[CodeGen][ObjC] Annotate calls to objc_retainAutoreleasedReturnValue with notail on x86-64

[ahatanak]

Summary

On x86-64, the epilogue code inserted before the tail jump blocks the autoreleased return optimization.

ARC optimizer unconditionally turns calls to objc_retainAutoreleasedReturnValue to tail calls. I plan to fix that too after this patch lands.

rdar://problem/38675807

[ahatanak]

There are existing test cases that test notail isn't added to calls on targets that aren't x86-64 (for example, test/CodeGenObjC/arc-arm.m), so I didn't write a new one.

[rjmccall]

lib/CodeGen/TargetInfo.h

161

shouldSuppressTailCallsOfRetainAutoreleasedReturnValue()?

[ahatanak]

Rename function.

[rjmccall]

LGTM.

lib/CodeGen/CGObjC.cpp

 static void setARCRuntimeFunctionLinkage(CodeGenModule &CGM,
                                          llvm::FunctionCallee RTF) {
   setARCRuntimeFunctionLinkage(CGM, RTF.getCallee());
 }
 
 /// Perform an operation having the signature
 ///   i8* (i8*)
 /// where a null input causes a no-op and returns null.
-static llvm::Value *
-emitARCValueOperation(CodeGenFunction &CGF, llvm::Value *value,
-                      llvm::Type *returnType, llvm::Function *&fn,
-                      llvm::Intrinsic::ID IntID, bool isTailCall = false) {
+static llvm::Value *emitARCValueOperation(
+    CodeGenFunction &CGF, llvm::Value *value, llvm::Type *returnType,
+    llvm::Function *&fn, llvm::Intrinsic::ID IntID,
+    llvm::CallInst::TailCallKind tailKind = llvm::CallInst::TCK_None) {
   if (isa<llvm::ConstantPointerNull>(value))
     return value;
 
   if (!fn) {
     fn = CGF.CGM.getIntrinsic(IntID);
     setARCRuntimeFunctionLinkage(CGF.CGM, fn);
   }
 
   // Cast the argument to 'id'.
   llvm::Type *origType = returnType ? returnType : value->getType();
   value = CGF.Builder.CreateBitCast(value, CGF.Int8PtrTy);
 
   // Call the function.
   llvm::CallInst *call = CGF.EmitNounwindRuntimeCall(fn, value);
-  if (isTailCall)
-    call->setTailCall();
+  call->setTailCallKind(tailKind);
 
   // Cast the result back to the original type.
   return CGF.Builder.CreateBitCast(call, origType);
 }
 
 /// Perform an operation having the following signature:
 ///   i8* (i8**)
 static llvm::Value *emitARCLoadOperation(CodeGenFunction &CGF, Address addr,

 /// Retain the given object which is the result of a function call.
 ///   call i8* \@objc_retainAutoreleasedReturnValue(i8* %value)
 ///
 /// Yes, this function name is one character away from a different
 /// call with completely different semantics.
 llvm::Value *
 CodeGenFunction::EmitARCRetainAutoreleasedReturnValue(llvm::Value *value) {
   emitAutoreleasedReturnValueMarker(*this);
-  return emitARCValueOperation(*this, value, nullptr,
-              CGM.getObjCEntrypoints().objc_retainAutoreleasedReturnValue,
-                           llvm::Intrinsic::objc_retainAutoreleasedReturnValue);
+  llvm::CallInst::TailCallKind tailKind =
+      CGM.getTargetCodeGenInfo()
+              .shouldSuppressTailCallsOfRetainAutoreleasedReturnValue()
+          ? llvm::CallInst::TCK_NoTail
+          : llvm::CallInst::TCK_None;
+  return emitARCValueOperation(
+      *this, value, nullptr,
+      llvm::Intrinsic::objc_retainAutoreleasedReturnValue, tailKind);
 }
 
 /// Claim a possibly-autoreleased return value at +0.  This is only
 /// valid to do in contexts which do not rely on the retain to keep
 /// the object valid for all of its uses; for example, when
 /// the value is ignored, or when it is being assigned to an
 /// __unsafe_unretained variable.
 ///

 
 /// Autorelease the given object.
 ///   call i8* \@objc_autoreleaseReturnValue(i8* %value)
 llvm::Value *
 CodeGenFunction::EmitARCAutoreleaseReturnValue(llvm::Value *value) {
   return emitARCValueOperation(*this, value, nullptr,
                             CGM.getObjCEntrypoints().objc_autoreleaseReturnValue,
                                llvm::Intrinsic::objc_autoreleaseReturnValue,
-                               /*isTailCall*/ true);
+                               llvm::CallInst::TCK_Tail);
 }
 
 /// Do a fused retain/autorelease of the given object.
 ///   call i8* \@objc_retainAutoreleaseReturnValue(i8* %value)
 llvm::Value *
 CodeGenFunction::EmitARCRetainAutoreleaseReturnValue(llvm::Value *value) {
   return emitARCValueOperation(*this, value, nullptr,
                      CGM.getObjCEntrypoints().objc_retainAutoreleaseReturnValue,
                              llvm::Intrinsic::objc_retainAutoreleaseReturnValue,
-                               /*isTailCall*/ true);
+                               llvm::CallInst::TCK_Tail);
 }
 
 /// Do a fused retain/autorelease of the given object.
 ///   call i8* \@objc_retainAutorelease(i8* %value)
 /// or
 ///   %retain = call i8* \@objc_retainBlock(i8* %value)
 ///   call i8* \@objc_autorelease(i8* %retain)
 llvm::Value *CodeGenFunction::EmitARCRetainAutorelease(QualType type,

lib/CodeGen/TargetInfo.h

public:
   /// relocation model, and so on some targets it instead sniffs for
   /// a particular instruction sequence.  This functions returns
   /// that instruction sequence in inline assembly, which will be
   /// empty if none is required.
   virtual StringRef getARCRetainAutoreleasedReturnValueMarker() const {
     return "";
   }
 
+  /// Determine whether a call to objc_retainAutoreleasedReturnValue should be
+  /// marked as 'notail'.
+  virtual bool shouldSuppressTailCallsOfRetainAutoreleasedReturnValue() const {
+    return false;
+  }
+
   /// Return a constant used by UBSan as a signature to identify functions
   /// possessing type information, or 0 if the platform is unsupported.
   virtual llvm::Constant *
   getUBSanFunctionSignature(CodeGen::CodeGenModule &CGM) const {
     return nullptr;
   }
 
   /// Determine whether a call to an unprototyped functions under

lib/CodeGen/TargetInfo.cpp

 public:
   X86_64TargetCodeGenInfo(CodeGen::CodeGenTypes &CGT, X86AVXABILevel AVXLevel)
       : TargetCodeGenInfo(new X86_64ABIInfo(CGT, AVXLevel)) {}
 
   const X86_64ABIInfo &getABIInfo() const {
     return static_cast<const X86_64ABIInfo&>(TargetCodeGenInfo::getABIInfo());
   }
 
+  /// Disable tail call on x86-64. The epilogue code before the tail jump blocks
+  /// the autoreleaseRV/retainRV optimization.
+  bool shouldSuppressTailCallsOfRetainAutoreleasedReturnValue() const override {
+    return true;
+  }
+
   int getDwarfEHStackPointer(CodeGen::CodeGenModule &CGM) const override {
     return 7;
   }
 
   bool initDwarfEHRegSizeTable(CodeGen::CodeGenFunction &CGF,
                                llvm::Value *Address) const override {
     llvm::Value *Eight8 = llvm::ConstantInt::get(CGF.Int8Ty, 8);
 

test/CodeGenObjC/arc-blocks.m

void test4(void) {
   // CHECK-LABEL:    define void @test4()
   // CHECK:      [[VAR:%.*]] = alloca [[BYREF_T:%.*]],
   // CHECK-NEXT: [[BLOCK:%.*]] = alloca [[BLOCK_T:<{.*}>]],
   // CHECK:      [[T0:%.*]] = getelementptr inbounds [[BYREF_T]], [[BYREF_T]]* [[VAR]], i32 0, i32 2
   // 0x02000000 - has copy/dispose helpers strong
   // CHECK-NEXT: store i32 838860800, i32* [[T0]]
   // CHECK:      [[SLOT:%.*]] = getelementptr inbounds [[BYREF_T]], [[BYREF_T]]* [[VAR]], i32 0, i32 6
   // CHECK-NEXT: [[T0:%.*]] = call i8* @test4_source()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: store i8* [[T1]], i8** [[SLOT]]
   // CHECK-NEXT: [[SLOT:%.*]] = getelementptr inbounds [[BYREF_T]], [[BYREF_T]]* [[VAR]], i32 0, i32 6
   // 0x42800000 - has signature, copy/dispose helpers, as well as BLOCK_HAS_EXTENDED_LAYOUT
   // CHECK:      store i32 -1040187392,
   // CHECK:      [[T0:%.*]] = bitcast [[BYREF_T]]* [[VAR]] to i8*
   // CHECK-NEXT: store i8* [[T0]], i8**
   // CHECK:      call void @test4_helper(
   // CHECK:      [[T0:%.*]] = bitcast [[BYREF_T]]* [[VAR]] to i8*
void test5(void) {
   test5_helper(^{ (void) var; });
 
