[X86] Add comment string for broadcast loads from the constant pool.

[craig.topper]

Summary

When broadcasting from the constant pool its useful to print out the final vector similar to what we do for normal moves from the constant pool.

I changed only a couple tests that were broadcast focused. One of them had been previously hand tweaked after running the script so that it could check the constant pool declaration. But I think this patch makes that unnecessary now since we can check the comment instead.

[spatel]

lib/Target/X86/X86MCInstLower.cpp

1863–1867 ↗

(On Diff #104980)

Could set NumElts above and then fall-through to avoid the switch-within-a-switch? Same for NumLanes below here.

1905–1909 ↗

(On Diff #104980)

Make a helper function, so we're not duplicating so much code?

test/CodeGen/X86/broadcast-elm-cross-splat-vec.ll

2 ↗

(On Diff #104980)

Extra/stale comment line.

[RKSimon]

test/CodeGen/X86/broadcast-elm-cross-splat-vec.ll

2039 ↗

(On Diff #104980)

Don't we have to be careful using NaNs in comments as different compilers/libs print them differently?

[craig.topper]

lib/Target/X86/X86MCInstLower.cpp

1863–1867 ↗

(On Diff #104980)

How do I fallthrough with different values for NumElts without using a goto?

[craig.topper]

Reduce code duplication.

I don't think we have to worry about NaN printing differences because we're going through APFloat's toString in this case which I think is stable.

The place we had NaN variation is when we go through raw_otream::operator<<(double) I think. Though that may have been fixed when an isnan check got added to llvm::write_double back in October. Previously I think we may have fallen into the platform's snprintf implementation.

[spatel]

LGTM.

lib/Target/X86/X86MCInstLower.cpp

1863–1867 ↗

(On Diff #104980)

Er, yes that would be difficult. :)
If there was a helper function for the remaining chunk of code, then we could pass NumElts to it as a param for each case and remove the switch-within-a-switch?

llvm/trunk/lib/Target/X86/X86MCInstLower.cpp

for (int i = 0, e = ShuffleMask.size(); i != e; ++i) {
     CS << ']';
     --i; // For loop increments element #.
   }
   CS.flush();
 
   return Comment;
 }
 
+static void printConstant(const Constant *COp, raw_ostream &CS) {
+  if (isa<UndefValue>(COp)) {
+    CS << "u";
+  } else if (auto *CI = dyn_cast<ConstantInt>(COp)) {
+    if (CI->getBitWidth() <= 64) {
+      CS << CI->getZExtValue();
+    } else {
+      // print multi-word constant as (w0,w1)
+      const auto &Val = CI->getValue();
+      CS << "(";
+      for (int i = 0, N = Val.getNumWords(); i < N; ++i) {
+        if (i > 0)
+          CS << ",";
+        CS << Val.getRawData()[i];
+      }
+      CS << ")";
+    }
+  } else if (auto *CF = dyn_cast<ConstantFP>(COp)) {
+    SmallString<32> Str;
+    CF->getValueAPF().toString(Str);
+    CS << Str;
+  } else {
+    CS << "?";
+  }
+}
+
 void X86AsmPrinter::EmitInstruction(const MachineInstr *MI) {
   X86MCInstLower MCInstLowering(*MF, *this);
   const X86RegisterInfo *RI = MF->getSubtarget<X86Subtarget>().getRegisterInfo();
 
   // Add a comment about EVEX-2-VEX compression for AVX-512 instrs that
   // are compressed from EVEX encoding to VEX encoding.
   if (TM.Options.MCOptions.ShowMCEncoding) {
     if (MI->getAsmPrinterFlags() & AC_EVEX_2_VEX)
#define CASE_ALL_MOV_RM()               \
   MOV_CASE(V, Y) /* AVX-256 */          \
   MOV_AVX512_CASE(Z)                    \
   MOV_AVX512_CASE(Z256)                 \
   MOV_AVX512_CASE(Z128)
 
   // For loads from a constant pool to a vector register, print the constant
   // loaded.
   CASE_ALL_MOV_RM()
+  case X86::VBROADCASTF128:
+  case X86::VBROADCASTI128:
+  case X86::VBROADCASTF32X4Z256rm:
+  case X86::VBROADCASTF32X4rm:
+  case X86::VBROADCASTF32X8rm:
+  case X86::VBROADCASTF64X2Z128rm:
+  case X86::VBROADCASTF64X2rm:
+  case X86::VBROADCASTF64X4rm:
+  case X86::VBROADCASTI32X4Z256rm:
+  case X86::VBROADCASTI32X4rm:
+  case X86::VBROADCASTI32X8rm:
+  case X86::VBROADCASTI64X2Z128rm:
+  case X86::VBROADCASTI64X2rm:
+  case X86::VBROADCASTI64X4rm:
     if (!OutStreamer->isVerboseAsm())
       break;
     if (MI->getNumOperands() <= 4)
       break;
     if (auto *C = getConstantFromPool(*MI, MI->getOperand(4))) {
+      int NumLanes = 1;
+      // Override NumLanes for the broadcast instructions.
+      switch (MI->getOpcode()) {
+      case X86::VBROADCASTF128:         NumLanes = 2;  break;
+      case X86::VBROADCASTI128:         NumLanes = 2;  break;
+      case X86::VBROADCASTF32X4Z256rm:  NumLanes = 2;  break;
+      case X86::VBROADCASTF32X4rm:      NumLanes = 4;  break;
+      case X86::VBROADCASTF32X8rm:      NumLanes = 2;  break;
+      case X86::VBROADCASTF64X2Z128rm:  NumLanes = 2;  break;
+      case X86::VBROADCASTF64X2rm:      NumLanes = 4;  break;
+      case X86::VBROADCASTF64X4rm:      NumLanes = 2;  break;
+      case X86::VBROADCASTI32X4Z256rm:  NumLanes = 2;  break;
+      case X86::VBROADCASTI32X4rm:      NumLanes = 4;  break;
+      case X86::VBROADCASTI32X8rm:      NumLanes = 2;  break;
+      case X86::VBROADCASTI64X2Z128rm:  NumLanes = 2;  break;
+      case X86::VBROADCASTI64X2rm:      NumLanes = 4;  break;
+      case X86::VBROADCASTI64X4rm:      NumLanes = 2;  break;
+      }
+
       std::string Comment;
       raw_string_ostream CS(Comment);
       const MachineOperand &DstOp = MI->getOperand(0);
       CS << X86ATTInstPrinter::getRegisterName(DstOp.getReg()) << " = ";
       if (auto *CDS = dyn_cast<ConstantDataSequential>(C)) {
         CS << "[";
+        for (int l = 0; l != NumLanes; ++l) {
           for (int i = 0, NumElements = CDS->getNumElements(); i < NumElements; ++i) {
-          if (i != 0)
+            if (i != 0 || l != 0)
               CS << ",";
             if (CDS->getElementType()->isIntegerTy())
               CS << CDS->getElementAsInteger(i);
             else if (CDS->getElementType()->isFloatTy())
               CS << CDS->getElementAsFloat(i);
             else if (CDS->getElementType()->isDoubleTy())
               CS << CDS->getElementAsDouble(i);
             else
               CS << "?";
           }
+        }
         CS << "]";
         OutStreamer->AddComment(CS.str(), !EnablePrintSchedInfo);
       } else if (auto *CV = dyn_cast<ConstantVector>(C)) {
         CS << "<";
+        for (int l = 0; l != NumLanes; ++l) {
           for (int i = 0, NumOperands = CV->getNumOperands(); i < NumOperands; ++i) {
-          if (i != 0)
+            if (i != 0 || l != 0)
               CS << ",";
-          Constant *COp = CV->getOperand(i);
-          if (isa<UndefValue>(COp)) {
-            CS << "u";
-          } else if (auto *CI = dyn_cast<ConstantInt>(COp)) {
-            if (CI->getBitWidth() <= 64) {
-              CS << CI->getZExtValue();
-            } else {
-              // print multi-word constant as (w0,w1)
-              const auto &Val = CI->getValue();
-              CS << "(";
-              for (int i = 0, N = Val.getNumWords(); i < N; ++i) {
-                if (i > 0)
-                  CS << ",";
-                CS << Val.getRawData()[i];
-              }
-              CS << ")";
-            }
-          } else if (auto *CF = dyn_cast<ConstantFP>(COp)) {
-            SmallString<32> Str;
-            CF->getValueAPF().toString(Str);
-            CS << Str;
-          } else {
-            CS << "?";
+            printConstant(CV->getOperand(i), CS);
           }
         }
         CS << ">";
         OutStreamer->AddComment(CS.str(), !EnablePrintSchedInfo);
       }
     }
     break;
+  case X86::VBROADCASTSSrm:
+  case X86::VBROADCASTSSYrm:
+  case X86::VBROADCASTSSZ128m:
+  case X86::VBROADCASTSSZ256m:
+  case X86::VBROADCASTSSZm:
+  case X86::VBROADCASTSDYrm:
+  case X86::VBROADCASTSDZ256m:
+  case X86::VBROADCASTSDZm:
+  case X86::VPBROADCASTBrm:
+  case X86::VPBROADCASTBYrm:
+  case X86::VPBROADCASTBZ128m:
+  case X86::VPBROADCASTBZ256m:
+  case X86::VPBROADCASTBZm:
+  case X86::VPBROADCASTDrm:
+  case X86::VPBROADCASTDYrm:
+  case X86::VPBROADCASTDZ128m:
+  case X86::VPBROADCASTDZ256m:
+  case X86::VPBROADCASTDZm:
+  case X86::VPBROADCASTQrm:
+  case X86::VPBROADCASTQYrm:
+  case X86::VPBROADCASTQZ128m:
+  case X86::VPBROADCASTQZ256m:
+  case X86::VPBROADCASTQZm:
+  case X86::VPBROADCASTWrm:
+  case X86::VPBROADCASTWYrm:
+  case X86::VPBROADCASTWZ128m:
+  case X86::VPBROADCASTWZ256m:
+  case X86::VPBROADCASTWZm:
+    if (!OutStreamer->isVerboseAsm())
+      break;
+    if (MI->getNumOperands() <= 4)
+      break;
+    if (auto *C = getConstantFromPool(*MI, MI->getOperand(4))) {
+      int NumElts;
+      switch (MI->getOpcode()) {
+      default: llvm_unreachable("Invalid opcode");
+      case X86::VBROADCASTSSrm:    NumElts = 4;  break;
+      case X86::VBROADCASTSSYrm:   NumElts = 8;  break;
+      case X86::VBROADCASTSSZ128m: NumElts = 4;  break;
+      case X86::VBROADCASTSSZ256m: NumElts = 8;  break;
+      case X86::VBROADCASTSSZm:    NumElts = 16; break;
+      case X86::VBROADCASTSDYrm:   NumElts = 4;  break;
+      case X86::VBROADCASTSDZ256m: NumElts = 4;  break;
+      case X86::VBROADCASTSDZm:    NumElts = 8;  break;
+      case X86::VPBROADCASTBrm:    NumElts = 16; break;
+      case X86::VPBROADCASTBYrm:   NumElts = 32; break;
+      case X86::VPBROADCASTBZ128m: NumElts = 16; break;
+      case X86::VPBROADCASTBZ256m: NumElts = 32; break;
+      case X86::VPBROADCASTBZm:    NumElts = 64; break;
+      case X86::VPBROADCASTDrm:    NumElts = 4;  break;
+      case X86::VPBROADCASTDYrm:   NumElts = 8;  break;
+      case X86::VPBROADCASTDZ128m: NumElts = 4;  break;
+      case X86::VPBROADCASTDZ256m: NumElts = 8;  break;
+      case X86::VPBROADCASTDZm:    NumElts = 16; break;
+      case X86::VPBROADCASTQrm:    NumElts = 2;  break;
+      case X86::VPBROADCASTQYrm:   NumElts = 4;  break;
+      case X86::VPBROADCASTQZ128m: NumElts = 2;  break;
+      case X86::VPBROADCASTQZ256m: NumElts = 4;  break;
+      case X86::VPBROADCASTQZm:    NumElts = 8;  break;
+      case X86::VPBROADCASTWrm:    NumElts = 8;  break;
+      case X86::VPBROADCASTWYrm:   NumElts = 16; break;
+      case X86::VPBROADCASTWZ128m: NumElts = 8;  break;
+      case X86::VPBROADCASTWZ256m: NumElts = 16; break;
+      case X86::VPBROADCASTWZm:    NumElts = 32; break;
+      }
+
+      std::string Comment;
+      raw_string_ostream CS(Comment);
+      const MachineOperand &DstOp = MI->getOperand(0);
+      CS << X86ATTInstPrinter::getRegisterName(DstOp.getReg()) << " = ";
+      CS << "[";
+      for (int i = 0; i != NumElts; ++i) {
+        if (i != 0)
+          CS << ",";
+        printConstant(C, CS);
+      }
+      CS << "]";
+      OutStreamer->AddComment(CS.str(), !EnablePrintSchedInfo);
+    }
   }
 
