ARM: implement low-level intrinsics for the atomic exclusive operations.

This adds three overloaded intrinsics to Clang:
    T __builtin_arm_ldrex(const volatile T *addr)
    int __builtin_arm_strex(T val, volatile T *addr)
    void __builtin_arm_clrex()

The intent is that these do what users would expect when given most sensible
types. Currently, "sensible" translates to ints, floats and pointers.

llvm-svn: 186394

clang/docs/LanguageExtensions.rst

@@ -1656,6 +1656,36 @@ C11's ``<stdatomic.h>`` header.  These builtins provide the semantics of the
 * ``__c11_atomic_fetch_or``
 * ``__c11_atomic_fetch_xor``
 
+Low-level ARM exclusive memory builtins
+---------------------------------------
+
+Clang provides overloaded builtins giving direct access to the three key ARM
+instructions for implementing atomic operations.
+
+.. code-block:: c
+  T __builtin_arm_ldrex(const volatile T *addr);
+  int __builtin_arm_strex(T val, volatile T *addr);
+  void __builtin_arm_clrex(void);
+
+The types ``T`` currently supported are:
+* Integer types with width at most 64 bits.
+* Floating-point types
+* Pointer types.
+
+Note that the compiler does not guarantee it will not insert stores which clear
+the exclusive monitor in between an ``ldrex`` and its paired ``strex``. In
+practice this is only usually a risk when the extra store is on the same cache
+line as the variable being modified and Clang will only insert stack stores on
+its own, so it is best not to use these operations on variables with automatic
+storage duration.
+
+Also, loads and stores may be implicit in code written between the ``ldrex`` and
+``strex``. Clang will not necessarily mitigate the effects of these either, so
+care should be exercised.
+
+For these reasons the higher level atomic primitives should be preferred where
+possible.
+
 Non-standard C++11 Attributes
 =============================
 

clang/include/clang/Basic/BuiltinsARM.def

@@ -24,10 +24,14 @@ BUILTIN(__builtin_arm_qsub, "iii", "nc")
 BUILTIN(__builtin_arm_ssat, "iiUi", "nc")
 BUILTIN(__builtin_arm_usat, "UiUiUi", "nc")
 
-// Store and load exclusive doubleword
+// Store and load exclusive
 BUILTIN(__builtin_arm_ldrexd, "LLUiv*", "")
 BUILTIN(__builtin_arm_strexd, "iLLUiv*", "")
 
+BUILTIN(__builtin_arm_ldrex, "v.", "t")
+BUILTIN(__builtin_arm_strex, "i.", "t")
+BUILTIN(__builtin_arm_clrex, "v", "")
+
 // VFP
 BUILTIN(__builtin_arm_get_fpscr, "Ui", "nc")
 BUILTIN(__builtin_arm_set_fpscr, "vUi", "nc")

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -5384,27 +5384,33 @@ def err_builtin_fn_use : Error<"builtin functions must be directly called">;
 def warn_call_wrong_number_of_arguments : Warning<
   "too %select{few|many}0 arguments in call to %1">;
 def err_atomic_builtin_must_be_pointer : Error<
-  "first argument to atomic builtin must be a pointer (%0 invalid)">;
+  "address argument to atomic builtin must be a pointer (%0 invalid)">;
 def err_atomic_builtin_must_be_pointer_intptr : Error<
-  "first argument to atomic builtin must be a pointer to integer or pointer"
+  "address argument to atomic builtin must be a pointer to integer or pointer"
   " (%0 invalid)">;
+def err_atomic_builtin_must_be_pointer_intfltptr : Error<
+  "address argument to atomic builtin must be a pointer to integer,"
+  " floating-point or pointer (%0 invalid)">;
 def err_atomic_builtin_pointer_size : Error<
-  "first argument to atomic builtin must be a pointer to 1,2,4,8 or 16 byte "
+  "address argument to atomic builtin must be a pointer to 1,2,4,8 or 16 byte "
   "type (%0 invalid)">;
+def err_atomic_exclusive_builtin_pointer_size : Error<
+  "address argument to load or store exclusive builtin must be a pointer to"
+  " 1,2,4 or 8 byte type (%0 invalid)">;
 def err_atomic_op_needs_atomic : Error<
-  "first argument to atomic operation must be a pointer to _Atomic "
+  "address argument to atomic operation must be a pointer to _Atomic "
   "type (%0 invalid)">;
 def err_atomic_op_needs_non_const_atomic : Error<
-  "first argument to atomic operation must be a pointer to non-const _Atomic "
+  "address argument to atomic operation must be a pointer to non-const _Atomic "
   "type (%0 invalid)">;
 def err_atomic_op_needs_trivial_copy : Error<
-  "first argument to atomic operation must be a pointer to a trivially-copyable"
-  " type (%0 invalid)">;
+  "address argument to atomic operation must be a pointer to a "
+  "trivially-copyable type (%0 invalid)">;
 def err_atomic_op_needs_atomic_int_or_ptr : Error<
-  "first argument to atomic operation must be a pointer to %select{|atomic }0"
+  "address argument to atomic operation must be a pointer to %select{|atomic }0"
   "integer or pointer (%1 invalid)">;
 def err_atomic_op_bitwise_needs_atomic_int : Error<
-  "first argument to bitwise atomic operation must be a pointer to "
+  "address argument to bitwise atomic operation must be a pointer to "
   "%select{|atomic }0integer (%1 invalid)">;
   
