[VectorCombine] scalarizeBinop - support an all-constant src vector operand

[RKSimon]

Summary

scalarizeBinop currently folds

vec_bo((inselt VecC0, V0, Index), (inselt VecC1, V1, Index))
->
inselt(vec_bo(VecC0, VecC1), scl_bo(V0,V1), Index)

This patch extends this to account for cases where one of the vec_bo operands is already all-constant and performs similar cost checks to determine if the scalar binop with a constant still makes sense:

vec_bo((inselt VecC0, V0, Index), VecC1)
->
inselt(vec_bo(VecC0, VecC1), scl_bo(V0,extractelt(V1,Index)), Index)

@spatel Have you any suggestions on specific test coverage you'd expect? So far I've just added the PR42174 test case (would you prefer this in pr42174.ll?) and haven't pre-committed the test case until I've got some initial feedback.

Fixes PR42174

[spatel]

I put the PR42174 test into PhaseOrdering, so we can verify that this transform is not undone or done elsewhere (was originally in instcombine with D50992).
I also added tests from D50992, but that still needs work to make it more applicable to x86 cost modeling:
rGe31f2a894a7b

The 1 logic adjustment I'd make here (at least initially) is to add a pattern match to see if the scalar variable is being loaded directly from memory. In that case, we should bail out. A possible enhancement would adjust the cost calc to account for the fact that there is (probably?) no transfer from GPR to vector register in that case.

[spatel]

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

383

Should assert or give up if the Builder result is not a constant. I don't think there's any danger from constant expressions or missing constant folding, but fuzzers may find a way.

[RKSimon]

In D80885#2065299, @spatel wrote:

The 1 logic adjustment I'd make here (at least initially) is to add a pattern match to see if the scalar variable is being loaded directly from memory. In that case, we should bail out. A possible enhancement would adjust the cost calc to account for the fact that there is (probably?) no transfer from GPR to vector register in that case.

Thanks @spatel, I'll update later today. Do you mean bail just for the 1 insertion load case or always bail if any loads+insertions are present?

I haven't checked but adding an Instruction arg to getVectorInstrCost would allow us to get more specific costs for load+insert instructions - what would you say if I investigate that instead?

[spatel]

In D80885#2065326, @RKSimon wrote:

In D80885#2065299, @spatel wrote:

The 1 logic adjustment I'd make here (at least initially) is to add a pattern match to see if the scalar variable is being loaded directly from memory. In that case, we should bail out. A possible enhancement would adjust the cost calc to account for the fact that there is (probably?) no transfer from GPR to vector register in that case.

Thanks @spatel, I'll update later today. Do you mean bail just for the 1 insertion load case or always bail if any loads+insertions are present?

I was only imagining the 1 insertion case. If both operands are insertions, then it seems unlikely that the vector op would win...but I guess we can add tests for that too if we're going to enhance the cost model.

I haven't checked but adding an Instruction arg to getVectorInstrCost would allow us to get more specific costs for load+insert instructions - what would you say if I investigate that instead?

Sure, if we can build this into the cost model, that would be good. Not sure if the vectorizers would benefit too, but that can be another patch.

[RKSimon]

Bail for load+insertion for single op cases, or if we somehow fail to extract a constant a constant vector.

[RKSimon]

llvm/test/Transforms/VectorCombine/X86/insert-binop-with-constant.ll

684

Do we need this or can we just rely on the phaseordering test?

[spatel]

Code looks good, but let's pre-commit some changes to the tests to exercise the cost model.

1. It's mostly <2 x i64> or <2 x double> - can we create a a negative test or show an SSE vs AVX diff by changing types?
2. Throw in some extra uses and see if that makes cost changes that inhibit the transform.

llvm/test/Transforms/VectorCombine/X86/insert-binop-with-constant.ll

684

I don't see any value in duplicating this test. We'll know if anything breaks meaningfully from the PhaseOrdering version of the test.

[RKSimon]

Rebased and removed duplicate test - I still need to add missing test coverage suggested by @spatel

[RKSimon]

rebase with mul oneuse test case to show SSE vs AVX diff

[RKSimon]

rebase with shl oneuse test case to show SSE vs AVX diff

[lebedev.ri]

I'm having a hard time following scalarizeBinop() now.
Perhaps it can be refactored somehow?
At the very least i'd suggest bool LhsIsConst = !V0 and so on.

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

385

ConstantExpr::getExtractElement() ?

[RKSimon]

In D80885#2073312, @lebedev.ri wrote:

I'm having a hard time following scalarizeBinop() now.
Perhaps it can be refactored somehow?

I'm hesitant to start refactoring yet as an imminent change will be to add cmp support (PR37463) will require further changes that aren't clear yet. @spatel - any suggestions?

[spatel]

In D80885#2073650, @RKSimon wrote:

In D80885#2073312, @lebedev.ri wrote:

I'm having a hard time following scalarizeBinop() now.
Perhaps it can be refactored somehow?

I'm hesitant to start refactoring yet as an imminent change will be to add cmp support (PR37463) will require further changes that aren't clear yet. @spatel - any suggestions?

I haven't looked at how to cram cmp into this yet. As we can see, the existing code for extract/extract patterns ended up being rather big. Not sure how much of that we can avoid going in this direction.

I think this is a good implementation, but as suggested, adding some bool predicates for readability will help.

I had an early draft version of this patch where the constant operand case was just split off as a separate function, so that duplicated a lot of code and didn't seem worth the cost.

