diff --git a/llvm/include/llvm/CodeGen/TargetInstrInfo.h b/llvm/include/llvm/CodeGen/TargetInstrInfo.h
--- a/llvm/include/llvm/CodeGen/TargetInstrInfo.h
+++ b/llvm/include/llvm/CodeGen/TargetInstrInfo.h
@@ -1044,6 +1044,16 @@
     return isCopyInstrImpl(MI);
   }
 
+  bool isFullCopyInstr(const MachineInstr &MI) const {
+    auto DestSrc = isCopyInstr(MI);
+    if (!DestSrc)
+      return false;
+
+    const MachineOperand *DestRegOp = DestSrc->Destination;
+    const MachineOperand *SrcRegOp = DestSrc->Source;
+    return !DestRegOp->getSubReg() && !SrcRegOp->getSubReg();
+  }
+
   /// If the specific machine instruction is an instruction that adds an
   /// immediate value and a physical register, and stores the result in
   /// the given physical register \c Reg, return a pair of the source
@@ -1954,6 +1964,13 @@
     return false;
   }
 
+  /// Allows targets to use appropriate copy instruction while spilitting live
+  /// range of a register in register allocation.
+  virtual unsigned getLiveRangeSplitOpcode(Register reg,
+                                           MachineRegisterInfo &MRI) const {
+    return TargetOpcode::COPY;
+  }
+
   /// During PHI eleimination lets target to make necessary checks and
   /// insert the copy to the PHI destination register in a target specific
   /// manner.
diff --git a/llvm/lib/CodeGen/CalcSpillWeights.cpp b/llvm/lib/CodeGen/CalcSpillWeights.cpp
--- a/llvm/lib/CodeGen/CalcSpillWeights.cpp
+++ b/llvm/lib/CodeGen/CalcSpillWeights.cpp
@@ -97,7 +97,7 @@
     // Trace copies introduced by live range splitting.  The inline
     // spiller can rematerialize through these copies, so the spill
     // weight must reflect this.
-    while (MI->isFullCopy()) {
+    while (TII.isFullCopyInstr(*MI)) {
       // The copy destination must match the interval register.
       if (MI->getOperand(0).getReg() != Reg)
         return false;
@@ -224,7 +224,16 @@
       continue;
 
     NumInstr++;
-    if (MI->isIdentityCopy() || MI->isImplicitDef())
+    bool identityCopy = false;
+    auto DestSrc = TII.isCopyInstr(*MI);
+    if (DestSrc) {
+      const MachineOperand *DestRegOp = DestSrc->Destination;
+      const MachineOperand *SrcRegOp = DestSrc->Source;
+      identityCopy = DestRegOp->getReg() == SrcRegOp->getReg() &&
+                     DestRegOp->getSubReg() == SrcRegOp->getSubReg();
+    }
+
+    if (identityCopy || MI->isImplicitDef())
       continue;
     if (!Visited.insert(MI).second)
       continue;
@@ -258,7 +267,7 @@
     }
 
     // Get allocation hints from copies.
-    if (!MI->isCopy())
+    if (!TII.isCopyInstr(*MI))
       continue;
     Register HintReg = copyHint(MI, LI.reg(), TRI, MRI);
     if (!HintReg)
diff --git a/llvm/lib/CodeGen/InlineSpiller.cpp b/llvm/lib/CodeGen/InlineSpiller.cpp
--- a/llvm/lib/CodeGen/InlineSpiller.cpp
+++ b/llvm/lib/CodeGen/InlineSpiller.cpp
@@ -257,8 +257,9 @@
 
 /// isFullCopyOf - If MI is a COPY to or from Reg, return the other register,
 /// otherwise return 0.
-static Register isFullCopyOf(const MachineInstr &MI, Register Reg) {
-  if (!MI.isFullCopy())
+static Register isFullCopyOf(const MachineInstr &MI, Register Reg,
+                             const TargetInstrInfo &TII) {
+  if (!TII.isFullCopyInstr(MI))
     return Register();
   if (MI.getOperand(0).getReg() == Reg)
     return MI.getOperand(1).getReg();
@@ -314,7 +315,7 @@
     MachineInstr &MI = *RI++;
 
