Add MV-index to GlobalDecl, spread GlobalDecl through CodeGen (CPU-Dispatch cleanup)

[erichkeane]

Summary

As brought up by Richard Smith on https://reviews.llvm.org/D51650:

Sorry, I don't think it's acceptable from a design perspective to have
mutable state in an Attr object, even if you can ensure that only one
client of the Attr will be using the state at a time. CodeGen is going
to need to track its own index into the attribute's list of clones.

The proposed fix was to put it into GlobalDecl, which this patch does.
However, doing this also requires making GlobalDecl be passed around
more often instead of FunctionDecl, so this does that as well.

cfe/trunk/include/clang/AST/GlobalDecl.h

 namespace clang {
 
 /// GlobalDecl - represents a global declaration. This can either be a
 /// CXXConstructorDecl and the constructor type (Base, Complete).
 /// a CXXDestructorDecl and the destructor type (Base, Complete) or
 /// a VarDecl, a FunctionDecl or a BlockDecl.
 class GlobalDecl {
   llvm::PointerIntPair<const Decl *, 2> Value;
+  unsigned MultiVersionIndex = 0;
 
   void Init(const Decl *D) {
     assert(!isa<CXXConstructorDecl>(D) && "Use other ctor with ctor decls!");
     assert(!isa<CXXDestructorDecl>(D) && "Use other ctor with dtor decls!");
 
     Value.setPointer(D);
   }
 
 public:
   GlobalDecl() = default;
   GlobalDecl(const VarDecl *D) { Init(D);}
-  GlobalDecl(const FunctionDecl *D) { Init(D); }
+  GlobalDecl(const FunctionDecl *D, unsigned MVIndex = 0)
+      : MultiVersionIndex(MVIndex) {
+    Init(D);
+  }
   GlobalDecl(const BlockDecl *D) { Init(D); }
   GlobalDecl(const CapturedDecl *D) { Init(D); }
   GlobalDecl(const ObjCMethodDecl *D) { Init(D); }
   GlobalDecl(const OMPDeclareReductionDecl *D) { Init(D); }
   GlobalDecl(const CXXConstructorDecl *D, CXXCtorType Type) : Value(D, Type) {}
   GlobalDecl(const CXXDestructorDecl *D, CXXDtorType Type) : Value(D, Type) {}
 
   GlobalDecl getCanonicalDecl() const {
     GlobalDecl CanonGD;
     CanonGD.Value.setPointer(Value.getPointer()->getCanonicalDecl());
     CanonGD.Value.setInt(Value.getInt());
+    CanonGD.MultiVersionIndex = MultiVersionIndex;
 
     return CanonGD;
   }
 
   const Decl *getDecl() const { return Value.getPointer(); }
 
   CXXCtorType getCtorType() const {
     assert(isa<CXXConstructorDecl>(getDecl()) && "Decl is not a ctor!");
     return static_cast<CXXCtorType>(Value.getInt());
   }
 
   CXXDtorType getDtorType() const {
     assert(isa<CXXDestructorDecl>(getDecl()) && "Decl is not a dtor!");
     return static_cast<CXXDtorType>(Value.getInt());
   }
 
+  unsigned getMultiVersionIndex() const {
+    assert(isa<FunctionDecl>(getDecl()) &&
+           !isa<CXXConstructorDecl>(getDecl()) &&
+           !isa<CXXDestructorDecl>(getDecl()) &&
+           "Decl is not a plain FunctionDecl!");
+    return MultiVersionIndex;
+  }
+
   friend bool operator==(const GlobalDecl &LHS, const GlobalDecl &RHS) {
-    return LHS.Value == RHS.Value;
+    return LHS.Value == RHS.Value &&
+           LHS.MultiVersionIndex == RHS.MultiVersionIndex;
   }
 
   void *getAsOpaquePtr() const { return Value.getOpaqueValue(); }
 
   static GlobalDecl getFromOpaquePtr(void *P) {
     GlobalDecl GD;
     GD.Value.setFromOpaqueValue(P);
     return GD;
   }
 
   GlobalDecl getWithDecl(const Decl *D) {
     GlobalDecl Result(*this);
     Result.Value.setPointer(D);
     return Result;
   }
+
+  GlobalDecl getWithMultiVersionIndex(unsigned Index) {
+    assert(isa<FunctionDecl>(getDecl()) &&
+           !isa<CXXConstructorDecl>(getDecl()) &&
+           !isa<CXXDestructorDecl>(getDecl()) &&
+           "Decl is not a plain FunctionDecl!");
+    GlobalDecl Result(*this);
+    Result.MultiVersionIndex = Index;
+    return Result;
+  }
 };
 
 } // namespace clang
 
 namespace llvm {
 
   template<> struct DenseMapInfo<clang::GlobalDecl> {
     static inline clang::GlobalDecl getEmptyKey() {

cfe/trunk/include/clang/Basic/Attr.td

 }
 
 def CPUSpecific : InheritableAttr {
   let Spellings = [Clang<"cpu_specific">, Declspec<"cpu_specific">];
   let Args = [VariadicIdentifierArgument<"Cpus">];
   let Subjects = SubjectList<[Function]>;
   let Documentation = [CPUSpecificCPUDispatchDocs];
   let AdditionalMembers = [{
-    unsigned ActiveArgIndex = 0;
-
-    IdentifierInfo *getCurCPUName() const {
-      return *(cpus_begin() + ActiveArgIndex);
+    IdentifierInfo *getCPUName(unsigned Index) const {
+      return *(cpus_begin() + Index);
     }
   }];
 }
 
 def CPUDispatch : InheritableAttr {
   let Spellings = [Clang<"cpu_dispatch">, Declspec<"cpu_dispatch">];
   let Args = [VariadicIdentifierArgument<"Cpus">];
   let Subjects = SubjectList<[Function]>;

cfe/trunk/lib/CodeGen/CGBlocks.cpp

static void setBlockHelperAttributesVisibility(bool CapturesNonExternalType,
                                                llvm::Function *Fn,
                                                const CGFunctionInfo &FI,
                                                CodeGenModule &CGM) {
   if (CapturesNonExternalType) {
     CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FI);
   } else {
     Fn->setVisibility(llvm::GlobalValue::HiddenVisibility);
     Fn->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
-    CGM.SetLLVMFunctionAttributes(nullptr, FI, Fn);
+    CGM.SetLLVMFunctionAttributes(GlobalDecl(), FI, Fn);
     CGM.SetLLVMFunctionAttributesForDefinition(nullptr, Fn);
   }
 }
 /// Generate the copy-helper function for a block closure object:
 ///   static void block_copy_helper(block_t *dst, block_t *src);
 /// The runtime will have previously initialized 'dst' by doing a
 /// bit-copy of 'src'.
 ///

cfe/trunk/lib/CodeGen/CGBuiltin.cpp

if (Ty->isPPC_FP128Ty()) {
     V = CGF.Builder.CreateTrunc(V, IntTy);
   }
   Value *Zero = llvm::Constant::getNullValue(IntTy);
   return CGF.Builder.CreateICmpSLT(V, Zero);
 }
 
 static RValue emitLibraryCall(CodeGenFunction &CGF, const FunctionDecl *FD,
                               const CallExpr *E, llvm::Constant *calleeValue) {
-  CGCallee callee = CGCallee::forDirect(calleeValue, FD);
+  CGCallee callee = CGCallee::forDirect(calleeValue, GlobalDecl(FD));
   return CGF.EmitCall(E->getCallee()->getType(), callee, E, ReturnValueSlot());
 }
 
 /// Emit a call to llvm.{sadd,uadd,ssub,usub,smul,umul}.with.overflow.*
 /// depending on IntrinsicID.
 ///
 /// \arg CGF The current codegen function.
 /// \arg IntrinsicID The ID for the Intrinsic we wish to generate.
llvm::Function *CodeGenFunction::generateBuiltinOSLogHelperFunction(
   // identical functions. To ensure the merging always happens, 'noinline' is
   // attached to the function when compiling with -Oz.
   const CGFunctionInfo &FI =
       CGM.getTypes().arrangeBuiltinFunctionDeclaration(ReturnTy, Args);
   llvm::FunctionType *FuncTy = CGM.getTypes().GetFunctionType(FI);
   llvm::Function *Fn = llvm::Function::Create(
       FuncTy, llvm::GlobalValue::LinkOnceODRLinkage, Name, &CGM.getModule());
   Fn->setVisibility(llvm::GlobalValue::HiddenVisibility);
-  CGM.SetLLVMFunctionAttributes(nullptr, FI, Fn);
+  CGM.SetLLVMFunctionAttributes(GlobalDecl(), FI, Fn);
   CGM.SetLLVMFunctionAttributesForDefinition(nullptr, Fn);
 
   // Attach 'noinline' at -Oz.
   if (CGM.getCodeGenOpts().OptimizeSize == 2)
     Fn->addFnAttr(llvm::Attribute::NoInline);
 
   auto NL = ApplyDebugLocation::CreateEmpty(*this);
   IdentifierInfo *II = &Ctx.Idents.get(Name);
RValue CodeGenFunction::emitRotate(const CallExpr *E, bool IsRotateRight) {
   ShiftAmt = Builder.CreateIntCast(ShiftAmt, Ty, false);
 
