP4]], i32 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = fcmp olt float [[TMP6]], [[TMP5]]
-; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x float> zeroinitializer, float 0.000000e+00, i64 0
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <2 x float> undef, <2 x float> [[TMP3]], <2 x i32> <i32 2, i32 1>
+; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x float> zeroinitializer, float 0.000000e+00, i64 0
 ; CHECK-NEXT:    store <2 x float> zeroinitializer, ptr null, align 4
-; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP2]], <2 x i32> <i32 3, i32 1>
+; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x float> [[TMP8]], <2 x float> [[TMP2]], <2 x i32> <i32 3, i32 1>
 ; CHECK-NEXT:    store <2 x float> zeroinitializer, ptr null, align 4
 ; CHECK-NEXT:    ret void
 ;
   %1 = getelementptr inbounds float, ptr undef, i32 2
   %2 = load float, ptr %1, align 4
   %3 = load float, ptr undef, align 4
   %4 = fsub float %2, %3
   %5 = getelementptr inbounds float, ptr undef, i32 3

 ; CHECK-LABEL: @test1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr undef, i32 2
 ; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, ptr undef, align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = fsub <2 x float> [[TMP2]], [[TMP3]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x float> [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x float> [[TMP4]], i32 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = fcmp olt float [[TMP6]], [[TMP5]]
-; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr null, align 4
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP2]], <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    store <2 x float> [[TMP8]], ptr null, align 4
-; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP2]], <2 x i32> <i32 0, i32 3>
-; CHECK-NEXT:    store <2 x float> [[TMP9]], ptr null, align 4
 ; CHECK-NEXT:    ret void
 ;
   %1 = getelementptr inbounds float, ptr undef, i32 2
   %2 = load float, ptr %1, align 4
   %3 = load float, ptr undef, align 4
   %4 = fsub float %2, %3
   %5 = getelementptr inbounds float, ptr undef, i32 3
   %6 = load float, ptr %5, align 4
   %7 = getelementptr inbounds float, ptr undef, i32 1

llvm/test/Transforms/SLPVectorizer/X86/cmp-as-alternate-ops.ll

 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i32> <i32 poison, i32 0, i32 poison, i32 0>, i32 [[CONV_I32_I_I_I:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32> [[TMP0]], i32 [[CONV_I32_I_I_I1]], i32 2
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp sgt <4 x i32> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp slt <4 x i32> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i1> [[TMP2]], <4 x i1> [[TMP3]], <4 x i32> <i32 0, i32 1, i32 6, i32 3>
 ; CHECK-NEXT:    [[TMP5:%.*]] = select <4 x i1> [[TMP4]], <4 x float> zeroinitializer, <4 x float> zeroinitializer
 ; CHECK-NEXT:    [[TMP6:%.*]] = fadd <4 x float> [[TMP5]], zeroinitializer
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x float> [[TMP6]], <4 x float> poison, <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <2 x float> zeroinitializer, <2 x float> [[TMP7]], <2 x i32> <i32 2, i32 3>
-; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <4 x float> [[TMP6]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
-; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x float> zeroinitializer, <2 x float> [[TMP9]], <2 x i32> <i32 2, i32 3>
-; CHECK-NEXT:    [[DOTFCA_0_INSERT:%.*]] = insertvalue { <2 x float>, <2 x float> } zeroinitializer, <2 x float> [[TMP8]], 0
-; CHECK-NEXT:    [[DOTFCA_1_INSERT:%.*]] = insertvalue { <2 x float>, <2 x float> } [[DOTFCA_0_INSERT]], <2 x float> [[TMP10]], 1
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float> [[TMP6]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[DOTFCA_0_INSERT:%.*]] = insertvalue { <2 x float>, <2 x float> } zeroinitializer, <2 x float> [[TMP7]], 0
+; CHECK-NEXT:    [[DOTFCA_1_INSERT:%.*]] = insertvalue { <2 x float>, <2 x float> } [[DOTFCA_0_INSERT]], <2 x float> [[TMP8]], 1
 ; CHECK-NEXT:    ret { <2 x float>, <2 x float> } zeroinitializer
 ;
 entry:
   %cmp.i.i.i.i.i = icmp slt i32 0, 0
   %cond.i.i.i.i = select i1 %cmp.i.i.i.i.i, float 0.000000e+00, float 0.000000e+00
   %conv.i32.i.i.i1 = fptosi float 0.000000e+00 to i32
   %cmp.i.i34.i.i.i = icmp slt i32 %conv.i32.i.i.i1, 0
   %cond.i35.i.i.i = select i1 %cmp.i.i34.i.i.i, float 0.000000e+00, float 0.000000e+00

llvm/test/Transforms/SLPVectorizer/X86/commutativity.ll

 ; AVX-LABEL: @same_opcode_on_one_side(
 ; AVX-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32> poison, i32 [[C:%.*]], i32 0
 ; AVX-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; AVX-NEXT:    [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 [[A:%.*]], i32 0
 ; AVX-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; AVX-NEXT:    [[TMP3:%.*]] = add <4 x i32> [[SHUFFLE]], [[SHUFFLE1]]
 ; AVX-NEXT:    [[TMP4:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[B:%.*]], i32 1
 ; AVX-NEXT:    [[TMP5:%.*]] = insertelement <4 x i32> [[TMP4]], i32 [[C]], i32 2
-; AVX-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32> [[TMP5]], i32 [[A]], i32 3
-; AVX-NEXT:    [[TMP7:%.*]] = xor <4 x i32> [[TMP3]], [[TMP6]]
-; AVX-NEXT:    store <4 x i32> [[TMP7]], <4 x i32>* bitcast ([32 x i32]* @cle32 to <4 x i32>*), align 16
+; AVX-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 0>
+; AVX-NEXT:    [[TMP6:%.*]] = xor <4 x i32> [[TMP3]], [[SHUFFLE2]]
+; AVX-NEXT:    store <4 x i32> [[TMP6]], <4 x i32>* bitcast ([32 x i32]* @cle32 to <4 x i32>*), align 16
 ; AVX-NEXT:    ret void
 ;
   %add1 = add i32 %c, %a
   %add2 = add i32 %c, %a
   %add3 = add i32 %a, %c
   %add4 = add i32 %c, %a
   %1 = xor i32 %add1, %a
   store i32 %1, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @cle32, i64 0, i64 0), align 16
   %2 = xor i32 %b, %add2
   store i32 %2, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @cle32, i64 0, i64 1)
   %3 = xor i32 %c, %add3
   store i32 %3, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @cle32, i64 0, i64 2)
   %4 = xor i32 %a, %add4
   store i32 %4, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @cle32, i64 0, i64 3)
   ret void
 }

llvm/test/Transforms/SLPVectorizer/X86/crash_7zip.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -basic-aa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | 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"
 target triple = "x86_64-apple-macosx10.8.0"
 
