[MS] Fix double evaluation of MSVC builtin arguments

This code got quite twisted because we consider some MSVC builtins to be
target agnostic, and some to be target specific. Target specific
intrinsics have a pattern of doing up-front argument evaluation, while
general intrinsics do not evaluate their arguments up front. As we tried
to share codepaths between the target-specific and target-agnostic
handling, we ended up doing double evaluation.

Instead, have each target handle MSVC intrinsics consistently before up
front argument evaluation. This requires passing less data around and is
more consistent with target independent intrinsic handling.

See D50979 for past examples of this bug. I noticed this while looking
into adding some more intrinsics.

Differential Revision: https://reviews.llvm.org/D92061

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index e98ec3e..b2bfc3c 100644 (file)
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -1011,16 +1011,342 @@ enum class CodeGenFunction::MSVCIntrin {
   __fastfail,
 };
 
   __fastfail,
 };
 

+static Optional<CodeGenFunction::MSVCIntrin>
+translateArmToMsvcIntrin(unsigned BuiltinID) {
+  using MSVCIntrin = CodeGenFunction::MSVCIntrin;
+  switch (BuiltinID) {
+  default:
+    return None;
+  case ARM::BI_BitScanForward:
+  case ARM::BI_BitScanForward64:
+    return MSVCIntrin::_BitScanForward;
+  case ARM::BI_BitScanReverse:
+  case ARM::BI_BitScanReverse64:
+    return MSVCIntrin::_BitScanReverse;
+  case ARM::BI_InterlockedAnd64:
+    return MSVCIntrin::_InterlockedAnd;
+  case ARM::BI_InterlockedExchange64:
+    return MSVCIntrin::_InterlockedExchange;
+  case ARM::BI_InterlockedExchangeAdd64:
+    return MSVCIntrin::_InterlockedExchangeAdd;
+  case ARM::BI_InterlockedExchangeSub64:
+    return MSVCIntrin::_InterlockedExchangeSub;
+  case ARM::BI_InterlockedOr64:
+    return MSVCIntrin::_InterlockedOr;
+  case ARM::BI_InterlockedXor64:
+    return MSVCIntrin::_InterlockedXor;
+  case ARM::BI_InterlockedDecrement64:
+    return MSVCIntrin::_InterlockedDecrement;
+  case ARM::BI_InterlockedIncrement64:
+    return MSVCIntrin::_InterlockedIncrement;
+  case ARM::BI_InterlockedExchangeAdd8_acq:
+  case ARM::BI_InterlockedExchangeAdd16_acq:
+  case ARM::BI_InterlockedExchangeAdd_acq:
+  case ARM::BI_InterlockedExchangeAdd64_acq:
+    return MSVCIntrin::_InterlockedExchangeAdd_acq;
+  case ARM::BI_InterlockedExchangeAdd8_rel:
+  case ARM::BI_InterlockedExchangeAdd16_rel:
+  case ARM::BI_InterlockedExchangeAdd_rel:
+  case ARM::BI_InterlockedExchangeAdd64_rel:
+    return MSVCIntrin::_InterlockedExchangeAdd_rel;
+  case ARM::BI_InterlockedExchangeAdd8_nf:
+  case ARM::BI_InterlockedExchangeAdd16_nf:
+  case ARM::BI_InterlockedExchangeAdd_nf:
+  case ARM::BI_InterlockedExchangeAdd64_nf:
+    return MSVCIntrin::_InterlockedExchangeAdd_nf;
+  case ARM::BI_InterlockedExchange8_acq:
+  case ARM::BI_InterlockedExchange16_acq:
+  case ARM::BI_InterlockedExchange_acq:
+  case ARM::BI_InterlockedExchange64_acq:
+    return MSVCIntrin::_InterlockedExchange_acq;
+  case ARM::BI_InterlockedExchange8_rel:
+  case ARM::BI_InterlockedExchange16_rel:
+  case ARM::BI_InterlockedExchange_rel:
+  case ARM::BI_InterlockedExchange64_rel:
+    return MSVCIntrin::_InterlockedExchange_rel;
+  case ARM::BI_InterlockedExchange8_nf:
+  case ARM::BI_InterlockedExchange16_nf:
+  case ARM::BI_InterlockedExchange_nf:
+  case ARM::BI_InterlockedExchange64_nf:
+    return MSVCIntrin::_InterlockedExchange_nf;
+  case ARM::BI_InterlockedCompareExchange8_acq:
+  case ARM::BI_InterlockedCompareExchange16_acq:
+  case ARM::BI_InterlockedCompareExchange_acq:
+  case ARM::BI_InterlockedCompareExchange64_acq:
+    return MSVCIntrin::_InterlockedCompareExchange_acq;
+  case ARM::BI_InterlockedCompareExchange8_rel:
+  case ARM::BI_InterlockedCompareExchange16_rel:
+  case ARM::BI_InterlockedCompareExchange_rel:
+  case ARM::BI_InterlockedCompareExchange64_rel:
+    return MSVCIntrin::_InterlockedCompareExchange_rel;
+  case ARM::BI_InterlockedCompareExchange8_nf:
+  case ARM::BI_InterlockedCompareExchange16_nf:
+  case ARM::BI_InterlockedCompareExchange_nf:
+  case ARM::BI_InterlockedCompareExchange64_nf:
+    return MSVCIntrin::_InterlockedCompareExchange_nf;
+  case ARM::BI_InterlockedOr8_acq:
+  case ARM::BI_InterlockedOr16_acq:
+  case ARM::BI_InterlockedOr_acq:
+  case ARM::BI_InterlockedOr64_acq:
+    return MSVCIntrin::_InterlockedOr_acq;
+  case ARM::BI_InterlockedOr8_rel:
+  case ARM::BI_InterlockedOr16_rel:
+  case ARM::BI_InterlockedOr_rel:
+  case ARM::BI_InterlockedOr64_rel:
+    return MSVCIntrin::_InterlockedOr_rel;
+  case ARM::BI_InterlockedOr8_nf:
+  case ARM::BI_InterlockedOr16_nf:
+  case ARM::BI_InterlockedOr_nf:
+  case ARM::BI_InterlockedOr64_nf:
+    return MSVCIntrin::_InterlockedOr_nf;
+  case ARM::BI_InterlockedXor8_acq:
+  case ARM::BI_InterlockedXor16_acq:
+  case ARM::BI_InterlockedXor_acq:
+  case ARM::BI_InterlockedXor64_acq:
+    return MSVCIntrin::_InterlockedXor_acq;
+  case ARM::BI_InterlockedXor8_rel:
+  case ARM::BI_InterlockedXor16_rel:
+  case ARM::BI_InterlockedXor_rel:
+  case ARM::BI_InterlockedXor64_rel:
+    return MSVCIntrin::_InterlockedXor_rel;
+  case ARM::BI_InterlockedXor8_nf:
+  case ARM::BI_InterlockedXor16_nf:
+  case ARM::BI_InterlockedXor_nf:
+  case ARM::BI_InterlockedXor64_nf:
+    return MSVCIntrin::_InterlockedXor_nf;
+  case ARM::BI_InterlockedAnd8_acq:
+  case ARM::BI_InterlockedAnd16_acq:
+  case ARM::BI_InterlockedAnd_acq:
+  case ARM::BI_InterlockedAnd64_acq:
+    return MSVCIntrin::_InterlockedAnd_acq;
+  case ARM::BI_InterlockedAnd8_rel:
+  case ARM::BI_InterlockedAnd16_rel:
+  case ARM::BI_InterlockedAnd_rel:
+  case ARM::BI_InterlockedAnd64_rel:
+    return MSVCIntrin::_InterlockedAnd_rel;
+  case ARM::BI_InterlockedAnd8_nf:
+  case ARM::BI_InterlockedAnd16_nf:
+  case ARM::BI_InterlockedAnd_nf:
+  case ARM::BI_InterlockedAnd64_nf:
+    return MSVCIntrin::_InterlockedAnd_nf;
+  case ARM::BI_InterlockedIncrement16_acq:
+  case ARM::BI_InterlockedIncrement_acq:
+  case ARM::BI_InterlockedIncrement64_acq:
+    return MSVCIntrin::_InterlockedIncrement_acq;
+  case ARM::BI_InterlockedIncrement16_rel:
+  case ARM::BI_InterlockedIncrement_rel:
+  case ARM::BI_InterlockedIncrement64_rel:
+    return MSVCIntrin::_InterlockedIncrement_rel;
+  case ARM::BI_InterlockedIncrement16_nf:
+  case ARM::BI_InterlockedIncrement_nf:
+  case ARM::BI_InterlockedIncrement64_nf:
+    return MSVCIntrin::_InterlockedIncrement_nf;
+  case ARM::BI_InterlockedDecrement16_acq:
+  case ARM::BI_InterlockedDecrement_acq:
+  case ARM::BI_InterlockedDecrement64_acq:
+    return MSVCIntrin::_InterlockedDecrement_acq;
+  case ARM::BI_InterlockedDecrement16_rel:
+  case ARM::BI_InterlockedDecrement_rel:
+  case ARM::BI_InterlockedDecrement64_rel:
+    return MSVCIntrin::_InterlockedDecrement_rel;
+  case ARM::BI_InterlockedDecrement16_nf:
+  case ARM::BI_InterlockedDecrement_nf:
+  case ARM::BI_InterlockedDecrement64_nf:
+    return MSVCIntrin::_InterlockedDecrement_nf;
+  }
+  llvm_unreachable("must return from switch");
+}
+
+static Optional<CodeGenFunction::MSVCIntrin>
+translateAarch64ToMsvcIntrin(unsigned BuiltinID) {
+  using MSVCIntrin = CodeGenFunction::MSVCIntrin;
+  switch (BuiltinID) {
+  default:
+    return None;
+  case AArch64::BI_BitScanForward:
+  case AArch64::BI_BitScanForward64:
+    return MSVCIntrin::_BitScanForward;
+  case AArch64::BI_BitScanReverse:
+  case AArch64::BI_BitScanReverse64:
+    return MSVCIntrin::_BitScanReverse;
+  case AArch64::BI_InterlockedAnd64:
+    return MSVCIntrin::_InterlockedAnd;
+  case AArch64::BI_InterlockedExchange64:
+    return MSVCIntrin::_InterlockedExchange;
+  case AArch64::BI_InterlockedExchangeAdd64:
+    return MSVCIntrin::_InterlockedExchangeAdd;
+  case AArch64::BI_InterlockedExchangeSub64:
+    return MSVCIntrin::_InterlockedExchangeSub;
+  case AArch64::BI_InterlockedOr64:
+    return MSVCIntrin::_InterlockedOr;
+  case AArch64::BI_InterlockedXor64:
+    return MSVCIntrin::_InterlockedXor;
+  case AArch64::BI_InterlockedDecrement64:
+    return MSVCIntrin::_InterlockedDecrement;
+  case AArch64::BI_InterlockedIncrement64:
+    return MSVCIntrin::_InterlockedIncrement;
+  case AArch64::BI_InterlockedExchangeAdd8_acq:
+  case AArch64::BI_InterlockedExchangeAdd16_acq:
+  case AArch64::BI_InterlockedExchangeAdd_acq:
+  case AArch64::BI_InterlockedExchangeAdd64_acq:
+    return MSVCIntrin::_InterlockedExchangeAdd_acq;
+  case AArch64::BI_InterlockedExchangeAdd8_rel:
+  case AArch64::BI_InterlockedExchangeAdd16_rel:
+  case AArch64::BI_InterlockedExchangeAdd_rel:
+  case AArch64::BI_InterlockedExchangeAdd64_rel:
+    return MSVCIntrin::_InterlockedExchangeAdd_rel;
+  case AArch64::BI_InterlockedExchangeAdd8_nf:
+  case AArch64::BI_InterlockedExchangeAdd16_nf:
+  case AArch64::BI_InterlockedExchangeAdd_nf:
+  case AArch64::BI_InterlockedExchangeAdd64_nf:
+    return MSVCIntrin::_InterlockedExchangeAdd_nf;
+  case AArch64::BI_InterlockedExchange8_acq:
+  case AArch64::BI_InterlockedExchange16_acq:
+  case AArch64::BI_InterlockedExchange_acq:
+  case AArch64::BI_InterlockedExchange64_acq:
+    return MSVCIntrin::_InterlockedExchange_acq;
+  case AArch64::BI_InterlockedExchange8_rel:
+  case AArch64::BI_InterlockedExchange16_rel:
+  case AArch64::BI_InterlockedExchange_rel:
+  case AArch64::BI_InterlockedExchange64_rel:
+    return MSVCIntrin::_InterlockedExchange_rel;
+  case AArch64::BI_InterlockedExchange8_nf:
+  case AArch64::BI_InterlockedExchange16_nf:
+  case AArch64::BI_InterlockedExchange_nf:
+  case AArch64::BI_InterlockedExchange64_nf:
+    return MSVCIntrin::_InterlockedExchange_nf;
+  case AArch64::BI_InterlockedCompareExchange8_acq:
+  case AArch64::BI_InterlockedCompareExchange16_acq:
+  case AArch64::BI_InterlockedCompareExchange_acq:
+  case AArch64::BI_InterlockedCompareExchange64_acq:
+    return MSVCIntrin::_InterlockedCompareExchange_acq;
+  case AArch64::BI_InterlockedCompareExchange8_rel:
+  case AArch64::BI_InterlockedCompareExchange16_rel:
+  case AArch64::BI_InterlockedCompareExchange_rel:
+  case AArch64::BI_InterlockedCompareExchange64_rel:
+    return MSVCIntrin::_InterlockedCompareExchange_rel;
+  case AArch64::BI_InterlockedCompareExchange8_nf:
+  case AArch64::BI_InterlockedCompareExchange16_nf:
+  case AArch64::BI_InterlockedCompareExchange_nf:
+  case AArch64::BI_InterlockedCompareExchange64_nf:
+    return MSVCIntrin::_InterlockedCompareExchange_nf;
+  case AArch64::BI_InterlockedOr8_acq:
+  case AArch64::BI_InterlockedOr16_acq:
+  case AArch64::BI_InterlockedOr_acq:
+  case AArch64::BI_InterlockedOr64_acq:
+    return MSVCIntrin::_InterlockedOr_acq;
+  case AArch64::BI_InterlockedOr8_rel:
+  case AArch64::BI_InterlockedOr16_rel:
+  case AArch64::BI_InterlockedOr_rel:
+  case AArch64::BI_InterlockedOr64_rel:
+    return MSVCIntrin::_InterlockedOr_rel;
+  case AArch64::BI_InterlockedOr8_nf:
+  case AArch64::BI_InterlockedOr16_nf:
+  case AArch64::BI_InterlockedOr_nf:
+  case AArch64::BI_InterlockedOr64_nf:
+    return MSVCIntrin::_InterlockedOr_nf;
+  case AArch64::BI_InterlockedXor8_acq:
+  case AArch64::BI_InterlockedXor16_acq:
+  case AArch64::BI_InterlockedXor_acq:
+  case AArch64::BI_InterlockedXor64_acq:
+    return MSVCIntrin::_InterlockedXor_acq;
+  case AArch64::BI_InterlockedXor8_rel:
+  case AArch64::BI_InterlockedXor16_rel:
+  case AArch64::BI_InterlockedXor_rel:
+  case AArch64::BI_InterlockedXor64_rel:
+    return MSVCIntrin::_InterlockedXor_rel;
+  case AArch64::BI_InterlockedXor8_nf:
+  case AArch64::BI_InterlockedXor16_nf:
+  case AArch64::BI_InterlockedXor_nf:
+  case AArch64::BI_InterlockedXor64_nf:
+    return MSVCIntrin::_InterlockedXor_nf;
+  case AArch64::BI_InterlockedAnd8_acq:
+  case AArch64::BI_InterlockedAnd16_acq:
+  case AArch64::BI_InterlockedAnd_acq:
+  case AArch64::BI_InterlockedAnd64_acq:
+    return MSVCIntrin::_InterlockedAnd_acq;
+  case AArch64::BI_InterlockedAnd8_rel:
+  case AArch64::BI_InterlockedAnd16_rel:
+  case AArch64::BI_InterlockedAnd_rel:
+  case AArch64::BI_InterlockedAnd64_rel:
+    return MSVCIntrin::_InterlockedAnd_rel;
+  case AArch64::BI_InterlockedAnd8_nf:
+  case AArch64::BI_InterlockedAnd16_nf:
+  case AArch64::BI_InterlockedAnd_nf:
+  case AArch64::BI_InterlockedAnd64_nf:
+    return MSVCIntrin::_InterlockedAnd_nf;
+  case AArch64::BI_InterlockedIncrement16_acq:
+  case AArch64::BI_InterlockedIncrement_acq:
+  case AArch64::BI_InterlockedIncrement64_acq:
+    return MSVCIntrin::_InterlockedIncrement_acq;
+  case AArch64::BI_InterlockedIncrement16_rel:
+  case AArch64::BI_InterlockedIncrement_rel:
+  case AArch64::BI_InterlockedIncrement64_rel:
+    return MSVCIntrin::_InterlockedIncrement_rel;
+  case AArch64::BI_InterlockedIncrement16_nf:
+  case AArch64::BI_InterlockedIncrement_nf:
+  case AArch64::BI_InterlockedIncrement64_nf:
+    return MSVCIntrin::_InterlockedIncrement_nf;
+  case AArch64::BI_InterlockedDecrement16_acq:
+  case AArch64::BI_InterlockedDecrement_acq:
+  case AArch64::BI_InterlockedDecrement64_acq:
+    return MSVCIntrin::_InterlockedDecrement_acq;
+  case AArch64::BI_InterlockedDecrement16_rel:
+  case AArch64::BI_InterlockedDecrement_rel:
+  case AArch64::BI_InterlockedDecrement64_rel:
+    return MSVCIntrin::_InterlockedDecrement_rel;
+  case AArch64::BI_InterlockedDecrement16_nf:
+  case AArch64::BI_InterlockedDecrement_nf:
+  case AArch64::BI_InterlockedDecrement64_nf:
+    return MSVCIntrin::_InterlockedDecrement_nf;
+  }
+  llvm_unreachable("must return from switch");
+}
+
+static Optional<CodeGenFunction::MSVCIntrin>
+translateX86ToMsvcIntrin(unsigned BuiltinID) {
+  using MSVCIntrin = CodeGenFunction::MSVCIntrin;
+  switch (BuiltinID) {
+  default:
+    return None;
+  case clang::X86::BI_BitScanForward:
+  case clang::X86::BI_BitScanForward64:
+    return MSVCIntrin::_BitScanForward;
+  case clang::X86::BI_BitScanReverse:
+  case clang::X86::BI_BitScanReverse64:
+    return MSVCIntrin::_BitScanReverse;
+  case clang::X86::BI_InterlockedAnd64:
+    return MSVCIntrin::_InterlockedAnd;
+  case clang::X86::BI_InterlockedExchange64:
+    return MSVCIntrin::_InterlockedExchange;
+  case clang::X86::BI_InterlockedExchangeAdd64:
+    return MSVCIntrin::_InterlockedExchangeAdd;
+  case clang::X86::BI_InterlockedExchangeSub64:
+    return MSVCIntrin::_InterlockedExchangeSub;
+  case clang::X86::BI_InterlockedOr64:
+    return MSVCIntrin::_InterlockedOr;
+  case clang::X86::BI_InterlockedXor64:
+    return MSVCIntrin::_InterlockedXor;
+  case clang::X86::BI_InterlockedDecrement64:
+    return MSVCIntrin::_InterlockedDecrement;
+  case clang::X86::BI_InterlockedIncrement64:
+    return MSVCIntrin::_InterlockedIncrement;
+  }
+  llvm_unreachable("must return from switch");
+}
+
+// Emit an MSVC intrinsic. Assumes that arguments have *not* been evaluated.