[RKSimon]

address feedback

[lebedev.ri]

SGTM, but please wait for @spatel.

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

385

not done

[spatel]

LGTM - with ConstantExpr and grammar nit fixed.

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

344

its -> it is

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

Value *Shuf =
       Builder.CreateShuffleVector(CastV, UndefValue::get(DestTy), NewMask);
   I.replaceAllUsesWith(Shuf);
   return true;
 }
 
 /// Match a vector binop instruction with inserted scalar operands and convert
 /// to scalar binop followed by insertelement.
 static bool scalarizeBinop(Instruction &I, const TargetTransformInfo &TTI) {
-  Instruction *Ins0, *Ins1;
-  if (!match(&I, m_BinOp(m_Instruction(Ins0), m_Instruction(Ins1))))
+  Value *Ins0, *Ins1;
+  if (!match(&I, m_BinOp(m_Value(Ins0), m_Value(Ins1))))
     return false;
 
+  // Match against one or both scalar values being inserted into constant
+  // vectors:
+  // vec_bo VecC0, (inselt VecC1, V1, Index)
+  // vec_bo (inselt VecC0, V0, Index), VecC1
+  // vec_bo (inselt VecC0, V0, Index), (inselt VecC1, V1, Index)
   // TODO: Deal with mismatched index constants and variable indexes?
   Constant *VecC0, *VecC1;
   Value *V0, *V1;
-  uint64_t Index;
+  uint64_t Index0, Index1;
   if (!match(Ins0, m_InsertElt(m_Constant(VecC0), m_Value(V0),
-                               m_ConstantInt(Index))) ||
-      !match(Ins1, m_InsertElt(m_Constant(VecC1), m_Value(V1),
-                               m_SpecificInt(Index))))
+                               m_ConstantInt(Index0)))) {
+    V0 = nullptr;
+    if (!match(Ins0, m_Constant(VecC0)))
+      return false;
+  }
+  if (!match(Ins1, m_InsertElt(m_Constant(VecC1), m_Value(V1),
+                               m_ConstantInt(Index1)))) {
+    V1 = nullptr;
+    if (!match(Ins1, m_Constant(VecC1)))
+      return false;
+  }
+  if (!V0 && !V1)
+    return false;
+  if (V0 && V1 && Index0 != Index1)
+    return false;
+
+  // Bail for single insertion if its a load.
+  // TODO: Handle this once getVectorInstrCost can cost for load/stores.
+  auto *I0 = dyn_cast_or_null<Instruction>(V0);
+  auto *I1 = dyn_cast_or_null<Instruction>(V1);
+  if ((!V0 && I1 && I1->mayReadFromMemory()) ||
+      (!V1 && I0 && I0->mayReadFromMemory()))
     return false;
 
-  Type *ScalarTy = V0->getType();
+  uint64_t Index = V0 ? Index0 : Index1;
+  Type *ScalarTy = V0 ? V0->getType() : V1->getType();
   Type *VecTy = I.getType();
-  assert(VecTy->isVectorTy() && ScalarTy == V1->getType() &&
+  assert(VecTy->isVectorTy() &&
+         (!V0 || !V1 || V0->getType() == V1->getType()) &&
          (ScalarTy->isIntegerTy() || ScalarTy->isFloatingPointTy()) &&
          "Unexpected types for insert into binop");
 
   Instruction::BinaryOps Opcode = cast<BinaryOperator>(&I)->getOpcode();
   int ScalarOpCost = TTI.getArithmeticInstrCost(Opcode, ScalarTy);
   int VectorOpCost = TTI.getArithmeticInstrCost(Opcode, VecTy);
 
   // Get cost estimate for the insert element. This cost will factor into
   // both sequences.
   int InsertCost =
       TTI.getVectorInstrCost(Instruction::InsertElement, VecTy, Index);
-  int OldCost = InsertCost + InsertCost + VectorOpCost;
+  int OldCost = (V0 ? InsertCost : 0) + (V1 ? InsertCost : 0) + VectorOpCost;
   int NewCost = ScalarOpCost + InsertCost +
-                !Ins0->hasOneUse() * InsertCost +
-                !Ins1->hasOneUse() * InsertCost;
+                (V0 ? !Ins0->hasOneUse() * InsertCost : 0) +
+                (V1 ? !Ins1->hasOneUse() * InsertCost : 0);
 
   // We want to scalarize unless the vector variant actually has lower cost.
   if (OldCost < NewCost)
     return false;
 
   // vec_bo (inselt VecC0, V0, Index), (inselt VecC1, V1, Index) -->
   // inselt NewVecC, (scalar_bo V0, V1), Index
   ++NumScalarBO;
   IRBuilder<> Builder(&I);
+
+  // For constant cases, extract the scalar element, this should constant fold.
+  if (!V0) {
+    V0 = Builder.CreateExtractElement(VecC0, Index);
+    if (!isa_and_nonnull<Constant>(V0))
+      return false;
+  }
+  if (!V1) {
+    V1 = Builder.CreateExtractElement(VecC1, Index);
+    if (!isa_and_nonnull<Constant>(V1))
+      return false;
+  }
+
   Value *Scalar = Builder.CreateBinOp(Opcode, V0, V1, I.getName() + ".scalar");
 
   // All IR flags are safe to back-propagate. There is no potential for extra
   // poison to be created by the scalar instruction.
   if (auto *ScalarInst = dyn_cast<Instruction>(Scalar))
     ScalarInst->copyIRFlags(&I);
 
   // Fold the vector constants in the original vectors into a new base vector.