     // Allow copies to/from Reg.
-    if (isFullCopyOf(MI, Reg))
+    if (isFullCopyOf(MI, Reg, TII))
       continue;
 
     // Allow stack slot loads.
@@ -353,7 +354,7 @@
 
   for (MachineInstr &MI :
        llvm::make_early_inc_range(MRI.reg_instructions(Reg))) {
-    Register SnipReg = isFullCopyOf(MI, Reg);
+    Register SnipReg = isFullCopyOf(MI, Reg, TII);
     if (!isSibling(SnipReg))
       continue;
     LiveInterval &SnipLI = LIS.getInterval(SnipReg);
@@ -476,14 +477,14 @@
     // Find all spills and copies of VNI.
     for (MachineInstr &MI :
          llvm::make_early_inc_range(MRI.use_nodbg_instructions(Reg))) {
-      if (!MI.isCopy() && !MI.mayStore())
+      if (!TII.isCopyInstr(MI) && !MI.mayStore())
         continue;
       SlotIndex Idx = LIS.getInstructionIndex(MI);
       if (LI->getVNInfoAt(Idx) != VNI)
         continue;
 
       // Follow sibling copies down the dominator tree.
-      if (Register DstReg = isFullCopyOf(MI, Reg)) {
+      if (Register DstReg = isFullCopyOf(MI, Reg, TII)) {
         if (isSibling(DstReg)) {
            LiveInterval &DstLI = LIS.getInterval(DstReg);
            VNInfo *DstVNI = DstLI.getVNInfoAt(Idx.getRegSlot());
@@ -826,7 +827,7 @@
   if (Ops.back().first != MI || MI->isBundled())
     return false;
 
-  bool WasCopy = MI->isCopy();
+  bool WasCopy = TII.isCopyInstr(*MI) ? true : false;
   Register ImpReg;
 
   // TII::foldMemoryOperand will do what we need here for statepoint
@@ -1111,7 +1112,7 @@
         Idx = VNI->def;
 
     // Check for a sibling copy.
-    Register SibReg = isFullCopyOf(MI, Reg);
+    Register SibReg = isFullCopyOf(MI, Reg, TII);
     if (SibReg && isSibling(SibReg)) {
       // This may actually be a copy between snippets.
       if (isRegToSpill(SibReg)) {
diff --git a/llvm/lib/CodeGen/LiveRangeEdit.cpp b/llvm/lib/CodeGen/LiveRangeEdit.cpp
--- a/llvm/lib/CodeGen/LiveRangeEdit.cpp
+++ b/llvm/lib/CodeGen/LiveRangeEdit.cpp
@@ -348,7 +348,8 @@
     // unlikely to change anything. We typically don't want to shrink the
     // PIC base register that has lots of uses everywhere.
     // Always shrink COPY uses that probably come from live range splitting.
-    if ((MI->readsVirtualRegister(Reg) && (MI->isCopy() || MO.isDef())) ||
+    if ((MI->readsVirtualRegister(Reg) &&
+         (TII.isCopyInstr(*MI) || MO.isDef())) ||
         (MO.readsReg() && (MRI.hasOneNonDBGUse(Reg) || useIsKill(LI, MO))))
       ToShrink.insert(&LI);
     else if (MO.readsReg())
diff --git a/llvm/lib/CodeGen/LiveRangeShrink.cpp b/llvm/lib/CodeGen/LiveRangeShrink.cpp
--- a/llvm/lib/CodeGen/LiveRangeShrink.cpp
+++ b/llvm/lib/CodeGen/LiveRangeShrink.cpp
@@ -23,6 +23,7 @@
 #include "llvm/CodeGen/MachineInstr.h"
 #include "llvm/CodeGen/MachineOperand.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGen/TargetInstrInfo.h"
 #include "llvm/InitializePasses.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/Debug.h"
@@ -197,7 +198,8 @@
           // is because it needs more accurate model to handle register
           // pressure correctly.
           MachineInstr &DefInstr = *MRI.def_instr_begin(Reg);
-          if (!DefInstr.isCopy())
+          const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
+          if (!TII.isCopyInstr(DefInstr))
             NumEligibleUse++;
           Insert = FindDominatedInstruction(DefInstr, Insert, IOM);
         } else {
diff --git a/llvm/lib/CodeGen/RegAllocGreedy.cpp b/llvm/lib/CodeGen/RegAllocGreedy.cpp
--- a/llvm/lib/CodeGen/RegAllocGreedy.cpp
+++ b/llvm/lib/CodeGen/RegAllocGreedy.cpp
@@ -1286,10 +1286,12 @@
 /// VirtReg.
 static bool readsLaneSubset(const MachineRegisterInfo &MRI,
                             const MachineInstr *MI, const LiveInterval &VirtReg,
-                            const TargetRegisterInfo *TRI, SlotIndex Use) {
+                            const TargetRegisterInfo *TRI, SlotIndex Use,
+                            const TargetInstrInfo *TII) {
   // Early check the common case.
-  if (MI->isCopy() &&
-      MI->getOperand(0).getSubReg() == MI->getOperand(1).getSubReg())
+  auto DestSrc = TII->isCopyInstr(*MI);
+  if (DestSrc &&
+      DestSrc->Destination->getSubReg() == DestSrc->Source->getSubReg())
     return false;
 
