diff --git a/clang/include/clang/Sema/SemaOpenACC.h b/clang/include/clang/Sema/SemaOpenACC.h
@@ -0,0 +1,74 @@
+//===----- SemaOpenACC.h - Semantic Analysis for OpenACC constructs -------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+/// \file
+/// This file declares semantic analysis for OpenACC constructs and
+/// clauses.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_SEMA_SEMAOPENACC_H
+#define LLVM_CLANG_SEMA_SEMAOPENACC_H
+
+#include "clang/AST/DeclGroup.h"
+#include "clang/Basic/OpenACCKinds.h"
+#include "clang/Basic/SourceLocation.h"
+#include "clang/Sema/Ownership.h"
+
+namespace clang {
+
+class ASTContext;
+class DiagnosticEngine;
+class LangOptions;
+class Sema;
+
+class SemaOpenACC {
+public:
+  SemaOpenACC(Sema &S);
+
+  ASTContext &getASTContext() const;
+  DiagnosticsEngine &getDiagnostics() const;
+  const LangOptions &getLangOpts() const;
+
+  Sema &SemaRef;
+
+  /// Called after parsing an OpenACC Clause so that it can be checked.
+  bool ActOnClause(OpenACCClauseKind ClauseKind, SourceLocation StartLoc);
+
+  /// Called after the construct has been parsed, but clauses haven't been
+  /// parsed.  This allows us to diagnose not-implemented, as well as set up any
+  /// state required for parsing the clauses.
+  void ActOnConstruct(OpenACCDirectiveKind K, SourceLocation StartLoc);
+
+  /// Called after the directive, including its clauses, have been parsed and
+  /// parsing has consumed the 'annot_pragma_openacc_end' token. This DOES
+  /// happen before any associated declarations or statements have been parsed.
+  /// This function is only called when we are parsing a 'statement' context.
+  bool ActOnStartStmtDirective(OpenACCDirectiveKind K, SourceLocation StartLoc);
+
+  /// Called after the directive, including its clauses, have been parsed and
+  /// parsing has consumed the 'annot_pragma_openacc_end' token. This DOES
+  /// happen before any associated declarations or statements have been parsed.
+  /// This function is only called when we are parsing a 'Decl' context.
+  bool ActOnStartDeclDirective(OpenACCDirectiveKind K, SourceLocation StartLoc);
+  /// Called when we encounter an associated statement for our construct, this
+  /// should check legality of the statement as it appertains to this Construct.
+  StmtResult ActOnAssociatedStmt(OpenACCDirectiveKind K, StmtResult AssocStmt);
+
+  /// Called after the directive has been completely parsed, including the
+  /// declaration group or associated statement.
+  StmtResult ActOnEndStmtDirective(OpenACCDirectiveKind K,
+                                   SourceLocation StartLoc,
+                                   SourceLocation EndLoc, StmtResult AssocStmt);
+  /// Called after the directive has been completely parsed, including the
+  /// declaration group or associated statement.
+  DeclGroupRef ActOnEndDeclDirective();
+};
+
+} // namespace clang
+
+#endif // LLVM_CLANG_SEMA_SEMAOPENACC_H
diff --git a/clang/include/clang/Serialization/TypeBitCodes.def b/clang/include/clang/Serialization/TypeBitCodes.def
@@ -66,5 +66,6 @@ TYPE_BIT_CODE(Using, USING, 54)
 TYPE_BIT_CODE(BTFTagAttributed, BTFTAG_ATTRIBUTED, 55)
 TYPE_BIT_CODE(PackIndexing, PACK_INDEXING, 56)
 TYPE_BIT_CODE(CountAttributed, COUNT_ATTRIBUTED, 57)
+TYPE_BIT_CODE(ArrayParameter, ARRAY_PARAMETER, 58)
 
 #undef TYPE_BIT_CODE
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
@@ -879,7 +879,8 @@ ASTContext::ASTContext(LangOptions &LOpts, SourceManager &SM,
       TemplateSpecializationTypes(this_()),
       DependentTemplateSpecializationTypes(this_()), AutoTypes(this_()),
       DependentBitIntTypes(this_()), SubstTemplateTemplateParmPacks(this_()),
-      CanonTemplateTemplateParms(this_()), SourceMgr(SM), LangOpts(LOpts),
+      ArrayParameterTypes(this_()), CanonTemplateTemplateParms(this_()),
+      SourceMgr(SM), LangOpts(LOpts),
       NoSanitizeL(new NoSanitizeList(LangOpts.NoSanitizeFiles, SM)),
       XRayFilter(new XRayFunctionFilter(LangOpts.XRayAlwaysInstrumentFiles,
                                         LangOpts.XRayNeverInstrumentFiles,
@@ -1906,7 +1907,8 @@ TypeInfo ASTContext::getTypeInfoImpl(const Type *T) const {
 
   case Type::IncompleteArray:
   case Type::VariableArray:
-  case Type::ConstantArray: {
+  case Type::ConstantArray:
+  case Type::ArrayParameter: {
     // Model non-constant sized arrays as size zero, but track the alignment.
     uint64_t Size = 0;
     if (const auto *CAT = dyn_cast<ConstantArrayType>(T))
@@ -3396,6 +3398,37 @@ QualType ASTContext::getDecayedType(QualType T) const {
   return getDecayedType(T, Decayed);
 }
 
+QualType ASTContext::getArrayParameterType(QualType Ty) const {
+  if (Ty->isArrayParameterType())
+    return Ty;
+  assert(Ty->isConstantArrayType() && "Ty must be an array type.");
+  const auto *ATy = cast<ConstantArrayType>(Ty);
+  llvm::FoldingSetNodeID ID;
+  ATy->Profile(ID, *this, ATy->getElementType(), ATy->getZExtSize(),
+               ATy->getSizeExpr(), ATy->getSizeModifier(),
+               ATy->getIndexTypeQualifiers().getAsOpaqueValue());
+  void *InsertPos = nullptr;
+  ArrayParameterType *AT =
+      ArrayParameterTypes.FindNodeOrInsertPos(ID, InsertPos);
+  if (AT)
+    return QualType(AT, 0);
+
+  QualType Canonical;
+  if (!Ty.isCanonical()) {
+    Canonical = getArrayParameterType(getCanonicalType(Ty));
+
+    // Get the new insert position for the node we care about.
+    AT = ArrayParameterTypes.FindNodeOrInsertPos(ID, InsertPos);
+    assert(!AT && "Shouldn't be in the map!");
+  }
+
+  AT = new (*this, alignof(ArrayParameterType))
+      ArrayParameterType(ATy, Canonical);
+  Types.push_back(AT);
+  ArrayParameterTypes.InsertNode(AT, InsertPos);
+  return QualType(AT, 0);
+}
+
 /// getBlockPointerType - Return the uniqued reference to the type for
 /// a pointer to the specified block.
 QualType ASTContext::getBlockPointerType(QualType T) const {
@@ -3642,6 +3675,7 @@ QualType ASTContext::getVariableArrayDecayedType(QualType type) const {
   case Type::PackIndexing:
   case Type::BitInt:
   case Type::DependentBitInt:
+  case Type::ArrayParameter:
     llvm_unreachable("type should never be variably-modified");
 
   // These types can be variably-modified but should never need to
@@ -6051,7 +6085,9 @@ CanQualType ASTContext::getCanonicalParamType(QualType T) const {
   T = getVariableArrayDecayedType(T);
   const Type *Ty = T.getTypePtr();
   QualType Result;
-  if (isa<ArrayType>(Ty)) {
+  if (getLangOpts().HLSL && isa<ConstantArrayType>(Ty)) {
+    Result = getArrayParameterType(QualType(Ty, 0));
+  } else if (isa<ArrayType>(Ty)) {
     Result = getArrayDecayedType(QualType(Ty,0));
   } else if (isa<FunctionType>(Ty)) {
     Result = getPointerType(QualType(Ty, 0));
@@ -6973,6 +7009,8 @@ const ArrayType *ASTContext::getAsArrayType(QualType T) const {
 }
 
 QualType ASTContext::getAdjustedParameterType(QualType T) const {
+  if (getLangOpts().HLSL && T->isConstantArrayType())
+    return getArrayParameterType(T);
   if (T->isArrayType() || T->isFunctionType())
     return getDecayedType(T);
   return T;
@@ -8583,6 +8621,7 @@ void ASTContext::getObjCEncodingForTypeImpl(QualType T, std::string &S,
   case Type::DeducedTemplateSpecialization:
     return;
 
+  case Type::ArrayParameter:
   case Type::Pipe:
 #define ABSTRACT_TYPE(KIND, BASE)
 #define TYPE(KIND, BASE)
@@ -10926,6 +10965,10 @@ QualType ASTContext::mergeTypes(QualType LHS, QualType RHS, bool OfBlockPointer,
     assert(LHS != RHS &&
            "Equivalent pipe types should have already been handled!");
     return {};
+  case Type::ArrayParameter:
+    assert(LHS != RHS &&
+           "Equivalent ArrayParameter types should have already been handled!");
+    return {};
   case Type::BitInt: {
     // Merge two bit-precise int types, while trying to preserve typedef info.
     bool LHSUnsigned = LHS->castAs<BitIntType>()->isUnsigned();
@@ -12817,6 +12860,18 @@ static QualType getCommonNonSugarTypeNode(ASTContext &Ctx, const Type *X,
         getCommonArrayElementType(Ctx, AX, QX, AY, QY), AX->getSize(), SizeExpr,
         getCommonSizeModifier(AX, AY), getCommonIndexTypeCVRQualifiers(AX, AY));
   }
+  case Type::ArrayParameter: {
+    const auto *AX = cast<ArrayParameterType>(X),
+               *AY = cast<ArrayParameterType>(Y);
+    assert(AX->getSize() == AY->getSize());
+    const Expr *SizeExpr = Ctx.hasSameExpr(AX->getSizeExpr(), AY->getSizeExpr())
+                               ? AX->getSizeExpr()
+                               : nullptr;
+    auto ArrayTy = Ctx.getConstantArrayType(
+        getCommonArrayElementType(Ctx, AX, QX, AY, QY), AX->getSize(), SizeExpr,
+        getCommonSizeModifier(AX, AY), getCommonIndexTypeCVRQualifiers(AX, AY));
+    return Ctx.getArrayParameterType(ArrayTy);
+  }
   case Type::Atomic: {
     const auto *AX = cast<AtomicType>(X), *AY = cast<AtomicType>(Y);
     return Ctx.getAtomicType(
@@ -13078,6 +13133,7 @@ static QualType getCommonSugarTypeNode(ASTContext &Ctx, const Type *X,
     CANONICAL_TYPE(Builtin)
     CANONICAL_TYPE(Complex)
     CANONICAL_TYPE(ConstantArray)
+    CANONICAL_TYPE(ArrayParameter)
     CANONICAL_TYPE(ConstantMatrix)
     CANONICAL_TYPE(Enum)
     CANONICAL_TYPE(ExtVector)

diff --git a/clang/lib/AST/ASTImporter.cpp b/clang/lib/AST/ASTImporter.cpp
@@ -1218,6 +1218,15 @@ ASTNodeImporter::VisitConstantArrayType(const ConstantArrayType *T) {
       T->getIndexTypeCVRQualifiers());
 }
 
+ExpectedType
+ASTNodeImporter::VisitArrayParameterType(const ArrayParameterType *T) {
+  ExpectedType ToArrayTypeOrErr = VisitConstantArrayType(T);
+  if (!ToArrayTypeOrErr)
+    return ToArrayTypeOrErr.takeError();
+
+  return Importer.getToContext().getArrayParameterType(*ToArrayTypeOrErr);
+}
+
 ExpectedType
 ASTNodeImporter::VisitIncompleteArrayType(const IncompleteArrayType *T) {
   ExpectedType ToElementTypeOrErr = import(T->getElementType());

diff --git a/clang/lib/AST/ASTStructuralEquivalence.cpp b/clang/lib/AST/ASTStructuralEquivalence.cpp
@@ -840,6 +840,7 @@ static bool IsStructurallyEquivalent(StructuralEquivalenceContext &Context,
 
   case Type::Adjusted:
   case Type::Decayed:
+  case Type::ArrayParameter:
     if (!IsStructurallyEquivalent(Context,
                                   cast<AdjustedType>(T1)->getOriginalType(),
                                   cast<AdjustedType>(T2)->getOriginalType()))

diff --git a/clang/lib/AST/Expr.cpp b/clang/lib/AST/Expr.cpp
@@ -1948,6 +1948,7 @@ bool CastExpr::CastConsistency() const {
   case CK_UserDefinedConversion:    // operator bool()
   case CK_BuiltinFnToFnPtr:
   case CK_FixedPointToBoolean:
+  case CK_HLSLArrayRValue:
   CheckNoBasePath:
     assert(path_empty() && "Cast kind should not have a base path!");
     break;

diff --git a/clang/lib/AST/ExprConstant.cpp b/clang/lib/AST/ExprConstant.cpp
@@ -11699,6 +11699,7 @@ GCCTypeClass EvaluateBuiltinClassifyType(QualType T,
   case Type::IncompleteArray:
   case Type::FunctionNoProto:
   case Type::FunctionProto:
+  case Type::ArrayParameter:
     return GCCTypeClass::Pointer;
 
   case Type::MemberPointer:
@@ -14085,6 +14086,7 @@ bool IntExprEvaluator::VisitCastExpr(const CastExpr *E) {
   case CK_AtomicToNonAtomic:
   case CK_NoOp:
   case CK_LValueToRValueBitCast:
+  case CK_HLSLArrayRValue:
     return ExprEvaluatorBaseTy::VisitCastExpr(E);
 
   case CK_MemberPointerToBoolean:
@@ -14913,6 +14915,7 @@ bool ComplexExprEvaluator::VisitCastExpr(const CastExpr *E) {
   case CK_AtomicToNonAtomic:
   case CK_NoOp:
   case CK_LValueToRValueBitCast:
+  case CK_HLSLArrayRValue:
     return ExprEvaluatorBaseTy::VisitCastExpr(E);
 
   case CK_Dependent:

diff --git a/clang/lib/AST/ItaniumMangle.cpp b/clang/lib/AST/ItaniumMangle.cpp
@@ -2398,6 +2398,7 @@ bool CXXNameMangler::mangleUnresolvedTypeOrSimpleId(QualType Ty,
   case Type::Complex:
   case Type::Adjusted:
   case Type::Decayed:
+  case Type::ArrayParameter:
   case Type::Pointer:
   case Type::BlockPointer:
   case Type::LValueReference:
@@ -4446,6 +4447,10 @@ void CXXNameMangler::mangleType(const DependentBitIntType *T) {
   Out << "_";
 }
 
+void CXXNameMangler::mangleType(const ArrayParameterType *T) {
+  mangleType(cast<ConstantArrayType>(T));
+}
+
 void CXXNameMangler::mangleIntegerLiteral(QualType T,
                                           const llvm::APSInt &Value) {
   //  <expr-primary> ::= L <type> <value number> E # integer literal

diff --git a/clang/lib/AST/MicrosoftMangle.cpp b/clang/lib/AST/MicrosoftMangle.cpp
@@ -3079,6 +3079,11 @@ void MicrosoftCXXNameMangler::mangleArrayType(const ArrayType *T) {
   mangleType(ElementTy, SourceRange(), QMM_Escape);
 }
 
+void MicrosoftCXXNameMangler::mangleType(const ArrayParameterType *T,
+                                         Qualifiers, SourceRange) {
+  mangleArrayType(cast<ConstantArrayType>(T));
+}
+
 // <type>                   ::= <pointer-to-member-type>
 // <pointer-to-member-type> ::= <pointer-cvr-qualifiers> <cvr-qualifiers>
 //                                                          <class name> <type>

diff --git a/clang/lib/AST/ODRHash.cpp b/clang/lib/AST/ODRHash.cpp
@@ -944,6 +944,10 @@ class ODRTypeVisitor : public TypeVisitor<ODRTypeVisitor> {
     VisitArrayType(T);
   }
 
+  void VisitArrayParameterType(const ArrayParameterType *T) {
+    VisitConstantArrayType(T);
+  }
+
   void VisitDependentSizedArrayType(const DependentSizedArrayType *T) {
     AddStmt(T->getSizeExpr());
     VisitArrayType(T);

diff --git a/clang/lib/AST/Type.cpp b/clang/lib/AST/Type.cpp
@@ -1197,6 +1197,14 @@ struct SimpleTransformVisitor : public TypeVisitor<Derived, QualType> {
     return Ctx.getDecayedType(originalType);
   }
 
