diff --git a/clang/lib/AST/Interp/Interp.h b/clang/lib/AST/Interp/Interp.h
@@ -199,6 +199,8 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const Function *F,
 bool InterpretOffsetOf(InterpState &S, CodePtr OpPC, const OffsetOfExpr *E,
                        llvm::ArrayRef<int64_t> ArrayIndices, int64_t &Result);
 
+inline bool Invalid(InterpState &S, CodePtr OpPC);
+
 enum class ArithOp { Add, Sub };
 
 //===----------------------------------------------------------------------===//
@@ -522,6 +524,11 @@ bool IncDecHelper(InterpState &S, CodePtr OpPC, const Pointer &Ptr) {
   if (Ptr.isDummy())
     return false;
 
+  if constexpr (std::is_same_v<T, Boolean>) {
+    if (!S.getLangOpts().CPlusPlus14)
+      return Invalid(S, OpPC);
+  }
+
   const T &Value = Ptr.deref<T>();
   T Result;
 
@@ -572,7 +579,8 @@ bool IncDecHelper(InterpState &S, CodePtr OpPC, const Pointer &Ptr) {
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool Inc(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Increment))
     return false;
 
@@ -585,7 +593,8 @@ bool Inc(InterpState &S, CodePtr OpPC) {
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool IncPop(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Increment))
     return false;
 
@@ -599,7 +608,8 @@ bool IncPop(InterpState &S, CodePtr OpPC) {
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool Dec(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Decrement))
     return false;
 
@@ -612,7 +622,8 @@ bool Dec(InterpState &S, CodePtr OpPC) {
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool DecPop(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Decrement))
     return false;
 
@@ -641,7 +652,8 @@ bool IncDecFloatHelper(InterpState &S, CodePtr OpPC, const Pointer &Ptr,
 
 inline bool Incf(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Increment))
     return false;
 
@@ -650,7 +662,8 @@ inline bool Incf(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
 
 inline bool IncfPop(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Increment))
     return false;
 
@@ -660,6 +673,9 @@ inline bool IncfPop(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
 inline bool Decf(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
 
+  if (Ptr.isDummy())
+    return false;
+
   if (!CheckInitialized(S, OpPC, Ptr, AK_Decrement))
     return false;
 
@@ -669,6 +685,8 @@ inline bool Decf(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
 inline bool DecfPop(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
 
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Decrement))
     return false;
 
@@ -774,9 +792,9 @@ inline bool CmpHelperEQ<Pointer>(InterpState &S, CodePtr OpPC, CompareFn Fn) {
     // element in the same array are NOT equal. They have the same Base value,
     // but a different Offset. This is a pretty rare case, so we fix this here
     // by comparing pointers to the first elements.
-    if (LHS.isArrayRoot())
+    if (!LHS.isDummy() && LHS.isArrayRoot())
       VL = LHS.atIndex(0).getByteOffset();
-    if (RHS.isArrayRoot())
+    if (!RHS.isDummy() && RHS.isArrayRoot())
       VR = RHS.atIndex(0).getByteOffset();
 
     S.Stk.push<BoolT>(BoolT::from(Fn(Compare(VL, VR))));
@@ -1680,7 +1698,7 @@ bool CastFloatingIntegral(InterpState &S, CodePtr OpPC) {
     auto Status = F.convertToInteger(Result);
 
     // Float-to-Integral overflow check.
-    if ((Status & APFloat::opStatus::opInvalidOp) && F.isFinite()) {
+    if ((Status & APFloat::opStatus::opInvalidOp)) {
       const Expr *E = S.Current->getExpr(OpPC);
       QualType Type = E->getType();
 
@@ -1895,7 +1913,7 @@ inline bool ArrayElemPtr(InterpState &S, CodePtr OpPC) {
   const T &Offset = S.Stk.pop<T>();
   const Pointer &Ptr = S.Stk.peek<Pointer>();
 
-  if (!CheckDummy(S, OpPC, Ptr))
+  if (Ptr.isDummy())
     return true;
 
   if (!OffsetHelper<T, ArithOp::Add>(S, OpPC, Offset, Ptr))
@@ -1909,7 +1927,7 @@ inline bool ArrayElemPtrPop(InterpState &S, CodePtr OpPC) {
   const T &Offset = S.Stk.pop<T>();
   const Pointer &Ptr = S.Stk.pop<Pointer>();
 
-  if (!CheckDummy(S, OpPC, Ptr)) {
+  if (Ptr.isDummy()) {
     S.Stk.push<Pointer>(Ptr);
     return true;
   }
@@ -1933,7 +1951,7 @@ inline bool ArrayElemPop(InterpState &S, CodePtr OpPC, uint32_t Index) {
 inline bool ArrayDecay(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
 
-  if (Ptr.isDummy()) {
+  if (Ptr.isZero() || Ptr.isDummy()) {
     S.Stk.push<Pointer>(Ptr);
     return true;
   }
@@ -2056,8 +2074,12 @@ inline bool CallVirt(InterpState &S, CodePtr OpPC, const Function *Func,
   size_t ThisOffset = ArgSize - (Func->hasRVO() ? primSize(PT_Ptr) : 0);
   Pointer &ThisPtr = S.Stk.peek<Pointer>(ThisOffset);
 
-  const CXXRecordDecl *DynamicDecl =
-      ThisPtr.getDeclDesc()->getType()->getAsCXXRecordDecl();
+  QualType DynamicType = ThisPtr.getDeclDesc()->getType();
+  const CXXRecordDecl *DynamicDecl;
+  if (DynamicType->isPointerType() || DynamicType->isReferenceType())
+    DynamicDecl = DynamicType->getPointeeCXXRecordDecl();
+  else
+    DynamicDecl = ThisPtr.getDeclDesc()->getType()->getAsCXXRecordDecl();
   const auto *StaticDecl = cast<CXXRecordDecl>(Func->getParentDecl());
   const auto *InitialFunction = cast<CXXMethodDecl>(Func->getDecl());
   const CXXMethodDecl *Overrider = S.getContext().getOverridingFunction(

diff --git a/clang/lib/AST/Interp/Opcodes.td b/clang/lib/AST/Interp/Opcodes.td
@@ -563,10 +563,10 @@ def Inv: Opcode {
 }
 
 // Increment and decrement.
-def Inc: IntegerOpcode;
-def IncPop : IntegerOpcode;
-def Dec: IntegerOpcode;
-def DecPop: IntegerOpcode;
+def Inc: AluOpcode;
+def IncPop : AluOpcode;
+def Dec: AluOpcode;
+def DecPop: AluOpcode;
 
 // Float increment and decrement.
 def Incf: FloatOpcode;

diff --git a/clang/lib/AST/Interp/Pointer.h b/clang/lib/AST/Interp/Pointer.h
@@ -215,7 +215,6 @@ class Pointer {
       assert(Offset == PastEndMark && "cannot get base of a block");
       return Pointer(Pointee, Base, 0);
     }
-    assert(Offset == Base && "not an inner field");
     unsigned NewBase = Base - getInlineDesc()->Offset;
     return Pointer(Pointee, NewBase, NewBase);
   }

diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -361,8 +361,8 @@ getFieldsGlobalsAndFuncs(const Stmt &S, FieldSet &Fields,
     if (const auto *FD = dyn_cast<FieldDecl>(VD))
       Fields.insert(FD);
   } else if (auto *InitList = dyn_cast<InitListExpr>(&S)) {
-    if (RecordDecl *RD = InitList->getType()->getAsRecordDecl())
-      for (const auto *FD : getFieldsForInitListExpr(RD))
+    if (InitList->getType()->isRecordType())
+      for (const auto *FD : getFieldsForInitListExpr(InitList))
         Fields.insert(FD);
   }
 }
@@ -983,7 +983,7 @@ StorageLocation &Environment::createObjectInternal(const ValueDecl *D,
   }
 
   Value *Val = nullptr;
-  if (InitExpr)
+  if (InitExpr) {
     // In the (few) cases where an expression is intentionally
     // "uninterpreted", `InitExpr` is not associated with a value.  There are
     // two ways to handle this situation: propagate the status, so that
@@ -998,6 +998,11 @@ StorageLocation &Environment::createObjectInternal(const ValueDecl *D,
     // default value (assuming we don't update the environment API to return
     // references).
     Val = getValue(*InitExpr);
+
+    if (!Val && isa<ImplicitValueInitExpr>(InitExpr) &&
+        InitExpr->getType()->isPointerType())
+      Val = &getOrCreateNullPointerValue(InitExpr->getType()->getPointeeType());
+  }
   if (!Val)
     Val = createValue(Ty);
 
@@ -1104,12 +1109,22 @@ RecordStorageLocation *getBaseObjectLocation(const MemberExpr &ME,
   return Env.get<RecordStorageLocation>(*Base);
 }
 
-std::vector<FieldDecl *> getFieldsForInitListExpr(const RecordDecl *RD) {
+std::vector<const FieldDecl *>
+getFieldsForInitListExpr(const InitListExpr *InitList) {
+  const RecordDecl *RD = InitList->getType()->getAsRecordDecl();
+  assert(RD != nullptr);
+
+  std::vector<const FieldDecl *> Fields;
+
+  if (InitList->getType()->isUnionType()) {
+    Fields.push_back(InitList->getInitializedFieldInUnion());
+    return Fields;
+  }
+
   // Unnamed bitfields are only used for padding and do not appear in
   // `InitListExpr`'s inits. However, those fields do appear in `RecordDecl`'s
   // field list, and we thus need to remove them before mapping inits to
   // fields to avoid mapping inits to the wrongs fields.
-  std::vector<FieldDecl *> Fields;
   llvm::copy_if(
       RD->fields(), std::back_inserter(Fields),
       [](const FieldDecl *Field) { return !Field->isUnnamedBitfield(); });

diff --git a/clang/lib/Analysis/FlowSensitive/Transfer.cpp b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
@@ -663,14 +663,7 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
   void VisitInitListExpr(const InitListExpr *S) {
     QualType Type = S->getType();
 
-    if (Type->isUnionType()) {
-      // FIXME: Initialize unions properly.
-      if (auto *Val = Env.createValue(Type))
-        Env.setValue(*S, *Val);
-      return;
-    }
-
-    if (!Type->isStructureOrClassType()) {
+    if (!Type->isRecordType()) {
       // Until array initialization is implemented, we skip arrays and don't
       // need to care about cases where `getNumInits() > 1`.
       if (!Type->isArrayType() && S->getNumInits() == 1)
@@ -688,14 +681,26 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     llvm::DenseMap<const ValueDecl *, StorageLocation *> FieldLocs;
 
     // This only contains the direct fields for the given type.
-    std::vector<FieldDecl *> FieldsForInit =
-        getFieldsForInitListExpr(Type->getAsRecordDecl());
+    std::vector<const FieldDecl *> FieldsForInit = getFieldsForInitListExpr(S);
 
-    // `S->inits()` contains all the initializer epressions, including the
+    // `S->inits()` contains all the initializer expressions, including the
     // ones for direct base classes.
-    auto Inits = S->inits();
+    ArrayRef<Expr *> Inits = S->inits();
     size_t InitIdx = 0;
 
+    // Unions initialized with an empty initializer list need special treatment.
+    // For structs/classes initialized with an empty initializer list, Clang
+    // puts `ImplicitValueInitExpr`s in `InitListExpr::inits()`, but for unions,
+    // it doesn't do this -- so we create an `ImplicitValueInitExpr` ourselves.
+    std::optional<ImplicitValueInitExpr> ImplicitValueInitForUnion;
+    SmallVector<Expr *> InitsForUnion;
+    if (S->getType()->isUnionType() && Inits.empty()) {
+      assert(FieldsForInit.size() == 1);
+      ImplicitValueInitForUnion.emplace(FieldsForInit.front()->getType());
+      InitsForUnion.push_back(&*ImplicitValueInitForUnion);
+      Inits = InitsForUnion;
+    }
+
     // Initialize base classes.
     if (auto* R = S->getType()->getAsCXXRecordDecl()) {
       assert(FieldsForInit.size() + R->getNumBases() == Inits.size());
@@ -731,6 +736,17 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       FieldLocs.insert({Field, &Loc});
     }
 
+    // In the case of a union, we don't in general have initializers for all
+    // of the fields. Create storage locations for the remaining fields (but
+    // don't associate them with values).
+    if (Type->isUnionType()) {
+      for (const FieldDecl *Field :
+           Env.getDataflowAnalysisContext().getModeledFields(Type)) {
+        if (auto [it, inserted] = FieldLocs.insert({Field, nullptr}); inserted)
+          it->second = &Env.createStorageLocation(Field->getType());
+      }
+    }
+
     // Check that we satisfy the invariant that a `RecordStorageLoation`
     // contains exactly the set of modeled fields for that type.
     // `ModeledFields` includes fields from all the bases, but only the

diff --git a/clang/lib/Analysis/UnsafeBufferUsage.cpp b/clang/lib/Analysis/UnsafeBufferUsage.cpp
diff --git a/clang/lib/CodeGen/ABIInfo.cpp b/clang/lib/CodeGen/ABIInfo.cpp
@@ -184,6 +184,58 @@ ABIArgInfo ABIInfo::getNaturalAlignIndirectInReg(QualType Ty,
                                       /*ByVal*/ false, Realign);
 }
 