llvm/test/Transforms/PhaseOrdering/X86/scalarization.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -O3                   -S < %s  | FileCheck %s
 ; RUN: opt -passes='default<O3>' -S < %s  | FileCheck %s
 
 target triple = "x86_64--"
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 
 ; PR42174 - https://bugs.llvm.org/show_bug.cgi?id=42174
 ; This test should match the IR produced by clang after running -mem2reg.
 ; All math before the final 'add' should be scalarized.
 
 define <4 x i32> @square(<4 x i32> %num, i32 %y, i32 %x, i32 %h, i32 %k, i32 %w, i32 %p, i32 %j, i32 %u) {
 ; CHECK-LABEL: @square(
 ; CHECK-NEXT:    [[DIV:%.*]] = sdiv i32 [[K:%.*]], 2
-; CHECK-NEXT:    [[SPLATINSERT:%.*]] = insertelement <4 x i32> undef, i32 [[DIV]], i32 0
 ; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[P:%.*]], 6234
-; CHECK-NEXT:    [[SPLATINSERT2:%.*]] = insertelement <4 x i32> undef, i32 [[MUL]], i32 0
 ; CHECK-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[H:%.*]], 75
-; CHECK-NEXT:    [[SPLATINSERT6:%.*]] = insertelement <4 x i32> undef, i32 [[MUL5]], i32 0
 ; CHECK-NEXT:    [[DIV9:%.*]] = sdiv i32 [[J:%.*]], 3452
-; CHECK-NEXT:    [[SPLATINSERT10:%.*]] = insertelement <4 x i32> undef, i32 [[DIV9]], i32 0
 ; CHECK-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[W:%.*]], 53
-; CHECK-NEXT:    [[SPLATINSERT14:%.*]] = insertelement <4 x i32> undef, i32 [[MUL13]], i32 0
 ; CHECK-NEXT:    [[DIV17:%.*]] = sdiv i32 [[X:%.*]], 820
-; CHECK-NEXT:    [[SPLATINSERT18:%.*]] = insertelement <4 x i32> undef, i32 [[DIV17]], i32 0
 ; CHECK-NEXT:    [[MUL21:%.*]] = shl nsw i32 [[U:%.*]], 2
-; CHECK-NEXT:    [[SPLATINSERT22:%.*]] = insertelement <4 x i32> undef, i32 [[MUL21]], i32 0
-; CHECK-NEXT:    [[SPLATINSERT25:%.*]] = insertelement <4 x i32> undef, i32 [[Y:%.*]], i32 0
-; CHECK-NEXT:    [[TMP1:%.*]] = add <4 x i32> [[SPLATINSERT25]], <i32 1, i32 undef, i32 undef, i32 undef>
-; CHECK-NEXT:    [[TMP2:%.*]] = add <4 x i32> [[TMP1]], [[SPLATINSERT18]]
-; CHECK-NEXT:    [[TMP3:%.*]] = add <4 x i32> [[TMP2]], [[SPLATINSERT6]]
-; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i32> [[TMP3]], [[SPLATINSERT]]
-; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], [[SPLATINSERT14]]
-; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i32> [[TMP5]], [[SPLATINSERT2]]
-; CHECK-NEXT:    [[TMP7:%.*]] = add <4 x i32> [[TMP6]], [[SPLATINSERT10]]
-; CHECK-NEXT:    [[TMP8:%.*]] = add <4 x i32> [[TMP7]], [[SPLATINSERT22]]
-; CHECK-NEXT:    [[TMP9:%.*]] = add <4 x i32> [[TMP8]], <i32 317425, i32 undef, i32 undef, i32 undef>
-; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <4 x i32> [[TMP9]], <4 x i32> undef, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[ADD29:%.*]] = add <4 x i32> [[TMP10]], [[NUM:%.*]]
+; CHECK-NEXT:    [[DOTSCALAR:%.*]] = add i32 [[Y:%.*]], 1
+; CHECK-NEXT:    [[DOTSCALAR1:%.*]] = add i32 [[DOTSCALAR]], [[DIV17]]
+; CHECK-NEXT:    [[DOTSCALAR2:%.*]] = add i32 [[DOTSCALAR1]], [[MUL5]]
+; CHECK-NEXT:    [[DOTSCALAR3:%.*]] = add i32 [[DOTSCALAR2]], [[DIV]]
+; CHECK-NEXT:    [[DOTSCALAR4:%.*]] = add i32 [[DOTSCALAR3]], [[MUL13]]
+; CHECK-NEXT:    [[DOTSCALAR5:%.*]] = add i32 [[DOTSCALAR4]], [[MUL]]
+; CHECK-NEXT:    [[DOTSCALAR6:%.*]] = add i32 [[DOTSCALAR5]], [[DIV9]]
+; CHECK-NEXT:    [[DOTSCALAR7:%.*]] = add i32 [[DOTSCALAR6]], [[MUL21]]
+; CHECK-NEXT:    [[DOTSCALAR8:%.*]] = add i32 [[DOTSCALAR7]], 317425
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32> undef, i32 [[DOTSCALAR8]], i64 0
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> undef, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[ADD29:%.*]] = add <4 x i32> [[TMP2]], [[NUM:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[ADD29]]
 ;
   %add = add <4 x i32> %num, <i32 1, i32 1, i32 1, i32 1>
   %div = sdiv i32 %k, 2
   %splatinsert = insertelement <4 x i32> undef, i32 %div, i32 0
   %splat = shufflevector <4 x i32> %splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
   %add1 = add <4 x i32> %add, %splat
   %mul = mul nsw i32 %p, 6234

