Index: include/clang/Basic/Builtins.def
===================================================================
--- include/clang/Basic/Builtins.def
+++ include/clang/Basic/Builtins.def
@@ -756,6 +756,7 @@
 LANGBUILTIN(_InterlockedXor8,  "ccD*c",       "n", ALL_MS_LANGUAGES)
 LANGBUILTIN(_InterlockedXor16, "ssD*s",       "n", ALL_MS_LANGUAGES)
 LANGBUILTIN(_InterlockedXor,   "LiLiD*Li",    "n", ALL_MS_LANGUAGES)
+LANGBUILTIN(_interlockedbittestandset, "UcLiD*Li", "n", ALL_MS_LANGUAGES)
 LANGBUILTIN(__noop,           "i.",  "n", ALL_MS_LANGUAGES)
 LANGBUILTIN(__popcnt16, "UsUs",     "nc", ALL_MS_LANGUAGES)
 LANGBUILTIN(__popcnt,   "UiUi",     "nc", ALL_MS_LANGUAGES)
Index: lib/CodeGen/CGBuiltin.cpp
===================================================================
--- lib/CodeGen/CGBuiltin.cpp
+++ lib/CodeGen/CGBuiltin.cpp
@@ -492,6 +492,7 @@
   _InterlockedIncrement,
   _InterlockedOr,
   _InterlockedXor,
+  _interlockedbittestandset,
   __fastfail,
 };
 
@@ -559,6 +560,20 @@
   case MSVCIntrin::_InterlockedXor:
     return MakeBinaryAtomicValue(*this, AtomicRMWInst::Xor, E);
 
+  case MSVCIntrin::_interlockedbittestandset: {
+    llvm::Value *Addr = EmitScalarExpr(E->getArg(0));
+    llvm::Value *Bit = EmitScalarExpr(E->getArg(1));
+    AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
+        AtomicRMWInst::Or, Addr,
+        Builder.CreateShl(ConstantInt::get(Bit->getType(), 1), Bit),
+        llvm::AtomicOrdering::SequentiallyConsistent);
+    llvm::Value *Shifted = Builder.CreateLShr(RMWI, Bit);
+    llvm::Value *Truncated =
+        Builder.CreateTrunc(Shifted, ConvertType(E->getType()));
+    return Builder.CreateAnd(Truncated,
+                             ConstantInt::get(Truncated->getType(), 1));
+  }
+
   case MSVCIntrin::_InterlockedDecrement: {
     llvm::Type *IntTy = ConvertType(E->getType());
     AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
@@ -2238,6 +2253,9 @@
   case Builtin::BI_InterlockedXor16:
   case Builtin::BI_InterlockedXor:
     return RValue::get(EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor, E));
+  case Builtin::BI_interlockedbittestandset:
+    return RValue::get(
+        EmitMSVCBuiltinExpr(MSVCIntrin::_interlockedbittestandset, E));
 
   case Builtin::BI__exception_code:
   case Builtin::BI_exception_code:
@@ -2309,10 +2327,8 @@
     break;
   }
 
-  case Builtin::BI__fastfail: {
+  case Builtin::BI__fastfail:
     return RValue::get(EmitMSVCBuiltinExpr(MSVCIntrin::__fastfail, E));
-    break;
-  }
 
   case Builtin::BI__builtin_coro_size: {
     auto & Context = getContext();
Index: lib/Headers/intrin.h
===================================================================
--- lib/Headers/intrin.h
+++ lib/Headers/intrin.h
@@ -173,7 +173,6 @@
 void __cdecl _enable(void);
 long _InterlockedAddLargeStatistic(__int64 volatile *_Addend, long _Value);
 unsigned char _interlockedbittestandreset(long volatile *, long);
-static __inline__
 unsigned char _interlockedbittestandset(long volatile *, long);
 long _InterlockedCompareExchange_HLEAcquire(long volatile *, long, long);
 long _InterlockedCompareExchange_HLERelease(long volatile *, long, long);
@@ -369,11 +368,6 @@
   *_BitBase = *_BitBase | (1 << _BitPos);
   return _Res;
 }
-static __inline__ unsigned char __DEFAULT_FN_ATTRS
-_interlockedbittestandset(long volatile *_BitBase, long _BitPos) {
-  long _PrevVal = __atomic_fetch_or(_BitBase, 1l << _BitPos, __ATOMIC_SEQ_CST);
-  return (_PrevVal >> _BitPos) & 1;
-}
 #if defined(__arm__) || defined(__aarch64__)
 static __inline__ unsigned char __DEFAULT_FN_ATTRS
 _interlockedbittestandset_acq(long volatile *_BitBase, long _BitPos) {
Index: test/CodeGen/ms-intrinsics.c
===================================================================
--- test/CodeGen/ms-intrinsics.c
+++ test/CodeGen/ms-intrinsics.c
@@ -434,6 +434,17 @@
 
 #endif
 
+unsigned char test_interlockedbittestandset(volatile long *ptr, long bit) {
+  return _interlockedbittestandset(ptr, bit);
+}
+// CHECK-LABEL: define{{.*}} i8 @test_interlockedbittestandset
+// CHECK: %0 = shl i32 1, %bit
+// CHECK: %1 = atomicrmw or i32* %ptr, i32 %0 seq_cst
+// CHECK: %2 = lshr i32 %1, %bit
+// CHECK: %3 = trunc i32 %2 to i8
+// CHECK: %4 = and i8 %3, 1
+// CHECK: ret i8 %4
+
 void test__fastfail() {
   __fastfail(42);
 }