[SLP][COST][X86]Improve cost model for masked gather.

[ABataev]

Summary

Revived D101297 in its original form + added some changes in X86
legalization cehcking for masked gathers.

This solution is the most stable and the most correct one. We have to
check the legality before trying to build the masked gather in SLP.
Without this check we have incorrect cost (for SLP) in case if the masked gather
is not legal/slower than the gather. And we're missing some
vectorization opportunities.

This can be fixed in the cost model, but in this case we need to add
special checks for the cost of GEPs for ScatterVectorize node, add
special check for small trees, etc., i.e. there are a lot of corner
cases here and there, which insrease code base and make it harder to
maintain the code.

Can't we rely on cost model to deal with this? This can be profitable for futher vectorization, when we can start from such gather loads as seed.

The question from D101297. Actually, no, it can't. Actually, simple
gather may give us better result, especially after we started
vectorization of insertelements. Plus, like I said before, the cost for
non-legal masked gathers leads to missed vectorization opportunities.

[ABataev]

Rebase

[anton-afanasyev]

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

3015

Maybe it's worth to use CommonAlignment for getEntryCost() as well? This patch looks good one to take this code from D57059.

[RKSimon]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

4780

I'm really not sure about this - by moving this into isLegalMaskedGather we're affecting codegen as well as IR.

While we agreed that we'd avoid this on AVX2 targets (where legal gathers are actually REALLY bad), more minor cost differences on AVX512 targets its not so clear.

[ABataev]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

4780

I think this change is ok. Actually, we saw perf degradations with masked gathers even for AVX512 and had to change to use only for at least 8 elements, even for 4 elements simple gathering might be more profitable than the masked gathering. I would agree, that this requires some extra fine-tuning but I believe this kind of check should be in legality rather than in the cost. Without this change, we return the cost of loads+gathering, which is not the cost of the MaskedGather op. And I think this is good for codegen too since it relies on the legality check when chooses between masked gathering and simple gathering.

[ABataev]

Address comment

[RKSimon]

LGTM - if you improve the VLX codegen test coverage this should be OK, cheers

llvm/test/CodeGen/X86/masked_gather.ll

5

Please can you add AVX512VL coverage:

; RUN: llc -mtriple=x86_64-unknown-linux-gnu  -mattr=+avx512f < %s | FileCheck %s --check-prefixes=AVX512,AVX512F
; RUN: llc -mtriple=x86_64-unknown-linux-gnu  -mattr=+avx512f,+avx512vl < %s | FileCheck %s --check-prefixes=AVX512,AVX512VL

[ABataev]

In D105042#2864723, @RKSimon wrote:

LGTM - if you improve the VLX codegen test coverage this should be OK, cheers

Sure, will add.

[vtjnash]

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

4090

Was this supposed to be CommonAlignment here? It seems like that added computation is currently unused

[ABataev]

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

4090

Yes, missed it, will fix, thanks!

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

if ((Opcode == Instruction::Load &&
         (Opcode == Instruction::Store &&
          isLegalMaskedScatter(SrcVTy, Align(Alignment))))
       return 1;
     return BaseT::getGatherScatterOpCost(Opcode, SrcVTy, Ptr, VariableMask,
                                          Alignment, CostKind, I);
   }
 
   assert(SrcVTy->isVectorTy() && "Unexpected data type for Gather/Scatter");
-  unsigned VF = cast<FixedVectorType>(SrcVTy)->getNumElements();
   PointerType *PtrTy = dyn_cast<PointerType>(Ptr->getType());
   if (!PtrTy && Ptr->getType()->isVectorTy())
     PtrTy = dyn_cast<PointerType>(
         cast<VectorType>(Ptr->getType())->getElementType());
   assert(PtrTy && "Unexpected type for Ptr argument");
   unsigned AddressSpace = PtrTy->getAddressSpace();
 
-  bool Scalarize = false;
   if ((Opcode == Instruction::Load &&
        !isLegalMaskedGather(SrcVTy, Align(Alignment))) ||
       (Opcode == Instruction::Store &&
        !isLegalMaskedScatter(SrcVTy, Align(Alignment))))
-    Scalarize = true;
-  // Gather / Scatter for vector 2 is not profitable on KNL / SKX
-  // Vector-4 of gather/scatter instruction does not exist on KNL.
-  // We can extend it to 8 elements, but zeroing upper bits of
-  // the mask vector will add more instructions. Right now we give the scalar
-  // cost of vector-4 for KNL. TODO: Check, maybe the gather/scatter instruction
-  // is better in the VariableMask case.
-  if (ST->hasAVX512() && (VF == 2 || (VF == 4 && !ST->hasVLX())))
-    Scalarize = true;
-
-  if (Scalarize)
     return getGSScalarCost(Opcode, SrcVTy, VariableMask, Alignment,
                            AddressSpace);
 
   return getGSVectorCost(Opcode, SrcVTy, Ptr, Alignment, AddressSpace);
 }
 
 bool X86TTIImpl::isLSRCostLess(TargetTransformInfo::LSRCost &C1,
                                TargetTransformInfo::LSRCost &C2) {
bool X86TTIImpl::isLegalMaskedGather(Type *DataTy, Align Alignment) {
   // The Scalarizer asks again about legality. It sends a vector type.
   // In this case we can reject non-power-of-2 vectors.
   // We also reject single element vectors as the type legalizer can't
   // scalarize it.
   if (auto *DataVTy = dyn_cast<FixedVectorType>(DataTy)) {
     unsigned NumElts = DataVTy->getNumElements();
     if (NumElts == 1)
       return false;
+    // Gather / Scatter for vector 2 is not profitable on KNL / SKX
+    // Vector-4 of gather/scatter instruction does not exist on KNL.
+    // We can extend it to 8 elements, but zeroing upper bits of
+    // the mask vector will add more instructions. Right now we give the scalar
+    // cost of vector-4 for KNL. TODO: Check, maybe the gather/scatter
+    // instruction is better in the VariableMask case.
+    if (ST->hasAVX512() && (NumElts == 2 || (NumElts == 4 && !ST->hasVLX())))
+      return false;
   }
   Type *ScalarTy = DataTy->getScalarType();
   if (ScalarTy->isPointerTy())
     return true;
 
   if (ScalarTy->isFloatTy() || ScalarTy->isDoubleTy())
     return true;
 

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

case Instruction::Load: {
                              ReuseShuffleIndicies, CurrentOrder);
             TE->setOperandsInOrder();
             LLVM_DEBUG(dbgs() << "SLP: added a vector of jumbled loads.\n");
             findRootOrder(CurrentOrder);
             ++NumOpsWantToKeepOrder[CurrentOrder];
           }
           return;
         }
+        Align CommonAlignment = cast<LoadInst>(VL0)->getAlign();
+        for (Value *V : VL)
+          CommonAlignment =
+              commonAlignment(CommonAlignment, cast<LoadInst>(V)->getAlign());
+        if (TTI->isLegalMaskedGather(FixedVectorType::get(ScalarTy, VL.size()),
+                                     CommonAlignment)) {
           // Vectorizing non-consecutive loads with `llvm.masked.gather`.
-        TreeEntry *TE = newTreeEntry(VL, TreeEntry::ScatterVectorize, Bundle, S,
-                                     UserTreeIdx, ReuseShuffleIndicies);
+          TreeEntry *TE = newTreeEntry(VL, TreeEntry::ScatterVectorize, Bundle,
+                                       S, UserTreeIdx, ReuseShuffleIndicies);
           TE->setOperandsInOrder();
           buildTree_rec(PointerOps, Depth + 1, {TE, 0});
-        LLVM_DEBUG(dbgs() << "SLP: added a vector of non-consecutive loads.\n");
+          LLVM_DEBUG(dbgs()
+                     << "SLP: added a vector of non-consecutive loads.\n");
           return;
         }
+      }
 
       LLVM_DEBUG(dbgs() << "SLP: Gathering non-consecutive loads.\n");
       BS.cancelScheduling(VL, VL0);
       newTreeEntry(VL, None /*not vectorized*/, S, UserTreeIdx,
                    ReuseShuffleIndicies);
       return;
     }
     case Instruction::ZExt:
case Instruction::Load: {
       InstructionCost VecLdCost;
       if (E->State == TreeEntry::Vectorize) {
         VecLdCost = TTI->getMemoryOpCost(Instruction::Load, VecTy, alignment, 0,
                                          CostKind, VL0);
       } else {
         assert(E->State == TreeEntry::ScatterVectorize && "Unknown EntryState");
         VecLdCost = TTI->getGatherScatterOpCost(
             Instruction::Load, VecTy, cast<LoadInst>(VL0)->getPointerOperand(),
             /*VariableMask=*/false, alignment, CostKind, VL0);
       }
       if (!NeedToShuffleReuses && !E->ReorderIndices.empty()) {
         SmallVector<int> NewMask;
         inversePermutation(E->ReorderIndices, NewMask);
         VecLdCost += TTI->getShuffleCost(
             TargetTransformInfo::SK_PermuteSingleSrc, VecTy, NewMask);
       }
       LLVM_DEBUG(dumpTreeCosts(E, ReuseShuffleCost, VecLdCost, ScalarLdCost));

llvm/test/CodeGen/X86/masked_gather.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=x86_64-unknown-linux-gnu  -mattr=+sse4.2 < %s | FileCheck %s --check-prefix=SSE
 ; RUN: llc -mtriple=x86_64-unknown-linux-gnu  -mattr=+avx < %s | FileCheck %s --check-prefix=AVX1
 ; RUN: llc -mtriple=x86_64-unknown-linux-gnu  -mattr=+avx2 < %s | FileCheck %s --check-prefix=AVX2
 ; RUN: llc -mtriple=x86_64-unknown-linux-gnu  -mattr=+avx512f < %s | FileCheck %s --check-prefix=AVX512
 
 ;
 ; vXf32
 ;
 
 %struct.a = type { [4 x i32], [4 x i8], %struct.b, i32 }
 %struct.b = type { i32, i32 }
 @c = external dso_local global %struct.a, align 4

 ; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
 ; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
 ; AVX2-NEXT:    vmovaps %xmm2, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: gather_v4f32_ptr_v4i32:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
 ; AVX512-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
 ; AVX512-NEXT:    vptestnmd %zmm1, %zmm1, %k0
-; AVX512-NEXT:    kshiftlw $12, %k0, %k0
-; AVX512-NEXT:    kshiftrw $12, %k0, %k1
-; AVX512-NEXT:    vgatherqps (,%zmm0), %ymm2 {%k1}
+; AVX512-NEXT:    kmovw %k0, %eax
+; AVX512-NEXT:    testb $1, %al
+; AVX512-NEXT:    je .LBB0_2
+; AVX512-NEXT:  # %bb.1: # %cond.load
+; AVX512-NEXT:    vmovq %xmm0, %rcx
+; AVX512-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX512-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; AVX512-NEXT:  .LBB0_2: # %else
+; AVX512-NEXT:    testb $2, %al
+; AVX512-NEXT:    je .LBB0_4
+; AVX512-NEXT:  # %bb.3: # %cond.load1
+; AVX512-NEXT:    vpextrq $1, %xmm0, %rcx
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
+; AVX512-NEXT:  .LBB0_4: # %else2
+; AVX512-NEXT:    testb $4, %al
+; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX512-NEXT:    jne .LBB0_5
+; AVX512-NEXT:  # %bb.6: # %else5
+; AVX512-NEXT:    testb $8, %al
+; AVX512-NEXT:    jne .LBB0_7
+; AVX512-NEXT:  .LBB0_8: # %else8
+; AVX512-NEXT:    vmovaps %xmm2, %xmm0
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
+; AVX512-NEXT:  .LBB0_5: # %cond.load4
+; AVX512-NEXT:    vmovq %xmm0, %rcx
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],mem[0],xmm2[3]
+; AVX512-NEXT:    testb $8, %al
+; AVX512-NEXT:    je .LBB0_8
+; AVX512-NEXT:  .LBB0_7: # %cond.load7
+; AVX512-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
 ; AVX512-NEXT:    vmovaps %xmm2, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %mask = icmp eq <4 x i32> %trigger, zeroinitializer
   %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %ptr, i32 4, <4 x i1> %mask, <4 x float> %passthru)
   ret <4 x float> %res
 }
 

 ; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
 ; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
 ; AVX2-NEXT:    vmovaps %xmm2, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: gather_v4f32_v4i32_v4i32:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
 ; AVX512-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512-NEXT:    vmovq %rdi, %xmm3
+; AVX512-NEXT:    vpbroadcastq %xmm3, %ymm3
+; AVX512-NEXT:    vpmovsxdq %xmm0, %ymm0
+; AVX512-NEXT:    vpsllq $2, %ymm0, %ymm0
+; AVX512-NEXT:    vpaddq %ymm0, %ymm3, %ymm0
 ; AVX512-NEXT:    vptestnmd %zmm1, %zmm1, %k0
-; AVX512-NEXT:    kshiftlw $12, %k0, %k0
-; AVX512-NEXT:    kshiftrw $12, %k0, %k1
-; AVX512-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm2 {%k1}
+; AVX512-NEXT:    kmovw %k0, %eax
+; AVX512-NEXT:    testb $1, %al
+; AVX512-NEXT:    je .LBB1_2
+; AVX512-NEXT:  # %bb.1: # %cond.load
+; AVX512-NEXT:    vmovq %xmm0, %rcx
+; AVX512-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX512-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; AVX512-NEXT:  .LBB1_2: # %else
+; AVX512-NEXT:    testb $2, %al
+; AVX512-NEXT:    je .LBB1_4
+; AVX512-NEXT:  # %bb.3: # %cond.load1
+; AVX512-NEXT:    vpextrq $1, %xmm0, %rcx
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
+; AVX512-NEXT:  .LBB1_4: # %else2
+; AVX512-NEXT:    testb $4, %al
+; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX512-NEXT:    jne .LBB1_5
+; AVX512-NEXT:  # %bb.6: # %else5
+; AVX512-NEXT:    testb $8, %al
+; AVX512-NEXT:    jne .LBB1_7
+; AVX512-NEXT:  .LBB1_8: # %else8
+; AVX512-NEXT:    vmovaps %xmm2, %xmm0
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
+; AVX512-NEXT:  .LBB1_5: # %cond.load4
+; AVX512-NEXT:    vmovq %xmm0, %rcx
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],mem[0],xmm2[3]
+; AVX512-NEXT:    testb $8, %al
+; AVX512-NEXT:    je .LBB1_8
+; AVX512-NEXT:  .LBB1_7: # %cond.load7
+; AVX512-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
 ; AVX512-NEXT:    vmovaps %xmm2, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %vptr0 = insertelement <4 x float*> undef, float* %base, i32 0
   %vptr1 = shufflevector <4 x float*> %vptr0, <4 x float*> undef, <4 x i32> zeroinitializer
   %vptr2 = getelementptr float, <4 x float*> %vptr1, <4 x i32> %idx
 
   %mask = icmp eq <4 x i32> %trigger, zeroinitializer

 ; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
 ; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
 ; AVX2-NEXT:    vmovaps %xmm2, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: gather_v4f32_v4i64_v4i32:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
 ; AVX512-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512-NEXT:    vmovq %rdi, %xmm3
+; AVX512-NEXT:    vpbroadcastq %xmm3, %ymm3
+; AVX512-NEXT:    vpsllq $2, %ymm0, %ymm0
+; AVX512-NEXT:    vpaddq %ymm0, %ymm3, %ymm0
 ; AVX512-NEXT:    vptestnmd %zmm1, %zmm1, %k0
-; AVX512-NEXT:    kshiftlw $12, %k0, %k0
-; AVX512-NEXT:    kshiftrw $12, %k0, %k1
-; AVX512-NEXT:    vgatherqps (%rdi,%zmm0,4), %ymm2 {%k1}
+; AVX512-NEXT:    kmovw %k0, %eax
+; AVX512-NEXT:    testb $1, %al
+; AVX512-NEXT:    je .LBB2_2
+; AVX512-NEXT:  # %bb.1: # %cond.load
+; AVX512-NEXT:    vmovq %xmm0, %rcx
+; AVX512-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX512-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; AVX512-NEXT:  .LBB2_2: # %else
+; AVX512-NEXT:    testb $2, %al
+; AVX512-NEXT:    je .LBB2_4
+; AVX512-NEXT:  # %bb.3: # %cond.load1
+; AVX512-NEXT:    vpextrq $1, %xmm0, %rcx
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
+; AVX512-NEXT:  .LBB2_4: # %else2
+; AVX512-NEXT:    testb $4, %al
+; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX512-NEXT:    jne .LBB2_5
+; AVX512-NEXT:  # %bb.6: # %else5
+; AVX512-NEXT:    testb $8, %al
+; AVX512-NEXT:    jne .LBB2_7
+; AVX512-NEXT:  .LBB2_8: # %else8
+; AVX512-NEXT:    vmovaps %xmm2, %xmm0
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
+; AVX512-NEXT:  .LBB2_5: # %cond.load4
+; AVX512-NEXT:    vmovq %xmm0, %rcx
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],mem[0],xmm2[3]
+; AVX512-NEXT:    testb $8, %al
+; AVX512-NEXT:    je .LBB2_8
+; AVX512-NEXT:  .LBB2_7: # %cond.load7
+; AVX512-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
 ; AVX512-NEXT:    vmovaps %xmm2, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %vptr0 = insertelement <4 x float*> undef, float* %base, i32 0
   %vptr1 = shufflevector <4 x float*> %vptr0, <4 x float*> undef, <4 x i32> zeroinitializer
   %vptr2 = getelementptr float, <4 x float*> %vptr1, <4 x i64> %idx
 
   %mask = icmp eq <4 x i32> %trigger, zeroinitializer

llvm/test/CodeGen/X86/masked_gather_scatter.ll

 declare <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*>, i32, <4 x i1>, <4 x float>)
 declare <4 x double> @llvm.masked.gather.v4f64.v4p0f64(<4 x double*>, i32, <4 x i1>, <4 x double>)
 declare <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*>, i32, <2 x i1>, <2 x double>)
 
 ; Gather smaller than existing instruction
 define <4 x float> @test15(float* %base, <4 x i32> %ind, <4 x i1> %mask) {
 ; KNL_64-LABEL: test15:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_64-NEXT:    vpslld $31, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
-; KNL_64-NEXT:    vmovaps %xmm1, %xmm0
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %ymm0
+; KNL_64-NEXT:    vpsllq $2, %ymm0, %ymm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %ymm1
+; KNL_64-NEXT:    vpaddq %ymm0, %ymm1, %ymm1
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    # implicit-def: $xmm0
+; KNL_64-NEXT:    je .LBB14_2
+; KNL_64-NEXT:  # %bb.1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; KNL_64-NEXT:  .LBB14_2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB14_4
+; KNL_64-NEXT:  # %bb.3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rcx
+; KNL_64-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
+; KNL_64-NEXT:  .LBB14_4: # %else2
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; KNL_64-NEXT:    jne .LBB14_5
+; KNL_64-NEXT:  # %bb.6: # %else5
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    jne .LBB14_7
+; KNL_64-NEXT:  .LBB14_8: # %else8
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB14_5: # %cond.load4
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB14_8
+; KNL_64-NEXT:  .LBB14_7: # %cond.load7
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rax
+; KNL_64-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test15:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpslld $31, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
-; KNL_32-NEXT:    vmovaps %xmm1, %xmm0
+; KNL_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm1
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    # implicit-def: $xmm0
+; KNL_32-NEXT:    jne .LBB14_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB14_3
+; KNL_32-NEXT:  .LBB14_4: # %else2
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    jne .LBB14_5
+; KNL_32-NEXT:  .LBB14_6: # %else5
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    jne .LBB14_7
+; KNL_32-NEXT:  .LBB14_8: # %else8
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB14_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm1, %ecx
+; KNL_32-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB14_4
+; KNL_32-NEXT:  .LBB14_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm1, %ecx
+; KNL_32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    je .LBB14_6
+; KNL_32-NEXT:  .LBB14_5: # %cond.load4
+; KNL_32-NEXT:    vpextrd $2, %xmm1, %ecx
+; KNL_32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    je .LBB14_8
+; KNL_32-NEXT:  .LBB14_7: # %cond.load7
+; KNL_32-NEXT:    vpextrd $3, %xmm1, %eax
+; KNL_32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test15:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm1, %xmm1
 ; SKX-NEXT:    vpmovd2m %xmm1, %k1
 ; SKX-NEXT:    vgatherdps (%rdi,%xmm0,4), %xmm1 {%k1}
; SKX_32-NEXT:    retl
   %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.random, i32 4, <4 x i1> %mask, <4 x float> undef)
   ret <4 x float>%res
 }
 
 ; Gather smaller than existing instruction
 define <4 x double> @test16(double* %base, <4 x i32> %ind, <4 x i1> %mask, <4 x double> %src0) {
 ; KNL_64-LABEL: test16:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $ymm2 killed $ymm2 def $zmm2
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_64-NEXT:    vpslld $31, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_64-NEXT:    vgatherdpd (%rdi,%ymm0,8), %zmm2 {%k1}
-; KNL_64-NEXT:    vmovapd %ymm2, %ymm0
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %ymm0
+; KNL_64-NEXT:    vpsllq $3, %ymm0, %ymm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %ymm1
+; KNL_64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    je .LBB15_2
+; KNL_64-NEXT:  # %bb.1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; KNL_64-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; KNL_64-NEXT:  .LBB15_2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB15_4
+; KNL_64-NEXT:  # %bb.3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vmovhps {{.*#+}} xmm1 = xmm2[0,1],mem[0,1]
+; KNL_64-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; KNL_64-NEXT:  .LBB15_4: # %else2
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; KNL_64-NEXT:    jne .LBB15_5
+; KNL_64-NEXT:  # %bb.6: # %else5
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    jne .LBB15_7
+; KNL_64-NEXT:  .LBB15_8: # %else8
+; KNL_64-NEXT:    vmovdqa %ymm2, %ymm0
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB15_5: # %cond.load4
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vextracti128 $1, %ymm2, %xmm1
+; KNL_64-NEXT:    vmovlps {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
+; KNL_64-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm2
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB15_8
+; KNL_64-NEXT:  .LBB15_7: # %cond.load7
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vextracti128 $1, %ymm2, %xmm0
+; KNL_64-NEXT:    vmovhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]
+; KNL_64-NEXT:    vinserti128 $1, %xmm0, %ymm2, %ymm2
+; KNL_64-NEXT:    vmovdqa %ymm2, %ymm0
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test16:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $ymm2 killed $ymm2 def $zmm2
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_32-NEXT:    vpslld $31, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vgatherdpd (%eax,%ymm0,8), %zmm2 {%k1}
-; KNL_32-NEXT:    vmovapd %ymm2, %ymm0
+; KNL_32-NEXT:    vpslld $3, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB15_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB15_3
+; KNL_32-NEXT:  .LBB15_4: # %else2
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    jne .LBB15_5
+; KNL_32-NEXT:  .LBB15_6: # %else5
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    jne .LBB15_7
+; KNL_32-NEXT:  .LBB15_8: # %else8
+; KNL_32-NEXT:    vmovdqa %ymm2, %ymm0
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB15_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; KNL_32-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB15_4
+; KNL_32-NEXT:  .LBB15_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; KNL_32-NEXT:    vmovhps {{.*#+}} xmm1 = xmm2[0,1],mem[0,1]
+; KNL_32-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    je .LBB15_6
+; KNL_32-NEXT:  .LBB15_5: # %cond.load4
+; KNL_32-NEXT:    vpextrd $2, %xmm0, %ecx
+; KNL_32-NEXT:    vextracti128 $1, %ymm2, %xmm1
+; KNL_32-NEXT:    vmovlps {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
+; KNL_32-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm2
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    je .LBB15_8
+; KNL_32-NEXT:  .LBB15_7: # %cond.load7
+; KNL_32-NEXT:    vpextrd $3, %xmm0, %eax
+; KNL_32-NEXT:    vextracti128 $1, %ymm2, %xmm0
+; KNL_32-NEXT:    vmovhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]
+; KNL_32-NEXT:    vinserti128 $1, %xmm0, %ymm2, %ymm2
+; KNL_32-NEXT:    vmovdqa %ymm2, %ymm0
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test16:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm1, %xmm1
 ; SKX-NEXT:    vpmovd2m %xmm1, %k1
 ; SKX-NEXT:    vgatherdpd (%rdi,%xmm0,8), %ymm2 {%k1}
 ; SKX-NEXT:    vmovapd %ymm2, %ymm0
; SKX_32-NEXT:    retl
   %gep.random = getelementptr double, double* %base, <4 x i64> %sext_ind
   %res = call <4 x double> @llvm.masked.gather.v4f64.v4p0f64(<4 x double*> %gep.random, i32 4, <4 x i1> %mask, <4 x double> %src0)
   ret <4 x double>%res
 }
 
 define <2 x double> @test17(double* %base, <2 x i32> %ind, <2 x i1> %mask, <2 x double> %src0) {
 ; KNL_64-LABEL: test17:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vgatherdpd (%rdi,%ymm0,8), %zmm2 {%k1}
-; KNL_64-NEXT:    vmovapd %xmm2, %xmm0
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %xmm0
+; KNL_64-NEXT:    vpsllq $3, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB16_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB16_3
+; KNL_64-NEXT:  .LBB16_4: # %else2
+; KNL_64-NEXT:    vmovaps %xmm2, %xmm0
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB16_1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovlps {{.*#+}} xmm2 = mem[0,1],xmm2[2,3]
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB16_4
+; KNL_64-NEXT:  .LBB16_3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vmovhps {{.*#+}} xmm2 = xmm2[0,1],mem[0,1]
+; KNL_64-NEXT:    vmovaps %xmm2, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test17:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vgatherdpd (%eax,%ymm0,8), %zmm2 {%k1}
-; KNL_32-NEXT:    vmovapd %xmm2, %xmm0
+; KNL_32-NEXT:    vpslld $3, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB16_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB16_3
+; KNL_32-NEXT:  .LBB16_4: # %else2
+; KNL_32-NEXT:    vmovaps %xmm2, %xmm0
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB16_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vmovlps {{.*#+}} xmm2 = mem[0,1],xmm2[2,3]
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB16_4
+; KNL_32-NEXT:  .LBB16_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vmovhps {{.*#+}} xmm2 = xmm2[0,1],mem[0,1]
+; KNL_32-NEXT:    vmovaps %xmm2, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test17:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX-NEXT:    vpmovq2m %xmm1, %k1
-; SKX-NEXT:    vgatherdpd (%rdi,%xmm0,8), %xmm2 {%k1}
-; SKX-NEXT:    vmovapd %xmm2, %xmm0
+; SKX-NEXT:    vpmovq2m %xmm1, %k0
+; SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; SKX-NEXT:    vpsllq $3, %xmm0, %xmm0
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB16_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB16_3
+; SKX-NEXT:  .LBB16_4: # %else2
+; SKX-NEXT:    vmovaps %xmm2, %xmm0
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB16_1: # %cond.load
+; SKX-NEXT:    vmovq %xmm0, %rcx
+; SKX-NEXT:    vmovlps {{.*#+}} xmm2 = mem[0,1],xmm2[2,3]
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB16_4
+; SKX-NEXT:  .LBB16_3: # %cond.load1
+; SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX-NEXT:    vmovhps {{.*#+}} xmm2 = xmm2[0,1],mem[0,1]
+; SKX-NEXT:    vmovaps %xmm2, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test17:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX_32-NEXT:    vpmovq2m %xmm1, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vgatherdpd (%eax,%xmm0,8), %xmm2 {%k1}
-; SKX_32-NEXT:    vmovapd %xmm2, %xmm0
+; SKX_32-NEXT:    vpmovq2m %xmm1, %k0
+; SKX_32-NEXT:    vpslld $3, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB16_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB16_3
+; SKX_32-NEXT:  .LBB16_4: # %else2
+; SKX_32-NEXT:    vmovaps %xmm2, %xmm0
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB16_1: # %cond.load
+; SKX_32-NEXT:    vmovd %xmm0, %ecx
+; SKX_32-NEXT:    vmovlps {{.*#+}} xmm2 = mem[0,1],xmm2[2,3]
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB16_4
+; SKX_32-NEXT:  .LBB16_3: # %cond.load1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vmovhps {{.*#+}} xmm2 = xmm2[0,1],mem[0,1]
+; SKX_32-NEXT:    vmovaps %xmm2, %xmm0
 ; SKX_32-NEXT:    retl
 
   %sext_ind = sext <2 x i32> %ind to <2 x i64>
   %gep.random = getelementptr double, double* %base, <2 x i64> %sext_ind
   %res = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %gep.random, i32 4, <2 x i1> %mask, <2 x double> %src0)
   ret <2 x double>%res
 }
 
 declare void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> , <4 x i32*> , i32 , <4 x i1> )
 declare void @llvm.masked.scatter.v4f64.v4p0f64(<4 x double> , <4 x double*> , i32 , <4 x i1> )
 declare void @llvm.masked.scatter.v2i64.v2p0i64(<2 x i64> , <2 x i64*> , i32 , <2 x i1> )
 declare void @llvm.masked.scatter.v2i32.v2p0i32(<2 x i32> , <2 x i32*> , i32 , <2 x i1> )
 declare void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> , <2 x float*> , i32 , <2 x i1> )
 
 define void @test18(<4 x i32>%a1, <4 x i32*> %ptr, <4 x i1>%mask) {
 ; KNL_64-LABEL: test18:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $ymm1 killed $ymm1 def $zmm1
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_64-NEXT:    vpslld $31, %xmm2, %xmm2
 ; KNL_64-NEXT:    vptestmd %zmm2, %zmm2, %k0
-; KNL_64-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_64-NEXT:    vpscatterqd %ymm0, (,%zmm1) {%k1}
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    je .LBB17_2
+; KNL_64-NEXT:  # %bb.1: # %cond.store
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vmovss %xmm0, (%rcx)
+; KNL_64-NEXT:  .LBB17_2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB17_4
+; KNL_64-NEXT:  # %bb.3: # %cond.store1
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rcx
+; KNL_64-NEXT:    vextractps $1, %xmm0, (%rcx)
+; KNL_64-NEXT:  .LBB17_4: # %else2
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; KNL_64-NEXT:    jne .LBB17_5
+; KNL_64-NEXT:  # %bb.6: # %else4
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    jne .LBB17_7
+; KNL_64-NEXT:  .LBB17_8: # %else6
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB17_5: # %cond.store3
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vextractps $2, %xmm0, (%rcx)
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB17_8
+; KNL_64-NEXT:  .LBB17_7: # %cond.store5
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rax
+; KNL_64-NEXT:    vextractps $3, %xmm0, (%rax)
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test18:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpslld $31, %xmm2, %xmm2
 ; KNL_32-NEXT:    vptestmd %zmm2, %zmm2, %k0
-; KNL_32-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_32-NEXT:    vpscatterdd %zmm0, (,%zmm1) {%k1}
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB17_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB17_3
+; KNL_32-NEXT:  .LBB17_4: # %else2
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    jne .LBB17_5
+; KNL_32-NEXT:  .LBB17_6: # %else4
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    jne .LBB17_7
+; KNL_32-NEXT:  .LBB17_8: # %else6
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB17_1: # %cond.store
+; KNL_32-NEXT:    vmovd %xmm1, %ecx
+; KNL_32-NEXT:    vmovss %xmm0, (%ecx)
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB17_4
+; KNL_32-NEXT:  .LBB17_3: # %cond.store1
+; KNL_32-NEXT:    vpextrd $1, %xmm1, %ecx
+; KNL_32-NEXT:    vextractps $1, %xmm0, (%ecx)
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    je .LBB17_6
+; KNL_32-NEXT:  .LBB17_5: # %cond.store3
+; KNL_32-NEXT:    vpextrd $2, %xmm1, %ecx
+; KNL_32-NEXT:    vextractps $2, %xmm0, (%ecx)
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    je .LBB17_8
+; KNL_32-NEXT:  .LBB17_7: # %cond.store5
+; KNL_32-NEXT:    vpextrd $3, %xmm1, %eax
+; KNL_32-NEXT:    vextractps $3, %xmm0, (%eax)
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test18:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm2, %xmm2
 ; SKX-NEXT:    vpmovd2m %xmm2, %k1
 ; SKX-NEXT:    vpscatterqd %xmm0, (,%ymm1) {%k1}
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test18:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpslld $31, %xmm2, %xmm2
 ; SKX_32-NEXT:    vpmovd2m %xmm2, %k1
 ; SKX_32-NEXT:    vpscatterdd %xmm0, (,%xmm1) {%k1}
 ; SKX_32-NEXT:    retl
   call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> %a1, <4 x i32*> %ptr, i32 4, <4 x i1> %mask)
   ret void
 }
 
 define void @test19(<4 x double>%a1, double* %ptr, <4 x i1>%mask, <4 x i64> %ind) {
 ; KNL_64-LABEL: test19:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $ymm2 killed $ymm2 def $zmm2
-; KNL_64-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
 ; KNL_64-NEXT:    vpslld $31, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_64-NEXT:    vscatterqpd %zmm0, (%rdi,%zmm2,8) {%k1}
+; KNL_64-NEXT:    vpsllq $3, %ymm2, %ymm1
+; KNL_64-NEXT:    vmovq %rdi, %xmm2
+; KNL_64-NEXT:    vpbroadcastq %xmm2, %ymm2
+; KNL_64-NEXT:    vpaddq %ymm1, %ymm2, %ymm1
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    je .LBB18_2
+; KNL_64-NEXT:  # %bb.1: # %cond.store
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vmovlps %xmm0, (%rcx)
+; KNL_64-NEXT:  .LBB18_2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB18_4
+; KNL_64-NEXT:  # %bb.3: # %cond.store1
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rcx
+; KNL_64-NEXT:    vmovhps %xmm0, (%rcx)
+; KNL_64-NEXT:  .LBB18_4: # %else2
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; KNL_64-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; KNL_64-NEXT:    jne .LBB18_5
+; KNL_64-NEXT:  # %bb.6: # %else4
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    jne .LBB18_7
+; KNL_64-NEXT:  .LBB18_8: # %else6
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB18_5: # %cond.store3
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vmovlps %xmm0, (%rcx)
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB18_8
+; KNL_64-NEXT:  .LBB18_7: # %cond.store5
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rax
+; KNL_64-NEXT:    vmovhps %xmm0, (%rax)
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test19:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    # kill: def $ymm2 killed $ymm2 def $zmm2
-; KNL_32-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
 ; KNL_32-NEXT:    vpslld $31, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vscatterqpd %zmm0, (%eax,%zmm2,8) {%k1}
+; KNL_32-NEXT:    vpmovqd %zmm2, %ymm1
+; KNL_32-NEXT:    vpslld $3, %xmm1, %xmm1
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm2
+; KNL_32-NEXT:    vpaddd %xmm1, %xmm2, %xmm1
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    je .LBB18_2
+; KNL_32-NEXT:  # %bb.1: # %cond.store
+; KNL_32-NEXT:    vmovd %xmm1, %ecx
+; KNL_32-NEXT:    vmovlps %xmm0, (%ecx)
+; KNL_32-NEXT:  .LBB18_2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB18_4
+; KNL_32-NEXT:  # %bb.3: # %cond.store1
+; KNL_32-NEXT:    vpextrd $1, %xmm1, %ecx
+; KNL_32-NEXT:    vmovhps %xmm0, (%ecx)
+; KNL_32-NEXT:  .LBB18_4: # %else2
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; KNL_32-NEXT:    jne .LBB18_5
+; KNL_32-NEXT:  # %bb.6: # %else4
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    jne .LBB18_7
+; KNL_32-NEXT:  .LBB18_8: # %else6
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB18_5: # %cond.store3
+; KNL_32-NEXT:    vpextrd $2, %xmm1, %ecx
+; KNL_32-NEXT:    vmovlps %xmm0, (%ecx)
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    je .LBB18_8
+; KNL_32-NEXT:  .LBB18_7: # %cond.store5
+; KNL_32-NEXT:    vpextrd $3, %xmm1, %eax
+; KNL_32-NEXT:    vmovhps %xmm0, (%eax)
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test19:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm1, %xmm1
 ; SKX-NEXT:    vpmovd2m %xmm1, %k1
 ; SKX-NEXT:    vscatterqpd %ymm0, (%rdi,%ymm2,8) {%k1}
; SKX_32-NEXT:    retl
   call void @llvm.masked.scatter.v4f64.v4p0f64(<4 x double> %a1, <4 x double*> %gep, i32 8, <4 x i1> %mask)
   ret void
 }
 