 def err_atomic_load_store_uses_lib : Error<

clang/include/clang/Sema/Sema.h

@@ -7499,7 +7499,10 @@ private:
   bool CheckObjCString(Expr *Arg);
 
   ExprResult CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall);
+
+  bool CheckARMBuiltinExclusiveCall(unsigned BuiltinID, CallExpr *TheCall);
   bool CheckARMBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall);
+
   bool CheckMipsBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall);
 
   bool SemaBuiltinVAStart(CallExpr *TheCall);

clang/lib/CodeGen/CGBuiltin.cpp

@@ -1737,11 +1737,14 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
     return EmitNounwindRuntimeCall(CGM.CreateRuntimeFunction(FTy, Name), Ops);
   }
 
-  if (BuiltinID == ARM::BI__builtin_arm_ldrexd) {
+  if (BuiltinID == ARM::BI__builtin_arm_ldrexd ||
+      (BuiltinID == ARM::BI__builtin_arm_ldrex &&
+       getContext().getTypeSize(E->getType()) == 64)) {
     Function *F = CGM.getIntrinsic(Intrinsic::arm_ldrexd);
 
     Value *LdPtr = EmitScalarExpr(E->getArg(0));
-    Value *Val = Builder.CreateCall(F, LdPtr, "ldrexd");
+    Value *Val = Builder.CreateCall(F, Builder.CreateBitCast(LdPtr, Int8PtrTy),
+                                    "ldrexd");
 
     Value *Val0 = Builder.CreateExtractValue(Val, 1);
     Value *Val1 = Builder.CreateExtractValue(Val, 0);
@@ -1750,15 +1753,39 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
 
     Value *ShiftCst = llvm::ConstantInt::get(Int64Ty, 32);
     Val = Builder.CreateShl(Val0, ShiftCst, "shl", true /* nuw */);
-    return Builder.CreateOr(Val, Val1);
+    Val = Builder.CreateOr(Val, Val1);
+    return Builder.CreateBitCast(Val, ConvertType(E->getType()));
   }
 
-  if (BuiltinID == ARM::BI__builtin_arm_strexd) {
+  if (BuiltinID == ARM::BI__builtin_arm_ldrex) {
+    Value *LoadAddr = EmitScalarExpr(E->getArg(0));
+
+    QualType Ty = E->getType();
+    llvm::Type *RealResTy = ConvertType(Ty);
+    llvm::Type *IntResTy = llvm::IntegerType::get(getLLVMContext(),
+                                                  getContext().getTypeSize(Ty));
+    LoadAddr = Builder.CreateBitCast(LoadAddr, IntResTy->getPointerTo());
+
+    Function *F = CGM.getIntrinsic(Intrinsic::arm_ldrex, LoadAddr->getType());
+    Value *Val = Builder.CreateCall(F, LoadAddr, "ldrex");
+
+    if (RealResTy->isPointerTy())
+      return Builder.CreateIntToPtr(Val, RealResTy);
+    else {
+      Val = Builder.CreateTruncOrBitCast(Val, IntResTy);
+      return Builder.CreateBitCast(Val, RealResTy);
+    }
+  }
+
+  if (BuiltinID == ARM::BI__builtin_arm_strexd ||
+      (BuiltinID == ARM::BI__builtin_arm_strex &&
+       getContext().getTypeSize(E->getArg(0)->getType()) == 64)) {
     Function *F = CGM.getIntrinsic(Intrinsic::arm_strexd);
     llvm::Type *STy = llvm::StructType::get(Int32Ty, Int32Ty, NULL);
 
