MS ABI: Implement thread_local for global variables

[majnemer]

Summary

This add support for the C++11 feature, thread_local global variables.
The ABI Clang implements is an improvement of the MSVC ABI. Sadly,
further improvements could be made but not without sacrificing ABI
compatibility.

The feature is implemented as follows:

• All thread_local initialization routines are pointed to from the .CRT$XDU section.
• All non-weak thread_local variables have their initialization routines call from a single function instead of getting their own .CRT$XDU section entry. This is done to open up optimization opportunities to the compiler.
• All weak thread_local variables have their own .CRT$XDU section entry. This entry is in a COMDAT with the global variable it is initializing; this ensures that we will initialize the global exactly once.
• Destructors are registered in the initialization function using __tlregdtor.

[rnk]

lib/CodeGen/CGCXXABI.h

449–450

This method appears dead now.

490–491

Make this pure and delete the generic CGCXXABI implementation now?

497–498

Can you update or delete the doxygen? IIRC there are clang warnings that detect parameter name mismatches like this.

lib/CodeGen/ItaniumCXXABI.cpp

237–238

dead

lib/CodeGen/MicrosoftCXXABI.cpp

274

dead

1778

"Nondiscardable" seems like the wrong name. Perhaps "NonComdatInits" or "TUOrderedInits"?

1799–1801

One possible alternative is to build a single global in the .CRT$XDU section that is an ordered array of all the thread local initializers that need to be run. I guess you considered and rejected this, because the single function allows for more optimization?

test/CodeGenCXX/ms-thread_local.cpp

22

Maybe add some code CHECKs like:

// CHECK-LABEL: define internal void @__tls_init()
// CHECK: call void @"\01??__E?b@@3UA@@A"()

[majnemer]

• Address review comments.

[majnemer]

• Fix a bug involving references.

[rnk]

lgtm

[majnemer]

Closed by commit rL219074 (authored by @majnemer).

lib/CodeGen/CGCXXABI.h

public:
   virtual void ReadArrayCookie(CodeGenFunction &CGF, llvm::Value *Ptr,
                                const CXXDeleteExpr *expr,
                                QualType ElementType, llvm::Value *&NumElements,
                                llvm::Value *&AllocPtr, CharUnits &CookieSize);
 
   /// Return whether the given global decl needs a VTT parameter.
   virtual bool NeedsVTTParameter(GlobalDecl GD);
 
 protected:
   /// Returns the extra size required in order to store the array
   /// cookie for the given type.  Assumes that an array cookie is
   /// required.
   virtual CharUnits getArrayCookieSizeImpl(QualType elementType);
 
   /// Reads the array cookie for an allocation which is known to have one.
   /// This is called by the standard implementation of ReadArrayCookie.
   ///
   /// \param ptr - a pointer to the allocation made for an array, as a char*
public:
   ///   - a static data member of a class template instantiation
   virtual void EmitGuardedInit(CodeGenFunction &CGF, const VarDecl &D,
                                llvm::GlobalVariable *DeclPtr,
                                bool PerformInit) = 0;
 
   /// Emit code to force the execution of a destructor during global
   /// teardown.  The default implementation of this uses atexit.
   ///
-  /// \param dtor - a function taking a single pointer argument
-  /// \param addr - a pointer to pass to the destructor function.
+  /// \param Dtor - a function taking a single pointer argument
+  /// \param Addr - a pointer to pass to the destructor function.
   virtual void registerGlobalDtor(CodeGenFunction &CGF, const VarDecl &D,
-                                  llvm::Constant *dtor, llvm::Constant *addr);
+                                  llvm::Constant *Dtor,
+                                  llvm::Constant *Addr) = 0;
 
   /*************************** thread_local initialization ********************/
 
   /// Emits ABI-required functions necessary to initialize thread_local
   /// variables in this translation unit.
   ///
-  /// \param Decls The thread_local declarations in this translation unit.
-  /// \param InitFunc If this translation unit contains any non-constant
-  ///        initialization or non-trivial destruction for thread_local
-  ///        variables, a function to perform the initialization. Otherwise, 0.
+  /// \param CXXThreadLocals - The thread_local declarations in this translation
+  ///        unit.
+  /// \param CXXThreadLocalInits - If this translation unit contains any
+  ///        non-constant initialization or non-trivial destruction for
+  ///        thread_local variables, a list of functions to perform the
+  ///        initialization.
   virtual void EmitThreadLocalInitFuncs(
-      ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *> > Decls,
-      llvm::Function *InitFunc);
+      CodeGenModule &CGM,
+      ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *>>
+          CXXThreadLocals,
+      ArrayRef<llvm::Function *> CXXThreadLocalInits,
+      ArrayRef<llvm::GlobalVariable *> CXXThreadLocalInitVars) = 0;
 
