[COFF, ARM64] Implement InterlockedExchange*_* builtins

[mgrang]

Summary

Windows SDK needs these intrinsics to be proper builtins. This is second in a series of patches to move intrinsic defintions out of intrin.h.

[efriedma]

LGTM

[rnk]

lgtm

include/clang/Basic/BuiltinsAArch64.def

 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd8_nf, "ccD*c", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd16_acq, "ssD*s", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd16_rel, "ssD*s", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd16_nf, "ssD*s", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd64_acq, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd64_rel, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd64_nf, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 
+TARGET_HEADER_BUILTIN(_InterlockedExchange8_acq,  "ccD*c",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange8_nf,   "ccD*c",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange8_rel,  "ccD*c",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange16_acq, "ssD*s",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange16_nf,  "ssD*s",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange16_rel, "ssD*s",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange_acq,   "LiLiD*Li",    "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange_nf,    "LiLiD*Li",    "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange_rel,   "LiLiD*Li",    "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange64_acq, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange64_nf,  "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange64_rel, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+
 TARGET_HEADER_BUILTIN(_ReadWriteBarrier, "v", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(__getReg, "ULLii", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_ReadStatusReg,  "ii",  "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_WriteStatusReg, "vii", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_AddressOfReturnAddress, "v*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 
 #undef BUILTIN
 #undef LANGBUILTIN
 #undef TARGET_HEADER_BUILTIN

include/clang/Basic/BuiltinsARM.def

 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd8_nf, "ccD*c", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd16_acq, "ssD*s", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd16_rel, "ssD*s", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd16_nf, "ssD*s", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd64_acq, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd64_rel, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedExchangeAdd64_nf, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 
+TARGET_HEADER_BUILTIN(_InterlockedExchange8_acq,  "ccD*c",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange8_nf,   "ccD*c",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange8_rel,  "ccD*c",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange16_acq, "ssD*s",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange16_nf,  "ssD*s",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange16_rel, "ssD*s",       "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange_acq,   "LiLiD*Li",    "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange_nf,    "LiLiD*Li",    "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange_rel,   "LiLiD*Li",    "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange64_acq, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange64_nf,  "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedExchange64_rel, "LLiLLiD*LLi", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+
 #undef BUILTIN
 #undef LANGBUILTIN
 #undef TARGET_HEADER_BUILTIN

lib/CodeGen/CGBuiltin.cpp

enum class CodeGenFunction::MSVCIntrin {
   _InterlockedExchangeAdd,
   _InterlockedExchangeSub,
   _InterlockedIncrement,
   _InterlockedOr,
   _InterlockedXor,
   _InterlockedExchangeAdd_acq,
   _InterlockedExchangeAdd_rel,
   _InterlockedExchangeAdd_nf,
+  _InterlockedExchange_acq,
+  _InterlockedExchange_rel,
+  _InterlockedExchange_nf,
   __fastfail,
 };
 
 Value *CodeGenFunction::EmitMSVCBuiltinExpr(MSVCIntrin BuiltinID,
                                             const CallExpr *E) {
   switch (BuiltinID) {
   case MSVCIntrin::_BitScanForward:
   case MSVCIntrin::_BitScanReverse: {
case MSVCIntrin::_InterlockedExchangeAdd_acq:
     return MakeBinaryAtomicValue(*this, AtomicRMWInst::Add, E,
                                  AtomicOrdering::Acquire);
   case MSVCIntrin::_InterlockedExchangeAdd_rel:
     return MakeBinaryAtomicValue(*this, AtomicRMWInst::Add, E,
                                  AtomicOrdering::Release);
   case MSVCIntrin::_InterlockedExchangeAdd_nf:
     return MakeBinaryAtomicValue(*this, AtomicRMWInst::Add, E,
                                  AtomicOrdering::Monotonic);
+  case MSVCIntrin::_InterlockedExchange_acq:
+    return MakeBinaryAtomicValue(*this, AtomicRMWInst::Xchg, E,
+                                 AtomicOrdering::Acquire);
+  case MSVCIntrin::_InterlockedExchange_rel:
+    return MakeBinaryAtomicValue(*this, AtomicRMWInst::Xchg, E,
+                                 AtomicOrdering::Release);
+  case MSVCIntrin::_InterlockedExchange_nf:
+    return MakeBinaryAtomicValue(*this, AtomicRMWInst::Xchg, E,
+                                 AtomicOrdering::Monotonic);
 