   // CHECK-LABEL:    define void @test5()
   // CHECK:      [[VAR:%.*]] = alloca i8*
   // CHECK-NEXT: [[BLOCK:%.*]] = alloca [[BLOCK_T:<{.*}>]],
   // CHECK-NEXT: [[VARPTR1:%.*]] = bitcast i8** [[VAR]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[VARPTR1]])
   // CHECK: [[T0:%.*]] = call i8* @test5_source()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: store i8* [[T1]], i8** [[VAR]],
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   // 0x40800000 - has signature but no copy/dispose, as well as BLOCK_HAS_EXTENDED_LAYOUT
   // CHECK:      store i32 -1073741824, i32*
   // CHECK:      [[CAPTURE:%.*]] = getelementptr inbounds [[BLOCK_T]], [[BLOCK_T]]* [[BLOCK]], i32 0, i32 5
   // CHECK-NEXT: [[T0:%.*]] = load i8*, i8** [[VAR]]
   // CHECK-NEXT: store i8* [[T0]], i8** [[CAPTURE]]
   // CHECK-NEXT: [[T0:%.*]] = bitcast [[BLOCK_T]]* [[BLOCK]] to
void test6(void) {
   // CHECK-NEXT: [[BLOCK:%.*]] = alloca [[BLOCK_T:<{.*}>]],
   // CHECK-NEXT: [[VARPTR1:%.*]] = bitcast [[BYREF_T]]* [[VAR]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 48, i8* [[VARPTR1]])
   // CHECK:      [[T0:%.*]] = getelementptr inbounds [[BYREF_T]], [[BYREF_T]]* [[VAR]], i32 0, i32 2
   // 0x02000000 - has copy/dispose helpers weak
   // CHECK-NEXT: store i32 1107296256, i32* [[T0]]
   // CHECK:      [[SLOT:%.*]] = getelementptr inbounds [[BYREF_T]], [[BYREF_T]]* [[VAR]], i32 0, i32 6
   // CHECK-NEXT: [[T0:%.*]] = call i8* @test6_source()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: call i8* @llvm.objc.initWeak(i8** [[SLOT]], i8* [[T1]])
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   // CHECK-NEXT: [[SLOT:%.*]] = getelementptr inbounds [[BYREF_T]], [[BYREF_T]]* [[VAR]], i32 0, i32 6
   // 0x42800000 - has signature, copy/dispose helpers, as well as BLOCK_HAS_EXTENDED_LAYOUT
   // CHECK:      store i32 -1040187392,
   // CHECK: %[[BLOCK_DESCRIPTOR:.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %[[STRUCT_BLOCK_DESCRIPTOR]]*, i8* }>, <{ i8*, i32, i32, i8*, %[[STRUCT_BLOCK_DESCRIPTOR]]*, i8* }>* %{{.*}}, i32 0, i32 4
 // CHECK: store %[[STRUCT_BLOCK_DESCRIPTOR]]* bitcast ({ i64, i64, i8*, i8*, i8*, i64 }* @[[BLOCK_DESCRIPTOR_TMP9]] to %[[STRUCT_BLOCK_DESCRIPTOR]]*), %[[STRUCT_BLOCK_DESCRIPTOR]]** %[[BLOCK_DESCRIPTOR]], align 8
   // CHECK:      [[T0:%.*]] = bitcast [[BYREF_T]]* [[VAR]] to i8*
void test7(void) {
   void test7_consume(id);
   __weak id var = test7_source();
   test7_helper(^{ test7_consume(var); });
 
   // CHECK-LABEL:    define void @test7()
   // CHECK:      [[VAR:%.*]] = alloca i8*,
   // CHECK-NEXT: [[BLOCK:%.*]] = alloca [[BLOCK_T:<{.*}>]],
   // CHECK:      [[T0:%.*]] = call i8* @test7_source()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: call i8* @llvm.objc.initWeak(i8** [[VAR]], i8* [[T1]])
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   // 0x42800000 - has signature, copy/dispose helpers, as well as BLOCK_HAS_EXTENDED_LAYOUT
   // CHECK:      store i32 -1040187392,
   // CHECK:      [[SLOT:%.*]] = getelementptr inbounds [[BLOCK_T]], [[BLOCK_T]]* [[BLOCK]], i32 0, i32 5
   // CHECK-NEXT: call void @llvm.objc.copyWeak(i8** [[SLOT]], i8** [[VAR]])
   // CHECK:      call void @test7_helper(
   // CHECK-NEXT: call void @llvm.objc.destroyWeak(i8** {{%.*}})

test/CodeGenObjC/arc-foreach.m

void test2(Test2 *a) {
   for (id x in a.array) {
     use(x);
   }
 }
 
 // CHECK-LP64-LABEL:    define void @test2(
 // CHECK-LP64:      [[T0:%.*]] = call [[ARRAY_T]]* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to [[ARRAY_T]]* (i8*, i8*)*)(
 // CHECK-LP64-NEXT: [[T1:%.*]] = bitcast [[ARRAY_T]]* [[T0]] to i8*
-// CHECK-LP64-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+// CHECK-LP64-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 // CHECK-LP64-NEXT: [[COLL:%.*]] = bitcast i8* [[T2]] to [[ARRAY_T]]*
 
 // Make sure it's not immediately released before starting the iteration.
 // CHECK-LP64-NEXT: load i8*, i8** @OBJC_SELECTOR_REFERENCES_
 // CHECK-LP64-NEXT: [[T0:%.*]] = bitcast [[ARRAY_T]]* [[COLL]] to i8*
 // CHECK-LP64-NEXT: @objc_msgSend
 
 // This bitcast is for the mutation check.

test/CodeGenObjC/arc-literals.m

void test_array(id a, id b) {
   // CHECK-NEXT: [[V1:%.*]] = load i8*, i8** [[B]],
   // CHECK-NEXT: store i8* [[V1]], i8** [[T0]]
 
   // CHECK-NEXT: [[T0:%.*]] = load [[CLASS_T:%.*]]*, [[CLASS_T:%.*]]** @"OBJC_CLASSLIST
   // CHECK-NEXT: [[SEL:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[CLASS_T]]* [[T0]] to i8*
   // CHECK-NEXT: [[T2:%.*]] = bitcast [2 x i8*]* [[OBJECTS]] to i8**
   // CHECK-NEXT: [[T3:%.*]] = call i8* bitcast ({{.*@objc_msgSend.*}})(i8* [[T1]], i8* [[SEL]], i8** [[T2]], i64 2)
-  // CHECK-NEXT: [[T4:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T3]])
+  // CHECK-NEXT: [[T4:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T3]])
   // CHECK: call void (...) @llvm.objc.clang.arc.use(i8* [[V0]], i8* [[V1]])
   id arr = @[a, b];
 
   // CHECK: call void @llvm.objc.release
   // CHECK: call void @llvm.objc.release
   // CHECK: call void @llvm.objc.release
   // CHECK-NEXT: ret void
 }
void test_dictionary(id k1, id o1, id k2, id o2) {
 
   // Constructing the dictionary
   // CHECK-NEXT: [[T0:%.*]] = load [[CLASS_T:%.*]]*, [[CLASS_T:%.*]]** @"OBJC_CLASSLIST
   // CHECK-NEXT: [[SEL:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[CLASS_T]]* [[T0]] to i8*
   // CHECK-NEXT: [[T2:%.*]] = bitcast [2 x i8*]* [[OBJECTS]] to i8**
   // CHECK-NEXT: [[T3:%.*]] = bitcast [2 x i8*]* [[KEYS]] to i8**
   // CHECK-NEXT: [[T4:%.*]] = call i8* bitcast ({{.*@objc_msgSend.*}})(i8* [[T1]], i8* [[SEL]], i8** [[T2]], i8** [[T3]], i64 2)
-  // CHECK-NEXT: [[T5:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
+  // CHECK-NEXT: [[T5:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
   // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.use(i8* [[V0]], i8* [[V1]], i8* [[V2]], i8* [[V3]])
 
   id dict = @{ k1 : o1, k2 : o2 };
 
