[DAG] MatchRotate - Add funnel shift by immediate support

[RKSimon]

Summary

This patch reuses the existing MatchRotate ROTL/ROTR rotation pattern code to also recognize the more general FSHL/FSHR funnel shift patterns when we have constant shift amounts.

[lebedev.ri]

llvm/test/CodeGen/X86/avg.ll

2708–2736

All this doesn't look like an improvement,
but i suspect this is an unrelated codegen issue
that is only being exposed by the change.

llvm/test/CodeGen/X86/bitreverse.ll

1114

Technically this implies more registers are being used here,
not sure what that says about the actual asm change.

llvm/test/CodeGen/X86/known-bits.ll

123

Demandedbits failure? (no longer narrowed into 32-bit)

[RKSimon]

llvm/lib/Target/Hexagon/HexagonPatterns.td

1098

@kparzysz Any comments on this? Should I split off a patch with a fshl specific test case?

[RKSimon]

llvm/test/CodeGen/X86/shift-combine.ll

270

This should be addressed by D75624

[RKSimon]

rebased

[kparzysz]

The Hexagon changes LGTM.

llvm/lib/Target/Hexagon/HexagonPatterns.td

1098

You are correct. This should be A2_combine_lh.

[RKSimon]

llvm/lib/Target/Hexagon/HexagonPatterns.td

1098

Cheers, I'll create a test and commit the fix shortly.

[RKSimon]

Rebased now that the Hexagon bugfix has landed - thanks @kparzysz

[RKSimon]

rebase - now that D75748 has landed, most of the regressions have been solved.

[lebedev.ri]

Looks good to me.
Anyone else?

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

6410–6411

/// ... where C1+C2 == bitwidth(x)

llvm/test/CodeGen/X86/known-bits.ll

123

Still applicable

[RKSimon]

llvm/test/CodeGen/X86/known-bits.ll

123

I'm still looking at the best way to handle overflow ops with SimplifyDemandedBits/ShrinkDemandedOp

[spatel]

LGTM - see inline for a nit.

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

6417–6418

Nit: the code structure of these checks seems odd to me.
If we're ok with using the "?:" to choose L/R for rotate, do the same for fsh?

if (IsRotate && (HasROTL || HasROTR))
  Res = DAG.getNode(HasROTL ? ISD::ROTL : ISD::ROTR, DL, VT, LHSShiftArg,
                    HasROTL ? LHSShiftAmt : RHSShiftAmt);
else
  Res = DAG.getNode(HasFSHL ? ISD::FSHL : ISD::FSHR, DL, VT, LHSShiftArg,
                    RHSShiftArg, HasFSHL ? LHSShiftAmt : RHSShiftAmt);

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

return DAG.getNode(HasPos ? PosOpcode : NegOpcode, DL, VT, Shifted,
                        HasPos ? Pos : Neg);
   }
 
   return SDValue();
 }
 
 // MatchRotate - Handle an 'or' of two operands.  If this is one of the many
 // idioms for rotate, and if the target supports rotation instructions, generate
-// a rot[lr].
+// a rot[lr]. This also matches funnel shift patterns, similar to rotation but
+// with different shifted sources.
 SDValue DAGCombiner::MatchRotate(SDValue LHS, SDValue RHS, const SDLoc &DL) {
   // Must be a legal type.  Expanded 'n promoted things won't work with rotates.
   EVT VT = LHS.getValueType();
   if (!TLI.isTypeLegal(VT))
     return SDValue();
 
-  // The target must have at least one rotate flavor.
+  // The target must have at least one rotate/funnel flavor.
   bool HasROTL = hasOperation(ISD::ROTL, VT);
   bool HasROTR = hasOperation(ISD::ROTR, VT);
-  if (!HasROTL && !HasROTR)
+  bool HasFSHL = hasOperation(ISD::FSHL, VT);
+  bool HasFSHR = hasOperation(ISD::FSHR, VT);
+  if (!HasROTL && !HasROTR && !HasFSHL && !HasFSHR)
     return SDValue();
 
   // Check for truncated rotate.
   if (LHS.getOpcode() == ISD::TRUNCATE && RHS.getOpcode() == ISD::TRUNCATE &&
       LHS.getOperand(0).getValueType() == RHS.getOperand(0).getValueType()) {
     assert(LHS.getValueType() == RHS.getValueType());
     if (SDValue Rot = MatchRotate(LHS.getOperand(0), RHS.getOperand(0), DL)) {
       return DAG.getNode(ISD::TRUNCATE, SDLoc(LHS), LHS.getValueType(), Rot);
if (SDValue NewLHSShift =
       LHSShift = NewLHSShift;
 
   // If a side is still missing, nothing else we can do.
   if (!RHSShift || !LHSShift)
     return SDValue();
 
   // At this point we've matched or extracted a shift op on each side.
 
-  if (LHSShift.getOperand(0) != RHSShift.getOperand(0))
-    return SDValue(); // Not shifting the same value.
-
   if (LHSShift.getOpcode() == RHSShift.getOpcode())
     return SDValue(); // Shifts must disagree.
 
+  bool IsRotate = LHSShift.getOperand(0) == RHSShift.getOperand(0);
+  if (!IsRotate && !(HasFSHL || HasFSHR))
+    return SDValue(); // Requires funnel shift support.
+
   // Canonicalize shl to left side in a shl/srl pair.
   if (RHSShift.getOpcode() == ISD::SHL) {
     std::swap(LHS, RHS);
     std::swap(LHSShift, RHSShift);
     std::swap(LHSMask, RHSMask);
   }
 
   unsigned EltSizeInBits = VT.getScalarSizeInBits();
   SDValue LHSShiftArg = LHSShift.getOperand(0);
   SDValue LHSShiftAmt = LHSShift.getOperand(1);
   SDValue RHSShiftArg = RHSShift.getOperand(0);
   SDValue RHSShiftAmt = RHSShift.getOperand(1);
 
   // fold (or (shl x, C1), (srl x, C2)) -> (rotl x, C1)
   // fold (or (shl x, C1), (srl x, C2)) -> (rotr x, C2)
+  // fold (or (shl x, C1), (srl y, C2)) -> (fshl x, y, C1)
+  // fold (or (shl x, C1), (srl y, C2)) -> (fshr x, y, C2)
   auto MatchRotateSum = [EltSizeInBits](ConstantSDNode *LHS,
                                         ConstantSDNode *RHS) {
     return (LHS->getAPIntValue() + RHS->getAPIntValue()) == EltSizeInBits;
   };
   if (ISD::matchBinaryPredicate(LHSShiftAmt, RHSShiftAmt, MatchRotateSum)) {
-    SDValue Rot = DAG.getNode(HasROTL ? ISD::ROTL : ISD::ROTR, DL, VT,
-                              LHSShiftArg, HasROTL ? LHSShiftAmt : RHSShiftAmt);
+    SDValue Res;
+    if (IsRotate && (HasROTL || HasROTR))
+      Res = DAG.getNode(HasROTL ? ISD::ROTL : ISD::ROTR, DL, VT,
+    else if (HasFSHL)
+      Res = DAG.getNode(ISD::FSHL, DL, VT, LHSShiftArg, RHSShiftArg,
+                        LHSShiftAmt);
+    else
+      Res = DAG.getNode(ISD::FSHR, DL, VT, LHSShiftArg, RHSShiftArg,
+                        RHSShiftAmt);
 
     // If there is an AND of either shifted operand, apply it to the result.
     if (LHSMask.getNode() || RHSMask.getNode()) {
       SDValue AllOnes = DAG.getAllOnesConstant(DL, VT);
       SDValue Mask = AllOnes;
 
       if (LHSMask.getNode()) {
         SDValue RHSBits = DAG.getNode(ISD::SRL, DL, VT, AllOnes, RHSShiftAmt);
         Mask = DAG.getNode(ISD::AND, DL, VT, Mask,
                            DAG.getNode(ISD::OR, DL, VT, LHSMask, RHSBits));
       }
       if (RHSMask.getNode()) {
         SDValue LHSBits = DAG.getNode(ISD::SHL, DL, VT, AllOnes, LHSShiftAmt);
         Mask = DAG.getNode(ISD::AND, DL, VT, Mask,
                            DAG.getNode(ISD::OR, DL, VT, RHSMask, LHSBits));
       }
 
-      Rot = DAG.getNode(ISD::AND, DL, VT, Rot, Mask);
+      Res = DAG.getNode(ISD::AND, DL, VT, Res, Mask);
     }
 
-    return Rot;
+    return Res;
   }
 
+  // TODO: Handle variable funnel shifts.
+  if (!IsRotate)
+    return SDValue();
+
   // If there is a mask here, and we have a variable shift, we can't be sure
   // that we're masking out the right stuff.
   if (LHSMask.getNode() || RHSMask.getNode())
     return SDValue();
 
   // If the shift amount is sign/zext/any-extended just peel it off.
   SDValue LExtOp0 = LHSShiftAmt;
   SDValue RExtOp0 = RHSShiftAmt;

llvm/lib/Target/Hexagon/HexagonPatterns.td

 def Divu64_8: SDNodeXForm<imm, [{
   return CurDAG->getTargetConstant((64 - N->getSExtValue()) / 8,
                                    SDLoc(N), MVT::i32);
 }]>;
 
 // Special cases:
 let AddedComplexity = 100 in {
   def: Pat<(fshl I32:$Rs, I32:$Rt, (i32 16)),
-           (A2_combine_hl I32:$Rs, I32:$Rt)>;
+           (A2_combine_hl I32:$Rs, I32:$Rt)>; // A2_combine_lh?
   def: Pat<(fshl I64:$Rs, I64:$Rt, IsMul8_U3:$S),
            (S2_valignib I64:$Rs, I64:$Rt, (Divu64_8 $S))>;
 }
 
 let Predicates = [HasV60], AddedComplexity = 50 in {
   def: OpR_RI_pat<S6_rol_i_r, Rol, i32, I32, u5_0ImmPred>;
   def: OpR_RI_pat<S6_rol_i_p, Rol, i64, I64, u6_0ImmPred>;
 }

llvm/test/CodeGen/Hexagon/isel-combine-half.ll

 ; RUN: llc -march=hexagon < %s | FileCheck %s
 
 ; CHECK-LABEL: A2_combine_ll:
 ; CHECK: combine(r1.l,r0.l)
 define i32 @A2_combine_ll(i32 %a0, i32 %a1) #0 {
 b2:
   %v3 = and i32 %a0, 65535
   %v4 = shl i32 %a1, 16
   %v5 = or i32 %v3, %v4
   ret i32 %v5
 }
 
 ; CHECK-LABEL: A2_combine_lh:
-; CHECK: combine(r1.l,r0.h)
+; CHECK: combine(r1.h,r0.l)
 define i32 @A2_combine_lh(i32 %a0, i32 %a1) #0 {
 b2:
   %v3 = lshr i32 %a0, 16
   %v4 = shl i32 %a1, 16
   %v5 = or i32 %v4, %v3
   ret i32 %v5
 }
 