   MCInst TmpInst;
   MCInstLowering.Lower(MI, TmpInst);
 
   // Stackmap shadows cannot include branch targets, so we can count the bytes
   // in a call towards the shadow, but must ensure that the no thread returns
   // in to the stackmap shadow.  The only way to achieve this is if the call

llvm/trunk/test/CodeGen/X86/avg.ll

 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: avg_v64i8_const:
 ; AVX512F:       # BB#0:
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero,mem[8],zero,zero,zero,mem[9],zero,zero,zero,mem[10],zero,zero,zero,mem[11],zero,zero,zero,mem[12],zero,zero,zero,mem[13],zero,zero,zero,mem[14],zero,zero,zero,mem[15],zero,zero,zero
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero,mem[8],zero,zero,zero,mem[9],zero,zero,zero,mem[10],zero,zero,zero,mem[11],zero,zero,zero,mem[12],zero,zero,zero,mem[13],zero,zero,zero,mem[14],zero,zero,zero,mem[15],zero,zero,zero
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm2 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero,mem[8],zero,zero,zero,mem[9],zero,zero,zero,mem[10],zero,zero,zero,mem[11],zero,zero,zero,mem[12],zero,zero,zero,mem[13],zero,zero,zero,mem[14],zero,zero,zero,mem[15],zero,zero,zero
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm3 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero,mem[8],zero,zero,zero,mem[9],zero,zero,zero,mem[10],zero,zero,zero,mem[11],zero,zero,zero,mem[12],zero,zero,zero,mem[13],zero,zero,zero,mem[14],zero,zero,zero,mem[15],zero,zero,zero
-; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [1,2,3,4,5,6,7,8,1,2,3,4,5,6,7,8]
+; AVX512F-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
 ; AVX512F-NEXT:    vpaddd %zmm4, %zmm3, %zmm3
 ; AVX512F-NEXT:    vpaddd %zmm4, %zmm2, %zmm2
 ; AVX512F-NEXT:    vpaddd %zmm4, %zmm1, %zmm1
 ; AVX512F-NEXT:    vpaddd %zmm4, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpsrld $1, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpsrld $1, %zmm1, %zmm1
 ; AVX512F-NEXT:    vpsrld $1, %zmm2, %zmm2
 ; AVX512F-NEXT:    vpsrld $1, %zmm3, %zmm3

 ; AVX2-NEXT:    vmovdqu %ymm2, (%rax)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: avg_v32i16_const:
 ; AVX512F:       # BB#0:
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
-; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm2 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm2 = [1,2,3,4,5,6,7,8,1,2,3,4,5,6,7,8]
+; AVX512F-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3]
 ; AVX512F-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
 ; AVX512F-NEXT:    vpaddd %zmm2, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpsrld $1, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpsrld $1, %zmm1, %zmm1
 ; AVX512F-NEXT:    vpmovdw %zmm1, (%rax)
 ; AVX512F-NEXT:    vpmovdw %zmm0, (%rax)
 ; AVX512F-NEXT:    vzeroupper
 ; AVX512F-NEXT:    retq

llvm/trunk/test/CodeGen/X86/avx2-vbroadcast.ll

entry:
   %vecinit.i = insertelement <2 x double> undef, double %q, i32 0
   %vecinit2.i = insertelement <2 x double> %vecinit.i, double %q, i32 1
   ret <2 x double> %vecinit2.i
 }
 
 define <8 x i32> @V111(<8 x i32> %in) nounwind uwtable readnone ssp {
 ; X32-AVX2-LABEL: V111:
 ; X32-AVX2:       ## BB#0: ## %entry
-; X32-AVX2-NEXT:    vpbroadcastd LCPI29_0, %ymm1
+; X32-AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
 ; X32-AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
 ; X32-AVX2-NEXT:    retl
 ;
 ; X64-AVX2-LABEL: V111:
 ; X64-AVX2:       ## BB#0: ## %entry
-; X64-AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm1
+; X64-AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
 ; X64-AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
 ; X64-AVX2-NEXT:    retq
 ;
 ; X32-AVX512VL-LABEL: V111:
 ; X32-AVX512VL:       ## BB#0: ## %entry
 ; X32-AVX512VL-NEXT:    vpaddd LCPI29_0{1to8}, %ymm0, %ymm0
 ; X32-AVX512VL-NEXT:    retl
 ;
 ; X64-AVX512VL-LABEL: V111:
 ; X64-AVX512VL:       ## BB#0: ## %entry
 ; X64-AVX512VL-NEXT:    vpaddd {{.*}}(%rip){1to8}, %ymm0, %ymm0
 ; X64-AVX512VL-NEXT:    retq
 entry:
   %g = add <8 x i32> %in, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
   ret <8 x i32> %g
 }
 
 define <8 x float> @V113(<8 x float> %in) nounwind uwtable readnone ssp {
 ; X32-AVX2-LABEL: V113:
 ; X32-AVX2:       ## BB#0: ## %entry
-; X32-AVX2-NEXT:    vbroadcastss LCPI30_0, %ymm1
+; X32-AVX2-NEXT:    vbroadcastss {{.*#+}} ymm1 = [-0.0078125,-0.0078125,-0.0078125,-0.0078125,-0.0078125,-0.0078125,-0.0078125,-0.0078125]
 ; X32-AVX2-NEXT:    vaddps %ymm1, %ymm0, %ymm0
 ; X32-AVX2-NEXT:    retl
 ;
 ; X64-AVX2-LABEL: V113:
 ; X64-AVX2:       ## BB#0: ## %entry
-; X64-AVX2-NEXT:    vbroadcastss {{.*}}(%rip), %ymm1
+; X64-AVX2-NEXT:    vbroadcastss {{.*#+}} ymm1 = [-0.0078125,-0.0078125,-0.0078125,-0.0078125,-0.0078125,-0.0078125,-0.0078125,-0.0078125]
 ; X64-AVX2-NEXT:    vaddps %ymm1, %ymm0, %ymm0
 ; X64-AVX2-NEXT:    retq
 ;
 ; X32-AVX512VL-LABEL: V113:
 ; X32-AVX512VL:       ## BB#0: ## %entry
 ; X32-AVX512VL-NEXT:    vaddps LCPI30_0{1to8}, %ymm0, %ymm0
 ; X32-AVX512VL-NEXT:    retl
 ;
 ; X64-AVX512VL-LABEL: V113:
 ; X64-AVX512VL:       ## BB#0: ## %entry
 ; X64-AVX512VL-NEXT:    vaddps {{.*}}(%rip){1to8}, %ymm0, %ymm0
 ; X64-AVX512VL-NEXT:    retq
 entry:
   %g = fadd <8 x float> %in, <float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000>
   ret <8 x float> %g
 }
 
 define <4 x float> @_e2(float* %ptr) nounwind uwtable readnone ssp {
 ; X32-LABEL: _e2:
 ; X32:       ## BB#0:
-; X32-NEXT:    vbroadcastss LCPI31_0, %xmm0
+; X32-NEXT:    vbroadcastss {{.*#+}} xmm0 = [-0.0078125,-0.0078125,-0.0078125,-0.0078125]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: _e2:
 ; X64:       ## BB#0:
-; X64-NEXT:    vbroadcastss {{.*}}(%rip), %xmm0
+; X64-NEXT:    vbroadcastss {{.*#+}} xmm0 = [-0.0078125,-0.0078125,-0.0078125,-0.0078125]
 ; X64-NEXT:    retq
   %vecinit.i = insertelement <4 x float> undef, float        0xbf80000000000000, i32 0
   %vecinit2.i = insertelement <4 x float> %vecinit.i, float  0xbf80000000000000, i32 1
   %vecinit4.i = insertelement <4 x float> %vecinit2.i, float 0xbf80000000000000, i32 2
   %vecinit6.i = insertelement <4 x float> %vecinit4.i, float 0xbf80000000000000, i32 3
   ret <4 x float> %vecinit6.i
 }
 

llvm/trunk/test/CodeGen/X86/broadcast-elm-cross-splat-vec.ll

-; NOTE: Assertions have been simpilfied MANUALLY after running utils/update_llc_test_checks.py
-;       Assertions for constant pools have been added MANUALLY.
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX
-; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx2 | FileCheck %s -check-prefix=ALL -check-prefix=ALL32 -check-prefix=NO-AVX512BW -check-prefix=AVX2 
-; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx512f | FileCheck %s -check-prefix=ALL -check-prefix=ALL32 -check-prefix=NO-AVX512BW -check-prefix=AVX512 
-; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx512f,+avx512bw | FileCheck %s -check-prefix=ALL -check-prefix=ALL32 -check-prefix=AVX512BW -check-prefix=AVX512 
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2 | FileCheck %s -check-prefix=ALL -check-prefix=ALL64 -check-prefix=NO-AVX512BW -check-prefix=AVX2 -check-prefix=AVX2-64
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx512f | FileCheck %s -check-prefix=ALL -check-prefix=ALL64 -check-prefix=NO-AVX512BW -check-prefix=AVX512 -check-prefix=AVX512F-64
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx512f,+avx512bw | FileCheck %s -check-prefix=ALL -check-prefix=ALL64 -check-prefix=AVX512BW -check-prefix=AVX512 -check-prefix=AVX512BW-64
+; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx2 | FileCheck %s -check-prefix=ALL32 -check-prefix=NO-AVX512BW -check-prefix=AVX2
+; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx512f | FileCheck %s -check-prefix=ALL32 -check-prefix=NO-AVX512BW -check-prefix=AVX512
+; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx512f,+avx512bw | FileCheck %s -check-prefix=ALL32 -check-prefix=AVX512 -check-prefix=AVX512BW
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX-64
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2 | FileCheck %s -check-prefix=ALL64 -check-prefix=NO-AVX512BW-64 -check-prefix=AVX2-64
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx512f | FileCheck %s -check-prefix=ALL64 -check-prefix=NO-AVX512BW-64 -check-prefix=AVX512F-64
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx512f,+avx512bw | FileCheck %s -check-prefix=ALL64 -check-prefix=AVX512F-64 -check-prefix=AVX512BW-64
 
 ;===-----------------------------------------------------------------------------===
 ;    This test checks the ability to recognize a cross element pattern of
 ;    constants and perform the load via broadcasting a smaller constant
 ;    vector.
 ;    For example:
 ;    <i32 0, i32 1, i32 0, i32 1> => broadcast of the constant vector <i32 0, i32 1>
 ;===-----------------------------------------------------------------------------===
 
-; ALL:       LCPI0
-; ALL-NEXT:  .short 256                     # 0x100
-
 define <16 x i8> @f16xi8_i16(<16 x i8> %a) {
+; AVX-LABEL: f16xi8_i16:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f16xi8_i16:
 ; ALL32:       # BB#0:
-; ALL32-NEXT:    vpbroadcastw {{\.LCPI.*}}, %xmm1
+; ALL32-NEXT:    vpbroadcastw {{.*#+}} xmm1 = [256,256,256,256,256,256,256,256]
 ; ALL32-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f16xi8_i16:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-64-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f16xi8_i16:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastw {{.*}}(%rip), %xmm1
+; ALL64-NEXT:    vpbroadcastw {{.*#+}} xmm1 = [256,256,256,256,256,256,256,256]
 ; ALL64-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    retq
   %res1 = add <16 x i8> <i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1>, %a
   %res2 = and <16 x i8> <i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1>, %res1
   ret <16 x i8> %res2
 }
 
 
-; ALL:       .LCPI1
-; ALL-NEXT:  .long  50462976                # 0x3020100
-
-; AVX:       .LCPI1
-; AVX-NEXT   .long  50462976                # float 3.82047143E-37
-
 define <16 x i8> @f16xi8_i32(<16 x i8> %a) {
+; AVX-LABEL: f16xi8_i32:
+; AVX:       # BB#0:
+; AVX-NEXT:    vbroadcastss {{.*#+}} xmm1 = [3.82047143E-37,3.82047143E-37,3.82047143E-37,3.82047143E-37]
+; AVX-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f16xi8_i32:
 ; ALL32:       # BB#0:
-; ALL32-NEXT:    vpbroadcastd {{\.LCPI.*}}, %xmm1
+; ALL32-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [50462976,50462976,50462976,50462976]
 ; ALL32-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f16xi8_i32:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastss {{.*#+}} xmm1 = [3.82047143E-37,3.82047143E-37,3.82047143E-37,3.82047143E-37]
+; AVX-64-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f16xi8_i32:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastd {{.*}}(%rip), %xmm1
+; ALL64-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [50462976,50462976,50462976,50462976]
 ; ALL64-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f16xi8_i32:
-; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastss {{\.LCPI.*}}, %xmm1
-; AVX-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
   %res1 = add <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3>, %a
   %res2 = and <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3>, %res1
   ret <16 x i8> %res2
 }
 