+void ABIInfo::appendAttributeMangling(TargetAttr *Attr,
+                                      raw_ostream &Out) const {
+  if (Attr->isDefaultVersion())
+    return;
+  appendAttributeMangling(Attr->getFeaturesStr(), Out);
+}
+
+void ABIInfo::appendAttributeMangling(TargetVersionAttr *Attr,
+                                      raw_ostream &Out) const {
+  appendAttributeMangling(Attr->getNamesStr(), Out);
+}
+
+void ABIInfo::appendAttributeMangling(TargetClonesAttr *Attr, unsigned Index,
+                                      raw_ostream &Out) const {
+  appendAttributeMangling(Attr->getFeatureStr(Index), Out);
+  Out << '.' << Attr->getMangledIndex(Index);
+}
+
+void ABIInfo::appendAttributeMangling(StringRef AttrStr,
+                                      raw_ostream &Out) const {
+  if (AttrStr == "default") {
+    Out << ".default";
+    return;
+  }
+
+  Out << '.';
+  const TargetInfo &TI = CGT.getTarget();
+  ParsedTargetAttr Info = TI.parseTargetAttr(AttrStr);
+
+  llvm::sort(Info.Features, [&TI](StringRef LHS, StringRef RHS) {
+    // Multiversioning doesn't allow "no-${feature}", so we can
+    // only have "+" prefixes here.
+    assert(LHS.starts_with("+") && RHS.starts_with("+") &&
+           "Features should always have a prefix.");
+    return TI.multiVersionSortPriority(LHS.substr(1)) >
+           TI.multiVersionSortPriority(RHS.substr(1));
+  });
+
+  bool IsFirst = true;
+  if (!Info.CPU.empty()) {
+    IsFirst = false;
+    Out << "arch_" << Info.CPU;
+  }
+
+  for (StringRef Feat : Info.Features) {
+    if (!IsFirst)
+      Out << '_';
+    IsFirst = false;
+    Out << Feat.substr(1);
+  }
+}
+
 // Pin the vtable to this file.
 SwiftABIInfo::~SwiftABIInfo() = default;
 

diff --git a/clang/lib/CodeGen/ABIInfo.h b/clang/lib/CodeGen/ABIInfo.h
@@ -9,6 +9,7 @@
 #ifndef LLVM_CLANG_LIB_CODEGEN_ABIINFO_H
 #define LLVM_CLANG_LIB_CODEGEN_ABIINFO_H
 
+#include "clang/AST/Attr.h"
 #include "clang/AST/CharUnits.h"
 #include "clang/AST/Type.h"
 #include "llvm/IR/CallingConv.h"
@@ -111,6 +112,15 @@ class ABIInfo {
 
   CodeGen::ABIArgInfo getNaturalAlignIndirectInReg(QualType Ty,
                                                    bool Realign = false) const;
+
+  virtual void appendAttributeMangling(TargetAttr *Attr,
+                                       raw_ostream &Out) const;
+  virtual void appendAttributeMangling(TargetVersionAttr *Attr,
+                                       raw_ostream &Out) const;
+  virtual void appendAttributeMangling(TargetClonesAttr *Attr, unsigned Index,
+                                       raw_ostream &Out) const;
+  virtual void appendAttributeMangling(StringRef AttrStr,
+                                       raw_ostream &Out) const;
 };
 
 /// Target specific hooks for defining how a type should be passed or returned

diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp
@@ -186,6 +186,14 @@ class EmitAssemblyHelper {
            TargetTriple.getVendor() != llvm::Triple::Apple;
   }
 
+  /// Check whether we should emit a flag for UnifiedLTO.
+  /// The UnifiedLTO module flag should be set when UnifiedLTO is enabled for
+  /// ThinLTO or Full LTO with module summaries.
+  bool shouldEmitUnifiedLTOModueFlag() const {
+    return CodeGenOpts.UnifiedLTO &&
+           (CodeGenOpts.PrepareForThinLTO || shouldEmitRegularLTOSummary());
+  }
+
 public:
   EmitAssemblyHelper(DiagnosticsEngine &_Diags,
                      const HeaderSearchOptions &HeaderSearchOpts,
@@ -1036,7 +1044,8 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
   if (!actionRequiresCodeGen(Action) && CodeGenOpts.VerifyModule)
     MPM.addPass(VerifierPass());
 
-  if (Action == Backend_EmitBC || Action == Backend_EmitLL) {
+  if (Action == Backend_EmitBC || Action == Backend_EmitLL ||
+      CodeGenOpts.FatLTO) {
     if (CodeGenOpts.PrepareForThinLTO && !CodeGenOpts.DisableLLVMPasses) {
       if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
         TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
@@ -1047,11 +1056,9 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
           if (!ThinLinkOS)
             return;
         }
-        if (CodeGenOpts.UnifiedLTO)
-          TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
         MPM.addPass(ThinLTOBitcodeWriterPass(
             *OS, ThinLinkOS ? &ThinLinkOS->os() : nullptr));
-      } else {
+      } else if (Action == Backend_EmitLL) {
         MPM.addPass(PrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists,
                                     /*EmitLTOSummary=*/true));
       }
@@ -1065,24 +1072,17 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
         if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
           TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
                                    uint32_t(1));
-        if (CodeGenOpts.UnifiedLTO)
-          TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
       }
-      if (Action == Backend_EmitBC)
+      if (Action == Backend_EmitBC) {
         MPM.addPass(BitcodeWriterPass(*OS, CodeGenOpts.EmitLLVMUseLists,
                                       EmitLTOSummary));
-      else
+      } else if (Action == Backend_EmitLL) {
         MPM.addPass(PrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists,
                                     EmitLTOSummary));
+      }
     }
-  }
-  if (CodeGenOpts.FatLTO) {
-    // Set the EnableSplitLTOUnit and UnifiedLTO module flags, since FatLTO
-    // uses a different action than Backend_EmitBC or Backend_EmitLL.
-    if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
-      TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
-                               uint32_t(CodeGenOpts.EnableSplitLTOUnit));
-    if (CodeGenOpts.UnifiedLTO && !TheModule->getModuleFlag("UnifiedLTO"))
+
+    if (shouldEmitUnifiedLTOModueFlag())
       TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
   }
 

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -13952,6 +13952,8 @@ Value *CodeGenFunction::EmitX86CpuIs(StringRef CPUStr) {
 Value *CodeGenFunction::EmitX86CpuSupports(const CallExpr *E) {
   const Expr *FeatureExpr = E->getArg(0)->IgnoreParenCasts();
   StringRef FeatureStr = cast<StringLiteral>(FeatureExpr)->getString();
+  if (!getContext().getTargetInfo().validateCpuSupports(FeatureStr))
+    return Builder.getFalse();
   return EmitX86CpuSupports(FeatureStr);
 }
 
@@ -14041,6 +14043,8 @@ Value *CodeGenFunction::EmitAArch64CpuSupports(const CallExpr *E) {
   ArgStr.split(Features, "+");
   for (auto &Feature : Features) {
     Feature = Feature.trim();
+    if (!llvm::AArch64::parseArchExtension(Feature))
+      return Builder.getFalse();
     if (Feature != "default")
       Features.push_back(Feature);
   }
@@ -16639,7 +16643,8 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
   .Case(Name, {FA_WORD, Bitmask})
 #include "llvm/TargetParser/PPCTargetParser.def"
             .Default({0, 0});
-    assert(BitMask && "Invalid target feature string. Missed by SemaChecking?");
+    if (!BitMask)
+      return Builder.getFalse();
     Value *Op0 = llvm::ConstantInt::get(Int32Ty, FeatureWord);
     llvm::Function *F = CGM.getIntrinsic(Intrinsic::ppc_fixed_addr_ld);
     Value *TheCall = Builder.CreateCall(F, {Op0}, "cpu_supports");
@@ -18007,6 +18012,51 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
         /*ReturnType*/ T0->getScalarType(), Intrinsic::dx_dot,
         ArrayRef<Value *>{Op0, Op1}, nullptr, "dx.dot");
   } break;
+  case Builtin::BI__builtin_hlsl_lerp: {
+    Value *X = EmitScalarExpr(E->getArg(0));
+    Value *Y = EmitScalarExpr(E->getArg(1));
+    Value *S = EmitScalarExpr(E->getArg(2));
+    llvm::Type *Xty = X->getType();
+    llvm::Type *Yty = Y->getType();
+    llvm::Type *Sty = S->getType();
+    if (!Xty->isVectorTy() && !Yty->isVectorTy() && !Sty->isVectorTy()) {
+      if (Xty->isFloatingPointTy()) {
+        auto V = Builder.CreateFSub(Y, X);
+        V = Builder.CreateFMul(S, V);
+        return Builder.CreateFAdd(X, V, "dx.lerp");
+      }
+      llvm_unreachable("Scalar Lerp is only supported on floats.");
+    }
+    // A VectorSplat should have happened
+    assert(Xty->isVectorTy() && Yty->isVectorTy() && Sty->isVectorTy() &&
+           "Lerp of vector and scalar is not supported.");
+
+    [[maybe_unused]] auto *XVecTy =
+        E->getArg(0)->getType()->getAs<VectorType>();
+    [[maybe_unused]] auto *YVecTy =
+        E->getArg(1)->getType()->getAs<VectorType>();
+    [[maybe_unused]] auto *SVecTy =
+        E->getArg(2)->getType()->getAs<VectorType>();
+    // A HLSLVectorTruncation should have happend
+    assert(XVecTy->getNumElements() == YVecTy->getNumElements() &&
+           XVecTy->getNumElements() == SVecTy->getNumElements() &&
+           "Lerp requires vectors to be of the same size.");
+    assert(XVecTy->getElementType()->isRealFloatingType() &&
+           XVecTy->getElementType() == YVecTy->getElementType() &&
+           XVecTy->getElementType() == SVecTy->getElementType() &&
+           "Lerp requires float vectors to be of the same type.");
+    return Builder.CreateIntrinsic(
+        /*ReturnType*/ Xty, Intrinsic::dx_lerp, ArrayRef<Value *>{X, Y, S},
+        nullptr, "dx.lerp");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_frac: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("frac operand must have a float representation");
+    return Builder.CreateIntrinsic(
+        /*ReturnType*/ Op0->getType(), Intrinsic::dx_frac,
+        ArrayRef<Value *>{Op0}, nullptr, "dx.frac");
+  }
   }
   return nullptr;
 }

diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
@@ -3221,12 +3221,11 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
 
       llvm::StructType *STy =
           dyn_cast<llvm::StructType>(ArgI.getCoerceToType());
-      llvm::TypeSize StructSize;
-      llvm::TypeSize PtrElementSize;
       if (ArgI.isDirect() && !ArgI.getCanBeFlattened() && STy &&
           STy->getNumElements() > 1) {
-        StructSize = CGM.getDataLayout().getTypeAllocSize(STy);
-        PtrElementSize =
+        [[maybe_unused]] llvm::TypeSize StructSize =
+            CGM.getDataLayout().getTypeAllocSize(STy);
+        [[maybe_unused]] llvm::TypeSize PtrElementSize =
             CGM.getDataLayout().getTypeAllocSize(ConvertTypeForMem(Ty));
         if (STy->containsHomogeneousScalableVectorTypes()) {
           assert(StructSize == PtrElementSize &&
@@ -5310,12 +5309,12 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
 
       llvm::StructType *STy =
           dyn_cast<llvm::StructType>(ArgInfo.getCoerceToType());
-      llvm::Type *SrcTy = ConvertTypeForMem(I->Ty);
-      llvm::TypeSize SrcTypeSize;
-      llvm::TypeSize DstTypeSize;
       if (STy && ArgInfo.isDirect() && !ArgInfo.getCanBeFlattened()) {
-        SrcTypeSize = CGM.getDataLayout().getTypeAllocSize(SrcTy);
-        DstTypeSize = CGM.getDataLayout().getTypeAllocSize(STy);
+        llvm::Type *SrcTy = ConvertTypeForMem(I->Ty);
+        [[maybe_unused]] llvm::TypeSize SrcTypeSize =
+            CGM.getDataLayout().getTypeAllocSize(SrcTy);
+        [[maybe_unused]] llvm::TypeSize DstTypeSize =
+            CGM.getDataLayout().getTypeAllocSize(STy);
         if (STy->containsHomogeneousScalableVectorTypes()) {
           assert(SrcTypeSize == DstTypeSize &&
                  "Only allow non-fractional movement of structure with "

diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -7023,31 +7023,47 @@ void CodeGenFunction::EmitOMPInteropDirective(const OMPInteropDirective &S) {
                                      S.getSingleClause<OMPUseClause>())) &&
          "OMPNowaitClause clause is used separately in OMPInteropDirective.");
 