 %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334 = type { %struct._CLzmaProps.0.27.54.81.102.123.144.165.180.195.228.258.333, i16*, i8*, i8*, i32, i32, i64, i64, i32, i32, i32, [4 x i32], i32, i32, i32, i32, i32, [20 x i8] }
 %struct._CLzmaProps.0.27.54.81.102.123.144.165.180.195.228.258.333 = type { i32, i32, i32, i32 }
 
 define fastcc void @LzmaDec_DecodeReal2(%struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p) {
 ; CHECK-LABEL: @LzmaDec_DecodeReal2(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[RANGE20_I:%.*]] = getelementptr inbounds [[STRUCT_CLZMADEC_1_28_55_82_103_124_145_166_181_196_229_259_334:%.*]], %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* [[P:%.*]], i64 0, i32 4
 ; CHECK-NEXT:    br label [[DO_BODY66_I:%.*]]
 ; CHECK:       do.body66.i:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi <2 x i32> [ [[TMP3:%.*]], [[DO_COND_I:%.*]] ], [ undef, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = phi <2 x i32> [ [[TMP4:%.*]], [[DO_COND_I:%.*]] ], [ undef, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = select <2 x i1> undef, <2 x i32> undef, <2 x i32> [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i32> <i32 undef, i32 poison>, <2 x i32> [[TMP1]], <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    br i1 undef, label [[DO_COND_I]], label [[IF_ELSE_I:%.*]]
 ; CHECK:       if.else.i:
-; CHECK-NEXT:    [[TMP2:%.*]] = sub <2 x i32> [[TMP1]], undef
+; CHECK-NEXT:    [[TMP3:%.*]] = sub <2 x i32> [[TMP1]], undef
 ; CHECK-NEXT:    br label [[DO_COND_I]]
 ; CHECK:       do.cond.i:
-; CHECK-NEXT:    [[TMP3]] = phi <2 x i32> [ [[TMP2]], [[IF_ELSE_I]] ], [ [[TMP1]], [[DO_BODY66_I]] ]
+; CHECK-NEXT:    [[TMP4]] = phi <2 x i32> [ [[TMP3]], [[IF_ELSE_I]] ], [ [[TMP2]], [[DO_BODY66_I]] ]
 ; CHECK-NEXT:    br i1 undef, label [[DO_BODY66_I]], label [[DO_END1006_I:%.*]]
 ; CHECK:       do.end1006.i:
-; CHECK-NEXT:    [[TMP4:%.*]] = select <2 x i1> undef, <2 x i32> undef, <2 x i32> [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i32* [[RANGE20_I]] to <2 x i32>*
-; CHECK-NEXT:    store <2 x i32> [[TMP4]], <2 x i32>* [[TMP5]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = select <2 x i1> undef, <2 x i32> undef, <2 x i32> [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i32* [[RANGE20_I]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP5]], <2 x i32>* [[TMP6]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %range20.i = getelementptr inbounds %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334, %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p, i64 0, i32 4
   %code21.i = getelementptr inbounds %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334, %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p, i64 0, i32 5
   br label %do.body66.i
 
 do.body66.i:                                      ; preds = %do.cond.i, %entry

llvm/test/Transforms/SLPVectorizer/X86/crash_lencod.ll

 }
 
 define fastcc void @dct36(double* %inbuf) {
 ; CHECK-LABEL: @dct36(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[ARRAYIDX44:%.*]] = getelementptr inbounds double, double* [[INBUF:%.*]], i64 1
 ; CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[INBUF]] to <2 x double>*
 ; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> poison, <2 x i32> <i32 1, i32 undef>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x double> <double poison, double undef>, <2 x double> [[TMP1]], <2 x i32> <i32 3, i32 1>
 ; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x double> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = bitcast double* [[ARRAYIDX44]] to <2 x double>*
 ; CHECK-NEXT:    store <2 x double> [[TMP3]], <2 x double>* [[TMP4]], align 8
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %arrayidx41 = getelementptr inbounds double, double* %inbuf, i64 2
   %arrayidx44 = getelementptr inbounds double, double* %inbuf, i64 1
   %0 = load double, double* %arrayidx44, align 8
   %add46 = fadd double %0, undef
   store double %add46, double* %arrayidx41, align 8
   %1 = load double, double* %inbuf, align 8
   %add49 = fadd double %1, %0
   store double %add49, double* %arrayidx44, align 8
   ret void
 }

llvm/test/Transforms/SLPVectorizer/X86/crash_scheduling.ll

 ; CHECK-NEXT:    [[ADD:%.*]] = fadd double [[MUL20]], 8.192000e+03
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x double> poison, double [[P1:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double [[P2:%.*]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = fmul <2 x double> [[TMP1]], <double 1.638400e+04, double 1.638400e+04>
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> <double 0.000000e+00, double poison>, double [[ADD]], i32 1
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV266:%.*]] = phi i64 [ 0, [[BB1]] ], [ [[INDVARS_IV_NEXT267:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP4:%.*]] = phi <2 x double> [ [[TMP3]], [[BB1]] ], [ [[TMP6:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[X13:%.*]] = tail call i32 @_xfn(<2 x double> [[TMP4]])
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <2 x double> [ [[TMP3]], [[BB1]] ], [ [[TMP7:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <2 x double> undef, <2 x double> [[TMP4]], <2 x i32> <i32 2, i32 1>
+; CHECK-NEXT:    [[X13:%.*]] = tail call i32 @_xfn(<2 x double> [[TMP5]])
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [256 x i32], [256 x i32]* [[TAB1]], i64 0, i64 [[INDVARS_IV266]]
 ; CHECK-NEXT:    store i32 [[X13]], i32* [[ARRAYIDX]], align 4, !tbaa [[TBAA0:![0-9]+]]
-; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <2 x double> [[TMP4]], <2 x double> poison, <2 x i32> <i32 1, i32 undef>
-; CHECK-NEXT:    [[X14:%.*]] = tail call i32 @_xfn(<2 x double> [[TMP5]])
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x double> undef, <2 x double> [[TMP4]], <2 x i32> <i32 3, i32 1>
+; CHECK-NEXT:    [[X14:%.*]] = tail call i32 @_xfn(<2 x double> [[TMP6]])
 ; CHECK-NEXT:    [[ARRAYIDX26:%.*]] = getelementptr inbounds [256 x i32], [256 x i32]* [[TAB2]], i64 0, i64 [[INDVARS_IV266]]
 ; CHECK-NEXT:    store i32 [[X14]], i32* [[ARRAYIDX26]], align 4, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[TMP6]] = fadd <2 x double> [[TMP2]], [[TMP4]]
+; CHECK-NEXT:    [[TMP7]] = fadd <2 x double> [[TMP2]], [[TMP4]]
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT267]] = add nuw nsw i64 [[INDVARS_IV266]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT267]], 256
 ; CHECK-NEXT:    br i1 [[EXITCOND]], label [[RETURN:%.*]], label [[FOR_BODY]]
 ; CHECK:       return:
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %tab1 = alloca [256 x i32], align 16

