vendor/lld/lld-trunk-r233088

26 files changed, 4004 insertions, 0 deletions

diff --git a/include/lld/Core/AbsoluteAtom.h b/include/lld/Core/AbsoluteAtom.h
new file mode 100644
index 000000000000..ed25297cea81
--- /dev/null
+++ b/include/lld/Core/AbsoluteAtom.h
@@ -0,0 +1,43 @@
+//===- Core/AbsoluteAtom.h - An absolute Atom -----------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_ABSOLUTE_ATOM_H
+#define LLD_CORE_ABSOLUTE_ATOM_H
+
+#include "lld/Core/Atom.h"
+
+namespace lld {
+
+/// An AbsoluteAtom has no content.
+/// It exists to represent content at fixed addresses in memory.
+class AbsoluteAtom : public Atom {
+public:
+
+  virtual uint64_t value() const = 0;
+
+  /// scope - The visibility of this atom to other atoms.  C static functions
+  /// have scope scopeTranslationUnit.  Regular C functions have scope
+  /// scopeGlobal.  Functions compiled with visibility=hidden have scope
+  /// scopeLinkageUnit so they can be see by other atoms being linked but not
+  /// by the OS loader.
+  virtual Scope scope() const = 0;
+
+  static bool classof(const Atom *a) {
+    return a->definition() == definitionAbsolute;
+  }
+
+  static bool classof(const AbsoluteAtom *) { return true; }
+
+protected:
+  AbsoluteAtom() : Atom(definitionAbsolute) {}
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_ABSOLUTE_ATOM_H
diff --git a/include/lld/Core/Alias.h b/include/lld/Core/Alias.h
new file mode 100644
index 000000000000..610022525ecb
--- /dev/null
+++ b/include/lld/Core/Alias.h
@@ -0,0 +1,102 @@
+//===- lld/Core/Alias.h - Alias atoms -------------------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// \brief Provide alias atoms.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_ALIAS_H
+#define LLD_CORE_ALIAS_H
+
+#include "lld/Core/LLVM.h"
+#include "lld/Core/Simple.h"
+#include "llvm/ADT/Optional.h"
+#include <string>
+
+namespace lld {
+
+// An AliasAtom is a zero-size atom representing an alias for other atom. It has
+// a LayoutAfter reference to the target atom, so that this atom and the target
+// atom will be laid out at the same location in the final result. Initially
+// the target atom is an undefined atom. Resolver will replace it with a defined
+// one.
+//
+// It does not have attributes itself. Most member function calls are forwarded
+// to the target atom.
+class AliasAtom : public SimpleDefinedAtom {
+public:
+  AliasAtom(const File &file, StringRef name)
+      : SimpleDefinedAtom(file), _target(nullptr), _name(name),
+        _merge(DefinedAtom::mergeNo), _deadStrip(DefinedAtom::deadStripNormal) {
+  }
+
+  StringRef name() const override { return _name; }
+  uint64_t size() const override { return 0; }
+  ArrayRef<uint8_t> rawContent() const override { return ArrayRef<uint8_t>(); }
+
+  Scope scope() const override {
+    getTarget();
+    return _target ? _target->scope() : scopeLinkageUnit;
+  }
+
+  Merge merge() const override {
+    if (_merge.hasValue())
+      return _merge.getValue();
+    getTarget();
+    return _target ? _target->merge() : mergeNo;
+  }
+
+  void setMerge(Merge val) { _merge = val; }
+
+  ContentType contentType() const override {
+    getTarget();
+    return _target ? _target->contentType() : typeUnknown;
+  }
+
+  Interposable interposable() const override {
+    getTarget();
+    return _target ? _target->interposable() : interposeNo;
+  }
+
+  SectionChoice sectionChoice() const override {
+    getTarget();
+    return _target ? _target->sectionChoice() : sectionBasedOnContent;
+  }
+
+  StringRef customSectionName() const override {
+    getTarget();
+    return _target ? _target->customSectionName() : StringRef("");
+  }
+
+  DeadStripKind deadStrip() const override { return _deadStrip; }
+  void setDeadStrip(DeadStripKind val) { _deadStrip = val; }
+
+private:
+  void getTarget() const {
+    if (_target)
+      return;
+    for (const Reference *r : *this) {
+      if (r->kindNamespace() == lld::Reference::KindNamespace::all &&
+          r->kindValue() == lld::Reference::kindLayoutAfter) {
+        _target = dyn_cast<DefinedAtom>(r->target());
+        return;
+      }
+    }
+  }
+
+  mutable const DefinedAtom *_target;
+  std::string _name;
+  llvm::Optional<Merge> _merge;
+  DeadStripKind _deadStrip;
+};
+
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/ArchiveLibraryFile.h b/include/lld/Core/ArchiveLibraryFile.h
new file mode 100644
index 000000000000..ff379d4f3ecf
--- /dev/null
+++ b/include/lld/Core/ArchiveLibraryFile.h
@@ -0,0 +1,60 @@
+//===- Core/ArchiveLibraryFile.h - Models static library ------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_ARCHIVE_LIBRARY_FILE_H
+#define LLD_CORE_ARCHIVE_LIBRARY_FILE_H
+
+#include "lld/Core/File.h"
+#include "lld/Core/Parallel.h"
+#include <set>
+
+namespace lld {
+
+///
+/// The ArchiveLibraryFile subclass of File is used to represent unix
+/// static library archives.  These libraries provide no atoms to the
+/// initial set of atoms linked.  Instead, when the Resolver will query
+/// ArchiveLibraryFile instances for specific symbols names using the
+/// find() method.  If the archive contains an object file which has a
+/// DefinedAtom whose scope is not translationUnit, then that entire
+/// object file File is returned.
+///
+class ArchiveLibraryFile : public File {
+public:
+  static bool classof(const File *f) {
+    return f->kind() == kindArchiveLibrary;
+  }
+
+  /// Check if any member of the archive contains an Atom with the
+  /// specified name and return the File object for that member, or nullptr.
+  virtual File *find(StringRef name, bool dataSymbolOnly) = 0;
+
+  virtual std::error_code
+  parseAllMembers(std::vector<std::unique_ptr<File>> &result) = 0;
+
+  // Parses a member file containing a given symbol, so that when you
+  // need the file find() can return that immediately. Calling this function
+  // has no side effect other than pre-instantiating a file. Calling this
+  // function doesn't affect correctness.
+  virtual void preload(TaskGroup &group, StringRef symbolName) {}
+
+  /// Returns a set of all defined symbols in the archive, i.e. all
+  /// resolvable symbol using this file.
+  virtual std::set<StringRef> getDefinedSymbols() {
+    return std::set<StringRef>();
+  }
+
+protected:
+  /// only subclasses of ArchiveLibraryFile can be instantiated
+  ArchiveLibraryFile(StringRef path) : File(path, kindArchiveLibrary) {}
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_ARCHIVE_LIBRARY_FILE_H
diff --git a/include/lld/Core/Atom.h b/include/lld/Core/Atom.h
new file mode 100644
index 000000000000..27fdde022ba7
--- /dev/null
+++ b/include/lld/Core/Atom.h
@@ -0,0 +1,76 @@
+//===- Core/Atom.h - A node in linking graph ------------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_ATOM_H
+#define LLD_CORE_ATOM_H
+
+#include "lld/Core/LLVM.h"
+
+namespace lld {
+
+class File;
+
+///
+/// The linker has a Graph Theory model of linking. An object file is seen
+/// as a set of Atoms with References to other Atoms.  Each Atom is a node
+/// and each Reference is an edge. An Atom can be a DefinedAtom which has
+/// content or a UndefinedAtom which is a placeholder and represents an
+/// undefined symbol (extern declaration).
+///
+class Atom {
+public:
+  /// Whether this atom is defined or a proxy for an undefined symbol
+  enum Definition {
+    definitionRegular,      ///< Normal C/C++ function or global variable.
+    definitionAbsolute,     ///< Asm-only (foo = 10). Not tied to any content.
+    definitionUndefined,    ///< Only in .o files to model reference to undef.
+    definitionSharedLibrary ///< Only in shared libraries to model export.
+  };
+
+  /// The scope in which this atom is acessible to other atoms.
+  enum Scope {
+    scopeTranslationUnit,  ///< Accessible only to atoms in the same translation
+                           ///  unit (e.g. a C static).
+    scopeLinkageUnit,      ///< Accessible to atoms being linked but not visible
+                           ///  to runtime loader (e.g. visibility=hidden).
+    scopeGlobal            ///< Accessible to all atoms and visible to runtime
+                           ///  loader (e.g. visibility=default).
+  };
+
+
+  /// file - returns the File that produced/owns this Atom
+  virtual const File& file() const = 0;
+
+  /// name - The name of the atom. For a function atom, it is the (mangled)
+  /// name of the function.
+  virtual StringRef name() const = 0;
+
+  /// definition - Whether this atom is a definition or represents an undefined
+  /// symbol.
+  Definition definition() const { return _definition; }
+
+  static bool classof(const Atom *a) { return true; }
+
+protected:
+  /// Atom is an abstract base class.  Only subclasses can access constructor.
+  explicit Atom(Definition def) : _definition(def) {}
+
+  /// The memory for Atom objects is always managed by the owning File
+  /// object.  Therefore, no one but the owning File object should call
+  /// delete on an Atom.  In fact, some File objects may bulk allocate
+  /// an array of Atoms, so they cannot be individually deleted by anyone.
+  virtual ~Atom() {}
+
+private:
+  Definition _definition;
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_ATOM_H
diff --git a/include/lld/Core/DefinedAtom.h b/include/lld/Core/DefinedAtom.h
new file mode 100644
index 000000000000..86d880c659b4
--- /dev/null
+++ b/include/lld/Core/DefinedAtom.h
@@ -0,0 +1,368 @@
+//===- Core/DefinedAtom.h - An Atom with content --------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_DEFINED_ATOM_H
+#define LLD_CORE_DEFINED_ATOM_H
+
+#include "lld/Core/Atom.h"
+#include "lld/Core/LLVM.h"
+
+namespace lld {
+class File;
+class Reference;
+
+/// \brief The fundamental unit of linking.
+///
+/// A C function or global variable is an atom.  An atom has content and
+/// attributes. The content of a function atom is the instructions that
+/// implement the function.  The content of a global variable atom is its
+/// initial bytes.
+///
+/// Here are some example attribute sets for common atoms. If a particular
+/// attribute is not listed, the default values are:  definition=regular,
+/// sectionChoice=basedOnContent, scope=translationUnit, merge=no,
+/// deadStrip=normal, interposable=no
+///
+///  C function:  void foo() {} <br>
+///    name=foo, type=code, perm=r_x, scope=global
+///
+///  C static function:  staic void func() {} <br>
+///    name=func, type=code, perm=r_x
+///
+///  C global variable:  int count = 1; <br>
+///    name=count, type=data, perm=rw_, scope=global
+///
+///  C tentative definition:  int bar; <br>
+///    name=bar, type=zerofill, perm=rw_, scope=global,
+///    merge=asTentative, interposable=yesAndRuntimeWeak
+///
+///  Uninitialized C static variable:  static int stuff; <br>
+///    name=stuff, type=zerofill, perm=rw_
+///
+///  Weak C function:  __attribute__((weak)) void foo() {} <br>
+///    name=foo, type=code, perm=r_x, scope=global, merge=asWeak
+///
+///  Hidden C function:  __attribute__((visibility("hidden"))) void foo() {}<br>
+///    name=foo, type=code, perm=r_x, scope=linkageUnit
+///
+///  No-dead-strip function:  __attribute__((used)) void foo() {} <br>
+///    name=foo, type=code, perm=r_x, scope=global, deadStrip=never
+///
+///  Non-inlined C++ inline method:  inline void Foo::doit() {} <br>
+///    name=_ZN3Foo4doitEv, type=code, perm=r_x, scope=global,
+///    mergeDupes=asWeak
+///
+///  Non-inlined C++ inline method whose address is taken:
+///     inline void Foo::doit() {} <br>
+///    name=_ZN3Foo4doitEv, type=code, perm=r_x, scope=global,
+///    mergeDupes=asAddressedWeak
+///
+///  literal c-string:  "hello" <br>
+///    name="" type=cstring, perm=r__, scope=linkageUnit
+///
+///  literal double:  1.234 <br>
+///    name="" type=literal8, perm=r__, scope=linkageUnit
+///
+///  constant:  { 1,2,3 } <br>
+///    name="" type=constant, perm=r__, scope=linkageUnit
+///
+///  Pointer to initializer function:  <br>
+///    name="" type=initializer, perm=rw_l,
+///    sectionChoice=customRequired
+///
+///  C function place in custom section:  __attribute__((section("__foo")))
+///                                       void foo() {} <br>
+///    name=foo, type=code, perm=r_x, scope=global,
+///    sectionChoice=customRequired, customSectionName=__foo
+///
+class DefinedAtom : public Atom {
+public:
+  enum Interposable {
+    interposeNo,            // linker can directly bind uses of this atom
+    interposeYes,           // linker must indirect (through GOT) uses
+    interposeYesAndRuntimeWeak // must indirect and mark symbol weak in final
+                               // linked image
+  };
+
+  enum Merge {
+    mergeNo,                // Another atom with same name is error
+    mergeAsTentative,       // Is ANSI C tentative definition, can be coalesced
+    mergeAsWeak,            // Is C++ inline definition that was not inlined,
+                            // but address was not taken, so atom can be hidden
+                            // by linker
+    mergeAsWeakAndAddressUsed, // Is C++ definition inline definition whose
+                               // address was taken.
+    mergeSameNameAndSize,   // Another atom with different size is error
+    mergeByLargestSection,  // Choose an atom whose section is the largest.
+    mergeByContent,         // Merge with other constants with same content.
+  };
+
+  enum ContentType {
+    typeUnknown,            // for use with definitionUndefined
+    typeCode,               // executable code
+    typeResolver,           // function which returns address of target
+    typeBranchIsland,       // linker created for large binaries
+    typeBranchShim,         // linker created to switch thumb mode
+    typeStub,               // linker created for calling external function
+    typeStubHelper,         // linker created for initial stub binding
+    typeConstant,           // a read-only constant
+    typeCString,            // a zero terminated UTF8 C string
+    typeUTF16String,        // a zero terminated UTF16 string
+    typeCFI,                // a FDE or CIE from dwarf unwind info
+    typeLSDA,               // extra unwinding info
+    typeLiteral4,           // a four-btye read-only constant
+    typeLiteral8,           // an eight-btye read-only constant
+    typeLiteral16,          // a sixteen-btye read-only constant
+    typeData,               // read-write data
+    typeDataFast,           // allow data to be quickly accessed
+    typeZeroFill,           // zero-fill data
+    typeZeroFillFast,       // allow zero-fill data to be quicky accessed
+    typeConstData,          // read-only data after dynamic linker is done
+    typeObjC1Class,         // ObjC1 class [Darwin]
+    typeLazyPointer,        // pointer through which a stub jumps
+    typeLazyDylibPointer,   // pointer through which a stub jumps [Darwin]
+    typeCFString,           // NS/CFString object [Darwin]
+    typeGOT,                // pointer to external symbol
+    typeInitializerPtr,     // pointer to initializer function
+    typeTerminatorPtr,      // pointer to terminator function
+    typeCStringPtr,         // pointer to UTF8 C string [Darwin]
+    typeObjCClassPtr,       // pointer to ObjC class [Darwin]
+    typeObjC2CategoryList,  // pointers to ObjC category [Darwin]
+    typeDTraceDOF,          // runtime data for Dtrace [Darwin]
+    typeInterposingTuples,  // tuples of interposing info for dyld [Darwin]
+    typeTempLTO,            // temporary atom for bitcode reader
+    typeCompactUnwindInfo,  // runtime data for unwinder [Darwin]
+    typeProcessedUnwindInfo,// compressed compact unwind info [Darwin]
+    typeThunkTLV,           // thunk used to access a TLV [Darwin]
+    typeTLVInitialData,     // initial data for a TLV [Darwin]
+    typeTLVInitialZeroFill, // TLV initial zero fill data [Darwin]
+    typeTLVInitializerPtr,  // pointer to thread local initializer [Darwin]
+    typeMachHeader,         // atom representing mach_header [Darwin]
+    typeThreadZeroFill,     // Uninitialized thread local data(TBSS) [ELF]
+    typeThreadData,         // Initialized thread local data(TDATA) [ELF]
+    typeRONote,             // Identifies readonly note sections [ELF]
+    typeRWNote,             // Identifies readwrite note sections [ELF]
+    typeNoAlloc,            // Identifies non allocatable sections [ELF]
+    typeGroupComdat,        // Identifies a section group [ELF, COFF]
+    typeGnuLinkOnce,        // Identifies a gnu.linkonce section [ELF]
+  };
+
+  // Permission bits for atoms and segments. The order of these values are
+  // important, because the layout pass may sort atoms by permission if other
+  // attributes are the same.
