[COFF, ARM64] Implement InterlockedDecrement*_* builtins

[mgrang]

Summary

This is eight in a series of patches to move intrinsic definitions out of intrin.h.

[efriedma]

LGTM

CodeGen/CGBuiltin.cpp

static Value *EmitAtomicIncrementValue(CodeGenFunction &CGF, const CallExpr *E,
   auto *Result = CGF.Builder.CreateAtomicRMW(
                    AtomicRMWInst::Add,
                    CGF.EmitScalarExpr(E->getArg(0)),
                    ConstantInt::get(IntTy, 1),
                    Ordering);
   return CGF.Builder.CreateAdd(Result, ConstantInt::get(IntTy, 1));
 }
 
+static Value *EmitAtomicDecrementValue(CodeGenFunction &CGF, const CallExpr *E,
+    AtomicOrdering Ordering = AtomicOrdering::SequentiallyConsistent) {
+  assert(E->getArg(0)->getType()->isPointerType());
+
+  auto *IntTy = CGF.ConvertType(E->getType());
+  auto *Result = CGF.Builder.CreateAtomicRMW(
+                   AtomicRMWInst::Sub,
+                   CGF.EmitScalarExpr(E->getArg(0)),
+                   ConstantInt::get(IntTy, 1),
+                   Ordering);
+  return CGF.Builder.CreateSub(Result, ConstantInt::get(IntTy, 1));
+}
+
 // Emit a simple mangled intrinsic that has 1 argument and a return type
 // matching the argument type.
 static Value *emitUnaryBuiltin(CodeGenFunction &CGF,
                                const CallExpr *E,
                                unsigned IntrinsicID) {
   llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
 
   Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
enum class CodeGenFunction::MSVCIntrin {
   _InterlockedXor_rel,
   _InterlockedXor_nf,
   _InterlockedAnd_acq,
   _InterlockedAnd_rel,
   _InterlockedAnd_nf,
   _InterlockedIncrement_acq,
   _InterlockedIncrement_rel,
   _InterlockedIncrement_nf,
+  _InterlockedDecrement_acq,
+  _InterlockedDecrement_rel,
+  _InterlockedDecrement_nf,
   __fastfail,
 };
 
 Value *CodeGenFunction::EmitMSVCBuiltinExpr(MSVCIntrin BuiltinID,
                                             const CallExpr *E) {
   switch (BuiltinID) {
   case MSVCIntrin::_BitScanForward:
   case MSVCIntrin::_BitScanReverse: {
case MSVCIntrin::_InterlockedAnd_nf:
     return MakeBinaryAtomicValue(*this, AtomicRMWInst::And, E,
                                  AtomicOrdering::Monotonic);
   case MSVCIntrin::_InterlockedIncrement_acq:
     return EmitAtomicIncrementValue(*this, E, AtomicOrdering::Acquire);
   case MSVCIntrin::_InterlockedIncrement_rel:
     return EmitAtomicIncrementValue(*this, E, AtomicOrdering::Release);
   case MSVCIntrin::_InterlockedIncrement_nf:
     return EmitAtomicIncrementValue(*this, E, AtomicOrdering::Monotonic);
+  case MSVCIntrin::_InterlockedDecrement_acq:
+    return EmitAtomicDecrementValue(*this, E, AtomicOrdering::Acquire);
+  case MSVCIntrin::_InterlockedDecrement_rel:
+    return EmitAtomicDecrementValue(*this, E, AtomicOrdering::Release);
+  case MSVCIntrin::_InterlockedDecrement_nf:
+    return EmitAtomicDecrementValue(*this, 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);
-    return Builder.CreateSub(RMWI, ConstantInt::get(IntTy, 1));
-  }
+  case MSVCIntrin::_InterlockedDecrement:
+    return EmitAtomicDecrementValue(*this, E);
   case MSVCIntrin::_InterlockedIncrement:
     return EmitAtomicIncrementValue(*this, E);
 