-  if (const auto *C = S.getSingleClause<OMPInitClause>()) {
-    llvm::Value *InteropvarPtr =
-        EmitLValue(C->getInteropVar()).getPointer(*this);
-    llvm::omp::OMPInteropType InteropType = llvm::omp::OMPInteropType::Unknown;
-    if (C->getIsTarget()) {
-      InteropType = llvm::omp::OMPInteropType::Target;
-    } else {
-      assert(C->getIsTargetSync() && "Expected interop-type target/targetsync");
-      InteropType = llvm::omp::OMPInteropType::TargetSync;
+  auto ItOMPInitClause = S.getClausesOfKind<OMPInitClause>();
+  if (!ItOMPInitClause.empty()) {
+    // Look at the multiple init clauses
+    for (const OMPInitClause *C : ItOMPInitClause) {
+      llvm::Value *InteropvarPtr =
+          EmitLValue(C->getInteropVar()).getPointer(*this);
+      llvm::omp::OMPInteropType InteropType =
+          llvm::omp::OMPInteropType::Unknown;
+      if (C->getIsTarget()) {
+        InteropType = llvm::omp::OMPInteropType::Target;
+      } else {
+        assert(C->getIsTargetSync() &&
+               "Expected interop-type target/targetsync");
+        InteropType = llvm::omp::OMPInteropType::TargetSync;
+      }
+      OMPBuilder.createOMPInteropInit(Builder, InteropvarPtr, InteropType,
+                                      Device, NumDependences, DependenceList,
+                                      Data.HasNowaitClause);
+    }
+  }
+  auto ItOMPDestroyClause = S.getClausesOfKind<OMPDestroyClause>();
+  if (!ItOMPDestroyClause.empty()) {
+    // Look at the multiple destroy clauses
+    for (const OMPDestroyClause *C : ItOMPDestroyClause) {
+      llvm::Value *InteropvarPtr =
+          EmitLValue(C->getInteropVar()).getPointer(*this);
+      OMPBuilder.createOMPInteropDestroy(Builder, InteropvarPtr, Device,
+                                         NumDependences, DependenceList,
+                                         Data.HasNowaitClause);
+    }
+  }
+  auto ItOMPUseClause = S.getClausesOfKind<OMPUseClause>();
+  if (!ItOMPUseClause.empty()) {
+    // Look at the multiple use clauses
+    for (const OMPUseClause *C : ItOMPUseClause) {
+      llvm::Value *InteropvarPtr =
+          EmitLValue(C->getInteropVar()).getPointer(*this);
+      OMPBuilder.createOMPInteropUse(Builder, InteropvarPtr, Device,
+                                     NumDependences, DependenceList,
+                                     Data.HasNowaitClause);
     }
-    OMPBuilder.createOMPInteropInit(Builder, InteropvarPtr, InteropType, Device,
-                                    NumDependences, DependenceList,
-                                    Data.HasNowaitClause);
-  } else if (const auto *C = S.getSingleClause<OMPDestroyClause>()) {
-    llvm::Value *InteropvarPtr =
-        EmitLValue(C->getInteropVar()).getPointer(*this);
-    OMPBuilder.createOMPInteropDestroy(Builder, InteropvarPtr, Device,
-                                       NumDependences, DependenceList,
-                                       Data.HasNowaitClause);
-  } else if (const auto *C = S.getSingleClause<OMPUseClause>()) {
-    llvm::Value *InteropvarPtr =
-        EmitLValue(C->getInteropVar()).getPointer(*this);
-    OMPBuilder.createOMPInteropUse(Builder, InteropvarPtr, Device,
-                                   NumDependences, DependenceList,
-                                   Data.HasNowaitClause);
   }
 }
 

diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -397,8 +397,8 @@ CodeGenModule::CodeGenModule(ASTContext &C,
   // Enable TBAA unless it's suppressed. ThreadSanitizer needs TBAA even at O0.
   if (LangOpts.Sanitize.has(SanitizerKind::Thread) ||
       (!CodeGenOpts.RelaxedAliasing && CodeGenOpts.OptimizationLevel > 0))
-    TBAA.reset(new CodeGenTBAA(Context, TheModule, CodeGenOpts, getLangOpts(),
-                               getCXXABI().getMangleContext()));
+    TBAA.reset(new CodeGenTBAA(Context, getTypes(), TheModule, CodeGenOpts,
+                               getLangOpts(), getCXXABI().getMangleContext()));
 
   // If debug info or coverage generation is enabled, create the CGDebugInfo
   // object.
@@ -1727,59 +1727,6 @@ static void AppendCPUSpecificCPUDispatchMangling(const CodeGenModule &CGM,
     Out << ".resolver";
 }
 
-static void AppendTargetVersionMangling(const CodeGenModule &CGM,
-                                        const TargetVersionAttr *Attr,
-                                        raw_ostream &Out) {
-  if (Attr->isDefaultVersion()) {
-    Out << ".default";
-    return;
-  }
-  Out << "._";
-  const TargetInfo &TI = CGM.getTarget();
-  llvm::SmallVector<StringRef, 8> Feats;
-  Attr->getFeatures(Feats);
-  llvm::stable_sort(Feats, [&TI](const StringRef FeatL, const StringRef FeatR) {
-    return TI.multiVersionSortPriority(FeatL) <
-           TI.multiVersionSortPriority(FeatR);
-  });
-  for (const auto &Feat : Feats) {
-    Out << 'M';
-    Out << Feat;
-  }
-}
-
-static void AppendTargetMangling(const CodeGenModule &CGM,
-                                 const TargetAttr *Attr, raw_ostream &Out) {
-  if (Attr->isDefaultVersion())
-    return;
-
-  Out << '.';
-  const TargetInfo &Target = CGM.getTarget();
-  ParsedTargetAttr Info = Target.parseTargetAttr(Attr->getFeaturesStr());
-  llvm::sort(Info.Features, [&Target](StringRef LHS, StringRef RHS) {
-    // Multiversioning doesn't allow "no-${feature}", so we can
-    // only have "+" prefixes here.
-    assert(LHS.starts_with("+") && RHS.starts_with("+") &&
-           "Features should always have a prefix.");
-    return Target.multiVersionSortPriority(LHS.substr(1)) >
-           Target.multiVersionSortPriority(RHS.substr(1));
-  });
-
-  bool IsFirst = true;
-
-  if (!Info.CPU.empty()) {
-    IsFirst = false;
-    Out << "arch_" << Info.CPU;
-  }
-
-  for (StringRef Feat : Info.Features) {
-    if (!IsFirst)
-      Out << '_';
-    IsFirst = false;
-    Out << Feat.substr(1);
-  }
-}
-
 // Returns true if GD is a function decl with internal linkage and
 // needs a unique suffix after the mangled name.
 static bool isUniqueInternalLinkageDecl(GlobalDecl GD,
@@ -1789,41 +1736,6 @@ static bool isUniqueInternalLinkageDecl(GlobalDecl GD,
          (CGM.getFunctionLinkage(GD) == llvm::GlobalValue::InternalLinkage);
 }
 
-static void AppendTargetClonesMangling(const CodeGenModule &CGM,
-                                       const TargetClonesAttr *Attr,
-                                       unsigned VersionIndex,
-                                       raw_ostream &Out) {
-  const TargetInfo &TI = CGM.getTarget();
-  if (TI.getTriple().isAArch64()) {
-    StringRef FeatureStr = Attr->getFeatureStr(VersionIndex);
-    if (FeatureStr == "default") {
-      Out << ".default";
-      return;
-    }
-    Out << "._";
-    SmallVector<StringRef, 8> Features;
-    FeatureStr.split(Features, "+");
-    llvm::stable_sort(Features,
-                      [&TI](const StringRef FeatL, const StringRef FeatR) {
-                        return TI.multiVersionSortPriority(FeatL) <
-                               TI.multiVersionSortPriority(FeatR);
-                      });
-    for (auto &Feat : Features) {
-      Out << 'M';
-      Out << Feat;
-    }
-  } else {
-    Out << '.';
-    StringRef FeatureStr = Attr->getFeatureStr(VersionIndex);
-    if (FeatureStr.starts_with("arch="))
-      Out << "arch_" << FeatureStr.substr(sizeof("arch=") - 1);
-    else
-      Out << FeatureStr;
-
-    Out << '.' << Attr->getMangledIndex(VersionIndex);
-  }
-}
-
 static std::string getMangledNameImpl(CodeGenModule &CGM, GlobalDecl GD,
                                       const NamedDecl *ND,
                                       bool OmitMultiVersionMangling = false) {
@@ -1877,16 +1789,25 @@ static std::string getMangledNameImpl(CodeGenModule &CGM, GlobalDecl GD,
                                              FD->getAttr<CPUSpecificAttr>(),
                                              GD.getMultiVersionIndex(), Out);
         break;
-      case MultiVersionKind::Target:
-        AppendTargetMangling(CGM, FD->getAttr<TargetAttr>(), Out);
+      case MultiVersionKind::Target: {
+        auto *Attr = FD->getAttr<TargetAttr>();
+        const ABIInfo &Info = CGM.getTargetCodeGenInfo().getABIInfo();
+        Info.appendAttributeMangling(Attr, Out);
         break;
-      case MultiVersionKind::TargetVersion:
-        AppendTargetVersionMangling(CGM, FD->getAttr<TargetVersionAttr>(), Out);
+      }
+      case MultiVersionKind::TargetVersion: {
+        auto *Attr = FD->getAttr<TargetVersionAttr>();
+        const ABIInfo &Info = CGM.getTargetCodeGenInfo().getABIInfo();
+        Info.appendAttributeMangling(Attr, Out);
         break;
-      case MultiVersionKind::TargetClones:
-        AppendTargetClonesMangling(CGM, FD->getAttr<TargetClonesAttr>(),
-                                   GD.getMultiVersionIndex(), Out);
+      }
+      case MultiVersionKind::TargetClones: {
+        auto *Attr = FD->getAttr<TargetClonesAttr>();
+        unsigned Index = GD.getMultiVersionIndex();
+        const ABIInfo &Info = CGM.getTargetCodeGenInfo().getABIInfo();
+        Info.appendAttributeMangling(Attr, Index, Out);
         break;
+      }
       case MultiVersionKind::None:
         llvm_unreachable("None multiversion type isn't valid here");
       }

diff --git a/clang/lib/CodeGen/CodeGenTBAA.cpp b/clang/lib/CodeGen/CodeGenTBAA.cpp
@@ -15,6 +15,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "CodeGenTBAA.h"
+#include "CGRecordLayout.h"
+#include "CodeGenTypes.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/AST/Attr.h"
 #include "clang/AST/Mangle.h"
@@ -26,16 +28,16 @@
 #include "llvm/IR/Metadata.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/Type.h"
+#include "llvm/Support/Debug.h"
 using namespace clang;
 using namespace CodeGen;
 
-CodeGenTBAA::CodeGenTBAA(ASTContext &Ctx, llvm::Module &M,
-                         const CodeGenOptions &CGO,
+CodeGenTBAA::CodeGenTBAA(ASTContext &Ctx, CodeGenTypes &CGTypes,
+                         llvm::Module &M, const CodeGenOptions &CGO,
                          const LangOptions &Features, MangleContext &MContext)
-  : Context(Ctx), Module(M), CodeGenOpts(CGO),
-    Features(Features), MContext(MContext), MDHelper(M.getContext()),
-    Root(nullptr), Char(nullptr)
-{}
+    : Context(Ctx), CGTypes(CGTypes), Module(M), CodeGenOpts(CGO),
+      Features(Features), MContext(MContext), MDHelper(M.getContext()),
+      Root(nullptr), Char(nullptr) {}
 
 CodeGenTBAA::~CodeGenTBAA() {
 }
@@ -294,14 +296,34 @@ CodeGenTBAA::CollectFields(uint64_t BaseOffset,
         return false;
 
     const ASTRecordLayout &Layout = Context.getASTRecordLayout(RD);
+    const CGRecordLayout &CGRL = CGTypes.getCGRecordLayout(RD);
 
     unsigned idx = 0;
-    for (RecordDecl::field_iterator i = RD->field_begin(),
-         e = RD->field_end(); i != e; ++i, ++idx) {
-      if ((*i)->isZeroSize(Context) || (*i)->isUnnamedBitfield())
+    for (RecordDecl::field_iterator i = RD->field_begin(), e = RD->field_end();
+         i != e; ++i, ++idx) {
+      if ((*i)->isZeroSize(Context))
         continue;
-      uint64_t Offset = BaseOffset +
-                        Layout.getFieldOffset(idx) / Context.getCharWidth();
+
+      uint64_t Offset =
+          BaseOffset + Layout.getFieldOffset(idx) / Context.getCharWidth();
+
+      // Create a single field for consecutive named bitfields using char as
+      // base type.
+      if ((*i)->isBitField()) {
+        const CGBitFieldInfo &Info = CGRL.getBitFieldInfo(*i);
+        if (Info.Offset != 0)
+          continue;
+        unsigned CurrentBitFieldSize = Info.StorageSize;
+        uint64_t Size =
+            llvm::divideCeil(CurrentBitFieldSize, Context.getCharWidth());
+        llvm::MDNode *TBAAType = getChar();
+        llvm::MDNode *TBAATag =
+            getAccessTagInfo(TBAAAccessInfo(TBAAType, Size));
+        Fields.push_back(
+            llvm::MDBuilder::TBAAStructField(Offset, Size, TBAATag));
+        continue;
+      }
+
       QualType FieldQTy = i->getType();
       if (!CollectFields(Offset, FieldQTy, Fields,
                          MayAlias || TypeHasMayAlias(FieldQTy)))

diff --git a/clang/lib/CodeGen/CodeGenTBAA.h b/clang/lib/CodeGen/CodeGenTBAA.h
@@ -29,6 +29,7 @@ namespace clang {
   class Type;
 
 namespace CodeGen {
+class CodeGenTypes;
 
 // TBAAAccessKind - A kind of TBAA memory access descriptor.
 enum class TBAAAccessKind : unsigned {
@@ -115,6 +116,7 @@ struct TBAAAccessInfo {
 /// while lowering AST types to LLVM types.
 class CodeGenTBAA {
   ASTContext &Context;
+  CodeGenTypes &CGTypes;
   llvm::Module &Module;
   const CodeGenOptions &CodeGenOpts;
   const LangOptions &Features;
@@ -167,8 +169,9 @@ class CodeGenTBAA {
   llvm::MDNode *getBaseTypeInfoHelper(const Type *Ty);
 
 public:
-  CodeGenTBAA(ASTContext &Ctx, llvm::Module &M, const CodeGenOptions &CGO,
-              const LangOptions &Features, MangleContext &MContext);
+  CodeGenTBAA(ASTContext &Ctx, CodeGenTypes &CGTypes, llvm::Module &M,
+              const CodeGenOptions &CGO, const LangOptions &Features,
+              MangleContext &MContext);
   ~CodeGenTBAA();
 