+  enum ContentPermissions {
+    perm___  = 0,           // mapped as unaccessible
+    permR__  = 8,           // mapped read-only
+    permRW_  = 8 + 2,       // mapped readable and writable
+    permRW_L = 8 + 2 + 1,   // initially mapped r/w, then made read-only
+                            // loader writable
+    permR_X  = 8 + 4,       // mapped readable and executable
+    permRWX  = 8 + 2 + 4,   // mapped readable and writable and executable
+    permUnknown = 16        // unknown or invalid permissions
+  };
+
+  enum SectionChoice {
+    sectionBasedOnContent,  // linker infers final section based on content
+    sectionCustomPreferred, // linker may place in specific section
+    sectionCustomRequired   // linker must place in specific section
+  };
+
+  enum DeadStripKind {
+    deadStripNormal,        // linker may dead strip this atom
+    deadStripNever,         // linker must never dead strip this atom
+    deadStripAlways         // linker must remove this atom if unused
+  };
+
+  enum DynamicExport {
+    /// \brief The linker may or may not export this atom dynamically depending
+    ///   on the output type and other context of the link.
+    dynamicExportNormal,
+    /// \brief The linker will always export this atom dynamically.
+    dynamicExportAlways,
+  };
+
+  // Attributes describe a code model used by the atom.
+  enum CodeModel {
+    codeNA,           // no specific code model
+    codeMipsPIC,      // PIC function in a PIC / non-PIC mixed file
+    codeMipsMicro,    // microMIPS instruction encoding
+    codeMipsMicroPIC, // microMIPS instruction encoding + PIC
+    codeMips16,       // MIPS-16 instruction encoding
+    codeARMThumb,     // ARM Thumb instruction set
+  };
+
+  struct Alignment {
+    Alignment(int p2, int m = 0)
+      : powerOf2(p2)
+      , modulus(m) {}
+
+    uint16_t powerOf2;
+    uint16_t modulus;
+
+    bool operator==(const Alignment &rhs) const {
+      return (powerOf2 == rhs.powerOf2) && (modulus == rhs.modulus);
+    }
+  };
+
+  /// \brief returns a value for the order of this Atom within its file.
+  ///
+  /// This is used by the linker to order the layout of Atoms so that the
+  /// resulting image is stable and reproducible.
+  ///
+  /// Note that this should not be confused with ordinals of exported symbols in
+  /// Windows DLLs. In Windows terminology, ordinals are symbols' export table
+  /// indices (small integers) which can be used instead of symbol names to
+  /// refer items in a DLL.
+  virtual uint64_t ordinal() const = 0;
+
+  /// \brief the number of bytes of space this atom's content will occupy in the
+  /// final linked image.
+  ///
+  /// For a function atom, it is the number of bytes of code in the function.
+  virtual uint64_t size() const = 0;
+
+  /// \brief The size of the section from which the atom is instantiated.
+  ///
+  /// Merge::mergeByLargestSection is defined in terms of section size
+  /// and not in terms of atom size, so we need this function separate
+  /// from size().
+  virtual uint64_t sectionSize() const { return 0; }
+
+  /// \brief The visibility of this atom to other atoms.
+  ///
+  /// C static functions have scope scopeTranslationUnit.  Regular C functions
+  /// have scope scopeGlobal.  Functions compiled with visibility=hidden have
+  /// scope scopeLinkageUnit so they can be see by other atoms being linked but
+  /// not by the OS loader.
+  virtual Scope scope() const = 0;
+
+  /// \brief Whether the linker should use direct or indirect access to this
+  /// atom.
+  virtual Interposable interposable() const = 0;
+
+  /// \brief how the linker should handle if multiple atoms have the same name.
+  virtual Merge merge() const = 0;
+
+  /// \brief The type of this atom, such as code or data.
+  virtual ContentType contentType() const = 0;
+
+  /// \brief The alignment constraints on how this atom must be laid out in the
+  /// final linked image (e.g. 16-byte aligned).
+  virtual Alignment alignment() const = 0;
+
+  /// \brief Whether this atom must be in a specially named section in the final
+  /// linked image, or if the linker can infer the section based on the
+  /// contentType().
+  virtual SectionChoice sectionChoice() const = 0;
+
+  /// \brief If sectionChoice() != sectionBasedOnContent, then this return the
+  /// name of the section the atom should be placed into.
+  virtual StringRef customSectionName() const = 0;
+
+  /// \brief constraints on whether the linker may dead strip away this atom.
+  virtual DeadStripKind deadStrip() const = 0;
+
+  /// \brief Under which conditions should this atom be dynamically exported.
+  virtual DynamicExport dynamicExport() const {
+    return dynamicExportNormal;
+  }
+
+  /// \brief Code model used by the atom.
+  virtual CodeModel codeModel() const { return codeNA; }
+
+  /// \brief Returns the OS memory protections required for this atom's content
+  /// at runtime.
+  ///
+  /// A function atom is R_X, a global variable is RW_, and a read-only constant
+  /// is R__.
+  virtual ContentPermissions permissions() const;
+
+  /// \brief returns a reference to the raw (unrelocated) bytes of this Atom's
+  /// content.
+  virtual ArrayRef<uint8_t> rawContent() const = 0;
+
+  /// This class abstracts iterating over the sequence of References
+  /// in an Atom.  Concrete instances of DefinedAtom must implement
+  /// the derefIterator() and incrementIterator() methods.
+  class reference_iterator {
+  public:
+    reference_iterator(const DefinedAtom &a, const void *it)
+      : _atom(a), _it(it) { }
+
+    const Reference *operator*() const {
+      return _atom.derefIterator(_it);
+    }
+
+    const Reference *operator->() const {
+      return _atom.derefIterator(_it);
+    }
+
+    bool operator!=(const reference_iterator &other) const {
+      return _it != other._it;
+    }
+
+    reference_iterator &operator++() {
+      _atom.incrementIterator(_it);
+      return *this;
+    }
+  private:
+    const DefinedAtom &_atom;
+    const void *_it;
+  };
+
+  /// \brief Returns an iterator to the beginning of this Atom's References.
+  virtual reference_iterator begin() const = 0;
+
+  /// \brief Returns an iterator to the end of this Atom's References.
+  virtual reference_iterator end() const = 0;
+
+  static bool classof(const Atom *a) {
+    return a->definition() == definitionRegular;
+  }
+
+  /// Utility for deriving permissions from content type
+  static ContentPermissions permissions(ContentType type);
+
+  /// Utility function to check if the atom occupies file space
+  bool occupiesDiskSpace() const {
+    ContentType atomContentType = contentType();
+    return !(atomContentType == DefinedAtom::typeZeroFill ||
+             atomContentType == DefinedAtom::typeZeroFillFast ||
+             atomContentType == DefinedAtom::typeTLVInitialZeroFill ||
+             atomContentType == DefinedAtom::typeThreadZeroFill);
+  }
+
+  /// Utility function to check if the atom belongs to a group section
+  /// that represents section groups or .gnu.linkonce sections.
+  bool isGroupParent() const {
+    ContentType atomContentType = contentType();
+    return (atomContentType == DefinedAtom::typeGroupComdat ||
+            atomContentType == DefinedAtom::typeGnuLinkOnce);
+  }
+
+  // Returns true if lhs should be placed before rhs in the final output.
+  static bool compareByPosition(const DefinedAtom *lhs,
+                                const DefinedAtom *rhs);
+
+protected:
+  // DefinedAtom is an abstract base class. Only subclasses can access
+  // constructor.
+  DefinedAtom() : Atom(definitionRegular) { }
+
+  /// \brief Returns a pointer to the Reference object that the abstract
+  /// iterator "points" to.
+  virtual const Reference *derefIterator(const void *iter) const = 0;
+
+  /// \brief Adjusts the abstract iterator to "point" to the next Reference
+  /// object for this Atom.
+  virtual void incrementIterator(const void *&iter) const = 0;
+};
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/Error.h b/include/lld/Core/Error.h
new file mode 100644
index 000000000000..7caa25018f40
--- /dev/null
+++ b/include/lld/Core/Error.h
@@ -0,0 +1,82 @@
+//===- Error.h - system_error extensions for lld ----------------*- C++ -*-===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This declares a new error_category for the lld library.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_ERROR_H
+#define LLD_CORE_ERROR_H
+
+#include "lld/Core/LLVM.h"
+#include <system_error>
+
+namespace lld {
+
+const std::error_category &native_reader_category();
+
+enum class NativeReaderError {
+  success = 0,
+  unknown_file_format,
+  file_too_short,
+  file_malformed,
+  unknown_chunk_type,
+  memory_error,
+  conflicting_target_machine,
+};
+
+inline std::error_code make_error_code(NativeReaderError e) {
+  return std::error_code(static_cast<int>(e), native_reader_category());
+}
+
+const std::error_category &YamlReaderCategory();
+
+enum class YamlReaderError {
+  success = 0,
+  unknown_keyword,
+  illegal_value
+};
+
+inline std::error_code make_error_code(YamlReaderError e) {
+  return std::error_code(static_cast<int>(e), YamlReaderCategory());
+}
+
+const std::error_category &LinkerScriptReaderCategory();
+
+enum class LinkerScriptReaderError {
+  success = 0,
+  parse_error,
+  unknown_symbol_in_expr,
+  unrecognized_function_in_expr
+};
+
+inline std::error_code make_error_code(LinkerScriptReaderError e) {
+  return std::error_code(static_cast<int>(e), LinkerScriptReaderCategory());
+}
+
+/// Creates an error_code object that has associated with it an arbitrary
+/// error messsage.  The value() of the error_code will always be non-zero
+/// but its value is meaningless. The messsage() will be (a copy of) the
+/// supplied error string.
+/// Note:  Once ErrorOr<> is updated to work with errors other than error_code,
+/// this can be updated to return some other kind of error.
+std::error_code make_dynamic_error_code(StringRef msg);
+std::error_code make_dynamic_error_code(const Twine &msg);
+
+} // end namespace lld
+
+namespace std {
+template <>
+struct is_error_code_enum<lld::NativeReaderError> : std::true_type {};
+template <> struct is_error_code_enum<lld::YamlReaderError> : std::true_type {};
+template <>
+struct is_error_code_enum<lld::LinkerScriptReaderError> : std::true_type {};
+}
+
+#endif
diff --git a/include/lld/Core/File.h b/include/lld/Core/File.h
new file mode 100644
index 000000000000..25b177ec879c
--- /dev/null
+++ b/include/lld/Core/File.h
@@ -0,0 +1,324 @@
+//===- Core/File.h - A Container of Atoms ---------------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_FILE_H
+#define LLD_CORE_FILE_H
+
+#include "lld/Core/AbsoluteAtom.h"
+#include "lld/Core/DefinedAtom.h"
+#include "lld/Core/SharedLibraryAtom.h"
+#include "lld/Core/UndefinedAtom.h"
+#include "lld/Core/range.h"
+#include "llvm/ADT/Optional.h"
+#include "llvm/ADT/Twine.h"
+#include "llvm/Support/ErrorHandling.h"
+#include <functional>
+#include <memory>
+#include <mutex>
+#include <vector>
+
+namespace lld {
+
+class LinkingContext;
+
+/// Every Atom is owned by some File. A common scenario is for a single
+/// object file (.o) to be parsed by some reader and produce a single
+/// File object that represents the content of that object file.
+///
+/// To iterate through the Atoms in a File there are four methods that
+/// return collections.  For instance to iterate through all the DefinedAtoms
+/// in a File object use:
+///      for (const DefinedAtoms *atom : file->defined()) {
+///      }
+///
+/// The Atom objects in a File are owned by the File object.  The Atom objects
+/// are destroyed when the File object is destroyed.
+class File {
+public:
+  virtual ~File();
+
+  /// \brief Kinds of files that are supported.
+  enum Kind {
+    kindObject,        ///< object file (.o)
+    kindSharedLibrary, ///< shared library (.so)
+    kindArchiveLibrary ///< archive (.a)
+  };
+
+  /// \brief Returns file kind.  Need for dyn_cast<> on File objects.
+  Kind kind() const {
+    return _kind;
+  }
+
+  /// This returns the path to the file which was used to create this object
+  /// (e.g. "/tmp/foo.o"). If the file is a member of an archive file, the
+  /// returned string includes the archive file name.
+  StringRef path() const {
+    if (_archivePath.empty())
+      return _path;
+    if (_archiveMemberPath.empty())
+      _archiveMemberPath = (_archivePath + "(" + _path + ")").str();
+    return _archiveMemberPath;
+  }
+
+  /// Returns the path of the archive file name if this file is instantiated
+  /// from an archive file. Otherwise returns the empty string.
+  StringRef archivePath() const { return _archivePath; }
+  void setArchivePath(StringRef path) { _archivePath = path; }
+
+  /// Returns the path name of this file. It doesn't include archive file name.
+  StringRef memberPath() const { return _path; }
+
+  /// Returns the command line order of the file.
+  uint64_t ordinal() const {
+    assert(_ordinal != UINT64_MAX);
+    return _ordinal;
+  }
+
+  /// Returns true/false depending on whether an ordinal has been set.
+  bool hasOrdinal() const { return (_ordinal != UINT64_MAX); }
+
+  /// Sets the command line order of the file.
+  void setOrdinal(uint64_t ordinal) const { _ordinal = ordinal; }
+
+  template <typename T> class atom_iterator; // forward reference
+
+  /// For allocating any objects owned by this File.
+  llvm::BumpPtrAllocator &allocator() const {
+    return _allocator;
+  }
+
+  /// \brief For use interating over DefinedAtoms in this File.
+  typedef atom_iterator<DefinedAtom>  defined_iterator;
+
+  /// \brief For use interating over UndefinedAtoms in this File.
+  typedef atom_iterator<UndefinedAtom> undefined_iterator;
+
+  /// \brief For use interating over SharedLibraryAtoms in this File.
+  typedef atom_iterator<SharedLibraryAtom> shared_library_iterator;
+
+  /// \brief For use interating over AbsoluteAtoms in this File.
+  typedef atom_iterator<AbsoluteAtom> absolute_iterator;
+
+  /// \brief Different object file readers may instantiate and manage atoms with
+  /// different data structures.  This class is a collection abstraction.
+  /// Each concrete File instance must implement these atom_collection
+  /// methods to enable clients to interate the File's atoms.
+  template <typename T>
+  class atom_collection {
+  public:
+    virtual ~atom_collection() { }
+    virtual atom_iterator<T> begin() const = 0;
+    virtual atom_iterator<T> end() const = 0;
+    virtual const T *deref(const void *it) const = 0;
+    virtual void next(const void *&it) const = 0;
+    virtual uint64_t size() const = 0;
+    bool empty() const { return size() == 0; }
+  };
+
+  /// \brief The class is the iterator type used to iterate through a File's
+  /// Atoms. This iterator delegates the work to the associated atom_collection
+  /// object. There are four kinds of Atoms, so this iterator is templated on
+  /// the four base Atom kinds.
+  template <typename T>
+  class atom_iterator : public std::iterator<std::forward_iterator_tag, T> {
+  public:
+    atom_iterator(const atom_collection<T> &c, const void *it)
+              : _collection(&c), _it(it) { }
+
+    const T *operator*() const {
+      return _collection->deref(_it);
+    }
+    const T *operator->() const {
+      return _collection->deref(_it);
+    }
+
+    friend bool operator==(const atom_iterator<T> &lhs, const atom_iterator<T> &rhs)  {
+      return lhs._it == rhs._it;
+    }
+
+    friend bool operator!=(const atom_iterator<T> &lhs, const atom_iterator<T> &rhs)  {
+      return !(lhs == rhs);
+    }
+
+    atom_iterator<T> &operator++() {
+      _collection->next(_it);
+      return *this;
+    }
+  private:
+    const atom_collection<T> *_collection;
+    const void               *_it;
+  };
+
+  /// \brief Must be implemented to return the atom_collection object for
+  /// all DefinedAtoms in this File.