llvm/test/Transforms/VectorCombine/X86/insert-binop-with-constant.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -vector-combine -S -mtriple=x86_64-- -mattr=SSE2 | FileCheck %s --check-prefixes=CHECK,SSE
 ; RUN: opt < %s -vector-combine -S -mtriple=x86_64-- -mattr=AVX2 | FileCheck %s --check-prefixes=CHECK,AVX
 
 define <2 x i64> @add_constant(i64 %x) {
 ; CHECK-LABEL: @add_constant(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = add <2 x i64> [[INS]], <i64 42, i64 undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = add i64 [[X:%.*]], 42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = add <2 x i64> %ins, <i64 42, i64 undef>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @add_constant_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @add_constant_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = add <2 x i64> [[INS]], <i64 42, i64 -42>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = add i64 [[X:%.*]], 42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = add <2 x i64> %ins, <i64 42, i64 -42>
   ret <2 x i64> %bo
 }
 
+define <2 x i64> @add_constant_load(i64* %p) {
+; CHECK-LABEL: @add_constant_load(
+; CHECK-NEXT:    [[LD:%.*]] = load i64, i64* [[P:%.*]], align 4
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[LD]], i32 0
+; CHECK-NEXT:    [[BO:%.*]] = add <2 x i64> [[INS]], <i64 42, i64 -42>
+; CHECK-NEXT:    ret <2 x i64> [[BO]]
+;
+  %ld = load i64, i64* %p
+  %ins = insertelement <2 x i64> undef, i64 %ld, i32 0
+  %bo = add <2 x i64> %ins, <i64 42, i64 -42>
+  ret <2 x i64> %bo
+}
+
 ; IR flags are not required, but they should propagate.
 
 define <4 x i32> @sub_constant_op0(i32 %x) {
 ; CHECK-LABEL: @sub_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <4 x i32> undef, i32 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = sub nuw nsw <4 x i32> <i32 undef, i32 -42, i32 undef, i32 undef>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = sub nuw nsw i32 -42, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <4 x i32> undef, i32 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <4 x i32> [[BO]]
 ;
   %ins = insertelement <4 x i32> undef, i32 %x, i32 1
   %bo = sub nsw nuw <4 x i32> <i32 undef, i32 -42, i32 undef, i32 undef>, %ins
   ret <4 x i32> %bo
 }
 
 define <4 x i32> @sub_constant_op0_not_undef_lane(i32 %x) {
 ; CHECK-LABEL: @sub_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <4 x i32> undef, i32 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = sub nuw <4 x i32> <i32 1, i32 42, i32 42, i32 -42>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = sub nuw i32 42, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <4 x i32> undef, i32 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <4 x i32> [[BO]]
 ;
   %ins = insertelement <4 x i32> undef, i32 %x, i32 1
   %bo = sub nuw <4 x i32> <i32 1, i32 42, i32 42, i32 -42>, %ins
   ret <4 x i32> %bo
 }
 
 define <8 x i16> @sub_constant_op1(i16 %x) {
 ; CHECK-LABEL: @sub_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <8 x i16> undef, i16 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = sub nuw <8 x i16> [[INS]], <i16 42, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = sub nuw i16 [[X:%.*]], 42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <8 x i16> undef, i16 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <8 x i16> [[BO]]
 ;
   %ins = insertelement <8 x i16> undef, i16 %x, i32 0
   %bo = sub nuw <8 x i16> %ins, <i16 42, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef, i16 undef>
   ret <8 x i16> %bo
 }
 
 define <8 x i16> @sub_constant_op1_not_undef_lane(i16 %x) {
 ; CHECK-LABEL: @sub_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <8 x i16> undef, i16 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = sub nuw <8 x i16> [[INS]], <i16 42, i16 -42, i16 0, i16 1, i16 -2, i16 3, i16 -4, i16 5>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = sub nuw i16 [[X:%.*]], 42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <8 x i16> undef, i16 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <8 x i16> [[BO]]
 ;
   %ins = insertelement <8 x i16> undef, i16 %x, i32 0
   %bo = sub nuw <8 x i16> %ins, <i16 42, i16 -42, i16 0, i16 1, i16 -2, i16 3, i16 -4, i16 5>
   ret <8 x i16> %bo
 }
 
 define <16 x i8> @mul_constant(i8 %x) {
 ; CHECK-LABEL: @mul_constant(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <16 x i8> undef, i8 [[X:%.*]], i32 2
-; CHECK-NEXT:    [[BO:%.*]] = mul <16 x i8> [[INS]], <i8 undef, i8 undef, i8 -42, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = mul i8 [[X:%.*]], -42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <16 x i8> <i8 undef, i8 undef, i8 0, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef>, i8 [[BO_SCALAR]], i64 2
 ; CHECK-NEXT:    ret <16 x i8> [[BO]]
 ;
   %ins = insertelement <16 x i8> undef, i8 %x, i32 2
   %bo = mul <16 x i8> %ins, <i8 undef, i8 undef, i8 -42, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef>
   ret <16 x i8> %bo
 }
 