 
-; ALL64:       .LCPI2
-; ALL64-NEXT:  .quad  506097522914230528      # 0x706050403020100
-
-; AVX:         .LCPI2
-; AVX-NEXT:    .quad  506097522914230528      # double 7.9499288951273625E-275
-
 define <16 x i8> @f16xi8_i64(<16 x i8> %a) {
+; AVX-LABEL: f16xi8_i64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f16xi8_i64:
 ; ALL32:       # BB#0:
 ; ALL32-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
 ; ALL32-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f16xi8_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX-64-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f16xi8_i64:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastq {{.*}}(%rip), %xmm1
+; ALL64-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [506097522914230528,506097522914230528]
 ; ALL64-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f16xi8_i64:
-; AVX:       # BB#0:
-; AVX-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
   %res1 = add <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, %a
   %res2 = and <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, %res1
   ret <16 x i8> %res2
 }
 
 
-; ALL:       .LCPI3
-; ALL-NEXT:  .short 256                     # 0x100
-
 define <32 x i8> @f32xi8_i16(<32 x i8> %a) {
+; AVX-LABEL: f32xi8_i16:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f32xi8_i16:
 ; ALL32:       # BB#0:
-; ALL32-NEXT:    vpbroadcastw {{\.LCPI.*}}, %ymm1
+; ALL32-NEXT:    vpbroadcastw {{.*#+}} ymm1 = [256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256]
 ; ALL32-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f32xi8_i16:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-64-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f32xi8_i16:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastw {{.*}}(%rip), %ymm1
+; ALL64-NEXT:    vpbroadcastw {{.*#+}} ymm1 = [256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256]
 ; ALL64-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; ALL64-NEXT:    retq
   %res1 = add <32 x i8> <i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1>, %a
   %res2 = and <32 x i8> <i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1>, %res1
   ret <32 x i8> %res2
 }
 
 
-; ALL:       .LCPI4
-; ALL-NEXT:  .long  50462976                # 0x3020100
-
-; AVX:       .LCPI4
-; AVX-NEXT:  .long  50462976                # float 3.82047143E-37
-
 define <32 x i8> @f32xi8_i32(<32 x i8> %a) {
+; AVX-LABEL: f32xi8_i32:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vbroadcastss {{.*#+}} xmm2 = [3.82047143E-37,3.82047143E-37,3.82047143E-37,3.82047143E-37]
+; AVX-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f32xi8_i32:
 ; ALL32:       # BB#0:
-; ALL32-NEXT:    vpbroadcastd {{\.LCPI.*}}, %ymm1
+; ALL32-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976]
 ; ALL32-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f32xi8_i32:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vbroadcastss {{.*#+}} xmm2 = [3.82047143E-37,3.82047143E-37,3.82047143E-37,3.82047143E-37]
+; AVX-64-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f32xi8_i32:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm1
+; ALL64-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976]
 ; ALL64-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f32xi8_i32:
-; AVX:       # BB#0:
-; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vbroadcastss {{\.LCPI.*}}, %xmm2
-; AVX-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
   %res1 = add <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3>, %a
   %res2 = and <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3>, %res1
   ret <32 x i8> %res2
 }
 
 
-; ALL64:       .LCPI5
-; ALL64-NEXT:  .quad  506097522914230528      # 0x706050403020100
-
-; AVX:         .LCPI5
-; AVX-NEXT:    .quad  506097522914230528      # double 7.9499288951273625E-275
-
 define <32 x i8> @f32xi8_i64(<32 x i8> %a) {
+; AVX-LABEL: f32xi8_i64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f32xi8_i64:
 ; ALL32:       # BB#0:
-; ALL32-NEXT:    vpbroadcastq {{\.LCPI.*}}, %ymm1
+; ALL32-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275]
 ; ALL32-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f32xi8_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX-64-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f32xi8_i64:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastq {{.*}}(%rip), %ymm1
+; ALL64-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [506097522914230528,506097522914230528,506097522914230528,506097522914230528]
 ; ALL64-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f32xi8_i64:
-; AVX:       # BB#0:
-; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
   %res1 = add <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, %a
   %res2 = and <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, %res1
   ret <32 x i8> %res2
 }
 
 
-; ALL:       .LCPI6
-; ALL-NEXT:  .byte  0                       # 0x0
-; ALL-NEXT:  .byte  1                       # 0x1
-; ALL-NEXT:  .byte  2                       # 0x2
-; ALL-NEXT:  .byte  3                       # 0x3
-; ALL-NEXT:  .byte  4                       # 0x4
-; ALL-NEXT:  .byte  5                       # 0x5
-; ALL-NEXT:  .byte  6                       # 0x6
-; ALL-NEXT:  .byte  7                       # 0x7
-; ALL-NEXT:  .byte  8                       # 0x8
-; ALL-NEXT:  .byte  9                       # 0x9
-; ALL-NEXT:  .byte  10                      # 0xa
-; ALL-NEXT:  .byte  11                      # 0xb
-; ALL-NEXT:  .byte  12                      # 0xc
-; ALL-NEXT:  .byte  13                      # 0xd
-; ALL-NEXT:  .byte  14                      # 0xe
-; ALL-NEXT:  .byte  15                      # 0xf
-; ALL-NOT:   .byte
-
 define <32 x i8> @f32xi8_i128(<32 x i8> %a) {
-; ALL-LABEL: f32xi8_i128:
-; ALL:       # BB#0:
-; ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1]
-; ALL-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX-LABEL: f32xi8_i128:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f32xi8_i128:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; ALL32-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL32-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f32xi8_i128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX-64-NEXT:    vpaddb %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f32xi8_i128:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; ALL64-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL64-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    retq
   %res1 = add <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %a
   %res2 = and <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %res1
   ret <32 x i8> %res2
 }
 
 
-; ALL:       .LCPI7
-; ALL-NEXT:  .short 256                     # 0x100
-
 define <64 x i8> @f64xi8_i16(<64 x i8> %a) {
+; AVX-LABEL: f64xi8_i16:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    retl
+;
 ; NO-AVX512BW-LABEL: f64xi8_i16:
 ; NO-AVX512BW:       # BB#0:
-; NO-AVX512BW-NEXT:    vpbroadcastw {{\.LCPI.*}}, %ymm2
+; NO-AVX512BW-NEXT:    vpbroadcastw {{.*#+}} ymm2 = [256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256]
 ; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
 ; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
 ;
 ; AVX512BW-LABEL: f64xi8_i16:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpbroadcastw {{\.LCPI.*}}, %zmm1
+; AVX512BW-NEXT:    vpbroadcastw {{.*#+}} zmm1 = [256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256]
 ; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
+;
+; AVX-64-LABEL: f64xi8_i16:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f64xi8_i16:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vpbroadcastw {{.*#+}} ymm2 = [256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256]
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f64xi8_i16:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vpbroadcastw {{.*#+}} zmm1 = [256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256]
+; AVX512BW-64-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
   %res1 = add <64 x i8> <i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1>, %a
   %res2 = and <64 x i8> <i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1>, %res1
   ret <64 x i8> %res2
 }
 
 
-; ALL:       .LCPI8
-; ALL-NEXT:  .long  50462976                # 0x3020100
-
-; AVX:       .LCPI8
-; AVX-NEXT:  .long  50462976                # float 3.82047143E-37
-
 define <64 x i8> @f64i8_i32(<64 x i8> %a) {
+; AVX-LABEL: f64i8_i32:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT:    vbroadcastss {{.*#+}} xmm3 = [3.82047143E-37,3.82047143E-37,3.82047143E-37,3.82047143E-37]
+; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    retl
+;
 ; NO-AVX512BW-LABEL: f64i8_i32:
 ; NO-AVX512BW:       # BB#0:
-; NO-AVX512BW-NEXT:    vpbroadcastd {{\.LCPI.*}}, %ymm2
+; NO-AVX512BW-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976]
 ; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
 ; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
 ;
 ; AVX512BW-LABEL: f64i8_i32:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpbroadcastd {{\.LCPI.*}}, %zmm1
+; AVX512BW-NEXT:    vpbroadcastd {{.*#+}} zmm1 = [50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976]
 ; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
 ;
-; AVX-LABEL: f64i8_i32:
+; AVX-64-LABEL: f64i8_i32:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vbroadcastss {{.*#+}} xmm3 = [3.82047143E-37,3.82047143E-37,3.82047143E-37,3.82047143E-37]
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f64i8_i32:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976]
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f64i8_i32:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vpbroadcastd {{.*#+}} zmm1 = [50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976,50462976]
+; AVX512BW-64-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
+  %res1 = add <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3>, %a
+  %res2 = and <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3>, %res1
+  ret <64 x i8> %res2
+}
+
+
+define <64 x i8> @f64xi8_i64(<64 x i8> %a) {
+; AVX-LABEL: f64xi8_i64:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX-NEXT:    vbroadcastss {{\.LCPI.*}}, %xmm3
+; AVX-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
 ; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
 ; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-  %res1 = add <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3>, %a
-  %res2 = and <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3, i8 0, i8 1, i8 2, i8 3>, %res1
-  ret <64 x i8> %res2
-}
-
-
-; ALL64:         .LCPI9
-; ALL64-NEXT:    .quad  506097522914230528      # 0x706050403020100
-
-; ALL32:         .LCPI9
-; ALL32-NEXT:    .quad  506097522914230528      # double 7.9499288951273625E-275
-
-; AVX:           .LCPI9
-; AVX-NEXT:      .quad  506097522914230528      # double 7.9499288951273625E-275
-
-define <64 x i8> @f64xi8_i64(<64 x i8> %a) {
+; AVX-NEXT:    retl
+;
 ; NO-AVX512BW-LABEL: f64xi8_i64:
 ; NO-AVX512BW:       # BB#0:
-; NO-AVX512BW-NEXT:    vpbroadcastq {{.*}}, %ymm2
+; NO-AVX512BW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275]
 ; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
 ; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
 ;
 ; AVX512BW-LABEL: f64xi8_i64:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpbroadcastq {{.*}}, %zmm1
+; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275,7.9499288951273625E-275]
 ; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
 ;
-; AVX-LABEL: f64xi8_i64:
+; AVX-64-LABEL: f64xi8_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f64xi8_i64:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [506097522914230528,506097522914230528,506097522914230528,506097522914230528]
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f64xi8_i64:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [506097522914230528,506097522914230528,506097522914230528,506097522914230528,506097522914230528,506097522914230528,506097522914230528,506097522914230528]
+; AVX512BW-64-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
+  %res1 = add <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, %a
+  %res2 = and <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, %res1
+  ret <64 x i8> %res2
+}
+
+
+define <64 x i8> @f64xi8_i128(<64 x i8> %a) {
+; AVX-LABEL: f64xi8_i128:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
 ; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
 ; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-  %res1 = add <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, %a
-  %res2 = and <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, %res1
-  ret <64 x i8> %res2
-}
-
-
-; ALL:       .LCPI10
-; ALL-NEXT:  .byte  0                       # 0x0
-; ALL-NEXT:  .byte  1                       # 0x1
-; ALL-NEXT:  .byte  2                       # 0x2
-; ALL-NEXT:  .byte  3                       # 0x3
-; ALL-NEXT:  .byte  4                       # 0x4
-; ALL-NEXT:  .byte  5                       # 0x5
-; ALL-NEXT:  .byte  6                       # 0x6
-; ALL-NEXT:  .byte  7                       # 0x7
-; ALL-NEXT:  .byte  8                       # 0x8
-; ALL-NEXT:  .byte  9                       # 0x9
-; ALL-NEXT:  .byte  10                      # 0xa
-; ALL-NEXT:  .byte  11                      # 0xb
-; ALL-NEXT:  .byte  12                      # 0xc
-; ALL-NEXT:  .byte  13                      # 0xd
-; ALL-NEXT:  .byte  14                      # 0xe
-; ALL-NEXT:  .byte  15                      # 0xf
-; ALL-NOT:   .byte
-
-define <64 x i8> @f64xi8_i128(<64 x i8> %a) {
+; AVX-NEXT:    retl
+;
 ; NO-AVX512BW-LABEL: f64xi8_i128:
 ; NO-AVX512BW:       # BB#0:
-; NO-AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; NO-AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; NO-AVX512BW-NEXT:    # ymm2 = mem[0,1,0,1]
 ; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
 ; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
 ;
 ; AVX512BW-LABEL: f64xi8_i128:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512BW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
 ; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
+;
+; AVX-64-LABEL: f64xi8_i128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f64xi8_i128:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; NO-AVX512BW-64-NEXT:    # ymm2 = mem[0,1,0,1]
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f64xi8_i128:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512BW-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-64-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
   %res1 = add <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %a
   %res2 = and <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %res1
   ret <64 x i8> %res2
 }
 