+  virtual const atom_collection<DefinedAtom> &defined() const = 0;
+
+  /// \brief Must be implemented to return the atom_collection object for
+  /// all UndefinedAtomw in this File.
+  virtual const atom_collection<UndefinedAtom> &undefined() const = 0;
+
+  /// \brief Must be implemented to return the atom_collection object for
+  /// all SharedLibraryAtoms in this File.
+  virtual const atom_collection<SharedLibraryAtom> &sharedLibrary() const = 0;
+
+  /// \brief Must be implemented to return the atom_collection object for
+  /// all AbsoluteAtoms in this File.
+  virtual const atom_collection<AbsoluteAtom> &absolute() const = 0;
+
+  /// \brief If a file is parsed using a different method than doParse(),
+  /// one must use this method to set the last error status, so that
+  /// doParse will not be called twice. Only YAML reader uses this
+  /// (because YAML reader does not read blobs but structured data).
+  void setLastError(std::error_code err) { _lastError = err; }
+
+  std::error_code parse();
+
+  // This function is called just before the core linker tries to use
+  // a file. Currently the PECOFF reader uses this to trigger the
+  // driver to parse .drectve section (which contains command line options).
+  // If you want to do something having side effects, don't do that in
+  // doParse() because a file could be pre-loaded speculatively.
+  // Use this hook instead.
+  virtual void beforeLink() {}
+
+  // Usually each file owns a std::unique_ptr<MemoryBuffer>.
+  // However, there's one special case. If a file is an archive file,
+  // the archive file and its children all shares the same memory buffer.
+  // This method is used by the ArchiveFile to give its children
+  // co-ownership of the buffer.
+  void setSharedMemoryBuffer(std::shared_ptr<MemoryBuffer> mb) {
+    _sharedMemoryBuffer = mb;
+  }
+
+protected:
+  /// \brief only subclasses of File can be instantiated
+  File(StringRef p, Kind kind)
+      : _path(p), _kind(kind), _ordinal(UINT64_MAX) {}
+
+  /// \brief Subclasses should override this method to parse the
+  /// memory buffer passed to this file's constructor.
+  virtual std::error_code doParse() { return std::error_code(); }
+
+  /// \brief This is a convenience class for File subclasses which manage their
+  /// atoms as a simple std::vector<>.
+  template <typename T>
+  class atom_collection_vector : public atom_collection<T> {
+  public:
+    atom_iterator<T> begin() const override {
+      auto *it = _atoms.empty() ? nullptr
+                                : reinterpret_cast<const void *>(_atoms.data());
+      return atom_iterator<T>(*this, it);
+    }
+
+    atom_iterator<T> end() const override {
+      auto *it = _atoms.empty() ? nullptr : reinterpret_cast<const void *>(
+                                                _atoms.data() + _atoms.size());
+      return atom_iterator<T>(*this, it);
+    }
+
+    const T *deref(const void *it) const override {
+      return *reinterpret_cast<const T *const *>(it);
+    }
+
+    void next(const void *&it) const override {
+      const T *const *p = reinterpret_cast<const T *const *>(it);
+      ++p;
+      it = reinterpret_cast<const void*>(p);
+    }
+
+    uint64_t size() const override { return _atoms.size(); }
+
+    std::vector<const T *> _atoms;
+  };
+
+  /// \brief This is a convenience class for File subclasses which need to
+  /// return an empty collection.
+  template <typename T>
+  class atom_collection_empty : public atom_collection<T> {
+  public:
+    atom_iterator<T> begin() const override {
+      return atom_iterator<T>(*this, nullptr);
+    }
+    atom_iterator<T> end() const override {
+      return atom_iterator<T>(*this, nullptr);
+    }
+    const T *deref(const void *it) const override {
+      llvm_unreachable("empty collection should never be accessed");
+    }
+    void next(const void *&it) const override {}
+    uint64_t size() const override { return 0; }
+  };
+
+  static atom_collection_empty<DefinedAtom>       _noDefinedAtoms;
+  static atom_collection_empty<UndefinedAtom>     _noUndefinedAtoms;
+  static atom_collection_empty<SharedLibraryAtom> _noSharedLibraryAtoms;
+  static atom_collection_empty<AbsoluteAtom>      _noAbsoluteAtoms;
+  mutable llvm::BumpPtrAllocator                  _allocator;
+
+private:
+  StringRef _path;
+  std::string _archivePath;
+  mutable std::string _archiveMemberPath;
+  Kind              _kind;
+  mutable uint64_t  _ordinal;
+  std::shared_ptr<MemoryBuffer> _sharedMemoryBuffer;
+  llvm::Optional<std::error_code> _lastError;
+  std::mutex _parseMutex;
+};
+
+/// \brief A mutable File.
+class MutableFile : public File {
+public:
+  /// \brief Add an atom to the file. Invalidates iterators for all returned
+  /// containters.
+  virtual void addAtom(const Atom&) = 0;
+
+  typedef range<std::vector<const DefinedAtom *>::iterator> DefinedAtomRange;
+  virtual DefinedAtomRange definedAtoms() = 0;
+
+  virtual void
+  removeDefinedAtomsIf(std::function<bool(const DefinedAtom *)> pred) = 0;
+
+protected:
+  /// \brief only subclasses of MutableFile can be instantiated
+  MutableFile(StringRef p) : File(p, kindObject) {}
+};
+
+/// An ErrorFile represents a file that doesn't exist.
+/// If you try to parse a file which doesn't exist, an instance of this
+/// class will be returned. That's parse method always returns an error.
+/// This is useful to delay erroring on non-existent files, so that we
+/// can do unit testing a driver using non-existing file paths.
+class ErrorFile : public File {
+public:
+  ErrorFile(StringRef path, std::error_code ec)
+      : File(path, kindObject), _ec(ec) {}
+
+  std::error_code doParse() override { return _ec; }
+
+  const atom_collection<DefinedAtom> &defined() const override {
+    llvm_unreachable("internal error");
+  }
+  const atom_collection<UndefinedAtom> &undefined() const override {
+    llvm_unreachable("internal error");
+  }
+  const atom_collection<SharedLibraryAtom> &sharedLibrary() const override {
+    llvm_unreachable("internal error");
+  }
+  const atom_collection<AbsoluteAtom> &absolute() const override {
+    llvm_unreachable("internal error");
+  }
+
+private:
+  std::error_code _ec;
+};
+
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/Instrumentation.h b/include/lld/Core/Instrumentation.h
new file mode 100644
index 000000000000..162375905e17
--- /dev/null
+++ b/include/lld/Core/Instrumentation.h
@@ -0,0 +1,132 @@
+//===- include/Core/Instrumentation.h - Instrumentation API ---------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// \brief Provide an Instrumentation API that optionally uses VTune interfaces.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_INSTRUMENTATION_H
+#define LLD_CORE_INSTRUMENTATION_H
+
+#include "llvm/Support/Compiler.h"
+#include <utility>
+
+#ifdef LLD_HAS_VTUNE
+# include <ittnotify.h>
+#endif
+
+namespace lld {
+#ifdef LLD_HAS_VTUNE
+/// \brief A unique global scope for instrumentation data.
+///
+/// Domains last for the lifetime of the application and cannot be destroyed.
+/// Multiple Domains created with the same name represent the same domain.
+class Domain {
+  __itt_domain *_domain;
+
+public:
+  explicit Domain(const char *name) : _domain(__itt_domain_createA(name)) {}
+
+  operator __itt_domain *() const { return _domain; }
+  __itt_domain *operator->() const { return _domain; }
+};
+
+/// \brief A global reference to a string constant.
+///
+/// These are uniqued by the ITT runtime and cannot be deleted. They are not
+/// specific to a domain.
+///
+/// Prefer reusing a single StringHandle over passing a ntbs when the same
+/// string will be used often.
+class StringHandle {
+  __itt_string_handle *_handle;
+
+public:
+  StringHandle(const char *name) : _handle(__itt_string_handle_createA(name)) {}
+
+  operator __itt_string_handle *() const { return _handle; }
+};
+
+/// \brief A task on a single thread. Nests within other tasks.
+///
+/// Each thread has its own task stack and tasks nest recursively on that stack.
+/// A task cannot transfer threads.
+///
+/// SBRM is used to ensure task starts and ends are ballanced. The lifetime of
+/// a task is either the lifetime of this object, or until end is called.
+class ScopedTask {
+  __itt_domain *_domain;
+
+  ScopedTask(const ScopedTask &) = delete;
+  ScopedTask &operator=(const ScopedTask &) = delete;
+
+public:
+  /// \brief Create a task in Domain \p d named \p s.
+  ScopedTask(const Domain &d, const StringHandle &s) : _domain(d) {
+    __itt_task_begin(d, __itt_null, __itt_null, s);
+  }
+
+  ScopedTask(ScopedTask &&other) {
+    *this = std::move(other);
+  }
+
+  ScopedTask &operator=(ScopedTask &&other) {
+    _domain = other._domain;
+    other._domain = nullptr;
+    return *this;
+  }
+
+  /// \brief Prematurely end this task.
+  void end() {
+    if (_domain)
+      __itt_task_end(_domain);
+    _domain = nullptr;
+  }
+
+  ~ScopedTask() { end(); }
+};
+
+/// \brief A specific point in time. Allows metadata to be associated.
+class Marker {
+public:
+  Marker(const Domain &d, const StringHandle &s) {
+    __itt_marker(d, __itt_null, s, __itt_scope_global);
+  }
+};
+#else
+class Domain {
+public:
+  Domain(const char *name) {}
+};
+
+class StringHandle {
+public:
+  StringHandle(const char *name) {}
+};
+
+class ScopedTask {
+public:
+  ScopedTask(const Domain &d, const StringHandle &s) {}
+  void end() {}
+};
+
+class Marker {
+public:
+  Marker(const Domain &d, const StringHandle &s) {}
+};
+#endif
+
+inline const Domain &getDefaultDomain() {
+  static Domain domain("org.llvm.lld");
+  return domain;
+}
+} // end namespace lld.
+
+#endif
diff --git a/include/lld/Core/LLVM.h b/include/lld/Core/LLVM.h
new file mode 100644
index 000000000000..1bc1173bd48b
--- /dev/null
+++ b/include/lld/Core/LLVM.h
@@ -0,0 +1,75 @@
+//===--- LLVM.h - Import various common LLVM datatypes ----------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file forward declares and imports various common LLVM datatypes that
+// lld wants to use unqualified.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_LLVM_H
+#define LLD_CORE_LLVM_H
+
+// This should be the only #include, force #includes of all the others on
+// clients.
+#include "llvm/ADT/Hashing.h"
+#include "llvm/Support/Casting.h"
+#include <utility>
+
+namespace llvm {
+  // ADT's.
+  class StringRef;
+  class Twine;
+  class MemoryBuffer;
+  template<typename T> class ArrayRef;
+  template<unsigned InternalLen> class SmallString;
+  template<typename T, unsigned N> class SmallVector;
+  template<typename T> class SmallVectorImpl;
+
+  template<typename T>
+  struct SaveAndRestore;
+
+  template<typename T>
+  class ErrorOr;
+
+  class raw_ostream;
+  // TODO: DenseMap, ...
+}
+
+namespace lld {
+  // Casting operators.
+  using llvm::isa;
+  using llvm::cast;
+  using llvm::dyn_cast;
+  using llvm::dyn_cast_or_null;
+  using llvm::cast_or_null;
+
+  // ADT's.
+  using llvm::StringRef;
+  using llvm::Twine;
+  using llvm::MemoryBuffer;
+  using llvm::ArrayRef;
+  using llvm::SmallString;
+  using llvm::SmallVector;
+  using llvm::SmallVectorImpl;
+  using llvm::SaveAndRestore;
+  using llvm::ErrorOr;
+
+  using llvm::raw_ostream;
+} // end namespace lld.
+
+namespace std {
+template <> struct hash<llvm::StringRef> {
+public:
+  size_t operator()(const llvm::StringRef &s) const {
+    return llvm::hash_value(s);
+  }
+};
+}
+
+#endif
diff --git a/include/lld/Core/LinkingContext.h b/include/lld/Core/LinkingContext.h
new file mode 100644
index 000000000000..81a3b4b4eb71
--- /dev/null
+++ b/include/lld/Core/LinkingContext.h
@@ -0,0 +1,364 @@
+//===- lld/Core/LinkingContext.h - Linker Target Info Interface -----------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_LINKING_CONTEXT_H
+#define LLD_CORE_LINKING_CONTEXT_H
+
+#include "lld/Core/Error.h"
+#include "lld/Core/LLVM.h"
+#include "lld/Core/Node.h"
+#include "lld/Core/Parallel.h"
+#include "lld/Core/Reference.h"
+#include "lld/Core/range.h"
+#include "lld/Core/Reader.h"
+#include "llvm/Support/ErrorOr.h"
+#include "llvm/Support/raw_ostream.h"
+#include <string>
+#include <vector>
+
+namespace lld {
+class PassManager;
+class File;
+class Writer;
+class Node;
+class SharedLibraryFile;
+
+/// \brief The LinkingContext class encapsulates "what and how" to link.
+///
+/// The base class LinkingContext contains the options needed by core linking.
+/// Subclasses of LinkingContext have additional options needed by specific
+/// Writers. For example, ELFLinkingContext has methods that supplies
+/// options to the ELF Writer and ELF Passes.
+class LinkingContext {
+public:
+  /// \brief The types of output file that the linker creates.
+  enum class OutputFileType : uint8_t {
+    Default, // The default output type for this target
+    YAML,    // The output type is set to YAML
+    Native   // The output file format is Native (Atoms)
+  };
+
+  virtual ~LinkingContext();
+
+  /// \name Methods needed by core linking
+  /// @{
+
+  /// Name of symbol linker should use as "entry point" to program,
+  /// usually "main" or "start".
+  virtual StringRef entrySymbolName() const { return _entrySymbolName; }
+
+  /// Whether core linking should remove Atoms not reachable by following
+  /// References from the entry point Atom or from all global scope Atoms
+  /// if globalsAreDeadStripRoots() is true.
+  bool deadStrip() const { return _deadStrip; }
+
+  /// Only used if deadStrip() returns true.  Means all global scope Atoms
+  /// should be marked live (along with all Atoms they reference).  Usually
+  /// this method returns false for main executables, but true for dynamic
+  /// shared libraries.
+  bool globalsAreDeadStripRoots() const { return _globalsAreDeadStripRoots; };
+
+  /// Only used if deadStrip() returns true.  This method returns the names
+  /// of DefinedAtoms that should be marked live (along with all Atoms they
+  /// reference). Only Atoms with scope scopeLinkageUnit or scopeGlobal can
+  /// be kept live using this method.
+  const std::vector<StringRef> &deadStripRoots() const {
+    return _deadStripRoots;
+  }
+
+  /// Add the given symbol name to the dead strip root set. Only used if
+  /// deadStrip() returns true.
+  void addDeadStripRoot(StringRef symbolName) {
+    assert(!symbolName.empty() && "Empty symbol cannot be a dead strip root");
+    _deadStripRoots.push_back(symbolName);
+  }
+
+  /// Archive files (aka static libraries) are normally lazily loaded.  That is,
+  /// object files within an archive are only loaded and linked in, if the
+  /// object file contains a DefinedAtom which will replace an existing
+  /// UndefinedAtom.  If this method returns true, core linking will also look
+  /// for archive members to replace existing tentative definitions in addition
+  /// to replacing undefines. Note: a "tentative definition" (also called a
+  /// "common" symbols) is a C (but not C++) concept. They are modeled in lld
+  /// as a DefinedAtom with merge() of mergeAsTentative.
+  bool searchArchivesToOverrideTentativeDefinitions() const {
+    return _searchArchivesToOverrideTentativeDefinitions;
+  }
+
+  /// Normally core linking will turn a tentative definition into a real
+  /// definition if not replaced by a real DefinedAtom from some object file.
+  /// If this method returns true, core linking will search all supplied
+  /// dynamic shared libraries for symbol names that match remaining tentative
+  /// definitions.  If any are found, the corresponding tentative definition
+  /// atom is replaced with SharedLibraryAtom.
+  bool searchSharedLibrariesToOverrideTentativeDefinitions() const {
+    return _searchSharedLibrariesToOverrideTentativeDefinitions;
+  }
+
+  /// Normally, every UndefinedAtom must be replaced by a DefinedAtom or a
+  /// SharedLibraryAtom for the link to be successful.  This method controls
+  /// whether core linking prints out a list of remaining UndefinedAtoms.