   // CHECK: call void @llvm.objc.release
   // CHECK: call void @llvm.objc.release
   // CHECK: call void @llvm.objc.release
   // CHECK: call void @llvm.objc.release
void test_property(B *b) {
 
   // CHECK:      [[T0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[OBJECTS:%.*]], i64 0, i64 0
 
   // Invoke 'prop'
   // CHECK:      [[SEL:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES
   // CHECK-NEXT: [[T1:%.*]] = bitcast
   // CHECK-NEXT: [[T2:%.*]] = call [[B:%.*]]* bitcast ({{.*}} @objc_msgSend to {{.*}})(i8* [[T1]], i8* [[SEL]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast [[B]]* [[T2]] to i8*
-  // CHECK-NEXT: [[T4:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T3]])
+  // CHECK-NEXT: [[T4:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T3]])
   // CHECK-NEXT: [[V0:%.*]] = bitcast i8* [[T4]] to [[B]]*
   // CHECK-NEXT: [[V1:%.*]] = bitcast [[B]]* [[V0]] to i8*
 
   // Store to array.
   // CHECK-NEXT: store i8* [[V1]], i8** [[T0]]
 
   // Invoke arrayWithObjects:count:
   // CHECK-NEXT: [[T0:%.*]] = load [[CLASS_T]]*, [[CLASS_T]]** @"OBJC_CLASSLIST

test/CodeGenObjC/arc-precise-lifetime.m

 // CHECK-LABEL: define void @test1a_message()
 void test1a_message(void) {
   // CHECK:      [[PTR:%.*]] = alloca [[PTR_T:%.*]]*, align 8
   // CHECK:      [[C:%.*]] = alloca i8*, align 8
   // CHECK:      [[PTRPTR1:%.*]] = bitcast [[PTR_T]]** [[PTR]] to i8*
   // CHECK:      call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTRPTR1]])
   // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper()
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
   // CHECK-NEXT: store [[TEST1]]* [[T3]]
   // CHECK-NEXT: [[CPTR1:%.*]] = bitcast i8** [[C]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[CPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = load [[TEST1]]*, [[TEST1]]**
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
   // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutorelease(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*

 // CHECK-LABEL: define void @test1a_property()
 void test1a_property(void) {
   // CHECK:      [[PTR:%.*]] = alloca [[PTR_T:%.*]]*, align 8
   // CHECK:      [[C:%.*]] = alloca i8*, align 8
   // CHECK:      [[PTRPTR1:%.*]] = bitcast [[PTR_T]]** [[PTR]] to i8*
   // CHECK:      call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTRPTR1]])
   // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper()
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
   // CHECK-NEXT: store [[TEST1]]* [[T3]]
   // CHECK-NEXT: [[CPTR1:%.*]] = bitcast i8** [[C]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[CPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = load [[TEST1]]*, [[TEST1]]**
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
   // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutorelease(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*

 // CHECK-LABEL: define void @test1b_message()
 void test1b_message(void) {
   // CHECK:      [[PTR:%.*]] = alloca [[PTR_T:%.*]]*, align 8
   // CHECK:      [[C:%.*]] = alloca i8*, align 8
   // CHECK:      [[PTRPTR1:%.*]] = bitcast [[PTR_T]]** [[PTR]] to i8*
   // CHECK:      call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTRPTR1]])
   // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper()
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
   // CHECK-NEXT: store [[TEST1]]* [[T3]]
   // CHECK-NEXT: [[CPTR1:%.*]] = bitcast i8** [[C]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[CPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = load [[TEST1]]*, [[TEST1]]**
   // CHECK-NEXT: [[T1:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES_
   // CHECK-NEXT: [[T2:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
   // CHECK-NEXT: [[T3:%.*]] = call i8* bitcast

 // CHECK-LABEL: define void @test1b_property()
 void test1b_property(void) {
   // CHECK:      [[PTR:%.*]] = alloca [[PTR_T:%.*]]*, align 8
   // CHECK:      [[C:%.*]] = alloca i8*, align 8
   // CHECK:      [[PTRPTR1:%.*]] = bitcast [[PTR_T]]** [[PTR]] to i8*
   // CHECK:      call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTRPTR1]])
   // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper()
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
   // CHECK-NEXT: store [[TEST1]]* [[T3]]
   // CHECK-NEXT: [[CPTR1:%.*]] = bitcast i8** [[C]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[CPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = load [[TEST1]]*, [[TEST1]]**
   // CHECK-NEXT: [[T1:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES_
   // CHECK-NEXT: [[T2:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
   // CHECK-NEXT: [[T3:%.*]] = call i8* bitcast

 // CHECK-LABEL: define void @test1c_message()
 void test1c_message(void) {
   // CHECK:      [[PTR:%.*]] = alloca [[PTR_T:%.*]]*, align 8
   // CHECK:      [[PC:%.*]] = alloca i8*, align 8
   // CHECK:      [[PTRPTR1:%.*]] = bitcast [[PTR_T]]** [[PTR]] to i8*
   // CHECK:      call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTRPTR1]])
   // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper()
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
   // CHECK-NEXT: store [[TEST1]]* [[T3]]
   // CHECK-NEXT: [[PCPTR1:%.*]] = bitcast i8** [[PC]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PCPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = load [[TEST1]]*, [[TEST1]]**
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
   // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutorelease(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*

 // CHECK-LABEL: define void @test1c_property()
 void test1c_property(void) {
   // CHECK:      [[PTR:%.*]] = alloca [[PTR_T:%.*]]*, align 8
   // CHECK:      [[PC:%.*]] = alloca i8*, align 8
   // CHECK:      [[PTRPTR1:%.*]] = bitcast [[PTR_T]]** [[PTR]] to i8*
   // CHECK:      call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTRPTR1]])
   // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper()
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
   // CHECK-NEXT: store [[TEST1]]* [[T3]]
   // CHECK-NEXT: [[PCPTR1:%.*]] = bitcast i8** [[PC]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PCPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = load [[TEST1]]*, [[TEST1]]**
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
   // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutorelease(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*

 // CHECK-LABEL: define void @test1d_message()
 void test1d_message(void) {
   // CHECK:      [[PTR:%.*]] = alloca [[PTR_T:%.*]]*, align 8
   // CHECK:      [[PC:%.*]] = alloca i8*, align 8
   // CHECK:      [[PTRPTR1:%.*]] = bitcast [[PTR_T]]** [[PTR]] to i8*
   // CHECK:      call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTRPTR1]])
   // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper()
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
   // CHECK-NEXT: store [[TEST1]]* [[T3]]
   // CHECK-NEXT: [[PCPTR1:%.*]] = bitcast i8** [[PC]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PCPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = load [[TEST1]]*, [[TEST1]]**
   // CHECK-NEXT: [[SEVEN:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES_
   // CHECK-NEXT: [[EIGHT:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
   // CHECK-NEXT: [[CALL1:%.*]] = call i8* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to i8* (i8*, i8*)*)(i8* [[EIGHT]], i8* [[SEVEN]])

 // CHECK-LABEL: define void @test1d_property()
 void test1d_property(void) {
   // CHECK:      [[PTR:%.*]] = alloca [[PTR_T:%.*]]*, align 8
   // CHECK:      [[PC:%.*]] = alloca i8*, align 8
   // CHECK:      [[PTRPTR1:%.*]] = bitcast [[PTR_T]]** [[PTR]] to i8*
   // CHECK:      call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTRPTR1]])
   // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper()
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
   // CHECK-NEXT: store [[TEST1]]* [[T3]]
   // CHECK-NEXT: [[PCPTR1:%.*]] = bitcast i8** [[PC]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PCPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = load [[TEST1]]*, [[TEST1]]**
   // CHECK-NEXT: [[SEVEN:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES_
   // CHECK-NEXT: [[EIGHT:%.*]] = bitcast [[TEST1]]* [[T0]] to i8*
   // CHECK-NEXT: [[CALL1:%.*]] = call i8* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to i8* (i8*, i8*)*)(i8* [[EIGHT]], i8* [[SEVEN]])

test/CodeGenObjC/arc-property.m

 
 @implementation Test3
 - (id) copyMachine {
   extern id test3_helper(void);
   return test3_helper();
 }
 // CHECK:    define internal i8* @"\01-[Test3 copyMachine]"(
 // CHECK:      [[T0:%.*]] = call i8* @test3_helper()
-// CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+// CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
 // CHECK-NEXT: ret i8* [[T1]]
 - (void) setCopyMachine: (id) x {}
 @end
 
