diff --git a/include/llvm/CodeGen/FastISel.h b/include/llvm/CodeGen/FastISel.h
index 5cb7d8a6f40..b9c1456c191 100644
--- a/include/llvm/CodeGen/FastISel.h
+++ b/include/llvm/CodeGen/FastISel.h
@@ -381,6 +381,10 @@ private:
 
   /// hasTrivialKill - Test whether the given value has exactly one use.
   bool hasTrivialKill(const Value *V) const;
+
+  /// removeDeadCode - Remove all dead instructions between the I and E.
+  void removeDeadCode(MachineBasicBlock::iterator I,
+                      MachineBasicBlock::iterator E);
 };
 
 }
diff --git a/lib/CodeGen/SelectionDAG/FastISel.cpp b/lib/CodeGen/SelectionDAG/FastISel.cpp
index 2087a1ba403..b4946ec5ee8 100644
--- a/lib/CodeGen/SelectionDAG/FastISel.cpp
+++ b/lib/CodeGen/SelectionDAG/FastISel.cpp
@@ -66,6 +66,7 @@ STATISTIC(NumFastIselSuccessIndependent, "Number of insts selected by "
           "target-independent selector");
 STATISTIC(NumFastIselSuccessTarget, "Number of insts selected by "
           "target-specific selector");
+STATISTIC(NumFastIselDead, "Number of dead insts removed on failure");
 
 /// startNewBlock - Set the current block to which generated machine
 /// instructions will be appended, and clear the local CSE map.
@@ -309,6 +310,18 @@ void FastISel::recomputeInsertPt() {
     ++FuncInfo.InsertPt;
 }
 
+void FastISel::removeDeadCode(MachineBasicBlock::iterator I,
+                              MachineBasicBlock::iterator E) {
+  assert (I && E && std::distance(I, E) > 0 && "Invalid iterator!");
+  while (I != E) {
+    MachineInstr *Dead = &*I;
+    ++I;
+    Dead->eraseFromParent();
+    ++NumFastIselDead;
+  }
+  recomputeInsertPt();
+}
+
 FastISel::SavePoint FastISel::enterLocalValueArea() {
   MachineBasicBlock::iterator OldInsertPt = FuncInfo.InsertPt;
   DebugLoc OldDL = DL;
@@ -794,19 +807,33 @@ FastISel::SelectInstruction(const Instruction *I) {
 
   DL = I->getDebugLoc();
 
+  MachineBasicBlock::iterator SavedInsertPt = FuncInfo.InsertPt;
+
   // First, try doing target-independent selection.
   if (SelectOperator(I, I->getOpcode())) {
     ++NumFastIselSuccessIndependent;
     DL = DebugLoc();
     return true;
   }
+  // Remove dead code.  However, ignore call instructions since we've flushed 
+  // the local value map and recomputed the insert point.
+  if (!isa<CallInst>(I)) {
+    recomputeInsertPt();
+    if (SavedInsertPt != FuncInfo.InsertPt)
+      removeDeadCode(FuncInfo.InsertPt, SavedInsertPt);
+  }
 
   // Next, try calling the target to attempt to handle the instruction.
+  SavedInsertPt = FuncInfo.InsertPt;
   if (TargetSelectInstruction(I)) {
     ++NumFastIselSuccessTarget;
     DL = DebugLoc();
     return true;
   }
+  // Check for dead code and remove as necessary.
+  recomputeInsertPt();
+  if (SavedInsertPt != FuncInfo.InsertPt)
+    removeDeadCode(FuncInfo.InsertPt, SavedInsertPt);
 
   DL = DebugLoc();
   return false;
diff --git a/test/CodeGen/ARM/fast-isel-deadcode.ll b/test/CodeGen/ARM/fast-isel-deadcode.ll
new file mode 100644
index 00000000000..028d94014fe
--- /dev/null
+++ b/test/CodeGen/ARM/fast-isel-deadcode.ll
@@ -0,0 +1,23 @@
+; RUN: llc < %s -O0 -relocation-model=dynamic-no-pic -mtriple=thumbv7-apple-darwin | FileCheck %s --check-prefix=THUMB
+
+; Target-specific selector can't properly handle the double because it isn't
+; being passed via a register, so the materialized arguments become dead code.
+
+define i32 @main(i32 %argc, i8** %argv) nounwind {
+entry:
+; THUMB: main
+  call void @printArgsNoRet(i32 1, float 0x4000CCCCC0000000, i8 signext 99, double 4.100000e+00)
+; THUMB: blx _printArgsNoRet
+; THUMB-NOT: ldr
+; THUMB-NOT: vldr
+; THUMB-NOT: vmov
+; THUMB-NOT: ldr
+; THUMB-NOT: sxtb
+; THUMB: movs r0, #0
+; THUMB: movt r0, #0
+; THUMB: add sp, #32
+; THUMb: pop {r7, pc}
+  ret i32 0
+}
+
+declare void @printArgsNoRet(i32 %a1, float %a2, i8 signext %a3, double %a4)