   // FIXME: We're only considering uses, but should be consider defs too?
@@ -1348,14 +1350,14 @@
   // the allocation.
   for (const SlotIndex Use : Uses) {
     if (const MachineInstr *MI = Indexes->getInstructionFromIndex(Use)) {
-      if (MI->isFullCopy() ||
+      if (TII->isFullCopyInstr(*MI) ||
           (SplitSubClass &&
            SuperRCNumAllocatableRegs ==
                getNumAllocatableRegsForConstraints(MI, VirtReg.reg(), SuperRC,
                                                    TII, TRI, RegClassInfo)) ||
           // TODO: Handle split for subranges with subclass constraints?
           (!SplitSubClass && VirtReg.hasSubRanges() &&
-           !readsLaneSubset(*MRI, MI, VirtReg, TRI, Use))) {
+           !readsLaneSubset(*MRI, MI, VirtReg, TRI, Use, TII))) {
         LLVM_DEBUG(dbgs() << "    skip:\t" << Use << '\t' << *MI);
         continue;
       }
@@ -2142,7 +2144,7 @@
 /// \p Out is not cleared before being populated.
 void RAGreedy::collectHintInfo(Register Reg, HintsInfo &Out) {
   for (const MachineInstr &Instr : MRI->reg_nodbg_instructions(Reg)) {
-    if (!Instr.isFullCopy())
+    if (!TII->isFullCopyInstr(Instr))
       continue;
     // Look for the other end of the copy.
     Register OtherReg = Instr.getOperand(0).getReg();
@@ -2457,22 +2459,23 @@
            MI.getOpcode() == TargetOpcode::STATEPOINT;
   };
   for (MachineInstr &MI : MBB) {
-    if (MI.isCopy()) {
-      const MachineOperand &Dest = MI.getOperand(0);
-      const MachineOperand &Src = MI.getOperand(1);
-      Register SrcReg = Src.getReg();
-      Register DestReg = Dest.getReg();
+    auto DestSrc = TII->isCopyInstr(MI);
+    if (DestSrc) {
+      const MachineOperand *Dest = DestSrc->Destination;
+      const MachineOperand *Src = DestSrc->Source;
+      Register SrcReg = Src->getReg();
+      Register DestReg = Dest->getReg();
       // Only count `COPY`s with a virtual register as source or destination.
       if (SrcReg.isVirtual() || DestReg.isVirtual()) {
         if (SrcReg.isVirtual()) {
           SrcReg = VRM->getPhys(SrcReg);
-          if (Src.getSubReg())
-            SrcReg = TRI->getSubReg(SrcReg, Src.getSubReg());
+          if (Src->getSubReg())
+            SrcReg = TRI->getSubReg(SrcReg, Src->getSubReg());
         }
         if (DestReg.isVirtual()) {
           DestReg = VRM->getPhys(DestReg);
-          if (Dest.getSubReg())
-            DestReg = TRI->getSubReg(DestReg, Dest.getSubReg());
+          if (Dest->getSubReg())
+            DestReg = TRI->getSubReg(DestReg, Dest->getSubReg());
         }
         if (SrcReg != DestReg)
           ++Stats.Copies;
diff --git a/llvm/lib/CodeGen/SplitKit.cpp b/llvm/lib/CodeGen/SplitKit.cpp
--- a/llvm/lib/CodeGen/SplitKit.cpp
+++ b/llvm/lib/CodeGen/SplitKit.cpp
@@ -536,7 +536,7 @@
 SlotIndex SplitEditor::buildCopy(Register FromReg, Register ToReg,
     LaneBitmask LaneMask, MachineBasicBlock &MBB,
     MachineBasicBlock::iterator InsertBefore, bool Late, unsigned RegIdx) {
-  const MCInstrDesc &Desc = TII.get(TargetOpcode::COPY);
+  const MCInstrDesc &Desc = TII.get(TII.getLiveRangeSplitOpcode(FromReg, MRI));
   SlotIndexes &Indexes = *LIS.getSlotIndexes();
   if (LaneMask.all() || LaneMask == MRI.getMaxLaneMaskForVReg(FromReg)) {
     // The full vreg is copied.
@@ -1584,7 +1584,9 @@
   if (BI.LiveIn && BI.LiveOut)
     return true;
   // No point in isolating a copy. It has no register class constraints.
-  if (LIS.getInstructionFromIndex(BI.FirstInstr)->isCopyLike())
+  MachineInstr *MI = LIS.getInstructionFromIndex(BI.FirstInstr);
+  bool copyLike = TII.isCopyInstr(*MI) || MI->isSubregToReg();
+  if (copyLike)
     return false;
   // Finally, don't isolate an end point that was created by earlier splits.
   return isOriginalEndpoint(BI.FirstInstr);
diff --git a/llvm/lib/CodeGen/TargetInstrInfo.cpp b/llvm/lib/CodeGen/TargetInstrInfo.cpp
--- a/llvm/lib/CodeGen/TargetInstrInfo.cpp
+++ b/llvm/lib/CodeGen/TargetInstrInfo.cpp
@@ -440,8 +440,9 @@
 // If the COPY instruction in MI can be folded to a stack operation, return
 // the register class to use.
 static const TargetRegisterClass *canFoldCopy(const MachineInstr &MI,
+                                              const TargetInstrInfo &TII,
                                               unsigned FoldIdx) {
-  assert(MI.isCopy() && "MI must be a COPY instruction");
+  assert(TII.isCopyInstr(MI) && "MI must be a COPY instruction");
   if (MI.getNumOperands() != 2)
     return nullptr;
   assert(FoldIdx<2 && "FoldIdx refers no nonexistent operand");
@@ -630,10 +631,10 @@
   }
 