 // rdar://31579994
 // When synthesizing a property that's declared in multiple protocols, ensure
 // that the setter is emitted if any of these declarations is readwrite.
 @protocol ABC

test/CodeGenObjC/arc-related-result-type.m

 // CHECK-NEXT: store [[TEST0]]* null
 // CHECK-NEXT: bitcast
 // CHECK-NEXT: bitcast
 // CHECK-NEXT: call void @llvm.objc.storeStrong(
 // CHECK-NEXT: load [[TEST0]]*, [[TEST0]]** [[VAL]],
 // CHECK-NEXT: load
 // CHECK-NEXT: bitcast
 // CHECK-NEXT: [[T0:%.*]] = call i8* bitcast (
-// CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+// CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
 // CHECK-NEXT: [[T2:%.*]] = bitcast i8* [[T1]] to [[TEST0]]*
 // CHECK-NEXT: store [[TEST0]]* [[T2]], [[TEST0]]** [[X]]
 // CHECK-NEXT: [[T0:%.*]] = bitcast [[TEST0]]** [[X]] to i8**
 // CHECK-NEXT: call void @llvm.objc.storeStrong(i8** [[T0]], i8* null)
 // CHECK-NEXT: [[T0:%.*]] = bitcast [[TEST0]]** [[VAL]] to i8**
 // CHECK-NEXT: call void @llvm.objc.storeStrong(i8** [[T0]], i8* null)
 // CHECK-NEXT: ret void
 }

test/CodeGenObjC/arc-ternary-op.m

void test2(int cond) {
   // CHECK:      [[RUN_CLEANUP:%.*]] = alloca i1
   //   Evaluate condition; cleanup disabled by default.
   // CHECK:      [[T0:%.*]] = load i32, i32* [[COND]],
   // CHECK-NEXT: icmp ne i32 [[T0]], 0
   // CHECK-NEXT: store i1 false, i1* [[RUN_CLEANUP]]
   // CHECK-NEXT: br i1
   //   Within true branch, cleanup enabled.
   // CHECK:      [[T0:%.*]] = call i8* @test2_producer()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: store i8* [[T1]], i8** [[CLEANUP_SAVE]]
   // CHECK-NEXT: store i1 true, i1* [[RUN_CLEANUP]]
   // CHECK-NEXT: br label
   //   Join point for conditional operator; retain immediately.
   // CHECK:      [[T0:%.*]] = phi i8* [ [[T1]], {{%.*}} ], [ null, {{%.*}} ]
   // CHECK-NEXT: [[RESULT:%.*]] = call i8* @llvm.objc.retain(i8* [[T0]])
   //   Leaving full-expression; run conditional cleanup.
   // CHECK-NEXT: [[T0:%.*]] = load i1, i1* [[RUN_CLEANUP]]

test/CodeGenObjC/arc-unsafeclaim.m

 // CHECK-OPTIMIZED-NEXT: bitcast
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT:           ret void
 
 // DISABLED-LABEL:     define void @test_assign()
 // DISABLED:             [[T0:%.*]] = call [[A:.*]]* @makeA()
 // DISABLED-MARKED-NEXT: call void asm sideeffect
 // DISABLED-NEXT:        [[T1:%.*]] = bitcast [[A]]* [[T0]] to i8*
-// DISABLED-NEXT:        [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+// DISABLED-NEXT:        [[T2:%.*]] = {{.*}}call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 
 void test_assign_assign() {
   __unsafe_unretained id x, y;
   x = y = makeA();
 }
 // CHECK-LABEL:        define void @test_assign_assign()
 // CHECK:                [[X:%.*]] = alloca i8*
 // CHECK:                [[Y:%.*]] = alloca i8*
void test_strong_assign_assign() {
   x = y = makeA();
 }
 // CHECK-LABEL:        define void @test_strong_assign_assign()
 // CHECK:                [[X:%.*]] = alloca i8*
 // CHECK:                [[Y:%.*]] = alloca i8*
 // CHECK:                [[T0:%.*]] = call [[A]]* @makeA()
 // CHECK-MARKED-NEXT:    call void asm sideeffect
 // CHECK-NEXT:           [[T1:%.*]] = bitcast [[A]]* [[T0]] to i8*
-// CHECK-NEXT:           [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+// CHECK-NEXT:           [[T2:%.*]] = {{.*}}call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 // CHECK-NEXT:           [[T3:%.*]] = bitcast i8* [[T2]] to [[A]]*
 // CHECK-NEXT:           [[T4:%.*]] = bitcast [[A]]* [[T3]] to i8*
 // CHECK-NEXT:           store i8* [[T4]], i8** [[Y]]
 // CHECK-NEXT:           [[OLD:%.*]] = load i8*, i8** [[X]]
 // CHECK-NEXT:           store i8* [[T4]], i8** [[X]]
 // CHECK-NEXT:           call void @llvm.objc.release(i8* [[OLD]]
 // CHECK-OPTIMIZED-NEXT: bitcast
 // CHECK-OPTIMIZED-NEXT: lifetime.end
void test_assign_strong_assign() {
   x = y = makeA();
 }
 // CHECK-LABEL:        define void @test_assign_strong_assign()
 // CHECK:                [[X:%.*]] = alloca i8*
 // CHECK:                [[Y:%.*]] = alloca i8*
 // CHECK:                [[T0:%.*]] = call [[A]]* @makeA()
 // CHECK-MARKED-NEXT:    call void asm sideeffect
 // CHECK-NEXT:           [[T1:%.*]] = bitcast [[A]]* [[T0]] to i8*
-// CHECK-NEXT:           [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+// CHECK-NEXT:           [[T2:%.*]] = {{.*}}call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 // CHECK-NEXT:           [[T3:%.*]] = bitcast i8* [[T2]] to [[A]]*
 // CHECK-NEXT:           [[T4:%.*]] = bitcast [[A]]* [[T3]] to i8*
 // CHECK-NEXT:           [[OLD:%.*]] = load i8*, i8** [[Y]]
 // CHECK-NEXT:           store i8* [[T4]], i8** [[Y]]
 // CHECK-NEXT:           call void @llvm.objc.release(i8* [[OLD]]
 // CHECK-NEXT:           store i8* [[T4]], i8** [[X]]
 // CHECK-UNOPTIMIZED-NEXT: call void @llvm.objc.storeStrong(i8** [[Y]], i8* null)
 // CHECK-OPTIMIZED-NEXT: [[T0:%.*]] = load i8*, i8** [[Y]]
void test_strong_init_assignment() {
   __strong id y = x = makeA();
 }
 // CHECK-LABEL:        define void @test_strong_init_assignment()
 // CHECK:                [[X:%.*]] = alloca i8*
 // CHECK:                [[Y:%.*]] = alloca i8*
 // CHECK:                [[T0:%.*]] = call [[A]]* @makeA()
 // CHECK-MARKED-NEXT:    call void asm sideeffect
 // CHECK-NEXT:           [[T1:%.*]] = bitcast [[A]]* [[T0]] to i8*
-// CHECK-NEXT:           [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+// CHECK-NEXT:           [[T2:%.*]] = {{.*}}call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 // CHECK-NEXT:           [[T3:%.*]] = bitcast i8* [[T2]] to [[A]]*
 // CHECK-NEXT:           [[T4:%.*]] = bitcast [[A]]* [[T3]] to i8*
 // CHECK-NEXT:           store i8* [[T4]], i8** [[X]]
 // CHECK-NEXT:           store i8* [[T4]], i8** [[Y]]
 // CHECK-UNOPTIMIZED-NEXT: call void @llvm.objc.storeStrong(i8** [[Y]], i8* null)
 // CHECK-OPTIMIZED-NEXT: [[T0:%.*]] = load i8*, i8** [[Y]]
 // CHECK-OPTIMIZED-NEXT: call void @llvm.objc.release(i8* [[T0]])
 // CHECK-OPTIMIZED-NEXT: bitcast
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-OPTIMIZED-NEXT: bitcast
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT: ret void
 