   /// getTypeInfo - Get metadata used to describe accesses to objects of the

diff --git a/clang/lib/CodeGen/Targets/AArch64.cpp b/clang/lib/CodeGen/Targets/AArch64.cpp
@@ -9,6 +9,7 @@
 #include "ABIInfoImpl.h"
 #include "TargetInfo.h"
 #include "clang/Basic/DiagnosticFrontend.h"
+#include "llvm/TargetParser/AArch64TargetParser.h"
 
 using namespace clang;
 using namespace clang::CodeGen;
@@ -75,6 +76,12 @@ class AArch64ABIInfo : public ABIInfo {
   bool allowBFloatArgsAndRet() const override {
     return getTarget().hasBFloat16Type();
   }
+
+  using ABIInfo::appendAttributeMangling;
+  void appendAttributeMangling(TargetClonesAttr *Attr, unsigned Index,
+                               raw_ostream &Out) const override;
+  void appendAttributeMangling(StringRef AttrStr,
+                               raw_ostream &Out) const override;
 };
 
 class AArch64SwiftABIInfo : public SwiftABIInfo {
@@ -125,8 +132,7 @@ class AArch64TargetCodeGenInfo : public TargetCodeGenInfo {
     assert(Error.empty());
 
     auto *Fn = cast<llvm::Function>(GV);
-    static const char *SignReturnAddrStr[] = {"none", "non-leaf", "all"};
-    Fn->addFnAttr("sign-return-address", SignReturnAddrStr[static_cast<int>(BPI.SignReturnAddr)]);
+    Fn->addFnAttr("sign-return-address", BPI.getSignReturnAddrStr());
 
     if (BPI.SignReturnAddr != LangOptions::SignReturnAddressScopeKind::None) {
       Fn->addFnAttr("sign-return-address-key",
@@ -857,6 +863,34 @@ void AArch64TargetCodeGenInfo::checkFunctionCallABI(
           << Callee->getDeclName();
 }
 
+void AArch64ABIInfo::appendAttributeMangling(TargetClonesAttr *Attr,
+                                             unsigned Index,
+                                             raw_ostream &Out) const {
+  appendAttributeMangling(Attr->getFeatureStr(Index), Out);
+}
+
+void AArch64ABIInfo::appendAttributeMangling(StringRef AttrStr,
+                                             raw_ostream &Out) const {
+  if (AttrStr == "default") {
+    Out << ".default";
+    return;
+  }
+
+  Out << "._";
+  SmallVector<StringRef, 8> Features;
+  AttrStr.split(Features, "+");
+  for (auto &Feat : Features)
+    Feat = Feat.trim();
+
+  llvm::sort(Features, [](const StringRef LHS, const StringRef RHS) {
+    return LHS.compare(RHS) < 0;
+  });
+
+  for (auto &Feat : Features)
+    if (auto Ext = llvm::AArch64::parseArchExtension(Feat))
+      Out << 'M' << Ext->Name;
+}
+
 std::unique_ptr<TargetCodeGenInfo>
 CodeGen::createAArch64TargetCodeGenInfo(CodeGenModule &CGM,
                                         AArch64ABIKind Kind) {

diff --git a/clang/lib/CodeGen/Targets/ARM.cpp b/clang/lib/CodeGen/Targets/ARM.cpp
@@ -152,13 +152,7 @@ class ARMTargetCodeGenInfo : public TargetCodeGenInfo {
               diag::warn_target_unsupported_branch_protection_attribute)
               << Arch;
         } else {
-          static const char *SignReturnAddrStr[] = {"none", "non-leaf", "all"};
-          assert(static_cast<unsigned>(BPI.SignReturnAddr) <= 2 &&
-                 "Unexpected SignReturnAddressScopeKind");
-          Fn->addFnAttr(
-              "sign-return-address",
-              SignReturnAddrStr[static_cast<int>(BPI.SignReturnAddr)]);
-
+          Fn->addFnAttr("sign-return-address", BPI.getSignReturnAddrStr());
           Fn->addFnAttr("branch-target-enforcement",
                         BPI.BranchTargetEnforcement ? "true" : "false");
         }

diff --git a/clang/lib/Driver/ToolChains/Arch/AArch64.cpp b/clang/lib/Driver/ToolChains/Arch/AArch64.cpp
@@ -165,11 +165,14 @@ getAArch64MicroArchFeaturesFromMtune(const Driver &D, StringRef Mtune,
   // Handle CPU name is 'native'.
   if (MtuneLowerCase == "native")
     MtuneLowerCase = std::string(llvm::sys::getHostCPUName());
+
+  // 'cyclone' and later have zero-cycle register moves and zeroing.
   if (MtuneLowerCase == "cyclone" ||
       StringRef(MtuneLowerCase).starts_with("apple")) {
     Features.push_back("+zcm");
     Features.push_back("+zcz");
   }
+
   return true;
 }
 

diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -4479,10 +4479,9 @@ renderDebugOptions(const ToolChain &TC, const Driver &D, const llvm::Triple &T,
                       options::OPT_gpubnames, options::OPT_gno_pubnames);
   if (DwarfFission != DwarfFissionKind::None ||
       (PubnamesArg && checkDebugInfoOption(PubnamesArg, Args, D, TC)))
-    if (DebuggerTuning != llvm::DebuggerKind::LLDB &&
-        (!PubnamesArg ||
-         (!PubnamesArg->getOption().matches(options::OPT_gno_gnu_pubnames) &&
-          !PubnamesArg->getOption().matches(options::OPT_gno_pubnames))))
+    if (!PubnamesArg ||
+        (!PubnamesArg->getOption().matches(options::OPT_gno_gnu_pubnames) &&
+         !PubnamesArg->getOption().matches(options::OPT_gno_pubnames)))
       CmdArgs.push_back(PubnamesArg && PubnamesArg->getOption().matches(
                                            options::OPT_gpubnames)
                             ? "-gpubnames"
@@ -5959,7 +5958,7 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
 
   if (Arg *A = Args.getLastArg(options::OPT_fbasic_block_address_map,
                                options::OPT_fno_basic_block_address_map)) {
-    if (Triple.isX86() && Triple.isOSBinFormatELF()) {
+    if ((Triple.isX86() || Triple.isAArch64()) && Triple.isOSBinFormatELF()) {
       if (A->getOption().matches(options::OPT_fbasic_block_address_map))
         A->render(Args, CmdArgs);
     } else {

diff --git a/clang/lib/Driver/ToolChains/CommonArgs.cpp b/clang/lib/Driver/ToolChains/CommonArgs.cpp
@@ -2763,14 +2763,10 @@ void tools::addOpenMPDeviceRTL(const Driver &D,
                                const llvm::opt::ArgList &DriverArgs,
                                llvm::opt::ArgStringList &CC1Args,
                                StringRef BitcodeSuffix,
-                               const llvm::Triple &Triple) {
+                               const llvm::Triple &Triple,
+                               const ToolChain &HostTC) {
   SmallVector<StringRef, 8> LibraryPaths;
 
-  // Add path to clang lib / lib64 folder.
-  SmallString<256> DefaultLibPath = llvm::sys::path::parent_path(D.Dir);
-  llvm::sys::path::append(DefaultLibPath, CLANG_INSTALL_LIBDIR_BASENAME);
-  LibraryPaths.emplace_back(DefaultLibPath.c_str());
-
   // Add user defined library paths from LIBRARY_PATH.
   std::optional<std::string> LibPath =
       llvm::sys::Process::GetEnv("LIBRARY_PATH");
@@ -2782,6 +2778,10 @@ void tools::addOpenMPDeviceRTL(const Driver &D,
       LibraryPaths.emplace_back(Path.trim());
   }
 
+  // Check all of the standard library search paths used by the compiler.
+  for (const auto &LibPath : HostTC.getFilePaths())
+    LibraryPaths.emplace_back(LibPath);
+
   OptSpecifier LibomptargetBCPathOpt =
       Triple.isAMDGCN() ? options::OPT_libomptarget_amdgpu_bc_path_EQ
                         : options::OPT_libomptarget_nvptx_bc_path_EQ;

diff --git a/clang/lib/Driver/ToolChains/CommonArgs.h b/clang/lib/Driver/ToolChains/CommonArgs.h
@@ -214,7 +214,8 @@ void addMachineOutlinerArgs(const Driver &D, const llvm::opt::ArgList &Args,
 
 void addOpenMPDeviceRTL(const Driver &D, const llvm::opt::ArgList &DriverArgs,
                         llvm::opt::ArgStringList &CC1Args,
-                        StringRef BitcodeSuffix, const llvm::Triple &Triple);
+                        StringRef BitcodeSuffix, const llvm::Triple &Triple,
+                        const ToolChain &HostTC);
 
 void addOutlineAtomicsArgs(const Driver &D, const ToolChain &TC,
                            const llvm::opt::ArgList &Args,

diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -903,7 +903,7 @@ void CudaToolChain::addClangTargetOptions(
       return;
 
     addOpenMPDeviceRTL(getDriver(), DriverArgs, CC1Args, GpuArch.str(),
-                       getTriple());
+                       getTriple(), HostTC);
   }
 }
 

diff --git a/clang/lib/Driver/ToolChains/Linux.cpp b/clang/lib/Driver/ToolChains/Linux.cpp
@@ -237,11 +237,18 @@ Linux::Linux(const Driver &D, const llvm::Triple &Triple, const ArgList &Args)
     ExtraOpts.push_back("relro");
   }
 
-  // Android ARM/AArch64 use max-page-size=4096 to reduce VMA usage. Note, lld
-  // from 11 onwards default max-page-size to 65536 for both ARM and AArch64.
-  if ((Triple.isARM() || Triple.isAArch64()) && Triple.isAndroid()) {
-    ExtraOpts.push_back("-z");
-    ExtraOpts.push_back("max-page-size=4096");
+  // Note, lld from 11 onwards default max-page-size to 65536 for both ARM and
+  // AArch64.
+  if (Triple.isAndroid()) {
+    if (Triple.isARM()) {
+      // Android ARM uses max-page-size=4096 to reduce VMA usage.
+      ExtraOpts.push_back("-z");
+      ExtraOpts.push_back("max-page-size=4096");
+    } else if (Triple.isAArch64()) {
+      // Android AArch64 uses max-page-size=16384 to support 4k/16k page sizes.
+      ExtraOpts.push_back("-z");
+      ExtraOpts.push_back("max-page-size=16384");
+    }
   }
 
   if (GCCInstallation.getParentLibPath().contains("opt/rh/"))

diff --git a/clang/lib/Frontend/CompilerInstance.cpp b/clang/lib/Frontend/CompilerInstance.cpp
@@ -1061,30 +1061,7 @@ bool CompilerInstance::ExecuteAction(FrontendAction &Act) {
     }
   }
 
-  if (getDiagnosticOpts().ShowCarets) {
-    // We can have multiple diagnostics sharing one diagnostic client.
-    // Get the total number of warnings/errors from the client.
-    unsigned NumWarnings = getDiagnostics().getClient()->getNumWarnings();
-    unsigned NumErrors = getDiagnostics().getClient()->getNumErrors();
-
-    if (NumWarnings)
-      OS << NumWarnings << " warning" << (NumWarnings == 1 ? "" : "s");
-    if (NumWarnings && NumErrors)
-      OS << " and ";
-    if (NumErrors)
-      OS << NumErrors << " error" << (NumErrors == 1 ? "" : "s");
-    if (NumWarnings || NumErrors) {
-      OS << " generated";
-      if (getLangOpts().CUDA) {
-        if (!getLangOpts().CUDAIsDevice) {
-          OS << " when compiling for host";
-        } else {
-          OS << " when compiling for " << getTargetOpts().CPU;
-        }
-      }
-      OS << ".\n";
-    }
-  }
+  printDiagnosticStats();
 
   if (getFrontendOpts().ShowStats) {
     if (hasFileManager()) {
@@ -1112,6 +1089,36 @@ bool CompilerInstance::ExecuteAction(FrontendAction &Act) {
   return !getDiagnostics().getClient()->getNumErrors();
 }
 
+void CompilerInstance::printDiagnosticStats() {
+  if (!getDiagnosticOpts().ShowCarets)
+    return;
+
+  raw_ostream &OS = getVerboseOutputStream();
+
+  // We can have multiple diagnostics sharing one diagnostic client.
+  // Get the total number of warnings/errors from the client.
+  unsigned NumWarnings = getDiagnostics().getClient()->getNumWarnings();
+  unsigned NumErrors = getDiagnostics().getClient()->getNumErrors();
+
+  if (NumWarnings)
+    OS << NumWarnings << " warning" << (NumWarnings == 1 ? "" : "s");
+  if (NumWarnings && NumErrors)
+    OS << " and ";
+  if (NumErrors)
+    OS << NumErrors << " error" << (NumErrors == 1 ? "" : "s");
+  if (NumWarnings || NumErrors) {
+    OS << " generated";
+    if (getLangOpts().CUDA) {
+      if (!getLangOpts().CUDAIsDevice) {
+        OS << " when compiling for host";
+      } else {
+        OS << " when compiling for " << getTargetOpts().CPU;
+      }
+    }
+    OS << ".\n";
+  }
+}
+
 void CompilerInstance::LoadRequestedPlugins() {
   // Load any requested plugins.
   for (const std::string &Path : getFrontendOpts().Plugins) {

diff --git a/clang/lib/Frontend/CompilerInvocation.cpp b/clang/lib/Frontend/CompilerInvocation.cpp
@@ -191,6 +191,17 @@ CompilerInvocationBase::shallow_copy_assign(const CompilerInvocationBase &X) {
   return *this;
 }
 