 
-; AVX512BW:       .LCPI11
-; AVX512BW-NEXT:  .byte 0                       # 0x0
-; AVX512BW-NEXT:  .byte 1                       # 0x1
-; AVX512BW-NEXT:  .byte 2                       # 0x2
-; AVX512BW-NEXT:  .byte 3                       # 0x3
-; AVX512BW-NEXT:  .byte 4                       # 0x4
-; AVX512BW-NEXT:  .byte 5                       # 0x5
-; AVX512BW-NEXT:  .byte 6                       # 0x6
-; AVX512BW-NEXT:  .byte 7                       # 0x7
-; AVX512BW-NEXT:  .byte 8                       # 0x8
-; AVX512BW-NEXT:  .byte 9                       # 0x9
-; AVX512BW-NEXT:  .byte 10                      # 0xa
-; AVX512BW-NEXT:  .byte 11                      # 0xb
-; AVX512BW-NEXT:  .byte 12                      # 0xc
-; AVX512BW-NEXT:  .byte 13                      # 0xd
-; AVX512BW-NEXT:  .byte 14                      # 0xe
-; AVX512BW-NEXT:  .byte 15                      # 0xf
-; AVX512BW-NEXT:  .byte 16                      # 0x10
-; AVX512BW-NEXT:  .byte 17                      # 0x11
-; AVX512BW-NEXT:  .byte 18                      # 0x12
-; AVX512BW-NEXT:  .byte 19                      # 0x13
-; AVX512BW-NEXT:  .byte 20                      # 0x14
-; AVX512BW-NEXT:  .byte 21                      # 0x15
-; AVX512BW-NEXT:  .byte 22                      # 0x16
-; AVX512BW-NEXT:  .byte 23                      # 0x17
-; AVX512BW-NEXT:  .byte 24                      # 0x18
-; AVX512BW-NEXT:  .byte 25                      # 0x19
-; AVX512BW-NEXT:  .byte 26                      # 0x1a
-; AVX512BW-NEXT:  .byte 27                      # 0x1b
-; AVX512BW-NEXT:  .byte 28                      # 0x1c
-; AVX512BW-NEXT:  .byte 29                      # 0x1d
-; AVX512BW-NEXT:  .byte 30                      # 0x1e
-; AVX512BW-NEXT:  .byte 31                      # 0x1f
-; AVX512BW-NOT:   .byte
-
 define <64 x i8> @f64xi8_i256(<64 x i8> %a) {
+; AVX-LABEL: f64xi8_i256:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
+; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX-NEXT:    vpaddb %xmm4, %xmm1, %xmm1
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
+; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    retl
+;
+; NO-AVX512BW-LABEL: f64xi8_i256:
+; NO-AVX512BW:       # BB#0:
+; NO-AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
+; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
+;
 ; AVX512BW-LABEL: f64xi8_i256:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
+; AVX512BW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
 ; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
+;
+; AVX-64-LABEL: f64xi8_i256:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm3 = [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX-64-NEXT:    vpaddb %xmm4, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f64xi8_i256:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f64xi8_i256:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
+; AVX512BW-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-64-NEXT:    vpaddb %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
   %res1 = add <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23, i8 24, i8 25, i8 26, i8 27, i8 28, i8 29, i8 30, i8 31, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23, i8 24, i8 25, i8 26, i8 27, i8 28, i8 29, i8 30, i8 31>, %a
   %res2 = and <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23, i8 24, i8 25, i8 26, i8 27, i8 28, i8 29, i8 30, i8 31, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23, i8 24, i8 25, i8 26, i8 27, i8 28, i8 29, i8 30, i8 31>, %res1
   ret <64 x i8> %res2
 }
 
 
-; ALL:       .LCPI12
-; ALL-NEXT:  .long  65536                   # 0x10000
-
-; AVX:       .LCPI12
-; AVX-NEXT:  .long  65536                   # float 9.18354962E-41
-
 define <8 x i16> @f8xi16_i32(<8 x i16> %a) {
+; AVX-LABEL: f8xi16_i32:
+; AVX:       # BB#0:
+; AVX-NEXT:    vbroadcastss {{.*#+}} xmm1 = [9.18354962E-41,9.18354962E-41,9.18354962E-41,9.18354962E-41]
+; AVX-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f8xi16_i32:
 ; ALL32:       # BB#0:
-; ALL32-NEXT:    vpbroadcastd {{\.LCPI.*}}, %xmm1
+; ALL32-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [65536,65536,65536,65536]
 ; ALL32-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f8xi16_i32:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastss {{.*#+}} xmm1 = [9.18354962E-41,9.18354962E-41,9.18354962E-41,9.18354962E-41]
+; AVX-64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f8xi16_i32:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastd {{.*}}(%rip), %xmm1
+; ALL64-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [65536,65536,65536,65536]
 ; ALL64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f8xi16_i32:
-; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastss {{\.LCPI.*}}, %xmm1
-; AVX-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
   %res1 = add <8 x i16> <i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1>, %a
   %res2 = and <8 x i16> <i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1>, %res1
   ret <8 x i16> %res2
 }
 
 
-; ALL64:       .LCPI13
-; ALL64-NEXT:  .quad  844433520132096         # 0x3000200010000
-
-; ALL32:       .LCPI13
-; ALL32-NEXT:  .quad  844433520132096         # double 4.1720559249406128E-309
-
-; AVX:         .LCPI13
-; AVX-NEXT:    .quad  844433520132096         # double 4.1720559249406128E-309
-
 define <8 x i16> @f8xi16_i64(<8 x i16> %a) {
+; AVX-LABEL: f8xi16_i64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f8xi16_i64:
 ; ALL32:       # BB#0:
 ; ALL32-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
 ; ALL32-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f8xi16_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX-64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f8xi16_i64:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastq {{.*}}(%rip), %xmm1
+; ALL64-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [844433520132096,844433520132096]
 ; ALL64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f8xi16_i64:
-; AVX:       # BB#0:
-; AVX-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
   %res1 = add <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3>, %a
   %res2 = and <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3>, %res1
   ret <8 x i16> %res2
 }
 
 
-; ALL:       .LCPI14
-; ALL-NEXT:  .long  65536                   # 0x10000
-
-; AVX:       .LCPI14
-; AVX-NEXT:  .long  65536                   # float 9.18354962E-41
-
 define <16 x i16> @f16xi16_i32(<16 x i16> %a) {
-; ALL-LABEL: f16xi16_i32:
-; ALL:       # BB#0:
-; ALL-NEXT:    vpbroadcastd {{\.LCPI.*}}, %ymm1
-; ALL-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vpand %ymm1, %ymm0, %ymm0
-;
 ; AVX-LABEL: f16xi16_i32:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vbroadcastss {{\.LCPI.*}}, %xmm2
+; AVX-NEXT:    vbroadcastss {{.*#+}} xmm2 = [9.18354962E-41,9.18354962E-41,9.18354962E-41,9.18354962E-41]
 ; AVX-NEXT:    vpaddw %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vpaddw %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f16xi16_i32:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [65536,65536,65536,65536,65536,65536,65536,65536]
+; ALL32-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f16xi16_i32:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vbroadcastss {{.*#+}} xmm2 = [9.18354962E-41,9.18354962E-41,9.18354962E-41,9.18354962E-41]
+; AVX-64-NEXT:    vpaddw %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f16xi16_i32:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [65536,65536,65536,65536,65536,65536,65536,65536]
+; ALL64-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    retq
   %res1 = add <16 x i16> <i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1>, %a
   %res2 = and <16 x i16> <i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1>, %res1
   ret <16 x i16> %res2
 }
 
 
-; ALL64:       .LCPI15
-; ALL64-NEXT:  .quad  844433520132096         # 0x3000200010000
-
-; ALL32:       .LCPI15
-; ALL32-NEXT:  .quad  844433520132096         # double 4.1720559249406128E-309
-
-; AVX:         .LCPI15
-; AVX-NEXT:    .quad  844433520132096         # double 4.1720559249406128E-309
-
 define <16 x i16> @f16xi16_i64(<16 x i16> %a) {
-; ALL-LABEL: f16xi16_i64:
-; ALL:       # BB#0:
-; ALL-NEXT:    vpbroadcastq {{.*}}, %ymm1
-; ALL-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vpand %ymm1, %ymm0, %ymm0
-;
 ; AVX-LABEL: f16xi16_i64:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
 ; AVX-NEXT:    vpaddw %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vpaddw %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f16xi16_i64:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309]
+; ALL32-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f16xi16_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX-64-NEXT:    vpaddw %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f16xi16_i64:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [844433520132096,844433520132096,844433520132096,844433520132096]
+; ALL64-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    retq
   %res1 = add <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3>, %a
   %res2 = and <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3>, %res1
   ret <16 x i16> %res2
 }
 
 
-; ALL:       .LCPI16
-; ALL-NEXT:  .short 0                       # 0x0
-; ALL-NEXT:  .short 1                       # 0x1
-; ALL-NEXT:  .short 2                       # 0x2
-; ALL-NEXT:  .short 3                       # 0x3
-; ALL-NEXT:  .short 4                       # 0x4
-; ALL-NEXT:  .short 5                       # 0x5
-; ALL-NEXT:  .short 6                       # 0x6
-; ALL-NEXT:  .short 7                       # 0x7
-; ALL-NOT:   .short
-
 define <16 x i16> @f16xi16_i128(<16 x i16> %a) {
-; ALL-LABEL: f16xi16_i128:
-; ALL:       # BB#0:
-; ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1]
-; ALL-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX-LABEL: f16xi16_i128:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7]
+; AVX-NEXT:    vpaddw %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vpaddw %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f16xi16_i128:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; ALL32-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL32-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f16xi16_i128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7]
+; AVX-64-NEXT:    vpaddw %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f16xi16_i128:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; ALL64-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL64-NEXT:    vpaddw %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    retq
   %res1 = add <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, %a
   %res2 = and <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, %res1
   ret <16 x i16> %res2
 }
 
 
-; ALL:       .LCPI17
-; ALL-NEXT:  .long  65536                   # 0x10000
-
-; AVX:       .LCPI17
-; AVX-NEXT:  .long  65536                   # float 9.18354962E-41
-
 define <32 x i16> @f32xi16_i32(<32 x i16> %a) {
+; AVX-LABEL: f32xi16_i32:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT:    vbroadcastss {{.*#+}} xmm3 = [9.18354962E-41,9.18354962E-41,9.18354962E-41,9.18354962E-41]
+; AVX-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddw %xmm3, %xmm1, %xmm1
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddw %xmm3, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    retl
+;
 ; NO-AVX512BW-LABEL: f32xi16_i32:
 ; NO-AVX512BW:       # BB#0:
-; NO-AVX512BW-NEXT:    vpbroadcastd {{\.LCPI.*}}, %ymm2
+; NO-AVX512BW-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [65536,65536,65536,65536,65536,65536,65536,65536]
 ; NO-AVX512BW-NEXT:    vpaddw %ymm2, %ymm1, %ymm1
 ; NO-AVX512BW-NEXT:    vpaddw %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
 ;
 ; AVX512BW-LABEL: f32xi16_i32:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpbroadcastd {{\.LCPI.*}}, %zmm1
+; AVX512BW-NEXT:    vpbroadcastd {{.*#+}} zmm1 = [65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536]
 ; AVX512BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
 ;
-; AVX-LABEL: f32xi16_i32:
+; AVX-64-LABEL: f32xi16_i32:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vbroadcastss {{.*#+}} xmm3 = [9.18354962E-41,9.18354962E-41,9.18354962E-41,9.18354962E-41]
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f32xi16_i32:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [65536,65536,65536,65536,65536,65536,65536,65536]
+; NO-AVX512BW-64-NEXT:    vpaddw %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f32xi16_i32:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vpbroadcastd {{.*#+}} zmm1 = [65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536,65536]
+; AVX512BW-64-NEXT:    vpaddw %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
+  %res1 = add <32 x i16> <i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1>, %a
+  %res2 = and <32 x i16> <i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1>, %res1
+  ret <32 x i16> %res2
+}
+
+
+define <32 x i16> @f32xi16_i64(<32 x i16> %a) {
+; AVX-LABEL: f32xi16_i64:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX-NEXT:    vbroadcastss {{\.LCPI.*}}, %xmm3
+; AVX-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
 ; AVX-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddw %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddw %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
 ; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-  %res1 = add <32 x i16> <i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1>, %a
-  %res2 = and <32 x i16> <i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1, i16 0, i16 1>, %res1
-  ret <32 x i16> %res2
-}
-
-
-; ALL64:         .LCPI18
-; ALL64-NEXT:    .quad  844433520132096         # 0x3000200010000
-
-; ALL32:         .LCPI18
-; ALL32-NEXT:    .quad  844433520132096         # double 4.1720559249406128E-309
-
-; AVX:           .LCPI18
-; AVX-NEXT:      .quad  844433520132096         # double 4.1720559249406128E-309
-
-define <32 x i16> @f32xi16_i64(<32 x i16> %a) {
+; AVX-NEXT:    retl
+;
 ; NO-AVX512BW-LABEL: f32xi16_i64:
 ; NO-AVX512BW:       # BB#0:
-; NO-AVX512BW-NEXT:    vpbroadcastq {{.*}}, %ymm2
+; NO-AVX512BW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309]
 ; NO-AVX512BW-NEXT:    vpaddw %ymm2, %ymm1, %ymm1
 ; NO-AVX512BW-NEXT:    vpaddw %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
 ;
 ; AVX512BW-LABEL: f32xi16_i64:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpbroadcastq {{.*}}, %zmm1
+; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309,4.1720559249406128E-309]
 ; AVX512BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
 ;
-; AVX-LABEL: f32xi16_i64:
+; AVX-64-LABEL: f32xi16_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f32xi16_i64:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [844433520132096,844433520132096,844433520132096,844433520132096]
+; NO-AVX512BW-64-NEXT:    vpaddw %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f32xi16_i64:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [844433520132096,844433520132096,844433520132096,844433520132096,844433520132096,844433520132096,844433520132096,844433520132096]
+; AVX512BW-64-NEXT:    vpaddw %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
+  %res1 = add <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3>, %a
+  %res2 = and <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3>, %res1
+  ret <32 x i16> %res2
+}
+
+
+define <32 x i16> @f32xi16_i128(<32 x i16> %a) {
+; AVX-LABEL: f32xi16_i128:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,2,3,4,5,6,7]
 ; AVX-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddw %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddw %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
 ; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-  %res1 = add <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3>, %a
-  %res2 = and <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3, i16 0, i16 1, i16 2, i16 3>, %res1
-  ret <32 x i16> %res2
-}
-
-
-; ALL:       .LCPI19
-; ALL-NEXT:  .short 0                       # 0x0
-; ALL-NEXT:  .short 1                       # 0x1
-; ALL-NEXT:  .short 2                       # 0x2
-; ALL-NEXT:  .short 3                       # 0x3
-; ALL-NEXT:  .short 4                       # 0x4
-; ALL-NEXT:  .short 5                       # 0x5
-; ALL-NEXT:  .short 6                       # 0x6
-; ALL-NEXT:  .short 7                       # 0x7
-; ALL-NOT:   .short
-
-define <32 x i16> @f32xi16_i128(<32 x i16> %a) {
+; AVX-NEXT:    retl
+;
 ; NO-AVX512BW-LABEL: f32xi16_i128:
 ; NO-AVX512BW:       # BB#0:
-; NO-AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; NO-AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; NO-AVX512BW-NEXT:    # ymm2 = mem[0,1,0,1]
 ; NO-AVX512BW-NEXT:    vpaddw %ymm2, %ymm1, %ymm1
 ; NO-AVX512BW-NEXT:    vpaddw %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
 ;
 ; AVX512BW-LABEL: f32xi16_i128:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; AVX512BW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
 ; AVX512BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
+;
+; AVX-64-LABEL: f32xi16_i128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,2,3,4,5,6,7]
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f32xi16_i128:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; NO-AVX512BW-64-NEXT:    # ymm2 = mem[0,1,0,1]
+; NO-AVX512BW-64-NEXT:    vpaddw %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f32xi16_i128:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; AVX512BW-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-64-NEXT:    vpaddw %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
   %res1 = add <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, %a
   %res2 = and <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, %res1
   ret <32 x i16> %res2
 }
 