   // Rotate is a special case of LLVM funnel shift - 1st 2 args are the same.
   unsigned IID = IsRotateRight ? Intrinsic::fshr : Intrinsic::fshl;
   Value *F = CGM.getIntrinsic(IID, Ty);
   return RValue::get(Builder.CreateCall(F, { Src, Src, ShiftAmt }));
 }
 
-RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
-                                        unsigned BuiltinID, const CallExpr *E,
+RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
+                                        const CallExpr *E,
                                         ReturnValueSlot ReturnValue) {
+  const FunctionDecl *FD = GD.getDecl()->getAsFunction();
   // See if we can constant fold this builtin.  If so, don't emit it at all.
   Expr::EvalResult Result;
   if (E->EvaluateAsRValue(Result, CGM.getContext()) &&
       !Result.hasSideEffects()) {
     if (Result.Val.isInt())
       return RValue::get(llvm::ConstantInt::get(getLLVMContext(),
                                                 Result.Val.getInt()));
     if (Result.Val.isFloat())

cfe/trunk/lib/CodeGen/CGCXX.cpp

static CGCallee BuildAppleKextVirtualCall(CodeGenFunction &CGF,
   VTableLayout::AddressPointLocation AddressPoint =
       VTLayout.getAddressPoint(BaseSubobject(RD, CharUnits::Zero()));
   VTableIndex += VTLayout.getVTableOffset(AddressPoint.VTableIndex) +
                  AddressPoint.AddressPointIndex;
   llvm::Value *VFuncPtr =
     CGF.Builder.CreateConstInBoundsGEP1_64(VTable, VTableIndex, "vfnkxt");
   llvm::Value *VFunc =
     CGF.Builder.CreateAlignedLoad(VFuncPtr, CGF.PointerAlignInBytes);
-  CGCallee Callee(GD.getDecl()->getCanonicalDecl(), VFunc);
+  CGCallee Callee(GD, VFunc);
   return Callee;
 }
 
 /// BuildAppleKextVirtualCall - This routine is to support gcc's kext ABI making
 /// indirect call to virtual functions. It makes the call through indexing
 /// into the vtable.
 CGCallee
 CodeGenFunction::BuildAppleKextVirtualCall(const CXXMethodDecl *MD,

cfe/trunk/lib/CodeGen/CGCall.h

 
 namespace CodeGen {
 
 /// Abstract information about a function or function prototype.
 class CGCalleeInfo {
   /// The function prototype of the callee.
   const FunctionProtoType *CalleeProtoTy;
   /// The function declaration of the callee.
-  const Decl *CalleeDecl;
+  GlobalDecl CalleeDecl;
 
 public:
-  explicit CGCalleeInfo() : CalleeProtoTy(nullptr), CalleeDecl(nullptr) {}
-  CGCalleeInfo(const FunctionProtoType *calleeProtoTy, const Decl *calleeDecl)
+  explicit CGCalleeInfo() : CalleeProtoTy(nullptr), CalleeDecl() {}
+  CGCalleeInfo(const FunctionProtoType *calleeProtoTy, GlobalDecl calleeDecl)
       : CalleeProtoTy(calleeProtoTy), CalleeDecl(calleeDecl) {}
   CGCalleeInfo(const FunctionProtoType *calleeProtoTy)
-      : CalleeProtoTy(calleeProtoTy), CalleeDecl(nullptr) {}
-  CGCalleeInfo(const Decl *calleeDecl)
+      : CalleeProtoTy(calleeProtoTy), CalleeDecl() {}
+  CGCalleeInfo(GlobalDecl calleeDecl)
       : CalleeProtoTy(nullptr), CalleeDecl(calleeDecl) {}
 
   const FunctionProtoType *getCalleeFunctionProtoType() const {
     return CalleeProtoTy;
   }
-  const Decl *getCalleeDecl() const { return CalleeDecl; }
+  const GlobalDecl getCalleeDecl() const { return CalleeDecl; }
   };
 
   /// All available information about a concrete callee.
   class CGCallee {
     enum class SpecialKind : uintptr_t {
       Invalid,
       Builtin,
       PseudoDestructor,
const CXXPseudoDestructorExpr *getPseudoDestructorExpr() const {
       return PseudoDestructorInfo.Expr;
     }
 
     bool isOrdinary() const {
       return uintptr_t(KindOrFunctionPointer) > uintptr_t(SpecialKind::Last);
     }
     CGCalleeInfo getAbstractInfo() const {
       if (isVirtual())
-        return VirtualInfo.MD.getDecl();
+        return VirtualInfo.MD;
       assert(isOrdinary());
       return AbstractInfo;
     }
     llvm::Value *getFunctionPointer() const {
       assert(isOrdinary());
       return reinterpret_cast<llvm::Value*>(uintptr_t(KindOrFunctionPointer));
     }
     void setFunctionPointer(llvm::Value *functionPtr) {

cfe/trunk/lib/CodeGen/CGCall.cpp

void CodeGenModule::ConstructAttributeList(
   if (FI.isNoReturn())
     FuncAttrs.addAttribute(llvm::Attribute::NoReturn);
 
   // If we have information about the function prototype, we can learn
   // attributes from there.
   AddAttributesFromFunctionProtoType(getContext(), FuncAttrs,
                                      CalleeInfo.getCalleeFunctionProtoType());
 
-  const Decl *TargetDecl = CalleeInfo.getCalleeDecl();
+  const Decl *TargetDecl = CalleeInfo.getCalleeDecl().getDecl();
 
   bool HasOptnone = false;
   // FIXME: handle sseregparm someday...
   if (TargetDecl) {
     if (TargetDecl->hasAttr<ReturnsTwiceAttr>())
       FuncAttrs.addAttribute(llvm::Attribute::ReturnsTwice);
     if (TargetDecl->hasAttr<NoThrowAttr>())
       FuncAttrs.addAttribute(llvm::Attribute::NoUnwind);
else if (TargetDecl) {
         if (const auto *BD = dyn_cast<BlockDecl>(TargetDecl))
           if (!BD->doesNotEscape())
             DisableTailCalls = true;
       }
     }
 
     FuncAttrs.addAttribute("disable-tail-calls",
                            llvm::toStringRef(DisableTailCalls));
-    GetCPUAndFeaturesAttributes(TargetDecl, FuncAttrs);
+    GetCPUAndFeaturesAttributes(CalleeInfo.getCalleeDecl(), FuncAttrs);
   }
 
   ClangToLLVMArgMapping IRFunctionArgs(getContext(), FI);
 
   QualType RetTy = FI.getReturnType();
   const ABIArgInfo &RetAI = FI.getReturnInfo();
   switch (RetAI.getKind()) {
   case ABIArgInfo::Extend:
CGM.ConstructAttributeList(CalleePtr->getName(), CallInfo,
                              /*AttrOnCallSite=*/true);
 
   // Apply some call-site-specific attributes.
   // TODO: work this into building the attribute set.
 
   // Apply always_inline to all calls within flatten functions.
   // FIXME: should this really take priority over __try, below?
   if (CurCodeDecl && CurCodeDecl->hasAttr<FlattenAttr>() &&
-      !(Callee.getAbstractInfo().getCalleeDecl() &&
-        Callee.getAbstractInfo().getCalleeDecl()->hasAttr<NoInlineAttr>())) {
+      !(Callee.getAbstractInfo().getCalleeDecl().getDecl() &&
+        Callee.getAbstractInfo()
+            .getCalleeDecl()
+            .getDecl()
+            ->hasAttr<NoInlineAttr>())) {
     Attrs =
         Attrs.addAttribute(getLLVMContext(), llvm::AttributeList::FunctionIndex,
                            llvm::Attribute::AlwaysInline);
   }
 
   // Disable inlining inside SEH __try blocks.
   if (isSEHTryScope()) {
     Attrs =
#endif
 
   // In ObjC ARC mode with no ObjC ARC exception safety, tell the ARC
   // optimizer it can aggressively ignore unwind edges.
   if (CGM.getLangOpts().ObjCAutoRefCount)
     AddObjCARCExceptionMetadata(CI);
 
   // Suppress tail calls if requested.
   if (llvm::CallInst *Call = dyn_cast<llvm::CallInst>(CI)) {
-    const Decl *TargetDecl = Callee.getAbstractInfo().getCalleeDecl();
+    const Decl *TargetDecl = Callee.getAbstractInfo().getCalleeDecl().getDecl();
     if (TargetDecl && TargetDecl->hasAttr<NotTailCalledAttr>())
       Call->setTailCallKind(llvm::CallInst::TCK_NoTail);
   }
 
   // 4. Finish the call.
 
