diff --git a/clang/lib/CodeGen/CGObjC.cpp b/clang/lib/CodeGen/CGObjC.cpp
--- a/clang/lib/CodeGen/CGObjC.cpp
+++ b/clang/lib/CodeGen/CGObjC.cpp
@@ -2354,9 +2354,12 @@
                  : llvm::Intrinsic::objc_unsafeClaimAutoreleasedReturnValue;
   EP = getARCIntrinsic(IID, CGF.CGM);
 
-  // FIXME: Do this when the target isn't aarch64.
+  llvm::Triple::ArchType Arch = CGF.CGM.getTriple().getArch();
+
+  // FIXME: Do this on all targets and at -O0 too. This can be enabled only if
+  // the target backend knows how to handle the operand bundle.
   if (CGF.CGM.getCodeGenOpts().OptimizationLevel > 0 &&
-      CGF.CGM.getTarget().getTriple().isAArch64()) {
+      (Arch == llvm::Triple::aarch64 || Arch == llvm::Triple::x86_64)) {
     llvm::Value *bundleArgs[] = {EP};
     llvm::OperandBundleDef OB("clang.arc.attachedcall", bundleArgs);
     auto *oldCall = cast<llvm::CallBase>(value);
diff --git a/clang/test/CodeGenObjC/arc-blocks.m b/clang/test/CodeGenObjC/arc-blocks.m
--- a/clang/test/CodeGenObjC/arc-blocks.m
+++ b/clang/test/CodeGenObjC/arc-blocks.m
@@ -126,9 +126,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
-  // CHECK-NEXT: store i8* [[T1]], i8** [[SLOT]]
+  // CHECK-NEXT: [[T0:%.*]] = call i8* @test4_source() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T0]])
+  // CHECK-NEXT: store i8* [[T0]], 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,
@@ -180,8 +180,8 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK: [[T1:%.*]] = call i8* @test5_source() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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
@@ -211,8 +211,8 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = call i8* @test6_source() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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
@@ -257,8 +257,8 @@
   // CHECK-LABEL:    define{{.*}} void @test7()
   // CHECK:      [[VAR:%.*]] = alloca i8*,
   // CHECK-NEXT: [[BLOCK:%.*]] = alloca [[BLOCK_T:<{.*}>]],
-  // CHECK:      [[T0:%.*]] = call i8* @test7_source()
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK:      [[T1:%.*]] = call i8* @test7_source() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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
@@ -325,8 +325,8 @@
 // CHECK-NEXT: tail call i8* @llvm.objc.autoreleaseReturnValue
 // CHECK-NEXT: ret i8*
 
-// CHECK:      call i8* @test9_produce()
-// CHECK-NEXT: call i8* @llvm.objc.retain
+// CHECK:      call i8* @test9_produce() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(
 // CHECK-NEXT: ret i8*
 }
 
diff --git a/clang/test/CodeGenObjC/arc-bridged-cast.m b/clang/test/CodeGenObjC/arc-bridged-cast.m
--- a/clang/test/CodeGenObjC/arc-bridged-cast.m
+++ b/clang/test/CodeGenObjC/arc-bridged-cast.m
@@ -40,9 +40,8 @@
 void bridge_from_cf(int *i) {
   // CHECK: store i32 7
   *i = 7;
-  // CHECK: call i8* @CFCreateSomething()
+  // CHECK: call i8* @CFCreateSomething() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   id obj1 = (__bridge id)CFCreateSomething();
-  // CHECK: llvm.objc.retainAutoreleasedReturnValue
   // CHECK: store i32 11
   *i = 11;
   // CHECK: call i8* @CFCreateSomething()
@@ -60,14 +59,12 @@
 // CHECK-LABEL: define{{.*}} void @bridge_retained_of_cf
 void bridge_retained_of_cf(int *i) {
   *i = 7;
-  // CHECK: call i8* @CreateSomething()
+  // CHECK: call i8* @CreateSomething() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   CFTypeRef cf1 = (__bridge_retained CFTypeRef)CreateSomething();
-  // CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue
   // CHECK: store i32 11
   *i = 11;
-  // CHECK: call i8* @CreateSomething()
+  // CHECK: call i8* @CreateSomething() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   (__bridge_retained CFTypeRef)CreateSomething(), *i = 13;
-  // CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue
   // CHECK: store i32 13
   // CHECK: store i32 17
   *i = 17;
@@ -99,8 +96,7 @@
 
 // CHECK-LABEL: define{{.*}} %struct.__CFString* @bridge_of_paren_expr()
 CFStringRef bridge_of_paren_expr() {
-  // CHECK-NOT: call i8* @llvm.objc.retainAutoreleasedReturnValue(
-  // CHECK-NOT: call void @llvm.objc.release(
+  // CHECK-NOT: "@llvm.objc"
   CFStringRef r = (__bridge CFStringRef)(CreateNSString());
   r = (__bridge CFStringRef)((NSString *)(CreateNSString()));
   return r;
diff --git a/clang/test/CodeGenObjC/arc-literals.m b/clang/test/CodeGenObjC/arc-literals.m
--- a/clang/test/CodeGenObjC/arc-literals.m
+++ b/clang/test/CodeGenObjC/arc-literals.m
@@ -14,17 +14,13 @@
 