 define <3 x i64> @mul_constant_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @mul_constant_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <3 x i64> undef, i64 [[X:%.*]], i32 2
-; CHECK-NEXT:    [[BO:%.*]] = mul <3 x i64> [[INS]], <i64 42, i64 undef, i64 -42>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = mul i64 [[X:%.*]], -42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <3 x i64> <i64 0, i64 undef, i64 0>, i64 [[BO_SCALAR]], i64 2
 ; CHECK-NEXT:    ret <3 x i64> [[BO]]
 ;
   %ins = insertelement <3 x i64> undef, i64 %x, i32 2
   %bo = mul <3 x i64> %ins, <i64 42, i64 undef, i64 -42>
   ret <3 x i64> %bo
 }
 
 define <2 x i64> @shl_constant_op0(i64 %x) {
 ; CHECK-LABEL: @shl_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = shl <2 x i64> <i64 undef, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = shl i64 2, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = shl <2 x i64> <i64 undef, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @shl_constant_op0_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @shl_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = shl <2 x i64> <i64 5, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = shl i64 2, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = shl <2 x i64> <i64 5, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
+define <2 x i64> @shl_constant_op0_load(i64* %p) {
+; CHECK-LABEL: @shl_constant_op0_load(
+; CHECK-NEXT:    [[LD:%.*]] = load i64, i64* [[P:%.*]], align 4
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[LD]], i32 1
+; CHECK-NEXT:    [[BO:%.*]] = shl <2 x i64> <i64 undef, i64 2>, [[INS]]
+; CHECK-NEXT:    ret <2 x i64> [[BO]]
+;
+  %ld = load i64, i64* %p
+  %ins = insertelement <2 x i64> undef, i64 %ld, i32 1
+  %bo = shl <2 x i64> <i64 undef, i64 2>, %ins
+  ret <2 x i64> %bo
+}
+
 define <2 x i64> @shl_constant_op1(i64 %x) {
 ; CHECK-LABEL: @shl_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = shl nuw <2 x i64> [[INS]], <i64 5, i64 undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = shl nuw i64 [[X:%.*]], 5
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 0, i64 undef>, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = shl nuw <2 x i64> %ins, <i64 5, i64 undef>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @shl_constant_op1_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @shl_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = shl nuw <2 x i64> [[INS]], <i64 5, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = shl nuw i64 [[X:%.*]], 5
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> zeroinitializer, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = shl nuw <2 x i64> %ins, <i64 5, i64 2>
   ret <2 x i64> %bo
 }
 
