lib/LTO/LTO.cpp

//===-LTO.cpp - LLVM Link Time Optimizer ----------------------------------===//
//
// 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
//
//===----------------------------------------------------------------------===//
//
// This file implements functions and classes used to support LTO.
//
//===----------------------------------------------------------------------===//

#include "llvm/LTO/LTO.h"
#include "llvm/ADT/ScopeExit.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
#include "llvm/Analysis/StackSafetyAnalysis.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Bitcode/BitcodeReader.h"
#include "llvm/Bitcode/BitcodeWriter.h"
#include "llvm/CodeGen/Analysis.h"
#include "llvm/Config/llvm-config.h"
#include "llvm/IR/AutoUpgrade.h"
#include "llvm/IR/DiagnosticPrinter.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/LLVMRemarkStreamer.h"
#include "llvm/IR/LegacyPassManager.h"
#include "llvm/IR/Mangler.h"
#include "llvm/IR/Metadata.h"
#include "llvm/LTO/LTOBackend.h"
#include "llvm/LTO/SummaryBasedOptimizations.h"
#include "llvm/Linker/IRMover.h"
#include "llvm/MC/TargetRegistry.h"
#include "llvm/Object/IRObjectFile.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Error.h"
#include "llvm/Support/FileSystem.h"
#include "llvm/Support/ManagedStatic.h"
#include "llvm/Support/MemoryBuffer.h"
#include "llvm/Support/Path.h"
#include "llvm/Support/SHA1.h"
#include "llvm/Support/SourceMgr.h"
#include "llvm/Support/ThreadPool.h"
#include "llvm/Support/Threading.h"
#include "llvm/Support/TimeProfiler.h"
#include "llvm/Support/ToolOutputFile.h"
#include "llvm/Support/VCSRevision.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/Transforms/IPO.h"
#include "llvm/Transforms/IPO/MemProfContextDisambiguation.h"
#include "llvm/Transforms/IPO/WholeProgramDevirt.h"
#include "llvm/Transforms/Utils/FunctionImportUtils.h"
#include "llvm/Transforms/Utils/SplitModule.h"

#include <optional>
#include <set>

using namespace llvm;
using namespace lto;
using namespace object;

#define DEBUG_TYPE "lto"

static cl::opt<bool>
    DumpThinCGSCCs("dump-thin-cg-sccs", cl::init(false), cl::Hidden,
                   cl::desc("Dump the SCCs in the ThinLTO index's callgraph"));

namespace llvm {
/// Enable global value internalization in LTO.
cl::opt<bool> EnableLTOInternalization(
    "enable-lto-internalization", cl::init(true), cl::Hidden,
    cl::desc("Enable global value internalization in LTO"));
}

/// Indicate we are linking with an allocator that supports hot/cold operator
/// new interfaces.
extern cl::opt<bool> SupportsHotColdNew;

/// Enable MemProf context disambiguation for thin link.
extern cl::opt<bool> EnableMemProfContextDisambiguation;

// Computes a unique hash for the Module considering the current list of
// export/import and other global analysis results.
// The hash is produced in \p Key.
void llvm::computeLTOCacheKey(
    SmallString<40> &Key, const Config &Conf, const ModuleSummaryIndex &Index,
    StringRef ModuleID, const FunctionImporter::ImportMapTy &ImportList,
    const FunctionImporter::ExportSetTy &ExportList,
    const std::map<GlobalValue::GUID, GlobalValue::LinkageTypes> &ResolvedODR,
    const GVSummaryMapTy &DefinedGlobals,
    const std::set<GlobalValue::GUID> &CfiFunctionDefs,
    const std::set<GlobalValue::GUID> &CfiFunctionDecls) {
  // Compute the unique hash for this entry.
  // This is based on the current compiler version, the module itself, the
  // export list, the hash for every single module in the import list, the
  // list of ResolvedODR for the module, and the list of preserved symbols.
  SHA1 Hasher;

  // Start with the compiler revision
  Hasher.update(LLVM_VERSION_STRING);
#ifdef LLVM_REVISION
  Hasher.update(LLVM_REVISION);
#endif

  // Include the parts of the LTO configuration that affect code generation.
  auto AddString = [&](StringRef Str) {
    Hasher.update(Str);
    Hasher.update(ArrayRef<uint8_t>{0});
  };
  auto AddUnsigned = [&](unsigned I) {
    uint8_t Data[4];
    support::endian::write32le(Data, I);
    Hasher.update(ArrayRef<uint8_t>{Data, 4});
  };
  auto AddUint64 = [&](uint64_t I) {
    uint8_t Data[8];
    support::endian::write64le(Data, I);
    Hasher.update(ArrayRef<uint8_t>{Data, 8});
  };
  AddString(Conf.CPU);
  // FIXME: Hash more of Options. For now all clients initialize Options from
  // command-line flags (which is unsupported in production), but may set
  // RelaxELFRelocations. The clang driver can also pass FunctionSections,
  // DataSections and DebuggerTuning via command line flags.
  AddUnsigned(Conf.Options.RelaxELFRelocations);
  AddUnsigned(Conf.Options.FunctionSections);
  AddUnsigned(Conf.Options.DataSections);
  AddUnsigned((unsigned)Conf.Options.DebuggerTuning);
  for (auto &A : Conf.MAttrs)
    AddString(A);
  if (Conf.RelocModel)
    AddUnsigned(*Conf.RelocModel);
  else
    AddUnsigned(-1);
  if (Conf.CodeModel)
    AddUnsigned(*Conf.CodeModel);
  else
    AddUnsigned(-1);
  for (const auto &S : Conf.MllvmArgs)
    AddString(S);
  AddUnsigned(Conf.CGOptLevel);
  AddUnsigned(Conf.CGFileType);
  AddUnsigned(Conf.OptLevel);
  AddUnsigned(Conf.Freestanding);
  AddString(Conf.OptPipeline);
  AddString(Conf.AAPipeline);
  AddString(Conf.OverrideTriple);
  AddString(Conf.DefaultTriple);
  AddString(Conf.DwoDir);

  // Include the hash for the current module
  auto ModHash = Index.getModuleHash(ModuleID);
  Hasher.update(ArrayRef<uint8_t>((uint8_t *)&ModHash[0], sizeof(ModHash)));

  std::vector<uint64_t> ExportsGUID;
  ExportsGUID.reserve(ExportList.size());
  for (const auto &VI : ExportList) {
    auto GUID = VI.getGUID();
    ExportsGUID.push_back(GUID);
  }

  // Sort the export list elements GUIDs.
  llvm::sort(ExportsGUID);
  for (uint64_t GUID : ExportsGUID) {
    // The export list can impact the internalization, be conservative here
    Hasher.update(ArrayRef<uint8_t>((uint8_t *)&GUID, sizeof(GUID)));
  }

  // Include the hash for every module we import functions from. The set of
  // imported symbols for each module may affect code generation and is
  // sensitive to link order, so include that as well.
  using ImportMapIteratorTy = FunctionImporter::ImportMapTy::const_iterator;
  struct ImportModule {
    ImportMapIteratorTy ModIt;
    const ModuleSummaryIndex::ModuleInfo *ModInfo;

    StringRef getIdentifier() const { return ModIt->getKey(); }
    const FunctionImporter::FunctionsToImportTy &getFunctions() const {
      return ModIt->second;
    }

    const ModuleHash &getHash() const { return ModInfo->second.second; }
  };

  std::vector<ImportModule> ImportModulesVector;
  ImportModulesVector.reserve(ImportList.size());

  for (ImportMapIteratorTy It = ImportList.begin(); It != ImportList.end();
       ++It) {
    ImportModulesVector.push_back({It, Index.getModule(It->getKey())});
  }
  // Order using module hash, to be both independent of module name and
  // module order.
  llvm::sort(ImportModulesVector,
             [](const ImportModule &Lhs, const ImportModule &Rhs) -> bool {
               return Lhs.getHash() < Rhs.getHash();
             });
  for (const ImportModule &Entry : ImportModulesVector) {
    auto ModHash = Entry.getHash();
    Hasher.update(ArrayRef<uint8_t>((uint8_t *)&ModHash[0], sizeof(ModHash)));

    AddUint64(Entry.getFunctions().size());
    for (auto &Fn : Entry.getFunctions())
      AddUint64(Fn);
  }

  // Include the hash for the resolved ODR.
  for (auto &Entry : ResolvedODR) {
    Hasher.update(ArrayRef<uint8_t>((const uint8_t *)&Entry.first,
                                    sizeof(GlobalValue::GUID)));
    Hasher.update(ArrayRef<uint8_t>((const uint8_t *)&Entry.second,
                                    sizeof(GlobalValue::LinkageTypes)));
  }

  // Members of CfiFunctionDefs and CfiFunctionDecls that are referenced or
  // defined in this module.
  std::set<GlobalValue::GUID> UsedCfiDefs;
  std::set<GlobalValue::GUID> UsedCfiDecls;

  // Typeids used in this module.
  std::set<GlobalValue::GUID> UsedTypeIds;

  auto AddUsedCfiGlobal = [&](GlobalValue::GUID ValueGUID) {
    if (CfiFunctionDefs.count(ValueGUID))
      UsedCfiDefs.insert(ValueGUID);
    if (CfiFunctionDecls.count(ValueGUID))
      UsedCfiDecls.insert(ValueGUID);
  };

  auto AddUsedThings = [&](GlobalValueSummary *GS) {
    if (!GS) return;
    AddUnsigned(GS->getVisibility());
    AddUnsigned(GS->isLive());
    AddUnsigned(GS->canAutoHide());
    for (const ValueInfo &VI : GS->refs()) {
      AddUnsigned(VI.isDSOLocal(Index.withDSOLocalPropagation()));
      AddUsedCfiGlobal(VI.getGUID());
    }
    if (auto *GVS = dyn_cast<GlobalVarSummary>(GS)) {
      AddUnsigned(GVS->maybeReadOnly());
      AddUnsigned(GVS->maybeWriteOnly());
    }
    if (auto *FS = dyn_cast<FunctionSummary>(GS)) {
      for (auto &TT : FS->type_tests())
        UsedTypeIds.insert(TT);
      for (auto &TT : FS->type_test_assume_vcalls())
        UsedTypeIds.insert(TT.GUID);
      for (auto &TT : FS->type_checked_load_vcalls())
        UsedTypeIds.insert(TT.GUID);
      for (auto &TT : FS->type_test_assume_const_vcalls())
        UsedTypeIds.insert(TT.VFunc.GUID);
      for (auto &TT : FS->type_checked_load_const_vcalls())
        UsedTypeIds.insert(TT.VFunc.GUID);
      for (auto &ET : FS->calls()) {
        AddUnsigned(ET.first.isDSOLocal(Index.withDSOLocalPropagation()));
        AddUsedCfiGlobal(ET.first.getGUID());
      }
    }
  };