 // CHECK-LABEL: define{{.*}} void @test_numeric()
 void test_numeric() {
-  // CHECK: {{call.*objc_msgSend.*i32 17}}
-  // CHECK: call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK: {{call.*objc_msgSend.*i32 17.* [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]}}
   id ilit = @17;
-  // CHECK: {{call.*objc_msgSend.*i32 25}}
-  // CHECK: call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK: {{call.*objc_msgSend.*i32 25.* [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]}}
   id ulit = @25u;
-  // CHECK: {{call.*objc_msgSend.*i64 42}}
-  // CHECK: call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK: {{call.*objc_msgSend.*i64 42.* [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]}}
   id ulllit = @42ull;
-  // CHECK: {{call.*objc_msgSend.*i8 signext 97}}
-  // CHECK: call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK: {{call.*objc_msgSend.*i8 signext 97.* [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]}}
   id charlit = @'a';
   // CHECK: call void @llvm.objc.release
   // CHECK: call void @llvm.lifetime.end
@@ -58,8 +54,7 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T3]])
+  // CHECK-NEXT: [[T3:%.*]] = call i8* bitcast ({{.*@objc_msgSend.*}})(i8* [[T1]], i8* [[SEL]], i8** [[T2]], i64 2) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   // CHECK: call void (...) @llvm.objc.clang.arc.use(i8* [[V0]], i8* [[V1]])
   id arr = @[a, b];
 
@@ -102,8 +97,8 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
+  // CHECK-NEXT: [[T4:%.*]] = call i8* bitcast ({{.*@objc_msgSend.*}})(i8* [[T1]], i8* [[SEL]], i8** [[T2]], i8** [[T3]], i64 2) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(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 };
@@ -133,10 +128,8 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T3]])
-  // CHECK-NEXT: [[V0:%.*]] = bitcast i8* [[T4]] to [[B]]*
+  // CHECK-NEXT: [[V0:%.*]] = call [[B:%.*]]* bitcast ({{.*}} @objc_msgSend to {{.*}})(i8* [[T1]], i8* [[SEL]]) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[B]]* [[V0]])
   // CHECK-NEXT: [[V1:%.*]] = bitcast [[B]]* [[V0]] to i8*
 
   // Store to array.
@@ -147,8 +140,8 @@
   // CHECK-NEXT: [[SEL:%.*]] = load i8*, i8** @OBJC_SELECTOR_REFERENCES
   // CHECK-NEXT: [[T1:%.*]] = bitcast [[CLASS_T]]* [[T0]] to i8*
   // CHECK-NEXT: [[T2:%.*]] = bitcast [1 x i8*]* [[OBJECTS]] to i8**
-  // CHECK-NEXT: [[T3:%.*]] = call i8* bitcast ({{.*}} @objc_msgSend to {{.*}}(i8* [[T1]], i8* [[SEL]], i8** [[T2]], i64 1)
-  // CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T3]])
+  // CHECK-NEXT: [[T3:%.*]] = call i8* bitcast ({{.*}} @objc_msgSend to {{.*}}(i8* [[T1]], i8* [[SEL]], i8** [[T2]], i64 1) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T3]])
   // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.use(i8* [[V1]])
   // CHECK-NEXT: bitcast
   // CHECK-NEXT: bitcast
diff --git a/clang/test/CodeGenObjC/arc-precise-lifetime.m b/clang/test/CodeGenObjC/arc-precise-lifetime.m
--- a/clang/test/CodeGenObjC/arc-precise-lifetime.m
+++ b/clang/test/CodeGenObjC/arc-precise-lifetime.m
@@ -41,11 +41,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
-  // CHECK-NEXT: store [[TEST1]]* [[T3]]
+  // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST1]]* [[T0]])
+  // CHECK-NEXT: store [[TEST1]]* [[T0]]
   // 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]]**
@@ -75,11 +73,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
-  // CHECK-NEXT: store [[TEST1]]* [[T3]]
+  // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST1]]* [[T0]])
+  // CHECK-NEXT: store [[TEST1]]* [[T0]]
   // 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]]**
@@ -109,11 +105,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
-  // CHECK-NEXT: store [[TEST1]]* [[T3]]
+  // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST1]]* [[T0]])
+  // CHECK-NEXT: store [[TEST1]]* [[T0]]
   // 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]]**
@@ -140,11 +134,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
-  // CHECK-NEXT: store [[TEST1]]* [[T3]]
+  // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST1]]* [[T0]])
+  // CHECK-NEXT: store [[TEST1]]* [[T0]]
   // 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]]**
@@ -171,11 +163,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
-  // CHECK-NEXT: store [[TEST1]]* [[T3]]
+  // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST1]]* [[T0]])
+  // CHECK-NEXT: store [[TEST1]]* [[T0]]
   // 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]]**
@@ -204,11 +194,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
-  // CHECK-NEXT: store [[TEST1]]* [[T3]]
+  // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST1]]* [[T0]])
+  // CHECK-NEXT: store [[TEST1]]* [[T0]]
   // 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]]**
@@ -237,11 +225,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
-  // CHECK-NEXT: store [[TEST1]]* [[T3]]
+  // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST1]]* [[T0]])
+  // CHECK-NEXT: store [[TEST1]]* [[T0]]
   // 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]]**
@@ -267,11 +253,9 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST1]]*
-  // CHECK-NEXT: store [[TEST1]]* [[T3]]
+  // CHECK:      [[T0:%.*]] = call [[TEST1:%.*]]* @test1_helper() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST1]]* [[T0]])
+  // CHECK-NEXT: store [[TEST1]]* [[T0]]
   // 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]]**
diff --git a/clang/test/CodeGenObjC/arc-rv-attr.m b/clang/test/CodeGenObjC/arc-rv-attr.m
--- a/clang/test/CodeGenObjC/arc-rv-attr.m
+++ b/clang/test/CodeGenObjC/arc-rv-attr.m
@@ -1,4 +1,5 @@
 // RUN: %clang_cc1 -triple arm64-apple-ios9 -fobjc-runtime=ios-9.0 -fobjc-arc -O -disable-llvm-passes -emit-llvm -o - %s | FileCheck %s -check-prefix=CHECK
+// RUN: %clang_cc1 -triple x86_64-apple-macosx10 -fobjc-runtime=ios-9.0 -fobjc-arc -O -disable-llvm-passes -emit-llvm -o - %s | FileCheck %s -check-prefix=CHECK
 