 Value *CodeGenFunction::EmitMSVCBuiltinExpr(MSVCIntrin BuiltinID,
                                             const CallExpr *E) {
   switch (BuiltinID) {
   case MSVCIntrin::_BitScanForward:
   case MSVCIntrin::_BitScanReverse: {
 Value *CodeGenFunction::EmitMSVCBuiltinExpr(MSVCIntrin BuiltinID,
                                             const CallExpr *E) {
   switch (BuiltinID) {
   case MSVCIntrin::_BitScanForward:
   case MSVCIntrin::_BitScanReverse: {

+    Address IndexAddress(EmitPointerWithAlignment(E->getArg(0)));

     Value *ArgValue = EmitScalarExpr(E->getArg(1));
 
     llvm::Type *ArgType = ArgValue->getType();
     llvm::Type *IndexType =
     Value *ArgValue = EmitScalarExpr(E->getArg(1));
 
     llvm::Type *ArgType = ArgValue->getType();
     llvm::Type *IndexType =

-      EmitScalarExpr(E->getArg(0))->getType()->getPointerElementType();
+        IndexAddress.getPointer()->getType()->getPointerElementType();

     llvm::Type *ResultType = ConvertType(E->getType());
 
     Value *ArgZero = llvm::Constant::getNullValue(ArgType);
     llvm::Type *ResultType = ConvertType(E->getType());
 
     Value *ArgZero = llvm::Constant::getNullValue(ArgType);


@@ -1039,7 +1365,6 @@ Value *CodeGenFunction::EmitMSVCBuiltinExpr(MSVCIntrin BuiltinID,
     Result->addIncoming(ResZero, Begin);
 
     Builder.SetInsertPoint(NotZero);
     Result->addIncoming(ResZero, Begin);
 
     Builder.SetInsertPoint(NotZero);

-    Address IndexAddress = EmitPointerWithAlignment(E->getArg(0));

 
     if (BuiltinID == MSVCIntrin::_BitScanForward) {
       Function *F = CGM.getIntrinsic(Intrinsic::cttz, ArgType);
 
     if (BuiltinID == MSVCIntrin::_BitScanForward) {
       Function *F = CGM.getIntrinsic(Intrinsic::cttz, ArgType);


@@ -6847,6 +7172,11 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
                                       AccessKind);
   }
 
                                       AccessKind);
   }
 

+  // Handle MSVC intrinsics before argument evaluation to prevent double
+  // evaluation.
+  if (Optional<MSVCIntrin> MsvcIntId = translateArmToMsvcIntrin(BuiltinID))
+    return EmitMSVCBuiltinExpr(*MsvcIntId, E);
+