+CompilerInvocation::CompilerInvocation(const CowCompilerInvocation &X)
+    : CompilerInvocationBase(EmptyConstructor{}) {
+  CompilerInvocationBase::deep_copy_assign(X);
+}
+
+CompilerInvocation &
+CompilerInvocation::operator=(const CowCompilerInvocation &X) {
+  CompilerInvocationBase::deep_copy_assign(X);
+  return *this;
+}
+
 namespace {
 template <typename T>
 T &ensureOwned(std::shared_ptr<T> &Storage) {
@@ -1975,14 +1986,6 @@ bool CompilerInvocation::ParseCodeGenArgs(CodeGenOptions &Opts, ArgList &Args,
     Opts.LinkBitcodeFiles.push_back(F);
   }
 
-  if (Arg *A = Args.getLastArg(OPT_ftlsmodel_EQ)) {
-    if (T.isOSAIX()) {
-      StringRef Name = A->getValue();
-      if (Name == "local-dynamic")
-        Diags.Report(diag::err_aix_unsupported_tls_model) << Name;
-    }
-  }
-
   if (Arg *A = Args.getLastArg(OPT_fdenormal_fp_math_EQ)) {
     StringRef Val = A->getValue();
     Opts.FPDenormalMode = llvm::parseDenormalFPAttribute(Val);

diff --git a/clang/lib/Headers/__clang_hip_math.h b/clang/lib/Headers/__clang_hip_math.h
@@ -1306,75 +1306,15 @@ float min(float __x, float __y) { return __builtin_fminf(__x, __y); }
 __DEVICE__
 double min(double __x, double __y) { return __builtin_fmin(__x, __y); }
 
-// Define host min/max functions.
-#if !defined(__HIPCC_RTC__) && !defined(__OPENMP_AMDGCN__) &&                  \
-    !defined(__HIP_NO_HOST_MIN_MAX_IN_GLOBAL_NAMESPACE__)
-
-#pragma push_macro("DEFINE_MIN_MAX_FUNCTIONS")
-#pragma push_macro("DEFINE_MIN_MAX_FUNCTIONS")
-#define DEFINE_MIN_MAX_FUNCTIONS(ret_type, type1, type2)                       \
-  inline ret_type min(const type1 __a, const type2 __b) {                      \
-    return (__a < __b) ? __a : __b;                                            \
-  }                                                                            \
-  inline ret_type max(const type1 __a, const type2 __b) {                      \
-    return (__a > __b) ? __a : __b;                                            \
-  }
-
-// Define min and max functions for same type comparisons
-DEFINE_MIN_MAX_FUNCTIONS(int, int, int)
-DEFINE_MIN_MAX_FUNCTIONS(unsigned int, unsigned int, unsigned int)
-DEFINE_MIN_MAX_FUNCTIONS(long, long, long)
-DEFINE_MIN_MAX_FUNCTIONS(unsigned long, unsigned long, unsigned long)
-DEFINE_MIN_MAX_FUNCTIONS(long long, long long, long long)
-DEFINE_MIN_MAX_FUNCTIONS(unsigned long long, unsigned long long,
-                         unsigned long long)
-
-// The host min/max functions below accept mixed signed/unsigned integer
-// parameters and perform unsigned comparisons, which may produce unexpected
-// results if a signed integer was passed unintentionally. To avoid this
-// happening silently, these overloaded functions are not defined by default.
-// However, for compatibility with CUDA, they will be defined if users define
-// __HIP_DEFINE_MIXED_HOST_MIN_MAX__.
-#ifdef __HIP_DEFINE_MIXED_HOST_MIN_MAX__
-DEFINE_MIN_MAX_FUNCTIONS(unsigned int, int, unsigned int)
-DEFINE_MIN_MAX_FUNCTIONS(unsigned int, unsigned int, int)
-DEFINE_MIN_MAX_FUNCTIONS(unsigned long, long, unsigned long)
-DEFINE_MIN_MAX_FUNCTIONS(unsigned long, unsigned long, long)
-DEFINE_MIN_MAX_FUNCTIONS(unsigned long long, long long, unsigned long long)
-DEFINE_MIN_MAX_FUNCTIONS(unsigned long long, unsigned long long, long long)
-#endif // ifdef __HIP_DEFINE_MIXED_HOST_MIN_MAX__
-
-// Floating-point comparisons using built-in functions
-inline float min(float const __a, float const __b) {
-  return __builtin_fminf(__a, __b);
-}
-inline double min(double const __a, double const __b) {
-  return __builtin_fmin(__a, __b);
-}
-inline double min(float const __a, double const __b) {
-  return __builtin_fmin(__a, __b);
-}
-inline double min(double const __a, float const __b) {
-  return __builtin_fmin(__a, __b);
+#if !defined(__HIPCC_RTC__) && !defined(__OPENMP_AMDGCN__)
+__host__ inline static int min(int __arg1, int __arg2) {
+  return __arg1 < __arg2 ? __arg1 : __arg2;
 }
 
-inline float max(float const __a, float const __b) {
-  return __builtin_fmaxf(__a, __b);
-}
-inline double max(double const __a, double const __b) {
-  return __builtin_fmax(__a, __b);
-}
-inline double max(float const __a, double const __b) {
-  return __builtin_fmax(__a, __b);
+__host__ inline static int max(int __arg1, int __arg2) {
+  return __arg1 > __arg2 ? __arg1 : __arg2;
 }
-inline double max(double const __a, float const __b) {
-  return __builtin_fmax(__a, __b);
-}
-
-#pragma pop_macro("DEFINE_MIN_MAX_FUNCTIONS")
-
-#endif // !defined(__HIPCC_RTC__) && !defined(__OPENMP_AMDGCN__) &&
-       // !defined(__HIP_NO_HOST_MIN_MAX_IN_GLOBAL_NAMESPACE__)
+#endif // !defined(__HIPCC_RTC__) && !defined(__OPENMP_AMDGCN__)
 #endif
 
 #pragma pop_macro("__DEVICE__")

diff --git a/clang/lib/Headers/emmintrin.h b/clang/lib/Headers/emmintrin.h
@@ -2099,9 +2099,11 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi64(__m128i __a,
 }
 
 /// Adds, with saturation, the corresponding elements of two 128-bit
-///    signed [16 x i8] vectors, saving each sum in the corresponding element of
-///    a 128-bit result vector of [16 x i8]. Positive sums greater than 0x7F are
-///    saturated to 0x7F. Negative sums less than 0x80 are saturated to 0x80.
+///    signed [16 x i8] vectors, saving each sum in the corresponding element
+///    of a 128-bit result vector of [16 x i8].
+///
+///    Positive sums greater than 0x7F are saturated to 0x7F. Negative sums
+///    less than 0x80 are saturated to 0x80.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -2119,10 +2121,11 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epi8(__m128i __a,
 }
 
 /// Adds, with saturation, the corresponding elements of two 128-bit
-///    signed [8 x i16] vectors, saving each sum in the corresponding element of
-///    a 128-bit result vector of [8 x i16]. Positive sums greater than 0x7FFF
-///    are saturated to 0x7FFF. Negative sums less than 0x8000 are saturated to
-///    0x8000.
+///    signed [8 x i16] vectors, saving each sum in the corresponding element
+///    of a 128-bit result vector of [8 x i16].
+///
+///    Positive sums greater than 0x7FFF are saturated to 0x7FFF. Negative sums
+///    less than 0x8000 are saturated to 0x8000.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -2141,8 +2144,10 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epi16(__m128i __a,
 
 /// Adds, with saturation, the corresponding elements of two 128-bit
 ///    unsigned [16 x i8] vectors, saving each sum in the corresponding element
-///    of a 128-bit result vector of [16 x i8]. Positive sums greater than 0xFF
-///    are saturated to 0xFF. Negative sums are saturated to 0x00.
+///    of a 128-bit result vector of [16 x i8].
+///
+///    Positive sums greater than 0xFF are saturated to 0xFF. Negative sums are
+///    saturated to 0x00.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -2161,8 +2166,10 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epu8(__m128i __a,
 
 /// Adds, with saturation, the corresponding elements of two 128-bit
 ///    unsigned [8 x i16] vectors, saving each sum in the corresponding element
-///    of a 128-bit result vector of [8 x i16]. Positive sums greater than
-///    0xFFFF are saturated to 0xFFFF. Negative sums are saturated to 0x0000.
+///    of a 128-bit result vector of [8 x i16].
+///
+///    Positive sums greater than 0xFFFF are saturated to 0xFFFF. Negative sums
+///    are saturated to 0x0000.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -2518,10 +2525,12 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi64(__m128i __a,
   return (__m128i)((__v2du)__a - (__v2du)__b);
 }
 
-/// Subtracts corresponding 8-bit signed integer values in the input and
-///    returns the differences in the corresponding bytes in the destination.
-///    Differences greater than 0x7F are saturated to 0x7F, and differences less
-///    than 0x80 are saturated to 0x80.
+/// Subtracts, with saturation, corresponding 8-bit signed integer values in
+///    the input and returns the differences in the corresponding bytes in the
+///    destination.
+///
+///    Differences greater than 0x7F are saturated to 0x7F, and differences
+///    less than 0x80 are saturated to 0x80.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -2538,8 +2547,10 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epi8(__m128i __a,
   return (__m128i)__builtin_elementwise_sub_sat((__v16qs)__a, (__v16qs)__b);
 }
 
-/// Subtracts corresponding 16-bit signed integer values in the input and
-///    returns the differences in the corresponding bytes in the destination.
+/// Subtracts, with saturation, corresponding 16-bit signed integer values in
+///    the input and returns the differences in the corresponding bytes in the
+///    destination.
+///
 ///    Differences greater than 0x7FFF are saturated to 0x7FFF, and values less
 ///    than 0x8000 are saturated to 0x8000.
 ///
@@ -2558,9 +2569,11 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epi16(__m128i __a,
   return (__m128i)__builtin_elementwise_sub_sat((__v8hi)__a, (__v8hi)__b);
 }
 
-/// Subtracts corresponding 8-bit unsigned integer values in the input
-///    and returns the differences in the corresponding bytes in the
-///    destination. Differences less than 0x00 are saturated to 0x00.
+/// Subtracts, with saturation, corresponding 8-bit unsigned integer values in
+///    the input and returns the differences in the corresponding bytes in the
+///    destination.
+///
+///    Differences less than 0x00 are saturated to 0x00.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -2577,9 +2590,11 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epu8(__m128i __a,
   return (__m128i)__builtin_elementwise_sub_sat((__v16qu)__a, (__v16qu)__b);
 }
 
-/// Subtracts corresponding 16-bit unsigned integer values in the input
-///    and returns the differences in the corresponding bytes in the
-///    destination. Differences less than 0x0000 are saturated to 0x0000.
+/// Subtracts, with saturation, corresponding 16-bit unsigned integer values in
+///    the input and returns the differences in the corresponding bytes in the
+///    destination.
+///
+///    Differences less than 0x0000 are saturated to 0x0000.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -4050,80 +4065,68 @@ void _mm_mfence(void);
 } // extern "C"
 #endif
 
-/// Converts 16-bit signed integers from both 128-bit integer vector
-///    operands into 8-bit signed integers, and packs the results into the
-///    destination. Positive values greater than 0x7F are saturated to 0x7F.
-///    Negative values less than 0x80 are saturated to 0x80.
+/// Converts, with saturation, 16-bit signed integers from both 128-bit integer
+///    vector operands into 8-bit signed integers, and packs the results into
+///    the destination.
+///
+///    Positive values greater than 0x7F are saturated to 0x7F. Negative values
+///    less than 0x80 are saturated to 0x80.
 ///
 /// \headerfile <x86intrin.h>
 ///
 /// This intrinsic corresponds to the <c> VPACKSSWB / PACKSSWB </c> instruction.
 ///
 /// \param __a
-///   A 128-bit integer vector of [8 x i16]. Each 16-bit element is treated as
-///   a signed integer and is converted to a 8-bit signed integer with
-///   saturation. Values greater than 0x7F are saturated to 0x7F. Values less
-///   than 0x80 are saturated to 0x80. The converted [8 x i8] values are
+///   A 128-bit integer vector of [8 x i16]. The converted [8 x i8] values are
 ///   written to the lower 64 bits of the result.
 /// \param __b
-///   A 128-bit integer vector of [8 x i16]. Each 16-bit element is treated as
-///   a signed integer and is converted to a 8-bit signed integer with
-///   saturation. Values greater than 0x7F are saturated to 0x7F. Values less
-///   than 0x80 are saturated to 0x80. The converted [8 x i8] values are
+///   A 128-bit integer vector of [8 x i16]. The converted [8 x i8] values are
 ///   written to the higher 64 bits of the result.
 /// \returns A 128-bit vector of [16 x i8] containing the converted values.
 static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_packs_epi16(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)__builtin_ia32_packsswb128((__v8hi)__a, (__v8hi)__b);
 }
 