llvm/test/CodeGen/X86/avg.ll

 ; SSE2-LABEL: not_avg_v16i8_wide_constants:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pushq %rbp
 ; SSE2-NEXT:    pushq %r15
 ; SSE2-NEXT:    pushq %r14
 ; SSE2-NEXT:    pushq %r13
 ; SSE2-NEXT:    pushq %r12
 ; SSE2-NEXT:    pushq %rbx
+; SSE2-NEXT:    subq $56, %rsp
 ; SSE2-NEXT:    movaps (%rdi), %xmm1
 ; SSE2-NEXT:    movaps (%rsi), %xmm0
 ; SSE2-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %eax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %eax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %eax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %eax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %eax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %edi
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ecx
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebp
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %edx
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r13d
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r12d
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r15d
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r11d
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r8d
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r10d
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r15d
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebx
+; SSE2-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebx
+; SSE2-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebx
+; SSE2-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebx
+; SSE2-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebx
+; SSE2-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebx
+; SSE2-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r9d
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r8d
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %edx
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ecx
+; SSE2-NEXT:    addq %rax, %r9
+; SSE2-NEXT:    movq %r9, %rbx
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %eax
+; SSE2-NEXT:    addq %rdi, %rax
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %edi
+; SSE2-NEXT:    addq %rcx, %rdi
+; SSE2-NEXT:    movq %rdi, %r14
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %edi
+; SSE2-NEXT:    addq %rbp, %rdi
+; SSE2-NEXT:    movq %rdi, %r11
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r12d
+; SSE2-NEXT:    addq %rsi, %r12
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebp
-; SSE2-NEXT:    addq %r11, %rbp
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r14d
-; SSE2-NEXT:    addq %r10, %r14
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ebx
-; SSE2-NEXT:    addq %r9, %rbx
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r11d
-; SSE2-NEXT:    addq %r8, %r11
+; SSE2-NEXT:    addq %rdx, %rsi
+; SSE2-NEXT:    movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %edx
+; SSE2-NEXT:    addq %r13, %rdx
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ecx
+; SSE2-NEXT:    addq %r8, %rcx
+; SSE2-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ecx
+; SSE2-NEXT:    addq %r10, %rcx
+; SSE2-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ecx
+; SSE2-NEXT:    addq %r15, %rcx
+; SSE2-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ecx
+; SSE2-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; SSE2-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %ecx
+; SSE2-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r10d
-; SSE2-NEXT:    addq %rdx, %r10
+; SSE2-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %r10 # 8-byte Folded Reload
+; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r9d
+; SSE2-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %r9 # 8-byte Folded Reload
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %r8d
-; SSE2-NEXT:    addq %rcx, %r8
+; SSE2-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %r8 # 8-byte Folded Reload
 ; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %edi
-; SSE2-NEXT:    addq %rax, %rdi
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %edx
-; SSE2-NEXT:    addq %rsi, %rdx
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    leaq -1(%r15,%rsi), %rax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    leaq -1(%r12,%rsi), %rax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    leaq -1(%r13,%rsi), %rax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; SSE2-NEXT:    leaq -1(%rax,%rsi), %rax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; SSE2-NEXT:    leaq -1(%rax,%rsi), %rax
-; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; SSE2-NEXT:    leaq -1(%rax,%rsi), %rax
+; SSE2-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Folded Reload
+; SSE2-NEXT:    addq $-1, %rbx
+; SSE2-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movl $0, %ebp
+; SSE2-NEXT:    adcq $-1, %rbp
+; SSE2-NEXT:    addq $-1, %rax
 ; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; SSE2-NEXT:    leaq -1(%rax,%rsi), %rsi
-; SSE2-NEXT:    movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    movzbl -{{[0-9]+}}(%rsp), %esi
-; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; SSE2-NEXT:    leaq -1(%rax,%rsi), %rsi
-; SSE2-NEXT:    movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    addq $-1, %rbp
-; SSE2-NEXT:    movl $0, %r9d
-; SSE2-NEXT:    adcq $-1, %r9
+; SSE2-NEXT:    movl $0, %ebx
+; SSE2-NEXT:    adcq $-1, %rbx
 ; SSE2-NEXT:    addq $-1, %r14
-; SSE2-NEXT:    movl $0, %esi
-; SSE2-NEXT:    adcq $-1, %rsi
-; SSE2-NEXT:    addq $-1, %rbx
+; SSE2-NEXT:    movq %r14, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movl $0, %r13d
+; SSE2-NEXT:    adcq $-1, %r13
+; SSE2-NEXT:    addq $-1, %r11
+; SSE2-NEXT:    movq %r11, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movl $0, %r11d
+; SSE2-NEXT:    adcq $-1, %r11
+; SSE2-NEXT:    addq $-1, %r12
+; SSE2-NEXT:    movq %r12, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; SSE2-NEXT:    movl $0, %eax
 ; SSE2-NEXT:    adcq $-1, %rax
 ; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; SSE2-NEXT:    addq $-1, %r11
+; SSE2-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; SSE2-NEXT:    movl $0, %r15d
+; SSE2-NEXT:    adcq $-1, %r15
+; SSE2-NEXT:    addq $-1, %rdx
+; SSE2-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; SSE2-NEXT:    movl $0, %r12d
 ; SSE2-NEXT:    adcq $-1, %r12
+; SSE2-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; SSE2-NEXT:    movl $0, %r14d
+; SSE2-NEXT:    adcq $-1, %r14
+; SSE2-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; SSE2-NEXT:    movl $0, %eax
+; SSE2-NEXT:    adcq $-1, %rax
+; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; SSE2-NEXT:    movl $0, %eax
+; SSE2-NEXT:    adcq $-1, %rax
+; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; SSE2-NEXT:    movl $0, %eax
+; SSE2-NEXT:    adcq $-1, %rax
+; SSE2-NEXT:    movq %rax, (%rsp) # 8-byte Spill
+; SSE2-NEXT:    addq $-1, %rcx
+; SSE2-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    movl $0, %eax
+; SSE2-NEXT:    adcq $-1, %rax
+; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; SSE2-NEXT:    addq $-1, %r10
-; SSE2-NEXT:    movl $0, %r13d
-; SSE2-NEXT:    adcq $-1, %r13
+; SSE2-NEXT:    movl $0, %esi
+; SSE2-NEXT:    adcq $-1, %rsi
+; SSE2-NEXT:    addq $-1, %r9
+; SSE2-NEXT:    movl $0, %edx
+; SSE2-NEXT:    adcq $-1, %rdx
 ; SSE2-NEXT:    addq $-1, %r8
-; SSE2-NEXT:    movl $0, %r15d
-; SSE2-NEXT:    adcq $-1, %r15
-; SSE2-NEXT:    addq $-1, %rdi
 ; SSE2-NEXT:    movl $0, %ecx
 ; SSE2-NEXT:    adcq $-1, %rcx
-; SSE2-NEXT:    addq $-1, %rdx
+; SSE2-NEXT:    addq $-1, %rdi
 ; SSE2-NEXT:    movl $0, %eax
 ; SSE2-NEXT:    adcq $-1, %rax
-; SSE2-NEXT:    shldq $63, %rdx, %rax
-; SSE2-NEXT:    shldq $63, %rdi, %rcx
-; SSE2-NEXT:    movq %rcx, %rdx
-; SSE2-NEXT:    shldq $63, %r8, %r15
-; SSE2-NEXT:    shldq $63, %r10, %r13
-; SSE2-NEXT:    shldq $63, %r11, %r12
+; SSE2-NEXT:    shldq $63, %rdi, %rax
+; SSE2-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    shldq $63, %r8, %rcx
+; SSE2-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; SSE2-NEXT:    shldq $63, %r9, %rdx
+; SSE2-NEXT:    movq %rdx, %r9
+; SSE2-NEXT:    shldq $63, %r10, %rsi
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r10 # 8-byte Reload
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; SSE2-NEXT:    shldq $63, %rax, %r10
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NEXT:    movq (%rsp), %r8 # 8-byte Reload
+; SSE2-NEXT:    shldq $63, %rcx, %r8
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
 ; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
-; SSE2-NEXT:    shldq $63, %rbx, %rdi
-; SSE2-NEXT:    shldq $63, %r14, %rsi
-; SSE2-NEXT:    shldq $63, %rbp, %r9
-; SSE2-NEXT:    movq %r9, %xmm8
-; SSE2-NEXT:    movq %rsi, %xmm15
+; SSE2-NEXT:    shldq $63, %rcx, %rdi
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; SSE2-NEXT:    shldq $63, %rcx, %rdx
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; SSE2-NEXT:    shldq $63, %rcx, %r14
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; SSE2-NEXT:    shldq $63, %rax, %r12
 ; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
-; SSE2-NEXT:    shrq %rcx
-; SSE2-NEXT:    movq %rcx, %xmm9
+; SSE2-NEXT:    shldq $63, %rcx, %r15
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
 ; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
-; SSE2-NEXT:    shrq %rcx
-; SSE2-NEXT:    movq %rcx, %xmm2
+; SSE2-NEXT:    shldq $63, %rcx, %rax
 ; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
-; SSE2-NEXT:    shrq %rcx
-; SSE2-NEXT:    movq %rcx, %xmm10
+; SSE2-NEXT:    shldq $63, %rcx, %r11
 ; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
-; SSE2-NEXT:    shrq %rcx
-; SSE2-NEXT:    movq %rcx, %xmm4
+; SSE2-NEXT:    shldq $63, %rcx, %r13
 ; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
-; SSE2-NEXT:    shrq %rcx
-; SSE2-NEXT:    movq %rcx, %xmm11
+; SSE2-NEXT:    shldq $63, %rcx, %rbx
 ; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
-; SSE2-NEXT:    shrq %rcx
-; SSE2-NEXT:    movq %rcx, %xmm7
-; SSE2-NEXT:    movq %rdi, %xmm12
-; SSE2-NEXT:    movq %r12, %xmm0
-; SSE2-NEXT:    movq %r13, %xmm13
-; SSE2-NEXT:    movq %r15, %xmm6
-; SSE2-NEXT:    movq %rdx, %xmm14
-; SSE2-NEXT:    movq %rax, %xmm5
-; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; SSE2-NEXT:    shrq %rax
-; SSE2-NEXT:    movq %rax, %xmm3
-; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; SSE2-NEXT:    shrq %rax
-; SSE2-NEXT:    movq %rax, %xmm1
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3],xmm15[4],xmm8[4],xmm15[5],xmm8[5],xmm15[6],xmm8[6],xmm15[7],xmm8[7]
+; SSE2-NEXT:    shldq $63, %rcx, %rbp
+; SSE2-NEXT:    movq %rbp, %xmm8
+; SSE2-NEXT:    movq %rbx, %xmm0
+; SSE2-NEXT:    movq %r13, %xmm9
+; SSE2-NEXT:    movq %r11, %xmm2
+; SSE2-NEXT:    movq %rax, %xmm10
+; SSE2-NEXT:    movq %r15, %xmm4
+; SSE2-NEXT:    movq %r12, %xmm11
+; SSE2-NEXT:    movq %r14, %xmm7
+; SSE2-NEXT:    movq %rdx, %xmm12
+; SSE2-NEXT:    movq %rdi, %xmm1
+; SSE2-NEXT:    movq %r8, %xmm13
+; SSE2-NEXT:    movq %r10, %xmm6
+; SSE2-NEXT:    movq %rsi, %xmm14
+; SSE2-NEXT:    movq %r9, %xmm5
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 8-byte Folded Reload
+; SSE2-NEXT:    # xmm15 = mem[0],zero
+; SSE2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 8-byte Folded Reload
+; SSE2-NEXT:    # xmm3 = mem[0],zero
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm15[0,1,2,0]
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm8
-; SSE2-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1]
-; SSE2-NEXT:    por %xmm8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE2-NEXT:    pand %xmm8, %xmm0
+; SSE2-NEXT:    pslld $16, %xmm2
+; SSE2-NEXT:    pandn %xmm2, %xmm8
+; SSE2-NEXT:    por %xmm0, %xmm8
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm10[0],xmm4[1],xmm10[1],xmm4[2],xmm10[2],xmm4[3],xmm10[3],xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
-; SSE2-NEXT:    pslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1,2,3,4,5]
+; SSE2-NEXT:    psllq $48, %xmm4
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm11[0],xmm7[1],xmm11[1],xmm7[2],xmm11[2],xmm7[3],xmm11[3],xmm7[4],xmm11[4],xmm7[5],xmm11[5],xmm7[6],xmm11[6],xmm7[7],xmm11[7]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,0,1]
-; SSE2-NEXT:    pand %xmm8, %xmm7
-; SSE2-NEXT:    pandn %xmm4, %xmm8
-; SSE2-NEXT:    por %xmm7, %xmm8
-; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,1,2,2]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3],xmm6[4],xmm13[4],xmm6[5],xmm13[5],xmm6[6],xmm13[6],xmm6[7],xmm13[7]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE2-NEXT:    pand %xmm2, %xmm0
-; SSE2-NEXT:    pslld $16, %xmm6
-; SSE2-NEXT:    pandn %xmm6, %xmm2
-; SSE2-NEXT:    por %xmm0, %xmm2
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1],xmm5[2],xmm14[2],xmm5[3],xmm14[3],xmm5[4],xmm14[4],xmm5[5],xmm14[5],xmm5[6],xmm14[6],xmm5[7],xmm14[7]
-; SSE2-NEXT:    psllq $48, %xmm5
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE2-NEXT:    pand %xmm0, %xmm1
-; SSE2-NEXT:    pandn %xmm5, %xmm0
-; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,0,1,1]
+; SSE2-NEXT:    pand %xmm0, %xmm2
+; SSE2-NEXT:    pandn %xmm4, %xmm0
+; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3]
-; SSE2-NEXT:    movups %xmm2, (%rax)
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3],xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3],xmm6[4],xmm13[4],xmm6[5],xmm13[5],xmm6[6],xmm13[6],xmm6[7],xmm13[7]
+; SSE2-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,1,2,0]
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1],xmm5[2],xmm14[2],xmm5[3],xmm14[3],xmm5[4],xmm14[4],xmm5[5],xmm14[5],xmm5[6],xmm14[6],xmm5[7],xmm14[7]
+; SSE2-NEXT:    pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1],xmm3[2],xmm15[2],xmm3[3],xmm15[3],xmm3[4],xmm15[4],xmm3[5],xmm15[5],xmm3[6],xmm15[6],xmm3[7],xmm15[7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,1]
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    pandn %xmm5, %xmm1
+; SSE2-NEXT:    por %xmm2, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
+; SSE2-NEXT:    movupd %xmm1, (%rax)
+; SSE2-NEXT:    addq $56, %rsp
 ; SSE2-NEXT:    popq %rbx
 ; SSE2-NEXT:    popq %r12
 ; SSE2-NEXT:    popq %r13
 ; SSE2-NEXT:    popq %r14
 ; SSE2-NEXT:    popq %r15
 ; SSE2-NEXT:    popq %rbp
 ; SSE2-NEXT:    retq
 ;
 ; AVX1-LABEL: not_avg_v16i8_wide_constants:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    pushq %rbp
 ; AVX1-NEXT:    pushq %r15
 ; AVX1-NEXT:    pushq %r14
 ; AVX1-NEXT:    pushq %r13
 ; AVX1-NEXT:    pushq %r12
 ; AVX1-NEXT:    pushq %rbx
+; AVX1-NEXT:    pushq %rax
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm3 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm5 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm5[2],xmm3[2],xmm5[3],xmm3[3]
-; AVX1-NEXT:    vmovq %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
-; AVX1-NEXT:    vpextrq $1, %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
-; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm6 = xmm5[2],xmm3[2],xmm5[3],xmm3[3]
-; AVX1-NEXT:    vmovq %xmm6, %r10
-; AVX1-NEXT:    vpextrq $1, %xmm6, %r9
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm7 = xmm6[0],zero,xmm6[1],zero
-; AVX1-NEXT:    vmovq %xmm7, %r8
-; AVX1-NEXT:    vpextrq $1, %xmm7, %rdi
-; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
-; AVX1-NEXT:    vpextrq $1, %xmm6, %rcx
-; AVX1-NEXT:    vmovq %xmm6, %r14
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm6 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
-; AVX1-NEXT:    vpextrq $1, %xmm6, %rax
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm6 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; AVX1-NEXT:    vpextrq $1, %xmm6, %rdi
 ; AVX1-NEXT:    vmovq %xmm6, %rbp
 ; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero
-; AVX1-NEXT:    vpextrq $1, %xmm5, %r11
-; AVX1-NEXT:    vmovq %xmm5, %r15
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm8 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; AVX1-NEXT:    vpextrq $1, %xmm5, %rbx
+; AVX1-NEXT:    vmovq %xmm5, %rsi
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm5 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; AVX1-NEXT:    vpextrq $1, %xmm5, %rdx
+; AVX1-NEXT:    vmovq %xmm5, %rcx
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm6 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; AVX1-NEXT:    vpextrq $1, %xmm6, %r13
+; AVX1-NEXT:    vmovq %xmm6, %r12
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero
+; AVX1-NEXT:    vpextrq $1, %xmm5, %r10
+; AVX1-NEXT:    vmovq %xmm5, %r14
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm5 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; AVX1-NEXT:    vpextrq $1, %xmm5, %r9
+; AVX1-NEXT:    vmovq %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
 ; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
-; AVX1-NEXT:    vpextrq $1, %xmm4, %rbx
-; AVX1-NEXT:    vmovq %xmm4, %rdx
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm4 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
-; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm4 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm7[2],xmm3[2],xmm7[3],xmm3[3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
-; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm6 = xmm0[0],zero,xmm0[1],zero
-; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rsi
-; AVX1-NEXT:    addq %rcx, %rsi
-; AVX1-NEXT:    vmovq %xmm0, %r13
-; AVX1-NEXT:    addq %r14, %r13
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
-; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX1-NEXT:    vpextrq $1, %xmm0, %r12
-; AVX1-NEXT:    addq %rax, %r12
-; AVX1-NEXT:    vmovq %xmm0, %r14
-; AVX1-NEXT:    addq %rbp, %r14
-; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm7[0],zero,xmm7[1],zero
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rbp
-; AVX1-NEXT:    addq %r11, %rbp
-; AVX1-NEXT:    vmovq %xmm0, %r11
-; AVX1-NEXT:    addq %r15, %r11
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
-; AVX1-NEXT:    vpextrq $1, %xmm0, %r15
-; AVX1-NEXT:    addq %rbx, %r15
-; AVX1-NEXT:    vmovq %xmm0, %rbx
-; AVX1-NEXT:    addq %rdx, %rbx
-; AVX1-NEXT:    vpextrq $1, %xmm6, %rax
-; AVX1-NEXT:    leaq -1(%rdi,%rax), %rax
-; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; AVX1-NEXT:    vmovq %xmm6, %rax
-; AVX1-NEXT:    leaq -1(%r8,%rax), %rax
-; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    vpextrq $1, %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX1-NEXT:    vmovq %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
 ; AVX1-NEXT:    vpextrq $1, %xmm5, %rax
-; AVX1-NEXT:    leaq -1(%r9,%rax), %rax
-; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    addq %rdi, %rax
+; AVX1-NEXT:    movq %rax, %rdi
 ; AVX1-NEXT:    vmovq %xmm5, %rax
-; AVX1-NEXT:    leaq -1(%r10,%rax), %rax
-; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    addq %rbp, %rax
+; AVX1-NEXT:    movq %rax, %rbp
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX1-NEXT:    vpextrq $1, %xmm4, %r15
+; AVX1-NEXT:    addq %rbx, %r15
+; AVX1-NEXT:    vmovq %xmm4, %r11
+; AVX1-NEXT:    addq %rsi, %r11
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm4 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
 ; AVX1-NEXT:    vpextrq $1, %xmm4, %rax
-; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
-; AVX1-NEXT:    leaq -1(%rcx,%rax), %rax
+; AVX1-NEXT:    addq %rdx, %rax
+; AVX1-NEXT:    movq %rax, %rdx
+; AVX1-NEXT:    vmovq %xmm4, %r8
+; AVX1-NEXT:    addq %rcx, %r8
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; AVX1-NEXT:    vpextrq $1, %xmm5, %rsi
+; AVX1-NEXT:    addq %r13, %rsi
+; AVX1-NEXT:    vmovq %xmm5, %rax
+; AVX1-NEXT:    addq %r12, %rax
 ; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
+; AVX1-NEXT:    vpextrq $1, %xmm4, %rax
+; AVX1-NEXT:    addq %r10, %rax
+; AVX1-NEXT:    movq %rax, %r10
 ; AVX1-NEXT:    vmovq %xmm4, %rax
-; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
-; AVX1-NEXT:    leaq -1(%rcx,%rax), %rax
+; AVX1-NEXT:    addq %r14, %rax
 ; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; AVX1-NEXT:    vpextrq $1, %xmm8, %rax
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rcx
-; AVX1-NEXT:    leaq -1(%rax,%rcx), %rax
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm4 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; AVX1-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX1-NEXT:    addq %r9, %rax
+; AVX1-NEXT:    movq %rax, %r13
+; AVX1-NEXT:    vmovq %xmm2, %rbx
+; AVX1-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Folded Reload
+; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
+; AVX1-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX1-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
+; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    vmovq %xmm1, %rax
+; AVX1-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
 ; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; AVX1-NEXT:    vmovq %xmm8, %rax
+; AVX1-NEXT:    vpextrq $1, %xmm3, %rax
+; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; AVX1-NEXT:    vpextrq $1, %xmm0, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    vmovq %xmm3, %rax
 ; AVX1-NEXT:    vmovq %xmm0, %rcx
-; AVX1-NEXT:    leaq -1(%rax,%rcx), %rax
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    movq %rcx, %r9
+; AVX1-NEXT:    addq $-1, %rdi
+; AVX1-NEXT:    movq %rdi, (%rsp) # 8-byte Spill
+; AVX1-NEXT:    movl $0, %eax
+; AVX1-NEXT:    adcq $-1, %rax
+; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    addq $-1, %rbp
+; AVX1-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    movl $0, %eax
+; AVX1-NEXT:    adcq $-1, %rax
+; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    addq $-1, %r15
+; AVX1-NEXT:    movq %r15, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    movl $0, %eax
+; AVX1-NEXT:    adcq $-1, %rax
 ; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; AVX1-NEXT:    xorl %r10d, %r10d
+; AVX1-NEXT:    addq $-1, %r11
+; AVX1-NEXT:    movq %r11, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    movl $0, %eax
+; AVX1-NEXT:    adcq $-1, %rax
+; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    addq $-1, %rdx
+; AVX1-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    movl $0, %eax
+; AVX1-NEXT:    adcq $-1, %rax
+; AVX1-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    addq $-1, %r8
+; AVX1-NEXT:    movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    movl $0, %r12d
+; AVX1-NEXT:    adcq $-1, %r12
 ; AVX1-NEXT:    addq $-1, %rsi
 ; AVX1-NEXT:    movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; AVX1-NEXT:    movl $0, %ecx
-; AVX1-NEXT:    adcq $-1, %rcx
-; AVX1-NEXT:    addq $-1, %r13
 ; AVX1-NEXT:    movl $0, %eax
 ; AVX1-NEXT:    adcq $-1, %rax
-; AVX1-NEXT:    addq $-1, %r12
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; AVX1-NEXT:    addq $-1, %rax
+; AVX1-NEXT:    movl $0, %edx
+; AVX1-NEXT:    adcq $-1, %rdx
+; AVX1-NEXT:    movq %rdx, %rsi
+; AVX1-NEXT:    addq $-1, %r10
+; AVX1-NEXT:    movq %r10, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    movl $0, %edx
+; AVX1-NEXT:    adcq $-1, %rdx
+; AVX1-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX1-NEXT:    movl $0, %r15d
+; AVX1-NEXT:    adcq $-1, %r15
+; AVX1-NEXT:    addq $-1, %r13
+; AVX1-NEXT:    movl $0, %r14d
+; AVX1-NEXT:    adcq $-1, %r14
+; AVX1-NEXT:    addq $-1, %rbx
+; AVX1-NEXT:    movl $0, %r11d
+; AVX1-NEXT:    adcq $-1, %r11
+; AVX1-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX1-NEXT:    movl $0, %r8d
+; AVX1-NEXT:    adcq $-1, %r8
+; AVX1-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
 ; AVX1-NEXT:    movl $0, %edi
 ; AVX1-NEXT:    adcq $-1, %rdi
-; AVX1-NEXT:    addq $-1, %r14
-; AVX1-NEXT:    movl $0, %esi
-; AVX1-NEXT:    adcq $-1, %rsi
+; AVX1-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX1-NEXT:    movl $0, %r10d
+; AVX1-NEXT:    adcq $-1, %r10
+; AVX1-NEXT:    movq %r9, %rbp
 ; AVX1-NEXT:    addq $-1, %rbp
 ; AVX1-NEXT:    movl $0, %r9d
 ; AVX1-NEXT:    adcq $-1, %r9
-; AVX1-NEXT:    addq $-1, %r11
-; AVX1-NEXT:    movl $0, %r8d
-; AVX1-NEXT:    adcq $-1, %r8
-; AVX1-NEXT:    addq $-1, %r15
-; AVX1-NEXT:    movl $0, %edx
-; AVX1-NEXT:    adcq $-1, %rdx
-; AVX1-NEXT:    addq $-1, %rbx
-; AVX1-NEXT:    adcq $-1, %r10
-; AVX1-NEXT:    shldq $63, %r11, %r8
-; AVX1-NEXT:    shldq $63, %rbp, %r9
-; AVX1-NEXT:    shldq $63, %r14, %rsi
-; AVX1-NEXT:    shldq $63, %r12, %rdi
-; AVX1-NEXT:    shldq $63, %r13, %rax
-; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Reload
-; AVX1-NEXT:    shldq $63, %rbp, %rcx
-; AVX1-NEXT:    shldq $63, %rbx, %r10
-; AVX1-NEXT:    shldq $63, %r15, %rdx
-; AVX1-NEXT:    vmovq %rcx, %xmm8
-; AVX1-NEXT:    vmovq %rax, %xmm9
-; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX1-NEXT:    shrq %rax
-; AVX1-NEXT:    vmovq %rax, %xmm0
-; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX1-NEXT:    shrq %rax
-; AVX1-NEXT:    vmovq %rax, %xmm11
-; AVX1-NEXT:    vmovq %rdi, %xmm12
-; AVX1-NEXT:    vmovq %rsi, %xmm13
-; AVX1-NEXT:    vmovq %rdx, %xmm14
-; AVX1-NEXT:    vmovq %r10, %xmm15
-; AVX1-NEXT:    vmovq %r9, %xmm10
-; AVX1-NEXT:    vmovq %r8, %xmm1
+; AVX1-NEXT:    shldq $63, %rbx, %r11
+; AVX1-NEXT:    shldq $63, %r13, %r14
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Reload
+; AVX1-NEXT:    shldq $63, %rbx, %r15
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Reload
+; AVX1-NEXT:    shldq $63, %rbx, %rdx
+; AVX1-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX1-NEXT:    shldq $63, %rax, %rsi
+; AVX1-NEXT:    movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX1-NEXT:    shrq %rax
-; AVX1-NEXT:    vmovq %rax, %xmm2
+; AVX1-NEXT:    shldq $63, %rax, %rcx
+; AVX1-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX1-NEXT:    shrq %rax
-; AVX1-NEXT:    vmovq %rax, %xmm3
+; AVX1-NEXT:    shldq $63, %rax, %r12
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
 ; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX1-NEXT:    shrq %rax
-; AVX1-NEXT:    vmovq %rax, %xmm4
+; AVX1-NEXT:    shldq $63, %rax, %rsi
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
 ; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX1-NEXT:    shrq %rax
-; AVX1-NEXT:    vmovq %rax, %xmm5
+; AVX1-NEXT:    shldq $63, %rax, %rdx
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
 ; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX1-NEXT:    shrq %rax
-; AVX1-NEXT:    vmovq %rax, %xmm6
+; AVX1-NEXT:    shldq $63, %rax, %rcx
 ; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX1-NEXT:    shrq %rax
-; AVX1-NEXT:    vmovq %rax, %xmm7
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Reload
+; AVX1-NEXT:    shldq $63, %rbx, %rax
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Reload
+; AVX1-NEXT:    movq (%rsp), %r13 # 8-byte Reload
+; AVX1-NEXT:    shldq $63, %r13, %rbx
+; AVX1-NEXT:    shldq $63, %rbp, %r9
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Reload
+; AVX1-NEXT:    shldq $63, %rbp, %r10
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Reload
+; AVX1-NEXT:    shldq $63, %rbp, %rdi
+; AVX1-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Reload
+; AVX1-NEXT:    shldq $63, %rbp, %r8
+; AVX1-NEXT:    vmovq %rbx, %xmm8
+; AVX1-NEXT:    vmovq %rax, %xmm9
+; AVX1-NEXT:    vmovq %rcx, %xmm0
+; AVX1-NEXT:    vmovq %rdx, %xmm1
+; AVX1-NEXT:    vmovq %rsi, %xmm12
+; AVX1-NEXT:    vmovq %r12, %xmm13
+; AVX1-NEXT:    vmovq %r8, %xmm14
+; AVX1-NEXT:    vmovq %rdi, %xmm15
+; AVX1-NEXT:    vmovq {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 8-byte Folded Reload
+; AVX1-NEXT:    # xmm10 = mem[0],zero
+; AVX1-NEXT:    vmovq {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 8-byte Folded Reload
+; AVX1-NEXT:    # xmm11 = mem[0],zero
+; AVX1-NEXT:    vmovq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 8-byte Folded Reload
+; AVX1-NEXT:    # xmm2 = mem[0],zero
+; AVX1-NEXT:    vmovq %r15, %xmm3
+; AVX1-NEXT:    vmovq %r14, %xmm4
+; AVX1-NEXT:    vmovq %r11, %xmm5
+; AVX1-NEXT:    vmovq %r10, %xmm6
+; AVX1-NEXT:    vmovq %r9, %xmm7
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3],xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm9 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3],xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7]
-; AVX1-NEXT:    vpsllq $48, %xmm8, %xmm8
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm9[0,0,1,1]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm8 = xmm0[0,1,2],xmm8[3],xmm0[4,5,6,7]
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm9 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; AVX1-NEXT:    vpslldq {{.*#+}} xmm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm8[0,1]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm9[0,1,2,0]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm8 = xmm0[0,1,2,3,4,5,6],xmm8[7]
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3],xmm13[4],xmm12[4],xmm13[5],xmm12[5],xmm13[6],xmm12[6],xmm13[7],xmm12[7]
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm9 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3],xmm15[4],xmm14[4],xmm15[5],xmm14[5],xmm15[6],xmm14[6],xmm15[7],xmm14[7]
-; AVX1-NEXT:    vpslld $16, %xmm0, %xmm0
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm9[0],xmm0[1],xmm9[2,3,4,5,6,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,3],xmm0[4,5,6,7]
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3],xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm9[0,1,0,1]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5],xmm1[6,7]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm8[6,7]
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3],xmm11[4],xmm10[4],xmm11[5],xmm10[5],xmm11[6],xmm10[6],xmm11[7],xmm10[7]
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; AVX1-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],xmm2[7]
+; AVX1-NEXT:    vpsllq $48, %xmm1, %xmm1
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3],xmm2[4,5,6,7]
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3],xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
-; AVX1-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5]
+; AVX1-NEXT:    vpslld $16, %xmm2, %xmm2
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm3 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3],xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,1]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3,4],xmm2[5],xmm3[6,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3,4,5,6,7]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5,6,7]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
 ; AVX1-NEXT:    vmovdqu %xmm0, (%rax)
+; AVX1-NEXT:    addq $8, %rsp
 ; AVX1-NEXT:    popq %rbx
 ; AVX1-NEXT:    popq %r12
 ; AVX1-NEXT:    popq %r13
 ; AVX1-NEXT:    popq %r14
 ; AVX1-NEXT:    popq %r15
 ; AVX1-NEXT:    popq %rbp
 ; AVX1-NEXT:    retq
 ;