   // Deal with MVE builtins
   if (Value *Result = EmitARMMVEBuiltinExpr(BuiltinID, E, ReturnValue, Arch))
     return Result;
   // Deal with MVE builtins
   if (Value *Result = EmitARMMVEBuiltinExpr(BuiltinID, E, ReturnValue, Arch))
     return Result;


@@ -7007,143 +7337,6 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
     return Builder.CreateCall(F, {Ops[1], Ops[2], Ops[0],
                                   Ops[3], Ops[4], Ops[5]});
   }
     return Builder.CreateCall(F, {Ops[1], Ops[2], Ops[0],
                                   Ops[3], Ops[4], Ops[5]});
   }

-  case ARM::BI_BitScanForward:
-  case ARM::BI_BitScanForward64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_BitScanForward, E);
-  case ARM::BI_BitScanReverse:
-  case ARM::BI_BitScanReverse64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_BitScanReverse, E);
-
-  case ARM::BI_InterlockedAnd64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd, E);
-  case ARM::BI_InterlockedExchange64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange, E);
-  case ARM::BI_InterlockedExchangeAdd64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd, E);
-  case ARM::BI_InterlockedExchangeSub64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeSub, E);
-  case ARM::BI_InterlockedOr64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr, E);
-  case ARM::BI_InterlockedXor64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor, E);
-  case ARM::BI_InterlockedDecrement64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement, E);
-  case ARM::BI_InterlockedIncrement64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement, E);
-  case ARM::BI_InterlockedExchangeAdd8_acq:
-  case ARM::BI_InterlockedExchangeAdd16_acq:
-  case ARM::BI_InterlockedExchangeAdd_acq:
-  case ARM::BI_InterlockedExchangeAdd64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd_acq, E);
-  case ARM::BI_InterlockedExchangeAdd8_rel:
-  case ARM::BI_InterlockedExchangeAdd16_rel:
-  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);
-  case ARM::BI_InterlockedCompareExchange8_acq:
-  case ARM::BI_InterlockedCompareExchange16_acq:
-  case ARM::BI_InterlockedCompareExchange_acq:
-  case ARM::BI_InterlockedCompareExchange64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedCompareExchange_acq, E);
-  case ARM::BI_InterlockedCompareExchange8_rel:
-  case ARM::BI_InterlockedCompareExchange16_rel:
-  case ARM::BI_InterlockedCompareExchange_rel:
-  case ARM::BI_InterlockedCompareExchange64_rel:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedCompareExchange_rel, E);
-  case ARM::BI_InterlockedCompareExchange8_nf:
-  case ARM::BI_InterlockedCompareExchange16_nf:
-  case ARM::BI_InterlockedCompareExchange_nf:
-  case ARM::BI_InterlockedCompareExchange64_nf:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedCompareExchange_nf, E);
-  case ARM::BI_InterlockedOr8_acq:
-  case ARM::BI_InterlockedOr16_acq:
-  case ARM::BI_InterlockedOr_acq:
-  case ARM::BI_InterlockedOr64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr_acq, E);
-  case ARM::BI_InterlockedOr8_rel:
-  case ARM::BI_InterlockedOr16_rel:
-  case ARM::BI_InterlockedOr_rel:
-  case ARM::BI_InterlockedOr64_rel:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr_rel, E);
-  case ARM::BI_InterlockedOr8_nf:
-  case ARM::BI_InterlockedOr16_nf:
-  case ARM::BI_InterlockedOr_nf:
-  case ARM::BI_InterlockedOr64_nf:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr_nf, E);
-  case ARM::BI_InterlockedXor8_acq:
-  case ARM::BI_InterlockedXor16_acq:
-  case ARM::BI_InterlockedXor_acq:
-  case ARM::BI_InterlockedXor64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor_acq, E);
-  case ARM::BI_InterlockedXor8_rel:
-  case ARM::BI_InterlockedXor16_rel:
-  case ARM::BI_InterlockedXor_rel:
-  case ARM::BI_InterlockedXor64_rel:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor_rel, E);
-  case ARM::BI_InterlockedXor8_nf:
-  case ARM::BI_InterlockedXor16_nf:
-  case ARM::BI_InterlockedXor_nf:
-  case ARM::BI_InterlockedXor64_nf:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor_nf, E);
-  case ARM::BI_InterlockedAnd8_acq:
-  case ARM::BI_InterlockedAnd16_acq:
-  case ARM::BI_InterlockedAnd_acq:
-  case ARM::BI_InterlockedAnd64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd_acq, E);
-  case ARM::BI_InterlockedAnd8_rel:
-  case ARM::BI_InterlockedAnd16_rel:
-  case ARM::BI_InterlockedAnd_rel:
-  case ARM::BI_InterlockedAnd64_rel:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd_rel, E);
-  case ARM::BI_InterlockedAnd8_nf:
-  case ARM::BI_InterlockedAnd16_nf:
-  case ARM::BI_InterlockedAnd_nf:
-  case ARM::BI_InterlockedAnd64_nf:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd_nf, E);
-  case ARM::BI_InterlockedIncrement16_acq:
-  case ARM::BI_InterlockedIncrement_acq:
-  case ARM::BI_InterlockedIncrement64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement_acq, E);
-  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.
   }
 
   // Get the last argument, which specifies the vector type.