   /// Emit a reference to a non-local thread_local variable (including
   /// triggering the initialization of all thread_local variables in its
   /// translation unit).
   virtual LValue EmitThreadLocalVarDeclLValue(CodeGenFunction &CGF,
                                               const VarDecl *VD,
-                                              QualType LValType);
+                                              QualType LValType) = 0;
 
   /// Emit a single constructor/destructor with the given type from a C++
   /// constructor Decl.
   virtual void emitCXXStructor(const CXXMethodDecl *MD, StructorType Type) = 0;
 };
 
 // Create an instance of a C++ ABI class:
 

lib/CodeGen/CGCXXABI.cpp

 
 llvm::Value *CGCXXABI::readArrayCookieImpl(CodeGenFunction &CGF,
                                            llvm::Value *ptr,
                                            CharUnits cookieSize) {
   ErrorUnsupportedABI(CGF, "reading a new[] cookie");
   return llvm::ConstantInt::get(CGF.SizeTy, 0);
 }
 
-void CGCXXABI::registerGlobalDtor(CodeGenFunction &CGF,
-                                  const VarDecl &D,
-                                  llvm::Constant *dtor,
-                                  llvm::Constant *addr) {
-  if (D.getTLSKind())
-    CGM.ErrorUnsupported(&D, "non-trivial TLS destruction");
-
-  // The default behavior is to use atexit.
-  CGF.registerGlobalDtorWithAtExit(D, dtor, addr);
-}
-
 /// Returns the adjustment, in bytes, required for the given
 /// member-pointer operation.  Returns null if no adjustment is
 /// required.
 llvm::Constant *CGCXXABI::getMemberPointerAdjustment(const CastExpr *E) {
   assert(E->getCastKind() == CK_DerivedToBaseMemberPointer ||
          E->getCastKind() == CK_BaseToDerivedMemberPointer);
 
   QualType derivedType;
CGCXXABI::EmitCtorCompleteObjectHandler(CodeGenFunction &CGF,
                                         const CXXRecordDecl *RD) {
   if (CGM.getTarget().getCXXABI().hasConstructorVariants())
     llvm_unreachable("shouldn't be called in this ABI");
 
   ErrorUnsupportedABI(CGF, "complete object detection in ctor");
   return nullptr;
 }
 
-void CGCXXABI::EmitThreadLocalInitFuncs(
-    ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *> > Decls,
-    llvm::Function *InitFunc) {
-}
-
-LValue CGCXXABI::EmitThreadLocalVarDeclLValue(CodeGenFunction &CGF,
-                                              const VarDecl *VD,
-                                              QualType LValType) {
-  ErrorUnsupportedABI(CGF, "odr-use of thread_local global");
-  return LValue();
-}
-
 bool CGCXXABI::NeedsVTTParameter(GlobalDecl GD) {
   return false;
 }

lib/CodeGen/CGDeclCXX.cpp

void CodeGenFunction::EmitCXXGlobalVarDeclInit(const VarDecl &D,
 
   assert(PerformInit && "cannot have constant initializer which needs "
          "destruction for reference");
   unsigned Alignment = getContext().getDeclAlign(&D).getQuantity();
   RValue RV = EmitReferenceBindingToExpr(Init);
   EmitStoreOfScalar(RV.getScalarVal(), DeclPtr, false, Alignment, T);
 }
 
-static llvm::Function *
-CreateGlobalInitOrDestructFunction(CodeGenModule &CGM,
-                                   llvm::FunctionType *ty,
-                                   const Twine &name,
-                                   bool TLS = false);
-
 /// Create a stub function, suitable for being passed to atexit,
 /// which passes the given address to the given destructor function.
-static llvm::Constant *createAtExitStub(CodeGenModule &CGM, const VarDecl &VD,
+llvm::Constant *CodeGenFunction::createAtExitStub(const VarDecl &VD,
                                                   llvm::Constant *dtor,
                                                   llvm::Constant *addr) {
   // Get the destructor function type, void(*)(void).
   llvm::FunctionType *ty = llvm::FunctionType::get(CGM.VoidTy, false);
   SmallString<256> FnName;
   {
     llvm::raw_svector_ostream Out(FnName);
     CGM.getCXXABI().getMangleContext().mangleDynamicAtExitDestructor(&VD, Out);
   }
-  llvm::Function *fn =
-      CreateGlobalInitOrDestructFunction(CGM, ty, FnName.str());
+  llvm::Function *fn = CGM.CreateGlobalInitOrDestructFunction(ty, FnName.str());
 
   CodeGenFunction CGF(CGM);
 
   CGF.StartFunction(&VD, CGM.getContext().VoidTy, fn,
                     CGM.getTypes().arrangeNullaryFunction(), FunctionArgList());
 
   llvm::CallInst *call = CGF.Builder.CreateCall(dtor, addr);
  
  // Make sure the call and the callee agree on calling convention.
   if (llvm::Function *dtorFn =
         dyn_cast<llvm::Function>(dtor->stripPointerCasts()))
     call->setCallingConv(dtorFn->getCallingConv());
 
