diff --git a/lib/Transforms/Scalar/CodeGenPrepare.cpp b/lib/Transforms/Scalar/CodeGenPrepare.cpp
index e951d98c0c8..37b97830716 100644
--- a/lib/Transforms/Scalar/CodeGenPrepare.cpp
+++ b/lib/Transforms/Scalar/CodeGenPrepare.cpp
@@ -594,15 +594,12 @@ bool CodeGenPrepare::DupRetToEnableTailCallOpts(ReturnInst *RI) {
     return false;
 
   Value *V = RI->getReturnValue();
-  if (!V)
-    return false;
-
-  PHINode *PN = dyn_cast<PHINode>(V);
-  if (!PN)
+  PHINode *PN = V ? dyn_cast<PHINode>(V) : NULL;
+  if (V && !PN)
     return false;
 
   BasicBlock *BB = RI->getParent();
-  if (PN->getParent() != BB)
+  if (PN && PN->getParent() != BB)
     return false;
 
   // It's not safe to eliminate the sign / zero extension of the return value.
@@ -612,21 +609,44 @@ bool CodeGenPrepare::DupRetToEnableTailCallOpts(ReturnInst *RI) {
   if ((CallerRetAttr & Attribute::ZExt) || (CallerRetAttr & Attribute::SExt))
     return false;
 
-  // Make sure there are no instructions between PHI and return.
-  BasicBlock::iterator BI = PN;
-  do { ++BI; } while (isa<DbgInfoIntrinsic>(BI));
-  if (&*BI != RI)
-    return false;
+  // Make sure there are no instructions between the PHI and return, or that the
+  // return is the first instruction in the block.
+  if (PN) {
+    BasicBlock::iterator BI = BB->begin();
+    do { ++BI; } while (isa<DbgInfoIntrinsic>(BI));
+    if (&*BI != RI)
+      return false;
+  } else {
+    if (&*BB->begin() != RI)
+      return false;
+  }
 
   /// Only dup the ReturnInst if the CallInst is likely to be emitted as a tail
   /// call.
   SmallVector<CallInst*, 4> TailCalls;
-  for (unsigned I = 0, E = PN->getNumIncomingValues(); I != E; ++I) {
-    CallInst *CI = dyn_cast<CallInst>(PN->getIncomingValue(I));
-    // Make sure the phi value is indeed produced by the tail call.
-    if (CI && CI->hasOneUse() && CI->getParent() == PN->getIncomingBlock(I) &&
-        TLI->mayBeEmittedAsTailCall(CI))
-      TailCalls.push_back(CI);
+  if (PN) {
+    for (unsigned I = 0, E = PN->getNumIncomingValues(); I != E; ++I) {
+      CallInst *CI = dyn_cast<CallInst>(PN->getIncomingValue(I));
+      // Make sure the phi value is indeed produced by the tail call.
+      if (CI && CI->hasOneUse() && CI->getParent() == PN->getIncomingBlock(I) &&
+          TLI->mayBeEmittedAsTailCall(CI))
+        TailCalls.push_back(CI);
+    }
+  } else {
+    SmallPtrSet<BasicBlock*, 4> VisitedBBs;
+    for (pred_iterator PI = pred_begin(BB), PE = pred_end(BB); PI != PE; ++PI) {
+      if (!VisitedBBs.insert(*PI))
+        continue;
+
+      BasicBlock::InstListType &InstList = (*PI)->getInstList();
+      BasicBlock::InstListType::reverse_iterator RI = InstList.rbegin();
+      BasicBlock::InstListType::reverse_iterator RE = InstList.rend();
+      if (++RI == RE)
+        continue;
+      CallInst *CI = dyn_cast<CallInst>(&*RI);
+      if (CI && CI->getType()->isVoidTy() && TLI->mayBeEmittedAsTailCall(CI))
+        TailCalls.push_back(CI);
+    }
   }
 
   bool Changed = false;
diff --git a/test/CodeGen/X86/tailcall-returndup-void.ll b/test/CodeGen/X86/tailcall-returndup-void.ll
new file mode 100644
index 00000000000..c1d631225ec
--- /dev/null
+++ b/test/CodeGen/X86/tailcall-returndup-void.ll
@@ -0,0 +1,37 @@
+; RUN: llc < %s -march=x86-64 | FileCheck %s
+; CHECK: rBM_info
+; CHECK-NOT: ret
+
+@sES_closure = external global [0 x i64]
+declare cc10 void @sEH_info(i64* noalias nocapture, i64* noalias nocapture, i64* noalias nocapture, i64, i64, i64) align 8
+
+define cc10 void @rBM_info(i64* noalias nocapture %Base_Arg, i64* noalias nocapture %Sp_Arg, i64* noalias nocapture %Hp_Arg, i64 %R1_Arg, i64 %R2_Arg, i64 %R3_Arg) nounwind align 8 {
+c263:
+  %ln265 = getelementptr inbounds i64* %Sp_Arg, i64 -2
+  %ln266 = ptrtoint i64* %ln265 to i64
+  %ln268 = icmp ult i64 %ln266, %R3_Arg
+  br i1 %ln268, label %c26a, label %n26p
+
+n26p:                                             ; preds = %c263
+  br i1 icmp ne (i64 and (i64 ptrtoint ([0 x i64]* @sES_closure to i64), i64 7), i64 0), label %c1ZP.i, label %n1ZQ.i
+
+n1ZQ.i:                                           ; preds = %n26p
+  %ln1ZT.i = load i64* getelementptr inbounds ([0 x i64]* @sES_closure, i64 0, i64 0), align 8
+  %ln1ZU.i = inttoptr i64 %ln1ZT.i to void (i64*, i64*, i64*, i64, i64, i64)*
+  tail call cc10 void %ln1ZU.i(i64* %Base_Arg, i64* %Sp_Arg, i64* %Hp_Arg, i64 ptrtoint ([0 x i64]* @sES_closure to i64), i64 ptrtoint ([0 x i64]* @sES_closure to i64), i64 %R3_Arg) nounwind
+  br label %rBL_info.exit
+
+c1ZP.i:                                           ; preds = %n26p
+  tail call cc10 void @sEH_info(i64* %Base_Arg, i64* %Sp_Arg, i64* %Hp_Arg, i64 ptrtoint ([0 x i64]* @sES_closure to i64), i64 ptrtoint ([0 x i64]* @sES_closure to i64), i64 %R3_Arg) nounwind
+  br label %rBL_info.exit
+
+rBL_info.exit:                                    ; preds = %c1ZP.i, %n1ZQ.i
+  ret void
+
+c26a:                                             ; preds = %c263
+  %ln27h = getelementptr inbounds i64* %Base_Arg, i64 -2
+  %ln27j = load i64* %ln27h, align 8
+  %ln27k = inttoptr i64 %ln27j to void (i64*, i64*, i64*, i64, i64, i64)*
+  tail call cc10 void %ln27k(i64* %Base_Arg, i64* %Sp_Arg, i64* %Hp_Arg, i64 %R1_Arg, i64 %R2_Arg, i64 %R3_Arg) nounwind
+  ret void
+}