 @class A;
 
diff --git a/clang/test/CodeGenObjC/arc-ternary-op.m b/clang/test/CodeGenObjC/arc-ternary-op.m
--- a/clang/test/CodeGenObjC/arc-ternary-op.m
+++ b/clang/test/CodeGenObjC/arc-ternary-op.m
@@ -131,8 +131,8 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK:      [[T1:%.*]] = call i8* @test2_producer() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // CHECK-NEXT: store i8* [[T1]], i8** [[CLEANUP_SAVE]]
   // CHECK-NEXT: store i1 true, i1* [[RUN_CLEANUP]]
   // CHECK-NEXT: br label
diff --git a/clang/test/CodeGenObjC/arc-unsafeclaim.m b/clang/test/CodeGenObjC/arc-unsafeclaim.m
--- a/clang/test/CodeGenObjC/arc-unsafeclaim.m
+++ b/clang/test/CodeGenObjC/arc-unsafeclaim.m
@@ -1,6 +1,8 @@
 //   Make sure it works on x86-64.
 // RUN: %clang_cc1 -triple x86_64-apple-darwin11 -fobjc-runtime=macosx-10.11 -fobjc-arc -emit-llvm -o - %s | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-UNOPTIMIZED -check-prefix=NOTAIL-CALL
 
+// RUN: %clang_cc1 -triple x86_64-apple-darwin11 -fobjc-runtime=macosx-10.11 -fobjc-arc -emit-llvm -O2 -disable-llvm-passes -o - %s | FileCheck %s -check-prefix=ATTACHED-CALL
+
 //   Make sure it works on x86-32.
 // RUN: %clang_cc1 -triple i386-apple-darwin11 -fobjc-runtime=macosx-fragile-10.11 -fobjc-arc -emit-llvm -o - %s | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-UNOPTIMIZED -check-prefix=CHECK-MARKED -check-prefix=CALL
 
@@ -43,6 +45,10 @@
 // DISABLED-NEXT:        [[T1:%.*]] = bitcast [[A]]* [[T0]] to i8*
 // DISABLED-NEXT:        [[T2:%.*]] = {{.*}}call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_assign()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.unsafeClaimAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_assign_assign() {
   __unsafe_unretained id x, y;
   x = y = makeA();
@@ -65,6 +71,10 @@
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT:           ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_assign_assign()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.unsafeClaimAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_strong_assign_assign() {
   __strong id x;
   __unsafe_unretained id y;
@@ -92,6 +102,10 @@
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT:           ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_strong_assign_assign()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_assign_strong_assign() {
   __unsafe_unretained id x;
   __strong id y;
@@ -119,6 +133,10 @@
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT:           ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_assign_strong_assign()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_init() {
   __unsafe_unretained id x = makeA();
 }
@@ -136,6 +154,10 @@
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT:           ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_init()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.unsafeClaimAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_init_assignment() {
   __unsafe_unretained id x;
   __unsafe_unretained id y = x = makeA();
@@ -158,6 +180,10 @@
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT: ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_init_assignment()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.unsafeClaimAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_strong_init_assignment() {
   __unsafe_unretained id x;
   __strong id y = x = makeA();
@@ -182,6 +208,10 @@
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT: ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_strong_init_assignment()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_init_strong_assignment() {
   __strong id x;
   __unsafe_unretained id y = x = makeA();
@@ -208,6 +238,10 @@
 // CHECK-OPTIMIZED-NEXT: lifetime.end
 // CHECK-NEXT: ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_init_strong_assignment()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_ignored() {
   makeA();
 }
@@ -220,6 +254,10 @@
 // CHECK-NEXT:        bitcast i8* [[T2]] to [[A]]*
 // CHECK-NEXT:        ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_ignored()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.unsafeClaimAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
+
 void test_cast_to_void() {
   (void) makeA();
 }
@@ -232,6 +270,9 @@
 // CHECK-NEXT:        bitcast i8* [[T2]] to [[A]]*
 // CHECK-NEXT:        ret void
 
+// ATTACHED-CALL-LABEL:      define{{.*}} void @test_cast_to_void()
+// ATTACHED-CALL:              [[T0:%.*]] = call [[A:.*]]* @makeA() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.unsafeClaimAutoreleasedReturnValue) ],
+// ATTACHED-CALL:              call void (...) @llvm.objc.clang.arc.noop.use([[A]]* [[T0]])
 