   CGF.FinishFunction();
 
   return fn;
 }
 
 /// Register a global destructor using the C atexit runtime function.
 void CodeGenFunction::registerGlobalDtorWithAtExit(const VarDecl &VD,
                                                    llvm::Constant *dtor,
                                                    llvm::Constant *addr) {
   // Create a function which calls the destructor.
-  llvm::Constant *dtorStub = createAtExitStub(CGM, VD, dtor, addr);
+  llvm::Constant *dtorStub = createAtExitStub(VD, dtor, addr);
 
   // extern "C" int atexit(void (*f)(void));
   llvm::FunctionType *atexitTy =
     llvm::FunctionType::get(IntTy, dtorStub->getType(), false);
 
   llvm::Constant *atexit =
     CGM.CreateRuntimeFunction(atexitTy, "atexit");
   if (llvm::Function *atexitFn = dyn_cast<llvm::Function>(atexit))
void CodeGenFunction::EmitCXXGuardedInit(const VarDecl &D,
   if (CGM.getCodeGenOpts().ForbidGuardVariables)
     CGM.Error(D.getLocation(),
               "this initialization requires a guard variable, which "
               "the kernel does not support");
 
   CGM.getCXXABI().EmitGuardedInit(*this, D, DeclPtr, PerformInit);
 }
 
-static llvm::Function *
-CreateGlobalInitOrDestructFunction(CodeGenModule &CGM,
-                                   llvm::FunctionType *FTy,
+llvm::Function *
+CodeGenModule::CreateGlobalInitOrDestructFunction(llvm::FunctionType *FTy,
                                                   const Twine &Name, bool TLS) {
   llvm::Function *Fn =
     llvm::Function::Create(FTy, llvm::GlobalValue::InternalLinkage,
-                           Name, &CGM.getModule());
-  if (!CGM.getLangOpts().AppleKext && !TLS) {
+                           Name, &getModule());
+  if (!getLangOpts().AppleKext && !TLS) {
     // Set the section if needed.
-    if (const char *Section = 
-          CGM.getTarget().getStaticInitSectionSpecifier())
+    if (const char *Section = getTarget().getStaticInitSectionSpecifier())
       Fn->setSection(Section);
   }
 
-  Fn->setCallingConv(CGM.getRuntimeCC());
+  Fn->setCallingConv(getRuntimeCC());
 
-  if (!CGM.getLangOpts().Exceptions)
+  if (!getLangOpts().Exceptions)
     Fn->setDoesNotThrow();
 
-  if (!CGM.getSanitizerBlacklist().isIn(*Fn)) {
-    if (CGM.getLangOpts().Sanitize.Address)
+  if (!getSanitizerBlacklist().isIn(*Fn)) {
+    if (getLangOpts().Sanitize.Address)
       Fn->addFnAttr(llvm::Attribute::SanitizeAddress);
-    if (CGM.getLangOpts().Sanitize.Thread)
+    if (getLangOpts().Sanitize.Thread)
       Fn->addFnAttr(llvm::Attribute::SanitizeThread);
-    if (CGM.getLangOpts().Sanitize.Memory)
+    if (getLangOpts().Sanitize.Memory)
       Fn->addFnAttr(llvm::Attribute::SanitizeMemory);
   }
 
   return Fn;
 }
 
 /// Create a global pointer to a function that will initialize a global
 /// variable.  The user has requested that this pointer be emitted in a specific
CodeGenModule::EmitCXXGlobalVarDeclInitFunc(const VarDecl *D,
   llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false);
   SmallString<256> FnName;
   {
     llvm::raw_svector_ostream Out(FnName);
     getCXXABI().getMangleContext().mangleDynamicInitializer(D, Out);
   }
 
   // Create a variable initialization function.
-  llvm::Function *Fn =
-      CreateGlobalInitOrDestructFunction(*this, FTy, FnName.str());
+  llvm::Function *Fn = CreateGlobalInitOrDestructFunction(FTy, FnName.str());
 
   auto *ISA = D->getAttr<InitSegAttr>();
   CodeGenFunction(*this).GenerateCXXGlobalVarDeclInitFunc(Fn, D, Addr,
                                                           PerformInit);
 
   llvm::GlobalVariable *Key = supportsCOMDAT() ? Addr : nullptr;
 