 ; AVX512F-LABEL: not_avg_v16i8_wide_constants:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    pushq %rbp
 ; AVX512F-NEXT:    pushq %r15
 ; AVX512F-NEXT:    pushq %r14
 ; AVX512F-NEXT:    pushq %r13
 ; AVX512F-NEXT:    pushq %r12
 ; AVX512F-NEXT:    pushq %rbx
+; AVX512F-NEXT:    subq $16, %rsp
 ; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm1 = 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:    vpmovzxbw {{.*#+}} ymm3 = 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:    vpmovzxbw {{.*#+}} ymm0 = 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 {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm3 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX512F-NEXT:    vpextrq $1, %xmm3, %rbx
+; AVX512F-NEXT:    vmovq %xmm3, %rbp
+; AVX512F-NEXT:    vextracti128 $1, %ymm3, %xmm3
+; AVX512F-NEXT:    vmovq %xmm3, %rdi
+; AVX512F-NEXT:    vpextrq $1, %xmm3, %rsi
+; AVX512F-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX512F-NEXT:    vmovq %xmm2, %rcx
+; AVX512F-NEXT:    vpextrq $1, %xmm2, %r9
+; AVX512F-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX512F-NEXT:    vmovq %xmm2, %r13
+; AVX512F-NEXT:    vpextrq $1, %xmm2, %r11
 ; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm4
-; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm4, %xmm5
-; AVX512F-NEXT:    vpextrq $1, %xmm5, %rdx
-; AVX512F-NEXT:    vmovq %xmm5, %rcx
-; AVX512F-NEXT:    vpextrq $1, %xmm4, %rax
-; AVX512F-NEXT:    vmovq %xmm4, %rbx
-; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm4
-; AVX512F-NEXT:    vpextrq $1, %xmm4, %rdi
-; AVX512F-NEXT:    vmovq %xmm4, %rsi
-; AVX512F-NEXT:    vpextrq $1, %xmm1, %r13
-; AVX512F-NEXT:    vmovq %xmm1, %r15
-; AVX512F-NEXT:    vextracti128 $1, %ymm2, %xmm1
-; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX512F-NEXT:    vpextrq $1, %xmm2, %r12
+; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
 ; AVX512F-NEXT:    vmovq %xmm2, %r14
-; AVX512F-NEXT:    vpextrq $1, %xmm1, %r11
+; AVX512F-NEXT:    vpextrq $1, %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX512F-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX512F-NEXT:    vmovq %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX512F-NEXT:    vpextrq $1, %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
 ; AVX512F-NEXT:    vmovq %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
-; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512F-NEXT:    vpextrq $1, %xmm1, %r10
-; AVX512F-NEXT:    vmovq %xmm1, %r9
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} ymm2 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
-; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm1 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX512F-NEXT:    vpextrq $1, %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} ymm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm3 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX512F-NEXT:    vpextrq $1, %xmm3, %rax
+; AVX512F-NEXT:    addq %rbx, %rax
+; AVX512F-NEXT:    movq %rax, %rbx
+; AVX512F-NEXT:    vmovq %xmm3, %r8
+; AVX512F-NEXT:    addq %rbp, %r8
 ; AVX512F-NEXT:    vextracti128 $1, %ymm3, %xmm3
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm3, %xmm4
-; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm4, %xmm5
-; AVX512F-NEXT:    vpextrq $1, %xmm5, %rbp
-; AVX512F-NEXT:    leal -1(%rdx,%rbp), %edx
-; AVX512F-NEXT:    movl %edx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
-; AVX512F-NEXT:    vmovq %xmm5, %rbp
-; AVX512F-NEXT:    leal -1(%rcx,%rbp), %ecx
-; AVX512F-NEXT:    movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
-; AVX512F-NEXT:    vpextrq $1, %xmm4, %rbp
-; AVX512F-NEXT:    leal -1(%rax,%rbp), %eax
-; AVX512F-NEXT:    movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
-; AVX512F-NEXT:    vmovq %xmm4, %rbp
-; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm3, %xmm4
-; AVX512F-NEXT:    leal -1(%rbx,%rbp), %r8d
-; AVX512F-NEXT:    vpextrq $1, %xmm4, %rbp
-; AVX512F-NEXT:    leal -1(%rdi,%rbp), %edi
-; AVX512F-NEXT:    vmovq %xmm4, %rbp
-; AVX512F-NEXT:    leal -1(%rsi,%rbp), %esi
-; AVX512F-NEXT:    vpextrq $1, %xmm3, %rbp
-; AVX512F-NEXT:    leal -1(%r13,%rbp), %r13d
-; AVX512F-NEXT:    vmovq %xmm3, %rbp
+; AVX512F-NEXT:    vmovq %xmm3, %rax
+; AVX512F-NEXT:    addq %rdi, %rax
+; AVX512F-NEXT:    movq %rax, %rdi
+; AVX512F-NEXT:    vpextrq $1, %xmm3, %r12
+; AVX512F-NEXT:    addq %rsi, %r12
 ; AVX512F-NEXT:    vextracti128 $1, %ymm2, %xmm2
 ; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm2, %xmm3
-; AVX512F-NEXT:    leal -1(%r15,%rbp), %r15d
-; AVX512F-NEXT:    vpextrq $1, %xmm3, %rbp
-; AVX512F-NEXT:    leal -1(%r12,%rbp), %r12d
-; AVX512F-NEXT:    vmovq %xmm3, %rbp
-; AVX512F-NEXT:    leal -1(%r14,%rbp), %r14d
+; AVX512F-NEXT:    vmovq %xmm2, %rax
+; AVX512F-NEXT:    addq %rcx, %rax
+; AVX512F-NEXT:    movq %rax, %rcx
 ; AVX512F-NEXT:    vpextrq $1, %xmm2, %rdx
-; AVX512F-NEXT:    leal -1(%r11,%rdx), %r11d
-; AVX512F-NEXT:    vmovq %xmm2, %rbp
-; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
-; AVX512F-NEXT:    leal -1(%rax,%rbp), %ebp
-; AVX512F-NEXT:    vpextrq $1, %xmm2, %rcx
-; AVX512F-NEXT:    leal -1(%r10,%rcx), %ecx
+; AVX512F-NEXT:    addq %r9, %rdx
+; AVX512F-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX512F-NEXT:    vmovq %xmm2, %r10
+; AVX512F-NEXT:    addq %r13, %r10
+; AVX512F-NEXT:    vpextrq $1, %xmm2, %r15
+; AVX512F-NEXT:    addq %r11, %r15
+; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; AVX512F-NEXT:    vmovq %xmm2, %rax
+; AVX512F-NEXT:    addq %r14, %rax
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX512F-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    vextracti128 $1, %ymm2, %xmm2
 ; AVX512F-NEXT:    vmovq %xmm2, %rax
-; AVX512F-NEXT:    leal -1(%r9,%rax), %eax
-; AVX512F-NEXT:    vpextrq $1, %xmm0, %rdx
-; AVX512F-NEXT:    vpextrq $1, %xmm1, %r10
-; AVX512F-NEXT:    leal -1(%rdx,%r10), %edx
-; AVX512F-NEXT:    vmovq %xmm0, %r10
-; AVX512F-NEXT:    vmovq %xmm1, %r9
-; AVX512F-NEXT:    leaq -1(%r10,%r9), %rbx
-; AVX512F-NEXT:    shrq %rbx
-; AVX512F-NEXT:    vmovd %ebx, %xmm0
-; AVX512F-NEXT:    shrl %edx
-; AVX512F-NEXT:    vpinsrb $1, %edx, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %eax
-; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %ecx
-; AVX512F-NEXT:    vpinsrb $3, %ecx, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %ebp
-; AVX512F-NEXT:    vpinsrb $4, %ebp, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %r11d
-; AVX512F-NEXT:    vpinsrb $5, %r11d, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %r14d
-; AVX512F-NEXT:    vpinsrb $6, %r14d, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %r12d
+; AVX512F-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX512F-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
+; AVX512F-NEXT:    movq %rax, %r9
+; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX512F-NEXT:    vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; AVX512F-NEXT:    vmovq %xmm0, %rbp
+; AVX512F-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Folded Reload
+; AVX512F-NEXT:    vpextrq $1, %xmm0, %r13
+; AVX512F-NEXT:    addq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Folded Reload
+; AVX512F-NEXT:    vmovq %xmm1, %rax
+; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX512F-NEXT:    vmovq %xmm0, %r11
+; AVX512F-NEXT:    addq %rax, %r11
+; AVX512F-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX512F-NEXT:    vpextrq $1, %xmm0, %rsi
+; AVX512F-NEXT:    addq %rax, %rsi
+; AVX512F-NEXT:    addq $-1, %rbx
+; AVX512F-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    movl $0, %eax
+; AVX512F-NEXT:    adcq $-1, %rax
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    addq $-1, %r8
+; AVX512F-NEXT:    movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    movl $0, %eax
+; AVX512F-NEXT:    adcq $-1, %rax
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    addq $-1, %rdi
+; AVX512F-NEXT:    movq %rdi, (%rsp) # 8-byte Spill
+; AVX512F-NEXT:    movl $0, %eax
+; AVX512F-NEXT:    adcq $-1, %rax
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    addq $-1, %r12
+; AVX512F-NEXT:    movq %r12, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    movl $0, %eax
+; AVX512F-NEXT:    adcq $-1, %rax
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    addq $-1, %rcx
+; AVX512F-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    movl $0, %eax
+; AVX512F-NEXT:    adcq $-1, %rax
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    addq $-1, %rdx
+; AVX512F-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    movl $0, %eax
+; AVX512F-NEXT:    adcq $-1, %rax
+; AVX512F-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    addq $-1, %r10
+; AVX512F-NEXT:    movq %r10, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    movl $0, %ecx
+; AVX512F-NEXT:    adcq $-1, %rcx
+; AVX512F-NEXT:    addq $-1, %r15
+; AVX512F-NEXT:    movq %r15, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    movl $0, %r12d
+; AVX512F-NEXT:    adcq $-1, %r12
+; AVX512F-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX512F-NEXT:    movl $0, %r15d
+; AVX512F-NEXT:    adcq $-1, %r15
+; AVX512F-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX512F-NEXT:    movl $0, %r14d
+; AVX512F-NEXT:    adcq $-1, %r14
+; AVX512F-NEXT:    addq $-1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
+; AVX512F-NEXT:    movl $0, %ebx
+; AVX512F-NEXT:    adcq $-1, %rbx
+; AVX512F-NEXT:    movq %r9, %rax
+; AVX512F-NEXT:    addq $-1, %rax
+; AVX512F-NEXT:    movl $0, %r10d
+; AVX512F-NEXT:    adcq $-1, %r10
+; AVX512F-NEXT:    addq $-1, %rbp
+; AVX512F-NEXT:    movl $0, %r9d
+; AVX512F-NEXT:    adcq $-1, %r9
+; AVX512F-NEXT:    addq $-1, %r13
+; AVX512F-NEXT:    movl $0, %r8d
+; AVX512F-NEXT:    adcq $-1, %r8
+; AVX512F-NEXT:    addq $-1, %r11
+; AVX512F-NEXT:    movl $0, %edi
+; AVX512F-NEXT:    adcq $-1, %rdi
+; AVX512F-NEXT:    addq $-1, %rsi
+; AVX512F-NEXT:    movl $0, %edx
+; AVX512F-NEXT:    adcq $-1, %rdx
+; AVX512F-NEXT:    shldq $63, %rsi, %rdx
+; AVX512F-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    shldq $63, %r11, %rdi
+; AVX512F-NEXT:    shldq $63, %r13, %r8
+; AVX512F-NEXT:    shldq $63, %rbp, %r9
+; AVX512F-NEXT:    shldq $63, %rax, %r10
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rsi, %rbx
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rsi, %r14
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rsi, %r15
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rax, %r12
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rax, %rcx
+; AVX512F-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rax, %rdx
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rax, %rcx
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Reload
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rax, %r13
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r11 # 8-byte Reload
+; AVX512F-NEXT:    movq (%rsp), %rax # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rax, %r11
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rsi, %rax
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
+; AVX512F-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Reload
+; AVX512F-NEXT:    shldq $63, %rbp, %rsi
+; AVX512F-NEXT:    vmovd %eax, %xmm0
+; AVX512F-NEXT:    vpinsrb $1, %esi, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $2, %r11d, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $3, %r13d, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $4, %ecx, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $5, %edx, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $6, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload
 ; AVX512F-NEXT:    vpinsrb $7, %r12d, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %r15d
 ; AVX512F-NEXT:    vpinsrb $8, %r15d, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %r13d
-; AVX512F-NEXT:    vpinsrb $9, %r13d, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %esi
-; AVX512F-NEXT:    vpinsrb $10, %esi, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %edi
-; AVX512F-NEXT:    vpinsrb $11, %edi, %xmm0, %xmm0
-; AVX512F-NEXT:    shrl %r8d
-; AVX512F-NEXT:    vpinsrb $12, %r8d, %xmm0, %xmm0
-; AVX512F-NEXT:    movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
-; AVX512F-NEXT:    shrl %eax
-; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
-; AVX512F-NEXT:    shrl %eax
-; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
-; AVX512F-NEXT:    shrl %eax
-; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $9, %r14d, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $10, %ebx, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $11, %r10d, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $12, %r9d, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $13, %r8d, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0
+; AVX512F-NEXT:    vpinsrb $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload
 ; AVX512F-NEXT:    vmovdqu %xmm0, (%rax)
+; AVX512F-NEXT:    addq $16, %rsp
 ; AVX512F-NEXT:    popq %rbx
 ; AVX512F-NEXT:    popq %r12
 ; AVX512F-NEXT:    popq %r13
 ; AVX512F-NEXT:    popq %r14
 ; AVX512F-NEXT:    popq %r15
 ; AVX512F-NEXT:    popq %rbp
 ; AVX512F-NEXT:    vzeroupper
 ; AVX512F-NEXT:    retq

llvm/test/CodeGen/X86/avx512vbmi2-funnel-shifts.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vbmi2 | FileCheck %s --check-prefixes=CHECK,X86
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi2 -| FileCheck %s --check-prefixes=CHECK,X64
 
 define <8 x i64> @avx512_funnel_shift_q_512(<8 x i64> %a0, <8 x i64> %a1) {
 ; X86-LABEL: avx512_funnel_shift_q_512:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvq {{\.LCPI.*}}, %zmm0, %zmm0
-; X86-NEXT:    vpsrlvq {{\.LCPI.*}}, %zmm1, %zmm1
-; X86-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; X86-NEXT:    vpshldvq {{\.LCPI.*}}, %zmm1, %zmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_q_512:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvq {{.*}}(%rip), %zmm0, %zmm0
-; X64-NEXT:    vpsrlvq {{.*}}(%rip), %zmm1, %zmm1
-; X64-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; X64-NEXT:    vpshldvq {{.*}}(%rip), %zmm1, %zmm0
 ; X64-NEXT:    retq
   %1 =  shl <8 x i64> %a0, <i64 31, i64 33, i64 31, i64 33, i64 31, i64 33, i64 31, i64 33>
   %2 = lshr <8 x i64> %a1, <i64 33, i64 31, i64 33, i64 31, i64 33, i64 31, i64 33, i64 31>
   %3 = or <8 x i64> %1, %2
   ret <8 x i64> %3
 }
 
 define <8 x i64> @avx512_funnel_shift_q_512_splat(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_q_512_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpsllq $31, %zmm0, %zmm0
-; CHECK-NEXT:    vpsrlq $33, %zmm1, %zmm1
-; CHECK-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    vpshldq $31, %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <8 x i64> %a0, <i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31>
   %2 = lshr <8 x i64> %a1, <i64 33, i64 33, i64 33, i64 33, i64 33, i64 33, i64 33, i64 33>
   %3 = or <8 x i64> %1, %2
   ret <8 x i64> %3
 }
 
 define <16 x i32> @avx512_funnel_shift_d_512(<16 x i32> %a0, <16 x i32> %a1) {
 ; X86-LABEL: avx512_funnel_shift_d_512:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvd {{\.LCPI.*}}, %zmm0, %zmm0
-; X86-NEXT:    vpsrlvd {{\.LCPI.*}}, %zmm1, %zmm1
-; X86-NEXT:    vpord %zmm1, %zmm0, %zmm0
+; X86-NEXT:    vpshldvd {{\.LCPI.*}}, %zmm1, %zmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_d_512:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvd {{.*}}(%rip), %zmm0, %zmm0
-; X64-NEXT:    vpsrlvd {{.*}}(%rip), %zmm1, %zmm1
-; X64-NEXT:    vpord %zmm1, %zmm0, %zmm0
+; X64-NEXT:    vpshldvd {{.*}}(%rip), %zmm1, %zmm0
 ; X64-NEXT:    retq
   %1 =  shl <16 x i32> %a0, <i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17>
   %2 = lshr <16 x i32> %a1, <i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15>
   %3 = or <16 x i32> %1, %2
   ret <16 x i32> %3
 }
 
 define <16 x i32> @avx512_funnel_shift_d_512_splat(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_d_512_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpslld $15, %zmm0, %zmm0
-; CHECK-NEXT:    vpsrld $17, %zmm1, %zmm1
-; CHECK-NEXT:    vpord %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    vpshldd $15, %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <16 x i32> %a0, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
   %2 = lshr <16 x i32> %a1, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
   %3 = or <16 x i32> %1, %2
   ret <16 x i32> %3
 }
 
 define <32 x i16> @avx512_funnel_shift_w_512(<32 x i16> %a0, <32 x i16> %a1) {
 ; X86-LABEL: avx512_funnel_shift_w_512:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvw {{\.LCPI.*}}, %zmm0, %zmm0
-; X86-NEXT:    vpsrlvw {{\.LCPI.*}}, %zmm1, %zmm1
-; X86-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; X86-NEXT:    vpshldvw {{\.LCPI.*}}, %zmm1, %zmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_w_512:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvw {{.*}}(%rip), %zmm0, %zmm0
-; X64-NEXT:    vpsrlvw {{.*}}(%rip), %zmm1, %zmm1
-; X64-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; X64-NEXT:    vpshldvw {{.*}}(%rip), %zmm1, %zmm0
 ; X64-NEXT:    retq
   %1 =  shl <32 x i16> %a0, <i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9>
   %2 = lshr <32 x i16> %a1, <i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7>
   %3 = or <32 x i16> %1, %2
   ret <32 x i16> %3
 }
 
 define <32 x i16> @avx512_funnel_shift_w_512_splat(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_w_512_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpsllw $7, %zmm0, %zmm0
-; CHECK-NEXT:    vpsrlw $9, %zmm1, %zmm1
-; CHECK-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    vpshldw $7, %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <32 x i16> %a0, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
   %2 = lshr <32 x i16> %a1, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
   %3 = or <32 x i16> %1, %2
   ret <32 x i16> %3
 }
 