llvm/test/Transforms/SLPVectorizer/X86/extract-scalar-from-undef.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -slp-vectorizer -mtriple=x86_64-apple-macosx -mattr=+avx2 < %s | FileCheck %s
 
 define i64 @foo(i32 %tmp7) {
 ; CHECK-LABEL: @foo(
 ; CHECK-NEXT:  bb:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i32> <i32 0, i32 0, i32 poison, i32 0>, i32 [[TMP7:%.*]], i32 2
 ; CHECK-NEXT:    [[TMP1:%.*]] = sub <4 x i32> [[TMP0]], zeroinitializer
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <8 x i32> [[TMP2]], i32 undef, i32 4
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i32> [[TMP3]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 2, i32 3, i32 undef, i32 4, i32 undef>
-; CHECK-NEXT:    [[TMP4:%.*]] = sub nsw <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 undef, i32 0, i32 undef, i32 0>, [[SHUFFLE]]
-; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 undef, i32 0, i32 undef, i32 0>, [[SHUFFLE]]
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i32> [[TMP4]], <8 x i32> [[TMP5]], <8 x i32> <i32 0, i32 9, i32 2, i32 3, i32 12, i32 13, i32 6, i32 7>
-; CHECK-NEXT:    [[TMP7:%.*]] = add <8 x i32> zeroinitializer, [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = xor <8 x i32> [[TMP7]], zeroinitializer
-; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP8]])
-; CHECK-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> zeroinitializer)
-; CHECK-NEXT:    [[OP_RDX:%.*]] = add i32 [[TMP9]], [[TMP10]]
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 2, i32 3, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> <i32 poison, i32 poison, i32 undef, i32 poison, i32 poison, i32 undef, i32 poison, i32 undef>, <8 x i32> [[TMP2]], <8 x i32> <i32 8, i32 9, i32 undef, i32 11, i32 12, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <8 x i32> [[TMP3]], i32 undef, i32 6
+; CHECK-NEXT:    [[TMP5:%.*]] = sub nsw <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 undef, i32 0, i32 undef, i32 0>, [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 undef, i32 0, i32 undef, i32 0>, [[TMP4]]
+; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <8 x i32> [[TMP5]], <8 x i32> [[TMP6]], <8 x i32> <i32 0, i32 9, i32 2, i32 3, i32 12, i32 13, i32 6, i32 7>
+; CHECK-NEXT:    [[TMP8:%.*]] = add <8 x i32> zeroinitializer, [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = xor <8 x i32> [[TMP8]], zeroinitializer
+; CHECK-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP9]])
+; CHECK-NEXT:    [[TMP11:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[OP_RDX:%.*]] = add i32 [[TMP10]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP64:%.*]] = zext i32 [[OP_RDX]] to i64
 ; CHECK-NEXT:    ret i64 [[TMP64]]
 ;
 bb:
   %tmp = sub i32 0, 0
   %tmp2 = sub nsw i32 0, %tmp
   %tmp3 = add i32 0, %tmp2
   %tmp4 = xor i32 %tmp3, 0

llvm/test/Transforms/SLPVectorizer/X86/insert-element-build-vector-inseltpoison.ll

 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp ne <2 x i32> [[TMP10]], zeroinitializer
 ; CHECK-NEXT:    [[TMP12:%.*]] = insertelement <2 x float> poison, float [[A2]], i32 0
 ; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <2 x float> [[TMP12]], float [[A3]], i32 1
 ; CHECK-NEXT:    [[TMP14:%.*]] = insertelement <2 x float> poison, float [[B2]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <2 x float> [[TMP14]], float [[B3]], i32 1
 ; CHECK-NEXT:    [[TMP16:%.*]] = select <2 x i1> [[TMP11]], <2 x float> [[TMP13]], <2 x float> [[TMP15]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = shufflevector <2 x float> [[TMP8]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
 ; CHECK-NEXT:    [[TMP18:%.*]] = shufflevector <2 x float> [[TMP16]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
-; CHECK-NEXT:    [[RD1:%.*]] = shufflevector <4 x float> poison, <4 x float> [[TMP18]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    [[RD1:%.*]] = shufflevector <4 x float> poison, <4 x float> [[TMP18]], <4 x i32> <i32 undef, i32 undef, i32 4, i32 5>
 ; CHECK-NEXT:    ret <4 x float> [[RD1]]
 ;
   %c0 = extractelement <4 x i32> %c, i32 0
   %c1 = extractelement <4 x i32> %c, i32 1
   %c2 = extractelement <4 x i32> %c, i32 2
   %c3 = extractelement <4 x i32> %c, i32 3
   %a0 = extractelement <4 x float> %a, i32 0
   %a1 = extractelement <4 x float> %a, i32 1

llvm/test/Transforms/SLPVectorizer/X86/insert-element-build-vector.ll

;
   ret <4 x float> %rd
 }
 
 define <8 x float> @simple_select2(<4 x float> %a, <4 x float> %b, <4 x i32> %c) #0 {
 ; CHECK-LABEL: @simple_select2(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne <4 x i32> [[C:%.*]], zeroinitializer
 ; CHECK-NEXT:    [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x float> [[A:%.*]], <4 x float> [[B:%.*]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x float> [[TMP2]], <4 x float> poison, <8 x i32> <i32 0, i32 undef, i32 1, i32 undef, i32 2, i32 undef, i32 undef, i32 3>
-; CHECK-NEXT:    [[RD1:%.*]] = shufflevector <8 x float> undef, <8 x float> [[TMP3]], <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 6, i32 15>
+; CHECK-NEXT:    [[RD1:%.*]] = shufflevector <8 x float> undef, <8 x float> [[TMP3]], <8 x i32> <i32 8, i32 undef, i32 10, i32 undef, i32 12, i32 undef, i32 undef, i32 15>
 ; CHECK-NEXT:    ret <8 x float> [[RD1]]
 ;
   %c0 = extractelement <4 x i32> %c, i32 0
   %c1 = extractelement <4 x i32> %c, i32 1
   %c2 = extractelement <4 x i32> %c, i32 2
   %c3 = extractelement <4 x i32> %c, i32 3
   %a0 = extractelement <4 x float> %a, i32 0
   %a1 = extractelement <4 x float> %a, i32 1