 void test_init_strong_assignment() {
   __strong id x;
   __unsafe_unretained id y = x = makeA();
 }
 // CHECK-LABEL:        define void @test_init_strong_assignment()
 // CHECK:                [[X:%.*]] = alloca i8*
 // CHECK:                [[Y:%.*]] = alloca i8*
 // CHECK:                [[T0:%.*]] = call [[A]]* @makeA()
 // CHECK-MARKED-NEXT:    call void asm sideeffect
 // CHECK-NEXT:           [[T1:%.*]] = bitcast [[A]]* [[T0]] to i8*
-// CHECK-NEXT:           [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+// CHECK-NEXT:           [[T2:%.*]] = {{.*}}call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 // CHECK-NEXT:           [[T3:%.*]] = bitcast i8* [[T2]] to [[A]]*
 // CHECK-NEXT:           [[T4:%.*]] = bitcast [[A]]* [[T3]] to i8*
 // CHECK-NEXT:           [[OLD:%.*]] = load i8*, i8** [[X]]
 // CHECK-NEXT:           store i8* [[T4]], i8** [[X]]
 // CHECK-NEXT:           call void @llvm.objc.release(i8* [[OLD]])
 // CHECK-NEXT:           store i8* [[T4]], i8** [[Y]]
 // CHECK-OPTIMIZED-NEXT: bitcast
 // CHECK-OPTIMIZED-NEXT: lifetime.end

test/CodeGenObjC/arc-with-atthrow.m

 // RUN: %clang_cc1 -triple x86_64-apple-darwin10 -emit-llvm -fobjc-arc -fobjc-exceptions -o - %s | FileCheck %s
 // pr10411
 // rdar://10042689
 
 id make(void);
 void test() { 
   @throw make();
 }
 
 // TODO: We should probably emit this specific pattern without the reclaim.
 
 // CHECK-LABEL:    define void @test()
 // CHECK:      [[T0:%.*]] = call i8* @make()
-// CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+// CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
 // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.autorelease(i8* [[T1]])
 // CHECK-NEXT: call void @objc_exception_throw(i8* [[T2]]) [[NR:#[0-9]+]]
 // CHECK-NEXT: unreachable
 
 // CHECK: attributes [[NR]] = { noreturn }

test/CodeGenObjC/arc.m

void test10() {
   // CHECK-NEXT: store [[TEST10]]* null, [[TEST10]]** [[X]]
   // CHECK-NEXT: [[YPTR1:%.*]] = bitcast i8** [[Y]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[YPTR1]])
   // CHECK-NEXT: load [[TEST10]]*, [[TEST10]]** [[X]], align
   // CHECK-NEXT: load i8*, i8** @OBJC_SELECTOR_REFERENCES_{{[0-9]*}}
   // CHECK-NEXT: bitcast
   // CHECK-NEXT: [[T0:%.*]] = call [[TEST10]]* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST10]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[V:%.*]] = bitcast i8* [[T2]] to [[TEST10]]*
   // CHECK-NEXT: load i8*, i8** @OBJC_SELECTOR_REFERENCES_{{[0-9]*}}
   // CHECK-NEXT: bitcast
   // CHECK-NEXT: [[T0:%.*]] = call [[TEST10]]* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST10]]* [[T0]] to i8*
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST10]]*
   // CHECK-NEXT: [[T4:%.*]] = bitcast [[TEST10]]* [[T3]] to i8*
   // CHECK-NEXT: store i8* [[T4]], i8** [[Y]]
   // CHECK-NEXT: [[T0:%.*]] = bitcast [[TEST10]]* [[V]] to i8*
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T0]])
   // CHECK-NEXT: [[T0:%.*]] = load i8*, i8** [[Y]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T0]])
   // CHECK-NEXT: [[YPTR2:%.*]] = bitcast i8** [[Y]] to i8*
void test12(void) {
   // CHECK-LABEL:      define void @test12()
   // CHECK:      [[X:%.*]] = alloca i8*, align
   // CHECK-NEXT: [[Y:%.*]] = alloca i8*, align
 
   __weak id x = test12_helper();
   // CHECK-NEXT: [[XPTR1:%.*]] = bitcast i8** [[X]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[XPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = call i8* @test12_helper()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: call i8* @llvm.objc.initWeak(i8** [[X]], i8* [[T1]])
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
 
   x = test12_helper();
   // CHECK-NEXT: [[T0:%.*]] = call i8* @test12_helper()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: call i8* @llvm.objc.storeWeak(i8** [[X]], i8* [[T1]])
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
 
   id y = x;
   // CHECK-NEXT: [[YPTR1:%.*]] = bitcast i8** [[Y]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[YPTR1]])
   // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.loadWeakRetained(i8** [[X]])
   // CHECK-NEXT: store i8* [[T2]], i8** [[Y]], align
void test19() {
   // CHECK-NEXT: [[T0:%.*]] = bitcast [5 x i8*]* [[X]] to i8*
   // CHECK: call void @llvm.memset.p0i8.i64(i8* align 16 [[T0]], i8 0, i64 40, i1 false)
   id x[5];
 
   extern id test19_helper(void);
   x[2] = test19_helper();
 
   // CHECK-NEXT: [[CALL:%.*]] = call i8* @test19_helper()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]]) [[NUW]]
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]]) [[NUW]]
   // CHECK-NEXT: [[SLOT:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[X]], i64 0, i64 2
   // CHECK-NEXT: [[T0:%.*]] = load i8*, i8** [[SLOT]]
   // CHECK-NEXT: store i8* [[T1]], i8** [[SLOT]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T0]]) [[NUW]]
 
   // CHECK-NEXT: [[BEGIN:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[X]], i32 0, i32 0
   // CHECK-NEXT: [[END:%.*]] = getelementptr inbounds i8*, i8** [[BEGIN]], i64 5
   // CHECK-NEXT: br label
// CHECK-NEXT: ret [[TEST30_HELPER]]* null
   return 0;
 }
 
 @end
 
 __attribute__((ns_returns_retained)) id test32(void) {
 // CHECK-LABEL:    define i8* @test32()
 // CHECK:      [[CALL:%.*]] = call i8* @test32_helper()
-// CHECK-NEXT: [[T0:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
+// CHECK-NEXT: [[T0:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
 // CHECK-NEXT: ret i8* [[T0]]
   extern id test32_helper(void);
   return test32_helper();
 }
 
 @class Test33_a;
 @interface Test33
 - (void) give: (Test33_a **) x;

 }
 
 @interface Test43 @end
 @implementation Test43
 - (id) test __attribute__((ns_returns_retained)) {
   extern id test43_produce(void);
   return test43_produce();
   // CHECK:      call i8* @test43_produce()
-  // CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue(
+  // CHECK-NEXT: notail call i8* @llvm.objc.retainAutoreleasedReturnValue(
   // CHECK-NEXT: ret 
 }
 @end
 
 @interface Test45
 @property (retain) id x;
 @end
 @implementation Test45
 @synthesize x;
 @end
 // CHECK:    define internal i8* @"\01-[Test45 x]"(
 // CHECK:      [[CALL:%.*]] = tail call i8* @objc_getProperty(
 // CHECK-NEXT: ret i8* [[CALL]]
 
 // rdar://problem/9315552
 void test46(__weak id *wp, __weak volatile id *wvp) {
   extern id test46_helper(void);
 
   // TODO: this is sub-optimal, we should retain at the actual call site.
 
   // CHECK:      [[T0:%.*]] = call i8* @test46_helper()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: [[T2:%.*]] = load i8**, i8*** {{%.*}}, align 8
   // CHECK-NEXT: [[T3:%.*]] = call i8* @llvm.objc.storeWeak(i8** [[T2]], i8* [[T1]])
   // CHECK-NEXT: [[T4:%.*]] = call i8* @llvm.objc.retain(i8* [[T3]])
   // CHECK-NEXT: store i8* [[T4]], i8**
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   id x = *wp = test46_helper();
 
   // CHECK:      [[T0:%.*]] = call i8* @test46_helper()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: [[T2:%.*]] = load i8**, i8*** {{%.*}}, align 8
   // CHECK-NEXT: [[T3:%.*]] = call i8* @llvm.objc.storeWeak(i8** [[T2]], i8* [[T1]])
   // CHECK-NEXT: [[T4:%.*]] = call i8* @llvm.objc.retain(i8* [[T3]])
   // CHECK-NEXT: store i8* [[T4]], i8**
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   id y = *wvp = test46_helper();
 }
 
 // rdar://problem/9378887
 void test47(void) {
   extern id test47_helper(void);
   id x = x = test47_helper();
 
