[TTI] Fix uses of SK_ExtractSubvector shuffle costs (PR39368)

[RKSimon]

Summary

Correct costings of SK_ExtractSubvector requires the SubTy argument to indicate the type/size of the extracted subvector.

Unlike the rest of the shuffle kinds this means that the main Ty argument represents the source vector type not the destination - if anyone wants to 'flip' this now is the time to ask before we start adding more reduction support!

I've done my best to fix a number of vectorizer uses:

SLP - AFAICT the reduction epilogue costs should be using a SK_PermuteSingleSrc shuffle as these all occur at the hardware vector width - we're not extracting (illegal) subvector types. This is causing the cost model diffs as SK_ExtractSubvector costs are poorly handled and tend to just return 1 at the moment.

LV - I'm not clear on what the SK_ExtractSubvector should represents for recurrences - I've used a <1 x ?> subvector extraction as that seems to match the VF delta but I need help here (@fhahn? @mssimpso?).

[RKSimon]

ping

[ABataev]

LG

include/llvm/Analysis/TargetTransformInfo.h

int getArithmeticInstrCost(
       unsigned Opcode, Type *Ty, OperandValueKind Opd1Info = OK_AnyValue,
       OperandValueKind Opd2Info = OK_AnyValue,
       OperandValueProperties Opd1PropInfo = OP_None,
       OperandValueProperties Opd2PropInfo = OP_None,
       ArrayRef<const Value *> Args = ArrayRef<const Value *>()) const;
 
   /// \return The cost of a shuffle instruction of kind Kind and of type Tp.
   /// The index and subtype parameters are used by the subvector insertion and
-  /// extraction shuffle kinds.
+  /// extraction shuffle kinds to show the insert/extract point and the type of
+  /// the subvector being inserted/extracted. 
+  /// NOTE: For subvector extractions Tp represents the source type.
   int getShuffleCost(ShuffleKind Kind, Type *Tp, int Index = 0,
                      Type *SubTp = nullptr) const;
 
   /// \return The expected cost of cast instructions, such as bitcast, trunc,
   /// zext, etc. If there is an existing instruction that holds Opcode, it
   /// may be passed in the 'I' parameter.
   int getCastInstrCost(unsigned Opcode, Type *Dst, Type *Src,
                        const Instruction *I = nullptr) const;

include/llvm/CodeGen/BasicTTIImpl.h

unsigned getArithmeticReductionCost(unsigned Opcode, Type *Ty,
     auto *ConcreteTTI = static_cast<T *>(this);
     std::pair<unsigned, MVT> LT =
         ConcreteTTI->getTLI()->getTypeLegalizationCost(DL, Ty);
     unsigned LongVectorCount = 0;
     unsigned MVTLen =
         LT.second.isVector() ? LT.second.getVectorNumElements() : 1;
     while (NumVecElts > MVTLen) {
       NumVecElts /= 2;
+      Type *SubTy = VectorType::get(ScalarTy, NumVecElts);
       // Assume the pairwise shuffles add a cost.
       ShuffleCost += (IsPairwise + 1) *
                      ConcreteTTI->getShuffleCost(TTI::SK_ExtractSubvector, Ty,
-                                                 NumVecElts, Ty);
+                                                 NumVecElts, SubTy);
       ArithCost += ConcreteTTI->getArithmeticInstrCost(Opcode, Ty);
-      Ty = VectorType::get(ScalarTy, NumVecElts);
+      Ty = SubTy;
       ++LongVectorCount;
     }
     // The minimal length of the vector is limited by the real length of vector
     // operations performed on the current platform. That's why several final
     // reduction operations are performed on the vectors with the same
     // architecture-dependent length.
     ShuffleCost += (NumReduxLevels - LongVectorCount) * (IsPairwise + 1) *
-                   ConcreteTTI->getShuffleCost(TTI::SK_ExtractSubvector, Ty,
-                                               NumVecElts, Ty);
+                   ConcreteTTI->getShuffleCost(TTI::SK_PermuteSingleSrc, Ty,
+                                               0, Ty);
     ArithCost += (NumReduxLevels - LongVectorCount) *
                  ConcreteTTI->getArithmeticInstrCost(Opcode, Ty);
     return ShuffleCost + ArithCost + getScalarizationOverhead(Ty, false, true);
   }
 