 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp ne <2 x i32> [[TMP10]], zeroinitializer
 ; CHECK-NEXT:    [[TMP12:%.*]] = insertelement <2 x float> poison, float [[A2]], i32 0
 ; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <2 x float> [[TMP12]], float [[A3]], i32 1
 ; CHECK-NEXT:    [[TMP14:%.*]] = insertelement <2 x float> poison, float [[B2]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <2 x float> [[TMP14]], float [[B3]], i32 1
 ; CHECK-NEXT:    [[TMP16:%.*]] = select <2 x i1> [[TMP11]], <2 x float> [[TMP13]], <2 x float> [[TMP15]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = shufflevector <2 x float> [[TMP8]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
 ; CHECK-NEXT:    [[TMP18:%.*]] = shufflevector <2 x float> [[TMP16]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
-; CHECK-NEXT:    [[RD1:%.*]] = shufflevector <4 x float> undef, <4 x float> [[TMP18]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    [[RD1:%.*]] = shufflevector <4 x float> undef, <4 x float> [[TMP18]], <4 x i32> <i32 undef, i32 undef, i32 4, i32 5>
 ; CHECK-NEXT:    ret <4 x float> [[RD1]]
 ;
   %c0 = extractelement <4 x i32> %c, i32 0
   %c1 = extractelement <4 x i32> %c, i32 1
   %c2 = extractelement <4 x i32> %c, i32 2
   %c3 = extractelement <4 x i32> %c, i32 3
   %a0 = extractelement <4 x float> %a, i32 0
   %a1 = extractelement <4 x float> %a, i32 1

llvm/test/Transforms/SLPVectorizer/X86/insert-shuffle.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -slp-vectorizer -S -o - -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
 
 %struct.sw = type { float, float, float, float }
 
 define { <2 x float>, <2 x float> } @foo(%struct.sw* %v) {
 ; CHECK-LABEL: @foo(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* undef, align 4
 ; CHECK-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_SW:%.*]], %struct.sw* [[V:%.*]], i64 0, i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* undef, align 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[X]] to <2 x float>*
 ; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[TMP2]], align 16
 ; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 1>
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> [[TMP4]], float [[TMP1]], i32 1
-; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> poison, <4 x i32> <i32 0, i32 undef, i32 1, i32 undef>
-; CHECK-NEXT:    [[TMP6:%.*]] = fmul <4 x float> [[SHUFFLE]], [[SHUFFLE1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> <float poison, float undef, float poison, float undef>, float [[TMP0]], i32 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x float> [[TMP4]], float [[TMP1]], i32 2
+; CHECK-NEXT:    [[TMP6:%.*]] = fmul <4 x float> [[SHUFFLE]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = fadd <4 x float> [[TMP6]], undef
 ; CHECK-NEXT:    [[TMP8:%.*]] = fadd <4 x float> [[TMP7]], undef
 ; CHECK-NEXT:    [[TMP9:%.*]] = fadd <4 x float> [[TMP8]], undef
 ; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> poison, <2 x i32> <i32 0, i32 1>
 ; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> poison, <2 x i32> <i32 2, i32 3>
 ; CHECK-NEXT:    [[INS1:%.*]] = insertvalue { <2 x float>, <2 x float> } undef, <2 x float> [[TMP10]], 0
 ; CHECK-NEXT:    [[INS2:%.*]] = insertvalue { <2 x float>, <2 x float> } [[INS1]], <2 x float> [[TMP11]], 1
 ; CHECK-NEXT:    ret { <2 x float>, <2 x float> } [[INS2]]

llvm/test/Transforms/SLPVectorizer/X86/jumbled-load-multiuse.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -slp-vectorizer -S -mtriple=x86_64-unknown-linux -mattr=+sse4.2 | FileCheck %s
 
 @a = common local_unnamed_addr global [4 x i32] zeroinitializer, align 4
 @b = common local_unnamed_addr global [4 x i32] zeroinitializer, align 4
 
 define i32 @fn1() {
 ; CHECK-LABEL: @fn1(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([4 x i32]* @b to <4 x i32>*), align 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp sgt <4 x i32> [[TMP0]], zeroinitializer
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> <i32 8, i32 poison, i32 ptrtoint (i32 ()* @fn1 to i32), i32 ptrtoint (i32 ()* @fn1 to i32)>, <4 x i32> [[TMP0]], <4 x i32> <i32 0, i32 5, i32 2, i32 3>
-; CHECK-NEXT:    [[TMP3:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> <i32 0, i32 6, i32 0, i32 0>
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> poison, <4 x i32> <i32 1, i32 2, i32 3, i32 0>
-; CHECK-NEXT:    store <4 x i32> [[SHUFFLE]], <4 x i32>* bitcast ([4 x i32]* @a to <4 x i32>*), align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> <i32 8, i32 poison, i32 ptrtoint (i32 ()* @fn1 to i32), i32 poison>, <4 x i32> [[TMP0]], <4 x i32> <i32 0, i32 5, i32 2, i32 3>
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 2>
+; CHECK-NEXT:    [[TMP3:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[SHUFFLE]], <4 x i32> <i32 0, i32 6, i32 0, i32 0>
+; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> poison, <4 x i32> <i32 1, i32 2, i32 3, i32 0>
+; CHECK-NEXT:    store <4 x i32> [[SHUFFLE1]], <4 x i32>* bitcast ([4 x i32]* @a to <4 x i32>*), align 4
 ; CHECK-NEXT:    ret i32 0
 ;
   entry:
   %0 = load i32, i32* getelementptr ([4 x i32], [4 x i32]* @b, i64 0, i32 0), align 4
   %cmp = icmp sgt i32 %0, 0
   %cond = select i1 %cmp, i32 8, i32 0
   store i32 %cond, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i32 3), align 4
   %1 = load i32, i32* getelementptr ([4 x i32], [4 x i32]* @b, i64 0, i32 1), align 4