  // Include the hash for the linkage type to reflect internalization and weak
  // resolution, and collect any used type identifier resolutions.
  for (auto &GS : DefinedGlobals) {
    GlobalValue::LinkageTypes Linkage = GS.second->linkage();
    Hasher.update(
        ArrayRef<uint8_t>((const uint8_t *)&Linkage, sizeof(Linkage)));
    AddUsedCfiGlobal(GS.first);
    AddUsedThings(GS.second);
  }

  // Imported functions may introduce new uses of type identifier resolutions,
  // so we need to collect their used resolutions as well.
  for (const ImportModule &ImpM : ImportModulesVector)
    for (auto &ImpF : ImpM.getFunctions()) {
      GlobalValueSummary *S =
          Index.findSummaryInModule(ImpF, ImpM.getIdentifier());
      AddUsedThings(S);
      // If this is an alias, we also care about any types/etc. that the aliasee
      // may reference.
      if (auto *AS = dyn_cast_or_null<AliasSummary>(S))
        AddUsedThings(AS->getBaseObject());
    }

  auto AddTypeIdSummary = [&](StringRef TId, const TypeIdSummary &S) {
    AddString(TId);

    AddUnsigned(S.TTRes.TheKind);
    AddUnsigned(S.TTRes.SizeM1BitWidth);

    AddUint64(S.TTRes.AlignLog2);
    AddUint64(S.TTRes.SizeM1);
    AddUint64(S.TTRes.BitMask);
    AddUint64(S.TTRes.InlineBits);

    AddUint64(S.WPDRes.size());
    for (auto &WPD : S.WPDRes) {
      AddUnsigned(WPD.first);
      AddUnsigned(WPD.second.TheKind);
      AddString(WPD.second.SingleImplName);

      AddUint64(WPD.second.ResByArg.size());
      for (auto &ByArg : WPD.second.ResByArg) {
        AddUint64(ByArg.first.size());
        for (uint64_t Arg : ByArg.first)
          AddUint64(Arg);
        AddUnsigned(ByArg.second.TheKind);
        AddUint64(ByArg.second.Info);
        AddUnsigned(ByArg.second.Byte);
        AddUnsigned(ByArg.second.Bit);
      }
    }
  };

  // Include the hash for all type identifiers used by this module.
  for (GlobalValue::GUID TId : UsedTypeIds) {
    auto TidIter = Index.typeIds().equal_range(TId);
    for (auto It = TidIter.first; It != TidIter.second; ++It)
      AddTypeIdSummary(It->second.first, It->second.second);
  }

  AddUnsigned(UsedCfiDefs.size());
  for (auto &V : UsedCfiDefs)
    AddUint64(V);

  AddUnsigned(UsedCfiDecls.size());
  for (auto &V : UsedCfiDecls)
    AddUint64(V);

  if (!Conf.SampleProfile.empty()) {
    auto FileOrErr = MemoryBuffer::getFile(Conf.SampleProfile);
    if (FileOrErr) {
      Hasher.update(FileOrErr.get()->getBuffer());

      if (!Conf.ProfileRemapping.empty()) {
        FileOrErr = MemoryBuffer::getFile(Conf.ProfileRemapping);
        if (FileOrErr)
          Hasher.update(FileOrErr.get()->getBuffer());
      }
    }
  }

  Key = toHex(Hasher.result());
}

static void thinLTOResolvePrevailingGUID(
    const Config &C, ValueInfo VI,
    DenseSet<GlobalValueSummary *> &GlobalInvolvedWithAlias,
    function_ref<bool(GlobalValue::GUID, const GlobalValueSummary *)>
        isPrevailing,
    function_ref<void(StringRef, GlobalValue::GUID, GlobalValue::LinkageTypes)>
        recordNewLinkage,
    const DenseSet<GlobalValue::GUID> &GUIDPreservedSymbols) {
  GlobalValue::VisibilityTypes Visibility =
      C.VisibilityScheme == Config::ELF ? VI.getELFVisibility()
                                        : GlobalValue::DefaultVisibility;
  for (auto &S : VI.getSummaryList()) {
    GlobalValue::LinkageTypes OriginalLinkage = S->linkage();
    // Ignore local and appending linkage values since the linker
    // doesn't resolve them.
    if (GlobalValue::isLocalLinkage(OriginalLinkage) ||
        GlobalValue::isAppendingLinkage(S->linkage()))
      continue;
    // We need to emit only one of these. The prevailing module will keep it,
    // but turned into a weak, while the others will drop it when possible.
    // This is both a compile-time optimization and a correctness
    // transformation. This is necessary for correctness when we have exported
    // a reference - we need to convert the linkonce to weak to
    // ensure a copy is kept to satisfy the exported reference.
    // FIXME: We may want to split the compile time and correctness
    // aspects into separate routines.
    if (isPrevailing(VI.getGUID(), S.get())) {
      if (GlobalValue::isLinkOnceLinkage(OriginalLinkage)) {
        S->setLinkage(GlobalValue::getWeakLinkage(
            GlobalValue::isLinkOnceODRLinkage(OriginalLinkage)));
        // The kept copy is eligible for auto-hiding (hidden visibility) if all
        // copies were (i.e. they were all linkonce_odr global unnamed addr).
        // If any copy is not (e.g. it was originally weak_odr), then the symbol
        // must remain externally available (e.g. a weak_odr from an explicitly
        // instantiated template). Additionally, if it is in the
        // GUIDPreservedSymbols set, that means that it is visibile outside
        // the summary (e.g. in a native object or a bitcode file without
        // summary), and in that case we cannot hide it as it isn't possible to
        // check all copies.
        S->setCanAutoHide(VI.canAutoHide() &&
                          !GUIDPreservedSymbols.count(VI.getGUID()));
      }
      if (C.VisibilityScheme == Config::FromPrevailing)
        Visibility = S->getVisibility();
    }
    // Alias and aliasee can't be turned into available_externally.
    else if (!isa<AliasSummary>(S.get()) &&
             !GlobalInvolvedWithAlias.count(S.get()))
      S->setLinkage(GlobalValue::AvailableExternallyLinkage);

    // For ELF, set visibility to the computed visibility from summaries. We
    // don't track visibility from declarations so this may be more relaxed than
    // the most constraining one.
    if (C.VisibilityScheme == Config::ELF)
      S->setVisibility(Visibility);

    if (S->linkage() != OriginalLinkage)
      recordNewLinkage(S->modulePath(), VI.getGUID(), S->linkage());
  }

  if (C.VisibilityScheme == Config::FromPrevailing) {
    for (auto &S : VI.getSummaryList()) {
      GlobalValue::LinkageTypes OriginalLinkage = S->linkage();
      if (GlobalValue::isLocalLinkage(OriginalLinkage) ||
          GlobalValue::isAppendingLinkage(S->linkage()))
        continue;
      S->setVisibility(Visibility);
    }
  }
}

/// Resolve linkage for prevailing symbols in the \p Index.
//
// We'd like to drop these functions if they are no longer referenced in the
// current module. However there is a chance that another module is still
// referencing them because of the import. We make sure we always emit at least
// one copy.
void llvm::thinLTOResolvePrevailingInIndex(
    const Config &C, ModuleSummaryIndex &Index,
    function_ref<bool(GlobalValue::GUID, const GlobalValueSummary *)>
        isPrevailing,
    function_ref<void(StringRef, GlobalValue::GUID, GlobalValue::LinkageTypes)>
        recordNewLinkage,
    const DenseSet<GlobalValue::GUID> &GUIDPreservedSymbols) {
  // We won't optimize the globals that are referenced by an alias for now
  // Ideally we should turn the alias into a global and duplicate the definition
  // when needed.
  DenseSet<GlobalValueSummary *> GlobalInvolvedWithAlias;
  for (auto &I : Index)
    for (auto &S : I.second.SummaryList)
      if (auto AS = dyn_cast<AliasSummary>(S.get()))
        GlobalInvolvedWithAlias.insert(&AS->getAliasee());

  for (auto &I : Index)
    thinLTOResolvePrevailingGUID(C, Index.getValueInfo(I),
                                 GlobalInvolvedWithAlias, isPrevailing,
                                 recordNewLinkage, GUIDPreservedSymbols);
}

static void thinLTOInternalizeAndPromoteGUID(
    ValueInfo VI, function_ref<bool(StringRef, ValueInfo)> isExported,
    function_ref<bool(GlobalValue::GUID, const GlobalValueSummary *)>
        isPrevailing) {
  auto ExternallyVisibleCopies =
      llvm::count_if(VI.getSummaryList(),
                     [](const std::unique_ptr<GlobalValueSummary> &Summary) {
                       return !GlobalValue::isLocalLinkage(Summary->linkage());
                     });

  for (auto &S : VI.getSummaryList()) {
    // First see if we need to promote an internal value because it is not
    // exported.
    if (isExported(S->modulePath(), VI)) {
      if (GlobalValue::isLocalLinkage(S->linkage()))
        S->setLinkage(GlobalValue::ExternalLinkage);
      continue;
    }

    // Otherwise, see if we can internalize.
    if (!EnableLTOInternalization)
      continue;

    // Ignore local and appending linkage values since the linker
    // doesn't resolve them (and there is no need to internalize if this is
    // already internal).
    if (GlobalValue::isLocalLinkage(S->linkage()) ||
        S->linkage() == GlobalValue::AppendingLinkage)
      continue;

    // We can't internalize available_externally globals because this
    // can break function pointer equality.
    if (S->linkage() == GlobalValue::AvailableExternallyLinkage)
      continue;

    bool IsPrevailing = isPrevailing(VI.getGUID(), S.get());

    if (GlobalValue::isInterposableLinkage(S->linkage()) && !IsPrevailing)
      continue;

