diff --git a/llvm/include/llvm/CodeGen/MachineFunction.h b/llvm/include/llvm/CodeGen/MachineFunction.h
index 010d7032c516a607d50082991a9f43586a952736..da3838f249768c2b009890d52752ca2d4bac1709 100644
--- a/llvm/include/llvm/CodeGen/MachineFunction.h
+++ b/llvm/include/llvm/CodeGen/MachineFunction.h
@@ -625,14 +625,23 @@ public:
   MachineInstr *CreateMachineInstr(const MCInstrDesc &MCID, const DebugLoc &DL,
                                    bool NoImp = false);
 
-  /// CloneMachineInstr - Create a new MachineInstr which is a copy of the
-  /// 'Orig' instruction, identical in all ways except the instruction
-  /// has no parent, prev, or next.
+  /// Create a new MachineInstr which is a copy of \p Orig, identical in all
+  /// ways except the instruction has no parent, prev, or next. Bundling flags
+  /// are reset.
   ///
-  /// See also TargetInstrInfo::duplicate() for target-specific fixes to cloned
-  /// instructions.
+  /// Note: Clones a single instruction, not whole instruction bundles.
+  /// Does not perform target specific adjustments; consider using
+  /// TargetInstrInfo::duplicate() instead.
   MachineInstr *CloneMachineInstr(const MachineInstr *Orig);
 
+  /// Clones instruction or the whole instruction bundle \p Orig and insert
+  /// into \p MBB before \p InsertBefore.
+  ///
+  /// Note: Does not perform target specific adjustments; consider using
+  /// TargetInstrInfo::duplicate() intead.
+  MachineInstr &CloneMachineInstrBundle(MachineBasicBlock &MBB,
+      MachineBasicBlock::iterator InsertBefore, const MachineInstr &Orig);
+
   /// DeleteMachineInstr - Delete the given MachineInstr.
   void DeleteMachineInstr(MachineInstr *MI);
 
diff --git a/llvm/include/llvm/Target/TargetInstrInfo.h b/llvm/include/llvm/Target/TargetInstrInfo.h
index 612f0b601c831d1b83020b75e1fb1da39defa24a..6e0838ceec0aed362a0f75f0790bd75964533a09 100644
--- a/llvm/include/llvm/Target/TargetInstrInfo.h
+++ b/llvm/include/llvm/Target/TargetInstrInfo.h
@@ -324,13 +324,13 @@ public:
                              unsigned SubIdx, const MachineInstr &Orig,
                              const TargetRegisterInfo &TRI) const;
 
-  /// Create a duplicate of the Orig instruction in MF. This is like
-  /// MachineFunction::CloneMachineInstr(), but the target may update operands
+  /// \brief Clones instruction or the whole instruction bundle \p Orig and
+  /// insert into \p MBB before \p InsertBefore. The target may update operands
   /// that are required to be unique.
   ///
-  /// The instruction must be duplicable as indicated by isNotDuplicable().
-  virtual MachineInstr *duplicate(MachineInstr &Orig,
-                                  MachineFunction &MF) const;
+  /// \p Orig must not return true for MachineInstr::isNotDuplicable().
+  virtual MachineInstr &duplicate(MachineBasicBlock &MBB,
+      MachineBasicBlock::iterator InsertBefore, const MachineInstr &Orig) const;
 
   /// This method must be implemented by targets that
   /// set the M_CONVERTIBLE_TO_3_ADDR flag.  When this flag is set, the target
diff --git a/llvm/lib/CodeGen/MachineFunction.cpp b/llvm/lib/CodeGen/MachineFunction.cpp
index 742b095d955e8226d8c2e350abb64b27bf042b25..c4c0dff7bdf527c084208723173e239f39b86ed4 100644
--- a/llvm/lib/CodeGen/MachineFunction.cpp
+++ b/llvm/lib/CodeGen/MachineFunction.cpp
@@ -270,6 +270,26 @@ MachineFunction::CloneMachineInstr(const MachineInstr *Orig) {
              MachineInstr(*this, *Orig);
 }
 