 
-; AVX512BW:       .LCPI20
-; AVX512BW-NEXT:  .short  0                       # 0x0
-; AVX512BW-NEXT:  .short  1                       # 0x1
-; AVX512BW-NEXT:  .short  2                       # 0x2
-; AVX512BW-NEXT:  .short  3                       # 0x3
-; AVX512BW-NEXT:  .short  4                       # 0x4
-; AVX512BW-NEXT:  .short  5                       # 0x5
-; AVX512BW-NEXT:  .short  6                       # 0x6
-; AVX512BW-NEXT:  .short  7                       # 0x7
-; AVX512BW-NEXT:  .short  8                       # 0x8
-; AVX512BW-NEXT:  .short  9                       # 0x9
-; AVX512BW-NEXT:  .short  10                      # 0xa
-; AVX512BW-NEXT:  .short  11                      # 0xb
-; AVX512BW-NEXT:  .short  12                      # 0xc
-; AVX512BW-NEXT:  .short  13                      # 0xd
-; AVX512BW-NEXT:  .short  14                      # 0xe
-; AVX512BW-NEXT:  .short  15                      # 0xf
-; AVX512BW-NOT:   .short
-
 define <32 x i16> @f32xi16_i256(<32 x i16> %a) {
+; AVX-LABEL: f32xi16_i256:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,9,10,11,12,13,14,15]
+; AVX-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,2,3,4,5,6,7]
+; AVX-NEXT:    vpaddw %xmm4, %xmm1, %xmm1
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddw %xmm4, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    retl
+;
+; NO-AVX512BW-LABEL: f32xi16_i256:
+; NO-AVX512BW:       # BB#0:
+; NO-AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; NO-AVX512BW-NEXT:    vpaddw %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    vpaddw %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-NEXT:    retl
+;
 ; AVX512BW-LABEL: f32xi16_i256:
 ; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512BW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
 ; AVX512BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    retl
+;
+; AVX-64-LABEL: f32xi16_i256:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,9,10,11,12,13,14,15]
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,2,3,4,5,6,7]
+; AVX-64-NEXT:    vpaddw %xmm4, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddw %xmm4, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; NO-AVX512BW-64-LABEL: f32xi16_i256:
+; NO-AVX512BW-64:       # BB#0:
+; NO-AVX512BW-64-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; NO-AVX512BW-64-NEXT:    vpaddw %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; NO-AVX512BW-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; NO-AVX512BW-64-NEXT:    retq
+;
+; AVX512BW-64-LABEL: f32xi16_i256:
+; AVX512BW-64:       # BB#0:
+; AVX512BW-64-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512BW-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-64-NEXT:    vpaddw %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-64-NEXT:    retq
   %res1 = add <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15>, %a
   %res2 = and <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15>, %res1
   ret <32 x i16> %res2
 }
 
 
-; ALL64:       .LCPI21
-; ALL64-NEXT:  .quad  4294967296              # 0x100000000
-
-; ALL32:       .LCPI21
-; ALL32-NEXT:  .quad  4294967296              # double 2.1219957909652723E-314
-
-; AVX:         .LCPI21
-; AVX-NEXT:    .quad  4294967296              # double 2.1219957909652723E-314
 
 define <4 x i32> @f4xi32_i64(<4 x i32> %a) {
+; AVX-LABEL: f4xi32_i64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f4xi32_i64:
 ; ALL32:       # BB#0:
 ; ALL32-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
 ; ALL32-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f4xi32_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX-64-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f4xi32_i64:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastq {{.*}}(%rip), %xmm1
+; ALL64-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [4294967296,4294967296]
 ; ALL64-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f4xi32_i64:
-; AVX:       # BB#0:
-; AVX-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
   %res1 = add <4 x i32> <i32 0, i32 1, i32 0, i32 1>, %a
   %res2 = and <4 x i32> <i32 0, i32 1, i32 0, i32 1>, %res1
   ret <4 x i32> %res2
 }
 
 
-; ALL64:       .LCPI22
-; ALL64-NEXT:  .quad  4294967296              # 0x100000000
-
-; ALL32:       .LCPI22
-; ALL32-NEXT:  .quad  4294967296              # double 2.1219957909652723E-314
-
-; AVX:         .LCPI22
-; AVX-NEXT:    .quad  4294967296              # double 2.1219957909652723E-314
-
 define <8 x i32> @f8xi32_i64(<8 x i32> %a) {
-; ALL-LABEL: f8xi32_i64:
-; ALL:       # BB#0:
-; ALL-NEXT:    vpbroadcastq {{.*}}, %ymm1
-; ALL-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vpand %ymm1, %ymm0, %ymm0
-;
 ; AVX-LABEL: f8xi32_i64:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
 ; AVX-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f8xi32_i64:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314]
+; ALL32-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f8xi32_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX-64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f8xi32_i64:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4294967296,4294967296,4294967296,4294967296]
+; ALL64-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    retq
   %res1 = add <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>, %a
   %res2 = and <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>, %res1
   ret <8 x i32> %res2
 }
 
 
-; ALL:       .LCPI23
-; ALL-NEXT:  .long  0                       # 0x0
-; ALL-NEXT:  .long  1                       # 0x1
-; ALL-NEXT:  .long  2                       # 0x2
-; ALL-NEXT:  .long  3                       # 0x3
-; ALL-NOT:   .long
-
 define <8 x i32> @f8xi32_i128(<8 x i32> %a) {
-; ALL-LABEL: f8xi32_i128:
-; ALL:       # BB#0:
-; ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1]
-; ALL-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX-LABEL: f8xi32_i128:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,2,3]
+; AVX-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT:    vandps {{\.LCPI.*}}, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f8xi32_i128:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,1,2,3,0,1,2,3]
+; ALL32-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL32-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f8xi32_i128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,2,3]
+; AVX-64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f8xi32_i128:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,1,2,3,0,1,2,3]
+; ALL64-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL64-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    retq
   %res1 = add <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>, %a
   %res2 = and <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>, %res1
   ret <8 x i32> %res2
 }
 
 
-; ALL64:         .LCPI24
-; ALL64-NEXT:    .quad  4294967296              # 0x100000000
-
-; ALL32:         .LCPI24
-; ALL32-NEXT:    .quad  4294967296              # double 2.1219957909652723E-314
-
-; AVX:           .LCPI24
-; AVX-NEXT:      .quad  4294967296              # double 2.1219957909652723E-314
-
 define <16 x i32> @f16xi32_i64(<16 x i32> %a) {
+; AVX-LABEL: f16xi32_i64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddd %xmm3, %xmm1, %xmm1
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
+; AVX-NEXT:    vpaddd %xmm3, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,0,1,0,1,0,1]
+; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    retl
+;
 ; AVX2-LABEL: f16xi32_i64:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpbroadcastq {{.*}}, %ymm2
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314]
 ; AVX2-NEXT:    vpaddd %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    vpaddd %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    retl
 ;
 ; AVX512-LABEL: f16xi32_i64:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vpbroadcastq {{.*}}, %zmm1
+; AVX512-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314,2.1219957909652723E-314]
 ; AVX512-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512-NEXT:    retl
 ;
-; AVX-LABEL: f16xi32_i64:
+; AVX-64-LABEL: f16xi32_i64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX-64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddd %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,0,1,0,1,0,1]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; AVX2-64-LABEL: f16xi32_i64:
+; AVX2-64:       # BB#0:
+; AVX2-64-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4294967296,4294967296,4294967296,4294967296]
+; AVX2-64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    retq
+;
+; AVX512F-64-LABEL: f16xi32_i64:
+; AVX512F-64:       # BB#0:
+; AVX512F-64-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [4294967296,4294967296,4294967296,4294967296,4294967296,4294967296,4294967296,4294967296]
+; AVX512F-64-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    retq
+  %res1 = add <16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>, %a
+  %res2 = and <16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>, %res1
+  ret <16 x i32> %res2
+}
+
+
+define <16 x i32> @f16xi32_i128(<16 x i32> %a) {
+; AVX-LABEL: f16xi32_i128:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,2,3]
 ; AVX-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddd %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpaddd %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,0,1,0,1,0,1]
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3]
 ; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-  %res1 = add <16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>, %a
-  %res2 = and <16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>, %res1
-  ret <16 x i32> %res2
-}
-
-
-; ALL:       .LCPI25
-; ALL-NEXT:  .long  0                       # 0x0
-; ALL-NEXT:  .long  1                       # 0x1
-; ALL-NEXT:  .long  2                       # 0x2
-; ALL-NEXT:  .long  3                       # 0x3
-; ALL-NOT:   .long
-
-define <16 x i32> @f16xi32_i128(<16 x i32> %a) {
+; AVX-NEXT:    retl
+;
 ; AVX2-LABEL: f16xi32_i128:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3]
+; AVX2-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-NEXT:    vpaddd %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    vpaddd %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    retl
 ;
 ; AVX512-LABEL: f16xi32_i128:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
 ; AVX512-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vpandd %zmm1, %zmm0, %zmm0
+; AVX512-NEXT:    retl
+;
+; AVX-64-LABEL: f16xi32_i128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,2,3]
+; AVX-64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddd %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; AVX2-64-LABEL: f16xi32_i128:
+; AVX2-64:       # BB#0:
+; AVX2-64-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [0,1,2,3,0,1,2,3]
+; AVX2-64-NEXT:    # ymm2 = mem[0,1,0,1]
+; AVX2-64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    retq
+;
+; AVX512F-64-LABEL: f16xi32_i128:
+; AVX512F-64:       # BB#0:
+; AVX512F-64-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    vpandd %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    retq
   %res1 = add <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>, %a
   %res2 = and <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>, %res1
   ret <16 x i32> %res2
 }
 