    // Non-exported functions and variables with linkonce_odr or weak_odr
    // linkage can be internalized in certain cases. The minimum legality
    // requirements would be that they are not address taken to ensure that we
    // don't break pointer equality checks, and that variables are either read-
    // or write-only. For functions, this is the case if either all copies are
    // [local_]unnamed_addr, or we can propagate reference edge attributes
    // (which is how this is guaranteed for variables, when analyzing whether
    // they are read or write-only).
    //
    // However, we only get to this code for weak/linkonce ODR values in one of
    // two cases:
    // 1) The prevailing copy is not in IR (it is in native code).
    // 2) The prevailing copy in IR is not exported from its module.
    // Additionally, at least for the new LTO API, case 2 will only happen if
    // there is exactly one definition of the value (i.e. in exactly one
    // module), as duplicate defs are result in the value being marked exported.
    // Likely, users of the legacy LTO API are similar, however, currently there
    // are llvm-lto based tests of the legacy LTO API that do not mark
    // duplicate linkonce_odr copies as exported via the tool, so we need
    // to handle that case below by checking the number of copies.
    //
    // Generally, we only want to internalize a linkonce/weak ODR value in case
    // 2, because in case 1 we cannot see how the value is used to know if it
    // is read or write-only. We also don't want to bloat the binary with
    // multiple internalized copies of non-prevailing linkonce_odr functions.
    // Note if we don't internalize, we will convert non-prevailing copies to
    // available_externally anyway, so that we drop them after inlining. The
    // only reason to internalize such a function is if we indeed have a single
    // copy, because internalizing it won't increase binary size, and enables
    // use of inliner heuristics that are more aggressive in the face of a
    // single call to a static (local). For variables, internalizing a read or
    // write only variable can enable more aggressive optimization. However, we
    // already perform this elsewhere in the ThinLTO backend handling for
    // read or write-only variables (processGlobalForThinLTO).
    //
    // Therefore, only internalize linkonce/weak ODR if there is a single copy,
    // that is prevailing in this IR module. We can do so aggressively, without
    // requiring the address to be insignificant, or that a variable be read or
    // write-only.
    if ((S->linkage() == GlobalValue::WeakODRLinkage ||
         S->linkage() == GlobalValue::LinkOnceODRLinkage) &&
        // We can have only one copy in ThinLTO that isn't prevailing, if the
        // prevailing copy is in a native object.
        (!IsPrevailing || ExternallyVisibleCopies > 1))
      continue;

    S->setLinkage(GlobalValue::InternalLinkage);
  }
}

// Update the linkages in the given \p Index to mark exported values
// as external and non-exported values as internal.
void llvm::thinLTOInternalizeAndPromoteInIndex(
    ModuleSummaryIndex &Index,
    function_ref<bool(StringRef, ValueInfo)> isExported,
    function_ref<bool(GlobalValue::GUID, const GlobalValueSummary *)>
        isPrevailing) {
  for (auto &I : Index)
    thinLTOInternalizeAndPromoteGUID(Index.getValueInfo(I), isExported,
                                     isPrevailing);
}

// Requires a destructor for std::vector<InputModule>.
InputFile::~InputFile() = default;

Expected<std::unique_ptr<InputFile>> InputFile::create(MemoryBufferRef Object) {
  std::unique_ptr<InputFile> File(new InputFile);

  Expected<IRSymtabFile> FOrErr = readIRSymtab(Object);
  if (!FOrErr)
    return FOrErr.takeError();

  File->TargetTriple = FOrErr->TheReader.getTargetTriple();
  File->SourceFileName = FOrErr->TheReader.getSourceFileName();
  File->COFFLinkerOpts = FOrErr->TheReader.getCOFFLinkerOpts();
  File->DependentLibraries = FOrErr->TheReader.getDependentLibraries();
  File->ComdatTable = FOrErr->TheReader.getComdatTable();

  for (unsigned I = 0; I != FOrErr->Mods.size(); ++I) {
    size_t Begin = File->Symbols.size();
    for (const irsymtab::Reader::SymbolRef &Sym :
         FOrErr->TheReader.module_symbols(I))
      // Skip symbols that are irrelevant to LTO. Note that this condition needs
      // to match the one in Skip() in LTO::addRegularLTO().
      if (Sym.isGlobal() && !Sym.isFormatSpecific())
        File->Symbols.push_back(Sym);
    File->ModuleSymIndices.push_back({Begin, File->Symbols.size()});
  }

  File->Mods = FOrErr->Mods;
  File->Strtab = std::move(FOrErr->Strtab);
  return std::move(File);
}

StringRef InputFile::getName() const {
  return Mods[0].getModuleIdentifier();
}

BitcodeModule &InputFile::getSingleBitcodeModule() {
  assert(Mods.size() == 1 && "Expect only one bitcode module");
  return Mods[0];
}

LTO::RegularLTOState::RegularLTOState(unsigned ParallelCodeGenParallelismLevel,
                                      const Config &Conf)
    : ParallelCodeGenParallelismLevel(ParallelCodeGenParallelismLevel),
      Ctx(Conf), CombinedModule(std::make_unique<Module>("ld-temp.o", Ctx)),
      Mover(std::make_unique<IRMover>(*CombinedModule)) {}

LTO::ThinLTOState::ThinLTOState(ThinBackend Backend)
    : Backend(Backend), CombinedIndex(/*HaveGVs*/ false) {
  if (!Backend)
    this->Backend =
        createInProcessThinBackend(llvm::heavyweight_hardware_concurrency());
}

LTO::LTO(Config Conf, ThinBackend Backend,
         unsigned ParallelCodeGenParallelismLevel, LTOKind LTOMode)
    : Conf(std::move(Conf)),
      RegularLTO(ParallelCodeGenParallelismLevel, this->Conf),
      ThinLTO(std::move(Backend)), LTOMode(LTOMode) {}

// Requires a destructor for MapVector<BitcodeModule>.
LTO::~LTO() = default;

// Add the symbols in the given module to the GlobalResolutions map, and resolve
// their partitions.
void LTO::addModuleToGlobalRes(ArrayRef<InputFile::Symbol> Syms,
                               ArrayRef<SymbolResolution> Res,
                               unsigned Partition, bool InSummary) {
  auto *ResI = Res.begin();
  auto *ResE = Res.end();
  (void)ResE;
  const Triple TT(RegularLTO.CombinedModule->getTargetTriple());
  for (const InputFile::Symbol &Sym : Syms) {
    assert(ResI != ResE);
    SymbolResolution Res = *ResI++;

    StringRef Name = Sym.getName();
    // Strip the __imp_ prefix from COFF dllimport symbols (similar to the
    // way they are handled by lld), otherwise we can end up with two
    // global resolutions (one with and one for a copy of the symbol without).
    if (TT.isOSBinFormatCOFF() && Name.startswith("__imp_"))
      Name = Name.substr(strlen("__imp_"));
    auto &GlobalRes = GlobalResolutions[Name];
    GlobalRes.UnnamedAddr &= Sym.isUnnamedAddr();
    if (Res.Prevailing) {
      assert(!GlobalRes.Prevailing &&
             "Multiple prevailing defs are not allowed");
      GlobalRes.Prevailing = true;
      GlobalRes.IRName = std::string(Sym.getIRName());
    } else if (!GlobalRes.Prevailing && GlobalRes.IRName.empty()) {
      // Sometimes it can be two copies of symbol in a module and prevailing
      // symbol can have no IR name. That might happen if symbol is defined in
      // module level inline asm block. In case we have multiple modules with
      // the same symbol we want to use IR name of the prevailing symbol.
      // Otherwise, if we haven't seen a prevailing symbol, set the name so that
      // we can later use it to check if there is any prevailing copy in IR.
      GlobalRes.IRName = std::string(Sym.getIRName());
    }

    // In rare occasion, the symbol used to initialize GlobalRes has a different
    // IRName from the inspected Symbol. This can happen on macOS + iOS, when a
    // symbol is referenced through its mangled name, say @"\01_symbol" while
    // the IRName is @symbol (the prefix underscore comes from MachO mangling).
    // In that case, we have the same actual Symbol that can get two different
    // GUID, leading to some invalid internalization. Workaround this by marking
    // the GlobalRes external.

    // FIXME: instead of this check, it would be desirable to compute GUIDs
    // based on mangled name, but this requires an access to the Target Triple
    // and would be relatively invasive on the codebase.
    if (GlobalRes.IRName != Sym.getIRName()) {
      GlobalRes.Partition = GlobalResolution::External;
      GlobalRes.VisibleOutsideSummary = true;
    }

    // Set the partition to external if we know it is re-defined by the linker
    // with -defsym or -wrap options, used elsewhere, e.g. it is visible to a
    // regular object, is referenced from llvm.compiler.used/llvm.used, or was
    // already recorded as being referenced from a different partition.
    if (Res.LinkerRedefined || Res.VisibleToRegularObj || Sym.isUsed() ||
        (GlobalRes.Partition != GlobalResolution::Unknown &&
         GlobalRes.Partition != Partition)) {
      GlobalRes.Partition = GlobalResolution::External;
    } else
      // First recorded reference, save the current partition.
      GlobalRes.Partition = Partition;

    // Flag as visible outside of summary if visible from a regular object or
    // from a module that does not have a summary.
    GlobalRes.VisibleOutsideSummary |=
        (Res.VisibleToRegularObj || Sym.isUsed() || !InSummary);

    GlobalRes.ExportDynamic |= Res.ExportDynamic;
  }
}

static void writeToResolutionFile(raw_ostream &OS, InputFile *Input,
                                  ArrayRef<SymbolResolution> Res) {
  StringRef Path = Input->getName();
  OS << Path << '\n';
  auto ResI = Res.begin();
  for (const InputFile::Symbol &Sym : Input->symbols()) {
    assert(ResI != Res.end());
    SymbolResolution Res = *ResI++;

    OS << "-r=" << Path << ',' << Sym.getName() << ',';
    if (Res.Prevailing)
      OS << 'p';
    if (Res.FinalDefinitionInLinkageUnit)
      OS << 'l';
    if (Res.VisibleToRegularObj)
      OS << 'x';
    if (Res.LinkerRedefined)
      OS << 'r';
    OS << '\n';
  }
  OS.flush();
  assert(ResI == Res.end());
}