   // If the call doesn't return, finish the basic block and clear the
   // insertion point; this allows the rest of IRGen to discard
RValue Ret = [&] {
     case ABIArgInfo::Expand:
       llvm_unreachable("Invalid ABI kind for return argument");
     }
 
     llvm_unreachable("Unhandled ABIArgInfo::Kind");
   } ();
 
   // Emit the assume_aligned check on the return value.
-  const Decl *TargetDecl = Callee.getAbstractInfo().getCalleeDecl();
+  const Decl *TargetDecl = Callee.getAbstractInfo().getCalleeDecl().getDecl();
   if (Ret.isScalar() && TargetDecl) {
     if (const auto *AA = TargetDecl->getAttr<AssumeAlignedAttr>()) {
       llvm::Value *OffsetValue = nullptr;
       if (const auto *Offset = AA->getOffset())
         OffsetValue = EmitScalarExpr(Offset);
 
       llvm::Value *Alignment = EmitScalarExpr(AA->getAlignment());
       llvm::ConstantInt *AlignmentCI = cast<llvm::ConstantInt>(Alignment);

cfe/trunk/lib/CodeGen/CGClass.cpp

CGCXXABI::AddedStructorArgs ExtraArgs =
       CGM.getCXXABI().addImplicitConstructorArgs(*this, D, Type, ForVirtualBase,
                                                  Delegating, Args);
 
   // Emit the call.
   llvm::Constant *CalleePtr =
     CGM.getAddrOfCXXStructor(D, getFromCtorType(Type));
   const CGFunctionInfo &Info = CGM.getTypes().arrangeCXXConstructorCall(
       Args, D, Type, ExtraArgs.Prefix, ExtraArgs.Suffix, PassPrototypeArgs);
-  CGCallee Callee = CGCallee::forDirect(CalleePtr, D);
+  CGCallee Callee = CGCallee::forDirect(CalleePtr, GlobalDecl(D, Type));
   EmitCall(Info, Callee, ReturnValueSlot(), Args);
 
   // Generate vtable assumptions if we're constructing a complete object
   // with a vtable.  We don't do this for base subobjects for two reasons:
   // first, it's incorrect for classes with virtual bases, and second, we're
   // about to overwrite the vptrs anyway.
   // We also have to make sure if we can refer to vtable:
   // - Otherwise we can refer to vtable if it's safe to speculatively emit.
if (!resultType->isVoidType() &&
       !hasScalarEvaluationKind(calleeFnInfo.getReturnType()))
     returnSlot = ReturnValueSlot(ReturnValue, resultType.isVolatileQualified());
 
   // We don't need to separately arrange the call arguments because
   // the call can't be variadic anyway --- it's impossible to forward
   // variadic arguments.
 
   // Now emit our call.
-  auto callee = CGCallee::forDirect(calleePtr, callOperator);
+  auto callee = CGCallee::forDirect(calleePtr, GlobalDecl(callOperator));
   RValue RV = EmitCall(calleeFnInfo, callee, returnSlot, callArgs);
 
   // If necessary, copy the returned value into the slot.
   if (!resultType->isVoidType() && returnSlot.isNull()) {
     if (getLangOpts().ObjCAutoRefCount && resultType->isObjCRetainableType()) {
       RV = RValue::get(EmitARCRetainAutoreleasedReturnValue(RV.getScalarVal()));
     }
     EmitReturnOfRValue(RV, resultType);

cfe/trunk/lib/CodeGen/CGException.cpp

llvm::Function *Fn = llvm::Function::Create(
       FnTy, llvm::GlobalValue::InternalLinkage, Name.str(), &CGM.getModule());
 
   IsOutlinedSEHHelper = true;
 
   StartFunction(GlobalDecl(), RetTy, Fn, FnInfo, Args,
                 OutlinedStmt->getBeginLoc(), OutlinedStmt->getBeginLoc());
   CurSEHParent = ParentCGF.CurSEHParent;
 
-  CGM.SetLLVMFunctionAttributes(nullptr, FnInfo, CurFn);
+  CGM.SetLLVMFunctionAttributes(GlobalDecl(), FnInfo, CurFn);
   EmitCapturedLocals(ParentCGF, OutlinedStmt, IsFilter);
 }
 
 /// Create a stub filter function that will ultimately hold the code of the
 /// filter expression. The EH preparation passes in LLVM will outline the code
 /// from the main function body into this stub.
 llvm::Function *
 CodeGenFunction::GenerateSEHFilterFunction(CodeGenFunction &ParentCGF,

cfe/trunk/lib/CodeGen/CGExpr.cpp

 }
 
 static CGCallee EmitDirectCallee(CodeGenFunction &CGF, const FunctionDecl *FD) {
   if (auto builtinID = FD->getBuiltinID()) {
     return CGCallee::forBuiltin(builtinID, FD);
   }
 
   llvm::Constant *calleePtr = EmitFunctionDeclPointer(CGF.CGM, FD);
-  return CGCallee::forDirect(calleePtr, FD);
+  return CGCallee::forDirect(calleePtr, GlobalDecl(FD));
 }
 
 CGCallee CodeGenFunction::EmitCallee(const Expr *E) {
   E = E->IgnoreParens();
 
   // Look through function-to-pointer decay.
   if (auto ICE = dyn_cast<ImplicitCastExpr>(E)) {
     if (ICE->getCastKind() == CK_FunctionToPointerDecay ||
CGCallee CodeGenFunction::EmitCallee(const Expr *E) {
   if (auto ptrType = E->getType()->getAs<PointerType>()) {
     calleePtr = EmitScalarExpr(E);
     functionType = ptrType->getPointeeType();
   } else {
     functionType = E->getType();
     calleePtr = EmitLValue(E).getPointer();
   }
   assert(functionType->isFunctionType());
-  CGCalleeInfo calleeInfo(functionType->getAs<FunctionProtoType>(),
-                          E->getReferencedDeclOfCallee());
+
+  GlobalDecl GD;
+  if (const auto *VD =
+          dyn_cast_or_null<VarDecl>(E->getReferencedDeclOfCallee()))
+    GD = GlobalDecl(VD);
+
+  CGCalleeInfo calleeInfo(functionType->getAs<FunctionProtoType>(), GD);
   CGCallee callee(calleeInfo, calleePtr);
   return callee;
 }
 
 LValue CodeGenFunction::EmitBinaryOperatorLValue(const BinaryOperator *E) {
   // Comma expressions just emit their LHS then their RHS as an l-value.
   if (E->getOpcode() == BO_Comma) {
     EmitIgnoredExpr(E->getLHS());

 RValue CodeGenFunction::EmitCall(QualType CalleeType, const CGCallee &OrigCallee,
                                  const CallExpr *E, ReturnValueSlot ReturnValue,
                                  llvm::Value *Chain) {
   // Get the actual function type. The callee type will always be a pointer to
   // function type or a block pointer type.
   assert(CalleeType->isFunctionPointerType() &&
          "Call must have function pointer type!");
 
-  const Decl *TargetDecl = OrigCallee.getAbstractInfo().getCalleeDecl();
+  const Decl *TargetDecl =
+      OrigCallee.getAbstractInfo().getCalleeDecl().getDecl();
 
   if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(TargetDecl))
     // We can only guarantee that a function is called from the correct
     // context/function based on the appropriate target attributes,
     // so only check in the case where we have both always_inline and target
     // since otherwise we could be making a conditional call after a check for
     // the proper cpu features (and it won't cause code generation issues due to
     // function based code generation).

cfe/trunk/lib/CodeGen/CGExprCXX.cpp

RValue CodeGenFunction::EmitCXXMemberCallExpr(const CXXMemberCallExpr *CE,
   if (isa<BinaryOperator>(callee))
     return EmitCXXMemberPointerCallExpr(CE, ReturnValue);
 
   const MemberExpr *ME = cast<MemberExpr>(callee);
   const CXXMethodDecl *MD = cast<CXXMethodDecl>(ME->getMemberDecl());
 
   if (MD->isStatic()) {
     // The method is static, emit it as we would a regular call.
-    CGCallee callee = CGCallee::forDirect(CGM.GetAddrOfFunction(MD), MD);
+    CGCallee callee =
+        CGCallee::forDirect(CGM.GetAddrOfFunction(MD), GlobalDecl(MD));
     return EmitCall(getContext().getPointerType(MD->getType()), callee, CE,
                     ReturnValue);
   }
 
   bool HasQualifier = ME->hasQualifier();
   NestedNameSpecifier *Qualifier = HasQualifier ? ME->getQualifier() : nullptr;
   bool IsArrow = ME->isArrow();
   const Expr *Base = ME->getBase();
if (UseVirtualCall) {
           cast<CXXMemberCallExpr>(CE));
     } else {
       CGCallee Callee;
       if (getLangOpts().AppleKext && MD->isVirtual() && HasQualifier)
         Callee = BuildAppleKextVirtualCall(MD, Qualifier, Ty);
       else if (!DevirtualizedMethod)
         Callee = CGCallee::forDirect(
             CGM.getAddrOfCXXStructor(Dtor, StructorType::Complete, FInfo, Ty),
-                                     Dtor);
+            GlobalDecl(Dtor, Dtor_Complete));
       else {
         const CXXDestructorDecl *DDtor =
           cast<CXXDestructorDecl>(DevirtualizedMethod);
         Callee = CGCallee::forDirect(
-                  CGM.GetAddrOfFunction(GlobalDecl(DDtor, Dtor_Complete), Ty),
-                                     DDtor);
+            GlobalDecl(DDtor, Dtor_Complete));
       }
       EmitCXXMemberOrOperatorCall(
           CalleeDecl, Callee, ReturnValue, This.getPointer(),
           /*ImplicitParam=*/nullptr, QualType(), CE, nullptr);
     }
     return RValue::get(nullptr);
   }
 