   /// Try to calculate op costs for min/max reduction operations.
   /// \param CondTy Conditional type for the Select instruction.
   unsigned getMinMaxReductionCost(Type *Ty, Type *CondTy, bool IsPairwise,
unsigned getMinMaxReductionCost(Type *Ty, Type *CondTy, bool IsPairwise,
     auto *ConcreteTTI = static_cast<T *>(this);
     std::pair<unsigned, MVT> LT =
         ConcreteTTI->getTLI()->getTypeLegalizationCost(DL, Ty);
     unsigned LongVectorCount = 0;
     unsigned MVTLen =
         LT.second.isVector() ? LT.second.getVectorNumElements() : 1;
     while (NumVecElts > MVTLen) {
       NumVecElts /= 2;
+      Type *SubTy = VectorType::get(ScalarTy, NumVecElts);
       // Assume the pairwise shuffles add a cost.
       ShuffleCost += (IsPairwise + 1) *
                      ConcreteTTI->getShuffleCost(TTI::SK_ExtractSubvector, Ty,
-                                                 NumVecElts, Ty);
+                                                 NumVecElts, SubTy);
       MinMaxCost +=
           ConcreteTTI->getCmpSelInstrCost(CmpOpcode, Ty, CondTy, nullptr) +
           ConcreteTTI->getCmpSelInstrCost(Instruction::Select, Ty, CondTy,
                                           nullptr);
-      Ty = VectorType::get(ScalarTy, NumVecElts);
+      Ty = SubTy;
       CondTy = VectorType::get(ScalarCondTy, NumVecElts);
       ++LongVectorCount;
     }
     // The minimal length of the vector is limited by the real length of vector
     // operations performed on the current platform. That's why several final
     // reduction opertions are perfomed on the vectors with the same
     // architecture-dependent length.
     ShuffleCost += (NumReduxLevels - LongVectorCount) * (IsPairwise + 1) *
-                   ConcreteTTI->getShuffleCost(TTI::SK_ExtractSubvector, Ty,
-                                               NumVecElts, Ty);
+                   ConcreteTTI->getShuffleCost(TTI::SK_PermuteSingleSrc, Ty,
+                                               0, Ty);
     MinMaxCost +=
         (NumReduxLevels - LongVectorCount) *
         (ConcreteTTI->getCmpSelInstrCost(CmpOpcode, Ty, CondTy, nullptr) +
          ConcreteTTI->getCmpSelInstrCost(Instruction::Select, Ty, CondTy,
                                          nullptr));
     // Need 3 extractelement instructions for scalarization + an additional
     // scalar select instruction.
     return ShuffleCost + MinMaxCost +

lib/Transforms/Vectorize/LoopVectorize.cpp

case Instruction::Br: {
     // may decide in the future to call TTI for all branches.
   }
   case Instruction::PHI: {
     auto *Phi = cast<PHINode>(I);
 
     // First-order recurrences are replaced by vector shuffles inside the loop.
     if (VF > 1 && Legal->isFirstOrderRecurrence(Phi))
       return TTI.getShuffleCost(TargetTransformInfo::SK_ExtractSubvector,
-                                VectorTy, VF - 1, VectorTy);
+                                VectorTy, VF - 1, ToVectorTy(RetTy, 1));
 
     // Phi nodes in non-header blocks (not inductions, reductions, etc.) are
     // converted into select instructions. We require N - 1 selects per phi
     // node, where N is the number of incoming values.
     if (VF > 1 && Phi->getParent() != TheLoop->getHeader())
       return (Phi->getNumIncomingValues() - 1) *
              TTI.getCmpSelInstrCost(
                  Instruction::Select, ToVectorTy(Phi->getType(), VF),

test/Analysis/CostModel/AArch64/vector-reduce.ll

 ; CODE-LABEL: add.i32.v4i32
 ; CODE:       addv s0, v0.4s
 define i32 @add.i32.v4i32(<4 x i32> %v) {
   %r = call i32 @llvm.experimental.vector.reduce.add.i32.v4i32(<4 x i32> %v)
   ret i32 %r
 }
 
 ; COST-LABEL: umin.i8.v8i8
-; COST:       Found an estimated cost of 157 for instruction: %r = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> %v)
+; COST:       Found an estimated cost of 280 for instruction: %r = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> %v)
 ; CODE-LABEL: umin.i8.v8i8
 ; CODE:       uminv b0, v0.8b
 define i8 @umin.i8.v8i8(<8 x i8> %v) {
   %r = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> %v)
   ret i8 %r
 }
 
 ; COST-LABEL: umin.i8.v16i8
-; COST:       Found an estimated cost of 388 for instruction: %r = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> %v)
+; COST:       Found an estimated cost of 744 for instruction: %r = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> %v)
 ; CODE-LABEL: umin.i8.v16i8
 ; CODE:       uminv b0, v0.16b
 define i8 @umin.i8.v16i8(<16 x i8> %v) {
   %r = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> %v)
   ret i8 %r
 }
 
 ; COST-LABEL: umin.i16.v4i16
-; COST:       Found an estimated cost of 58 for instruction: %r = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> %v)
+; COST:       Found an estimated cost of 92 for instruction: %r = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> %v)
 ; CODE-LABEL: umin.i16.v4i16
 ; CODE:       uminv h0, v0.4h
 define i16 @umin.i16.v4i16(<4 x i16> %v) {
   %r = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> %v)
   ret i16 %r
 }
 
 ; COST-LABEL: umin.i16.v8i16
-; COST:       Found an estimated cost of 157 for instruction: %r = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> %v)
+; COST:       Found an estimated cost of 280 for instruction: %r = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> %v)
 ; CODE-LABEL: umin.i16.v8i16
 ; CODE:       uminv h0, v0.8h
 define i16 @umin.i16.v8i16(<8 x i16> %v) {
   %r = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> %v)
   ret i16 %r
 }
 