+  QualType VisitArrayParameterType(const ArrayParameterType *T) {
+    QualType ArrTy = VisitConstantArrayType(T);
+    if (ArrTy.isNull())
+      return {};
+
+    return Ctx.getArrayParameterType(ArrTy);
+  }
+
   SUGARED_TYPE_CLASS(TypeOfExpr)
   SUGARED_TYPE_CLASS(TypeOf)
   SUGARED_TYPE_CLASS(Decltype)
@@ -4454,6 +4462,7 @@ static CachedProperties computeCachedProperties(const Type *T) {
   case Type::ConstantArray:
   case Type::IncompleteArray:
   case Type::VariableArray:
+  case Type::ArrayParameter:
     return Cache::get(cast<ArrayType>(T)->getElementType());
   case Type::Vector:
   case Type::ExtVector:
@@ -4542,6 +4551,7 @@ LinkageInfo LinkageComputer::computeTypeLinkageInfo(const Type *T) {
   case Type::ConstantArray:
   case Type::IncompleteArray:
   case Type::VariableArray:
+  case Type::ArrayParameter:
     return computeTypeLinkageInfo(cast<ArrayType>(T)->getElementType());
   case Type::Vector:
   case Type::ExtVector:
@@ -4736,6 +4746,7 @@ bool Type::canHaveNullability(bool ResultIfUnknown) const {
   case Type::Pipe:
   case Type::BitInt:
   case Type::DependentBitInt:
+  case Type::ArrayParameter:
     return false;
   }
   llvm_unreachable("bad type kind!");

diff --git a/clang/lib/AST/TypePrinter.cpp b/clang/lib/AST/TypePrinter.cpp
@@ -268,6 +268,7 @@ bool TypePrinter::canPrefixQualifiers(const Type *T,
 
     case Type::Adjusted:
     case Type::Decayed:
+    case Type::ArrayParameter:
     case Type::Pointer:
     case Type::BlockPointer:
     case Type::LValueReference:
@@ -595,6 +596,16 @@ void TypePrinter::printDecayedBefore(const DecayedType *T, raw_ostream &OS) {
   printAdjustedBefore(T, OS);
 }
 
+void TypePrinter::printArrayParameterAfter(const ArrayParameterType *T,
+                                           raw_ostream &OS) {
+  printConstantArrayAfter(T, OS);
+}
+
+void TypePrinter::printArrayParameterBefore(const ArrayParameterType *T,
+                                            raw_ostream &OS) {
+  printConstantArrayBefore(T, OS);
+}
+
 void TypePrinter::printDecayedAfter(const DecayedType *T, raw_ostream &OS) {
   printAdjustedAfter(T, OS);
 }

diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
@@ -4719,7 +4719,8 @@ void CodeGenFunction::EmitCallArg(CallArgList &args, const Expr *E,
   }
 
   if (HasAggregateEvalKind && isa<ImplicitCastExpr>(E) &&
-      cast<CastExpr>(E)->getCastKind() == CK_LValueToRValue) {
+      cast<CastExpr>(E)->getCastKind() == CK_LValueToRValue &&
+      !type->isArrayParameterType()) {
     LValue L = EmitLValue(cast<CastExpr>(E)->getSubExpr());
     assert(L.isSimple());
     args.addUncopiedAggregate(L, type);

diff --git a/clang/lib/CodeGen/CGDebugInfo.cpp b/clang/lib/CodeGen/CGDebugInfo.cpp
@@ -3641,6 +3641,7 @@ llvm::DIType *CGDebugInfo::CreateTypeNode(QualType Ty, llvm::DIFile *Unit) {
   case Type::ConstantArray:
   case Type::VariableArray:
   case Type::IncompleteArray:
+  case Type::ArrayParameter:
     return CreateType(cast<ArrayType>(Ty), Unit);
 
   case Type::LValueReference:

diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
@@ -5190,6 +5190,7 @@ LValue CodeGenFunction::EmitCastLValue(const CastExpr *E) {
   case CK_IntegralToFixedPoint:
   case CK_MatrixCast:
   case CK_HLSLVectorTruncation:
+  case CK_HLSLArrayRValue:
     return EmitUnsupportedLValue(E, "unexpected cast lvalue");
 
   case CK_Dependent:

diff --git a/clang/lib/CodeGen/CGExprAgg.cpp b/clang/lib/CodeGen/CGExprAgg.cpp
@@ -883,6 +883,9 @@ void AggExprEmitter::VisitCastExpr(CastExpr *E) {
 
     [[fallthrough]];
 
+  case CK_HLSLArrayRValue:
+    Visit(E->getSubExpr());
+    break;
 
   case CK_NoOp:
   case CK_UserDefinedConversion:
@@ -1524,6 +1527,7 @@ static bool castPreservesZero(const CastExpr *CE) {
   case CK_LValueToRValue:
   case CK_LValueToRValueBitCast:
   case CK_UncheckedDerivedToBase:
+  case CK_HLSLArrayRValue:
     return false;
   }
   llvm_unreachable("Unhandled clang::CastKind enum");

diff --git a/clang/lib/CodeGen/CGExprComplex.cpp b/clang/lib/CodeGen/CGExprComplex.cpp
@@ -616,6 +616,7 @@ ComplexPairTy ComplexExprEmitter::EmitCast(CastKind CK, Expr *Op,
   case CK_IntegralToFixedPoint:
   case CK_MatrixCast:
   case CK_HLSLVectorTruncation:
+  case CK_HLSLArrayRValue:
     llvm_unreachable("invalid cast kind for complex value");
 
   case CK_FloatingRealToComplex:

diff --git a/clang/lib/CodeGen/CGExprConstant.cpp b/clang/lib/CodeGen/CGExprConstant.cpp
@@ -1226,6 +1226,7 @@ class ConstExprEmitter :
     case CK_ZeroToOCLOpaqueType:
     case CK_MatrixCast:
     case CK_HLSLVectorTruncation:
+    case CK_HLSLArrayRValue:
       return nullptr;
     }
     llvm_unreachable("Invalid CastKind");

diff --git a/clang/lib/CodeGen/CGExprScalar.cpp b/clang/lib/CodeGen/CGExprScalar.cpp
@@ -2329,6 +2329,7 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) {
   case CK_FloatingComplexToIntegralComplex:
   case CK_ConstructorConversion:
   case CK_ToUnion:
+  case CK_HLSLArrayRValue:
     llvm_unreachable("scalar cast to non-scalar value");
 
   case CK_LValueToRValue:

diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -276,6 +276,7 @@ TypeEvaluationKind CodeGenFunction::getEvaluationKind(QualType type) {
     case Type::Record:
     case Type::ObjCObject:
     case Type::ObjCInterface:
+    case Type::ArrayParameter:
       return TEK_Aggregate;
 
     // We operate on atomic values according to their underlying type.
@@ -2361,6 +2362,7 @@ void CodeGenFunction::EmitVariablyModifiedType(QualType type) {
       type = cast<MemberPointerType>(ty)->getPointeeType();
       break;
 
+    case Type::ArrayParameter:
     case Type::ConstantArray:
     case Type::IncompleteArray:
       // Losing element qualification here is fine.

diff --git a/clang/lib/CodeGen/CodeGenTypes.cpp b/clang/lib/CodeGen/CodeGenTypes.cpp
@@ -590,6 +590,7 @@ llvm::Type *CodeGenTypes::ConvertType(QualType T) {
     ResultType = llvm::ArrayType::get(ResultType, 0);
     break;
   }
+  case Type::ArrayParameter:
   case Type::ConstantArray: {
     const ConstantArrayType *A = cast<ConstantArrayType>(Ty);
     llvm::Type *EltTy = ConvertTypeForMem(A->getElementType());

diff --git a/clang/lib/CodeGen/ItaniumCXXABI.cpp b/clang/lib/CodeGen/ItaniumCXXABI.cpp
@@ -3584,6 +3584,9 @@ void ItaniumRTTIBuilder::BuildVTablePointer(const Type *Ty) {
   case Type::Pipe:
     llvm_unreachable("Pipe types shouldn't get here");
 
+  case Type::ArrayParameter:
+    llvm_unreachable("Array Parameter types should not get here.");
+
   case Type::Builtin:
   case Type::BitInt:
   // GCC treats vector and complex types as fundamental types.
@@ -3868,6 +3871,7 @@ llvm::Constant *ItaniumRTTIBuilder::BuildTypeInfo(
   case Type::ConstantArray:
   case Type::IncompleteArray:
   case Type::VariableArray:
+  case Type::ArrayParameter:
     // Itanium C++ ABI 2.9.5p5:
     // abi::__array_type_info adds no data members to std::type_info.
     break;

diff --git a/clang/lib/Edit/RewriteObjCFoundationAPI.cpp b/clang/lib/Edit/RewriteObjCFoundationAPI.cpp
@@ -1000,6 +1000,7 @@ static bool rewriteToNumericBoxedExpression(const ObjCMessageExpr *Msg,
     case CK_LValueToRValue:
     case CK_NoOp:
     case CK_UserDefinedConversion:
+    case CK_HLSLArrayRValue:
       break;
 
     case CK_IntegralCast: {

diff --git a/clang/lib/Parse/ParseOpenACC.cpp b/clang/lib/Parse/ParseOpenACC.cpp
@@ -14,6 +14,7 @@
 #include "clang/Parse/ParseDiagnostic.h"
 #include "clang/Parse/Parser.h"
 #include "clang/Parse/RAIIObjectsForParser.h"
+#include "clang/Sema/SemaOpenACC.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/StringSwitch.h"
 
@@ -777,7 +778,7 @@ bool Parser::ParseOpenACCClause(OpenACCDirectiveKind DirKind) {
   SourceLocation ClauseLoc = ConsumeToken();
 
   bool Result = ParseOpenACCClauseParams(DirKind, Kind);
-  getActions().ActOnOpenACCClause(Kind, ClauseLoc);
+  getActions().OpenACC().ActOnClause(Kind, ClauseLoc);
   return Result;
 }
 
@@ -1151,7 +1152,7 @@ Parser::OpenACCDirectiveParseInfo Parser::ParseOpenACCDirective() {
   SourceLocation StartLoc = getCurToken().getLocation();
   OpenACCDirectiveKind DirKind = ParseOpenACCDirectiveKind(*this);
 
-  getActions().ActOnOpenACCConstruct(DirKind, StartLoc);
+  getActions().OpenACC().ActOnConstruct(DirKind, StartLoc);
 
   // Once we've parsed the construct/directive name, some have additional
   // specifiers that need to be taken care of. Atomic has an 'atomic-clause'
@@ -1223,12 +1224,12 @@ Parser::DeclGroupPtrTy Parser::ParseOpenACCDirectiveDecl() {
 
   OpenACCDirectiveParseInfo DirInfo = ParseOpenACCDirective();
 
-  if (getActions().ActOnStartOpenACCDeclDirective(DirInfo.DirKind,
-                                                  DirInfo.StartLoc))
+  if (getActions().OpenACC().ActOnStartDeclDirective(DirInfo.DirKind,
+                                                     DirInfo.StartLoc))
     return nullptr;
 
   // TODO OpenACC: Do whatever decl parsing is required here.
-  return DeclGroupPtrTy::make(getActions().ActOnEndOpenACCDeclDirective());
+  return DeclGroupPtrTy::make(getActions().OpenACC().ActOnEndDeclDirective());
 }
 
 // Parse OpenACC Directive on a Statement.
@@ -1239,8 +1240,8 @@ StmtResult Parser::ParseOpenACCDirectiveStmt() {
   ConsumeAnnotationToken();
 
   OpenACCDirectiveParseInfo DirInfo = ParseOpenACCDirective();
-  if (getActions().ActOnStartOpenACCStmtDirective(DirInfo.DirKind,
-                                                  DirInfo.StartLoc))
+  if (getActions().OpenACC().ActOnStartStmtDirective(DirInfo.DirKind,
+                                                     DirInfo.StartLoc))
     return StmtError();
 
   StmtResult AssocStmt;
@@ -1249,10 +1250,10 @@ StmtResult Parser::ParseOpenACCDirectiveStmt() {
     ParsingOpenACCDirectiveRAII DirScope(*this, /*Value=*/false);
     ParseScope ACCScope(this, getOpenACCScopeFlags(DirInfo.DirKind));
 
-    AssocStmt = getActions().ActOnOpenACCAssociatedStmt(DirInfo.DirKind,
-                                                        ParseStatement());
+    AssocStmt = getActions().OpenACC().ActOnAssociatedStmt(DirInfo.DirKind,
+                                                           ParseStatement());
   }
 
-  return getActions().ActOnEndOpenACCStmtDirective(
+  return getActions().OpenACC().ActOnEndStmtDirective(
       DirInfo.DirKind, DirInfo.StartLoc, DirInfo.EndLoc, AssocStmt);
 }
diff --git a/clang/lib/Sema/JumpDiagnostics.cpp b/clang/lib/Sema/JumpDiagnostics.cpp
@@ -16,6 +16,7 @@
 #include "clang/AST/ExprCXX.h"
 #include "clang/AST/StmtCXX.h"
 #include "clang/AST/StmtObjC.h"
+#include "clang/AST/StmtOpenACC.h"
 #include "clang/AST/StmtOpenMP.h"
 #include "clang/Basic/SourceLocation.h"
 #include "clang/Sema/SemaInternal.h"

diff --git a/clang/lib/Sema/Sema.cpp b/clang/lib/Sema/Sema.cpp
@@ -43,6 +43,7 @@
 #include "clang/Sema/ScopeInfo.h"
 #include "clang/Sema/SemaConsumer.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenACC.h"
 #include "clang/Sema/TemplateDeduction.h"
 #include "clang/Sema/TemplateInstCallback.h"
 #include "clang/Sema/TypoCorrection.h"
@@ -196,7 +197,7 @@ Sema::Sema(Preprocessor &pp, ASTContext &ctxt, ASTConsumer &consumer,
       ThreadSafetyDeclCache(nullptr), LateTemplateParser(nullptr),
       LateTemplateParserCleanup(nullptr), OpaqueParser(nullptr),
       CurContext(nullptr), ExternalSource(nullptr), CurScope(nullptr),
-      Ident_super(nullptr),
+      Ident_super(nullptr), OpenACCPtr(std::make_unique<SemaOpenACC>(*this)),
       MSPointerToMemberRepresentationMethod(
           LangOpts.getMSPointerToMemberRepresentationMethod()),
       MSStructPragmaOn(false), VtorDispStack(LangOpts.getVtorDispMode()),
@@ -653,6 +654,7 @@ ExprResult Sema::ImpCastExprToType(Expr *E, QualType Ty,
     case CK_FunctionToPointerDecay:
     case CK_ToVoid:
     case CK_NonAtomicToAtomic:
+    case CK_HLSLArrayRValue:
       break;
     }
   }

diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
@@ -658,8 +658,9 @@ ExprResult Sema::DefaultLvalueConversion(Expr *E) {
   QualType T = E->getType();
   assert(!T.isNull() && "r-value conversion on typeless expression?");
 
-  // lvalue-to-rvalue conversion cannot be applied to function or array types.
-  if (T->isFunctionType() || T->isArrayType())
+  // lvalue-to-rvalue conversion cannot be applied to types that decay to
+  // pointers (i.e. function or array types).
+  if (T->canDecayToPointerType())
     return E;
 
   // We don't want to throw lvalue-to-rvalue casts on top of
@@ -4686,6 +4687,9 @@ static void captureVariablyModifiedType(ASTContext &Context, QualType T,
     case Type::Decayed:
       T = cast<DecayedType>(Ty)->getPointeeType();
       break;
+    case Type::ArrayParameter:
+      T = cast<ArrayParameterType>(Ty)->getElementType();
+      break;
     case Type::Pointer:
       T = cast<PointerType>(Ty)->getPointeeType();
       break;
@@ -12908,6 +12912,8 @@ static ImplicitConversionKind castKindToImplicitConversionKind(CastKind CK) {
   case CK_IntegralComplexToReal:
   case CK_IntegralRealToComplex:
     return ICK_Complex_Real;
+  case CK_HLSLArrayRValue:
+    return ICK_HLSL_Array_RValue;
   }
 }
 

diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
@@ -4416,6 +4416,13 @@ Sema::PerformImplicitConversion(Expr *From, QualType ToType,
                .get();
     break;
 
+  case ICK_HLSL_Array_RValue:
+    FromType = Context.getArrayParameterType(FromType);
+    From = ImpCastExprToType(From, FromType, CK_HLSLArrayRValue, VK_PRValue,
+                             /*BasePath=*/nullptr, CCK)
+               .get();
+    break;
+
   case ICK_Function_To_Pointer:
     FromType = Context.getPointerType(FromType);
     From = ImpCastExprToType(From, FromType, CK_FunctionToPointerDecay,
@@ -4793,6 +4800,7 @@ Sema::PerformImplicitConversion(Expr *From, QualType ToType,
   case ICK_Num_Conversion_Kinds:
   case ICK_C_Only_Conversion:
   case ICK_Incompatible_Pointer_Conversion:
+  case ICK_HLSL_Array_RValue:
     llvm_unreachable("Improper second standard conversion");
   }
 

diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
@@ -6269,7 +6269,10 @@ void InitializationSequence::InitializeFrom(Sema &S,
   //       initializer is a string literal, see 8.5.2.
   //     - Otherwise, if the destination type is an array, the program is
   //       ill-formed.
-  if (const ArrayType *DestAT = Context.getAsArrayType(DestType)) {
+  //     - Except in HLSL, where non-decaying array parameters behave like
+  //       non-array types for initialization.
+  if (DestType->isArrayType() && !DestType->isArrayParameterType()) {
+    const ArrayType *DestAT = Context.getAsArrayType(DestType);
     if (Initializer && isa<VariableArrayType>(DestAT)) {
       SetFailed(FK_VariableLengthArrayHasInitializer);
       return;

diff --git a/clang/lib/Sema/SemaLookup.cpp b/clang/lib/Sema/SemaLookup.cpp
@@ -3243,6 +3243,10 @@ addAssociatedClassesAndNamespaces(AssociatedLookup &Result, QualType Ty) {
     case Type::Pipe:
       T = cast<PipeType>(T)->getElementType().getTypePtr();
       continue;
+
+    // Array parameter types are treated as fundamental types.
+    case Type::ArrayParameter:
+      break;
     }
 
     if (Queue.empty())

diff --git a/clang/lib/Sema/SemaOpenACC.cpp b/clang/lib/Sema/SemaOpenACC.cpp
@@ -11,14 +11,15 @@
 ///
 //===----------------------------------------------------------------------===//
 