   // CHECK-LABEL:    define void @test47()
   // CHECK:      [[X:%.*]] = alloca i8*
   // CHECK-NEXT: [[XPTR1:%.*]] = bitcast i8** [[X]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[XPTR1]])
   // CHECK-NEXT: store i8* null, i8** [[X]]
   // CHECK-NEXT: [[CALL:%.*]] = call i8* @test47_helper()
-  // CHECK-NEXT: [[T0:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
+  // CHECK-NEXT: [[T0:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
   // CHECK-NEXT: [[T1:%.*]] = load i8*, i8** [[X]]
   // CHECK-NEXT: store i8* [[T0]], i8** [[X]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retain(i8* [[T0]])
   // CHECK-NEXT: [[T3:%.*]] = load i8*, i8** [[X]]
   // CHECK-NEXT: store i8* [[T2]], i8** [[X]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T3]])
   // CHECK-NEXT: [[T4:%.*]] = load i8*, i8** [[X]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T4]])
   // CHECK-NEXT: [[XPTR2:%.*]] = bitcast i8** [[X]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* [[XPTR2]])
   // CHECK-NEXT: ret void
 }
 
 void test48(void) {
   extern id test48_helper(void);
   __weak id x = x = test48_helper();
   // CHECK-LABEL:    define void @test48()
   // CHECK:      [[X:%.*]] = alloca i8*
   // CHECK-NEXT: [[XPTR1:%.*]] = bitcast i8** [[X]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[XPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = call i8* @llvm.objc.initWeak(i8** [[X]], i8* null)
   // CHECK-NEXT: [[T1:%.*]] = call i8* @test48_helper()
-  // CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
   // CHECK-NEXT: [[T3:%.*]] = call i8* @llvm.objc.storeWeak(i8** [[X]], i8* [[T2]])
   // CHECK-NEXT: [[T4:%.*]] = call i8* @llvm.objc.storeWeak(i8** [[X]], i8* [[T3]])
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T2]])
   // CHECK-NEXT: call void @llvm.objc.destroyWeak(i8** [[X]])
   // CHECK-NEXT: [[XPTR2:%.*]] = bitcast i8** [[X]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* [[XPTR2]])
   // CHECK-NEXT: ret void
 }
 
 void test49(void) {
   extern id test49_helper(void);
   __autoreleasing id x = x = test49_helper();
   // CHECK-LABEL:    define void @test49()
   // CHECK:      [[X:%.*]] = alloca i8*
   // CHECK-NEXT: [[XPTR1:%.*]] = bitcast i8** [[X]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[XPTR1]])
   // CHECK-NEXT: store i8* null, i8** [[X]]
   // CHECK-NEXT: [[CALL:%.*]] = call i8* @test49_helper()
-  // CHECK-NEXT: [[T0:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
+  // CHECK-NEXT: [[T0:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
   // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.autorelease(i8* [[T0]])
   // CHECK-NEXT: store i8* [[T2]], i8** [[X]]
   // CHECK-NEXT: [[T3:%.*]] = call i8* @llvm.objc.retainAutorelease(i8* [[T1]])
   // CHECK-NEXT: store i8* [[T3]], i8** [[X]]
   // CHECK-NEXT: [[XPTR2:%.*]] = bitcast i8** [[X]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* [[XPTR2]])
   // CHECK-NEXT: ret void
 }

 // CHECK:      [[X:%.*]] = alloca i8*,
 // CHECK-NEXT: [[Y:%.*]] = alloca i8*,
 // CHECK-NEXT: [[TMPALLOCA:%.*]] = alloca i8*,
 // CHECK-NEXT: [[XPTR1:%.*]] = bitcast i8** [[X]] to i8*
 // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[XPTR1]])
 // CHECK-NEXT: [[YPTR1:%.*]] = bitcast i8** [[Y]] to i8*
 // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[YPTR1]])
 // CHECK-NEXT: [[T0:%.*]] = call i8* @test53_helper()
-// CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+// CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
 // CHECK-NEXT: store i8* [[T1]], i8** [[Y]],
 // CHECK-NEXT: [[T0:%.*]] = load i8*, i8** [[Y]],
 // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retain(i8* [[T0]])
 // CHECK-NEXT: store i8* [[T1]], i8** [[TMPALLOCA]]
 // CHECK-NEXT: [[T2:%.*]] = load i8*, i8** [[Y]]
 // CHECK-NEXT: call void @llvm.objc.release(i8* [[T2]])
 // CHECK-NEXT: [[YPTR2:%.*]] = bitcast i8** [[Y]] to i8*
 // CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* [[YPTR2]])

 + (id) make __attribute__((ns_returns_retained));
 @end
 @interface Test56<Test56Protocol> @end
 @implementation Test56
 // CHECK: define internal i8* @"\01+[Test56 make]"(
 + (id) make {
   extern id test56_helper(void);
   // CHECK:      [[T0:%.*]] = call i8* @test56_helper()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: ret i8* [[T1]]
   return test56_helper();
 }
 @end
 void test56_test(void) {
   id x = [Test56 make];
   // CHECK-LABEL: define void @test56_test()
   // CHECK:      [[X:%.*]] = alloca i8*, align 8
void test59(void) {
   extern id test59_getlock(void);
   extern void test59_body(void);
   @synchronized (test59_getlock()) {
     test59_body();
   }
 
   // CHECK-LABEL:    define void @test59()
   // CHECK:      [[T0:%.*]] = call i8* @test59_getlock()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: call i32 @objc_sync_enter(i8* [[T1]])
   // CHECK-NEXT: call void @test59_body()
   // CHECK-NEXT: call i32 @objc_sync_exit(i8* [[T1]])
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   // CHECK-NEXT: ret void
 }
 
 // Verify that we don't try to reclaim the result of performSelector.
 // rdar://problem/9887545
 @interface Test61
 - (id) performSelector: (SEL) selector;
 - (void) test61_void;
 - (id) test61_id;
 @end
 void test61(void) {
   // CHECK-LABEL:    define void @test61()
   // CHECK:      [[Y:%.*]] = alloca i8*, align 8
 
   extern id test61_make(void);
 
   // CHECK-NEXT: [[T0:%.*]] = call i8* @test61_make()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: [[T2:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES_
   // CHECK-NEXT: [[T3:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES_
   // CHECK-NEXT: [[T4:%.*]] = call i8* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to i8* (i8*, i8*, i8*)*)(i8* [[T1]], i8* [[T3]], i8* [[T2]])
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   [test61_make() performSelector: @selector(test61_void)];
 
   // CHECK-NEXT: [[YPTR1:%.*]] = bitcast i8** [[Y]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[YPTR1]])
   // CHECK-NEXT: [[T0:%.*]] = call i8* @test61_make()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: [[T2:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES_
   // CHECK-NEXT: [[T3:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES_
   // CHECK-NEXT: [[T4:%.*]] = call i8* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to i8* (i8*, i8*, i8*)*)(i8* [[T1]], i8* [[T3]], i8* [[T2]])
-  // CHECK-NEXT: [[T5:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
+  // CHECK-NEXT: [[T5:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
   // CHECK-NEXT: store i8* [[T5]], i8** [[Y]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   id y = [test61_make() performSelector: @selector(test61_id)];
 
   // CHECK-NEXT: [[T0:%.*]] = load i8*, i8** [[Y]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T0]])
   // CHECK-NEXT: [[YPTR2:%.*]] = bitcast i8** [[Y]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* [[YPTR2]])
void test62(void) {
   // CHECK-NEXT: br i1 [[T1]],
 
   for (unsigned i = 0; i != 20; ++i) {
     // CHECK:      [[T0:%.*]] = load i32, i32* [[I]], align 4
     // CHECK-NEXT: [[T1:%.*]] = icmp ne i32 [[T0]], 0
     // CHECK-NEXT: store i1 false, i1* [[CLEANUP_REQUIRED]]
     // CHECK-NEXT: br i1 [[T1]],
     // CHECK:      [[T0:%.*]] = call i8* @test62_make()
-    // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+    // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
     // CHECK-NEXT: store i8* [[T1]], i8** [[CLEANUP_VALUE]]
     // CHECK-NEXT: store i1 true, i1* [[CLEANUP_REQUIRED]]
     // CHECK-NEXT: [[T2:%.*]] = icmp ne i8* [[T1]], null
     // CHECK-NEXT: br label
     // CHECK:      [[COND:%.*]] = phi i1 [ false, {{%.*}} ], [ [[T2]], {{%.*}} ]
     // CHECK-NEXT: [[T0:%.*]] = load i1, i1* [[CLEANUP_REQUIRED]]
     // CHECK-NEXT: br i1 [[T0]],
     // CHECK:      [[T0:%.*]] = load i8*, i8** [[CLEANUP_VALUE]]