 

llvm/test/CodeGen/X86/avx512vbmi2vl-funnel-shifts.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vbmi2,+avx512vl | FileCheck %s --check-prefixes=CHECK,X86
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi2,+avx512vl -| FileCheck %s --check-prefixes=CHECK,X64
 
 define <2 x i64> @avx512_funnel_shift_q_128(<2 x i64> %a0, <2 x i64> %a1) {
 ; X86-LABEL: avx512_funnel_shift_q_128:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvq {{\.LCPI.*}}, %xmm0, %xmm0
-; X86-NEXT:    vpsrlvq {{\.LCPI.*}}, %xmm1, %xmm1
-; X86-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; X86-NEXT:    vpshldvq {{\.LCPI.*}}, %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_q_128:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvq {{.*}}(%rip), %xmm0, %xmm0
-; X64-NEXT:    vpsrlvq {{.*}}(%rip), %xmm1, %xmm1
-; X64-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; X64-NEXT:    vpshldvq {{.*}}(%rip), %xmm1, %xmm0
 ; X64-NEXT:    retq
   %1 =  shl <2 x i64> %a0, <i64 31, i64 33>
   %2 = lshr <2 x i64> %a1, <i64 33, i64 31>
   %3 = or <2 x i64> %1, %2
   ret <2 x i64> %3
 }
 
 define <4 x i64> @avx512_funnel_shift_q_256(<4 x i64> %a0, <4 x i64> %a1) {
 ; X86-LABEL: avx512_funnel_shift_q_256:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvq {{\.LCPI.*}}, %ymm0, %ymm0
-; X86-NEXT:    vpsrlvq {{\.LCPI.*}}, %ymm1, %ymm1
-; X86-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; X86-NEXT:    vpshldvq {{\.LCPI.*}}, %ymm1, %ymm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_q_256:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvq {{.*}}(%rip), %ymm0, %ymm0
-; X64-NEXT:    vpsrlvq {{.*}}(%rip), %ymm1, %ymm1
-; X64-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; X64-NEXT:    vpshldvq {{.*}}(%rip), %ymm1, %ymm0
 ; X64-NEXT:    retq
   %1 =  shl <4 x i64> %a0, <i64 31, i64 33, i64 31, i64 33>
   %2 = lshr <4 x i64> %a1, <i64 33, i64 31, i64 33, i64 31>
   %3 = or <4 x i64> %1, %2
   ret <4 x i64> %3
 }
 
 define <2 x i64> @avx512_funnel_shift_q_128_splat(<2 x i64> %a0, <2 x i64> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_q_128_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpsllq $31, %xmm0, %xmm0
-; CHECK-NEXT:    vpsrlq $33, %xmm1, %xmm1
-; CHECK-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vpshldq $31, %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <2 x i64> %a0, <i64 31, i64 31>
   %2 = lshr <2 x i64> %a1, <i64 33, i64 33>
   %3 = or <2 x i64> %1, %2
   ret <2 x i64> %3
 }
 
 define <4 x i64> @avx512_funnel_shift_q_256_splat(<4 x i64> %a0, <4 x i64> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_q_256_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpsllq $31, %ymm0, %ymm0
-; CHECK-NEXT:    vpsrlq $33, %ymm1, %ymm1
-; CHECK-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    vpshldq $31, %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <4 x i64> %a0, <i64 31, i64 31, i64 31, i64 31>
   %2 = lshr <4 x i64> %a1, <i64 33, i64 33, i64 33, i64 33>
   %3 = or <4 x i64> %1, %2
   ret <4 x i64> %3
 }
 
 define <4 x i32> @avx512_funnel_shift_d_128(<4 x i32> %a0, <4 x i32> %a1) {
 ; X86-LABEL: avx512_funnel_shift_d_128:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvd {{\.LCPI.*}}, %xmm0, %xmm0
-; X86-NEXT:    vpsrlvd {{\.LCPI.*}}, %xmm1, %xmm1
-; X86-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; X86-NEXT:    vpshldvd {{\.LCPI.*}}, %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_d_128:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvd {{.*}}(%rip), %xmm0, %xmm0
-; X64-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; X64-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; X64-NEXT:    vpshldvd {{.*}}(%rip), %xmm1, %xmm0
 ; X64-NEXT:    retq
   %1 =  shl <4 x i32> %a0, <i32 15, i32 17, i32 15, i32 17>
   %2 = lshr <4 x i32> %a1, <i32 17, i32 15, i32 17, i32 15>
   %3 = or <4 x i32> %1, %2
   ret <4 x i32> %3
 }
 
 define <8 x i32> @avx512_funnel_shift_d_256(<8 x i32> %a0, <8 x i32> %a1) {
 ; X86-LABEL: avx512_funnel_shift_d_256:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvd {{\.LCPI.*}}, %ymm0, %ymm0
-; X86-NEXT:    vpsrlvd {{\.LCPI.*}}, %ymm1, %ymm1
-; X86-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; X86-NEXT:    vpshldvd {{\.LCPI.*}}, %ymm1, %ymm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_d_256:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvd {{.*}}(%rip), %ymm0, %ymm0
-; X64-NEXT:    vpsrlvd {{.*}}(%rip), %ymm1, %ymm1
-; X64-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; X64-NEXT:    vpshldvd {{.*}}(%rip), %ymm1, %ymm0
 ; X64-NEXT:    retq
   %1 =  shl <8 x i32> %a0, <i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17>
   %2 = lshr <8 x i32> %a1, <i32 17, i32 15, i32 17, i32 15, i32 17, i32 15, i32 17, i32 15>
   %3 = or <8 x i32> %1, %2
   ret <8 x i32> %3
 }
 
 define <4 x i32> @avx512_funnel_shift_d_128_splat(<4 x i32> %a0, <4 x i32> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_d_128_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpslld $15, %xmm0, %xmm0
-; CHECK-NEXT:    vpsrld $17, %xmm1, %xmm1
-; CHECK-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vpshldd $15, %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <4 x i32> %a0, <i32 15, i32 15, i32 15, i32 15>
   %2 = lshr <4 x i32> %a1, <i32 17, i32 17, i32 17, i32 17>
   %3 = or <4 x i32> %1, %2
   ret <4 x i32> %3
 }
 
 define <8 x i32> @avx512_funnel_shift_d_256_splat(<8 x i32> %a0, <8 x i32> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_d_256_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpslld $15, %ymm0, %ymm0
-; CHECK-NEXT:    vpsrld $17, %ymm1, %ymm1
-; CHECK-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    vpshldd $15, %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <8 x i32> %a0, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
   %2 = lshr <8 x i32> %a1, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
   %3 = or <8 x i32> %1, %2
   ret <8 x i32> %3
 }
 
 define <8 x i16> @avx512_funnel_shift_w_128(<8 x i16> %a0, <8 x i16> %a1) {
 ; X86-LABEL: avx512_funnel_shift_w_128:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvw {{\.LCPI.*}}, %xmm0, %xmm0
-; X86-NEXT:    vpsrlvw {{\.LCPI.*}}, %xmm1, %xmm1
-; X86-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; X86-NEXT:    vpshldvw {{\.LCPI.*}}, %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_w_128:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvw {{.*}}(%rip), %xmm0, %xmm0
-; X64-NEXT:    vpsrlvw {{.*}}(%rip), %xmm1, %xmm1
-; X64-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; X64-NEXT:    vpshldvw {{.*}}(%rip), %xmm1, %xmm0
 ; X64-NEXT:    retq
   %1 =  shl <8 x i16> %a0, <i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9>
   %2 = lshr <8 x i16> %a1, <i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7>
   %3 = or <8 x i16> %1, %2
   ret <8 x i16> %3
 }
 
 define <16 x i16> @avx512_funnel_shift_w_256(<16 x i16> %a0, <16 x i16> %a1) {
 ; X86-LABEL: avx512_funnel_shift_w_256:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpsllvw {{\.LCPI.*}}, %ymm0, %ymm0
-; X86-NEXT:    vpsrlvw {{\.LCPI.*}}, %ymm1, %ymm1
-; X86-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; X86-NEXT:    vpshldvw {{\.LCPI.*}}, %ymm1, %ymm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: avx512_funnel_shift_w_256:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpsllvw {{.*}}(%rip), %ymm0, %ymm0
-; X64-NEXT:    vpsrlvw {{.*}}(%rip), %ymm1, %ymm1
-; X64-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; X64-NEXT:    vpshldvw {{.*}}(%rip), %ymm1, %ymm0
 ; X64-NEXT:    retq
   %1 =  shl <16 x i16> %a0, <i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9>
   %2 = lshr <16 x i16> %a1, <i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7, i16 9, i16 7>
   %3 = or <16 x i16> %1, %2
   ret <16 x i16> %3
 }
 
 define <8 x i16> @avx512_funnel_shift_w_128_splat(<8 x i16> %a0, <8 x i16> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_w_128_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpsllw $7, %xmm0, %xmm0
-; CHECK-NEXT:    vpsrlw $9, %xmm1, %xmm1
-; CHECK-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vpshldw $7, %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <8 x i16> %a0, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
   %2 = lshr <8 x i16> %a1, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
   %3 = or <8 x i16> %1, %2
   ret <8 x i16> %3
 }
 
 define <16 x i16> @avx512_funnel_shift_w_256_splat(<16 x i16> %a0, <16 x i16> %a1) {
 ; CHECK-LABEL: avx512_funnel_shift_w_256_splat:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpsllw $7, %ymm0, %ymm0
-; CHECK-NEXT:    vpsrlw $9, %ymm1, %ymm1
-; CHECK-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    vpshldw $7, %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 =  shl <16 x i16> %a0, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
   %2 = lshr <16 x i16> %a1, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
   %3 = or <16 x i16> %1, %2
   ret <16 x i16> %3
 }
 
 

llvm/test/CodeGen/X86/bitreverse.ll

 ; X86-NEXT:    shrl $4, %ebx
 ; X86-NEXT:    orl %ebp, %ebx
 ; X86-NEXT:    movl %ebx, %ebp
 ; X86-NEXT:    andl $858993459, %ebp # imm = 0x33333333
 ; X86-NEXT:    andl $-858993460, %ebx # imm = 0xCCCCCCCC
 ; X86-NEXT:    shrl $2, %ebx
 ; X86-NEXT:    leal (%ebx,%ebp,4), %ebx
 ; X86-NEXT:    movl %ebx, %ebp
-; X86-NEXT:    andl $1431633920, %ebp # imm = 0x55550000
-; X86-NEXT:    andl $-1431699456, %ebx # imm = 0xAAAA0000
+; X86-NEXT:    andl $1431655765, %ebp # imm = 0x55555555
+; X86-NEXT:    andl $-1431655766, %ebx # imm = 0xAAAAAAAA
 ; X86-NEXT:    shrl %ebx
 ; X86-NEXT:    leal (%ebx,%ebp,2), %ebx
 ; X86-NEXT:    movl %ebx, (%esp) # 4-byte Spill
 ; X86-NEXT:    bswapl %edi
 ; X86-NEXT:    movl %edi, %ebx
 ; X86-NEXT:    andl $252645135, %ebx # imm = 0xF0F0F0F
 ; X86-NEXT:    shll $4, %ebx
 ; X86-NEXT:    andl $-252645136, %edi # imm = 0xF0F0F0F0

 ; X64-LABEL: large_promotion:
 ; X64:       # %bb.0:
 ; X64-NEXT:    pushq %rbp
 ; X64-NEXT:    pushq %r15
 ; X64-NEXT:    pushq %r14
 ; X64-NEXT:    pushq %r13
 ; X64-NEXT:    pushq %r12
 ; X64-NEXT:    pushq %rbx
-; X64-NEXT:    movq %rdi, %r12
+; X64-NEXT:    movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
 ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rbx
 ; X64-NEXT:    bswapq %rbx
-; X64-NEXT:    movabsq $1085102592571150095, %r13 # imm = 0xF0F0F0F0F0F0F0F
+; X64-NEXT:    movabsq $1085102592571150095, %rdi # imm = 0xF0F0F0F0F0F0F0F
 ; X64-NEXT:    movq %rbx, %r10
-; X64-NEXT:    andq %r13, %r10
+; X64-NEXT:    andq %rdi, %r10
 ; X64-NEXT:    shlq $4, %r10
 ; X64-NEXT:    movabsq $-1085102592571150096, %rax # imm = 0xF0F0F0F0F0F0F0F0
 ; X64-NEXT:    andq %rax, %rbx
 ; X64-NEXT:    shrq $4, %rbx
 ; X64-NEXT:    orq %r10, %rbx
-; X64-NEXT:    movabsq $3689348814741910323, %r11 # imm = 0x3333333333333333
+; X64-NEXT:    movabsq $3689348814741910323, %r14 # imm = 0x3333333333333333
 ; X64-NEXT:    movq %rbx, %r10
-; X64-NEXT:    andq %r11, %r10
-; X64-NEXT:    movabsq $-3689348814741910324, %r14 # imm = 0xCCCCCCCCCCCCCCCC
-; X64-NEXT:    andq %r14, %rbx
+; X64-NEXT:    andq %r14, %r10
+; X64-NEXT:    movabsq $-3689348814741910324, %r11 # imm = 0xCCCCCCCCCCCCCCCC
+; X64-NEXT:    andq %r11, %rbx
+; X64-NEXT:    movq %r11, %r12
 ; X64-NEXT:    shrq $2, %rbx
 ; X64-NEXT:    leaq (%rbx,%r10,4), %r10
-; X64-NEXT:    movabsq $6148820866244280320, %rbx # imm = 0x5555000000000000
-; X64-NEXT:    andq %r10, %rbx
-; X64-NEXT:    movabsq $-6149102341220990976, %rdi # imm = 0xAAAA000000000000
-; X64-NEXT:    andq %r10, %rdi
-; X64-NEXT:    shrq %rdi
-; X64-NEXT:    leaq (%rdi,%rbx,2), %rdi
-; X64-NEXT:    movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT:    movabsq $6148914691236517205, %r11 # imm = 0x5555555555555555
+; X64-NEXT:    movq %r10, %r13
+; X64-NEXT:    andq %r11, %r13
+; X64-NEXT:    movabsq $-6148914691236517206, %rbx # imm = 0xAAAAAAAAAAAAAAAA
+; X64-NEXT:    andq %rbx, %r10
+; X64-NEXT:    movq %rbx, %r15
+; X64-NEXT:    shrq %r10
+; X64-NEXT:    leaq (%r10,%r13,2), %rbx
+; X64-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    bswapq %rbp
-; X64-NEXT:    movq %rbp, %rdi
-; X64-NEXT:    andq %r13, %rdi
-; X64-NEXT:    shlq $4, %rdi
+; X64-NEXT:    movq %rbp, %r10
+; X64-NEXT:    andq %rdi, %r10
+; X64-NEXT:    shlq $4, %r10
 ; X64-NEXT:    andq %rax, %rbp
 ; X64-NEXT:    shrq $4, %rbp
-; X64-NEXT:    orq %rdi, %rbp
-; X64-NEXT:    movq %rbp, %rdi
-; X64-NEXT:    andq %r11, %rdi
-; X64-NEXT:    andq %r14, %rbp
+; X64-NEXT:    orq %r10, %rbp
+; X64-NEXT:    movq %rbp, %r10
+; X64-NEXT:    andq %r14, %r10
+; X64-NEXT:    movq %r12, %rbx
+; X64-NEXT:    andq %r12, %rbp
 ; X64-NEXT:    shrq $2, %rbp
-; X64-NEXT:    leaq (%rbp,%rdi,4), %rbp
-; X64-NEXT:    movabsq $6148914691236517205, %rbx # imm = 0x5555555555555555
+; X64-NEXT:    leaq (%rbp,%r10,4), %rbp
 ; X64-NEXT:    movq %rbp, %r10
-; X64-NEXT:    andq %rbx, %r10
-; X64-NEXT:    movabsq $-6148914691236517206, %rdi # imm = 0xAAAAAAAAAAAAAAAA
-; X64-NEXT:    andq %rdi, %rbp
+; X64-NEXT:    andq %r11, %r10
+; X64-NEXT:    andq %r15, %rbp
 ; X64-NEXT:    shrq %rbp
 ; X64-NEXT:    leaq (%rbp,%r10,2), %rbp
 ; X64-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
 ; X64-NEXT:    bswapq %rbp
 ; X64-NEXT:    movq %rbp, %r10
-; X64-NEXT:    andq %r13, %r10
+; X64-NEXT:    andq %rdi, %r10
 ; X64-NEXT:    shlq $4, %r10
 ; X64-NEXT:    andq %rax, %rbp
-; X64-NEXT:    movq %rax, %r15
+; X64-NEXT:    movq %rax, %r12
 ; X64-NEXT:    shrq $4, %rbp
 ; X64-NEXT:    orq %r10, %rbp
 ; X64-NEXT:    movq %rbp, %r10
-; X64-NEXT:    andq %r11, %r10
-; X64-NEXT:    andq %r14, %rbp
+; X64-NEXT:    andq %r14, %r10
+; X64-NEXT:    andq %rbx, %rbp
+; X64-NEXT:    movq %rbx, %r13
 ; X64-NEXT:    shrq $2, %rbp
 ; X64-NEXT:    leaq (%rbp,%r10,4), %rbp
 ; X64-NEXT:    movq %rbp, %r10
-; X64-NEXT:    andq %rbx, %r10
-; X64-NEXT:    andq %rdi, %rbp
+; X64-NEXT:    andq %r11, %r10
+; X64-NEXT:    andq %r15, %rbp
 ; X64-NEXT:    shrq %rbp
 ; X64-NEXT:    leaq (%rbp,%r10,2), %rbp
 ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %r10
 ; X64-NEXT:    bswapq %r10
 ; X64-NEXT:    movq %r10, %rax
-; X64-NEXT:    andq %r13, %rax
+; X64-NEXT:    andq %rdi, %rax
 ; X64-NEXT:    shlq $4, %rax
-; X64-NEXT:    movq %r15, %rdi
-; X64-NEXT:    andq %r15, %r10
+; X64-NEXT:    movq %r12, %rbx
+; X64-NEXT:    andq %r12, %r10
 ; X64-NEXT:    shrq $4, %r10
 ; X64-NEXT:    orq %rax, %r10
 ; X64-NEXT:    movq %r10, %rax
-; X64-NEXT:    andq %r11, %rax
-; X64-NEXT:    andq %r14, %r10
+; X64-NEXT:    andq %r14, %rax
+; X64-NEXT:    movq %r13, %r15
+; X64-NEXT:    andq %r13, %r10
 ; X64-NEXT:    shrq $2, %r10
 ; X64-NEXT:    leaq (%r10,%rax,4), %rax
 ; X64-NEXT:    movq %rax, %r10
-; X64-NEXT:    andq %rbx, %r10
-; X64-NEXT:    movabsq $-6148914691236517206, %r15 # imm = 0xAAAAAAAAAAAAAAAA
-; X64-NEXT:    andq %r15, %rax
+; X64-NEXT:    andq %r11, %r10
+; X64-NEXT:    movabsq $-6148914691236517206, %r12 # imm = 0xAAAAAAAAAAAAAAAA
+; X64-NEXT:    andq %r12, %rax
 ; X64-NEXT:    shrq %rax
 ; X64-NEXT:    leaq (%rax,%r10,2), %r10
 ; X64-NEXT:    bswapq %r9
 ; X64-NEXT:    movq %r9, %rax
-; X64-NEXT:    andq %r13, %rax
+; X64-NEXT:    andq %rdi, %rax
 ; X64-NEXT:    shlq $4, %rax
-; X64-NEXT:    andq %rdi, %r9
+; X64-NEXT:    andq %rbx, %r9
+; X64-NEXT:    movq %rbx, %r13
 ; X64-NEXT:    shrq $4, %r9
 ; X64-NEXT:    orq %rax, %r9
 ; X64-NEXT:    movq %r9, %rax
-; X64-NEXT:    andq %r11, %rax
-; X64-NEXT:    andq %r14, %r9
+; X64-NEXT:    andq %r14, %rax
+; X64-NEXT:    andq %r15, %r9
 ; X64-NEXT:    shrq $2, %r9
 ; X64-NEXT:    leaq (%r9,%rax,4), %rax
 ; X64-NEXT:    movq %rax, %r9
-; X64-NEXT:    andq %rbx, %r9
-; X64-NEXT:    andq %r15, %rax
+; X64-NEXT:    andq %r11, %r9
+; X64-NEXT:    andq %r12, %rax
 ; X64-NEXT:    shrq %rax
 ; X64-NEXT:    leaq (%rax,%r9,2), %r9
 ; X64-NEXT:    bswapq %r8
 ; X64-NEXT:    movq %r8, %rax
-; X64-NEXT:    andq %r13, %rax
+; X64-NEXT:    andq %rdi, %rax
 ; X64-NEXT:    shlq $4, %rax
-; X64-NEXT:    andq %rdi, %r8
+; X64-NEXT:    andq %rbx, %r8
 ; X64-NEXT:    shrq $4, %r8
 ; X64-NEXT:    orq %rax, %r8
 ; X64-NEXT:    movq %r8, %rax
-; X64-NEXT:    andq %r11, %rax
-; X64-NEXT:    andq %r14, %r8
+; X64-NEXT:    andq %r14, %rax
+; X64-NEXT:    andq %r15, %r8
 ; X64-NEXT:    shrq $2, %r8
 ; X64-NEXT:    leaq (%r8,%rax,4), %rax
 ; X64-NEXT:    movq %rax, %r8
-; X64-NEXT:    andq %rbx, %r8
-; X64-NEXT:    andq %r15, %rax
+; X64-NEXT:    andq %r11, %r8
+; X64-NEXT:    andq %r12, %rax
 ; X64-NEXT:    shrq %rax
 ; X64-NEXT:    leaq (%rax,%r8,2), %r8
 ; X64-NEXT:    bswapq %rcx
 ; X64-NEXT:    movq %rcx, %rax
-; X64-NEXT:    andq %r13, %rax
+; X64-NEXT:    andq %rdi, %rax
 ; X64-NEXT:    shlq $4, %rax
-; X64-NEXT:    andq %rdi, %rcx
+; X64-NEXT:    andq %rbx, %rcx
 ; X64-NEXT:    shrq $4, %rcx
 ; X64-NEXT:    orq %rax, %rcx
 ; X64-NEXT:    movq %rcx, %rax
-; X64-NEXT:    andq %r11, %rax
-; X64-NEXT:    andq %r14, %rcx
+; X64-NEXT:    andq %r14, %rax
+; X64-NEXT:    andq %r15, %rcx
 ; X64-NEXT:    shrq $2, %rcx
 ; X64-NEXT:    leaq (%rcx,%rax,4), %rax
 ; X64-NEXT:    movq %rax, %rcx
-; X64-NEXT:    andq %rbx, %rcx
-; X64-NEXT:    andq %r15, %rax
+; X64-NEXT:    andq %r11, %rcx
+; X64-NEXT:    andq %r12, %rax
+; X64-NEXT:    movq %r12, %rbx
 ; X64-NEXT:    shrq %rax
-; X64-NEXT:    leaq (%rax,%rcx,2), %rcx
+; X64-NEXT:    leaq (%rax,%rcx,2), %r12
 ; X64-NEXT:    bswapq %rdx
 ; X64-NEXT:    movq %rdx, %rax
-; X64-NEXT:    andq %r13, %rax
+; X64-NEXT:    andq %rdi, %rax
 ; X64-NEXT:    shlq $4, %rax
-; X64-NEXT:    andq %rdi, %rdx
+; X64-NEXT:    andq %r13, %rdx
 ; X64-NEXT:    shrq $4, %rdx
 ; X64-NEXT:    orq %rax, %rdx
 ; X64-NEXT:    movq %rdx, %rax
-; X64-NEXT:    andq %r11, %rax
-; X64-NEXT:    andq %r14, %rdx
+; X64-NEXT:    andq %r14, %rax
+; X64-NEXT:    andq %r15, %rdx
 ; X64-NEXT:    shrq $2, %rdx
 ; X64-NEXT:    leaq (%rdx,%rax,4), %rax
 ; X64-NEXT:    movq %rax, %rdx
-; X64-NEXT:    andq %rbx, %rdx
-; X64-NEXT:    andq %r15, %rax
+; X64-NEXT:    andq %r11, %rdx
+; X64-NEXT:    andq %rbx, %rax
 ; X64-NEXT:    shrq %rax
-; X64-NEXT:    leaq (%rax,%rdx,2), %rax
+; X64-NEXT:    leaq (%rax,%rdx,2), %rcx
 ; X64-NEXT:    bswapq %rsi
-; X64-NEXT:    andq %rsi, %r13
-; X64-NEXT:    andq %rdi, %rsi
-; X64-NEXT:    shlq $4, %r13
+; X64-NEXT:    andq %rsi, %rdi
+; X64-NEXT:    andq %r13, %rsi
+; X64-NEXT:    shlq $4, %rdi
 ; X64-NEXT:    shrq $4, %rsi
-; X64-NEXT:    orq %r13, %rsi
-; X64-NEXT:    andq %rsi, %r11
-; X64-NEXT:    andq %r14, %rsi
+; X64-NEXT:    orq %rdi, %rsi
+; X64-NEXT:    andq %rsi, %r14
+; X64-NEXT:    andq %r15, %rsi
 ; X64-NEXT:    shrq $2, %rsi
-; X64-NEXT:    leaq (%rsi,%r11,4), %rdx
-; X64-NEXT:    andq %rdx, %rbx
-; X64-NEXT:    andq %r15, %rdx
+; X64-NEXT:    leaq (%rsi,%r14,4), %rdx
+; X64-NEXT:    andq %rdx, %r11
+; X64-NEXT:    andq %rbx, %rdx
 ; X64-NEXT:    shrq %rdx
-; X64-NEXT:    leaq (%rdx,%rbx,2), %rdx
+; X64-NEXT:    leaq (%rdx,%r11,2), %rdx
 ; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
-; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
-; X64-NEXT:    shrdq $48, %rdi, %rsi
-; X64-NEXT:    shrdq $48, %rbp, %rdi
+; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; X64-NEXT:    shrdq $48, %rax, %rsi
+; X64-NEXT:    shrdq $48, %rbp, %rax
+; X64-NEXT:    movq %rax, %rdi
 ; X64-NEXT:    shrdq $48, %r10, %rbp
 ; X64-NEXT:    shrdq $48, %r9, %r10
 ; X64-NEXT:    shrdq $48, %r8, %r9
-; X64-NEXT:    shrdq $48, %rcx, %r8
-; X64-NEXT:    shrdq $48, %rax, %rcx
-; X64-NEXT:    shrdq $48, %rdx, %rax
-; X64-NEXT:    movq %rax, 56(%r12)
-; X64-NEXT:    movq %rcx, 48(%r12)
-; X64-NEXT:    movq %r8, 40(%r12)
-; X64-NEXT:    movq %r9, 32(%r12)
-; X64-NEXT:    movq %r10, 24(%r12)
-; X64-NEXT:    movq %rbp, 16(%r12)
-; X64-NEXT:    movq %rdi, 8(%r12)
-; X64-NEXT:    movq %rsi, (%r12)
+; X64-NEXT:    shrdq $48, %r12, %r8
+; X64-NEXT:    shrdq $48, %rcx, %r12
+; X64-NEXT:    shrdq $48, %rdx, %rcx
+; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; X64-NEXT:    movq %rcx, 56(%rax)
+; X64-NEXT:    movq %r12, 48(%rax)
+; X64-NEXT:    movq %r8, 40(%rax)
+; X64-NEXT:    movq %r9, 32(%rax)
+; X64-NEXT:    movq %r10, 24(%rax)
+; X64-NEXT:    movq %rbp, 16(%rax)
+; X64-NEXT:    movq %rdi, 8(%rax)
+; X64-NEXT:    movq %rsi, (%rax)
 ; X64-NEXT:    shrq $48, %rdx
-; X64-NEXT:    movw %dx, 64(%r12)
-; X64-NEXT:    movq %r12, %rax
+; X64-NEXT:    movw %dx, 64(%rax)
 ; X64-NEXT:    popq %rbx
 ; X64-NEXT:    popq %r12
 ; X64-NEXT:    popq %r13
 ; X64-NEXT:    popq %r14
 ; X64-NEXT:    popq %r15
 ; X64-NEXT:    popq %rbp
 ; X64-NEXT:    retq
   %Z = call i528 @llvm.bitreverse.i528(i528 %A)
   ret i528 %Z
 }
 declare i528 @llvm.bitreverse.i528(i528)