   if (D->getTLSKind()) {
     // FIXME: Should we support init_priority for thread_local?
     // FIXME: Ideally, initialization of instantiated thread_local static data
     // members of class templates should not trigger initialization of other
     // entities in the TU.
     // FIXME: We only need to register one __cxa_thread_atexit function for the
     // entire TU.
     CXXThreadLocalInits.push_back(Fn);
+    CXXThreadLocalInitVars.push_back(Addr);
   } else if (PerformInit && ISA) {
     EmitPointerToInitFunc(D, Addr, Fn, ISA);
     DelayedCXXInitPosition.erase(D);
   } else if (auto *IPA = D->getAttr<InitPriorityAttr>()) {
     OrderGlobalInits Key(IPA->getPriority(), PrioritizedCXXGlobalInits.size());
     PrioritizedCXXGlobalInits.push_back(std::make_pair(Key, Fn));
     DelayedCXXInitPosition.erase(D);
   } else if (D->getTemplateSpecializationKind() != TSK_ExplicitSpecialization &&
if (I == DelayedCXXInitPosition.end()) {
       assert(CXXGlobalInits[I->second] == nullptr);
       CXXGlobalInits[I->second] = Fn;
       DelayedCXXInitPosition.erase(I);
     }
   }
 }
 
 void CodeGenModule::EmitCXXThreadLocalInitFunc() {
-  llvm::Function *InitFn = nullptr;
-  if (!CXXThreadLocalInits.empty()) {
-    // Generate a guarded initialization function.
-    llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false);
-    InitFn = CreateGlobalInitOrDestructFunction(*this, FTy, "__tls_init",
-                                                /*TLS*/ true);
-    llvm::GlobalVariable *Guard = new llvm::GlobalVariable(
-        getModule(), Int8Ty, false, llvm::GlobalVariable::InternalLinkage,
-        llvm::ConstantInt::get(Int8Ty, 0), "__tls_guard");
-    Guard->setThreadLocal(true);
-    CodeGenFunction(*this)
-        .GenerateCXXGlobalInitFunc(InitFn, CXXThreadLocalInits, Guard);
-  }
-
-  getCXXABI().EmitThreadLocalInitFuncs(CXXThreadLocals, InitFn);
+  getCXXABI().EmitThreadLocalInitFuncs(
+      *this, CXXThreadLocals, CXXThreadLocalInits, CXXThreadLocalInitVars);
 
   CXXThreadLocalInits.clear();
+  CXXThreadLocalInitVars.clear();
   CXXThreadLocals.clear();
 }
 
 void
 CodeGenModule::EmitCXXGlobalInitFunc() {
   while (!CXXGlobalInits.empty() && !CXXGlobalInits.back())
     CXXGlobalInits.pop_back();
 
   if (CXXGlobalInits.empty() && PrioritizedCXXGlobalInits.empty())
     return;
 
   llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false);
 
 
   // Create our global initialization function.
   if (!PrioritizedCXXGlobalInits.empty()) {
-    SmallVector<llvm::Constant*, 8> LocalCXXGlobalInits;
+    SmallVector<llvm::Function *, 8> LocalCXXGlobalInits;
     llvm::array_pod_sort(PrioritizedCXXGlobalInits.begin(), 
                          PrioritizedCXXGlobalInits.end());
     // Iterate over "chunks" of ctors with same priority and emit each chunk
     // into separate function. Note - everything is sorted first by priority,
     // second - by lex order, so we emit ctor functions in proper order.
     for (SmallVectorImpl<GlobalInitData >::iterator
            I = PrioritizedCXXGlobalInits.begin(),
            E = PrioritizedCXXGlobalInits.end(); I != E; ) {
       SmallVectorImpl<GlobalInitData >::iterator
         PrioE = std::upper_bound(I + 1, E, *I, GlobalInitPriorityCmp());
 
       LocalCXXGlobalInits.clear();
       unsigned Priority = I->first.priority;
       // Compute the function suffix from priority. Prepend with zeroes to make
       // sure the function names are also ordered as priorities.
       std::string PrioritySuffix = llvm::utostr(Priority);
       // Priority is always <= 65535 (enforced by sema).
       PrioritySuffix = std::string(6-PrioritySuffix.size(), '0')+PrioritySuffix;
-      llvm::Function *Fn = 
-        CreateGlobalInitOrDestructFunction(*this, FTy,
-                                           "_GLOBAL__I_" + PrioritySuffix);
-      
+      llvm::Function *Fn = CreateGlobalInitOrDestructFunction(
+          FTy, "_GLOBAL__I_" + PrioritySuffix);
       for (; I < PrioE; ++I)
         LocalCXXGlobalInits.push_back(I->second);
 
       CodeGenFunction(*this).GenerateCXXGlobalInitFunc(Fn, LocalCXXGlobalInits);
       AddGlobalCtor(Fn, Priority);
     }
   }
 
CodeGenModule::EmitCXXGlobalInitFunc() {
   for (size_t i = 0; i < FileName.size(); ++i) {
     // Replace everything that's not [a-zA-Z0-9._] with a _. This set happens
     // to be the set of C preprocessing numbers.
     if (!isPreprocessingNumberBody(FileName[i]))
       FileName[i] = '_';
   }
 
   llvm::Function *Fn = CreateGlobalInitOrDestructFunction(
-      *this, FTy, llvm::Twine("_GLOBAL__sub_I_", FileName));
+      FTy, llvm::Twine("_GLOBAL__sub_I_", FileName));
 