 
-; ALL64:       .LCPI26
-; ALL64-NEXT:  .quad  0                       # 0x0
-; ALL64-NEXT:  .quad  1                       # 0x1
-; ALL64-NOT:   .quad
-
 define <4 x i64> @f4xi64_i128(<4 x i64> %a) {
+; AVX-LABEL: f4xi64_i128:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,1,0,0,0,1,0]
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm3
+; AVX-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
+; AVX-NEXT:    vpaddq %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm1, %ymm0, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f4xi64_i128:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,1,0,0,0,1,0]
+; ALL32-NEXT:    vpaddq %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f4xi64_i128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-64-NEXT:    movl $1, %eax
+; AVX-64-NEXT:    vmovq %rax, %xmm2
+; AVX-64-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6,7]
+; AVX-64-NEXT:    vpaddq %xmm2, %xmm1, %xmm1
+; AVX-64-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f4xi64_i128:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1]
+; ALL64-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,1,0,1]
+; ALL64-NEXT:    # ymm1 = mem[0,1,0,1]
 ; ALL64-NEXT:    vpaddq %ymm1, %ymm0, %ymm0
 ; ALL64-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; ALL64-NEXT:    retq
   %res1 = add <4 x i64> <i64 0, i64 1, i64 0, i64 1>, %a
   %res2 = and <4 x i64> <i64 0, i64 1, i64 0, i64 1>, %res1
   ret <4 x i64> %res2
 }
 
 
-; ALL64:       .LCPI27
-; ALL64-NEXT:  .quad  0                       # 0x0
-; ALL64-NEXT:  .quad  1                       # 0x1
-; ALL64-NOT:   .quad
-
 define <8 x i64> @f8xi64_i128(<8 x i64> %a) {
+; AVX-LABEL: f8xi64_i128:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,1,0,0,0,1,0]
+; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX-NEXT:    vpaddq %xmm3, %xmm4, %xmm4
+; AVX-NEXT:    vpaddq %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm4
+; AVX-NEXT:    vpaddq %xmm3, %xmm4, %xmm3
+; AVX-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    retl
+;
+; AVX2-LABEL: f8xi64_i128:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,1,0,0,0,1,0]
+; AVX2-NEXT:    vpaddq %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddq %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    retl
+;
+; AVX512-LABEL: f8xi64_i128:
+; AVX512:       # BB#0:
+; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [0,0,1,0,0,0,1,0,0,0,1,0,0,0,1,0]
+; AVX512-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
+; AVX512-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512-NEXT:    retl
+;
+; AVX-64-LABEL: f8xi64_i128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    movl $1, %eax
+; AVX-64-NEXT:    vmovq %rax, %xmm3
+; AVX-64-NEXT:    vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4,5,6,7]
+; AVX-64-NEXT:    vpaddq %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddq %xmm3, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddq %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [0,1,0,1]
+; AVX-64-NEXT:    # ymm2 = mem[0,1,0,1]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
 ; AVX2-64-LABEL: f8xi64_i128:
 ; AVX2-64:       # BB#0:
-; AVX2-64-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; AVX2-64-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [0,1,0,1]
+; AVX2-64-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-64-NEXT:    vpaddq %ymm2, %ymm1, %ymm1
 ; AVX2-64-NEXT:    vpaddq %ymm2, %ymm0, %ymm0
 ; AVX2-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
 ; AVX2-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
 ; AVX2-64-NEXT:    retq
 ;
 ; AVX512F-64-LABEL: f8xi64_i128:
 ; AVX512F-64:       # BB#0:
-; AVX512F-64-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [0,1,0,1,0,1,0,1]
+; AVX512F-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
 ; AVX512F-64-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
 ; AVX512F-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
 ; AVX512F-64-NEXT:    retq
-;
-; AVX512BW-64-LABEL: f8xi64_i128:
-; AVX512BW-64:       # BB#0:
-; AVX512BW-64-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
-; AVX512BW-64-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
-; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
-; AVX512BW-64-NEXT:    retq
   %res1 = add <8 x i64> <i64 0, i64 1, i64 0, i64 1, i64 0, i64 1, i64 0, i64 1>, %a
   %res2 = and <8 x i64> <i64 0, i64 1, i64 0, i64 1, i64 0, i64 1, i64 0, i64 1>, %res1
   ret <8 x i64> %res2
 }
 
 
-; ALL64:            .LCPI28
-; ALL64-NEXT:       .quad 0                       # 0x0
-; ALL64-NEXT:       .quad 1                       # 0x1
-; ALL64-NEXT:       .quad 2                       # 0x2
-; ALL64-NEXT:       .quad 3                       # 0x3
-; ALL64-NOT:        .quad
-
 define <8 x i64> @f8xi64_i256(<8 x i64> %a) {
+; AVX-LABEL: f8xi64_i256:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,1,0,2,0,3,0]
+; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX-NEXT:    vpaddq %xmm3, %xmm4, %xmm4
+; AVX-NEXT:    vpaddq %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm4
+; AVX-NEXT:    vpaddq %xmm3, %xmm4, %xmm3
+; AVX-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    retl
+;
+; AVX2-LABEL: f8xi64_i256:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,1,0,2,0,3,0]
+; AVX2-NEXT:    vpaddq %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddq %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    retl
+;
+; AVX512-LABEL: f8xi64_i256:
+; AVX512:       # BB#0:
+; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [0,0,1,0,2,0,3,0,0,0,1,0,2,0,3,0]
+; AVX512-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
+; AVX512-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512-NEXT:    retl
+;
+; AVX-64-LABEL: f8xi64_i256:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-64-NEXT:    vmovdqa {{.*#+}} xmm3 = [2,3]
+; AVX-64-NEXT:    vpaddq %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    movl $1, %eax
+; AVX-64-NEXT:    vmovq %rax, %xmm4
+; AVX-64-NEXT:    vpslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1,2,3,4,5,6,7]
+; AVX-64-NEXT:    vpaddq %xmm4, %xmm1, %xmm1
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX-64-NEXT:    vpaddq %xmm3, %xmm2, %xmm2
+; AVX-64-NEXT:    vpaddq %xmm4, %xmm0, %xmm0
+; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,2,3]
+; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    retq
+;
+; AVX2-64-LABEL: f8xi64_i256:
+; AVX2-64:       # BB#0:
+; AVX2-64-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,2,3]
+; AVX2-64-NEXT:    vpaddq %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vpaddq %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    retq
+;
 ; AVX512F-64-LABEL: f8xi64_i256:
 ; AVX512F-64:       # BB#0:
-; AVX512F-64-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
 ; AVX512F-64-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
 ; AVX512F-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
 ; AVX512F-64-NEXT:    retq
-;
-; AVX512BW-64-LABEL: f8xi64_i256:
-; AVX512BW-64:       # BB#0:
-; AVX512BW-64-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3]
-; AVX512BW-64-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
-; AVX512BW-64-NEXT:    vpandq %zmm1, %zmm0, %zmm0
-; AVX512BW-64-NEXT:    retq
   %res1 = add <8 x i64> <i64 0, i64 1, i64 2, i64 3, i64 0, i64 1, i64 2, i64 3>, %a
   %res2 = and <8 x i64> <i64 0, i64 1, i64 2, i64 3, i64 0, i64 1, i64 2, i64 3>, %res1
   ret <8 x i64> %res2
 }
 
 
-; ALL:       .LCPI29
-; ALL-NEXT:  .quad  4575657222482165760     
-
-; AVX:       .LCPI29
-; AVX-NEXT:  .quad  4575657222482165760     # double 0.0078125018626451492
-
 define <4 x float> @f4xf32_f64(<4 x float> %a) {
+; AVX-LABEL: f4xf32_f64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vdivps %xmm0, %xmm1, %xmm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f4xf32_f64:
 ; ALL32:       # BB#0:
 ; ALL32-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
 ; ALL32-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    vdivps %xmm0, %xmm1, %xmm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f4xf32_f64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX-64-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    vdivps %xmm0, %xmm1, %xmm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f4xf32_f64:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastq {{.*}}(%rip), %xmm1
+; ALL64-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [4575657222482165760,4575657222482165760]
 ; ALL64-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    vdivps %xmm0, %xmm1, %xmm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f4xf32_f64:
-; AVX:       # BB#0:
-; AVX-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX-NEXT:    vaddps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vdivps %xmm0, %xmm1, %xmm0
   %res1 = fadd <4 x float> <float 2.0, float 1.0, float 2.0, float 1.0>, %a
   %res2 = fdiv <4 x float> <float 2.0, float 1.0, float 2.0, float 1.0>, %res1
   ret <4 x float> %res2
 }
 
 
-; ALL64:       .LCPI30
-; ALL64-NEXT:  .quad  4575657222482165760     # 0x3f80000040000000
-
-; ALL32:         .LCPI30
-; ALL32-NEXT:    .quad  4575657222482165760     # double 0.0078125018626451492
-
-; AVX:         .LCPI30
-; AVX-NEXT:    .quad  4575657222482165760     # double 0.0078125018626451492
-
 define <8 x float> @f8xf32_f64(<8 x float> %a) {
-; ALL-LABEL: f8xf32_f64:
-; ALL:       # BB#0:
-; ALL-NEXT:    vbroadcastsd {{.*}}, %ymm1
-; ALL-NEXT:    vaddps %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vdivps %ymm0, %ymm1, %ymm0
-;
 ; AVX-LABEL: f8xf32_f64:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastsd {{\.LCPI.*}}, %ymm1
+; AVX-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492]
 ; AVX-NEXT:    vaddps %ymm1, %ymm0, %ymm0
 ; AVX-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f8xf32_f64:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492]
+; ALL32-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f8xf32_f64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492]
+; AVX-64-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f8xf32_f64:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [4575657222482165760,4575657222482165760,4575657222482165760,4575657222482165760]
+; ALL64-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; ALL64-NEXT:    retq
   %res1 = fadd <8 x float> <float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0>, %a
   %res2 = fdiv <8 x float> <float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0>, %res1
   ret <8 x float> %res2
 }
 
 
-; ALL:       .LCPI31
-; ALL-NEXT:  .long  1082130432              # float 4
-; ALL-NEXT:  .long  1065353216              # float 1
-; ALL-NEXT:  .long  1073741824              # float 2
-; ALL-NEXT:  .long  1077936128              # float 3
-; ALL-NOT:   .long
-
 define <8 x float> @f8xf32_f128(<8 x float> %a) {
-; ALL-LABEL: f8xf32_f128:
-; ALL:       # BB#0:
-; ALL-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
-; ALL-NEXT:    vaddps %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vdivps %ymm0, %ymm1, %ymm0
-;
 ; AVX-LABEL: f8xf32_f128:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
+; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX-NEXT:    # ymm1 = mem[0,1,0,1]
 ; AVX-NEXT:    vaddps %ymm1, %ymm0, %ymm0
 ; AVX-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f8xf32_f128:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; ALL32-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL32-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f8xf32_f128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX-64-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX-64-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f8xf32_f128:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; ALL64-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL64-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; ALL64-NEXT:    retq
   %res1 = fadd <8 x float> <float 4.0, float 1.0, float 2.0, float 3.0, float 4.0, float 1.0, float 2.0, float 3.0>, %a
   %res2 = fdiv <8 x float> <float 4.0, float 1.0, float 2.0, float 3.0, float 4.0, float 1.0, float 2.0, float 3.0>, %res1
   ret <8 x float> %res2
 }
 
 
-; ALL64:       .LCPI32
-; ALL64-NEXT:  .quad  4575657222482165760     # 0x3f80000040000000
-
-; ALL32:       .LCPI32
-; ALL32-NEXT:  .quad  4575657222482165760     # double 0.0078125018626451492
-
-; AVX:       .LCPI32
-; AVX-NEXT:  .quad  4575657222482165760     # double 0.0078125018626451492
-
 define <16 x float> @f16xf32_f64(<16 x float> %a) {
+; AVX-LABEL: f16xf32_f64:
+; AVX:       # BB#0:
+; AVX-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492]
+; AVX-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX-NEXT:    retl
+;
 ; AVX2-LABEL: f16xf32_f64:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vbroadcastsd {{.*}}, %ymm2
+; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492]
 ; AVX2-NEXT:    vaddps %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    vaddps %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vdivps %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    retl
 ;
 ; AVX512-LABEL: f16xf32_f64:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcastsd {{.*}}, %zmm1
+; AVX512-NEXT:    vbroadcastsd {{.*#+}} zmm1 = [0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492]
 ; AVX512-NEXT:    vaddps %zmm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vdivps %zmm0, %zmm1, %zmm0
+; AVX512-NEXT:    retl
 ;
-; AVX-LABEL: f16xf32_f64:
-; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastsd {{\.LCPI.*}}, %ymm2
-; AVX-NEXT:    vaddps %ymm2, %ymm1, %ymm1
-; AVX-NEXT:    vaddps %ymm2, %ymm0, %ymm0
-; AVX-NEXT:    vdivps %ymm0, %ymm2, %ymm0
-; AVX-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX-64-LABEL: f16xf32_f64:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [0.0078125018626451492,0.0078125018626451492,0.0078125018626451492,0.0078125018626451492]
+; AVX-64-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX-64-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX-64-NEXT:    retq
+;
+; AVX2-64-LABEL: f16xf32_f64:
+; AVX2-64:       # BB#0:
+; AVX2-64-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4575657222482165760,4575657222482165760,4575657222482165760,4575657222482165760]
+; AVX2-64-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX2-64-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX2-64-NEXT:    retq
+;
+; AVX512F-64-LABEL: f16xf32_f64:
+; AVX512F-64:       # BB#0:
+; AVX512F-64-NEXT:    vbroadcastsd {{.*#+}} zmm1 = [4575657222482165760,4575657222482165760,4575657222482165760,4575657222482165760,4575657222482165760,4575657222482165760,4575657222482165760,4575657222482165760]
+; AVX512F-64-NEXT:    vaddps %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    vdivps %zmm0, %zmm1, %zmm0
+; AVX512F-64-NEXT:    retq
   %res1 = fadd <16 x float> <float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0>, %a
   %res2 = fdiv <16 x float> <float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0, float 2.0, float 1.0>, %res1
   ret <16 x float> %res2
 }
 