Error LTO::add(std::unique_ptr<InputFile> Input,
               ArrayRef<SymbolResolution> Res) {
  assert(!CalledGetMaxTasks);

  if (Conf.ResolutionFile)
    writeToResolutionFile(*Conf.ResolutionFile, Input.get(), Res);

  if (RegularLTO.CombinedModule->getTargetTriple().empty()) {
    RegularLTO.CombinedModule->setTargetTriple(Input->getTargetTriple());
    if (Triple(Input->getTargetTriple()).isOSBinFormatELF())
      Conf.VisibilityScheme = Config::ELF;
  }

  const SymbolResolution *ResI = Res.begin();
  for (unsigned I = 0; I != Input->Mods.size(); ++I)
    if (Error Err = addModule(*Input, I, ResI, Res.end()))
      return Err;

  assert(ResI == Res.end());
  return Error::success();
}

Error LTO::addModule(InputFile &Input, unsigned ModI,
                     const SymbolResolution *&ResI,
                     const SymbolResolution *ResE) {
  Expected<BitcodeLTOInfo> LTOInfo = Input.Mods[ModI].getLTOInfo();
  if (!LTOInfo)
    return LTOInfo.takeError();

  if (EnableSplitLTOUnit) {
    // If only some modules were split, flag this in the index so that
    // we can skip or error on optimizations that need consistently split
    // modules (whole program devirt and lower type tests).
    if (*EnableSplitLTOUnit != LTOInfo->EnableSplitLTOUnit)
      ThinLTO.CombinedIndex.setPartiallySplitLTOUnits();
  } else
    EnableSplitLTOUnit = LTOInfo->EnableSplitLTOUnit;

  BitcodeModule BM = Input.Mods[ModI];

  if ((LTOMode == LTOK_UnifiedRegular || LTOMode == LTOK_UnifiedThin) &&
      !LTOInfo->UnifiedLTO)
    return make_error<StringError>(
        "unified LTO compilation must use "
        "compatible bitcode modules (use -funified-lto)",
        inconvertibleErrorCode());

  if (LTOInfo->UnifiedLTO && LTOMode == LTOK_Default)
    LTOMode = LTOK_UnifiedThin;

  bool IsThinLTO = LTOInfo->IsThinLTO && (LTOMode != LTOK_UnifiedRegular);

  auto ModSyms = Input.module_symbols(ModI);
  addModuleToGlobalRes(ModSyms, {ResI, ResE},
                       IsThinLTO ? ThinLTO.ModuleMap.size() + 1 : 0,
                       LTOInfo->HasSummary);

  if (IsThinLTO)
    return addThinLTO(BM, ModSyms, ResI, ResE);

  RegularLTO.EmptyCombinedModule = false;
  Expected<RegularLTOState::AddedModule> ModOrErr =
      addRegularLTO(BM, ModSyms, ResI, ResE);
  if (!ModOrErr)
    return ModOrErr.takeError();

  if (!LTOInfo->HasSummary)
    return linkRegularLTO(std::move(*ModOrErr), /*LivenessFromIndex=*/false);

  // Regular LTO module summaries are added to a dummy module that represents
  // the combined regular LTO module.
  if (Error Err = BM.readSummary(ThinLTO.CombinedIndex, "", -1ull))
    return Err;
  RegularLTO.ModsWithSummaries.push_back(std::move(*ModOrErr));
  return Error::success();
}

// Checks whether the given global value is in a non-prevailing comdat
// (comdat containing values the linker indicated were not prevailing,
// which we then dropped to available_externally), and if so, removes
// it from the comdat. This is called for all global values to ensure the
// comdat is empty rather than leaving an incomplete comdat. It is needed for
// regular LTO modules, in case we are in a mixed-LTO mode (both regular
// and thin LTO modules) compilation. Since the regular LTO module will be
// linked first in the final native link, we want to make sure the linker
// doesn't select any of these incomplete comdats that would be left
// in the regular LTO module without this cleanup.
static void
handleNonPrevailingComdat(GlobalValue &GV,
                          std::set<const Comdat *> &NonPrevailingComdats) {
  Comdat *C = GV.getComdat();
  if (!C)
    return;

  if (!NonPrevailingComdats.count(C))
    return;

  // Additionally need to drop all global values from the comdat to
  // available_externally, to satisfy the COMDAT requirement that all members
  // are discarded as a unit. The non-local linkage global values avoid
  // duplicate definition linker errors.
  GV.setLinkage(GlobalValue::AvailableExternallyLinkage);

  if (auto GO = dyn_cast<GlobalObject>(&GV))
    GO->setComdat(nullptr);
}

// Add a regular LTO object to the link.
// The resulting module needs to be linked into the combined LTO module with
// linkRegularLTO.
Expected<LTO::RegularLTOState::AddedModule>
LTO::addRegularLTO(BitcodeModule BM, ArrayRef<InputFile::Symbol> Syms,
                   const SymbolResolution *&ResI,
                   const SymbolResolution *ResE) {
  RegularLTOState::AddedModule Mod;
  Expected<std::unique_ptr<Module>> MOrErr =
      BM.getLazyModule(RegularLTO.Ctx, /*ShouldLazyLoadMetadata*/ true,
                       /*IsImporting*/ false);
  if (!MOrErr)
    return MOrErr.takeError();
  Module &M = **MOrErr;
  Mod.M = std::move(*MOrErr);

  if (Error Err = M.materializeMetadata())
    return std::move(Err);

  // If cfi.functions is present and we are in regular LTO mode, LowerTypeTests
  // will rename local functions in the merged module as "<function name>.1".
  // This causes linking errors, since other parts of the module expect the
  // original function name.
  if (LTOMode == LTOK_UnifiedRegular)
    if (NamedMDNode *CfiFunctionsMD = M.getNamedMetadata("cfi.functions"))
      M.eraseNamedMetadata(CfiFunctionsMD);

  UpgradeDebugInfo(M);

  ModuleSymbolTable SymTab;
  SymTab.addModule(&M);

  for (GlobalVariable &GV : M.globals())
    if (GV.hasAppendingLinkage())
      Mod.Keep.push_back(&GV);

  DenseSet<GlobalObject *> AliasedGlobals;
  for (auto &GA : M.aliases())
    if (GlobalObject *GO = GA.getAliaseeObject())
      AliasedGlobals.insert(GO);

  // In this function we need IR GlobalValues matching the symbols in Syms
  // (which is not backed by a module), so we need to enumerate them in the same
  // order. The symbol enumeration order of a ModuleSymbolTable intentionally
  // matches the order of an irsymtab, but when we read the irsymtab in
  // InputFile::create we omit some symbols that are irrelevant to LTO. The
  // Skip() function skips the same symbols from the module as InputFile does
  // from the symbol table.
  auto MsymI = SymTab.symbols().begin(), MsymE = SymTab.symbols().end();
  auto Skip = [&]() {
    while (MsymI != MsymE) {
      auto Flags = SymTab.getSymbolFlags(*MsymI);
      if ((Flags & object::BasicSymbolRef::SF_Global) &&
          !(Flags & object::BasicSymbolRef::SF_FormatSpecific))
        return;
      ++MsymI;
    }
  };
  Skip();

  std::set<const Comdat *> NonPrevailingComdats;
  SmallSet<StringRef, 2> NonPrevailingAsmSymbols;
  for (const InputFile::Symbol &Sym : Syms) {
    assert(ResI != ResE);
    SymbolResolution Res = *ResI++;

    assert(MsymI != MsymE);
    ModuleSymbolTable::Symbol Msym = *MsymI++;
    Skip();

    if (GlobalValue *GV = dyn_cast_if_present<GlobalValue *>(Msym)) {
      if (Res.Prevailing) {
        if (Sym.isUndefined())
          continue;
        Mod.Keep.push_back(GV);
        // For symbols re-defined with linker -wrap and -defsym options,
        // set the linkage to weak to inhibit IPO. The linkage will be
        // restored by the linker.
        if (Res.LinkerRedefined)
          GV->setLinkage(GlobalValue::WeakAnyLinkage);

        GlobalValue::LinkageTypes OriginalLinkage = GV->getLinkage();
        if (GlobalValue::isLinkOnceLinkage(OriginalLinkage))
          GV->setLinkage(GlobalValue::getWeakLinkage(
              GlobalValue::isLinkOnceODRLinkage(OriginalLinkage)));
      } else if (isa<GlobalObject>(GV) &&
                 (GV->hasLinkOnceODRLinkage() || GV->hasWeakODRLinkage() ||
                  GV->hasAvailableExternallyLinkage()) &&
                 !AliasedGlobals.count(cast<GlobalObject>(GV))) {
        // Any of the above three types of linkage indicates that the
        // chosen prevailing symbol will have the same semantics as this copy of
        // the symbol, so we may be able to link it with available_externally
        // linkage. We will decide later whether to do that when we link this
        // module (in linkRegularLTO), based on whether it is undefined.
        Mod.Keep.push_back(GV);
        GV->setLinkage(GlobalValue::AvailableExternallyLinkage);
        if (GV->hasComdat())
          NonPrevailingComdats.insert(GV->getComdat());
        cast<GlobalObject>(GV)->setComdat(nullptr);
      }

      // Set the 'local' flag based on the linker resolution for this symbol.
      if (Res.FinalDefinitionInLinkageUnit) {
        GV->setDSOLocal(true);
        if (GV->hasDLLImportStorageClass())
          GV->setDLLStorageClass(GlobalValue::DLLStorageClassTypes::
                                 DefaultStorageClass);
      }
    } else if (auto *AS =
                   dyn_cast_if_present<ModuleSymbolTable::AsmSymbol *>(Msym)) {
      // Collect non-prevailing symbols.
      if (!Res.Prevailing)
        NonPrevailingAsmSymbols.insert(AS->first);
    } else {
      llvm_unreachable("unknown symbol type");
    }

    // Common resolution: collect the maximum size/alignment over all commons.
    // We also record if we see an instance of a common as prevailing, so that
    // if none is prevailing we can ignore it later.
    if (Sym.isCommon()) {
      // FIXME: We should figure out what to do about commons defined by asm.
      // For now they aren't reported correctly by ModuleSymbolTable.
      auto &CommonRes = RegularLTO.Commons[std::string(Sym.getIRName())];
      CommonRes.Size = std::max(CommonRes.Size, Sym.getCommonSize());
      if (uint32_t SymAlignValue = Sym.getCommonAlignment()) {
        CommonRes.Alignment =
            std::max(Align(SymAlignValue), CommonRes.Alignment);
      }
      CommonRes.Prevailing |= Res.Prevailing;
    }
  }

  if (!M.getComdatSymbolTable().empty())
    for (GlobalValue &GV : M.global_values())
      handleNonPrevailingComdat(GV, NonPrevailingComdats);

