diff --git a/lib/CodeGen/VirtRegMap.cpp b/lib/CodeGen/VirtRegMap.cpp
index 7a7ea69e27f..226b78f7bcb 100644
--- a/lib/CodeGen/VirtRegMap.cpp
+++ b/lib/CodeGen/VirtRegMap.cpp
@@ -260,6 +260,8 @@ void VirtRegMap::rewrite(SlotIndexes *Indexes) {
                << "********** Function: "
                << MF->getFunction()->getName() << '\n');
   DEBUG(dump());
+  SmallVector<unsigned, 8> SuperDeads;
+  SmallVector<unsigned, 8> SuperDefs;
   SmallVector<unsigned, 8> SuperKills;
 
   for (MachineFunction::iterator MBBI = MF->begin(), MBBE = MF->end();
@@ -283,12 +285,13 @@ void VirtRegMap::rewrite(SlotIndexes *Indexes) {
         if (MO.getSubReg()) {
           // A virtual register kill refers to the whole register, so we may
           // have to add <imp-use,kill> operands for the super-register.
-          if (MO.isUse() && MO.isKill() && !MO.isUndef())
-            SuperKills.push_back(PhysReg);
-
-          // We don't have to deal with sub-register defs because
-          // LiveIntervalAnalysis already added the necessary <imp-def>
-          // operands.
+          if (MO.isUse()) {
+            if (MO.isKill() && !MO.isUndef())
+              SuperKills.push_back(PhysReg);
+          } else if (MO.isDead())
+            SuperDeads.push_back(PhysReg);
+          else
+            SuperDefs.push_back(PhysReg);
 
           // PhysReg operands cannot have subregister indexes.
           PhysReg = TRI->getSubReg(PhysReg, MO.getSubReg());
@@ -305,6 +308,12 @@ void VirtRegMap::rewrite(SlotIndexes *Indexes) {
       while (!SuperKills.empty())
         MI->addRegisterKilled(SuperKills.pop_back_val(), TRI, true);
 
+      while (!SuperDeads.empty())
+        MI->addRegisterDead(SuperDeads.pop_back_val(), TRI, true);
+
+      while (!SuperDefs.empty())
+        MI->addRegisterDefined(SuperDefs.pop_back_val(), TRI);
+
       DEBUG(dbgs() << "> " << *MI);
 
       // Finally, remove any identity copies.
diff --git a/test/CodeGen/ARM/crash-greedy.ll b/test/CodeGen/ARM/crash-greedy.ll
index 0b229b27d2d..8a865e23d0a 100644
--- a/test/CodeGen/ARM/crash-greedy.ll
+++ b/test/CodeGen/ARM/crash-greedy.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -regalloc=greedy -mcpu=cortex-a8 -relocation-model=pic -disable-fp-elim | FileCheck %s
+; RUN: llc < %s -regalloc=greedy -mcpu=cortex-a8 -relocation-model=pic -disable-fp-elim -verify-machineinstrs | FileCheck %s
 ;
 ; ARM tests that crash or fail with the greedy register allocator.
 
@@ -59,3 +59,26 @@ for.end:                                          ; preds = %cond.end
   ret void
 }
 
+; CHECK: insert_elem
+; This test has a sub-register copy with a kill flag:
+;   %vreg6:ssub_3<def> = COPY %vreg6:ssub_2<kill>; QPR_VFP2:%vreg6
+; The rewriter must do something sensible with that, or the scavenger crashes.
+define void @insert_elem() nounwind {
+entry:
+  br i1 undef, label %if.end251, label %if.then84
+
+if.then84:                                        ; preds = %entry
+  br i1 undef, label %if.end251, label %if.then195
+
+if.then195:                                       ; preds = %if.then84
+  %div = fdiv float 1.000000e+00, undef
+  %vecinit207 = insertelement <4 x float> undef, float %div, i32 1
+  %vecinit208 = insertelement <4 x float> %vecinit207, float 1.000000e+00, i32 2
+  %vecinit209 = insertelement <4 x float> %vecinit208, float 1.000000e+00, i32 3
+  %mul216 = fmul <4 x float> zeroinitializer, %vecinit209
+  store <4 x float> %mul216, <4 x float>* undef, align 16
+  br label %if.end251
+
+if.end251:                                        ; preds = %if.then195, %if.then84, %entry
+  ret void
+}