llvm/test/Transforms/SLPVectorizer/X86/landing_pad.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -passes=slp-vectorizer,verify -slp-threshold=-99999 -S | FileCheck %s
 
 target triple = "x86_64-unknown-linux-gnu"
 
 define void @foo() personality i32* ()* @bar {
 ; CHECK-LABEL: @foo(
 ; CHECK-NEXT:  bb1:
 ; CHECK-NEXT:    br label [[BB3:%.*]]
 ; CHECK:       bb2.loopexit:
 ; CHECK-NEXT:    br label [[BB2:%.*]]
 ; CHECK:       bb2:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi <4 x i32> [ [[SHUFFLE:%.*]], [[BB9:%.*]] ], [ poison, [[BB2_LOOPEXIT:%.*]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = phi <4 x i32> [ [[TMP10:%.*]], [[BB9:%.*]] ], [ poison, [[BB2_LOOPEXIT:%.*]] ]
 ; CHECK-NEXT:    ret void
 ; CHECK:       bb3:
 ; CHECK-NEXT:    [[TMP1:%.*]] = phi <2 x i32> [ [[TMP5:%.*]], [[BB6:%.*]] ], [ poison, [[BB1:%.*]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = invoke i32 poison(i8 addrspace(1)* nonnull poison, i32 0, i32 0, i32 poison) [ "deopt"() ]
 ; CHECK-NEXT:    to label [[BB4:%.*]] unwind label [[BB10:%.*]]
 ; CHECK:       bb4:
 ; CHECK-NEXT:    br i1 poison, label [[BB11:%.*]], label [[BB5:%.*]]
 ; CHECK:       bb5:
 ; CHECK-NEXT:    br label [[BB7:%.*]]
 ; CHECK:       bb6:
 ; CHECK-NEXT:    [[TMP3:%.*]] = phi <2 x i32> [ <i32 0, i32 poison>, [[BB8:%.*]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[TMP3]], i32 1
 ; CHECK-NEXT:    [[TMP5]] = insertelement <2 x i32> poison, i32 [[TMP4]], i32 1
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb7:
 ; CHECK-NEXT:    [[LOCAL_5_84111:%.*]] = phi i32 [ poison, [[BB8]] ], [ poison, [[BB5]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x i32> poison, i32 [[LOCAL_5_84111]], i32 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = invoke i32 poison(i8 addrspace(1)* nonnull poison, i32 poison, i32 poison, i32 poison) [ "deopt"() ]
 ; CHECK-NEXT:    to label [[BB8]] unwind label [[BB12:%.*]]
 ; CHECK:       bb8:
 ; CHECK-NEXT:    br i1 poison, label [[BB7]], label [[BB6]]
 ; CHECK:       bb9:
 ; CHECK-NEXT:    [[INDVARS_IV528799:%.*]] = phi i64 [ poison, [[BB10]] ], [ poison, [[BB12]] ]
-; CHECK-NEXT:    [[TMP8:%.*]] = phi <2 x i32> [ [[SHUFFLE1:%.*]], [[BB10]] ], [ [[TMP11:%.*]], [[BB12]] ]
+; CHECK-NEXT:    [[TMP8:%.*]] = phi <2 x i32> [ [[SHUFFLE:%.*]], [[BB10]] ], [ [[TMP12:%.*]], [[BB12]] ]
 ; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x i32> [[TMP8]], <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
-; CHECK-NEXT:    [[SHUFFLE]] = shufflevector <4 x i32> [[TMP9]], <4 x i32> poison, <4 x i32> <i32 undef, i32 undef, i32 0, i32 1>
+; CHECK-NEXT:    [[TMP10]] = shufflevector <4 x i32> poison, <4 x i32> [[TMP9]], <4 x i32> <i32 undef, i32 undef, i32 4, i32 5>
 ; CHECK-NEXT:    br label [[BB2]]
 ; CHECK:       bb10:
-; CHECK-NEXT:    [[TMP10:%.*]] = phi <2 x i32> [ [[TMP1]], [[BB3]] ]
+; CHECK-NEXT:    [[TMP11:%.*]] = phi <2 x i32> [ [[TMP1]], [[BB3]] ]
 ; CHECK-NEXT:    [[LANDING_PAD68:%.*]] = landingpad { i8*, i32 }
 ; CHECK-NEXT:    cleanup
-; CHECK-NEXT:    [[SHUFFLE1]] = shufflevector <2 x i32> [[TMP10]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
+; CHECK-NEXT:    [[SHUFFLE]] = shufflevector <2 x i32> [[TMP11]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
 ; CHECK-NEXT:    br label [[BB9]]
 ; CHECK:       bb11:
 ; CHECK-NEXT:    ret void
 ; CHECK:       bb12:
-; CHECK-NEXT:    [[TMP11]] = phi <2 x i32> [ [[TMP6]], [[BB7]] ]
+; CHECK-NEXT:    [[TMP12]] = phi <2 x i32> [ [[TMP6]], [[BB7]] ]
 ; CHECK-NEXT:    [[LANDING_PAD149:%.*]] = landingpad { i8*, i32 }
 ; CHECK-NEXT:    cleanup
 ; CHECK-NEXT:    br label [[BB9]]
 ;
 bb1:
   br label %bb3
 
 bb2.loopexit:

llvm/test/Transforms/SLPVectorizer/X86/load-partial-vector-shuffle.ll

 ;
 ; AVX-LABEL: @load_00123456(
 ; AVX-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, ptr [[DATA:%.*]], i64 2
 ; AVX-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, ptr [[DATA]], i64 3
 ; AVX-NEXT:    [[TMP1:%.*]] = load <2 x i16>, ptr [[DATA]], align 2
 ; AVX-NEXT:    [[T2:%.*]] = load i16, ptr [[ARRAYIDX2]], align 2
 ; AVX-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[ARRAYIDX3]], align 2
 ; AVX-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i16> [[TMP1]], <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[VECINIT2_I_I2:%.*]] = shufflevector <8 x i16> [[TMP3]], <8 x i16> [[TMP3]], <8 x i32> <i32 0, i32 8, i32 9, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i16> undef, <8 x i16> [[TMP3]], <8 x i32> <i32 8, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX-NEXT:    [[VECINIT2_I_I2:%.*]] = shufflevector <8 x i16> [[TMP4]], <8 x i16> [[TMP3]], <8 x i32> <i32 0, i32 8, i32 9, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; AVX-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I_I2]], i16 [[T2]], i64 3
-; AVX-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[TMP2]], <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[VECINIT7_I_I1:%.*]] = shufflevector <8 x i16> [[VECINIT3_I_I]], <8 x i16> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+; AVX-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i16> [[TMP2]], <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX-NEXT:    [[VECINIT7_I_I1:%.*]] = shufflevector <8 x i16> [[VECINIT3_I_I]], <8 x i16> [[TMP5]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
 ; AVX-NEXT:    [[T7:%.*]] = bitcast <8 x i16> [[VECINIT7_I_I1]] to <2 x i64>
 ; AVX-NEXT:    ret <2 x i64> [[T7]]
 ;
   %arrayidx1 = getelementptr inbounds i16, ptr %data, i64 1
   %arrayidx2 = getelementptr inbounds i16, ptr %data, i64 2
   %arrayidx3 = getelementptr inbounds i16, ptr %data, i64 3
   %arrayidx4 = getelementptr inbounds i16, ptr %data, i64 4
   %arrayidx5 = getelementptr inbounds i16, ptr %data, i64 5

llvm/test/Transforms/SLPVectorizer/X86/matched-shuffled-entries.ll

 ; CHECK-NEXT:    [[SUB102_1:%.*]] = sub nsw i32 undef, undef
 ; CHECK-NEXT:    [[ADD78_2:%.*]] = add nsw i32 undef, undef
 ; CHECK-NEXT:    [[SUB102_3:%.*]] = sub nsw i32 undef, undef
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <16 x i32> poison, i32 [[SUB102_1]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <16 x i32> [[TMP0]], i32 [[ADD94_1]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <16 x i32> [[TMP1]], i32 [[ADD78_1]], i32 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <16 x i32> [[TMP2]], i32 [[SUB86_1]], i32 3
 ; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <16 x i32> [[TMP3]], i32 [[ADD78_2]], i32 4
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <16 x i32> [[TMP4]], <16 x i32> poison, <16 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 0, i32 1, i32 2, i32 3, i32 undef, i32 4, i32 4, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <16 x i32> [[TMP4]], <16 x i32> poison, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 3, i32 4, i32 0, i32 5, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
 ; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <16 x i32> poison, i32 [[SUB86_1]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <16 x i32> [[TMP5]], i32 [[ADD78_1]], i32 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <16 x i32> [[TMP6]], i32 [[ADD94_1]], i32 2
 ; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <16 x i32> [[TMP7]], i32 [[SUB102_1]], i32 3
 ; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <16 x i32> [[TMP8]], i32 [[SUB102_3]], i32 4
-; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <16 x i32> [[TMP9]], <16 x i32> poison, <16 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 4, i32 undef, i32 undef, i32 4>
+; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <16 x i32> [[TMP9]], <16 x i32> poison, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 3, i32 4, i32 0, i32 0, i32 0, i32 0, i32 5, i32 0, i32 0, i32 5>
 ; CHECK-NEXT:    [[TMP10:%.*]] = add nsw <16 x i32> [[SHUFFLE]], [[SHUFFLE1]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = sub nsw <16 x i32> [[SHUFFLE]], [[SHUFFLE1]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = shufflevector <16 x i32> [[TMP10]], <16 x i32> [[TMP11]], <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 4, i32 5, i32 22, i32 23, i32 8, i32 9, i32 26, i32 27, i32 12, i32 13, i32 30, i32 31>
 ; CHECK-NEXT:    [[TMP13:%.*]] = lshr <16 x i32> [[TMP12]], <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
 ; CHECK-NEXT:    [[TMP14:%.*]] = and <16 x i32> [[TMP13]], <i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537>
 ; CHECK-NEXT:    [[TMP15:%.*]] = mul nuw <16 x i32> [[TMP14]], <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
 ; CHECK-NEXT:    [[TMP16:%.*]] = add <16 x i32> [[TMP15]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = xor <16 x i32> [[TMP16]], [[TMP15]]

llvm/test/Transforms/SLPVectorizer/X86/partail.ll

 ; CHECK-NEXT:    unreachable
 ; CHECK:       if.end:
 ; CHECK-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[Y_POS:%.*]], undef
 ; CHECK-NEXT:    [[SHR15:%.*]] = ashr i32 [[SUB14]], 2
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i32> poison, i32 [[SHR15]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32> [[TMP0]], i32 [[SUB14]], i32 1
 ; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp sgt <4 x i32> [[SHUFFLE]], <i32 0, i32 -1, i32 -5, i32 -9>
-; CHECK-NEXT:    [[TMP3:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP0]], <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp slt <4 x i32> [[TMP3]], undef
-; CHECK-NEXT:    [[TMP5:%.*]] = select <4 x i1> [[TMP4]], <4 x i32> [[TMP3]], <4 x i32> undef
-; CHECK-NEXT:    [[TMP6:%.*]] = sext <4 x i32> [[TMP5]] to <4 x i64>
-; CHECK-NEXT:    [[TMP7:%.*]] = trunc <4 x i64> [[TMP6]] to <4 x i32>
-; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <4 x i32> [[TMP7]], i32 0
-; CHECK-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
-; CHECK-NEXT:    [[ARRAYIDX31:%.*]] = getelementptr inbounds i16*, i16** undef, i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i32> [[TMP7]], i32 1
-; CHECK-NEXT:    [[TMP11:%.*]] = sext i32 [[TMP10]] to i64
-; CHECK-NEXT:    [[ARRAYIDX31_1:%.*]] = getelementptr inbounds i16*, i16** undef, i64 [[TMP11]]
-; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <4 x i32> [[TMP7]], i32 2
-; CHECK-NEXT:    [[TMP13:%.*]] = sext i32 [[TMP12]] to i64
-; CHECK-NEXT:    [[ARRAYIDX31_2:%.*]] = getelementptr inbounds i16*, i16** undef, i64 [[TMP13]]
-; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <4 x i32> [[TMP7]], i32 3
-; CHECK-NEXT:    [[TMP15:%.*]] = sext i32 [[TMP14]] to i64
-; CHECK-NEXT:    [[ARRAYIDX31_3:%.*]] = getelementptr inbounds i16*, i16** undef, i64 [[TMP15]]
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32> <i32 poison, i32 undef, i32 undef, i32 undef>, i32 [[SHR15]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp slt <4 x i32> [[TMP4]], undef
+; CHECK-NEXT:    [[TMP6:%.*]] = select <4 x i1> [[TMP5]], <4 x i32> [[TMP4]], <4 x i32> undef
+; CHECK-NEXT:    [[TMP7:%.*]] = sext <4 x i32> [[TMP6]] to <4 x i64>
+; CHECK-NEXT:    [[TMP8:%.*]] = trunc <4 x i64> [[TMP7]] to <4 x i32>
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <4 x i32> [[TMP8]], i32 0
+; CHECK-NEXT:    [[TMP10:%.*]] = sext i32 [[TMP9]] to i64
+; CHECK-NEXT:    [[ARRAYIDX31:%.*]] = getelementptr inbounds i16*, i16** undef, i64 [[TMP10]]
+; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <4 x i32> [[TMP8]], i32 1
+; CHECK-NEXT:    [[TMP12:%.*]] = sext i32 [[TMP11]] to i64
+; CHECK-NEXT:    [[ARRAYIDX31_1:%.*]] = getelementptr inbounds i16*, i16** undef, i64 [[TMP12]]
+; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <4 x i32> [[TMP8]], i32 2
+; CHECK-NEXT:    [[TMP14:%.*]] = sext i32 [[TMP13]] to i64
+; CHECK-NEXT:    [[ARRAYIDX31_2:%.*]] = getelementptr inbounds i16*, i16** undef, i64 [[TMP14]]
+; CHECK-NEXT:    [[TMP15:%.*]] = extractelement <4 x i32> [[TMP8]], i32 3
+; CHECK-NEXT:    [[TMP16:%.*]] = sext i32 [[TMP15]] to i64
+; CHECK-NEXT:    [[ARRAYIDX31_3:%.*]] = getelementptr inbounds i16*, i16** undef, i64 [[TMP16]]
 ; CHECK-NEXT:    unreachable
 ;
 entry:
   br label %land.lhs.true
 