@@ -8949,6 +9142,11 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
     return Builder.CreateCall(F);
   }
 
     return Builder.CreateCall(F);
   }
 

+  // Handle MSVC intrinsics before argument evaluation to prevent double
+  // evaluation.
+  if (Optional<MSVCIntrin> MsvcIntId = translateAarch64ToMsvcIntrin(BuiltinID))
+    return EmitMSVCBuiltinExpr(*MsvcIntId, E);
+

   // Find out if any arguments are required to be integer constant
   // expressions.
   unsigned ICEArguments = 0;
   // Find out if any arguments are required to be integer constant
   // expressions.
   unsigned ICEArguments = 0;


@@ -9650,142 +9848,6 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
     return Builder.CreateExtractElement(Ops[0], EmitScalarExpr(E->getArg(1)),
                                         "vgetq_lane");
   }
     return Builder.CreateExtractElement(Ops[0], EmitScalarExpr(E->getArg(1)),
                                         "vgetq_lane");
   }

-  case AArch64::BI_BitScanForward:
-  case AArch64::BI_BitScanForward64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_BitScanForward, E);
-  case AArch64::BI_BitScanReverse:
-  case AArch64::BI_BitScanReverse64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_BitScanReverse, E);
-  case AArch64::BI_InterlockedAnd64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd, E);
-  case AArch64::BI_InterlockedExchange64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange, E);
-  case AArch64::BI_InterlockedExchangeAdd64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd, E);
-  case AArch64::BI_InterlockedExchangeSub64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeSub, E);
-  case AArch64::BI_InterlockedOr64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr, E);
-  case AArch64::BI_InterlockedXor64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor, E);
-  case AArch64::BI_InterlockedDecrement64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement, E);
-  case AArch64::BI_InterlockedIncrement64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement, E);
-  case AArch64::BI_InterlockedExchangeAdd8_acq:
-  case AArch64::BI_InterlockedExchangeAdd16_acq:
-  case AArch64::BI_InterlockedExchangeAdd_acq:
-  case AArch64::BI_InterlockedExchangeAdd64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd_acq, E);
-  case AArch64::BI_InterlockedExchangeAdd8_rel:
-  case AArch64::BI_InterlockedExchangeAdd16_rel:
-  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_InterlockedCompareExchange8_acq:
-  case AArch64::BI_InterlockedCompareExchange16_acq:
-  case AArch64::BI_InterlockedCompareExchange_acq:
-  case AArch64::BI_InterlockedCompareExchange64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedCompareExchange_acq, E);
-  case AArch64::BI_InterlockedCompareExchange8_rel:
-  case AArch64::BI_InterlockedCompareExchange16_rel:
-  case AArch64::BI_InterlockedCompareExchange_rel:
-  case AArch64::BI_InterlockedCompareExchange64_rel:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedCompareExchange_rel, E);
-  case AArch64::BI_InterlockedCompareExchange8_nf:
-  case AArch64::BI_InterlockedCompareExchange16_nf:
-  case AArch64::BI_InterlockedCompareExchange_nf:
-  case AArch64::BI_InterlockedCompareExchange64_nf:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedCompareExchange_nf, E);
-  case AArch64::BI_InterlockedOr8_acq:
-  case AArch64::BI_InterlockedOr16_acq:
-  case AArch64::BI_InterlockedOr_acq:
-  case AArch64::BI_InterlockedOr64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr_acq, E);
-  case AArch64::BI_InterlockedOr8_rel:
-  case AArch64::BI_InterlockedOr16_rel:
-  case AArch64::BI_InterlockedOr_rel:
-  case AArch64::BI_InterlockedOr64_rel:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr_rel, E);
-  case AArch64::BI_InterlockedOr8_nf:
-  case AArch64::BI_InterlockedOr16_nf:
-  case AArch64::BI_InterlockedOr_nf:
-  case AArch64::BI_InterlockedOr64_nf:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr_nf, E);
-  case AArch64::BI_InterlockedXor8_acq:
-  case AArch64::BI_InterlockedXor16_acq:
-  case AArch64::BI_InterlockedXor_acq:
-  case AArch64::BI_InterlockedXor64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor_acq, E);
-  case AArch64::BI_InterlockedXor8_rel:
-  case AArch64::BI_InterlockedXor16_rel:
-  case AArch64::BI_InterlockedXor_rel:
-  case AArch64::BI_InterlockedXor64_rel:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor_rel, E);
-  case AArch64::BI_InterlockedXor8_nf:
-  case AArch64::BI_InterlockedXor16_nf:
-  case AArch64::BI_InterlockedXor_nf:
-  case AArch64::BI_InterlockedXor64_nf:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor_nf, E);
-  case AArch64::BI_InterlockedAnd8_acq:
-  case AArch64::BI_InterlockedAnd16_acq:
-  case AArch64::BI_InterlockedAnd_acq:
-  case AArch64::BI_InterlockedAnd64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd_acq, E);
-  case AArch64::BI_InterlockedAnd8_rel:
-  case AArch64::BI_InterlockedAnd16_rel:
-  case AArch64::BI_InterlockedAnd_rel:
-  case AArch64::BI_InterlockedAnd64_rel:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd_rel, E);
-  case AArch64::BI_InterlockedAnd8_nf:
-  case AArch64::BI_InterlockedAnd16_nf:
-  case AArch64::BI_InterlockedAnd_nf:
-  case AArch64::BI_InterlockedAnd64_nf:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd_nf, E);
-  case AArch64::BI_InterlockedIncrement16_acq:
-  case AArch64::BI_InterlockedIncrement_acq:
-  case AArch64::BI_InterlockedIncrement64_acq:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement_acq, E);
-  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));
 
   case AArch64::BI_InterlockedAdd: {
     Value *Arg0 = EmitScalarExpr(E->getArg(0));


@@ -11759,6 +11821,11 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
   if (BuiltinID == X86::BI__builtin_cpu_init)
     return EmitX86CpuInit();
 
   if (BuiltinID == X86::BI__builtin_cpu_init)
     return EmitX86CpuInit();
 