   CodeGenFunction(*this).GenerateCXXGlobalInitFunc(Fn, CXXGlobalInits);
   AddGlobalCtor(Fn);
 
   CXXGlobalInits.clear();
   PrioritizedCXXGlobalInits.clear();
 }
 
 void CodeGenModule::EmitCXXGlobalDtorFunc() {
   if (CXXGlobalDtors.empty())
     return;
 
   llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false);
 
   // Create our global destructor function.
-  llvm::Function *Fn =
-    CreateGlobalInitOrDestructFunction(*this, FTy, "_GLOBAL__D_a");
+  llvm::Function *Fn = CreateGlobalInitOrDestructFunction(FTy, "_GLOBAL__D_a");
 
   CodeGenFunction(*this).GenerateCXXGlobalDtorsFunc(Fn, CXXGlobalDtors);
   AddGlobalDtor(Fn);
 }
 
 /// Emit the code necessary to initialize the given global variable.
 void CodeGenFunction::GenerateCXXGlobalVarDeclInitFunc(llvm::Function *Fn,
                                                        const VarDecl *D,
if (Addr->hasWeakLinkage() || Addr->hasLinkOnceLinkage()) {
     EmitCXXGlobalVarDeclInit(*D, Addr, PerformInit);
   }
 
   FinishFunction();
 }
 
 void
 CodeGenFunction::GenerateCXXGlobalInitFunc(llvm::Function *Fn,
-                                           ArrayRef<llvm::Constant *> Decls,
+                                           ArrayRef<llvm::Function *> Decls,
                                            llvm::GlobalVariable *Guard) {
   {
     ArtificialLocation AL(*this, Builder);
     StartFunction(GlobalDecl(), getContext().VoidTy, Fn,
                   getTypes().arrangeNullaryFunction(), FunctionArgList());
     // Emit an artificial location for this function.
     AL.Emit();
 
llvm::Function *CodeGenFunction::generateDestroyHelper(
   FunctionArgList args;
   ImplicitParamDecl dst(getContext(), nullptr, SourceLocation(), nullptr,
                         getContext().VoidPtrTy);
   args.push_back(&dst);
 
   const CGFunctionInfo &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
       getContext().VoidTy, args, FunctionType::ExtInfo(), /*variadic=*/false);
   llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
-  llvm::Function *fn = 
-    CreateGlobalInitOrDestructFunction(CGM, FTy, "__cxx_global_array_dtor");
+  llvm::Function *fn =
+      CGM.CreateGlobalInitOrDestructFunction(FTy, "__cxx_global_array_dtor");
 
   StartFunction(VD, getContext().VoidTy, fn, FI, args);
 
   emitDestroy(addr, type, destroyer, useEHCleanupForArray);
   
   FinishFunction();
   
   return fn;
 }

lib/CodeGen/CodeGenFunction.h

AddInitializerToStaticVarDecl(const VarDecl &D,
                                 llvm::GlobalVariable *GV);
 
 
   /// EmitCXXGlobalVarDeclInit - Create the initializer for a C++
   /// variable with global storage.
   void EmitCXXGlobalVarDeclInit(const VarDecl &D, llvm::Constant *DeclPtr,
                                 bool PerformInit);
 
+  llvm::Constant *createAtExitStub(const VarDecl &VD, llvm::Constant *Dtor,
+                                   llvm::Constant *Addr);
+
   /// Call atexit() with a function that passes the given argument to
   /// the given function.
   void registerGlobalDtorWithAtExit(const VarDecl &D, llvm::Constant *fn,
                                     llvm::Constant *addr);
 
   /// Emit code in this function to perform a guarded variable
   /// initialization.  Guarded initializations are used when it's not
   /// possible to prove that an initialization will be done exactly
   /// once, e.g. with a static local variable or a static data member
   /// of a class template.
   void EmitCXXGuardedInit(const VarDecl &D, llvm::GlobalVariable *DeclPtr,
                           bool PerformInit);
 
   /// GenerateCXXGlobalInitFunc - Generates code for initializing global
   /// variables.
   void GenerateCXXGlobalInitFunc(llvm::Function *Fn,
-                                 ArrayRef<llvm::Constant *> Decls,
+                                 ArrayRef<llvm::Function *> CXXThreadLocals,
                                  llvm::GlobalVariable *Guard = nullptr);
 
   /// GenerateCXXGlobalDtorsFunc - Generates code for destroying global
   /// variables.
   void GenerateCXXGlobalDtorsFunc(llvm::Function *Fn,
                                   const std::vector<std::pair<llvm::WeakVH,
                                   llvm::Constant*> > &DtorsAndObjects);
 

lib/CodeGen/CodeGenModule.h

class CodeGenModule : public CodeGenTypeCache {
   StaticExternCMap StaticExternCValues;
 