 ; Data type requires widening
 define void @test20(<2 x float>%a1, <2 x float*> %ptr, <2 x i1> %mask) {
 ; KNL_64-LABEL: test20:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; KNL_64-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vscatterqps %ymm0, (,%zmm1) {%k1}
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB19_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB19_3
+; KNL_64-NEXT:  .LBB19_4: # %else2
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB19_1: # %cond.store
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vmovd %xmm0, (%rcx)
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB19_4
+; KNL_64-NEXT:  .LBB19_3: # %cond.store1
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rax
+; KNL_64-NEXT:    vextractps $1, %xmm0, (%rax)
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test20:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; KNL_32-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    vscatterdps %zmm0, (,%zmm1) {%k1}
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB19_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB19_3
+; KNL_32-NEXT:  .LBB19_4: # %else2
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB19_1: # %cond.store
+; KNL_32-NEXT:    vmovd %xmm1, %ecx
+; KNL_32-NEXT:    vmovd %xmm0, (%ecx)
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB19_4
+; KNL_32-NEXT:  .LBB19_3: # %cond.store1
+; KNL_32-NEXT:    vpextrd $1, %xmm1, %eax
+; KNL_32-NEXT:    vextractps $1, %xmm0, (%eax)
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test20:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm2, %xmm2
-; SKX-NEXT:    vpmovq2m %xmm2, %k1
-; SKX-NEXT:    vscatterqps %xmm0, (,%xmm1) {%k1}
+; SKX-NEXT:    vpmovq2m %xmm2, %k0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB19_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB19_3
+; SKX-NEXT:  .LBB19_4: # %else2
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB19_1: # %cond.store
+; SKX-NEXT:    vmovq %xmm1, %rcx
+; SKX-NEXT:    vmovd %xmm0, (%rcx)
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB19_4
+; SKX-NEXT:  .LBB19_3: # %cond.store1
+; SKX-NEXT:    vpextrq $1, %xmm1, %rax
+; SKX-NEXT:    vextractps $1, %xmm0, (%rax)
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test20:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm2, %xmm2
-; SKX_32-NEXT:    vpmovq2m %xmm2, %k1
-; SKX_32-NEXT:    vscatterdps %xmm0, (,%xmm1) {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm2, %k0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB19_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB19_3
+; SKX_32-NEXT:  .LBB19_4: # %else2
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB19_1: # %cond.store
+; SKX_32-NEXT:    vmovd %xmm1, %ecx
+; SKX_32-NEXT:    vmovd %xmm0, (%ecx)
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB19_4
+; SKX_32-NEXT:  .LBB19_3: # %cond.store1
+; SKX_32-NEXT:    vpextrd $1, %xmm1, %eax
+; SKX_32-NEXT:    vextractps $1, %xmm0, (%eax)
 ; SKX_32-NEXT:    retl
   call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> %a1, <2 x float*> %ptr, i32 4, <2 x i1> %mask)
   ret void
 }
 
 ; Data type requires promotion
 define void @test21(<2 x i32>%a1, <2 x i32*> %ptr, <2 x i1>%mask) {
 ; KNL_64-LABEL: test21:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; KNL_64-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vpscatterqd %ymm0, (,%zmm1) {%k1}
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB20_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB20_3
+; KNL_64-NEXT:  .LBB20_4: # %else2
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB20_1: # %cond.store
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vmovss %xmm0, (%rcx)
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB20_4
+; KNL_64-NEXT:  .LBB20_3: # %cond.store1
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rax
+; KNL_64-NEXT:    vextractps $1, %xmm0, (%rax)
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test21:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; KNL_32-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    vpscatterdd %zmm0, (,%zmm1) {%k1}
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB20_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB20_3
+; KNL_32-NEXT:  .LBB20_4: # %else2
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB20_1: # %cond.store
+; KNL_32-NEXT:    vmovd %xmm1, %ecx
+; KNL_32-NEXT:    vmovss %xmm0, (%ecx)
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB20_4
+; KNL_32-NEXT:  .LBB20_3: # %cond.store1
+; KNL_32-NEXT:    vpextrd $1, %xmm1, %eax
+; KNL_32-NEXT:    vextractps $1, %xmm0, (%eax)
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test21:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm2, %xmm2
-; SKX-NEXT:    vpmovq2m %xmm2, %k1
-; SKX-NEXT:    vpscatterqd %xmm0, (,%xmm1) {%k1}
+; SKX-NEXT:    vpmovq2m %xmm2, %k0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB20_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB20_3
+; SKX-NEXT:  .LBB20_4: # %else2
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB20_1: # %cond.store
+; SKX-NEXT:    vmovq %xmm1, %rcx
+; SKX-NEXT:    vmovss %xmm0, (%rcx)
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB20_4
+; SKX-NEXT:  .LBB20_3: # %cond.store1
+; SKX-NEXT:    vpextrq $1, %xmm1, %rax
+; SKX-NEXT:    vextractps $1, %xmm0, (%rax)
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test21:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm2, %xmm2
-; SKX_32-NEXT:    vpmovq2m %xmm2, %k1
-; SKX_32-NEXT:    vpscatterdd %xmm0, (,%xmm1) {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm2, %k0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB20_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB20_3
+; SKX_32-NEXT:  .LBB20_4: # %else2
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB20_1: # %cond.store
+; SKX_32-NEXT:    vmovd %xmm1, %ecx
+; SKX_32-NEXT:    vmovss %xmm0, (%ecx)
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB20_4
+; SKX_32-NEXT:  .LBB20_3: # %cond.store1
+; SKX_32-NEXT:    vpextrd $1, %xmm1, %eax
+; SKX_32-NEXT:    vextractps $1, %xmm0, (%eax)
 ; SKX_32-NEXT:    retl
   call void @llvm.masked.scatter.v2i32.v2p0i32(<2 x i32> %a1, <2 x i32*> %ptr, i32 4, <2 x i1> %mask)
   ret void
 }
 
 ; The result type requires widening
 declare <2 x float> @llvm.masked.gather.v2f32.v2p0f32(<2 x float*>, i32, <2 x i1>, <2 x float>)
 
 define <2 x float> @test22(float* %base, <2 x i32> %ind, <2 x i1> %mask, <2 x float> %src0) {
 ; KNL_64-LABEL: test22:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm2 {%k1}
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %xmm0
+; KNL_64-NEXT:    vpsllq $2, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB21_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB21_3
+; KNL_64-NEXT:  .LBB21_4: # %else2
+; KNL_64-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB21_1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; KNL_64-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB21_4
+; KNL_64-NEXT:  .LBB21_3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; KNL_64-NEXT:    vmovaps %xmm2, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test22:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm2 {%k1}
+; KNL_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB21_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB21_3
+; KNL_32-NEXT:  .LBB21_4: # %else2
+; KNL_32-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB21_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; KNL_32-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB21_4
+; KNL_32-NEXT:  .LBB21_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; KNL_32-NEXT:    vmovaps %xmm2, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test22:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX-NEXT:    vpmovq2m %xmm1, %k1
-; SKX-NEXT:    vgatherdps (%rdi,%xmm0,4), %xmm2 {%k1}
+; SKX-NEXT:    vpmovq2m %xmm1, %k0
+; SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; SKX-NEXT:    vpsllq $2, %xmm0, %xmm0
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB21_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB21_3
+; SKX-NEXT:  .LBB21_4: # %else2
+; SKX-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB21_1: # %cond.load
+; SKX-NEXT:    vmovq %xmm0, %rcx
+; SKX-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SKX-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB21_4
+; SKX-NEXT:  .LBB21_3: # %cond.load1
+; SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; SKX-NEXT:    vmovaps %xmm2, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test22:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX_32-NEXT:    vpmovq2m %xmm1, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vgatherdps (%eax,%xmm0,4), %xmm2 {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm1, %k0
+; SKX_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB21_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB21_3
+; SKX_32-NEXT:  .LBB21_4: # %else2
+; SKX_32-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB21_1: # %cond.load
+; SKX_32-NEXT:    vmovd %xmm0, %ecx
+; SKX_32-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SKX_32-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB21_4
+; SKX_32-NEXT:  .LBB21_3: # %cond.load1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; SKX_32-NEXT:    vmovaps %xmm2, %xmm0
 ; SKX_32-NEXT:    retl
   %sext_ind = sext <2 x i32> %ind to <2 x i64>
   %gep.random = getelementptr float, float* %base, <2 x i64> %sext_ind
   %res = call <2 x float> @llvm.masked.gather.v2f32.v2p0f32(<2 x float*> %gep.random, i32 4, <2 x i1> %mask, <2 x float> %src0)
   ret <2 x float>%res
 }
 
 define <2 x float> @test22a(float* %base, <2 x i64> %ind, <2 x i1> %mask, <2 x float> %src0) {
 ; KNL_64-LABEL: test22a:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vgatherqps (%rdi,%zmm0,4), %ymm2 {%k1}
+; KNL_64-NEXT:    vpsllq $2, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB22_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB22_3
+; KNL_64-NEXT:  .LBB22_4: # %else2
+; KNL_64-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB22_1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; KNL_64-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB22_4
+; KNL_64-NEXT:  .LBB22_3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; KNL_64-NEXT:    vmovaps %xmm2, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test22a:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vgatherqps (%eax,%zmm0,4), %ymm2 {%k1}
+; KNL_32-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; KNL_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB22_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB22_3
+; KNL_32-NEXT:  .LBB22_4: # %else2
+; KNL_32-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB22_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; KNL_32-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB22_4
+; KNL_32-NEXT:  .LBB22_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; KNL_32-NEXT:    vmovaps %xmm2, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test22a:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX-NEXT:    vpmovq2m %xmm1, %k1
-; SKX-NEXT:    vgatherqps (%rdi,%xmm0,4), %xmm2 {%k1}
+; SKX-NEXT:    vpmovq2m %xmm1, %k0
+; SKX-NEXT:    vpsllq $2, %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB22_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB22_3
+; SKX-NEXT:  .LBB22_4: # %else2
+; SKX-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB22_1: # %cond.load
+; SKX-NEXT:    vmovq %xmm0, %rcx
+; SKX-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SKX-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB22_4
+; SKX-NEXT:  .LBB22_3: # %cond.load1
+; SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; SKX-NEXT:    vmovaps %xmm2, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test22a:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX_32-NEXT:    vpmovq2m %xmm1, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vgatherqps (%eax,%xmm0,4), %xmm2 {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm1, %k0
+; SKX_32-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SKX_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB22_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB22_3
+; SKX_32-NEXT:  .LBB22_4: # %else2
+; SKX_32-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB22_1: # %cond.load
+; SKX_32-NEXT:    vmovd %xmm0, %ecx
+; SKX_32-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SKX_32-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB22_4
+; SKX_32-NEXT:  .LBB22_3: # %cond.load1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; SKX_32-NEXT:    vmovaps %xmm2, %xmm0
 ; SKX_32-NEXT:    retl
   %gep.random = getelementptr float, float* %base, <2 x i64> %ind
   %res = call <2 x float> @llvm.masked.gather.v2f32.v2p0f32(<2 x float*> %gep.random, i32 4, <2 x i1> %mask, <2 x float> %src0)
   ret <2 x float>%res
 }
 
 declare <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*>, i32, <2 x i1>, <2 x i32>)
 declare <2 x i64> @llvm.masked.gather.v2i64.v2p0i64(<2 x i64*>, i32, <2 x i1>, <2 x i64>)
 
 define <2 x i32> @test23(i32* %base, <2 x i32> %ind, <2 x i1> %mask, <2 x i32> %src0) {
 ; KNL_64-LABEL: test23:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm2 {%k1}
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %xmm0
+; KNL_64-NEXT:    vpsllq $2, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB23_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB23_3
+; KNL_64-NEXT:  .LBB23_4: # %else2
+; KNL_64-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB23_1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrd $0, (%rcx), %xmm2, %xmm2
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB23_4
+; KNL_64-NEXT:  .LBB23_3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vpinsrd $1, (%rax), %xmm2, %xmm2
 ; KNL_64-NEXT:    vmovdqa %xmm2, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test23:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vpgatherdd (%eax,%zmm0,4), %zmm2 {%k1}
+; KNL_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB23_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB23_3
+; KNL_32-NEXT:  .LBB23_4: # %else2
+; KNL_32-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB23_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vpinsrd $0, (%ecx), %xmm2, %xmm2
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB23_4
+; KNL_32-NEXT:  .LBB23_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vpinsrd $1, (%eax), %xmm2, %xmm2
 ; KNL_32-NEXT:    vmovdqa %xmm2, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test23:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX-NEXT:    vpmovq2m %xmm1, %k1
-; SKX-NEXT:    vpgatherdd (%rdi,%xmm0,4), %xmm2 {%k1}
+; SKX-NEXT:    vpmovq2m %xmm1, %k0
+; SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; SKX-NEXT:    vpsllq $2, %xmm0, %xmm0
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB23_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB23_3
+; SKX-NEXT:  .LBB23_4: # %else2
+; SKX-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB23_1: # %cond.load
+; SKX-NEXT:    vmovq %xmm0, %rcx
+; SKX-NEXT:    vpinsrd $0, (%rcx), %xmm2, %xmm2
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB23_4
+; SKX-NEXT:  .LBB23_3: # %cond.load1
+; SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX-NEXT:    vpinsrd $1, (%rax), %xmm2, %xmm2
 ; SKX-NEXT:    vmovdqa %xmm2, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test23:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX_32-NEXT:    vpmovq2m %xmm1, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vpgatherdd (%eax,%xmm0,4), %xmm2 {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm1, %k0
+; SKX_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB23_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB23_3
+; SKX_32-NEXT:  .LBB23_4: # %else2
+; SKX_32-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB23_1: # %cond.load
+; SKX_32-NEXT:    vmovd %xmm0, %ecx
+; SKX_32-NEXT:    vpinsrd $0, (%ecx), %xmm2, %xmm2
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB23_4
+; SKX_32-NEXT:  .LBB23_3: # %cond.load1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vpinsrd $1, (%eax), %xmm2, %xmm2
 ; SKX_32-NEXT:    vmovdqa %xmm2, %xmm0
 ; SKX_32-NEXT:    retl
   %sext_ind = sext <2 x i32> %ind to <2 x i64>
   %gep.random = getelementptr i32, i32* %base, <2 x i64> %sext_ind
   %res = call <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*> %gep.random, i32 4, <2 x i1> %mask, <2 x i32> %src0)
   ret <2 x i32>%res
 }
 
 define <2 x i32> @test23b(i32* %base, <2 x i64> %ind, <2 x i1> %mask, <2 x i32> %src0) {
 ; KNL_64-LABEL: test23b:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vpgatherqd (%rdi,%zmm0,4), %ymm2 {%k1}
+; KNL_64-NEXT:    vpsllq $2, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB24_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB24_3
+; KNL_64-NEXT:  .LBB24_4: # %else2
+; KNL_64-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB24_1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrd $0, (%rcx), %xmm2, %xmm2
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB24_4
+; KNL_64-NEXT:  .LBB24_3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vpinsrd $1, (%rax), %xmm2, %xmm2
 ; KNL_64-NEXT:    vmovdqa %xmm2, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test23b:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vpgatherqd (%eax,%zmm0,4), %ymm2 {%k1}
+; KNL_32-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; KNL_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB24_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB24_3
+; KNL_32-NEXT:  .LBB24_4: # %else2
+; KNL_32-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB24_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vpinsrd $0, (%ecx), %xmm2, %xmm2
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB24_4
+; KNL_32-NEXT:  .LBB24_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vpinsrd $1, (%eax), %xmm2, %xmm2
 ; KNL_32-NEXT:    vmovdqa %xmm2, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test23b:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX-NEXT:    vpmovq2m %xmm1, %k1
-; SKX-NEXT:    vpgatherqd (%rdi,%xmm0,4), %xmm2 {%k1}
+; SKX-NEXT:    vpmovq2m %xmm1, %k0
+; SKX-NEXT:    vpsllq $2, %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB24_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB24_3
+; SKX-NEXT:  .LBB24_4: # %else2
+; SKX-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB24_1: # %cond.load
+; SKX-NEXT:    vmovq %xmm0, %rcx
+; SKX-NEXT:    vpinsrd $0, (%rcx), %xmm2, %xmm2
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB24_4
+; SKX-NEXT:  .LBB24_3: # %cond.load1
+; SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX-NEXT:    vpinsrd $1, (%rax), %xmm2, %xmm2
 ; SKX-NEXT:    vmovdqa %xmm2, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test23b:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX_32-NEXT:    vpmovq2m %xmm1, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vpgatherqd (%eax,%xmm0,4), %xmm2 {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm1, %k0
+; SKX_32-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SKX_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB24_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB24_3
+; SKX_32-NEXT:  .LBB24_4: # %else2
+; SKX_32-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB24_1: # %cond.load
+; SKX_32-NEXT:    vmovd %xmm0, %ecx
+; SKX_32-NEXT:    vpinsrd $0, (%ecx), %xmm2, %xmm2
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB24_4
+; SKX_32-NEXT:  .LBB24_3: # %cond.load1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vpinsrd $1, (%eax), %xmm2, %xmm2
 ; SKX_32-NEXT:    vmovdqa %xmm2, %xmm0
 ; SKX_32-NEXT:    retl
   %gep.random = getelementptr i32, i32* %base, <2 x i64> %ind
   %res = call <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*> %gep.random, i32 4, <2 x i1> %mask, <2 x i32> %src0)
   ret <2 x i32>%res
 }
 
 define <2 x i32> @test24(i32* %base, <2 x i32> %ind) {
 ; KNL_64-LABEL: test24:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; KNL_64-NEXT:    movw $3, %ax
-; KNL_64-NEXT:    kmovw %eax, %k1
-; KNL_64-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm1 {%k1}
-; KNL_64-NEXT:    vmovdqa %xmm1, %xmm0
-; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %xmm0
+; KNL_64-NEXT:    vpsllq $2, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; KNL_64-NEXT:    vmovq %xmm0, %rax
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; KNL_64-NEXT:    vpinsrd $1, (%rcx), %xmm0, %xmm0
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test24:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    movw $3, %cx
-; KNL_32-NEXT:    kmovw %ecx, %k1
-; KNL_32-NEXT:    vpgatherdd (%eax,%zmm0,4), %zmm1 {%k1}
-; KNL_32-NEXT:    vmovdqa %xmm1, %xmm0
-; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    vmovd %xmm0, %eax
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; KNL_32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; KNL_32-NEXT:    vpinsrd $1, (%ecx), %xmm0, %xmm0
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test24:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    movb $3, %al
-; SKX-NEXT:    kmovw %eax, %k1
-; SKX-NEXT:    vpgatherdd (%rdi,%xmm0,4), %xmm1 {%k1}
-; SKX-NEXT:    vmovdqa %xmm1, %xmm0
+; SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; SKX-NEXT:    vpsllq $2, %xmm0, %xmm0
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; SKX-NEXT:    vmovq %xmm0, %rax
+; SKX-NEXT:    vpextrq $1, %xmm0, %rcx
+; SKX-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SKX-NEXT:    vpinsrd $1, (%rcx), %xmm0, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test24:
 ; SKX_32:       # %bb.0:
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    movb $3, %cl
-; SKX_32-NEXT:    kmovw %ecx, %k1
-; SKX_32-NEXT:    vpgatherdd (%eax,%xmm0,4), %xmm1 {%k1}
-; SKX_32-NEXT:    vmovdqa %xmm1, %xmm0
+; SKX_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    vmovd %xmm0, %eax
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; SKX_32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SKX_32-NEXT:    vpinsrd $1, (%ecx), %xmm0, %xmm0
 ; SKX_32-NEXT:    retl
   %sext_ind = sext <2 x i32> %ind to <2 x i64>
   %gep.random = getelementptr i32, i32* %base, <2 x i64> %sext_ind
   %res = call <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*> %gep.random, i32 4, <2 x i1> <i1 true, i1 true>, <2 x i32> undef)
   ret <2 x i32>%res
 }
 
 define <2 x i64> @test25(i64* %base, <2 x i32> %ind, <2 x i1> %mask, <2 x i64> %src0) {
 ; KNL_64-LABEL: test25:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vpgatherdq (%rdi,%ymm0,8), %zmm2 {%k1}
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %xmm0
+; KNL_64-NEXT:    vpsllq $3, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB26_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB26_3
+; KNL_64-NEXT:  .LBB26_4: # %else2
+; KNL_64-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB26_1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrq $0, (%rcx), %xmm2, %xmm2
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB26_4
+; KNL_64-NEXT:  .LBB26_3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vpinsrq $1, (%rax), %xmm2, %xmm2
 ; KNL_64-NEXT:    vmovdqa %xmm2, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test25:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vpgatherdq (%eax,%ymm0,8), %zmm2 {%k1}
+; KNL_32-NEXT:    vpslld $3, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB26_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB26_3
+; KNL_32-NEXT:  .LBB26_4: # %else2
+; KNL_32-NEXT:    vmovdqa %xmm2, %xmm0
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB26_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vpinsrd $0, (%ecx), %xmm2, %xmm1
+; KNL_32-NEXT:    vpinsrd $1, 4(%ecx), %xmm1, %xmm2
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB26_4
+; KNL_32-NEXT:  .LBB26_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vpinsrd $2, (%eax), %xmm2, %xmm0
+; KNL_32-NEXT:    vpinsrd $3, 4(%eax), %xmm0, %xmm2
 ; KNL_32-NEXT:    vmovdqa %xmm2, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test25:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX-NEXT:    vpmovq2m %xmm1, %k1
-; SKX-NEXT:    vpgatherdq (%rdi,%xmm0,8), %xmm2 {%k1}
+; SKX-NEXT:    vpmovq2m %xmm1, %k0
+; SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; SKX-NEXT:    vpsllq $3, %xmm0, %xmm0
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB26_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB26_3
+; SKX-NEXT:  .LBB26_4: # %else2
+; SKX-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB26_1: # %cond.load
+; SKX-NEXT:    vmovq %xmm0, %rcx
+; SKX-NEXT:    vpinsrq $0, (%rcx), %xmm2, %xmm2
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB26_4
+; SKX-NEXT:  .LBB26_3: # %cond.load1
+; SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX-NEXT:    vpinsrq $1, (%rax), %xmm2, %xmm2
 ; SKX-NEXT:    vmovdqa %xmm2, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test25:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm1, %xmm1
-; SKX_32-NEXT:    vpmovq2m %xmm1, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vpgatherdq (%eax,%xmm0,8), %xmm2 {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm1, %k0
+; SKX_32-NEXT:    vpslld $3, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB26_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB26_3
+; SKX_32-NEXT:  .LBB26_4: # %else2
+; SKX_32-NEXT:    vmovdqa %xmm2, %xmm0
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB26_1: # %cond.load
+; SKX_32-NEXT:    vmovd %xmm0, %ecx
+; SKX_32-NEXT:    vpinsrd $0, (%ecx), %xmm2, %xmm1
+; SKX_32-NEXT:    vpinsrd $1, 4(%ecx), %xmm1, %xmm2
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB26_4
+; SKX_32-NEXT:  .LBB26_3: # %cond.load1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vpinsrd $2, (%eax), %xmm2, %xmm0
+; SKX_32-NEXT:    vpinsrd $3, 4(%eax), %xmm0, %xmm2
 ; SKX_32-NEXT:    vmovdqa %xmm2, %xmm0
 ; SKX_32-NEXT:    retl
   %sext_ind = sext <2 x i32> %ind to <2 x i64>
   %gep.random = getelementptr i64, i64* %base, <2 x i64> %sext_ind
   %res = call <2 x i64> @llvm.masked.gather.v2i64.v2p0i64(<2 x i64*> %gep.random, i32 8, <2 x i1> %mask, <2 x i64> %src0)
   ret <2 x i64>%res
 }
 
 define <2 x i64> @test26(i64* %base, <2 x i32> %ind, <2 x i64> %src0) {
 ; KNL_64-LABEL: test26:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; KNL_64-NEXT:    movb $3, %al
-; KNL_64-NEXT:    kmovw %eax, %k1
-; KNL_64-NEXT:    vpgatherdq (%rdi,%ymm0,8), %zmm1 {%k1}
-; KNL_64-NEXT:    vmovdqa %xmm1, %xmm0
-; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %xmm0
+; KNL_64-NEXT:    vpsllq $3, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm2
+; KNL_64-NEXT:    vpbroadcastq %xmm2, %xmm2
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; KNL_64-NEXT:    vmovq %xmm0, %rax
+; KNL_64-NEXT:    vpinsrq $0, (%rax), %xmm1, %xmm1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vpinsrq $1, (%rax), %xmm1, %xmm0
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test26:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    movb $3, %cl
-; KNL_32-NEXT:    kmovw %ecx, %k1
-; KNL_32-NEXT:    vpgatherdq (%eax,%ymm0,8), %zmm1 {%k1}
-; KNL_32-NEXT:    vmovdqa %xmm1, %xmm0
-; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    vpslld $3, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm2
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm2, %xmm0
+; KNL_32-NEXT:    vmovd %xmm0, %eax
+; KNL_32-NEXT:    vpinsrd $0, (%eax), %xmm1, %xmm1
+; KNL_32-NEXT:    vpinsrd $1, 4(%eax), %xmm1, %xmm1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vpinsrd $2, (%eax), %xmm1, %xmm0
+; KNL_32-NEXT:    vpinsrd $3, 4(%eax), %xmm0, %xmm0
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test26:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    kxnorw %k0, %k0, %k1
-; SKX-NEXT:    vpgatherdq (%rdi,%xmm0,8), %xmm1 {%k1}
-; SKX-NEXT:    vmovdqa %xmm1, %xmm0
+; SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm2
+; SKX-NEXT:    vpsllq $3, %xmm0, %xmm0
+; SKX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; SKX-NEXT:    vmovq %xmm0, %rax
+; SKX-NEXT:    vpinsrq $0, (%rax), %xmm1, %xmm1
+; SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX-NEXT:    vpinsrq $1, (%rax), %xmm1, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test26:
 ; SKX_32:       # %bb.0:
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
-; SKX_32-NEXT:    vpgatherdq (%eax,%xmm0,8), %xmm1 {%k1}
-; SKX_32-NEXT:    vmovdqa %xmm1, %xmm0
+; SKX_32-NEXT:    vpslld $3, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    vmovd %xmm0, %eax
+; SKX_32-NEXT:    vpinsrd $0, (%eax), %xmm1, %xmm1
+; SKX_32-NEXT:    vpinsrd $1, 4(%eax), %xmm1, %xmm1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vpinsrd $2, (%eax), %xmm1, %xmm0
+; SKX_32-NEXT:    vpinsrd $3, 4(%eax), %xmm0, %xmm0
 ; SKX_32-NEXT:    retl
   %sext_ind = sext <2 x i32> %ind to <2 x i64>
   %gep.random = getelementptr i64, i64* %base, <2 x i64> %sext_ind
   %res = call <2 x i64> @llvm.masked.gather.v2i64.v2p0i64(<2 x i64*> %gep.random, i32 8, <2 x i1> <i1 true, i1 true>, <2 x i64> %src0)
   ret <2 x i64>%res
 }
 
 ; Result type requires widening; all-ones mask
 define <2 x float> @test27(float* %base, <2 x i32> %ind) {
 ; KNL_64-LABEL: test27:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; KNL_64-NEXT:    movw $3, %ax
-; KNL_64-NEXT:    kmovw %eax, %k1
-; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
-; KNL_64-NEXT:    vmovaps %xmm1, %xmm0
-; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    vpmovsxdq %xmm0, %xmm0
+; KNL_64-NEXT:    vpsllq $2, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; KNL_64-NEXT:    vmovq %xmm0, %rax
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; KNL_64-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test27:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    movw $3, %cx
-; KNL_32-NEXT:    kmovw %ecx, %k1
-; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
-; KNL_32-NEXT:    vmovaps %xmm1, %xmm0
-; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; KNL_32-NEXT:    vmovd %xmm0, %eax
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; KNL_32-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; KNL_32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test27:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    movb $3, %al
-; SKX-NEXT:    kmovw %eax, %k1
-; SKX-NEXT:    vgatherdps (%rdi,%xmm0,4), %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %xmm1, %xmm0
+; SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; SKX-NEXT:    vpsllq $2, %xmm0, %xmm0
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; SKX-NEXT:    vmovq %xmm0, %rax
+; SKX-NEXT:    vpextrq $1, %xmm0, %rcx
+; SKX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SKX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test27:
 ; SKX_32:       # %bb.0:
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    movb $3, %cl
-; SKX_32-NEXT:    kmovw %ecx, %k1
-; SKX_32-NEXT:    vgatherdps (%eax,%xmm0,4), %xmm1 {%k1}
-; SKX_32-NEXT:    vmovaps %xmm1, %xmm0
+; SKX_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    vmovd %xmm0, %eax
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; SKX_32-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SKX_32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; SKX_32-NEXT:    retl
   %sext_ind = sext <2 x i32> %ind to <2 x i64>
   %gep.random = getelementptr float, float* %base, <2 x i64> %sext_ind
   %res = call <2 x float> @llvm.masked.gather.v2f32.v2p0f32(<2 x float*> %gep.random, i32 4, <2 x i1> <i1 true, i1 true>, <2 x float> undef)
   ret <2 x float>%res
 }
 
 ; Data type requires promotion, mask is all-ones
 define void @test28(<2 x i32>%a1, <2 x i32*> %ptr) {
 ; KNL_64-LABEL: test28:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; KNL_64-NEXT:    movb $3, %al
-; KNL_64-NEXT:    kmovw %eax, %k1
-; KNL_64-NEXT:    vpscatterqd %ymm0, (,%zmm1) {%k1}
-; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    vmovq %xmm1, %rax
+; KNL_64-NEXT:    vmovss %xmm0, (%rax)
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rax
+; KNL_64-NEXT:    vextractps $1, %xmm0, (%rax)
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test28:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; KNL_32-NEXT:    movw $3, %ax
-; KNL_32-NEXT:    kmovw %eax, %k1
-; KNL_32-NEXT:    vpscatterdd %zmm0, (,%zmm1) {%k1}
-; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    vmovd %xmm1, %eax
+; KNL_32-NEXT:    vmovss %xmm0, (%eax)
+; KNL_32-NEXT:    vpextrd $1, %xmm1, %eax
+; KNL_32-NEXT:    vextractps $1, %xmm0, (%eax)
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test28:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    kxnorw %k0, %k0, %k1
-; SKX-NEXT:    vpscatterqd %xmm0, (,%xmm1) {%k1}
+; SKX-NEXT:    vmovq %xmm1, %rax
+; SKX-NEXT:    vmovss %xmm0, (%rax)
+; SKX-NEXT:    vpextrq $1, %xmm1, %rax
+; SKX-NEXT:    vextractps $1, %xmm0, (%rax)
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test28:
 ; SKX_32:       # %bb.0:
-; SKX_32-NEXT:    movb $3, %al
-; SKX_32-NEXT:    kmovw %eax, %k1
-; SKX_32-NEXT:    vpscatterdd %xmm0, (,%xmm1) {%k1}
+; SKX_32-NEXT:    vmovd %xmm1, %eax
+; SKX_32-NEXT:    vmovss %xmm0, (%eax)
+; SKX_32-NEXT:    vpextrd $1, %xmm1, %eax
+; SKX_32-NEXT:    vextractps $1, %xmm0, (%eax)
 ; SKX_32-NEXT:    retl
   call void @llvm.masked.scatter.v2i32.v2p0i32(<2 x i32> %a1, <2 x i32*> %ptr, i32 4, <2 x i1> <i1 true, i1 true>)
   ret void
 }
 