+define <2 x i64> @shl_constant_op1_load(i64* %p) {
+; CHECK-LABEL: @shl_constant_op1_load(
+; CHECK-NEXT:    [[LD:%.*]] = load i64, i64* [[P:%.*]], align 4
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[LD]], i32 0
+; CHECK-NEXT:    [[BO:%.*]] = shl nuw <2 x i64> [[INS]], <i64 5, i64 2>
+; CHECK-NEXT:    ret <2 x i64> [[BO]]
+;
+  %ld = load i64, i64* %p
+  %ins = insertelement <2 x i64> undef, i64 %ld, i32 0
+  %bo = shl nuw <2 x i64> %ins, <i64 5, i64 2>
+  ret <2 x i64> %bo
+}
+
 define <2 x i64> @ashr_constant_op0(i64 %x) {
 ; CHECK-LABEL: @ashr_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = ashr exact <2 x i64> <i64 undef, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = ashr exact i64 2, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = ashr exact <2 x i64> <i64 undef, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @ashr_constant_op0_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @ashr_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = ashr exact <2 x i64> <i64 5, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = ashr exact i64 2, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = ashr exact <2 x i64> <i64 5, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @ashr_constant_op1(i64 %x) {
 ; CHECK-LABEL: @ashr_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = ashr <2 x i64> [[INS]], <i64 5, i64 undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = ashr i64 [[X:%.*]], 5
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 0, i64 undef>, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = ashr <2 x i64> %ins, <i64 5, i64 undef>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @ashr_constant_op1_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @ashr_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = ashr <2 x i64> [[INS]], <i64 5, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = ashr i64 [[X:%.*]], 5
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> zeroinitializer, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = ashr <2 x i64> %ins, <i64 5, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @lshr_constant_op0(i64 %x) {
 ; CHECK-LABEL: @lshr_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = lshr <2 x i64> <i64 5, i64 undef>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = lshr i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = lshr <2 x i64> <i64 5, i64 undef>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @lshr_constant_op0_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @lshr_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = lshr <2 x i64> <i64 5, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = lshr i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = lshr <2 x i64> <i64 5, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @lshr_constant_op1(i64 %x) {
 ; CHECK-LABEL: @lshr_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = lshr exact <2 x i64> [[INS]], <i64 undef, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = lshr exact i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 undef, i64 0>, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = lshr exact <2 x i64> %ins, <i64 undef, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @lshr_constant_op1_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @lshr_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = lshr exact <2 x i64> [[INS]], <i64 5, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = lshr exact i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> zeroinitializer, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = lshr exact <2 x i64> %ins, <i64 5, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @urem_constant_op0(i64 %x) {
 ; CHECK-LABEL: @urem_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = urem <2 x i64> <i64 5, i64 undef>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = urem i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = urem <2 x i64> <i64 5, i64 undef>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @urem_constant_op0_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @urem_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = urem <2 x i64> <i64 5, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = urem i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = urem <2 x i64> <i64 5, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @urem_constant_op1(i64 %x) {
 ; CHECK-LABEL: @urem_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = urem <2 x i64> [[INS]], <i64 undef, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = urem i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 undef, i64 0>, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = urem <2 x i64> %ins, <i64 undef, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @urem_constant_op1_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @urem_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = urem <2 x i64> [[INS]], <i64 5, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = urem i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> zeroinitializer, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = urem <2 x i64> %ins, <i64 5, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @srem_constant_op0(i64 %x) {
 ; CHECK-LABEL: @srem_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = srem <2 x i64> <i64 5, i64 undef>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = srem i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = srem <2 x i64> <i64 5, i64 undef>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @srem_constant_op0_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @srem_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = srem <2 x i64> <i64 5, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = srem i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = srem <2 x i64> <i64 5, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @srem_constant_op1(i64 %x) {
 ; CHECK-LABEL: @srem_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = srem <2 x i64> [[INS]], <i64 undef, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = srem i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 undef, i64 0>, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = srem <2 x i64> %ins, <i64 undef, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @srem_constant_op1_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @srem_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = srem <2 x i64> [[INS]], <i64 5, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = srem i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> zeroinitializer, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = srem <2 x i64> %ins, <i64 5, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @udiv_constant_op0(i64 %x) {
 ; CHECK-LABEL: @udiv_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = udiv exact <2 x i64> <i64 5, i64 undef>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = udiv exact i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = udiv exact <2 x i64> <i64 5, i64 undef>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @udiv_constant_op0_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @udiv_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = udiv exact <2 x i64> <i64 5, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = udiv exact i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = udiv exact <2 x i64> <i64 5, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @udiv_constant_op1(i64 %x) {
 ; CHECK-LABEL: @udiv_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = udiv <2 x i64> [[INS]], <i64 undef, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = udiv i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 undef, i64 0>, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = udiv <2 x i64> %ins, <i64 undef, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @udiv_constant_op1_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @udiv_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = udiv <2 x i64> [[INS]], <i64 5, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = udiv i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> zeroinitializer, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = udiv <2 x i64> %ins, <i64 5, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @sdiv_constant_op0(i64 %x) {
 ; CHECK-LABEL: @sdiv_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = sdiv <2 x i64> <i64 5, i64 undef>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = sdiv i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = sdiv <2 x i64> <i64 5, i64 undef>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @sdiv_constant_op0_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @sdiv_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = sdiv <2 x i64> <i64 5, i64 2>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = sdiv i64 5, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = sdiv <2 x i64> <i64 5, i64 2>, %ins
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @sdiv_constant_op1(i64 %x) {
 ; CHECK-LABEL: @sdiv_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = sdiv exact <2 x i64> [[INS]], <i64 undef, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = sdiv exact i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 undef, i64 0>, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = sdiv exact <2 x i64> %ins, <i64 undef, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @sdiv_constant_op1_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @sdiv_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = sdiv exact <2 x i64> [[INS]], <i64 5, i64 2>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = sdiv exact i64 [[X:%.*]], 2
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> zeroinitializer, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = sdiv exact <2 x i64> %ins, <i64 5, i64 2>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @and_constant(i64 %x) {
 ; CHECK-LABEL: @and_constant(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = and <2 x i64> [[INS]], <i64 42, i64 undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = and i64 [[X:%.*]], 42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 0, i64 undef>, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = and <2 x i64> %ins, <i64 42, i64 undef>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @and_constant_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @and_constant_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = and <2 x i64> [[INS]], <i64 42, i64 -42>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = and i64 [[X:%.*]], 42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> zeroinitializer, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = and <2 x i64> %ins, <i64 42, i64 -42>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @or_constant(i64 %x) {
 ; CHECK-LABEL: @or_constant(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = or <2 x i64> [[INS]], <i64 undef, i64 -42>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = or i64 [[X:%.*]], -42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 undef, i64 -1>, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = or <2 x i64> %ins, <i64 undef, i64 -42>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @or_constant_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @or_constant_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = or <2 x i64> [[INS]], <i64 42, i64 -42>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = or i64 [[X:%.*]], -42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 -1, i64 -1>, i64 [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 1
   %bo = or <2 x i64> %ins, <i64 42, i64 -42>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @xor_constant(i64 %x) {
 ; CHECK-LABEL: @xor_constant(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = xor <2 x i64> [[INS]], <i64 42, i64 undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = xor i64 [[X:%.*]], 42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> <i64 undef, i64 0>, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = xor <2 x i64> %ins, <i64 42, i64 undef>
   ret <2 x i64> %bo
 }
 
 define <2 x i64> @xor_constant_not_undef_lane(i64 %x) {
 ; CHECK-LABEL: @xor_constant_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i64> undef, i64 [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = xor <2 x i64> [[INS]], <i64 42, i64 -42>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = xor i64 [[X:%.*]], 42
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x i64> undef, i64 [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x i64> [[BO]]
 ;
   %ins = insertelement <2 x i64> undef, i64 %x, i32 0
   %bo = xor <2 x i64> %ins, <i64 42, i64 -42>
   ret <2 x i64> %bo
 }
 
 define <2 x double> @fadd_constant(double %x) {
 ; CHECK-LABEL: @fadd_constant(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = fadd <2 x double> [[INS]], <double 4.200000e+01, double undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fadd double [[X:%.*]], 4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double undef>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = fadd <2 x double> %ins, <double 42.0, double undef>
   ret <2 x double> %bo
 }
 
 define <2 x double> @fadd_constant_not_undef_lane(double %x) {
 ; CHECK-LABEL: @fadd_constant_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = fadd <2 x double> [[INS]], <double 4.200000e+01, double -4.200000e+01>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fadd double [[X:%.*]], -4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 1
   %bo = fadd <2 x double> %ins, <double 42.0, double -42.0>
   ret <2 x double> %bo
 }
 