   // Straight COPY may fold as load/store.
-  if (!MI.isCopy() || Ops.size() != 1)
+  if (!isCopyInstr(MI) || Ops.size() != 1)
     return nullptr;
 
-  const TargetRegisterClass *RC = canFoldCopy(MI, Ops[0]);
+  const TargetRegisterClass *RC = canFoldCopy(MI, *this, Ops[0]);
   if (!RC)
     return nullptr;
 
diff --git a/llvm/test/CodeGen/Mips/madd-msub.ll b/llvm/test/CodeGen/Mips/madd-msub.ll
--- a/llvm/test/CodeGen/Mips/madd-msub.ll
+++ b/llvm/test/CodeGen/Mips/madd-msub.ll
@@ -42,22 +42,22 @@
 ;
 ; 64-LABEL: madd1:
 ; 64:       # %bb.0: # %entry
-; 64-NEXT:    sll $1, $4, 0
-; 64-NEXT:    sll $2, $5, 0
-; 64-NEXT:    dmult $2, $1
+; 64-NEXT:    sll $4, $4, 0
+; 64-NEXT:    sll $5, $5, 0
+; 64-NEXT:    dmult $5, $4
 ; 64-NEXT:    mflo $1
-; 64-NEXT:    sll $2, $6, 0
+; 64-NEXT:    sll $6, $6, 0
 ; 64-NEXT:    jr $ra
-; 64-NEXT:    daddu $2, $1, $2
+; 64-NEXT:    daddu $2, $1, $6
 ;
 ; 64R6-LABEL: madd1:
 ; 64R6:       # %bb.0: # %entry
-; 64R6-NEXT:    sll $1, $4, 0
-; 64R6-NEXT:    sll $2, $5, 0
-; 64R6-NEXT:    dmul $1, $2, $1
-; 64R6-NEXT:    sll $2, $6, 0
+; 64R6-NEXT:    sll $4, $4, 0
+; 64R6-NEXT:    sll $5, $5, 0
+; 64R6-NEXT:    dmul $1, $5, $4
+; 64R6-NEXT:    sll $6, $6, 0
 ; 64R6-NEXT:    jr $ra
-; 64R6-NEXT:    daddu $2, $1, $2
+; 64R6-NEXT:    daddu $2, $1, $6
 ;
 ; 16-LABEL: madd1:
 ; 16:       # %bb.0: # %entry
@@ -173,18 +173,18 @@
 ;
 ; 64-LABEL: madd3:
 ; 64:       # %bb.0: # %entry
-; 64-NEXT:    sll $1, $4, 0
-; 64-NEXT:    sll $2, $5, 0
-; 64-NEXT:    dmult $2, $1
+; 64-NEXT:    sll $4, $4, 0
+; 64-NEXT:    sll $5, $5, 0
+; 64-NEXT:    dmult $5, $4
 ; 64-NEXT:    mflo $1
 ; 64-NEXT:    jr $ra
 ; 64-NEXT:    daddu $2, $1, $6
 ;
 ; 64R6-LABEL: madd3:
 ; 64R6:       # %bb.0: # %entry
-; 64R6-NEXT:    sll $1, $4, 0
-; 64R6-NEXT:    sll $2, $5, 0
-; 64R6-NEXT:    dmul $1, $2, $1
+; 64R6-NEXT:    sll $4, $4, 0
+; 64R6-NEXT:    sll $5, $5, 0
+; 64R6-NEXT:    dmul $1, $5, $4
 ; 64R6-NEXT:    jr $ra
 ; 64R6-NEXT:    daddu $2, $1, $6
 ;
@@ -291,22 +291,22 @@
 ;
 ; 64-LABEL: msub1:
 ; 64:       # %bb.0: # %entry
-; 64-NEXT:    sll $1, $4, 0
-; 64-NEXT:    sll $2, $5, 0