 ; SCALAR-LABEL: test29
 ; SCALAR:      extractelement <16 x float*>
 ; SCALAR-NEXT: load float
; SKX_32-NEXT:    retl
   call void @llvm.masked.scatter.v16f64.v16p0f64(<16 x double> %src0, <16 x double*> %ptrs, i32 4, <16 x i1> %mask)
   ret void
 }
 declare void @llvm.masked.scatter.v16f64.v16p0f64(<16 x double> %src0, <16 x double*> %ptrs, i32, <16 x i1> %mask)
 
 define <4 x i64> @test_pr28312(<4 x i64*> %p1, <4 x i1> %k, <4 x i1> %k2,<4 x i64> %d) {
 ; KNL_64-LABEL: test_pr28312:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
 ; KNL_64-NEXT:    vpslld $31, %xmm1, %xmm1
 ; KNL_64-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL_64-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_64-NEXT:    vpgatherqq (,%zmm0), %zmm1 {%k1}
-; KNL_64-NEXT:    vpaddq %ymm1, %ymm1, %ymm0
-; KNL_64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0
-; KNL_64-NEXT:    retq
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    # implicit-def: $ymm1
+; KNL_64-NEXT:    je .LBB42_2
+; KNL_64-NEXT:  # %bb.1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; KNL_64-NEXT:  .LBB42_2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB42_4
+; KNL_64-NEXT:  # %bb.3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrq $1, (%rcx), %xmm1, %xmm2
+; KNL_64-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; KNL_64-NEXT:  .LBB42_4: # %else2
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; KNL_64-NEXT:    je .LBB42_6
+; KNL_64-NEXT:  # %bb.5: # %cond.load4
+; KNL_64-NEXT:    vmovq %xmm2, %rcx
+; KNL_64-NEXT:    vextracti128 $1, %ymm1, %xmm3
+; KNL_64-NEXT:    vpinsrq $0, (%rcx), %xmm3, %xmm3
+; KNL_64-NEXT:    vinserti128 $1, %xmm3, %ymm1, %ymm1
+; KNL_64-NEXT:  .LBB42_6: # %else5
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB42_8
+; KNL_64-NEXT:  # %bb.7: # %cond.load7
+; KNL_64-NEXT:    vpextrq $1, %xmm2, %rax
+; KNL_64-NEXT:    vextracti128 $1, %ymm1, %xmm3
+; KNL_64-NEXT:    vpinsrq $1, (%rax), %xmm3, %xmm3
+; KNL_64-NEXT:    vinserti128 $1, %xmm3, %ymm1, %ymm1
+; KNL_64-NEXT:  .LBB42_8: # %else8
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    # implicit-def: $ymm3
+; KNL_64-NEXT:    jne .LBB42_9
+; KNL_64-NEXT:  # %bb.10: # %else15
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB42_11
+; KNL_64-NEXT:  .LBB42_12: # %else21
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    jne .LBB42_13
+; KNL_64-NEXT:  .LBB42_14: # %else27
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB42_16
+; KNL_64-NEXT:  .LBB42_15: # %cond.load29
+; KNL_64-NEXT:    vpextrq $1, %xmm2, %rax
+; KNL_64-NEXT:    vextracti128 $1, %ymm3, %xmm4
+; KNL_64-NEXT:    vpinsrq $1, (%rax), %xmm4, %xmm4
+; KNL_64-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm3
+; KNL_64-NEXT:  .LBB42_16: # %else33
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    # implicit-def: $ymm4
+; KNL_64-NEXT:    jne .LBB42_17
+; KNL_64-NEXT:  # %bb.18: # %else40
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB42_19
+; KNL_64-NEXT:  .LBB42_20: # %else46
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    jne .LBB42_21
+; KNL_64-NEXT:  .LBB42_22: # %else52
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB42_24
+; KNL_64-NEXT:  .LBB42_23: # %cond.load54
+; KNL_64-NEXT:    vpextrq $1, %xmm2, %rax
+; KNL_64-NEXT:    vextracti128 $1, %ymm4, %xmm0
+; KNL_64-NEXT:    vpinsrq $1, (%rax), %xmm0, %xmm0
+; KNL_64-NEXT:    vinserti128 $1, %xmm0, %ymm4, %ymm4
+; KNL_64-NEXT:  .LBB42_24: # %else58
+; KNL_64-NEXT:    vpaddq %ymm3, %ymm1, %ymm0
+; KNL_64-NEXT:    vpaddq %ymm4, %ymm0, %ymm0
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB42_9: # %cond.load11
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovq {{.*#+}} xmm3 = mem[0],zero
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB42_12
+; KNL_64-NEXT:  .LBB42_11: # %cond.load17
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrq $1, (%rcx), %xmm3, %xmm4
+; KNL_64-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    je .LBB42_14
+; KNL_64-NEXT:  .LBB42_13: # %cond.load23
+; KNL_64-NEXT:    vmovq %xmm2, %rcx
+; KNL_64-NEXT:    vextracti128 $1, %ymm3, %xmm4
+; KNL_64-NEXT:    vpinsrq $0, (%rcx), %xmm4, %xmm4
+; KNL_64-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm3
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    jne .LBB42_15
+; KNL_64-NEXT:    jmp .LBB42_16
+; KNL_64-NEXT:  .LBB42_17: # %cond.load36
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovq {{.*#+}} xmm4 = mem[0],zero
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB42_20
+; KNL_64-NEXT:  .LBB42_19: # %cond.load42
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrq $1, (%rcx), %xmm4, %xmm0
+; KNL_64-NEXT:    vpblendd {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    je .LBB42_22
+; KNL_64-NEXT:  .LBB42_21: # %cond.load48
+; KNL_64-NEXT:    vmovq %xmm2, %rcx
+; KNL_64-NEXT:    vextracti128 $1, %ymm4, %xmm0
+; KNL_64-NEXT:    vpinsrq $0, (%rcx), %xmm0, %xmm0
+; KNL_64-NEXT:    vinserti128 $1, %xmm0, %ymm4, %ymm4
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    jne .LBB42_23
+; KNL_64-NEXT:    jmp .LBB42_24
 ;
 ; KNL_32-LABEL: test_pr28312:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    pushl %ebp
 ; KNL_32-NEXT:    .cfi_def_cfa_offset 8
 ; KNL_32-NEXT:    .cfi_offset %ebp, -8
 ; KNL_32-NEXT:    movl %esp, %ebp
 ; KNL_32-NEXT:    .cfi_def_cfa_register %ebp
+; KNL_32-NEXT:    pushl %ebx
+; KNL_32-NEXT:    pushl %esi
 ; KNL_32-NEXT:    andl $-32, %esp
 ; KNL_32-NEXT:    subl $32, %esp
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
+; KNL_32-NEXT:    .cfi_offset %esi, -16
+; KNL_32-NEXT:    .cfi_offset %ebx, -12
 ; KNL_32-NEXT:    vpslld $31, %xmm1, %xmm1
 ; KNL_32-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; KNL_32-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_32-NEXT:    vpgatherdq (,%ymm0), %zmm1 {%k1}
-; KNL_32-NEXT:    vpaddq %ymm1, %ymm1, %ymm0
+; KNL_32-NEXT:    kmovw %k0, %ebx
+; KNL_32-NEXT:    testb $1, %bl
+; KNL_32-NEXT:    vmovd %xmm0, %eax
+; KNL_32-NEXT:    # implicit-def: $ymm1
+; KNL_32-NEXT:    je .LBB42_2
+; KNL_32-NEXT:  # %bb.1: # %cond.load
+; KNL_32-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; KNL_32-NEXT:  .LBB42_2: # %else
+; KNL_32-NEXT:    testb $2, %bl
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; KNL_32-NEXT:    je .LBB42_4
+; KNL_32-NEXT:  # %bb.3: # %cond.load1
+; KNL_32-NEXT:    vpinsrd $2, (%ecx), %xmm1, %xmm2
+; KNL_32-NEXT:    vpinsrd $3, 4(%ecx), %xmm2, %xmm2
+; KNL_32-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; KNL_32-NEXT:  .LBB42_4: # %else2
+; KNL_32-NEXT:    testb $4, %bl
+; KNL_32-NEXT:    vpextrd $2, %xmm0, %edx
+; KNL_32-NEXT:    je .LBB42_6
+; KNL_32-NEXT:  # %bb.5: # %cond.load4
+; KNL_32-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; KNL_32-NEXT:    vpinsrd $0, (%edx), %xmm2, %xmm2
+; KNL_32-NEXT:    vpinsrd $1, 4(%edx), %xmm2, %xmm2
+; KNL_32-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
+; KNL_32-NEXT:  .LBB42_6: # %else5
+; KNL_32-NEXT:    testb $8, %bl
+; KNL_32-NEXT:    vpextrd $3, %xmm0, %esi
+; KNL_32-NEXT:    je .LBB42_8
+; KNL_32-NEXT:  # %bb.7: # %cond.load7
+; KNL_32-NEXT:    vextracti128 $1, %ymm1, %xmm0
+; KNL_32-NEXT:    vpinsrd $2, (%esi), %xmm0, %xmm0
+; KNL_32-NEXT:    vpinsrd $3, 4(%esi), %xmm0, %xmm0
+; KNL_32-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm1
+; KNL_32-NEXT:  .LBB42_8: # %else8
+; KNL_32-NEXT:    kmovw %k0, %ebx
+; KNL_32-NEXT:    testb $1, %bl
+; KNL_32-NEXT:    # implicit-def: $ymm0
+; KNL_32-NEXT:    jne .LBB42_9
+; KNL_32-NEXT:  # %bb.10: # %else15
+; KNL_32-NEXT:    testb $2, %bl
+; KNL_32-NEXT:    jne .LBB42_11
+; KNL_32-NEXT:  .LBB42_12: # %else21
+; KNL_32-NEXT:    testb $4, %bl
+; KNL_32-NEXT:    jne .LBB42_13
+; KNL_32-NEXT:  .LBB42_14: # %else27
+; KNL_32-NEXT:    testb $8, %bl
+; KNL_32-NEXT:    je .LBB42_16
+; KNL_32-NEXT:  .LBB42_15: # %cond.load29
+; KNL_32-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; KNL_32-NEXT:    vpinsrd $2, (%esi), %xmm2, %xmm2
+; KNL_32-NEXT:    vpinsrd $3, 4(%esi), %xmm2, %xmm2
+; KNL_32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
+; KNL_32-NEXT:  .LBB42_16: # %else33
+; KNL_32-NEXT:    kmovw %k0, %ebx
+; KNL_32-NEXT:    testb $1, %bl
+; KNL_32-NEXT:    # implicit-def: $ymm2
+; KNL_32-NEXT:    jne .LBB42_17
+; KNL_32-NEXT:  # %bb.18: # %else40
+; KNL_32-NEXT:    testb $2, %bl
+; KNL_32-NEXT:    jne .LBB42_19
+; KNL_32-NEXT:  .LBB42_20: # %else46
+; KNL_32-NEXT:    testb $4, %bl
+; KNL_32-NEXT:    jne .LBB42_21
+; KNL_32-NEXT:  .LBB42_22: # %else52
+; KNL_32-NEXT:    testb $8, %bl
+; KNL_32-NEXT:    je .LBB42_24
+; KNL_32-NEXT:  .LBB42_23: # %cond.load54
+; KNL_32-NEXT:    vextracti128 $1, %ymm2, %xmm3
+; KNL_32-NEXT:    vpinsrd $2, (%esi), %xmm3, %xmm3
+; KNL_32-NEXT:    vpinsrd $3, 4(%esi), %xmm3, %xmm3
+; KNL_32-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; KNL_32-NEXT:  .LBB42_24: # %else58
 ; KNL_32-NEXT:    vpaddq %ymm0, %ymm1, %ymm0
-; KNL_32-NEXT:    movl %ebp, %esp
+; KNL_32-NEXT:    vpaddq %ymm2, %ymm0, %ymm0
+; KNL_32-NEXT:    leal -8(%ebp), %esp
+; KNL_32-NEXT:    popl %esi
+; KNL_32-NEXT:    popl %ebx
 ; KNL_32-NEXT:    popl %ebp
 ; KNL_32-NEXT:    .cfi_def_cfa %esp, 4
 ; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB42_9: # %cond.load11
+; KNL_32-NEXT:    .cfi_def_cfa %ebp, 8
+; KNL_32-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; KNL_32-NEXT:    testb $2, %bl
+; KNL_32-NEXT:    je .LBB42_12
+; KNL_32-NEXT:  .LBB42_11: # %cond.load17
+; KNL_32-NEXT:    vpinsrd $2, (%ecx), %xmm0, %xmm2
+; KNL_32-NEXT:    vpinsrd $3, 4(%ecx), %xmm2, %xmm2
+; KNL_32-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; KNL_32-NEXT:    testb $4, %bl
+; KNL_32-NEXT:    je .LBB42_14
+; KNL_32-NEXT:  .LBB42_13: # %cond.load23
+; KNL_32-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; KNL_32-NEXT:    vpinsrd $0, (%edx), %xmm2, %xmm2
+; KNL_32-NEXT:    vpinsrd $1, 4(%edx), %xmm2, %xmm2
+; KNL_32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
+; KNL_32-NEXT:    testb $8, %bl
+; KNL_32-NEXT:    jne .LBB42_15
+; KNL_32-NEXT:    jmp .LBB42_16
+; KNL_32-NEXT:  .LBB42_17: # %cond.load36
+; KNL_32-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
+; KNL_32-NEXT:    testb $2, %bl
+; KNL_32-NEXT:    je .LBB42_20
+; KNL_32-NEXT:  .LBB42_19: # %cond.load42
+; KNL_32-NEXT:    vpinsrd $2, (%ecx), %xmm2, %xmm3
+; KNL_32-NEXT:    vpinsrd $3, 4(%ecx), %xmm3, %xmm3
+; KNL_32-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; KNL_32-NEXT:    testb $4, %bl
+; KNL_32-NEXT:    je .LBB42_22
+; KNL_32-NEXT:  .LBB42_21: # %cond.load48
+; KNL_32-NEXT:    vextracti128 $1, %ymm2, %xmm3
+; KNL_32-NEXT:    vpinsrd $0, (%edx), %xmm3, %xmm3
+; KNL_32-NEXT:    vpinsrd $1, 4(%edx), %xmm3, %xmm3
+; KNL_32-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; KNL_32-NEXT:    testb $8, %bl
+; KNL_32-NEXT:    jne .LBB42_23
+; KNL_32-NEXT:    jmp .LBB42_24
 ;
 ; SKX-LABEL: test_pr28312:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm1, %xmm1
 ; SKX-NEXT:    vpmovd2m %xmm1, %k1
 ; SKX-NEXT:    vpgatherqq (,%ymm0), %ymm1 {%k1}
 ; SKX-NEXT:    vpaddq %ymm1, %ymm1, %ymm0
 ; SKX-NEXT:    vpaddq %ymm0, %ymm1, %ymm0

 }
 declare <1 x i32> @llvm.masked.gather.v1i32.v1p0i32(<1 x i32*>, i32, <1 x i1>, <1 x i32>)
 
 ; Make sure we don't crash when the index element type is larger than i64 and we need to widen the result
 ; This experienced a bad interaction when we widened and then tried to split.
 define <2 x float> @large_index(float* %base, <2 x i128> %ind, <2 x i1> %mask, <2 x float> %src0) {
 ; KNL_64-LABEL: large_index:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $ymm1
 ; KNL_64-NEXT:    vpsllq $63, %xmm0, %xmm0
 ; KNL_64-NEXT:    vptestmq %zmm0, %zmm0, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
 ; KNL_64-NEXT:    vmovq %rcx, %xmm0
 ; KNL_64-NEXT:    vmovq %rsi, %xmm2
 ; KNL_64-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
-; KNL_64-NEXT:    vgatherqps (%rdi,%zmm0,4), %ymm1 {%k1}
+; KNL_64-NEXT:    vpsllq $2, %xmm0, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm2
+; KNL_64-NEXT:    vpbroadcastq %xmm2, %xmm2
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB47_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB47_3
+; KNL_64-NEXT:  .LBB47_4: # %else2
+; KNL_64-NEXT:    vmovdqa %xmm1, %xmm0
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB47_1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovd {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; KNL_64-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB47_4
+; KNL_64-NEXT:  .LBB47_3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[2,3]
 ; KNL_64-NEXT:    vmovaps %xmm1, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: large_index:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $ymm1
 ; KNL_32-NEXT:    vpsllq $63, %xmm0, %xmm0
 ; KNL_32-NEXT:    vptestmq %zmm0, %zmm0, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; KNL_32-NEXT:    vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0
-; KNL_32-NEXT:    vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0
-; KNL_32-NEXT:    vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0
-; KNL_32-NEXT:    vgatherqps (%eax,%zmm0,4), %ymm1 {%k1}
+; KNL_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm2
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm2, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB47_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB47_3
+; KNL_32-NEXT:  .LBB47_4: # %else2
+; KNL_32-NEXT:    vmovdqa %xmm1, %xmm0
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB47_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vmovd {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; KNL_32-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB47_4
+; KNL_32-NEXT:  .LBB47_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[2,3]
 ; KNL_32-NEXT:    vmovaps %xmm1, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: large_index:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm0, %xmm0
-; SKX-NEXT:    vpmovq2m %xmm0, %k1
+; SKX-NEXT:    vpmovq2m %xmm0, %k0
 ; SKX-NEXT:    vmovq %rcx, %xmm0
 ; SKX-NEXT:    vmovq %rsi, %xmm2
 ; SKX-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
-; SKX-NEXT:    vgatherqps (%rdi,%xmm0,4), %xmm1 {%k1}
+; SKX-NEXT:    vpsllq $2, %xmm0, %xmm0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm2
+; SKX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB47_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB47_3
+; SKX-NEXT:  .LBB47_4: # %else2
+; SKX-NEXT:    vmovdqa %xmm1, %xmm0
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB47_1: # %cond.load
+; SKX-NEXT:    vmovq %xmm0, %rcx
+; SKX-NEXT:    vmovd {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SKX-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB47_4
+; SKX-NEXT:  .LBB47_3: # %cond.load1
+; SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[2,3]
 ; SKX-NEXT:    vmovaps %xmm1, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: large_index:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm0, %xmm0
-; SKX_32-NEXT:    vpmovq2m %xmm0, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; SKX_32-NEXT:    vpmovq2m %xmm0, %k0
 ; SKX_32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SKX_32-NEXT:    vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0
-; SKX_32-NEXT:    vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0
-; SKX_32-NEXT:    vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0
-; SKX_32-NEXT:    vgatherqps (%eax,%xmm0,4), %xmm1 {%k1}
+; SKX_32-NEXT:    vpslld $2, %xmm0, %xmm0
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB47_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB47_3
+; SKX_32-NEXT:  .LBB47_4: # %else2
+; SKX_32-NEXT:    vmovaps %xmm1, %xmm0
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB47_1: # %cond.load
+; SKX_32-NEXT:    vmovd %xmm0, %ecx
+; SKX_32-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SKX_32-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB47_4
+; SKX_32-NEXT:  .LBB47_3: # %cond.load1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[2,3]
 ; SKX_32-NEXT:    vmovaps %xmm1, %xmm0
 ; SKX_32-NEXT:    retl
   %gep.random = getelementptr float, float* %base, <2 x i128> %ind
   %res = call <2 x float> @llvm.masked.gather.v2f32.v2p0f32(<2 x float*> %gep.random, i32 4, <2 x i1> %mask, <2 x float> %src0)
   ret <2 x float>%res
 }
 
 ; Make sure we allow index to be sign extended from a smaller than i32 element size.
; SKX_32-NEXT:    retl
   %res = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> %gep.random, i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef)
   ret <8 x float>%res
 }
 
 ; Index requires promotion
 define void @test_scatter_2i32_index(<2 x double> %a1, double* %base, <2 x i32> %ind, <2 x i1> %mask) {
 ; KNL_64-LABEL: test_scatter_2i32_index:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $ymm1
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_64-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; KNL_64-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vscatterdpd %zmm0, (%rdi,%ymm1,8) {%k1}
+; KNL_64-NEXT:    vpmovsxdq %xmm1, %xmm1
+; KNL_64-NEXT:    vpsllq $3, %xmm1, %xmm1
+; KNL_64-NEXT:    vmovq %rdi, %xmm2
+; KNL_64-NEXT:    vpbroadcastq %xmm2, %xmm2
+; KNL_64-NEXT:    vpaddq %xmm1, %xmm2, %xmm1
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB52_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB52_3
+; KNL_64-NEXT:  .LBB52_4: # %else2
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB52_1: # %cond.store
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vmovlps %xmm0, (%rcx)
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB52_4
+; KNL_64-NEXT:  .LBB52_3: # %cond.store1
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rax
+; KNL_64-NEXT:    vmovhps %xmm0, (%rax)
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test_scatter_2i32_index:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $ymm1
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; KNL_32-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; KNL_32-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vscatterdpd %zmm0, (%eax,%ymm1,8) {%k1}
+; KNL_32-NEXT:    vpslld $3, %xmm1, %xmm1
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm2
+; KNL_32-NEXT:    vpaddd %xmm1, %xmm2, %xmm1
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB52_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB52_3
+; KNL_32-NEXT:  .LBB52_4: # %else2
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB52_1: # %cond.store
+; KNL_32-NEXT:    vmovd %xmm1, %ecx
+; KNL_32-NEXT:    vmovlps %xmm0, (%ecx)
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB52_4
+; KNL_32-NEXT:  .LBB52_3: # %cond.store1
+; KNL_32-NEXT:    vpextrd $1, %xmm1, %eax
+; KNL_32-NEXT:    vmovhps %xmm0, (%eax)
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test_scatter_2i32_index:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpsllq $63, %xmm2, %xmm2
-; SKX-NEXT:    vpmovq2m %xmm2, %k1
-; SKX-NEXT:    vscatterdpd %xmm0, (%rdi,%xmm1,8) {%k1}
+; SKX-NEXT:    vpmovq2m %xmm2, %k0
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm2
+; SKX-NEXT:    vpmovsxdq %xmm1, %xmm1
+; SKX-NEXT:    vpsllq $3, %xmm1, %xmm1
+; SKX-NEXT:    vpaddq %xmm1, %xmm2, %xmm1
+; SKX-NEXT:    kmovw %k0, %eax
+; SKX-NEXT:    testb $1, %al
+; SKX-NEXT:    jne .LBB52_1
+; SKX-NEXT:  # %bb.2: # %else
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    jne .LBB52_3
+; SKX-NEXT:  .LBB52_4: # %else2
+; SKX-NEXT:    retq
+; SKX-NEXT:  .LBB52_1: # %cond.store
+; SKX-NEXT:    vmovq %xmm1, %rcx
+; SKX-NEXT:    vmovlps %xmm0, (%rcx)
+; SKX-NEXT:    testb $2, %al
+; SKX-NEXT:    je .LBB52_4
+; SKX-NEXT:  .LBB52_3: # %cond.store1
+; SKX-NEXT:    vpextrq $1, %xmm1, %rax
+; SKX-NEXT:    vmovhps %xmm0, (%rax)
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test_scatter_2i32_index:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm2, %xmm2
-; SKX_32-NEXT:    vpmovq2m %xmm2, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vscatterdpd %xmm0, (%eax,%xmm1,8) {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm2, %k0
+; SKX_32-NEXT:    vpslld $3, %xmm1, %xmm1
+; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm1, %xmm1
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB52_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB52_3
+; SKX_32-NEXT:  .LBB52_4: # %else2
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB52_1: # %cond.store
+; SKX_32-NEXT:    vmovd %xmm1, %ecx
+; SKX_32-NEXT:    vmovlps %xmm0, (%ecx)
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB52_4
+; SKX_32-NEXT:  .LBB52_3: # %cond.store1
+; SKX_32-NEXT:    vpextrd $1, %xmm1, %eax
+; SKX_32-NEXT:    vmovhps %xmm0, (%eax)
 ; SKX_32-NEXT:    retl
   %gep = getelementptr double, double *%base, <2 x i32> %ind
   call void @llvm.masked.scatter.v2f64.v2p0f64(<2 x double> %a1, <2 x double*> %gep, i32 4, <2 x i1> %mask)
   ret void
 }
 declare void @llvm.masked.scatter.v2f64.v2p0f64(<2 x double>, <2 x double*>, i32, <2 x i1>)
 
 define <16 x float> @zext_index(float* %base, <16 x i32> %ind) {
; ALL-NEXT:    ret{{[l|q]}}
   ret void
 }
 
 define <2 x i64> @gather_2i64_constant_indices(i64* %ptr, <2 x i1> %mask) {
 ; KNL_64-LABEL: gather_2i64_constant_indices:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    vpsllq $63, %xmm0, %xmm0
 ; KNL_64-NEXT:    vptestmq %zmm0, %zmm0, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vmovdqa {{.*#+}} ymm1 = <0,4294967294,u,u,u,u,u,u>
+; KNL_64-NEXT:    vmovq %rdi, %xmm0
+; KNL_64-NEXT:    vpbroadcastq %xmm0, %xmm0
+; KNL_64-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; KNL_64-NEXT:    kmovw %k0, %eax
 ; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; KNL_64-NEXT:    vpgatherdq (%rdi,%ymm1,8), %zmm0 {%k1}
-; KNL_64-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB58_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB58_3
+; KNL_64-NEXT:  .LBB58_4: # %else2
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB58_1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm1, %rcx
+; KNL_64-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB58_4
+; KNL_64-NEXT:  .LBB58_3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm1, %rax
+; KNL_64-NEXT:    vpinsrq $1, (%rax), %xmm0, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: gather_2i64_constant_indices:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    vpsllq $63, %xmm0, %xmm0
 ; KNL_32-NEXT:    vptestmq %zmm0, %zmm0, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vmovdqa {{.*#+}} ymm1 = <0,4294967294,u,u,u,u,u,u>
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm0
+; KNL_32-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm1
+; KNL_32-NEXT:    kmovw %k0, %eax
 ; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; KNL_32-NEXT:    vpgatherdq (%eax,%ymm1,8), %zmm0 {%k1}
-; KNL_32-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB58_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB58_3
+; KNL_32-NEXT:  .LBB58_4: # %else2
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB58_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm1, %ecx
+; KNL_32-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB58_4
+; KNL_32-NEXT:  .LBB58_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm1, %eax
+; KNL_32-NEXT:    vpinsrd $2, (%eax), %xmm0, %xmm0
+; KNL_32-NEXT:    vpinsrd $3, 4(%eax), %xmm0, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX_SMALL-LABEL: gather_2i64_constant_indices:
 ; SKX_SMALL:       # %bb.0:
 ; SKX_SMALL-NEXT:    vpsllq $63, %xmm0, %xmm0
-; SKX_SMALL-NEXT:    vpmovq2m %xmm0, %k1
-; SKX_SMALL-NEXT:    vmovdqa {{.*#+}} xmm1 = <0,4294967294,u,u>
+; SKX_SMALL-NEXT:    vpmovq2m %xmm0, %k0
+; SKX_SMALL-NEXT:    vpbroadcastq %rdi, %xmm0
+; SKX_SMALL-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; SKX_SMALL-NEXT:    kmovw %k0, %eax
 ; SKX_SMALL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; SKX_SMALL-NEXT:    vpgatherdq (%rdi,%xmm1,8), %xmm0 {%k1}
+; SKX_SMALL-NEXT:    testb $1, %al
+; SKX_SMALL-NEXT:    jne .LBB58_1
+; SKX_SMALL-NEXT:  # %bb.2: # %else
+; SKX_SMALL-NEXT:    testb $2, %al
+; SKX_SMALL-NEXT:    jne .LBB58_3
+; SKX_SMALL-NEXT:  .LBB58_4: # %else2
+; SKX_SMALL-NEXT:    retq
+; SKX_SMALL-NEXT:  .LBB58_1: # %cond.load
+; SKX_SMALL-NEXT:    vmovq %xmm1, %rcx
+; SKX_SMALL-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; SKX_SMALL-NEXT:    testb $2, %al
+; SKX_SMALL-NEXT:    je .LBB58_4
+; SKX_SMALL-NEXT:  .LBB58_3: # %cond.load1
+; SKX_SMALL-NEXT:    vpextrq $1, %xmm1, %rax
+; SKX_SMALL-NEXT:    vpinsrq $1, (%rax), %xmm0, %xmm0
 ; SKX_SMALL-NEXT:    retq
 ;
 ; SKX_LARGE-LABEL: gather_2i64_constant_indices:
 ; SKX_LARGE:       # %bb.0:
 ; SKX_LARGE-NEXT:    vpsllq $63, %xmm0, %xmm0
-; SKX_LARGE-NEXT:    vpmovq2m %xmm0, %k1
+; SKX_LARGE-NEXT:    vpmovq2m %xmm0, %k0
+; SKX_LARGE-NEXT:    vpbroadcastq %rdi, %xmm0
 ; SKX_LARGE-NEXT:    movabsq ${{\.?LCPI[0-9]+_[0-9]+}}, %rax
-; SKX_LARGE-NEXT:    vmovdqa (%rax), %xmm1
+; SKX_LARGE-NEXT:    vpaddq (%rax), %xmm0, %xmm1
+; SKX_LARGE-NEXT:    kmovw %k0, %eax
 ; SKX_LARGE-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; SKX_LARGE-NEXT:    vpgatherdq (%rdi,%xmm1,8), %xmm0 {%k1}
+; SKX_LARGE-NEXT:    testb $1, %al
+; SKX_LARGE-NEXT:    jne .LBB58_1
+; SKX_LARGE-NEXT:  # %bb.2: # %else
+; SKX_LARGE-NEXT:    testb $2, %al
+; SKX_LARGE-NEXT:    jne .LBB58_3
+; SKX_LARGE-NEXT:  .LBB58_4: # %else2
+; SKX_LARGE-NEXT:    retq
+; SKX_LARGE-NEXT:  .LBB58_1: # %cond.load
+; SKX_LARGE-NEXT:    vmovq %xmm1, %rcx
+; SKX_LARGE-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; SKX_LARGE-NEXT:    testb $2, %al
+; SKX_LARGE-NEXT:    je .LBB58_4
+; SKX_LARGE-NEXT:  .LBB58_3: # %cond.load1
+; SKX_LARGE-NEXT:    vpextrq $1, %xmm1, %rax
+; SKX_LARGE-NEXT:    vpinsrq $1, (%rax), %xmm0, %xmm0
 ; SKX_LARGE-NEXT:    retq
 ;
 ; SKX_32-LABEL: gather_2i64_constant_indices:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm0, %xmm0
-; SKX_32-NEXT:    vpmovq2m %xmm0, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vmovdqa {{.*#+}} xmm1 = <0,4294967294,u,u>
+; SKX_32-NEXT:    vpmovq2m %xmm0, %k0
+; SKX_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm0
+; SKX_32-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm1
+; SKX_32-NEXT:    kmovw %k0, %eax
 ; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; SKX_32-NEXT:    vpgatherdq (%eax,%xmm1,8), %xmm0 {%k1}
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB58_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB58_3
+; SKX_32-NEXT:  .LBB58_4: # %else2
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB58_1: # %cond.load
+; SKX_32-NEXT:    vmovd %xmm1, %ecx
+; SKX_32-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB58_4
+; SKX_32-NEXT:  .LBB58_3: # %cond.load1
+; SKX_32-NEXT:    vpextrd $1, %xmm1, %eax
+; SKX_32-NEXT:    vpinsrd $2, (%eax), %xmm0, %xmm0
+; SKX_32-NEXT:    vpinsrd $3, 4(%eax), %xmm0, %xmm0
 ; SKX_32-NEXT:    retl
   %gep = getelementptr i64, i64* %ptr, <2 x i64> <i64 0, i64 -2>
   %res = tail call <2 x i64> @llvm.masked.gather.v2i64.v2p0i64(<2 x i64*> %gep, i32 8, <2 x i1> %mask, <2 x i64> zeroinitializer) #1
   ret <2 x i64> %res
 }
 
 define <16 x i32> @gather_16i64_constant_indices(i32* %ptr, <16 x i1> %mask) {
 ; KNL_64-LABEL: gather_16i64_constant_indices:
; SKX_32-NEXT:    retl
   %gep = getelementptr i32, i32* %ptr, <16 x i64> <i64 0, i64 -2, i64 1, i64 -8, i64 10, i64 20, i64 50, i64 65536, i64 16777215, i64 2147483647, i64 100, i64 -2000, i64 -2147483648, i64 76897723, i64 7, i64 -67897687>
   %res = tail call <16 x i32> @llvm.masked.gather.v16i32.v16p0i32(<16 x i32*> %gep, i32 4, <16 x i1> %mask, <16 x i32> zeroinitializer) #1
   ret <16 x i32> %res
 }
 
 define void @scatter_2i64_constant_indices(i32* %ptr, <2 x i1> %mask, <2 x i32> %src0)  {
 ; KNL_64-LABEL: scatter_2i64_constant_indices:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
 ; KNL_64-NEXT:    vpsllq $63, %xmm0, %xmm0
 ; KNL_64-NEXT:    vptestmq %zmm0, %zmm0, %k0
-; KNL_64-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_64-NEXT:    vmovdqa64 {{.*#+}} zmm0 = <0,4294967294,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; KNL_64-NEXT:    vpscatterdd %zmm1, (%rdi,%zmm0,4) {%k1}
+; KNL_64-NEXT:    vmovq %rdi, %xmm0
+; KNL_64-NEXT:    vpbroadcastq %xmm0, %xmm0
+; KNL_64-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    jne .LBB60_1
+; KNL_64-NEXT:  # %bb.2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    jne .LBB60_3
+; KNL_64-NEXT:  .LBB60_4: # %else2
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB60_1: # %cond.store
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovss %xmm1, (%rcx)
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB60_4
+; KNL_64-NEXT:  .LBB60_3: # %cond.store1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vextractps $1, %xmm1, (%rax)
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: scatter_2i64_constant_indices:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
 ; KNL_32-NEXT:    vpsllq $63, %xmm0, %xmm0
 ; KNL_32-NEXT:    vptestmq %zmm0, %zmm0, %k0
-; KNL_32-NEXT:    kshiftlw $14, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $14, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vmovdqa64 {{.*#+}} zmm0 = <0,4294967294,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; KNL_32-NEXT:    vpscatterdd %zmm1, (%eax,%zmm0,4) {%k1}
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm0
+; KNL_32-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB60_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB60_3
+; KNL_32-NEXT:  .LBB60_4: # %else2
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB60_1: # %cond.store
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vmovss %xmm1, (%ecx)
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB60_4
+; KNL_32-NEXT:  .LBB60_3: # %cond.store1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
+; KNL_32-NEXT:    vextractps $1, %xmm1, (%eax)
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX_SMALL-LABEL: scatter_2i64_constant_indices:
 ; SKX_SMALL:       # %bb.0:
 ; SKX_SMALL-NEXT:    vpsllq $63, %xmm0, %xmm0
-; SKX_SMALL-NEXT:    vpmovq2m %xmm0, %k1
-; SKX_SMALL-NEXT:    vmovdqa {{.*#+}} xmm0 = <0,4294967294,u,u>
-; SKX_SMALL-NEXT:    vpscatterdd %xmm1, (%rdi,%xmm0,4) {%k1}
+; SKX_SMALL-NEXT:    vpmovq2m %xmm0, %k0
+; SKX_SMALL-NEXT:    vpbroadcastq %rdi, %xmm0
+; SKX_SMALL-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; SKX_SMALL-NEXT:    kmovw %k0, %eax
+; SKX_SMALL-NEXT:    testb $1, %al
+; SKX_SMALL-NEXT:    jne .LBB60_1
+; SKX_SMALL-NEXT:  # %bb.2: # %else
+; SKX_SMALL-NEXT:    testb $2, %al
+; SKX_SMALL-NEXT:    jne .LBB60_3
+; SKX_SMALL-NEXT:  .LBB60_4: # %else2
+; SKX_SMALL-NEXT:    retq
+; SKX_SMALL-NEXT:  .LBB60_1: # %cond.store
+; SKX_SMALL-NEXT:    vmovq %xmm0, %rcx
+; SKX_SMALL-NEXT:    vmovss %xmm1, (%rcx)
+; SKX_SMALL-NEXT:    testb $2, %al
+; SKX_SMALL-NEXT:    je .LBB60_4
+; SKX_SMALL-NEXT:  .LBB60_3: # %cond.store1
+; SKX_SMALL-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX_SMALL-NEXT:    vextractps $1, %xmm1, (%rax)
 ; SKX_SMALL-NEXT:    retq
 ;
 ; SKX_LARGE-LABEL: scatter_2i64_constant_indices:
 ; SKX_LARGE:       # %bb.0:
 ; SKX_LARGE-NEXT:    vpsllq $63, %xmm0, %xmm0
-; SKX_LARGE-NEXT:    vpmovq2m %xmm0, %k1
+; SKX_LARGE-NEXT:    vpmovq2m %xmm0, %k0
+; SKX_LARGE-NEXT:    vpbroadcastq %rdi, %xmm0
 ; SKX_LARGE-NEXT:    movabsq ${{\.?LCPI[0-9]+_[0-9]+}}, %rax
-; SKX_LARGE-NEXT:    vmovdqa (%rax), %xmm0
-; SKX_LARGE-NEXT:    vpscatterdd %xmm1, (%rdi,%xmm0,4) {%k1}
+; SKX_LARGE-NEXT:    vpaddq (%rax), %xmm0, %xmm0
+; SKX_LARGE-NEXT:    kmovw %k0, %eax
+; SKX_LARGE-NEXT:    testb $1, %al
+; SKX_LARGE-NEXT:    jne .LBB60_1
+; SKX_LARGE-NEXT:  # %bb.2: # %else
+; SKX_LARGE-NEXT:    testb $2, %al
+; SKX_LARGE-NEXT:    jne .LBB60_3
+; SKX_LARGE-NEXT:  .LBB60_4: # %else2
+; SKX_LARGE-NEXT:    retq
+; SKX_LARGE-NEXT:  .LBB60_1: # %cond.store
+; SKX_LARGE-NEXT:    vmovq %xmm0, %rcx
+; SKX_LARGE-NEXT:    vmovss %xmm1, (%rcx)
+; SKX_LARGE-NEXT:    testb $2, %al
+; SKX_LARGE-NEXT:    je .LBB60_4
+; SKX_LARGE-NEXT:  .LBB60_3: # %cond.store1
+; SKX_LARGE-NEXT:    vpextrq $1, %xmm0, %rax
+; SKX_LARGE-NEXT:    vextractps $1, %xmm1, (%rax)
 ; SKX_LARGE-NEXT:    retq
 ;
 ; SKX_32-LABEL: scatter_2i64_constant_indices:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    vpsllq $63, %xmm0, %xmm0
-; SKX_32-NEXT:    vpmovq2m %xmm0, %k1
-; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; SKX_32-NEXT:    vmovdqa {{.*#+}} xmm0 = <0,4294967294,u,u>
-; SKX_32-NEXT:    vpscatterdd %xmm1, (%eax,%xmm0,4) {%k1}
+; SKX_32-NEXT:    vpmovq2m %xmm0, %k0
+; SKX_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm0
+; SKX_32-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; SKX_32-NEXT:    kmovw %k0, %eax
+; SKX_32-NEXT:    testb $1, %al
+; SKX_32-NEXT:    jne .LBB60_1
+; SKX_32-NEXT:  # %bb.2: # %else
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    jne .LBB60_3
+; SKX_32-NEXT:  .LBB60_4: # %else2
+; SKX_32-NEXT:    retl
+; SKX_32-NEXT:  .LBB60_1: # %cond.store
+; SKX_32-NEXT:    vmovd %xmm0, %ecx
+; SKX_32-NEXT:    vmovss %xmm1, (%ecx)
+; SKX_32-NEXT:    testb $2, %al
+; SKX_32-NEXT:    je .LBB60_4
+; SKX_32-NEXT:  .LBB60_3: # %cond.store1
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
+; SKX_32-NEXT:    vextractps $1, %xmm1, (%eax)
 ; SKX_32-NEXT:    retl
   %gep = getelementptr i32, i32* %ptr, <2 x i64> <i64 0, i64 -2>
   call void @llvm.masked.scatter.v2i32.v2p0i32(<2 x i32> %src0, <2 x i32*> %gep, i32 4, <2 x i1> %mask)
   ret void
 }
 