 
 // This is always at the end of the module.
diff --git a/clang/test/CodeGenObjC/arc.m b/clang/test/CodeGenObjC/arc.m
--- a/clang/test/CodeGenObjC/arc.m
+++ b/clang/test/CodeGenObjC/arc.m
@@ -316,16 +316,12 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[V:%.*]] = bitcast i8* [[T2]] to [[TEST10]]*
+  // CHECK-NEXT: [[V:%.*]] = call [[TEST10]]* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(%3* [[V]])
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-  // CHECK-NEXT: [[T3:%.*]] = bitcast i8* [[T2]] to [[TEST10]]*
+  // CHECK-NEXT: [[T3:%.*]] = call [[TEST10]]* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(%3* [[T3]])
   // CHECK-NEXT: [[T4:%.*]] = bitcast [[TEST10]]* [[T3]] to i8*
   // CHECK-NEXT: store i8* [[T4]], i8** [[Y]]
   // CHECK-NEXT: [[T0:%.*]] = bitcast [[TEST10]]* [[V]] to i8*
@@ -370,14 +366,14 @@
   __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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = call i8* @test12_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = call i8* @test12_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // CHECK-NEXT: call i8* @llvm.objc.storeWeak(i8** [[X]], i8* [[T1]])
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
 
@@ -513,8 +509,8 @@
   extern id test19_helper(void);
   x[2] = test19_helper();
 
-  // CHECK-NEXT: [[CALL:%.*]] = call i8* @test19_helper()
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]]) [[NUW]]
+  // CHECK-NEXT: [[T1:%.*]] = call i8* @test19_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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]]
@@ -875,8 +871,8 @@
 
 __attribute__((ns_returns_retained)) id test32(void) {
 // CHECK-LABEL:    define{{.*}} i8* @test32()
-// CHECK:      [[CALL:%.*]] = call i8* @test32_helper()
-// CHECK-NEXT: [[T0:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
+// CHECK:      [[T0:%.*]] = call i8* @test32_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T0]])
 // CHECK-NEXT: ret i8* [[T0]]
   extern id test32_helper(void);
   return test32_helper();
@@ -1044,8 +1040,8 @@
 - (id) test __attribute__((ns_returns_retained)) {
   extern id test43_produce(void);
   return test43_produce();
-  // CHECK:      call i8* @test43_produce()
-  // CHECK-NEXT: notail call i8* @llvm.objc.retainAutoreleasedReturnValue(
+  // CHECK:      call i8* @test43_produce(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(
   // CHECK-NEXT: ret
 }
 @end
@@ -1066,8 +1062,8 @@
 
   // TODO: this is sub-optimal, we should retain at the actual call site.
 
-  // CHECK:      [[T0:%.*]] = call i8* @test46_helper()
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK:      [[T1:%.*]] = call i8* @test46_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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]])
@@ -1075,8 +1071,8 @@
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   id x = *wp = test46_helper();
 
-  // CHECK:      [[T0:%.*]] = call i8* @test46_helper()
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK:      [[T1:%.*]] = call i8* @test46_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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]])
@@ -1095,8 +1091,8 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
+  // CHECK-NEXT: [[T0:%.*]] = call i8* @test47_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T0]])
   // CHECK-NEXT: [[T1:%.*]] = load i8*, i8** [[X]]
   // CHECK-NEXT: store i8* [[T0]], i8** [[X]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
@@ -1119,8 +1115,8 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
+  // CHECK-NEXT: [[T2:%.*]] = call i8* @test48_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T2]])
   // 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]])
@@ -1138,10 +1134,10 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[CALL]])
+  // CHECK-NEXT: [[T0:%.*]] = call i8* @test49_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T0]])
   // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.autorelease(i8* [[T0]])
-  // CHECK-NEXT: store i8* [[T2]], i8** [[X]]
+  // CHECK-NEXT: store i8* [[T1]], 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*
@@ -1207,8 +1203,8 @@
 // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+// CHECK-NEXT: [[T1:%.*]] = call i8* @test53_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
 // CHECK-NEXT: store i8* [[T1]], i8** [[Y]],
 // CHECK-NEXT: [[T0:%.*]] = load i8*, i8** [[Y]],
 // CHECK-NEXT: [[T1:%.*]] = call i8* @llvm.objc.retain(i8* [[T0]])
@@ -1260,8 +1256,8 @@
 // CHECK: define internal i8* @"\01+[Test56 make]"(
 + (id) make {
   extern id test56_helper(void);
-  // CHECK:      [[T0:%.*]] = call i8* @test56_helper()
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK:      [[T1:%.*]] = call i8* @test56_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // CHECK-NEXT: ret i8* [[T1]]
   return test56_helper();
 }
@@ -1327,8 +1323,8 @@
   }
 
   // CHECK-LABEL:    define{{.*}} void @test59()
-  // CHECK:      [[T0:%.*]] = call i8* @test59_getlock()
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK:      [[T1:%.*]] = call i8* @test59_getlock(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // CHECK-NEXT: call i32 @objc_sync_enter(i8* [[T1]])
   // CHECK-NEXT: call void @test59_body()
   // CHECK-NEXT: call i32 @objc_sync_exit(i8* [[T1]])
@@ -1349,8 +1345,8 @@
 
   extern id test61_make(void);
 
-  // CHECK-NEXT: [[T0:%.*]] = call i8* @test61_make()
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T0:%.*]] = call i8* @test61_make(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(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]])
@@ -1359,12 +1355,12 @@
 