     Value *One = llvm::ConstantInt::get(Int32Ty, 1);
-    Value *Tmp = Builder.CreateAlloca(Int64Ty, One);
+    Value *Tmp = Builder.CreateAlloca(ConvertType(E->getArg(0)->getType()),
+                                      One);
     Value *Val = EmitScalarExpr(E->getArg(0));
     Builder.CreateStore(Val, Tmp);
 
@@ -1767,10 +1794,35 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
 
     Value *Arg0 = Builder.CreateExtractValue(Val, 0);
     Value *Arg1 = Builder.CreateExtractValue(Val, 1);
-    Value *StPtr = EmitScalarExpr(E->getArg(1));
+    Value *StPtr = Builder.CreateBitCast(EmitScalarExpr(E->getArg(1)), Int8PtrTy);
     return Builder.CreateCall3(F, Arg0, Arg1, StPtr, "strexd");
   }
 
+  if (BuiltinID == ARM::BI__builtin_arm_strex) {
+    Value *StoreVal = EmitScalarExpr(E->getArg(0));
+    Value *StoreAddr = EmitScalarExpr(E->getArg(1));
+
+    QualType Ty = E->getArg(0)->getType();
+    llvm::Type *StoreTy = llvm::IntegerType::get(getLLVMContext(),
+                                                 getContext().getTypeSize(Ty));
+    StoreAddr = Builder.CreateBitCast(StoreAddr, StoreTy->getPointerTo());
+
+    if (StoreVal->getType()->isPointerTy())
+      StoreVal = Builder.CreatePtrToInt(StoreVal, Int32Ty);
+    else {
+      StoreVal = Builder.CreateBitCast(StoreVal, StoreTy);
+      StoreVal = Builder.CreateZExtOrBitCast(StoreVal, Int32Ty);
+    }
+
+    Function *F = CGM.getIntrinsic(Intrinsic::arm_strex, StoreAddr->getType());
+    return Builder.CreateCall2(F, StoreVal, StoreAddr, "strex");
+  }
+
+  if (BuiltinID == ARM::BI__builtin_arm_clrex) {
+    Function *F = CGM.getIntrinsic(Intrinsic::arm_clrex);
+    return Builder.CreateCall(F);
+  }
+
   SmallVector<Value*, 4> Ops;
   llvm::Value *Align = 0;
   for (unsigned i = 0, e = E->getNumArgs() - 1; i != e; i++) {

clang/lib/Sema/SemaChecking.cpp

@@ -371,9 +371,117 @@ static QualType getNeonEltType(NeonTypeFlags Flags, ASTContext &Context) {
   llvm_unreachable("Invalid NeonTypeFlag!");
 }
 