   case MSVCIntrin::_InterlockedDecrement: {
     llvm::Type *IntTy = ConvertType(E->getType());
     AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
       AtomicRMWInst::Sub,
       EmitScalarExpr(E->getArg(0)),
       ConstantInt::get(IntTy, 1),
       llvm::AtomicOrdering::SequentiallyConsistent);
Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
   case ARM::BI_InterlockedExchangeAdd_rel:
   case ARM::BI_InterlockedExchangeAdd64_rel:
     return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd_rel, E);
   case ARM::BI_InterlockedExchangeAdd8_nf:
   case ARM::BI_InterlockedExchangeAdd16_nf:
   case ARM::BI_InterlockedExchangeAdd_nf:
   case ARM::BI_InterlockedExchangeAdd64_nf:
     return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd_nf, E);
+  case ARM::BI_InterlockedExchange8_acq:
+  case ARM::BI_InterlockedExchange16_acq:
+  case ARM::BI_InterlockedExchange_acq:
+  case ARM::BI_InterlockedExchange64_acq:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange_acq, E);
+  case ARM::BI_InterlockedExchange8_rel:
+  case ARM::BI_InterlockedExchange16_rel:
+  case ARM::BI_InterlockedExchange_rel:
+  case ARM::BI_InterlockedExchange64_rel:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange_rel, E);
+  case ARM::BI_InterlockedExchange8_nf:
+  case ARM::BI_InterlockedExchange16_nf:
+  case ARM::BI_InterlockedExchange_nf:
+  case ARM::BI_InterlockedExchange64_nf:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange_nf, E);
   }
 
   // Get the last argument, which specifies the vector type.
   assert(HasExtraArg);
   llvm::APSInt Result;
   const Expr *Arg = E->getArg(E->getNumArgs()-1);
   if (!Arg->isIntegerConstantExpr(Result, getContext()))
     return nullptr;

   case AArch64::BI_InterlockedExchangeAdd_rel:
   case AArch64::BI_InterlockedExchangeAdd64_rel:
     return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd_rel, E);
   case AArch64::BI_InterlockedExchangeAdd8_nf:
   case AArch64::BI_InterlockedExchangeAdd16_nf:
   case AArch64::BI_InterlockedExchangeAdd_nf:
   case AArch64::BI_InterlockedExchangeAdd64_nf:
     return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd_nf, E);
+  case AArch64::BI_InterlockedExchange8_acq:
+  case AArch64::BI_InterlockedExchange16_acq:
+  case AArch64::BI_InterlockedExchange_acq:
+  case AArch64::BI_InterlockedExchange64_acq:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange_acq, E);
+  case AArch64::BI_InterlockedExchange8_rel:
+  case AArch64::BI_InterlockedExchange16_rel:
+  case AArch64::BI_InterlockedExchange_rel:
+  case AArch64::BI_InterlockedExchange64_rel:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange_rel, E);
+  case AArch64::BI_InterlockedExchange8_nf:
+  case AArch64::BI_InterlockedExchange16_nf:
+  case AArch64::BI_InterlockedExchange_nf:
+  case AArch64::BI_InterlockedExchange64_nf:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange_nf, E);
 
   case AArch64::BI_InterlockedAdd: {
     Value *Arg0 = EmitScalarExpr(E->getArg(0));
     Value *Arg1 = EmitScalarExpr(E->getArg(1));
     AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
       AtomicRMWInst::Add, Arg0, Arg1,
       llvm::AtomicOrdering::SequentiallyConsistent);
     return Builder.CreateAdd(RMWI, Arg1);