 define void @scatter_16i64_constant_indices(i32* %ptr, <16 x i1> %mask, <16 x i32> %src0)  {
 ; KNL_64-LABEL: scatter_16i64_constant_indices:
; SKX_32-NEXT:    retl
   %gep = getelementptr i32, i32* %ptr, <16 x i64> <i64 0, i64 -2, i64 1, i64 -8, i64 10, i64 20, i64 50, i64 65536, i64 16777215, i64 2147483647, i64 100, i64 -2000, i64 -2147483648, i64 76897723, i64 7, i64 -67897687>
   call void @llvm.masked.scatter.v16i32.v16p0i32(<16 x i32> %src0, <16 x i32*> %gep, i32 4, <16 x i1> %mask)
   ret void
 }
 
 define <4 x i32> @splat_ptr_gather(i32* %ptr, <4 x i1> %mask, <4 x i32> %passthru) {
 ; KNL_64-LABEL: splat_ptr_gather:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
 ; KNL_64-NEXT:    vpslld $31, %xmm0, %xmm0
 ; KNL_64-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; KNL_64-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; KNL_64-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm1 {%k1}
+; KNL_64-NEXT:    vmovq %rdi, %xmm0
+; KNL_64-NEXT:    vpbroadcastq %xmm0, %ymm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    je .LBB62_2
+; KNL_64-NEXT:  # %bb.1: # %cond.load
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrd $0, (%rcx), %xmm1, %xmm1
+; KNL_64-NEXT:  .LBB62_2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB62_4
+; KNL_64-NEXT:  # %bb.3: # %cond.load1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrd $1, (%rcx), %xmm1, %xmm1
+; KNL_64-NEXT:  .LBB62_4: # %else2
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; KNL_64-NEXT:    jne .LBB62_5
+; KNL_64-NEXT:  # %bb.6: # %else5
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    jne .LBB62_7
+; KNL_64-NEXT:  .LBB62_8: # %else8
+; KNL_64-NEXT:    vmovdqa %xmm1, %xmm0
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB62_5: # %cond.load4
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vpinsrd $2, (%rcx), %xmm1, %xmm1
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB62_8
+; KNL_64-NEXT:  .LBB62_7: # %cond.load7
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vpinsrd $3, (%rax), %xmm1, %xmm1
 ; KNL_64-NEXT:    vmovdqa %xmm1, %xmm0
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: splat_ptr_gather:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
 ; KNL_32-NEXT:    vpslld $31, %xmm0, %xmm0
 ; KNL_32-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; KNL_32-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; KNL_32-NEXT:    vpgatherdd (%eax,%zmm0,4), %zmm1 {%k1}
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB62_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB62_3
+; KNL_32-NEXT:  .LBB62_4: # %else2
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    jne .LBB62_5
+; KNL_32-NEXT:  .LBB62_6: # %else5
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    jne .LBB62_7
+; KNL_32-NEXT:  .LBB62_8: # %else8
+; KNL_32-NEXT:    vmovdqa %xmm1, %xmm0
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB62_1: # %cond.load
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vpinsrd $0, (%ecx), %xmm1, %xmm1
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB62_4
+; KNL_32-NEXT:  .LBB62_3: # %cond.load1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; KNL_32-NEXT:    vpinsrd $1, (%ecx), %xmm1, %xmm1
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    je .LBB62_6
+; KNL_32-NEXT:  .LBB62_5: # %cond.load4
+; KNL_32-NEXT:    vpextrd $2, %xmm0, %ecx
+; KNL_32-NEXT:    vpinsrd $2, (%ecx), %xmm1, %xmm1
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    je .LBB62_8
+; KNL_32-NEXT:  .LBB62_7: # %cond.load7
+; KNL_32-NEXT:    vpextrd $3, %xmm0, %eax
+; KNL_32-NEXT:    vpinsrd $3, (%eax), %xmm1, %xmm1
 ; KNL_32-NEXT:    vmovdqa %xmm1, %xmm0
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: splat_ptr_gather:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm0, %xmm0
 ; SKX-NEXT:    vpmovd2m %xmm0, %k1
; SKX_32-NEXT:    retl
   %3 = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %2, i32 4, <4 x i1> %mask, <4 x i32> %passthru)
   ret <4 x i32> %3
 }
 declare  <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*>, i32, <4 x i1>, <4 x i32>)
 
 define void @splat_ptr_scatter(i32* %ptr, <4 x i1> %mask, <4 x i32> %val) {
 ; KNL_64-LABEL: splat_ptr_scatter:
 ; KNL_64:       # %bb.0:
-; KNL_64-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
 ; KNL_64-NEXT:    vpslld $31, %xmm0, %xmm0
 ; KNL_64-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; KNL_64-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_64-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; KNL_64-NEXT:    vpscatterdd %zmm1, (%rdi,%zmm0,4) {%k1}
+; KNL_64-NEXT:    vmovq %rdi, %xmm0
+; KNL_64-NEXT:    vpbroadcastq %xmm0, %ymm0
+; KNL_64-NEXT:    kmovw %k0, %eax
+; KNL_64-NEXT:    testb $1, %al
+; KNL_64-NEXT:    je .LBB63_2
+; KNL_64-NEXT:  # %bb.1: # %cond.store
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vmovss %xmm1, (%rcx)
+; KNL_64-NEXT:  .LBB63_2: # %else
+; KNL_64-NEXT:    testb $2, %al
+; KNL_64-NEXT:    je .LBB63_4
+; KNL_64-NEXT:  # %bb.3: # %cond.store1
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vextractps $1, %xmm1, (%rcx)
+; KNL_64-NEXT:  .LBB63_4: # %else2
+; KNL_64-NEXT:    testb $4, %al
+; KNL_64-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; KNL_64-NEXT:    jne .LBB63_5
+; KNL_64-NEXT:  # %bb.6: # %else4
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    jne .LBB63_7
+; KNL_64-NEXT:  .LBB63_8: # %else6
+; KNL_64-NEXT:    vzeroupper
+; KNL_64-NEXT:    retq
+; KNL_64-NEXT:  .LBB63_5: # %cond.store3
+; KNL_64-NEXT:    vmovq %xmm0, %rcx
+; KNL_64-NEXT:    vextractps $2, %xmm1, (%rcx)
+; KNL_64-NEXT:    testb $8, %al
+; KNL_64-NEXT:    je .LBB63_8
+; KNL_64-NEXT:  .LBB63_7: # %cond.store5
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
+; KNL_64-NEXT:    vextractps $3, %xmm1, (%rax)
 ; KNL_64-NEXT:    vzeroupper
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: splat_ptr_scatter:
 ; KNL_32:       # %bb.0:
-; KNL_32-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
 ; KNL_32-NEXT:    vpslld $31, %xmm0, %xmm0
 ; KNL_32-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; KNL_32-NEXT:    kshiftlw $12, %k0, %k0
-; KNL_32-NEXT:    kshiftrw $12, %k0, %k1
-; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; KNL_32-NEXT:    vpscatterdd %zmm1, (%eax,%zmm0,4) {%k1}
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm0
+; KNL_32-NEXT:    kmovw %k0, %eax
+; KNL_32-NEXT:    testb $1, %al
+; KNL_32-NEXT:    jne .LBB63_1
+; KNL_32-NEXT:  # %bb.2: # %else
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    jne .LBB63_3
+; KNL_32-NEXT:  .LBB63_4: # %else2
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    jne .LBB63_5
+; KNL_32-NEXT:  .LBB63_6: # %else4
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    jne .LBB63_7
+; KNL_32-NEXT:  .LBB63_8: # %else6
+; KNL_32-NEXT:    vzeroupper
+; KNL_32-NEXT:    retl
+; KNL_32-NEXT:  .LBB63_1: # %cond.store
+; KNL_32-NEXT:    vmovd %xmm0, %ecx
+; KNL_32-NEXT:    vmovss %xmm1, (%ecx)
+; KNL_32-NEXT:    testb $2, %al
+; KNL_32-NEXT:    je .LBB63_4
+; KNL_32-NEXT:  .LBB63_3: # %cond.store1
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; KNL_32-NEXT:    vextractps $1, %xmm1, (%ecx)
+; KNL_32-NEXT:    testb $4, %al
+; KNL_32-NEXT:    je .LBB63_6
+; KNL_32-NEXT:  .LBB63_5: # %cond.store3
+; KNL_32-NEXT:    vpextrd $2, %xmm0, %ecx
+; KNL_32-NEXT:    vextractps $2, %xmm1, (%ecx)
+; KNL_32-NEXT:    testb $8, %al
+; KNL_32-NEXT:    je .LBB63_8
+; KNL_32-NEXT:  .LBB63_7: # %cond.store5
+; KNL_32-NEXT:    vpextrd $3, %xmm0, %eax
+; KNL_32-NEXT:    vextractps $3, %xmm1, (%eax)
 ; KNL_32-NEXT:    vzeroupper
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: splat_ptr_scatter:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm0, %xmm0
 ; SKX-NEXT:    vpmovd2m %xmm0, %k1
 ; SKX-NEXT:    vpxor %xmm0, %xmm0, %xmm0

llvm/test/CodeGen/X86/masked_gather_scatter_widen.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512vl -mattr=+avx512dq < %s | FileCheck %s --check-prefix=WIDEN_SKX
 ; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f < %s | FileCheck %s --check-prefix=WIDEN_KNL
 ; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mcpu=skylake < %s | FileCheck %s --check-prefix=WIDEN_AVX2
 
 define <2 x double> @test_gather_v2i32_index(double* %base, <2 x i32> %ind, <2 x i1> %mask, <2 x double> %src0) {
 ; WIDEN_SKX-LABEL: test_gather_v2i32_index:
 ; WIDEN_SKX:       # %bb.0:
 ; WIDEN_SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; WIDEN_SKX-NEXT:    vpmovq2m %xmm1, %k1
-; WIDEN_SKX-NEXT:    vgatherdpd (%rdi,%xmm0,8), %xmm2 {%k1}
-; WIDEN_SKX-NEXT:    vmovapd %xmm2, %xmm0
+; WIDEN_SKX-NEXT:    vpmovq2m %xmm1, %k0
+; WIDEN_SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; WIDEN_SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; WIDEN_SKX-NEXT:    vpsllq $3, %xmm0, %xmm0
+; WIDEN_SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; WIDEN_SKX-NEXT:    kmovw %k0, %eax
+; WIDEN_SKX-NEXT:    testb $1, %al
+; WIDEN_SKX-NEXT:    jne .LBB0_1
+; WIDEN_SKX-NEXT:  # %bb.2: # %else
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    jne .LBB0_3
+; WIDEN_SKX-NEXT:  .LBB0_4: # %else2
+; WIDEN_SKX-NEXT:    vmovaps %xmm2, %xmm0
+; WIDEN_SKX-NEXT:    retq
+; WIDEN_SKX-NEXT:  .LBB0_1: # %cond.load
+; WIDEN_SKX-NEXT:    vmovq %xmm0, %rcx
+; WIDEN_SKX-NEXT:    vmovlps {{.*#+}} xmm2 = mem[0,1],xmm2[2,3]
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    je .LBB0_4
+; WIDEN_SKX-NEXT:  .LBB0_3: # %cond.load1
+; WIDEN_SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; WIDEN_SKX-NEXT:    vmovhps {{.*#+}} xmm2 = xmm2[0,1],mem[0,1]
+; WIDEN_SKX-NEXT:    vmovaps %xmm2, %xmm0
 ; WIDEN_SKX-NEXT:    retq
 ;
 ; WIDEN_KNL-LABEL: test_gather_v2i32_index:
 ; WIDEN_KNL:       # %bb.0:
-; WIDEN_KNL-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; WIDEN_KNL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; WIDEN_KNL-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; WIDEN_KNL-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; WIDEN_KNL-NEXT:    kshiftlw $14, %k0, %k0
-; WIDEN_KNL-NEXT:    kshiftrw $14, %k0, %k1
-; WIDEN_KNL-NEXT:    vgatherdpd (%rdi,%ymm0,8), %zmm2 {%k1}
-; WIDEN_KNL-NEXT:    vmovapd %xmm2, %xmm0
+; WIDEN_KNL-NEXT:    vpmovsxdq %xmm0, %xmm0
+; WIDEN_KNL-NEXT:    vpsllq $3, %xmm0, %xmm0
+; WIDEN_KNL-NEXT:    vmovq %rdi, %xmm1
+; WIDEN_KNL-NEXT:    vpbroadcastq %xmm1, %xmm1
+; WIDEN_KNL-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; WIDEN_KNL-NEXT:    kmovw %k0, %eax
+; WIDEN_KNL-NEXT:    testb $1, %al
+; WIDEN_KNL-NEXT:    jne .LBB0_1
+; WIDEN_KNL-NEXT:  # %bb.2: # %else
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    jne .LBB0_3
+; WIDEN_KNL-NEXT:  .LBB0_4: # %else2
+; WIDEN_KNL-NEXT:    vmovaps %xmm2, %xmm0
+; WIDEN_KNL-NEXT:    vzeroupper
+; WIDEN_KNL-NEXT:    retq
+; WIDEN_KNL-NEXT:  .LBB0_1: # %cond.load
+; WIDEN_KNL-NEXT:    vmovq %xmm0, %rcx
+; WIDEN_KNL-NEXT:    vmovlps {{.*#+}} xmm2 = mem[0,1],xmm2[2,3]
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    je .LBB0_4
+; WIDEN_KNL-NEXT:  .LBB0_3: # %cond.load1
+; WIDEN_KNL-NEXT:    vpextrq $1, %xmm0, %rax
+; WIDEN_KNL-NEXT:    vmovhps {{.*#+}} xmm2 = xmm2[0,1],mem[0,1]
+; WIDEN_KNL-NEXT:    vmovaps %xmm2, %xmm0
 ; WIDEN_KNL-NEXT:    vzeroupper
 ; WIDEN_KNL-NEXT:    retq
 ;
 ; WIDEN_AVX2-LABEL: test_gather_v2i32_index:
 ; WIDEN_AVX2:       # %bb.0:
 ; WIDEN_AVX2-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vgatherdpd %xmm1, (%rdi,%xmm0,8), %xmm2
 ; WIDEN_AVX2-NEXT:    vmovapd %xmm2, %xmm0
 ; WIDEN_AVX2-NEXT:    retq
   %gep.random = getelementptr double, double* %base, <2 x i32> %ind
   %res = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %gep.random, i32 4, <2 x i1> %mask, <2 x double> %src0)
   ret <2 x double> %res
 }
 
 define void @test_scatter_v2i32_index(<2 x double> %a1, double* %base, <2 x i32> %ind, <2 x i1> %mask) {
 ; WIDEN_SKX-LABEL: test_scatter_v2i32_index:
 ; WIDEN_SKX:       # %bb.0:
 ; WIDEN_SKX-NEXT:    vpsllq $63, %xmm2, %xmm2
-; WIDEN_SKX-NEXT:    vpmovq2m %xmm2, %k1
-; WIDEN_SKX-NEXT:    vscatterdpd %xmm0, (%rdi,%xmm1,8) {%k1}
+; WIDEN_SKX-NEXT:    vpmovq2m %xmm2, %k0
+; WIDEN_SKX-NEXT:    vpbroadcastq %rdi, %xmm2
+; WIDEN_SKX-NEXT:    vpmovsxdq %xmm1, %xmm1
+; WIDEN_SKX-NEXT:    vpsllq $3, %xmm1, %xmm1
+; WIDEN_SKX-NEXT:    vpaddq %xmm1, %xmm2, %xmm1
+; WIDEN_SKX-NEXT:    kmovw %k0, %eax
+; WIDEN_SKX-NEXT:    testb $1, %al
+; WIDEN_SKX-NEXT:    jne .LBB1_1
+; WIDEN_SKX-NEXT:  # %bb.2: # %else
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    jne .LBB1_3
+; WIDEN_SKX-NEXT:  .LBB1_4: # %else2
+; WIDEN_SKX-NEXT:    retq
+; WIDEN_SKX-NEXT:  .LBB1_1: # %cond.store
+; WIDEN_SKX-NEXT:    vmovq %xmm1, %rcx
+; WIDEN_SKX-NEXT:    vmovlps %xmm0, (%rcx)
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    je .LBB1_4
+; WIDEN_SKX-NEXT:  .LBB1_3: # %cond.store1
+; WIDEN_SKX-NEXT:    vpextrq $1, %xmm1, %rax
+; WIDEN_SKX-NEXT:    vmovhps %xmm0, (%rax)
 ; WIDEN_SKX-NEXT:    retq
 ;
 ; WIDEN_KNL-LABEL: test_scatter_v2i32_index:
 ; WIDEN_KNL:       # %bb.0:
-; WIDEN_KNL-NEXT:    # kill: def $xmm1 killed $xmm1 def $ymm1
-; WIDEN_KNL-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; WIDEN_KNL-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; WIDEN_KNL-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; WIDEN_KNL-NEXT:    kshiftlw $14, %k0, %k0
-; WIDEN_KNL-NEXT:    kshiftrw $14, %k0, %k1
-; WIDEN_KNL-NEXT:    vscatterdpd %zmm0, (%rdi,%ymm1,8) {%k1}
+; WIDEN_KNL-NEXT:    vpmovsxdq %xmm1, %xmm1
+; WIDEN_KNL-NEXT:    vpsllq $3, %xmm1, %xmm1
+; WIDEN_KNL-NEXT:    vmovq %rdi, %xmm2
+; WIDEN_KNL-NEXT:    vpbroadcastq %xmm2, %xmm2
+; WIDEN_KNL-NEXT:    vpaddq %xmm1, %xmm2, %xmm1
+; WIDEN_KNL-NEXT:    kmovw %k0, %eax
+; WIDEN_KNL-NEXT:    testb $1, %al
+; WIDEN_KNL-NEXT:    jne .LBB1_1
+; WIDEN_KNL-NEXT:  # %bb.2: # %else
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    jne .LBB1_3
+; WIDEN_KNL-NEXT:  .LBB1_4: # %else2
+; WIDEN_KNL-NEXT:    vzeroupper
+; WIDEN_KNL-NEXT:    retq
+; WIDEN_KNL-NEXT:  .LBB1_1: # %cond.store
+; WIDEN_KNL-NEXT:    vmovq %xmm1, %rcx
+; WIDEN_KNL-NEXT:    vmovlps %xmm0, (%rcx)
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    je .LBB1_4
+; WIDEN_KNL-NEXT:  .LBB1_3: # %cond.store1
+; WIDEN_KNL-NEXT:    vpextrq $1, %xmm1, %rax
+; WIDEN_KNL-NEXT:    vmovhps %xmm0, (%rax)
 ; WIDEN_KNL-NEXT:    vzeroupper
 ; WIDEN_KNL-NEXT:    retq
 ;
 ; WIDEN_AVX2-LABEL: test_scatter_v2i32_index:
 ; WIDEN_AVX2:       # %bb.0:
 ; WIDEN_AVX2-NEXT:    vpmovsxdq %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vpsllq $3, %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vmovq %rdi, %xmm3
; WIDEN_AVX2-NEXT:    retq
   call void @llvm.masked.scatter.v2f64.v2p0f64(<2 x double> %a1, <2 x double*> %gep, i32 4, <2 x i1> %mask)
   ret void
 }
 
 define <2 x i32> @test_gather_v2i32_data(<2 x i32*> %ptr, <2 x i1> %mask, <2 x i32> %src0) {
 ; WIDEN_SKX-LABEL: test_gather_v2i32_data:
 ; WIDEN_SKX:       # %bb.0:
 ; WIDEN_SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; WIDEN_SKX-NEXT:    vpmovq2m %xmm1, %k1
-; WIDEN_SKX-NEXT:    vpgatherqd (,%xmm0), %xmm2 {%k1}
+; WIDEN_SKX-NEXT:    vpmovq2m %xmm1, %k0
+; WIDEN_SKX-NEXT:    kmovw %k0, %eax
+; WIDEN_SKX-NEXT:    testb $1, %al
+; WIDEN_SKX-NEXT:    jne .LBB2_1
+; WIDEN_SKX-NEXT:  # %bb.2: # %else
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    jne .LBB2_3
+; WIDEN_SKX-NEXT:  .LBB2_4: # %else2
+; WIDEN_SKX-NEXT:    vmovdqa %xmm2, %xmm0
+; WIDEN_SKX-NEXT:    retq
+; WIDEN_SKX-NEXT:  .LBB2_1: # %cond.load
+; WIDEN_SKX-NEXT:    vmovq %xmm0, %rcx
+; WIDEN_SKX-NEXT:    vpinsrd $0, (%rcx), %xmm2, %xmm2
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    je .LBB2_4
+; WIDEN_SKX-NEXT:  .LBB2_3: # %cond.load1
+; WIDEN_SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; WIDEN_SKX-NEXT:    vpinsrd $1, (%rax), %xmm2, %xmm2
 ; WIDEN_SKX-NEXT:    vmovdqa %xmm2, %xmm0
 ; WIDEN_SKX-NEXT:    retq
 ;
 ; WIDEN_KNL-LABEL: test_gather_v2i32_data:
 ; WIDEN_KNL:       # %bb.0:
-; WIDEN_KNL-NEXT:    # kill: def $xmm2 killed $xmm2 def $ymm2
-; WIDEN_KNL-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; WIDEN_KNL-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; WIDEN_KNL-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; WIDEN_KNL-NEXT:    kshiftlw $14, %k0, %k0
-; WIDEN_KNL-NEXT:    kshiftrw $14, %k0, %k1
-; WIDEN_KNL-NEXT:    vpgatherqd (,%zmm0), %ymm2 {%k1}
+; WIDEN_KNL-NEXT:    kmovw %k0, %eax
+; WIDEN_KNL-NEXT:    testb $1, %al
+; WIDEN_KNL-NEXT:    jne .LBB2_1
+; WIDEN_KNL-NEXT:  # %bb.2: # %else
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    jne .LBB2_3
+; WIDEN_KNL-NEXT:  .LBB2_4: # %else2
+; WIDEN_KNL-NEXT:    vmovdqa %xmm2, %xmm0
+; WIDEN_KNL-NEXT:    vzeroupper
+; WIDEN_KNL-NEXT:    retq
+; WIDEN_KNL-NEXT:  .LBB2_1: # %cond.load
+; WIDEN_KNL-NEXT:    vmovq %xmm0, %rcx
+; WIDEN_KNL-NEXT:    vpinsrd $0, (%rcx), %xmm2, %xmm2
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    je .LBB2_4
+; WIDEN_KNL-NEXT:  .LBB2_3: # %cond.load1
+; WIDEN_KNL-NEXT:    vpextrq $1, %xmm0, %rax
+; WIDEN_KNL-NEXT:    vpinsrd $1, (%rax), %xmm2, %xmm2
 ; WIDEN_KNL-NEXT:    vmovdqa %xmm2, %xmm0
 ; WIDEN_KNL-NEXT:    vzeroupper
 ; WIDEN_KNL-NEXT:    retq
 ;
 ; WIDEN_AVX2-LABEL: test_gather_v2i32_data:
 ; WIDEN_AVX2:       # %bb.0:
 ; WIDEN_AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; WIDEN_AVX2-NEXT:    vpslld $31, %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vpgatherqd %xmm1, (,%xmm0), %xmm2
 ; WIDEN_AVX2-NEXT:    vmovdqa %xmm2, %xmm0
 ; WIDEN_AVX2-NEXT:    retq
   %res = call <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*> %ptr, i32 4, <2 x i1> %mask, <2 x i32> %src0)
   ret <2 x i32>%res
 }
 
 define void @test_scatter_v2i32_data(<2 x i32>%a1, <2 x i32*> %ptr, <2 x i1>%mask) {
 ; WIDEN_SKX-LABEL: test_scatter_v2i32_data:
 ; WIDEN_SKX:       # %bb.0:
 ; WIDEN_SKX-NEXT:    vpsllq $63, %xmm2, %xmm2
-; WIDEN_SKX-NEXT:    vpmovq2m %xmm2, %k1
-; WIDEN_SKX-NEXT:    vpscatterqd %xmm0, (,%xmm1) {%k1}
+; WIDEN_SKX-NEXT:    vpmovq2m %xmm2, %k0
+; WIDEN_SKX-NEXT:    kmovw %k0, %eax
+; WIDEN_SKX-NEXT:    testb $1, %al
+; WIDEN_SKX-NEXT:    jne .LBB3_1
+; WIDEN_SKX-NEXT:  # %bb.2: # %else
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    jne .LBB3_3
+; WIDEN_SKX-NEXT:  .LBB3_4: # %else2
+; WIDEN_SKX-NEXT:    retq
+; WIDEN_SKX-NEXT:  .LBB3_1: # %cond.store
+; WIDEN_SKX-NEXT:    vmovq %xmm1, %rcx
+; WIDEN_SKX-NEXT:    vmovss %xmm0, (%rcx)
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    je .LBB3_4
+; WIDEN_SKX-NEXT:  .LBB3_3: # %cond.store1
+; WIDEN_SKX-NEXT:    vpextrq $1, %xmm1, %rax
+; WIDEN_SKX-NEXT:    vextractps $1, %xmm0, (%rax)
 ; WIDEN_SKX-NEXT:    retq
 ;
 ; WIDEN_KNL-LABEL: test_scatter_v2i32_data:
 ; WIDEN_KNL:       # %bb.0:
-; WIDEN_KNL-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; WIDEN_KNL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; WIDEN_KNL-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; WIDEN_KNL-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; WIDEN_KNL-NEXT:    kshiftlw $14, %k0, %k0
-; WIDEN_KNL-NEXT:    kshiftrw $14, %k0, %k1
-; WIDEN_KNL-NEXT:    vpscatterqd %ymm0, (,%zmm1) {%k1}
+; WIDEN_KNL-NEXT:    kmovw %k0, %eax
+; WIDEN_KNL-NEXT:    testb $1, %al
+; WIDEN_KNL-NEXT:    jne .LBB3_1
+; WIDEN_KNL-NEXT:  # %bb.2: # %else
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    jne .LBB3_3
+; WIDEN_KNL-NEXT:  .LBB3_4: # %else2
+; WIDEN_KNL-NEXT:    vzeroupper
+; WIDEN_KNL-NEXT:    retq
+; WIDEN_KNL-NEXT:  .LBB3_1: # %cond.store
+; WIDEN_KNL-NEXT:    vmovq %xmm1, %rcx
+; WIDEN_KNL-NEXT:    vmovss %xmm0, (%rcx)
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    je .LBB3_4
+; WIDEN_KNL-NEXT:  .LBB3_3: # %cond.store1
+; WIDEN_KNL-NEXT:    vpextrq $1, %xmm1, %rax
+; WIDEN_KNL-NEXT:    vextractps $1, %xmm0, (%rax)
 ; WIDEN_KNL-NEXT:    vzeroupper
 ; WIDEN_KNL-NEXT:    retq
 ;
 ; WIDEN_AVX2-LABEL: test_scatter_v2i32_data:
 ; WIDEN_AVX2:       # %bb.0:
 ; WIDEN_AVX2-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; WIDEN_AVX2-NEXT:    vmovmskpd %xmm2, %eax
 ; WIDEN_AVX2-NEXT:    testb $1, %al
; WIDEN_AVX2-NEXT:    retq
   call void @llvm.masked.scatter.v2i32.v2p0i32(<2 x i32> %a1, <2 x i32*> %ptr, i32 4, <2 x i1> %mask)
   ret void
 }
 
 define <2 x i32> @test_gather_v2i32_data_index(i32* %base, <2 x i32> %ind, <2 x i1> %mask, <2 x i32> %src0) {
 ; WIDEN_SKX-LABEL: test_gather_v2i32_data_index:
 ; WIDEN_SKX:       # %bb.0:
 ; WIDEN_SKX-NEXT:    vpsllq $63, %xmm1, %xmm1
-; WIDEN_SKX-NEXT:    vpmovq2m %xmm1, %k1
-; WIDEN_SKX-NEXT:    vpgatherdd (%rdi,%xmm0,4), %xmm2 {%k1}
+; WIDEN_SKX-NEXT:    vpmovq2m %xmm1, %k0
+; WIDEN_SKX-NEXT:    vpbroadcastq %rdi, %xmm1
+; WIDEN_SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
+; WIDEN_SKX-NEXT:    vpsllq $2, %xmm0, %xmm0
+; WIDEN_SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; WIDEN_SKX-NEXT:    kmovw %k0, %eax
+; WIDEN_SKX-NEXT:    testb $1, %al
+; WIDEN_SKX-NEXT:    jne .LBB4_1
+; WIDEN_SKX-NEXT:  # %bb.2: # %else
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    jne .LBB4_3
+; WIDEN_SKX-NEXT:  .LBB4_4: # %else2
+; WIDEN_SKX-NEXT:    vmovdqa %xmm2, %xmm0
+; WIDEN_SKX-NEXT:    retq
+; WIDEN_SKX-NEXT:  .LBB4_1: # %cond.load
+; WIDEN_SKX-NEXT:    vmovq %xmm0, %rcx
+; WIDEN_SKX-NEXT:    vpinsrd $0, (%rcx), %xmm2, %xmm2
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    je .LBB4_4
+; WIDEN_SKX-NEXT:  .LBB4_3: # %cond.load1
+; WIDEN_SKX-NEXT:    vpextrq $1, %xmm0, %rax
+; WIDEN_SKX-NEXT:    vpinsrd $1, (%rax), %xmm2, %xmm2
 ; WIDEN_SKX-NEXT:    vmovdqa %xmm2, %xmm0
 ; WIDEN_SKX-NEXT:    retq
 ;
 ; WIDEN_KNL-LABEL: test_gather_v2i32_data_index:
 ; WIDEN_KNL:       # %bb.0:
-; WIDEN_KNL-NEXT:    # kill: def $xmm2 killed $xmm2 def $zmm2
-; WIDEN_KNL-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; WIDEN_KNL-NEXT:    vpsllq $63, %xmm1, %xmm1
 ; WIDEN_KNL-NEXT:    vptestmq %zmm1, %zmm1, %k0
-; WIDEN_KNL-NEXT:    kshiftlw $14, %k0, %k0
-; WIDEN_KNL-NEXT:    kshiftrw $14, %k0, %k1
-; WIDEN_KNL-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm2 {%k1}
+; WIDEN_KNL-NEXT:    vpmovsxdq %xmm0, %xmm0
+; WIDEN_KNL-NEXT:    vpsllq $2, %xmm0, %xmm0
+; WIDEN_KNL-NEXT:    vmovq %rdi, %xmm1
+; WIDEN_KNL-NEXT:    vpbroadcastq %xmm1, %xmm1
+; WIDEN_KNL-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; WIDEN_KNL-NEXT:    kmovw %k0, %eax
+; WIDEN_KNL-NEXT:    testb $1, %al
+; WIDEN_KNL-NEXT:    jne .LBB4_1
+; WIDEN_KNL-NEXT:  # %bb.2: # %else
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    jne .LBB4_3
+; WIDEN_KNL-NEXT:  .LBB4_4: # %else2
+; WIDEN_KNL-NEXT:    vmovdqa %xmm2, %xmm0
+; WIDEN_KNL-NEXT:    vzeroupper
+; WIDEN_KNL-NEXT:    retq
+; WIDEN_KNL-NEXT:  .LBB4_1: # %cond.load
+; WIDEN_KNL-NEXT:    vmovq %xmm0, %rcx
+; WIDEN_KNL-NEXT:    vpinsrd $0, (%rcx), %xmm2, %xmm2
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    je .LBB4_4
+; WIDEN_KNL-NEXT:  .LBB4_3: # %cond.load1
+; WIDEN_KNL-NEXT:    vpextrq $1, %xmm0, %rax
+; WIDEN_KNL-NEXT:    vpinsrd $1, (%rax), %xmm2, %xmm2
 ; WIDEN_KNL-NEXT:    vmovdqa %xmm2, %xmm0
 ; WIDEN_KNL-NEXT:    vzeroupper
 ; WIDEN_KNL-NEXT:    retq
 ;
 ; WIDEN_AVX2-LABEL: test_gather_v2i32_data_index:
 ; WIDEN_AVX2:       # %bb.0:
 ; WIDEN_AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,2],zero,zero
 ; WIDEN_AVX2-NEXT:    vpslld $31, %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vpgatherdd %xmm1, (%rdi,%xmm0,4), %xmm2
 ; WIDEN_AVX2-NEXT:    vmovdqa %xmm2, %xmm0
 ; WIDEN_AVX2-NEXT:    retq
   %gep.random = getelementptr i32, i32* %base, <2 x i32> %ind
   %res = call <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*> %gep.random, i32 4, <2 x i1> %mask, <2 x i32> %src0)
   ret <2 x i32> %res
 }
 
 define void @test_scatter_v2i32_data_index(<2 x i32> %a1, i32* %base, <2 x i32> %ind, <2 x i1> %mask) {
 ; WIDEN_SKX-LABEL: test_scatter_v2i32_data_index:
 ; WIDEN_SKX:       # %bb.0:
 ; WIDEN_SKX-NEXT:    vpsllq $63, %xmm2, %xmm2
-; WIDEN_SKX-NEXT:    vpmovq2m %xmm2, %k1
-; WIDEN_SKX-NEXT:    vpscatterdd %xmm0, (%rdi,%xmm1,4) {%k1}
+; WIDEN_SKX-NEXT:    vpmovq2m %xmm2, %k0
+; WIDEN_SKX-NEXT:    vpbroadcastq %rdi, %xmm2
+; WIDEN_SKX-NEXT:    vpmovsxdq %xmm1, %xmm1
+; WIDEN_SKX-NEXT:    vpsllq $2, %xmm1, %xmm1
+; WIDEN_SKX-NEXT:    vpaddq %xmm1, %xmm2, %xmm1
+; WIDEN_SKX-NEXT:    kmovw %k0, %eax
+; WIDEN_SKX-NEXT:    testb $1, %al
+; WIDEN_SKX-NEXT:    jne .LBB5_1
+; WIDEN_SKX-NEXT:  # %bb.2: # %else
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    jne .LBB5_3
+; WIDEN_SKX-NEXT:  .LBB5_4: # %else2
+; WIDEN_SKX-NEXT:    retq
+; WIDEN_SKX-NEXT:  .LBB5_1: # %cond.store
+; WIDEN_SKX-NEXT:    vmovq %xmm1, %rcx
+; WIDEN_SKX-NEXT:    vmovss %xmm0, (%rcx)
+; WIDEN_SKX-NEXT:    testb $2, %al
+; WIDEN_SKX-NEXT:    je .LBB5_4
+; WIDEN_SKX-NEXT:  .LBB5_3: # %cond.store1
+; WIDEN_SKX-NEXT:    vpextrq $1, %xmm1, %rax
+; WIDEN_SKX-NEXT:    vextractps $1, %xmm0, (%rax)
 ; WIDEN_SKX-NEXT:    retq
 ;
 ; WIDEN_KNL-LABEL: test_scatter_v2i32_data_index:
 ; WIDEN_KNL:       # %bb.0:
-; WIDEN_KNL-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
-; WIDEN_KNL-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; WIDEN_KNL-NEXT:    vpsllq $63, %xmm2, %xmm2
 ; WIDEN_KNL-NEXT:    vptestmq %zmm2, %zmm2, %k0
-; WIDEN_KNL-NEXT:    kshiftlw $14, %k0, %k0
-; WIDEN_KNL-NEXT:    kshiftrw $14, %k0, %k1
-; WIDEN_KNL-NEXT:    vpscatterdd %zmm0, (%rdi,%zmm1,4) {%k1}
+; WIDEN_KNL-NEXT:    vpmovsxdq %xmm1, %xmm1
+; WIDEN_KNL-NEXT:    vpsllq $2, %xmm1, %xmm1
+; WIDEN_KNL-NEXT:    vmovq %rdi, %xmm2
+; WIDEN_KNL-NEXT:    vpbroadcastq %xmm2, %xmm2
+; WIDEN_KNL-NEXT:    vpaddq %xmm1, %xmm2, %xmm1
+; WIDEN_KNL-NEXT:    kmovw %k0, %eax
+; WIDEN_KNL-NEXT:    testb $1, %al
+; WIDEN_KNL-NEXT:    jne .LBB5_1
+; WIDEN_KNL-NEXT:  # %bb.2: # %else
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    jne .LBB5_3
+; WIDEN_KNL-NEXT:  .LBB5_4: # %else2
+; WIDEN_KNL-NEXT:    vzeroupper
+; WIDEN_KNL-NEXT:    retq
+; WIDEN_KNL-NEXT:  .LBB5_1: # %cond.store
+; WIDEN_KNL-NEXT:    vmovq %xmm1, %rcx
+; WIDEN_KNL-NEXT:    vmovss %xmm0, (%rcx)
+; WIDEN_KNL-NEXT:    testb $2, %al
+; WIDEN_KNL-NEXT:    je .LBB5_4
+; WIDEN_KNL-NEXT:  .LBB5_3: # %cond.store1
+; WIDEN_KNL-NEXT:    vpextrq $1, %xmm1, %rax
+; WIDEN_KNL-NEXT:    vextractps $1, %xmm0, (%rax)
 ; WIDEN_KNL-NEXT:    vzeroupper
 ; WIDEN_KNL-NEXT:    retq
 ;
 ; WIDEN_AVX2-LABEL: test_scatter_v2i32_data_index:
 ; WIDEN_AVX2:       # %bb.0:
 ; WIDEN_AVX2-NEXT:    vpmovsxdq %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vpsllq $2, %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vmovq %rdi, %xmm3