+bool Sema::CheckARMBuiltinExclusiveCall(unsigned BuiltinID, CallExpr *TheCall) {
+  assert((BuiltinID == ARM::BI__builtin_arm_ldrex ||
+          BuiltinID == ARM::BI__builtin_arm_strex) &&
+         "unexpected ARM builtin");
+  bool IsLdrex = BuiltinID == ARM::BI__builtin_arm_ldrex;
+
+  DeclRefExpr *DRE =cast<DeclRefExpr>(TheCall->getCallee()->IgnoreParenCasts());
+
+  // Ensure that we have the proper number of arguments.
+  if (checkArgCount(*this, TheCall, IsLdrex ? 1 : 2))
+    return true;
+
+  // Inspect the pointer argument of the atomic builtin.  This should always be
+  // a pointer type, whose element is an integral scalar or pointer type.
+  // Because it is a pointer type, we don't have to worry about any implicit
+  // casts here.
+  Expr *PointerArg = TheCall->getArg(IsLdrex ? 0 : 1);
+  ExprResult PointerArgRes = DefaultFunctionArrayLvalueConversion(PointerArg);
+  if (PointerArgRes.isInvalid())
+    return true;
+  PointerArg = PointerArgRes.take();
+
+  const PointerType *pointerType = PointerArg->getType()->getAs<PointerType>();
+  if (!pointerType) {
+    Diag(DRE->getLocStart(), diag::err_atomic_builtin_must_be_pointer)
+      << PointerArg->getType() << PointerArg->getSourceRange();
+    return true;
+  }
+
+  // ldrex takes a "const volatile T*" and strex takes a "volatile T*". Our next
+  // task is to insert the appropriate casts into the AST. First work out just
+  // what the appropriate type is.
+  QualType ValType = pointerType->getPointeeType();
+  QualType AddrType = ValType.getUnqualifiedType().withVolatile();
+  if (IsLdrex)
+    AddrType.addConst();
+
+  // Issue a warning if the cast is dodgy.
+  CastKind CastNeeded = CK_NoOp;
+  if (!AddrType.isAtLeastAsQualifiedAs(ValType)) {
+    CastNeeded = CK_BitCast;
+    Diag(DRE->getLocStart(), diag::ext_typecheck_convert_discards_qualifiers)
+      << PointerArg->getType()
+      << Context.getPointerType(AddrType)
+      << AA_Passing << PointerArg->getSourceRange();
+  }
+
+  // Finally, do the cast and replace the argument with the corrected version.
+  AddrType = Context.getPointerType(AddrType);
+  PointerArgRes = ImpCastExprToType(PointerArg, AddrType, CastNeeded);
+  if (PointerArgRes.isInvalid())
+    return true;
+  PointerArg = PointerArgRes.take();
+
+  TheCall->setArg(IsLdrex ? 0 : 1, PointerArg);
+
+  // In general, we allow ints, floats and pointers to be loaded and stored.
+  if (!ValType->isIntegerType() && !ValType->isAnyPointerType() &&
+      !ValType->isBlockPointerType() && !ValType->isFloatingType()) {
+    Diag(DRE->getLocStart(), diag::err_atomic_builtin_must_be_pointer_intfltptr)
+      << PointerArg->getType() << PointerArg->getSourceRange();
+    return true;
+  }
+
+  // But ARM doesn't have instructions to deal with 128-bit versions.
+  if (Context.getTypeSize(ValType) > 64) {
+    Diag(DRE->getLocStart(), diag::err_atomic_exclusive_builtin_pointer_size)
+      << PointerArg->getType() << PointerArg->getSourceRange();
+    return true;
+  }
+
+  switch (ValType.getObjCLifetime()) {
+  case Qualifiers::OCL_None:
+  case Qualifiers::OCL_ExplicitNone:
+    // okay
+    break;
+
+  case Qualifiers::OCL_Weak:
+  case Qualifiers::OCL_Strong:
+  case Qualifiers::OCL_Autoreleasing:
+    Diag(DRE->getLocStart(), diag::err_arc_atomic_ownership)
+      << ValType << PointerArg->getSourceRange();
+    return true;
+  }
+
+
+  if (IsLdrex) {
+    TheCall->setType(ValType);
+    return false;
+  }
+
+  // Initialize the argument to be stored.
+  ExprResult ValArg = TheCall->getArg(0);
+  InitializedEntity Entity = InitializedEntity::InitializeParameter(
+      Context, ValType, /*consume*/ false);
+  ValArg = PerformCopyInitialization(Entity, SourceLocation(), ValArg);
+  if (ValArg.isInvalid())
+    return true;
+
+  TheCall->setArg(0, ValArg.get());
+  return false;
+}
+
 bool Sema::CheckARMBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
   llvm::APSInt Result;
 