 ; COST-LABEL: umin.i32.v4i32
-; COST:       Found an estimated cost of 58 for instruction: %r = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> %v)
+; COST:       Found an estimated cost of 62 for instruction: %r = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> %v)
 ; CODE-LABEL: umin.i32.v4i32
 ; CODE:       uminv s0, v0.4s
 define i32 @umin.i32.v4i32(<4 x i32> %v) {
   %r = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> %v)
   ret i32 %r
 }
 
 ; COST-LABEL: umax.i8.v8i8
-; COST:       Found an estimated cost of 157 for instruction: %r = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> %v)
+; COST:       Found an estimated cost of 280 for instruction: %r = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> %v)
 ; CODE-LABEL: umax.i8.v8i8
 ; CODE:       umaxv b0, v0.8b
 define i8 @umax.i8.v8i8(<8 x i8> %v) {
   %r = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> %v)
   ret i8 %r
 }
 
 ; COST-LABEL: umax.i8.v16i8
-; COST:       Found an estimated cost of 388 for instruction: %r = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> %v)
+; COST:       Found an estimated cost of 744 for instruction: %r = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> %v)
 ; CODE-LABEL: umax.i8.v16i8
 ; CODE:       umaxv b0, v0.16b
 define i8 @umax.i8.v16i8(<16 x i8> %v) {
   %r = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> %v)
   ret i8 %r
 }
 
 ; COST-LABEL: umax.i16.v4i16
-; COST:       Found an estimated cost of 58 for instruction: %r = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> %v)
+; COST:       Found an estimated cost of 92 for instruction: %r = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> %v)
 ; CODE-LABEL: umax.i16.v4i16
 ; CODE:       umaxv h0, v0.4h
 define i16 @umax.i16.v4i16(<4 x i16> %v) {
   %r = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> %v)
   ret i16 %r
 }
 
 ; COST-LABEL: umax.i16.v8i16
-; COST:       Found an estimated cost of 157 for instruction: %r = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> %v)
+; COST:       Found an estimated cost of 280 for instruction: %r = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> %v)
 ; CODE-LABEL: umax.i16.v8i16
 ; CODE:       umaxv h0, v0.8h
 define i16 @umax.i16.v8i16(<8 x i16> %v) {
   %r = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> %v)
   ret i16 %r
 }
 
 ; COST-LABEL: umax.i32.v4i32
-; COST:       Found an estimated cost of 58 for instruction: %r = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> %v)
+; COST:       Found an estimated cost of 62 for instruction: %r = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> %v)
 ; CODE-LABEL: umax.i32.v4i32
 ; CODE:       umaxv s0, v0.4s
 define i32 @umax.i32.v4i32(<4 x i32> %v) {
   %r = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> %v)
   ret i32 %r
 }
 
 ; COST-LABEL: smin.i8.v8i8
-; COST:       Found an estimated cost of 157 for instruction: %r = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> %v)
+; COST:       Found an estimated cost of 280 for instruction: %r = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> %v)
 ; CODE-LABEL: smin.i8.v8i8
 ; CODE:       sminv b0, v0.8b
 define i8 @smin.i8.v8i8(<8 x i8> %v) {
   %r = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> %v)
   ret i8 %r
 }
 
 ; COST-LABEL: smin.i8.v16i8
-; COST:       Found an estimated cost of 388 for instruction: %r = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> %v)
+; COST:       Found an estimated cost of 744 for instruction: %r = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> %v)
 ; CODE-LABEL: smin.i8.v16i8
 ; CODE:       sminv b0, v0.16b
 define i8 @smin.i8.v16i8(<16 x i8> %v) {
   %r = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> %v)
   ret i8 %r
 }
 
 ; COST-LABEL: smin.i16.v4i16
-; COST:       Found an estimated cost of 58 for instruction: %r = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> %v)
+; COST:       Found an estimated cost of 92 for instruction: %r = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> %v)
 ; CODE-LABEL: smin.i16.v4i16
 ; CODE:       sminv h0, v0.4h
 define i16 @smin.i16.v4i16(<4 x i16> %v) {
   %r = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> %v)
   ret i16 %r
 }
 
 ; COST-LABEL: smin.i16.v8i16
-; COST:       Found an estimated cost of 157 for instruction: %r = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> %v)
+; COST:       Found an estimated cost of 280 for instruction: %r = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> %v)
 ; CODE-LABEL: smin.i16.v8i16
 ; CODE:       sminv h0, v0.8h
 define i16 @smin.i16.v8i16(<8 x i16> %v) {
   %r = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> %v)
   ret i16 %r
 }
 
 ; COST-LABEL: smin.i32.v4i32
-; COST:       Found an estimated cost of 58 for instruction: %r = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> %v)
+; COST:       Found an estimated cost of 62 for instruction: %r = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> %v)
 ; CODE-LABEL: smin.i32.v4i32
 ; CODE:       sminv s0, v0.4s
 define i32 @smin.i32.v4i32(<4 x i32> %v) {
   %r = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> %v)
   ret i32 %r
 }
 
 ; COST-LABEL: smax.i8.v8i8
-; COST:       Found an estimated cost of 157 for instruction: %r = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> %v)
+; COST:       Found an estimated cost of 280 for instruction: %r = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> %v)
 ; CODE-LABEL: smax.i8.v8i8
 ; CODE:       smaxv b0, v0.8b
 define i8 @smax.i8.v8i8(<8 x i8> %v) {
   %r = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> %v)
   ret i8 %r
 }
 