llvm/test/CodeGen/X86/const-shift-of-constmasked.ll

 ; X64-NEXT:    retq
   %t0 = and i64 %a0, 140737488289792
   %t1 = lshr i64 %t0, 15
   ret i64 %t1
 }
 define i64 @test_i64_140737488289792_mask_lshr_16(i64 %a0) {
 ; X32-LABEL: test_i64_140737488289792_mask_lshr_16:
 ; X32:       # %bb.0:
-; X32-NEXT:    movzwl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    movl $32767, %eax # imm = 0x7FFF
 ; X32-NEXT:    andl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    shll $16, %eax
-; X32-NEXT:    orl %ecx, %eax
+; X32-NEXT:    shldl $16, %ecx, %eax
 ; X32-NEXT:    xorl %edx, %edx
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_i64_140737488289792_mask_lshr_16:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movq %rdi, %rax
 ; X64-NEXT:    shrq $16, %rax
 ; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF

 ; X64-NEXT:    retq
   %t0 = and i64 %a0, 140737488289792
   %t1 = ashr i64 %t0, 15
   ret i64 %t1
 }
 define i64 @test_i64_140737488289792_mask_ashr_16(i64 %a0) {
 ; X32-LABEL: test_i64_140737488289792_mask_ashr_16:
 ; X32:       # %bb.0:
-; X32-NEXT:    movzwl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    movl $32767, %eax # imm = 0x7FFF
 ; X32-NEXT:    andl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    shll $16, %eax
-; X32-NEXT:    orl %ecx, %eax
+; X32-NEXT:    shldl $16, %ecx, %eax
 ; X32-NEXT:    xorl %edx, %edx
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_i64_140737488289792_mask_ashr_16:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movq %rdi, %rax
 ; X64-NEXT:    shrq $16, %rax
 ; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF

llvm/test/CodeGen/X86/fshl.ll

; X64-NEXT:    retq
   %p1 = getelementptr i32, i32* %p, i32 3
   %ld0 = load i32, i32 *%p0
   %ld1 = load i32, i32 *%p1
   %res = call i32 @llvm.fshl.i32(i32 %ld1, i32 %ld0, i32 8)
   ret i32 %res
 }
 
 define i64 @combine_fshl_load_i64(i64* %p) nounwind {
-; X86-FAST-LABEL: combine_fshl_load_i64:
-; X86-FAST:       # %bb.0:
-; X86-FAST-NEXT:    pushl %esi
-; X86-FAST-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-FAST-NEXT:    movl 12(%ecx), %eax
-; X86-FAST-NEXT:    movl 16(%ecx), %esi
-; X86-FAST-NEXT:    movl 20(%ecx), %edx
-; X86-FAST-NEXT:    shldl $24, %esi, %edx
-; X86-FAST-NEXT:    shrdl $8, %esi, %eax
-; X86-FAST-NEXT:    popl %esi
-; X86-FAST-NEXT:    retl
-;
-; X86-SLOW-LABEL: combine_fshl_load_i64:
-; X86-SLOW:       # %bb.0:
-; X86-SLOW-NEXT:    pushl %esi
-; X86-SLOW-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-SLOW-NEXT:    movl 20(%eax), %edx
-; X86-SLOW-NEXT:    movl 12(%eax), %ecx
-; X86-SLOW-NEXT:    movl 16(%eax), %esi
-; X86-SLOW-NEXT:    shrl $8, %ecx
-; X86-SLOW-NEXT:    movl %esi, %eax
-; X86-SLOW-NEXT:    shll $24, %eax
-; X86-SLOW-NEXT:    orl %ecx, %eax
-; X86-SLOW-NEXT:    shrl $8, %esi
-; X86-SLOW-NEXT:    shll $24, %edx
-; X86-SLOW-NEXT:    orl %esi, %edx
-; X86-SLOW-NEXT:    popl %esi
-; X86-SLOW-NEXT:    retl
+; X86-LABEL: combine_fshl_load_i64:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl 13(%ecx), %eax
+; X86-NEXT:    movl 17(%ecx), %edx
+; X86-NEXT:    retl
 ;
 ; X64-LABEL: combine_fshl_load_i64:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movq 13(%rdi), %rax
 ; X64-NEXT:    retq
   %p0 = getelementptr i64, i64* %p, i64 1
   %p1 = getelementptr i64, i64* %p, i64 2
   %ld0 = load i64, i64 *%p0

llvm/test/CodeGen/X86/fshr.ll

 ; X86-FAST-NEXT:    shldl $25, %ecx, %edx
 ; X86-FAST-NEXT:    shrdl $7, %ecx, %eax
 ; X86-FAST-NEXT:    retl
 ;
 ; X86-SLOW-LABEL: const_shift_i64:
 ; X86-SLOW:       # %bb.0:
 ; X86-SLOW-NEXT:    pushl %esi
 ; X86-SLOW-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-SLOW-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-SLOW-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-SLOW-NEXT:    movl %ecx, %esi
-; X86-SLOW-NEXT:    shll $25, %esi
-; X86-SLOW-NEXT:    shrl $7, %eax
-; X86-SLOW-NEXT:    orl %esi, %eax
+; X86-SLOW-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; X86-SLOW-NEXT:    shrl $7, %ecx
+; X86-SLOW-NEXT:    movl %esi, %eax
+; X86-SLOW-NEXT:    shll $25, %eax
+; X86-SLOW-NEXT:    orl %ecx, %eax
+; X86-SLOW-NEXT:    shrl $7, %esi
 ; X86-SLOW-NEXT:    shll $25, %edx
-; X86-SLOW-NEXT:    orl %ecx, %edx
+; X86-SLOW-NEXT:    orl %esi, %edx
 ; X86-SLOW-NEXT:    popl %esi
 ; X86-SLOW-NEXT:    retl
 ;
 ; X64-FAST-LABEL: const_shift_i64:
 ; X64-FAST:       # %bb.0:
 ; X64-FAST-NEXT:    movq %rdi, %rax
 ; X64-FAST-NEXT:    shldq $57, %rsi, %rax
 ; X64-FAST-NEXT:    retq
; X64-NEXT:    retq
   %p1 = getelementptr i32, i32* %p, i32 3
   %ld0 = load i32, i32 *%p0
   %ld1 = load i32, i32 *%p1
   %res = call i32 @llvm.fshr.i32(i32 %ld1, i32 %ld0, i32 8)
   ret i32 %res
 }
 
 define i64 @combine_fshr_load_i64(i64* %p) nounwind {
-; X86-FAST-LABEL: combine_fshr_load_i64:
-; X86-FAST:       # %bb.0:
-; X86-FAST-NEXT:    pushl %esi
-; X86-FAST-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-FAST-NEXT:    movzbl 11(%eax), %ecx
-; X86-FAST-NEXT:    movl 12(%eax), %esi
-; X86-FAST-NEXT:    movl 16(%eax), %edx
-; X86-FAST-NEXT:    shldl $8, %esi, %edx
-; X86-FAST-NEXT:    movl %esi, %eax
-; X86-FAST-NEXT:    shll $8, %eax
-; X86-FAST-NEXT:    orl %ecx, %eax
-; X86-FAST-NEXT:    popl %esi
-; X86-FAST-NEXT:    retl
-;
-; X86-SLOW-LABEL: combine_fshr_load_i64:
-; X86-SLOW:       # %bb.0:
-; X86-SLOW-NEXT:    pushl %esi
-; X86-SLOW-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-SLOW-NEXT:    movzbl 11(%eax), %ecx
-; X86-SLOW-NEXT:    movl 12(%eax), %esi
-; X86-SLOW-NEXT:    movl 16(%eax), %edx
-; X86-SLOW-NEXT:    movl %esi, %eax
-; X86-SLOW-NEXT:    shll $8, %eax
-; X86-SLOW-NEXT:    orl %ecx, %eax
-; X86-SLOW-NEXT:    shrl $24, %esi
-; X86-SLOW-NEXT:    shll $8, %edx
-; X86-SLOW-NEXT:    orl %esi, %edx
-; X86-SLOW-NEXT:    popl %esi
-; X86-SLOW-NEXT:    retl
+; X86-LABEL: combine_fshr_load_i64:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl 11(%ecx), %eax
+; X86-NEXT:    movl 15(%ecx), %edx
+; X86-NEXT:    retl
 ;
 ; X64-LABEL: combine_fshr_load_i64:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movq 11(%rdi), %rax
 ; X64-NEXT:    retq
   %p0 = getelementptr i64, i64* %p, i64 1
   %p1 = getelementptr i64, i64* %p, i64 2
   %ld0 = load i64, i64 *%p0

llvm/test/CodeGen/X86/known-bits.ll

 ; X32-NEXT:    popl %edi
 ; X32-NEXT:    retl $4
 ;
 ; X64-LABEL: knownbits_mask_addc_shl:
 ; X64:       # %bb.0:
 ; X64-NEXT:    andq $-1024, %rdi # imm = 0xFC00
 ; X64-NEXT:    andq $-1024, %rsi # imm = 0xFC00
 ; X64-NEXT:    addq %rdi, %rsi
-; X64-NEXT:    adcl $0, %edx
+; X64-NEXT:    adcq $0, %rdx
 ; X64-NEXT:    shldq $54, %rsi, %rdx
 ; X64-NEXT:    xorl %eax, %eax
 ; X64-NEXT:    retq
   %1 = and i64 %a0, -1024
   %2 = zext i64 %1 to i128
   %3 = and i64 %a1, -1024
   %4 = zext i64 %3 to i128
   %5 = add i128 %2, %4

llvm/test/CodeGen/X86/pr43820.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
 