+  // Handle MSVC intrinsics before argument evaluation to prevent double
+  // evaluation.
+  if (Optional<MSVCIntrin> MsvcIntId = translateX86ToMsvcIntrin(BuiltinID))
+    return EmitMSVCBuiltinExpr(*MsvcIntId, E);
+

   SmallVector<Value*, 4> Ops;
   bool IsMaskFCmp = false;
 
   SmallVector<Value*, 4> Ops;
   bool IsMaskFCmp = false;
 


@@ -13965,29 +14032,6 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
     return Builder.CreateFence(llvm::AtomicOrdering::SequentiallyConsistent,
                                llvm::SyncScope::SingleThread);
   }
     return Builder.CreateFence(llvm::AtomicOrdering::SequentiallyConsistent,
                                llvm::SyncScope::SingleThread);
   }

-  case X86::BI_BitScanForward:
-  case X86::BI_BitScanForward64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_BitScanForward, E);
-  case X86::BI_BitScanReverse:
-  case X86::BI_BitScanReverse64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_BitScanReverse, E);
-
-  case X86::BI_InterlockedAnd64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedAnd, E);
-  case X86::BI_InterlockedExchange64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchange, E);
-  case X86::BI_InterlockedExchangeAdd64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeAdd, E);
-  case X86::BI_InterlockedExchangeSub64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedExchangeSub, E);
-  case X86::BI_InterlockedOr64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedOr, E);
-  case X86::BI_InterlockedXor64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedXor, E);
-  case X86::BI_InterlockedDecrement64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedDecrement, E);
-  case X86::BI_InterlockedIncrement64:
-    return EmitMSVCBuiltinExpr(MSVCIntrin::_InterlockedIncrement, E);

   case X86::BI_InterlockedCompareExchange128: {
     // InterlockedCompareExchange128 doesn't directly refer to 128bit ints,
     // instead it takes pointers to 64bit ints for Destination and
   case X86::BI_InterlockedCompareExchange128: {
     // InterlockedCompareExchange128 doesn't directly refer to 128bit ints,
     // instead it takes pointers to 64bit ints for Destination and

diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index 40efa6d..95e7d77 100644 (file)
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -4120,10 +4120,7 @@ public:
                                llvm::AtomicOrdering &AO,
                                llvm::SyncScope::ID &SSID);
 
                                llvm::AtomicOrdering &AO,
                                llvm::SyncScope::ID &SSID);
 

-private:

   enum class MSVCIntrin;
   enum class MSVCIntrin;

-
-public:

   llvm::Value *EmitMSVCBuiltinExpr(MSVCIntrin BuiltinID, const CallExpr *E);
 
   llvm::Value *EmitBuiltinAvailable(const VersionTuple &Version);
   llvm::Value *EmitMSVCBuiltinExpr(MSVCIntrin BuiltinID, const CallExpr *E);
 
   llvm::Value *EmitBuiltinAvailable(const VersionTuple &Version);

diff --git a/clang/test/CodeGen/ms-intrinsics.c b/clang/test/CodeGen/ms-intrinsics.c
index 5182ae4..14e591e 100644 (file)
--- a/clang/test/CodeGen/ms-intrinsics.c
+++ b/clang/test/CodeGen/ms-intrinsics.c
@@ -147,7 +147,7 @@ void *test_AddressOfReturnAddress() {
 #endif
 
 unsigned char test_BitScanForward(unsigned long *Index, unsigned long Mask) {
 #endif
 
 unsigned char test_BitScanForward(unsigned long *Index, unsigned long Mask) {

-  return _BitScanForward(Index, Mask);
+  return _BitScanForward(++Index, Mask);

 }
 // CHECK: define{{.*}}i8 @test_BitScanForward(i32* {{[a-z_ ]*}}%Index, i32 {{[a-z_ ]*}}%Mask){{.*}}{
 // CHECK:   [[ISNOTZERO:%[a-z0-9._]+]] = icmp eq i32 %Mask, 0
 }
 // CHECK: define{{.*}}i8 @test_BitScanForward(i32* {{[a-z_ ]*}}%Index, i32 {{[a-z_ ]*}}%Mask){{.*}}{
 // CHECK:   [[ISNOTZERO:%[a-z0-9._]+]] = icmp eq i32 %Mask, 0


@@ -156,12 +156,13 @@ unsigned char test_BitScanForward(unsigned long *Index, unsigned long Mask) {
 // CHECK:   [[RESULT:%[a-z0-9._]+]] = phi i8 [ 0, %[[ISZERO_LABEL:[a-z0-9._]+]] ], [ 1, %[[ISNOTZERO_LABEL]] ]
 // CHECK:   ret i8 [[RESULT]]
 // CHECK:   [[ISNOTZERO_LABEL]]:
 // CHECK:   [[RESULT:%[a-z0-9._]+]] = phi i8 [ 0, %[[ISZERO_LABEL:[a-z0-9._]+]] ], [ 1, %[[ISNOTZERO_LABEL]] ]
 // CHECK:   ret i8 [[RESULT]]
 // CHECK:   [[ISNOTZERO_LABEL]]:

+// CHECK:   [[IDXGEP:%[a-z0-9._]+]] = getelementptr inbounds i32, i32* %Index, {{i64|i32}} 1

 // CHECK:   [[INDEX:%[0-9]+]] = tail call i32 @llvm.cttz.i32(i32 %Mask, i1 true)
 // CHECK:   [[INDEX:%[0-9]+]] = tail call i32 @llvm.cttz.i32(i32 %Mask, i1 true)

-// CHECK:   store i32 [[INDEX]], i32* %Index, align 4
+// CHECK:   store i32 [[INDEX]], i32* [[IDXGEP]], align 4