 define <2 x double> @fsub_constant_op0(double %x) {
 ; CHECK-LABEL: @fsub_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = fsub fast <2 x double> <double 4.200000e+01, double undef>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fsub fast double 4.200000e+01, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double undef>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = fsub fast <2 x double> <double 42.0, double undef>, %ins
   ret <2 x double> %bo
 }
 
 define <2 x double> @fsub_constant_op0_not_undef_lane(double %x) {
 ; CHECK-LABEL: @fsub_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = fsub nsz <2 x double> <double 4.200000e+01, double -4.200000e+01>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fsub nsz double -4.200000e+01, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 1
   %bo = fsub nsz <2 x double> <double 42.0, double -42.0>, %ins
   ret <2 x double> %bo
 }
 
 define <2 x double> @fsub_constant_op1(double %x) {
 ; CHECK-LABEL: @fsub_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = fsub <2 x double> [[INS]], <double undef, double 4.200000e+01>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fsub double [[X:%.*]], 4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double undef, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 1
   %bo = fsub <2 x double> %ins, <double undef, double 42.0>
   ret <2 x double> %bo
 }
 
 define <2 x double> @fsub_constant_op1_not_undef_lane(double %x) {
 ; CHECK-LABEL: @fsub_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = fsub <2 x double> [[INS]], <double 4.200000e+01, double -4.200000e+01>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fsub double [[X:%.*]], 4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = fsub <2 x double> %ins, <double 42.0, double -42.0>
   ret <2 x double> %bo
 }
 
 define <2 x double> @fmul_constant(double %x) {
 ; CHECK-LABEL: @fmul_constant(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = fmul reassoc <2 x double> [[INS]], <double 4.200000e+01, double undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fmul reassoc double [[X:%.*]], 4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double undef>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = fmul reassoc <2 x double> %ins, <double 42.0, double undef>
   ret <2 x double> %bo
 }
 
 define <2 x double> @fmul_constant_not_undef_lane(double %x) {
 ; CHECK-LABEL: @fmul_constant_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = fmul <2 x double> [[INS]], <double 4.200000e+01, double -4.200000e+01>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fmul double [[X:%.*]], -4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 1
   %bo = fmul <2 x double> %ins, <double 42.0, double -42.0>
   ret <2 x double> %bo
 }
 
 define <2 x double> @fdiv_constant_op0(double %x) {
 ; CHECK-LABEL: @fdiv_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = fdiv nnan <2 x double> <double undef, double 4.200000e+01>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fdiv nnan double 4.200000e+01, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double undef, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 1
   %bo = fdiv nnan <2 x double> <double undef, double 42.0>, %ins
   ret <2 x double> %bo
 }
 
 define <2 x double> @fdiv_constant_op0_not_undef_lane(double %x) {
 ; CHECK-LABEL: @fdiv_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = fdiv ninf <2 x double> <double 4.200000e+01, double -4.200000e+01>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fdiv ninf double 4.200000e+01, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = fdiv ninf <2 x double> <double 42.0, double -42.0>, %ins
   ret <2 x double> %bo
 }
 
 define <2 x double> @fdiv_constant_op1(double %x) {
 ; CHECK-LABEL: @fdiv_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = fdiv <2 x double> [[INS]], <double 4.200000e+01, double undef>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fdiv double [[X:%.*]], 4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double undef>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = fdiv <2 x double> %ins, <double 42.0, double undef>
   ret <2 x double> %bo
 }
 
 define <2 x double> @fdiv_constant_op1_not_undef_lane(double %x) {
 ; CHECK-LABEL: @fdiv_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = fdiv <2 x double> [[INS]], <double 4.200000e+01, double -4.200000e+01>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = fdiv double [[X:%.*]], 4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = fdiv <2 x double> %ins, <double 42.0, double -42.0>
   ret <2 x double> %bo
 }
 
 define <2 x double> @frem_constant_op0(double %x) {
 ; CHECK-LABEL: @frem_constant_op0(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = frem fast <2 x double> <double 4.200000e+01, double undef>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = frem fast double 4.200000e+01, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double undef>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = frem fast <2 x double> <double 42.0, double undef>, %ins
   ret <2 x double> %bo
 }
 
 define <2 x double> @frem_constant_op0_not_undef_lane(double %x) {
 ; CHECK-LABEL: @frem_constant_op0_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = frem <2 x double> <double 4.200000e+01, double -4.200000e+01>, [[INS]]
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = frem double -4.200000e+01, [[X:%.*]]
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 1
   %bo = frem <2 x double> <double 42.0, double -42.0>, %ins
   ret <2 x double> %bo
 }
 
 define <2 x double> @frem_constant_op1(double %x) {
 ; CHECK-LABEL: @frem_constant_op1(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 1
-; CHECK-NEXT:    [[BO:%.*]] = frem ninf <2 x double> [[INS]], <double undef, double 4.200000e+01>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = frem ninf double [[X:%.*]], 4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double undef, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 1
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 1
   %bo = frem ninf <2 x double> %ins, <double undef, double 42.0>
   ret <2 x double> %bo
 }
 