+  ///
+  /// \todo This should be a method core linking calls with a list of the
+  /// UndefinedAtoms so that different drivers can format the error message
+  /// as needed.
+  bool printRemainingUndefines() const { return _printRemainingUndefines; }
+
+  /// Normally, every UndefinedAtom must be replaced by a DefinedAtom or a
+  /// SharedLibraryAtom for the link to be successful.  This method controls
+  /// whether core linking considers remaining undefines to be an error.
+  bool allowRemainingUndefines() const { return _allowRemainingUndefines; }
+
+  /// In the lld model, a SharedLibraryAtom is a proxy atom for something
+  /// that will be found in a dynamic shared library when the program runs.
+  /// A SharedLibraryAtom optionally contains the name of the shared library
+  /// in which to find the symbol name at runtime.  Core linking may merge
+  /// two SharedLibraryAtom with the same name.  If this method returns true,
+  /// when merging core linking will also verify that they both have the same
+  /// loadName() and if not print a warning.
+  ///
+  /// \todo This should be a method core linking calls so that drivers can
+  /// format the warning as needed.
+  bool warnIfCoalesableAtomsHaveDifferentLoadName() const {
+    return _warnIfCoalesableAtomsHaveDifferentLoadName;
+  }
+
+  /// In C/C++ you can mark a function's prototype with
+  /// __attribute__((weak_import)) or __attribute__((weak)) to say the function
+  /// may not be available at runtime and/or build time and in which case its
+  /// address will evaluate to NULL. In lld this is modeled using the
+  /// UndefinedAtom::canBeNull() method.  During core linking, UndefinedAtom
+  /// with the same name are automatically merged.  If this method returns
+  /// true, core link also verfies that the canBeNull() value for merged
+  /// UndefinedAtoms are the same and warns if not.
+  ///
+  /// \todo This should be a method core linking calls so that drivers can
+  /// format the warning as needed.
+  bool warnIfCoalesableAtomsHaveDifferentCanBeNull() const {
+    return _warnIfCoalesableAtomsHaveDifferentCanBeNull;
+  }
+
+  /// Normally, every UndefinedAtom must be replaced by a DefinedAtom or a
+  /// SharedLibraryAtom for the link to be successful.  This method controls
+  /// whether core linking considers remaining undefines from the shared library
+  /// to be an error.
+  bool allowShlibUndefines() const { return _allowShlibUndefines; }
+
+  /// If true, core linking will write the path to each input file to stdout
+  /// (i.e. llvm::outs()) as it is used.  This is used to implement the -t
+  /// linker option.
+  ///
+  /// \todo This should be a method core linking calls so that drivers can
+  /// format the line as needed.
+  bool logInputFiles() const { return _logInputFiles; }
+
+  /// Parts of LLVM use global variables which are bound to command line
+  /// options (see llvm::cl::Options). This method returns "command line"
+  /// options which are used to configure LLVM's command line settings.
+  /// For instance the -debug-only XXX option can be used to dynamically
+  /// trace different parts of LLVM and lld.
+  const std::vector<const char *> &llvmOptions() const { return _llvmOptions; }
+
+  /// \name Methods used by Drivers to configure TargetInfo
+  /// @{
+  void setOutputPath(StringRef str) { _outputPath = str; }
+
+  // Set the entry symbol name. You may also need to call addDeadStripRoot() for
+  // the symbol if your platform supports dead-stripping, so that the symbol
+  // will not be removed from the output.
+  void setEntrySymbolName(StringRef name) {
+    _entrySymbolName = name;
+  }
+
+  void setDeadStripping(bool enable) { _deadStrip = enable; }
+  void setAllowDuplicates(bool enable) { _allowDuplicates = enable; }
+  void setGlobalsAreDeadStripRoots(bool v) { _globalsAreDeadStripRoots = v; }
+  void setSearchArchivesToOverrideTentativeDefinitions(bool search) {
+    _searchArchivesToOverrideTentativeDefinitions = search;
+  }
+  void setSearchSharedLibrariesToOverrideTentativeDefinitions(bool search) {
+    _searchSharedLibrariesToOverrideTentativeDefinitions = search;
+  }
+  void setWarnIfCoalesableAtomsHaveDifferentCanBeNull(bool warn) {
+    _warnIfCoalesableAtomsHaveDifferentCanBeNull = warn;
+  }
+  void setWarnIfCoalesableAtomsHaveDifferentLoadName(bool warn) {
+    _warnIfCoalesableAtomsHaveDifferentLoadName = warn;
+  }
+  void setPrintRemainingUndefines(bool print) {
+    _printRemainingUndefines = print;
+  }
+  void setAllowRemainingUndefines(bool allow) {
+    _allowRemainingUndefines = allow;
+  }
+  void setAllowShlibUndefines(bool allow) { _allowShlibUndefines = allow; }
+  void setLogInputFiles(bool log) { _logInputFiles = log; }
+
+  // Returns true if multiple definitions should not be treated as a
+  // fatal error.
+  bool getAllowDuplicates() const { return _allowDuplicates; }
+
+  void appendLLVMOption(const char *opt) { _llvmOptions.push_back(opt); }
+
+  void addAlias(StringRef from, StringRef to) { _aliasSymbols[from] = to; }
+  const std::map<std::string, std::string> &getAliases() const {
+    return _aliasSymbols;
+  }
+
+  std::vector<std::unique_ptr<Node>> &getNodes() { return _nodes; }
+  const std::vector<std::unique_ptr<Node>> &getNodes() const { return _nodes; }
+
+  /// Notify the LinkingContext when the symbol table found a name collision.
+  /// The useNew parameter specifies which the symbol table plans to keep,
+  /// but that can be changed by the LinkingContext.  This is also an
+  /// opportunity for flavor specific processing.
+  virtual void notifySymbolTableCoalesce(const Atom *existingAtom,
+                                         const Atom *newAtom, bool &useNew) {}
+
+  /// This method adds undefined symbols specified by the -u option to the to
+  /// the list of undefined symbols known to the linker. This option essentially
+  /// forces an undefined symbol to be created. You may also need to call
+  /// addDeadStripRoot() for the symbol if your platform supports dead
+  /// stripping, so that the symbol will not be removed from the output.
+  void addInitialUndefinedSymbol(StringRef symbolName) {
+    _initialUndefinedSymbols.push_back(symbolName);
+  }
+
+  /// Iterators for symbols that appear on the command line.
+  typedef std::vector<StringRef> StringRefVector;
+  typedef StringRefVector::iterator StringRefVectorIter;
+  typedef StringRefVector::const_iterator StringRefVectorConstIter;
+
+  /// Create linker internal files containing atoms for the linker to include
+  /// during link. Flavors can override this function in their LinkingContext
+  /// to add more internal files. These internal files are positioned before
+  /// the actual input files.
+  virtual void createInternalFiles(std::vector<std::unique_ptr<File> > &) const;
+
+  /// Return the list of undefined symbols that are specified in the
+  /// linker command line, using the -u option.
+  range<const StringRef *> initialUndefinedSymbols() const {
+    return _initialUndefinedSymbols;
+  }
+
+  /// After all set* methods are called, the Driver calls this method
+  /// to validate that there are no missing options or invalid combinations
+  /// of options.  If there is a problem, a description of the problem
+  /// is written to the supplied stream.
+  ///
+  /// \returns true if there is an error with the current settings.
+  bool validate(raw_ostream &diagnostics);
+
+  /// Formats symbol name for use in error messages.
+  virtual std::string demangle(StringRef symbolName) const {
+    return symbolName;
+  }
+
+  /// @}
+  /// \name Methods used by Driver::link()
+  /// @{
+
+  /// Returns the file system path to which the linked output should be written.
+  ///
+  /// \todo To support in-memory linking, we need an abstraction that allows
+  /// the linker to write to an in-memory buffer.
+  StringRef outputPath() const { return _outputPath; }
+
+  /// Set the various output file types that the linker would
+  /// create
+  bool setOutputFileType(StringRef outputFileType) {
+    if (outputFileType.equals_lower("yaml"))
+      _outputFileType = OutputFileType::YAML;
+    else if (outputFileType.equals_lower("native"))
+      _outputFileType = OutputFileType::YAML;
+    else
+      return false;
+    return true;
+  }
+
+  /// Returns the output file type that that the linker needs to create.
+  OutputFileType outputFileType() const { return _outputFileType; }
+
+  /// Accessor for Register object embedded in LinkingContext.
+  const Registry &registry() const { return _registry; }
+  Registry &registry() { return _registry; }
+
+  /// This method is called by core linking to give the Writer a chance
+  /// to add file format specific "files" to set of files to be linked. This is
+  /// how file format specific atoms can be added to the link.
+  virtual bool createImplicitFiles(std::vector<std::unique_ptr<File> > &);
+
+  /// This method is called by core linking to build the list of Passes to be
+  /// run on the merged/linked graph of all input files.
+  virtual void addPasses(PassManager &pm);
+
+  /// Calls through to the writeFile() method on the specified Writer.
+  ///
+  /// \param linkedFile This is the merged/linked graph of all input file Atoms.
+  virtual std::error_code writeFile(const File &linkedFile) const;
+
+  /// Return the next ordinal and Increment it.
+  virtual uint64_t getNextOrdinalAndIncrement() const { return _nextOrdinal++; }
+
+  // This function is called just before the Resolver kicks in.
+  // Derived classes may use it to change the list of input files.
+  virtual void finalizeInputFiles() {}
+
+  TaskGroup &getTaskGroup() { return _taskGroup; }
+
+  /// @}
+protected:
+  LinkingContext(); // Must be subclassed
+
+  /// Abstract method to lazily instantiate the Writer.
+  virtual Writer &writer() const = 0;
+
+  /// Method to create an internal file for the entry symbol
+  virtual std::unique_ptr<File> createEntrySymbolFile() const;
+  std::unique_ptr<File> createEntrySymbolFile(StringRef filename) const;
+
+  /// Method to create an internal file for an undefined symbol
+  virtual std::unique_ptr<File> createUndefinedSymbolFile() const;
+  std::unique_ptr<File> createUndefinedSymbolFile(StringRef filename) const;
+
+  /// Method to create an internal file for alias symbols
+  std::unique_ptr<File> createAliasSymbolFile() const;
+
+  StringRef _outputPath;
+  StringRef _entrySymbolName;
+  bool _deadStrip;
+  bool _allowDuplicates;
+  bool _globalsAreDeadStripRoots;
+  bool _searchArchivesToOverrideTentativeDefinitions;
+  bool _searchSharedLibrariesToOverrideTentativeDefinitions;
+  bool _warnIfCoalesableAtomsHaveDifferentCanBeNull;
+  bool _warnIfCoalesableAtomsHaveDifferentLoadName;
+  bool _printRemainingUndefines;
+  bool _allowRemainingUndefines;
+  bool _logInputFiles;
+  bool _allowShlibUndefines;
+  OutputFileType _outputFileType;
+  std::vector<StringRef> _deadStripRoots;
+  std::map<std::string, std::string> _aliasSymbols;
+  std::vector<const char *> _llvmOptions;
+  StringRefVector _initialUndefinedSymbols;
+  std::vector<std::unique_ptr<Node>> _nodes;
+  mutable llvm::BumpPtrAllocator _allocator;
+  mutable uint64_t _nextOrdinal;
+  Registry _registry;
+
+private:
+  /// Validate the subclass bits. Only called by validate.
+  virtual bool validateImpl(raw_ostream &diagnostics) = 0;
+  TaskGroup _taskGroup;
+};
+
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/Node.h b/include/lld/Core/Node.h
new file mode 100644
index 000000000000..cd38fbd4a482
--- /dev/null
+++ b/include/lld/Core/Node.h
@@ -0,0 +1,78 @@
+//===- lld/Core/Node.h - Input file class ---------------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+///
+/// The classes in this file represents inputs to the linker.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_NODE_H
+#define LLD_CORE_NODE_H
+
+#include "lld/Core/File.h"
+#include "llvm/Option/ArgList.h"
+#include <memory>
+#include <vector>
+
+namespace lld {
+
+// A Node represents a FileNode or other type of Node. In the latter case,
+// the node contains meta information about the input file list.
+// Currently only GroupEnd node is defined as a meta node.
+class Node {
+public:
+  enum class Kind { File, GroupEnd };
+  explicit Node(Kind type) : _kind(type) {}
+  virtual ~Node() {}
+  virtual Kind kind() const { return _kind; }
+
+private:
+  Kind _kind;
+};
+
+// This is a marker for --end-group. getSize() returns the number of
+// files between the corresponding --start-group and this marker.
+class GroupEnd : public Node {
+public:
+  explicit GroupEnd(int size) : Node(Kind::GroupEnd), _size(size) {}
+
+  int getSize() const { return _size; }
+
+  static bool classof(const Node *a) {
+    return a->kind() == Kind::GroupEnd;
+  }
+
+private:
+  int _size;
+};
+
+// A container of File.
+class FileNode : public Node {
+public:
+  explicit FileNode(std::unique_ptr<File> f)
+      : Node(Node::Kind::File), _file(std::move(f)), _asNeeded(false) {}
+
+  static bool classof(const Node *a) {
+    return a->kind() == Node::Kind::File;
+  }
+
+  File *getFile() { return _file.get(); }
+
+  void setAsNeeded(bool val) { _asNeeded = val; }
+  bool asNeeded() const { return _asNeeded; }
+
+protected:
+  std::unique_ptr<File> _file;
+  bool _asNeeded;
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_NODE_H
diff --git a/include/lld/Core/Parallel.h b/include/lld/Core/Parallel.h
new file mode 100644
index 000000000000..65176ac2b04d
--- /dev/null
+++ b/include/lld/Core/Parallel.h
@@ -0,0 +1,309 @@
+//===- lld/Core/Parallel.h - Parallel utilities ---------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_PARALLEL_H
+#define LLD_CORE_PARALLEL_H
+
+#include "lld/Core/Instrumentation.h"
+#include "lld/Core/LLVM.h"
+#include "lld/Core/range.h"
+#include "llvm/Support/MathExtras.h"
+
+#ifdef _MSC_VER
+// concrt.h depends on eh.h for __uncaught_exception declaration
+// even if we disable exceptions.
+#include <eh.h>
+#endif
+
+#include <algorithm>
+#include <atomic>
+#include <condition_variable>
+#include <mutex>
+#include <thread>
+#include <stack>
+
+#ifdef _MSC_VER
+#include <concrt.h>
+#include <ppl.h>
+#endif
+
+namespace lld {
+/// \brief Allows one or more threads to wait on a potentially unknown number of
+///   events.
+///
+/// A latch starts at \p count. inc() increments this, and dec() decrements it.
+/// All calls to sync() will block while the count is not 0.
+///
+/// Calling dec() on a Latch with a count of 0 has undefined behaivor.
+class Latch {
+  uint32_t _count;
+  mutable std::mutex _condMut;
+  mutable std::condition_variable _cond;
+
+public:
+  explicit Latch(uint32_t count = 0) : _count(count) {}
+  ~Latch() { sync(); }
+
+  void inc() {
+    std::unique_lock<std::mutex> lock(_condMut);
+    ++_count;
+  }
+
+  void dec() {
+    std::unique_lock<std::mutex> lock(_condMut);
+    if (--_count == 0)
+      _cond.notify_all();
+  }
+
+  void sync() const {
+    std::unique_lock<std::mutex> lock(_condMut);
+    _cond.wait(lock, [&] {
+      return _count == 0;
+    });
+  }
+};
+
+/// \brief An implementation of future. std::future and std::promise in
+/// old libstdc++ have a threading bug; there is a small chance that a
+/// call of future::get throws an exception in the normal use case.
+/// We want to use our own future implementation until we drop support
+/// of old versions of libstdc++.
+/// https://gcc.gnu.org/ml/gcc-patches/2014-05/msg01389.html
+template<typename T> class Future {
+public:
+  Future() : _hasValue(false) {}
+
+  void set(T &&val) {
+    assert(!_hasValue);
+    {
+      std::unique_lock<std::mutex> lock(_mutex);
+      _val = val;
+      _hasValue = true;
+    }
+    _cond.notify_all();
+  }
+
+  T &get() {
+    std::unique_lock<std::mutex> lock(_mutex);
+    if (_hasValue)
+      return _val;
+    _cond.wait(lock, [&] { return _hasValue; });
+    return _val;
+  }
+
+private:
+  T _val;
+  bool _hasValue;
+  std::mutex _mutex;
+  std::condition_variable _cond;
+};
+
+/// \brief An abstract class that takes closures and runs them asynchronously.