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

 ; FVW2-NEXT:    [[FOUND_CONFLICT12:%.*]] = and i1 [[BOUND010]], [[BOUND111]]
 ; FVW2-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT12]]
 ; FVW2-NEXT:    br i1 [[CONFLICT_RDX]], label [[FOR_BODY_PREHEADER]], label [[VECTOR_PH:%.*]]
 ; FVW2:       vector.ph:
 ; FVW2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP3]], 9223372036854775804
 ; FVW2-NEXT:    [[IND_END:%.*]] = getelementptr float, float* [[PTR]], i64 [[N_VEC]]
 ; FVW2-NEXT:    [[TMP12:%.*]] = shl i64 [[N_VEC]], 4
 ; FVW2-NEXT:    [[IND_END14:%.*]] = getelementptr float, float* [[DEST]], i64 [[TMP12]]
-; FVW2-NEXT:    [[TMP13:%.*]] = add nsw i64 [[N_VEC]], -4
-; FVW2-NEXT:    [[TMP14:%.*]] = lshr exact i64 [[TMP13]], 2
-; FVW2-NEXT:    [[TMP15:%.*]] = add nuw nsw i64 [[TMP14]], 1
-; FVW2-NEXT:    [[XTRAITER:%.*]] = and i64 [[TMP15]], 3
-; FVW2-NEXT:    [[TMP16:%.*]] = icmp ult i64 [[TMP13]], 12
-; FVW2-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK_UNR_LCSSA:%.*]], label [[VECTOR_PH_NEW:%.*]]
-; FVW2:       vector.ph.new:
-; FVW2-NEXT:    [[UNROLL_ITER:%.*]] = and i64 [[TMP15]], 9223372036854775804
 ; FVW2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; FVW2:       vector.body:
-; FVW2-NEXT:    [[POINTER_PHI:%.*]] = phi float* [ [[DEST]], [[VECTOR_PH_NEW]] ], [ [[PTR_IND_3:%.*]], [[VECTOR_BODY]] ]
-; FVW2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH_NEW]] ], [ [[INDEX_NEXT_3:%.*]], [[VECTOR_BODY]] ]
-; FVW2-NEXT:    [[NITER:%.*]] = phi i64 [ [[UNROLL_ITER]], [[VECTOR_PH_NEW]] ], [ [[NITER_NSUB_3:%.*]], [[VECTOR_BODY]] ]
+; FVW2-NEXT:    [[POINTER_PHI:%.*]] = phi float* [ [[DEST]], [[VECTOR_PH]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
+; FVW2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FVW2-NEXT:    [[NEXT_GEP:%.*]] = getelementptr float, float* [[PTR]], i64 [[INDEX]]
-; FVW2-NEXT:    [[TMP17:%.*]] = getelementptr float, float* [[POINTER_PHI]], <2 x i64> <i64 0, i64 16>
-; FVW2-NEXT:    [[TMP18:%.*]] = getelementptr float, float* [[POINTER_PHI]], <2 x i64> <i64 32, i64 48>
-; FVW2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, float* [[NEXT_GEP]], i64 [[IDXPROM]]
-; FVW2-NEXT:    [[TMP20:%.*]] = bitcast float* [[TMP19]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP20]], align 4, !alias.scope !7
-; FVW2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 2
-; FVW2-NEXT:    [[TMP22:%.*]] = bitcast float* [[TMP21]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD16:%.*]] = load <2 x float>, <2 x float>* [[TMP22]], align 4, !alias.scope !7
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD]], <2 x float*> [[TMP17]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD16]], <2 x float*> [[TMP18]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[TMP23:%.*]] = bitcast float* [[NEXT_GEP]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD17:%.*]] = load <2 x float>, <2 x float>* [[TMP23]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP24:%.*]] = getelementptr float, float* [[NEXT_GEP]], i64 2
-; FVW2-NEXT:    [[TMP25:%.*]] = bitcast float* [[TMP24]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD18:%.*]] = load <2 x float>, <2 x float>* [[TMP25]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP17]], i64 1
-; FVW2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP18]], i64 1
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD17]], <2 x float*> [[TMP26]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD18]], <2 x float*> [[TMP27]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[INDEX_NEXT:%.*]] = or i64 [[INDEX]], 4
-; FVW2-NEXT:    [[PTR_IND:%.*]] = getelementptr float, float* [[POINTER_PHI]], i64 64
-; FVW2-NEXT:    [[NEXT_GEP_1:%.*]] = getelementptr float, float* [[PTR]], i64 [[INDEX_NEXT]]
-; FVW2-NEXT:    [[TMP28:%.*]] = getelementptr float, float* [[PTR_IND]], <2 x i64> <i64 0, i64 16>
-; FVW2-NEXT:    [[TMP29:%.*]] = getelementptr float, float* [[PTR_IND]], <2 x i64> <i64 32, i64 48>
-; FVW2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, float* [[NEXT_GEP_1]], i64 [[IDXPROM]]
-; FVW2-NEXT:    [[TMP31:%.*]] = bitcast float* [[TMP30]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD_1:%.*]] = load <2 x float>, <2 x float>* [[TMP31]], align 4, !alias.scope !7
-; FVW2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP30]], i64 2
-; FVW2-NEXT:    [[TMP33:%.*]] = bitcast float* [[TMP32]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD16_1:%.*]] = load <2 x float>, <2 x float>* [[TMP33]], align 4, !alias.scope !7
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD_1]], <2 x float*> [[TMP28]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD16_1]], <2 x float*> [[TMP29]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[TMP34:%.*]] = bitcast float* [[NEXT_GEP_1]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD17_1:%.*]] = load <2 x float>, <2 x float>* [[TMP34]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP35:%.*]] = getelementptr float, float* [[NEXT_GEP_1]], i64 2
-; FVW2-NEXT:    [[TMP36:%.*]] = bitcast float* [[TMP35]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD18_1:%.*]] = load <2 x float>, <2 x float>* [[TMP36]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP28]], i64 1
-; FVW2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP29]], i64 1
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD17_1]], <2 x float*> [[TMP37]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD18_1]], <2 x float*> [[TMP38]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[INDEX_NEXT_1:%.*]] = or i64 [[INDEX]], 8
-; FVW2-NEXT:    [[PTR_IND_1:%.*]] = getelementptr float, float* [[POINTER_PHI]], i64 128
-; FVW2-NEXT:    [[NEXT_GEP_2:%.*]] = getelementptr float, float* [[PTR]], i64 [[INDEX_NEXT_1]]
-; FVW2-NEXT:    [[TMP39:%.*]] = getelementptr float, float* [[PTR_IND_1]], <2 x i64> <i64 0, i64 16>
-; FVW2-NEXT:    [[TMP40:%.*]] = getelementptr float, float* [[PTR_IND_1]], <2 x i64> <i64 32, i64 48>
-; FVW2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, float* [[NEXT_GEP_2]], i64 [[IDXPROM]]
-; FVW2-NEXT:    [[TMP42:%.*]] = bitcast float* [[TMP41]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD_2:%.*]] = load <2 x float>, <2 x float>* [[TMP42]], align 4, !alias.scope !7
-; FVW2-NEXT:    [[TMP43:%.*]] = getelementptr inbounds float, float* [[TMP41]], i64 2
-; FVW2-NEXT:    [[TMP44:%.*]] = bitcast float* [[TMP43]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD16_2:%.*]] = load <2 x float>, <2 x float>* [[TMP44]], align 4, !alias.scope !7
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD_2]], <2 x float*> [[TMP39]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD16_2]], <2 x float*> [[TMP40]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[TMP45:%.*]] = bitcast float* [[NEXT_GEP_2]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD17_2:%.*]] = load <2 x float>, <2 x float>* [[TMP45]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP46:%.*]] = getelementptr float, float* [[NEXT_GEP_2]], i64 2
-; FVW2-NEXT:    [[TMP47:%.*]] = bitcast float* [[TMP46]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD18_2:%.*]] = load <2 x float>, <2 x float>* [[TMP47]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP39]], i64 1
-; FVW2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP40]], i64 1
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD17_2]], <2 x float*> [[TMP48]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD18_2]], <2 x float*> [[TMP49]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[INDEX_NEXT_2:%.*]] = or i64 [[INDEX]], 12
-; FVW2-NEXT:    [[PTR_IND_2:%.*]] = getelementptr float, float* [[POINTER_PHI]], i64 192
-; FVW2-NEXT:    [[NEXT_GEP_3:%.*]] = getelementptr float, float* [[PTR]], i64 [[INDEX_NEXT_2]]
-; FVW2-NEXT:    [[TMP50:%.*]] = getelementptr float, float* [[PTR_IND_2]], <2 x i64> <i64 0, i64 16>
-; FVW2-NEXT:    [[TMP51:%.*]] = getelementptr float, float* [[PTR_IND_2]], <2 x i64> <i64 32, i64 48>
-; FVW2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds float, float* [[NEXT_GEP_3]], i64 [[IDXPROM]]
-; FVW2-NEXT:    [[TMP53:%.*]] = bitcast float* [[TMP52]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD_3:%.*]] = load <2 x float>, <2 x float>* [[TMP53]], align 4, !alias.scope !7
-; FVW2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds float, float* [[TMP52]], i64 2
-; FVW2-NEXT:    [[TMP55:%.*]] = bitcast float* [[TMP54]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD16_3:%.*]] = load <2 x float>, <2 x float>* [[TMP55]], align 4, !alias.scope !7
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD_3]], <2 x float*> [[TMP50]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD16_3]], <2 x float*> [[TMP51]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[TMP56:%.*]] = bitcast float* [[NEXT_GEP_3]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD17_3:%.*]] = load <2 x float>, <2 x float>* [[TMP56]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP57:%.*]] = getelementptr float, float* [[NEXT_GEP_3]], i64 2
-; FVW2-NEXT:    [[TMP58:%.*]] = bitcast float* [[TMP57]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD18_3:%.*]] = load <2 x float>, <2 x float>* [[TMP58]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP50]], i64 1
-; FVW2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP51]], i64 1
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD17_3]], <2 x float*> [[TMP59]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD18_3]], <2 x float*> [[TMP60]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[INDEX_NEXT_3]] = add nuw i64 [[INDEX]], 16
-; FVW2-NEXT:    [[PTR_IND_3]] = getelementptr float, float* [[POINTER_PHI]], i64 256
-; FVW2-NEXT:    [[NITER_NSUB_3]] = add i64 [[NITER]], -4
-; FVW2-NEXT:    [[NITER_NCMP_3:%.*]] = icmp eq i64 [[NITER_NSUB_3]], 0
-; FVW2-NEXT:    br i1 [[NITER_NCMP_3]], label [[MIDDLE_BLOCK_UNR_LCSSA]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
-; FVW2:       middle.block.unr-lcssa:
-; FVW2-NEXT:    [[POINTER_PHI_UNR:%.*]] = phi float* [ [[DEST]], [[VECTOR_PH]] ], [ [[PTR_IND_3]], [[VECTOR_BODY]] ]
-; FVW2-NEXT:    [[INDEX_UNR:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT_3]], [[VECTOR_BODY]] ]
-; FVW2-NEXT:    [[LCMP_MOD_NOT:%.*]] = icmp eq i64 [[XTRAITER]], 0
-; FVW2-NEXT:    br i1 [[LCMP_MOD_NOT]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY_EPIL:%.*]]
-; FVW2:       vector.body.epil:
-; FVW2-NEXT:    [[POINTER_PHI_EPIL:%.*]] = phi float* [ [[PTR_IND_EPIL:%.*]], [[VECTOR_BODY_EPIL]] ], [ [[POINTER_PHI_UNR]], [[MIDDLE_BLOCK_UNR_LCSSA]] ]
-; FVW2-NEXT:    [[INDEX_EPIL:%.*]] = phi i64 [ [[INDEX_NEXT_EPIL:%.*]], [[VECTOR_BODY_EPIL]] ], [ [[INDEX_UNR]], [[MIDDLE_BLOCK_UNR_LCSSA]] ]
-; FVW2-NEXT:    [[EPIL_ITER:%.*]] = phi i64 [ [[EPIL_ITER_SUB:%.*]], [[VECTOR_BODY_EPIL]] ], [ [[XTRAITER]], [[MIDDLE_BLOCK_UNR_LCSSA]] ]
-; FVW2-NEXT:    [[NEXT_GEP_EPIL:%.*]] = getelementptr float, float* [[PTR]], i64 [[INDEX_EPIL]]
-; FVW2-NEXT:    [[TMP61:%.*]] = getelementptr float, float* [[POINTER_PHI_EPIL]], <2 x i64> <i64 0, i64 16>
-; FVW2-NEXT:    [[TMP62:%.*]] = getelementptr float, float* [[POINTER_PHI_EPIL]], <2 x i64> <i64 32, i64 48>
-; FVW2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds float, float* [[NEXT_GEP_EPIL]], i64 [[IDXPROM]]
-; FVW2-NEXT:    [[TMP64:%.*]] = bitcast float* [[TMP63]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD_EPIL:%.*]] = load <2 x float>, <2 x float>* [[TMP64]], align 4, !alias.scope !7
-; FVW2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds float, float* [[TMP63]], i64 2
-; FVW2-NEXT:    [[TMP66:%.*]] = bitcast float* [[TMP65]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD16_EPIL:%.*]] = load <2 x float>, <2 x float>* [[TMP66]], align 4, !alias.scope !7
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD_EPIL]], <2 x float*> [[TMP61]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD16_EPIL]], <2 x float*> [[TMP62]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[TMP67:%.*]] = bitcast float* [[NEXT_GEP_EPIL]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD17_EPIL:%.*]] = load <2 x float>, <2 x float>* [[TMP67]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP68:%.*]] = getelementptr float, float* [[NEXT_GEP_EPIL]], i64 2
-; FVW2-NEXT:    [[TMP69:%.*]] = bitcast float* [[TMP68]] to <2 x float>*
-; FVW2-NEXT:    [[WIDE_LOAD18_EPIL:%.*]] = load <2 x float>, <2 x float>* [[TMP69]], align 4, !alias.scope !14
-; FVW2-NEXT:    [[TMP70:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP61]], i64 1
-; FVW2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP62]], i64 1
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD17_EPIL]], <2 x float*> [[TMP70]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD18_EPIL]], <2 x float*> [[TMP71]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
-; FVW2-NEXT:    [[INDEX_NEXT_EPIL]] = add nuw i64 [[INDEX_EPIL]], 4
-; FVW2-NEXT:    [[PTR_IND_EPIL]] = getelementptr float, float* [[POINTER_PHI_EPIL]], i64 64
-; FVW2-NEXT:    [[EPIL_ITER_SUB]] = add i64 [[EPIL_ITER]], -1
-; FVW2-NEXT:    [[EPIL_ITER_CMP_NOT:%.*]] = icmp eq i64 [[EPIL_ITER_SUB]], 0
-; FVW2-NEXT:    br i1 [[EPIL_ITER_CMP_NOT]], label [[MIDDLE_BLOCK]], label [[VECTOR_BODY_EPIL]], !llvm.loop [[LOOP16:![0-9]+]]
+; FVW2-NEXT:    [[TMP13:%.*]] = getelementptr float, float* [[POINTER_PHI]], <2 x i64> <i64 0, i64 16>
+; FVW2-NEXT:    [[TMP14:%.*]] = getelementptr float, float* [[POINTER_PHI]], <2 x i64> <i64 32, i64 48>
+; FVW2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, float* [[NEXT_GEP]], i64 [[IDXPROM]]
+; FVW2-NEXT:    [[TMP16:%.*]] = bitcast float* [[TMP15]] to <2 x float>*
+; FVW2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP16]], align 4, !alias.scope !7
+; FVW2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, float* [[TMP15]], i64 2
+; FVW2-NEXT:    [[TMP18:%.*]] = bitcast float* [[TMP17]] to <2 x float>*
+; FVW2-NEXT:    [[WIDE_LOAD16:%.*]] = load <2 x float>, <2 x float>* [[TMP18]], align 4, !alias.scope !7
+; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD]], <2 x float*> [[TMP13]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
+; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD16]], <2 x float*> [[TMP14]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
+; FVW2-NEXT:    [[TMP19:%.*]] = bitcast float* [[NEXT_GEP]] to <2 x float>*
+; FVW2-NEXT:    [[WIDE_LOAD17:%.*]] = load <2 x float>, <2 x float>* [[TMP19]], align 4, !alias.scope !14
+; FVW2-NEXT:    [[TMP20:%.*]] = getelementptr float, float* [[NEXT_GEP]], i64 2
+; FVW2-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP20]] to <2 x float>*
+; FVW2-NEXT:    [[WIDE_LOAD18:%.*]] = load <2 x float>, <2 x float>* [[TMP21]], align 4, !alias.scope !14
+; FVW2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP13]], i64 1
+; FVW2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, <2 x float*> [[TMP14]], i64 1
+; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD17]], <2 x float*> [[TMP22]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
+; FVW2-NEXT:    call void @llvm.masked.scatter.v2f32.v2p0f32(<2 x float> [[WIDE_LOAD18]], <2 x float*> [[TMP23]], i32 4, <2 x i1> <i1 true, i1 true>), !alias.scope !10, !noalias !12
+; FVW2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; FVW2-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; FVW2-NEXT:    [[PTR_IND]] = getelementptr float, float* [[POINTER_PHI]], i64 64
+; FVW2-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; FVW2:       middle.block:
 ; FVW2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
 ; FVW2-NEXT:    br i1 [[CMP_N]], label [[FOR_END]], label [[FOR_BODY_PREHEADER]]
 ; FVW2:       for.body.preheader:
 ; FVW2-NEXT:    [[PTR_ADDR_012_PH:%.*]] = phi float* [ [[PTR]], [[VECTOR_MEMCHECK]] ], [ [[PTR]], [[FOR_BODY_LR_PH]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
 ; FVW2-NEXT:    [[DEST_ADDR_011_PH:%.*]] = phi float* [ [[DEST]], [[VECTOR_MEMCHECK]] ], [ [[DEST]], [[FOR_BODY_LR_PH]] ], [ [[IND_END14]], [[MIDDLE_BLOCK]] ]
 ; FVW2-NEXT:    br label [[FOR_BODY:%.*]]
 ; FVW2:       for.body:
 ; FVW2-NEXT:    [[PTR_ADDR_012:%.*]] = phi float* [ [[INCDEC_PTR:%.*]], [[FOR_BODY]] ], [ [[PTR_ADDR_012_PH]], [[FOR_BODY_PREHEADER]] ]
 ; FVW2-NEXT:    [[DEST_ADDR_011:%.*]] = phi float* [ [[ADD_PTR6:%.*]], [[FOR_BODY]] ], [ [[DEST_ADDR_011_PH]], [[FOR_BODY_PREHEADER]] ]
 ; FVW2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[PTR_ADDR_012]], i64 [[IDXPROM]]
-; FVW2-NEXT:    [[TMP72:%.*]] = load float, float* [[ARRAYIDX]], align 4
-; FVW2-NEXT:    store float [[TMP72]], float* [[DEST_ADDR_011]], align 4
-; FVW2-NEXT:    [[TMP73:%.*]] = load float, float* [[PTR_ADDR_012]], align 4
+; FVW2-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX]], align 4
+; FVW2-NEXT:    store float [[TMP25]], float* [[DEST_ADDR_011]], align 4
+; FVW2-NEXT:    [[TMP26:%.*]] = load float, float* [[PTR_ADDR_012]], align 4
 ; FVW2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, float* [[DEST_ADDR_011]], i64 1
-; FVW2-NEXT:    store float [[TMP73]], float* [[ARRAYIDX5]], align 4
+; FVW2-NEXT:    store float [[TMP26]], float* [[ARRAYIDX5]], align 4
 ; FVW2-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds float, float* [[PTR_ADDR_012]], i64 1
 ; FVW2-NEXT:    [[ADD_PTR6]] = getelementptr inbounds float, float* [[DEST_ADDR_011]], i64 16
 ; FVW2-NEXT:    [[CMP_NOT:%.*]] = icmp eq float* [[INCDEC_PTR]], [[ADD_PTR]]
-; FVW2-NEXT:    br i1 [[CMP_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; FVW2-NEXT:    br i1 [[CMP_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; FVW2:       for.end:
 ; FVW2-NEXT:    ret void
 ;
 entry:
   %idx.ext = sext i32 %d to i64
   %add.ptr = getelementptr inbounds float, float* %ptr, i64 %idx.ext
   %cmp.not10 = icmp eq i32 %d, 0
   br i1 %cmp.not10, label %for.end, label %for.body.lr.ph

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

 ;         S[0]                 S[1]
 ;
 ; SLP should reorder the operands of the RHS add taking into consideration the cost of external uses.
 ; It is more profitable to reorder the operands of the RHS add, because A[1] has an external use.
 
 define void @lookahead_external_uses(double* %A, double *%B, double *%C, double *%D, double *%S, double *%Ext1, double *%Ext2) {
 ; CHECK-LABEL: @lookahead_external_uses(
 ; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[IDXA0:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 0
 ; CHECK-NEXT:    [[IDXB0:%.*]] = getelementptr inbounds double, double* [[B:%.*]], i64 0
 ; CHECK-NEXT:    [[IDXC0:%.*]] = getelementptr inbounds double, double* [[C:%.*]], i64 0
 ; CHECK-NEXT:    [[IDXD0:%.*]] = getelementptr inbounds double, double* [[D:%.*]], i64 0
-; CHECK-NEXT:    [[IDXA1:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 1
+; CHECK-NEXT:    [[IDXA1:%.*]] = getelementptr inbounds double, double* [[A]], i64 1
 ; CHECK-NEXT:    [[IDXB2:%.*]] = getelementptr inbounds double, double* [[B]], i64 2
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x double*> poison, double* [[A]], i32 0
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double*> [[TMP0]], double* [[A]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr double, <2 x double*> [[TMP1]], <2 x i64> <i64 0, i64 2>
+; CHECK-NEXT:    [[IDXA2:%.*]] = getelementptr inbounds double, double* [[A]], i64 2
 ; CHECK-NEXT:    [[IDXB1:%.*]] = getelementptr inbounds double, double* [[B]], i64 1
+; CHECK-NEXT:    [[A0:%.*]] = load double, double* [[IDXA0]], align 8
 ; CHECK-NEXT:    [[C0:%.*]] = load double, double* [[IDXC0]], align 8
 ; CHECK-NEXT:    [[D0:%.*]] = load double, double* [[IDXD0]], align 8
 ; CHECK-NEXT:    [[A1:%.*]] = load double, double* [[IDXA1]], align 8
 ; CHECK-NEXT:    [[B2:%.*]] = load double, double* [[IDXB2]], align 8
-; CHECK-NEXT:    [[TMP3:%.*]] = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> [[TMP2]], i32 8, <2 x i1> <i1 true, i1 true>, <2 x double> undef)
-; CHECK-NEXT:    [[TMP4:%.*]] = bitcast double* [[IDXB0]] to <2 x double>*
-; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[TMP4]], align 8
-; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x double> poison, double [[C0]], i32 0
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x double> [[TMP6]], double [[A1]], i32 1
-; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x double> poison, double [[D0]], i32 0
-; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x double> [[TMP8]], double [[B2]], i32 1
-; CHECK-NEXT:    [[TMP10:%.*]] = fsub fast <2 x double> [[TMP7]], [[TMP9]]
-; CHECK-NEXT:    [[TMP11:%.*]] = fsub fast <2 x double> [[TMP3]], [[TMP5]]
-; CHECK-NEXT:    [[TMP12:%.*]] = fadd fast <2 x double> [[TMP11]], [[TMP10]]
+; CHECK-NEXT:    [[A2:%.*]] = load double, double* [[IDXA2]], align 8
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[IDXB0]] to <2 x double>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> poison, double [[C0]], i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> [[TMP2]], double [[A1]], i32 1
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x double> poison, double [[D0]], i32 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x double> [[TMP4]], double [[B2]], i32 1
+; CHECK-NEXT:    [[TMP6:%.*]] = fsub fast <2 x double> [[TMP3]], [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x double> poison, double [[A0]], i32 0
+; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x double> [[TMP7]], double [[A2]], i32 1
+; CHECK-NEXT:    [[TMP9:%.*]] = fsub fast <2 x double> [[TMP8]], [[TMP1]]
+; CHECK-NEXT:    [[TMP10:%.*]] = fadd fast <2 x double> [[TMP9]], [[TMP6]]
 ; CHECK-NEXT:    [[IDXS0:%.*]] = getelementptr inbounds double, double* [[S:%.*]], i64 0
 ; CHECK-NEXT:    [[IDXS1:%.*]] = getelementptr inbounds double, double* [[S]], i64 1
-; CHECK-NEXT:    [[TMP13:%.*]] = bitcast double* [[IDXS0]] to <2 x double>*
-; CHECK-NEXT:    store <2 x double> [[TMP12]], <2 x double>* [[TMP13]], align 8
+; CHECK-NEXT:    [[TMP11:%.*]] = bitcast double* [[IDXS0]] to <2 x double>*
+; CHECK-NEXT:    store <2 x double> [[TMP10]], <2 x double>* [[TMP11]], align 8
 ; CHECK-NEXT:    store double [[A1]], double* [[EXT1:%.*]], align 8
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %IdxA0 = getelementptr inbounds double, double* %A, i64 0
   %IdxB0 = getelementptr inbounds double, double* %B, i64 0
   %IdxC0 = getelementptr inbounds double, double* %C, i64 0
   %IdxD0 = getelementptr inbounds double, double* %D, i64 0

 ; over A[1] (with 3 external users).
 ; The result is that the operands are of the Add not reordered and the loads
 ; from A get vectorized instead of the loads from B.
 ;
 define void @lookahead_limit_users_budget(double* %A, double *%B, double *%C, double *%D, double *%S, double *%Ext1, double *%Ext2, double *%Ext3, double *%Ext4, double *%Ext5) {
 ; CHECK-LABEL: @lookahead_limit_users_budget(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[IDXA0:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 0
+; CHECK-NEXT:    [[IDXB0:%.*]] = getelementptr inbounds double, double* [[B:%.*]], i64 0
 ; CHECK-NEXT:    [[IDXC0:%.*]] = getelementptr inbounds double, double* [[C:%.*]], i64 0
 ; CHECK-NEXT:    [[IDXD0:%.*]] = getelementptr inbounds double, double* [[D:%.*]], i64 0
 ; CHECK-NEXT:    [[IDXA1:%.*]] = getelementptr inbounds double, double* [[A]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x double*> poison, double* [[B:%.*]], i32 0
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x double*> [[TMP0]], double* [[B]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr double, <2 x double*> [[TMP1]], <2 x i64> <i64 0, i64 2>
+; CHECK-NEXT:    [[IDXB2:%.*]] = getelementptr inbounds double, double* [[B]], i64 2
 ; CHECK-NEXT:    [[IDXA2:%.*]] = getelementptr inbounds double, double* [[A]], i64 2
 ; CHECK-NEXT:    [[IDXB1:%.*]] = getelementptr inbounds double, double* [[B]], i64 1
+; CHECK-NEXT:    [[B0:%.*]] = load double, double* [[IDXB0]], align 8
 ; CHECK-NEXT:    [[C0:%.*]] = load double, double* [[IDXC0]], align 8
 ; CHECK-NEXT:    [[D0:%.*]] = load double, double* [[IDXD0]], align 8
-; CHECK-NEXT:    [[TMP3:%.*]] = bitcast double* [[IDXA0]] to <2 x double>*
-; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x double>, <2 x double>* [[TMP3]], align 8
-; CHECK-NEXT:    [[TMP5:%.*]] = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> [[TMP2]], i32 8, <2 x i1> <i1 true, i1 true>, <2 x double> undef)
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[IDXA0]] to <2 x double>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT:    [[B2:%.*]] = load double, double* [[IDXB2]], align 8
 ; CHECK-NEXT:    [[A2:%.*]] = load double, double* [[IDXA2]], align 8
 ; CHECK-NEXT:    [[B1:%.*]] = load double, double* [[IDXB1]], align 8
-; CHECK-NEXT:    [[TMP6:%.*]] = fsub fast <2 x double> [[TMP4]], [[TMP5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x double> poison, double [[C0]], i32 0
-; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x double> [[TMP7]], double [[A2]], i32 1
-; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x double> poison, double [[D0]], i32 0
-; CHECK-NEXT:    [[TMP10:%.*]] = insertelement <2 x double> [[TMP9]], double [[B1]], i32 1
-; CHECK-NEXT:    [[TMP11:%.*]] = fsub fast <2 x double> [[TMP8]], [[TMP10]]
-; CHECK-NEXT:    [[TMP12:%.*]] = fadd fast <2 x double> [[TMP6]], [[TMP11]]
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x double> poison, double [[B0]], i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> [[TMP2]], double [[B2]], i32 1
+; CHECK-NEXT:    [[TMP4:%.*]] = fsub fast <2 x double> [[TMP1]], [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x double> poison, double [[C0]], i32 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x double> [[TMP5]], double [[A2]], i32 1
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x double> poison, double [[D0]], i32 0
+; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x double> [[TMP7]], double [[B1]], i32 1
+; CHECK-NEXT:    [[TMP9:%.*]] = fsub fast <2 x double> [[TMP6]], [[TMP8]]
+; CHECK-NEXT:    [[TMP10:%.*]] = fadd fast <2 x double> [[TMP4]], [[TMP9]]
 ; CHECK-NEXT:    [[IDXS0:%.*]] = getelementptr inbounds double, double* [[S:%.*]], i64 0
 ; CHECK-NEXT:    [[IDXS1:%.*]] = getelementptr inbounds double, double* [[S]], i64 1
-; CHECK-NEXT:    [[TMP13:%.*]] = bitcast double* [[IDXS0]] to <2 x double>*
-; CHECK-NEXT:    store <2 x double> [[TMP12]], <2 x double>* [[TMP13]], align 8
-; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
-; CHECK-NEXT:    store double [[TMP14]], double* [[EXT1:%.*]], align 8
-; CHECK-NEXT:    store double [[TMP14]], double* [[EXT2:%.*]], align 8
-; CHECK-NEXT:    store double [[TMP14]], double* [[EXT3:%.*]], align 8
+; CHECK-NEXT:    [[TMP11:%.*]] = bitcast double* [[IDXS0]] to <2 x double>*
+; CHECK-NEXT:    store <2 x double> [[TMP10]], <2 x double>* [[TMP11]], align 8
+; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
+; CHECK-NEXT:    store double [[TMP12]], double* [[EXT1:%.*]], align 8
+; CHECK-NEXT:    store double [[TMP12]], double* [[EXT2:%.*]], align 8
+; CHECK-NEXT:    store double [[TMP12]], double* [[EXT3:%.*]], align 8
 ; CHECK-NEXT:    store double [[B1]], double* [[EXT4:%.*]], align 8
 ; CHECK-NEXT:    store double [[B1]], double* [[EXT5:%.*]], align 8
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %IdxA0 = getelementptr inbounds double, double* %A, i64 0
   %IdxB0 = getelementptr inbounds double, double* %B, i64 0
   %IdxC0 = getelementptr inbounds double, double* %C, i64 0

llvm/test/Transforms/SLPVectorizer/X86/memory-runtime-checks.ll

entry:
   store i32 %xor.3, i32* %arrayidx2.3, align 4
   ret void
 }
 
 declare void @use(<8 x float>)
 define void @delete_pointer_bound(float* %a, float* %b, i1 %c) #0 {
 ; CHECK-LABEL: @delete_pointer_bound(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x float*> poison, float* [[B:%.*]], i32 0
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x float*> [[TMP0]], float* [[B]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr float, <2 x float*> [[TMP1]], <2 x i64> <i64 10, i64 14>
+; CHECK-NEXT:    [[B_10:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 10
+; CHECK-NEXT:    [[B_14:%.*]] = getelementptr inbounds float, float* [[B]], i64 14
 ; CHECK-NEXT:    br i1 [[C:%.*]], label [[THEN:%.*]], label [[ELSE:%.*]]
 ; CHECK:       else:
-; CHECK-NEXT:    [[TMP3:%.*]] = call <2 x float> @llvm.masked.gather.v2f32.v2p0f32(<2 x float*> [[TMP2]], i32 4, <2 x i1> <i1 true, i1 true>, <2 x float> undef)
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> poison, <4 x i32> <i32 0, i32 0, i32 1, i32 1>
-; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x float> [[SHUFFLE]], <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 3, i32 undef, i32 undef>
-; CHECK-NEXT:    [[I71:%.*]] = shufflevector <8 x float> undef, <8 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 10, i32 5, i32 6, i32 13>
-; CHECK-NEXT:    call void @use(<8 x float> [[I71]])
+; CHECK-NEXT:    [[L0:%.*]] = load float, float* [[B_10]], align 4
+; CHECK-NEXT:    [[L1:%.*]] = load float, float* [[B_14]], align 4
+; CHECK-NEXT:    [[I2:%.*]] = insertelement <8 x float> undef, float [[L0]], i32 2
+; CHECK-NEXT:    [[I3:%.*]] = insertelement <8 x float> [[I2]], float [[L0]], i32 3
+; CHECK-NEXT:    [[I4:%.*]] = insertelement <8 x float> [[I3]], float [[L1]], i32 4
+; CHECK-NEXT:    [[I7:%.*]] = insertelement <8 x float> [[I4]], float [[L1]], i32 7
+; CHECK-NEXT:    call void @use(<8 x float> [[I7]])
 ; CHECK-NEXT:    ret void
 ; CHECK:       then:
 ; CHECK-NEXT:    [[A_8:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 8
 ; CHECK-NEXT:    store float 0.000000e+00, float* [[A_8]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x float*> [[TMP2]], i32 1
-; CHECK-NEXT:    [[L6:%.*]] = load float, float* [[TMP5]], align 4
+; CHECK-NEXT:    [[L6:%.*]] = load float, float* [[B_14]], align 4
 ; CHECK-NEXT:    [[A_5:%.*]] = getelementptr inbounds float, float* [[A]], i64 5
 ; CHECK-NEXT:    store float [[L6]], float* [[A_5]], align 4
 ; CHECK-NEXT:    [[A_6:%.*]] = getelementptr inbounds float, float* [[A]], i64 6
 ; CHECK-NEXT:    store float 0.000000e+00, float* [[A_6]], align 4
 ; CHECK-NEXT:    [[A_7:%.*]] = getelementptr inbounds float, float* [[A]], i64 7
 ; CHECK-NEXT:    store float 0.000000e+00, float* [[A_7]], align 4
 ; CHECK-NEXT:    ret void
 ;

llvm/test/Transforms/SLPVectorizer/X86/pr47623.ll

 ; SSE-NEXT:    store i32 [[TMP2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 3), align 4
 ; SSE-NEXT:    store i32 [[TMP1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 4), align 16
 ; SSE-NEXT:    store i32 [[TMP2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 5), align 4
 ; SSE-NEXT:    store i32 [[TMP1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 6), align 8
 ; SSE-NEXT:    store i32 [[TMP2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 7), align 4
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @foo(
-; AVX-NEXT:    [[TMP1:%.*]] = call <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*> <i32* getelementptr inbounds ([8 x i32], [8 x i32]* @b, i64 0, i64 0), i32* getelementptr inbounds ([8 x i32], [8 x i32]* @b, i64 0, i64 2)>, i32 8, <2 x i1> <i1 true, i1 true>, <2 x i32> undef)
-; AVX-NEXT:    [[SHUFFLE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> poison, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>
+; AVX-NEXT:    [[TMP1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @b, i64 0, i64 0), align 16
+; AVX-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @b, i64 0, i64 2), align 8
+; AVX-NEXT:    [[TMP3:%.*]] = insertelement <8 x i32> poison, i32 [[TMP1]], i32 0
+; AVX-NEXT:    [[TMP4:%.*]] = insertelement <8 x i32> [[TMP3]], i32 [[TMP2]], i32 1
+; AVX-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x i32> [[TMP4]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>
 ; AVX-NEXT:    store <8 x i32> [[SHUFFLE]], <8 x i32>* bitcast ([8 x i32]* @a to <8 x i32>*), align 16
 ; AVX-NEXT:    ret void
 ;
   %1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @b, i64 0, i64 0), align 16
   store i32 %1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 0), align 16
   %2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @b, i64 0, i64 2), align 8
   store i32 %2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 1), align 4
   store i32 %1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 2), align 8
   store i32 %2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 3), align 4
   store i32 %1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 4), align 16
   store i32 %2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 5), align 4
   store i32 %1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 6), align 8
   store i32 %2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @a, i64 0, i64 7), align 4
   ret void
 }

llvm/test/Transforms/SLPVectorizer/X86/pr47629-inseltpoison.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+sse2     | FileCheck %s --check-prefixes=CHECK,SSE
 ; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx      | FileCheck %s --check-prefixes=CHECK,AVX
 ; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx2     | FileCheck %s --check-prefixes=CHECK,AVX2
-; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx512f  | FileCheck %s --check-prefixes=CHECK,AVX512
-; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512
+; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx512f  | FileCheck %s --check-prefixes=CHECK,AVX512F
+; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512VL
 
 define void @gather_load(i32* noalias nocapture %0, i32* noalias nocapture readonly %1) {
 ; CHECK-LABEL: @gather_load(
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP1]], align 4, !tbaa [[TBAA0:![0-9]+]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
 ; CHECK-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4