lib/Headers/intrin.h

 _InterlockedXor64_rel(__int64 volatile *_Value, __int64 _Mask) {
   return __atomic_fetch_xor(_Value, _Mask, __ATOMIC_RELEASE);
 }
 #endif
 /*----------------------------------------------------------------------------*\
 |* Interlocked Exchange
 \*----------------------------------------------------------------------------*/
 #if defined(__arm__) || defined(__aarch64__)
-static __inline__ char __DEFAULT_FN_ATTRS
-_InterlockedExchange8_acq(char volatile *_Target, char _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_ACQUIRE);
-  return _Value;
-}
-static __inline__ char __DEFAULT_FN_ATTRS
-_InterlockedExchange8_nf(char volatile *_Target, char _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_RELAXED);
-  return _Value;
-}
-static __inline__ char __DEFAULT_FN_ATTRS
-_InterlockedExchange8_rel(char volatile *_Target, char _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_RELEASE);
-  return _Value;
-}
-static __inline__ short __DEFAULT_FN_ATTRS
-_InterlockedExchange16_acq(short volatile *_Target, short _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_ACQUIRE);
-  return _Value;
-}
-static __inline__ short __DEFAULT_FN_ATTRS
-_InterlockedExchange16_nf(short volatile *_Target, short _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_RELAXED);
-  return _Value;
-}
-static __inline__ short __DEFAULT_FN_ATTRS
-_InterlockedExchange16_rel(short volatile *_Target, short _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_RELEASE);
-  return _Value;
-}
-static __inline__ long __DEFAULT_FN_ATTRS
-_InterlockedExchange_acq(long volatile *_Target, long _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_ACQUIRE);
-  return _Value;
-}
-static __inline__ long __DEFAULT_FN_ATTRS
-_InterlockedExchange_nf(long volatile *_Target, long _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_RELAXED);
-  return _Value;
-}
-static __inline__ long __DEFAULT_FN_ATTRS
-_InterlockedExchange_rel(long volatile *_Target, long _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_RELEASE);
-  return _Value;
-}
-static __inline__ __int64 __DEFAULT_FN_ATTRS
-_InterlockedExchange64_acq(__int64 volatile *_Target, __int64 _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_ACQUIRE);
-  return _Value;
-}
-static __inline__ __int64 __DEFAULT_FN_ATTRS
-_InterlockedExchange64_nf(__int64 volatile *_Target, __int64 _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_RELAXED);
-  return _Value;
-}
-static __inline__ __int64 __DEFAULT_FN_ATTRS
-_InterlockedExchange64_rel(__int64 volatile *_Target, __int64 _Value) {
-  __atomic_exchange(_Target, &_Value, &_Value, __ATOMIC_RELEASE);
-  return _Value;
-}
+char _InterlockedExchange8_acq(char volatile *_Target, char _Value);
+char _InterlockedExchange8_nf(char volatile *_Target, char _Value);
+char _InterlockedExchange8_rel(char volatile *_Target, char _Value);
+short _InterlockedExchange16_acq(short volatile *_Target, short _Value);
+short _InterlockedExchange16_nf(short volatile *_Target, short _Value);
+short _InterlockedExchange16_rel(short volatile *_Target, short _Value);
+long _InterlockedExchange_acq(long volatile *_Target, long _Value);
+long _InterlockedExchange_nf(long volatile *_Target, long _Value);
+long _InterlockedExchange_rel(long volatile *_Target, long _Value);
+__int64 _InterlockedExchange64_acq(__int64 volatile *_Target, __int64 _Value);
+__int64 _InterlockedExchange64_nf(__int64 volatile *_Target, __int64 _Value);
+__int64 _InterlockedExchange64_rel(__int64 volatile *_Target, __int64 _Value);
 #endif
 /*----------------------------------------------------------------------------*\
 |* Interlocked Compare Exchange
 \*----------------------------------------------------------------------------*/
 #if defined(__arm__) || defined(__aarch64__)
 static __inline__ char __DEFAULT_FN_ATTRS
 _InterlockedCompareExchange8_acq(char volatile *_Destination,
                              char _Exchange, char _Comparand) {

test/CodeGen/ms-intrinsics.c

 __int64 test_InterlockedExchangeAdd64_nf(__int64 volatile *value, __int64 mask) {
   return _InterlockedExchangeAdd64_nf(value, mask);
 }
 // CHECK-ARM-ARM64: define{{.*}}i64 @test_InterlockedExchangeAdd64_nf(i64*{{[a-z_ ]*}}%value, i64{{[a-z_ ]*}}%mask){{.*}}{
 // CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw add i64* %value, i64 %mask monotonic
 // CHECK-ARM-ARM64:   ret i64 [[RESULT:%[0-9]+]]
 // CHECK-ARM-ARM64: }
 