  // Prepend ".lto_discard <sym>, <sym>*" directive to each module inline asm
  // block.
  if (!M.getModuleInlineAsm().empty()) {
    std::string NewIA = ".lto_discard";
    if (!NonPrevailingAsmSymbols.empty()) {
      // Don't dicard a symbol if there is a live .symver for it.
      ModuleSymbolTable::CollectAsmSymvers(
          M, [&](StringRef Name, StringRef Alias) {
            if (!NonPrevailingAsmSymbols.count(Alias))
              NonPrevailingAsmSymbols.erase(Name);
          });
      NewIA += " " + llvm::join(NonPrevailingAsmSymbols, ", ");
    }
    NewIA += "\n";
    M.setModuleInlineAsm(NewIA + M.getModuleInlineAsm());
  }

  assert(MsymI == MsymE);
  return std::move(Mod);
}

Error LTO::linkRegularLTO(RegularLTOState::AddedModule Mod,
                          bool LivenessFromIndex) {
  std::vector<GlobalValue *> Keep;
  for (GlobalValue *GV : Mod.Keep) {
    if (LivenessFromIndex && !ThinLTO.CombinedIndex.isGUIDLive(GV->getGUID())) {
      if (Function *F = dyn_cast<Function>(GV)) {
        if (DiagnosticOutputFile) {
          if (Error Err = F->materialize())
            return Err;
          OptimizationRemarkEmitter ORE(F, nullptr);
          ORE.emit(OptimizationRemark(DEBUG_TYPE, "deadfunction", F)
                   << ore::NV("Function", F)
                   << " not added to the combined module ");
        }
      }
      continue;
    }

    if (!GV->hasAvailableExternallyLinkage()) {
      Keep.push_back(GV);
      continue;
    }

    // Only link available_externally definitions if we don't already have a
    // definition.
    GlobalValue *CombinedGV =
        RegularLTO.CombinedModule->getNamedValue(GV->getName());
    if (CombinedGV && !CombinedGV->isDeclaration())
      continue;

    Keep.push_back(GV);
  }

  return RegularLTO.Mover->move(std::move(Mod.M), Keep, nullptr,
                                /* IsPerformingImport */ false);
}

// Add a ThinLTO module to the link.
Error LTO::addThinLTO(BitcodeModule BM, ArrayRef<InputFile::Symbol> Syms,
                      const SymbolResolution *&ResI,
                      const SymbolResolution *ResE) {
  const SymbolResolution *ResITmp = ResI;
  for (const InputFile::Symbol &Sym : Syms) {
    assert(ResITmp != ResE);
    SymbolResolution Res = *ResITmp++;

    if (!Sym.getIRName().empty()) {
      auto GUID = GlobalValue::getGUID(GlobalValue::getGlobalIdentifier(
          Sym.getIRName(), GlobalValue::ExternalLinkage, ""));
      if (Res.Prevailing)
        ThinLTO.PrevailingModuleForGUID[GUID] = BM.getModuleIdentifier();
    }
  }

  uint64_t ModuleId = ThinLTO.ModuleMap.size();
  if (Error Err =
          BM.readSummary(ThinLTO.CombinedIndex, BM.getModuleIdentifier(),
                         ModuleId, [&](GlobalValue::GUID GUID) {
                           return ThinLTO.PrevailingModuleForGUID[GUID] ==
                                  BM.getModuleIdentifier();
                         }))
    return Err;
  LLVM_DEBUG(dbgs() << "Module " << ModuleId << ": " << BM.getModuleIdentifier()
                    << "\n");

  for (const InputFile::Symbol &Sym : Syms) {
    assert(ResI != ResE);
    SymbolResolution Res = *ResI++;

    if (!Sym.getIRName().empty()) {
      auto GUID = GlobalValue::getGUID(GlobalValue::getGlobalIdentifier(
          Sym.getIRName(), GlobalValue::ExternalLinkage, ""));
      if (Res.Prevailing) {
        assert(ThinLTO.PrevailingModuleForGUID[GUID] ==
               BM.getModuleIdentifier());

        // For linker redefined symbols (via --wrap or --defsym) we want to
        // switch the linkage to `weak` to prevent IPOs from happening.
        // Find the summary in the module for this very GV and record the new
        // linkage so that we can switch it when we import the GV.
        if (Res.LinkerRedefined)
          if (auto S = ThinLTO.CombinedIndex.findSummaryInModule(
                  GUID, BM.getModuleIdentifier()))
            S->setLinkage(GlobalValue::WeakAnyLinkage);
      }

      // If the linker resolved the symbol to a local definition then mark it
      // as local in the summary for the module we are adding.
      if (Res.FinalDefinitionInLinkageUnit) {
        if (auto S = ThinLTO.CombinedIndex.findSummaryInModule(
                GUID, BM.getModuleIdentifier())) {
          S->setDSOLocal(true);
        }
      }
    }
  }

  if (!ThinLTO.ModuleMap.insert({BM.getModuleIdentifier(), BM}).second)
    return make_error<StringError>(
        "Expected at most one ThinLTO module per bitcode file",
        inconvertibleErrorCode());

  if (!Conf.ThinLTOModulesToCompile.empty()) {
    if (!ThinLTO.ModulesToCompile)
      ThinLTO.ModulesToCompile = ModuleMapType();
    // This is a fuzzy name matching where only modules with name containing the
    // specified switch values are going to be compiled.
    for (const std::string &Name : Conf.ThinLTOModulesToCompile) {
      if (BM.getModuleIdentifier().contains(Name)) {
        ThinLTO.ModulesToCompile->insert({BM.getModuleIdentifier(), BM});
        llvm::errs() << "[ThinLTO] Selecting " << BM.getModuleIdentifier()
                     << " to compile\n";
      }
    }
  }

  return Error::success();
}

unsigned LTO::getMaxTasks() const {
  CalledGetMaxTasks = true;
  auto ModuleCount = ThinLTO.ModulesToCompile ? ThinLTO.ModulesToCompile->size()
                                              : ThinLTO.ModuleMap.size();
  return RegularLTO.ParallelCodeGenParallelismLevel + ModuleCount;
}

// If only some of the modules were split, we cannot correctly handle
// code that contains type tests or type checked loads.
Error LTO::checkPartiallySplit() {
  if (!ThinLTO.CombinedIndex.partiallySplitLTOUnits())
    return Error::success();

  Function *TypeTestFunc = RegularLTO.CombinedModule->getFunction(
      Intrinsic::getName(Intrinsic::type_test));
  Function *TypeCheckedLoadFunc = RegularLTO.CombinedModule->getFunction(
      Intrinsic::getName(Intrinsic::type_checked_load));
  Function *TypeCheckedLoadRelativeFunc =
      RegularLTO.CombinedModule->getFunction(
          Intrinsic::getName(Intrinsic::type_checked_load_relative));

  // First check if there are type tests / type checked loads in the
  // merged regular LTO module IR.
  if ((TypeTestFunc && !TypeTestFunc->use_empty()) ||
      (TypeCheckedLoadFunc && !TypeCheckedLoadFunc->use_empty()) ||
      (TypeCheckedLoadRelativeFunc &&
       !TypeCheckedLoadRelativeFunc->use_empty()))
    return make_error<StringError>(
        "inconsistent LTO Unit splitting (recompile with -fsplit-lto-unit)",
        inconvertibleErrorCode());

  // Otherwise check if there are any recorded in the combined summary from the
  // ThinLTO modules.
  for (auto &P : ThinLTO.CombinedIndex) {
    for (auto &S : P.second.SummaryList) {
      auto *FS = dyn_cast<FunctionSummary>(S.get());
      if (!FS)
        continue;
      if (!FS->type_test_assume_vcalls().empty() ||
          !FS->type_checked_load_vcalls().empty() ||
          !FS->type_test_assume_const_vcalls().empty() ||
          !FS->type_checked_load_const_vcalls().empty() ||
          !FS->type_tests().empty())
        return make_error<StringError>(
            "inconsistent LTO Unit splitting (recompile with -fsplit-lto-unit)",
            inconvertibleErrorCode());
    }
  }
  return Error::success();
}

Error LTO::run(AddStreamFn AddStream, FileCache Cache) {
  // Compute "dead" symbols, we don't want to import/export these!
  DenseSet<GlobalValue::GUID> GUIDPreservedSymbols;
  DenseMap<GlobalValue::GUID, PrevailingType> GUIDPrevailingResolutions;
  for (auto &Res : GlobalResolutions) {
    // Normally resolution have IR name of symbol. We can do nothing here
    // otherwise. See comments in GlobalResolution struct for more details.
    if (Res.second.IRName.empty())
      continue;

    GlobalValue::GUID GUID = GlobalValue::getGUID(
        GlobalValue::dropLLVMManglingEscape(Res.second.IRName));

    if (Res.second.VisibleOutsideSummary && Res.second.Prevailing)
      GUIDPreservedSymbols.insert(GUID);

    if (Res.second.ExportDynamic)
      DynamicExportSymbols.insert(GUID);

    GUIDPrevailingResolutions[GUID] =
        Res.second.Prevailing ? PrevailingType::Yes : PrevailingType::No;
  }

  auto isPrevailing = [&](GlobalValue::GUID G) {
    auto It = GUIDPrevailingResolutions.find(G);
    if (It == GUIDPrevailingResolutions.end())
      return PrevailingType::Unknown;
    return It->second;
  };
  computeDeadSymbolsWithConstProp(ThinLTO.CombinedIndex, GUIDPreservedSymbols,
                                  isPrevailing, Conf.OptLevel > 0);

  // Setup output file to emit statistics.
  auto StatsFileOrErr = setupStatsFile(Conf.StatsFile);
  if (!StatsFileOrErr)
    return StatsFileOrErr.takeError();
  std::unique_ptr<ToolOutputFile> StatsFile = std::move(StatsFileOrErr.get());

  // TODO: Ideally this would be controlled automatically by detecting that we
  // are linking with an allocator that supports these interfaces, rather than
  // an internal option (which would still be needed for tests, however). For
  // example, if the library exported a symbol like __malloc_hot_cold the linker
  // could recognize that and set a flag in the lto::Config.
  if (SupportsHotColdNew)
    ThinLTO.CombinedIndex.setWithSupportsHotColdNew();

  Error Result = runRegularLTO(AddStream);
  if (!Result)
    Result = runThinLTO(AddStream, Cache, GUIDPreservedSymbols);

  if (StatsFile)
    PrintStatisticsJSON(StatsFile->os());

  return Result;
}

void lto::updateMemProfAttributes(Module &Mod,
                                  const ModuleSummaryIndex &Index) {
  if (Index.withSupportsHotColdNew())
    return;