-; 64-NEXT:    dmult $2, $1
+; 64-NEXT:    sll $4, $4, 0
+; 64-NEXT:    sll $5, $5, 0
+; 64-NEXT:    dmult $5, $4
 ; 64-NEXT:    mflo $1
-; 64-NEXT:    sll $2, $6, 0
+; 64-NEXT:    sll $6, $6, 0
 ; 64-NEXT:    jr $ra
-; 64-NEXT:    dsubu $2, $2, $1
+; 64-NEXT:    dsubu $2, $6, $1
 ;
 ; 64R6-LABEL: msub1:
 ; 64R6:       # %bb.0: # %entry
-; 64R6-NEXT:    sll $1, $4, 0
-; 64R6-NEXT:    sll $2, $5, 0
-; 64R6-NEXT:    dmul $1, $2, $1
-; 64R6-NEXT:    sll $2, $6, 0
+; 64R6-NEXT:    sll $4, $4, 0
+; 64R6-NEXT:    sll $5, $5, 0
+; 64R6-NEXT:    dmul $1, $5, $4
+; 64R6-NEXT:    sll $6, $6, 0
 ; 64R6-NEXT:    jr $ra
-; 64R6-NEXT:    dsubu $2, $2, $1
+; 64R6-NEXT:    dsubu $2, $6, $1
 ;
 ; 16-LABEL: msub1:
 ; 16:       # %bb.0: # %entry
@@ -424,18 +424,18 @@
 ;
 ; 64-LABEL: msub3:
 ; 64:       # %bb.0: # %entry
-; 64-NEXT:    sll $1, $4, 0
-; 64-NEXT:    sll $2, $5, 0
-; 64-NEXT:    dmult $2, $1
+; 64-NEXT:    sll $4, $4, 0
+; 64-NEXT:    sll $5, $5, 0
+; 64-NEXT:    dmult $5, $4
 ; 64-NEXT:    mflo $1
 ; 64-NEXT:    jr $ra
 ; 64-NEXT:    dsubu $2, $6, $1
 ;
 ; 64R6-LABEL: msub3:
 ; 64R6:       # %bb.0: # %entry
-; 64R6-NEXT:    sll $1, $4, 0
-; 64R6-NEXT:    sll $2, $5, 0
-; 64R6-NEXT:    dmul $1, $2, $1
+; 64R6-NEXT:    sll $4, $4, 0
+; 64R6-NEXT:    sll $5, $5, 0
+; 64R6-NEXT:    dmul $1, $5, $4
 ; 64R6-NEXT:    jr $ra
 ; 64R6-NEXT:    dsubu $2, $6, $1
 ;
@@ -546,22 +546,22 @@
 ;
 ; 64-LABEL: msub5:
 ; 64:       # %bb.0: # %entry
-; 64-NEXT:    sll $1, $4, 0
-; 64-NEXT:    sll $2, $5, 0
-; 64-NEXT:    dmult $2, $1
+; 64-NEXT:    sll $4, $4, 0
+; 64-NEXT:    sll $5, $5, 0
+; 64-NEXT:    dmult $5, $4
 ; 64-NEXT:    mflo $1
-; 64-NEXT:    sll $2, $6, 0
+; 64-NEXT:    sll $6, $6, 0
 ; 64-NEXT:    jr $ra
-; 64-NEXT:    dsubu $2, $1, $2
+; 64-NEXT:    dsubu $2, $1, $6
 ;
 ; 64R6-LABEL: msub5:
 ; 64R6:       # %bb.0: # %entry
-; 64R6-NEXT:    sll $1, $4, 0
-; 64R6-NEXT:    sll $2, $5, 0
-; 64R6-NEXT:    dmul $1, $2, $1
-; 64R6-NEXT:    sll $2, $6, 0
+; 64R6-NEXT:    sll $4, $4, 0
+; 64R6-NEXT:    sll $5, $5, 0
+; 64R6-NEXT:    dmul $1, $5, $4
+; 64R6-NEXT:    sll $6, $6, 0
 ; 64R6-NEXT:    jr $ra