 ; COST-LABEL: smax.i8.v16i8
-; COST:       Found an estimated cost of 388 for instruction: %r = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> %v)
+; COST:       Found an estimated cost of 744 for instruction: %r = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> %v)
 ; CODE-LABEL: smax.i8.v16i8
 ; CODE:       smaxv b0, v0.16b
 define i8 @smax.i8.v16i8(<16 x i8> %v) {
   %r = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> %v)
   ret i8 %r
 }
 
 ; COST-LABEL: smax.i16.v4i16
-; COST:       Found an estimated cost of 58 for instruction: %r = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> %v)
+; COST:       Found an estimated cost of 92 for instruction: %r = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> %v)
 ; CODE-LABEL: smax.i16.v4i16
 ; CODE:       smaxv h0, v0.4h
 define i16 @smax.i16.v4i16(<4 x i16> %v) {
   %r = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> %v)
   ret i16 %r
 }
 
 ; COST-LABEL: smax.i16.v8i16
-; COST:       Found an estimated cost of 157 for instruction: %r = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> %v)
+; COST:       Found an estimated cost of 280 for instruction: %r = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> %v)
 ; CODE-LABEL: smax.i16.v8i16
 ; CODE:       smaxv h0, v0.8h
 define i16 @smax.i16.v8i16(<8 x i16> %v) {
   %r = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> %v)
   ret i16 %r
 }
 
 ; COST-LABEL: smax.i32.v4i32
-; COST:       Found an estimated cost of 58 for instruction: %r = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> %v)
+; COST:       Found an estimated cost of 62 for instruction: %r = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> %v)
 ; CODE-LABEL: smax.i32.v4i32
 ; CODE:       smaxv s0, v0.4s
 define i32 @smax.i32.v4i32(<4 x i32> %v) {
   %r = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> %v)
   ret i32 %r
 }
 
 ; COST-LABEL: fmin.f32.v4f32
-; COST:       Found an estimated cost of 58 for instruction: %r = call nnan float @llvm.experimental.vector.reduce.fmin.f32.v4f32(<4 x float> %v)
+; COST:       Found an estimated cost of 62 for instruction: %r = call nnan float @llvm.experimental.vector.reduce.fmin.f32.v4f32(<4 x float> %v)
 ; CODE-LABEL: fmin.f32.v4f32
 ; CODE:       fminnmv s0, v0.4s
 define float @fmin.f32.v4f32(<4 x float> %v) {
   %r = call nnan float @llvm.experimental.vector.reduce.fmin.f32.v4f32(<4 x float> %v)
   ret float %r
 }
 
 ; COST-LABEL: fmax.f32.v4f32
-; COST:       Found an estimated cost of 58 for instruction: %r = call nnan float @llvm.experimental.vector.reduce.fmax.f32.v4f32(<4 x float> %v)
+; COST:       Found an estimated cost of 62 for instruction: %r = call nnan float @llvm.experimental.vector.reduce.fmax.f32.v4f32(<4 x float> %v)
 ; CODE-LABEL: fmax.f32.v4f32
 ; CODE:       fmaxnmv s0, v0.4s
 define float @fmax.f32.v4f32(<4 x float> %v) {
   %r = call nnan float @llvm.experimental.vector.reduce.fmax.f32.v4f32(<4 x float> %v)
   ret float %r
 }
 
 declare i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8>)

test/Analysis/CostModel/X86/reduce-add.ll

;
   %V16 = call i32 @llvm.experimental.vector.reduce.add.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.add.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V4  = call i16 @llvm.experimental.vector.reduce.add.i16.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.experimental.vector.reduce.add.i16.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.experimental.vector.reduce.add.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.add.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.add.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 101 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 79 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 121 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V8   = call i8 @llvm.experimental.vector.reduce.add.i8.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.experimental.vector.reduce.add.i8.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.experimental.vector.reduce.add.i8.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.experimental.vector.reduce.add.i8.v64i8(<64 x i8> undef)
   %V128 = call i8 @llvm.experimental.vector.reduce.add.i8.v128i8(<128 x i8> undef)
   ret i32 undef

test/Analysis/CostModel/X86/reduce-and.ll

 ; SSE-LABEL: 'reduce_i64'
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.and.i64.v1i64(<1 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.and.i64.v2i64(<2 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.and.i64.v4i64(<4 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.and.i64.v8i64(<8 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.and.i64.v16i64(<16 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
-; AVX-LABEL: 'reduce_i64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.and.i64.v1i64(<1 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.and.i64.v2i64(<2 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.and.i64.v4i64(<4 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.and.i64.v8i64(<8 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.and.i64.v16i64(<16 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; AVX1-LABEL: 'reduce_i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.and.i64.v1i64(<1 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.and.i64.v2i64(<2 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.and.i64.v4i64(<4 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.and.i64.v8i64(<8 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.and.i64.v16i64(<16 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.and.i64.v1i64(<1 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.and.i64.v2i64(<2 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.and.i64.v4i64(<4 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.and.i64.v8i64(<8 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.and.i64.v16i64(<16 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'reduce_i64'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.and.i64.v1i64(<1 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.and.i64.v2i64(<2 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.and.i64.v4i64(<4 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.and.i64.v8i64(<8 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.and.i64.v16i64(<16 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef

 ; SSE-LABEL: 'reduce_i32'
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.and.i32.v2i32(<2 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.and.i32.v4i32(<4 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.and.i32.v8i32(<8 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.and.i32.v16i32(<16 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.and.i32.v32i32(<32 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
-; AVX-LABEL: 'reduce_i32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.and.i32.v2i32(<2 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.and.i32.v4i32(<4 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.and.i32.v8i32(<8 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.and.i32.v16i32(<16 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.and.i32.v32i32(<32 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; AVX1-LABEL: 'reduce_i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.and.i32.v2i32(<2 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.and.i32.v4i32(<4 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.and.i32.v8i32(<8 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.and.i32.v16i32(<16 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.and.i32.v32i32(<32 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.and.i32.v2i32(<2 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.and.i32.v4i32(<4 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.and.i32.v8i32(<8 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.and.i32.v16i32(<16 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.and.i32.v32i32(<32 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'reduce_i32'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.and.i32.v2i32(<2 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.and.i32.v4i32(<4 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.and.i32.v8i32(<8 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.and.i32.v16i32(<16 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.and.i32.v32i32(<32 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i32 @llvm.experimental.vector.reduce.and.i32.v2i32(<2 x i32> undef)
   %V4  = call i32 @llvm.experimental.vector.reduce.and.i32.v4i32(<4 x i32> undef)
   %V8  = call i32 @llvm.experimental.vector.reduce.and.i32.v8i32(<8 x i32> undef)
   %V16 = call i32 @llvm.experimental.vector.reduce.and.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.and.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
-; SSE-LABEL: 'reduce_i16'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i16'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i16'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i16'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V4  = call i16 @llvm.experimental.vector.reduce.and.i16.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.experimental.vector.reduce.and.i16.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.experimental.vector.reduce.and.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.and.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.and.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
-; SSE-LABEL: 'reduce_i8'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i8'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i8'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i8'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 79 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 121 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V8   = call i8 @llvm.experimental.vector.reduce.and.i8.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.experimental.vector.reduce.and.i8.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.experimental.vector.reduce.and.i8.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.experimental.vector.reduce.and.i8.v64i8(<64 x i8> undef)
   %V128 = call i8 @llvm.experimental.vector.reduce.and.i8.v128i8(<128 x i8> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i1(i32 %arg) {
-; SSE-LABEL: 'reduce_i1'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i1'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i1'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i1'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i1'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i1'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i1'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i1'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 151 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 165 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i1'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 79 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 357 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 838 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 841 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i1'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 151 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.and.i1.v64i1(<64 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 165 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.and.i1.v128i1(<128 x i1> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V1   = call i1 @llvm.experimental.vector.reduce.and.i1.v1i1(<1 x i1> undef)
   %V2   = call i1 @llvm.experimental.vector.reduce.and.i1.v2i1(<2 x i1> undef)
   %V4   = call i1 @llvm.experimental.vector.reduce.and.i1.v4i1(<4 x i1> undef)
   %V8   = call i1 @llvm.experimental.vector.reduce.and.i1.v8i1(<8 x i1> undef)
   %V16  = call i1 @llvm.experimental.vector.reduce.and.i1.v16i1(<16 x i1> undef)
   %V32  = call i1 @llvm.experimental.vector.reduce.and.i1.v32i1(<32 x i1> undef)

test/Analysis/CostModel/X86/reduce-mul.ll

 ; SSE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.mul.i64.v4i64(<4 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.mul.i64.v8i64(<8 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.mul.i64.v16i64(<16 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.mul.i64.v1i64(<1 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.mul.i64.v2i64(<2 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.mul.i64.v4i64(<4 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 79 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.mul.i64.v8i64(<8 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 152 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.mul.i64.v16i64(<16 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.mul.i64.v4i64(<4 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.mul.i64.v8i64(<8 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 154 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.mul.i64.v16i64(<16 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.mul.i64.v1i64(<1 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.mul.i64.v2i64(<2 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.mul.i64.v4i64(<4 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.mul.i64.v8i64(<8 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 72 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.mul.i64.v16i64(<16 x i64> undef)