+class Executor {
+public:
+  virtual ~Executor() {}
+  virtual void add(std::function<void()> func) = 0;
+};
+
+/// \brief An implementation of an Executor that runs closures on a thread pool
+///   in filo order.
+class ThreadPoolExecutor : public Executor {
+public:
+  explicit ThreadPoolExecutor(unsigned threadCount =
+                                  std::thread::hardware_concurrency())
+      : _stop(false), _done(threadCount) {
+    // Spawn all but one of the threads in another thread as spawning threads
+    // can take a while.
+    std::thread([&, threadCount] {
+      for (std::size_t i = 1; i < threadCount; ++i) {
+        std::thread([=] {
+          work();
+        }).detach();
+      }
+      work();
+    }).detach();
+  }
+
+  ~ThreadPoolExecutor() {
+    std::unique_lock<std::mutex> lock(_mutex);
+    _stop = true;
+    lock.unlock();
+    _cond.notify_all();
+    // Wait for ~Latch.
+  }
+
+  void add(std::function<void()> f) override {
+    std::unique_lock<std::mutex> lock(_mutex);
+    _workStack.push(f);
+    lock.unlock();
+    _cond.notify_one();
+  }
+
+private:
+  void work() {
+    while (true) {
+      std::unique_lock<std::mutex> lock(_mutex);
+      _cond.wait(lock, [&] {
+        return _stop || !_workStack.empty();
+      });
+      if (_stop)
+        break;
+      auto task = _workStack.top();
+      _workStack.pop();
+      lock.unlock();
+      task();
+    }
+    _done.dec();
+  }
+
+  std::atomic<bool> _stop;
+  std::stack<std::function<void()>> _workStack;
+  std::mutex _mutex;
+  std::condition_variable _cond;
+  Latch _done;
+};
+
+#ifdef _MSC_VER
+/// \brief An Executor that runs tasks via ConcRT.
+class ConcRTExecutor : public Executor {
+  struct Taskish {
+    Taskish(std::function<void()> task) : _task(task) {}
+
+    std::function<void()> _task;
+
+    static void run(void *p) {
+      Taskish *self = static_cast<Taskish *>(p);
+      self->_task();
+      concurrency::Free(self);
+    }
+  };
+
+public:
+  virtual void add(std::function<void()> func) {
+    Concurrency::CurrentScheduler::ScheduleTask(Taskish::run,
+        new (concurrency::Alloc(sizeof(Taskish))) Taskish(func));
+  }
+};
+
+inline Executor *getDefaultExecutor() {
+  static ConcRTExecutor exec;
+  return &exec;
+}
+#else
+inline Executor *getDefaultExecutor() {
+  static ThreadPoolExecutor exec;
+  return &exec;
+}
+#endif
+
+/// \brief Allows launching a number of tasks and waiting for them to finish
+///   either explicitly via sync() or implicitly on destruction.
+class TaskGroup {
+  Latch _latch;
+
+public:
+  void spawn(std::function<void()> f) {
+    _latch.inc();
+    getDefaultExecutor()->add([&, f] {
+      f();
+      _latch.dec();
+    });
+  }
+
+  void sync() const { _latch.sync(); }
+};
+
+#ifdef _MSC_VER
+// Use ppl parallel_sort on Windows.
+template <class RandomAccessIterator, class Comp>
+void parallel_sort(
+    RandomAccessIterator start, RandomAccessIterator end,
+    const Comp &comp = std::less<
+        typename std::iterator_traits<RandomAccessIterator>::value_type>()) {
+  concurrency::parallel_sort(start, end, comp);
+}
+#else
+namespace detail {
+const ptrdiff_t minParallelSize = 1024;
+
+/// \brief Inclusive median.
+template <class RandomAccessIterator, class Comp>
+RandomAccessIterator medianOf3(RandomAccessIterator start,
+                               RandomAccessIterator end, const Comp &comp) {
+  RandomAccessIterator mid = start + (std::distance(start, end) / 2);
+  return comp(*start, *(end - 1))
+         ? (comp(*mid, *(end - 1)) ? (comp(*start, *mid) ? mid : start)
+                                   : end - 1)
+         : (comp(*mid, *start) ? (comp(*(end - 1), *mid) ? mid : end - 1)
+                               : start);
+}
+
+template <class RandomAccessIterator, class Comp>
+void parallel_quick_sort(RandomAccessIterator start, RandomAccessIterator end,
+                         const Comp &comp, TaskGroup &tg, size_t depth) {
+  // Do a sequential sort for small inputs.
+  if (std::distance(start, end) < detail::minParallelSize || depth == 0) {
+    std::sort(start, end, comp);
+    return;
+  }
+
+  // Partition.
+  auto pivot = medianOf3(start, end, comp);
+  // Move pivot to end.
+  std::swap(*(end - 1), *pivot);
+  pivot = std::partition(start, end - 1, [&comp, end](decltype(*start) v) {
+    return comp(v, *(end - 1));
+  });
+  // Move pivot to middle of partition.
+  std::swap(*pivot, *(end - 1));
+
+  // Recurse.
+  tg.spawn([=, &comp, &tg] {
+    parallel_quick_sort(start, pivot, comp, tg, depth - 1);
+  });
+  parallel_quick_sort(pivot + 1, end, comp, tg, depth - 1);
+}
+}
+
+template <class RandomAccessIterator, class Comp>
+void parallel_sort(
+    RandomAccessIterator start, RandomAccessIterator end,
+    const Comp &comp = std::less<
+        typename std::iterator_traits<RandomAccessIterator>::value_type>()) {
+  TaskGroup tg;
+  detail::parallel_quick_sort(start, end, comp, tg,
+                              llvm::Log2_64(std::distance(start, end)) + 1);
+}
+#endif
+
+template <class T> void parallel_sort(T *start, T *end) {
+  parallel_sort(start, end, std::less<T>());
+}
+
+#ifdef _MSC_VER
+// Use ppl parallel_for_each on Windows.
+template <class Iterator, class Func>
+void parallel_for_each(Iterator begin, Iterator end, Func func) {
+  concurrency::parallel_for_each(begin, end, func);
+}
+#else
+template <class Iterator, class Func>
+void parallel_for_each(Iterator begin, Iterator end, Func func) {
+  TaskGroup tg;
+  ptrdiff_t taskSize = 1024;
+  while (taskSize <= std::distance(begin, end)) {
+    tg.spawn([=, &func] { std::for_each(begin, begin + taskSize, func); });
+    begin += taskSize;
+  }
+  std::for_each(begin, end, func);
+}
+#endif
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/Pass.h b/include/lld/Core/Pass.h
new file mode 100644
index 000000000000..7a9d2453f482
--- /dev/null
+++ b/include/lld/Core/Pass.h
@@ -0,0 +1,46 @@
+//===------ Core/Pass.h - Base class for linker passes --------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_PASS_H
+#define LLD_CORE_PASS_H
+
+#include "lld/Core/Atom.h"
+#include "lld/Core/File.h"
+#include "lld/Core/Reference.h"
+#include "lld/Core/range.h"
+#include <vector>
+
+namespace lld {
+class MutableFile;
+
+/// Once the core linking is done (which resolves references, coalesces atoms
+/// and produces a complete Atom graph), the linker runs a series of passes
+/// on the Atom graph. The graph is modeled as a File, which means the pass
+/// has access to all the atoms and to File level attributes. Each pass does
+/// a particular transformation to the Atom graph or to the File attributes.
+///
+/// This is the abstract base class for all passes.  A Pass does its
+/// actual work in it perform() method.  It can iterator over Atoms in the
+/// graph using the *begin()/*end() atom iterator of the File.  It can add
+/// new Atoms to the graph using the File's addAtom() method.
+class Pass {
+public:
+  virtual ~Pass() { }
+
+  /// Do the actual work of the Pass.
+  virtual void perform(std::unique_ptr<MutableFile> &mergedFile) = 0;
+
+protected:
+  // Only subclassess can be instantiated.
+  Pass() { }
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_PASS_H
diff --git a/include/lld/Core/PassManager.h b/include/lld/Core/PassManager.h
new file mode 100644
index 000000000000..65fc4d806ceb
--- /dev/null
+++ b/include/lld/Core/PassManager.h
@@ -0,0 +1,46 @@
+//===- lld/Core/PassManager.h - Manage linker passes ----------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_PASS_MANAGER_H
+#define LLD_CORE_PASS_MANAGER_H
+
+#include "lld/Core/LLVM.h"
+#include "lld/Core/Pass.h"
+#include <memory>
+#include <vector>
+
+namespace lld {
+class MutableFile;
+class Pass;
+
+/// \brief Owns and runs a collection of passes.
+///
+/// This class is currently just a container for passes and a way to run them.
+///
+/// In the future this should handle timing pass runs, running parallel passes,
+/// and validate/satisfy pass dependencies.
+class PassManager {
+public:
+  void add(std::unique_ptr<Pass> pass) {
+    _passes.push_back(std::move(pass));
+  }
+
+  std::error_code runOnFile(std::unique_ptr<MutableFile> &file) {
+    for (std::unique_ptr<Pass> &pass : _passes)
+      pass->perform(file);
+    return std::error_code();
+  }
+
+private:
+  /// \brief Passes in the order they should run.
+  std::vector<std::unique_ptr<Pass>> _passes;
+};
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/Reader.h b/include/lld/Core/Reader.h
new file mode 100644
index 000000000000..ac90c5a7e85c
--- /dev/null
+++ b/include/lld/Core/Reader.h
@@ -0,0 +1,169 @@
+//===- lld/Core/Reader.h - Abstract File Format Reading Interface ---------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_READER_H
+#define LLD_CORE_READER_H
+
+#include "lld/Core/LLVM.h"
+#include "lld/Core/Reference.h"
+#include "llvm/Support/FileSystem.h"
+#include "llvm/Support/YAMLTraits.h"
+#include <functional>
+#include <memory>
+#include <vector>
+
+using llvm::sys::fs::file_magic;
+
+namespace llvm {
+namespace yaml {
+class IO;
+}
+}
+
+namespace lld {
+class ELFLinkingContext;
+class File;
+class LinkingContext;
+class PECOFFLinkingContext;
+class TargetHandlerBase;
+class MachOLinkingContext;
+
+/// \brief An abstract class for reading object files, library files, and
+/// executable files.
+///
+/// Each file format (e.g. ELF, mach-o, PECOFF, native, etc) have a concrete
+/// subclass of Reader.
+class Reader {
+public:
+  virtual ~Reader() {}
+
+  /// Sniffs the file to determine if this Reader can parse it.
+  /// The method is called with:
+  /// 1) the file_magic enumeration returned by identify_magic()
+  /// 2) the file extension (e.g. ".obj")
+  /// 3) the whole file content buffer if the above is not enough.
+  virtual bool canParse(file_magic magic, StringRef fileExtension,
+                        const MemoryBuffer &mb) const = 0;
+
+  /// \brief Parse a supplied buffer (already filled with the contents of a
+  /// file) and create a File object.
+  /// The resulting File object takes ownership of the MemoryBuffer.
+  virtual std::error_code
+  loadFile(std::unique_ptr<MemoryBuffer> mb, const class Registry &,
+           std::vector<std::unique_ptr<File>> &result) const = 0;
+};
+
+
+/// \brief An abstract class for handling alternate yaml representations
+/// of object files.
+///
+/// The YAML syntax allows "tags" which are used to specify the type of
+/// the YAML node.  In lld, top level YAML documents can be in many YAML
+/// representations (e.g mach-o encoded as yaml, etc).  A tag is used to
+/// specify which representation is used in the following YAML document.
+/// To work, there must be a YamlIOTaggedDocumentHandler registered that
+/// handles each tag type.
+class YamlIOTaggedDocumentHandler {
+public:
+  virtual ~YamlIOTaggedDocumentHandler();
+
+  /// This method is called on each registered YamlIOTaggedDocumentHandler
+  /// until one returns true.  If the subclass handles tag type !xyz, then
+  /// this method should call io.mapTag("!xzy") to see if that is the current
+  /// document type, and if so, process the rest of the document using
+  /// YAML I/O, then convert the result into an lld::File* and return it.
+  virtual bool handledDocTag(llvm::yaml::IO &io, const lld::File *&f) const = 0;
+};
+
+
+/// A registry to hold the list of currently registered Readers and
+/// tables which map Reference kind values to strings.
+/// The linker does not directly invoke Readers.  Instead, it registers
+/// Readers based on it configuration and command line options, then calls
+/// the Registry object to parse files.
+class Registry {
+public:
+  Registry();
+
+  /// Walk the list of registered Readers and find one that can parse the
+  /// supplied file and parse it.
+  std::error_code loadFile(std::unique_ptr<MemoryBuffer> mb,
+                           std::vector<std::unique_ptr<File>> &result) const;
+
+  /// Walk the list of registered kind tables to convert a Reference Kind
+  /// name to a value.
+  bool referenceKindFromString(StringRef inputStr, Reference::KindNamespace &ns,
+                               Reference::KindArch &a,
+                               Reference::KindValue &value) const;
+
+  /// Walk the list of registered kind tables to convert a Reference Kind
+  /// value to a string.
+  bool referenceKindToString(Reference::KindNamespace ns, Reference::KindArch a,
+                             Reference::KindValue value, StringRef &) const;
+
+  /// Walk the list of registered tag handlers and have the one that handles
+  /// the current document type process the yaml into an lld::File*.
+  bool handleTaggedDoc(llvm::yaml::IO &io, const lld::File *&file) const;
+
+  // These methods are called to dynamically add support for various file
+  // formats. The methods are also implemented in the appropriate lib*.a
+  // library, so that the code for handling a format is only linked in, if this
+  // method is used.  Any options that a Reader might need must be passed
+  // as parameters to the addSupport*() method.
+  void addSupportArchives(bool logLoading);
+  void addSupportYamlFiles();
+  void addSupportNativeObjects();
+  void addSupportCOFFObjects(PECOFFLinkingContext &);
+  void addSupportCOFFImportLibraries(PECOFFLinkingContext &);
+  void addSupportMachOObjects(MachOLinkingContext &);
+  void addSupportELFObjects(ELFLinkingContext &);
+  void addSupportELFDynamicSharedObjects(ELFLinkingContext &);
+
+  /// To convert between kind values and names, the registry walks the list
+  /// of registered kind tables. Each table is a zero terminated array of
+  /// KindStrings elements.
+  struct KindStrings {
+    Reference::KindValue  value;
+    StringRef             name;
+  };
+
+  /// A Reference Kind value is a tuple of <namespace, arch, value>.  All
+  /// entries in a conversion table have the same <namespace, arch>.  The
+  /// array then contains the value/name pairs.
+  void addKindTable(Reference::KindNamespace ns, Reference::KindArch arch,
+                    const KindStrings array[]);
+
+
+private:
+  struct KindEntry {
+    Reference::KindNamespace  ns;
+    Reference::KindArch       arch;
+    const KindStrings        *array;
+  };
+
+  void add(std::unique_ptr<Reader>);
+  void add(std::unique_ptr<YamlIOTaggedDocumentHandler>);
+
+  std::vector<std::unique_ptr<Reader>>                       _readers;
+  std::vector<std::unique_ptr<YamlIOTaggedDocumentHandler>>  _yamlHandlers;
+  std::vector<KindEntry>                                     _kindEntries;
+};
+
+// Utilities for building a KindString table.  For instance:
+//   static const Registry::KindStrings table[] = {
+//      LLD_KIND_STRING_ENTRY(R_VAX_ADDR16),
+//      LLD_KIND_STRING_ENTRY(R_VAX_DATA16),
+//      LLD_KIND_STRING_END
+//   };
+#define LLD_KIND_STRING_ENTRY(name) { name, #name }
+#define LLD_KIND_STRING_END         { 0,    "" }
+
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/Reference.h b/include/lld/Core/Reference.h
new file mode 100644
index 000000000000..7a804c31e182
--- /dev/null
+++ b/include/lld/Core/Reference.h
@@ -0,0 +1,125 @@
+//===- Core/References.h - A Reference to Another Atom --------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_REFERENCES_H
+#define LLD_CORE_REFERENCES_H
+
+#include "lld/Core/LLVM.h"
+#include "llvm/ADT/StringSwitch.h"
+
+namespace lld {
+class Atom;
+
+///
+/// The linker has a Graph Theory model of linking. An object file is seen
+/// as a set of Atoms with References to other Atoms.  Each Atom is a node
+/// and each Reference is an edge.