+#include "clang/AST/StmtOpenACC.h"
+#include "clang/Sema/SemaOpenACC.h"
 #include "clang/Basic/DiagnosticSema.h"
-#include "clang/Basic/OpenACCKinds.h"
 #include "clang/Sema/Sema.h"
 
 using namespace clang;
 
 namespace {
-bool diagnoseConstructAppertainment(Sema &S, OpenACCDirectiveKind K,
+bool diagnoseConstructAppertainment(SemaOpenACC &S, OpenACCDirectiveKind K,
                                     SourceLocation StartLoc, bool IsStmt) {
   switch (K) {
   default:
@@ -30,24 +31,32 @@ bool diagnoseConstructAppertainment(Sema &S, OpenACCDirectiveKind K,
   case OpenACCDirectiveKind::Serial:
   case OpenACCDirectiveKind::Kernels:
     if (!IsStmt)
-      return S.Diag(StartLoc, diag::err_acc_construct_appertainment) << K;
+      return S.SemaRef.Diag(StartLoc, diag::err_acc_construct_appertainment)
+             << K;
     break;
   }
   return false;
 }
 } // namespace
 
-bool Sema::ActOnOpenACCClause(OpenACCClauseKind ClauseKind,
+SemaOpenACC::SemaOpenACC(Sema &S) : SemaRef(S) {}
+
+ASTContext &SemaOpenACC::getASTContext() const { return SemaRef.Context; }
+DiagnosticsEngine &SemaOpenACC::getDiagnostics() const { return SemaRef.Diags; }
+const LangOptions &SemaOpenACC::getLangOpts() const { return SemaRef.LangOpts; }
+
+bool SemaOpenACC::ActOnClause(OpenACCClauseKind ClauseKind,
                               SourceLocation StartLoc) {
   if (ClauseKind == OpenACCClauseKind::Invalid)
     return false;
   // For now just diagnose that it is unsupported and leave the parsing to do
   // whatever it can do. This function will eventually need to start returning
   // some sort of Clause AST type, but for now just return true/false based on
   // success.
-  return Diag(StartLoc, diag::warn_acc_clause_unimplemented) << ClauseKind;
+  return SemaRef.Diag(StartLoc, diag::warn_acc_clause_unimplemented)
+         << ClauseKind;
 }
-void Sema::ActOnOpenACCConstruct(OpenACCDirectiveKind K,
+void SemaOpenACC::ActOnConstruct(OpenACCDirectiveKind K,
                                  SourceLocation StartLoc) {
   switch (K) {
   case OpenACCDirectiveKind::Invalid:
@@ -63,17 +72,17 @@ void Sema::ActOnOpenACCConstruct(OpenACCDirectiveKind K,
     // here as these constructs do not take any arguments.
     break;
   default:
-    Diag(StartLoc, diag::warn_acc_construct_unimplemented) << K;
+    SemaRef.Diag(StartLoc, diag::warn_acc_construct_unimplemented) << K;
     break;
   }
 }
 
-bool Sema::ActOnStartOpenACCStmtDirective(OpenACCDirectiveKind K,
+bool SemaOpenACC::ActOnStartStmtDirective(OpenACCDirectiveKind K,
                                           SourceLocation StartLoc) {
   return diagnoseConstructAppertainment(*this, K, StartLoc, /*IsStmt=*/true);
 }
 
-StmtResult Sema::ActOnEndOpenACCStmtDirective(OpenACCDirectiveKind K,
+StmtResult SemaOpenACC::ActOnEndStmtDirective(OpenACCDirectiveKind K,
                                               SourceLocation StartLoc,
                                               SourceLocation EndLoc,
                                               StmtResult AssocStmt) {
@@ -92,7 +101,7 @@ StmtResult Sema::ActOnEndOpenACCStmtDirective(OpenACCDirectiveKind K,
   llvm_unreachable("Unhandled case in directive handling?");
 }
 
-StmtResult Sema::ActOnOpenACCAssociatedStmt(OpenACCDirectiveKind K,
+StmtResult SemaOpenACC::ActOnAssociatedStmt(OpenACCDirectiveKind K,
                                             StmtResult AssocStmt) {
   switch (K) {
   default:
@@ -114,9 +123,9 @@ StmtResult Sema::ActOnOpenACCAssociatedStmt(OpenACCDirectiveKind K,
   llvm_unreachable("Invalid associated statement application");
 }
 
-bool Sema::ActOnStartOpenACCDeclDirective(OpenACCDirectiveKind K,
+bool SemaOpenACC::ActOnStartDeclDirective(OpenACCDirectiveKind K,
                                           SourceLocation StartLoc) {
   return diagnoseConstructAppertainment(*this, K, StartLoc, /*IsStmt=*/false);
 }
 
-DeclGroupRef Sema::ActOnEndOpenACCDeclDirective() { return DeclGroupRef{}; }
+DeclGroupRef SemaOpenACC::ActOnEndDeclDirective() { return DeclGroupRef{}; }
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
@@ -160,6 +160,7 @@ ImplicitConversionRank clang::GetConversionRank(ImplicitConversionKind Kind) {
       ICR_C_Conversion_Extension,
       ICR_Conversion,
       ICR_Conversion,
+      ICR_Conversion,
   };
   static_assert(std::size(Rank) == (int)ICK_Num_Conversion_Kinds);
   return Rank[(int)Kind];
@@ -201,6 +202,7 @@ static const char *GetImplicitConversionName(ImplicitConversionKind Kind) {
       "Incompatible pointer conversion",
       "Fixed point conversion",
       "HLSL vector truncation",
+      "Non-decaying array conversion",
   };
   static_assert(std::size(Name) == (int)ICK_Num_Conversion_Kinds);
   return Name[Kind];
@@ -2131,8 +2133,7 @@ static bool IsStandardConversion(Sema &S, Expr* From, QualType ToType,
   //   A glvalue (3.10) of a non-function, non-array type T can
   //   be converted to a prvalue.
   bool argIsLValue = From->isGLValue();
-  if (argIsLValue &&
-      !FromType->isFunctionType() && !FromType->isArrayType() &&
+  if (argIsLValue && !FromType->canDecayToPointerType() &&
       S.Context.getCanonicalType(FromType) != S.Context.OverloadTy) {
     SCS.First = ICK_Lvalue_To_Rvalue;
 
@@ -2147,6 +2148,19 @@ static bool IsStandardConversion(Sema &S, Expr* From, QualType ToType,
     // is T (C++ 4.1p1). C++ can't get here with class types; in C, we
     // just strip the qualifiers because they don't matter.
     FromType = FromType.getUnqualifiedType();
+  } else if (S.getLangOpts().HLSL && FromType->isConstantArrayType() &&
+             ToType->isArrayParameterType()) {
+    // HLSL constant array parameters do not decay, so if the argument is a
+    // constant array and the parameter is an ArrayParameterType we have special
+    // handling here.
+    FromType = S.Context.getArrayParameterType(FromType);
+    if (S.Context.getCanonicalType(FromType) !=
+        S.Context.getCanonicalType(ToType))
+      return false;
+
+    SCS.First = ICK_HLSL_Array_RValue;
+    SCS.setAllToTypes(ToType);
+    return true;
   } else if (FromType->isArrayType()) {
     // Array-to-pointer conversion (C++ 4.2)
     SCS.First = ICK_Array_To_Pointer;
@@ -6100,6 +6114,7 @@ static bool CheckConvertedConstantConversions(Sema &S,
   case ICK_Lvalue_To_Rvalue:
   case ICK_Array_To_Pointer:
   case ICK_Function_To_Pointer:
+  case ICK_HLSL_Array_RValue:
     llvm_unreachable("found a first conversion kind in Second");
 
   case ICK_Function_Conversion:

diff --git a/clang/lib/Sema/SemaTemplate.cpp b/clang/lib/Sema/SemaTemplate.cpp
@@ -6898,6 +6898,11 @@ bool UnnamedLocalNoLinkageFinder::VisitBitIntType(const BitIntType *T) {
   return false;
 }
 
+bool UnnamedLocalNoLinkageFinder::VisitArrayParameterType(
+    const ArrayParameterType *T) {
+  return VisitConstantArrayType(T);
+}
+
 bool UnnamedLocalNoLinkageFinder::VisitDependentBitIntType(
     const DependentBitIntType *T) {
   return false;

diff --git a/clang/lib/Sema/SemaTemplateDeduction.cpp b/clang/lib/Sema/SemaTemplateDeduction.cpp
@@ -2277,6 +2277,7 @@ static TemplateDeductionResult DeduceTemplateArgumentsByTypeMatch(
     case Type::DependentTemplateSpecialization:
     case Type::PackExpansion:
     case Type::Pipe:
+    case Type::ArrayParameter:
       // No template argument deduction for these types
       return TemplateDeductionResult::Success;
 
@@ -6355,11 +6356,11 @@ MarkUsedTemplateParameters(ASTContext &Ctx, QualType T,
 
   case Type::ConstantArray:
   case Type::IncompleteArray:
+  case Type::ArrayParameter:
     MarkUsedTemplateParameters(Ctx,
                                cast<ArrayType>(T)->getElementType(),
                                OnlyDeduced, Depth, Used);
     break;
-
   case Type::Vector:
   case Type::ExtVector:
     MarkUsedTemplateParameters(Ctx,

diff --git a/clang/lib/Sema/SemaType.cpp b/clang/lib/Sema/SemaType.cpp
@@ -6505,6 +6505,9 @@ namespace {
     void VisitDecayedTypeLoc(DecayedTypeLoc TL) {
       llvm_unreachable("decayed type locs not expected here!");
     }
+    void VisitArrayParameterTypeLoc(ArrayParameterTypeLoc TL) {
+      llvm_unreachable("array parameter type locs not expected here!");
+    }
 
     void VisitAttributedTypeLoc(AttributedTypeLoc TL) {
       fillAttributedTypeLoc(TL, State);

diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
@@ -39,6 +39,7 @@
 #include "clang/Sema/ScopeInfo.h"
 #include "clang/Sema/SemaDiagnostic.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenACC.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/Support/ErrorHandling.h"
 #include <algorithm>
@@ -4000,16 +4001,16 @@ class TreeTransform {
                                             SourceLocation BeginLoc,
                                             SourceLocation EndLoc,
                                             StmtResult StrBlock) {
-    getSema().ActOnOpenACCConstruct(K, BeginLoc);
+    getSema().OpenACC().ActOnConstruct(K, BeginLoc);
 
     // TODO OpenACC: Include clauses.
-    if (getSema().ActOnStartOpenACCStmtDirective(K, BeginLoc))
+    if (getSema().OpenACC().ActOnStartStmtDirective(K, BeginLoc))
       return StmtError();
 
-    StrBlock = getSema().ActOnOpenACCAssociatedStmt(K, StrBlock);
+    StrBlock = getSema().OpenACC().ActOnAssociatedStmt(K, StrBlock);
 
-    return getSema().ActOnEndOpenACCStmtDirective(K, BeginLoc, EndLoc,
-                                                  StrBlock);
+    return getSema().OpenACC().ActOnEndStmtDirective(K, BeginLoc, EndLoc,
+                                                     StrBlock);
   }
 
 private:
@@ -5243,6 +5244,23 @@ QualType TreeTransform<Derived>::TransformDecayedType(TypeLocBuilder &TLB,
   return Result;
 }
 
+template <typename Derived>
+QualType
+TreeTransform<Derived>::TransformArrayParameterType(TypeLocBuilder &TLB,
+                                                    ArrayParameterTypeLoc TL) {
+  QualType OriginalType = getDerived().TransformType(TLB, TL.getElementLoc());
+  if (OriginalType.isNull())
+    return QualType();
+
+  QualType Result = TL.getType();
+  if (getDerived().AlwaysRebuild() ||
+      OriginalType != TL.getElementLoc().getType())
+    Result = SemaRef.Context.getArrayParameterType(OriginalType);
+  TLB.push<ArrayParameterTypeLoc>(Result);
+  // Nothing to set for ArrayParameterTypeLoc.
+  return Result;
+}
+
 template<typename Derived>
 QualType TreeTransform<Derived>::TransformPointerType(TypeLocBuilder &TLB,
                                                       PointerTypeLoc TL) {

diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
@@ -6810,6 +6810,10 @@ void TypeLocReader::VisitAdjustedTypeLoc(AdjustedTypeLoc TL) {
   // nothing to do
 }
 
+void TypeLocReader::VisitArrayParameterTypeLoc(ArrayParameterTypeLoc TL) {
+  // nothing to do
+}
+
 void TypeLocReader::VisitMacroQualifiedTypeLoc(MacroQualifiedTypeLoc TL) {
   TL.setExpansionLoc(readSourceLocation());
 }

diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
@@ -318,6 +318,10 @@ void TypeLocWriter::VisitAdjustedTypeLoc(AdjustedTypeLoc TL) {
   // nothing to do
 }
 
+void TypeLocWriter::VisitArrayParameterTypeLoc(ArrayParameterTypeLoc TL) {
+  // nothing to do
+}
+
 void TypeLocWriter::VisitBlockPointerTypeLoc(BlockPointerTypeLoc TL) {
   addSourceLocation(TL.getCaretLoc());
 }

diff --git a/clang/lib/StaticAnalyzer/Core/ExprEngineC.cpp b/clang/lib/StaticAnalyzer/Core/ExprEngineC.cpp
@@ -330,7 +330,8 @@ void ExprEngine::VisitCast(const CastExpr *CastE, const Expr *Ex,
       case CK_ConstructorConversion:
       case CK_UserDefinedConversion:
       case CK_FunctionToPointerDecay:
-      case CK_BuiltinFnToFnPtr: {
+      case CK_BuiltinFnToFnPtr:
+      case CK_HLSLArrayRValue: {
         // Copy the SVal of Ex to CastE.
         ProgramStateRef state = Pred->getState();
         const LocationContext *LCtx = Pred->getLocationContext();