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK-NEXT: [[T1:%.*]] = call i8* @test61_make(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
+  // CHECK-NEXT: [[T5:%.*]] = call i8* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to i8* (i8*, i8*, i8*)*)(i8* [[T1]], i8* [[T3]], i8* [[T2]]){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T5]])
   // CHECK-NEXT: store i8* [[T5]], i8** [[Y]]
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   id y = [test61_make() performSelector: @selector(test61_id)];
@@ -1399,8 +1395,8 @@
     // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+    // CHECK:      [[T1:%.*]] = call i8* @test62_make(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+    // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
     // CHECK-NEXT: store i8* [[T1]], i8** [[CLEANUP_VALUE]]
     // CHECK-NEXT: store i1 true, i1* [[CLEANUP_REQUIRED]]
     // CHECK-NEXT: [[T2:%.*]] = icmp ne i8* [[T1]], null
@@ -1453,19 +1449,17 @@
   [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:%.*]] = 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T4]])
+// CHECK:      [[T3:%.*]] = call [[TEST66:%.*]]* @test66_receiver(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use([[TEST66]]* [[T3]])
+// CHECK-NEXT: [[T4:%.*]] = call i8* @test66_arg(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(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: call void bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to void (i8*, i8*, i8*)*)(i8* [[T7]], i8* [[T6]], i8* [[T4]])
 // CHECK-NEXT: br label [[CONT:%.*]]
-// CHECK: call void @llvm.objc.release(i8* [[T5]]) [[NUW]]
+// CHECK: call void @llvm.objc.release(i8* [[T4]]) [[NUW]]
 // CHECK-NEXT: br label [[CONT:%.*]]
 // CHECK: [[T8:%.*]] = bitcast [[TEST66]]* [[T3]] to i8*
 // CHECK-NEXT: call void @llvm.objc.release(i8* [[T8]])
@@ -1494,8 +1488,8 @@
 // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+// CHECK-NEXT: [[T1:%.*]] = call i8* @test67_helper(){{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
 // 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]])
diff --git a/clang/test/CodeGenObjC/nsvalue-objc-boxable-mac-arc.m b/clang/test/CodeGenObjC/nsvalue-objc-boxable-mac-arc.m
--- a/clang/test/CodeGenObjC/nsvalue-objc-boxable-mac-arc.m
+++ b/clang/test/CodeGenObjC/nsvalue-objc-boxable-mac-arc.m
@@ -24,8 +24,7 @@
   // CHECK:      [[SEL:%.*]]        = load i8*, i8** [[VALUE_SEL]]
   // CHECK:      [[RECV:%.*]]       = bitcast %struct._class_t* [[RECV_PTR]] to i8*
   NSRange ns_range = { .location = 0, .length = 42 };
-  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8* {{.*}}[[RANGE_STR]]{{.*}})
-  // CHECK:      call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8* {{.*}}[[RANGE_STR]]{{.*}}) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   NSValue *range = @(ns_range);
   // CHECK:      call void @llvm.objc.release
   // CHECK:      ret void
@@ -43,8 +42,7 @@
   // CHECK:      [[SEL:%.*]]        = load i8*, i8** [[VALUE_SEL]]
   // CHECK:      [[RECV:%.*]]       = bitcast %struct._class_t* [[RECV_PTR]] to i8*
   NSPoint ns_point = { .x = 42, .y = 24 };
-  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8* {{.*}}[[POINT_STR]]{{.*}})
-  // CHECK:      call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8* {{.*}}[[POINT_STR]]{{.*}}) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   NSValue *point = @(ns_point);
   // CHECK:      call void @llvm.objc.release
   // CHECK:      ret void
@@ -62,8 +60,7 @@
   // CHECK:      [[SEL:%.*]]        = load i8*, i8** [[VALUE_SEL]]
   // CHECK:      [[RECV:%.*]]       = bitcast %struct._class_t* [[RECV_PTR]] to i8*
   NSSize ns_size = { .width = 42, .height = 24 };
-  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8* {{.*}}[[SIZE_STR]]{{.*}})
-  // CHECK:      call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8* {{.*}}[[SIZE_STR]]{{.*}}) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   NSValue *size = @(ns_size);
   // CHECK:      call void @llvm.objc.release
   // CHECK:      ret void
@@ -83,8 +80,7 @@
   NSPoint ns_point = { .x = 42, .y = 24 };
   NSSize ns_size = { .width = 42, .height = 24 };
   NSRect ns_rect = { .origin = ns_point, .size = ns_size };
-  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8*{{.*}}[[RECT_STR]]{{.*}})
-  // CHECK:      call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8*{{.*}}[[RECT_STR]]{{.*}}) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   NSValue *rect = @(ns_rect);
   // CHECK:      call void @llvm.objc.release
   // CHECK:      ret void
@@ -102,8 +98,7 @@
   // CHECK:      [[SEL:%.*]]        = load i8*, i8** [[VALUE_SEL]]
   // CHECK:      [[RECV:%.*]]       = bitcast %struct._class_t* [[RECV_PTR]] to i8*
   NSEdgeInsets ns_edge_insets;
-  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8*{{.*}}[[EDGE_STR]]{{.*}})
-  // CHECK:      call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK:      call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[PARAM_CAST]], i8*{{.*}}[[EDGE_STR]]{{.*}}) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   NSValue *edge_insets = @(ns_edge_insets);