   CGCallee Callee;
   if (const CXXConstructorDecl *Ctor = dyn_cast<CXXConstructorDecl>(MD)) {
     Callee = CGCallee::forDirect(
-                  CGM.GetAddrOfFunction(GlobalDecl(Ctor, Ctor_Complete), Ty),
-                                 Ctor);
+        GlobalDecl(Ctor, Ctor_Complete));
   } else if (UseVirtualCall) {
     Callee = CGCallee::forVirtual(CE, MD, This.getAddress(), Ty);
   } else {
     if (SanOpts.has(SanitizerKind::CFINVCall) &&
         MD->getParent()->isDynamicClass()) {
       llvm::Value *VTable;
       const CXXRecordDecl *RD;
       std::tie(VTable, RD) =
           CGM.getCXXABI().LoadVTablePtr(*this, This.getAddress(),
                                         MD->getParent());
       EmitVTablePtrCheckForCall(RD, VTable, CFITCK_NVCall, CE->getBeginLoc());
     }
 
     if (getLangOpts().AppleKext && MD->isVirtual() && HasQualifier)
       Callee = BuildAppleKextVirtualCall(MD, Qualifier, Ty);
     else if (!DevirtualizedMethod)
-      Callee = CGCallee::forDirect(CGM.GetAddrOfFunction(MD, Ty), MD);
+      Callee =
+          CGCallee::forDirect(CGM.GetAddrOfFunction(MD, Ty), GlobalDecl(MD));
     else {
-      Callee = CGCallee::forDirect(
-                                CGM.GetAddrOfFunction(DevirtualizedMethod, Ty),
-                                   DevirtualizedMethod);
+      Callee =
+          CGCallee::forDirect(CGM.GetAddrOfFunction(DevirtualizedMethod, Ty),
+                              GlobalDecl(DevirtualizedMethod));
     }
   }
 
   if (MD->isVirtual()) {
     Address NewThisAddr =
         CGM.getCXXABI().adjustThisArgumentForVirtualFunctionCall(
             *this, CalleeDecl, This.getAddress(), UseVirtualCall);
     This.setAddress(NewThisAddr);

 /// Emit a call to an operator new or operator delete function, as implicitly
 /// created by new-expressions and delete-expressions.
 static RValue EmitNewDeleteCall(CodeGenFunction &CGF,
                                 const FunctionDecl *CalleeDecl,
                                 const FunctionProtoType *CalleeType,
                                 const CallArgList &Args) {
   llvm::Instruction *CallOrInvoke;
   llvm::Constant *CalleePtr = CGF.CGM.GetAddrOfFunction(CalleeDecl);
-  CGCallee Callee = CGCallee::forDirect(CalleePtr, CalleeDecl);
+  CGCallee Callee = CGCallee::forDirect(CalleePtr, GlobalDecl(CalleeDecl));
   RValue RV =
       CGF.EmitCall(CGF.CGM.getTypes().arrangeFreeFunctionCall(
                        Args, CalleeType, /*chainCall=*/false),
                    Callee, ReturnValueSlot(), Args, &CallOrInvoke);
 
   /// C++1y [expr.new]p10:
   ///   [In a new-expression,] an implementation is allowed to omit a call
   ///   to a replaceable global allocation function.

cfe/trunk/lib/CodeGen/CGNonTrivialStruct.cpp

getFunction(StringRef FuncName, QualType QT, std::array<Address, N> Addrs,
     ASTContext &Ctx = CGM.getContext();
     FunctionArgList Args;
     const CGFunctionInfo &FI = getFunctionInfo<N>(CGM, Args);
     llvm::FunctionType *FuncTy = CGM.getTypes().GetFunctionType(FI);
     llvm::Function *F =
         llvm::Function::Create(FuncTy, llvm::GlobalValue::LinkOnceODRLinkage,
                                FuncName, &CGM.getModule());
     F->setVisibility(llvm::GlobalValue::HiddenVisibility);
-    CGM.SetLLVMFunctionAttributes(nullptr, FI, F);
+    CGM.SetLLVMFunctionAttributes(GlobalDecl(), FI, F);
     CGM.SetLLVMFunctionAttributesForDefinition(nullptr, F);
     IdentifierInfo *II = &Ctx.Idents.get(FuncName);
     FunctionDecl *FD = FunctionDecl::Create(
         Ctx, Ctx.getTranslationUnitDecl(), SourceLocation(), SourceLocation(),
         II, Ctx.getFunctionType(Ctx.VoidTy, llvm::None, {}), nullptr,
         SC_PrivateExtern, false, false);
     CodeGenFunction NewCGF(CGM);
     setCGF(&NewCGF);

cfe/trunk/lib/CodeGen/CGVTables.cpp

if (CurFnInfo->usesInAlloca() || IsUnprototyped) {
     if (Thunk && !Thunk->Return.isEmpty()) {
       if (IsUnprototyped)
         CGM.ErrorUnsupported(
             MD, "return-adjusting thunk with incomplete parameter type");
       else
         CGM.ErrorUnsupported(
             MD, "non-trivial argument copy for return-adjusting thunk");
     }
-    EmitMustTailThunk(MD, AdjustedThisPtr, CalleePtr);
+    EmitMustTailThunk(CurGD, AdjustedThisPtr, CalleePtr);
     return;
   }
 
   // Start building CallArgs.
   CallArgList CallArgs;
   QualType ThisType = MD->getThisType(getContext());
   CallArgs.add(RValue::get(AdjustedThisPtr), ThisType);
 
#endif
   ReturnValueSlot Slot;
   if (!ResultType->isVoidType() &&
       CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::Indirect &&
       !hasScalarEvaluationKind(CurFnInfo->getReturnType()))
     Slot = ReturnValueSlot(ReturnValue, ResultType.isVolatileQualified());
 
   // Now emit our call.
   llvm::Instruction *CallOrInvoke;
-  CGCallee Callee = CGCallee::forDirect(CalleePtr, MD);
+  CGCallee Callee = CGCallee::forDirect(CalleePtr, CurGD);
   RValue RV = EmitCall(*CurFnInfo, Callee, Slot, CallArgs, &CallOrInvoke);
 
   // Consider return adjustment if we have ThunkInfo.
   if (Thunk && !Thunk->Return.isEmpty())
     RV = PerformReturnAdjustment(*this, ResultType, RV, *Thunk);
   else if (llvm::CallInst* Call = dyn_cast<llvm::CallInst>(CallOrInvoke))
     Call->setTailCallKind(llvm::CallInst::TCK_Tail);
 
   // Emit return.
   if (!ResultType->isVoidType() && Slot.isNull())
     CGM.getCXXABI().EmitReturnFromThunk(*this, RV, ResultType);
 
   // Disable the final ARC autorelease.
   AutoreleaseResult = false;
 
   FinishThunk();
 }
 
-void CodeGenFunction::EmitMustTailThunk(const CXXMethodDecl *MD,
+void CodeGenFunction::EmitMustTailThunk(GlobalDecl GD,
                                         llvm::Value *AdjustedThisPtr,
                                         llvm::Value *CalleePtr) {
   // Emitting a musttail call thunk doesn't use any of the CGCall.cpp machinery
   // to translate AST arguments into LLVM IR arguments.  For thunks, we know
   // that the caller prototype more or less matches the callee prototype with
   // the exception of 'this'.
   SmallVector<llvm::Value *, 8> Args;
   for (llvm::Argument &A : CurFn->args())
void CodeGenFunction::EmitMustTailThunk(GlobalDecl GD,
   // Emit the musttail call manually.  Even if the prologue pushed cleanups, we
   // don't actually want to run them.
   llvm::CallInst *Call = Builder.CreateCall(CalleePtr, Args);
   Call->setTailCallKind(llvm::CallInst::TCK_MustTail);
 
   // Apply the standard set of call attributes.
   unsigned CallingConv;
   llvm::AttributeList Attrs;
-  CGM.ConstructAttributeList(CalleePtr->getName(), *CurFnInfo, MD, Attrs,
+  CGM.ConstructAttributeList(CalleePtr->getName(), *CurFnInfo, GD, Attrs,
                              CallingConv, /*AttrOnCallSite=*/true);
   Call->setAttributes(Attrs);
   Call->setCallingConv(static_cast<llvm::CallingConv::ID>(CallingConv));
 
   if (Call->getType()->isVoidTy())
     Builder.CreateRetVoid();
   else
     Builder.CreateRet(Call);

cfe/trunk/lib/CodeGen/CodeGenFunction.h

void StartThunk(llvm::Function *Fn, GlobalDecl GD,
                   const CGFunctionInfo &FnInfo, bool IsUnprototyped);
 
   void EmitCallAndReturnForThunk(llvm::Constant *Callee, const ThunkInfo *Thunk,
                                  bool IsUnprototyped);
 
   void FinishThunk();
 
   /// Emit a musttail call for a thunk with a potentially adjusted this pointer.
-  void EmitMustTailThunk(const CXXMethodDecl *MD, llvm::Value *AdjustedThisPtr,
+  void EmitMustTailThunk(GlobalDecl GD, llvm::Value *AdjustedThisPtr,
                          llvm::Value *Callee);
 