+MachineInstr &MachineFunction::CloneMachineInstrBundle(MachineBasicBlock &MBB,
+    MachineBasicBlock::iterator InsertBefore, const MachineInstr &Orig) {
+  MachineInstr *FirstClone = nullptr;
+  MachineBasicBlock::const_instr_iterator I = Orig.getIterator();
+  for (;;) {
+    MachineInstr *Cloned = CloneMachineInstr(&*I);
+    MBB.insert(InsertBefore, Cloned);
+    if (FirstClone == nullptr) {
+      FirstClone = Cloned;
+    } else {
+      Cloned->bundleWithPred();
+    }
+
+    if (!I->isBundledWithSucc())
+      break;
+    ++I;
+  }
+  return *FirstClone;
+}
+
 /// Delete the given MachineInstr.
 ///
 /// This function also serves as the MachineInstr destructor - the real
diff --git a/llvm/lib/CodeGen/TailDuplicator.cpp b/llvm/lib/CodeGen/TailDuplicator.cpp
index dc7265dcf6c244403670dfa01e91449599dca50c..0f22040f3aec5e292828225642e32c26ef0a7b65 100644
--- a/llvm/lib/CodeGen/TailDuplicator.cpp
+++ b/llvm/lib/CodeGen/TailDuplicator.cpp
@@ -369,10 +369,10 @@ void TailDuplicator::duplicateInstruction(
     MachineInstr *MI, MachineBasicBlock *TailBB, MachineBasicBlock *PredBB,
     DenseMap<unsigned, RegSubRegPair> &LocalVRMap,
     const DenseSet<unsigned> &UsedByPhi) {
-  MachineInstr *NewMI = TII->duplicate(*MI, *MF);
+  MachineInstr &NewMI = TII->duplicate(*PredBB, PredBB->end(), *MI);
   if (PreRegAlloc) {
-    for (unsigned i = 0, e = NewMI->getNumOperands(); i != e; ++i) {
-      MachineOperand &MO = NewMI->getOperand(i);
+    for (unsigned i = 0, e = NewMI.getNumOperands(); i != e; ++i) {
+      MachineOperand &MO = NewMI.getOperand(i);
       if (!MO.isReg())
         continue;
       unsigned Reg = MO.getReg();
@@ -443,7 +443,6 @@ void TailDuplicator::duplicateInstruction(
       }
     }
   }
-  PredBB->insert(PredBB->instr_end(), NewMI);
 }
 
 /// After FromBB is tail duplicated into its predecessor blocks, the successors
@@ -825,10 +824,8 @@ bool TailDuplicator::tailDuplicate(bool IsSimple, MachineBasicBlock *TailBB,
     // Clone the contents of TailBB into PredBB.
     DenseMap<unsigned, RegSubRegPair> LocalVRMap;
     SmallVector<std::pair<unsigned, RegSubRegPair>, 4> CopyInfos;
-    // Use instr_iterator here to properly handle bundles, e.g.
-    // ARM Thumb2 IT block.
-    MachineBasicBlock::instr_iterator I = TailBB->instr_begin();
-    while (I != TailBB->instr_end()) {
+    for (MachineBasicBlock::iterator I = TailBB->begin(), E = TailBB->end();
+         I != E; /* empty */) {
       MachineInstr *MI = &*I;
       ++I;
       if (MI->isPHI()) {
diff --git a/llvm/lib/CodeGen/TargetInstrInfo.cpp b/llvm/lib/CodeGen/TargetInstrInfo.cpp
index 14c5adc0d898bd0f33e51f6d5d6ee5409ed8be7b..5e92cbf1ee0d80b668f96ac133e1765964b4f79d 100644
--- a/llvm/lib/CodeGen/TargetInstrInfo.cpp
+++ b/llvm/lib/CodeGen/TargetInstrInfo.cpp
@@ -388,10 +388,11 @@ bool TargetInstrInfo::produceSameValue(const MachineInstr &MI0,
   return MI0.isIdenticalTo(MI1, MachineInstr::IgnoreVRegDefs);
 }
 