  // The profile matcher applies hotness attributes directly for allocations,
  // and those will cause us to generate calls to the hot/cold interfaces
  // unconditionally. If supports-hot-cold-new was not enabled in the LTO
  // link then assume we don't want these calls (e.g. not linking with
  // the appropriate library, or otherwise trying to disable this behavior).
  for (auto &F : Mod) {
    for (auto &BB : F) {
      for (auto &I : BB) {
        auto *CI = dyn_cast<CallBase>(&I);
        if (!CI)
          continue;
        if (CI->hasFnAttr("memprof"))
          CI->removeFnAttr("memprof");
        // Strip off all memprof metadata as it is no longer needed.
        // Importantly, this avoids the addition of new memprof attributes
        // after inlining propagation.
        // TODO: If we support additional types of MemProf metadata beyond hot
        // and cold, we will need to update the metadata based on the allocator
        // APIs supported instead of completely stripping all.
        CI->setMetadata(LLVMContext::MD_memprof, nullptr);
        CI->setMetadata(LLVMContext::MD_callsite, nullptr);
      }
    }
  }
}

Error LTO::runRegularLTO(AddStreamFn AddStream) {
  // Setup optimization remarks.
  auto DiagFileOrErr = lto::setupLLVMOptimizationRemarks(
      RegularLTO.CombinedModule->getContext(), Conf.RemarksFilename,
      Conf.RemarksPasses, Conf.RemarksFormat, Conf.RemarksWithHotness,
      Conf.RemarksHotnessThreshold);
  LLVM_DEBUG(dbgs() << "Running regular LTO\n");
  if (!DiagFileOrErr)
    return DiagFileOrErr.takeError();
  DiagnosticOutputFile = std::move(*DiagFileOrErr);

  // Finalize linking of regular LTO modules containing summaries now that
  // we have computed liveness information.
  for (auto &M : RegularLTO.ModsWithSummaries)
    if (Error Err = linkRegularLTO(std::move(M),
                                   /*LivenessFromIndex=*/true))
      return Err;

  // Ensure we don't have inconsistently split LTO units with type tests.
  // FIXME: this checks both LTO and ThinLTO. It happens to work as we take
  // this path both cases but eventually this should be split into two and
  // do the ThinLTO checks in `runThinLTO`.
  if (Error Err = checkPartiallySplit())
    return Err;

  // Make sure commons have the right size/alignment: we kept the largest from
  // all the prevailing when adding the inputs, and we apply it here.
  const DataLayout &DL = RegularLTO.CombinedModule->getDataLayout();
  for (auto &I : RegularLTO.Commons) {
    if (!I.second.Prevailing)
      // Don't do anything if no instance of this common was prevailing.
      continue;
    GlobalVariable *OldGV = RegularLTO.CombinedModule->getNamedGlobal(I.first);
    if (OldGV && DL.getTypeAllocSize(OldGV->getValueType()) == I.second.Size) {
      // Don't create a new global if the type is already correct, just make
      // sure the alignment is correct.
      OldGV->setAlignment(I.second.Alignment);
      continue;
    }
    ArrayType *Ty =
        ArrayType::get(Type::getInt8Ty(RegularLTO.Ctx), I.second.Size);
    auto *GV = new GlobalVariable(*RegularLTO.CombinedModule, Ty, false,
                                  GlobalValue::CommonLinkage,
                                  ConstantAggregateZero::get(Ty), "");
    GV->setAlignment(I.second.Alignment);
    if (OldGV) {
      OldGV->replaceAllUsesWith(ConstantExpr::getBitCast(GV, OldGV->getType()));
      GV->takeName(OldGV);
      OldGV->eraseFromParent();
    } else {
      GV->setName(I.first);
    }
  }

  updateMemProfAttributes(*RegularLTO.CombinedModule, ThinLTO.CombinedIndex);

  // If allowed, upgrade public vcall visibility metadata to linkage unit
  // visibility before whole program devirtualization in the optimizer.
  updateVCallVisibilityInModule(*RegularLTO.CombinedModule,
                                Conf.HasWholeProgramVisibility,
                                DynamicExportSymbols);
  updatePublicTypeTestCalls(*RegularLTO.CombinedModule,
                            Conf.HasWholeProgramVisibility);

  if (Conf.PreOptModuleHook &&
      !Conf.PreOptModuleHook(0, *RegularLTO.CombinedModule))
    return finalizeOptimizationRemarks(std::move(DiagnosticOutputFile));

  if (!Conf.CodeGenOnly) {
    for (const auto &R : GlobalResolutions) {
      GlobalValue *GV =
          RegularLTO.CombinedModule->getNamedValue(R.second.IRName);
      if (!R.second.isPrevailingIRSymbol())
        continue;
      if (R.second.Partition != 0 &&
          R.second.Partition != GlobalResolution::External)
        continue;

      // Ignore symbols defined in other partitions.
      // Also skip declarations, which are not allowed to have internal linkage.
      if (!GV || GV->hasLocalLinkage() || GV->isDeclaration())
        continue;

      // Symbols that are marked DLLImport or DLLExport should not be
      // internalized, as they are either externally visible or referencing
      // external symbols. Symbols that have AvailableExternally or Appending
      // linkage might be used by future passes and should be kept as is.
      // These linkages are seen in Unified regular LTO, because the process
      // of creating split LTO units introduces symbols with that linkage into
      // one of the created modules. Normally, only the ThinLTO backend would
      // compile this module, but Unified Regular LTO processes both
      // modules created by the splitting process as regular LTO modules.
      if ((LTOMode == LTOKind::LTOK_UnifiedRegular) &&
          ((GV->getDLLStorageClass() != GlobalValue::DefaultStorageClass) ||
           GV->hasAvailableExternallyLinkage() || GV->hasAppendingLinkage()))
        continue;

      GV->setUnnamedAddr(R.second.UnnamedAddr ? GlobalValue::UnnamedAddr::Global
                                              : GlobalValue::UnnamedAddr::None);
      if (EnableLTOInternalization && R.second.Partition == 0)
        GV->setLinkage(GlobalValue::InternalLinkage);
    }

    if (Conf.PostInternalizeModuleHook &&
        !Conf.PostInternalizeModuleHook(0, *RegularLTO.CombinedModule))
      return finalizeOptimizationRemarks(std::move(DiagnosticOutputFile));
  }

  if (!RegularLTO.EmptyCombinedModule || Conf.AlwaysEmitRegularLTOObj) {
    if (Error Err =
            backend(Conf, AddStream, RegularLTO.ParallelCodeGenParallelismLevel,
                    *RegularLTO.CombinedModule, ThinLTO.CombinedIndex))
      return Err;
  }

  return finalizeOptimizationRemarks(std::move(DiagnosticOutputFile));
}

static const char *libcallRoutineNames[] = {
#define HANDLE_LIBCALL(code, name) name,
#include "llvm/IR/RuntimeLibcalls.def"
#undef HANDLE_LIBCALL
};

ArrayRef<const char*> LTO::getRuntimeLibcallSymbols() {
  return ArrayRef(libcallRoutineNames);
}

/// This class defines the interface to the ThinLTO backend.
class lto::ThinBackendProc {
protected:
  const Config &Conf;
  ModuleSummaryIndex &CombinedIndex;
  const StringMap<GVSummaryMapTy> &ModuleToDefinedGVSummaries;
  lto::IndexWriteCallback OnWrite;
  bool ShouldEmitImportsFiles;

public:
  ThinBackendProc(const Config &Conf, ModuleSummaryIndex &CombinedIndex,
                  const StringMap<GVSummaryMapTy> &ModuleToDefinedGVSummaries,
                  lto::IndexWriteCallback OnWrite, bool ShouldEmitImportsFiles)
      : Conf(Conf), CombinedIndex(CombinedIndex),
        ModuleToDefinedGVSummaries(ModuleToDefinedGVSummaries),
        OnWrite(OnWrite), ShouldEmitImportsFiles(ShouldEmitImportsFiles) {}

  virtual ~ThinBackendProc() = default;
  virtual Error start(
      unsigned Task, BitcodeModule BM,
      const FunctionImporter::ImportMapTy &ImportList,
      const FunctionImporter::ExportSetTy &ExportList,
      const std::map<GlobalValue::GUID, GlobalValue::LinkageTypes> &ResolvedODR,
      MapVector<StringRef, BitcodeModule> &ModuleMap) = 0;
  virtual Error wait() = 0;
  virtual unsigned getThreadCount() = 0;

  // Write sharded indices and (optionally) imports to disk
  Error emitFiles(const FunctionImporter::ImportMapTy &ImportList,
                  llvm::StringRef ModulePath,
                  const std::string &NewModulePath) {
    std::map<std::string, GVSummaryMapTy> ModuleToSummariesForIndex;
    std::error_code EC;
    gatherImportedSummariesForModule(ModulePath, ModuleToDefinedGVSummaries,
                                     ImportList, ModuleToSummariesForIndex);

    raw_fd_ostream OS(NewModulePath + ".thinlto.bc", EC,
                      sys::fs::OpenFlags::OF_None);
    if (EC)
      return errorCodeToError(EC);
    writeIndexToFile(CombinedIndex, OS, &ModuleToSummariesForIndex);

    if (ShouldEmitImportsFiles) {
      EC = EmitImportsFiles(ModulePath, NewModulePath + ".imports",
                            ModuleToSummariesForIndex);
      if (EC)
        return errorCodeToError(EC);
    }
    return Error::success();
  }
};

namespace {
class InProcessThinBackend : public ThinBackendProc {
  ThreadPool BackendThreadPool;
  AddStreamFn AddStream;
  FileCache Cache;
  std::set<GlobalValue::GUID> CfiFunctionDefs;
  std::set<GlobalValue::GUID> CfiFunctionDecls;

  std::optional<Error> Err;
  std::mutex ErrMu;

  bool ShouldEmitIndexFiles;

public:
  InProcessThinBackend(
      const Config &Conf, ModuleSummaryIndex &CombinedIndex,
      ThreadPoolStrategy ThinLTOParallelism,
      const StringMap<GVSummaryMapTy> &ModuleToDefinedGVSummaries,
      AddStreamFn AddStream, FileCache Cache, lto::IndexWriteCallback OnWrite,
      bool ShouldEmitIndexFiles, bool ShouldEmitImportsFiles)
      : ThinBackendProc(Conf, CombinedIndex, ModuleToDefinedGVSummaries,
                        OnWrite, ShouldEmitImportsFiles),
        BackendThreadPool(ThinLTOParallelism), AddStream(std::move(AddStream)),
        Cache(std::move(Cache)), ShouldEmitIndexFiles(ShouldEmitIndexFiles) {
    for (auto &Name : CombinedIndex.cfiFunctionDefs())
      CfiFunctionDefs.insert(
          GlobalValue::getGUID(GlobalValue::dropLLVMManglingEscape(Name)));
    for (auto &Name : CombinedIndex.cfiFunctionDecls())
      CfiFunctionDecls.insert(
          GlobalValue::getGUID(GlobalValue::dropLLVMManglingEscape(Name)));
  }