+///
+/// For example if a function contains a call site to "malloc" 40 bytes into
+/// the Atom, then the function Atom will have a Reference of: offsetInAtom=40,
+/// kind=callsite, target=malloc, addend=0.
+///
+/// Besides supporting traditional "relocations", References are also used
+/// grouping atoms (group comdat), forcing layout (one atom must follow
+/// another), marking data-in-code (jump tables or ARM constants), etc.
+///
+/// The "kind" of a reference is a tuple of <namespace, arch, value>.  This
+/// enable us to re-use existing relocation types definded for various
+/// file formats and architectures.  For instance, in ELF the relocation type 10
+/// means R_X86_64_32 for x86_64, and R_386_GOTPC for i386. For PE/COFF
+/// relocation 10 means IMAGE_REL_AMD64_SECTION.
+///
+/// References and atoms form a directed graph. The dead-stripping pass
+/// traverses them starting from dead-strip root atoms to garbage collect
+/// unreachable ones.
+///
+/// References of any kind are considered as directed edges. In addition to
+/// that, references of some kind is considered as bidirected edges.
+class Reference {
+public:
+  /// Which universe defines the kindValue().
+  enum class KindNamespace {
+    all     = 0,
+    testing = 1,
+    ELF     = 2,
+    COFF    = 3,
+    mach_o  = 4,
+  };
+
+  KindNamespace kindNamespace() const { return (KindNamespace)_kindNamespace; }
+  void setKindNamespace(KindNamespace ns) { _kindNamespace = (uint8_t)ns; }
+
+  // Which architecture the kind value is for.
+  enum class KindArch { all, AArch64, ARM, Hexagon, Mips, x86, x86_64 };
+
+  KindArch kindArch() const { return (KindArch)_kindArch; }
+  void setKindArch(KindArch a) { _kindArch = (uint8_t)a; }
+
+  typedef uint16_t KindValue;
+
+  KindValue kindValue() const { return _kindValue; }
+
+  /// setKindValue() is needed because during linking, some optimizations may
+  /// change the codegen and hence the reference kind.
+  void setKindValue(KindValue value) {
+    _kindValue = value;
+  }
+
+  /// KindValues used with KindNamespace::all and KindArch::all.
+  enum {
+    // kindLayoutAfter is treated as a bidirected edge by the dead-stripping
+    // pass.
+    kindLayoutAfter = 1,
+    // kindGroupChild is treated as a bidirected edge too.
+    kindGroupChild,
+    kindAssociate,
+  };
+
+  // A value to be added to the value of a target
+  typedef int64_t Addend;
+
+  /// If the reference is a fixup in the Atom, then this returns the
+  /// byte offset into the Atom's content to do the fix up.
+  virtual uint64_t offsetInAtom() const = 0;
+
+  /// Returns the atom this reference refers to.
+  virtual const Atom *target() const = 0;
+
+  /// During linking, the linker may merge graphs which coalesces some nodes
+  /// (i.e. Atoms).  To switch the target of a reference, this method is called.
+  virtual void setTarget(const Atom *) = 0;
+
+  /// Some relocations require a symbol and a value (e.g. foo + 4).
+  virtual Addend addend() const = 0;
+
+  /// During linking, some optimzations may change addend value.
+  virtual void setAddend(Addend) = 0;
+
+  /// Returns target specific attributes of the reference.
+  virtual uint32_t tag() const { return 0; }
+
+protected:
+  /// Reference is an abstract base class.  Only subclasses can use constructor.
+  Reference(KindNamespace ns, KindArch a, KindValue value)
+      : _kindValue(value), _kindNamespace((uint8_t)ns), _kindArch((uint8_t)a) {}
+
+  /// The memory for Reference objects is always managed by the owning File
+  /// object.  Therefore, no one but the owning File object should call
+  /// delete on an Reference.  In fact, some File objects may bulk allocate
+  /// an array of References, so they cannot be individually deleted by anyone.
+  virtual ~Reference() {}
+
+  KindValue  _kindValue;
+  uint8_t    _kindNamespace;
+  uint8_t    _kindArch;
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_REFERENCES_H
diff --git a/include/lld/Core/Resolver.h b/include/lld/Core/Resolver.h
new file mode 100644
index 000000000000..e16c07b839fa
--- /dev/null
+++ b/include/lld/Core/Resolver.h
@@ -0,0 +1,119 @@
+//===- Core/Resolver.h - Resolves Atom References -------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_RESOLVER_H
+#define LLD_CORE_RESOLVER_H
+
+#include "lld/Core/ArchiveLibraryFile.h"
+#include "lld/Core/File.h"
+#include "lld/Core/SharedLibraryFile.h"
+#include "lld/Core/Simple.h"
+#include "lld/Core/SymbolTable.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/DenseSet.h"
+#include <set>
+#include <unordered_map>
+#include <unordered_set>
+#include <vector>
+
+namespace lld {
+
+class Atom;
+class LinkingContext;
+
+/// \brief The Resolver is responsible for merging all input object files
+/// and producing a merged graph.
+class Resolver {
+public:
+  Resolver(LinkingContext &ctx)
+      : _ctx(ctx), _symbolTable(ctx), _result(new MergedFile()),
+        _fileIndex(0) {}
+
+  // InputFiles::Handler methods
+  void doDefinedAtom(const DefinedAtom&);
+  bool doUndefinedAtom(const UndefinedAtom &);
+  void doSharedLibraryAtom(const SharedLibraryAtom &);
+  void doAbsoluteAtom(const AbsoluteAtom &);
+
+  // Handle files, this adds atoms from the current file thats
+  // being processed by the resolver
+  bool handleFile(File &);
+
+  // Handle an archive library file.
+  bool handleArchiveFile(File &);
+
+  // Handle a shared library file.
+  void handleSharedLibrary(File &);
+
+  /// @brief do work of merging and resolving and return list
+  bool resolve();
+
+  std::unique_ptr<MutableFile> resultFile() { return std::move(_result); }
+
+private:
+  typedef std::function<void(StringRef, bool)> UndefCallback;
+
+  bool undefinesAdded(int begin, int end);
+  File *getFile(int &index);
+
+  /// \brief Add section group/.gnu.linkonce if it does not exist previously.
+  void maybeAddSectionGroupOrGnuLinkOnce(const DefinedAtom &atom);
+
+  /// \brief The main function that iterates over the files to resolve
+  void updatePreloadArchiveMap();
+  bool resolveUndefines();
+  void updateReferences();
+  void deadStripOptimize();
+  bool checkUndefines();
+  void removeCoalescedAwayAtoms();
+  void checkDylibSymbolCollisions();
+  void forEachUndefines(File &file, bool searchForOverrides, UndefCallback callback);
+
+  void markLive(const Atom *atom);
+  void addAtoms(const std::vector<const DefinedAtom *>&);
+  void maybePreloadArchiveMember(StringRef sym);
+
+  class MergedFile : public SimpleFile {
+  public:
+    MergedFile() : SimpleFile("<linker-internal>") {}
+    void addAtoms(std::vector<const Atom*>& atoms);
+  };
+
+  LinkingContext &_ctx;
+  SymbolTable _symbolTable;
+  std::vector<const Atom *>     _atoms;
+  std::set<const Atom *>        _deadStripRoots;
+  llvm::DenseSet<const Atom *>  _liveAtoms;
+  llvm::DenseSet<const Atom *>  _deadAtoms;
+  std::unique_ptr<MergedFile>   _result;
+  std::unordered_multimap<const Atom *, const Atom *> _reverseRef;
+
+  // --start-group and --end-group
+  std::vector<File *> _files;
+  std::map<File *, bool> _newUndefinesAdded;
+  size_t _fileIndex;
+
+  // Preloading
+  llvm::StringMap<ArchiveLibraryFile *> _archiveMap;
+  llvm::DenseSet<ArchiveLibraryFile *> _archiveSeen;
+
+  // List of undefined symbols.
+  std::vector<StringRef> _undefines;
+
+  // Start position in _undefines for each archive/shared library file.
+  // Symbols from index 0 to the start position are already searched before.
+  // Searching them again would never succeed. When we look for undefined
+  // symbols from an archive/shared library file, start from its start
+  // position to save time.
+  std::map<File *, size_t> _undefineIndex;
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_RESOLVER_H
diff --git a/include/lld/Core/STDExtras.h b/include/lld/Core/STDExtras.h
new file mode 100644
index 000000000000..4a6183891844
--- /dev/null
+++ b/include/lld/Core/STDExtras.h
@@ -0,0 +1,29 @@
+//===- lld/Core/STDExtra.h - Helpers for the stdlib -----------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_STD_EXTRA_H
+#define LLD_CORE_STD_EXTRA_H
+
+namespace lld {
+/// \brief Deleter for smart pointers that only calls the destructor. Memory is
+/// managed elsewhere. A common use of this is for things allocated with a
+/// BumpPtrAllocator.
+template <class T>
+struct destruct_delete {
+  void operator ()(T *ptr) {
+    ptr->~T();
+  }
+};
+
+template <class T>
+using unique_bump_ptr = std::unique_ptr<T, destruct_delete<T>>;
+
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/SharedLibraryAtom.h b/include/lld/Core/SharedLibraryAtom.h
new file mode 100644
index 000000000000..1b0c37c41138
--- /dev/null
+++ b/include/lld/Core/SharedLibraryAtom.h
@@ -0,0 +1,53 @@
+//===- Core/SharedLibraryAtom.h - A Shared Library Atom -------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_SHARED_LIBRARY_ATOM_H
+#define LLD_CORE_SHARED_LIBRARY_ATOM_H
+
+#include "lld/Core/Atom.h"
+
+namespace lld {
+
+/// A SharedLibraryAtom has no content.
+/// It exists to represent a symbol which will be bound at runtime.
+class SharedLibraryAtom : public Atom {
+public:
+  enum class Type : uint32_t {
+    Unknown,
+    Code,
+    Data,
+  };
+
+  /// Returns shared library name used to load it at runtime.
+  /// On linux that is the DT_NEEDED name.
+  /// On Darwin it is the LC_DYLIB_LOAD dylib name.
+  /// On Windows it is the DLL name that to be referred from .idata section.
+  virtual StringRef loadName() const = 0;
+
+  /// Returns if shared library symbol can be missing at runtime and if
+  /// so the loader should silently resolve address of symbol to be nullptr.
+  virtual bool canBeNullAtRuntime() const = 0;
+
+  virtual Type type() const = 0;
+
+  virtual uint64_t size() const = 0;
+
+  static bool classof(const Atom *a) {
+    return a->definition() == definitionSharedLibrary;
+  }
+
+  static inline bool classof(const SharedLibraryAtom *) { return true; }
+
+protected:
+  SharedLibraryAtom() : Atom(definitionSharedLibrary) {}
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_SHARED_LIBRARY_ATOM_H
diff --git a/include/lld/Core/SharedLibraryFile.h b/include/lld/Core/SharedLibraryFile.h
new file mode 100644
index 000000000000..2f84624287d8
--- /dev/null
+++ b/include/lld/Core/SharedLibraryFile.h
@@ -0,0 +1,65 @@
+//===- Core/SharedLibraryFile.h - Models shared libraries as Atoms --------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_SHARED_LIBRARY_FILE_H
+#define LLD_CORE_SHARED_LIBRARY_FILE_H
+
+#include "lld/Core/File.h"
+
+namespace lld {
+
+///
+/// The SharedLibraryFile subclass of File is used to represent dynamic
+/// shared libraries being linked against.
+///
+class SharedLibraryFile : public File {
+public:
+  static bool classof(const File *f) {
+    return f->kind() == kindSharedLibrary;
+  }
+
+  /// Check if the shared library exports a symbol with the specified name.
+  /// If so, return a SharedLibraryAtom which represents that exported
+  /// symbol.  Otherwise return nullptr.
+  virtual const SharedLibraryAtom *exports(StringRef name,
+                                           bool dataSymbolOnly) const = 0;
+
+  // Returns DSO name. It's the soname (ELF), the install name (MachO) or
+  // the import name (Windows).
+  virtual StringRef getDSOName() const = 0;
+
+  const atom_collection<DefinedAtom> &defined() const override {
+    return _definedAtoms;
+  }
+
+  const atom_collection<UndefinedAtom> &undefined() const override {
+    return _undefinedAtoms;
+  }
+
+  const atom_collection<SharedLibraryAtom> &sharedLibrary() const override {
+    return _sharedLibraryAtoms;
+  }
+
+  const atom_collection<AbsoluteAtom> &absolute() const override {
+    return _absoluteAtoms;
+  }
+
+protected:
+  /// only subclasses of SharedLibraryFile can be instantiated
+  explicit SharedLibraryFile(StringRef path) : File(path, kindSharedLibrary) {}
+
+  atom_collection_vector<DefinedAtom>        _definedAtoms;
+  atom_collection_vector<UndefinedAtom>      _undefinedAtoms;
+  atom_collection_vector<SharedLibraryAtom>  _sharedLibraryAtoms;
+  atom_collection_vector<AbsoluteAtom>       _absoluteAtoms;
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_SHARED_LIBRARY_FILE_H
diff --git a/include/lld/Core/Simple.h b/include/lld/Core/Simple.h
new file mode 100644
index 000000000000..71d0c0702301
--- /dev/null
+++ b/include/lld/Core/Simple.h
@@ -0,0 +1,341 @@
+//===- lld/Core/Simple.h - Simple implementations of Atom and File --------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// \brief Provide simple implementations for Atoms and File.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_SIMPLE_H
+#define LLD_CORE_SIMPLE_H
+
+#include "lld/Core/DefinedAtom.h"
+#include "lld/Core/File.h"
+#include "lld/Core/ArchiveLibraryFile.h"
+#include "lld/Core/LinkingContext.h"
+#include "lld/Core/Reference.h"
+#include "lld/Core/UndefinedAtom.h"
+#include "llvm/ADT/ilist.h"
+#include "llvm/ADT/ilist_node.h"
+
+namespace lld {
+
+class SimpleFile : public MutableFile {
+public:
+  SimpleFile(StringRef path) : MutableFile(path) {}
+
+  void addAtom(const Atom &atom) override {
+    if (auto *defAtom = dyn_cast<DefinedAtom>(&atom)) {
+      _definedAtoms._atoms.push_back(defAtom);
+    } else if (auto *undefAtom = dyn_cast<UndefinedAtom>(&atom)) {
+      _undefinedAtoms._atoms.push_back(undefAtom);
+    } else if (auto *shlibAtom = dyn_cast<SharedLibraryAtom>(&atom)) {
+      _sharedLibraryAtoms._atoms.push_back(shlibAtom);
+    } else if (auto *absAtom = dyn_cast<AbsoluteAtom>(&atom)) {
+      _absoluteAtoms._atoms.push_back(absAtom);
+    } else {
+      llvm_unreachable("atom has unknown definition kind");
+    }
+  }
+
+  void
+  removeDefinedAtomsIf(std::function<bool(const DefinedAtom *)> pred) override {
+    auto &atoms = _definedAtoms._atoms;
+    auto newEnd = std::remove_if(atoms.begin(), atoms.end(), pred);
+    atoms.erase(newEnd, atoms.end());
+  }
+
+  const atom_collection<DefinedAtom> &defined() const override {
+    return _definedAtoms;
+  }
+
+  const atom_collection<UndefinedAtom> &undefined() const override {
+    return _undefinedAtoms;
+  }
+
+  const atom_collection<SharedLibraryAtom> &sharedLibrary() const override {
+    return _sharedLibraryAtoms;
+  }
+
+  const atom_collection<AbsoluteAtom> &absolute() const override {
+    return _absoluteAtoms;
+  }
+
+  DefinedAtomRange definedAtoms() override {
+    return make_range(_definedAtoms._atoms);
+  }
+
+private:
+  atom_collection_vector<DefinedAtom>        _definedAtoms;
+  atom_collection_vector<UndefinedAtom>      _undefinedAtoms;
+  atom_collection_vector<SharedLibraryAtom>  _sharedLibraryAtoms;
+  atom_collection_vector<AbsoluteAtom>       _absoluteAtoms;
+};
+
+/// \brief Archive library file that may be used as a virtual container
+/// for symbols that should be added dynamically in response to
+/// call to find() method.