   case MSVCIntrin::__fastfail: {
     // Request immediate process termination from the kernel. The instruction
     // sequences to do this are documented on MSDN:
     // https://msdn.microsoft.com/en-us/library/dn774154.aspx
     llvm::Triple::ArchType ISA = getTarget().getTriple().getArch();
Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
   case ARM::BI_InterlockedIncrement16_rel:
   case ARM::BI_InterlockedIncrement_rel:
   case ARM::BI_InterlockedIncrement64_rel:
     return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement_rel, E);
   case ARM::BI_InterlockedIncrement16_nf:
   case ARM::BI_InterlockedIncrement_nf:
   case ARM::BI_InterlockedIncrement64_nf:
     return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement_nf, E);
+  case ARM::BI_InterlockedDecrement16_acq:
+  case ARM::BI_InterlockedDecrement_acq:
+  case ARM::BI_InterlockedDecrement64_acq:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement_acq, E);
+  case ARM::BI_InterlockedDecrement16_rel:
+  case ARM::BI_InterlockedDecrement_rel:
+  case ARM::BI_InterlockedDecrement64_rel:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement_rel, E);
+  case ARM::BI_InterlockedDecrement16_nf:
+  case ARM::BI_InterlockedDecrement_nf:
+  case ARM::BI_InterlockedDecrement64_nf:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement_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_InterlockedIncrement16_rel:
   case AArch64::BI_InterlockedIncrement_rel:
   case AArch64::BI_InterlockedIncrement64_rel:
     return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement_rel, E);
   case AArch64::BI_InterlockedIncrement16_nf:
   case AArch64::BI_InterlockedIncrement_nf:
   case AArch64::BI_InterlockedIncrement64_nf:
     return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement_nf, E);
+  case AArch64::BI_InterlockedDecrement16_acq:
+  case AArch64::BI_InterlockedDecrement_acq:
+  case AArch64::BI_InterlockedDecrement64_acq:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement_acq, E);
+  case AArch64::BI_InterlockedDecrement16_rel:
+  case AArch64::BI_InterlockedDecrement_rel:
+  case AArch64::BI_InterlockedDecrement64_rel:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement_rel, E);
+  case AArch64::BI_InterlockedDecrement16_nf:
+  case AArch64::BI_InterlockedDecrement_nf:
+  case AArch64::BI_InterlockedDecrement64_nf:
+    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement_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);

CodeGen/ms-intrinsics.c

 __int64 test_InterlockedIncrement64_nf(__int64 volatile *Addend) {
   return _InterlockedIncrement64_nf(Addend);
 }
 // CHECK-ARM-ARM64: define{{.*}}i64 @test_InterlockedIncrement64_nf(i64*{{[a-z_ ]*}}%Addend){{.*}}{
 // CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw add i64* %Addend, i64 1 monotonic
 // CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i64 [[TMP]], 1
 // CHECK-ARM-ARM64: ret i64 [[RESULT]]
 // CHECK-ARM-ARM64: }
+
+short test_InterlockedDecrement16_acq(short volatile *Addend) {
+  return _InterlockedDecrement16_acq(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i16 @test_InterlockedDecrement16_acq(i16*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i16* %Addend, i16 1 acquire
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i16 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i16 [[RESULT]]
+// CHECK-ARM-ARM64: }
+
+short test_InterlockedDecrement16_rel(short volatile *Addend) {
+  return _InterlockedDecrement16_rel(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i16 @test_InterlockedDecrement16_rel(i16*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i16* %Addend, i16 1 release
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i16 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i16 [[RESULT]]
+// CHECK-ARM-ARM64: }
+
+short test_InterlockedDecrement16_nf(short volatile *Addend) {
+  return _InterlockedDecrement16_nf(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i16 @test_InterlockedDecrement16_nf(i16*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i16* %Addend, i16 1 monotonic
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i16 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i16 [[RESULT]]
+// CHECK-ARM-ARM64: }
+
+long test_InterlockedDecrement_acq(long volatile *Addend) {
+  return _InterlockedDecrement_acq(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i32 @test_InterlockedDecrement_acq(i32*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i32* %Addend, i32 1 acquire
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i32 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i32 [[RESULT]]
+// CHECK-ARM-ARM64: }
+
+long test_InterlockedDecrement_rel(long volatile *Addend) {
+  return _InterlockedDecrement_rel(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i32 @test_InterlockedDecrement_rel(i32*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i32* %Addend, i32 1 release
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i32 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i32 [[RESULT]]
+// CHECK-ARM-ARM64: }
+
+long test_InterlockedDecrement_nf(long volatile *Addend) {
+  return _InterlockedDecrement_nf(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i32 @test_InterlockedDecrement_nf(i32*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i32* %Addend, i32 1 monotonic
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i32 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i32 [[RESULT]]
+// CHECK-ARM-ARM64: }
+
+__int64 test_InterlockedDecrement64_acq(__int64 volatile *Addend) {
+  return _InterlockedDecrement64_acq(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i64 @test_InterlockedDecrement64_acq(i64*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i64* %Addend, i64 1 acquire
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i64 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i64 [[RESULT]]
+// CHECK-ARM-ARM64: }
+
+__int64 test_InterlockedDecrement64_rel(__int64 volatile *Addend) {
+  return _InterlockedDecrement64_rel(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i64 @test_InterlockedDecrement64_rel(i64*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i64* %Addend, i64 1 release
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i64 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i64 [[RESULT]]
+// CHECK-ARM-ARM64: }
+
+__int64 test_InterlockedDecrement64_nf(__int64 volatile *Addend) {
+  return _InterlockedDecrement64_nf(Addend);
+}
+// CHECK-ARM-ARM64: define{{.*}}i64 @test_InterlockedDecrement64_nf(i64*{{[a-z_ ]*}}%Addend){{.*}}{
+// CHECK-ARM-ARM64: [[TMP:%[0-9]+]] = atomicrmw sub i64* %Addend, i64 1 monotonic
+// CHECK-ARM-ARM64: [[RESULT:%[0-9]+]] = add i64 [[TMP]], -1
+// CHECK-ARM-ARM64: ret i64 [[RESULT]]
+// 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