   /// Generate a thunk for the given method.
   void generateThunk(llvm::Function *Fn, const CGFunctionInfo &FnInfo,
                      GlobalDecl GD, const ThunkInfo &Thunk,
                      bool IsUnprototyped);
 
   llvm::Function *GenerateVarArgsThunk(llvm::Function *Fn,
public:
   RValue EmitCXXPseudoDestructorExpr(const CXXPseudoDestructorExpr *E);
 
   RValue EmitCUDAKernelCallExpr(const CUDAKernelCallExpr *E,
                                 ReturnValueSlot ReturnValue);
 
   RValue EmitNVPTXDevicePrintfCallExpr(const CallExpr *E,
                                        ReturnValueSlot ReturnValue);
 
-  RValue EmitBuiltinExpr(const FunctionDecl *FD,
-                         unsigned BuiltinID, const CallExpr *E,
-                         ReturnValueSlot ReturnValue);
+  RValue EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
+                         const CallExpr *E, ReturnValueSlot ReturnValue);
 
   RValue emitRotate(const CallExpr *E, bool IsRotateRight);
 
   /// Emit IR for __builtin_os_log_format.
   RValue emitBuiltinOSLogFormat(const CallExpr &E);
 
   llvm::Function *generateBuiltinOSLogHelperFunction(
       const analyze_os_log::OSLogBufferLayout &Layout,

cfe/trunk/lib/CodeGen/CodeGenFunction.cpp

static bool hasRequiredFeatures(const SmallVectorImpl<StringRef> &ReqFeatures,
                                 std::string &FirstMissing) {
   // If there aren't any required features listed then go ahead and return.
   if (ReqFeatures.empty())
     return false;
 
   // Now build up the set of caller features and verify that all the required
   // features are there.
   llvm::StringMap<bool> CallerFeatureMap;
-  CGM.getFunctionFeatureMap(CallerFeatureMap, FD);
+  CGM.getFunctionFeatureMap(CallerFeatureMap, GlobalDecl().getWithDecl(FD));
 
   // If we have at least one of the features in the feature list return
   // true, otherwise return false.
   return std::all_of(
       ReqFeatures.begin(), ReqFeatures.end(), [&](StringRef Feature) {
         SmallVector<StringRef, 1> OrFeatures;
         Feature.split(OrFeatures, '|');
         return llvm::any_of(OrFeatures, [&](StringRef Feature) {

cfe/trunk/lib/CodeGen/CodeGenModule.h

public:
 
   /// Set the attributes on the LLVM function for the given decl and function
   /// info. This applies attributes necessary for handling the ABI as well as
   /// user specified attributes like section.
   void SetInternalFunctionAttributes(GlobalDecl GD, llvm::Function *F,
                                      const CGFunctionInfo &FI);
 
   /// Set the LLVM function attributes (sext, zext, etc).
-  void SetLLVMFunctionAttributes(const Decl *D,
-                                 const CGFunctionInfo &Info,
+  void SetLLVMFunctionAttributes(GlobalDecl GD, const CGFunctionInfo &Info,
                                  llvm::Function *F);
 
   /// Set the LLVM function attributes which only apply to a function
   /// definition.
   void SetLLVMFunctionAttributesForDefinition(const Decl *D, llvm::Function *F);
 
   /// Return true iff the given type uses 'sret' when used as a return type.
   bool ReturnTypeUsesSRet(const CGFunctionInfo &FI);
public:
   void AddDefaultFnAttrs(llvm::Function &F);
 
   /// Parses the target attributes passed in, and returns only the ones that are
   /// valid feature names.
   TargetAttr::ParsedTargetAttr filterFunctionTargetAttrs(const TargetAttr *TD);
 
   // Fills in the supplied string map with the set of target features for the
   // passed in function.
-  void getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
-                             const FunctionDecl *FD);
+  void getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap, GlobalDecl GD);
 
   StringRef getMangledName(GlobalDecl GD);
   StringRef getBlockMangledName(GlobalDecl GD, const BlockDecl *BD);
 
   void EmitTentativeDefinition(const VarDecl *D);
 
   void EmitVTable(CXXRecordDecl *Class);
 
private:
   void UpdateMultiVersionNames(GlobalDecl GD, const FunctionDecl *FD);
 
   llvm::Constant *GetOrCreateLLVMGlobal(StringRef MangledName,
                                         llvm::PointerType *PTy,
                                         const VarDecl *D,
                                         ForDefinition_t IsForDefinition
                                           = NotForDefinition);
 
-  bool GetCPUAndFeaturesAttributes(const Decl *D,
+  bool GetCPUAndFeaturesAttributes(GlobalDecl GD,
                                    llvm::AttrBuilder &AttrBuilder);
   void setNonAliasAttributes(GlobalDecl GD, llvm::GlobalObject *GO);
 
   /// Set function attributes for a function declaration.
   void SetFunctionAttributes(GlobalDecl GD, llvm::Function *F,
                              bool IsIncompleteFunction, bool IsThunk);
 
   void EmitGlobalDefinition(GlobalDecl D, llvm::GlobalValue *GV = nullptr);
 
   void EmitGlobalFunctionDefinition(GlobalDecl GD, llvm::GlobalValue *GV);
+  void EmitMultiVersionFunctionDefinition(GlobalDecl GD, llvm::GlobalValue *GV);
+
   void EmitGlobalVarDefinition(const VarDecl *D, bool IsTentative = false);
   void EmitAliasDefinition(GlobalDecl GD);
   void emitIFuncDefinition(GlobalDecl GD);
   void emitCPUDispatchDefinition(GlobalDecl GD);
   void EmitObjCPropertyImplementations(const ObjCImplementationDecl *D);
   void EmitObjCIvarInitializations(ObjCImplementationDecl *D);
 
   // C++ related functions.

cfe/trunk/lib/CodeGen/CodeGenModule.cpp

 static std::string getCPUSpecificMangling(const CodeGenModule &CGM,
                                           StringRef Name) {
   const TargetInfo &Target = CGM.getTarget();
   return (Twine('.') + Twine(Target.CPUSpecificManglingCharacter(Name))).str();
 }
 
 static void AppendCPUSpecificCPUDispatchMangling(const CodeGenModule &CGM,
                                                  const CPUSpecificAttr *Attr,
+                                                 unsigned CPUIndex,
                                                  raw_ostream &Out) {
   // cpu_specific gets the current name, dispatch gets the resolver if IFunc is
   // supported.
   if (Attr)
-    Out << getCPUSpecificMangling(CGM, Attr->getCurCPUName()->getName());
+    Out << getCPUSpecificMangling(CGM, Attr->getCPUName(CPUIndex)->getName());
   else if (CGM.getTarget().supportsIFunc())
     Out << ".resolver";
 }
 
 static void AppendTargetMangling(const CodeGenModule &CGM,
                                  const TargetAttr *Attr, raw_ostream &Out) {
   if (Attr->isDefaultVersion())
     return;
if (MC.shouldMangleDeclName(ND)) {
     } else {
       Out << II->getName();
     }
   }
 
   if (const auto *FD = dyn_cast<FunctionDecl>(ND))
     if (FD->isMultiVersion() && !OmitMultiVersionMangling) {
       if (FD->isCPUDispatchMultiVersion() || FD->isCPUSpecificMultiVersion())
-        AppendCPUSpecificCPUDispatchMangling(
-            CGM, FD->getAttr<CPUSpecificAttr>(), Out);
+        AppendCPUSpecificCPUDispatchMangling(CGM,
+                                             FD->getAttr<CPUSpecificAttr>(),
+                                             GD.getMultiVersionIndex(), Out);
       else
         AppendTargetMangling(CGM, FD->getAttr<TargetAttr>(), Out);
     }
 
   return Out.str();
 }
 
 void CodeGenModule::UpdateMultiVersionNames(GlobalDecl GD,
if (const auto *CD = dyn_cast<CXXConstructorDecl>(CanonicalGD.getDecl())) {
     if (!getTarget().getCXXABI().hasConstructorVariants()) {
       CXXCtorType OrigCtorType = GD.getCtorType();
       assert(OrigCtorType == Ctor_Base || OrigCtorType == Ctor_Complete);
       if (OrigCtorType == Ctor_Base)
         CanonicalGD = GlobalDecl(CD, Ctor_Complete);
     }
   }
 
-  const auto *FD = dyn_cast<FunctionDecl>(GD.getDecl());
-  // Since CPUSpecific can require multiple emits per decl, store the manglings
-  // separately.
-  if (FD &&
-      (FD->isCPUDispatchMultiVersion() || FD->isCPUSpecificMultiVersion())) {
-    const auto *SD = FD->getAttr<CPUSpecificAttr>();
-
-    std::pair<GlobalDecl, unsigned> SpecCanonicalGD{
-        CanonicalGD,
-        SD ? SD->ActiveArgIndex : std::numeric_limits<unsigned>::max()};
-
-    auto FoundName = CPUSpecificMangledDeclNames.find(SpecCanonicalGD);
-    if (FoundName != CPUSpecificMangledDeclNames.end())
-      return FoundName->second;
-
-    auto Result = CPUSpecificManglings.insert(
-        std::make_pair(getMangledNameImpl(*this, GD, FD), SpecCanonicalGD));
-    return CPUSpecificMangledDeclNames[SpecCanonicalGD] = Result.first->first();
-  }
-
   auto FoundName = MangledDeclNames.find(CanonicalGD);
   if (FoundName != MangledDeclNames.end())
     return FoundName->second;
 