+  if (BuiltinID == ARM::BI__builtin_arm_ldrex ||
+      BuiltinID == ARM::BI__builtin_arm_strex) {
+    return CheckARMBuiltinExclusiveCall(BuiltinID, TheCall);
+  }
+
   uint64_t mask = 0;
   unsigned TV = 0;
   int PtrArgNum = -1;

clang/test/CodeGen/builtins-arm-exclusive.c

@@ -0,0 +1,113 @@
+// REQUIRES: arm-registered-target
+// RUN: %clang_cc1 -Wall -Werror -triple thumbv7-linux-gnueabi -fno-signed-char -O3 -emit-llvm -o - %s | FileCheck %s
+
+// Make sure the canonical use works before going into smaller details:
+int atomic_inc(int *addr) {
+  int Failure, OldVal;
+  do {
+    OldVal = __builtin_arm_ldrex(addr);
+    Failure = __builtin_arm_strex(OldVal + 1, addr);
+  } while (Failure);
+
+  return OldVal;
+}
+
+// CHECK: @atomic_inc
+// CHECK:   [[OLDVAL:%.*]] = tail call i32 @llvm.arm.ldrex.p0i32(i32* %addr)
+// CHECK:   [[INC:%.*]] = add nsw i32 [[OLDVAL]], 1
+// CHECK:   [[FAILURE:%.*]] = tail call i32 @llvm.arm.strex.p0i32(i32 [[INC]], i32* %addr)
+// CHECK:   [[TST:%.*]] = icmp eq i32 [[FAILURE]], 0
+// CHECK:   br i1 [[TST]], label %[[LOOP_END:[a-zA-Z0-9.]+]], label {{%[a-zA-Z0-9.]+}}
+// CHECK: [[LOOP_END]]:
+
+struct Simple {
+  char a, b;
+};
+
+int test_ldrex(char *addr, long long *addr64, float *addrfloat) {
+// CHECK: @test_ldrex
+  int sum = 0;
+  sum += __builtin_arm_ldrex(addr);
+// CHECK: [[INTRES:%.*]] = tail call i32 @llvm.arm.ldrex.p0i8(i8* %addr)
+// CHECK: and i32 [[INTRES]], 255
+
+  sum += __builtin_arm_ldrex((short *)addr);
+// CHECK: [[ADDR16:%.*]] = bitcast i8* %addr to i16*
+// CHECK: [[INTRES:%.*]] = tail call i32 @llvm.arm.ldrex.p0i16(i16* [[ADDR16]])
+// CHECK: [[TMPSEXT:%.*]] = shl i32 [[INTRES]], 16
+// CHECK: ashr exact i32 [[TMPSEXT]], 16
+
+  sum += __builtin_arm_ldrex((int *)addr);
+// CHECK: [[ADDR32:%.*]] = bitcast i8* %addr to i32*
+// CHECK:  call i32 @llvm.arm.ldrex.p0i32(i32* [[ADDR32]])
+
+  sum += __builtin_arm_ldrex((long long *)addr);
+// CHECK: call { i32, i32 } @llvm.arm.ldrexd(i8* %addr)
+
+  sum += __builtin_arm_ldrex(addr64);
+// CHECK: [[ADDR64_AS8:%.*]] = bitcast i64* %addr64 to i8*
+// CHECK: call { i32, i32 } @llvm.arm.ldrexd(i8* [[ADDR64_AS8]])
+
+  sum += __builtin_arm_ldrex(addrfloat);
+// CHECK: [[INTADDR:%.*]] = bitcast float* %addrfloat to i32*
+// CHECK: [[INTRES:%.*]] = tail call i32 @llvm.arm.ldrex.p0i32(i32* [[INTADDR]])
+// CHECK: bitcast i32 [[INTRES]] to float
+
+  sum += __builtin_arm_ldrex((double *)addr);
+// CHECK: [[STRUCTRES:%.*]] = tail call { i32, i32 } @llvm.arm.ldrexd(i8* %addr)
+// CHECK: [[RESHI:%.*]] = extractvalue { i32, i32 } [[STRUCTRES]], 1
+// CHECK: [[RESLO:%.*]] = extractvalue { i32, i32 } [[STRUCTRES]], 0
+// CHECK: [[RESHI64:%.*]] = zext i32 [[RESHI]] to i64
+// CHECK: [[RESLO64:%.*]] = zext i32 [[RESLO]] to i64
+// CHECK: [[RESHIHI:%.*]] = shl nuw i64 [[RESHI64]], 32
+// CHECK: [[INTRES:%.*]] = or i64 [[RESHIHI]], [[RESLO64]]
+// CHECK: bitcast i64 [[INTRES]] to double
+
+  sum += *__builtin_arm_ldrex((int **)addr);
+// CHECK: [[INTRES:%.*]] = tail call i32 @llvm.arm.ldrex.p0i32(i32* [[ADDR32]])
+// CHECK: inttoptr i32 [[INTRES]] to i32*
+
+  sum += __builtin_arm_ldrex((struct Simple **)addr)->a;
+// CHECK: [[INTRES:%.*]] = tail call i32 @llvm.arm.ldrex.p0i32(i32* [[ADDR32]])
+// CHECK: inttoptr i32 [[INTRES]] to %struct.Simple*
+
+  return sum;
+}
+
+int test_strex(char *addr) {
+// CHECK: @test_strex
+  int res = 0;
+  struct Simple var = {0};
+  res |= __builtin_arm_strex(4, addr);
+// CHECK: call i32 @llvm.arm.strex.p0i8(i32 4, i8* %addr)
+
+  res |= __builtin_arm_strex(42, (short *)addr);
+// CHECK: [[ADDR16:%.*]] = bitcast i8* %addr to i16*
+// CHECK:  call i32 @llvm.arm.strex.p0i16(i32 42, i16* [[ADDR16]])
+
+  res |= __builtin_arm_strex(42, (int *)addr);
+// CHECK: [[ADDR32:%.*]] = bitcast i8* %addr to i32*
+// CHECK: call i32 @llvm.arm.strex.p0i32(i32 42, i32* [[ADDR32]])
+
+  res |= __builtin_arm_strex(42, (long long *)addr);
+// CHECK: call i32 @llvm.arm.strexd(i32 42, i32 0, i8* %addr)
+
+  res |= __builtin_arm_strex(2.71828f, (float *)addr);
+// CHECK: call i32 @llvm.arm.strex.p0i32(i32 1076754509, i32* [[ADDR32]])
+
+  res |= __builtin_arm_strex(3.14159, (double *)addr);
+// CHECK: call i32 @llvm.arm.strexd(i32 -266631570, i32 1074340345, i8* %addr)
+
+  res |= __builtin_arm_strex(&var, (struct Simple **)addr);
+// CHECK: [[INTVAL:%.*]] = ptrtoint i16* %var to i32
+// CHECK: call i32 @llvm.arm.strex.p0i32(i32 [[INTVAL]], i32* [[ADDR32]])
+
+  return res;
+}
+
+void test_clrex() {
+// CHECK: @test_clrex
+
+  __builtin_arm_clrex();
+// CHECK: call void @llvm.arm.clrex()
+}