Headers/intrin.h

 __int64 _InterlockedIncrement64_acq(__int64 volatile *_Value);
 __int64 _InterlockedIncrement64_nf(__int64 volatile *_Value);
 __int64 _InterlockedIncrement64_rel(__int64 volatile *_Value);
 #endif
 /*----------------------------------------------------------------------------*\
 |* Interlocked Decrement
 \*----------------------------------------------------------------------------*/
 #if defined(__arm__) || defined(__aarch64__)
-static __inline__ short __DEFAULT_FN_ATTRS
-_InterlockedDecrement16_acq(short volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_ACQUIRE);
-}
-static __inline__ short __DEFAULT_FN_ATTRS
-_InterlockedDecrement16_nf(short volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_RELAXED);
-}
-static __inline__ short __DEFAULT_FN_ATTRS
-_InterlockedDecrement16_rel(short volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_RELEASE);
-}
-static __inline__ long __DEFAULT_FN_ATTRS
-_InterlockedDecrement_acq(long volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_ACQUIRE);
-}
-static __inline__ long __DEFAULT_FN_ATTRS
-_InterlockedDecrement_nf(long volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_RELAXED);
-}
-static __inline__ long __DEFAULT_FN_ATTRS
-_InterlockedDecrement_rel(long volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_RELEASE);
-}
-static __inline__ __int64 __DEFAULT_FN_ATTRS
-_InterlockedDecrement64_acq(__int64 volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_ACQUIRE);
-}
-static __inline__ __int64 __DEFAULT_FN_ATTRS
-_InterlockedDecrement64_nf(__int64 volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_RELAXED);
-}
-static __inline__ __int64 __DEFAULT_FN_ATTRS
-_InterlockedDecrement64_rel(__int64 volatile *_Value) {
-  return __atomic_sub_fetch(_Value, 1, __ATOMIC_RELEASE);
-}
+short _InterlockedDecrement16_acq(short volatile *_Value);
+short _InterlockedDecrement16_nf(short volatile *_Value);
+short _InterlockedDecrement16_rel(short volatile *_Value);
+long _InterlockedDecrement_acq(long volatile *_Value);
+long _InterlockedDecrement_nf(long volatile *_Value);
+long _InterlockedDecrement_rel(long volatile *_Value);
+__int64 _InterlockedDecrement64_acq(__int64 volatile *_Value);
+__int64 _InterlockedDecrement64_nf(__int64 volatile *_Value);
+__int64 _InterlockedDecrement64_rel(__int64 volatile *_Value);
 #endif
 /*----------------------------------------------------------------------------*\
 |* Interlocked And
 \*----------------------------------------------------------------------------*/
 #if defined(__arm__) || defined(__aarch64__)
 char _InterlockedAnd8_acq(char volatile *_Value, char _Mask);
 char _InterlockedAnd8_nf(char volatile *_Value, char _Mask);
 char _InterlockedAnd8_rel(char volatile *_Value, char _Mask);

clang/Basic/BuiltinsAArch64.def

 TARGET_HEADER_BUILTIN(_InterlockedIncrement16_rel, "ssD*",     "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement_acq,   "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement_nf,    "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement_rel,   "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement64_acq, "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement64_nf,  "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement64_rel, "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 
+TARGET_HEADER_BUILTIN(_InterlockedDecrement16_acq, "ssD*",     "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement16_nf,  "ssD*",     "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement16_rel, "ssD*",     "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement_acq,   "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement_nf,    "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement_rel,   "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement64_acq, "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement64_nf,  "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement64_rel, "LLiLLiD*", "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

clang/Basic/BuiltinsARM.def

 TARGET_HEADER_BUILTIN(_InterlockedIncrement16_rel, "ssD*",     "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement_acq,   "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement_nf,    "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement_rel,   "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement64_acq, "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement64_nf,  "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 TARGET_HEADER_BUILTIN(_InterlockedIncrement64_rel, "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
 
+TARGET_HEADER_BUILTIN(_InterlockedDecrement16_acq, "ssD*",     "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement16_nf,  "ssD*",     "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement16_rel, "ssD*",     "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement_acq,   "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement_nf,    "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement_rel,   "LiLiD*",   "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement64_acq, "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement64_nf,  "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+TARGET_HEADER_BUILTIN(_InterlockedDecrement64_rel, "LLiLLiD*", "nh", "intrin.h", ALL_MS_LANGUAGES, "")
+
 #undef BUILTIN
 #undef LANGBUILTIN
 #undef TARGET_HEADER_BUILTIN