   // CHECK:      call void @llvm.objc.release
   // CHECK:      ret void
@@ -121,8 +116,7 @@
   // CHECK:     [[COERCE_CAST:%.*]]     = bitcast %struct._NSRange* [[COERCE]]{{.*}}
   // CHECK:     [[SEL:%.*]]             = load i8*, i8** [[VALUE_SEL]]
   // CHECK:     [[RECV:%.*]]            = bitcast %struct._class_t* [[RECV_PTR]] to i8*
-  // CHECK:     call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[COERCE_CAST]], i8* {{.*}}[[RANGE_STR]]{{.*}})
-  // CHECK:     call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK:     call {{.*objc_msgSend.*}}(i8* [[RECV]], i8* [[SEL]], i8* [[COERCE_CAST]], i8* {{.*}}[[RANGE_STR]]{{.*}}) [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   NSValue *range_rvalue = @(getRange());
   // CHECK:     call void @llvm.objc.release
   // CHECK:     ret void
diff --git a/clang/test/CodeGenObjC/os_log.m b/clang/test/CodeGenObjC/os_log.m
--- a/clang/test/CodeGenObjC/os_log.m
+++ b/clang/test/CodeGenObjC/os_log.m
@@ -26,10 +26,11 @@
 // CHECK-O2: %[[V0:.*]] = call i8* @llvm.objc.retain(
 // CHECK-O2: store i8* %[[V0]], i8** %[[A_ADDR]], align 8,
 // CHECK-O0: call void @llvm.objc.storeStrong(i8** %[[A_ADDR]], i8* %{{.*}})
-// CHECK: %[[CALL:.*]] = call %{{.*}}* (...) @GenString()
-// CHECK: %[[V2:.*]] = bitcast %{{.*}}* %[[CALL]] to i8*
-// CHECK: %[[V3:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[V2]])
-// CHECK: %[[V4:.*]] = bitcast i8* %[[V3]] to %{{.*}}*
+// CHECK-O2: %[[V4:.*]] = call %{{.*}}* (...) @GenString() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-O0: %[[CALL:.*]] = call %{{.*}}* (...) @GenString()
+// CHECK-O0: %[[V2:.*]] = bitcast %{{.*}}* %[[CALL]] to i8*
+// CHECK-O0: %[[V3:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[V2]])
+// CHECK-O0: %[[V4:.*]] = bitcast i8* %[[V3]] to %{{.*}}*
 // CHECK: %[[V5:.*]] = bitcast %{{.*}}* %[[V4]] to i8*
 // CHECK: %[[V6:.*]] = call i8* @llvm.objc.retain(i8* %[[V5]])
 // CHECK: %[[V7:.*]] = bitcast i8* %[[V6]] to %{{.*}}*
@@ -71,10 +72,11 @@
 // CHECK-LABEL: define{{.*}} void @test_builtin_os_log3(
 // CHECK: alloca i8*, align 8
 // CHECK: %[[OS_LOG_ARG:.*]] = alloca i8*, align 8
-// CHECK: %[[CALL:.*]] = call %{{.*}}* (...) @GenString()
-// CHECK: %[[V1:.*]] = bitcast %{{.*}}* %[[CALL]] to i8*
-// CHECK: %[[V2:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[V1]])
-// CHECK: %[[V3:.*]] = bitcast i8* %[[V2]] to %{{.*}}*
+// CHECK-O2: %[[V3:.*]] = call %{{.*}}* (...) @GenString() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-O0: %[[CALL:.*]] = call %{{.*}}* (...) @GenString()
+// CHECK-O0: %[[V1:.*]] = bitcast %{{.*}}* %[[CALL]] to i8*
+// CHECK-O0: %[[V2:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[V1]])
+// CHECK-O0: %[[V3:.*]] = bitcast i8* %[[V2]] to %{{.*}}*
 // CHECK: %[[V4:.*]] = bitcast %{{.*}}* %[[V3]] to i8*
 // CHECK: %[[V5:.*]] = call i8* @llvm.objc.retain(i8* %[[V4]])
 // CHECK: store i8* %[[V5]], i8** %[[OS_LOG_ARG]], align 8
@@ -97,13 +99,15 @@
 // CHECK: alloca i8*, align 8
 // CHECK: %[[OS_LOG_ARG:.*]] = alloca i8*, align 8
 // CHECK: %[[OS_LOG_ARG2:.*]] = alloca i8*, align 8
-// CHECK: %[[CALL:.*]] = call {{.*}} @objc_msgSend
-// CHECK: %[[V4:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[CALL]])
+// CHECK-O2: %[[V4:.*]] = call {{.*}} @objc_msgSend{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-O0: %[[CALL:.*]] = call {{.*}} @objc_msgSend
+// CHECK-O0: %[[V4:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[CALL]])
 // CHECK: %[[V5:.*]] = call i8* @llvm.objc.retain(i8* %[[V4]])
 // CHECK: store i8* %[[V5]], i8** %[[OS_LOG_ARG]], align 8
 // CHECK: %[[V6:.*]] = ptrtoint i8* %[[V5]] to i64
-// CHECK: %[[CALL1:.*]] = call {{.*}} @objc_msgSend
-// CHECK: %[[V10:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[CALL1]])
+// CHECK-O2: %[[V10:.*]] = call {{.*}} @objc_msgSend{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-O0: %[[CALL1:.*]] = call {{.*}} @objc_msgSend
+// CHECK-O0: %[[V10:.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* %[[CALL1]])
 // CHECK: %[[V11:.*]] = call i8* @llvm.objc.retain(i8* %[[V10]])
 // CHECK: store i8* %[[V11]], i8** %[[OS_LOG_ARG2]], align 8
 // CHECK: %[[V12:.*]] = ptrtoint i8* %[[V11]] to i64
diff --git a/clang/test/CodeGenObjCXX/arc-forwarded-lambda-call.mm b/clang/test/CodeGenObjCXX/arc-forwarded-lambda-call.mm
--- a/clang/test/CodeGenObjCXX/arc-forwarded-lambda-call.mm
+++ b/clang/test/CodeGenObjCXX/arc-forwarded-lambda-call.mm
@@ -4,9 +4,9 @@
   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$_0clEv"
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
-  // CHECK-NEXT: [[T2:%.*]] = tail call i8* @llvm.objc.autoreleaseReturnValue(i8* [[T1]])
+  // CHECK: [[T0:%.*]] = call i8* @"_ZZ5test0P11objc_objectENK3$_0clEv"{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T0]])
+  // CHECK-NEXT: [[T2:%.*]] = tail call i8* @llvm.objc.autoreleaseReturnValue(i8* [[T0]])
   // CHECK-NEXT: ret i8* [[T2]]
 }
 