 void test66(void) {
   extern Test66 *test66_receiver(void);
   extern id test66_arg(void);
   [test66_receiver() consume: test66_arg()];
 }
 // CHECK-LABEL:    define void @test66()
 // CHECK:      [[T0:%.*]] = call [[TEST66:%.*]]* @test66_receiver()
 // CHECK-NEXT: [[T1:%.*]] = bitcast [[TEST66]]* [[T0]] to i8*
-// CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+// CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST66]]*
 // CHECK-NEXT: [[T4:%.*]] = call i8* @test66_arg()
-// CHECK-NEXT: [[T5:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
+// CHECK-NEXT: [[T5:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
 // CHECK-NEXT: [[T6:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES
 // CHECK-NEXT: [[T7:%.*]] = bitcast [[TEST66]]* [[T3]] to i8*
 // CHECK-NEXT: [[SIX:%.*]] = icmp eq i8* [[T7]], null
 // CHECK-NEXT: br i1 [[SIX]], label [[NULINIT:%.*]], label [[CALL:%.*]]
 // CHECK: call void bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to void (i8*, i8*, i8*)*)(i8* [[T7]], i8* [[T6]], i8* [[T5]])
 // CHECK-NEXT: br label [[CONT:%.*]]
 // CHECK: call void @llvm.objc.release(i8* [[T5]]) [[NUW]]
 // CHECK-NEXT: br label [[CONT:%.*]]

 void test68(void) {
   __strong Class cl = test67_helper();
 }
 // CHECK-LABEL:    define void @test68()
 // CHECK:      [[CL:%.*]] = alloca i8*, align 8
 // CHECK-NEXT: [[CLPTR1:%.*]] = bitcast i8** [[CL]] to i8*
 // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[CLPTR1]])
 // CHECK-NEXT: [[T0:%.*]] = call i8* @test67_helper()
-// CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+// CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
 // CHECK-NEXT: store i8* [[T1]], i8** [[CL]], align 8
 // CHECK-NEXT: [[T2:%.*]] = load i8*, i8** [[CL]]
 // CHECK-NEXT: call void @llvm.objc.release(i8* [[T2]])
 // CHECK-NEXT: [[CLPTR2:%.*]] = bitcast i8** [[CL]] to i8*
 // CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* [[CLPTR2]])
 // CHECK-NEXT: ret void
 
 // rdar://problem/10564852

test/CodeGenObjC/objc-arc-container-subscripting.m

 // RUN: %clang_cc1 -fobjc-arc -emit-llvm -triple x86_64-apple-darwin -o - %s | FileCheck %s
 
 @interface NSMutableArray
 - (id)objectAtIndexedSubscript:(int)index;
 - (void)setObject:(id)object atIndexedSubscript:(int)index;
 @end
 
 id func() {
   NSMutableArray *array;
   array[3] = 0;
   return array[3];
 }
 
 // CHECK: [[call:%.*]] = call i8* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend
-// CHECK: [[SIX:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[call]]) [[NUW:#[0-9]+]]
+// CHECK: [[SIX:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[call]]) [[NUW:#[0-9]+]]
 // CHECK: [[ARRAY_CASTED:%.*]] = bitcast %0** {{%.*}} to i8**
 // CHECK: call void @llvm.objc.storeStrong(i8** [[ARRAY_CASTED]], i8* null)
 // CHECK: [[EIGHT:%.*]] = tail call i8* @llvm.objc.autoreleaseReturnValue(i8* [[SIX]]) [[NUW]]
 // CHECK: ret i8* [[EIGHT]]
 
 // CHECK: attributes [[NUW]] = { nounwind }

test/CodeGenObjC/os_log.m

 // CHECK-O0-LABEL: define i8* @test_builtin_os_log
 // CHECK: (i8* returned %[[BUF:.*]])
 // CHECK-O0: (i8* %[[BUF:.*]])
 void *test_builtin_os_log(void *buf) {
   return __builtin_os_log_format(buf, "capabilities: %@", GenString());
 
   // CHECK: %[[CALL:.*]] = tail call %[[TY0:.*]]* (...) @GenString()
   // CHECK: %[[V0:.*]] = bitcast %[[TY0]]* %[[CALL]] to i8*
-  // CHECK: %[[V1:.*]] = tail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[V0]])
+  // CHECK: %[[V1:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[V0]])
   // CHECK: %[[V2:.*]] = ptrtoint %[[TY0]]* %[[CALL]] to i64
   // CHECK: store i8 2, i8* %[[BUF]], align 1
   // CHECK: %[[NUMARGS_I:.*]] = getelementptr i8, i8* %[[BUF]], i64 1
   // CHECK: store i8 1, i8* %[[NUMARGS_I]], align 1
   // CHECK: %[[ARGDESCRIPTOR_I:.*]] = getelementptr i8, i8* %[[BUF]], i64 2
   // CHECK: store i8 64, i8* %[[ARGDESCRIPTOR_I]], align 1
   // CHECK: %[[ARGSIZE_I:.*]] = getelementptr i8, i8* %[[BUF]], i64 3
   // CHECK: store i8 8, i8* %[[ARGSIZE_I]], align 1
   // CHECK: %[[ARGDATA_I:.*]] = getelementptr i8, i8* %[[BUF]], i64 4
   // CHECK: %[[ARGDATACAST_I:.*]] = bitcast i8* %[[ARGDATA_I]] to i64*
   // CHECK: store i64 %[[V2]], i64* %[[ARGDATACAST_I]], align 1
   // CHECK: tail call void (...) @llvm.objc.clang.arc.use(%[[TY0]]* %[[CALL]])
   // CHECK: tail call void @llvm.objc.release(i8* %[[V0]])
   // CHECK: ret i8* %[[BUF]]
 
   // clang.arc.use is used and removed in IR optimizations. At O0, we should not
   // emit clang.arc.use, since it will not be removed and we will have a link
   // error.
   // CHECK-O0: %[[BUF_ADDR:.*]] = alloca i8*, align 8
   // CHECK-O0: store i8* %[[BUF]], i8** %[[BUF_ADDR]], align 8
   // CHECK-O0: %[[V0:.*]] = load i8*, i8** %[[BUF_ADDR]], align 8
   // CHECK-O0: %[[CALL:.*]] = call %[[TY0:.*]]* (...) @GenString()
   // CHECK-O0: %[[V1:.*]] = bitcast %[[TY0]]* %[[CALL]] to i8*
-  // CHECK-O0: %[[V2:.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[V1]])
+  // CHECK-O0: %[[V2:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[V1]])
   // CHECK-O0: %[[V3:.*]] = bitcast i8* %[[V2]] to %[[TY0]]*
   // CHECK-O0: %[[V4:.*]] = ptrtoint %[[TY0]]* %[[V3]] to i64
   // CHECK-O0: call void @__os_log_helper_1_2_1_8_64(i8* %[[V0]], i64 %[[V4]])
   // CHECK-O0: %[[V5:.*]] = bitcast %[[TY0]]* %[[V3]] to i8*
   // CHECK-O0-NOT: call void (...) @llvm.objc.clang.arc.use({{.*}}
   // CHECK-O0: call void @llvm.objc.release(i8* %[[V5]])
   // CHECK-O0: ret i8* %[[V0]]
 }

test/CodeGenObjCXX/arc-forwarded-lambda-call.mm

 // RUN: %clang_cc1 -triple x86_64-apple-macosx10.12.0 -emit-llvm -disable-llvm-passes -O3 -fblocks -fobjc-arc -fobjc-runtime-has-weak -std=c++11 -o - %s | FileCheck %s
 
 void test0(id x) {
   extern void test0_helper(id (^)(void));
   test0_helper([=]() { return x; });
   // CHECK-LABEL: define internal i8* @___Z5test0P11objc_object_block_invoke
   // CHECK: [[T0:%.*]] = call i8* @"_ZZ5test0P11objc_objectENK3$_0clEv"
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: [[T2:%.*]] = tail call i8* @llvm.objc.autoreleaseReturnValue(i8* [[T1]])
   // CHECK-NEXT: ret i8* [[T2]]
 }
 
 // Check that the delegating block invoke function doesn't destruct the Weak
 // object that is passed.
 