-; 64R6-NEXT:    dsubu $2, $1, $2
+; 64R6-NEXT:    dsubu $2, $1, $6
 ;
 ; 16-LABEL: msub5:
 ; 16:       # %bb.0: # %entry
diff --git a/llvm/test/CodeGen/Thumb2/mve-float32regloops.ll b/llvm/test/CodeGen/Thumb2/mve-float32regloops.ll
--- a/llvm/test/CodeGen/Thumb2/mve-float32regloops.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-float32regloops.ll
@@ -1614,32 +1614,33 @@
 ; CHECK-NEXT:    @ Parent Loop BB19_3 Depth=1
 ; CHECK-NEXT:    @ => This Inner Loop Header: Depth=2
 ; CHECK-NEXT:    vmov r7, s7
-; CHECK-NEXT:    vldrw.u32 q2, [r9, #16]
+; CHECK-NEXT:    vldr s0, [r1, #12]
 ; CHECK-NEXT:    vmov r11, s6
 ; CHECK-NEXT:    vldrw.u32 q1, [r9, #112]
-; CHECK-NEXT:    vmov r4, s1
-; CHECK-NEXT:    vldr s1, [r1, #12]
 ; CHECK-NEXT:    vmov r3, s3
 ; CHECK-NEXT:    vldr s3, [r1, #8]
 ; CHECK-NEXT:    vstrw.32 q1, [sp, #32] @ 16-byte Spill
 ; CHECK-NEXT:    vldrw.u32 q1, [r9]
-; CHECK-NEXT:    vmov r8, s1
+; CHECK-NEXT:    vmov r8, s0
+; CHECK-NEXT:    vldrw.u32 q2, [r9, #16]
 ; CHECK-NEXT:    ldr r6, [r1, #4]
 ; CHECK-NEXT:    vldrw.u32 q7, [r9, #32]
 ; CHECK-NEXT:    vmul.f32 q1, q1, r8
 ; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    vldrw.u32 q3, [r9, #48]
 ; CHECK-NEXT:    vfma.f32 q1, q2, r0
+; CHECK-NEXT:    vldrw.u32 q3, [r9, #48]
 ; CHECK-NEXT:    ldr r0, [r1], #16
 ; CHECK-NEXT:    vfma.f32 q1, q7, r6
+; CHECK-NEXT:    vmov r4, s1
 ; CHECK-NEXT:    vldrw.u32 q6, [r9, #64]
-; CHECK-NEXT:    vmov.f32 s2, s1
+; CHECK-NEXT:    vmov.f32 s1, s0
 ; CHECK-NEXT:    vfma.f32 q1, q3, r0
+; CHECK-NEXT:    vmov.f32 s2, s0
 ; CHECK-NEXT:    vldrw.u32 q5, [r9, #80]
 ; CHECK-NEXT:    vfma.f32 q1, q6, r4
 ; CHECK-NEXT:    vldrw.u32 q4, [r9, #96]
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #32] @ 16-byte Reload
 ; CHECK-NEXT:    vfma.f32 q1, q5, r3
+; CHECK-NEXT:    vldrw.u32 q2, [sp, #32] @ 16-byte Reload
 ; CHECK-NEXT:    vfma.f32 q1, q4, r7
 ; CHECK-NEXT:    vfma.f32 q1, q2, r11
 ; CHECK-NEXT:    vstrb.8 q1, [r5], #16
diff --git a/llvm/test/CodeGen/X86/GlobalISel/add-ext.ll b/llvm/test/CodeGen/X86/GlobalISel/add-ext.ll
--- a/llvm/test/CodeGen/X86/GlobalISel/add-ext.ll