 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.mul.i32.v8i32(<8 x i32> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.mul.i32.v16i32(<16 x i32> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.mul.i32.v32i32(<32 x i32> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i32'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.mul.i32.v2i32(<2 x i32> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.mul.i32.v4i32(<4 x i32> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.mul.i32.v8i32(<8 x i32> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.mul.i32.v16i32(<16 x i32> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.mul.i32.v32i32(<32 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.mul.i32.v8i32(<8 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.mul.i32.v16i32(<16 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.mul.i32.v32i32(<32 x i32> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i32'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.mul.i32.v2i32(<2 x i32> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.mul.i32.v4i32(<4 x i32> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.mul.i32.v8i32(<8 x i32> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.mul.i32.v16i32(<16 x i32> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.mul.i32.v32i32(<32 x i32> undef)
;
   %V16 = call i32 @llvm.experimental.vector.reduce.mul.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.mul.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V4  = call i16 @llvm.experimental.vector.reduce.mul.i16.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.experimental.vector.reduce.mul.i16.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.experimental.vector.reduce.mul.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.mul.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.mul.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
-; SSE-LABEL: 'reduce_i8'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 93 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 239 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 129 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 178 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 275 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i8'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 93 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 239 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i8'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 93 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 239 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 167 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 220 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 325 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 202 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 255 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 360 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 157 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 226 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 172 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 241 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 102 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 129 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 182 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 117 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 144 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 197 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 136 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 159 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 67 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 178 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 201 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 102 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 129 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 182 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 117 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 144 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 197 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V8   = call i8 @llvm.experimental.vector.reduce.mul.i8.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.experimental.vector.reduce.mul.i8.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.experimental.vector.reduce.mul.i8.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.experimental.vector.reduce.mul.i8.v64i8(<64 x i8> undef)
   %V128 = call i8 @llvm.experimental.vector.reduce.mul.i8.v128i8(<128 x i8> undef)
   ret i32 undef

test/Analysis/CostModel/X86/reduce-or.ll

 ; SSE-LABEL: 'reduce_i64'
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.or.i64.v1i64(<1 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.or.i64.v2i64(<2 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.or.i64.v4i64(<4 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.or.i64.v8i64(<8 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.or.i64.v16i64(<16 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
-; AVX-LABEL: 'reduce_i64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.or.i64.v1i64(<1 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.or.i64.v2i64(<2 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.or.i64.v4i64(<4 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.or.i64.v8i64(<8 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.or.i64.v16i64(<16 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; AVX1-LABEL: 'reduce_i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.or.i64.v1i64(<1 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.or.i64.v2i64(<2 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.or.i64.v4i64(<4 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.or.i64.v8i64(<8 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.or.i64.v16i64(<16 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.or.i64.v1i64(<1 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.or.i64.v2i64(<2 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.or.i64.v4i64(<4 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.or.i64.v8i64(<8 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.or.i64.v16i64(<16 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'reduce_i64'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.or.i64.v1i64(<1 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.or.i64.v2i64(<2 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.or.i64.v4i64(<4 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.or.i64.v8i64(<8 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.or.i64.v16i64(<16 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef

 ; SSE-LABEL: 'reduce_i32'
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.or.i32.v2i32(<2 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.or.i32.v4i32(<4 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.or.i32.v8i32(<8 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.or.i32.v16i32(<16 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.or.i32.v32i32(<32 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
-; AVX-LABEL: 'reduce_i32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.or.i32.v2i32(<2 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.or.i32.v4i32(<4 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.or.i32.v8i32(<8 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.or.i32.v16i32(<16 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.or.i32.v32i32(<32 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; AVX1-LABEL: 'reduce_i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.or.i32.v2i32(<2 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.or.i32.v4i32(<4 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.or.i32.v8i32(<8 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.or.i32.v16i32(<16 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.or.i32.v32i32(<32 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.or.i32.v2i32(<2 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.or.i32.v4i32(<4 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.or.i32.v8i32(<8 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.or.i32.v16i32(<16 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.or.i32.v32i32(<32 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'reduce_i32'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.or.i32.v2i32(<2 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.or.i32.v4i32(<4 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.or.i32.v8i32(<8 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.or.i32.v16i32(<16 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.or.i32.v32i32(<32 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i32 @llvm.experimental.vector.reduce.or.i32.v2i32(<2 x i32> undef)
   %V4  = call i32 @llvm.experimental.vector.reduce.or.i32.v4i32(<4 x i32> undef)
   %V8  = call i32 @llvm.experimental.vector.reduce.or.i32.v8i32(<8 x i32> undef)
   %V16 = call i32 @llvm.experimental.vector.reduce.or.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.or.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
-; SSE-LABEL: 'reduce_i16'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i16'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i16'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i16'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V4  = call i16 @llvm.experimental.vector.reduce.or.i16.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.experimental.vector.reduce.or.i16.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.experimental.vector.reduce.or.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.or.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.or.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
-; SSE-LABEL: 'reduce_i8'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i8'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i8'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i8'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 79 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 121 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V8   = call i8 @llvm.experimental.vector.reduce.or.i8.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.experimental.vector.reduce.or.i8.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.experimental.vector.reduce.or.i8.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.experimental.vector.reduce.or.i8.v64i8(<64 x i8> undef)
   %V128 = call i8 @llvm.experimental.vector.reduce.or.i8.v128i8(<128 x i8> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i1(i32 %arg) {
-; SSE-LABEL: 'reduce_i1'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i1'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i1'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i1'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i1'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i1'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i1'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i1'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 151 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 165 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i1'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 79 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 357 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 838 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 841 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i1'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 151 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.or.i1.v64i1(<64 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 165 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.or.i1.v128i1(<128 x i1> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V1   = call i1 @llvm.experimental.vector.reduce.or.i1.v1i1(<1 x i1> undef)
   %V2   = call i1 @llvm.experimental.vector.reduce.or.i1.v2i1(<2 x i1> undef)
   %V4   = call i1 @llvm.experimental.vector.reduce.or.i1.v4i1(<4 x i1> undef)
   %V8   = call i1 @llvm.experimental.vector.reduce.or.i1.v8i1(<8 x i1> undef)
   %V16  = call i1 @llvm.experimental.vector.reduce.or.i1.v16i1(<16 x i1> undef)
   %V32  = call i1 @llvm.experimental.vector.reduce.or.i1.v32i1(<32 x i1> undef)