 ; SSE-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    [[TMP17:%.*]] = add nsw i32 [[TMP16]], 4
 ; SSE-NEXT:    store i32 [[TMP17]], i32* [[TMP14]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @gather_load_2(
 ; AVX-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
 ; AVX-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP5:%.*]] = add nsw i32 [[TMP4]], 1
-; AVX-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
-; AVX-NEXT:    store i32 [[TMP5]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 10
+; AVX-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 10
+; AVX-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
 ; AVX-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP9:%.*]] = add nsw i32 [[TMP8]], 2
-; AVX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 2
-; AVX-NEXT:    store i32 [[TMP9]], i32* [[TMP6]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
-; AVX-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP13:%.*]] = add nsw i32 [[TMP12]], 3
-; AVX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 3
-; AVX-NEXT:    store i32 [[TMP13]], i32* [[TMP10]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 5
-; AVX-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP17:%.*]] = add nsw i32 [[TMP16]], 4
-; AVX-NEXT:    store i32 [[TMP17]], i32* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 5
+; AVX-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i32 0
+; AVX-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP6]], i32 1
+; AVX-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP8]], i32 2
+; AVX-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP10]], i32 3
+; AVX-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 1, i32 2, i32 3, i32 4>
+; AVX-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
+; AVX-NEXT:    store <4 x i32> [[TMP15]], <4 x i32>* [[TMP16]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    ret void
 ;
 ; AVX2-LABEL: @gather_load_2(
-; AVX2-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1:%.*]], i32 0
-; AVX2-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32*> [[TMP3]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr i32, <4 x i32*> [[TMP4]], <4 x i64> <i64 1, i64 10, i64 3, i64 5>
-; AVX2-NEXT:    [[TMP6:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP5]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 1, i32 2, i32 3, i32 4>
-; AVX2-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
-; AVX2-NEXT:    store <4 x i32> [[TMP7]], <4 x i32>* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
+; AVX2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 10
+; AVX2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
+; AVX2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 5
+; AVX2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i32 0
+; AVX2-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP6]], i32 1
+; AVX2-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP8]], i32 2
+; AVX2-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP10]], i32 3
+; AVX2-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 1, i32 2, i32 3, i32 4>
+; AVX2-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
+; AVX2-NEXT:    store <4 x i32> [[TMP15]], <4 x i32>* [[TMP16]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    ret void
 ;
-; AVX512-LABEL: @gather_load_2(
-; AVX512-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1:%.*]], i32 0
-; AVX512-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32*> [[TMP3]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr i32, <4 x i32*> [[TMP4]], <4 x i64> <i64 1, i64 10, i64 3, i64 5>
-; AVX512-NEXT:    [[TMP6:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP5]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 1, i32 2, i32 3, i32 4>
-; AVX512-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
-; AVX512-NEXT:    store <4 x i32> [[TMP7]], <4 x i32>* [[TMP8]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    ret void
+; AVX512F-LABEL: @gather_load_2(
+; AVX512F-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
+; AVX512F-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 10
+; AVX512F-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
+; AVX512F-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 5
+; AVX512F-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i32 0
+; AVX512F-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP6]], i32 1
+; AVX512F-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP8]], i32 2
+; AVX512F-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP10]], i32 3
+; AVX512F-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP15]], <4 x i32>* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    ret void
+;
+; AVX512VL-LABEL: @gather_load_2(
+; AVX512VL-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1:%.*]], i32 0
+; AVX512VL-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32*> [[TMP3]], <4 x i32*> poison, <4 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP5:%.*]] = getelementptr i32, <4 x i32*> [[TMP4]], <4 x i64> <i64 1, i64 10, i64 3, i64 5>
+; AVX512VL-NEXT:    [[TMP6:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP5]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512VL-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
+; AVX512VL-NEXT:    store <4 x i32> [[TMP7]], <4 x i32>* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    ret void
 ;
   %3 = getelementptr inbounds i32, i32* %1, i64 1
   %4 = load i32, i32* %3, align 4, !tbaa !2
   %5 = add nsw i32 %4, 1
   %6 = getelementptr inbounds i32, i32* %0, i64 1
   store i32 %5, i32* %0, align 4, !tbaa !2
   %7 = getelementptr inbounds i32, i32* %1, i64 10
   %8 = load i32, i32* %7, align 4, !tbaa !2

 ; SSE-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
 ; SSE-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    [[TMP32:%.*]] = add i32 [[TMP31]], 4
 ; SSE-NEXT:    store i32 [[TMP32]], i32* [[TMP29]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @gather_load_3(
 ; AVX-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], 1
-; AVX-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
-; AVX-NEXT:    store i32 [[TMP4]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
+; AVX-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
+; AVX-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
 ; AVX-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP8:%.*]] = add i32 [[TMP7]], 2
-; AVX-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 2
-; AVX-NEXT:    store i32 [[TMP8]], i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
+; AVX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 15
+; AVX-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 18
 ; AVX-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], 3
-; AVX-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 3
-; AVX-NEXT:    store i32 [[TMP12]], i32* [[TMP9]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 15
+; AVX-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
+; AVX-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
 ; AVX-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP16:%.*]] = add i32 [[TMP15]], 4
-; AVX-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 4
-; AVX-NEXT:    store i32 [[TMP16]], i32* [[TMP13]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 18
-; AVX-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP20:%.*]] = add i32 [[TMP19]], 1
-; AVX-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
-; AVX-NEXT:    store i32 [[TMP20]], i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
-; AVX-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP24:%.*]] = add i32 [[TMP23]], 2
-; AVX-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
-; AVX-NEXT:    store i32 [[TMP24]], i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
-; AVX-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP28:%.*]] = add i32 [[TMP27]], 3
-; AVX-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
-; AVX-NEXT:    store i32 [[TMP28]], i32* [[TMP25]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
-; AVX-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP32:%.*]] = add i32 [[TMP31]], 4
-; AVX-NEXT:    store i32 [[TMP32]], i32* [[TMP29]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
+; AVX-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP18:%.*]] = insertelement <8 x i32> poison, i32 [[TMP3]], i32 0
+; AVX-NEXT:    [[TMP19:%.*]] = insertelement <8 x i32> [[TMP18]], i32 [[TMP5]], i32 1
+; AVX-NEXT:    [[TMP20:%.*]] = insertelement <8 x i32> [[TMP19]], i32 [[TMP7]], i32 2
+; AVX-NEXT:    [[TMP21:%.*]] = insertelement <8 x i32> [[TMP20]], i32 [[TMP9]], i32 3
+; AVX-NEXT:    [[TMP22:%.*]] = insertelement <8 x i32> [[TMP21]], i32 [[TMP11]], i32 4
+; AVX-NEXT:    [[TMP23:%.*]] = insertelement <8 x i32> [[TMP22]], i32 [[TMP13]], i32 5
+; AVX-NEXT:    [[TMP24:%.*]] = insertelement <8 x i32> [[TMP23]], i32 [[TMP15]], i32 6
+; AVX-NEXT:    [[TMP25:%.*]] = insertelement <8 x i32> [[TMP24]], i32 [[TMP17]], i32 7
+; AVX-NEXT:    [[TMP26:%.*]] = add <8 x i32> [[TMP25]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
+; AVX-NEXT:    [[TMP27:%.*]] = bitcast i32* [[TMP0:%.*]] to <8 x i32>*
+; AVX-NEXT:    store <8 x i32> [[TMP26]], <8 x i32>* [[TMP27]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    ret void
 ;
 ; AVX2-LABEL: @gather_load_3(
 ; AVX2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], 1
-; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
-; AVX2-NEXT:    store i32 [[TMP4]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1]], i32 0
-; AVX2-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32*> [[TMP6]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP8:%.*]] = getelementptr i32, <4 x i32*> [[TMP7]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
-; AVX2-NEXT:    [[TMP9:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP8]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[TMP9]], <i32 2, i32 3, i32 4, i32 1>
-; AVX2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
-; AVX2-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
-; AVX2-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
-; AVX2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 2
-; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
-; AVX2-NEXT:    store i32 [[TMP15]], i32* [[TMP11]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
-; AVX2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP19:%.*]] = add i32 [[TMP18]], 3
-; AVX2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
-; AVX2-NEXT:    store i32 [[TMP19]], i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
-; AVX2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP23:%.*]] = add i32 [[TMP22]], 4
-; AVX2-NEXT:    store i32 [[TMP23]], i32* [[TMP20]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
+; AVX2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
+; AVX2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 15
+; AVX2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 18
+; AVX2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
+; AVX2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
+; AVX2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
+; AVX2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP18:%.*]] = insertelement <8 x i32> poison, i32 [[TMP3]], i32 0
+; AVX2-NEXT:    [[TMP19:%.*]] = insertelement <8 x i32> [[TMP18]], i32 [[TMP5]], i32 1
+; AVX2-NEXT:    [[TMP20:%.*]] = insertelement <8 x i32> [[TMP19]], i32 [[TMP7]], i32 2
+; AVX2-NEXT:    [[TMP21:%.*]] = insertelement <8 x i32> [[TMP20]], i32 [[TMP9]], i32 3
+; AVX2-NEXT:    [[TMP22:%.*]] = insertelement <8 x i32> [[TMP21]], i32 [[TMP11]], i32 4
+; AVX2-NEXT:    [[TMP23:%.*]] = insertelement <8 x i32> [[TMP22]], i32 [[TMP13]], i32 5
+; AVX2-NEXT:    [[TMP24:%.*]] = insertelement <8 x i32> [[TMP23]], i32 [[TMP15]], i32 6
+; AVX2-NEXT:    [[TMP25:%.*]] = insertelement <8 x i32> [[TMP24]], i32 [[TMP17]], i32 7
+; AVX2-NEXT:    [[TMP26:%.*]] = add <8 x i32> [[TMP25]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
+; AVX2-NEXT:    [[TMP27:%.*]] = bitcast i32* [[TMP0:%.*]] to <8 x i32>*
+; AVX2-NEXT:    store <8 x i32> [[TMP26]], <8 x i32>* [[TMP27]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    ret void
 ;
-; AVX512-LABEL: @gather_load_3(
-; AVX512-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], 1
-; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
-; AVX512-NEXT:    store i32 [[TMP4]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1]], i32 0
-; AVX512-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32*> [[TMP6]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP8:%.*]] = getelementptr i32, <4 x i32*> [[TMP7]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
-; AVX512-NEXT:    [[TMP9:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP8]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[TMP9]], <i32 2, i32 3, i32 4, i32 1>
-; AVX512-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
-; AVX512-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
-; AVX512-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
-; AVX512-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 2
-; AVX512-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
-; AVX512-NEXT:    store i32 [[TMP15]], i32* [[TMP11]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
-; AVX512-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP19:%.*]] = add i32 [[TMP18]], 3
-; AVX512-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
-; AVX512-NEXT:    store i32 [[TMP19]], i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
-; AVX512-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP23:%.*]] = add i32 [[TMP22]], 4
-; AVX512-NEXT:    store i32 [[TMP23]], i32* [[TMP20]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    ret void
+; AVX512F-LABEL: @gather_load_3(
+; AVX512F-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
+; AVX512F-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
+; AVX512F-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 15
+; AVX512F-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP10:%.*]] = insertelement <4 x i32> poison, i32 [[TMP3]], i32 0
+; AVX512F-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> [[TMP10]], i32 [[TMP5]], i32 1
+; AVX512F-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP7]], i32 2
+; AVX512F-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP9]], i32 3
+; AVX512F-NEXT:    [[TMP14:%.*]] = add <4 x i32> [[TMP13]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 4
+; AVX512F-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP0]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP14]], <4 x i32>* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 18
+; AVX512F-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
+; AVX512F-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
+; AVX512F-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
+; AVX512F-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP25:%.*]] = insertelement <4 x i32> poison, i32 [[TMP18]], i32 0
+; AVX512F-NEXT:    [[TMP26:%.*]] = insertelement <4 x i32> [[TMP25]], i32 [[TMP20]], i32 1
+; AVX512F-NEXT:    [[TMP27:%.*]] = insertelement <4 x i32> [[TMP26]], i32 [[TMP22]], i32 2
+; AVX512F-NEXT:    [[TMP28:%.*]] = insertelement <4 x i32> [[TMP27]], i32 [[TMP24]], i32 3
+; AVX512F-NEXT:    [[TMP29:%.*]] = add <4 x i32> [[TMP28]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP30:%.*]] = bitcast i32* [[TMP15]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP29]], <4 x i32>* [[TMP30]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    ret void
+;
+; AVX512VL-LABEL: @gather_load_3(
+; AVX512VL-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], 1
+; AVX512VL-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
+; AVX512VL-NEXT:    store i32 [[TMP4]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1]], i32 0
+; AVX512VL-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32*> [[TMP6]], <4 x i32*> poison, <4 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP8:%.*]] = getelementptr i32, <4 x i32*> [[TMP7]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
+; AVX512VL-NEXT:    [[TMP9:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP8]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[TMP9]], <i32 2, i32 3, i32 4, i32 1>
+; AVX512VL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
+; AVX512VL-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
+; AVX512VL-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
+; AVX512VL-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 2
+; AVX512VL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
+; AVX512VL-NEXT:    store i32 [[TMP15]], i32* [[TMP11]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
+; AVX512VL-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP19:%.*]] = add i32 [[TMP18]], 3
+; AVX512VL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
+; AVX512VL-NEXT:    store i32 [[TMP19]], i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
+; AVX512VL-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP23:%.*]] = add i32 [[TMP22]], 4
+; AVX512VL-NEXT:    store i32 [[TMP23]], i32* [[TMP20]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    ret void
 ;
   %3 = load i32, i32* %1, align 4, !tbaa !2
   %4 = add i32 %3, 1
   %5 = getelementptr inbounds i32, i32* %0, i64 1
   store i32 %4, i32* %0, align 4, !tbaa !2
   %6 = getelementptr inbounds i32, i32* %1, i64 11
   %7 = load i32, i32* %6, align 4, !tbaa !2
   %8 = add i32 %7, 2

 ; SSE-NEXT:    store i32 [[T16]], i32* [[T13]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    store i32 [[T20]], i32* [[T17]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @gather_load_4(
-; AVX-NEXT:    [[T5:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 1
 ; AVX-NEXT:    [[T6:%.*]] = getelementptr inbounds i32, i32* [[T1:%.*]], i64 11
-; AVX-NEXT:    [[T9:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 2
 ; AVX-NEXT:    [[T10:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 4
-; AVX-NEXT:    [[T13:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 3
 ; AVX-NEXT:    [[T14:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 15
-; AVX-NEXT:    [[T17:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 4
 ; AVX-NEXT:    [[T18:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 18
-; AVX-NEXT:    [[T21:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 5
 ; AVX-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
-; AVX-NEXT:    [[T25:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 6
 ; AVX-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
-; AVX-NEXT:    [[T29:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 7
 ; AVX-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
 ; AVX-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T7:%.*]] = load i32, i32* [[T6]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T11:%.*]] = load i32, i32* [[T10]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T15:%.*]] = load i32, i32* [[T14]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T19:%.*]] = load i32, i32* [[T18]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[T4:%.*]] = add i32 [[T3]], 1
-; AVX-NEXT:    [[T8:%.*]] = add i32 [[T7]], 2
-; AVX-NEXT:    [[T12:%.*]] = add i32 [[T11]], 3
-; AVX-NEXT:    [[T16:%.*]] = add i32 [[T15]], 4
-; AVX-NEXT:    [[T20:%.*]] = add i32 [[T19]], 1
-; AVX-NEXT:    [[T24:%.*]] = add i32 [[T23]], 2
-; AVX-NEXT:    [[T28:%.*]] = add i32 [[T27]], 3
-; AVX-NEXT:    [[T32:%.*]] = add i32 [[T31]], 4
-; AVX-NEXT:    store i32 [[T4]], i32* [[T0]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T8]], i32* [[T5]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T12]], i32* [[T9]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T16]], i32* [[T13]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T20]], i32* [[T17]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP1:%.*]] = insertelement <8 x i32> poison, i32 [[T3]], i32 0
+; AVX-NEXT:    [[TMP2:%.*]] = insertelement <8 x i32> [[TMP1]], i32 [[T7]], i32 1
+; AVX-NEXT:    [[TMP3:%.*]] = insertelement <8 x i32> [[TMP2]], i32 [[T11]], i32 2
+; AVX-NEXT:    [[TMP4:%.*]] = insertelement <8 x i32> [[TMP3]], i32 [[T15]], i32 3
+; AVX-NEXT:    [[TMP5:%.*]] = insertelement <8 x i32> [[TMP4]], i32 [[T19]], i32 4
+; AVX-NEXT:    [[TMP6:%.*]] = insertelement <8 x i32> [[TMP5]], i32 [[T23]], i32 5
+; AVX-NEXT:    [[TMP7:%.*]] = insertelement <8 x i32> [[TMP6]], i32 [[T27]], i32 6
+; AVX-NEXT:    [[TMP8:%.*]] = insertelement <8 x i32> [[TMP7]], i32 [[T31]], i32 7
+; AVX-NEXT:    [[TMP9:%.*]] = add <8 x i32> [[TMP8]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
+; AVX-NEXT:    [[TMP10:%.*]] = bitcast i32* [[T0:%.*]] to <8 x i32>*
+; AVX-NEXT:    store <8 x i32> [[TMP9]], <8 x i32>* [[TMP10]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    ret void
 ;
 ; AVX2-LABEL: @gather_load_4(
-; AVX2-NEXT:    [[T5:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 1
-; AVX2-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32*> poison, i32* [[T1:%.*]], i32 0
-; AVX2-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32*> [[TMP1]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr i32, <4 x i32*> [[TMP2]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
-; AVX2-NEXT:    [[T21:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 5
+; AVX2-NEXT:    [[T6:%.*]] = getelementptr inbounds i32, i32* [[T1:%.*]], i64 11
+; AVX2-NEXT:    [[T10:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 4
+; AVX2-NEXT:    [[T14:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 15
+; AVX2-NEXT:    [[T18:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 18
 ; AVX2-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
-; AVX2-NEXT:    [[T25:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 6
 ; AVX2-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
-; AVX2-NEXT:    [[T29:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 7
 ; AVX2-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
 ; AVX2-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP3]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[T7:%.*]] = load i32, i32* [[T6]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[T11:%.*]] = load i32, i32* [[T10]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[T15:%.*]] = load i32, i32* [[T14]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[T19:%.*]] = load i32, i32* [[T18]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[T4:%.*]] = add i32 [[T3]], 1
-; AVX2-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], <i32 2, i32 3, i32 4, i32 1>
-; AVX2-NEXT:    [[T24:%.*]] = add i32 [[T23]], 2
-; AVX2-NEXT:    [[T28:%.*]] = add i32 [[T27]], 3
-; AVX2-NEXT:    [[T32:%.*]] = add i32 [[T31]], 4
-; AVX2-NEXT:    store i32 [[T4]], i32* [[T0]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP6:%.*]] = bitcast i32* [[T5]] to <4 x i32>*
-; AVX2-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP1:%.*]] = insertelement <8 x i32> poison, i32 [[T3]], i32 0
+; AVX2-NEXT:    [[TMP2:%.*]] = insertelement <8 x i32> [[TMP1]], i32 [[T7]], i32 1
+; AVX2-NEXT:    [[TMP3:%.*]] = insertelement <8 x i32> [[TMP2]], i32 [[T11]], i32 2
+; AVX2-NEXT:    [[TMP4:%.*]] = insertelement <8 x i32> [[TMP3]], i32 [[T15]], i32 3
+; AVX2-NEXT:    [[TMP5:%.*]] = insertelement <8 x i32> [[TMP4]], i32 [[T19]], i32 4
+; AVX2-NEXT:    [[TMP6:%.*]] = insertelement <8 x i32> [[TMP5]], i32 [[T23]], i32 5
+; AVX2-NEXT:    [[TMP7:%.*]] = insertelement <8 x i32> [[TMP6]], i32 [[T27]], i32 6
+; AVX2-NEXT:    [[TMP8:%.*]] = insertelement <8 x i32> [[TMP7]], i32 [[T31]], i32 7
+; AVX2-NEXT:    [[TMP9:%.*]] = add <8 x i32> [[TMP8]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
+; AVX2-NEXT:    [[TMP10:%.*]] = bitcast i32* [[T0:%.*]] to <8 x i32>*
+; AVX2-NEXT:    store <8 x i32> [[TMP9]], <8 x i32>* [[TMP10]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    ret void
 ;
-; AVX512-LABEL: @gather_load_4(
-; AVX512-NEXT:    [[T5:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 1
-; AVX512-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32*> poison, i32* [[T1:%.*]], i32 0
-; AVX512-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32*> [[TMP1]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr i32, <4 x i32*> [[TMP2]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
-; AVX512-NEXT:    [[T21:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 5
-; AVX512-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
-; AVX512-NEXT:    [[T25:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 6
-; AVX512-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
-; AVX512-NEXT:    [[T29:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 7
-; AVX512-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
-; AVX512-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP3]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[T4:%.*]] = add i32 [[T3]], 1
-; AVX512-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], <i32 2, i32 3, i32 4, i32 1>
-; AVX512-NEXT:    [[T24:%.*]] = add i32 [[T23]], 2
-; AVX512-NEXT:    [[T28:%.*]] = add i32 [[T27]], 3
-; AVX512-NEXT:    [[T32:%.*]] = add i32 [[T31]], 4
-; AVX512-NEXT:    store i32 [[T4]], i32* [[T0]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP6:%.*]] = bitcast i32* [[T5]] to <4 x i32>*
-; AVX512-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    ret void
+; AVX512F-LABEL: @gather_load_4(
+; AVX512F-NEXT:    [[T6:%.*]] = getelementptr inbounds i32, i32* [[T1:%.*]], i64 11
+; AVX512F-NEXT:    [[T10:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 4
+; AVX512F-NEXT:    [[T14:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 15
+; AVX512F-NEXT:    [[T17:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 4
+; AVX512F-NEXT:    [[T18:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 18
+; AVX512F-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
+; AVX512F-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
+; AVX512F-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
+; AVX512F-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T7:%.*]] = load i32, i32* [[T6]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T11:%.*]] = load i32, i32* [[T10]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T15:%.*]] = load i32, i32* [[T14]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T19:%.*]] = load i32, i32* [[T18]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32> poison, i32 [[T3]], i32 0
+; AVX512F-NEXT:    [[TMP2:%.*]] = insertelement <4 x i32> [[TMP1]], i32 [[T7]], i32 1
+; AVX512F-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[T11]], i32 2
+; AVX512F-NEXT:    [[TMP4:%.*]] = insertelement <4 x i32> [[TMP3]], i32 [[T15]], i32 3
+; AVX512F-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32> poison, i32 [[T19]], i32 0
+; AVX512F-NEXT:    [[TMP7:%.*]] = insertelement <4 x i32> [[TMP6]], i32 [[T23]], i32 1
+; AVX512F-NEXT:    [[TMP8:%.*]] = insertelement <4 x i32> [[TMP7]], i32 [[T27]], i32 2
+; AVX512F-NEXT:    [[TMP9:%.*]] = insertelement <4 x i32> [[TMP8]], i32 [[T31]], i32 3
+; AVX512F-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[TMP9]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP11:%.*]] = bitcast i32* [[T0]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP11]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP12:%.*]] = bitcast i32* [[T17]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    ret void
+;
+; AVX512VL-LABEL: @gather_load_4(
+; AVX512VL-NEXT:    [[T5:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 1
+; AVX512VL-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32*> poison, i32* [[T1:%.*]], i32 0
+; AVX512VL-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32*> [[TMP1]], <4 x i32*> poison, <4 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP3:%.*]] = getelementptr i32, <4 x i32*> [[TMP2]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
+; AVX512VL-NEXT:    [[T21:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 5
+; AVX512VL-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
+; AVX512VL-NEXT:    [[T25:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 6
+; AVX512VL-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
+; AVX512VL-NEXT:    [[T29:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 7
+; AVX512VL-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
+; AVX512VL-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP3]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[T4:%.*]] = add i32 [[T3]], 1
+; AVX512VL-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], <i32 2, i32 3, i32 4, i32 1>
+; AVX512VL-NEXT:    [[T24:%.*]] = add i32 [[T23]], 2
+; AVX512VL-NEXT:    [[T28:%.*]] = add i32 [[T27]], 3
+; AVX512VL-NEXT:    [[T32:%.*]] = add i32 [[T31]], 4
+; AVX512VL-NEXT:    store i32 [[T4]], i32* [[T0]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP6:%.*]] = bitcast i32* [[T5]] to <4 x i32>*
+; AVX512VL-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    ret void
 ;
   %t5 = getelementptr inbounds i32, i32* %t0, i64 1
   %t6 = getelementptr inbounds i32, i32* %t1, i64 11
   %t9 = getelementptr inbounds i32, i32* %t0, i64 2
   %t10 = getelementptr inbounds i32, i32* %t1, i64 4
   %t13 = getelementptr inbounds i32, i32* %t0, i64 3
   %t14 = getelementptr inbounds i32, i32* %t1, i64 15
   %t17 = getelementptr inbounds i32, i32* %t0, i64 4
;
   store i32 %t32, i32* %t29, align 4, !tbaa !2
 