-/// Converts 32-bit signed integers from both 128-bit integer vector
-///    operands into 16-bit signed integers, and packs the results into the
-///    destination. Positive values greater than 0x7FFF are saturated to 0x7FFF.
-///    Negative values less than 0x8000 are saturated to 0x8000.
+/// Converts, with saturation, 32-bit signed integers from both 128-bit integer
+///    vector operands into 16-bit signed integers, and packs the results into
+///    the destination.
+///
+///    Positive values greater than 0x7FFF are saturated to 0x7FFF. Negative
+///    values less than 0x8000 are saturated to 0x8000.
 ///
 /// \headerfile <x86intrin.h>
 ///
 /// This intrinsic corresponds to the <c> VPACKSSDW / PACKSSDW </c> instruction.
 ///
 /// \param __a
-///    A 128-bit integer vector of [4 x i32]. Each 32-bit element is treated as
-///    a signed integer and is converted to a 16-bit signed integer with
-///    saturation. Values greater than 0x7FFF are saturated to 0x7FFF. Values
-///    less than 0x8000 are saturated to 0x8000. The converted [4 x i16] values
+///    A 128-bit integer vector of [4 x i32]. The converted [4 x i16] values
 ///    are written to the lower 64 bits of the result.
 /// \param __b
-///    A 128-bit integer vector of [4 x i32]. Each 32-bit element is treated as
-///    a signed integer and is converted to a 16-bit signed integer with
-///    saturation. Values greater than 0x7FFF are saturated to 0x7FFF. Values
-///    less than 0x8000 are saturated to 0x8000. The converted [4 x i16] values
+///    A 128-bit integer vector of [4 x i32]. The converted [4 x i16] values
 ///    are written to the higher 64 bits of the result.
 /// \returns A 128-bit vector of [8 x i16] containing the converted values.
 static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_packs_epi32(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)__builtin_ia32_packssdw128((__v4si)__a, (__v4si)__b);
 }
 
-/// Converts 16-bit signed integers from both 128-bit integer vector
-///    operands into 8-bit unsigned integers, and packs the results into the
-///    destination. Values greater than 0xFF are saturated to 0xFF. Values less
-///    than 0x00 are saturated to 0x00.
+/// Converts, with saturation, 16-bit signed integers from both 128-bit integer
+///    vector operands into 8-bit unsigned integers, and packs the results into
+///    the destination.
+///
+///    Values greater than 0xFF are saturated to 0xFF. Values less than 0x00
+///    are saturated to 0x00.
 ///
 /// \headerfile <x86intrin.h>
 ///
 /// This intrinsic corresponds to the <c> VPACKUSWB / PACKUSWB </c> instruction.
 ///
 /// \param __a
-///    A 128-bit integer vector of [8 x i16]. Each 16-bit element is treated as
-///    a signed integer and is converted to an 8-bit unsigned integer with
-///    saturation. Values greater than 0xFF are saturated to 0xFF. Values less
-///    than 0x00 are saturated to 0x00. The converted [8 x i8] values are
+///    A 128-bit integer vector of [8 x i16]. The converted [8 x i8] values are
 ///    written to the lower 64 bits of the result.
 /// \param __b
-///    A 128-bit integer vector of [8 x i16]. Each 16-bit element is treated as
-///    a signed integer and is converted to an 8-bit unsigned integer with
-///    saturation. Values greater than 0xFF are saturated to 0xFF. Values less
-///    than 0x00 are saturated to 0x00. The converted [8 x i8] values are
+///    A 128-bit integer vector of [8 x i16]. The converted [8 x i8] values are
 ///    written to the higher 64 bits of the result.
 /// \returns A 128-bit vector of [16 x i8] containing the converted values.
 static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_packus_epi16(__m128i __a,

diff --git a/clang/lib/Headers/fmaintrin.h b/clang/lib/Headers/fmaintrin.h
@@ -60,7 +60,8 @@ _mm_fmadd_pd(__m128d __A, __m128d __B, __m128d __C)
 
 /// Computes a scalar multiply-add of the single-precision values in the
 ///    low 32 bits of 128-bit vectors of [4 x float].
-/// \code
+///
+/// \code{.operation}
 /// result[31:0] = (__A[31:0] * __B[31:0]) + __C[31:0]
 /// result[127:32] = __A[127:32]
 /// \endcode
@@ -88,7 +89,8 @@ _mm_fmadd_ss(__m128 __A, __m128 __B, __m128 __C)
 
 /// Computes a scalar multiply-add of the double-precision values in the
 ///    low 64 bits of 128-bit vectors of [2 x double].
-/// \code
+///
+/// \code{.operation}
 /// result[63:0] = (__A[63:0] * __B[63:0]) + __C[63:0]
 /// result[127:64] = __A[127:64]
 /// \endcode
@@ -156,7 +158,8 @@ _mm_fmsub_pd(__m128d __A, __m128d __B, __m128d __C)
 
 /// Computes a scalar multiply-subtract of the single-precision values in
 ///    the low 32 bits of 128-bit vectors of [4 x float].
-/// \code
+///
+/// \code{.operation}
 /// result[31:0] = (__A[31:0] * __B[31:0]) - __C[31:0]
 /// result[127:32] = __A[127:32]
 /// \endcode
@@ -184,7 +187,8 @@ _mm_fmsub_ss(__m128 __A, __m128 __B, __m128 __C)
 
 /// Computes a scalar multiply-subtract of the double-precision values in
 ///    the low 64 bits of 128-bit vectors of [2 x double].
-/// \code
+///
+/// \code{.operation}
 /// result[63:0] = (__A[63:0] * __B[63:0]) - __C[63:0]
 /// result[127:64] = __A[127:64]
 /// \endcode
@@ -252,7 +256,8 @@ _mm_fnmadd_pd(__m128d __A, __m128d __B, __m128d __C)
 
 /// Computes a scalar negated multiply-add of the single-precision values in
 ///    the low 32 bits of 128-bit vectors of [4 x float].
-/// \code
+///
+/// \code{.operation}
 /// result[31:0] = -(__A[31:0] * __B[31:0]) + __C[31:0]
 /// result[127:32] = __A[127:32]
 /// \endcode
@@ -280,7 +285,8 @@ _mm_fnmadd_ss(__m128 __A, __m128 __B, __m128 __C)
 
 /// Computes a scalar negated multiply-add of the double-precision values
 ///    in the low 64 bits of 128-bit vectors of [2 x double].
-/// \code
+///
+/// \code{.operation}
 /// result[63:0] = -(__A[63:0] * __B[63:0]) + __C[63:0]
 /// result[127:64] = __A[127:64]
 /// \endcode
@@ -348,7 +354,8 @@ _mm_fnmsub_pd(__m128d __A, __m128d __B, __m128d __C)
 
 /// Computes a scalar negated multiply-subtract of the single-precision
 ///    values in the low 32 bits of 128-bit vectors of [4 x float].
-/// \code
+///
+/// \code{.operation}
 /// result[31:0] = -(__A[31:0] * __B[31:0]) - __C[31:0]
 /// result[127:32] = __A[127:32]
 /// \endcode
@@ -376,7 +383,8 @@ _mm_fnmsub_ss(__m128 __A, __m128 __B, __m128 __C)
 
 /// Computes a scalar negated multiply-subtract of the double-precision
 ///    values in the low 64 bits of 128-bit vectors of [2 x double].
-/// \code
+///
+/// \code{.operation}
 /// result[63:0] = -(__A[63:0] * __B[63:0]) - __C[63:0]
 /// result[127:64] = __A[127:64]
 /// \endcode
@@ -404,7 +412,8 @@ _mm_fnmsub_sd(__m128d __A, __m128d __B, __m128d __C)
 
 /// Computes a multiply with alternating add/subtract of 128-bit vectors of
 ///    [4 x float].
-/// \code
+///
+/// \code{.operation}
 /// result[31:0]  = (__A[31:0] * __B[31:0]) - __C[31:0]
 /// result[63:32] = (__A[63:32] * __B[63:32]) + __C[63:32]
 /// result[95:64] = (__A[95:64] * __B[95:64]) - __C[95:64]
@@ -430,7 +439,8 @@ _mm_fmaddsub_ps(__m128 __A, __m128 __B, __m128 __C)
 
 /// Computes a multiply with alternating add/subtract of 128-bit vectors of
 ///    [2 x double].
-/// \code
+///
+/// \code{.operation}
 /// result[63:0]  = (__A[63:0] * __B[63:0]) - __C[63:0]
 /// result[127:64] = (__A[127:64] * __B[127:64]) + __C[127:64]
 /// \endcode
@@ -454,7 +464,8 @@ _mm_fmaddsub_pd(__m128d __A, __m128d __B, __m128d __C)
 
 /// Computes a multiply with alternating add/subtract of 128-bit vectors of
 ///    [4 x float].
-/// \code
+///
+/// \code{.operation}
 /// result[31:0]  = (__A[31:0] * __B[31:0]) + __C[31:0]
 /// result[63:32] = (__A[63:32] * __B[63:32]) - __C[63:32]
 /// result[95:64] = (__A[95:64] * __B[95:64]) + __C[95:64]
@@ -480,7 +491,8 @@ _mm_fmsubadd_ps(__m128 __A, __m128 __B, __m128 __C)
 
 /// Computes a multiply with alternating add/subtract of 128-bit vectors of
 ///    [2 x double].
-/// \code
+///
+/// \code{.operation}
 /// result[63:0]  = (__A[63:0] * __B[63:0]) + __C[63:0]
 /// result[127:64] = (__A[127:64] * __B[127:64]) - __C[127:64]
 /// \endcode
@@ -664,7 +676,8 @@ _mm256_fnmsub_pd(__m256d __A, __m256d __B, __m256d __C)
 
 /// Computes a multiply with alternating add/subtract of 256-bit vectors of
 ///    [8 x float].
-/// \code
+///
+/// \code{.operation}
 /// result[31:0] = (__A[31:0] * __B[31:0]) - __C[31:0]
 /// result[63:32] = (__A[63:32] * __B[63:32]) + __C[63:32]
 /// result[95:64] = (__A[95:64] * __B[95:64]) - __C[95:64]
@@ -694,7 +707,8 @@ _mm256_fmaddsub_ps(__m256 __A, __m256 __B, __m256 __C)
 
 /// Computes a multiply with alternating add/subtract of 256-bit vectors of
 ///    [4 x double].
-/// \code
+///
+/// \code{.operation}
 /// result[63:0] = (__A[63:0] * __B[63:0]) - __C[63:0]
 /// result[127:64] = (__A[127:64] * __B[127:64]) + __C[127:64]
 /// result[191:128] = (__A[191:128] * __B[191:128]) - __C[191:128]
@@ -720,7 +734,8 @@ _mm256_fmaddsub_pd(__m256d __A, __m256d __B, __m256d __C)
 
 /// Computes a vector multiply with alternating add/subtract of 256-bit
 ///    vectors of [8 x float].
-/// \code
+///
+/// \code{.operation}
 /// result[31:0] = (__A[31:0] * __B[31:0]) + __C[31:0]
 /// result[63:32] = (__A[63:32] * __B[63:32]) - __C[63:32]
 /// result[95:64] = (__A[95:64] * __B[95:64]) + __C[95:64]
@@ -750,7 +765,8 @@ _mm256_fmsubadd_ps(__m256 __A, __m256 __B, __m256 __C)
 
 /// Computes a vector multiply with alternating add/subtract of 256-bit
 ///    vectors of [4 x double].
-/// \code
+///
+/// \code{.operation}
 /// result[63:0] = (__A[63:0] * __B[63:0]) + __C[63:0]
 /// result[127:64] = (__A[127:64] * __B[127:64]) - __C[127:64]
 /// result[191:128] = (__A[191:128] * __B[191:128]) + __C[191:128]

diff --git a/clang/lib/Headers/hlsl/hlsl_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
@@ -317,6 +317,74 @@ double3 floor(double3);
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 double4 floor(double4);
 
+//===----------------------------------------------------------------------===//
+// frac builtins
+//===----------------------------------------------------------------------===//
+
+/// \fn T frac(T x)
+/// \brief Returns the fractional (or decimal) part of x. \a x parameter.
+/// \param x The specified input value.
+///
+/// If \a the return value is greater than or equal to 0 and less than 1.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_frac)
+half frac(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_frac)
+half2 frac(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_frac)
+half3 frac(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_frac)
+half4 frac(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_frac)
+float frac(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_frac)
+float2 frac(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_frac)
+float3 frac(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_frac)
+float4 frac(float4);
+
+//===----------------------------------------------------------------------===//
+// lerp builtins
+//===----------------------------------------------------------------------===//
+
+/// \fn T lerp(T x, T y, T s)
+/// \brief Returns the linear interpolation of x to y by s.
+/// \param x [in] The first-floating point value.
+/// \param y [in] The second-floating point value.
+/// \param s [in] A value that linearly interpolates between the x parameter and
+/// the y parameter.
+///
+/// Linear interpolation is based on the following formula: x*(1-s) + y*s which
+/// can equivalently be written as x + s(y-x).
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_lerp)
+half lerp(half, half, half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_lerp)
+half2 lerp(half2, half2, half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_lerp)
+half3 lerp(half3, half3, half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_lerp)
+half4 lerp(half4, half4, half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_lerp)
+float lerp(float, float, float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_lerp)
+float2 lerp(float2, float2, float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_lerp)
+float3 lerp(float3, float3, float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_lerp)
+float4 lerp(float4, float4, float4);
+
 //===----------------------------------------------------------------------===//
 // log builtins
 //===----------------------------------------------------------------------===//
@@ -763,6 +831,40 @@ uint64_t3 reversebits(uint64_t3);
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_bitreverse)
 uint64_t4 reversebits(uint64_t4);
 