+class SimpleArchiveLibraryFile : public ArchiveLibraryFile {
+public:
+  SimpleArchiveLibraryFile(StringRef filename)
+      : ArchiveLibraryFile(filename) {}
+
+  const atom_collection<DefinedAtom> &defined() const override {
+    return _definedAtoms;
+  }
+
+  const atom_collection<UndefinedAtom> &undefined() const override {
+    return _undefinedAtoms;
+  }
+
+  const atom_collection<SharedLibraryAtom> &sharedLibrary() const override {
+    return _sharedLibraryAtoms;
+  }
+
+  const atom_collection<AbsoluteAtom> &absolute() const override {
+    return _absoluteAtoms;
+  }
+
+  File *find(StringRef sym, bool dataSymbolOnly) override {
+    // For descendants:
+    // do some checks here and return dynamically generated files with atoms.
+    return nullptr;
+  }
+
+  std::error_code
+  parseAllMembers(std::vector<std::unique_ptr<File>> &result) override {
+    return std::error_code();
+  }
+
+private:
+  atom_collection_vector<DefinedAtom> _definedAtoms;
+  atom_collection_vector<UndefinedAtom> _undefinedAtoms;
+  atom_collection_vector<SharedLibraryAtom> _sharedLibraryAtoms;
+  atom_collection_vector<AbsoluteAtom> _absoluteAtoms;
+};
+
+class SimpleReference : public Reference {
+public:
+  SimpleReference(Reference::KindNamespace ns, Reference::KindArch arch,
+                  Reference::KindValue value, uint64_t off, const Atom *t,
+                  Reference::Addend a)
+      : Reference(ns, arch, value), _target(t), _offsetInAtom(off), _addend(a),
+        _next(nullptr), _prev(nullptr) {
+  }
+  SimpleReference()
+      : Reference(Reference::KindNamespace::all, Reference::KindArch::all, 0),
+        _target(nullptr), _offsetInAtom(0), _addend(0), _next(nullptr),
+        _prev(nullptr) {
+  }
+
+  uint64_t offsetInAtom() const override { return _offsetInAtom; }
+
+  const Atom *target() const override {
+    assert(_target);
+    return _target;
+  }
+
+  Addend addend() const override { return _addend; }
+  void setAddend(Addend a) override { _addend = a; }
+  void setTarget(const Atom *newAtom) override { _target = newAtom; }
+  SimpleReference *getNext() const { return _next; }
+  SimpleReference *getPrev() const { return _prev; }
+  void setNext(SimpleReference *n) { _next = n; }
+  void setPrev(SimpleReference *p) { _prev = p; }
+
+private:
+  const Atom *_target;
+  uint64_t _offsetInAtom;
+  Addend _addend;
+  SimpleReference *_next;
+  SimpleReference *_prev;
+};
+
+}
+
+// ilist will lazily create a sentinal (so end() can return a node past the
+// end of the list). We need this trait so that the sentinal is allocated
+// via the BumpPtrAllocator.
+namespace llvm {
+template<>
+struct ilist_sentinel_traits<lld::SimpleReference> {
+
+  ilist_sentinel_traits() : _allocator(nullptr) { }
+
+  void setAllocator(llvm::BumpPtrAllocator *alloc) {
+    _allocator = alloc;
+  }
+
+  lld::SimpleReference *createSentinel() const {
+    return new (*_allocator) lld::SimpleReference();
+  }
+
+  static void destroySentinel(lld::SimpleReference*) {}
+
+  static lld::SimpleReference *provideInitialHead() { return nullptr; }
+
+  lld::SimpleReference *ensureHead(lld::SimpleReference *&head) const {
+    if (!head) {
+      head = createSentinel();
+      noteHead(head, head);
+      ilist_traits<lld::SimpleReference>::setNext(head, nullptr);
+      return head;
+    }
+    return ilist_traits<lld::SimpleReference>::getPrev(head);
+  }
+
+  void noteHead(lld::SimpleReference *newHead,
+                lld::SimpleReference *sentinel) const {
+    ilist_traits<lld::SimpleReference>::setPrev(newHead, sentinel);
+  }
+
+private:
+  mutable llvm::BumpPtrAllocator *_allocator;
+};
+}
+
+namespace lld {
+
+class SimpleDefinedAtom : public DefinedAtom {
+public:
+  explicit SimpleDefinedAtom(const File &f) : _file(f) {
+    static uint32_t lastOrdinal = 0;
+    _ordinal = lastOrdinal++;
+    _references.setAllocator(&f.allocator());
+  }
+
+  const File &file() const override { return _file; }
+
+  StringRef name() const override { return StringRef(); }
+
+  uint64_t ordinal() const override { return _ordinal; }
+
+  Scope scope() const override { return DefinedAtom::scopeLinkageUnit; }
+
+  Interposable interposable() const override {
+    return DefinedAtom::interposeNo;
+  }
+
+  Merge merge() const override { return DefinedAtom::mergeNo; }
+
+  Alignment alignment() const override { return Alignment(0, 0); }
+
+  SectionChoice sectionChoice() const override {
+    return DefinedAtom::sectionBasedOnContent;
+  }
+
+  StringRef customSectionName() const override { return StringRef(); }
+  DeadStripKind deadStrip() const override {
+    return DefinedAtom::deadStripNormal;
+  }
+
+  DefinedAtom::reference_iterator begin() const override {
+    const void *it = reinterpret_cast<const void *>(&*_references.begin());
+    return reference_iterator(*this, it);
+  }
+
+  DefinedAtom::reference_iterator end() const override {
+    const void *it = reinterpret_cast<const void *>(&*_references.end());
+    return reference_iterator(*this, it);
+  }
+
+  const Reference *derefIterator(const void *it) const override {
+    return reinterpret_cast<const Reference*>(it);
+  }
+
+  void incrementIterator(const void *&it) const override {
+    const SimpleReference* node = reinterpret_cast<const SimpleReference*>(it);
+    const SimpleReference* next = node->getNext();
+    it = reinterpret_cast<const void*>(next);
+  }
+
+  void addReference(Reference::KindNamespace ns, Reference::KindArch arch,
+                    Reference::KindValue kindValue, uint64_t off,
+                    const Atom *target, Reference::Addend a) {
+    assert(target && "trying to create reference to nothing");
+    auto node = new (_file.allocator())
+        SimpleReference(ns, arch, kindValue, off, target, a);
+    _references.push_back(node);
+  }
+
+  /// Sort references in a canonical order (by offset, then by kind).
+  void sortReferences() const {
+    // Cannot sort a linked  list, so move elements into a temporary vector,
+    // sort the vector, then reconstruct the list.
+    llvm::SmallVector<SimpleReference *, 16> elements;
+    for (SimpleReference &node : _references) {
+      elements.push_back(&node);
+    }
+    std::sort(elements.begin(), elements.end(),
+        [] (const SimpleReference *lhs, const SimpleReference *rhs) -> bool {
+          uint64_t lhsOffset = lhs->offsetInAtom();
+          uint64_t rhsOffset = rhs->offsetInAtom();
+          if (rhsOffset != lhsOffset)
+            return (lhsOffset < rhsOffset);
+          if (rhs->kindNamespace() != lhs->kindNamespace())
+            return (lhs->kindNamespace() < rhs->kindNamespace());
+          if (rhs->kindArch() != lhs->kindArch())
+            return (lhs->kindArch() < rhs->kindArch());
+          return (lhs->kindValue() < rhs->kindValue());
+        });
+    _references.clearAndLeakNodesUnsafely();
+    for (SimpleReference *node : elements) {
+      _references.push_back(node);
+    }
+  }
+  void setOrdinal(uint64_t ord) { _ordinal = ord; }
+
+private:
+  typedef llvm::ilist<SimpleReference> RefList;
+
+  const File &_file;
+  uint64_t _ordinal;
+  mutable RefList _references;
+};
+
+class SimpleUndefinedAtom : public UndefinedAtom {
+public:
+  SimpleUndefinedAtom(const File &f, StringRef name) : _file(f), _name(name) {
+    assert(!name.empty() && "UndefinedAtoms must have a name");
+  }
+
+  /// file - returns the File that produced/owns this Atom
+  const File &file() const override { return _file; }
+
+  /// name - The name of the atom. For a function atom, it is the (mangled)
+  /// name of the function.
+  StringRef name() const override { return _name; }
+
+  CanBeNull canBeNull() const override { return UndefinedAtom::canBeNullNever; }
+
+private:
+  const File &_file;
+  StringRef _name;
+};
+
+class SimpleAbsoluteAtom : public AbsoluteAtom {
+public:
+  SimpleAbsoluteAtom(const File &f, StringRef name, Scope s, uint64_t value)
+      : _file(f), _name(name), _scope(s), _value(value) {}
+
+  const File &file() const override { return _file; }
+  StringRef name() const override { return _name; }
+  uint64_t value() const override { return _value; }
+  Scope scope() const override { return _scope; }
+
+private:
+  const File &_file;
+  StringRef _name;
+  Scope _scope;
+  uint64_t _value;
+};
+
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/SymbolTable.h b/include/lld/Core/SymbolTable.h
new file mode 100644
index 000000000000..683ed65e3635
--- /dev/null
+++ b/include/lld/Core/SymbolTable.h
@@ -0,0 +1,117 @@
+//===- Core/SymbolTable.h - Main Symbol Table -----------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_SYMBOL_TABLE_H
+#define LLD_CORE_SYMBOL_TABLE_H
+
+#include "lld/Core/LLVM.h"
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/StringExtras.h"
+#include <cstring>
+#include <map>
+#include <vector>
+
+namespace lld {
+
+class AbsoluteAtom;
+class Atom;
+class DefinedAtom;
+class LinkingContext;
+class ResolverOptions;
+class SharedLibraryAtom;
+class UndefinedAtom;
+
+/// \brief The SymbolTable class is responsible for coalescing atoms.
+///
+/// All atoms coalescable by-name or by-content should be added.
+/// The method replacement() can be used to find the replacement atom
+/// if an atom has been coalesced away.
+class SymbolTable {
+public:
+  explicit SymbolTable(LinkingContext &);
+
+  /// @brief add atom to symbol table
+  bool add(const DefinedAtom &);
+
+  /// @brief add atom to symbol table
+  bool add(const UndefinedAtom &);
+
+  /// @brief add atom to symbol table
+  bool add(const SharedLibraryAtom &);
+
+  /// @brief add atom to symbol table
+  bool add(const AbsoluteAtom &);
+
+  /// @brief checks if name is in symbol table and if so atom is not
+  ///        UndefinedAtom
+  bool isDefined(StringRef sym);
+
+  /// @brief returns atom in symbol table for specified name (or nullptr)
+  const Atom *findByName(StringRef sym);
+
+  /// @brief returns vector of remaining UndefinedAtoms
+  std::vector<const UndefinedAtom *> undefines();
+
+  /// returns vector of tentative definitions
+  std::vector<StringRef> tentativeDefinitions();
+
+  /// @brief add atom to replacement table
+  void addReplacement(const Atom *replaced, const Atom *replacement);
+
+  /// @brief if atom has been coalesced away, return replacement, else return atom
+  const Atom *replacement(const Atom *);
+
+  /// @brief if atom has been coalesced away, return true
+  bool isCoalescedAway(const Atom *);
+
+  /// @brief Find a group atom.
+  const Atom *findGroup(StringRef name);
+
+  /// @brief Add a group atom and returns true/false depending on whether the
+  /// previously existed.
+  bool addGroup(const DefinedAtom &da);
+
+private:
+  typedef llvm::DenseMap<const Atom *, const Atom *> AtomToAtom;
+
+  struct StringRefMappingInfo {
+    static StringRef getEmptyKey() { return StringRef(); }
+    static StringRef getTombstoneKey() { return StringRef(" ", 1); }
+    static unsigned getHashValue(StringRef const val) {
+      return llvm::HashString(val);
+    }
+    static bool isEqual(StringRef const lhs, StringRef const rhs) {
+      return lhs.equals(rhs);
+    }
+  };
+  typedef llvm::DenseMap<StringRef, const Atom *,
+                                           StringRefMappingInfo> NameToAtom;
+
+  struct AtomMappingInfo {
+    static const DefinedAtom * getEmptyKey() { return nullptr; }
+    static const DefinedAtom * getTombstoneKey() { return (DefinedAtom*)(-1); }
+    static unsigned getHashValue(const DefinedAtom * const Val);
+    static bool isEqual(const DefinedAtom * const LHS,
+                        const DefinedAtom * const RHS);
+  };
+  typedef llvm::DenseSet<const DefinedAtom*, AtomMappingInfo> AtomContentSet;
+
+  bool addByName(const Atom &);
+  bool addByContent(const DefinedAtom &);
+
+  LinkingContext &_context;
+  AtomToAtom _replacedAtoms;
+  NameToAtom _nameTable;
+  NameToAtom _groupTable;
+  AtomContentSet _contentTable;
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_SYMBOL_TABLE_H
diff --git a/include/lld/Core/TODO.txt b/include/lld/Core/TODO.txt
new file mode 100644
index 000000000000..8888c763ef65
--- /dev/null
+++ b/include/lld/Core/TODO.txt
@@ -0,0 +1,17 @@
+include/lld/Core
+~~~~~~~~~~~~~~~~
+
+* The native/yaml reader/writer interfaces should be changed to return
+  an explanatory string if there is an error.  The existing error_code
+  abstraction only works for returning low level OS errors.  It does not
+  work for describing formatting issues.
+
+* We need to design a diagnostics interface.  It would be nice to share code
+  with Clang_ where possible.
+
+* We need to add more attributes to File.  In particular, we need cpu
+  and OS information (like target triples).  We should also provide explicit
+  support for `LLVM IR module flags metadata`__.
+
+.. __: http://llvm.org/docs/LangRef.html#module_flags
+.. _Clang: http://clang.llvm.org/docs/InternalsManual.html#Diagnostics
diff --git a/include/lld/Core/UndefinedAtom.h b/include/lld/Core/UndefinedAtom.h
new file mode 100644
index 000000000000..7a835a4ebaa8
--- /dev/null
+++ b/include/lld/Core/UndefinedAtom.h
@@ -0,0 +1,74 @@
+//===- Core/UndefinedAtom.h - An Undefined Atom ---------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_UNDEFINED_ATOM_H
+#define LLD_CORE_UNDEFINED_ATOM_H
+
+#include "lld/Core/Atom.h"
+
+namespace lld {
+
+/// An UndefinedAtom has no content.
+/// It exists as a placeholder for a future atom.
+class UndefinedAtom : public Atom {
+public:
+  /// Whether this undefined symbol needs to be resolved,
+  /// or whether it can just evaluate to nullptr.
+  /// This concept is often called "weak", but that term
+  /// is overloaded to mean other things too.
+  enum CanBeNull {
+    /// Normal symbols must be resolved at build time
+    canBeNullNever,
+
+    /// This symbol can be missing at runtime and will evalute to nullptr.
+    /// That is, the static linker still must find a definition (usually
+    /// is some shared library), but at runtime, the dynamic loader
+    /// will allow the symbol to be missing and resolved to nullptr.
+    ///
+    /// On Darwin this is generated using a function prototype with
+    /// __attribute__((weak_import)).
+    /// On linux this is generated using a function prototype with
+    ///  __attribute__((weak)).
+    /// On Windows this feature is not supported.
+    canBeNullAtRuntime,
+
+    /// This symbol can be missing at build time.
+    /// That is, the static linker will not error if a definition for
+    /// this symbol is not found at build time. Instead, the linker
+    /// will build an executable that lets the dynamic loader find the
+    /// symbol at runtime.
+    /// This feature is not supported on Darwin nor Windows.
+    /// On linux this is generated using a function prototype with
+    ///  __attribute__((weak)).
+    canBeNullAtBuildtime
+  };
+
+  virtual CanBeNull canBeNull() const = 0;
+
+  static bool classof(const Atom *a) {
+    return a->definition() == definitionUndefined;
+  }
+
+  static bool classof(const UndefinedAtom *) { return true; }
+
+  /// Returns an undefined atom if this undefined symbol has a synonym.  This is
+  /// mainly used in COFF. In COFF, an unresolved external symbol can have up to
+  /// one optional name (sym2) in addition to its regular name (sym1). If a
+  /// definition of sym1 exists, sym1 is resolved normally. Otherwise, all
+  /// references to sym1 refer to sym2 instead. In that case sym2 must be
+  /// resolved, or link will fail.