diff --git a/clang/test/C/C99/n570.c b/clang/test/C/C99/n570.c
@@ -0,0 +1,31 @@
+// RUN: %clang_cc1 -verify -std=c99 %s
+// RUN: %clang_cc1 -E -std=c99 %s | FileCheck %s
+// expected-no-diagnostics
+
+/* WG14 N570: Yes
+ * Empty macro arguments
+ *
+ * NB: the original paper is not available online anywhere, so the test
+ * coverage is coming from what could be gleaned from the C99 rationale
+ * document. In C89, it was UB to pass no arguments to a function-like macro,
+ * and that's now supported in C99.
+ */
+
+#define TEN 10
+#define U u
+#define I // expands into no preprocessing tokens
+#define L L
+#define glue(a, b) a ## b
+#define xglue(a, b) glue(a, b)
+
+const unsigned u = xglue(TEN, U);
+const int i = xglue(TEN, I);
+const long l = xglue(TEN, L);
+
+// CHECK: const unsigned u = 10u;
+// CHECK-NEXT: const int i = 10;
+// CHECK-NEXT: const long l = 10L;
+
+_Static_assert(u == 10U, "");
+_Static_assert(i == 10, "");
+_Static_assert(l == 10L, "");
diff --git a/clang/test/CodeGenHLSL/ArrayTemporary.hlsl b/clang/test/CodeGenHLSL/ArrayTemporary.hlsl
@@ -0,0 +1,104 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s
+
+void fn(float x[2]) { }
+
+// CHECK-LABEL: define void {{.*}}call{{.*}}
+// CHECK: [[Arr:%.*]] = alloca [2 x float]
+// CHECK: [[Tmp:%.*]] = alloca [2 x float]
+// CHECK: call void @llvm.memset.p0.i32(ptr align 4 [[Arr]], i8 0, i32 8, i1 false)
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[Tmp]], ptr align 4 [[Arr]], i32 8, i1 false)
+// CHECK: call void {{.*}}fn{{.*}}(ptr noundef byval([2 x float]) align 4 [[Tmp]])
+void call() {
+  float Arr[2] = {0, 0};
+  fn(Arr);
+}
+
+struct Obj {
+  float V;
+  int X;
+};
+
+void fn2(Obj O[4]) { }
+
+// CHECK-LABEL: define void {{.*}}call2{{.*}}
+// CHECK: [[Arr:%.*]] = alloca [4 x %struct.Obj]
+// CHECK: [[Tmp:%.*]] = alloca [4 x %struct.Obj]
+// CHECK: call void @llvm.memset.p0.i32(ptr align 4 [[Arr]], i8 0, i32 32, i1 false)
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[Tmp]], ptr align 4 [[Arr]], i32 32, i1 false)
+// CHECK: call void {{.*}}fn2{{.*}}(ptr noundef byval([4 x %struct.Obj]) align 4 [[Tmp]])
+void call2() {
+  Obj Arr[4] = {};
+  fn2(Arr);
+}
+
+
+void fn3(float x[2][2]) { }
+
+// CHECK-LABEL: define void {{.*}}call3{{.*}}
+// CHECK: [[Arr:%.*]] = alloca [2 x [2 x float]]
+// CHECK: [[Tmp:%.*]] = alloca [2 x [2 x float]]
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[Arr]], ptr align 4 {{.*}}, i32 16, i1 false)
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[Tmp]], ptr align 4 [[Arr]], i32 16, i1 false)
+// CHECK: call void {{.*}}fn3{{.*}}(ptr noundef byval([2 x [2 x float]]) align 4 [[Tmp]])
+void call3() {
+  float Arr[2][2] = {{0, 0}, {1,1}};
+  fn3(Arr);
+}
+
+// CHECK-LABEL: define void {{.*}}call4{{.*}}(ptr
+// CHECK-SAME: noundef byval([2 x [2 x float]]) align 4 [[Arr:%.*]])
+// CHECK: [[Tmp:%.*]] = alloca [2 x [2 x float]]
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[Tmp]], ptr align 4 [[Arr]], i32 16, i1 false)
+// CHECK: call void {{.*}}fn3{{.*}}(ptr noundef byval([2 x [2 x float]]) align 4 [[Tmp]])
+
+void call4(float Arr[2][2]) {
+  fn3(Arr);
+}
+
+// Verify that each template instantiation codegens to a unique and correctly
+// mangled function name.
+
+// CHECK-LABEL: define void {{.*}}template_call{{.*}}(ptr
+
+// CHECK-SAME: noundef byval([2 x float]) align 4 [[FA2:%[0-9A-Z]+]],
+// CHECK-SAME: ptr noundef byval([4 x float]) align 4 [[FA4:%[0-9A-Z]+]],
+// CHECK-SAME: ptr noundef byval([3 x i32]) align 4 [[IA3:%[0-9A-Z]+]]
+
+// CHECK: [[Tmp1:%.*]] = alloca [2 x float]
+// CHECK: [[Tmp2:%.*]] = alloca [4 x float]
+// CHECK: [[Tmp3:%.*]] = alloca [3 x i32]
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[Tmp1]], ptr align 4 [[FA2]], i32 8, i1 false)
+// CHECK: call void @"??$template_fn@$$BY01M@@YAXY01M@Z"(ptr noundef byval([2 x float]) align 4 [[Tmp1]])
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[Tmp2]], ptr align 4 [[FA4]], i32 16, i1 false)
+// CHECK: call void @"??$template_fn@$$BY03M@@YAXY03M@Z"(ptr noundef byval([4 x float]) align 4 [[Tmp2]])
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[Tmp3]], ptr align 4 [[IA3]], i32 12, i1 false)
+// CHECK: call void @"??$template_fn@$$BY02H@@YAXY02H@Z"(ptr noundef byval([3 x i32]) align 4 [[Tmp3]])
+
+template<typename T>
+void template_fn(T Val) {}
+
+void template_call(float FA2[2], float FA4[4], int IA3[3]) {
+  template_fn(FA2);
+  template_fn(FA4);
+  template_fn(IA3);
+}
+
+
+// Verify that Array parameter element access correctly codegens.
+// CHECK-LABEL: define void {{.*}}element_access{{.*}}(ptr
+// CHECK-SAME: noundef byval([2 x float]) align 4 [[FA2:%[0-9A-Z]+]]
+
+// CHECK: [[Addr:%.*]] = getelementptr inbounds [2 x float], ptr [[FA2]], i32 0, i32 0
+// CHECK: [[Tmp:%.*]] = load float, ptr [[Addr]]
+// CHECK: call void @"??$template_fn@M@@YAXM@Z"(float noundef [[Tmp]])
+
+// CHECK: [[Idx0:%.*]] = getelementptr inbounds [2 x float], ptr [[FA2]], i32 0, i32 0
+// CHECK: [[Val0:%.*]] = load float, ptr [[Idx0]]
+// CHECK: [[Sum:%.*]] = fadd float [[Val0]], 5.000000e+00
+// CHECK: [[Idx1:%.*]] = getelementptr inbounds [2 x float], ptr [[FA2]], i32 0, i32 1
+// CHECK: store float [[Sum]], ptr [[Idx1]]
+
+void element_access(float FA2[2]) {
+  template_fn(FA2[0]);
+  FA2[1] = FA2[0] + 5;
+}
diff --git a/clang/test/Driver/cl-outputs.c b/clang/test/Driver/cl-outputs.c
@@ -301,5 +301,8 @@
 // RUN: %clang_cl -fdebug-compilation-dir=. /Z7 /Foa.obj -### -- %s 2>&1 | FileCheck -check-prefix=RELATIVE_OBJPATH1 %s
 // RELATIVE_OBJPATH1: "-object-file-name=a.obj"
 
+// RUN: %clang_cl -fdebug-compilation-dir=. /Z7 /Fo:a.obj -### -- %s 2>&1 | FileCheck -check-prefix=RELATIVE_OBJPATH1_COLON %s
+// RELATIVE_OBJPATH1_COLON: "-object-file-name=a.obj"
+
 // RUN: %clang_cl -fdebug-compilation-dir=. /Z7 /Fofoo/a.obj -### -- %s 2>&1 | FileCheck -check-prefix=RELATIVE_OBJPATH2 %s
 // RELATIVE_OBJPATH2: "-object-file-name=foo\\a.obj"
diff --git a/clang/test/SemaHLSL/ArrayParams.hlsl b/clang/test/SemaHLSL/ArrayParams.hlsl
@@ -0,0 +1,29 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library %s -verify
+
+void fn(int I[5]);     // #fn
+void fn2(int I[3][3]); // #fn2
+
+void call() {
+  float F[5];
+  double D[4];
+  int Long[9];
+  int Short[4];
+  int Same[5];
+
+  fn(F);      // expected-error{{no matching function for call to 'fn'}}
+              // expected-note@#fn{{candidate function not viable: no known conversion from 'float[5]' to 'int[5]' for 1st argument}}
+
+  fn(D);      // expected-error{{no matching function for call to 'fn'}}
+              // expected-note@#fn{{candidate function not viable: no known conversion from 'double[4]' to 'int[5]' for 1st argument}}
+
+  fn(Long);  // expected-error{{no matching function for call to 'fn'}}
+             // expected-note@#fn{{candidate function not viable: no known conversion from 'int[9]' to 'int[5]' for 1st argument}}
+
+  fn(Short); // expected-error{{no matching function for call to 'fn'}}
+             // expected-note@#fn{{candidate function not viable: no known conversion from 'int[4]' to 'int[5]' for 1st argument}}
+
+  fn(Same);  // totally fine, nothing to see here.
+
+  fn2(Long); // expected-error{{no matching function for call to 'fn2'}}
+             // expected-note@#fn2{{candidate function not viable: no known conversion from 'int[9]' to 'int[3][3]' for 1st argument}}
+}
diff --git a/clang/test/SemaHLSL/ArrayTemporary.hlsl b/clang/test/SemaHLSL/ArrayTemporary.hlsl
@@ -0,0 +1,95 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -ast-dump %s | FileCheck %s
+
+void fn(float x[2]) { }
+
+// CHECK: CallExpr {{.*}} 'void'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float[2])' <FunctionToPointerDecay>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float[2])' lvalue Function {{.*}} 'fn' 'void (float[2])'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'float[2]' <HLSLArrayRValue>
+
+void call() {
+  float Arr[2] = {0, 0};
+  fn(Arr);
+}
+
+struct Obj {
+  float V;
+  int X;
+};
+
+void fn2(Obj O[4]) { }
+
+// CHECK: CallExpr {{.*}} 'void'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(Obj[4])' <FunctionToPointerDecay>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'void (Obj[4])' lvalue Function {{.*}} 'fn2' 'void (Obj[4])'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'Obj[4]' <HLSLArrayRValue>
+
+void call2() {
+  Obj Arr[4] = {};
+  fn2(Arr);
+}
+
+
+void fn3(float x[2][2]) { }
+
+// CHECK: CallExpr {{.*}} 'void'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float[2][2])' <FunctionToPointerDecay>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float[2][2])' lvalue Function {{.*}} 'fn3' 'void (float[2][2])'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'float[2][2]' <HLSLArrayRValue>
+
+void call3() {
+  float Arr[2][2] = {{0, 0}, {1,1}};
+  fn3(Arr);
+}
+
+// This template function should be instantiated 3 times for the different array
+// types and lengths.
+
+// CHECK: FunctionTemplateDecl {{.*}} template_fn
+// CHECK-NEXT: TemplateTypeParmDecl {{.*}} referenced typename depth 0 index 0 T
+// CHECK-NEXT: FunctionDecl {{.*}} template_fn 'void (T)'
+// CHECK-NEXT: ParmVarDecl {{.*}} Val 'T'
+
+// CHECK: FunctionDecl {{.*}} used template_fn 'void (float[2])' implicit_instantiation
+// CHECK-NEXT: TemplateArgument type 'float[2]'
+// CHECK-NEXT: ArrayParameterType {{.*}} 'float[2]' 2
+// CHECK-NEXT: BuiltinType {{.*}} 'float'
+// CHECK-NEXT: ParmVarDecl {{.*}} Val 'float[2]'
+
+// CHECK: FunctionDecl {{.*}} used template_fn 'void (float[4])' implicit_instantiation
+// CHECK-NEXT: TemplateArgument type 'float[4]'
+// CHECK-NEXT: ArrayParameterType {{.*}} 'float[4]' 4
+// CHECK-NEXT: BuiltinType {{.*}} 'float'
+// CHECK-NEXT: ParmVarDecl {{.*}} Val 'float[4]'
+
+// CHECK: FunctionDecl {{.*}} used template_fn 'void (int[3])' implicit_instantiation
+// CHECK-NEXT: TemplateArgument type 'int[3]'
+// CHECK-NEXT: ArrayParameterType {{.*}} 'int[3]' 3
+// CHECK-NEXT: BuiltinType {{.*}} 'int'
+// CHECK-NEXT: ParmVarDecl {{.*}} Val 'int[3]'
+
+template<typename T>
+void template_fn(T Val) {}
+
+// CHECK: FunctionDecl {{.*}} call 'void (float[2], float[4], int[3])'
+// CHECK: CallExpr {{.*}} 'void'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float[2])' <FunctionToPointerDecay>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float[2])' lvalue Function {{.*}} 'template_fn' 'void (float[2])' (FunctionTemplate {{.*}} 'template_fn')
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'float[2]' <LValueToRValue>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'float[2]' lvalue ParmVar {{.*}} 'FA2' 'float[2]'
+// CHECK-NEXT: CallExpr {{.*}} 'void'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float[4])' <FunctionToPointerDecay>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float[4])' lvalue Function {{.*}} 'template_fn' 'void (float[4])' (FunctionTemplate {{.*}} 'template_fn')
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'float[4]' <LValueToRValue>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'float[4]' lvalue ParmVar {{.*}} 'FA4' 'float[4]'
+// CHECK-NEXT: CallExpr {{.*}} 'void'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(int[3])' <FunctionToPointerDecay>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'void (int[3])' lvalue Function {{.*}} 'template_fn' 'void (int[3])' (FunctionTemplate {{.*}} 'template_fn')
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'int[3]' <LValueToRValue>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'int[3]' lvalue ParmVar {{.*}} 'IA3' 'int[3]'
+
+void call(float FA2[2], float FA4[4], int IA3[3]) {
+  template_fn(FA2);
+  template_fn(FA4);
+  template_fn(IA3);
+}
diff --git a/clang/test/SemaHLSL/ArrayTemporary.ll b/clang/test/SemaHLSL/ArrayTemporary.ll
@@ -0,0 +1,76 @@
+; ModuleID = '/Users/cbieneman/dev/llvm-project/clang/test/SemaHLSL/ArrayTemporary.hlsl'
+source_filename = "/Users/cbieneman/dev/llvm-project/clang/test/SemaHLSL/ArrayTemporary.hlsl"
+target datalayout = "e-m:e-p:32:32-i1:32-i8:8-i16:16-i32:32-i64:64-f16:16-f32:32-f64:64-n8:16:32:64"
+target triple = "dxil-pc-shadermodel6.3-library"
+
+%struct.Obj = type { float, i32 }
+
+@"__const.?call3@@YAXXZ.Arr" = private unnamed_addr constant [2 x [2 x float]] [[2 x float] zeroinitializer, [2 x float] [float 1.000000e+00, float 1.000000e+00]], align 4
+
+; Function Attrs: noinline nounwind optnone
+define void @"?fn@@YAXY01M@Z"(ptr noundef byval([2 x float]) align 4 %x) #0 {
+entry:
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define void @"?call@@YAXXZ"() #0 {
+entry:
+  %Arr = alloca [2 x float], align 4
+  %agg.tmp = alloca [2 x float], align 4
+  call void @llvm.memset.p0.i32(ptr align 4 %Arr, i8 0, i32 8, i1 false)
+  call void @llvm.memcpy.p0.p0.i32(ptr align 4 %agg.tmp, ptr align 4 %Arr, i32 8, i1 false)
+  call void @"?fn@@YAXY01M@Z"(ptr noundef byval([2 x float]) align 4 %agg.tmp)
+  ret void
+}
+
+; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: write)
+declare void @llvm.memset.p0.i32(ptr nocapture writeonly, i8, i32, i1 immarg) #1
+
+; Function Attrs: nocallback nofree nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.memcpy.p0.p0.i32(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i32, i1 immarg) #2
+
+; Function Attrs: noinline nounwind optnone
+define void @"?fn2@@YAXY03UObj@@@Z"(ptr noundef byval([4 x %struct.Obj]) align 4 %O) #0 {
+entry:
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define void @"?call2@@YAXXZ"() #0 {
+entry:
+  %Arr = alloca [4 x %struct.Obj], align 4
+  %agg.tmp = alloca [4 x %struct.Obj], align 4
+  call void @llvm.memset.p0.i32(ptr align 4 %Arr, i8 0, i32 32, i1 false)
+  call void @llvm.memcpy.p0.p0.i32(ptr align 4 %agg.tmp, ptr align 4 %Arr, i32 32, i1 false)
+  call void @"?fn2@@YAXY03UObj@@@Z"(ptr noundef byval([4 x %struct.Obj]) align 4 %agg.tmp)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define void @"?fn3@@YAXY111M@Z"(ptr noundef byval([2 x [2 x float]]) align 4 %x) #0 {
+entry:
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define void @"?call3@@YAXXZ"() #0 {
+entry:
+  %Arr = alloca [2 x [2 x float]], align 4
+  %agg.tmp = alloca [2 x [2 x float]], align 4
+  call void @llvm.memcpy.p0.p0.i32(ptr align 4 %Arr, ptr align 4 @"__const.?call3@@YAXXZ.Arr", i32 16, i1 false)
+  call void @llvm.memcpy.p0.p0.i32(ptr align 4 %agg.tmp, ptr align 4 %Arr, i32 16, i1 false)
+  call void @"?fn3@@YAXY111M@Z"(ptr noundef byval([2 x [2 x float]]) align 4 %agg.tmp)
+  ret void
+}
+
+attributes #0 = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+attributes #1 = { nocallback nofree nounwind willreturn memory(argmem: write) }
+attributes #2 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }
+
+!llvm.module.flags = !{!0, !1}
+!llvm.ident = !{!2}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 4, !"dx.disable_optimizations", i32 1}
+!2 = !{!"clang version 19.0.0git (git@github.com:llvm/llvm-project.git 64e1c15c520cf11114ef2ddd887e76560903db2b)"}
diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp
@@ -1923,6 +1923,7 @@ DEFAULT_TYPELOC_IMPL(ConstantArray, ArrayType)
 DEFAULT_TYPELOC_IMPL(IncompleteArray, ArrayType)
 DEFAULT_TYPELOC_IMPL(VariableArray, ArrayType)
 DEFAULT_TYPELOC_IMPL(DependentSizedArray, ArrayType)
+DEFAULT_TYPELOC_IMPL(ArrayParameter, ConstantArrayType)
 DEFAULT_TYPELOC_IMPL(DependentAddressSpace, Type)
 DEFAULT_TYPELOC_IMPL(DependentVector, Type)
 DEFAULT_TYPELOC_IMPL(DependentSizedExtVector, Type)

diff --git a/clang/www/c_status.html b/clang/www/c_status.html
@@ -300,7 +300,7 @@ <h2 id="c99">C99 implementation status</h2>
     <tr>
       <td>empty macro arguments</td>
       <td>N570</td>
-      <td class="unknown" align="center">Unknown</td>
+      <td class="full" align="center">Yes</td>
     </tr>
     <tr>
       <td>new structure type compatibility (tag compatibility)</td>

diff --git a/compiler-rt/lib/scudo/standalone/wrappers_c.inc b/compiler-rt/lib/scudo/standalone/wrappers_c.inc
@@ -247,6 +247,11 @@ void SCUDO_PREFIX(malloc_postinit)() {
 INTERFACE WEAK int SCUDO_PREFIX(mallopt)(int param, int value) {
   if (param == M_DECAY_TIME) {
     if (SCUDO_ANDROID) {
+      // Before changing the interval, reset the memory usage status by doing a
+      // M_PURGE call so that we can minimize the impact of any unreleased pages
+      // introduced by interval transition.
+      SCUDO_ALLOCATOR.releaseToOS(scudo::ReleaseToOS::Force);
+
       if (value == 0) {
         // Will set the release values to their minimum values.
         value = INT32_MIN;

diff --git a/compiler-rt/test/profile/Linux/instrprof-vtable-value-prof.cpp b/compiler-rt/test/profile/Linux/instrprof-vtable-value-prof.cpp
@@ -1,5 +1,10 @@
 // REQUIRES: lld-available
 