 define <2 x double> @frem_constant_op1_not_undef_lane(double %x) {
 ; CHECK-LABEL: @frem_constant_op1_not_undef_lane(
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x double> undef, double [[X:%.*]], i32 0
-; CHECK-NEXT:    [[BO:%.*]] = frem nnan <2 x double> [[INS]], <double 4.200000e+01, double -4.200000e+01>
+; CHECK-NEXT:    [[BO_SCALAR:%.*]] = frem nnan double [[X:%.*]], 4.200000e+01
+; CHECK-NEXT:    [[BO:%.*]] = insertelement <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>, double [[BO_SCALAR]], i64 0
 ; CHECK-NEXT:    ret <2 x double> [[BO]]
 ;
   %ins = insertelement <2 x double> undef, double %x, i32 0
   %bo = frem nnan <2 x double> %ins, <double 42.0, double -42.0>
   ret <2 x double> %bo
 }
+
+define <4 x i32> @PR42174(<4 x i32> %arg, i32 %arg1, i32 %arg2, i32 %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 %arg7, i32 %arg8) {
+; CHECK-LABEL: @PR42174(
+; CHECK-NEXT:    [[TMP:%.*]] = sdiv i32 [[ARG4:%.*]], 2
+; CHECK-NEXT:    [[TMP10:%.*]] = mul nsw i32 [[ARG6:%.*]], 6234
+; CHECK-NEXT:    [[TMP12:%.*]] = mul nsw i32 [[ARG3:%.*]], 75
+; CHECK-NEXT:    [[TMP14:%.*]] = sdiv i32 [[ARG7:%.*]], 3452
+; CHECK-NEXT:    [[TMP16:%.*]] = mul nsw i32 [[ARG5:%.*]], 53
+; CHECK-NEXT:    [[TMP18:%.*]] = sdiv i32 [[ARG2:%.*]], 820
+; CHECK-NEXT:    [[TMP20:%.*]] = shl nsw i32 [[ARG8:%.*]], 2
+; CHECK-NEXT:    [[TMP23_SCALAR:%.*]] = add i32 [[ARG1:%.*]], 1
+; CHECK-NEXT:    [[TMP24_SCALAR:%.*]] = add i32 [[TMP23_SCALAR]], [[TMP18]]
+; CHECK-NEXT:    [[TMP25_SCALAR:%.*]] = add i32 [[TMP24_SCALAR]], [[TMP12]]
+; CHECK-NEXT:    [[TMP26_SCALAR:%.*]] = add i32 [[TMP25_SCALAR]], [[TMP]]
+; CHECK-NEXT:    [[TMP27_SCALAR:%.*]] = add i32 [[TMP26_SCALAR]], [[TMP16]]
+; CHECK-NEXT:    [[TMP28_SCALAR:%.*]] = add i32 [[TMP27_SCALAR]], [[TMP10]]
+; CHECK-NEXT:    [[TMP29_SCALAR:%.*]] = add i32 [[TMP28_SCALAR]], [[TMP14]]
+; CHECK-NEXT:    [[TMP30_SCALAR:%.*]] = add i32 [[TMP29_SCALAR]], [[TMP20]]
+; CHECK-NEXT:    [[TMP31_SCALAR:%.*]] = add i32 [[TMP30_SCALAR]], 317425
+; CHECK-NEXT:    [[TMP31:%.*]] = insertelement <4 x i32> undef, i32 [[TMP31_SCALAR]], i64 0
+; CHECK-NEXT:    [[TMP32:%.*]] = shufflevector <4 x i32> [[TMP31]], <4 x i32> undef, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP33:%.*]] = add <4 x i32> [[TMP32]], [[ARG:%.*]]
+; CHECK-NEXT:    ret <4 x i32> [[TMP33]]
+;
+  %tmp = sdiv i32 %arg4, 2
+  %tmp9 = insertelement <4 x i32> undef, i32 %tmp, i32 0
+  %tmp10 = mul nsw i32 %arg6, 6234
+  %tmp11 = insertelement <4 x i32> undef, i32 %tmp10, i32 0
+  %tmp12 = mul nsw i32 %arg3, 75
+  %tmp13 = insertelement <4 x i32> undef, i32 %tmp12, i32 0
+  %tmp14 = sdiv i32 %arg7, 3452
+  %tmp15 = insertelement <4 x i32> undef, i32 %tmp14, i32 0
+  %tmp16 = mul nsw i32 %arg5, 53
+  %tmp17 = insertelement <4 x i32> undef, i32 %tmp16, i32 0
+  %tmp18 = sdiv i32 %arg2, 820
+  %tmp19 = insertelement <4 x i32> undef, i32 %tmp18, i32 0
+  %tmp20 = shl nsw i32 %arg8, 2
+  %tmp21 = insertelement <4 x i32> undef, i32 %tmp20, i32 0
+  %tmp22 = insertelement <4 x i32> undef, i32 %arg1, i32 0
+  %tmp23 = add <4 x i32> %tmp22, <i32 1, i32 undef, i32 undef, i32 undef>
+  %tmp24 = add <4 x i32> %tmp23, %tmp19
+  %tmp25 = add <4 x i32> %tmp24, %tmp13
+  %tmp26 = add <4 x i32> %tmp25, %tmp9
+  %tmp27 = add <4 x i32> %tmp26, %tmp17
+  %tmp28 = add <4 x i32> %tmp27, %tmp11
+  %tmp29 = add <4 x i32> %tmp28, %tmp15
+  %tmp30 = add <4 x i32> %tmp29, %tmp21
+  %tmp31 = add <4 x i32> %tmp30, <i32 317425, i32 undef, i32 undef, i32 undef>
+  %tmp32 = shufflevector <4 x i32> %tmp31, <4 x i32> undef, <4 x i32> zeroinitializer
+  %tmp33 = add <4 x i32> %tmp32, %arg
+  ret <4 x i32> %tmp33
+}