clang/test/Sema/atomic-ops.c

@@ -86,8 +86,8 @@ void f(_Atomic(int) *i, _Atomic(int*) *p, _Atomic(float) *d,
   __c11_atomic_init(I, 5); // expected-error {{pointer to _Atomic}}
   __c11_atomic_load(0); // expected-error {{too few arguments to function}}
   __c11_atomic_load(0,0,0); // expected-error {{too many arguments to function}}
-  __c11_atomic_store(0,0,0); // expected-error {{first argument to atomic builtin must be a pointer}}
-  __c11_atomic_store((int*)0,0,0); // expected-error {{first argument to atomic operation must be a pointer to _Atomic}}
+  __c11_atomic_store(0,0,0); // expected-error {{address argument to atomic builtin must be a pointer}}
+  __c11_atomic_store((int*)0,0,0); // expected-error {{address argument to atomic operation must be a pointer to _Atomic}}
 
   __c11_atomic_load(i, memory_order_seq_cst);
   __c11_atomic_load(p, memory_order_seq_cst);
@@ -169,9 +169,9 @@ void f(_Atomic(int) *i, _Atomic(int*) *p, _Atomic(float) *d,
   (int)__atomic_clear(&flag, memory_order_seq_cst); // expected-error {{operand of type 'void'}}
 
   const _Atomic(int) const_atomic;
-  __c11_atomic_init(&const_atomic, 0); // expected-error {{first argument to atomic operation must be a pointer to non-const _Atomic type ('const _Atomic(int) *' invalid)}}
-  __c11_atomic_store(&const_atomic, 0, memory_order_release); // expected-error {{first argument to atomic operation must be a pointer to non-const _Atomic type ('const _Atomic(int) *' invalid)}}
-  __c11_atomic_load(&const_atomic, memory_order_acquire); // expected-error {{first argument to atomic operation must be a pointer to non-const _Atomic type ('const _Atomic(int) *' invalid)}}
+  __c11_atomic_init(&const_atomic, 0); // expected-error {{address argument to atomic operation must be a pointer to non-const _Atomic type ('const _Atomic(int) *' invalid)}}
+  __c11_atomic_store(&const_atomic, 0, memory_order_release); // expected-error {{address argument to atomic operation must be a pointer to non-const _Atomic type ('const _Atomic(int) *' invalid)}}
+  __c11_atomic_load(&const_atomic, memory_order_acquire); // expected-error {{address argument to atomic operation must be a pointer to non-const _Atomic type ('const _Atomic(int) *' invalid)}}
 }
 