   /// \brief thread_local variables defined or used in this TU.
   std::vector<std::pair<const VarDecl *, llvm::GlobalVariable *> >
     CXXThreadLocals;
 
   /// \brief thread_local variables with initializers that need to run
   /// before any thread_local variable in this TU is odr-used.
-  std::vector<llvm::Constant*> CXXThreadLocalInits;
+  std::vector<llvm::Function *> CXXThreadLocalInits;
+  std::vector<llvm::GlobalVariable *> CXXThreadLocalInitVars;
 
   /// Global variables with initializers that need to run before main.
-  std::vector<llvm::Constant*> CXXGlobalInits;
+  std::vector<llvm::Function *> CXXGlobalInits;
 
   /// When a C++ decl with an initializer is deferred, null is
   /// appended to CXXGlobalInits, and the index of that null is placed
   /// here so that the initializer will be performed in the correct
   /// order.
   llvm::DenseMap<const Decl*, unsigned> DelayedCXXInitPosition;
   
   typedef std::pair<OrderGlobalInits, llvm::Function*> GlobalInitData;
public:
   /// Will return a global variable of the given type. If a variable with a
   /// different type already exists then a new  variable with the right type
   /// will be created and all uses of the old variable will be replaced with a
   /// bitcast to the new variable.
   llvm::GlobalVariable *
   CreateOrReplaceCXXRuntimeVariable(StringRef Name, llvm::Type *Ty,
                                     llvm::GlobalValue::LinkageTypes Linkage);
 
+  llvm::Function *CreateGlobalInitOrDestructFunction(llvm::FunctionType *ty,
+                                                     const Twine &name,
+                                                     bool TLS = false);
+
   /// Return the address space of the underlying global variable for D, as
   /// determined by its declaration. Normally this is the same as the address
   /// space of D's type, but in CUDA, address spaces are associated with
   /// declarations, not types.
   unsigned GetGlobalVarAddressSpace(const VarDecl *D, unsigned AddrSpace);
 
   /// Return the llvm::Constant for the address of the given global variable.
   /// If Ty is non-null and if the global doesn't exist, then it will be greated

lib/CodeGen/ItaniumCXXABI.cpp

llvm::Value *readArrayCookieImpl(CodeGenFunction &CGF,
                                    CharUnits cookieSize) override;
 
   void EmitGuardedInit(CodeGenFunction &CGF, const VarDecl &D,
                        llvm::GlobalVariable *DeclPtr,
                        bool PerformInit) override;
   void registerGlobalDtor(CodeGenFunction &CGF, const VarDecl &D,
                           llvm::Constant *dtor, llvm::Constant *addr) override;
 
   llvm::Function *getOrCreateThreadLocalWrapper(const VarDecl *VD,
                                                 llvm::Value *Val);
   void EmitThreadLocalInitFuncs(
-      ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *> > Decls,
-      llvm::Function *InitFunc) override;
+      CodeGenModule &CGM,
+      ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *>>
+          CXXThreadLocals,
+      ArrayRef<llvm::Function *> CXXThreadLocalInits,
+      ArrayRef<llvm::GlobalVariable *> CXXThreadLocalInitVars) override;
   LValue EmitThreadLocalVarDeclLValue(CodeGenFunction &CGF, const VarDecl *VD,
                                       QualType LValType) override;
 
   bool NeedsVTTParameter(GlobalDecl GD) override;
 
   /**************************** RTTI Uniqueness ******************************/
 
 protected:
llvm::Function *Wrapper =
                              WrapperName.str(), &CGM.getModule());
   // Always resolve references to the wrapper at link time.
   if (!Wrapper->hasLocalLinkage() && !isThreadWrapperReplaceable(VD, CGM))
     Wrapper->setVisibility(llvm::GlobalValue::HiddenVisibility);
   return Wrapper;
 }
 
 void ItaniumCXXABI::EmitThreadLocalInitFuncs(
-    ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *> > Decls,
-    llvm::Function *InitFunc) {
-  for (unsigned I = 0, N = Decls.size(); I != N; ++I) {
-    const VarDecl *VD = Decls[I].first;
-    llvm::GlobalVariable *Var = Decls[I].second;
+    CodeGenModule &CGM,
+    ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *>>
+        CXXThreadLocals, ArrayRef<llvm::Function *> CXXThreadLocalInits,
+    ArrayRef<llvm::GlobalVariable *> CXXThreadLocalInitVars) {
+  llvm::Function *InitFunc = nullptr;
+  if (!CXXThreadLocalInits.empty()) {
+    // Generate a guarded initialization function.
+    llvm::FunctionType *FTy =
+        llvm::FunctionType::get(CGM.VoidTy, /*isVarArg=*/false);
+    InitFunc = CGM.CreateGlobalInitOrDestructFunction(FTy, "__tls_init",
+                                                      /*TLS=*/true);
+    llvm::GlobalVariable *Guard = new llvm::GlobalVariable(
+        CGM.getModule(), CGM.Int8Ty, /*isConstant=*/false,
+        llvm::GlobalVariable::InternalLinkage,
+        llvm::ConstantInt::get(CGM.Int8Ty, 0), "__tls_guard");
+    Guard->setThreadLocal(true);
+    CodeGenFunction(CGM)
+        .GenerateCXXGlobalInitFunc(InitFunc, CXXThreadLocalInits, Guard);
+  }
+  for (unsigned I = 0, N = CXXThreadLocals.size(); I != N; ++I) {
+    const VarDecl *VD = CXXThreadLocals[I].first;
+    llvm::GlobalVariable *Var = CXXThreadLocals[I].second;
 