 define i1000 @square(i1000 %A) nounwind {
 ; CHECK-LABEL: square:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    pushq %rbp
 ; CHECK-NEXT:    pushq %r15
 ; CHECK-NEXT:    pushq %r14
 ; CHECK-NEXT:    pushq %r13
 ; CHECK-NEXT:    pushq %r12
 ; CHECK-NEXT:    pushq %rbx
+; CHECK-NEXT:    movq %rsi, %r12
 ; CHECK-NEXT:    movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
 ; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbx
 ; CHECK-NEXT:    bswapq %rbx
 ; CHECK-NEXT:    movabsq $1085102592571150095, %rdi # imm = 0xF0F0F0F0F0F0F0F
 ; CHECK-NEXT:    movq %rbx, %rbp
 ; CHECK-NEXT:    andq %rdi, %rbp
 ; CHECK-NEXT:    shlq $4, %rbp
-; CHECK-NEXT:    movabsq $-1085102592571150096, %r11 # imm = 0xF0F0F0F0F0F0F0F0
-; CHECK-NEXT:    andq %r11, %rbx
-; CHECK-NEXT:    movq %r11, %rax
+; CHECK-NEXT:    movabsq $-1085102592571150096, %rax # imm = 0xF0F0F0F0F0F0F0F0
+; CHECK-NEXT:    andq %rax, %rbx
+; CHECK-NEXT:    movq %rax, %rsi
 ; CHECK-NEXT:    shrq $4, %rbx
 ; CHECK-NEXT:    orq %rbp, %rbx
 ; CHECK-NEXT:    movabsq $3689348814741910323, %r11 # imm = 0x3333333333333333
 ; CHECK-NEXT:    movq %rbx, %r14
 ; CHECK-NEXT:    andq %r11, %r14
-; CHECK-NEXT:    movabsq $-3689348814741910324, %rbp # imm = 0xCCCCCCCCCCCCCCCC
-; CHECK-NEXT:    andq %rbp, %rbx
-; CHECK-NEXT:    movq %rbp, %r15
+; CHECK-NEXT:    movabsq $-3689348814741910324, %rax # imm = 0xCCCCCCCCCCCCCCCC
+; CHECK-NEXT:    andq %rax, %rbx
 ; CHECK-NEXT:    shrq $2, %rbx
 ; CHECK-NEXT:    leaq (%rbx,%r14,4), %r14
-; CHECK-NEXT:    movabsq $6148914691230924800, %rbx # imm = 0x5555555555000000
-; CHECK-NEXT:    andq %r14, %rbx
-; CHECK-NEXT:    movabsq $-6148914691247702016, %rbp # imm = 0xAAAAAAAAAA000000
-; CHECK-NEXT:    andq %r14, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%rbx,2), %rbx
-; CHECK-NEXT:    movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movabsq $6148914691236517205, %rbx # imm = 0x5555555555555555
+; CHECK-NEXT:    movq %r14, %r15
+; CHECK-NEXT:    andq %rbx, %r15
+; CHECK-NEXT:    movabsq $-6148914691236517206, %rbp # imm = 0xAAAAAAAAAAAAAAAA
+; CHECK-NEXT:    andq %rbp, %r14
+; CHECK-NEXT:    movq %rbp, %r13
+; CHECK-NEXT:    shrq %r14
+; CHECK-NEXT:    leaq (%r14,%r15,2), %rbp
+; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; CHECK-NEXT:    bswapq %r10
-; CHECK-NEXT:    movq %r10, %rbx
-; CHECK-NEXT:    andq %rdi, %rbx
-; CHECK-NEXT:    shlq $4, %rbx
-; CHECK-NEXT:    andq %rax, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rsi, %r10
 ; CHECK-NEXT:    shrq $4, %r10
-; CHECK-NEXT:    orq %rbx, %r10
-; CHECK-NEXT:    movq %r10, %rbx
-; CHECK-NEXT:    andq %r11, %rbx
-; CHECK-NEXT:    andq %r15, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    andq %rax, %r10
 ; CHECK-NEXT:    shrq $2, %r10
-; CHECK-NEXT:    leaq (%r10,%rbx,4), %rbp
-; CHECK-NEXT:    movabsq $6148914691236517205, %rbx # imm = 0x5555555555555555
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    movabsq $-6148914691236517206, %r13 # imm = 0xAAAAAAAAAAAAAAAA
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %rax, %rbp
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rbp
 ; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %rax, %rbp
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %rax, %rbp
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %rax, %rbp
-; CHECK-NEXT:    movq %rax, %r14
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rax
-; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %r14, %rbp
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %r14, %rbp
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %r14, %rbp
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %r14, %rbp
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    andq %rax, %r10
+; CHECK-NEXT:    movq %rax, %rbp
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    movq %rbp, %rax
+; CHECK-NEXT:    andq %rbp, %r10
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rbp
 ; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rbp
-; CHECK-NEXT:    bswapq %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %rdi, %r10
-; CHECK-NEXT:    shlq $4, %r10
-; CHECK-NEXT:    andq %r14, %rbp
-; CHECK-NEXT:    shrq $4, %rbp
-; CHECK-NEXT:    orq %r10, %rbp
-; CHECK-NEXT:    movq %rbp, %r10
-; CHECK-NEXT:    andq %r11, %r10
-; CHECK-NEXT:    andq %r15, %rbp
-; CHECK-NEXT:    shrq $2, %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r10
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    movq %rsi, %rbp
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    andq %rax, %r10
+; CHECK-NEXT:    movq %rax, %rsi
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rbp, %r10
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rbp, %r10
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rbp, %r10
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rbp, %r10
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rdi, %r14
+; CHECK-NEXT:    shlq $4, %r14
+; CHECK-NEXT:    andq %rbp, %r10
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %r14, %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %r11, %r14
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%r14,4), %r10
+; CHECK-NEXT:    movq %r10, %r14
+; CHECK-NEXT:    andq %rbx, %r14
+; CHECK-NEXT:    andq %r13, %r10
+; CHECK-NEXT:    shrq %r10
+; CHECK-NEXT:    leaq (%r10,%r14,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %r10
+; CHECK-NEXT:    bswapq %r10
+; CHECK-NEXT:    movq %r10, %rax
+; CHECK-NEXT:    andq %rdi, %rax
+; CHECK-NEXT:    shlq $4, %rax
+; CHECK-NEXT:    andq %rbp, %r10
+; CHECK-NEXT:    shrq $4, %r10
+; CHECK-NEXT:    orq %rax, %r10
+; CHECK-NEXT:    movq %r10, %rax
+; CHECK-NEXT:    andq %r11, %rax
+; CHECK-NEXT:    andq %rsi, %r10
+; CHECK-NEXT:    shrq $2, %r10
+; CHECK-NEXT:    leaq (%r10,%rax,4), %rax
+; CHECK-NEXT:    movq %rax, %r10
 ; CHECK-NEXT:    andq %rbx, %r10
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r10,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    andq %r13, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    leaq (%rax,%r10,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; CHECK-NEXT:    bswapq %r9
-; CHECK-NEXT:    movq %r9, %rbp
-; CHECK-NEXT:    andq %rdi, %rbp
-; CHECK-NEXT:    shlq $4, %rbp
-; CHECK-NEXT:    andq %r14, %r9
+; CHECK-NEXT:    movq %r9, %rax
+; CHECK-NEXT:    andq %rdi, %rax
+; CHECK-NEXT:    shlq $4, %rax
+; CHECK-NEXT:    andq %rbp, %r9
 ; CHECK-NEXT:    shrq $4, %r9
-; CHECK-NEXT:    orq %rbp, %r9
-; CHECK-NEXT:    movq %r9, %rbp
-; CHECK-NEXT:    andq %r11, %rbp
-; CHECK-NEXT:    andq %r15, %r9
+; CHECK-NEXT:    orq %rax, %r9
+; CHECK-NEXT:    movq %r9, %rax
+; CHECK-NEXT:    andq %r11, %rax
+; CHECK-NEXT:    andq %rsi, %r9
 ; CHECK-NEXT:    shrq $2, %r9
-; CHECK-NEXT:    leaq (%r9,%rbp,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r9
+; CHECK-NEXT:    leaq (%r9,%rax,4), %rax
+; CHECK-NEXT:    movq %rax, %r9
 ; CHECK-NEXT:    andq %rbx, %r9
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r9,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    andq %r13, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    leaq (%rax,%r9,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; CHECK-NEXT:    bswapq %r8
-; CHECK-NEXT:    movq %r8, %rbp
-; CHECK-NEXT:    andq %rdi, %rbp
-; CHECK-NEXT:    shlq $4, %rbp
-; CHECK-NEXT:    andq %r14, %r8
+; CHECK-NEXT:    movq %r8, %rax
+; CHECK-NEXT:    andq %rdi, %rax
+; CHECK-NEXT:    shlq $4, %rax
+; CHECK-NEXT:    andq %rbp, %r8
 ; CHECK-NEXT:    shrq $4, %r8
-; CHECK-NEXT:    orq %rbp, %r8
-; CHECK-NEXT:    movq %r8, %rbp
-; CHECK-NEXT:    andq %r11, %rbp
-; CHECK-NEXT:    andq %r15, %r8
-; CHECK-NEXT:    movq %r15, %r9
+; CHECK-NEXT:    orq %rax, %r8
+; CHECK-NEXT:    movq %r8, %rax
+; CHECK-NEXT:    andq %r11, %rax
+; CHECK-NEXT:    andq %rsi, %r8
 ; CHECK-NEXT:    shrq $2, %r8
-; CHECK-NEXT:    leaq (%r8,%rbp,4), %rbp
-; CHECK-NEXT:    movq %rbp, %r8
+; CHECK-NEXT:    leaq (%r8,%rax,4), %rax
+; CHECK-NEXT:    movq %rax, %r8
 ; CHECK-NEXT:    andq %rbx, %r8
-; CHECK-NEXT:    andq %r13, %rbp
-; CHECK-NEXT:    shrq %rbp
-; CHECK-NEXT:    leaq (%rbp,%r8,2), %rbp
-; CHECK-NEXT:    movq %rbp, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    andq %r13, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    leaq (%rax,%r8,2), %rax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; CHECK-NEXT:    bswapq %rcx
-; CHECK-NEXT:    movq %rcx, %rbp
-; CHECK-NEXT:    andq %rdi, %rbp
-; CHECK-NEXT:    shlq $4, %rbp
-; CHECK-NEXT:    andq %r14, %rcx
+; CHECK-NEXT:    movq %rcx, %rax
+; CHECK-NEXT:    andq %rdi, %rax
+; CHECK-NEXT:    shlq $4, %rax
+; CHECK-NEXT:    andq %rbp, %rcx
 ; CHECK-NEXT:    shrq $4, %rcx
-; CHECK-NEXT:    orq %rbp, %rcx
-; CHECK-NEXT:    movq %rcx, %rbp
-; CHECK-NEXT:    andq %r11, %rbp
-; CHECK-NEXT:    andq %r15, %rcx
+; CHECK-NEXT:    orq %rax, %rcx
+; CHECK-NEXT:    movq %rcx, %rax
+; CHECK-NEXT:    andq %r11, %rax
+; CHECK-NEXT:    andq %rsi, %rcx
 ; CHECK-NEXT:    shrq $2, %rcx
-; CHECK-NEXT:    leaq (%rcx,%rbp,4), %rcx
-; CHECK-NEXT:    movq %rcx, %rbp
-; CHECK-NEXT:    andq %rbx, %rbp
-; CHECK-NEXT:    andq %r13, %rcx
-; CHECK-NEXT:    shrq %rcx
-; CHECK-NEXT:    leaq (%rcx,%rbp,2), %r15
+; CHECK-NEXT:    leaq (%rcx,%rax,4), %rax
+; CHECK-NEXT:    movq %rax, %rcx
+; CHECK-NEXT:    andq %rbx, %rcx
+; CHECK-NEXT:    andq %r13, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    leaq (%rax,%rcx,2), %r15
 ; CHECK-NEXT:    bswapq %rdx
-; CHECK-NEXT:    movq %rdx, %rbp
-; CHECK-NEXT:    andq %rdi, %rbp
-; CHECK-NEXT:    shlq $4, %rbp
-; CHECK-NEXT:    andq %r14, %rdx
+; CHECK-NEXT:    movq %rdx, %rax
+; CHECK-NEXT:    andq %rdi, %rax
+; CHECK-NEXT:    shlq $4, %rax
+; CHECK-NEXT:    andq %rbp, %rdx
 ; CHECK-NEXT:    shrq $4, %rdx
-; CHECK-NEXT:    orq %rbp, %rdx
-; CHECK-NEXT:    movq %rdx, %rbp
-; CHECK-NEXT:    andq %r11, %rbp
-; CHECK-NEXT:    andq %r9, %rdx
+; CHECK-NEXT:    orq %rax, %rdx
+; CHECK-NEXT:    movq %rdx, %rax
+; CHECK-NEXT:    andq %r11, %rax
+; CHECK-NEXT:    andq %rsi, %rdx
 ; CHECK-NEXT:    shrq $2, %rdx
-; CHECK-NEXT:    leaq (%rdx,%rbp,4), %rdx
-; CHECK-NEXT:    movq %rdx, %rbp
-; CHECK-NEXT:    andq %rbx, %rbp
-; CHECK-NEXT:    andq %r13, %rdx
-; CHECK-NEXT:    shrq %rdx
-; CHECK-NEXT:    leaq (%rdx,%rbp,2), %rdx
-; CHECK-NEXT:    bswapq %rsi
-; CHECK-NEXT:    andq %rsi, %rdi
-; CHECK-NEXT:    andq %r14, %rsi
+; CHECK-NEXT:    leaq (%rdx,%rax,4), %rax
+; CHECK-NEXT:    movq %rax, %rdx
+; CHECK-NEXT:    andq %rbx, %rdx
+; CHECK-NEXT:    andq %r13, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    leaq (%rax,%rdx,2), %rdx
+; CHECK-NEXT:    bswapq %r12
+; CHECK-NEXT:    andq %r12, %rdi
+; CHECK-NEXT:    andq %rbp, %r12
 ; CHECK-NEXT:    shlq $4, %rdi
-; CHECK-NEXT:    shrq $4, %rsi
-; CHECK-NEXT:    orq %rdi, %rsi
-; CHECK-NEXT:    andq %rsi, %r11
-; CHECK-NEXT:    andq %r9, %rsi
-; CHECK-NEXT:    shrq $2, %rsi
-; CHECK-NEXT:    leaq (%rsi,%r11,4), %rsi
-; CHECK-NEXT:    andq %rsi, %rbx
-; CHECK-NEXT:    andq %r13, %rsi
-; CHECK-NEXT:    shrq %rsi
-; CHECK-NEXT:    leaq (%rsi,%rbx,2), %r13
+; CHECK-NEXT:    shrq $4, %r12
+; CHECK-NEXT:    orq %rdi, %r12
+; CHECK-NEXT:    andq %r12, %r11
+; CHECK-NEXT:    andq %rsi, %r12
+; CHECK-NEXT:    shrq $2, %r12
+; CHECK-NEXT:    leaq (%r12,%r11,4), %rax
+; CHECK-NEXT:    andq %rax, %rbx
+; CHECK-NEXT:    andq %r13, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    leaq (%rax,%rbx,2), %r12
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r11 # 8-byte Reload
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %rax, %r11
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %rcx, %rax
 ; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %rbp, %rcx
 ; CHECK-NEXT:    movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r12 # 8-byte Reload
-; CHECK-NEXT:    shrdq $24, %r12, %rbp
+; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Reload
+; CHECK-NEXT:    shrdq $24, %r13, %rbp
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r14 # 8-byte Reload
-; CHECK-NEXT:    shrdq $24, %r14, %r12
+; CHECK-NEXT:    shrdq $24, %r14, %r13
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %rbx, %r14
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r10 # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %r10, %rbx
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r9 # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %r9, %r10
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r8 # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %r8, %r9
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %rdi, %r8
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %rsi, %rdi
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
 ; CHECK-NEXT:    shrdq $24, %rax, %rsi
 ; CHECK-NEXT:    shrdq $24, %r15, %rax
 ; CHECK-NEXT:    movq %rax, %rcx
 ; CHECK-NEXT:    shrdq $24, %rdx, %r15
-; CHECK-NEXT:    shrdq $24, %r13, %rdx
+; CHECK-NEXT:    shrdq $24, %r12, %rdx
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
 ; CHECK-NEXT:    movq %rdx, 112(%rax)
 ; CHECK-NEXT:    movq %r15, 104(%rax)
 ; CHECK-NEXT:    movq %rcx, 96(%rax)
 ; CHECK-NEXT:    movq %rsi, 88(%rax)
 ; CHECK-NEXT:    movq %rdi, 80(%rax)
 ; CHECK-NEXT:    movq %r8, 72(%rax)
 ; CHECK-NEXT:    movq %r9, 64(%rax)
 ; CHECK-NEXT:    movq %r10, 56(%rax)
 ; CHECK-NEXT:    movq %rbx, 48(%rax)
 ; CHECK-NEXT:    movq %r14, 40(%rax)
-; CHECK-NEXT:    movq %r12, 32(%rax)
+; CHECK-NEXT:    movq %r13, 32(%rax)
 ; CHECK-NEXT:    movq %rbp, 24(%rax)
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
 ; CHECK-NEXT:    movq %rcx, 16(%rax)
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
 ; CHECK-NEXT:    movq %rcx, 8(%rax)
 ; CHECK-NEXT:    movq %r11, (%rax)
-; CHECK-NEXT:    movq %r13, %rcx
-; CHECK-NEXT:    shrq $56, %r13
-; CHECK-NEXT:    movb %r13b, 124(%rax)
+; CHECK-NEXT:    movq %r12, %rcx
+; CHECK-NEXT:    shrq $56, %r12
+; CHECK-NEXT:    movb %r12b, 124(%rax)
 ; CHECK-NEXT:    shrq $24, %rcx
 ; CHECK-NEXT:    movl %ecx, 120(%rax)
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    popq %r12
 ; CHECK-NEXT:    popq %r13
 ; CHECK-NEXT:    popq %r14
 ; CHECK-NEXT:    popq %r15
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    retq
   %Z = call i1000 @llvm.bitreverse.i1000(i1000 %A)
   ret i1000 %Z
 }
 
 declare i1000 @llvm.bitreverse.i1000(i1000)

llvm/test/CodeGen/X86/sdiv_fix.ll

 ; X64-LABEL: func5:
 ; X64:       # %bb.0:
 ; X64-NEXT:    pushq %rbp
 ; X64-NEXT:    pushq %r15
 ; X64-NEXT:    pushq %r14
 ; X64-NEXT:    pushq %r13
 ; X64-NEXT:    pushq %r12
 ; X64-NEXT:    pushq %rbx
-; X64-NEXT:    subq $24, %rsp
+; X64-NEXT:    pushq %rax
 ; X64-NEXT:    movq %rsi, %r14
 ; X64-NEXT:    movq %rdi, %r15
-; X64-NEXT:    movq %rdi, %rax
-; X64-NEXT:    shrq $33, %rax
 ; X64-NEXT:    movq %rdi, %rbx
 ; X64-NEXT:    sarq $63, %rbx
-; X64-NEXT:    shlq $31, %rbx
-; X64-NEXT:    orq %rax, %rbx
-; X64-NEXT:    sets {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Folded Spill
+; X64-NEXT:    shldq $31, %rdi, %rbx
 ; X64-NEXT:    shlq $31, %r15
 ; X64-NEXT:    movq %rsi, %r12
 ; X64-NEXT:    sarq $63, %r12
 ; X64-NEXT:    movq %r15, %rdi
 ; X64-NEXT:    movq %rbx, %rsi
 ; X64-NEXT:    movq %r14, %rdx
 ; X64-NEXT:    movq %r12, %rcx
 ; X64-NEXT:    callq __divti3
 ; X64-NEXT:    movq %rax, %r13
 ; X64-NEXT:    decq %rax
-; X64-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT:    movq %rax, (%rsp) # 8-byte Spill
+; X64-NEXT:    testq %rbx, %rbx
+; X64-NEXT:    sets %al
 ; X64-NEXT:    testq %r12, %r12
 ; X64-NEXT:    sets %bpl
-; X64-NEXT:    xorb {{[-0-9]+}}(%r{{[sb]}}p), %bpl # 1-byte Folded Reload
+; X64-NEXT:    xorb %al, %bpl
 ; X64-NEXT:    movq %r15, %rdi
 ; X64-NEXT:    movq %rbx, %rsi
 ; X64-NEXT:    movq %r14, %rdx
 ; X64-NEXT:    movq %r12, %rcx
 ; X64-NEXT:    callq __modti3
 ; X64-NEXT:    orq %rax, %rdx
 ; X64-NEXT:    setne %al
 ; X64-NEXT:    testb %bpl, %al
-; X64-NEXT:    cmovneq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Folded Reload
+; X64-NEXT:    cmovneq (%rsp), %r13 # 8-byte Folded Reload
 ; X64-NEXT:    movq %r13, %rax
-; X64-NEXT:    addq $24, %rsp
+; X64-NEXT:    addq $8, %rsp
 ; X64-NEXT:    popq %rbx
 ; X64-NEXT:    popq %r12
 ; X64-NEXT:    popq %r13
 ; X64-NEXT:    popq %r14
 ; X64-NEXT:    popq %r15
 ; X64-NEXT:    popq %rbp
 ; X64-NEXT:    retq
 ;

 ; X64-NEXT:    retq
 ;
 ; X86-LABEL: vec:
 ; X86:       # %bb.0:
 ; X86-NEXT:    pushl %ebp
 ; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %edi
 ; X86-NEXT:    pushl %esi
-; X86-NEXT:    subl $64, %esp
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; X86-NEXT:    subl $60, %esp
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    movl %ecx, %edx
-; X86-NEXT:    sarl $31, %edx
-; X86-NEXT:    movl %edi, %esi
-; X86-NEXT:    shll $31, %esi
-; X86-NEXT:    movl %ebx, %eax
-; X86-NEXT:    shrl %eax
-; X86-NEXT:    andl $-2147483648, %ebx # imm = 0x80000000
-; X86-NEXT:    orl %eax, %ebx
-; X86-NEXT:    sets {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
-; X86-NEXT:    movl %ebp, %eax
-; X86-NEXT:    shrl %eax
-; X86-NEXT:    andl $-2147483648, %ebp # imm = 0x80000000
-; X86-NEXT:    orl %eax, %ebp
-; X86-NEXT:    movl %ebp, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    sets {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    shrl %eax
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
-; X86-NEXT:    andl $-2147483648, %ebp # imm = 0x80000000
-; X86-NEXT:    orl %eax, %ebp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl %esi, %ebp
+; X86-NEXT:    sarl $31, %ebp
 ; X86-NEXT:    movl %ebp, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    sets (%esp) # 1-byte Folded Spill
-; X86-NEXT:    movl %edi, %eax
-; X86-NEXT:    shrl %eax
-; X86-NEXT:    andl $-2147483648, %edi # imm = 0x80000000
-; X86-NEXT:    orl %eax, %edi
-; X86-NEXT:    sets {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
-; X86-NEXT:    pushl %edx
-; X86-NEXT:    movl %edx, %ebp
+; X86-NEXT:    movl %ecx, %edx
+; X86-NEXT:    shll $31, %edx
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    pushl %ecx
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl %esi
-; X86-NEXT:    calll __moddi3
-; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    shrl $31, %eax
+; X86-NEXT:    shldl $31, %ecx, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    pushl %ebp
-; X86-NEXT:    pushl {{[0-9]+}}(%esp)
-; X86-NEXT:    pushl %edi
 ; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %edx
 ; X86-NEXT:    calll __divdi3
 ; X86-NEXT:    addl $16, %esp
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    shll $31, %ecx
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-NEXT:    movl %edx, %eax
+; X86-NEXT:    movl %edi, %eax
 ; X86-NEXT:    sarl $31, %eax
+; X86-NEXT:    movl %ebx, %ebp
+; X86-NEXT:    shll $31, %ebp
+; X86-NEXT:    movl %ebx, %ecx
+; X86-NEXT:    shrl $31, %ecx
+; X86-NEXT:    shldl $31, %ebx, %ecx
 ; X86-NEXT:    pushl %eax
 ; X86-NEXT:    movl %eax, %esi
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    pushl %edx
-; X86-NEXT:    movl %edx, %ebp
-; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
 ; X86-NEXT:    pushl %ecx
-; X86-NEXT:    movl %ecx, %edi
+; X86-NEXT:    movl %ecx, %ebx
+; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    pushl %ebp
 ; X86-NEXT:    calll __moddi3
 ; X86-NEXT:    addl $16, %esp
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    pushl %esi
-; X86-NEXT:    pushl %ebp
-; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %ebp
 ; X86-NEXT:    calll __divdi3
 ; X86-NEXT:    addl $16, %esp
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    shll $31, %eax
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    movl %ecx, %edx
-; X86-NEXT:    sarl $31, %edx
-; X86-NEXT:    pushl %edx
-; X86-NEXT:    movl %edx, %ebp
-; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %eax, %ecx
+; X86-NEXT:    sarl $31, %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    movl %edx, %ebx
+; X86-NEXT:    shll $31, %ebx
+; X86-NEXT:    movl %edx, %edi
+; X86-NEXT:    shrl $31, %edi
+; X86-NEXT:    shldl $31, %edx, %edi
 ; X86-NEXT:    pushl %ecx
-; X86-NEXT:    movl %ecx, %edi
-; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload
-; X86-NEXT:    pushl %ebx
+; X86-NEXT:    movl %ecx, %ebp
+; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    pushl %eax
 ; X86-NEXT:    movl %eax, %esi
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    calll __moddi3
 ; X86-NEXT:    addl $16, %esp
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl %esi
 ; X86-NEXT:    pushl %edi
 ; X86-NEXT:    pushl %ebx
-; X86-NEXT:    pushl %esi
 ; X86-NEXT:    calll __divdi3
 ; X86-NEXT:    addl $16, %esp
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    shll $31, %eax
-; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %eax, %ebx
+; X86-NEXT:    sarl $31, %ebx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    movl %ecx, %ebp
-; X86-NEXT:    sarl $31, %ebp
-; X86-NEXT:    pushl %ebp
-; X86-NEXT:    pushl %ecx
-; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
-; X86-NEXT:    pushl %esi
+; X86-NEXT:    shll $31, %ebp
+; X86-NEXT:    movl %ecx, %esi
+; X86-NEXT:    shrl $31, %esi
+; X86-NEXT:    shldl $31, %ecx, %esi
+; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %ebp
 ; X86-NEXT:    calll __moddi3
 ; X86-NEXT:    addl $16, %esp
-; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    movl %edx, %edi
-; X86-NEXT:    testl %ebp, %ebp
-; X86-NEXT:    sets %bl
-; X86-NEXT:    xorb (%esp), %bl # 1-byte Folded Reload
-; X86-NEXT:    pushl %ebp
+; X86-NEXT:    movl %eax, (%esp) # 4-byte Spill
+; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl {{[0-9]+}}(%esp)
 ; X86-NEXT:    pushl %esi
-; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    pushl %ebp
 ; X86-NEXT:    calll __divdi3
 ; X86-NEXT:    addl $16, %esp
-; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %edi # 4-byte Folded Reload
+; X86-NEXT:    testl %esi, %esi
+; X86-NEXT:    sets %cl
+; X86-NEXT:    testl %ebx, %ebx
+; X86-NEXT:    sets %dl
+; X86-NEXT:    xorb %cl, %dl
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
+; X86-NEXT:    orl (%esp), %ecx # 4-byte Folded Reload
 ; X86-NEXT:    setne %cl
-; X86-NEXT:    testb %bl, %cl
+; X86-NEXT:    testb %dl, %cl
 ; X86-NEXT:    leal -1(%eax), %ecx
 ; X86-NEXT:    cmovel %eax, %ecx
-; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    movl %ecx, (%esp) # 4-byte Spill
+; X86-NEXT:    testl %edi, %edi
 ; X86-NEXT:    sets %al
-; X86-NEXT:    xorb {{[-0-9]+}}(%e{{[sb]}}p), %al # 1-byte Folded Reload
-; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
-; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Folded Reload
-; X86-NEXT:    setne %dl
-; X86-NEXT:    testb %al, %dl
-; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
-; X86-NEXT:    leal -1(%eax), %edi
-; X86-NEXT:    cmovel %eax, %edi
 ; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
-; X86-NEXT:    sets %dl
-; X86-NEXT:    xorb {{[-0-9]+}}(%e{{[sb]}}p), %dl # 1-byte Folded Reload
+; X86-NEXT:    sets %cl
+; X86-NEXT:    xorb %al, %cl
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
-; X86-NEXT:    setne %dh
-; X86-NEXT:    testb %dl, %dh
+; X86-NEXT:    setne %al
+; X86-NEXT:    testb %cl, %al
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
-; X86-NEXT:    leal -1(%eax), %edx
-; X86-NEXT:    cmovel %eax, %edx
+; X86-NEXT:    leal -1(%eax), %ecx
+; X86-NEXT:    cmovel %eax, %ecx
+; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
-; X86-NEXT:    sets %bl
-; X86-NEXT:    xorb {{[-0-9]+}}(%e{{[sb]}}p), %bl # 1-byte Folded Reload
+; X86-NEXT:    sets %al
+; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    sets %cl
+; X86-NEXT:    xorb %al, %cl
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
-; X86-NEXT:    setne %bh
-; X86-NEXT:    testb %bl, %bh
+; X86-NEXT:    setne %al
+; X86-NEXT:    testb %cl, %al
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
-; X86-NEXT:    leal -1(%eax), %esi
-; X86-NEXT:    cmovel %eax, %esi
+; X86-NEXT:    leal -1(%eax), %ebp
+; X86-NEXT:    cmovel %eax, %ebp
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
+; X86-NEXT:    testl %edx, %edx
+; X86-NEXT:    sets %al
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
+; X86-NEXT:    testl %ecx, %ecx
+; X86-NEXT:    sets %bl
+; X86-NEXT:    xorb %al, %bl
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edi # 4-byte Reload
+; X86-NEXT:    leal -1(%edi), %esi
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl %edx
+; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    calll __moddi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    orl %eax, %edx
+; X86-NEXT:    setne %al
+; X86-NEXT:    testb %bl, %al
+; X86-NEXT:    cmovel %edi, %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %esi, 12(%eax)
-; X86-NEXT:    movl %edx, 8(%eax)
-; X86-NEXT:    movl %edi, 4(%eax)
+; X86-NEXT:    movl %ebp, 8(%eax)
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
+; X86-NEXT:    movl %ecx, 4(%eax)
+; X86-NEXT:    movl (%esp), %ecx # 4-byte Reload
 ; X86-NEXT:    movl %ecx, (%eax)
-; X86-NEXT:    addl $64, %esp
+; X86-NEXT:    addl $60, %esp
 ; X86-NEXT:    popl %esi
 ; X86-NEXT:    popl %edi
 ; X86-NEXT:    popl %ebx
 ; X86-NEXT:    popl %ebp
 ; X86-NEXT:    retl $4
   %tmp = call <4 x i32> @llvm.sdiv.fix.v4i32(<4 x i32> %x, <4 x i32> %y, i32 31)
   ret <4 x i32> %tmp
 }