 _Atomic(int*) PR12527_a;

clang/test/Sema/builtins-arm-exclusive.c

@@ -0,0 +1,61 @@
+// RUN: %clang_cc1 -triple armv7 -fsyntax-only -verify %s
+
+struct Simple {
+  char a, b;
+};
+
+int test_ldrex(char *addr) {
+  int sum = 0;
+  sum += __builtin_arm_ldrex(addr);
+  sum += __builtin_arm_ldrex((short *)addr);
+  sum += __builtin_arm_ldrex((int *)addr);
+  sum += __builtin_arm_ldrex((long long *)addr);
+  sum += __builtin_arm_ldrex((float *)addr);
+  sum += __builtin_arm_ldrex((double *)addr);
+  sum += *__builtin_arm_ldrex((int **)addr);
+  sum += __builtin_arm_ldrex((struct Simple **)addr)->a;
+  sum += __builtin_arm_ldrex((volatile char *)addr);
+  sum += __builtin_arm_ldrex((const volatile char *)addr);
+
+  // In principle this might be valid, but stick to ints and floats for scalar
+  // types at the moment.
+  sum += __builtin_arm_ldrex((struct Simple *)addr).a; // expected-error {{address argument to atomic builtin must be a pointer to}}
+
+  sum += __builtin_arm_ldrex((__int128 *)addr); // expected-error {{__int128 is not supported on this target}} expected-error {{address argument to load or store exclusive builtin must be a pointer to 1,2,4 or 8 byte type}}
+
+  __builtin_arm_ldrex(); // expected-error {{too few arguments to function call}}
+  __builtin_arm_ldrex(1, 2); // expected-error {{too many arguments to function call}}
+  return sum;
+}
+
+int test_strex(char *addr) {
+  int res = 0;
+  struct Simple var = {0};
+  res |= __builtin_arm_strex(4, addr);
+  res |= __builtin_arm_strex(42, (short *)addr);
+  res |= __builtin_arm_strex(42, (int *)addr);
+  res |= __builtin_arm_strex(42, (long long *)addr);
+  res |= __builtin_arm_strex(2.71828f, (float *)addr);
+  res |= __builtin_arm_strex(3.14159, (double *)addr);
+  res |= __builtin_arm_strex(&var, (struct Simple **)addr);
+
+  res |= __builtin_arm_strex(42, (volatile char *)addr);
+  res |= __builtin_arm_strex(42, (char *const)addr);
+  res |= __builtin_arm_strex(42, (const char *)addr); // expected-warning {{passing 'const char *' to parameter of type 'volatile char *' discards qualifiers}}
+
+
+  res |= __builtin_arm_strex(var, (struct Simple *)addr); // expected-error {{address argument to atomic builtin must be a pointer to}}
+  res |= __builtin_arm_strex(var, (struct Simple **)addr); // expected-error {{passing 'struct Simple' to parameter of incompatible type 'struct Simple *'}}
+  res |= __builtin_arm_strex(&var, (struct Simple **)addr).a; // expected-error {{is not a structure or union}}
+
+  res |= __builtin_arm_strex(1, (__int128 *)addr); // expected-error {{__int128 is not supported on this target}} expected-error {{address argument to load or store exclusive builtin must be a pointer to 1,2,4 or 8 byte type}}
+
+  __builtin_arm_strex(1); // expected-error {{too few arguments to function call}}
+  __builtin_arm_strex(1, 2, 3); // expected-error {{too many arguments to function call}}
+  return res;
+}
+
+void test_clrex() {
+  __builtin_arm_clrex();
+  __builtin_arm_clrex(1); // expected-error {{too many arguments to function call}}
+}