 land.lhs.true:                                    ; preds = %entry
   br i1 undef, label %if.then, label %if.end
 

llvm/test/Transforms/SLPVectorizer/X86/phi-undef-input.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -slp-vectorizer -slp-threshold=-1000 -mtriple=x86_64 -S | FileCheck %s
 
 ; The inputs to vector phi should remain undef.
 
 define i32 @phi3UndefInput(i1 %cond, i8 %arg0, i8 %arg1, i8 %arg2, i8 %arg3) {
 ; CHECK-LABEL: @phi3UndefInput(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br i1 [[COND:%.*]], label [[BB2:%.*]], label [[BB3:%.*]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i8> poison, i8 [[ARG0:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i8> [[TMP0]], i8 [[ARG1:%.*]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i8> [[TMP1]], i8 [[ARG2:%.*]], i32 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i8> [[TMP2]], i8 [[ARG3:%.*]], i32 3
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb3:
-; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 poison, i8 poison, i8 poison>, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 undef, i8 undef, i8 undef>, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext <4 x i8> [[TMP4]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP5]])
 ; CHECK-NEXT:    ret i32 [[TMP6]]
 ;
 entry:
   br i1 %cond, label %bb2, label %bb3
 
 bb2:

 ; CHECK-NEXT:    br i1 [[COND:%.*]], label [[BB2:%.*]], label [[BB3:%.*]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i8> poison, i8 [[ARG0:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i8> [[TMP0]], i8 [[ARG1:%.*]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i8> [[TMP1]], i8 [[ARG2:%.*]], i32 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i8> [[TMP2]], i8 [[ARG3:%.*]], i32 3
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb3:
-; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 0, i8 poison, i8 poison>, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 0, i8 undef, i8 undef>, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext <4 x i8> [[TMP4]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP5]])
 ; CHECK-NEXT:    ret i32 [[TMP6]]
 ;
 entry:
   br i1 %cond, label %bb2, label %bb3
 
 bb2:

 ; CHECK-NEXT:    br i1 [[COND:%.*]], label [[BB2:%.*]], label [[BB3:%.*]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i8> poison, i8 [[ARG0:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i8> [[TMP0]], i8 [[ARG1:%.*]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i8> [[TMP1]], i8 [[ARG2:%.*]], i32 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i8> [[TMP2]], i8 [[ARG3:%.*]], i32 3
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb3:
-; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 0, i8 0, i8 poison>, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 0, i8 0, i8 undef>, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext <4 x i8> [[TMP4]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP5]])
 ; CHECK-NEXT:    ret i32 [[TMP6]]
 ;
 entry:
   br i1 %cond, label %bb2, label %bb3
 
 bb2:

 ; CHECK-NEXT:    br i1 [[COND:%.*]], label [[BB2:%.*]], label [[BB3:%.*]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i8> poison, i8 [[ARG0:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i8> [[TMP0]], i8 [[ARG1:%.*]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i8> [[TMP1]], i8 [[ARG2:%.*]], i32 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i8> [[TMP2]], i8 [[ARG3:%.*]], i32 3
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb3:
-; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 0, i8 poison, i8 poison>, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 0, i8 poison, i8 undef>, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext <4 x i8> [[TMP4]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP5]])
 ; CHECK-NEXT:    ret i32 [[TMP6]]
 ;
 entry:
   br i1 %cond, label %bb2, label %bb3
 
 bb2:

 ; CHECK-NEXT:    br i1 [[COND:%.*]], label [[BB2:%.*]], label [[BB3:%.*]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i8> poison, i8 [[ARG1:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i8> [[TMP0]], i8 [[ARG0:%.*]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i8> [[TMP1]], i8 [[ARG2:%.*]], i32 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i8> [[TMP2]], i8 [[ARG3:%.*]], i32 3
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb3:
-; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 poison, i8 poison, i8 poison>, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 poison, i8 poison, i8 undef>, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext <4 x i8> [[TMP4]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP5]])
 ; CHECK-NEXT:    ret i32 [[TMP6]]
 ;
 entry:
   br i1 %cond, label %bb2, label %bb3
 
 bb2:

 ; CHECK-NEXT:    br i1 [[COND:%.*]], label [[BB2:%.*]], label [[BB3:%.*]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i8> poison, i8 [[ARG1:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i8> [[TMP0]], i8 [[ARG3:%.*]], i32 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i8> [[TMP1]], i8 [[ARG0:%.*]], i32 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i8> [[TMP2]], i8 [[ARG2:%.*]], i32 3
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb3:
-; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 0, i8 poison, i8 poison>, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i8> [ [[TMP3]], [[BB2]] ], [ <i8 0, i8 0, i8 poison, i8 undef>, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext <4 x i8> [[TMP4]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP5]])
 ; CHECK-NEXT:    ret i32 [[TMP6]]
 ;
 entry:
   br i1 %cond, label %bb2, label %bb3
 
 bb2:

llvm/test/Transforms/SLPVectorizer/X86/remark_extract_broadcast.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -mtriple=x86_64-pc-linux-gnu -mcpu=generic -mattr=sse2 -slp-vectorizer -pass-remarks-output=%t < %s -slp-threshold=-2 | FileCheck %s
 ; RUN: FileCheck --input-file=%t --check-prefix=YAML %s
 
 define void @fextr(i16* %ptr) {
 ; CHECK-LABEL: @fextr(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[LD:%.*]] = load <8 x i16>, <8 x i16>* undef, align 16
 ; CHECK-NEXT:    br label [[T:%.*]]
 ; CHECK:       t:
 ; CHECK-NEXT:    [[P0:%.*]] = getelementptr inbounds i16, i16* [[PTR:%.*]], i64 0
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i16> [[LD]], <8 x i16> poison, <8 x i32> <i32 0, i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i16> [[LD]], <8 x i16> poison, <8 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP0:%.*]] = add <8 x i16> [[LD]], [[SHUFFLE]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <8 x i16>*
 ; CHECK-NEXT:    store <8 x i16> [[TMP0]], <8 x i16>* [[TMP1]], align 2
 ; CHECK-NEXT:    ret void
 ;
 ; YAML:      Pass:            slp-vectorizer
 ; YAML-NEXT: Name:            StoresVectorized
 ; YAML-NEXT: Function:        fextr

llvm/test/Transforms/SLPVectorizer/X86/reused-undefs.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -slp-threshold=-1000 < %s | FileCheck %s
 
 define i32 @main(i32 %0) {
 ; CHECK-LABEL: @main(
 ; CHECK-NEXT:  for.cond.preheader:
 ; CHECK-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[FOR_INC_PREHEADER:%.*]]
 ; CHECK:       for.inc.preheader:
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 poison, i32 poison>, i32 [[TMP0:%.*]], i32 6
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 poison, i32 undef>, i32 [[TMP0:%.*]], i32 6
 ; CHECK-NEXT:    br i1 false, label [[FOR_END]], label [[L1_PREHEADER:%.*]]
 ; CHECK:       for.end:
 ; CHECK-NEXT:    [[DOTPR:%.*]] = phi i32 [ 0, [[FOR_INC_PREHEADER]] ], [ 0, [[FOR_COND_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <8 x i32> poison, i32 [[DOTPR]], i32 0
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i32> [[TMP2]], <8 x i32> poison, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0>
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i32> [[TMP2]], <8 x i32> poison, <8 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[L1_PREHEADER]]
 ; CHECK:       L1.preheader:
 ; CHECK-NEXT:    [[TMP3:%.*]] = phi <8 x i32> [ [[SHUFFLE]], [[FOR_END]] ], [ [[TMP1]], [[FOR_INC_PREHEADER]] ]
 ; CHECK-NEXT:    ret i32 0
 ;
 for.cond.preheader:
   br i1 false, label %for.end, label %for.inc.preheader
 

llvm/test/Transforms/SLPVectorizer/X86/vectorize-widest-phis.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -slp-vectorizer -S -mcpu=cascadelake -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
 
 define void @foo() {
 ; CHECK-LABEL: @foo(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[CONV:%.*]] = uitofp i16 undef to float
 ; CHECK-NEXT:    [[SUB:%.*]] = fsub float 6.553500e+04, undef
 ; CHECK-NEXT:    br label [[BB1:%.*]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x float> poison, float [[SUB]], i32 0
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x float> <float poison, float poison, float undef, float undef>, float [[SUB]], i32 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x float> [[TMP0]], float [[CONV]], i32 1
 ; CHECK-NEXT:    br label [[BB2:%.*]]
 ; CHECK:       bb2:
-; CHECK-NEXT:    [[TMP2:%.*]] = phi <4 x float> [ [[TMP1]], [[BB1]] ], [ [[TMP14:%.*]], [[BB3:%.*]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = phi <4 x float> [ [[TMP1]], [[BB1]] ], [ [[TMP15:%.*]], [[BB3:%.*]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = load double, double* undef, align 8
 ; CHECK-NEXT:    br i1 undef, label [[BB3]], label [[BB4:%.*]]
 ; CHECK:       bb4:
 ; CHECK-NEXT:    [[TMP4:%.*]] = fpext <4 x float> [[TMP2]] to <4 x double>
 ; CHECK-NEXT:    [[CONV2:%.*]] = uitofp i16 undef to double
 ; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x double> <double undef, double poison>, double [[TMP3]], i32 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x double> <double undef, double poison>, double [[CONV2]], i32 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = fsub <2 x double> [[TMP5]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = fadd <2 x double> [[TMP5]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x double> [[TMP7]], <2 x double> [[TMP8]], <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x double> [[TMP9]], <2 x double> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
-; CHECK-NEXT:    [[TMP11:%.*]] = fcmp ogt <4 x double> [[TMP10]], [[TMP4]]
-; CHECK-NEXT:    [[TMP12:%.*]] = fptrunc <4 x double> [[TMP10]] to <4 x float>
-; CHECK-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP11]], <4 x float> [[TMP2]], <4 x float> [[TMP12]]
+; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x double> <double poison, double poison, double undef, double undef>, <4 x double> [[TMP10]], <4 x i32> <i32 4, i32 5, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP12:%.*]] = fcmp ogt <4 x double> [[TMP11]], [[TMP4]]
+; CHECK-NEXT:    [[TMP13:%.*]] = fptrunc <4 x double> [[TMP10]] to <4 x float>
+; CHECK-NEXT:    [[TMP14:%.*]] = select <4 x i1> [[TMP12]], <4 x float> [[TMP2]], <4 x float> [[TMP13]]
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb3:
-; CHECK-NEXT:    [[TMP14]] = phi <4 x float> [ [[TMP13]], [[BB4]] ], [ [[TMP2]], [[BB2]] ]
+; CHECK-NEXT:    [[TMP15]] = phi <4 x float> [ [[TMP14]], [[BB4]] ], [ [[TMP2]], [[BB2]] ]
 ; CHECK-NEXT:    br label [[BB2]]
 ;
 entry:
   %conv = uitofp i16 undef to float
   %sub = fsub float 6.553500e+04, undef
   br label %bb1
 
 bb1: