diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 29033e5117f..2112992dd86 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -712,11 +712,12 @@ void ARMBaseInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
       unsigned Dst = TRI->getSubReg(DestReg, BeginIdx + i*Spacing);
       unsigned Src = TRI->getSubReg(SrcReg,  BeginIdx + i*Spacing);
       assert(Dst && Src && "Bad sub-register");
-      Mov = AddDefaultPred(BuildMI(MBB, I, I->getDebugLoc(), get(Opc), Dst)
-                             .addReg(Src));
+      Mov = BuildMI(MBB, I, I->getDebugLoc(), get(Opc), Dst)
+        .addReg(Src);
       // VORR takes two source operands.
       if (Opc == ARM::VORRq)
         Mov.addReg(Src);
+      Mov = AddDefaultPred(Mov);
     }
     // Add implicit super-register defs and kills to the last instruction.
     Mov->addRegisterDefined(DestReg, TRI);
diff --git a/test/CodeGen/ARM/2012-08-27-CopyPhysRegCrash.ll b/test/CodeGen/ARM/2012-08-27-CopyPhysRegCrash.ll
new file mode 100644
index 00000000000..ec7f72d7c2e
--- /dev/null
+++ b/test/CodeGen/ARM/2012-08-27-CopyPhysRegCrash.ll
@@ -0,0 +1,129 @@
+; RUN: llc < %s -mcpu=cortex-a8 -march=thumb
+; Test that this doesn't crash.
+; <rdar://problem/12183003>
+
+target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:32:64-v128:32:128-a0:0:32-n32-S32"
+target triple = "thumbv7-apple-ios5.1.0"
+
+declare { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld3.v16i8(i8*, i32) nounwind readonly
+
+declare void @llvm.arm.neon.vst1.v16i8(i8*, <16 x i8>, i32) nounwind
+
+define void @findEdges(i8*) nounwind ssp {
+  %2 = icmp sgt i32 undef, 0
+  br i1 %2, label %5, label %3
+
+; <label>:3                                       ; preds = %5, %1
+  %4 = phi i8* [ %0, %1 ], [ %19, %5 ]
+  ret void
+
+; <label>:5                                       ; preds = %5, %1
+  %6 = phi i8* [ %19, %5 ], [ %0, %1 ]
+  %7 = tail call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld3.v16i8(i8* null, i32 1)
+  %8 = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %7, 0
+  %9 = getelementptr inbounds i8* null, i32 3
+  %10 = tail call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld3.v16i8(i8* %9, i32 1)
+  %11 = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %10, 2
+  %12 = tail call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld3.v16i8(i8* %6, i32 1)
+  %13 = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %12, 0
+  %14 = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %12, 1
+  %15 = getelementptr inbounds i8* %6, i32 3
+  %16 = tail call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld3.v16i8(i8* %15, i32 1)
+  %17 = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %16, 1
+  %18 = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %16, 2
+  %19 = getelementptr inbounds i8* %6, i32 48
+  %20 = bitcast <16 x i8> %13 to <2 x i64>
+  %21 = bitcast <16 x i8> %8 to <2 x i64>
+  %22 = bitcast <16 x i8> %14 to <2 x i64>
+  %23 = shufflevector <2 x i64> %22, <2 x i64> undef, <1 x i32> zeroinitializer
+  %24 = bitcast <1 x i64> %23 to <8 x i8>
+  %25 = zext <8 x i8> %24 to <8 x i16>
+  %26 = sub <8 x i16> zeroinitializer, %25
+  %27 = bitcast <16 x i8> %17 to <2 x i64>
+  %28 = tail call <8 x i16> @llvm.arm.neon.vabs.v8i16(<8 x i16> %26) nounwind
+  %29 = mul <8 x i16> %28, %28
+  %30 = add <8 x i16> zeroinitializer, %29
+  %31 = tail call <8 x i16> @llvm.arm.neon.vmaxu.v8i16(<8 x i16> undef, <8 x i16> %30) nounwind
+  %32 = bitcast <16 x i8> %11 to <2 x i64>
+  %33 = shufflevector <2 x i64> %32, <2 x i64> undef, <1 x i32> zeroinitializer
+  %34 = bitcast <1 x i64> %33 to <8 x i8>
+  %35 = zext <8 x i8> %34 to <8 x i16>
+  %36 = sub <8 x i16> %35, zeroinitializer
+  %37 = bitcast <16 x i8> %18 to <2 x i64>
+  %38 = shufflevector <2 x i64> %37, <2 x i64> undef, <1 x i32> zeroinitializer
+  %39 = bitcast <1 x i64> %38 to <8 x i8>
+  %40 = zext <8 x i8> %39 to <8 x i16>
+  %41 = sub <8 x i16> zeroinitializer, %40
+  %42 = tail call <8 x i16> @llvm.arm.neon.vabs.v8i16(<8 x i16> %36) nounwind
+  %43 = tail call <8 x i16> @llvm.arm.neon.vabs.v8i16(<8 x i16> %41) nounwind
+  %44 = mul <8 x i16> %42, %42
+  %45 = mul <8 x i16> %43, %43
+  %46 = add <8 x i16> %45, %44
+  %47 = tail call <8 x i16> @llvm.arm.neon.vmaxu.v8i16(<8 x i16> %31, <8 x i16> %46) nounwind
+  %48 = bitcast <8 x i16> %47 to <2 x i64>
+  %49 = shufflevector <2 x i64> %48, <2 x i64> undef, <1 x i32> zeroinitializer
+  %50 = bitcast <1 x i64> %49 to <4 x i16>
+  %51 = tail call <4 x i32> @llvm.arm.neon.vmullu.v4i32(<4 x i16> %50, <4 x i16> undef) nounwind
+  %52 = tail call <4 x i16> @llvm.arm.neon.vqshiftnu.v4i16(<4 x i32> %51, <4 x i32> <i32 -6, i32 -6, i32 -6, i32 -6>)
+  %53 = bitcast <4 x i16> %52 to <1 x i64>
+  %54 = shufflevector <1 x i64> %53, <1 x i64> undef, <2 x i32> <i32 0, i32 1>
+  %55 = bitcast <2 x i64> %54 to <8 x i16>
+  %56 = tail call <8 x i8> @llvm.arm.neon.vshiftn.v8i8(<8 x i16> %55, <8 x i16> <i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8>)
+  %57 = shufflevector <2 x i64> %20, <2 x i64> undef, <1 x i32> <i32 1>
+  %58 = bitcast <1 x i64> %57 to <8 x i8>
+  %59 = zext <8 x i8> %58 to <8 x i16>
+  %60 = sub <8 x i16> zeroinitializer, %59
+  %61 = shufflevector <2 x i64> %21, <2 x i64> undef, <1 x i32> <i32 1>
+  %62 = bitcast <1 x i64> %61 to <8 x i8>
+  %63 = zext <8 x i8> %62 to <8 x i16>
+  %64 = sub <8 x i16> %63, zeroinitializer
+  %65 = tail call <8 x i16> @llvm.arm.neon.vabs.v8i16(<8 x i16> %60) nounwind
+  %66 = mul <8 x i16> %65, %65
+  %67 = add <8 x i16> zeroinitializer, %66
+  %68 = shufflevector <2 x i64> %27, <2 x i64> undef, <1 x i32> <i32 1>
+  %69 = bitcast <1 x i64> %68 to <8 x i8>
+  %70 = zext <8 x i8> %69 to <8 x i16>
+  %71 = sub <8 x i16> zeroinitializer, %70
+  %72 = tail call <8 x i16> @llvm.arm.neon.vabs.v8i16(<8 x i16> undef) nounwind
+  %73 = tail call <8 x i16> @llvm.arm.neon.vabs.v8i16(<8 x i16> %71) nounwind
+  %74 = mul <8 x i16> %72, %72
+  %75 = mul <8 x i16> %73, %73
+  %76 = add <8 x i16> %75, %74
+  %77 = shufflevector <2 x i64> %32, <2 x i64> undef, <1 x i32> <i32 1>
+  %78 = bitcast <1 x i64> %77 to <8 x i8>
+  %79 = zext <8 x i8> %78 to <8 x i16>
+  %80 = sub <8 x i16> %79, zeroinitializer
+  %81 = tail call <8 x i16> @llvm.arm.neon.vabs.v8i16(<8 x i16> %80) nounwind
+  %82 = mul <8 x i16> %81, %81
+  %83 = add <8 x i16> zeroinitializer, %82
+  %84 = tail call <8 x i16> @llvm.arm.neon.vmaxu.v8i16(<8 x i16> %76, <8 x i16> %83) nounwind
+  %85 = tail call <8 x i16> @llvm.arm.neon.vmaxu.v8i16(<8 x i16> %67, <8 x i16> %84) nounwind
+  %86 = bitcast <8 x i16> %85 to <2 x i64>
+  %87 = shufflevector <2 x i64> %86, <2 x i64> undef, <1 x i32> <i32 1>
+  %88 = bitcast <1 x i64> %87 to <4 x i16>
+  %89 = tail call <4 x i32> @llvm.arm.neon.vmullu.v4i32(<4 x i16> %88, <4 x i16> undef) nounwind
+  %90 = tail call <4 x i16> @llvm.arm.neon.vqrshiftnu.v4i16(<4 x i32> %89, <4 x i32> <i32 -6, i32 -6, i32 -6, i32 -6>)
+  %91 = bitcast <4 x i16> %90 to <1 x i64>
+  %92 = shufflevector <1 x i64> undef, <1 x i64> %91, <2 x i32> <i32 0, i32 1>
+  %93 = bitcast <2 x i64> %92 to <8 x i16>
+  %94 = tail call <8 x i8> @llvm.arm.neon.vshiftn.v8i8(<8 x i16> %93, <8 x i16> <i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8>)
+  %95 = bitcast <8 x i8> %56 to <1 x i64>
+  %96 = bitcast <8 x i8> %94 to <1 x i64>
+  %97 = shufflevector <1 x i64> %95, <1 x i64> %96, <2 x i32> <i32 0, i32 1>
+  %98 = bitcast <2 x i64> %97 to <16 x i8>
+  tail call void @llvm.arm.neon.vst1.v16i8(i8* null, <16 x i8> %98, i32 1)
+  %99 = icmp slt i32 undef, undef
+  br i1 %99, label %5, label %3
+}
+
+declare <4 x i16> @llvm.arm.neon.vqshiftnu.v4i16(<4 x i32>, <4 x i32>) nounwind readnone
+
+declare <8 x i8> @llvm.arm.neon.vshiftn.v8i8(<8 x i16>, <8 x i16>) nounwind readnone
+
+declare <4 x i16> @llvm.arm.neon.vqrshiftnu.v4i16(<4 x i32>, <4 x i32>) nounwind readnone
+
+declare <4 x i32> @llvm.arm.neon.vmullu.v4i32(<4 x i16>, <4 x i16>) nounwind readnone
+
+declare <8 x i16> @llvm.arm.neon.vmaxu.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
+
+declare <8 x i16> @llvm.arm.neon.vabs.v8i16(<8 x i16>) nounwind readnone