+//===----------------------------------------------------------------------===//
+// round builtins
+//===----------------------------------------------------------------------===//
+
+/// \fn T round(T x)
+/// \brief Rounds the specified value \a x to the nearest integer.
+/// \param x The specified input value.
+///
+/// The return value is the \a x parameter, rounded to the nearest integer
+/// within a floating-point type. Halfway cases are
+/// rounded to the nearest even value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_round)
+half round(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_round)
+half2 round(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_round)
+half3 round(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_round)
+half4 round(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_round)
+float round(float);
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_round)
+float2 round(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_round)
+float3 round(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_round)
+float4 round(float4);
+
 //===----------------------------------------------------------------------===//
 // sin builtins
 //===----------------------------------------------------------------------===//

diff --git a/clang/lib/Headers/mmintrin.h b/clang/lib/Headers/mmintrin.h
@@ -105,28 +105,23 @@ _mm_cvtm64_si64(__m64 __m)
     return (long long)__m;
 }
 
-/// Converts 16-bit signed integers from both 64-bit integer vector
-///    parameters of [4 x i16] into 8-bit signed integer values, and constructs
-///    a 64-bit integer vector of [8 x i8] as the result. Positive values
-///    greater than 0x7F are saturated to 0x7F. Negative values less than 0x80
-///    are saturated to 0x80.
+/// Converts, with saturation, 16-bit signed integers from both 64-bit integer
+///    vector parameters of [4 x i16] into 8-bit signed integer values, and
+///    constructs a 64-bit integer vector of [8 x i8] as the result.
+///
+///    Positive values greater than 0x7F are saturated to 0x7F. Negative values
+///    less than 0x80 are saturated to 0x80.
 ///
 /// \headerfile <x86intrin.h>
 ///
 /// This intrinsic corresponds to the <c> PACKSSWB </c> instruction.
 ///
 /// \param __m1
-///    A 64-bit integer vector of [4 x i16]. Each 16-bit element is treated as a
-///    16-bit signed integer and is converted to an 8-bit signed integer with
-///    saturation. Positive values greater than 0x7F are saturated to 0x7F.
-///    Negative values less than 0x80 are saturated to 0x80. The converted
-///    [4 x i8] values are written to the lower 32 bits of the result.
+///    A 64-bit integer vector of [4 x i16]. The converted [4 x i8] values are
+///    written to the lower 32 bits of the result.
 /// \param __m2
-///    A 64-bit integer vector of [4 x i16]. Each 16-bit element is treated as a
-///    16-bit signed integer and is converted to an 8-bit signed integer with
-///    saturation. Positive values greater than 0x7F are saturated to 0x7F.
-///    Negative values less than 0x80 are saturated to 0x80. The converted
-///    [4 x i8] values are written to the upper 32 bits of the result.
+///    A 64-bit integer vector of [4 x i16]. The converted [4 x i8] values are
+///    written to the upper 32 bits of the result.
 /// \returns A 64-bit integer vector of [8 x i8] containing the converted
 ///    values.
 static __inline__ __m64 __DEFAULT_FN_ATTRS
@@ -135,28 +130,23 @@ _mm_packs_pi16(__m64 __m1, __m64 __m2)
     return (__m64)__builtin_ia32_packsswb((__v4hi)__m1, (__v4hi)__m2);
 }
 
-/// Converts 32-bit signed integers from both 64-bit integer vector
-///    parameters of [2 x i32] into 16-bit signed integer values, and constructs
-///    a 64-bit integer vector of [4 x i16] as the result. Positive values
-///    greater than 0x7FFF are saturated to 0x7FFF. Negative values less than
-///    0x8000 are saturated to 0x8000.
+/// Converts, with saturation, 32-bit signed integers from both 64-bit integer
+///    vector parameters of [2 x i32] into 16-bit signed integer values, and
+///    constructs a 64-bit integer vector of [4 x i16] as the result.
+///
+///    Positive values greater than 0x7FFF are saturated to 0x7FFF. Negative
+///    values less than 0x8000 are saturated to 0x8000.
 ///
 /// \headerfile <x86intrin.h>
 ///
 /// This intrinsic corresponds to the <c> PACKSSDW </c> instruction.
 ///
 /// \param __m1
-///    A 64-bit integer vector of [2 x i32]. Each 32-bit element is treated as a
-///    32-bit signed integer and is converted to a 16-bit signed integer with
-///    saturation. Positive values greater than 0x7FFF are saturated to 0x7FFF.
-///    Negative values less than 0x8000 are saturated to 0x8000. The converted
-///    [2 x i16] values are written to the lower 32 bits of the result.
+///    A 64-bit integer vector of [2 x i32]. The converted [2 x i16] values are
+///    written to the lower 32 bits of the result.
 /// \param __m2
-///    A 64-bit integer vector of [2 x i32]. Each 32-bit element is treated as a
-///    32-bit signed integer and is converted to a 16-bit signed integer with
-///    saturation. Positive values greater than 0x7FFF are saturated to 0x7FFF.
-///    Negative values less than 0x8000 are saturated to 0x8000. The converted
-///    [2 x i16] values are written to the upper 32 bits of the result.
+///    A 64-bit integer vector of [2 x i32]. The converted [2 x i16] values are
+///    written to the upper 32 bits of the result.
 /// \returns A 64-bit integer vector of [4 x i16] containing the converted
 ///    values.
 static __inline__ __m64 __DEFAULT_FN_ATTRS
@@ -165,28 +155,23 @@ _mm_packs_pi32(__m64 __m1, __m64 __m2)
     return (__m64)__builtin_ia32_packssdw((__v2si)__m1, (__v2si)__m2);
 }
 
-/// Converts 16-bit signed integers from both 64-bit integer vector
-///    parameters of [4 x i16] into 8-bit unsigned integer values, and
-///    constructs a 64-bit integer vector of [8 x i8] as the result. Values
-///    greater than 0xFF are saturated to 0xFF. Values less than 0 are saturated
-///    to 0.
+/// Converts, with saturation, 16-bit signed integers from both 64-bit integer
+///    vector parameters of [4 x i16] into 8-bit unsigned integer values, and
+///    constructs a 64-bit integer vector of [8 x i8] as the result.
+///
+///    Values greater than 0xFF are saturated to 0xFF. Values less than 0 are
+///    saturated to 0.
 ///
 /// \headerfile <x86intrin.h>
 ///
 /// This intrinsic corresponds to the <c> PACKUSWB </c> instruction.
 ///
 /// \param __m1
-///    A 64-bit integer vector of [4 x i16]. Each 16-bit element is treated as a
-///    16-bit signed integer and is converted to an 8-bit unsigned integer with
-///    saturation. Values greater than 0xFF are saturated to 0xFF. Values less
-///    than 0 are saturated to 0. The converted [4 x i8] values are written to
-///    the lower 32 bits of the result.
+///    A 64-bit integer vector of [4 x i16]. The converted [4 x i8] values are
+///    written to the lower 32 bits of the result.
 /// \param __m2
-///    A 64-bit integer vector of [4 x i16]. Each 16-bit element is treated as a
-///    16-bit signed integer and is converted to an 8-bit unsigned integer with
-///    saturation. Values greater than 0xFF are saturated to 0xFF. Values less
-///    than 0 are saturated to 0. The converted [4 x i8] values are written to
-///    the upper 32 bits of the result.
+///    A 64-bit integer vector of [4 x i16]. The converted [4 x i8] values are
+///    written to the upper 32 bits of the result.
 /// \returns A 64-bit integer vector of [8 x i8] containing the converted
 ///    values.
 static __inline__ __m64 __DEFAULT_FN_ATTRS
@@ -400,11 +385,13 @@ _mm_add_pi32(__m64 __m1, __m64 __m2)
     return (__m64)__builtin_ia32_paddd((__v2si)__m1, (__v2si)__m2);
 }
 
-/// Adds each 8-bit signed integer element of the first 64-bit integer
-///    vector of [8 x i8] to the corresponding 8-bit signed integer element of
-///    the second 64-bit integer vector of [8 x i8]. Positive sums greater than
-///    0x7F are saturated to 0x7F. Negative sums less than 0x80 are saturated to
-///    0x80. The results are packed into a 64-bit integer vector of [8 x i8].
+/// Adds, with saturation, each 8-bit signed integer element of the first
+///    64-bit integer vector of [8 x i8] to the corresponding 8-bit signed
+///    integer element of the second 64-bit integer vector of [8 x i8].
+///
+///    Positive sums greater than 0x7F are saturated to 0x7F. Negative sums
+///    less than 0x80 are saturated to 0x80. The results are packed into a
+///    64-bit integer vector of [8 x i8].
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -422,12 +409,13 @@ _mm_adds_pi8(__m64 __m1, __m64 __m2)
     return (__m64)__builtin_ia32_paddsb((__v8qi)__m1, (__v8qi)__m2);
 }
 
-/// Adds each 16-bit signed integer element of the first 64-bit integer
-///    vector of [4 x i16] to the corresponding 16-bit signed integer element of
-///    the second 64-bit integer vector of [4 x i16]. Positive sums greater than
-///    0x7FFF are saturated to 0x7FFF. Negative sums less than 0x8000 are
-///    saturated to 0x8000. The results are packed into a 64-bit integer vector
-///    of [4 x i16].
+/// Adds, with saturation, each 16-bit signed integer element of the first
+///    64-bit integer vector of [4 x i16] to the corresponding 16-bit signed
+///    integer element of the second 64-bit integer vector of [4 x i16].
+///
+///    Positive sums greater than 0x7FFF are saturated to 0x7FFF. Negative sums
+///    less than 0x8000 are saturated to 0x8000. The results are packed into a
+///    64-bit integer vector of [4 x i16].
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -445,11 +433,12 @@ _mm_adds_pi16(__m64 __m1, __m64 __m2)
     return (__m64)__builtin_ia32_paddsw((__v4hi)__m1, (__v4hi)__m2);
 }
 
-/// Adds each 8-bit unsigned integer element of the first 64-bit integer
-///    vector of [8 x i8] to the corresponding 8-bit unsigned integer element of
-///    the second 64-bit integer vector of [8 x i8]. Sums greater than 0xFF are
-///    saturated to 0xFF. The results are packed into a 64-bit integer vector of
-///    [8 x i8].
+/// Adds, with saturation, each 8-bit unsigned integer element of the first
+///    64-bit integer vector of [8 x i8] to the corresponding 8-bit unsigned
+///    integer element of the second 64-bit integer vector of [8 x i8].
+///
+///    Sums greater than 0xFF are saturated to 0xFF. The results are packed
+///    into a 64-bit integer vector of [8 x i8].
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -467,11 +456,12 @@ _mm_adds_pu8(__m64 __m1, __m64 __m2)
     return (__m64)__builtin_ia32_paddusb((__v8qi)__m1, (__v8qi)__m2);
 }
 
-/// Adds each 16-bit unsigned integer element of the first 64-bit integer
-///    vector of [4 x i16] to the corresponding 16-bit unsigned integer element
-///    of the second 64-bit integer vector of [4 x i16]. Sums greater than
-///    0xFFFF are saturated to 0xFFFF. The results are packed into a 64-bit
-///    integer vector of [4 x i16].
+/// Adds, with saturation, each 16-bit unsigned integer element of the first
+///    64-bit integer vector of [4 x i16] to the corresponding 16-bit unsigned
+///    integer element of the second 64-bit integer vector of [4 x i16].
+///
+///    Sums greater than 0xFFFF are saturated to 0xFFFF. The results are packed
+///    into a 64-bit integer vector of [4 x i16].
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -552,12 +542,13 @@ _mm_sub_pi32(__m64 __m1, __m64 __m2)
     return (__m64)__builtin_ia32_psubd((__v2si)__m1, (__v2si)__m2);
 }
 
-/// Subtracts each 8-bit signed integer element of the second 64-bit
-///    integer vector of [8 x i8] from the corresponding 8-bit signed integer
-///    element of the first 64-bit integer vector of [8 x i8]. Positive results
-///    greater than 0x7F are saturated to 0x7F. Negative results less than 0x80
-///    are saturated to 0x80. The results are packed into a 64-bit integer
-///    vector of [8 x i8].
+/// Subtracts, with saturation, each 8-bit signed integer element of the second
+///    64-bit integer vector of [8 x i8] from the corresponding 8-bit signed
+///    integer element of the first 64-bit integer vector of [8 x i8].
+///
+///    Positive results greater than 0x7F are saturated to 0x7F. Negative
+///    results less than 0x80 are saturated to 0x80. The results are packed
+///    into a 64-bit integer vector of [8 x i8].
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -575,12 +566,13 @@ _mm_subs_pi8(__m64 __m1, __m64 __m2)
     return (__m64)__builtin_ia32_psubsb((__v8qi)__m1, (__v8qi)__m2);
 }
 
-/// Subtracts each 16-bit signed integer element of the second 64-bit
-///    integer vector of [4 x i16] from the corresponding 16-bit signed integer
-///    element of the first 64-bit integer vector of [4 x i16]. Positive results
-///    greater than 0x7FFF are saturated to 0x7FFF. Negative results less than
-///    0x8000 are saturated to 0x8000. The results are packed into a 64-bit
-///    integer vector of [4 x i16].
+/// Subtracts, with saturation, each 16-bit signed integer element of the
+///    second 64-bit integer vector of [4 x i16] from the corresponding 16-bit
+///    signed integer element of the first 64-bit integer vector of [4 x i16].
+///
+///    Positive results greater than 0x7FFF are saturated to 0x7FFF. Negative
+///    results less than 0x8000 are saturated to 0x8000. The results are packed
+///    into a 64-bit integer vector of [4 x i16].
 ///
 /// \headerfile <x86intrin.h>
 ///

diff --git a/clang/lib/Headers/prfchwintrin.h b/clang/lib/Headers/prfchwintrin.h
@@ -15,9 +15,10 @@
 #define __PRFCHWINTRIN_H
 