   ret void
 }
 
 
 define void @gather_load_div(float* noalias nocapture %0, float* noalias nocapture readonly %1) {
 ; SSE-LABEL: @gather_load_div(
-; SSE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
-; SSE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 3
-; SSE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 14
-; SSE-NEXT:    [[TMP6:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
-; SSE-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> [[TMP6]], float* [[TMP3]], i32 1
-; SSE-NEXT:    [[TMP8:%.*]] = insertelement <4 x float*> [[TMP7]], float* [[TMP4]], i32 2
-; SSE-NEXT:    [[TMP9:%.*]] = insertelement <4 x float*> [[TMP8]], float* [[TMP5]], i32 3
-; SSE-NEXT:    [[TMP10:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> [[TMP9]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef), !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP11:%.*]] = shufflevector <4 x float*> [[TMP6]], <4 x float*> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:    [[TMP12:%.*]] = getelementptr float, <4 x float*> [[TMP11]], <4 x i64> <i64 4, i64 13, i64 11, i64 44>
-; SSE-NEXT:    [[TMP13:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> [[TMP12]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef), !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP14:%.*]] = fdiv <4 x float> [[TMP10]], [[TMP13]]
-; SSE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, float* [[TMP0:%.*]], i64 4
-; SSE-NEXT:    [[TMP16:%.*]] = bitcast float* [[TMP0]] to <4 x float>*
-; SSE-NEXT:    store <4 x float> [[TMP14]], <4 x float>* [[TMP16]], align 4, !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP17:%.*]] = getelementptr float, <4 x float*> [[TMP11]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
-; SSE-NEXT:    [[TMP18:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> [[TMP17]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef), !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP19:%.*]] = getelementptr float, <4 x float*> [[TMP11]], <4 x i64> <i64 33, i64 30, i64 27, i64 23>
-; SSE-NEXT:    [[TMP20:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> [[TMP19]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef), !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP21:%.*]] = fdiv <4 x float> [[TMP18]], [[TMP20]]
-; SSE-NEXT:    [[TMP22:%.*]] = bitcast float* [[TMP15]] to <4 x float>*
-; SSE-NEXT:    store <4 x float> [[TMP21]], <4 x float>* [[TMP22]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 4
+; SSE-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 10
+; SSE-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 13
+; SSE-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 3
+; SSE-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP10]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 11
+; SSE-NEXT:    [[TMP13:%.*]] = load float, float* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 14
+; SSE-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 44
+; SSE-NEXT:    [[TMP17:%.*]] = load float, float* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP18:%.*]] = insertelement <4 x float> poison, float [[TMP3]], i32 0
+; SSE-NEXT:    [[TMP19:%.*]] = insertelement <4 x float> [[TMP18]], float [[TMP7]], i32 1
+; SSE-NEXT:    [[TMP20:%.*]] = insertelement <4 x float> [[TMP19]], float [[TMP11]], i32 2
+; SSE-NEXT:    [[TMP21:%.*]] = insertelement <4 x float> [[TMP20]], float [[TMP15]], i32 3
+; SSE-NEXT:    [[TMP22:%.*]] = insertelement <4 x float> poison, float [[TMP5]], i32 0
+; SSE-NEXT:    [[TMP23:%.*]] = insertelement <4 x float> [[TMP22]], float [[TMP9]], i32 1
+; SSE-NEXT:    [[TMP24:%.*]] = insertelement <4 x float> [[TMP23]], float [[TMP13]], i32 2
+; SSE-NEXT:    [[TMP25:%.*]] = insertelement <4 x float> [[TMP24]], float [[TMP17]], i32 3
+; SSE-NEXT:    [[TMP26:%.*]] = fdiv <4 x float> [[TMP21]], [[TMP25]]
+; SSE-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, float* [[TMP0:%.*]], i64 4
+; SSE-NEXT:    [[TMP28:%.*]] = bitcast float* [[TMP0]] to <4 x float>*
+; SSE-NEXT:    store <4 x float> [[TMP26]], <4 x float>* [[TMP28]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 17
+; SSE-NEXT:    [[TMP30:%.*]] = load float, float* [[TMP29]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP31:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 33
+; SSE-NEXT:    [[TMP32:%.*]] = load float, float* [[TMP31]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 8
+; SSE-NEXT:    [[TMP34:%.*]] = load float, float* [[TMP33]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 30
+; SSE-NEXT:    [[TMP36:%.*]] = load float, float* [[TMP35]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 5
+; SSE-NEXT:    [[TMP38:%.*]] = load float, float* [[TMP37]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 27
+; SSE-NEXT:    [[TMP40:%.*]] = load float, float* [[TMP39]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 20
+; SSE-NEXT:    [[TMP42:%.*]] = load float, float* [[TMP41]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP43:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 23
+; SSE-NEXT:    [[TMP44:%.*]] = load float, float* [[TMP43]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP45:%.*]] = insertelement <4 x float> poison, float [[TMP30]], i32 0
+; SSE-NEXT:    [[TMP46:%.*]] = insertelement <4 x float> [[TMP45]], float [[TMP34]], i32 1
+; SSE-NEXT:    [[TMP47:%.*]] = insertelement <4 x float> [[TMP46]], float [[TMP38]], i32 2
+; SSE-NEXT:    [[TMP48:%.*]] = insertelement <4 x float> [[TMP47]], float [[TMP42]], i32 3
+; SSE-NEXT:    [[TMP49:%.*]] = insertelement <4 x float> poison, float [[TMP32]], i32 0
+; SSE-NEXT:    [[TMP50:%.*]] = insertelement <4 x float> [[TMP49]], float [[TMP36]], i32 1
+; SSE-NEXT:    [[TMP51:%.*]] = insertelement <4 x float> [[TMP50]], float [[TMP40]], i32 2
+; SSE-NEXT:    [[TMP52:%.*]] = insertelement <4 x float> [[TMP51]], float [[TMP44]], i32 3
+; SSE-NEXT:    [[TMP53:%.*]] = fdiv <4 x float> [[TMP48]], [[TMP52]]
+; SSE-NEXT:    [[TMP54:%.*]] = bitcast float* [[TMP27]] to <4 x float>*
+; SSE-NEXT:    store <4 x float> [[TMP53]], <4 x float>* [[TMP54]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @gather_load_div(
-; AVX-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
-; AVX-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
-; AVX-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
-; AVX-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
-; AVX-NEXT:    [[TMP7:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 17, i64 8>
-; AVX-NEXT:    [[TMP8:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 5, i64 20>
-; AVX-NEXT:    [[TMP9:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
-; AVX-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> [[TMP9]], float* [[TMP3]], i32 1
-; AVX-NEXT:    [[TMP11:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP12:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> [[TMP11]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP13:%.*]] = shufflevector <2 x float*> [[TMP7]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP14:%.*]] = shufflevector <8 x float*> [[TMP12]], <8 x float*> [[TMP13]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP15:%.*]] = shufflevector <2 x float*> [[TMP8]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP16:%.*]] = shufflevector <8 x float*> [[TMP14]], <8 x float*> [[TMP15]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 8, i32 9>
-; AVX-NEXT:    [[TMP17:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP16]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP18:%.*]] = shufflevector <8 x float*> [[TMP9]], <8 x float*> poison, <8 x i32> zeroinitializer
-; AVX-NEXT:    [[TMP19:%.*]] = getelementptr float, <8 x float*> [[TMP18]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
-; AVX-NEXT:    [[TMP20:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP19]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP21:%.*]] = fdiv <8 x float> [[TMP17]], [[TMP20]]
-; AVX-NEXT:    [[TMP22:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
-; AVX-NEXT:    store <8 x float> [[TMP21]], <8 x float>* [[TMP22]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 4
+; AVX-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 10
+; AVX-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 13
+; AVX-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 3
+; AVX-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP10]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 11
+; AVX-NEXT:    [[TMP13:%.*]] = load float, float* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 14
+; AVX-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 44
+; AVX-NEXT:    [[TMP17:%.*]] = load float, float* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 17
+; AVX-NEXT:    [[TMP19:%.*]] = load float, float* [[TMP18]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 33
+; AVX-NEXT:    [[TMP21:%.*]] = load float, float* [[TMP20]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 8
+; AVX-NEXT:    [[TMP23:%.*]] = load float, float* [[TMP22]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 30
+; AVX-NEXT:    [[TMP25:%.*]] = load float, float* [[TMP24]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 5
+; AVX-NEXT:    [[TMP27:%.*]] = load float, float* [[TMP26]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 27
+; AVX-NEXT:    [[TMP29:%.*]] = load float, float* [[TMP28]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 20
+; AVX-NEXT:    [[TMP31:%.*]] = load float, float* [[TMP30]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 23
+; AVX-NEXT:    [[TMP33:%.*]] = load float, float* [[TMP32]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP34:%.*]] = insertelement <8 x float> poison, float [[TMP3]], i32 0
+; AVX-NEXT:    [[TMP35:%.*]] = insertelement <8 x float> [[TMP34]], float [[TMP7]], i32 1
+; AVX-NEXT:    [[TMP36:%.*]] = insertelement <8 x float> [[TMP35]], float [[TMP11]], i32 2
+; AVX-NEXT:    [[TMP37:%.*]] = insertelement <8 x float> [[TMP36]], float [[TMP15]], i32 3
+; AVX-NEXT:    [[TMP38:%.*]] = insertelement <8 x float> [[TMP37]], float [[TMP19]], i32 4
+; AVX-NEXT:    [[TMP39:%.*]] = insertelement <8 x float> [[TMP38]], float [[TMP23]], i32 5
+; AVX-NEXT:    [[TMP40:%.*]] = insertelement <8 x float> [[TMP39]], float [[TMP27]], i32 6
+; AVX-NEXT:    [[TMP41:%.*]] = insertelement <8 x float> [[TMP40]], float [[TMP31]], i32 7
+; AVX-NEXT:    [[TMP42:%.*]] = insertelement <8 x float> poison, float [[TMP5]], i32 0
+; AVX-NEXT:    [[TMP43:%.*]] = insertelement <8 x float> [[TMP42]], float [[TMP9]], i32 1
+; AVX-NEXT:    [[TMP44:%.*]] = insertelement <8 x float> [[TMP43]], float [[TMP13]], i32 2
+; AVX-NEXT:    [[TMP45:%.*]] = insertelement <8 x float> [[TMP44]], float [[TMP17]], i32 3
+; AVX-NEXT:    [[TMP46:%.*]] = insertelement <8 x float> [[TMP45]], float [[TMP21]], i32 4
+; AVX-NEXT:    [[TMP47:%.*]] = insertelement <8 x float> [[TMP46]], float [[TMP25]], i32 5
+; AVX-NEXT:    [[TMP48:%.*]] = insertelement <8 x float> [[TMP47]], float [[TMP29]], i32 6
+; AVX-NEXT:    [[TMP49:%.*]] = insertelement <8 x float> [[TMP48]], float [[TMP33]], i32 7
+; AVX-NEXT:    [[TMP50:%.*]] = fdiv <8 x float> [[TMP41]], [[TMP49]]
+; AVX-NEXT:    [[TMP51:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
+; AVX-NEXT:    store <8 x float> [[TMP50]], <8 x float>* [[TMP51]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    ret void
 ;
 ; AVX2-LABEL: @gather_load_div(
-; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
-; AVX2-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
-; AVX2-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
-; AVX2-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
-; AVX2-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float*> [[TMP7]], <4 x float*> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP9:%.*]] = getelementptr float, <4 x float*> [[TMP8]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
-; AVX2-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
-; AVX2-NEXT:    [[TMP11:%.*]] = insertelement <8 x float*> [[TMP10]], float* [[TMP3]], i32 1
-; AVX2-NEXT:    [[TMP12:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX2-NEXT:    [[TMP13:%.*]] = shufflevector <8 x float*> [[TMP11]], <8 x float*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX2-NEXT:    [[TMP14:%.*]] = shufflevector <4 x float*> [[TMP9]], <4 x float*> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX2-NEXT:    [[TMP15:%.*]] = shufflevector <8 x float*> [[TMP13]], <8 x float*> [[TMP14]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
-; AVX2-NEXT:    [[TMP16:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP15]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP17:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> poison, <8 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP18:%.*]] = getelementptr float, <8 x float*> [[TMP17]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
-; AVX2-NEXT:    [[TMP19:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP18]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP20:%.*]] = fdiv <8 x float> [[TMP16]], [[TMP19]]
-; AVX2-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
-; AVX2-NEXT:    store <8 x float> [[TMP20]], <8 x float>* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 4
+; AVX2-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 10
+; AVX2-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 13
+; AVX2-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 3
+; AVX2-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP10]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 11
+; AVX2-NEXT:    [[TMP13:%.*]] = load float, float* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 14
+; AVX2-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 44
+; AVX2-NEXT:    [[TMP17:%.*]] = load float, float* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 17
+; AVX2-NEXT:    [[TMP19:%.*]] = load float, float* [[TMP18]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 33
+; AVX2-NEXT:    [[TMP21:%.*]] = load float, float* [[TMP20]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 8
+; AVX2-NEXT:    [[TMP23:%.*]] = load float, float* [[TMP22]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 30
+; AVX2-NEXT:    [[TMP25:%.*]] = load float, float* [[TMP24]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 5
+; AVX2-NEXT:    [[TMP27:%.*]] = load float, float* [[TMP26]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 27
+; AVX2-NEXT:    [[TMP29:%.*]] = load float, float* [[TMP28]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 20
+; AVX2-NEXT:    [[TMP31:%.*]] = load float, float* [[TMP30]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 23
+; AVX2-NEXT:    [[TMP33:%.*]] = load float, float* [[TMP32]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP34:%.*]] = insertelement <8 x float> poison, float [[TMP3]], i32 0
+; AVX2-NEXT:    [[TMP35:%.*]] = insertelement <8 x float> [[TMP34]], float [[TMP7]], i32 1
+; AVX2-NEXT:    [[TMP36:%.*]] = insertelement <8 x float> [[TMP35]], float [[TMP11]], i32 2
+; AVX2-NEXT:    [[TMP37:%.*]] = insertelement <8 x float> [[TMP36]], float [[TMP15]], i32 3
+; AVX2-NEXT:    [[TMP38:%.*]] = insertelement <8 x float> [[TMP37]], float [[TMP19]], i32 4
+; AVX2-NEXT:    [[TMP39:%.*]] = insertelement <8 x float> [[TMP38]], float [[TMP23]], i32 5
+; AVX2-NEXT:    [[TMP40:%.*]] = insertelement <8 x float> [[TMP39]], float [[TMP27]], i32 6
+; AVX2-NEXT:    [[TMP41:%.*]] = insertelement <8 x float> [[TMP40]], float [[TMP31]], i32 7
+; AVX2-NEXT:    [[TMP42:%.*]] = insertelement <8 x float> poison, float [[TMP5]], i32 0
+; AVX2-NEXT:    [[TMP43:%.*]] = insertelement <8 x float> [[TMP42]], float [[TMP9]], i32 1
+; AVX2-NEXT:    [[TMP44:%.*]] = insertelement <8 x float> [[TMP43]], float [[TMP13]], i32 2
+; AVX2-NEXT:    [[TMP45:%.*]] = insertelement <8 x float> [[TMP44]], float [[TMP17]], i32 3
+; AVX2-NEXT:    [[TMP46:%.*]] = insertelement <8 x float> [[TMP45]], float [[TMP21]], i32 4
+; AVX2-NEXT:    [[TMP47:%.*]] = insertelement <8 x float> [[TMP46]], float [[TMP25]], i32 5
+; AVX2-NEXT:    [[TMP48:%.*]] = insertelement <8 x float> [[TMP47]], float [[TMP29]], i32 6
+; AVX2-NEXT:    [[TMP49:%.*]] = insertelement <8 x float> [[TMP48]], float [[TMP33]], i32 7
+; AVX2-NEXT:    [[TMP50:%.*]] = fdiv <8 x float> [[TMP41]], [[TMP49]]
+; AVX2-NEXT:    [[TMP51:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
+; AVX2-NEXT:    store <8 x float> [[TMP50]], <8 x float>* [[TMP51]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    ret void
 ;
-; AVX512-LABEL: @gather_load_div(
-; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
-; AVX512-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
-; AVX512-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
-; AVX512-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
-; AVX512-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float*> [[TMP7]], <4 x float*> poison, <4 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP9:%.*]] = getelementptr float, <4 x float*> [[TMP8]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
-; AVX512-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
-; AVX512-NEXT:    [[TMP11:%.*]] = insertelement <8 x float*> [[TMP10]], float* [[TMP3]], i32 1
-; AVX512-NEXT:    [[TMP12:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX512-NEXT:    [[TMP13:%.*]] = shufflevector <8 x float*> [[TMP11]], <8 x float*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX512-NEXT:    [[TMP14:%.*]] = shufflevector <4 x float*> [[TMP9]], <4 x float*> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX512-NEXT:    [[TMP15:%.*]] = shufflevector <8 x float*> [[TMP13]], <8 x float*> [[TMP14]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
-; AVX512-NEXT:    [[TMP16:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP15]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP17:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> poison, <8 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP18:%.*]] = getelementptr float, <8 x float*> [[TMP17]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
-; AVX512-NEXT:    [[TMP19:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP18]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP20:%.*]] = fdiv <8 x float> [[TMP16]], [[TMP19]]
-; AVX512-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
-; AVX512-NEXT:    store <8 x float> [[TMP20]], <8 x float>* [[TMP21]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    ret void
+; AVX512F-LABEL: @gather_load_div(
+; AVX512F-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
+; AVX512F-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
+; AVX512F-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
+; AVX512F-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
+; AVX512F-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
+; AVX512F-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float*> [[TMP7]], <4 x float*> poison, <4 x i32> zeroinitializer
+; AVX512F-NEXT:    [[TMP9:%.*]] = getelementptr float, <4 x float*> [[TMP8]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
+; AVX512F-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
+; AVX512F-NEXT:    [[TMP11:%.*]] = insertelement <8 x float*> [[TMP10]], float* [[TMP3]], i32 1
+; AVX512F-NEXT:    [[TMP12:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512F-NEXT:    [[TMP13:%.*]] = shufflevector <8 x float*> [[TMP11]], <8 x float*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512F-NEXT:    [[TMP14:%.*]] = shufflevector <4 x float*> [[TMP9]], <4 x float*> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512F-NEXT:    [[TMP15:%.*]] = shufflevector <8 x float*> [[TMP13]], <8 x float*> [[TMP14]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:    [[TMP16:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP15]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP17:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> poison, <8 x i32> zeroinitializer
+; AVX512F-NEXT:    [[TMP18:%.*]] = getelementptr float, <8 x float*> [[TMP17]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
+; AVX512F-NEXT:    [[TMP19:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP18]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP20:%.*]] = fdiv <8 x float> [[TMP16]], [[TMP19]]
+; AVX512F-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
+; AVX512F-NEXT:    store <8 x float> [[TMP20]], <8 x float>* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    ret void
+;
+; AVX512VL-LABEL: @gather_load_div(
+; AVX512VL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
+; AVX512VL-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
+; AVX512VL-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
+; AVX512VL-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
+; AVX512VL-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float*> [[TMP7]], <4 x float*> poison, <4 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP9:%.*]] = getelementptr float, <4 x float*> [[TMP8]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
+; AVX512VL-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
+; AVX512VL-NEXT:    [[TMP11:%.*]] = insertelement <8 x float*> [[TMP10]], float* [[TMP3]], i32 1
+; AVX512VL-NEXT:    [[TMP12:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512VL-NEXT:    [[TMP13:%.*]] = shufflevector <8 x float*> [[TMP11]], <8 x float*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512VL-NEXT:    [[TMP14:%.*]] = shufflevector <4 x float*> [[TMP9]], <4 x float*> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512VL-NEXT:    [[TMP15:%.*]] = shufflevector <8 x float*> [[TMP13]], <8 x float*> [[TMP14]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+; AVX512VL-NEXT:    [[TMP16:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP15]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP17:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> poison, <8 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP18:%.*]] = getelementptr float, <8 x float*> [[TMP17]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
+; AVX512VL-NEXT:    [[TMP19:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP18]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP20:%.*]] = fdiv <8 x float> [[TMP16]], [[TMP19]]
+; AVX512VL-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
+; AVX512VL-NEXT:    store <8 x float> [[TMP20]], <8 x float>* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    ret void
 ;
   %3 = load float, float* %1, align 4, !tbaa !2
   %4 = getelementptr inbounds float, float* %1, i64 4
   %5 = load float, float* %4, align 4, !tbaa !2
   %6 = fdiv float %3, %5
   %7 = getelementptr inbounds float, float* %0, i64 1
   store float %6, float* %0, align 4, !tbaa !2
   %8 = getelementptr inbounds float, float* %1, i64 10

llvm/test/Transforms/SLPVectorizer/X86/pr47629.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+sse2     | FileCheck %s --check-prefixes=CHECK,SSE
 ; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx      | FileCheck %s --check-prefixes=CHECK,AVX
 ; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx2     | FileCheck %s --check-prefixes=CHECK,AVX2
-; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx512f  | FileCheck %s --check-prefixes=CHECK,AVX512
-; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512
+; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx512f  | FileCheck %s --check-prefixes=CHECK,AVX512F
+; RUN:  opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux -mattr=+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512VL
 
 define void @gather_load(i32* noalias nocapture %0, i32* noalias nocapture readonly %1) {
 ; CHECK-LABEL: @gather_load(
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP1]], align 4, !tbaa [[TBAA0:![0-9]+]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
 ; CHECK-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4

 ; SSE-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    [[TMP17:%.*]] = add nsw i32 [[TMP16]], 4
 ; SSE-NEXT:    store i32 [[TMP17]], i32* [[TMP14]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @gather_load_2(
 ; AVX-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
 ; AVX-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP5:%.*]] = add nsw i32 [[TMP4]], 1
-; AVX-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
-; AVX-NEXT:    store i32 [[TMP5]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 10
+; AVX-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 10
+; AVX-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
 ; AVX-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP9:%.*]] = add nsw i32 [[TMP8]], 2
-; AVX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 2
-; AVX-NEXT:    store i32 [[TMP9]], i32* [[TMP6]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
-; AVX-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP13:%.*]] = add nsw i32 [[TMP12]], 3
-; AVX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 3
-; AVX-NEXT:    store i32 [[TMP13]], i32* [[TMP10]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 5
-; AVX-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP17:%.*]] = add nsw i32 [[TMP16]], 4
-; AVX-NEXT:    store i32 [[TMP17]], i32* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 5
+; AVX-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i32 0
+; AVX-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP6]], i32 1
+; AVX-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP8]], i32 2
+; AVX-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP10]], i32 3
+; AVX-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 1, i32 2, i32 3, i32 4>
+; AVX-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
+; AVX-NEXT:    store <4 x i32> [[TMP15]], <4 x i32>* [[TMP16]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    ret void
 ;
 ; AVX2-LABEL: @gather_load_2(
-; AVX2-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1:%.*]], i32 0
-; AVX2-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32*> [[TMP3]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr i32, <4 x i32*> [[TMP4]], <4 x i64> <i64 1, i64 10, i64 3, i64 5>
-; AVX2-NEXT:    [[TMP6:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP5]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 1, i32 2, i32 3, i32 4>
-; AVX2-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
-; AVX2-NEXT:    store <4 x i32> [[TMP7]], <4 x i32>* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
+; AVX2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 10
+; AVX2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
+; AVX2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 5
+; AVX2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i32 0
+; AVX2-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP6]], i32 1
+; AVX2-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP8]], i32 2
+; AVX2-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP10]], i32 3
+; AVX2-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 1, i32 2, i32 3, i32 4>
+; AVX2-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
+; AVX2-NEXT:    store <4 x i32> [[TMP15]], <4 x i32>* [[TMP16]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    ret void
 ;
-; AVX512-LABEL: @gather_load_2(
-; AVX512-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1:%.*]], i32 0
-; AVX512-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32*> [[TMP3]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr i32, <4 x i32*> [[TMP4]], <4 x i64> <i64 1, i64 10, i64 3, i64 5>
-; AVX512-NEXT:    [[TMP6:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP5]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 1, i32 2, i32 3, i32 4>
-; AVX512-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
-; AVX512-NEXT:    store <4 x i32> [[TMP7]], <4 x i32>* [[TMP8]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    ret void
+; AVX512F-LABEL: @gather_load_2(
+; AVX512F-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
+; AVX512F-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 10
+; AVX512F-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
+; AVX512F-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 5
+; AVX512F-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i32 0
+; AVX512F-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP6]], i32 1
+; AVX512F-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP8]], i32 2
+; AVX512F-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP10]], i32 3
+; AVX512F-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP15]], <4 x i32>* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    ret void
+;
+; AVX512VL-LABEL: @gather_load_2(
+; AVX512VL-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1:%.*]], i32 0
+; AVX512VL-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32*> [[TMP3]], <4 x i32*> poison, <4 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP5:%.*]] = getelementptr i32, <4 x i32*> [[TMP4]], <4 x i64> <i64 1, i64 10, i64 3, i64 5>
+; AVX512VL-NEXT:    [[TMP6:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP5]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512VL-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP0:%.*]] to <4 x i32>*
+; AVX512VL-NEXT:    store <4 x i32> [[TMP7]], <4 x i32>* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    ret void
 ;
   %3 = getelementptr inbounds i32, i32* %1, i64 1
   %4 = load i32, i32* %3, align 4, !tbaa !2
   %5 = add nsw i32 %4, 1
   %6 = getelementptr inbounds i32, i32* %0, i64 1
   store i32 %5, i32* %0, align 4, !tbaa !2
   %7 = getelementptr inbounds i32, i32* %1, i64 10
   %8 = load i32, i32* %7, align 4, !tbaa !2

 ; SSE-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
 ; SSE-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    [[TMP32:%.*]] = add i32 [[TMP31]], 4
 ; SSE-NEXT:    store i32 [[TMP32]], i32* [[TMP29]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @gather_load_3(
 ; AVX-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], 1
-; AVX-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
-; AVX-NEXT:    store i32 [[TMP4]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
+; AVX-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
+; AVX-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
 ; AVX-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP8:%.*]] = add i32 [[TMP7]], 2
-; AVX-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 2
-; AVX-NEXT:    store i32 [[TMP8]], i32* [[TMP5]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
+; AVX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 15
+; AVX-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 18
 ; AVX-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], 3
-; AVX-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 3
-; AVX-NEXT:    store i32 [[TMP12]], i32* [[TMP9]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 15
+; AVX-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
+; AVX-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
 ; AVX-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP16:%.*]] = add i32 [[TMP15]], 4
-; AVX-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 4
-; AVX-NEXT:    store i32 [[TMP16]], i32* [[TMP13]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 18
-; AVX-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP20:%.*]] = add i32 [[TMP19]], 1
-; AVX-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
-; AVX-NEXT:    store i32 [[TMP20]], i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
-; AVX-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP24:%.*]] = add i32 [[TMP23]], 2
-; AVX-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
-; AVX-NEXT:    store i32 [[TMP24]], i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
-; AVX-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP28:%.*]] = add i32 [[TMP27]], 3
-; AVX-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
-; AVX-NEXT:    store i32 [[TMP28]], i32* [[TMP25]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
-; AVX-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP32:%.*]] = add i32 [[TMP31]], 4
-; AVX-NEXT:    store i32 [[TMP32]], i32* [[TMP29]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
+; AVX-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP18:%.*]] = insertelement <8 x i32> poison, i32 [[TMP3]], i32 0
+; AVX-NEXT:    [[TMP19:%.*]] = insertelement <8 x i32> [[TMP18]], i32 [[TMP5]], i32 1
+; AVX-NEXT:    [[TMP20:%.*]] = insertelement <8 x i32> [[TMP19]], i32 [[TMP7]], i32 2
+; AVX-NEXT:    [[TMP21:%.*]] = insertelement <8 x i32> [[TMP20]], i32 [[TMP9]], i32 3
+; AVX-NEXT:    [[TMP22:%.*]] = insertelement <8 x i32> [[TMP21]], i32 [[TMP11]], i32 4
+; AVX-NEXT:    [[TMP23:%.*]] = insertelement <8 x i32> [[TMP22]], i32 [[TMP13]], i32 5
+; AVX-NEXT:    [[TMP24:%.*]] = insertelement <8 x i32> [[TMP23]], i32 [[TMP15]], i32 6
+; AVX-NEXT:    [[TMP25:%.*]] = insertelement <8 x i32> [[TMP24]], i32 [[TMP17]], i32 7
+; AVX-NEXT:    [[TMP26:%.*]] = add <8 x i32> [[TMP25]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
+; AVX-NEXT:    [[TMP27:%.*]] = bitcast i32* [[TMP0:%.*]] to <8 x i32>*
+; AVX-NEXT:    store <8 x i32> [[TMP26]], <8 x i32>* [[TMP27]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    ret void
 ;
 ; AVX2-LABEL: @gather_load_3(
 ; AVX2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], 1
-; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
-; AVX2-NEXT:    store i32 [[TMP4]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1]], i32 0
-; AVX2-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32*> [[TMP6]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP8:%.*]] = getelementptr i32, <4 x i32*> [[TMP7]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
-; AVX2-NEXT:    [[TMP9:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP8]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[TMP9]], <i32 2, i32 3, i32 4, i32 1>
-; AVX2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
-; AVX2-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
-; AVX2-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
-; AVX2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 2
-; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
-; AVX2-NEXT:    store i32 [[TMP15]], i32* [[TMP11]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
-; AVX2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP19:%.*]] = add i32 [[TMP18]], 3
-; AVX2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
-; AVX2-NEXT:    store i32 [[TMP19]], i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
-; AVX2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP23:%.*]] = add i32 [[TMP22]], 4
-; AVX2-NEXT:    store i32 [[TMP23]], i32* [[TMP20]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
+; AVX2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
+; AVX2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 15
+; AVX2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 18
+; AVX2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
+; AVX2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
+; AVX2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
+; AVX2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP18:%.*]] = insertelement <8 x i32> poison, i32 [[TMP3]], i32 0
+; AVX2-NEXT:    [[TMP19:%.*]] = insertelement <8 x i32> [[TMP18]], i32 [[TMP5]], i32 1
+; AVX2-NEXT:    [[TMP20:%.*]] = insertelement <8 x i32> [[TMP19]], i32 [[TMP7]], i32 2
+; AVX2-NEXT:    [[TMP21:%.*]] = insertelement <8 x i32> [[TMP20]], i32 [[TMP9]], i32 3
+; AVX2-NEXT:    [[TMP22:%.*]] = insertelement <8 x i32> [[TMP21]], i32 [[TMP11]], i32 4
+; AVX2-NEXT:    [[TMP23:%.*]] = insertelement <8 x i32> [[TMP22]], i32 [[TMP13]], i32 5
+; AVX2-NEXT:    [[TMP24:%.*]] = insertelement <8 x i32> [[TMP23]], i32 [[TMP15]], i32 6
+; AVX2-NEXT:    [[TMP25:%.*]] = insertelement <8 x i32> [[TMP24]], i32 [[TMP17]], i32 7
+; AVX2-NEXT:    [[TMP26:%.*]] = add <8 x i32> [[TMP25]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
+; AVX2-NEXT:    [[TMP27:%.*]] = bitcast i32* [[TMP0:%.*]] to <8 x i32>*
+; AVX2-NEXT:    store <8 x i32> [[TMP26]], <8 x i32>* [[TMP27]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    ret void
 ;
-; AVX512-LABEL: @gather_load_3(
-; AVX512-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], 1
-; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
-; AVX512-NEXT:    store i32 [[TMP4]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1]], i32 0
-; AVX512-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32*> [[TMP6]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP8:%.*]] = getelementptr i32, <4 x i32*> [[TMP7]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
-; AVX512-NEXT:    [[TMP9:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP8]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[TMP9]], <i32 2, i32 3, i32 4, i32 1>
-; AVX512-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
-; AVX512-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
-; AVX512-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
-; AVX512-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 2
-; AVX512-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
-; AVX512-NEXT:    store i32 [[TMP15]], i32* [[TMP11]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
-; AVX512-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP19:%.*]] = add i32 [[TMP18]], 3
-; AVX512-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
-; AVX512-NEXT:    store i32 [[TMP19]], i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
-; AVX512-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP23:%.*]] = add i32 [[TMP22]], 4
-; AVX512-NEXT:    store i32 [[TMP23]], i32* [[TMP20]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    ret void
+; AVX512F-LABEL: @gather_load_3(
+; AVX512F-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 11
+; AVX512F-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
+; AVX512F-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 15
+; AVX512F-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP10:%.*]] = insertelement <4 x i32> poison, i32 [[TMP3]], i32 0
+; AVX512F-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> [[TMP10]], i32 [[TMP5]], i32 1
+; AVX512F-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP7]], i32 2
+; AVX512F-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP9]], i32 3
+; AVX512F-NEXT:    [[TMP14:%.*]] = add <4 x i32> [[TMP13]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 4
+; AVX512F-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP0]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP14]], <4 x i32>* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 18
+; AVX512F-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
+; AVX512F-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
+; AVX512F-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
+; AVX512F-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP25:%.*]] = insertelement <4 x i32> poison, i32 [[TMP18]], i32 0
+; AVX512F-NEXT:    [[TMP26:%.*]] = insertelement <4 x i32> [[TMP25]], i32 [[TMP20]], i32 1
+; AVX512F-NEXT:    [[TMP27:%.*]] = insertelement <4 x i32> [[TMP26]], i32 [[TMP22]], i32 2
+; AVX512F-NEXT:    [[TMP28:%.*]] = insertelement <4 x i32> [[TMP27]], i32 [[TMP24]], i32 3
+; AVX512F-NEXT:    [[TMP29:%.*]] = add <4 x i32> [[TMP28]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP30:%.*]] = bitcast i32* [[TMP15]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP29]], <4 x i32>* [[TMP30]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    ret void
+;
+; AVX512VL-LABEL: @gather_load_3(
+; AVX512VL-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP4:%.*]] = add i32 [[TMP3]], 1
+; AVX512VL-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 1
+; AVX512VL-NEXT:    store i32 [[TMP4]], i32* [[TMP0]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32*> poison, i32* [[TMP1]], i32 0
+; AVX512VL-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32*> [[TMP6]], <4 x i32*> poison, <4 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP8:%.*]] = getelementptr i32, <4 x i32*> [[TMP7]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
+; AVX512VL-NEXT:    [[TMP9:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP8]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[TMP9]], <i32 2, i32 3, i32 4, i32 1>
+; AVX512VL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
+; AVX512VL-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
+; AVX512VL-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 9
+; AVX512VL-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 2
+; AVX512VL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
+; AVX512VL-NEXT:    store i32 [[TMP15]], i32* [[TMP11]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 6
+; AVX512VL-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP19:%.*]] = add i32 [[TMP18]], 3
+; AVX512VL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
+; AVX512VL-NEXT:    store i32 [[TMP19]], i32* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 21
+; AVX512VL-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP23:%.*]] = add i32 [[TMP22]], 4
+; AVX512VL-NEXT:    store i32 [[TMP23]], i32* [[TMP20]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    ret void
 ;
   %3 = load i32, i32* %1, align 4, !tbaa !2
   %4 = add i32 %3, 1
   %5 = getelementptr inbounds i32, i32* %0, i64 1
   store i32 %4, i32* %0, align 4, !tbaa !2
   %6 = getelementptr inbounds i32, i32* %1, i64 11
   %7 = load i32, i32* %6, align 4, !tbaa !2
   %8 = add i32 %7, 2