 // CHECK-LABEL: define internal void @___Z8testWeakv_block_invoke(
 // CHECK: call void @"_ZZ8testWeakvENK3$_2clE4Weak"(
 // CHECK-NEXT: ret void
 
 // CHECK-LABEL: define internal void @"_ZZ8testWeakvENK3$_2clE4Weak"(
 // CHECK: call void @_ZN4WeakD1Ev(
 // CHECK-NEXT: ret void
 
 id test1_rv;
 
 void test1() {
   extern void test1_helper(id (*)(void));
   test1_helper([](){ return test1_rv; });
   // CHECK-LABEL: define internal i8* @"_ZZ5test1vEN3$_18__invokeEv"
   // CHECK: [[T0:%.*]] = call i8* @"_ZZ5test1vENK3$_1clEv"
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: [[T2:%.*]] = tail call i8* @llvm.objc.autoreleaseReturnValue(i8* [[T1]])
   // CHECK-NEXT: ret i8* [[T2]]
 }
 
 struct Weak {
   __weak id x;
 };
 
 void testWeak() {
   extern void testWeak_helper(void (^)(Weak));
   testWeak_helper([](Weak){});
 }

test/CodeGenObjCXX/arc.mm

 // The analogous ObjC testcase test46 in arr.m.
 void test0(__weak id *wp, __weak volatile id *wvp) {
   extern id test0_helper(void);
 
   // TODO: this is sub-optimal, we should retain at the actual call site.
   // TODO: in the non-volatile case, we do not need to be reloading.
 
   // CHECK:      [[T0:%.*]] = call i8* @_Z12test0_helperv()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: [[T2:%.*]] = load i8**, i8*** {{%.*}}, align 8
   // CHECK-NEXT: [[T3:%.*]] = call i8* @llvm.objc.storeWeak(i8** [[T2]], i8* [[T1]])
   // CHECK-NEXT: [[T4:%.*]] = call i8* @llvm.objc.retain(i8* [[T3]])
   // CHECK-NEXT: store i8* [[T4]], i8**
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   id x = *wp = test0_helper();
 
   // CHECK:      [[T0:%.*]] = call i8* @_Z12test0_helperv()
-  // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
   // CHECK-NEXT: [[T2:%.*]] = load i8**, i8*** {{%.*}}, align 8
   // CHECK-NEXT: [[T3:%.*]] = call i8* @llvm.objc.storeWeak(i8** [[T2]], i8* [[T1]])
   // CHECK-NEXT: [[T4:%.*]] = call i8* @llvm.objc.loadWeakRetained(i8** [[T2]])
   // CHECK-NEXT: store i8* [[T4]], i8**
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   id y = *wvp = test0_helper();
 }
 
for (id x in a.array) {
     use(x);
   }
 }
 extern template void test37<Test37>(Test37 *a);
 template void test37<Test37>(Test37 *a);
 // CHECK-LABEL: define weak_odr void @_Z6test37I6Test37EvPT_(
 // CHECK:      [[T0:%.*]] = call [[NSARRAY]]* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to [[NSARRAY]]* (i8*, i8*)*)(
 // CHECK-NEXT: [[T1:%.*]] = bitcast [[NSARRAY]]* [[T0]] to i8*
-// CHECK-NEXT: [[T2:%.*]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+// CHECK-NEXT: [[T2:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 // CHECK-NEXT: [[COLL:%.*]] = bitcast i8* [[T2]] to [[NSARRAY]]*
 
 // Make sure it's not immediately released before starting the iteration.
 // CHECK-NEXT: load i8*, i8** @OBJC_SELECTOR_REFERENCES_
 // CHECK-NEXT: [[T0:%.*]] = bitcast [[NSARRAY]]* [[COLL]] to i8*
 // CHECK-NEXT: @objc_msgSend
 
 // This bitcast is for the mutation check.

test/CodeGenObjCXX/inheriting-constructor-cleanup.mm

 };
 
 id g(void);
 void f() {
   Inheritor({g()});
 }
 // CHECK-LABEL: define void @_Z1fv
 // CHECK:       %[[TMP:.*]] = call i8* @_Z1gv()
-// CHECK:       {{.*}} = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[TMP]])
+// CHECK:       {{.*}} = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[TMP]])
 // CHECK:       call void (%struct.Base*, i8*, ...) @_ZN4BaseC2E6Strongz(%struct.Base* {{.*}}, i8* {{.*}})
 // CHECK-NEXT:  call void @_ZN9InheritorD1Ev(%struct.Inheritor* {{.*}})
 
 // CHECK-LABEL: define linkonce_odr void @_ZN4BaseC2E6Strongz(%struct.Base* {{.*}}, i8* {{.*}}, ...)
 // CHECK:       call void @_ZN6StrongD1Ev(%struct.Strong* {{.*}})
 
 // CHECK-LABEL: define linkonce_odr void @_ZN9InheritorD1Ev(%struct.Inheritor* {{.*}})
 // CHECK:       call void @_ZN9InheritorD2Ev(%struct.Inheritor* {{.*}})

test/CodeGenObjCXX/literals.mm

void test_array() {
   // Initializing first element
   // CHECK: [[PTR1:%.*]] = bitcast i8** [[ARR]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTR1]])
   // CHECK: [[ELEMENT0:%[a-zA-Z0-9.]+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[OBJECTS]], i64 0, i64 0
   // CHECK-NEXT: [[TMP_CAST:%.*]] = bitcast {{.*}} [[TMPX]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 1, i8* [[TMP_CAST]])
   // CHECK-NEXT: call void @_ZN1XC1Ev({{.*}} [[TMPX]])
   // CHECK-NEXT: [[OBJECT0:%[a-zA-Z0-9.]+]] = invoke i8* @_ZNK1XcvP11objc_objectEv
-  // CHECK: [[RET0:%[a-zA-Z0-9.]+]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT0]])
+  // CHECK: [[RET0:%[a-zA-Z0-9.]+]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT0]])
   // CHECK: store i8* [[RET0]], i8** [[ELEMENT0]]
   
   // Initializing the second element
   // CHECK: [[ELEMENT1:%[a-zA-Z0-9.]+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[OBJECTS]], i64 0, i64 1
   // CHECK-NEXT: [[TMP_CAST:%.*]] = bitcast {{.*}} [[TMPY]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 1, i8* [[TMP_CAST]])
   // CHECK-NEXT: invoke void @_ZN1YC1Ev({{.*}} [[TMPY]])
   // CHECK: [[OBJECT1:%[a-zA-Z0-9.]+]] = invoke i8* @_ZNK1YcvP11objc_objectEv
-  // CHECK: [[RET1:%[a-zA-Z0-9.]+]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT1]])
+  // CHECK: [[RET1:%[a-zA-Z0-9.]+]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT1]])
   // CHECK: store i8* [[RET1]], i8** [[ELEMENT1]]
 
   // Build the array
   // CHECK: {{invoke.*@objc_msgSend}}
   // CHECK: call i8* @llvm.objc.retainAutoreleasedReturnValue
   id arr = @[ X(), Y() ];
 
   // Destroy temporaries
void test_array_instantiation() {
   // CHECK: [[OBJECTS:%[a-zA-Z0-9.]+]] = alloca [2 x i8*]
 
   // Initializing first element
   // CHECK:      [[PTR1:%.*]] = bitcast i8** [[ARR]] to i8*
   // CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[PTR1]])
   // CHECK: [[ELEMENT0:%[a-zA-Z0-9.]+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[OBJECTS]], i64 0, i64 0
   // CHECK: call void @_ZN1XC1Ev
   // CHECK-NEXT: [[OBJECT0:%[a-zA-Z0-9.]+]] = invoke i8* @_ZNK1XcvP11objc_objectEv
-  // CHECK: [[RET0:%[a-zA-Z0-9.]+]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT0]])
+  // CHECK: [[RET0:%[a-zA-Z0-9.]+]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT0]])
   // CHECK: store i8* [[RET0]], i8** [[ELEMENT0]]
   
   // Initializing the second element
   // CHECK: [[ELEMENT1:%[a-zA-Z0-9.]+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[OBJECTS]], i64 0, i64 1
   // CHECK: invoke void @_ZN1YC1Ev
   // CHECK: [[OBJECT1:%[a-zA-Z0-9.]+]] = invoke i8* @_ZNK1YcvP11objc_objectEv
-  // CHECK: [[RET1:%[a-zA-Z0-9.]+]] = call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT1]])
+  // CHECK: [[RET1:%[a-zA-Z0-9.]+]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT1]])
   // CHECK: store i8* [[RET1]], i8** [[ELEMENT1]]
 
   // Build the array
   // CHECK: {{invoke.*@objc_msgSend}}
   // CHECK: call i8* @llvm.objc.retainAutoreleasedReturnValue
   id arr = @[ X(), Y() ];
 
   // Destroy temporaries