   // Keep the first result in the case of a mangling collision.
   const auto *ND = cast<NamedDecl>(GD.getDecl());
   auto Result =
       Manglings.insert(std::make_pair(getMangledNameImpl(*this, GD, ND), GD));

 
 llvm::ConstantInt *CodeGenModule::CreateCrossDsoCfiTypeId(llvm::Metadata *MD) {
   llvm::MDString *MDS = dyn_cast<llvm::MDString>(MD);
   if (!MDS) return nullptr;
 
   return llvm::ConstantInt::get(Int64Ty, llvm::MD5Hash(MDS->getString()));
 }
 
-void CodeGenModule::SetLLVMFunctionAttributes(const Decl *D,
+void CodeGenModule::SetLLVMFunctionAttributes(GlobalDecl GD,
                                               const CGFunctionInfo &Info,
                                               llvm::Function *F) {
   unsigned CallingConv;
   llvm::AttributeList PAL;
-  ConstructAttributeList(F->getName(), Info, D, PAL, CallingConv, false);
+  ConstructAttributeList(F->getName(), Info, GD, PAL, CallingConv, false);
   F->setAttributes(PAL);
   F->setCallingConv(static_cast<llvm::CallingConv::ID>(CallingConv));
 }
 
 /// Determines whether the language options require us to model
 /// unwind exceptions.  We treat -fexceptions as mandating this
 /// except under the fragile ObjC ABI with only ObjC exceptions
 /// enabled.  This means, for example, that C with -fexceptions
void CodeGenModule::SetCommonAttributes(GlobalDecl GD, llvm::GlobalValue *GV) {
   if (CodeGenOpts.KeepStaticConsts && D && isa<VarDecl>(D)) {
     const auto *VD = cast<VarDecl>(D);
     if (VD->getType().isConstQualified() &&
         VD->getStorageDuration() == SD_Static)
       addUsedGlobal(GV);
   }
 }
 
-bool CodeGenModule::GetCPUAndFeaturesAttributes(const Decl *D,
+bool CodeGenModule::GetCPUAndFeaturesAttributes(GlobalDecl GD,
                                                 llvm::AttrBuilder &Attrs) {
   // Add target-cpu and target-features attributes to functions. If
   // we have a decl for the function and it has a target attribute then
   // parse that and add it to the feature set.
   StringRef TargetCPU = getTarget().getTargetOpts().CPU;
   std::vector<std::string> Features;
-  const auto *FD = dyn_cast_or_null<FunctionDecl>(D);
+  const auto *FD = dyn_cast_or_null<FunctionDecl>(GD.getDecl());
   FD = FD ? FD->getMostRecentDecl() : FD;
   const auto *TD = FD ? FD->getAttr<TargetAttr>() : nullptr;
   const auto *SD = FD ? FD->getAttr<CPUSpecificAttr>() : nullptr;
   bool AddedAttr = false;
   if (TD || SD) {
     llvm::StringMap<bool> FeatureMap;
-    getFunctionFeatureMap(FeatureMap, FD);
+    getFunctionFeatureMap(FeatureMap, GD);
 
     // Produce the canonical string for this set of features.
     for (const llvm::StringMap<bool>::value_type &Entry : FeatureMap)
       Features.push_back((Entry.getValue() ? "+" : "-") + Entry.getKey().str());
 
     // Now add the target-cpu and target-features to the function.
     // While we populated the feature map above, we still need to
     // get and parse the target attribute so we can get the cpu for
if (D) {
     }
 
     if (auto *F = dyn_cast<llvm::Function>(GO)) {
       if (auto *SA = D->getAttr<PragmaClangTextSectionAttr>())
         if (!D->getAttr<SectionAttr>())
           F->addFnAttr("implicit-section-name", SA->getName());
 
       llvm::AttrBuilder Attrs;
-      if (GetCPUAndFeaturesAttributes(D, Attrs)) {
+      if (GetCPUAndFeaturesAttributes(GD, Attrs)) {
         // We know that GetCPUAndFeaturesAttributes will always have the
         // newest set, since it has the newest possible FunctionDecl, so the
         // new ones should replace the old.
         F->removeFnAttr("target-cpu");
         F->removeFnAttr("target-features");
         F->addAttributes(llvm::AttributeList::FunctionIndex, Attrs);
       }
     }
 
     if (const auto *CSA = D->getAttr<CodeSegAttr>())
       GO->setSection(CSA->getName());
     else if (const auto *SA = D->getAttr<SectionAttr>())
       GO->setSection(SA->getName());
   }
 
   getTargetCodeGenInfo().setTargetAttributes(D, GO, *this);
 }
 
 void CodeGenModule::SetInternalFunctionAttributes(GlobalDecl GD,
                                                   llvm::Function *F,
                                                   const CGFunctionInfo &FI) {
   const Decl *D = GD.getDecl();
-  SetLLVMFunctionAttributes(D, FI, F);
+  SetLLVMFunctionAttributes(GD, FI, F);
   SetLLVMFunctionAttributesForDefinition(D, F);
 
   F->setLinkage(llvm::Function::InternalLinkage);
 
   setNonAliasAttributes(GD, F);
 }
 
 static void setLinkageForGV(llvm::GlobalValue *GV, const NamedDecl *ND) {
if (llvm::Intrinsic::ID IID = F->getIntrinsicID()) {
     // to the intrinsic's attributes.
     F->setAttributes(llvm::Intrinsic::getAttributes(getLLVMContext(), IID));
     return;
   }
 
   const auto *FD = cast<FunctionDecl>(GD.getDecl());
 
   if (!IsIncompleteFunction) {
-    SetLLVMFunctionAttributes(FD, getTypes().arrangeGlobalDeclaration(GD), F);
+    SetLLVMFunctionAttributes(GD, getTypes().arrangeGlobalDeclaration(GD), F);
     // Setup target-specific attributes.
     if (F->isDeclaration())
       getTargetCodeGenInfo().setTargetAttributes(FD, F, *this);
   }
 
   // Add the Returned attribute for "this", except for iOS 5 and earlier
   // where substantial code, including the libstdc++ dylib, was compiled with
   // GCC and does not actually return "this".
bool CodeGenModule::shouldEmitFunction(GlobalDecl GD) {
   // This happens in glibc's btowc and in some configure checks.
   return !isTriviallyRecursive(F);
 }
 
 bool CodeGenModule::shouldOpportunisticallyEmitVTables() {
   return CodeGenOpts.OptimizationLevel > 0;
 }
 
+void CodeGenModule::EmitMultiVersionFunctionDefinition(GlobalDecl GD,
+                                                       llvm::GlobalValue *GV) {
+  const auto *FD = cast<FunctionDecl>(GD.getDecl());
+
+  if (FD->isCPUSpecificMultiVersion()) {
+    auto *Spec = FD->getAttr<CPUSpecificAttr>();
+    for (unsigned I = 0; I < Spec->cpus_size(); ++I)
+      EmitGlobalFunctionDefinition(GD.getWithMultiVersionIndex(I), nullptr);
+    // Requires multiple emits.
+  } else
+    EmitGlobalFunctionDefinition(GD, GV);
+}
+
 void CodeGenModule::EmitGlobalDefinition(GlobalDecl GD, llvm::GlobalValue *GV) {
   const auto *D = cast<ValueDecl>(GD.getDecl());
 
   PrettyStackTraceDecl CrashInfo(const_cast<ValueDecl *>(D), D->getLocation(),
                                  Context.getSourceManager(),
                                  "Generating code for declaration");
 
-  if (isa<FunctionDecl>(D)) {
+  if (const auto *FD = dyn_cast<FunctionDecl>(D)) {
     // At -O0, don't generate IR for functions with available_externally
     // linkage.
     if (!shouldEmitFunction(GD))
       return;
 
     if (const auto *Method = dyn_cast<CXXMethodDecl>(D)) {
       // Make sure to emit the definition(s) before we emit the thunks.
       // This is necessary for the generation of certain thunks.
       if (const auto *CD = dyn_cast<CXXConstructorDecl>(Method))
         ABI->emitCXXStructor(CD, getFromCtorType(GD.getCtorType()));
       else if (const auto *DD = dyn_cast<CXXDestructorDecl>(Method))
         ABI->emitCXXStructor(DD, getFromDtorType(GD.getDtorType()));
+      else if (FD->isMultiVersion())
+        EmitMultiVersionFunctionDefinition(GD, GV);
       else
         EmitGlobalFunctionDefinition(GD, GV);
 
       if (Method->isVirtual())
         getVTables().EmitThunks(GD);
 
       return;
     }
 
+    if (FD->isMultiVersion())
+      return EmitMultiVersionFunctionDefinition(GD, GV);
     return EmitGlobalFunctionDefinition(GD, GV);
   }
 
   if (const auto *VD = dyn_cast<VarDecl>(D))
     return EmitGlobalVarDefinition(VD, !VD->hasDefinition());
 
   llvm_unreachable("Invalid argument to EmitGlobalDefinition()");
 }
else {
     ResolverGD = GD;
   }
 
   auto *ResolverFunc = cast<llvm::Function>(GetOrCreateLLVMFunction(
       ResolverName, ResolverType, ResolverGD, /*ForVTable=*/false));
 