     // Some targets require that all access to thread local variables go through
     // the thread wrapper.  This means that we cannot attempt to create a thread
     // wrapper or a thread helper.
     if (isThreadWrapperReplaceable(VD, CGM) && !VD->hasDefinition())
       continue;
 
     // Mangle the name for the thread_local initialization function.

lib/CodeGen/MicrosoftCXXABI.cpp

public:
   }
 
   llvm::Value *performThisAdjustment(CodeGenFunction &CGF, llvm::Value *This,
                                      const ThisAdjustment &TA) override;
 
   llvm::Value *performReturnAdjustment(CodeGenFunction &CGF, llvm::Value *Ret,
                                        const ReturnAdjustment &RA) override;
 
+  void EmitThreadLocalInitFuncs(
+      CodeGenModule &CGM,
+      ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *>>
+          CXXThreadLocals,
+      ArrayRef<llvm::Function *> CXXThreadLocalInits,
+      ArrayRef<llvm::GlobalVariable *> CXXThreadLocalInitVars) override;
+
+  LValue EmitThreadLocalVarDeclLValue(CodeGenFunction &CGF, const VarDecl *VD,
+                                      QualType LValType) override;
+
   void EmitGuardedInit(CodeGenFunction &CGF, const VarDecl &D,
                        llvm::GlobalVariable *DeclPtr,
                        bool PerformInit) override;
+  void registerGlobalDtor(CodeGenFunction &CGF, const VarDecl &D,
+                          llvm::Constant *Dtor, llvm::Constant *Addr) override;
 
   // ==== Notes on array cookies =========
   //
   // MSVC seems to only use cookies when the class has a destructor; a
   // two-argument usual array deallocation function isn't sufficient.
   //
   // For example, this code prints "100" and "1":
   //   struct A {
   //     char x;
   //     void *operator new[](size_t sz) {
llvm::Value* MicrosoftCXXABI::InitializeArrayCookie(CodeGenFunction &CGF,
   CGF.Builder.CreateStore(numElements, numElementsPtr);
 
   // Finally, compute a pointer to the actual data buffer by skipping
   // over the cookie completely.
   return CGF.Builder.CreateConstInBoundsGEP1_64(newPtr,
                                                 cookieSize.getQuantity());
 }
 