test/Analysis/CostModel/X86/reduce-smax.ll

 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 72 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smax.i64.v1i64(<1 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smax.i64.v1i64(<1 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
;
   %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 105 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 101 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 133 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
;
   %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 102 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 111 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 138 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 159 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 162 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 173 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 194 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 211 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 216 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 253 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 258 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)

test/Analysis/CostModel/X86/reduce-smin.ll

 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 72 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smin.i64.v1i64(<1 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smin.i64.v1i64(<1 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
;
   %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 105 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 101 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 133 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
;
   %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 102 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 111 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 138 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 159 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 162 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 173 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 194 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 211 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 216 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 253 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 258 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)

test/Analysis/CostModel/X86/reduce-umax.ll

 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umax.i64.v1i64(<1 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umax.i64.v1i64(<1 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
;
   %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 105 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 101 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 133 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
;
   %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 102 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 111 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 138 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 159 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 162 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 173 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 194 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 211 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 216 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 253 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 258 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)

test/Analysis/CostModel/X86/reduce-umin.ll

 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umin.i64.v1i64(<1 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umin.i64.v1i64(<1 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
;
   %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 105 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 101 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 133 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
;
   %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 102 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 111 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'reduce_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 138 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 159 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 162 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 173 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 194 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 211 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 216 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 253 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 258 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)

test/Analysis/CostModel/X86/reduce-xor.ll

 ; SSE-LABEL: 'reduce_i64'
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.xor.i64.v1i64(<1 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.xor.i64.v2i64(<2 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.xor.i64.v4i64(<4 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.xor.i64.v8i64(<8 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.xor.i64.v16i64(<16 x i64> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
-; AVX-LABEL: 'reduce_i64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.xor.i64.v1i64(<1 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.xor.i64.v2i64(<2 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.xor.i64.v4i64(<4 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.xor.i64.v8i64(<8 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.xor.i64.v16i64(<16 x i64> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; AVX1-LABEL: 'reduce_i64'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.xor.i64.v1i64(<1 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.xor.i64.v2i64(<2 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.xor.i64.v4i64(<4 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.xor.i64.v8i64(<8 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.xor.i64.v16i64(<16 x i64> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i64'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.xor.i64.v1i64(<1 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.xor.i64.v2i64(<2 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.xor.i64.v4i64(<4 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.xor.i64.v8i64(<8 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.xor.i64.v16i64(<16 x i64> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'reduce_i64'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.xor.i64.v1i64(<1 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.xor.i64.v2i64(<2 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.xor.i64.v4i64(<4 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.xor.i64.v8i64(<8 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.xor.i64.v16i64(<16 x i64> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef

 ; SSE-LABEL: 'reduce_i32'
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.xor.i32.v2i32(<2 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.xor.i32.v4i32(<4 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.xor.i32.v8i32(<8 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.xor.i32.v16i32(<16 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.xor.i32.v32i32(<32 x i32> undef)
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
-; AVX-LABEL: 'reduce_i32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.xor.i32.v2i32(<2 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.xor.i32.v4i32(<4 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.xor.i32.v8i32(<8 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.xor.i32.v16i32(<16 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.xor.i32.v32i32(<32 x i32> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; AVX1-LABEL: 'reduce_i32'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.xor.i32.v2i32(<2 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.xor.i32.v4i32(<4 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.xor.i32.v8i32(<8 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.xor.i32.v16i32(<16 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.xor.i32.v32i32(<32 x i32> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i32'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.xor.i32.v2i32(<2 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.xor.i32.v4i32(<4 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.xor.i32.v8i32(<8 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.xor.i32.v16i32(<16 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.xor.i32.v32i32(<32 x i32> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'reduce_i32'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.xor.i32.v2i32(<2 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.xor.i32.v4i32(<4 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.xor.i32.v8i32(<8 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.xor.i32.v16i32(<16 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.xor.i32.v32i32(<32 x i32> undef)
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i32 @llvm.experimental.vector.reduce.xor.i32.v2i32(<2 x i32> undef)
   %V4  = call i32 @llvm.experimental.vector.reduce.xor.i32.v4i32(<4 x i32> undef)
   %V8  = call i32 @llvm.experimental.vector.reduce.xor.i32.v8i32(<8 x i32> undef)
   %V16 = call i32 @llvm.experimental.vector.reduce.xor.i32.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.experimental.vector.reduce.xor.i32.v32i32(<32 x i32> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i16(i32 %arg) {
-; SSE-LABEL: 'reduce_i16'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i16'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i16'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i16'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V4  = call i16 @llvm.experimental.vector.reduce.xor.i16.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.experimental.vector.reduce.xor.i16.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.experimental.vector.reduce.xor.i16.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.experimental.vector.reduce.xor.i16.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.experimental.vector.reduce.xor.i16.v64i16(<64 x i16> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i8(i32 %arg) {
-; SSE-LABEL: 'reduce_i8'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i8'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i8'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i8'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 79 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 121 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V8   = call i8 @llvm.experimental.vector.reduce.xor.i8.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.experimental.vector.reduce.xor.i8.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.experimental.vector.reduce.xor.i8.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.experimental.vector.reduce.xor.i8.v64i8(<64 x i8> undef)
   %V128 = call i8 @llvm.experimental.vector.reduce.xor.i8.v128i8(<128 x i8> undef)
   ret i32 undef
 }
 