 ; SSE-NEXT:    store i32 [[T16]], i32* [[T13]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    store i32 [[T20]], i32* [[T17]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @gather_load_4(
-; AVX-NEXT:    [[T5:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 1
 ; AVX-NEXT:    [[T6:%.*]] = getelementptr inbounds i32, i32* [[T1:%.*]], i64 11
-; AVX-NEXT:    [[T9:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 2
 ; AVX-NEXT:    [[T10:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 4
-; AVX-NEXT:    [[T13:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 3
 ; AVX-NEXT:    [[T14:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 15
-; AVX-NEXT:    [[T17:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 4
 ; AVX-NEXT:    [[T18:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 18
-; AVX-NEXT:    [[T21:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 5
 ; AVX-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
-; AVX-NEXT:    [[T25:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 6
 ; AVX-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
-; AVX-NEXT:    [[T29:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 7
 ; AVX-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
 ; AVX-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T7:%.*]] = load i32, i32* [[T6]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T11:%.*]] = load i32, i32* [[T10]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T15:%.*]] = load i32, i32* [[T14]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T19:%.*]] = load i32, i32* [[T18]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    [[T4:%.*]] = add i32 [[T3]], 1
-; AVX-NEXT:    [[T8:%.*]] = add i32 [[T7]], 2
-; AVX-NEXT:    [[T12:%.*]] = add i32 [[T11]], 3
-; AVX-NEXT:    [[T16:%.*]] = add i32 [[T15]], 4
-; AVX-NEXT:    [[T20:%.*]] = add i32 [[T19]], 1
-; AVX-NEXT:    [[T24:%.*]] = add i32 [[T23]], 2
-; AVX-NEXT:    [[T28:%.*]] = add i32 [[T27]], 3
-; AVX-NEXT:    [[T32:%.*]] = add i32 [[T31]], 4
-; AVX-NEXT:    store i32 [[T4]], i32* [[T0]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T8]], i32* [[T5]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T12]], i32* [[T9]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T16]], i32* [[T13]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T20]], i32* [[T17]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
-; AVX-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP1:%.*]] = insertelement <8 x i32> poison, i32 [[T3]], i32 0
+; AVX-NEXT:    [[TMP2:%.*]] = insertelement <8 x i32> [[TMP1]], i32 [[T7]], i32 1
+; AVX-NEXT:    [[TMP3:%.*]] = insertelement <8 x i32> [[TMP2]], i32 [[T11]], i32 2
+; AVX-NEXT:    [[TMP4:%.*]] = insertelement <8 x i32> [[TMP3]], i32 [[T15]], i32 3
+; AVX-NEXT:    [[TMP5:%.*]] = insertelement <8 x i32> [[TMP4]], i32 [[T19]], i32 4
+; AVX-NEXT:    [[TMP6:%.*]] = insertelement <8 x i32> [[TMP5]], i32 [[T23]], i32 5
+; AVX-NEXT:    [[TMP7:%.*]] = insertelement <8 x i32> [[TMP6]], i32 [[T27]], i32 6
+; AVX-NEXT:    [[TMP8:%.*]] = insertelement <8 x i32> [[TMP7]], i32 [[T31]], i32 7
+; AVX-NEXT:    [[TMP9:%.*]] = add <8 x i32> [[TMP8]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
+; AVX-NEXT:    [[TMP10:%.*]] = bitcast i32* [[T0:%.*]] to <8 x i32>*
+; AVX-NEXT:    store <8 x i32> [[TMP9]], <8 x i32>* [[TMP10]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    ret void
 ;
 ; AVX2-LABEL: @gather_load_4(
-; AVX2-NEXT:    [[T5:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 1
-; AVX2-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32*> poison, i32* [[T1:%.*]], i32 0
-; AVX2-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32*> [[TMP1]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr i32, <4 x i32*> [[TMP2]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
-; AVX2-NEXT:    [[T21:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 5
+; AVX2-NEXT:    [[T6:%.*]] = getelementptr inbounds i32, i32* [[T1:%.*]], i64 11
+; AVX2-NEXT:    [[T10:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 4
+; AVX2-NEXT:    [[T14:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 15
+; AVX2-NEXT:    [[T18:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 18
 ; AVX2-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
-; AVX2-NEXT:    [[T25:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 6
 ; AVX2-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
-; AVX2-NEXT:    [[T29:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 7
 ; AVX2-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
 ; AVX2-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP3]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[T7:%.*]] = load i32, i32* [[T6]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[T11:%.*]] = load i32, i32* [[T10]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[T15:%.*]] = load i32, i32* [[T14]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[T19:%.*]] = load i32, i32* [[T18]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[T4:%.*]] = add i32 [[T3]], 1
-; AVX2-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], <i32 2, i32 3, i32 4, i32 1>
-; AVX2-NEXT:    [[T24:%.*]] = add i32 [[T23]], 2
-; AVX2-NEXT:    [[T28:%.*]] = add i32 [[T27]], 3
-; AVX2-NEXT:    [[T32:%.*]] = add i32 [[T31]], 4
-; AVX2-NEXT:    store i32 [[T4]], i32* [[T0]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP6:%.*]] = bitcast i32* [[T5]] to <4 x i32>*
-; AVX2-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
-; AVX2-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP1:%.*]] = insertelement <8 x i32> poison, i32 [[T3]], i32 0
+; AVX2-NEXT:    [[TMP2:%.*]] = insertelement <8 x i32> [[TMP1]], i32 [[T7]], i32 1
+; AVX2-NEXT:    [[TMP3:%.*]] = insertelement <8 x i32> [[TMP2]], i32 [[T11]], i32 2
+; AVX2-NEXT:    [[TMP4:%.*]] = insertelement <8 x i32> [[TMP3]], i32 [[T15]], i32 3
+; AVX2-NEXT:    [[TMP5:%.*]] = insertelement <8 x i32> [[TMP4]], i32 [[T19]], i32 4
+; AVX2-NEXT:    [[TMP6:%.*]] = insertelement <8 x i32> [[TMP5]], i32 [[T23]], i32 5
+; AVX2-NEXT:    [[TMP7:%.*]] = insertelement <8 x i32> [[TMP6]], i32 [[T27]], i32 6
+; AVX2-NEXT:    [[TMP8:%.*]] = insertelement <8 x i32> [[TMP7]], i32 [[T31]], i32 7
+; AVX2-NEXT:    [[TMP9:%.*]] = add <8 x i32> [[TMP8]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
+; AVX2-NEXT:    [[TMP10:%.*]] = bitcast i32* [[T0:%.*]] to <8 x i32>*
+; AVX2-NEXT:    store <8 x i32> [[TMP9]], <8 x i32>* [[TMP10]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    ret void
 ;
-; AVX512-LABEL: @gather_load_4(
-; AVX512-NEXT:    [[T5:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 1
-; AVX512-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32*> poison, i32* [[T1:%.*]], i32 0
-; AVX512-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32*> [[TMP1]], <4 x i32*> poison, <4 x i32> zeroinitializer
-; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr i32, <4 x i32*> [[TMP2]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
-; AVX512-NEXT:    [[T21:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 5
-; AVX512-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
-; AVX512-NEXT:    [[T25:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 6
-; AVX512-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
-; AVX512-NEXT:    [[T29:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 7
-; AVX512-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
-; AVX512-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP3]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[T4:%.*]] = add i32 [[T3]], 1
-; AVX512-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], <i32 2, i32 3, i32 4, i32 1>
-; AVX512-NEXT:    [[T24:%.*]] = add i32 [[T23]], 2
-; AVX512-NEXT:    [[T28:%.*]] = add i32 [[T27]], 3
-; AVX512-NEXT:    [[T32:%.*]] = add i32 [[T31]], 4
-; AVX512-NEXT:    store i32 [[T4]], i32* [[T0]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    [[TMP6:%.*]] = bitcast i32* [[T5]] to <4 x i32>*
-; AVX512-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
-; AVX512-NEXT:    ret void
+; AVX512F-LABEL: @gather_load_4(
+; AVX512F-NEXT:    [[T6:%.*]] = getelementptr inbounds i32, i32* [[T1:%.*]], i64 11
+; AVX512F-NEXT:    [[T10:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 4
+; AVX512F-NEXT:    [[T14:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 15
+; AVX512F-NEXT:    [[T17:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 4
+; AVX512F-NEXT:    [[T18:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 18
+; AVX512F-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
+; AVX512F-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
+; AVX512F-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
+; AVX512F-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T7:%.*]] = load i32, i32* [[T6]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T11:%.*]] = load i32, i32* [[T10]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T15:%.*]] = load i32, i32* [[T14]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T19:%.*]] = load i32, i32* [[T18]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32> poison, i32 [[T3]], i32 0
+; AVX512F-NEXT:    [[TMP2:%.*]] = insertelement <4 x i32> [[TMP1]], i32 [[T7]], i32 1
+; AVX512F-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[T11]], i32 2
+; AVX512F-NEXT:    [[TMP4:%.*]] = insertelement <4 x i32> [[TMP3]], i32 [[T15]], i32 3
+; AVX512F-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32> poison, i32 [[T19]], i32 0
+; AVX512F-NEXT:    [[TMP7:%.*]] = insertelement <4 x i32> [[TMP6]], i32 [[T23]], i32 1
+; AVX512F-NEXT:    [[TMP8:%.*]] = insertelement <4 x i32> [[TMP7]], i32 [[T27]], i32 2
+; AVX512F-NEXT:    [[TMP9:%.*]] = insertelement <4 x i32> [[TMP8]], i32 [[T31]], i32 3
+; AVX512F-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[TMP9]], <i32 1, i32 2, i32 3, i32 4>
+; AVX512F-NEXT:    [[TMP11:%.*]] = bitcast i32* [[T0]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP11]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP12:%.*]] = bitcast i32* [[T17]] to <4 x i32>*
+; AVX512F-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    ret void
+;
+; AVX512VL-LABEL: @gather_load_4(
+; AVX512VL-NEXT:    [[T5:%.*]] = getelementptr inbounds i32, i32* [[T0:%.*]], i64 1
+; AVX512VL-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32*> poison, i32* [[T1:%.*]], i32 0
+; AVX512VL-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32*> [[TMP1]], <4 x i32*> poison, <4 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP3:%.*]] = getelementptr i32, <4 x i32*> [[TMP2]], <4 x i64> <i64 11, i64 4, i64 15, i64 18>
+; AVX512VL-NEXT:    [[T21:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 5
+; AVX512VL-NEXT:    [[T22:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 9
+; AVX512VL-NEXT:    [[T25:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 6
+; AVX512VL-NEXT:    [[T26:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 6
+; AVX512VL-NEXT:    [[T29:%.*]] = getelementptr inbounds i32, i32* [[T0]], i64 7
+; AVX512VL-NEXT:    [[T30:%.*]] = getelementptr inbounds i32, i32* [[T1]], i64 21
+; AVX512VL-NEXT:    [[T3:%.*]] = load i32, i32* [[T1]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> [[TMP3]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[T23:%.*]] = load i32, i32* [[T22]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[T27:%.*]] = load i32, i32* [[T26]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[T31:%.*]] = load i32, i32* [[T30]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[T4:%.*]] = add i32 [[T3]], 1
+; AVX512VL-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], <i32 2, i32 3, i32 4, i32 1>
+; AVX512VL-NEXT:    [[T24:%.*]] = add i32 [[T23]], 2
+; AVX512VL-NEXT:    [[T28:%.*]] = add i32 [[T27]], 3
+; AVX512VL-NEXT:    [[T32:%.*]] = add i32 [[T31]], 4
+; AVX512VL-NEXT:    store i32 [[T4]], i32* [[T0]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP6:%.*]] = bitcast i32* [[T5]] to <4 x i32>*
+; AVX512VL-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    store i32 [[T24]], i32* [[T21]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    store i32 [[T28]], i32* [[T25]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    store i32 [[T32]], i32* [[T29]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    ret void
 ;
   %t5 = getelementptr inbounds i32, i32* %t0, i64 1
   %t6 = getelementptr inbounds i32, i32* %t1, i64 11
   %t9 = getelementptr inbounds i32, i32* %t0, i64 2
   %t10 = getelementptr inbounds i32, i32* %t1, i64 4
   %t13 = getelementptr inbounds i32, i32* %t0, i64 3
   %t14 = getelementptr inbounds i32, i32* %t1, i64 15
   %t17 = getelementptr inbounds i32, i32* %t0, i64 4
;
   store i32 %t32, i32* %t29, align 4, !tbaa !2
 
   ret void
 }
 
 
 define void @gather_load_div(float* noalias nocapture %0, float* noalias nocapture readonly %1) {
 ; SSE-LABEL: @gather_load_div(
-; SSE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
-; SSE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 3
-; SSE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 14
-; SSE-NEXT:    [[TMP6:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
-; SSE-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> [[TMP6]], float* [[TMP3]], i32 1
-; SSE-NEXT:    [[TMP8:%.*]] = insertelement <4 x float*> [[TMP7]], float* [[TMP4]], i32 2
-; SSE-NEXT:    [[TMP9:%.*]] = insertelement <4 x float*> [[TMP8]], float* [[TMP5]], i32 3
-; SSE-NEXT:    [[TMP10:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> [[TMP9]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef), !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP11:%.*]] = shufflevector <4 x float*> [[TMP6]], <4 x float*> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:    [[TMP12:%.*]] = getelementptr float, <4 x float*> [[TMP11]], <4 x i64> <i64 4, i64 13, i64 11, i64 44>
-; SSE-NEXT:    [[TMP13:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> [[TMP12]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef), !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP14:%.*]] = fdiv <4 x float> [[TMP10]], [[TMP13]]
-; SSE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, float* [[TMP0:%.*]], i64 4
-; SSE-NEXT:    [[TMP16:%.*]] = bitcast float* [[TMP0]] to <4 x float>*
-; SSE-NEXT:    store <4 x float> [[TMP14]], <4 x float>* [[TMP16]], align 4, !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP17:%.*]] = getelementptr float, <4 x float*> [[TMP11]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
-; SSE-NEXT:    [[TMP18:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> [[TMP17]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef), !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP19:%.*]] = getelementptr float, <4 x float*> [[TMP11]], <4 x i64> <i64 33, i64 30, i64 27, i64 23>
-; SSE-NEXT:    [[TMP20:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> [[TMP19]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef), !tbaa [[TBAA0]]
-; SSE-NEXT:    [[TMP21:%.*]] = fdiv <4 x float> [[TMP18]], [[TMP20]]
-; SSE-NEXT:    [[TMP22:%.*]] = bitcast float* [[TMP15]] to <4 x float>*
-; SSE-NEXT:    store <4 x float> [[TMP21]], <4 x float>* [[TMP22]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 4
+; SSE-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 10
+; SSE-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 13
+; SSE-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 3
+; SSE-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP10]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 11
+; SSE-NEXT:    [[TMP13:%.*]] = load float, float* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 14
+; SSE-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 44
+; SSE-NEXT:    [[TMP17:%.*]] = load float, float* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP18:%.*]] = insertelement <4 x float> poison, float [[TMP3]], i32 0
+; SSE-NEXT:    [[TMP19:%.*]] = insertelement <4 x float> [[TMP18]], float [[TMP7]], i32 1
+; SSE-NEXT:    [[TMP20:%.*]] = insertelement <4 x float> [[TMP19]], float [[TMP11]], i32 2
+; SSE-NEXT:    [[TMP21:%.*]] = insertelement <4 x float> [[TMP20]], float [[TMP15]], i32 3
+; SSE-NEXT:    [[TMP22:%.*]] = insertelement <4 x float> poison, float [[TMP5]], i32 0
+; SSE-NEXT:    [[TMP23:%.*]] = insertelement <4 x float> [[TMP22]], float [[TMP9]], i32 1
+; SSE-NEXT:    [[TMP24:%.*]] = insertelement <4 x float> [[TMP23]], float [[TMP13]], i32 2
+; SSE-NEXT:    [[TMP25:%.*]] = insertelement <4 x float> [[TMP24]], float [[TMP17]], i32 3
+; SSE-NEXT:    [[TMP26:%.*]] = fdiv <4 x float> [[TMP21]], [[TMP25]]
+; SSE-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, float* [[TMP0:%.*]], i64 4
+; SSE-NEXT:    [[TMP28:%.*]] = bitcast float* [[TMP0]] to <4 x float>*
+; SSE-NEXT:    store <4 x float> [[TMP26]], <4 x float>* [[TMP28]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 17
+; SSE-NEXT:    [[TMP30:%.*]] = load float, float* [[TMP29]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP31:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 33
+; SSE-NEXT:    [[TMP32:%.*]] = load float, float* [[TMP31]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 8
+; SSE-NEXT:    [[TMP34:%.*]] = load float, float* [[TMP33]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 30
+; SSE-NEXT:    [[TMP36:%.*]] = load float, float* [[TMP35]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 5
+; SSE-NEXT:    [[TMP38:%.*]] = load float, float* [[TMP37]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 27
+; SSE-NEXT:    [[TMP40:%.*]] = load float, float* [[TMP39]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 20
+; SSE-NEXT:    [[TMP42:%.*]] = load float, float* [[TMP41]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP43:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 23
+; SSE-NEXT:    [[TMP44:%.*]] = load float, float* [[TMP43]], align 4, !tbaa [[TBAA0]]
+; SSE-NEXT:    [[TMP45:%.*]] = insertelement <4 x float> poison, float [[TMP30]], i32 0
+; SSE-NEXT:    [[TMP46:%.*]] = insertelement <4 x float> [[TMP45]], float [[TMP34]], i32 1
+; SSE-NEXT:    [[TMP47:%.*]] = insertelement <4 x float> [[TMP46]], float [[TMP38]], i32 2
+; SSE-NEXT:    [[TMP48:%.*]] = insertelement <4 x float> [[TMP47]], float [[TMP42]], i32 3
+; SSE-NEXT:    [[TMP49:%.*]] = insertelement <4 x float> poison, float [[TMP32]], i32 0
+; SSE-NEXT:    [[TMP50:%.*]] = insertelement <4 x float> [[TMP49]], float [[TMP36]], i32 1
+; SSE-NEXT:    [[TMP51:%.*]] = insertelement <4 x float> [[TMP50]], float [[TMP40]], i32 2
+; SSE-NEXT:    [[TMP52:%.*]] = insertelement <4 x float> [[TMP51]], float [[TMP44]], i32 3
+; SSE-NEXT:    [[TMP53:%.*]] = fdiv <4 x float> [[TMP48]], [[TMP52]]
+; SSE-NEXT:    [[TMP54:%.*]] = bitcast float* [[TMP27]] to <4 x float>*
+; SSE-NEXT:    store <4 x float> [[TMP53]], <4 x float>* [[TMP54]], align 4, !tbaa [[TBAA0]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @gather_load_div(
-; AVX-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
-; AVX-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
-; AVX-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
-; AVX-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
-; AVX-NEXT:    [[TMP7:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 17, i64 8>
-; AVX-NEXT:    [[TMP8:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 5, i64 20>
-; AVX-NEXT:    [[TMP9:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
-; AVX-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> [[TMP9]], float* [[TMP3]], i32 1
-; AVX-NEXT:    [[TMP11:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP12:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> [[TMP11]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP13:%.*]] = shufflevector <2 x float*> [[TMP7]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP14:%.*]] = shufflevector <8 x float*> [[TMP12]], <8 x float*> [[TMP13]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP15:%.*]] = shufflevector <2 x float*> [[TMP8]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[TMP16:%.*]] = shufflevector <8 x float*> [[TMP14]], <8 x float*> [[TMP15]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 8, i32 9>
-; AVX-NEXT:    [[TMP17:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP16]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP18:%.*]] = shufflevector <8 x float*> [[TMP9]], <8 x float*> poison, <8 x i32> zeroinitializer
-; AVX-NEXT:    [[TMP19:%.*]] = getelementptr float, <8 x float*> [[TMP18]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
-; AVX-NEXT:    [[TMP20:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP19]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX-NEXT:    [[TMP21:%.*]] = fdiv <8 x float> [[TMP17]], [[TMP20]]
-; AVX-NEXT:    [[TMP22:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
-; AVX-NEXT:    store <8 x float> [[TMP21]], <8 x float>* [[TMP22]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 4
+; AVX-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 10
+; AVX-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 13
+; AVX-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 3
+; AVX-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP10]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 11
+; AVX-NEXT:    [[TMP13:%.*]] = load float, float* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 14
+; AVX-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 44
+; AVX-NEXT:    [[TMP17:%.*]] = load float, float* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 17
+; AVX-NEXT:    [[TMP19:%.*]] = load float, float* [[TMP18]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 33
+; AVX-NEXT:    [[TMP21:%.*]] = load float, float* [[TMP20]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 8
+; AVX-NEXT:    [[TMP23:%.*]] = load float, float* [[TMP22]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 30
+; AVX-NEXT:    [[TMP25:%.*]] = load float, float* [[TMP24]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 5
+; AVX-NEXT:    [[TMP27:%.*]] = load float, float* [[TMP26]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 27
+; AVX-NEXT:    [[TMP29:%.*]] = load float, float* [[TMP28]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 20
+; AVX-NEXT:    [[TMP31:%.*]] = load float, float* [[TMP30]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 23
+; AVX-NEXT:    [[TMP33:%.*]] = load float, float* [[TMP32]], align 4, !tbaa [[TBAA0]]
+; AVX-NEXT:    [[TMP34:%.*]] = insertelement <8 x float> poison, float [[TMP3]], i32 0
+; AVX-NEXT:    [[TMP35:%.*]] = insertelement <8 x float> [[TMP34]], float [[TMP7]], i32 1
+; AVX-NEXT:    [[TMP36:%.*]] = insertelement <8 x float> [[TMP35]], float [[TMP11]], i32 2
+; AVX-NEXT:    [[TMP37:%.*]] = insertelement <8 x float> [[TMP36]], float [[TMP15]], i32 3
+; AVX-NEXT:    [[TMP38:%.*]] = insertelement <8 x float> [[TMP37]], float [[TMP19]], i32 4
+; AVX-NEXT:    [[TMP39:%.*]] = insertelement <8 x float> [[TMP38]], float [[TMP23]], i32 5
+; AVX-NEXT:    [[TMP40:%.*]] = insertelement <8 x float> [[TMP39]], float [[TMP27]], i32 6
+; AVX-NEXT:    [[TMP41:%.*]] = insertelement <8 x float> [[TMP40]], float [[TMP31]], i32 7
+; AVX-NEXT:    [[TMP42:%.*]] = insertelement <8 x float> poison, float [[TMP5]], i32 0
+; AVX-NEXT:    [[TMP43:%.*]] = insertelement <8 x float> [[TMP42]], float [[TMP9]], i32 1
+; AVX-NEXT:    [[TMP44:%.*]] = insertelement <8 x float> [[TMP43]], float [[TMP13]], i32 2
+; AVX-NEXT:    [[TMP45:%.*]] = insertelement <8 x float> [[TMP44]], float [[TMP17]], i32 3
+; AVX-NEXT:    [[TMP46:%.*]] = insertelement <8 x float> [[TMP45]], float [[TMP21]], i32 4
+; AVX-NEXT:    [[TMP47:%.*]] = insertelement <8 x float> [[TMP46]], float [[TMP25]], i32 5
+; AVX-NEXT:    [[TMP48:%.*]] = insertelement <8 x float> [[TMP47]], float [[TMP29]], i32 6
+; AVX-NEXT:    [[TMP49:%.*]] = insertelement <8 x float> [[TMP48]], float [[TMP33]], i32 7
+; AVX-NEXT:    [[TMP50:%.*]] = fdiv <8 x float> [[TMP41]], [[TMP49]]
+; AVX-NEXT:    [[TMP51:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
+; AVX-NEXT:    store <8 x float> [[TMP50]], <8 x float>* [[TMP51]], align 4, !tbaa [[TBAA0]]
 ; AVX-NEXT:    ret void
 ;
 ; AVX2-LABEL: @gather_load_div(
-; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
-; AVX2-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
-; AVX2-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
-; AVX2-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
-; AVX2-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float*> [[TMP7]], <4 x float*> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP9:%.*]] = getelementptr float, <4 x float*> [[TMP8]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
-; AVX2-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
-; AVX2-NEXT:    [[TMP11:%.*]] = insertelement <8 x float*> [[TMP10]], float* [[TMP3]], i32 1
-; AVX2-NEXT:    [[TMP12:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX2-NEXT:    [[TMP13:%.*]] = shufflevector <8 x float*> [[TMP11]], <8 x float*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX2-NEXT:    [[TMP14:%.*]] = shufflevector <4 x float*> [[TMP9]], <4 x float*> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
-; AVX2-NEXT:    [[TMP15:%.*]] = shufflevector <8 x float*> [[TMP13]], <8 x float*> [[TMP14]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
-; AVX2-NEXT:    [[TMP16:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP15]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP17:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> poison, <8 x i32> zeroinitializer
-; AVX2-NEXT:    [[TMP18:%.*]] = getelementptr float, <8 x float*> [[TMP17]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
-; AVX2-NEXT:    [[TMP19:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP18]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
-; AVX2-NEXT:    [[TMP20:%.*]] = fdiv <8 x float> [[TMP16]], [[TMP19]]
-; AVX2-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
-; AVX2-NEXT:    store <8 x float> [[TMP20]], <8 x float>* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP1:%.*]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 4
+; AVX2-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP4]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 10
+; AVX2-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP6]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 13
+; AVX2-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP8]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 3
+; AVX2-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP10]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 11
+; AVX2-NEXT:    [[TMP13:%.*]] = load float, float* [[TMP12]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 14
+; AVX2-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP14]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 44
+; AVX2-NEXT:    [[TMP17:%.*]] = load float, float* [[TMP16]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 17
+; AVX2-NEXT:    [[TMP19:%.*]] = load float, float* [[TMP18]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 33
+; AVX2-NEXT:    [[TMP21:%.*]] = load float, float* [[TMP20]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 8
+; AVX2-NEXT:    [[TMP23:%.*]] = load float, float* [[TMP22]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 30
+; AVX2-NEXT:    [[TMP25:%.*]] = load float, float* [[TMP24]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 5
+; AVX2-NEXT:    [[TMP27:%.*]] = load float, float* [[TMP26]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 27
+; AVX2-NEXT:    [[TMP29:%.*]] = load float, float* [[TMP28]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 20
+; AVX2-NEXT:    [[TMP31:%.*]] = load float, float* [[TMP30]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP1]], i64 23
+; AVX2-NEXT:    [[TMP33:%.*]] = load float, float* [[TMP32]], align 4, !tbaa [[TBAA0]]
+; AVX2-NEXT:    [[TMP34:%.*]] = insertelement <8 x float> poison, float [[TMP3]], i32 0
+; AVX2-NEXT:    [[TMP35:%.*]] = insertelement <8 x float> [[TMP34]], float [[TMP7]], i32 1
+; AVX2-NEXT:    [[TMP36:%.*]] = insertelement <8 x float> [[TMP35]], float [[TMP11]], i32 2
+; AVX2-NEXT:    [[TMP37:%.*]] = insertelement <8 x float> [[TMP36]], float [[TMP15]], i32 3
+; AVX2-NEXT:    [[TMP38:%.*]] = insertelement <8 x float> [[TMP37]], float [[TMP19]], i32 4
+; AVX2-NEXT:    [[TMP39:%.*]] = insertelement <8 x float> [[TMP38]], float [[TMP23]], i32 5
+; AVX2-NEXT:    [[TMP40:%.*]] = insertelement <8 x float> [[TMP39]], float [[TMP27]], i32 6
+; AVX2-NEXT:    [[TMP41:%.*]] = insertelement <8 x float> [[TMP40]], float [[TMP31]], i32 7
+; AVX2-NEXT:    [[TMP42:%.*]] = insertelement <8 x float> poison, float [[TMP5]], i32 0
+; AVX2-NEXT:    [[TMP43:%.*]] = insertelement <8 x float> [[TMP42]], float [[TMP9]], i32 1
+; AVX2-NEXT:    [[TMP44:%.*]] = insertelement <8 x float> [[TMP43]], float [[TMP13]], i32 2
+; AVX2-NEXT:    [[TMP45:%.*]] = insertelement <8 x float> [[TMP44]], float [[TMP17]], i32 3
+; AVX2-NEXT:    [[TMP46:%.*]] = insertelement <8 x float> [[TMP45]], float [[TMP21]], i32 4
+; AVX2-NEXT:    [[TMP47:%.*]] = insertelement <8 x float> [[TMP46]], float [[TMP25]], i32 5
+; AVX2-NEXT:    [[TMP48:%.*]] = insertelement <8 x float> [[TMP47]], float [[TMP29]], i32 6
+; AVX2-NEXT:    [[TMP49:%.*]] = insertelement <8 x float> [[TMP48]], float [[TMP33]], i32 7
+; AVX2-NEXT:    [[TMP50:%.*]] = fdiv <8 x float> [[TMP41]], [[TMP49]]
+; AVX2-NEXT:    [[TMP51:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
+; AVX2-NEXT:    store <8 x float> [[TMP50]], <8 x float>* [[TMP51]], align 4, !tbaa [[TBAA0]]
 ; AVX2-NEXT:    ret void
 ;
+; AVX512F-LABEL: @gather_load_div(
+; AVX512F-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
+; AVX512F-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
+; AVX512F-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
+; AVX512F-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
+; AVX512F-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
+; AVX512F-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float*> [[TMP7]], <4 x float*> poison, <4 x i32> zeroinitializer
+; AVX512F-NEXT:    [[TMP9:%.*]] = getelementptr float, <4 x float*> [[TMP8]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
+; AVX512F-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
+; AVX512F-NEXT:    [[TMP11:%.*]] = insertelement <8 x float*> [[TMP10]], float* [[TMP3]], i32 1
+; AVX512F-NEXT:    [[TMP12:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512F-NEXT:    [[TMP13:%.*]] = shufflevector <8 x float*> [[TMP11]], <8 x float*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512F-NEXT:    [[TMP14:%.*]] = shufflevector <4 x float*> [[TMP9]], <4 x float*> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512F-NEXT:    [[TMP15:%.*]] = shufflevector <8 x float*> [[TMP13]], <8 x float*> [[TMP14]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:    [[TMP16:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP15]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP17:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> poison, <8 x i32> zeroinitializer
+; AVX512F-NEXT:    [[TMP18:%.*]] = getelementptr float, <8 x float*> [[TMP17]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
+; AVX512F-NEXT:    [[TMP19:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP18]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
+; AVX512F-NEXT:    [[TMP20:%.*]] = fdiv <8 x float> [[TMP16]], [[TMP19]]
+; AVX512F-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
+; AVX512F-NEXT:    store <8 x float> [[TMP20]], <8 x float>* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX512F-NEXT:    ret void
+;
+; AVX512VL-LABEL: @gather_load_div(
+; AVX512VL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
+; AVX512VL-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
+; AVX512VL-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
+; AVX512VL-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
+; AVX512VL-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float*> [[TMP7]], <4 x float*> poison, <4 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP9:%.*]] = getelementptr float, <4 x float*> [[TMP8]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
+; AVX512VL-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
+; AVX512VL-NEXT:    [[TMP11:%.*]] = insertelement <8 x float*> [[TMP10]], float* [[TMP3]], i32 1
+; AVX512VL-NEXT:    [[TMP12:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512VL-NEXT:    [[TMP13:%.*]] = shufflevector <8 x float*> [[TMP11]], <8 x float*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512VL-NEXT:    [[TMP14:%.*]] = shufflevector <4 x float*> [[TMP9]], <4 x float*> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+; AVX512VL-NEXT:    [[TMP15:%.*]] = shufflevector <8 x float*> [[TMP13]], <8 x float*> [[TMP14]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+; AVX512VL-NEXT:    [[TMP16:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP15]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP17:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> poison, <8 x i32> zeroinitializer
+; AVX512VL-NEXT:    [[TMP18:%.*]] = getelementptr float, <8 x float*> [[TMP17]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
+; AVX512VL-NEXT:    [[TMP19:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP18]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    [[TMP20:%.*]] = fdiv <8 x float> [[TMP16]], [[TMP19]]
+; AVX512VL-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
+; AVX512VL-NEXT:    store <8 x float> [[TMP20]], <8 x float>* [[TMP21]], align 4, !tbaa [[TBAA0]]
+; AVX512VL-NEXT:    ret void
+;
 ; AVX512-LABEL: @gather_load_div(
 ; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP1:%.*]], i64 10
 ; AVX512-NEXT:    [[TMP4:%.*]] = insertelement <2 x float*> poison, float* [[TMP1]], i32 0
 ; AVX512-NEXT:    [[TMP5:%.*]] = shufflevector <2 x float*> [[TMP4]], <2 x float*> poison, <2 x i32> zeroinitializer
 ; AVX512-NEXT:    [[TMP6:%.*]] = getelementptr float, <2 x float*> [[TMP5]], <2 x i64> <i64 3, i64 14>
 ; AVX512-NEXT:    [[TMP7:%.*]] = insertelement <4 x float*> poison, float* [[TMP1]], i32 0
 ; AVX512-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float*> [[TMP7]], <4 x float*> poison, <4 x i32> zeroinitializer
 ; AVX512-NEXT:    [[TMP9:%.*]] = getelementptr float, <4 x float*> [[TMP8]], <4 x i64> <i64 17, i64 8, i64 5, i64 20>
 ; AVX512-NEXT:    [[TMP10:%.*]] = insertelement <8 x float*> poison, float* [[TMP1]], i32 0
 ; AVX512-NEXT:    [[TMP11:%.*]] = insertelement <8 x float*> [[TMP10]], float* [[TMP3]], i32 1
 ; AVX512-NEXT:    [[TMP12:%.*]] = shufflevector <2 x float*> [[TMP6]], <2 x float*> poison, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; AVX512-NEXT:    [[TMP13:%.*]] = shufflevector <8 x float*> [[TMP11]], <8 x float*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 undef, i32 undef, i32 undef, i32 undef>
 ; AVX512-NEXT:    [[TMP14:%.*]] = shufflevector <4 x float*> [[TMP9]], <4 x float*> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
 ; AVX512-NEXT:    [[TMP15:%.*]] = shufflevector <8 x float*> [[TMP13]], <8 x float*> [[TMP14]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
 ; AVX512-NEXT:    [[TMP16:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP15]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
 ; AVX512-NEXT:    [[TMP17:%.*]] = shufflevector <8 x float*> [[TMP10]], <8 x float*> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:    [[TMP18:%.*]] = getelementptr float, <8 x float*> [[TMP17]], <8 x i64> <i64 4, i64 13, i64 11, i64 44, i64 33, i64 30, i64 27, i64 23>
 ; AVX512-NEXT:    [[TMP19:%.*]] = call <8 x float> @llvm.masked.gather.v8f32.v8p0f32(<8 x float*> [[TMP18]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x float> undef), !tbaa [[TBAA0]]
 ; AVX512-NEXT:    [[TMP20:%.*]] = fdiv <8 x float> [[TMP16]], [[TMP19]]
 ; AVX512-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP0:%.*]] to <8 x float>*
 ; AVX512-NEXT:    store <8 x float> [[TMP20]], <8 x float>* [[TMP21]], align 4, !tbaa [[TBAA0]]
 ; AVX512-NEXT:    ret void
-;
   %3 = load float, float* %1, align 4, !tbaa !2
   %4 = getelementptr inbounds float, float* %1, i64 4
   %5 = load float, float* %4, align 4, !tbaa !2
   %6 = fdiv float %3, %5
   %7 = getelementptr inbounds float, float* %0, i64 1
   store float %6, float* %0, align 4, !tbaa !2
   %8 = getelementptr inbounds float, float* %1, i64 10
   %9 = load float, float* %8, align 4, !tbaa !2

llvm/test/Transforms/SLPVectorizer/X86/stores-non-ordered.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -S -mtriple=x86_64-unknown -slp-vectorizer -slp-max-store-lookup=2 -slp-min-reg-size=64 -slp-threshold=-1000 | FileCheck %s
 
 define i32 @non-ordered-stores(i32* noalias nocapture %in, i32* noalias nocapture %inn, i32* noalias nocapture %out) {
 ; CHECK-LABEL: @non-ordered-stores(
 ; CHECK-NEXT:    [[IN_ADDR:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i64 0
 ; CHECK-NEXT:    [[LOAD_1:%.*]] = load i32, i32* [[IN_ADDR]], align 4
+; CHECK-NEXT:    [[GEP_1:%.*]] = getelementptr inbounds i32, i32* [[IN_ADDR]], i64 1
+; CHECK-NEXT:    [[LOAD_2:%.*]] = load i32, i32* [[GEP_1]], align 4
 ; CHECK-NEXT:    [[GEP_2:%.*]] = getelementptr inbounds i32, i32* [[IN_ADDR]], i64 2
 ; CHECK-NEXT:    [[LOAD_3:%.*]] = load i32, i32* [[GEP_2]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32*> poison, i32* [[IN_ADDR]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x i32*> [[TMP1]], i32* [[IN_ADDR]], i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i32, <2 x i32*> [[TMP2]], <2 x i64> <i64 1, i64 3>
-; CHECK-NEXT:    [[TMP4:%.*]] = call <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*> [[TMP3]], i32 4, <2 x i1> <i1 true, i1 true>, <2 x i32> undef)
+; CHECK-NEXT:    [[GEP_3:%.*]] = getelementptr inbounds i32, i32* [[IN_ADDR]], i64 3
+; CHECK-NEXT:    [[LOAD_4:%.*]] = load i32, i32* [[GEP_3]], align 4
 ; CHECK-NEXT:    [[INN_ADDR:%.*]] = getelementptr inbounds i32, i32* [[INN:%.*]], i64 0
 ; CHECK-NEXT:    [[LOAD_5:%.*]] = load i32, i32* [[INN_ADDR]], align 4
+; CHECK-NEXT:    [[GEP_4:%.*]] = getelementptr inbounds i32, i32* [[INN_ADDR]], i64 1
+; CHECK-NEXT:    [[LOAD_6:%.*]] = load i32, i32* [[GEP_4]], align 4
 ; CHECK-NEXT:    [[GEP_5:%.*]] = getelementptr inbounds i32, i32* [[INN_ADDR]], i64 2
 ; CHECK-NEXT:    [[LOAD_7:%.*]] = load i32, i32* [[GEP_5]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x i32*> poison, i32* [[INN_ADDR]], i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x i32*> [[TMP5]], i32* [[INN_ADDR]], i32 1
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i32, <2 x i32*> [[TMP6]], <2 x i64> <i64 1, i64 3>
-; CHECK-NEXT:    [[TMP8:%.*]] = call <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*> [[TMP7]], i32 4, <2 x i1> <i1 true, i1 true>, <2 x i32> undef)
+; CHECK-NEXT:    [[GEP_6:%.*]] = getelementptr inbounds i32, i32* [[INN_ADDR]], i64 3
+; CHECK-NEXT:    [[LOAD_8:%.*]] = load i32, i32* [[GEP_6]], align 4
 ; CHECK-NEXT:    [[MUL_1:%.*]] = mul i32 [[LOAD_1]], [[LOAD_5]]
 ; CHECK-NEXT:    [[MUL_3:%.*]] = mul i32 [[LOAD_3]], [[LOAD_7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = mul <2 x i32> [[TMP4]], [[TMP8]]
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> poison, i32 [[LOAD_2]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x i32> [[TMP1]], i32 [[LOAD_4]], i32 1
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i32> poison, i32 [[LOAD_6]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x i32> [[TMP3]], i32 [[LOAD_8]], i32 1
+; CHECK-NEXT:    [[TMP5:%.*]] = mul <2 x i32> [[TMP2]], [[TMP4]]
 ; CHECK-NEXT:    br label [[BLOCK1:%.*]]
 ; CHECK:       block1:
 ; CHECK-NEXT:    [[GEP_X:%.*]] = getelementptr inbounds i32, i32* [[INN_ADDR]], i64 5
 ; CHECK-NEXT:    [[LOAD_9:%.*]] = load i32, i32* [[GEP_X]], align 4
 ; CHECK-NEXT:    br label [[BLOCK2:%.*]]
 ; CHECK:       block2:
 ; CHECK-NEXT:    [[GEP_7:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i64 0
 ; CHECK-NEXT:    [[GEP_8:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 1
 ; CHECK-NEXT:    [[GEP_9:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 2
 ; CHECK-NEXT:    [[GEP_10:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 3
 ; CHECK-NEXT:    [[GEP_11:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 4
 ; CHECK-NEXT:    store i32 [[MUL_1]], i32* [[GEP_10]], align 4
 ; CHECK-NEXT:    store i32 [[LOAD_9]], i32* [[GEP_9]], align 4
 ; CHECK-NEXT:    store i32 [[MUL_3]], i32* [[GEP_11]], align 4
-; CHECK-NEXT:    [[TMP10:%.*]] = bitcast i32* [[GEP_7]] to <2 x i32>*
-; CHECK-NEXT:    store <2 x i32> [[TMP9]], <2 x i32>* [[TMP10]], align 4
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i32* [[GEP_7]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP5]], <2 x i32>* [[TMP6]], align 4
 ; CHECK-NEXT:    ret i32 undef
 ;
   %in.addr = getelementptr inbounds i32, i32* %in, i64 0
   %load.1 = load i32, i32* %in.addr, align 4
   %gep.1 = getelementptr inbounds i32, i32* %in.addr, i64 1
   %load.2 = load i32, i32* %gep.1, align 4
   %gep.2 = getelementptr inbounds i32, i32* %in.addr, i64 2
   %load.3 = load i32, i32* %gep.2, align 4