   SmallVector<CodeGenFunction::MultiVersionResolverOption, 10> Options;
   const TargetInfo &Target = getTarget();
+  unsigned Index = 0;
   for (const IdentifierInfo *II : DD->cpus()) {
     // Get the name of the target function so we can look it up/create it.
     std::string MangledName = getMangledNameImpl(*this, GD, FD, true) +
                               getCPUSpecificMangling(*this, II->getName());
-    llvm::Constant *Func = GetOrCreateLLVMFunction(
-        MangledName, DeclTy, GD, /*ForVTable=*/false, /*DontDefer=*/true,
+
+    llvm::Constant *Func = GetGlobalValue(MangledName);
+
+    if (!Func)
+      Func = GetOrCreateLLVMFunction(
+          MangledName, DeclTy, GD.getWithMultiVersionIndex(Index),
+          /*ForVTable=*/false, /*DontDefer=*/true,
           /*IsThunk=*/false, llvm::AttributeList(), ForDefinition);
+
     llvm::SmallVector<StringRef, 32> Features;
     Target.getCPUSpecificCPUDispatchFeatures(II->getName(), Features);
     llvm::transform(Features, Features.begin(),
                     [](StringRef Str) { return Str.substr(1); });
     Features.erase(std::remove_if(
         Features.begin(), Features.end(), [&Target](StringRef Feat) {
           return !Target.validateCpuSupports(Feat);
         }), Features.end());
     Options.emplace_back(cast<llvm::Function>(Func), StringRef{}, Features);
+    ++Index;
   }
 
   llvm::sort(
       Options, [](const CodeGenFunction::MultiVersionResolverOption &LHS,
                   const CodeGenFunction::MultiVersionResolverOption &RHS) {
         return CodeGenFunction::GetX86CpuSupportsMask(LHS.Conditions.Features) >
                CodeGenFunction::GetX86CpuSupportsMask(RHS.Conditions.Features);
       });
void CodeGenModule::EmitGlobalFunctionDefinition(GlobalDecl GD,
   SetLLVMFunctionAttributesForDefinition(D, Fn);
 
   if (const ConstructorAttr *CA = D->getAttr<ConstructorAttr>())
     AddGlobalCtor(Fn, CA->getPriority());
   if (const DestructorAttr *DA = D->getAttr<DestructorAttr>())
     AddGlobalDtor(Fn, DA->getPriority());
   if (D->hasAttr<AnnotateAttr>())
     AddGlobalAnnotations(D, Fn);
-
-  if (D->isCPUSpecificMultiVersion()) {
-    auto *Spec = D->getAttr<CPUSpecificAttr>();
-    // If there is another specific version we need to emit, do so here.
-    if (Spec->ActiveArgIndex + 1 < Spec->cpus_size()) {
-      ++Spec->ActiveArgIndex;
-      EmitGlobalFunctionDefinition(GD, nullptr);
-    }
-  }
 }
 
 void CodeGenModule::EmitAliasDefinition(GlobalDecl GD) {
   const auto *D = cast<ValueDecl>(GD.getDecl());
   const AliasAttr *AA = D->getAttr<AliasAttr>();
   assert(AA && "Not an alias?");
 
   StringRef MangledName = getMangledName(GD);
ParsedAttr.Features.erase(
       ParsedAttr.Features.end());
   return ParsedAttr;
 }
 
 
 // Fills in the supplied string map with the set of target features for the
 // passed in function.
 void CodeGenModule::getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
-                                          const FunctionDecl *FD) {
+                                          GlobalDecl GD) {
   StringRef TargetCPU = Target.getTargetOpts().CPU;
+  const FunctionDecl *FD = GD.getDecl()->getAsFunction();
   if (const auto *TD = FD->getAttr<TargetAttr>()) {
     TargetAttr::ParsedTargetAttr ParsedAttr = filterFunctionTargetAttrs(TD);
 
     // Make a copy of the features as passed on the command line into the
     // beginning of the additional features from the function to override.
     ParsedAttr.Features.insert(ParsedAttr.Features.begin(),
                             Target.getTargetOpts().FeaturesAsWritten.begin(),
                             Target.getTargetOpts().FeaturesAsWritten.end());
 
     if (ParsedAttr.Architecture != "" &&
         Target.isValidCPUName(ParsedAttr.Architecture))
       TargetCPU = ParsedAttr.Architecture;
 
     // Now populate the feature map, first with the TargetCPU which is either
     // the default or a new one from the target attribute string. Then we'll use
     // the passed in features (FeaturesAsWritten) along with the new ones from
     // the attribute.
     Target.initFeatureMap(FeatureMap, getDiags(), TargetCPU,
                           ParsedAttr.Features);
   } else if (const auto *SD = FD->getAttr<CPUSpecificAttr>()) {
     llvm::SmallVector<StringRef, 32> FeaturesTmp;
-    Target.getCPUSpecificCPUDispatchFeatures(SD->getCurCPUName()->getName(),
-                                             FeaturesTmp);
+    Target.getCPUSpecificCPUDispatchFeatures(
+        SD->getCPUName(GD.getMultiVersionIndex())->getName(), FeaturesTmp);
     std::vector<std::string> Features(FeaturesTmp.begin(), FeaturesTmp.end());
     Target.initFeatureMap(FeatureMap, getDiags(), TargetCPU, Features);
   } else {
     Target.initFeatureMap(FeatureMap, getDiags(), TargetCPU,
                           Target.getTargetOpts().Features);
   }
 }
 

cfe/trunk/lib/CodeGen/ItaniumCXXABI.cpp

void ItaniumCXXABI::EmitDestructorCall(CodeGenFunction &CGF,
   llvm::Value *VTT = CGF.GetVTTParameter(GD, ForVirtualBase, Delegating);
   QualType VTTTy = getContext().getPointerType(getContext().VoidPtrTy);
 
   CGCallee Callee;
   if (getContext().getLangOpts().AppleKext &&
       Type != Dtor_Base && DD->isVirtual())
     Callee = CGF.BuildAppleKextVirtualDestructorCall(DD, Type, DD->getParent());
   else
-    Callee =
-      CGCallee::forDirect(CGM.getAddrOfCXXStructor(DD, getFromDtorType(Type)),
-                          DD);
+    Callee = CGCallee::forDirect(
+        CGM.getAddrOfCXXStructor(DD, getFromDtorType(Type)), GD);
 
   CGF.EmitCXXMemberOrOperatorCall(DD, Callee, ReturnValueSlot(),
                                   This.getPointer(), VTT, VTTTy,
                                   nullptr, nullptr);
 }
 
 void ItaniumCXXABI::emitVTableDefinitions(CodeGenVTables &CGVT,
                                           const CXXRecordDecl *RD) {
if (CGM.getCodeGenOpts().OptimizationLevel > 0 &&
         CGM.getCodeGenOpts().StrictVTablePointers)
       VFuncLoad->setMetadata(
           llvm::LLVMContext::MD_invariant_load,
           llvm::MDNode::get(CGM.getLLVMContext(),
                             llvm::ArrayRef<llvm::Metadata *>()));
     VFunc = VFuncLoad;
   }
 
-  CGCallee Callee(MethodDecl->getCanonicalDecl(), VFunc);
+  CGCallee Callee(GD, VFunc);
   return Callee;
 }
 
 llvm::Value *ItaniumCXXABI::EmitVirtualDestructorCall(
     CodeGenFunction &CGF, const CXXDestructorDecl *Dtor, CXXDtorType DtorType,
     Address This, const CXXMemberCallExpr *CE) {
   assert(CE == nullptr || CE->arg_begin() == CE->arg_end());
   assert(DtorType == Dtor_Deleting || DtorType == Dtor_Complete);
ItaniumCXXABI::getOrCreateThreadLocalWrapper(const VarDecl *VD,
   const CGFunctionInfo &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
       getContext().getPointerType(RetQT), FunctionArgList());
 
   llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FI);
   llvm::Function *Wrapper =
       llvm::Function::Create(FnTy, getThreadLocalWrapperLinkage(VD, CGM),
                              WrapperName.str(), &CGM.getModule());
 
-  CGM.SetLLVMFunctionAttributes(nullptr, FI, Wrapper);
+  CGM.SetLLVMFunctionAttributes(GlobalDecl(), FI, Wrapper);
 
   if (VD->hasDefinition())
     CGM.SetLLVMFunctionAttributesForDefinition(nullptr, Wrapper);
 