+// Building the instrumented binary will fail because lld doesn't support
+// big-endian ELF for PPC (aka ABI 1).
+// ld.lld: error: /lib/../lib64/Scrt1.o: ABI version 1 is not supported
+// UNSUPPORTED: ppc && host-byteorder-big-endian
+
 // RUN: %clangxx_pgogen -fuse-ld=lld -O2 -g -fprofile-generate=. -mllvm -enable-vtable-value-profiling %s -o %t-test
 // RUN: env LLVM_PROFILE_FILE=%t-test.profraw %t-test
 

diff --git a/libc/config/baremetal/arm/entrypoints.txt b/libc/config/baremetal/arm/entrypoints.txt
@@ -207,6 +207,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.acoshf
     libc.src.math.asinf
     libc.src.math.asinhf
+    libc.src.math.atan2f
     libc.src.math.atanf
     libc.src.math.atanhf
     libc.src.math.ceil

diff --git a/libc/config/baremetal/riscv/entrypoints.txt b/libc/config/baremetal/riscv/entrypoints.txt
@@ -207,6 +207,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.acoshf
     libc.src.math.asinf
     libc.src.math.asinhf
+    libc.src.math.atan2f
     libc.src.math.atanf
     libc.src.math.atanhf
     libc.src.math.ceil

diff --git a/libc/config/darwin/arm/entrypoints.txt b/libc/config/darwin/arm/entrypoints.txt
@@ -118,6 +118,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.acoshf
     libc.src.math.asinf
     libc.src.math.asinhf
+    libc.src.math.atan2f
     libc.src.math.atanf
     libc.src.math.atanhf
     libc.src.math.copysign

diff --git a/libc/config/linux/aarch64/entrypoints.txt b/libc/config/linux/aarch64/entrypoints.txt
@@ -330,6 +330,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.acoshf
     libc.src.math.asinf
     libc.src.math.asinhf
+    libc.src.math.atan2f
     libc.src.math.atanf
     libc.src.math.atanhf
     libc.src.math.copysign

diff --git a/libc/config/linux/arm/entrypoints.txt b/libc/config/linux/arm/entrypoints.txt
@@ -198,6 +198,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.acoshf
     libc.src.math.asinf
     libc.src.math.asinhf
+    libc.src.math.atan2f
     libc.src.math.atanf
     libc.src.math.atanhf
     libc.src.math.ceil

diff --git a/libc/config/linux/riscv/entrypoints.txt b/libc/config/linux/riscv/entrypoints.txt
@@ -338,6 +338,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.acoshf
     libc.src.math.asinf
     libc.src.math.asinhf
+    libc.src.math.atan2f
     libc.src.math.atanf
     libc.src.math.atanhf
     libc.src.math.copysign

diff --git a/libc/config/linux/x86_64/entrypoints.txt b/libc/config/linux/x86_64/entrypoints.txt
@@ -348,6 +348,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.acoshf
     libc.src.math.asinf
     libc.src.math.asinhf
+    libc.src.math.atan2f
     libc.src.math.atanf
     libc.src.math.atanhf
     libc.src.math.canonicalize
@@ -681,6 +682,8 @@ if(LLVM_LIBC_FULL_BUILD)
     libc.src.stdio.fread_unlocked
     libc.src.stdio.fseek
     libc.src.stdio.ftell
+    libc.src.stdio.fseeko
+    libc.src.stdio.ftello
     libc.src.stdio.funlockfile
     libc.src.stdio.fwrite
     libc.src.stdio.fwrite_unlocked

diff --git a/libc/config/windows/entrypoints.txt b/libc/config/windows/entrypoints.txt
@@ -116,6 +116,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.acoshf
     libc.src.math.asinf
     libc.src.math.asinhf
+    libc.src.math.atan2f
     libc.src.math.atanf
     libc.src.math.atanhf
     libc.src.math.copysign

diff --git a/libc/docs/math/index.rst b/libc/docs/math/index.rst
@@ -422,7 +422,7 @@ Higher Math Functions
 +------------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+
 | atan2      |         |         |         |         |         |         |         |         |         |         |         |         |
 +------------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+
-| atan2f     |         |         |         |         |         |         |         |         |         |         |         |         |
+| atan2f     | |check| | |check| | |check| | |check| | |check| |         |         | |check| | |check| | |check| |         |         |
 +------------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+
 | atan2l     |         |         |         |         |         |         |         |         |         |         |         |         |
 +------------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+---------+
@@ -591,6 +591,7 @@ acosh          |check|
 asin           |check|
 asinh          |check|
 atan           |check|
+atan2          |check|
 atanh          |check|
 cos            |check|          large
 cosh           |check|

diff --git a/libc/spec/stdc.td b/libc/spec/stdc.td
@@ -620,10 +620,14 @@ def StdC : StandardSpec<"stdc"> {
           FunctionSpec<"tanhf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
 
           FunctionSpec<"acosf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
+
           FunctionSpec<"asinf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
           FunctionSpec<"asin", RetValSpec<DoubleType>, [ArgSpec<DoubleType>]>,
+
           FunctionSpec<"atanf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
 
+          FunctionSpec<"atan2f", RetValSpec<FloatType>, [ArgSpec<FloatType>, ArgSpec<FloatType>]>,
+
           FunctionSpec<"acoshf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
           FunctionSpec<"asinhf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
           FunctionSpec<"atanhf", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,

diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
@@ -2854,6 +2854,24 @@ add_entrypoint_object(
     -O3
 )
 
+add_entrypoint_object(
+  atan2f
+  SRCS
+    atan2f.cpp
+  HDRS
+    ../atan2f.h
+  COMPILE_OPTIONS
+    -O3
+  DEPENDS
+    .inv_trigf_utils
+    libc.src.__support.FPUtil.fp_bits
+    libc.src.__support.FPUtil.multiply_add
+    libc.src.__support.FPUtil.nearest_integer
+    libc.src.__support.FPUtil.polyeval
+    libc.src.__support.FPUtil.rounding_mode
+    libc.src.__support.macros.optimization
+)
+
 add_entrypoint_object(
   scalbn
   SRCS

diff --git a/libc/src/math/generic/atan2f.cpp b/libc/src/math/generic/atan2f.cpp
@@ -0,0 +1,306 @@
+//===-- Single-precision atan2f function ----------------------------------===//
+//
+// 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 "src/math/atan2f.h"
+#include "inv_trigf_utils.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/PolyEval.h"
+#include "src/__support/FPUtil/double_double.h"
+#include "src/__support/FPUtil/multiply_add.h"
+#include "src/__support/FPUtil/nearest_integer.h"
+#include "src/__support/FPUtil/rounding_mode.h"
+#include "src/__support/macros/optimization.h" // LIBC_UNLIKELY
+
+namespace LIBC_NAMESPACE {
+
+namespace {
+
+// Look up tables for accurate pass:
+
+// atan(i/16) with i = 0..16, generated by Sollya with:
+// > for i from 0 to 16 do {
+//     a = round(atan(i/16), D, RN);
+//     b = round(atan(i/16) - a, D, RN);
+//     print("{", b, ",", a, "},");
+//   };
+constexpr fputil::DoubleDouble ATAN_I[17] = {
+    {0.0, 0.0},
+    {-0x1.c934d86d23f1dp-60, 0x1.ff55bb72cfdeap-5},
+    {-0x1.cd37686760c17p-59, 0x1.fd5ba9aac2f6ep-4},
+    {0x1.347b0b4f881cap-58, 0x1.7b97b4bce5b02p-3},
+    {0x1.8ab6e3cf7afbdp-57, 0x1.f5b75f92c80ddp-3},
+    {-0x1.963a544b672d8p-57, 0x1.362773707ebccp-2},
+    {-0x1.c63aae6f6e918p-56, 0x1.6f61941e4def1p-2},
+    {-0x1.24dec1b50b7ffp-56, 0x1.a64eec3cc23fdp-2},
+    {0x1.a2b7f222f65e2p-56, 0x1.dac670561bb4fp-2},
+    {-0x1.d5b495f6349e6p-56, 0x1.0657e94db30dp-1},
+    {-0x1.928df287a668fp-58, 0x1.1e00babdefeb4p-1},
+    {0x1.1021137c71102p-55, 0x1.345f01cce37bbp-1},
+    {0x1.2419a87f2a458p-56, 0x1.4978fa3269ee1p-1},
+    {0x1.0028e4bc5e7cap-57, 0x1.5d58987169b18p-1},
+    {-0x1.8c34d25aadef6p-56, 0x1.700a7c5784634p-1},
+    {-0x1.bf76229d3b917p-56, 0x1.819d0b7158a4dp-1},
+    {0x1.1a62633145c07p-55, 0x1.921fb54442d18p-1},
+};
+
+// Taylor polynomial, generated by Sollya with:
+// > for i from 0 to 8 do {
+//     j = (-1)^(i + 1)/(2*i + 1);
+//     a = round(j, D, RN);
+//     b = round(j - a, D, RN);
+//     print("{", b, ",", a, "},");
+//   };
+constexpr fputil::DoubleDouble COEFFS[9] = {
+    {0.0, 1.0},                                      // 1
+    {-0x1.5555555555555p-56, -0x1.5555555555555p-2}, // -1/3
+    {-0x1.999999999999ap-57, 0x1.999999999999ap-3},  // 1/5
+    {-0x1.2492492492492p-57, -0x1.2492492492492p-3}, // -1/7
+    {0x1.c71c71c71c71cp-58, 0x1.c71c71c71c71cp-4},   // 1/9
+    {0x1.745d1745d1746p-59, -0x1.745d1745d1746p-4},  // -1/11
+    {-0x1.3b13b13b13b14p-58, 0x1.3b13b13b13b14p-4},  // 1/13
+    {-0x1.1111111111111p-60, -0x1.1111111111111p-4}, // -1/15
+    {0x1.e1e1e1e1e1e1ep-61, 0x1.e1e1e1e1e1e1ep-5},   // 1/17
+};
+
+// Veltkamp's splitting of a double precision into hi + lo, where the hi part is
+// slightly smaller than an even split, so that the product of
+//   hi * (s1 * k + s2) is exact,
+// where:
+//   s1, s2 are single precsion,
+//   1/16 <= s1/s2 <= 1
+//   1/16 <= k <= 1 is an integer.
+// So the maximal precision of (s1 * k + s2) is:
+//   prec(s1 * k + s2) = 2 + log2(msb(s2)) - log2(lsb(k_d * s1))
+//                     = 2 + log2(msb(s1)) + 4 - log2(lsb(k_d)) - log2(lsb(s1))
+//                     = 2 + log2(lsb(s1)) + 23 + 4 - (-4) - log2(lsb(s1))
+//                     = 33.
+// Thus, the Veltkamp splitting constant is C = 2^33 + 1.
+// This is used when FMA instruction is not available.
+[[maybe_unused]] constexpr fputil::DoubleDouble split_d(double a) {
+  fputil::DoubleDouble r{0.0, 0.0};
+  constexpr double C = 0x1.0p33 + 1.0;
+  double t1 = C * a;
+  double t2 = a - t1;
+  r.hi = t1 + t2;
+  r.lo = a - r.hi;
+  return r;
+}
+
+// Compute atan( num_d / den_d ) in double-double precision.
+//   num_d      = min(|x|, |y|)
+//   den_d      = max(|x|, |y|)
+//   q_d        = num_d / den_d
+//   idx, k_d   = round( 2^4 * num_d / den_d )
+//   final_sign = sign of the final result
+//   const_term = the constant term in the final expression.
+float atan2f_double_double(double num_d, double den_d, double q_d, int idx,
+                           double k_d, double final_sign,
+                           const fputil::DoubleDouble &const_term) {
+  fputil::DoubleDouble q;
+  double num_r, den_r;
+
+  if (idx != 0) {
+    // The following range reduction is accurate even without fma for
+    //   1/16 <= n/d <= 1.
+    // atan(n/d) - atan(idx/16) = atan((n/d - idx/16) / (1 + (n/d) * (idx/16)))
+    //                          = atan((n - d*(idx/16)) / (d + n*idx/16))
+    k_d *= 0x1.0p-4;
+    num_r = fputil::multiply_add(k_d, -den_d, num_d); // Exact
+    den_r = fputil::multiply_add(k_d, num_d, den_d);  // Exact
+    q.hi = num_r / den_r;
+  } else {
+    // For 0 < n/d < 1/16, we just need to calculate the lower part of their
+    // quotient.
+    q.hi = q_d;
+    num_r = num_d;
+    den_r = den_d;
+  }
+#ifdef LIBC_TARGET_CPU_HAS_FMA
+  q.lo = fputil::multiply_add(q.hi, -den_r, num_r) / den_r;
+#else
+  // Compute `(num_r - q.hi * den_r) / den_r` accurately without FMA
+  // instructions.
+  fputil::DoubleDouble q_hi_dd = split_d(q.hi);
+  double t1 = fputil::multiply_add(q_hi_dd.hi, -den_r, num_r); // Exact
+  double t2 = fputil::multiply_add(q_hi_dd.lo, -den_r, t1);
+  q.lo = t2 / den_r;
+#endif // LIBC_TARGET_CPU_HAS_FMA
+
+  // Taylor polynomial, evaluating using Horner's scheme:
+  //   P = x - x^3/3 + x^5/5 -x^7/7 + x^9/9 - x^11/11 + x^13/13 - x^15/15
+  //       + x^17/17
+  //     = x*(1 + x^2*(-1/3 + x^2*(1/5 + x^2*(-1/7 + x^2*(1/9 + x^2*
+  //          *(-1/11 + x^2*(1/13 + x^2*(-1/15 + x^2 * 1/17))))))))
+  fputil::DoubleDouble q2 = fputil::quick_mult(q, q);
+  fputil::DoubleDouble p_dd =
+      fputil::polyeval(q2, COEFFS[0], COEFFS[1], COEFFS[2], COEFFS[3],
+                       COEFFS[4], COEFFS[5], COEFFS[6], COEFFS[7], COEFFS[8]);
+  fputil::DoubleDouble r_dd =
+      fputil::add(const_term, fputil::multiply_add(q, p_dd, ATAN_I[idx]));
+  r_dd.hi *= final_sign;
+  r_dd.lo *= final_sign;
+
+  // Make sure the sum is normalized:
+  fputil::DoubleDouble rr = fputil::exact_add(r_dd.hi, r_dd.lo);
+  // Round to odd.
+  uint64_t rr_bits = cpp::bit_cast<uint64_t>(rr.hi);
+  if (LIBC_UNLIKELY(((rr_bits & 0xfff'ffff) == 0) && (rr.lo != 0.0))) {
+    Sign hi_sign = fputil::FPBits<double>(rr.hi).sign();
+    Sign lo_sign = fputil::FPBits<double>(rr.lo).sign();
+    if (hi_sign == lo_sign) {
+      ++rr_bits;
+    } else if ((rr_bits & fputil::FPBits<double>::FRACTION_MASK) > 0) {
+      --rr_bits;
+    }
+  }
+
+  return static_cast<float>(cpp::bit_cast<double>(rr_bits));
+}
+
+} // anonymous namespace
+
+// There are several range reduction steps we can take for atan2(y, x) as
+// follow:
+
+// * Range reduction 1: signness
+// atan2(y, x) will return a number between -PI and PI representing the angle
+// forming by the 0x axis and the vector (x, y) on the 0xy-plane.
+// In particular, we have that:
+//   atan2(y, x) = atan( y/x )         if x >= 0 and y >= 0 (I-quadrant)
+//               = pi + atan( y/x )    if x < 0 and y >= 0  (II-quadrant)
+//               = -pi + atan( y/x )   if x < 0 and y < 0   (III-quadrant)
+//               = atan( y/x )         if x >= 0 and y < 0  (IV-quadrant)
+// Since atan function is odd, we can use the formula:
+//   atan(-u) = -atan(u)
+// to adjust the above conditions a bit further:
+//   atan2(y, x) = atan( |y|/|x| )         if x >= 0 and y >= 0 (I-quadrant)
+//               = pi - atan( |y|/|x| )    if x < 0 and y >= 0  (II-quadrant)
+//               = -pi + atan( |y|/|x| )   if x < 0 and y < 0   (III-quadrant)
+//               = -atan( |y|/|x| )        if x >= 0 and y < 0  (IV-quadrant)
+// Which can be simplified to:
+//   atan2(y, x) = sign(y) * atan( |y|/|x| )             if x >= 0
+//               = sign(y) * (pi - atan( |y|/|x| ))      if x < 0
+
+// * Range reduction 2: reciprocal
+// Now that the argument inside atan is positive, we can use the formula:
+//   atan(1/x) = pi/2 - atan(x)
+// to make the argument inside atan <= 1 as follow:
+//   atan2(y, x) = sign(y) * atan( |y|/|x|)            if 0 <= |y| <= x
+//               = sign(y) * (pi/2 - atan( |x|/|y| )   if 0 <= x < |y|
+//               = sign(y) * (pi - atan( |y|/|x| ))    if 0 <= |y| <= -x
+//               = sign(y) * (pi/2 + atan( |x|/|y| ))  if 0 <= -x < |y|
+
+// * Range reduction 3: look up table.
+// After the previous two range reduction steps, we reduce the problem to
+// compute atan(u) with 0 <= u <= 1, or to be precise:
+//   atan( n / d ) where n = min(|x|, |y|) and d = max(|x|, |y|).
+// An accurate polynomial approximation for the whole [0, 1] input range will
+// require a very large degree.  To make it more efficient, we reduce the input
+// range further by finding an integer idx such that:
+//   | n/d - idx/16 | <= 1/32.
+// In particular,
+//   idx := 2^-4 * round(2^4 * n/d)
+// Then for the fast pass, we find a polynomial approximation for:
+//   atan( n/d ) ~ atan( idx/16 ) + (n/d - idx/16) * Q(n/d - idx/16)
+// For the accurate pass, we use the addition formula:
+//   atan( n/d ) - atan( idx/16 ) = atan( (n/d - idx/16)/(1 + (n*idx)/(16*d)) )
+//                                = atan( (n - d * idx/16)/(d + n * idx/16) )
+// And finally we use Taylor polynomial to compute the RHS in the accurate pass:
+//   atan(u) ~ P(u) = u - u^3/3 + u^5/5 - u^7/7 + u^9/9 - u^11/11 + u^13/13 -
+//                      - u^15/15 + u^17/17
+// It's error in double-double precision is estimated in Sollya to be:
+// > P = x - x^3/3 + x^5/5 -x^7/7 + x^9/9 - x^11/11 + x^13/13 - x^15/15
+//       + x^17/17;
+// > dirtyinfnorm(atan(x) - P, [-2^-5, 2^-5]);
+// 0x1.aec6f...p-100
+// which is about rounding errors of double-double (2^-104).
+
+LLVM_LIBC_FUNCTION(float, atan2f, (float y, float x)) {
+  using FPBits = typename fputil::FPBits<float>;
+  constexpr double IS_NEG[2] = {1.0, -1.0};
+  constexpr double PI = 0x1.921fb54442d18p1;
+  constexpr double PI_LO = 0x1.1a62633145c07p-53;
+  constexpr double PI_OVER_4 = 0x1.921fb54442d18p-1;
+  constexpr double PI_OVER_2 = 0x1.921fb54442d18p0;
+  constexpr double THREE_PI_OVER_4 = 0x1.2d97c7f3321d2p+1;
+  // Adjustment for constant term:
+  //   CONST_ADJ[x_sign][y_sign][recip]
+  constexpr fputil::DoubleDouble CONST_ADJ[2][2][2] = {
+      {{{0.0, 0.0}, {-PI_LO / 2, -PI_OVER_2}},
+       {{-0.0, -0.0}, {-PI_LO / 2, -PI_OVER_2}}},
+      {{{-PI_LO, -PI}, {PI_LO / 2, PI_OVER_2}},
+       {{-PI_LO, -PI}, {PI_LO / 2, PI_OVER_2}}}};
+
+  FPBits x_bits(x), y_bits(y);
+  bool x_sign = x_bits.sign().is_neg();
+  bool y_sign = y_bits.sign().is_neg();
+  x_bits.set_sign(Sign::POS);
+  y_bits.set_sign(Sign::POS);
+  uint32_t x_abs = x_bits.uintval();
+  uint32_t y_abs = y_bits.uintval();
+  uint32_t max_abs = x_abs > y_abs ? x_abs : y_abs;
+  uint32_t min_abs = x_abs <= y_abs ? x_abs : y_abs;
+
+  if (LIBC_UNLIKELY(max_abs >= 0x7f80'0000U || min_abs == 0U)) {
+    if (x_bits.is_nan() || y_bits.is_nan())
+      return FPBits::quiet_nan().get_val();
+    size_t x_except = x_abs == 0 ? 0 : (x_abs == 0x7f80'0000 ? 2 : 1);
+    size_t y_except = y_abs == 0 ? 0 : (y_abs == 0x7f80'0000 ? 2 : 1);
+
+    // Exceptional cases:
+    //   EXCEPT[y_except][x_except][x_is_neg]
+    // with x_except & y_except:
+    //   0: zero
+    //   1: finite, non-zero
+    //   2: infinity
+    constexpr double EXCEPTS[3][3][2] = {
+        {{0.0, PI}, {0.0, PI}, {0.0, PI}},
+        {{PI_OVER_2, PI_OVER_2}, {0.0, 0.0}, {0.0, PI}},
+        {{PI_OVER_2, PI_OVER_2},
+         {PI_OVER_2, PI_OVER_2},
+         {PI_OVER_4, THREE_PI_OVER_4}},
+    };
+
+    double r = IS_NEG[y_sign] * EXCEPTS[y_except][x_except][x_sign];
+
+    return static_cast<float>(r);
+  }
+
+  bool recip = x_abs < y_abs;
+  double final_sign = IS_NEG[(x_sign != y_sign) != recip];
+  fputil::DoubleDouble const_term = CONST_ADJ[x_sign][y_sign][recip];
+  double num_d = static_cast<double>(FPBits(min_abs).get_val());
+  double den_d = static_cast<double>(FPBits(max_abs).get_val());
+  double q_d = num_d / den_d;
+
+  double k_d = fputil::nearest_integer(q_d * 0x1.0p4f);
+  int idx = static_cast<int>(k_d);
+  q_d = fputil::multiply_add(k_d, -0x1.0p-4, q_d);
+
+  double p = atan_eval(q_d, idx);
+  double r = final_sign *
+             fputil::multiply_add(q_d, p, const_term.hi + ATAN_COEFFS[idx][0]);
+
+  constexpr uint32_t LOWER_ERR = 4;
+  // Mask sticky bits in double precision before rounding to single precision.
+  constexpr uint32_t MASK =
+      mask_trailing_ones<uint32_t, fputil::FPBits<double>::SIG_LEN -
+                                       FPBits::SIG_LEN - 1>();
+  constexpr uint32_t UPPER_ERR = MASK - LOWER_ERR;
+
+  uint32_t r_bits = static_cast<uint32_t>(cpp::bit_cast<uint64_t>(r)) & MASK;
+
+  // Ziv's rounding test.
+  if (LIBC_LIKELY(r_bits > LOWER_ERR && r_bits < UPPER_ERR))
+    return static_cast<float>(r);
+
+  return atan2f_double_double(num_d, den_d, q_d, idx, k_d, final_sign,
+                              const_term);
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/math/generic/atanf.cpp b/libc/src/math/generic/atanf.cpp
@@ -60,11 +60,15 @@ LLVM_LIBC_FUNCTION(float, atanf, (float x)) {
     }
     // Use Taylor polynomial:
     //   atan(x) ~ x * (1 - x^2 / 3 + x^4 / 5 - x^6 / 7 + x^8 / 9 - x^10 / 11).
+    constexpr double ATAN_TAYLOR[6] = {
+        0x1.0000000000000p+0,  -0x1.5555555555555p-2, 0x1.999999999999ap-3,
+        -0x1.2492492492492p-3, 0x1.c71c71c71c71cp-4,  -0x1.745d1745d1746p-4,
+    };
     double x2 = x_d * x_d;
     double x4 = x2 * x2;
-    double c0 = fputil::multiply_add(x2, ATAN_COEFFS[0][1], ATAN_COEFFS[0][0]);
-    double c1 = fputil::multiply_add(x2, ATAN_COEFFS[0][3], ATAN_COEFFS[0][2]);
-    double c2 = fputil::multiply_add(x2, ATAN_COEFFS[0][5], ATAN_COEFFS[0][4]);
+    double c0 = fputil::multiply_add(x2, ATAN_TAYLOR[1], ATAN_TAYLOR[0]);
+    double c1 = fputil::multiply_add(x2, ATAN_TAYLOR[3], ATAN_TAYLOR[2]);
+    double c2 = fputil::multiply_add(x2, ATAN_TAYLOR[5], ATAN_TAYLOR[4]);
     double p = fputil::polyeval(x4, c0, c1, c2);
     double r = fputil::multiply_add(x_d, p, const_term);
     return static_cast<float>(r);
@@ -81,11 +85,6 @@ LLVM_LIBC_FUNCTION(float, atanf, (float x)) {
   int idx;
 
