diff --git a/lib/CodeGen/LowerSubregs.cpp b/lib/CodeGen/LowerSubregs.cpp
index 85729956824..dfd666f96db 100644
--- a/lib/CodeGen/LowerSubregs.cpp
+++ b/lib/CodeGen/LowerSubregs.cpp
@@ -19,6 +19,7 @@
 #include "llvm/Function.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include "llvm/CodeGen/MachineInstr.h"
+#include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/Target/TargetRegisterInfo.h"
 #include "llvm/Target/TargetInstrInfo.h"
@@ -130,6 +131,7 @@ bool LowerSubregsInstructionPass::LowerExtract(MachineInstr *MI) {
            MII != MBB->end(); ++MII)
         if (MII->killsRegister(DstReg, &TRI)) {
           MII->addRegisterKilled(SuperReg, &TRI, /*AddIfNotFound=*/true);
+          DOUT << "\nsubreg: killed here: " << *MII;
           break;
         }
   } else {
@@ -231,7 +233,7 @@ bool LowerSubregsInstructionPass::LowerInsert(MachineInstr *MI) {
   assert(DstReg == SrcReg && "insert_subreg not a two-address instruction?");
   assert(SubIdx != 0 && "Invalid index for insert_subreg");
   unsigned DstSubReg = TRI.getSubReg(DstReg, SubIdx);
-  
+  assert(DstSubReg && "invalid subregister index for register");
   assert(TargetRegisterInfo::isPhysicalRegister(SrcReg) &&
          "Insert superreg source must be in a physical register");
   assert(TargetRegisterInfo::isPhysicalRegister(InsReg) &&
@@ -240,24 +242,45 @@ bool LowerSubregsInstructionPass::LowerInsert(MachineInstr *MI) {
   DOUT << "subreg: CONVERTING: " << *MI;
 
   if (DstSubReg == InsReg) {
-    // No need to insert an identify copy instruction.
-    DOUT << "subreg: eliminated!";
+    // No need to insert an identity copy instruction. If the SrcReg was
+    // <undef>, we need to make sure it is alive by inserting an IMPLICIT_DEF
+    if (MI->getOperand(1).isUndef() && !MI->getOperand(0).isDead()) {
+      BuildMI(*MBB, MI, MI->getDebugLoc(),
+              TII.get(TargetInstrInfo::IMPLICIT_DEF), DstReg)
+        .addReg(InsReg, RegState::ImplicitKill);
+    } else {
+      DOUT << "subreg: eliminated!\n";
+      MBB->erase(MI);
+      return true;
+    }
   } else {
     // Insert sub-register copy
     const TargetRegisterClass *TRC0= TRI.getPhysicalRegisterRegClass(DstSubReg);
     const TargetRegisterClass *TRC1= TRI.getPhysicalRegisterRegClass(InsReg);
     TII.copyRegToReg(*MBB, MI, DstSubReg, InsReg, TRC0, TRC1);
+    MachineBasicBlock::iterator CopyMI = MI;
+    --CopyMI;
+
     // Transfer the kill/dead flags, if needed.
-    if (MI->getOperand(0).isDead())
+    if (MI->getOperand(0).isDead()) {
       TransferDeadFlag(MI, DstSubReg, TRI);
-    if (MI->getOperand(1).isKill())
+      // Also add a SrcReg<imp-kill> of the super register.
+      CopyMI->addOperand(MachineOperand::CreateReg(DstReg, false, true, true));
+    } else if (MI->getOperand(1).isUndef()) {
+      // If SrcReg was marked <undef> we must make sure it is alive after this
+      // replacement.  Add a SrcReg<imp-def> operand.
+      CopyMI->addOperand(MachineOperand::CreateReg(DstReg, true, true));
+    }
+
+    // Make sure the inserted register gets killed
+    if (MI->getOperand(2).isKill())
       TransferKillFlag(MI, InsReg, TRI);
+  }
 
 #ifndef NDEBUG
-    MachineBasicBlock::iterator dMI = MI;
-    DOUT << "subreg: " << *(--dMI);
+  MachineBasicBlock::iterator dMI = MI;
+  DOUT << "subreg: " << *(--dMI);
 #endif
-  }
 
   DOUT << "\n";
   MBB->erase(MI);
diff --git a/lib/CodeGen/MachineInstr.cpp b/lib/CodeGen/MachineInstr.cpp
index 006a10a0ce6..c29a6e339e0 100644
--- a/lib/CodeGen/MachineInstr.cpp
+++ b/lib/CodeGen/MachineInstr.cpp
@@ -1041,6 +1041,8 @@ bool MachineInstr::addRegisterKilled(unsigned IncomingReg,
         if (MO.isKill())
           // The register is already marked kill.
           return true;
+        // This operand can no longer be undef since Reg is live-in.
+        MO.setIsUndef(false);
         if (isPhysReg && isRegTiedToDefOperand(i))
           // Two-address uses of physregs must not be marked kill.
           return true;
diff --git a/test/CodeGen/ARM/2009-08-02-RegScavengerAssert-Neon.ll b/test/CodeGen/ARM/2009-08-02-RegScavengerAssert-Neon.ll
new file mode 100644
index 00000000000..bdf5131ec32
--- /dev/null
+++ b/test/CodeGen/ARM/2009-08-02-RegScavengerAssert-Neon.ll
@@ -0,0 +1,29 @@
+; RUN: llvm-as < %s | llc -march=arm -mattr=+neon
+; PR4657
+
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:64:64-v128:128:128-a0:0:64"
+target triple = "armv7-apple-darwin9"
+
+define arm_apcscc <4 x i32> @scale(<4 x i32> %v, i32 %f) nounwind {
+entry:
+	%v_addr = alloca <4 x i32>		; <<4 x i32>*> [#uses=2]
+	%f_addr = alloca i32		; <i32*> [#uses=2]
+	%retval = alloca <4 x i32>		; <<4 x i32>*> [#uses=2]
+	%0 = alloca <4 x i32>		; <<4 x i32>*> [#uses=2]
+	%"alloca point" = bitcast i32 0 to i32		; <i32> [#uses=0]
+	store <4 x i32> %v, <4 x i32>* %v_addr
+	store i32 %f, i32* %f_addr
+	%1 = load <4 x i32>* %v_addr, align 16		; <<4 x i32>> [#uses=1]
+	%2 = load i32* %f_addr, align 4		; <i32> [#uses=1]
+	%3 = insertelement <4 x i32> undef, i32 %2, i32 0		; <<4 x i32>> [#uses=1]
+	%4 = shufflevector <4 x i32> %3, <4 x i32> undef, <4 x i32> zeroinitializer		; <<4 x i32>> [#uses=1]
+	%5 = mul <4 x i32> %1, %4		; <<4 x i32>> [#uses=1]
+	store <4 x i32> %5, <4 x i32>* %0, align 16
+	%6 = load <4 x i32>* %0, align 16		; <<4 x i32>> [#uses=1]
+	store <4 x i32> %6, <4 x i32>* %retval, align 16
+	br label %return
+
+return:		; preds = %entry
+	%retval1 = load <4 x i32>* %retval		; <<4 x i32>> [#uses=1]
+	ret <4 x i32> %retval1
+}