 
-; ALL:       .LCPI33
-; ALL-NEXT:  .long  1082130432              # float 4
-; ALL-NEXT:  .long  1065353216              # float 1
-; ALL-NEXT:  .long  1073741824              # float 2
-; ALL-NEXT:  .long  1077936128              # float 3
-; ALL-NOT:   .long
-
 define <16 x float> @f16xf32_f128(<16 x float> %a) {
+; AVX-LABEL: f16xf32_f128:
+; AVX:       # BB#0:
+; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX-NEXT:    # ymm2 = mem[0,1,0,1]
+; AVX-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX-NEXT:    retl
+;
 ; AVX2-LABEL: f16xf32_f128:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX2-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-NEXT:    vaddps %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    vaddps %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vdivps %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    retl
 ;
 ; AVX512-LABEL: f16xf32_f128:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcastf32x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vbroadcastf32x4 {{.*#+}} zmm1 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX512-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
 ; AVX512-NEXT:    vaddps %zmm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vdivps %zmm0, %zmm1, %zmm0
+; AVX512-NEXT:    retl
 ;
-; AVX-LABEL: f16xf32_f128:
-; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = mem[0,1,0,1]
-; AVX-NEXT:    vaddps %ymm2, %ymm1, %ymm1
-; AVX-NEXT:    vaddps %ymm2, %ymm0, %ymm0
-; AVX-NEXT:    vdivps %ymm0, %ymm2, %ymm0
-; AVX-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX-64-LABEL: f16xf32_f128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX-64-NEXT:    # ymm2 = mem[0,1,0,1]
+; AVX-64-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX-64-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX-64-NEXT:    retq
+;
+; AVX2-64-LABEL: f16xf32_f128:
+; AVX2-64:       # BB#0:
+; AVX2-64-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX2-64-NEXT:    # ymm2 = mem[0,1,0,1]
+; AVX2-64-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX2-64-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX2-64-NEXT:    retq
+;
+; AVX512F-64-LABEL: f16xf32_f128:
+; AVX512F-64:       # BB#0:
+; AVX512F-64-NEXT:    vbroadcastf32x4 {{.*#+}} zmm1 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX512F-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    vaddps %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    vdivps %zmm0, %zmm1, %zmm0
+; AVX512F-64-NEXT:    retq
   %res1 = fadd <16 x float> <float 4.0, float 1.0, float 2.0, float 3.0, float 4.0, float 1.0, float 2.0, float 3.0, float 4.0, float 1.0, float 2.0, float 3.0, float 4.0, float 1.0, float 2.0, float 3.0>, %a
   %res2 = fdiv <16 x float> <float 4.0, float 1.0, float 2.0, float 3.0, float 4.0, float 1.0, float 2.0, float 3.0, float 4.0, float 1.0, float 2.0, float 3.0, float 4.0, float 1.0, float 2.0, float 3.0>, %res1
   ret <16 x float> %res2
 }
 
 
-; AVX512:       .LCPI34
-; AVX512-NEXT:  .long 1090519040              # float 8
-; AVX512-NEXT:  .long 1065353216              # float 1
-; AVX512-NEXT:  .long 1073741824              # float 2
-; AVX512-NEXT:  .long 1077936128              # float 3
-; AVX512-NEXT:  .long 1082130432              # float 4
-; AVX512-NEXT:  .long 1084227584              # float 5
-; AVX512-NEXT:  .long 1086324736              # float 6
-; AVX512-NEXT:  .long 1088421888              # float 7
-; AVX512-NOT:   .long
-
 define <16 x float> @f16xf32_f256(<16 x float> %a) {
+; AVX-LABEL: f16xf32_f256:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovaps {{.*#+}} ymm2 = [8.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00]
+; AVX-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX-NEXT:    retl
+;
+; AVX2-LABEL: f16xf32_f256:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vmovaps {{.*#+}} ymm2 = [8.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00]
+; AVX2-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX2-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    retl
+;
 ; AVX512-LABEL: f16xf32_f256:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcastf64x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vbroadcastf64x4 {{.*#+}} zmm1 = [8.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00]
+; AVX512-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
 ; AVX512-NEXT:    vaddps %zmm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vdivps %zmm0, %zmm1, %zmm0
+; AVX512-NEXT:    retl
+;
+; AVX-64-LABEL: f16xf32_f256:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vmovaps {{.*#+}} ymm2 = [8.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00]
+; AVX-64-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX-64-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX-64-NEXT:    retq
+;
+; AVX2-64-LABEL: f16xf32_f256:
+; AVX2-64:       # BB#0:
+; AVX2-64-NEXT:    vmovaps {{.*#+}} ymm2 = [8.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00]
+; AVX2-64-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vaddps %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vdivps %ymm0, %ymm2, %ymm0
+; AVX2-64-NEXT:    vdivps %ymm1, %ymm2, %ymm1
+; AVX2-64-NEXT:    retq
+;
+; AVX512F-64-LABEL: f16xf32_f256:
+; AVX512F-64:       # BB#0:
+; AVX512F-64-NEXT:    vbroadcastf64x4 {{.*#+}} zmm1 = [8.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00]
+; AVX512F-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    vaddps %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    vdivps %zmm0, %zmm1, %zmm0
+; AVX512F-64-NEXT:    retq
   %res1 = fadd <16 x float> <float 8.0, float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0>, %a
   %res2 = fdiv <16 x float> <float 8.0, float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0>, %res1
   ret <16 x float> %res2
 }
 
 
-; ALL:       .LCPI35
-; ALL-NEXT:  .quad  4611686018427387904     # double 2
-; ALL-NEXT:  .quad  4607182418800017408     # double 1
-; ALL-NOT:   .quad
-
 define <4 x double> @f4xf64_f128(<4 x double> %a) {
-; ALL-LABEL: f4xf64_f128:
-; ALL:       # BB#0:
-; ALL-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
-; ALL-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
-; ALL-NEXT:    vdivpd %ymm0, %ymm1, %ymm0
-;
 ; AVX-LABEL: f4xf64_f128:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
+; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; AVX-NEXT:    # ymm1 = mem[0,1,0,1]
 ; AVX-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
 ; AVX-NEXT:    vdivpd %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    retl
+;
+; ALL32-LABEL: f4xf64_f128:
+; ALL32:       # BB#0:
+; ALL32-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; ALL32-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL32-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
+; ALL32-NEXT:    vdivpd %ymm0, %ymm1, %ymm0
+; ALL32-NEXT:    retl
+;
+; AVX-64-LABEL: f4xf64_f128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; AVX-64-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX-64-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
+; AVX-64-NEXT:    vdivpd %ymm0, %ymm1, %ymm0
+; AVX-64-NEXT:    retq
+;
+; ALL64-LABEL: f4xf64_f128:
+; ALL64:       # BB#0:
+; ALL64-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; ALL64-NEXT:    # ymm1 = mem[0,1,0,1]
+; ALL64-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
+; ALL64-NEXT:    vdivpd %ymm0, %ymm1, %ymm0
+; ALL64-NEXT:    retq
   %res1 = fadd <4 x double> <double 2.0, double 1.0, double 2.0, double 1.0>, %a
   %res2 = fdiv <4 x double> <double 2.0, double 1.0, double 2.0, double 1.0>, %res1
   ret <4 x double> %res2
 }
 
 
-; ALL:       .LCPI36
-; ALL-NEXT:  .quad  4611686018427387904     # double 2
-; ALL-NEXT:  .quad  4607182418800017408     # double 1
-; ALL-NOT:   .quad
-
 define <8 x double> @f8xf64_f128(<8 x double> %a) {
+; AVX-LABEL: f8xf64_f128:
+; AVX:       # BB#0:
+; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; AVX-NEXT:    # ymm2 = mem[0,1,0,1]
+; AVX-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
+; AVX-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX-NEXT:    retl
+;
 ; AVX2-LABEL: f8xf64_f128:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; AVX2-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    retl
 ;
 ; AVX512-LABEL: f8xf64_f128:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcastf32x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vbroadcastf32x4 {{.*#+}} zmm1 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; AVX512-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
 ; AVX512-NEXT:    vaddpd %zmm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vdivpd %zmm0, %zmm1, %zmm0
+; AVX512-NEXT:    retl
 ;
-; AVX-LABEL: f8xf64_f128:
-; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = mem[0,1,0,1]
-; AVX-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
-; AVX-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
-; AVX-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
-; AVX-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX-64-LABEL: f8xf64_f128:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; AVX-64-NEXT:    # ymm2 = mem[0,1,0,1]
+; AVX-64-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
+; AVX-64-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX-64-NEXT:    retq
+;
+; AVX2-64-LABEL: f8xf64_f128:
+; AVX2-64:       # BB#0:
+; AVX2-64-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; AVX2-64-NEXT:    # ymm2 = mem[0,1,0,1]
+; AVX2-64-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
+; AVX2-64-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX2-64-NEXT:    retq
+;
+; AVX512F-64-LABEL: f8xf64_f128:
+; AVX512F-64:       # BB#0:
+; AVX512F-64-NEXT:    vbroadcastf32x4 {{.*#+}} zmm1 = [2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00,2.000000e+00,1.000000e+00]
+; AVX512F-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    vaddpd %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    vdivpd %zmm0, %zmm1, %zmm0
+; AVX512F-64-NEXT:    retq
   %res1 = fadd <8 x double> <double 2.0, double 1.0, double 2.0, double 1.0, double 2.0, double 1.0, double 2.0, double 1.0>, %a
   %res2 = fdiv <8 x double> <double 2.0, double 1.0, double 2.0, double 1.0, double 2.0, double 1.0, double 2.0, double 1.0>, %res1
   ret <8 x double> %res2
 }
 
 
 ; AVX512:       .LCPI37
 ; AVX512-NEXT:  .quad 4616189618054758400     # double 4
 ; AVX512-NEXT:  .quad 4607182418800017408     # double 1
 ; AVX512-NEXT:  .quad 4611686018427387904     # double 2
 ; AVX512-NEXT:  .quad 4613937818241073152     # double 3
 ; AVX512-NOT:   .quad
 
 define <8 x double> @f8xf64_f256(<8 x double> %a) {
+; AVX-LABEL: f8xf64_f256:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovapd {{.*#+}} ymm2 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
+; AVX-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
+; AVX-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
+; AVX-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX-NEXT:    retl
+;
+; AVX2-LABEL: f8xf64_f256:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vmovapd {{.*#+}} ymm2 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX2-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
+; AVX2-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    retl
+;
 ; AVX512-LABEL: f8xf64_f256:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcastf64x4 {{.*#+}} zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vbroadcastf64x4 {{.*#+}} zmm1 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX512-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
 ; AVX512-NEXT:    vaddpd %zmm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vdivpd %zmm0, %zmm1, %zmm0
+; AVX512-NEXT:    retl
+;
+; AVX-64-LABEL: f8xf64_f256:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vmovapd {{.*#+}} ymm2 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX-64-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
+; AVX-64-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
+; AVX-64-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
+; AVX-64-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX-64-NEXT:    retq
+;
+; AVX2-64-LABEL: f8xf64_f256:
+; AVX2-64:       # BB#0:
+; AVX2-64-NEXT:    vmovapd {{.*#+}} ymm2 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX2-64-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vaddpd %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vdivpd %ymm0, %ymm2, %ymm0
+; AVX2-64-NEXT:    vdivpd %ymm1, %ymm2, %ymm1
+; AVX2-64-NEXT:    retq
+;
+; AVX512F-64-LABEL: f8xf64_f256:
+; AVX512F-64:       # BB#0:
+; AVX512F-64-NEXT:    vbroadcastf64x4 {{.*#+}} zmm1 = [4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,1.000000e+00,2.000000e+00,3.000000e+00]
+; AVX512F-64-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-64-NEXT:    vaddpd %zmm1, %zmm0, %zmm0
+; AVX512F-64-NEXT:    vdivpd %zmm0, %zmm1, %zmm0
+; AVX512F-64-NEXT:    retq
   %res1 = fadd <8 x double> <double 4.0, double 1.0, double 2.0, double 3.0, double 4.0, double 1.0, double 2.0, double 3.0>, %a
   %res2 = fdiv <8 x double> <double 4.0, double 1.0, double 2.0, double 3.0, double 4.0, double 1.0, double 2.0, double 3.0>, %res1
   ret <8 x double> %res2
 }
 