 define i32 @reduce_i1(i32 %arg) {
-; SSE-LABEL: 'reduce_i1'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
-;
-; AVX-LABEL: 'reduce_i1'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'reduce_i1'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'reduce_i1'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'reduce_i1'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'reduce_i1'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 77 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'reduce_i1'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'reduce_i1'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 151 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 165 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'reduce_i1'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 79 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 357 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 838 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 841 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512DQ-LABEL: 'reduce_i1'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1 = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2 = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4 = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V8 = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16 = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 151 for instruction: %V32 = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V64 = call i1 @llvm.experimental.vector.reduce.xor.i1.v64i1(<64 x i1> undef)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 165 for instruction: %V128 = call i1 @llvm.experimental.vector.reduce.xor.i1.v128i1(<128 x i1> undef)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V1   = call i1 @llvm.experimental.vector.reduce.xor.i1.v1i1(<1 x i1> undef)
   %V2   = call i1 @llvm.experimental.vector.reduce.xor.i1.v2i1(<2 x i1> undef)
   %V4   = call i1 @llvm.experimental.vector.reduce.xor.i1.v4i1(<4 x i1> undef)
   %V8   = call i1 @llvm.experimental.vector.reduce.xor.i1.v8i1(<8 x i1> undef)
   %V16  = call i1 @llvm.experimental.vector.reduce.xor.i1.v16i1(<16 x i1> undef)
   %V32  = call i1 @llvm.experimental.vector.reduce.xor.i1.v32i1(<32 x i1> undef)

test/Analysis/CostModel/X86/reduction.ll

 define fastcc i16 @no_pairwise_reduction8i16(<8 x i16> %rdx, i16 %f1) {
 ; SSE2-LABEL: 'no_pairwise_reduction8i16'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %rdx.shuf3 = shufflevector <8 x i16> %rdx, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx4 = add <8 x i16> %rdx, %rdx.shuf3
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %rdx.shuf = shufflevector <8 x i16> %bin.rdx4, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx = add <8 x i16> %bin.rdx4, %rdx.shuf
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %rdx.shuf7 = shufflevector <8 x i16> %bin.rdx, <8 x i16> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx8 = add <8 x i16> %bin.rdx, %rdx.shuf7
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %r = extractelement <8 x i16> %bin.rdx8, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %r = extractelement <8 x i16> %bin.rdx8, i32 0
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %r
 ;
 ; SSSE3-LABEL: 'no_pairwise_reduction8i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %rdx.shuf3 = shufflevector <8 x i16> %rdx, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx4 = add <8 x i16> %rdx, %rdx.shuf3
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %rdx.shuf = shufflevector <8 x i16> %bin.rdx4, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx = add <8 x i16> %bin.rdx4, %rdx.shuf
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %rdx.shuf7 = shufflevector <8 x i16> %bin.rdx, <8 x i16> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>

 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %rdx.shuf.0.1 = shufflevector <8 x i16> %rdx, <8 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx = add <8 x i16> %rdx.shuf.0.0, %rdx.shuf.0.1
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %rdx.shuf.1.0 = shufflevector <8 x i16> %bin.rdx, <8 x i16> undef, <8 x i32> <i32 0, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %rdx.shuf.1.1 = shufflevector <8 x i16> %bin.rdx, <8 x i16> undef, <8 x i32> <i32 1, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx8 = add <8 x i16> %rdx.shuf.1.0, %rdx.shuf.1.1
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %rdx.shuf.2.0 = shufflevector <8 x i16> %bin.rdx8, <8 x i16> undef, <8 x i32> <i32 0, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %rdx.shuf.2.1 = shufflevector <8 x i16> %bin.rdx8, <8 x i16> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx9 = add <8 x i16> %rdx.shuf.2.0, %rdx.shuf.2.1
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %r = extractelement <8 x i16> %bin.rdx9, i32 0
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %r = extractelement <8 x i16> %bin.rdx9, i32 0
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %r
 ;
 ; SSSE3-LABEL: 'pairwise_reduction8i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %rdx.shuf.0.0 = shufflevector <8 x i16> %rdx, <8 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %rdx.shuf.0.1 = shufflevector <8 x i16> %rdx, <8 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bin.rdx = add <8 x i16> %rdx.shuf.0.0, %rdx.shuf.0.1
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %rdx.shuf.1.0 = shufflevector <8 x i16> %bin.rdx, <8 x i16> undef, <8 x i32> <i32 0, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %rdx.shuf.1.1 = shufflevector <8 x i16> %bin.rdx, <8 x i16> undef, <8 x i32> <i32 1, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>