@@ -27,9 +27,9 @@
   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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
-  // CHECK-NEXT: [[T2:%.*]] = tail call i8* @llvm.objc.autoreleaseReturnValue(i8* [[T1]])
+  // CHECK: [[T0:%.*]] = call i8* @"_ZZ5test1vENK3$_1clEv"{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T0]])
+  // CHECK-NEXT: [[T2:%.*]] = tail call i8* @llvm.objc.autoreleaseReturnValue(i8* [[T0]])
   // CHECK-NEXT: ret i8* [[T2]]
 }
 
diff --git a/clang/test/CodeGenObjCXX/arc-globals.mm b/clang/test/CodeGenObjCXX/arc-globals.mm
--- a/clang/test/CodeGenObjCXX/arc-globals.mm
+++ b/clang/test/CodeGenObjCXX/arc-globals.mm
@@ -6,15 +6,15 @@
 id getObject();
 
 // CHECK-LABEL: define internal void @__cxx_global_var_init
-// CHECK: call i8* @_Z9getObjectv
-// CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue
+// CHECK: call i8* @_Z9getObjectv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use
 // CHECK-NEXT: {{store i8*.*@global_obj}}
 // CHECK-NEXT: ret void
 id global_obj = getObject();
 
 // CHECK-LABEL: define internal void @__cxx_global_var_init
-// CHECK: call i8* @_Z9getObjectv
-// CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue
+// CHECK: call i8* @_Z9getObjectv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use
 // CHECK-NEXT: {{store i8*.*@global_obj2}}
 // CHECK-NEXT: ret void
 id global_obj2 = getObject();
diff --git a/clang/test/CodeGenObjCXX/arc-references.mm b/clang/test/CodeGenObjCXX/arc-references.mm
--- a/clang/test/CodeGenObjCXX/arc-references.mm
+++ b/clang/test/CodeGenObjCXX/arc-references.mm
@@ -9,13 +9,13 @@
 // Lifetime extension for binding a reference to an rvalue
 // CHECK-LABEL: define{{.*}} void @_Z5test0v()
 void test0() {
-  // CHECK: call i8* @_Z9getObjectv
-  // CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK: call i8* @_Z9getObjectv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(
   const __strong id &ref1 = getObject();
   // CHECK: call void @_Z6calleev
   callee();
-  // CHECK: call i8* @_Z9getObjectv
-  // CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK: call i8* @_Z9getObjectv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(
   // CHECK-NEXT: call i8* @llvm.objc.autorelease
   const __autoreleasing id &ref2 = getObject();
   // CHECK: call void @_Z6calleev
@@ -84,8 +84,8 @@
 }
 
 // CHECK-LABEL: define internal void @__cxx_global_var_init(
-// CHECK: call i8* @_Z9getObjectv
-// CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue
+// CHECK: call i8* @_Z9getObjectv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(
 const __strong id &global_ref = getObject();
 
 // Note: we intentionally don't release the object.
diff --git a/clang/test/CodeGenObjCXX/arc.mm b/clang/test/CodeGenObjCXX/arc.mm
--- a/clang/test/CodeGenObjCXX/arc.mm
+++ b/clang/test/CodeGenObjCXX/arc.mm
@@ -19,8 +19,8 @@
   // 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK:      [[T1:%.*]] = call i8* @_Z12test0_helperv() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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]])
@@ -28,8 +28,8 @@
   // CHECK-NEXT: call void @llvm.objc.release(i8* [[T1]])
   id x = *wp = test0_helper();
 
-  // CHECK:      [[T0:%.*]] = call i8* @_Z12test0_helperv()
-  // CHECK-NEXT: [[T1:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T0]])
+  // CHECK:      [[T1:%.*]] = call i8* @_Z12test0_helperv() [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(i8* [[T1]])
   // 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]])
@@ -163,25 +163,20 @@
 // CHECK-LABEL: define{{.*}} void @_Z7test35b13Test35_HelperPS_
 void test35b(Test35_Helper x0, Test35_Helper *x0p) {
   // CHECK: call void @llvm.lifetime.start
-  // CHECK: call i8* @_ZN13Test35_Helper11makeObject3Ev
-  // CHECK: call i8* @llvm.objc.retain
+  // CHECK: call i8* @_ZN13Test35_Helper11makeObject3Ev{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   id obj1 = Test35_Helper::makeObject3();
   // CHECK: call void @llvm.lifetime.start
-  // CHECK: call i8* @_ZN13Test35_Helper11makeObject4Ev
-  // CHECK: call i8* @llvm.objc.retain
+  // CHECK: call i8* @_ZN13Test35_Helper11makeObject4Ev{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   id obj2 = x0.makeObject4();
   // CHECK: call void @llvm.lifetime.start
-  // CHECK: call i8* @_ZN13Test35_Helper11makeObject4Ev
-  // CHECK: call i8* @llvm.objc.retain
+  // CHECK: call i8* @_ZN13Test35_Helper11makeObject4Ev{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   id obj3 = x0p->makeObject4();
   id (Test35_Helper::*pmf)() = &Test35_Helper::makeObject4;
   // CHECK: call void @llvm.lifetime.start
-  // CHECK: call i8* %
-  // CHECK: call i8* @llvm.objc.retain
+  // CHECK: call i8* %{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   id obj4 = (x0.*pmf)();
   // CHECK: call void @llvm.lifetime.start
-  // CHECK: call i8* %
-  // CHECK: call i8* @llvm.objc.retain
+  // CHECK: call i8* %{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   id obj5 = (x0p->*pmf)();
 