 
 
-; ALL:       .LCPI38
-; ALL-NEXT:  .long  4290379776              # 0xffba0000
-
-; AVX:       .LCPI38
-; AVX-NEXT:  .long  4290379776              # float NaN
-
 define <8 x i16> @f8xi16_i32_NaN(<8 x i16> %a) {
+; AVX-LABEL: f8xi16_i32_NaN:
+; AVX:       # BB#0:
+; AVX-NEXT:    vbroadcastss {{.*#+}} xmm1 = [NaN,NaN,NaN,NaN]
+; AVX-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retl
+;
 ; ALL32-LABEL: f8xi16_i32_NaN:
 ; ALL32:       # BB#0:
-; ALL32-NEXT:    vpbroadcastd {{\.LCPI.*}}, %xmm1
+; ALL32-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4290379776,4290379776,4290379776,4290379776]
 ; ALL32-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL32-NEXT:    retl
 ;
+; AVX-64-LABEL: f8xi16_i32_NaN:
+; AVX-64:       # BB#0:
+; AVX-64-NEXT:    vbroadcastss {{.*#+}} xmm1 = [NaN,NaN,NaN,NaN]
+; AVX-64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-64-NEXT:    retq
+;
 ; ALL64-LABEL: f8xi16_i32_NaN:
 ; ALL64:       # BB#0:
-; ALL64-NEXT:    vpbroadcastd {{.*}}(%rip), %xmm1
+; ALL64-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4290379776,4290379776,4290379776,4290379776]
 ; ALL64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; ALL64-NEXT:    retq
-;
-; AVX-LABEL: f8xi16_i32_NaN:
-; AVX:       # BB#0:
-; AVX-NEXT:    vbroadcastss {{\.LCPI.*}}, %xmm1
-; AVX-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
   %res1 = add <8 x i16> <i16 0, i16 -70, i16 0, i16 -70, i16 0, i16 -70, i16 0, i16 -70>, %a
   %res2 = and <8 x i16> <i16 0, i16 -70, i16 0, i16 -70, i16 0, i16 -70, i16 0, i16 -70>, %res1
   ret <8 x i16> %res2
 }

llvm/trunk/test/CodeGen/X86/recip-fastmath.ll

 ; SANDY-LABEL: v4f32_no_estimate:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vmovaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
 ; SANDY-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v4f32_no_estimate:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm1 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_no_estimate:
 ; HASWELL-NO-FMA:       # BB#0:
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm1
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1,1,1,1]
 ; HASWELL-NO-FMA-NEXT:    vdivps %xmm0, %xmm1, %xmm0
 ; HASWELL-NO-FMA-NEXT:    retq
 ;
 ; AVX512-LABEL: v4f32_no_estimate:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcastss {{.*}}(%rip), %xmm1 # sched: [4:0.50]
+; AVX512-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1,1,1,1] sched: [4:0.50]
 ; AVX512-NEXT:    vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00]
 ; AVX512-NEXT:    retq # sched: [1:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <4 x float> %div
 }
 
 define <4 x float> @v4f32_one_step(<4 x float> %x) #1 {
 ; SSE-LABEL: v4f32_one_step:

 ; SANDY-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v4f32_one_step:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
 ; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_one_step:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm2, %xmm0
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0
 ; HASWELL-NO-FMA-NEXT:    retq
 ;
 ; KNL-LABEL: v4f32_one_step:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
 ; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v4f32_one_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1
 ; SKX-NEXT:    vfnmadd213ps {{.*}}(%rip){1to4}, %xmm1, %xmm0

 ; SANDY-NEXT:    vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v4f32_two_step:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
 ; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
 ; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
 ; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
 ; HASWELL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_two_step:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm2
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm3
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm3 = [1,1,1,1]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm2, %xmm3, %xmm2
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm2, %xmm1, %xmm2
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm2, %xmm1, %xmm1
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm3, %xmm0
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0
 ; HASWELL-NO-FMA-NEXT:    retq
 ;
 ; KNL-LABEL: v4f32_two_step:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; KNL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
 ; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
 ; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
 ; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
 ; KNL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v4f32_two_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1
-; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; SKX-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; SKX-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
 ; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
 ; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
 ; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
 ; SKX-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
 ; SKX-NEXT:    retq # sched: [1:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <4 x float> %div

 ; SANDY-LABEL: v8f32_no_estimate:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vmovaps {{.*#+}} ymm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50]
 ; SANDY-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [12:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v8f32_no_estimate:
 ; HASWELL:       # BB#0:
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm1 # sched: [5:1.00]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [19:2.00]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_no_estimate:
 ; HASWELL-NO-FMA:       # BB#0:
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm1
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
 ; HASWELL-NO-FMA-NEXT:    vdivps %ymm0, %ymm1, %ymm0
 ; HASWELL-NO-FMA-NEXT:    retq
 ;
 ; AVX512-LABEL: v8f32_no_estimate:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcastss {{.*}}(%rip), %ymm1 # sched: [5:1.00]
+; AVX512-NEXT:    vbroadcastss {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; AVX512-NEXT:    vdivps %ymm0, %ymm1, %ymm0 # sched: [19:2.00]
 ; AVX512-NEXT:    retq # sched: [1:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <8 x float> %div
 }
 
 define <8 x float> @v8f32_one_step(<8 x float> %x) #1 {
 ; SSE-LABEL: v8f32_one_step:

 ; SANDY-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v8f32_one_step:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
 ; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_one_step:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm2, %ymm0
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0
 ; HASWELL-NO-FMA-NEXT:    retq
 ;
 ; KNL-LABEL: v8f32_one_step:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; KNL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
 ; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v8f32_one_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1
 ; SKX-NEXT:    vfnmadd213ps {{.*}}(%rip){1to8}, %ymm1, %ymm0

 ; SANDY-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v8f32_two_step:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
 ; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
 ; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
 ; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
 ; HASWELL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_two_step:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm2
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm3
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm2, %ymm3, %ymm2
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm1, %ymm2
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm2, %ymm1, %ymm1
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm3, %ymm0
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0
 ; HASWELL-NO-FMA-NEXT:    retq
 ;
 ; KNL-LABEL: v8f32_two_step:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; KNL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; KNL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
 ; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
 ; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
 ; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
 ; KNL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v8f32_two_step:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1
-; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; SKX-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; SKX-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
 ; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
 ; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
 ; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
 ; SKX-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
 ; SKX-NEXT:    retq # sched: [1:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
   ret <8 x float> %div
 }
 
 attributes #0 = { "unsafe-fp-math"="true" "reciprocal-estimates"="!divf,!vec-divf" }
 attributes #1 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf,vec-divf" }
 attributes #2 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf:2,vec-divf:2" }
 

llvm/trunk/test/CodeGen/X86/recip-fastmath2.ll

 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v4f32_one_step2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
 ; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
 ; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_one_step2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
 ; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
 ;
 ; KNL-LABEL: v4f32_one_step2:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
 ; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
 ; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v4f32_one_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1

 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:1.00]
 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v4f32_one_step_2_divs:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
 ; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
 ; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
 ; HASWELL-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_one_step_2_divs:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
 ;
 ; KNL-LABEL: v4f32_one_step_2_divs:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0
 ; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm0
 ; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
 ; KNL-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v4f32_one_step_2_divs:
 ; SKX:       # BB#0:

 ; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v4f32_two_step2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
 ; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
 ; HASWELL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
 ; HASWELL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
 ; HASWELL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
 ; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v4f32_two_step2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm2 # sched: [5:0.50]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %xmm3 # sched: [4:0.50]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm3 = [1,1,1,1] sched: [4:0.50]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm2, %xmm1, %xmm2 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
 ; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
 ; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
 ;
 ; KNL-LABEL: v4f32_two_step2:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; KNL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; KNL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
 ; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
 ; KNL-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
 ; KNL-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
 ; KNL-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
 ; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v4f32_two_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %xmm0, %xmm1
-; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50]
+; SKX-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1,1,1,1] sched: [4:0.50]
 ; SKX-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
 ; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm0, %xmm3
 ; SKX-NEXT:    vfmadd132ps %xmm1, %xmm1, %xmm3
 ; SKX-NEXT:    vfnmadd213ps %xmm2, %xmm3, %xmm0
 ; SKX-NEXT:    vfmadd132ps %xmm3, %xmm3, %xmm0
 ; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
 ; SKX-NEXT:    retq # sched: [1:1.00]
   %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x

 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v8f32_one_step2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
 ; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
 ; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_one_step2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
 ;
 ; KNL-LABEL: v8f32_one_step2:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; KNL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
 ; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
 ; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v8f32_one_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1

 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v8f32_one_step_2_divs:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
 ; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
 ; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
 ; HASWELL-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_one_step_2_divs:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
 ;
 ; KNL-LABEL: v8f32_one_step_2_divs:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; KNL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0
 ; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm0
 ; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00]
 ; KNL-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v8f32_one_step_2_divs:
 ; SKX:       # BB#0:

 ; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; SANDY-NEXT:    retq # sched: [5:1.00]
 ;
 ; HASWELL-LABEL: v8f32_two_step2:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; HASWELL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
 ; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
 ; HASWELL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
 ; HASWELL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
 ; HASWELL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
 ; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-NO-FMA-LABEL: v8f32_two_step2:
 ; HASWELL-NO-FMA:       # BB#0:
 ; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00]
-; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*}}(%rip), %ymm3 # sched: [5:1.00]
+; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
 ; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; HASWELL-NO-FMA-NEXT:    retq # sched: [1:1.00]
 ;
 ; KNL-LABEL: v8f32_two_step2:
 ; KNL:       # BB#0:
 ; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
-; KNL-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; KNL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; KNL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
 ; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
 ; KNL-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
 ; KNL-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
 ; KNL-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
 ; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; KNL-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: v8f32_two_step2:
 ; SKX:       # BB#0:
 ; SKX-NEXT:    vrcp14ps %ymm0, %ymm1
-; SKX-NEXT:    vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00]
+; SKX-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00]
 ; SKX-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
 ; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm0, %ymm3
 ; SKX-NEXT:    vfmadd132ps %ymm1, %ymm1, %ymm3
 ; SKX-NEXT:    vfnmadd213ps %ymm2, %ymm3, %ymm0
 ; SKX-NEXT:    vfmadd132ps %ymm3, %ymm3, %ymm0
 ; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00]
 ; SKX-NEXT:    retq # sched: [1:1.00]
   %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x

llvm/trunk/test/CodeGen/X86/vec_shift6.ll

 ; SSE-NEXT:    pmullw %xmm4, %xmm0
 ; SSE-NEXT:    pmullw %xmm4, %xmm1
 ; SSE-NEXT:    pmullw %xmm4, %xmm2
 ; SSE-NEXT:    pmullw %xmm4, %xmm3
 ; SSE-NEXT:    retq
 ;
 ; AVX2-LABEL: test7:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [2,2,4,8,128,1,512,2048,2,2,4,8,128,1,512,2048]
+; AVX2-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-NEXT:    vpmullw %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vpmullw %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test7:
 ; AVX512:       # BB#0:
-; AVX512-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; AVX512-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [2,2,4,8,128,1,512,2048,2,2,4,8,128,1,512,2048]
+; AVX512-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX512-NEXT:    vpmullw %ymm2, %ymm0, %ymm0
 ; AVX512-NEXT:    vpmullw %ymm2, %ymm1, %ymm1
 ; AVX512-NEXT:    retq
   %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <32 x i16> %shl
 }
 
 ; Similar to test7; the difference is that with AVX512 support
 ; we only produce a single vpsllvd/vpsllvq instead of a pair of vpsllvd/vpsllvq.
 
 define <16 x i32> @test8(<16 x i32> %a) {
 ; SSE-LABEL: test8:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [2,2,4,8]
 ; SSE-NEXT:    pmulld %xmm4, %xmm0
 ; SSE-NEXT:    pmulld %xmm4, %xmm1
 ; SSE-NEXT:    pmulld %xmm4, %xmm2
 ; SSE-NEXT:    pmulld %xmm4, %xmm3
 ; SSE-NEXT:    retq
 ;
 ; AVX2-LABEL: test8:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = mem[0,1,0,1]
+; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [1,1,2,3,1,1,2,3]
+; AVX2-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-NEXT:    vpsllvd %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vpsllvd %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test8:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vpsllvd {{.*}}(%rip), %zmm0, %zmm0
 ; AVX512-NEXT:    retq