  Error runThinLTOBackendThread(
      AddStreamFn AddStream, FileCache Cache, unsigned Task, BitcodeModule BM,
      ModuleSummaryIndex &CombinedIndex,
      const FunctionImporter::ImportMapTy &ImportList,
      const FunctionImporter::ExportSetTy &ExportList,
      const std::map<GlobalValue::GUID, GlobalValue::LinkageTypes> &ResolvedODR,
      const GVSummaryMapTy &DefinedGlobals,
      MapVector<StringRef, BitcodeModule> &ModuleMap) {
    auto RunThinBackend = [&](AddStreamFn AddStream) {
      LTOLLVMContext BackendContext(Conf);
      Expected<std::unique_ptr<Module>> MOrErr = BM.parseModule(BackendContext);
      if (!MOrErr)
        return MOrErr.takeError();

      return thinBackend(Conf, Task, AddStream, **MOrErr, CombinedIndex,
                         ImportList, DefinedGlobals, &ModuleMap);
    };

    auto ModuleID = BM.getModuleIdentifier();

    if (ShouldEmitIndexFiles) {
      if (auto E = emitFiles(ImportList, ModuleID, ModuleID.str()))
        return E;
    }

    if (!Cache || !CombinedIndex.modulePaths().count(ModuleID) ||
        all_of(CombinedIndex.getModuleHash(ModuleID),
               [](uint32_t V) { return V == 0; }))
      // Cache disabled or no entry for this module in the combined index or
      // no module hash.
      return RunThinBackend(AddStream);

    SmallString<40> Key;
    // The module may be cached, this helps handling it.
    computeLTOCacheKey(Key, Conf, CombinedIndex, ModuleID, ImportList,
                       ExportList, ResolvedODR, DefinedGlobals, CfiFunctionDefs,
                       CfiFunctionDecls);
    Expected<AddStreamFn> CacheAddStreamOrErr = Cache(Task, Key, ModuleID);
    if (Error Err = CacheAddStreamOrErr.takeError())
      return Err;
    AddStreamFn &CacheAddStream = *CacheAddStreamOrErr;
    if (CacheAddStream)
      return RunThinBackend(CacheAddStream);

    return Error::success();
  }

  Error start(
      unsigned Task, BitcodeModule BM,
      const FunctionImporter::ImportMapTy &ImportList,
      const FunctionImporter::ExportSetTy &ExportList,
      const std::map<GlobalValue::GUID, GlobalValue::LinkageTypes> &ResolvedODR,
      MapVector<StringRef, BitcodeModule> &ModuleMap) override {
    StringRef ModulePath = BM.getModuleIdentifier();
    assert(ModuleToDefinedGVSummaries.count(ModulePath));
    const GVSummaryMapTy &DefinedGlobals =
        ModuleToDefinedGVSummaries.find(ModulePath)->second;
    BackendThreadPool.async(
        [=](BitcodeModule BM, ModuleSummaryIndex &CombinedIndex,
            const FunctionImporter::ImportMapTy &ImportList,
            const FunctionImporter::ExportSetTy &ExportList,
            const std::map<GlobalValue::GUID, GlobalValue::LinkageTypes>
                &ResolvedODR,
            const GVSummaryMapTy &DefinedGlobals,
            MapVector<StringRef, BitcodeModule> &ModuleMap) {
          if (LLVM_ENABLE_THREADS && Conf.TimeTraceEnabled)
            timeTraceProfilerInitialize(Conf.TimeTraceGranularity,
                                        "thin backend");
          Error E = runThinLTOBackendThread(
              AddStream, Cache, Task, BM, CombinedIndex, ImportList, ExportList,
              ResolvedODR, DefinedGlobals, ModuleMap);
          if (E) {
            std::unique_lock<std::mutex> L(ErrMu);
            if (Err)
              Err = joinErrors(std::move(*Err), std::move(E));
            else
              Err = std::move(E);
          }
          if (LLVM_ENABLE_THREADS && Conf.TimeTraceEnabled)
            timeTraceProfilerFinishThread();
        },
        BM, std::ref(CombinedIndex), std::ref(ImportList), std::ref(ExportList),
        std::ref(ResolvedODR), std::ref(DefinedGlobals), std::ref(ModuleMap));

    if (OnWrite)
      OnWrite(std::string(ModulePath));
    return Error::success();
  }

  Error wait() override {
    BackendThreadPool.wait();
    if (Err)
      return std::move(*Err);
    else
      return Error::success();
  }

  unsigned getThreadCount() override {
    return BackendThreadPool.getThreadCount();
  }
};
} // end anonymous namespace

ThinBackend lto::createInProcessThinBackend(ThreadPoolStrategy Parallelism,
                                            lto::IndexWriteCallback OnWrite,
                                            bool ShouldEmitIndexFiles,
                                            bool ShouldEmitImportsFiles) {
  return [=](const Config &Conf, ModuleSummaryIndex &CombinedIndex,
             const StringMap<GVSummaryMapTy> &ModuleToDefinedGVSummaries,
             AddStreamFn AddStream, FileCache Cache) {
    return std::make_unique<InProcessThinBackend>(
        Conf, CombinedIndex, Parallelism, ModuleToDefinedGVSummaries, AddStream,
        Cache, OnWrite, ShouldEmitIndexFiles, ShouldEmitImportsFiles);
  };
}

// Given the original \p Path to an output file, replace any path
// prefix matching \p OldPrefix with \p NewPrefix. Also, create the
// resulting directory if it does not yet exist.
std::string lto::getThinLTOOutputFile(StringRef Path, StringRef OldPrefix,
                                      StringRef NewPrefix) {
  if (OldPrefix.empty() && NewPrefix.empty())
    return std::string(Path);
  SmallString<128> NewPath(Path);
  llvm::sys::path::replace_path_prefix(NewPath, OldPrefix, NewPrefix);
  StringRef ParentPath = llvm::sys::path::parent_path(NewPath.str());
  if (!ParentPath.empty()) {
    // Make sure the new directory exists, creating it if necessary.
    if (std::error_code EC = llvm::sys::fs::create_directories(ParentPath))
      llvm::errs() << "warning: could not create directory '" << ParentPath
                   << "': " << EC.message() << '\n';
  }
  return std::string(NewPath.str());
}

namespace {
class WriteIndexesThinBackend : public ThinBackendProc {
  std::string OldPrefix, NewPrefix, NativeObjectPrefix;
  raw_fd_ostream *LinkedObjectsFile;

public:
  WriteIndexesThinBackend(
      const Config &Conf, ModuleSummaryIndex &CombinedIndex,
      const StringMap<GVSummaryMapTy> &ModuleToDefinedGVSummaries,
      std::string OldPrefix, std::string NewPrefix,
      std::string NativeObjectPrefix, bool ShouldEmitImportsFiles,
      raw_fd_ostream *LinkedObjectsFile, lto::IndexWriteCallback OnWrite)
      : ThinBackendProc(Conf, CombinedIndex, ModuleToDefinedGVSummaries,
                        OnWrite, ShouldEmitImportsFiles),
        OldPrefix(OldPrefix), NewPrefix(NewPrefix),
        NativeObjectPrefix(NativeObjectPrefix),
        LinkedObjectsFile(LinkedObjectsFile) {}

  Error start(
      unsigned Task, BitcodeModule BM,
      const FunctionImporter::ImportMapTy &ImportList,
      const FunctionImporter::ExportSetTy &ExportList,
      const std::map<GlobalValue::GUID, GlobalValue::LinkageTypes> &ResolvedODR,
      MapVector<StringRef, BitcodeModule> &ModuleMap) override {
    StringRef ModulePath = BM.getModuleIdentifier();
    std::string NewModulePath =
        getThinLTOOutputFile(ModulePath, OldPrefix, NewPrefix);

    if (LinkedObjectsFile) {
      std::string ObjectPrefix =
          NativeObjectPrefix.empty() ? NewPrefix : NativeObjectPrefix;
      std::string LinkedObjectsFilePath =
          getThinLTOOutputFile(ModulePath, OldPrefix, ObjectPrefix);
      *LinkedObjectsFile << LinkedObjectsFilePath << '\n';
    }

    if (auto E = emitFiles(ImportList, ModulePath, NewModulePath))
      return E;

    if (OnWrite)
      OnWrite(std::string(ModulePath));
    return Error::success();
  }

  Error wait() override { return Error::success(); }

  // WriteIndexesThinBackend should always return 1 to prevent module
  // re-ordering and avoid non-determinism in the final link.
  unsigned getThreadCount() override { return 1; }
};
} // end anonymous namespace

ThinBackend lto::createWriteIndexesThinBackend(
    std::string OldPrefix, std::string NewPrefix,
    std::string NativeObjectPrefix, bool ShouldEmitImportsFiles,
    raw_fd_ostream *LinkedObjectsFile, IndexWriteCallback OnWrite) {
  return [=](const Config &Conf, ModuleSummaryIndex &CombinedIndex,
             const StringMap<GVSummaryMapTy> &ModuleToDefinedGVSummaries,
             AddStreamFn AddStream, FileCache Cache) {
    return std::make_unique<WriteIndexesThinBackend>(
        Conf, CombinedIndex, ModuleToDefinedGVSummaries, OldPrefix, NewPrefix,
        NativeObjectPrefix, ShouldEmitImportsFiles, LinkedObjectsFile, OnWrite);
  };
}

Error LTO::runThinLTO(AddStreamFn AddStream, FileCache Cache,
                      const DenseSet<GlobalValue::GUID> &GUIDPreservedSymbols) {
  LLVM_DEBUG(dbgs() << "Running ThinLTO\n");
  ThinLTO.CombinedIndex.releaseTemporaryMemory();
  timeTraceProfilerBegin("ThinLink", StringRef(""));
  auto TimeTraceScopeExit = llvm::make_scope_exit([]() {
    if (llvm::timeTraceProfilerEnabled())
      llvm::timeTraceProfilerEnd();
  });
  if (ThinLTO.ModuleMap.empty())
    return Error::success();

  if (ThinLTO.ModulesToCompile && ThinLTO.ModulesToCompile->empty()) {
    llvm::errs() << "warning: [ThinLTO] No module compiled\n";
    return Error::success();
  }

  if (Conf.CombinedIndexHook &&
      !Conf.CombinedIndexHook(ThinLTO.CombinedIndex, GUIDPreservedSymbols))
    return Error::success();