+static void emitGlobalDtorWithTLRegDtor(CodeGenFunction &CGF, const VarDecl &VD,
+                                        llvm::Constant *Dtor,
+                                        llvm::Constant *Addr) {
+  // Create a function which calls the destructor.
+  llvm::Constant *DtorStub = CGF.createAtExitStub(VD, Dtor, Addr);
+
+  // extern "C" int __tlregdtor(void (*f)(void));
+  llvm::FunctionType *TLRegDtorTy = llvm::FunctionType::get(
+      CGF.IntTy, DtorStub->getType(), /*IsVarArg=*/false);
+
+  llvm::Constant *TLRegDtor =
+      CGF.CGM.CreateRuntimeFunction(TLRegDtorTy, "__tlregdtor");
+  if (llvm::Function *TLRegDtorFn = dyn_cast<llvm::Function>(TLRegDtor))
+    TLRegDtorFn->setDoesNotThrow();
+
+  CGF.EmitNounwindRuntimeCall(TLRegDtor, DtorStub);
+}
+
+void MicrosoftCXXABI::registerGlobalDtor(CodeGenFunction &CGF, const VarDecl &D,
+                                         llvm::Constant *Dtor,
+                                         llvm::Constant *Addr) {
+  if (D.getTLSKind())
+    return emitGlobalDtorWithTLRegDtor(CGF, D, Dtor, Addr);
+
+  // The default behavior is to use atexit.
+  CGF.registerGlobalDtorWithAtExit(D, Dtor, Addr);
+}
+
+void MicrosoftCXXABI::EmitThreadLocalInitFuncs(
+    CodeGenModule &CGM,
+    ArrayRef<std::pair<const VarDecl *, llvm::GlobalVariable *>>
+        CXXThreadLocals,
+    ArrayRef<llvm::Function *> CXXThreadLocalInits,
+    ArrayRef<llvm::GlobalVariable *> CXXThreadLocalInitVars) {
+  // This will create a GV in the .CRT$XDU section.  It will point to our
+  // initialization function.  The CRT will call all of these function
+  // pointers at start-up time and, eventually, at thread-creation time.
+  auto AddToXDU = [&CGM](llvm::Function *InitFunc) {
+    llvm::GlobalVariable *InitFuncPtr = new llvm::GlobalVariable(
+        CGM.getModule(), InitFunc->getType(), /*IsConstant=*/true,
+        llvm::GlobalVariable::InternalLinkage, InitFunc,
+        Twine(InitFunc->getName(), "$initializer$"));
+    InitFuncPtr->setSection(".CRT$XDU");
+    // This variable has discardable linkage, we have to add it to @llvm.used to
+    // ensure it won't get discarded.
+    CGM.addUsedGlobal(InitFuncPtr);
+    return InitFuncPtr;
+  };
+
+  std::vector<llvm::Function *> NonComdatInits;
+  for (size_t I = 0, E = CXXThreadLocalInitVars.size(); I != E; ++I) {
+    llvm::GlobalVariable *GV = CXXThreadLocalInitVars[I];
+    llvm::Function *F = CXXThreadLocalInits[I];
+
+    // If the GV is already in a comdat group, then we have to join it.
+    llvm::Comdat *C = GV->getComdat();
+
+    // LinkOnce and Weak linkage are lowered down to a single-member comdat
+    // group.
+    // Make an explicit group so we can join it.
+    if (!C && (GV->hasWeakLinkage() || GV->hasLinkOnceLinkage())) {
+      C = CGM.getModule().getOrInsertComdat(GV->getName());
+      GV->setComdat(C);
+      AddToXDU(F)->setComdat(C);
+    } else {
+      NonComdatInits.push_back(F);
+    }
+  }
+
+  if (!NonComdatInits.empty()) {
+    llvm::FunctionType *FTy =
+        llvm::FunctionType::get(CGM.VoidTy, /*isVarArg=*/false);
+    llvm::Function *InitFunc =
+        CGM.CreateGlobalInitOrDestructFunction(FTy, "__tls_init",
+                                               /*TLS=*/true);
+    CodeGenFunction(CGM).GenerateCXXGlobalInitFunc(InitFunc, NonComdatInits);
+
+    AddToXDU(InitFunc);
+  }
+}
+
+LValue MicrosoftCXXABI::EmitThreadLocalVarDeclLValue(CodeGenFunction &CGF,
+                                                     const VarDecl *VD,
+                                                     QualType LValType) {
+  QualType T = VD->getType();
+  llvm::Type *Ty = CGF.getTypes().ConvertTypeForMem(T);
+  llvm::Value *Val = CGF.CGM.GetAddrOfGlobalVar(VD, Ty);
+
+  LValue LV;
+  if (VD->getType()->isReferenceType())
+    LV = CGF.MakeNaturalAlignAddrLValue(Val, LValType);
+  else
+    LV = CGF.MakeAddrLValue(Val, LValType, CGF.getContext().getDeclAlign(VD));
+  return LV;
+}
+
 void MicrosoftCXXABI::EmitGuardedInit(CodeGenFunction &CGF, const VarDecl &D,
                                       llvm::GlobalVariable *GV,
                                       bool PerformInit) {
   // MSVC only uses guards for static locals.
   if (!D.isStaticLocal()) {
     assert(GV->hasWeakLinkage() || GV->hasLinkOnceLinkage());
     // GlobalOpt is allowed to discard the initializer, so use linkonce_odr.
     CGF.CurFn->setLinkage(llvm::GlobalValue::LinkOnceODRLinkage);

test/CodeGenCXX/ms-thread_local.cpp

+// RUN: %clang_cc1 %s -std=c++1y -triple=i686-pc-win32 -emit-llvm -o - | FileCheck %s
+
+struct A {
+  A();
+  ~A();
+};
+
+// CHECK-DAG: $"\01??$a@X@@3UA@@A" = comdat any
+// CHECK-DAG: @"\01??$a@X@@3UA@@A" = linkonce_odr thread_local global %struct.A zeroinitializer, comdat $"\01??$a@X@@3UA@@A"
+// CHECK-DAG: @"\01??__E?$a@X@@YAXXZ$initializer$" = internal constant void ()* @"\01??__E?$a@X@@YAXXZ", section ".CRT$XDU", comdat $"\01??$a@X@@3UA@@A"
+template <typename T>
+thread_local A a = A();
+
+// CHECK-DAG: @"\01?b@@3UA@@A" = thread_local global %struct.A zeroinitializer, align 1
+// CHECK-DAG: @"__tls_init$initializer$" = internal constant void ()* @__tls_init, section ".CRT$XDU"
+thread_local A b;
+
+// CHECK-LABEL: define internal void @__tls_init()
+// CHECK: call void @"\01??__Eb@@YAXXZ"
+
+void f() {
+  (void)a<void>;
+  (void)b;
+}