+++ b/llvm/test/CodeGen/X86/GlobalISel/add-ext.ll
@@ -205,8 +205,8 @@
 ; CHECK-NEXT:    addq %rdi, %rcx
 ; CHECK-NEXT:    movl (%rcx), %ecx
 ; CHECK-NEXT:    addl (%rax), %ecx
-; CHECK-NEXT:    movl %esi, %eax
-; CHECK-NEXT:    imulq $4, %rax, %rax
+; CHECK-NEXT:    movl %esi, %esi
+; CHECK-NEXT:    imulq $4, %rsi, %rax
 ; CHECK-NEXT:    addq %rdi, %rax
 ; CHECK-NEXT:    movl %ecx, (%rax)
 ; CHECK-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/dagcombine-cse.ll b/llvm/test/CodeGen/X86/dagcombine-cse.ll
--- a/llvm/test/CodeGen/X86/dagcombine-cse.ll
+++ b/llvm/test/CodeGen/X86/dagcombine-cse.ll
@@ -106,24 +106,24 @@
 ;
 ; X64-LABEL: square_high:
 ; X64:       ## %bb.0: ## %entry
-; X64-NEXT:    movl %esi, %ecx
-; X64-NEXT:    movq %rcx, %rax
+; X64-NEXT:    movl %esi, %esi
+; X64-NEXT:    movq %rsi, %rax
 ; X64-NEXT:    mulq %rdi
-; X64-NEXT:    movq %rdx, %rsi
+; X64-NEXT:    movq %rdx, %rcx
 ; X64-NEXT:    movq %rax, %r8
 ; X64-NEXT:    movq %rdi, %rax
 ; X64-NEXT:    mulq %rdi
 ; X64-NEXT:    addq %r8, %rdx
-; X64-NEXT:    movq %rsi, %rax
+; X64-NEXT:    movq %rcx, %rax
 ; X64-NEXT:    adcq $0, %rax
 ; X64-NEXT:    addq %rdx, %r8
-; X64-NEXT:    adcq %rsi, %rax
-; X64-NEXT:    imulq %rcx, %rcx
-; X64-NEXT:    addq %rax, %rcx
-; X64-NEXT:    shrdq $32, %rcx, %r8
-; X64-NEXT:    shrq $32, %rcx
+; X64-NEXT:    adcq %rcx, %rax
+; X64-NEXT:    imulq %rsi, %rsi
+; X64-NEXT:    addq %rax, %rsi
+; X64-NEXT:    shrdq $32, %rsi, %r8
+; X64-NEXT:    shrq $32, %rsi
 ; X64-NEXT:    movq %r8, %rax
-; X64-NEXT:    movq %rcx, %rdx
+; X64-NEXT:    movq %rsi, %rdx
 ; X64-NEXT:    retq
 entry:
   %conv = zext i96 %x to i192
diff --git a/llvm/test/CodeGen/X86/fold-and-shift-x86_64.ll b/llvm/test/CodeGen/X86/fold-and-shift-x86_64.ll
--- a/llvm/test/CodeGen/X86/fold-and-shift-x86_64.ll
+++ b/llvm/test/CodeGen/X86/fold-and-shift-x86_64.ll
@@ -34,8 +34,8 @@
 define i8 @t3(ptr %X, i64 %i) {
 ; CHECK-LABEL: t3:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    movl %esi, %eax
-; CHECK-NEXT:    movzbl (%rdi,%rax,4), %eax
+; CHECK-NEXT:    movl %esi, %esi
+; CHECK-NEXT:    movzbl (%rdi,%rsi,4), %eax
 ; CHECK-NEXT:    retq
 