   if (x_abs > 0x3f80'0000U) {
-    // Exceptional value:
-    if (LIBC_UNLIKELY(x_abs == 0x3ffe'2ec1U)) { // |x| = 0x1.fc5d82p+0
-      return sign.is_pos() ? fputil::round_result_slightly_up(0x1.1ab2fp0f)
-                           : fputil::round_result_slightly_down(-0x1.1ab2fp0f);
-    }
     // |x| > 1, we need to invert x, so we will perform range reduction in
     // double precision.
     x_d = 1.0 / static_cast<double>(x_bits.get_val());
@@ -98,10 +97,9 @@ LLVM_LIBC_FUNCTION(float, atanf, (float x)) {
                                       SIGNED_PI_OVER_2[sign.is_neg()]);
   } else {
     // Exceptional value:
-    if (LIBC_UNLIKELY(x_abs == 0x3dbb'6ac7U)) { // |x| = 0x1.76d58ep-4
-      return sign.is_pos()
-                 ? fputil::round_result_slightly_up(0x1.75cb06p-4f)
-                 : fputil::round_result_slightly_down(-0x1.75cb06p-4f);
+    if (LIBC_UNLIKELY(x_abs == 0x3d8d'6b23U)) { // |x| = 0x1.1ad646p-4
+      return sign.is_pos() ? fputil::round_result_slightly_down(0x1.1a6386p-4f)
+                           : fputil::round_result_slightly_up(-0x1.1a6386p-4f);
     }
     // Perform range reduction in single precision.
     float x_f = x_bits.get_val();