   // CHECK: call void @llvm.objc.release
@@ -222,26 +217,24 @@
 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:%.*]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[T1]])
-// CHECK-NEXT: [[COLL:%.*]] = bitcast i8* [[T2]] to [[NSARRAY]]*
+// CHECK:      [[T2:%.*]] = call [[NSARRAY]]* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend to [[NSARRAY]]* (i8*, i8*)*)({{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use({{.*}} [[T2]])
 
 // 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: [[T0:%.*]] = bitcast [[NSARRAY]]* [[T2]] to i8*
 // CHECK-NEXT: @objc_msgSend
 
 // This bitcast is for the mutation check.
-// CHECK:      [[T0:%.*]] = bitcast [[NSARRAY]]* [[COLL]] to i8*
+// CHECK:      [[T0:%.*]] = bitcast [[NSARRAY]]* [[T2]] to i8*
 // CHECK-NEXT: @objc_enumerationMutation
 
 // This bitcast is for the 'next' message send.
-// CHECK:      [[T0:%.*]] = bitcast [[NSARRAY]]* [[COLL]] to i8*
+// CHECK:      [[T0:%.*]] = bitcast [[NSARRAY]]* [[T2]] to i8*
 // CHECK-NEXT: @objc_msgSend
 
 // This bitcast is for the final release.
-// CHECK:      [[T0:%.*]] = bitcast [[NSARRAY]]* [[COLL]] to i8*
+// CHECK:      [[T0:%.*]] = bitcast [[NSARRAY]]* [[T2]] to i8*
 // CHECK-NEXT: call void @llvm.objc.release(i8* [[T0]])
 
 template<typename T>
@@ -250,10 +243,8 @@
 }
 
 // CHECK-LABEL: define weak_odr void @_Z12send_releaseIiEvv(
-// CHECK: call %0* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend
-// CHECK-NEXT: bitcast
-// CHECK-NEXT: call i8* @llvm.objc.retainAutoreleasedReturnValue
-// CHECK-NEXT: bitcast
+// CHECK: call %0* bitcast (i8* (i8*, i8*, ...)* @objc_msgSend{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+// CHECK-NEXT: call void (...) @llvm.objc.clang.arc.noop.use(
 // CHECK-NEXT: bitcast
 // CHECK-NEXT: call void @llvm.objc.release
 // CHECK-NEXT: ret void
diff --git a/clang/test/CodeGenObjCXX/literals.mm b/clang/test/CodeGenObjCXX/literals.mm
--- a/clang/test/CodeGenObjCXX/literals.mm
+++ b/clang/test/CodeGenObjCXX/literals.mm
@@ -28,22 +28,19 @@
   // 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.]+]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT0]])
-  // CHECK: store i8* [[RET0]], i8** [[ELEMENT0]]
+  // CHECK-NEXT: [[OBJECT0:%[a-zA-Z0-9.]+]] = invoke i8* @_ZNK1XcvP11objc_objectEv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK: store i8* [[OBJECT0]], 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.]+]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT1]])
-  // CHECK: store i8* [[RET1]], i8** [[ELEMENT1]]
+  // CHECK: [[OBJECT1:%[a-zA-Z0-9.]+]] = invoke i8* @_ZNK1YcvP11objc_objectEv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK: store i8* [[OBJECT1]], i8** [[ELEMENT1]]
 
   // Build the array
-  // CHECK: {{invoke.*@objc_msgSend}}
-  // CHECK: call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK: {{invoke.*@objc_msgSend}}{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   id arr = @[ X(), Y() ];
 
   // Destroy temporaries
@@ -82,20 +79,17 @@
   // 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.]+]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT0]])
-  // CHECK: store i8* [[RET0]], i8** [[ELEMENT0]]
+  // CHECK-NEXT: [[OBJECT0:%[a-zA-Z0-9.]+]] = invoke i8* @_ZNK1XcvP11objc_objectEv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK: store i8* [[OBJECT0]], 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.]+]] = notail call i8* @llvm.objc.retainAutoreleasedReturnValue(i8* [[OBJECT1]])
-  // CHECK: store i8* [[RET1]], i8** [[ELEMENT1]]
+  // CHECK: [[OBJECT1:%[a-zA-Z0-9.]+]] = invoke i8* @_ZNK1YcvP11objc_objectEv{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
+  // CHECK: store i8* [[OBJECT1]], i8** [[ELEMENT1]]
 
   // Build the array
-  // CHECK: {{invoke.*@objc_msgSend}}
-  // CHECK: call i8* @llvm.objc.retainAutoreleasedReturnValue
+  // CHECK: {{invoke.*@objc_msgSend}}{{.*}} [ "clang.arc.attachedcall"(i8* (i8*)* @llvm.objc.retainAutoreleasedReturnValue) ]
   id arr = @[ X(), Y() ];
 
   // Destroy temporaries