llvm/test/CodeGen/X86/sdiv_fix_sat.ll

 ; X64-LABEL: func5:
 ; X64:       # %bb.0:
 ; X64-NEXT:    pushq %rbp
 ; X64-NEXT:    pushq %r15
 ; X64-NEXT:    pushq %r14
 ; X64-NEXT:    pushq %r13
 ; X64-NEXT:    pushq %r12
 ; X64-NEXT:    pushq %rbx
-; X64-NEXT:    subq $40, %rsp
+; X64-NEXT:    subq $24, %rsp
+; X64-NEXT:    movq %rsi, %rdx
+; X64-NEXT:    movq %rsi, (%rsp) # 8-byte Spill
 ; X64-NEXT:    movq %rdi, %r15
 ; X64-NEXT:    leaq (%rdi,%rdi), %rax
-; X64-NEXT:    shrq $33, %rax
 ; X64-NEXT:    movq %rdi, %r12
 ; X64-NEXT:    sarq $63, %r12
-; X64-NEXT:    shlq $31, %r12
-; X64-NEXT:    orq %rax, %r12
-; X64-NEXT:    sets {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Folded Spill
+; X64-NEXT:    shldq $31, %rax, %r12
 ; X64-NEXT:    shlq $32, %r15
-; X64-NEXT:    movq %rsi, %rdx
-; X64-NEXT:    movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    movq %rsi, %r13
 ; X64-NEXT:    sarq $63, %r13
 ; X64-NEXT:    movq %r15, %rdi
 ; X64-NEXT:    movq %r12, %rsi
 ; X64-NEXT:    movq %r13, %rcx
 ; X64-NEXT:    callq __divti3
 ; X64-NEXT:    movq %rax, %rbx
 ; X64-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    movq %rdx, %rbp
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    subq $1, %rbx
 ; X64-NEXT:    sbbq $0, %rbp
+; X64-NEXT:    testq %r12, %r12
+; X64-NEXT:    sets %al
 ; X64-NEXT:    testq %r13, %r13
 ; X64-NEXT:    sets %r14b
-; X64-NEXT:    xorb {{[-0-9]+}}(%r{{[sb]}}p), %r14b # 1-byte Folded Reload
+; X64-NEXT:    xorb %al, %r14b
 ; X64-NEXT:    movq %r15, %rdi
 ; X64-NEXT:    movq %r12, %rsi
-; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; X64-NEXT:    movq (%rsp), %rdx # 8-byte Reload
 ; X64-NEXT:    movq %r13, %rcx
 ; X64-NEXT:    callq __modti3
 ; X64-NEXT:    orq %rax, %rdx
 ; X64-NEXT:    setne %al
 ; X64-NEXT:    testb %r14b, %al
 ; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Folded Reload
 ; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Folded Reload
 ; X64-NEXT:    cmpq $-1, %rbx

 ; X64-NEXT:    cmovaq %rbx, %rcx
 ; X64-NEXT:    testq %rbp, %rbp
 ; X64-NEXT:    cmovnsq %rbp, %rax
 ; X64-NEXT:    cmovsq %rdx, %rbx
 ; X64-NEXT:    cmpq $-1, %rbp
 ; X64-NEXT:    cmoveq %rcx, %rbx
 ; X64-NEXT:    shrdq $1, %rax, %rbx
 ; X64-NEXT:    movq %rbx, %rax
-; X64-NEXT:    addq $40, %rsp
+; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    popq %rbx
 ; X64-NEXT:    popq %r12
 ; X64-NEXT:    popq %r13
 ; X64-NEXT:    popq %r14
 ; X64-NEXT:    popq %r15
 ; X64-NEXT:    popq %rbp
 ; X64-NEXT:    retq
 ;

 ; X64-NEXT:    subq $104, %rsp
 ; X64-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; X64-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; X64-NEXT:    pxor %xmm2, %xmm2
 ; X64-NEXT:    pcmpgtd %xmm0, %xmm2
 ; X64-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; X64-NEXT:    paddq %xmm0, %xmm0
 ; X64-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; X64-NEXT:    movq %xmm0, %rbp
-; X64-NEXT:    movq %rbp, %r12
-; X64-NEXT:    shrq $33, %r12
-; X64-NEXT:    movq %rbp, %r14
-; X64-NEXT:    sarq $63, %r14
-; X64-NEXT:    shlq $31, %r14
-; X64-NEXT:    orq %r14, %r12
+; X64-NEXT:    movq %xmm0, %rbx
+; X64-NEXT:    movq %rbx, %r15
+; X64-NEXT:    sarq $63, %r15
+; X64-NEXT:    movq %r15, %r12
+; X64-NEXT:    shldq $31, %rbx, %r12
 ; X64-NEXT:    pxor %xmm0, %xmm0
 ; X64-NEXT:    pcmpgtd %xmm1, %xmm0
 ; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; X64-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; X64-NEXT:    movq %xmm1, %rdx
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; X64-NEXT:    movq %rdx, %rbx
-; X64-NEXT:    sarq $63, %rbx
-; X64-NEXT:    shlq $31, %rbp
-; X64-NEXT:    movq %rbp, %rdi
+; X64-NEXT:    movq %rdx, %r14
+; X64-NEXT:    sarq $63, %r14
+; X64-NEXT:    shlq $31, %rbx
+; X64-NEXT:    movq %rbx, %rdi
 ; X64-NEXT:    movq %r12, %rsi
-; X64-NEXT:    movq %rbx, %rcx
+; X64-NEXT:    movq %r14, %rcx
 ; X64-NEXT:    callq __divti3
 ; X64-NEXT:    movq %rax, %r13
 ; X64-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; X64-NEXT:    movq %rdx, %r15
+; X64-NEXT:    movq %rdx, %rbp
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    subq $1, %r13
-; X64-NEXT:    sbbq $0, %r15
-; X64-NEXT:    shrq $63, %r14
-; X64-NEXT:    xorl %ebx, %r14d
-; X64-NEXT:    movq %rbp, %rdi
+; X64-NEXT:    sbbq $0, %rbp
+; X64-NEXT:    movq %rbx, %rdi
 ; X64-NEXT:    movq %r12, %rsi
 ; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
-; X64-NEXT:    movq %rbx, %rcx
+; X64-NEXT:    movq %r14, %rcx
 ; X64-NEXT:    callq __modti3
 ; X64-NEXT:    orq %rax, %rdx
 ; X64-NEXT:    setne %al
-; X64-NEXT:    testb %r14b, %al
-; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r15 # 8-byte Folded Reload
+; X64-NEXT:    shlq $31, %r15
+; X64-NEXT:    shrq $63, %r15
+; X64-NEXT:    xorl %r14d, %r15d
+; X64-NEXT:    testb %r15b, %al
+; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %rbp # 8-byte Folded Reload
 ; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Folded Reload
 ; X64-NEXT:    movl $4294967295, %edx # imm = 0xFFFFFFFF
 ; X64-NEXT:    cmpq %rdx, %r13
 ; X64-NEXT:    movl $4294967295, %eax # imm = 0xFFFFFFFF
 ; X64-NEXT:    cmovbq %r13, %rax
 ; X64-NEXT:    xorl %ecx, %ecx
-; X64-NEXT:    testq %r15, %r15
+; X64-NEXT:    testq %rbp, %rbp
 ; X64-NEXT:    cmovnsq %rdx, %r13
 ; X64-NEXT:    cmoveq %rax, %r13
-; X64-NEXT:    cmovnsq %rcx, %r15
+; X64-NEXT:    cmovnsq %rcx, %rbp
 ; X64-NEXT:    movabsq $-4294967296, %rcx # imm = 0xFFFFFFFF00000000
 ; X64-NEXT:    cmpq %rcx, %r13
 ; X64-NEXT:    movq %rcx, %rax
 ; X64-NEXT:    cmovaq %r13, %rax
-; X64-NEXT:    testq %r15, %r15
+; X64-NEXT:    testq %rbp, %rbp
 ; X64-NEXT:    cmovsq %rcx, %r13
-; X64-NEXT:    cmpq $-1, %r15
+; X64-NEXT:    cmpq $-1, %rbp
 ; X64-NEXT:    cmoveq %rax, %r13
 ; X64-NEXT:    movq %r13, %xmm0
 ; X64-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; X64-NEXT:    pshufd $78, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    # xmm0 = mem[2,3,0,1]
-; X64-NEXT:    movq %xmm0, %r13
-; X64-NEXT:    movq %r13, %rbx
-; X64-NEXT:    shrq $33, %rbx
-; X64-NEXT:    movq %r13, %r14
-; X64-NEXT:    sarq $63, %r14
-; X64-NEXT:    shlq $31, %r14
-; X64-NEXT:    orq %r14, %rbx
+; X64-NEXT:    movq %xmm0, %rbp
+; X64-NEXT:    movq %rbp, %rbx
+; X64-NEXT:    sarq $63, %rbx
+; X64-NEXT:    movq %rbx, %r13
+; X64-NEXT:    shldq $31, %rbp, %r13
 ; X64-NEXT:    pshufd $78, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    # xmm0 = mem[2,3,0,1]
 ; X64-NEXT:    movq %xmm0, %rdx
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; X64-NEXT:    movq %rdx, %rbp
-; X64-NEXT:    sarq $63, %rbp
-; X64-NEXT:    shlq $31, %r13
-; X64-NEXT:    movq %r13, %rdi
-; X64-NEXT:    movq %rbx, %rsi
-; X64-NEXT:    movq %rbp, %rcx
+; X64-NEXT:    movq %rdx, %r14
+; X64-NEXT:    sarq $63, %r14
+; X64-NEXT:    shlq $31, %rbp
+; X64-NEXT:    movq %rbp, %rdi
+; X64-NEXT:    movq %r13, %rsi
+; X64-NEXT:    movq %r14, %rcx
 ; X64-NEXT:    callq __divti3
 ; X64-NEXT:    movq %rax, %r12
 ; X64-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    movq %rdx, %r15
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    subq $1, %r12
 ; X64-NEXT:    sbbq $0, %r15
-; X64-NEXT:    shrq $63, %r14
-; X64-NEXT:    xorl %ebp, %r14d
-; X64-NEXT:    movq %r13, %rdi
-; X64-NEXT:    movq %rbx, %rsi
+; X64-NEXT:    movq %rbp, %rdi
+; X64-NEXT:    movq %r13, %rsi
 ; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
-; X64-NEXT:    movq %rbp, %rcx
+; X64-NEXT:    movq %r14, %rcx
 ; X64-NEXT:    callq __modti3
 ; X64-NEXT:    orq %rax, %rdx
 ; X64-NEXT:    setne %al
-; X64-NEXT:    testb %r14b, %al
+; X64-NEXT:    shlq $31, %rbx
+; X64-NEXT:    shrq $63, %rbx
+; X64-NEXT:    xorl %r14d, %ebx
+; X64-NEXT:    testb %bl, %al
 ; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r15 # 8-byte Folded Reload
 ; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r12 # 8-byte Folded Reload
 ; X64-NEXT:    movl $4294967295, %ecx # imm = 0xFFFFFFFF
 ; X64-NEXT:    cmpq %rcx, %r12
 ; X64-NEXT:    movl $4294967295, %eax # imm = 0xFFFFFFFF
 ; X64-NEXT:    cmovbq %r12, %rax
 ; X64-NEXT:    testq %r15, %r15
 ; X64-NEXT:    cmovnsq %rcx, %r12

 ; X64-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; X64-NEXT:    pshufd $78, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; X64-NEXT:    # xmm1 = mem[2,3,0,1]
 ; X64-NEXT:    pxor %xmm0, %xmm0
 ; X64-NEXT:    pcmpgtd %xmm1, %xmm0
 ; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; X64-NEXT:    paddq %xmm1, %xmm1
 ; X64-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; X64-NEXT:    movq %xmm1, %r12
-; X64-NEXT:    movq %r12, %rbx
-; X64-NEXT:    shrq $33, %rbx
-; X64-NEXT:    movq %r12, %r14
+; X64-NEXT:    movq %xmm1, %rbp
+; X64-NEXT:    movq %rbp, %r14
 ; X64-NEXT:    sarq $63, %r14
-; X64-NEXT:    shlq $31, %r14
-; X64-NEXT:    orq %r14, %rbx
+; X64-NEXT:    movq %r14, %r13
+; X64-NEXT:    shldq $31, %rbp, %r13
 ; X64-NEXT:    pshufd $78, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; X64-NEXT:    # xmm1 = mem[2,3,0,1]
 ; X64-NEXT:    pxor %xmm0, %xmm0
 ; X64-NEXT:    pcmpgtd %xmm1, %xmm0
 ; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; X64-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; X64-NEXT:    movq %xmm1, %rdx
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; X64-NEXT:    movq %rdx, %rbp
-; X64-NEXT:    sarq $63, %rbp
-; X64-NEXT:    shlq $31, %r12
-; X64-NEXT:    movq %r12, %rdi
-; X64-NEXT:    movq %rbx, %rsi
-; X64-NEXT:    movq %rbp, %rcx
+; X64-NEXT:    movq %rdx, %rbx
+; X64-NEXT:    sarq $63, %rbx
+; X64-NEXT:    shlq $31, %rbp
+; X64-NEXT:    movq %rbp, %rdi
+; X64-NEXT:    movq %r13, %rsi
+; X64-NEXT:    movq %rbx, %rcx
 ; X64-NEXT:    callq __divti3
-; X64-NEXT:    movq %rax, %r13
+; X64-NEXT:    movq %rax, %r12
 ; X64-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    movq %rdx, %r15
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; X64-NEXT:    subq $1, %r13
+; X64-NEXT:    subq $1, %r12
 ; X64-NEXT:    sbbq $0, %r15
-; X64-NEXT:    shrq $63, %r14
-; X64-NEXT:    xorl %ebp, %r14d
-; X64-NEXT:    movq %r12, %rdi
-; X64-NEXT:    movq %rbx, %rsi
+; X64-NEXT:    movq %rbp, %rdi
+; X64-NEXT:    movq %r13, %rsi
 ; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
-; X64-NEXT:    movq %rbp, %rcx
+; X64-NEXT:    movq %rbx, %rcx
 ; X64-NEXT:    callq __modti3
 ; X64-NEXT:    orq %rax, %rdx
 ; X64-NEXT:    setne %al
+; X64-NEXT:    shlq $31, %r14
+; X64-NEXT:    shrq $63, %r14
+; X64-NEXT:    xorl %ebx, %r14d
 ; X64-NEXT:    testb %r14b, %al
 ; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r15 # 8-byte Folded Reload
-; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Folded Reload
+; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r12 # 8-byte Folded Reload
 ; X64-NEXT:    movl $4294967295, %ecx # imm = 0xFFFFFFFF
-; X64-NEXT:    cmpq %rcx, %r13
+; X64-NEXT:    cmpq %rcx, %r12
 ; X64-NEXT:    movl $4294967295, %eax # imm = 0xFFFFFFFF
-; X64-NEXT:    cmovbq %r13, %rax
+; X64-NEXT:    cmovbq %r12, %rax
 ; X64-NEXT:    testq %r15, %r15
-; X64-NEXT:    cmovnsq %rcx, %r13
-; X64-NEXT:    cmoveq %rax, %r13
+; X64-NEXT:    cmovnsq %rcx, %r12
+; X64-NEXT:    cmoveq %rax, %r12
 ; X64-NEXT:    movl $0, %eax
 ; X64-NEXT:    cmovnsq %rax, %r15
 ; X64-NEXT:    movabsq $-4294967296, %rcx # imm = 0xFFFFFFFF00000000
-; X64-NEXT:    cmpq %rcx, %r13
+; X64-NEXT:    cmpq %rcx, %r12
 ; X64-NEXT:    movq %rcx, %rax
-; X64-NEXT:    cmovaq %r13, %rax
+; X64-NEXT:    cmovaq %r12, %rax
 ; X64-NEXT:    testq %r15, %r15
-; X64-NEXT:    cmovsq %rcx, %r13
+; X64-NEXT:    cmovsq %rcx, %r12
 ; X64-NEXT:    cmpq $-1, %r15
-; X64-NEXT:    cmoveq %rax, %r13
-; X64-NEXT:    movq %r13, %xmm0
+; X64-NEXT:    cmoveq %rax, %r12
+; X64-NEXT:    movq %r12, %xmm0
 ; X64-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; X64-NEXT:    pshufd $78, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    # xmm0 = mem[2,3,0,1]
-; X64-NEXT:    movq %xmm0, %r13
-; X64-NEXT:    movq %r13, %rbx
-; X64-NEXT:    shrq $33, %rbx
-; X64-NEXT:    movq %r13, %r14
-; X64-NEXT:    sarq $63, %r14
-; X64-NEXT:    shlq $31, %r14
-; X64-NEXT:    orq %r14, %rbx
+; X64-NEXT:    movq %xmm0, %rbp
+; X64-NEXT:    movq %rbp, %rbx
+; X64-NEXT:    sarq $63, %rbx
+; X64-NEXT:    movq %rbx, %r13
+; X64-NEXT:    shldq $31, %rbp, %r13
 ; X64-NEXT:    pshufd $78, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    # xmm0 = mem[2,3,0,1]
 ; X64-NEXT:    movq %xmm0, %rdx
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
-; X64-NEXT:    movq %rdx, %rbp
-; X64-NEXT:    sarq $63, %rbp
-; X64-NEXT:    shlq $31, %r13
-; X64-NEXT:    movq %r13, %rdi
-; X64-NEXT:    movq %rbx, %rsi
-; X64-NEXT:    movq %rbp, %rcx
+; X64-NEXT:    movq %rdx, %r14
+; X64-NEXT:    sarq $63, %r14
+; X64-NEXT:    shlq $31, %rbp
+; X64-NEXT:    movq %rbp, %rdi
+; X64-NEXT:    movq %r13, %rsi
+; X64-NEXT:    movq %r14, %rcx
 ; X64-NEXT:    callq __divti3
 ; X64-NEXT:    movq %rax, %r12
 ; X64-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    movq %rdx, %r15
 ; X64-NEXT:    movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; X64-NEXT:    subq $1, %r12
 ; X64-NEXT:    sbbq $0, %r15
-; X64-NEXT:    shrq $63, %r14
-; X64-NEXT:    xorl %ebp, %r14d
-; X64-NEXT:    movq %r13, %rdi
-; X64-NEXT:    movq %rbx, %rsi
+; X64-NEXT:    movq %rbp, %rdi
+; X64-NEXT:    movq %r13, %rsi
 ; X64-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
-; X64-NEXT:    movq %rbp, %rcx
+; X64-NEXT:    movq %r14, %rcx
 ; X64-NEXT:    callq __modti3
 ; X64-NEXT:    orq %rax, %rdx
 ; X64-NEXT:    setne %al
-; X64-NEXT:    testb %r14b, %al
+; X64-NEXT:    shlq $31, %rbx
+; X64-NEXT:    shrq $63, %rbx
+; X64-NEXT:    xorl %r14d, %ebx
+; X64-NEXT:    testb %bl, %al
 ; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r15 # 8-byte Folded Reload
 ; X64-NEXT:    cmoveq {{[-0-9]+}}(%r{{[sb]}}p), %r12 # 8-byte Folded Reload
 ; X64-NEXT:    movl $4294967295, %ecx # imm = 0xFFFFFFFF
 ; X64-NEXT:    cmpq %rcx, %r12
 ; X64-NEXT:    movl $4294967295, %eax # imm = 0xFFFFFFFF
 ; X64-NEXT:    cmovbq %r12, %rax
 ; X64-NEXT:    testq %r15, %r15
 ; X64-NEXT:    cmovnsq %rcx, %r12