diff --git a/libc/src/math/generic/inv_trigf_utils.cpp b/libc/src/math/generic/inv_trigf_utils.cpp
@@ -16,65 +16,70 @@ namespace LIBC_NAMESPACE {
 // Generated by Sollya with:
 // > for i from 1 to 16 do {
 //     mid_point = i/16;
-//     P = fpminimax(atan(mid_point + x), 7, [|D...|], [-1/32, 1/32]);
+//     P = fpminimax(atan(mid_point + x), 8, [|D...|], [-1/32, 1/32]);
 //     print("{", coeff(P, 0), ",", coeff(P, 1), ",", coeff(P, 2), ",",
 //           coeff(P, 3), ",", coeff(P, 4), ",", coeff(P, 5), ",", coeff(P, 6),
-//           ",", coeff(P, 7), "},");
+//           ",", coeff(P, 7), ",", coeff(P, 8), "},");
 //   };
-// For i = 0, ATAN_COEFFS[0][j] = (-1)^j * (1/(2*j + 1)) is the odd coefficients
-// of the Taylor polynomial of atan(x).
-double ATAN_COEFFS[17][8] = {
-    {0x1.0000000000000p+0, -0x1.5555555555555p-2, 0x1.999999999999ap-3,
-     -0x1.2492492492492p-3, 0x1.c71c71c71c71cp-4, -0x1.745d1745d1746p-4,
-     0x1.3b13b13b13b14p-4, -0x1.1111111111111p-4},
-    {0x1.ff55bb72cfdb1p-5, 0x1.fe01fe01fe1bp-1, -0x1.fc05f80821d1ap-5,
-     -0x1.4d6930419fc5fp-2, 0x1.f61b9f6d69313p-5, 0x1.8208a32f4346cp-3,
-     -0x1.ecb8fc53d04efp-5, -0x1.060710cb59cbcp-3},
-    {0x1.fd5ba9aac2f3cp-4, 0x1.f81f81f81f96ap-1, -0x1.f05e09cf4c1b2p-4,
-     -0x1.368c3aac7543ep-2, 0x1.d9b14bddfac55p-4, 0x1.4048e55ec725ep-3,
-     -0x1.b98ca3c1594b5p-4, -0x1.664eabaabbc16p-4},
-    {0x1.7b97b4bce5ae7p-3, 0x1.ee9c7f8458f06p-1, -0x1.665c226c8dc69p-3,
-     -0x1.1344bb77961b7p-2, 0x1.42ac97745d3ccp-3, 0x1.c32e142047ec1p-4,
-     -0x1.137ae41ab96cbp-3, -0x1.1a6ae8c09a4b6p-5},
-    {0x1.f5b75f92c80c6p-3, 0x1.e1e1e1e1e1ed4p-1, -0x1.c5894d101ad4p-3,
-     -0x1.ce6de02b38c38p-3, 0x1.78a3920c336b9p-3, 0x1.dd5ff94a9d499p-5,
-     -0x1.1ac2d3f9d072ep-3, 0x1.0af9735dff373p-6},
-    {0x1.362773707ebc5p-2, 0x1.d272ca3fc5b8bp-1, -0x1.0997e8ae90cb6p-2,
-     -0x1.6cf6667146798p-3, 0x1.8dd1dff17f3d3p-3, 0x1.24860eced656fp-7,
-     -0x1.f4220e8f18ed5p-4, 0x1.b700aed7cdc34p-5},
-    {0x1.6f61941e4deeep-2, 0x1.c0e070381c115p-1, -0x1.2726dd1347c7ep-2,
-     -0x1.09f37b3ad010dp-3, 0x1.85eaca5196f5cp-3, -0x1.04d640117852ap-5,
-     -0x1.802c2956871c7p-4, 0x1.2992b45df0ee7p-4},
-    {0x1.a64eec3cc23fep-2, 0x1.adbe87f94906bp-1, -0x1.3b9d8eab5eae5p-2,
-     -0x1.57c09646faabbp-4, 0x1.6795330e73aep-3, -0x1.f2d89a702a652p-5,
-     -0x1.f3afd90a9d4d7p-5, 0x1.3261723d3f153p-4},
-    {0x1.dac670561bb53p-2, 0x1.999999999998fp-1, -0x1.47ae147afd8cap-2,
-     -0x1.5d867c3dfd72ap-5, 0x1.3a92a76cba833p-3, -0x1.3ec460286928ap-4,
-     -0x1.ed02ff86892acp-6, 0x1.0a674c8f05727p-4},
-    {0x1.0657e94db30d2p-1, 0x1.84f00c27805ffp-1, -0x1.4c62cb564f677p-2,
-     -0x1.e6495b262dfe7p-8, 0x1.063c34eca262bp-3, -0x1.58b78dc79b5aep-4,
-     -0x1.4623815233be1p-8, 0x1.93afe94328089p-5},
-    {0x1.1e00babdefeb6p-1, 0x1.702e05c0b8159p-1, -0x1.4af2b78236bd6p-2,
-     0x1.5d0b7ea46ed08p-6, 0x1.a124870236935p-4, -0x1.519e1ec133a88p-4,
-     0x1.a54632a3f48c7p-7, 0x1.099ca0945096dp-5},
-    {0x1.345f01cce37bdp-1, 0x1.5babcc647fa7ep-1, -0x1.449db09443a67p-2,
-     0x1.655caac78a0fcp-5, 0x1.3bbbdb0d09efap-4, -0x1.34a306c27e021p-4,
-     0x1.83fe749c7966p-6, 0x1.2057cc96d9edcp-6},
-    {0x1.4978fa3269ee2p-1, 0x1.47ae147ae146bp-1, -0x1.3a92a305652e1p-2,
-     0x1.ec21b5172657fp-5, 0x1.c2f8c45d2f4eep-5, -0x1.0ba99c4aeb8acp-4,
-     0x1.d716a4af4d1d6p-6, 0x1.97fba0a9696dep-8},
-    {0x1.5d58987169b19p-1, 0x1.34679ace0133cp-1, -0x1.2ddfb03920e2fp-2,
-     0x1.2491307c0fa0bp-4, 0x1.29c7eca0136fp-5, -0x1.bca792caa6f1cp-5,
-     0x1.e5d92545576bcp-6, -0x1.8ca76fcf5ccd2p-10},
-    {0x1.700a7c5784634p-1, 0x1.21fb78121fb71p-1, -0x1.1f6a8499ea541p-2,
-     0x1.41b15e5e77bcfp-4, 0x1.59bc9bf54fb02p-6, -0x1.63b54ff058e0fp-5,
-     0x1.c8da01221306fp-6, -0x1.906b2c274c39cp-8},
-    {0x1.819d0b7158a4dp-1, 0x1.107fbbe01107cp-1, -0x1.0feeb40897d4ep-2,
-     0x1.50e5afb95f5d6p-4, 0x1.2a7c2f0c7495dp-7, -0x1.12bd2bb5062cdp-5,
-     0x1.93e8ceb89afebp-6, -0x1.10da9b8c6b731p-7},
-    {0x1.921fb54442d18p-1, 0x1.fffffffffffebp-2, -0x1.fffffffffcbbcp-3,
-     0x1.555555564e2fep-4, -0x1.20b17d5dd89dcp-30, -0x1.9999c5ad71711p-6,
-     0x1.5558b76e7aaf9p-6, -0x1.236e803c6c1f6p-7},
+// For i = 0, the polynomial is generated by:
+// > P = fpminimax(atan(x)/x, 7, [|1, D...|], [0, 1/32]);
+// > dirtyinfnorm((atan(x) - x*P)/x, [0, 1/32]);
+//   0x1.feb2fcdba66447ccbe28a1a0f935b51678a718fb1p-59
+// Notice that degree-7 is good enough for atanf, but degree-8 helps reduce the
+// error bounds for atan2f's fast pass 16 times, and it does not affect the
+// performance of atanf much.
+double ATAN_COEFFS[17][9] = {
+    {0.0, 1.0, 0x1.3f8d76d26d61bp-47, -0x1.5555555574cd8p-2,
+     0x1.0dde5d06878eap-29, 0x1.99997738acc77p-3, 0x1.2c43eac9797cap-16,
+     -0x1.25fb020007dbdp-3, 0x1.c1b6c31d7b0aep-7},
+    {0x1.ff55bb72cfde9p-5, 0x1.fe01fe01fe007p-1, -0x1.fc05f809ed8dap-5,
+     -0x1.4d69303afe04ep-2, 0x1.f61bc3e8349cp-5, 0x1.820839278756bp-3,
+     -0x1.eda4de1c6bf3fp-5, -0x1.0514d42d64a63p-3, 0x1.db3746a442dcbp-5},
+    {0x1.fd5ba9aac2f6ep-4, 0x1.f81f81f81f813p-1, -0x1.f05e09d0dc378p-4,
+     -0x1.368c3aa719215p-2, 0x1.d9b16b33ff9c9p-4, 0x1.40488f9c6262ap-3,
+     -0x1.ba55933e62ea5p-4, -0x1.64c6a15cd9116p-4, 0x1.9273d5939a75ap-4},
+    {0x1.7b97b4bce5b02p-3, 0x1.ee9c7f8458e05p-1, -0x1.665c226d6961p-3,
+     -0x1.1344bb7391703p-2, 0x1.42aca8b0081b9p-3, 0x1.c32d9381d7c03p-4,
+     -0x1.13e970672e246p-3, -0x1.181ed934dd733p-5, 0x1.bad81ea190c08p-4},
+    {0x1.f5b75f92c80ddp-3, 0x1.e1e1e1e1e1e2cp-1, -0x1.c5894d10d363dp-3,
+     -0x1.ce6de025f9f5ep-3, 0x1.78a3a07c8dd7fp-3, 0x1.dd5f5180f386ep-5,
+     -0x1.1b1f513c4536bp-3, 0x1.0df852e58c43cp-6, 0x1.722e7a7e42505p-4},
+    {0x1.362773707ebccp-2, 0x1.d272ca3fc5b2ep-1, -0x1.0997e8aeca8fbp-2,
+     -0x1.6cf6666e5e693p-3, 0x1.8dd1e907e88adp-3, 0x1.24849ac0caa5dp-7,
+     -0x1.f496be486229dp-4, 0x1.b7d54b8e759ecp-5, 0x1.d39c0d39c3922p-5},
+    {0x1.6f61941e4def1p-2, 0x1.c0e070381c0f2p-1, -0x1.2726dd135d9eep-2,
+     -0x1.09f37b39b70e4p-3, 0x1.85eacdaadd712p-3, -0x1.04d66340d5b9p-5,
+     -0x1.8056b15a22b98p-4, 0x1.29baf494ad3ddp-4, 0x1.52d5881322a7ap-6},
+    {0x1.a64eec3cc23fdp-2, 0x1.adbe87f94906ap-1, -0x1.3b9d8eab55addp-2,
+     -0x1.57c09646eb7p-4, 0x1.6795319e3b8dfp-3, -0x1.f2d89b5ef31bep-5,
+     -0x1.f38aac26203cap-5, 0x1.3262802235e3fp-4, -0x1.2afd6b9a57d66p-7},
+    {0x1.dac670561bb4fp-2, 0x1.99999999999ap-1, -0x1.47ae147adff11p-2,
+     -0x1.5d867c40188b7p-5, 0x1.3a92a2df85e7ap-3, -0x1.3ec457c46e851p-4,
+     -0x1.ec1b9777e2e5bp-6, 0x1.0a542992a821ep-4, -0x1.ccffbe2f0d945p-6},
+    {0x1.0657e94db30dp-1, 0x1.84f00c2780615p-1, -0x1.4c62cb562defap-2,
+     -0x1.e6495b3c14e03p-8, 0x1.063c2fa617bfcp-3, -0x1.58b782d9907aap-4,
+     -0x1.41e6ff524b7fp-8, 0x1.937dfff3205a7p-5, -0x1.0fb1fd1c729dp-5},
+    {0x1.1e00babdefeb4p-1, 0x1.702e05c0b816ep-1, -0x1.4af2b78215fbep-2,
+     0x1.5d0b7e9f36997p-6, 0x1.a1247cb978debp-4, -0x1.519e1457734cap-4,
+     0x1.a755cf86b5bfbp-7, 0x1.096d174284564p-5, -0x1.081adf539ad58p-5},
+    {0x1.345f01cce37bbp-1, 0x1.5babcc647fa8ep-1, -0x1.449db09426a6dp-2,
+     0x1.655caac5896dap-5, 0x1.3bbbd22d05a61p-4, -0x1.34a2febee042fp-4,
+     0x1.84df9c8269e34p-6, 0x1.200e8176c899ap-6, -0x1.c00b23c3ce222p-6},
+    {0x1.4978fa3269ee1p-1, 0x1.47ae147ae1477p-1, -0x1.3a92a3055231ap-2,
+     0x1.ec21b515a4a2p-5, 0x1.c2f8b81f9a0d2p-5, -0x1.0ba9964125453p-4,
+     0x1.d7b5614777a05p-6, 0x1.971e91ed73595p-8, -0x1.3fc375a78dc74p-6},
+    {0x1.5d58987169b18p-1, 0x1.34679ace01343p-1, -0x1.2ddfb039136e5p-2,
+     0x1.2491307b9fb73p-4, 0x1.29c7e4886dc22p-5, -0x1.bca78bcca83ap-5,
+     0x1.e63efd7cbe1ddp-6, -0x1.8ea6c4f03b42dp-10, -0x1.9385b5c3a6997p-7},
+    {0x1.700a7c5784634p-1, 0x1.21fb78121fb76p-1, -0x1.1f6a8499e5d1ap-2,
+     0x1.41b15e5e29423p-4, 0x1.59bc953163345p-6, -0x1.63b54b13184ddp-5,
+     0x1.c9086666d213p-6, -0x1.90c3b4ad8d4bcp-8, -0x1.80f08ed9f6f57p-8},
+    {0x1.819d0b7158a4dp-1, 0x1.107fbbe01107ep-1, -0x1.0feeb4089670ep-2,
+     0x1.50e5afb93f5cbp-4, 0x1.2a7c2adffeffbp-7, -0x1.12bd29b4f1b43p-5,
+     0x1.93f71f0eb00eap-6, -0x1.10ece5ad30e28p-7, -0x1.db1a76bcd2b9cp-10},
+    {0x1.921fb54442d18p-1, 0x1.ffffffffffffep-2, -0x1.fffffffffc51cp-3,
+     0x1.555555557002ep-4, -0x1.a88260c338e75p-30, -0x1.99999f9a7614fp-6,
+     0x1.555e31a1e15e9p-6, -0x1.245240d65e629p-7, -0x1.fa9ba66478903p-11},
 };
 
 } // namespace LIBC_NAMESPACE
diff --git a/libc/src/math/generic/inv_trigf_utils.h b/libc/src/math/generic/inv_trigf_utils.h
@@ -19,20 +19,21 @@ namespace LIBC_NAMESPACE {
 constexpr double M_MATH_PI = 0x1.921fb54442d18p+1;
 constexpr double M_MATH_PI_2 = 0x1.921fb54442d18p+0;
 
-extern double ATAN_COEFFS[17][8];
+extern double ATAN_COEFFS[17][9];
 
-// For |x| <= 1/32 and 1 <= i <= 16, return Q(x) such that:
+// For |x| <= 1/32 and 0 <= i <= 16, return Q(x) such that:
 //   Q(x) ~ (atan(x + i/16) - atan(i/16)) / x.
 LIBC_INLINE double atan_eval(double x, int i) {
   double x2 = x * x;
 
   double c0 = fputil::multiply_add(x, ATAN_COEFFS[i][2], ATAN_COEFFS[i][1]);
   double c1 = fputil::multiply_add(x, ATAN_COEFFS[i][4], ATAN_COEFFS[i][3]);
   double c2 = fputil::multiply_add(x, ATAN_COEFFS[i][6], ATAN_COEFFS[i][5]);
+  double c3 = fputil::multiply_add(x, ATAN_COEFFS[i][8], ATAN_COEFFS[i][7]);
 
   double x4 = x2 * x2;
   double d1 = fputil::multiply_add(x2, c1, c0);
-  double d2 = fputil::multiply_add(x2, ATAN_COEFFS[i][7], c2);
+  double d2 = fputil::multiply_add(x2, c3, c2);
   double p = fputil::multiply_add(x4, d2, d1);
   return p;
 }

diff --git a/libc/src/stdio/CMakeLists.txt b/libc/src/stdio/CMakeLists.txt
@@ -270,6 +270,8 @@ add_stdio_entrypoint_object(ferror)
 add_stdio_entrypoint_object(ferror_unlocked)
 add_stdio_entrypoint_object(fseek)
 add_stdio_entrypoint_object(ftell)
+add_stdio_entrypoint_object(fseeko)
+add_stdio_entrypoint_object(ftello)
 add_stdio_entrypoint_object(fflush)
 add_stdio_entrypoint_object(clearerr)
 add_stdio_entrypoint_object(clearerr_unlocked)

diff --git a/libc/src/stdio/fseeko.h b/libc/src/stdio/fseeko.h
@@ -0,0 +1,21 @@
+//===-- Implementation header of fseeko -------------------------*- 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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_STDIO_FSEEKO_H
+#define LLVM_LIBC_SRC_STDIO_FSEEKO_H
+
+#include <stdio.h>
+#include <unistd.h>
+
+namespace LIBC_NAMESPACE {
+
+int fseeko(::FILE *stream, off_t offset, int whence);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_STDIO_FSEEKO_H
diff --git a/libc/src/stdio/ftello.h b/libc/src/stdio/ftello.h
@@ -0,0 +1,21 @@
+//===-- Implementation header of ftello -------------------------*- 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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_STDIO_FTELLO_H
+#define LLVM_LIBC_SRC_STDIO_FTELLO_H
+
+#include <stdio.h>
+#include <unistd.h>
+
+namespace LIBC_NAMESPACE {
+
+off_t ftello(::FILE *f);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_STDIO_FTELLO_H
diff --git a/libc/src/stdio/generic/CMakeLists.txt b/libc/src/stdio/generic/CMakeLists.txt
@@ -103,7 +103,6 @@ add_entrypoint_object(
     ../fseek.h
   DEPENDS
     libc.src.errno.errno
-    libc.include.stdio
     libc.src.__support.File.file
     libc.src.__support.File.platform_file
 )
@@ -116,7 +115,30 @@ add_entrypoint_object(
     ../ftell.h
   DEPENDS
     libc.src.errno.errno
-    libc.include.stdio
+    libc.src.__support.File.file
+    libc.src.__support.File.platform_file
+)
+
+add_entrypoint_object(
+  fseeko
+  SRCS
+    fseeko.cpp
+  HDRS
+    ../fseeko.h
+  DEPENDS
+    libc.src.errno.errno
+    libc.src.__support.File.file
+    libc.src.__support.File.platform_file
+)
+
+add_entrypoint_object(
+  ftello
+  SRCS
+    ftello.cpp
+  HDRS
+    ../ftello.h
+  DEPENDS
+    libc.src.errno.errno
     libc.src.__support.File.file
     libc.src.__support.File.platform_file
 )

diff --git a/libc/src/stdio/generic/fseek.cpp b/libc/src/stdio/generic/fseek.cpp
@@ -10,7 +10,6 @@
 #include "src/__support/File/file.h"
 
 #include "src/errno/libc_errno.h"
-#include <stdio.h>
 
 namespace LIBC_NAMESPACE {
 

diff --git a/libc/src/stdio/generic/fseeko.cpp b/libc/src/stdio/generic/fseeko.cpp
@@ -0,0 +1,26 @@
+//===-- Implementation of fseeko ------------------------------------------===//
+//
+// 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 "src/stdio/fseeko.h"
+#include "src/__support/File/file.h"
+
+#include "src/errno/libc_errno.h"
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, fseeko, (::FILE * stream, off_t offset, int whence)) {
+  auto result =
+      reinterpret_cast<LIBC_NAMESPACE::File *>(stream)->seek(offset, whence);
+  if (!result.has_value()) {
+    libc_errno = result.error();
+    return -1;
+  }
+  return 0;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/stdio/generic/ftell.cpp b/libc/src/stdio/generic/ftell.cpp
@@ -10,7 +10,6 @@
 #include "src/__support/File/file.h"
 
 #include "src/errno/libc_errno.h"
-#include <stdio.h>
 
 namespace LIBC_NAMESPACE {
 

diff --git a/libc/src/stdio/generic/ftello.cpp b/libc/src/stdio/generic/ftello.cpp
@@ -0,0 +1,25 @@
+//===-- Implementation of ftello ------------------------------------------===//
+//
+// 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 "src/stdio/ftello.h"
+#include "src/__support/File/file.h"
+
+#include "src/errno/libc_errno.h"
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(off_t, ftello, (::FILE * stream)) {
+  auto result = reinterpret_cast<LIBC_NAMESPACE::File *>(stream)->tell();
+  if (!result.has_value()) {
+    libc_errno = result.error();
+    return -1;
+  }
+  return result.value();
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/stdio/printf_core/CMakeLists.txt b/libc/src/stdio/printf_core/CMakeLists.txt
@@ -42,6 +42,7 @@ add_header_library(
     libc.src.__support.arg_list
     libc.src.__support.ctype_utils
     libc.src.__support.str_to_integer
+    libc.src.__support.CPP.algorithm
     libc.src.__support.CPP.bit
     libc.src.__support.CPP.optional
     libc.src.__support.CPP.string_view

diff --git a/libc/src/stdio/printf_core/parser.h b/libc/src/stdio/printf_core/parser.h
@@ -10,6 +10,7 @@
 #define LLVM_LIBC_SRC_STDIO_PRINTF_CORE_PARSER_H
 
 #include "include/llvm-libc-macros/stdfix-macros.h"
+#include "src/__support/CPP/algorithm.h" // max
 #include "src/__support/CPP/optional.h"
 #include "src/__support/CPP/type_traits.h"
 #include "src/__support/str_to_integer.h"

diff --git a/libc/test/src/math/CMakeLists.txt b/libc/test/src/math/CMakeLists.txt
@@ -1719,6 +1719,19 @@ add_fp_unittest(
     libc.src.__support.FPUtil.fp_bits
 )
 