 // CHECK:   br label %[[END_LABEL]]
 
 unsigned char test_BitScanReverse(unsigned long *Index, unsigned long Mask) {
 // CHECK:   br label %[[END_LABEL]]
 
 unsigned char test_BitScanReverse(unsigned long *Index, unsigned long Mask) {

-  return _BitScanReverse(Index, Mask);
+  return _BitScanReverse(++Index, Mask);

 }
 // CHECK: define{{.*}}i8 @test_BitScanReverse(i32* {{[a-z_ ]*}}%Index, i32 {{[a-z_ ]*}}%Mask){{.*}}{
 // CHECK:   [[ISNOTZERO:%[0-9]+]] = icmp eq i32 %Mask, 0
 }
 // CHECK: define{{.*}}i8 @test_BitScanReverse(i32* {{[a-z_ ]*}}%Index, i32 {{[a-z_ ]*}}%Mask){{.*}}{
 // CHECK:   [[ISNOTZERO:%[0-9]+]] = icmp eq i32 %Mask, 0


@@ -170,9 +171,10 @@ unsigned char test_BitScanReverse(unsigned long *Index, unsigned long Mask) {
 // CHECK:   [[RESULT:%[a-z0-9._]+]] = phi i8 [ 0, %[[ISZERO_LABEL:[a-z0-9._]+]] ], [ 1, %[[ISNOTZERO_LABEL]] ]
 // CHECK:   ret i8 [[RESULT]]
 // CHECK:   [[ISNOTZERO_LABEL]]:
 // CHECK:   [[RESULT:%[a-z0-9._]+]] = phi i8 [ 0, %[[ISZERO_LABEL:[a-z0-9._]+]] ], [ 1, %[[ISNOTZERO_LABEL]] ]
 // CHECK:   ret i8 [[RESULT]]
 // CHECK:   [[ISNOTZERO_LABEL]]:

+// CHECK:   [[IDXGEP:%[a-z0-9._]+]] = getelementptr inbounds i32, i32* %Index, {{i64|i32}} 1

 // CHECK:   [[REVINDEX:%[0-9]+]] = tail call i32 @llvm.ctlz.i32(i32 %Mask, i1 true)
 // CHECK:   [[INDEX:%[0-9]+]] = xor i32 [[REVINDEX]], 31
 // CHECK:   [[REVINDEX:%[0-9]+]] = tail call i32 @llvm.ctlz.i32(i32 %Mask, i1 true)
 // CHECK:   [[INDEX:%[0-9]+]] = xor i32 [[REVINDEX]], 31

-// CHECK:   store i32 [[INDEX]], i32* %Index, align 4
+// CHECK:   store i32 [[INDEX]], i32* [[IDXGEP]], align 4

 // CHECK:   br label %[[END_LABEL]]
 
 #if defined(__x86_64__) || defined(__arm__) || defined(__aarch64__)
 // CHECK:   br label %[[END_LABEL]]
 
 #if defined(__x86_64__) || defined(__arm__) || defined(__aarch64__)


@@ -459,19 +461,21 @@ unsigned char test_InterlockedCompareExchange128(
 #endif
 
 short test_InterlockedIncrement16(short volatile *Addend) {
 #endif
 
 short test_InterlockedIncrement16(short volatile *Addend) {

-  return _InterlockedIncrement16(Addend);
+  return _InterlockedIncrement16(++Addend);

 }
 // CHECK: define{{.*}}i16 @test_InterlockedIncrement16(i16*{{[a-z_ ]*}}%Addend){{.*}}{
 }
 // CHECK: define{{.*}}i16 @test_InterlockedIncrement16(i16*{{[a-z_ ]*}}%Addend){{.*}}{

-// CHECK: [[TMP:%[0-9]+]] = atomicrmw add i16* %Addend, i16 1 seq_cst
+// CHECK: %incdec.ptr = getelementptr inbounds i16, i16* %Addend, {{i64|i32}} 1
+// CHECK: [[TMP:%[0-9]+]] = atomicrmw add i16* %incdec.ptr, i16 1 seq_cst

 // CHECK: [[RESULT:%[0-9]+]] = add i16 [[TMP]], 1
 // CHECK: ret i16 [[RESULT]]
 // CHECK: }
 
 long test_InterlockedIncrement(long volatile *Addend) {
 // CHECK: [[RESULT:%[0-9]+]] = add i16 [[TMP]], 1
 // CHECK: ret i16 [[RESULT]]
 // CHECK: }
 
 long test_InterlockedIncrement(long volatile *Addend) {

-  return _InterlockedIncrement(Addend);
+  return _InterlockedIncrement(++Addend);

 }
 // CHECK: define{{.*}}i32 @test_InterlockedIncrement(i32*{{[a-z_ ]*}}%Addend){{.*}}{
 }
 // CHECK: define{{.*}}i32 @test_InterlockedIncrement(i32*{{[a-z_ ]*}}%Addend){{.*}}{

-// CHECK: [[TMP:%[0-9]+]] = atomicrmw add i32* %Addend, i32 1 seq_cst
+// CHECK: %incdec.ptr = getelementptr inbounds i32, i32* %Addend, {{i64|i32}} 1
+// CHECK: [[TMP:%[0-9]+]] = atomicrmw add i32* %incdec.ptr, i32 1 seq_cst

 // CHECK: [[RESULT:%[0-9]+]] = add i32 [[TMP]], 1
 // CHECK: ret i32 [[RESULT]]
 // CHECK: }
 // CHECK: [[RESULT:%[0-9]+]] = add i32 [[TMP]], 1
 // CHECK: ret i32 [[RESULT]]
 // CHECK: }