 ; X86:       # %bb.0:
 ; X86-NEXT:    pushl %ebp
 ; X86-NEXT:    movl %esp, %ebp
 ; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %edi
 ; X86-NEXT:    pushl %esi
 ; X86-NEXT:    andl $-8, %esp
 ; X86-NEXT:    subl $256, %esp # imm = 0x100
-; X86-NEXT:    movl 24(%ebp), %ecx
-; X86-NEXT:    movl 40(%ebp), %ebx
-; X86-NEXT:    movl %ebx, %edx
-; X86-NEXT:    sarl $31, %edx
-; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    movl 24(%ebp), %edx
+; X86-NEXT:    movl 40(%ebp), %edi
+; X86-NEXT:    leal {{[0-9]+}}(%esp), %ebx
+; X86-NEXT:    movl %edi, %eax
 ; X86-NEXT:    sarl $31, %eax
-; X86-NEXT:    addl %ecx, %ecx
-; X86-NEXT:    adcl %eax, %eax
-; X86-NEXT:    movl %ecx, %esi
-; X86-NEXT:    andl $1, %eax
-; X86-NEXT:    movl %eax, %edi
-; X86-NEXT:    shll $31, %eax
-; X86-NEXT:    shrl %ecx
-; X86-NEXT:    subl %eax, %ecx
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %edx, %ecx
+; X86-NEXT:    sarl $31, %ecx
+; X86-NEXT:    addl %edx, %edx
+; X86-NEXT:    adcl %ecx, %ecx
+; X86-NEXT:    andl $1, %ecx
+; X86-NEXT:    negl %ecx
 ; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    shll $31, %esi
+; X86-NEXT:    movl %ecx, %esi
+; X86-NEXT:    shldl $31, %edx, %esi
 ; X86-NEXT:    movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    negl %edi
+; X86-NEXT:    shll $31, %edx
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    pushl %edx
-; X86-NEXT:    pushl %edx
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %esi
 ; X86-NEXT:    pushl %edx
 ; X86-NEXT:    pushl %ebx
+; X86-NEXT:    calll __modti3
+; X86-NEXT:    addl $32, %esp
+; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl 36(%ebp), %edx
+; X86-NEXT:    movl %edx, %edi
+; X86-NEXT:    sarl $31, %edi
+; X86-NEXT:    movl 20(%ebp), %ecx
+; X86-NEXT:    movl %ecx, %esi
+; X86-NEXT:    sarl $31, %esi
+; X86-NEXT:    addl %ecx, %ecx
+; X86-NEXT:    adcl %esi, %esi
+; X86-NEXT:    andl $1, %esi
+; X86-NEXT:    negl %esi
+; X86-NEXT:    movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %esi, %ebx
+; X86-NEXT:    shldl $31, %ecx, %ebx
+; X86-NEXT:    movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    shll $31, %ecx
+; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl %edi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    pushl %edi
 ; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %edx
 ; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl %eax
 ; X86-NEXT:    calll __modti3
 ; X86-NEXT:    addl $32, %esp
-; X86-NEXT:    movl 36(%ebp), %edi
-; X86-NEXT:    movl %edi, %edx
+; X86-NEXT:    leal {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    movl 28(%ebp), %ebx
+; X86-NEXT:    movl %ebx, %edx
 ; X86-NEXT:    sarl $31, %edx
-; X86-NEXT:    movl 20(%ebp), %ecx
-; X86-NEXT:    movl %ecx, %eax
-; X86-NEXT:    sarl $31, %eax
-; X86-NEXT:    addl %ecx, %ecx
-; X86-NEXT:    adcl %eax, %eax
-; X86-NEXT:    movl %ecx, %esi
-; X86-NEXT:    andl $1, %eax
-; X86-NEXT:    movl %eax, %ebx
-; X86-NEXT:    shll $31, %eax
-; X86-NEXT:    shrl %ecx
-; X86-NEXT:    subl %eax, %ecx
+; X86-NEXT:    movl 12(%ebp), %esi
+; X86-NEXT:    movl %esi, %ecx
+; X86-NEXT:    sarl $31, %ecx
+; X86-NEXT:    addl %esi, %esi
+; X86-NEXT:    adcl %ecx, %ecx
+; X86-NEXT:    andl $1, %ecx
+; X86-NEXT:    negl %ecx
 ; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    shldl $31, %esi, %eax
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    shll $31, %esi
 ; X86-NEXT:    movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    negl %ebx
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    pushl %edx
 ; X86-NEXT:    pushl %edx
 ; X86-NEXT:    pushl %edx
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %ecx
-; X86-NEXT:    pushl %esi
-; X86-NEXT:    pushl %eax
-; X86-NEXT:    calll __modti3
-; X86-NEXT:    addl $32, %esp
-; X86-NEXT:    movl 28(%ebp), %edi
-; X86-NEXT:    movl %edi, %ebx
-; X86-NEXT:    sarl $31, %ebx
-; X86-NEXT:    movl 12(%ebp), %ecx
-; X86-NEXT:    movl %ecx, %eax
-; X86-NEXT:    sarl $31, %eax
-; X86-NEXT:    addl %ecx, %ecx
-; X86-NEXT:    adcl %eax, %eax
-; X86-NEXT:    movl %ecx, %edx
-; X86-NEXT:    andl $1, %eax
-; X86-NEXT:    movl %eax, %esi
-; X86-NEXT:    shll $31, %eax
-; X86-NEXT:    shrl %ecx
-; X86-NEXT:    subl %eax, %ecx
-; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    shll $31, %edx
-; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    negl %esi
-; X86-NEXT:    movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    pushl %ebx
-; X86-NEXT:    pushl %ebx
-; X86-NEXT:    pushl %ebx
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    pushl %esi
-; X86-NEXT:    pushl %esi
 ; X86-NEXT:    pushl %ecx
-; X86-NEXT:    pushl %edx
 ; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %edi
 ; X86-NEXT:    calll __divti3
 ; X86-NEXT:    addl $32, %esp
+; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl 32(%ebp), %edx
-; X86-NEXT:    movl %edx, %edi
-; X86-NEXT:    sarl $31, %edi
+; X86-NEXT:    movl %edx, %esi
+; X86-NEXT:    sarl $31, %esi
 ; X86-NEXT:    movl 16(%ebp), %ecx
-; X86-NEXT:    movl %ecx, %eax
-; X86-NEXT:    sarl $31, %eax
-; X86-NEXT:    addl %ecx, %ecx
-; X86-NEXT:    adcl %eax, %eax
 ; X86-NEXT:    movl %ecx, %ebx
-; X86-NEXT:    andl $1, %eax
-; X86-NEXT:    movl %eax, %esi
-; X86-NEXT:    shll $31, %eax
-; X86-NEXT:    shrl %ecx
-; X86-NEXT:    subl %eax, %ecx
+; X86-NEXT:    sarl $31, %ebx
+; X86-NEXT:    addl %ecx, %ecx
+; X86-NEXT:    adcl %ebx, %ebx
+; X86-NEXT:    andl $1, %ebx
+; X86-NEXT:    negl %ebx
+; X86-NEXT:    movl %ebx, %edi
+; X86-NEXT:    shldl $31, %ecx, %edi
+; X86-NEXT:    shll $31, %ecx
 ; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    shll $31, %ebx
-; X86-NEXT:    negl %esi
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl %edx
 ; X86-NEXT:    pushl %esi
 ; X86-NEXT:    pushl %esi
-; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %edx
+; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl %eax
 ; X86-NEXT:    calll __modti3
 ; X86-NEXT:    addl $32, %esp
 ; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl 32(%ebp)
 ; X86-NEXT:    movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    pushl %esi
 ; X86-NEXT:    pushl %esi
-; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl 32(%ebp)
 ; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
 ; X86-NEXT:    pushl %eax
 ; X86-NEXT:    calll __divti3
 ; X86-NEXT:    addl $32, %esp
 ; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
 ; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl %ecx

 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
 ; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl 36(%ebp)
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
 ; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl %ecx
-; X86-NEXT:    movl %ecx, %ebx
+; X86-NEXT:    movl %ecx, %edi
 ; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
 ; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
 ; X86-NEXT:    pushl %eax
 ; X86-NEXT:    calll __divti3
 ; X86-NEXT:    addl $32, %esp
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    subl $1, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; X86-NEXT:    movl %esi, %ecx
 ; X86-NEXT:    sbbl $0, %ecx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    sbbl $0, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    sbbl $0, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    testl %ebx, %ebx
-; X86-NEXT:    sets %bl
+; X86-NEXT:    testl %edi, %edi
+; X86-NEXT:    sets %al
 ; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
-; X86-NEXT:    sets %bh
-; X86-NEXT:    xorb %bl, %bh
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    orl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    sets %ah
+; X86-NEXT:    xorb %al, %ah
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    orl {{[0-9]+}}(%esp), %edi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    orl {{[0-9]+}}(%esp), %edx
-; X86-NEXT:    orl %eax, %edx
+; X86-NEXT:    orl %edi, %edx
 ; X86-NEXT:    setne %al
-; X86-NEXT:    testb %bh, %al
+; X86-NEXT:    testb %ah, %al
 ; X86-NEXT:    cmovel %esi, %ecx
 ; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    subl $1, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    movl %edx, %ecx
 ; X86-NEXT:    sbbl $0, %ecx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    sbbl $0, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X86-NEXT:    movl %esi, %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    movl %edi, %eax
 ; X86-NEXT:    sbbl $0, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    testl %ebx, %ebx
 ; X86-NEXT:    sets %bl
-; X86-NEXT:    testl %edi, %edi
+; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
 ; X86-NEXT:    sets %bh
 ; X86-NEXT:    xorb %bl, %bh
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
-; X86-NEXT:    orl {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    orl {{[0-9]+}}(%esp), %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    orl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    orl %edi, %eax
+; X86-NEXT:    orl %esi, %eax
 ; X86-NEXT:    setne %al
 ; X86-NEXT:    testb %bh, %al
 ; X86-NEXT:    cmovel %edx, %ecx
 ; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
-; X86-NEXT:    cmovel %esi, %eax
+; X86-NEXT:    cmovel %edi, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    subl $1, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X86-NEXT:    movl %esi, %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    movl %edi, %eax
 ; X86-NEXT:    sbbl $0, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    sbbl $0, %eax
-; X86-NEXT:    movl %eax, %edi
+; X86-NEXT:    movl %eax, %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    sbbl $0, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
-; X86-NEXT:    testl %edx, %edx
-; X86-NEXT:    sets %al
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
 ; X86-NEXT:    testl %ecx, %ecx
+; X86-NEXT:    sets %al
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
+; X86-NEXT:    testl %edx, %edx
 ; X86-NEXT:    sets %bl
 ; X86-NEXT:    xorb %al, %bl
 ; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    pushl %ecx
-; X86-NEXT:    pushl %ecx
-; X86-NEXT:    pushl %ecx
-; X86-NEXT:    pushl 28(%ebp)
 ; X86-NEXT:    pushl %edx
 ; X86-NEXT:    pushl %edx
+; X86-NEXT:    pushl %edx
+; X86-NEXT:    pushl 28(%ebp)
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %ecx
 ; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
 ; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
 ; X86-NEXT:    pushl %eax
 ; X86-NEXT:    calll __modti3
 ; X86-NEXT:    addl $32, %esp
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    orl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    orl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    orl %eax, %ecx
 ; X86-NEXT:    setne %al
 ; X86-NEXT:    testb %bl, %al
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
-; X86-NEXT:    cmovel %esi, %eax
+; X86-NEXT:    cmovel %edi, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %edi # 4-byte Folded Reload
-; X86-NEXT:    movl %edi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Folded Reload
+; X86-NEXT:    movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    subl $1, %edx
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
-; X86-NEXT:    movl %edi, %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl %esi, %eax
 ; X86-NEXT:    sbbl $0, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    sbbl $0, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    sbbl $0, %ecx
 ; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
 ; X86-NEXT:    sets %bl
 ; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
 ; X86-NEXT:    sets %bh
 ; X86-NEXT:    xorb %bl, %bh
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    orl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X86-NEXT:    orl {{[0-9]+}}(%esp), %esi
-; X86-NEXT:    orl %eax, %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    orl {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    orl %eax, %edi
 ; X86-NEXT:    setne %al
 ; X86-NEXT:    testb %bh, %al
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
-; X86-NEXT:    cmovel %edi, %eax
+; X86-NEXT:    cmovel %esi, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Folded Reload
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Folded Reload
 ; X86-NEXT:    testl %ecx, %ecx

 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl $0, %edx
 ; X86-NEXT:    cmovsl %eax, %edx
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl $-1, %eax
 ; X86-NEXT:    cmovsl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
-; X86-NEXT:    movl %eax, %esi
-; X86-NEXT:    sarl $31, %esi
-; X86-NEXT:    movl %esi, %edx
+; X86-NEXT:    movl %eax, %edi
+; X86-NEXT:    sarl $31, %edi
+; X86-NEXT:    movl %edi, %edx
 ; X86-NEXT:    andl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Folded Reload
 ; X86-NEXT:    testl %eax, %eax
 ; X86-NEXT:    cmovel %eax, %edx
 ; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
-; X86-NEXT:    movl $0, %edi
-; X86-NEXT:    cmovsl %eax, %edi
+; X86-NEXT:    movl $0, %esi
+; X86-NEXT:    cmovsl %eax, %esi
 ; X86-NEXT:    movl $-1, %eax
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
 ; X86-NEXT:    cmovsl %edx, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl %ecx, %eax
 ; X86-NEXT:    sarl $31, %eax
 ; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill

 ; X86-NEXT:    cmpl $1, %edx
 ; X86-NEXT:    movl $0, %eax
 ; X86-NEXT:    sbbl %eax, %eax
 ; X86-NEXT:    notl %eax
 ; X86-NEXT:    orl %ecx, %eax
 ; X86-NEXT:    testl %edx, %edx
 ; X86-NEXT:    movl $0, %ecx
 ; X86-NEXT:    cmovbl %edx, %ecx
-; X86-NEXT:    andl %edx, %esi
+; X86-NEXT:    andl %edx, %edi
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
 ; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Folded Reload
-; X86-NEXT:    cmovel %ecx, %esi
+; X86-NEXT:    cmovel %ecx, %edi
 ; X86-NEXT:    cmovnel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
 ; X86-NEXT:    testl %eax, %eax
 ; X86-NEXT:    movl $0, %ecx
 ; X86-NEXT:    cmoval %eax, %ecx
-; X86-NEXT:    cmpl $-1, %esi
+; X86-NEXT:    cmpl $-1, %edi
 ; X86-NEXT:    movl $0, %edx
 ; X86-NEXT:    cmovnel %edx, %ecx
-; X86-NEXT:    testl %edi, %edi
+; X86-NEXT:    testl %esi, %esi
 ; X86-NEXT:    movl $-1, %edx
-; X86-NEXT:    cmovsl %edx, %esi
+; X86-NEXT:    cmovsl %edx, %edi
 ; X86-NEXT:    movl $0, %edx
 ; X86-NEXT:    cmovsl %edx, %eax
-; X86-NEXT:    andl {{[-0-9]+}}(%e{{[sb]}}p), %edi # 4-byte Folded Reload
-; X86-NEXT:    cmpl $-1, %edi
+; X86-NEXT:    andl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Folded Reload
+; X86-NEXT:    cmpl $-1, %esi
 ; X86-NEXT:    cmovel %ecx, %eax
-; X86-NEXT:    cmovnel %esi, %edi
-; X86-NEXT:    shldl $31, %eax, %edi
-; X86-NEXT:    movl %edi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    cmovnel %edi, %esi
+; X86-NEXT:    shldl $31, %eax, %esi
+; X86-NEXT:    movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
 ; X86-NEXT:    cmpl $-1, %eax
 ; X86-NEXT:    movl $-1, %ecx
 ; X86-NEXT:    cmovael %ecx, %eax
 ; X86-NEXT:    movl %eax, %ecx
 ; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
 ; X86-NEXT:    cmpl $1, %esi
 ; X86-NEXT:    movl $0, %eax

llvm/test/CodeGen/X86/shift-combine.ll

; X64-NEXT:    retq
   %conv1 = ashr i64 %sext, 36
   ret i64 %conv1
 }
 
 define i64 @ashr_add_shl_mismatch_shifts1(i64 %r) nounwind {
 ; X32-LABEL: ashr_add_shl_mismatch_shifts1:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    incl %eax
 ; X32-NEXT:    movl %eax, %edx
 ; X32-NEXT:    sarl $31, %edx
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: ashr_add_shl_mismatch_shifts1:
 ; X64:       # %bb.0:
 ; X64-NEXT:    shlq $8, %rdi
 ; X64-NEXT:    movabsq $4294967296, %rax # imm = 0x100000000
 ; X64-NEXT:    addq %rdi, %rax
 ; X64-NEXT:    sarq $32, %rax
 ; X64-NEXT:    retq
   %conv = shl i64 %r, 8
   %sext = add i64 %conv, 4294967296
   %conv1 = ashr i64 %sext, 32
   ret i64 %conv1
 }
 
 define i64 @ashr_add_shl_mismatch_shifts2(i64 %r) nounwind {
 ; X32-LABEL: ashr_add_shl_mismatch_shifts2:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X32-NEXT:    shrdl $8, %edx, %eax
 ; X32-NEXT:    shrl $8, %edx
 ; X32-NEXT:    incl %edx
 ; X32-NEXT:    shrdl $8, %edx, %eax
 ; X32-NEXT:    shrl $8, %edx
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: ashr_add_shl_mismatch_shifts2:
 ; X64:       # %bb.0:

llvm/test/CodeGen/X86/udiv_fix_sat.ll

 }
 
 define i64 @func5(i64 %x, i64 %y) nounwind {
 ; X64-LABEL: func5:
 ; X64:       # %bb.0:
 ; X64-NEXT:    pushq %rbx
 ; X64-NEXT:    movq %rsi, %rdx
 ; X64-NEXT:    leaq (%rdi,%rdi), %rsi
-; X64-NEXT:    shrq $33, %rsi
 ; X64-NEXT:    movq %rdi, %rax
-; X64-NEXT:    shrq $32, %rax
-; X64-NEXT:    andl $-2147483648, %eax # imm = 0x80000000
-; X64-NEXT:    orq %rax, %rsi
+; X64-NEXT:    shrq $63, %rax
+; X64-NEXT:    shrdq $33, %rax, %rsi
 ; X64-NEXT:    shlq $32, %rdi
 ; X64-NEXT:    xorl %ebx, %ebx
 ; X64-NEXT:    xorl %ecx, %ecx
 ; X64-NEXT:    callq __udivti3
 ; X64-NEXT:    cmpq $-1, %rax
 ; X64-NEXT:    movq $-1, %rcx
 ; X64-NEXT:    cmovbq %rax, %rcx
 ; X64-NEXT:    cmpq $1, %rdx

llvm/test/CodeGen/X86/x86-64-double-precision-shift-left.ll

 ;uint64_t lshift7(uint64_t a, uint64_t b)
 ;{
 ;    return (a << 7) | (b >> 57);
 ;}
 
 define i64 @lshift7(i64 %a, i64 %b) nounwind readnone uwtable {
 ; CHECK-LABEL: lshift7:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    shlq $7, %rdi
 ; CHECK-NEXT:    shrq $57, %rsi
-; CHECK-NEXT:    leaq (%rsi,%rdi), %rax
+; CHECK-NEXT:    shlq $7, %rdi
+; CHECK-NEXT:    leaq (%rdi,%rsi), %rax
 ; CHECK-NEXT:    retq
 entry:
   %shl = shl i64 %a, 7
   %shr = lshr i64 %b, 57
   %or = or i64 %shr, %shl
   ret i64 %or
 }
 
 ;uint64_t lshift63(uint64_t a, uint64_t b)
 ;{
 ;    return (a << 63) | (b >> 1);
 ;}
 
 define i64 @lshift63(i64 %a, i64 %b) nounwind readnone uwtable {
 ; CHECK-LABEL: lshift63:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    shlq $63, %rdi
 ; CHECK-NEXT:    shrq %rsi
-; CHECK-NEXT:    leaq (%rsi,%rdi), %rax
+; CHECK-NEXT:    shlq $63, %rdi
+; CHECK-NEXT:    leaq (%rdi,%rsi), %rax
 ; CHECK-NEXT:    retq
 entry:
   %shl = shl i64 %a, 63
   %shr = lshr i64 %b, 1
   %or = or i64 %shr, %shl
   ret i64 %or
 }

llvm/test/CodeGen/X86/x86-64-double-shifts-Oz-Os-O2.ll

 ; {
 ;     return (a << 12) | (b >> 52);
 ; }
 
 ; Function Attrs: nounwind optsize readnone uwtable
 define i64 @_Z8lshift12mm(i64 %a, i64 %b) #2 {
 ; CHECK-LABEL: _Z8lshift12mm:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    shlq $12, %rdi
 ; CHECK-NEXT:    shrq $52, %rsi
-; CHECK-NEXT:    leaq (%rsi,%rdi), %rax
+; CHECK-NEXT:    shlq $12, %rdi
+; CHECK-NEXT:    leaq (%rdi,%rsi), %rax
 ; CHECK-NEXT:    retq
 entry:
   %shl = shl i64 %a, 12
   %shr = lshr i64 %b, 52
   %or = or i64 %shr, %shl
   ret i64 %or
 }