  // Collect for each module the list of function it defines (GUID ->
  // Summary).
  StringMap<GVSummaryMapTy>
      ModuleToDefinedGVSummaries(ThinLTO.ModuleMap.size());
  ThinLTO.CombinedIndex.collectDefinedGVSummariesPerModule(
      ModuleToDefinedGVSummaries);
  // Create entries for any modules that didn't have any GV summaries
  // (either they didn't have any GVs to start with, or we suppressed
  // generation of the summaries because they e.g. had inline assembly
  // uses that couldn't be promoted/renamed on export). This is so
  // InProcessThinBackend::start can still launch a backend thread, which
  // is passed the map of summaries for the module, without any special
  // handling for this case.
  for (auto &Mod : ThinLTO.ModuleMap)
    if (!ModuleToDefinedGVSummaries.count(Mod.first))
      ModuleToDefinedGVSummaries.try_emplace(Mod.first);

  // Synthesize entry counts for functions in the CombinedIndex.
  computeSyntheticCounts(ThinLTO.CombinedIndex);

  StringMap<FunctionImporter::ImportMapTy> ImportLists(
      ThinLTO.ModuleMap.size());
  StringMap<FunctionImporter::ExportSetTy> ExportLists(
      ThinLTO.ModuleMap.size());
  StringMap<std::map<GlobalValue::GUID, GlobalValue::LinkageTypes>> ResolvedODR;

  if (DumpThinCGSCCs)
    ThinLTO.CombinedIndex.dumpSCCs(outs());

  std::set<GlobalValue::GUID> ExportedGUIDs;

  if (hasWholeProgramVisibility(Conf.HasWholeProgramVisibility))
    ThinLTO.CombinedIndex.setWithWholeProgramVisibility();
  // If allowed, upgrade public vcall visibility to linkage unit visibility in
  // the summaries before whole program devirtualization below.
  updateVCallVisibilityInIndex(ThinLTO.CombinedIndex,
                               Conf.HasWholeProgramVisibility,
                               DynamicExportSymbols);

  // Perform index-based WPD. This will return immediately if there are
  // no index entries in the typeIdMetadata map (e.g. if we are instead
  // performing IR-based WPD in hybrid regular/thin LTO mode).
  std::map<ValueInfo, std::vector<VTableSlotSummary>> LocalWPDTargetsMap;
  runWholeProgramDevirtOnIndex(ThinLTO.CombinedIndex, ExportedGUIDs,
                               LocalWPDTargetsMap);

  auto isPrevailing = [&](GlobalValue::GUID GUID, const GlobalValueSummary *S) {
    return ThinLTO.PrevailingModuleForGUID[GUID] == S->modulePath();
  };
  if (EnableMemProfContextDisambiguation) {
    MemProfContextDisambiguation ContextDisambiguation;
    ContextDisambiguation.run(ThinLTO.CombinedIndex, isPrevailing);
  }

  if (Conf.OptLevel > 0)
    ComputeCrossModuleImport(ThinLTO.CombinedIndex, ModuleToDefinedGVSummaries,
                             isPrevailing, ImportLists, ExportLists);

  // Figure out which symbols need to be internalized. This also needs to happen
  // at -O0 because summary-based DCE is implemented using internalization, and
  // we must apply DCE consistently with the full LTO module in order to avoid
  // undefined references during the final link.
  for (auto &Res : GlobalResolutions) {
    // If the symbol does not have external references or it is not prevailing,
    // then not need to mark it as exported from a ThinLTO partition.
    if (Res.second.Partition != GlobalResolution::External ||
        !Res.second.isPrevailingIRSymbol())
      continue;
    auto GUID = GlobalValue::getGUID(
        GlobalValue::dropLLVMManglingEscape(Res.second.IRName));
    // Mark exported unless index-based analysis determined it to be dead.
    if (ThinLTO.CombinedIndex.isGUIDLive(GUID))
      ExportedGUIDs.insert(GUID);
  }

  // Any functions referenced by the jump table in the regular LTO object must
  // be exported.
  for (auto &Def : ThinLTO.CombinedIndex.cfiFunctionDefs())
    ExportedGUIDs.insert(
        GlobalValue::getGUID(GlobalValue::dropLLVMManglingEscape(Def)));
  for (auto &Decl : ThinLTO.CombinedIndex.cfiFunctionDecls())
    ExportedGUIDs.insert(
        GlobalValue::getGUID(GlobalValue::dropLLVMManglingEscape(Decl)));

  auto isExported = [&](StringRef ModuleIdentifier, ValueInfo VI) {
    const auto &ExportList = ExportLists.find(ModuleIdentifier);
    return (ExportList != ExportLists.end() && ExportList->second.count(VI)) ||
           ExportedGUIDs.count(VI.getGUID());
  };

  // Update local devirtualized targets that were exported by cross-module
  // importing or by other devirtualizations marked in the ExportedGUIDs set.
  updateIndexWPDForExports(ThinLTO.CombinedIndex, isExported,
                           LocalWPDTargetsMap);

  thinLTOInternalizeAndPromoteInIndex(ThinLTO.CombinedIndex, isExported,
                                      isPrevailing);

  auto recordNewLinkage = [&](StringRef ModuleIdentifier,
                              GlobalValue::GUID GUID,
                              GlobalValue::LinkageTypes NewLinkage) {
    ResolvedODR[ModuleIdentifier][GUID] = NewLinkage;
  };
  thinLTOResolvePrevailingInIndex(Conf, ThinLTO.CombinedIndex, isPrevailing,
                                  recordNewLinkage, GUIDPreservedSymbols);

  thinLTOPropagateFunctionAttrs(ThinLTO.CombinedIndex, isPrevailing);

  generateParamAccessSummary(ThinLTO.CombinedIndex);

  if (llvm::timeTraceProfilerEnabled())
    llvm::timeTraceProfilerEnd();

  TimeTraceScopeExit.release();

  std::unique_ptr<ThinBackendProc> BackendProc =
      ThinLTO.Backend(Conf, ThinLTO.CombinedIndex, ModuleToDefinedGVSummaries,
                      AddStream, Cache);

  auto &ModuleMap =
      ThinLTO.ModulesToCompile ? *ThinLTO.ModulesToCompile : ThinLTO.ModuleMap;

  auto ProcessOneModule = [&](int I) -> Error {
    auto &Mod = *(ModuleMap.begin() + I);
    // Tasks 0 through ParallelCodeGenParallelismLevel-1 are reserved for
    // combined module and parallel code generation partitions.
    return BackendProc->start(RegularLTO.ParallelCodeGenParallelismLevel + I,
                              Mod.second, ImportLists[Mod.first],
                              ExportLists[Mod.first], ResolvedODR[Mod.first],
                              ThinLTO.ModuleMap);
  };

  if (BackendProc->getThreadCount() == 1) {
    // Process the modules in the order they were provided on the command-line.
    // It is important for this codepath to be used for WriteIndexesThinBackend,
    // to ensure the emitted LinkedObjectsFile lists ThinLTO objects in the same
    // order as the inputs, which otherwise would affect the final link order.
    for (int I = 0, E = ModuleMap.size(); I != E; ++I)
      if (Error E = ProcessOneModule(I))
        return E;
  } else {
    // When executing in parallel, process largest bitsize modules first to
    // improve parallelism, and avoid starving the thread pool near the end.
    // This saves about 15 sec on a 36-core machine while link `clang.exe` (out
    // of 100 sec).
    std::vector<BitcodeModule *> ModulesVec;
    ModulesVec.reserve(ModuleMap.size());
    for (auto &Mod : ModuleMap)
      ModulesVec.push_back(&Mod.second);
    for (int I : generateModulesOrdering(ModulesVec))
      if (Error E = ProcessOneModule(I))
        return E;
  }
  return BackendProc->wait();
}

Expected<std::unique_ptr<ToolOutputFile>> lto::setupLLVMOptimizationRemarks(
    LLVMContext &Context, StringRef RemarksFilename, StringRef RemarksPasses,
    StringRef RemarksFormat, bool RemarksWithHotness,
    std::optional<uint64_t> RemarksHotnessThreshold, int Count) {
  std::string Filename = std::string(RemarksFilename);
  // For ThinLTO, file.opt.<format> becomes
  // file.opt.<format>.thin.<num>.<format>.
  if (!Filename.empty() && Count != -1)
    Filename =
        (Twine(Filename) + ".thin." + llvm::utostr(Count) + "." + RemarksFormat)
            .str();

  auto ResultOrErr = llvm::setupLLVMOptimizationRemarks(
      Context, Filename, RemarksPasses, RemarksFormat, RemarksWithHotness,
      RemarksHotnessThreshold);
  if (Error E = ResultOrErr.takeError())
    return std::move(E);

  if (*ResultOrErr)
    (*ResultOrErr)->keep();

  return ResultOrErr;
}

Expected<std::unique_ptr<ToolOutputFile>>
lto::setupStatsFile(StringRef StatsFilename) {
  // Setup output file to emit statistics.
  if (StatsFilename.empty())
    return nullptr;

  llvm::EnableStatistics(false);
  std::error_code EC;
  auto StatsFile =
      std::make_unique<ToolOutputFile>(StatsFilename, EC, sys::fs::OF_None);
  if (EC)
    return errorCodeToError(EC);

  StatsFile->keep();
  return std::move(StatsFile);
}

// Compute the ordering we will process the inputs: the rough heuristic here
// is to sort them per size so that the largest module get schedule as soon as
// possible. This is purely a compile-time optimization.
std::vector<int> lto::generateModulesOrdering(ArrayRef<BitcodeModule *> R) {
  auto Seq = llvm::seq<int>(0, R.size());
  std::vector<int> ModulesOrdering(Seq.begin(), Seq.end());
  llvm::sort(ModulesOrdering, [&](int LeftIndex, int RightIndex) {
    auto LSize = R[LeftIndex]->getBuffer().size();
    auto RSize = R[RightIndex]->getBuffer().size();
    return LSize > RSize;
  });
  return ModulesOrdering;
}