+char test_InterlockedExchange8_acq(char volatile *value, char mask) {
+  return _InterlockedExchange8_acq(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i8 @test_InterlockedExchange8_acq(i8*{{[a-z_ ]*}}%value, i8{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i8* %value, i8 %mask acquire
+// CHECK-ARM-ARM64:   ret i8 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+char test_InterlockedExchange8_rel(char volatile *value, char mask) {
+  return _InterlockedExchange8_rel(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i8 @test_InterlockedExchange8_rel(i8*{{[a-z_ ]*}}%value, i8{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i8* %value, i8 %mask release
+// CHECK-ARM-ARM64:   ret i8 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+char test_InterlockedExchange8_nf(char volatile *value, char mask) {
+  return _InterlockedExchange8_nf(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i8 @test_InterlockedExchange8_nf(i8*{{[a-z_ ]*}}%value, i8{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i8* %value, i8 %mask monotonic
+// CHECK-ARM-ARM64:   ret i8 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+short test_InterlockedExchange16_acq(short volatile *value, short mask) {
+  return _InterlockedExchange16_acq(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i16 @test_InterlockedExchange16_acq(i16*{{[a-z_ ]*}}%value, i16{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i16* %value, i16 %mask acquire
+// CHECK-ARM-ARM64:   ret i16 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+short test_InterlockedExchange16_rel(short volatile *value, short mask) {
+  return _InterlockedExchange16_rel(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i16 @test_InterlockedExchange16_rel(i16*{{[a-z_ ]*}}%value, i16{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i16* %value, i16 %mask release
+// CHECK-ARM-ARM64:   ret i16 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+short test_InterlockedExchange16_nf(short volatile *value, short mask) {
+  return _InterlockedExchange16_nf(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i16 @test_InterlockedExchange16_nf(i16*{{[a-z_ ]*}}%value, i16{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i16* %value, i16 %mask monotonic
+// CHECK-ARM-ARM64:   ret i16 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+long test_InterlockedExchange_acq(long volatile *value, long mask) {
+  return _InterlockedExchange_acq(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i32 @test_InterlockedExchange_acq(i32*{{[a-z_ ]*}}%value, i32{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i32* %value, i32 %mask acquire
+// CHECK-ARM-ARM64:   ret i32 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+long test_InterlockedExchange_rel(long volatile *value, long mask) {
+  return _InterlockedExchange_rel(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i32 @test_InterlockedExchange_rel(i32*{{[a-z_ ]*}}%value, i32{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i32* %value, i32 %mask release
+// CHECK-ARM-ARM64:   ret i32 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+long test_InterlockedExchange_nf(long volatile *value, long mask) {
+  return _InterlockedExchange_nf(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i32 @test_InterlockedExchange_nf(i32*{{[a-z_ ]*}}%value, i32{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i32* %value, i32 %mask monotonic
+// CHECK-ARM-ARM64:   ret i32 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+__int64 test_InterlockedExchange64_acq(__int64 volatile *value, __int64 mask) {
+  return _InterlockedExchange64_acq(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i64 @test_InterlockedExchange64_acq(i64*{{[a-z_ ]*}}%value, i64{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i64* %value, i64 %mask acquire
+// CHECK-ARM-ARM64:   ret i64 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+__int64 test_InterlockedExchange64_rel(__int64 volatile *value, __int64 mask) {
+  return _InterlockedExchange64_rel(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i64 @test_InterlockedExchange64_rel(i64*{{[a-z_ ]*}}%value, i64{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i64* %value, i64 %mask release
+// CHECK-ARM-ARM64:   ret i64 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
+__int64 test_InterlockedExchange64_nf(__int64 volatile *value, __int64 mask) {
+  return _InterlockedExchange64_nf(value, mask);
+}
+// CHECK-ARM-ARM64: define{{.*}}i64 @test_InterlockedExchange64_nf(i64*{{[a-z_ ]*}}%value, i64{{[a-z_ ]*}}%mask){{.*}}{
+// CHECK-ARM-ARM64:   [[RESULT:%[0-9]+]] = atomicrmw xchg i64* %value, i64 %mask monotonic
+// CHECK-ARM-ARM64:   ret i64 [[RESULT:%[0-9]+]]
+// CHECK-ARM-ARM64: }
 #endif
 
 #if !defined(__aarch64__)
 void test__fastfail() {
   __fastfail(42);
 }
 // CHECK-LABEL: define{{.*}} void @test__fastfail()
 // CHECK-ARM: call void asm sideeffect "udf #251", "{r0}"(i32 42) #[[NORETURN:[0-9]+]]
 // CHECK-INTEL: call void asm sideeffect "int $$0x29", "{cx}"(i32 42) #[[NORETURN]]
 
 // Attributes come last.
 
 // CHECK: attributes #[[NORETURN]] = { noreturn{{.*}} }
 #endif