 entry:
diff --git a/llvm/test/CodeGen/X86/unfold-masked-merge-scalar-constmask-lowhigh.ll b/llvm/test/CodeGen/X86/unfold-masked-merge-scalar-constmask-lowhigh.ll
--- a/llvm/test/CodeGen/X86/unfold-masked-merge-scalar-constmask-lowhigh.ll
+++ b/llvm/test/CodeGen/X86/unfold-masked-merge-scalar-constmask-lowhigh.ll
@@ -78,18 +78,18 @@
 define i64 @out64_constmask(i64 %x, i64 %y) {
 ; CHECK-NOBMI-LABEL: out64_constmask:
 ; CHECK-NOBMI:       # %bb.0:
-; CHECK-NOBMI-NEXT:    movl %edi, %ecx
+; CHECK-NOBMI-NEXT:    movl %edi, %edi
 ; CHECK-NOBMI-NEXT:    movabsq $-4294967296, %rax # imm = 0xFFFFFFFF00000000
 ; CHECK-NOBMI-NEXT:    andq %rsi, %rax
-; CHECK-NOBMI-NEXT:    orq %rcx, %rax
+; CHECK-NOBMI-NEXT:    orq %rdi, %rax
 ; CHECK-NOBMI-NEXT:    retq
 ;
 ; CHECK-BMI-LABEL: out64_constmask:
 ; CHECK-BMI:       # %bb.0:
-; CHECK-BMI-NEXT:    movl %edi, %ecx
+; CHECK-BMI-NEXT:    movl %edi, %edi
 ; CHECK-BMI-NEXT:    movabsq $-4294967296, %rax # imm = 0xFFFFFFFF00000000
 ; CHECK-BMI-NEXT:    andq %rsi, %rax
-; CHECK-BMI-NEXT:    orq %rcx, %rax
+; CHECK-BMI-NEXT:    orq %rdi, %rax
 ; CHECK-BMI-NEXT:    retq
   %mx = and i64 %x, 4294967295
   %my = and i64 %y, -4294967296