 /// Loads a memory sequence containing the specified memory address into
-///    all data cache levels. The cache-coherency state is set to exclusive.
-///    Data can be read from and written to the cache line without additional
-///    delay.
+///    all data cache levels.
+///
+///    The cache-coherency state is set to exclusive. Data can be read from
+///    and written to the cache line without additional delay.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -32,10 +33,11 @@ _m_prefetch(void *__P)
 }
 
 /// Loads a memory sequence containing the specified memory address into
-///    the L1 data cache and sets the cache-coherency to modified. This
-///    provides a hint to the processor that the cache line will be modified.
-///    It is intended for use when the cache line will be written to shortly
-///    after the prefetch is performed.
+///    the L1 data cache and sets the cache-coherency state to modified.
+///
+///    This provides a hint to the processor that the cache line will be
+///    modified. It is intended for use when the cache line will be written to
+///    shortly after the prefetch is performed.
 ///
 ///    Note that the effect of this intrinsic is dependent on the processor
 ///    implementation.

diff --git a/clang/lib/Headers/smmintrin.h b/clang/lib/Headers/smmintrin.h
@@ -1431,8 +1431,10 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu32_epi64(__m128i __V) {
 }
 
 /* SSE4 Pack with Unsigned Saturation.  */
-/// Converts 32-bit signed integers from both 128-bit integer vector
-///    operands into 16-bit unsigned integers, and returns the packed result.
+/// Converts, with saturation, 32-bit signed integers from both 128-bit integer
+///    vector operands into 16-bit unsigned integers, and returns the packed
+///    result.
+///
 ///    Values greater than 0xFFFF are saturated to 0xFFFF. Values less than
 ///    0x0000 are saturated to 0x0000.
 ///
@@ -1441,17 +1443,11 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu32_epi64(__m128i __V) {
 /// This intrinsic corresponds to the <c> VPACKUSDW / PACKUSDW </c> instruction.
 ///
 /// \param __V1
-///    A 128-bit vector of [4 x i32]. Each 32-bit element is treated as a
-///    signed integer and is converted to a 16-bit unsigned integer with
-///    saturation. Values greater than 0xFFFF are saturated to 0xFFFF. Values
-///    less than 0x0000 are saturated to 0x0000. The converted [4 x i16] values
-///    are written to the lower 64 bits of the result.
+///    A 128-bit vector of [4 x i32]. The converted [4 x i16] values are
+///    written to the lower 64 bits of the result.
 /// \param __V2
-///    A 128-bit vector of [4 x i32]. Each 32-bit element is treated as a
-///    signed integer and is converted to a 16-bit unsigned integer with
-///    saturation. Values greater than 0xFFFF are saturated to 0xFFFF. Values
-///    less than 0x0000 are saturated to 0x0000. The converted [4 x i16] values
-///    are written to the higher 64 bits of the result.
+///    A 128-bit vector of [4 x i32]. The converted [4 x i16] values are
+///    written to the higher 64 bits of the result.
 /// \returns A 128-bit vector of [8 x i16] containing the converted values.
 static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_packus_epi32(__m128i __V1,
                                                               __m128i __V2) {

diff --git a/clang/lib/Headers/tmmintrin.h b/clang/lib/Headers/tmmintrin.h
@@ -271,10 +271,11 @@ _mm_hadd_pi32(__m64 __a, __m64 __b)
     return (__m64)__builtin_ia32_phaddd((__v2si)__a, (__v2si)__b);
 }
 
-/// Horizontally adds the adjacent pairs of values contained in 2 packed
-///    128-bit vectors of [8 x i16]. Positive sums greater than 0x7FFF are
-///    saturated to 0x7FFF. Negative sums less than 0x8000 are saturated to
-///    0x8000.
+/// Horizontally adds, with saturation, the adjacent pairs of values contained
+///    in two packed 128-bit vectors of [8 x i16].
+///
+///    Positive sums greater than 0x7FFF are saturated to 0x7FFF. Negative sums
+///    less than 0x8000 are saturated to 0x8000.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -296,10 +297,11 @@ _mm_hadds_epi16(__m128i __a, __m128i __b)
     return (__m128i)__builtin_ia32_phaddsw128((__v8hi)__a, (__v8hi)__b);
 }
 
-/// Horizontally adds the adjacent pairs of values contained in 2 packed
-///    64-bit vectors of [4 x i16]. Positive sums greater than 0x7FFF are
-///    saturated to 0x7FFF. Negative sums less than 0x8000 are saturated to
-///    0x8000.
+/// Horizontally adds, with saturation, the adjacent pairs of values contained
+///    in two packed 64-bit vectors of [4 x i16].
+///
+///    Positive sums greater than 0x7FFF are saturated to 0x7FFF. Negative sums
+///    less than 0x8000 are saturated to 0x8000.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -413,10 +415,11 @@ _mm_hsub_pi32(__m64 __a, __m64 __b)
     return (__m64)__builtin_ia32_phsubd((__v2si)__a, (__v2si)__b);
 }
 
-/// Horizontally subtracts the adjacent pairs of values contained in 2
-///    packed 128-bit vectors of [8 x i16]. Positive differences greater than
-///    0x7FFF are saturated to 0x7FFF. Negative differences less than 0x8000 are
-///    saturated to 0x8000.
+/// Horizontally subtracts, with saturation, the adjacent pairs of values
+///    contained in two packed 128-bit vectors of [8 x i16].
+///
+///    Positive differences greater than 0x7FFF are saturated to 0x7FFF.
+///    Negative differences less than 0x8000 are saturated to 0x8000.
 ///
 /// \headerfile <x86intrin.h>
 ///
@@ -438,10 +441,11 @@ _mm_hsubs_epi16(__m128i __a, __m128i __b)
     return (__m128i)__builtin_ia32_phsubsw128((__v8hi)__a, (__v8hi)__b);
 }
 
-/// Horizontally subtracts the adjacent pairs of values contained in 2
-///    packed 64-bit vectors of [4 x i16]. Positive differences greater than
-///    0x7FFF are saturated to 0x7FFF. Negative differences less than 0x8000 are
-///    saturated to 0x8000.
+/// Horizontally subtracts, with saturation, the adjacent pairs of values
+///    contained in two packed 64-bit vectors of [4 x i16].
+///
+///    Positive differences greater than 0x7FFF are saturated to 0x7FFF.
+///    Negative differences less than 0x8000 are saturated to 0x8000.
 ///
 /// \headerfile <x86intrin.h>
 ///

diff --git a/clang/lib/Index/IndexSymbol.cpp b/clang/lib/Index/IndexSymbol.cpp
@@ -552,8 +552,7 @@ StringRef index::getSymbolSubKindString(SymbolSubKind K) {
   case SymbolSubKind::AccessorSetter: return "acc-set";
   case SymbolSubKind::UsingTypename: return "using-typename";
   case SymbolSubKind::UsingValue: return "using-value";
-  case SymbolSubKind::UsingEnum:
-    return "using-enum";
+  case SymbolSubKind::UsingEnum: return "using-enum";
   }
   llvm_unreachable("invalid symbol subkind");
 }

diff --git a/clang/lib/Index/IndexingAction.cpp b/clang/lib/Index/IndexingAction.cpp
@@ -199,7 +199,7 @@ index::createIndexingAction(std::shared_ptr<IndexDataConsumer> DataConsumer,
 }
 
 static bool topLevelDeclVisitor(void *context, const Decl *D) {
-  IndexingContext &IndexCtx = *static_cast<IndexingContext*>(context);
+  IndexingContext &IndexCtx = *static_cast<IndexingContext *>(context);
   return IndexCtx.indexTopLevelDecl(D);
 }
 

diff --git a/clang/lib/InstallAPI/CMakeLists.txt b/clang/lib/InstallAPI/CMakeLists.txt
@@ -1,13 +1,17 @@
 set(LLVM_LINK_COMPONENTS
   Support
   TextAPI
+  Core
   )
 
 add_clang_library(clangInstallAPI
   FileList.cpp
+  Frontend.cpp
   HeaderFile.cpp
+  Visitor.cpp
 
   LINK_LIBS
   clangAST
   clangBasic
+  clangLex
   )
diff --git a/clang/lib/InstallAPI/Frontend.cpp b/clang/lib/InstallAPI/Frontend.cpp
@@ -0,0 +1,129 @@
+//===- Frontend.cpp ---------------------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/InstallAPI/Frontend.h"
+#include "clang/AST/Availability.h"
+#include "llvm/ADT/SmallString.h"
+#include "llvm/ADT/StringRef.h"
+
+using namespace llvm;
+using namespace llvm::MachO;
+
+namespace clang::installapi {
+
+GlobalRecord *FrontendRecordsSlice::addGlobal(
+    StringRef Name, RecordLinkage Linkage, GlobalRecord::Kind GV,
+    const clang::AvailabilityInfo Avail, const Decl *D, const HeaderType Access,
+    SymbolFlags Flags) {
+
+  auto *GR = llvm::MachO::RecordsSlice::addGlobal(Name, Linkage, GV, Flags);
+  FrontendRecords.insert({GR, FrontendAttrs{Avail, D, Access}});
+  return GR;
+}
+
+ObjCInterfaceRecord *FrontendRecordsSlice::addObjCInterface(
+    StringRef Name, RecordLinkage Linkage, const clang::AvailabilityInfo Avail,
+    const Decl *D, HeaderType Access, bool IsEHType) {
+  ObjCIFSymbolKind SymType =
+      ObjCIFSymbolKind::Class | ObjCIFSymbolKind::MetaClass;
+  if (IsEHType)
+    SymType |= ObjCIFSymbolKind::EHType;
+  auto *ObjCR =
+      llvm::MachO::RecordsSlice::addObjCInterface(Name, Linkage, SymType);
+  FrontendRecords.insert({ObjCR, FrontendAttrs{Avail, D, Access}});
+  return ObjCR;
+}
+
+std::optional<HeaderType>
+InstallAPIContext::findAndRecordFile(const FileEntry *FE,
+                                     const Preprocessor &PP) {
+  if (!FE)
+    return std::nullopt;
+
+  // Check if header has been looked up already and whether it is something
+  // installapi should use.
+  auto It = KnownFiles.find(FE);
+  if (It != KnownFiles.end()) {
+    if (It->second != HeaderType::Unknown)
+      return It->second;
+    else
+      return std::nullopt;
+  }
+
+  // If file was not found, search by how the header was
+  // included. This is primarily to resolve headers found
+  // in a different location than what passed directly as input.
+  StringRef IncludeName = PP.getHeaderSearchInfo().getIncludeNameForHeader(FE);
+  auto BackupIt = KnownIncludes.find(IncludeName.str());
+  if (BackupIt != KnownIncludes.end()) {
+    KnownFiles[FE] = BackupIt->second;
+    return BackupIt->second;
+  }
+
+  // Record that the file was found to avoid future string searches for the
+  // same file.
+  KnownFiles.insert({FE, HeaderType::Unknown});
+  return std::nullopt;
+}
+
+void InstallAPIContext::addKnownHeader(const HeaderFile &H) {
+  auto FE = FM->getFile(H.getPath());
+  if (!FE)
+    return; // File does not exist.
+  KnownFiles[*FE] = H.getType();
+
+  if (!H.useIncludeName())
+    return;
+
+  KnownIncludes[H.getIncludeName()] = H.getType();
+}
+
+static StringRef getFileExtension(clang::Language Lang) {
+  switch (Lang) {
+  default:
+    llvm_unreachable("Unexpected language option.");
+  case clang::Language::C:
+    return ".c";
+  case clang::Language::CXX:
+    return ".cpp";
+  case clang::Language::ObjC:
+    return ".m";
+  case clang::Language::ObjCXX:
+    return ".mm";
+  }
+}
+
+std::unique_ptr<MemoryBuffer> createInputBuffer(InstallAPIContext &Ctx) {
+  assert(Ctx.Type != HeaderType::Unknown &&
+         "unexpected access level for parsing");
+  SmallString<4096> Contents;
+  raw_svector_ostream OS(Contents);
+  for (const HeaderFile &H : Ctx.InputHeaders) {
+    if (H.getType() != Ctx.Type)
+      continue;
+    if (Ctx.LangMode == Language::C || Ctx.LangMode == Language::CXX)
+      OS << "#include ";
+    else
+      OS << "#import ";
+    if (H.useIncludeName())
+      OS << "<" << H.getIncludeName() << ">";
+    else
+      OS << "\"" << H.getPath() << "\"";
+
+    Ctx.addKnownHeader(H);
+  }
+  if (Contents.empty())
+    return nullptr;
+
+  SmallString<64> BufferName(
+      {"installapi-includes-", Ctx.Slice->getTriple().str(), "-",
+       getName(Ctx.Type), getFileExtension(Ctx.LangMode)});
+  return llvm::MemoryBuffer::getMemBufferCopy(Contents, BufferName);
+}
+
+} // namespace clang::installapi
-Original file line number
+Diff line change
@@ Expand Up / @@ -903,7 +903,7 @@ void CudaToolChain::addClangTargetOptions( @@
           return;
         addOpenMPDeviceRTL(getDriver(), DriverArgs, CC1Args, GpuArch.str(),
-                           getTriple());
+                           getTriple(), HostTC);
       }
     }
@@ Expand Down @@