   // Always resolve references to the wrapper at link time.
   if (!Wrapper->hasLocalLinkage() && !(isThreadWrapperReplaceable(VD, CGM) &&
       !llvm::GlobalVariable::isLinkOnceLinkage(Wrapper->getLinkage()) &&
       !llvm::GlobalVariable::isWeakODRLinkage(Wrapper->getLinkage())))
if (VD->hasDefinition()) {
       // This function will not exist if the TU defining the thread_local
       // variable in question does not need any dynamic initialization for
       // its thread_local variables.
       llvm::FunctionType *FnTy = llvm::FunctionType::get(CGM.VoidTy, false);
       Init = llvm::Function::Create(FnTy,
                                     llvm::GlobalVariable::ExternalWeakLinkage,
                                     InitFnName.str(), &CGM.getModule());
       const CGFunctionInfo &FI = CGM.getTypes().arrangeNullaryFunction();
-      CGM.SetLLVMFunctionAttributes(nullptr, FI, cast<llvm::Function>(Init));
+      CGM.SetLLVMFunctionAttributes(GlobalDecl(), FI,
+                                    cast<llvm::Function>(Init));
     }
 
     if (Init) {
       Init->setVisibility(Var->getVisibility());
       Init->setDSOLocal(Var->isDSOLocal());
     }
 
     llvm::LLVMContext &Context = CGM.getModule().getContext();

cfe/trunk/lib/CodeGen/MicrosoftCXXABI.cpp

void MicrosoftCXXABI::EmitDestructorCall(CodeGenFunction &CGF,
                                          CXXDtorType Type, bool ForVirtualBase,
                                          bool Delegating, Address This) {
   // Use the base destructor variant in place of the complete destructor variant
   // if the class has no virtual bases. This effectively implements some of the
   // -mconstructor-aliases optimization, but as part of the MS C++ ABI.
   if (Type == Dtor_Complete && DD->getParent()->getNumVBases() == 0)
     Type = Dtor_Base;
 
-  CGCallee Callee = CGCallee::forDirect(
-                          CGM.getAddrOfCXXStructor(DD, getFromDtorType(Type)),
-                                        DD);
+  CGCallee Callee =
+      CGCallee::forDirect(CGM.getAddrOfCXXStructor(DD, getFromDtorType(Type)),
+                          GlobalDecl(DD, Type));
 
   if (DD->isVirtual()) {
     assert(Type != CXXDtorType::Dtor_Deleting &&
            "The deleting destructor should only be called via a virtual call");
     This = adjustThisArgumentForVirtualFunctionCall(CGF, GlobalDecl(DD, Type),
                                                     This, false);
   }
 
if (CGF.ShouldEmitVTableTypeCheckedLoad(MethodDecl->getParent())) {
     if (CGM.getCodeGenOpts().PrepareForLTO)
       CGF.EmitTypeMetadataCodeForVCall(getObjectWithVPtr(), VTable, Loc);
 
     llvm::Value *VFuncPtr =
         Builder.CreateConstInBoundsGEP1_64(VTable, ML.Index, "vfn");
     VFunc = Builder.CreateAlignedLoad(VFuncPtr, CGF.getPointerAlign());
   }
 
-  CGCallee Callee(MethodDecl->getCanonicalDecl(), VFunc);
+  CGCallee Callee(GD, VFunc);
   return Callee;
 }
 
 llvm::Value *MicrosoftCXXABI::EmitVirtualDestructorCall(
     CodeGenFunction &CGF, const CXXDestructorDecl *Dtor, CXXDtorType DtorType,
     Address This, const CXXMemberCallExpr *CE) {
   assert(CE == nullptr || CE->arg_begin() == CE->arg_end());
   assert(DtorType == Dtor_Deleting || DtorType == Dtor_Complete);
MicrosoftCXXABI::getAddrOfCXXCtorClosure(const CXXConstructorDecl *CD,
   // Insert any ABI-specific implicit constructor arguments.
   AddedStructorArgs ExtraArgs =
       addImplicitConstructorArgs(CGF, CD, Ctor_Complete,
                                  /*ForVirtualBase=*/false,
                                  /*Delegating=*/false, Args);
   // Call the destructor with our arguments.
   llvm::Constant *CalleePtr =
     CGM.getAddrOfCXXStructor(CD, StructorType::Complete);
-  CGCallee Callee = CGCallee::forDirect(CalleePtr, CD);
+  CGCallee Callee =
+      CGCallee::forDirect(CalleePtr, GlobalDecl(CD, Ctor_Complete));
   const CGFunctionInfo &CalleeInfo = CGM.getTypes().arrangeCXXConstructorCall(
       Args, CD, Ctor_Complete, ExtraArgs.Prefix, ExtraArgs.Suffix);
   CGF.EmitCall(CalleeInfo, Callee, ReturnValueSlot(), Args);
 
   Cleanups.ForceCleanup();
 
   // Emit the ret instruction, remove any temporary instructions created for the
   // aid of CodeGen.

cfe/trunk/lib/Parse/ParseDeclCXX.cpp

 /// there was an inline keyword, it has already been parsed.
 ///
 ///       namespace-definition: [C++: namespace.def]
 ///         named-namespace-definition
 ///         unnamed-namespace-definition
 ///         nested-namespace-definition
 ///
 ///       named-namespace-definition:
-///         'inline'[opt] 'namespace' attributes[opt] identifier '{' namespace-body '}'
+///         'inline'[opt] 'namespace' attributes[opt] identifier '{'
+///         namespace-body '}'
 ///
 ///       unnamed-namespace-definition:
 ///         'inline'[opt] 'namespace' attributes[opt] '{' namespace-body '}'
 ///
 ///       nested-namespace-definition:
-///         'namespace' enclosing-namespace-specifier '::' 'inline'[opt] identifier '{' namespace-body '}'
+///         'namespace' enclosing-namespace-specifier '::' 'inline'[opt]
+///         identifier '{' namespace-body '}'
 ///
 ///       enclosing-namespace-specifier:
 ///         identifier
 ///         enclosing-namespace-specifier '::' 'inline'[opt] identifier
 ///
 ///       namespace-alias-definition:  [C++ 7.3.2: namespace.alias]
 ///         'namespace' identifier '=' qualified-namespace-specifier ';'
 ///
void Parser::ParseInnerNamespace(const InnerNamespaceInfoList &InnerNSs,
   UsingDirectiveDecl *ImplicitUsingDirectiveDecl = nullptr;
   Decl *NamespcDecl = Actions.ActOnStartNamespaceDef(
       getCurScope(), InnerNSs[index].InlineLoc, InnerNSs[index].NamespaceLoc,
       InnerNSs[index].IdentLoc, InnerNSs[index].Ident,
       Tracker.getOpenLocation(), attrs, ImplicitUsingDirectiveDecl);
   assert(!ImplicitUsingDirectiveDecl &&
          "nested namespace definition cannot define anonymous namespace");
 
-  ParseInnerNamespace(InnerNSs, ++index, InlineLoc,
-                      attrs, Tracker);
+  ParseInnerNamespace(InnerNSs, ++index, InlineLoc, attrs, Tracker);
 
   NamespaceScope.Exit();
   Actions.ActOnFinishNamespaceDef(NamespcDecl, Tracker.getCloseLocation());
 }
 
 /// ParseNamespaceAlias - Parse the part after the '=' in a namespace
 /// alias definition.
 ///

cfe/trunk/test/CodeGen/attr-cpuspecific.c

 // WINDOWS: call void @__cpu_indicator_init
 // WINDOWS: %[[RET:.+]] = musttail call i32 @GenericAndPentium.O(i32 %0, double %1)
 // WINDOWS-NEXT: ret i32 %[[RET]]
 // WINDOWS: %[[RET:.+]] = musttail call i32 @GenericAndPentium.B(i32 %0, double %1)
 // WINDOWS-NEXT: ret i32 %[[RET]]
 // WINDOWS-NOT: call i32 @GenericAndPentium.A
 // WINDOWS-NOT: call void @llvm.trap
 
+ATTR(cpu_dispatch(atom, pentium))
+int DispatchFirst(void);
+// LINUX: define i32 ()* @DispatchFirst.resolver
+// LINUX: ret i32 ()* @DispatchFirst.O
+// LINUX: ret i32 ()* @DispatchFirst.B
+
+// WINDOWS: define dso_local i32 @DispatchFirst()
+// WINDOWS: %[[RET:.+]] = musttail call i32 @DispatchFirst.O()
+// WINDOWS-NEXT: ret i32 %[[RET]]
+// WINDOWS: %[[RET:.+]] = musttail call i32 @DispatchFirst.B()
+// WINDOWS-NEXT: ret i32 %[[RET]]
+
+ATTR(cpu_specific(atom))
+int DispatchFirst(void) {return 0;}
+// LINUX: define i32 @DispatchFirst.O
+// LINUX: ret i32 0
+
+// WINDOWS: define dso_local i32 @DispatchFirst.O()
+// WINDOWS: ret i32 0
+
+ATTR(cpu_specific(pentium))
+int DispatchFirst(void) {return 1;}
+// LINUX: define i32 @DispatchFirst.B
+// LINUX: ret i32 1
+
+// WINDOWS: define dso_local i32 @DispatchFirst.B
+// WINDOWS: ret i32 1
+
 // CHECK: attributes #[[S]] = {{.*}}"target-features"="+avx,+cmov,+f16c,+mmx,+popcnt,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87,+xsave"
 // CHECK: attributes #[[K]] = {{.*}}"target-features"="+adx,+avx,+avx2,+avx512cd,+avx512er,+avx512f,+avx512pf,+bmi,+cmov,+f16c,+fma,+lzcnt,+mmx,+movbe,+popcnt,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87,+xsave"
 // CHECK: attributes #[[O]] = {{.*}}"target-features"="+cmov,+mmx,+movbe,+sse,+sse2,+sse3,+ssse3,+x87"