+add_fp_unittest(
+  atan2f_test
+  NEED_MPFR
+  SUITE
+    libc-math-unittests
+  SRCS
+    atan2f_test.cpp
+  DEPENDS
+    libc.include.math
+    libc.src.math.atan2f
+    libc.src.__support.FPUtil.fp_bits
+)
+
 add_subdirectory(generic)
 add_subdirectory(smoke)
 

diff --git a/libc/test/src/math/atan2f_test.cpp b/libc/test/src/math/atan2f_test.cpp
@@ -0,0 +1,133 @@
+//===-- Unittests for atan2f ----------------------------------------------===//
+//
+// 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 "include/llvm-libc-macros/math-macros.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/math/atan2f.h"
+#include "test/UnitTest/FPMatcher.h"
+#include "test/UnitTest/Test.h"
+#include "utils/MPFRWrapper/MPFRUtils.h"
+
+using LlvmLibcAtan2fTest = LIBC_NAMESPACE::testing::FPTest<float>;
+using LIBC_NAMESPACE::testing::tlog;
+
+namespace mpfr = LIBC_NAMESPACE::testing::mpfr;
+
+TEST_F(LlvmLibcAtan2fTest, TrickyInputs) {
+  constexpr int N = 17;
+  mpfr::BinaryInput<float> INPUTS[N] = {
+      {0x1.0cb3a4p+20f, 0x1.4ebacp+22f},   {0x1.12215p+1f, 0x1.4fabfcp+22f},
+      {-0x1.13baaep+41f, 0x1.5bd22ep+23f}, {0x1.1ff7dcp+41f, 0x1.aec0a6p+23f},
+      {0x1.2bc794p+23f, 0x1.0bc0c6p+23f},  {0x1.2fba3ap+42f, 0x1.f99456p+23f},
+      {0x1.5ea1f8p+27f, 0x1.f2a1aep+23f},  {0x1.7a931p+44f, 0x1.352ac4p+22f},
+      {0x1.8802bcp+21f, 0x1.8f130ap+23f},  {0x1.658ef8p+17f, 0x1.3c00f4p+22f},
+      {0x1.69fb0cp+21f, 0x1.39e4c4p+23f},  {0x1.8eb24cp+11f, 0x1.36518p+23f},
+      {0x1.9e7ebp+30f, 0x1.d80522p+23f},   {0x1.b4bdeep+19f, 0x1.c19b4p+23f},
+      {0x1.bc201p+43f, 0x1.617346p+23f},   {0x1.c96c3cp+20f, 0x1.c01d1ep+23f},
+      {0x1.781fcp+28f, 0x1.dcb3cap+23f},
+  };
+
+  for (int i = 0; i < N; ++i) {
+    float x = INPUTS[i].x;
+    float y = INPUTS[i].y;
+    ASSERT_MPFR_MATCH_ALL_ROUNDING(mpfr::Operation::Atan2, INPUTS[i],
+                                   LIBC_NAMESPACE::atan2f(x, y), 0.5);
+    INPUTS[i].x = -INPUTS[i].x;
+    ASSERT_MPFR_MATCH_ALL_ROUNDING(mpfr::Operation::Atan2, INPUTS[i],
+                                   LIBC_NAMESPACE::atan2f(-x, y), 0.5);
+    INPUTS[i].y = -INPUTS[i].y;
+    ASSERT_MPFR_MATCH_ALL_ROUNDING(mpfr::Operation::Atan2, INPUTS[i],
+                                   LIBC_NAMESPACE::atan2f(-x, -y), 0.5);
+    INPUTS[i].x = -INPUTS[i].x;
+    ASSERT_MPFR_MATCH_ALL_ROUNDING(mpfr::Operation::Atan2, INPUTS[i],
+                                   LIBC_NAMESPACE::atan2f(x, -y), 0.5);
+  }
+}
+
+TEST_F(LlvmLibcAtan2fTest, InFloatRange) {
+  constexpr uint32_t X_COUNT = 1'23;
+  constexpr uint32_t X_START = FPBits(0.25f).uintval();
+  constexpr uint32_t X_STOP = FPBits(4.0f).uintval();
+  constexpr uint32_t X_STEP = (X_STOP - X_START) / X_COUNT;
+
+  constexpr uint32_t Y_COUNT = 1'37;
+  constexpr uint32_t Y_START = FPBits(0.25f).uintval();
+  constexpr uint32_t Y_STOP = FPBits(4.0f).uintval();
+  constexpr uint32_t Y_STEP = (Y_STOP - Y_START) / Y_COUNT;
+
+  auto test = [&](mpfr::RoundingMode rounding_mode) {
+    mpfr::ForceRoundingMode __r(rounding_mode);
+    if (!__r.success)
+      return;
+
+    uint64_t fails = 0;
+    uint64_t finite_count = 0;
+    uint64_t total_count = 0;
+    float failed_x, failed_y, failed_r = 0.0;
+    double tol = 0.5;
+
+    for (uint32_t i = 0, v = X_START; i <= X_COUNT; ++i, v += X_STEP) {
+      float x = FPBits(v).get_val();
+      if (isnan(x) || isinf(x) || x < 0.0)
+        continue;
+
+      for (uint32_t j = 0, w = Y_START; j <= Y_COUNT; ++j, w += Y_STEP) {
+        float y = FPBits(w).get_val();
+        if (isnan(y) || isinf(y))
+          continue;
+
+        LIBC_NAMESPACE::libc_errno = 0;
+        float result = LIBC_NAMESPACE::atan2f(x, y);
+        ++total_count;
+        if (isnan(result) || isinf(result))
+          continue;
+
+        ++finite_count;
+        mpfr::BinaryInput<float> inputs{x, y};
+
+        if (!TEST_MPFR_MATCH_ROUNDING_SILENTLY(mpfr::Operation::Atan2, inputs,
+                                               result, 0.5, rounding_mode)) {
+          ++fails;
+          while (!TEST_MPFR_MATCH_ROUNDING_SILENTLY(
+              mpfr::Operation::Atan2, inputs, result, tol, rounding_mode)) {
+            failed_x = x;
+            failed_y = y;
+            failed_r = result;
+
+            if (tol > 1000.0)
+              break;
+
+            tol *= 2.0;
+          }
+        }
+      }
+    }
+    if (fails || (finite_count < total_count)) {
+      tlog << " Atan2f failed: " << fails << "/" << finite_count << "/"
+           << total_count << " tests.\n"
+           << "   Max ULPs is at most: " << static_cast<uint64_t>(tol) << ".\n";
+    }
+    if (fails) {
+      mpfr::BinaryInput<float> inputs{failed_x, failed_y};
+      EXPECT_MPFR_MATCH(mpfr::Operation::Atan2, inputs, failed_r, 0.5,
+                        rounding_mode);
+    }
+  };
+
+  tlog << " Test Rounding To Nearest...\n";
+  test(mpfr::RoundingMode::Nearest);
+
+  tlog << " Test Rounding Downward...\n";
+  test(mpfr::RoundingMode::Downward);
+
+  tlog << " Test Rounding Upward...\n";
+  test(mpfr::RoundingMode::Upward);
+
+  tlog << " Test Rounding Toward Zero...\n";
+  test(mpfr::RoundingMode::TowardZero);
+}
diff --git a/libc/test/src/math/atanf_test.cpp b/libc/test/src/math/atanf_test.cpp
@@ -55,12 +55,15 @@ TEST_F(LlvmLibcAtanfTest, InFloatRange) {
 TEST_F(LlvmLibcAtanfTest, SpecialValues) {
   uint32_t val_arr[] = {
       0x3d8d6b23U, // x = 0x1.1ad646p-4f
+      0x3dbb6ac7U, // x = 0x1.76d58ep-4f
       0x3feefcfbU, // x = 0x1.ddf9f6p+0f
+      0x3ffe2ec1U, // x = 0x1.fc5d82p+0f
       0xbd8d6b23U, // x = -0x1.1ad646p-4f
+      0xbdbb6ac7U, // x = -0x1.76d58ep-4f
       0xbfeefcfbU, // x = -0x1.ddf9f6p+0f
+      0xbffe2ec1U, // x = -0x1.fc5d82p+0
       0x7F800000U, // x = +Inf
       0xFF800000U, // x = -Inf
-      0xbffe2ec1U, // x = -0x1.fc5d82p+0f
   };
   for (uint32_t v : val_arr) {
     float x = FPBits(v).get_val();

diff --git a/libc/test/src/math/smoke/CMakeLists.txt b/libc/test/src/math/smoke/CMakeLists.txt
@@ -2755,6 +2755,18 @@ add_fp_unittest(
     libc.src.__support.FPUtil.fp_bits
 )
 
+add_fp_unittest(
+  atan2f_test
+  SUITE
+    libc-math-smoke-tests
+  SRCS
+    atan2f_test.cpp
+  DEPENDS
+    libc.src.errno.errno
+    libc.src.math.atan2f
+    libc.src.__support.FPUtil.fp_bits
+)
+
 add_fp_unittest(
   scalbn_test
   SUITE

diff --git a/libc/test/src/math/smoke/atan2f_test.cpp b/libc/test/src/math/smoke/atan2f_test.cpp
@@ -0,0 +1,50 @@
+//===-- Unittests for atan2f ----------------------------------------------===//
+//
+// 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 "include/llvm-libc-macros/math-macros.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/errno/libc_errno.h"
+#include "src/math/atan2f.h"
+#include "test/UnitTest/FPMatcher.h"
+#include "test/UnitTest/Test.h"
+
+using LlvmLibcAtan2fTest = LIBC_NAMESPACE::testing::FPTest<float>;
+
+TEST_F(LlvmLibcAtan2fTest, SpecialNumbers) {
+  LIBC_NAMESPACE::libc_errno = 0;
+
+  LIBC_NAMESPACE::fputil::clear_except(FE_ALL_EXCEPT);
+  EXPECT_FP_EQ_ALL_ROUNDING(aNaN, LIBC_NAMESPACE::atan2f(aNaN, zero));
+  EXPECT_FP_EXCEPTION(0);
+  EXPECT_MATH_ERRNO(0);
+
+  LIBC_NAMESPACE::fputil::clear_except(FE_ALL_EXCEPT);
+  EXPECT_FP_EQ_ALL_ROUNDING(aNaN, LIBC_NAMESPACE::atan2f(1.0f, aNaN));
+  EXPECT_FP_EXCEPTION(0);
+  EXPECT_MATH_ERRNO(0);
+
+  LIBC_NAMESPACE::fputil::clear_except(FE_ALL_EXCEPT);
+  EXPECT_FP_EQ_ALL_ROUNDING(0.0f, LIBC_NAMESPACE::atan2f(zero, zero));
+  EXPECT_FP_EXCEPTION(0);
+  EXPECT_MATH_ERRNO(0);
+
+  LIBC_NAMESPACE::fputil::clear_except(FE_ALL_EXCEPT);
+  EXPECT_FP_EQ_ALL_ROUNDING(-0.0f, LIBC_NAMESPACE::atan2f(-0.0f, zero));
+  EXPECT_FP_EXCEPTION(0);
+  EXPECT_MATH_ERRNO(0);
+
+  LIBC_NAMESPACE::fputil::clear_except(FE_ALL_EXCEPT);
+  EXPECT_FP_EQ_ALL_ROUNDING(0.0f, LIBC_NAMESPACE::atan2f(1.0f, inf));
+  EXPECT_FP_EXCEPTION(0);
+  EXPECT_MATH_ERRNO(0);
+
+  LIBC_NAMESPACE::fputil::clear_except(FE_ALL_EXCEPT);
+  EXPECT_FP_EQ_ALL_ROUNDING(-0.0f, LIBC_NAMESPACE::atan2f(-1.0f, inf));
+  EXPECT_FP_EXCEPTION(0);
+  EXPECT_MATH_ERRNO(0);
+}
diff --git a/libc/test/src/stdio/CMakeLists.txt b/libc/test/src/stdio/CMakeLists.txt
@@ -442,6 +442,8 @@ add_libc_test(
     libc.src.stdio.fread
     libc.src.stdio.fseek
     libc.src.stdio.ftell
+    libc.src.stdio.fseeko
+    libc.src.stdio.ftello
     libc.src.stdio.fwrite
     libc.src.stdio.setvbuf
 )

diff --git a/libc/test/src/stdio/ftell_test.cpp b/libc/test/src/stdio/ftell_test.cpp
@@ -10,7 +10,9 @@
 #include "src/stdio/fopen.h"
 #include "src/stdio/fread.h"
 #include "src/stdio/fseek.h"
+#include "src/stdio/fseeko.h"
 #include "src/stdio/ftell.h"
+#include "src/stdio/ftello.h"
 #include "src/stdio/fwrite.h"
 #include "src/stdio/setvbuf.h"
 #include "test/UnitTest/Test.h"
@@ -37,6 +39,13 @@ class LlvmLibcFTellTest : public LIBC_NAMESPACE::testing::Test {
     // still return the correct effective offset.
     ASSERT_EQ(size_t(LIBC_NAMESPACE::ftell(file)), WRITE_SIZE);
 
+    off_t offseto = 42;
+    ASSERT_EQ(0, LIBC_NAMESPACE::fseeko(file, offseto, SEEK_SET));
+    ASSERT_EQ(LIBC_NAMESPACE::ftello(file), offseto);
+    ASSERT_EQ(0, LIBC_NAMESPACE::fseeko(file, -offseto, SEEK_END));
+    ASSERT_EQ(size_t(LIBC_NAMESPACE::ftello(file)),
+              size_t(WRITE_SIZE - offseto));
+
     long offset = 5;
     ASSERT_EQ(0, LIBC_NAMESPACE::fseek(file, offset, SEEK_SET));
     ASSERT_EQ(LIBC_NAMESPACE::ftell(file), offset);

diff --git a/libc/utils/MPFRWrapper/MPFRUtils.cpp b/libc/utils/MPFRWrapper/MPFRUtils.cpp
@@ -181,6 +181,12 @@ class MPFRNumber {
     return result;
   }
 
+  MPFRNumber atan2(const MPFRNumber &b) {
+    MPFRNumber result(*this);
+    mpfr_atan2(result.value, value, b.value, mpfr_rounding);
+    return result;
+  }
+
   MPFRNumber atanh() const {
     MPFRNumber result(*this);
     mpfr_atanh(result.value, value, mpfr_rounding);
@@ -623,6 +629,8 @@ binary_operation_one_output(Operation op, InputType x, InputType y,
   MPFRNumber inputX(x, precision, rounding);
   MPFRNumber inputY(y, precision, rounding);
   switch (op) {
+  case Operation::Atan2:
+    return inputX.atan2(inputY);
   case Operation::Fmod:
     return inputX.fmod(inputY);
   case Operation::Hypot:

diff --git a/libc/utils/MPFRWrapper/MPFRUtils.h b/libc/utils/MPFRWrapper/MPFRUtils.h
@@ -68,6 +68,7 @@ enum class Operation : int {
   // input and produce a single floating point number of the same type as
   // output.
   BeginBinaryOperationsSingleOutput,
+  Atan2,
   Fmod,
   Hypot,
   Pow,

diff --git a/libcxx/benchmarks/algorithms/mismatch.bench.cpp b/libcxx/benchmarks/algorithms/mismatch.bench.cpp
@@ -37,4 +37,20 @@ BENCHMARK(bm_mismatch<char>)->Apply(BenchmarkSizes);
 BENCHMARK(bm_mismatch<short>)->Apply(BenchmarkSizes);
 BENCHMARK(bm_mismatch<int>)->Apply(BenchmarkSizes);
 
+template <class T>
+static void bm_mismatch_two_range_overload(benchmark::State& state) {
+  std::vector<T> vec1(state.range(), '1');
+  std::vector<T> vec2(state.range(), '1');
+  std::mt19937_64 rng(std::random_device{}());
+
+  vec1.back() = '2';
+  for (auto _ : state) {
+    benchmark::DoNotOptimize(vec1);
+    benchmark::DoNotOptimize(std::mismatch(vec1.begin(), vec1.end(), vec2.begin(), vec2.end()));
+  }
+}
+BENCHMARK(bm_mismatch_two_range_overload<char>)->DenseRange(1, 8)->Range(16, 1 << 20);
+BENCHMARK(bm_mismatch_two_range_overload<short>)->DenseRange(1, 8)->Range(16, 1 << 20);
+BENCHMARK(bm_mismatch_two_range_overload<int>)->DenseRange(1, 8)->Range(16, 1 << 20);
+
 BENCHMARK_MAIN();
diff --git a/libcxx/docs/index.rst b/libcxx/docs/index.rst
@@ -66,6 +66,25 @@ Getting Started with libc++
 Current Status
 ==============
 
+libc++ has become the default C++ Standard Library implementation for many major platforms, including Apple's macOS,
+iOS, watchOS, and tvOS, Google Search, the Android operating system, and FreeBSD. As a result, libc++ has an estimated
+user base of over 1 billion daily active users.
+
+Since its inception, libc++ has focused on delivering high performance, standards-conformance, and portability. It has
+been extensively tested and optimized, making it robust and production ready. libc++ fully implements C++11 and C++14, 
+with C++17, C++20, C++23, and C++26 features being actively developed and making steady progress.
+
+libc++ is continuously integrated and tested on a wide range of platforms and configurations, ensuring its reliability
+and compatibility across various systems. The library's extensive test suite and rigorous quality assurance process have
+made it a top choice for platform providers looking to offer their users a robust and efficient C++ Standard Library.
+
+As an open-source project, libc++ benefits from a vibrant community of contributors who work together to improve the
+library and add new features. This ongoing development and support ensure that libc++ remains at the forefront of
+C++ standardization efforts and continues to meet the evolving needs of C++ developers worldwide.
+
+
+History
+-------
 After its initial introduction, many people have asked "why start a new
 library instead of contributing to an existing library?" (like Apache's
 libstdcxx, GNU's libstdc++, STLport, etc).  There are many contributing
-Original file line number
+Diff line change
@@ Expand Up / @@ -68,6 +68,7 @@ enum class Operation : int { @@
       // input and produce a single floating point number of the same type as
       // output.
       BeginBinaryOperationsSingleOutput,
+      Atan2,
       Fmod,
       Hypot,
       Pow,
@@ Expand Down @@