-MachineInstr *TargetInstrInfo::duplicate(MachineInstr &Orig,
-                                         MachineFunction &MF) const {
+MachineInstr &TargetInstrInfo::duplicate(MachineBasicBlock &MBB,
+    MachineBasicBlock::iterator InsertBefore, const MachineInstr &Orig) const {
   assert(!Orig.isNotDuplicable() && "Instruction cannot be duplicated");
-  return MF.CloneMachineInstr(&Orig);
+  MachineFunction &MF = *MBB.getParent();
+  return MF.CloneMachineInstrBundle(MBB, InsertBefore, Orig);
 }
 
 // If the COPY instruction in MI can be folded to a stack operation, return
diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 3cf5950a1918d9583004fb6aa1cd36d91525ff18..3688db943d54161ca649be32a9890b31dcbd8975 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -1550,20 +1550,29 @@ void ARMBaseInstrInfo::reMaterialize(MachineBasicBlock &MBB,
   }
 }
 
-MachineInstr *ARMBaseInstrInfo::duplicate(MachineInstr &Orig,
-                                          MachineFunction &MF) const {
-  MachineInstr *MI = TargetInstrInfo::duplicate(Orig, MF);
-  switch (Orig.getOpcode()) {
-  case ARM::tLDRpci_pic:
-  case ARM::t2LDRpci_pic: {
-    unsigned CPI = Orig.getOperand(1).getIndex();
-    unsigned PCLabelId = duplicateCPV(MF, CPI);
-    Orig.getOperand(1).setIndex(CPI);
-    Orig.getOperand(2).setImm(PCLabelId);
-    break;
-  }
+MachineInstr &
+ARMBaseInstrInfo::duplicate(MachineBasicBlock &MBB,
+    MachineBasicBlock::iterator InsertBefore,
+    const MachineInstr &Orig) const {
+  MachineInstr &Cloned = TargetInstrInfo::duplicate(MBB, InsertBefore, Orig);
+  MachineBasicBlock::instr_iterator I = Cloned.getIterator();
+  for (;;) {
+    switch (I->getOpcode()) {
+    case ARM::tLDRpci_pic:
+    case ARM::t2LDRpci_pic: {
+      MachineFunction &MF = *MBB.getParent();
+      unsigned CPI = I->getOperand(1).getIndex();
+      unsigned PCLabelId = duplicateCPV(MF, CPI);
+      I->getOperand(1).setIndex(CPI);
+      I->getOperand(2).setImm(PCLabelId);
+      break;
+    }
+    }
+    if (!I->isBundledWithSucc())
+      break;
+    ++I;
   }
-  return MI;
+  return Cloned;
 }
 
 bool ARMBaseInstrInfo::produceSameValue(const MachineInstr &MI0,
diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.h b/llvm/lib/Target/ARM/ARMBaseInstrInfo.h
index c52e572786d4823d3e22e157f9ba1c5a5d3bd9ec..9f168acd56769d04efd56d7fc3ea700a14ba9a81 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.h
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.h
@@ -220,8 +220,9 @@ public:
                      const MachineInstr &Orig,
                      const TargetRegisterInfo &TRI) const override;
 
-  MachineInstr *duplicate(MachineInstr &Orig,
-                          MachineFunction &MF) const override;
+  MachineInstr &
+  duplicate(MachineBasicBlock &MBB, MachineBasicBlock::iterator InsertBefore,
+            const MachineInstr &Orig) const override;
 
   const MachineInstrBuilder &AddDReg(MachineInstrBuilder &MIB, unsigned Reg,
                                      unsigned SubIdx, unsigned State,