+  virtual const UndefinedAtom *fallback() const { return nullptr; }
+
+protected:
+  UndefinedAtom() : Atom(definitionUndefined) {}
+};
+
+} // namespace lld
+
+#endif // LLD_CORE_UNDEFINED_ATOM_H
diff --git a/include/lld/Core/Writer.h b/include/lld/Core/Writer.h
new file mode 100644
index 000000000000..94c75d8d019f
--- /dev/null
+++ b/include/lld/Core/Writer.h
@@ -0,0 +1,52 @@
+//===- lld/Core/Writer.h - Abstract File Format Interface -----------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_WRITER_H
+#define LLD_CORE_WRITER_H
+
+#include "lld/Core/LLVM.h"
+#include <memory>
+#include <vector>
+
+namespace lld {
+class File;
+class ELFLinkingContext;
+class MachOLinkingContext;
+class PECOFFLinkingContext;
+class LinkingContext;
+class TargetHandlerBase;
+
+/// \brief The Writer is an abstract class for writing object files, shared
+/// library files, and executable files.  Each file format (e.g. ELF, mach-o,
+/// PECOFF, native, etc) have a concrete subclass of Writer.
+class Writer {
+public:
+  virtual ~Writer();
+
+  /// \brief Write a file from the supplied File object
+  virtual std::error_code writeFile(const File &linkedFile, StringRef path) = 0;
+
+  /// \brief This method is called by Core Linking to give the Writer a chance
+  /// to add file format specific "files" to set of files to be linked. This is
+  /// how file format specific atoms can be added to the link.
+  virtual bool createImplicitFiles(std::vector<std::unique_ptr<File> > &);
+
+protected:
+  // only concrete subclasses can be instantiated
+  Writer();
+};
+
+std::unique_ptr<Writer> createWriterELF(TargetHandlerBase *handler);
+std::unique_ptr<Writer> createWriterMachO(const MachOLinkingContext &);
+std::unique_ptr<Writer> createWriterPECOFF(const PECOFFLinkingContext &);
+std::unique_ptr<Writer> createWriterNative();
+std::unique_ptr<Writer> createWriterYAML(const LinkingContext &);
+} // end namespace lld
+
+#endif
diff --git a/include/lld/Core/range.h b/include/lld/Core/range.h
new file mode 100644
index 000000000000..614c9672955c
--- /dev/null
+++ b/include/lld/Core/range.h
@@ -0,0 +1,738 @@
+//===-- lld/Core/range.h - Iterator ranges ----------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// \brief Iterator range type based on c++1y range proposal.
+///
+/// See http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3350.html
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_RANGE_H
+#define LLD_CORE_RANGE_H
+
+#include "llvm/Support/Compiler.h"
+#include <array>
+#include <cassert>
+#include <iterator>
+#include <string>
+#include <type_traits>
+#include <utility>
+#include <vector>
+
+namespace lld {
+// Nothing in this namespace is part of the exported interface.
+namespace detail {
+using std::begin;
+using std::end;
+/// Used as the result type of undefined functions.
+struct undefined {};
+
+template <typename R> class begin_result {
+  template <typename T> static auto check(T &&t) -> decltype(begin(t));
+  static undefined check(...);
+public:
+  typedef decltype(check(std::declval<R>())) type;
+};
+
+template <typename R> class end_result {
+  template <typename T> static auto check(T &&t) -> decltype(end(t));
+  static undefined check(...);
+public:
+  typedef decltype(check(std::declval<R>())) type;
+};
+
+// Things that begin and end work on, in compatible ways, are
+// ranges. [stmt.ranged]
+template <typename R>
+struct is_range : std::is_same<typename detail::begin_result<R>::type,
+                               typename detail::end_result<R>::type> {};
+
+// This currently requires specialization and doesn't work for
+// detecting \c range<>s or iterators.  We should add
+// \c contiguous_iterator_tag to fix that.
+template <typename R> struct is_contiguous_range : std::false_type {};
+template <typename R>
+struct is_contiguous_range<R &> : is_contiguous_range<R> {};
+template <typename R>
+struct is_contiguous_range <R &&> : is_contiguous_range<R> {};
+template <typename R>
+struct is_contiguous_range<const R> : is_contiguous_range<R> {};
+
+template <typename T, size_t N>
+struct is_contiguous_range<T[N]> : std::true_type {};
+template <typename T, size_t N>
+struct is_contiguous_range<const T[N]> : std::true_type {};
+template <typename T, size_t N>
+struct is_contiguous_range<std::array<T, N> > : std::true_type {};
+template <typename charT, typename traits, typename Allocator>
+struct is_contiguous_range<
+    std::basic_string<charT, traits, Allocator> > : std::true_type {};
+template <typename T, typename Allocator>
+struct is_contiguous_range<std::vector<T, Allocator> > : std::true_type {};
+
+// Removes cv qualifiers from all levels of a multi-level pointer
+// type, not just the type level.
+template <typename T> struct remove_all_cv_ptr {
+  typedef T type;
+};
+template <typename T> struct remove_all_cv_ptr<T *> {
+  typedef typename remove_all_cv_ptr<T>::type *type;
+};
+template <typename T> struct remove_all_cv_ptr<const T> {
+  typedef typename remove_all_cv_ptr<T>::type type;
+};
+template <typename T> struct remove_all_cv_ptr<volatile T> {
+  typedef typename remove_all_cv_ptr<T>::type type;
+};
+template <typename T> struct remove_all_cv_ptr<const volatile T> {
+  typedef typename remove_all_cv_ptr<T>::type type;
+};
+
+template <typename From, typename To>
+struct conversion_preserves_array_indexing : std::false_type {};
+
+template <typename FromVal, typename ToVal>
+struct conversion_preserves_array_indexing<FromVal *,
+                                           ToVal *> : std::integral_constant<
+    bool, std::is_convertible<FromVal *, ToVal *>::value &&
+    std::is_same<typename remove_all_cv_ptr<FromVal>::type,
+                 typename remove_all_cv_ptr<ToVal>::type>::value> {};
+
+template <typename T>
+LLVM_CONSTEXPR auto adl_begin(T &&t) -> decltype(begin(t)) {
+  return begin(std::forward<T>(t));
+}
+
+template <typename T> LLVM_CONSTEXPR auto adl_end(T &&t) -> decltype(end(t)) {
+  return end(std::forward<T>(t));
+}
+} // end namespace detail
+
+/// A \c std::range<Iterator> represents a half-open iterator range
+/// built from two iterators, \c 'begin', and \c 'end'.  If \c end is
+/// not reachable from \c begin, the behavior is undefined.
+///
+/// The mutability of elements of the range is controlled by the
+/// Iterator argument.  Instantiate
+/// <code>range<<var>Foo</var>::iterator></code> or
+/// <code>range<<var>T</var>*></code>, or call
+/// <code>make_range(<var>non_const_container</var>)</code>, and you
+/// get a mutable range.  Instantiate
+/// <code>range<<var>Foo</var>::const_iterator></code> or
+/// <code>range<const <var>T</var>*></code>, or call
+/// <code>make_range(<var>const_container</var>)</code>, and you get a
+/// constant range.
+///
+/// \todo Inherit from std::pair<Iterator, Iterator>?
+///
+/// \todo This interface contains some functions that could be
+/// provided as free algorithms rather than member functions, and all
+/// of the <code>pop_*()</code> functions could be replaced by \c
+/// slice() at the cost of some extra iterator copies.  This makes
+/// them more awkward to use, but makes it easier for users to write
+/// their own types that follow the same interface. On the other hand,
+/// a \c range_facade could be provided to help users write new
+/// ranges, and it could provide the members.  Such functions are
+/// marked with a note in their documentation.  (Of course, all of
+/// these member functions could be provided as free functions using
+/// the iterator access methods, but one goal here is to allow people
+/// to program without touching iterators at all.)
+template <typename Iterator> class range {
+  Iterator begin_, end_;
+public:
+  /// \name types
+  /// @{
+
+  /// The iterator category of \c Iterator.
+  /// \todo Consider defining range categories. If they don't add
+  /// anything over the corresponding iterator categories, then
+  /// they're probably not worth defining.
+  typedef typename std::iterator_traits<
+      Iterator>::iterator_category iterator_category;
+  /// The type of elements of the range. Not cv-qualified.
+  typedef typename std::iterator_traits<Iterator>::value_type value_type;
+  /// The type of the size of the range and offsets within the range.
+  typedef typename std::iterator_traits<
+      Iterator>::difference_type difference_type;
+  /// The return type of element access methods: \c front(), \c back(), etc.
+  typedef typename std::iterator_traits<Iterator>::reference reference;
+  typedef typename std::iterator_traits<Iterator>::pointer pointer;
+  /// @}
+
+  /// \name constructors
+  /// @{
+
+  /// Creates a range of default-constructed (<em>not</em>
+  /// value-initialized) iterators.  For most \c Iterator types, this
+  /// will be an invalid range.
+  range() : begin_(), end_() {}
+
+  /// \pre \c end is reachable from \c begin.
+  /// \post <code>this->begin() == begin && this->end() == end</code>
+  LLVM_CONSTEXPR range(Iterator begin, Iterator end)
+      : begin_(begin), end_(end) {}
+
+  /// \par Participates in overload resolution if:
+  ///   - \c Iterator is not a pointer type,
+  ///   - \c begin(r) and \c end(r) return the same type, and
+  ///   - that type is convertible to \c Iterator.
+  ///
+  /// \todo std::begin and std::end are overloaded between T& and
+  /// const T&, which means that if a container has only a non-const
+  /// begin or end method, then it's ill-formed to pass an rvalue to
+  /// the free function.  To avoid that problem, we don't use
+  /// std::forward<> here, so begin() and end() are always called with
+  /// an lvalue.  Another option would be to insist that rvalue
+  /// arguments to range() must have const begin() and end() methods.
+  template <typename R> LLVM_CONSTEXPR range(
+      R &&r,
+      typename std::enable_if<
+        !std::is_pointer<Iterator>::value &&
+        detail::is_range<R>::value &&
+        std::is_convertible<typename detail::begin_result<R>::type,
+                            Iterator>::value>::type* = 0)
+      : begin_(detail::adl_begin(r)), end_(detail::adl_end(r)) {}
+
+  /// This constructor creates a \c range<T*> from any range with
+  /// contiguous iterators. Because dereferencing a past-the-end
+  /// iterator can be undefined behavior, empty ranges get initialized
+  /// with \c nullptr rather than \c &*begin().
+  ///
+  /// \par Participates in overload resolution if:
+  ///   - \c Iterator is a pointer type \c T*,
+  ///   - \c begin(r) and \c end(r) return the same type,
+  ///   - elements \c i of that type satisfy the invariant
+  ///     <code>&*(i + N) == (&*i) + N</code>, and
+  ///   - The result of <code>&*begin()</code> is convertible to \c T*
+  ///     using only qualification conversions [conv.qual] (since
+  ///     pointer conversions stop the pointer from pointing to an
+  ///     array element).
+  ///
+  /// \todo The <code>&*(i + N) == (&*i) + N</code> invariant is
+  /// currently impossible to check for user-defined types.  We need a
+  /// \c contiguous_iterator_tag to let users assert it.
+  template <typename R> LLVM_CONSTEXPR range(
+      R &&r,
+      typename std::enable_if<
+        std::is_pointer<Iterator>::value &&
+        detail::is_contiguous_range<R>::value
+      // MSVC returns false for this in this context, but not if we lift it out of the
+      // constructor.
+#ifndef _MSC_VER
+        && detail::conversion_preserves_array_indexing<
+             decltype(&*detail::adl_begin(r)), Iterator>::value
+#endif
+      >::type* = 0)
+      : begin_((detail::adl_begin(r) == detail::adl_end(r) &&
+                !std::is_pointer<decltype(detail::adl_begin(r))>::value)
+               // For non-pointers, &*begin(r) is only defined behavior
+               // if there's an element there.  Otherwise, use nullptr
+               // since the user can't dereference it anyway.  This _is_
+               // detectable.
+               ? nullptr : &*detail::adl_begin(r)),
+        end_(begin_ + (detail::adl_end(r) - detail::adl_begin(r))) {}
+
+  /// @}
+
+  /// \name iterator access
+  /// @{
+  LLVM_CONSTEXPR Iterator begin() const { return begin_; }
+  LLVM_CONSTEXPR Iterator end() const { return end_; }
+  /// @}
+
+  /// \name element access
+  /// @{
+
+  /// \par Complexity:
+  /// O(1)
+  /// \pre \c !empty()
+  /// \returns a reference to the element at the front of the range.
+  LLVM_CONSTEXPR reference front() const { return *begin(); }
+
+  /// \par Ill-formed unless:
+  /// \c iterator_category is convertible to \c
+  /// std::bidirectional_iterator_tag.
+  ///
+  /// \par Complexity:
+  /// O(2) (Involves copying and decrementing an iterator, so not
+  /// quite as cheap as \c front())
+  ///
+  /// \pre \c !empty()
+  /// \returns a reference to the element at the front of the range.
+  LLVM_CONSTEXPR reference back() const {
+    static_assert(
+        std::is_convertible<iterator_category,
+                            std::bidirectional_iterator_tag>::value,
+        "Can only retrieve the last element of a bidirectional range.");
+    using std::prev;
+    return *prev(end());
+  }
+
+  /// This method is drawn from scripting language indexing.  It
+  /// indexes std::forward from the beginning of the range if the argument
+  /// is positive, or backwards from the end of the array if the
+  /// argument is negative.
+  ///
+  /// \par Ill-formed unless:
+  /// \c iterator_category is convertible to \c
+  /// std::random_access_iterator_tag.
+  ///
+  /// \par Complexity:
+  /// O(1)
+  ///
+  /// \pre <code>abs(index) < size() || index == -size()</code>
+  ///
+  /// \returns if <code>index >= 0</code>, a reference to the
+  /// <code>index</code>'th element in the range. Otherwise, a
+  /// reference to the <code>size()+index</code>'th element.
+  LLVM_CONSTEXPR reference operator[](difference_type index) const {
+    static_assert(std::is_convertible<iterator_category,
+                                      std::random_access_iterator_tag>::value,
+                  "Can only index into a random-access range.");
+    // Less readable construction for constexpr support.
+    return index < 0 ? end()[index]
+                     : begin()[index];
+  }
+  /// @}
+
+  /// \name size
+  /// @{
+
+  /// \par Complexity:
+  /// O(1)
+  /// \returns \c true if the range contains no elements.
+  LLVM_CONSTEXPR bool empty() const { return begin() == end(); }
+
+  /// \par Ill-formed unless:
+  /// \c iterator_category is convertible to
+  /// \c std::forward_iterator_tag.
+  ///
+  /// \par Complexity:
+  /// O(1) if \c iterator_category is convertible to \c
+  /// std::random_access_iterator_tag. O(<code>size()</code>)
+  /// otherwise.
+  ///
+  /// \returns the number of times \c pop_front() can be called before
+  /// \c empty() becomes true.
+  LLVM_CONSTEXPR difference_type size() const {
+    static_assert(std::is_convertible<iterator_category,
+                                      std::forward_iterator_tag>::value,
+                  "Calling size on an input range would destroy the range.");
+    return dispatch_size(iterator_category());
+  }
+  /// @}
+
+  /// \name traversal from the beginning of the range
+  /// @{
+
+  /// Advances the beginning of the range by one element.
+  /// \pre \c !empty()
+  void pop_front() { ++begin_; }
+
+  /// Advances the beginning of the range by \c n elements.
+  ///
+  /// \par Complexity:
+  /// O(1) if \c iterator_category is convertible to \c
+  /// std::random_access_iterator_tag, O(<code>n</code>) otherwise.
+  ///
+  /// \pre <code>n >= 0</code>, and there must be at least \c n
+  /// elements in the range.
+  void pop_front(difference_type n) { advance(begin_, n); }
+
+  /// Advances the beginning of the range by at most \c n elements,
+  /// stopping if the range becomes empty.  A negative argument causes
+  /// no change.
+  ///
+  /// \par Complexity:
+  /// O(1) if \c iterator_category is convertible to \c
+  /// std::random_access_iterator_tag, O(<code>min(n,
+  /// <var>#-elements-in-range</var>)</code>) otherwise.
+  ///
+  /// \note Could be provided as a free function with little-to-no
+  /// loss in efficiency.
+  void pop_front_upto(difference_type n) {
+    advance_upto(begin_, std::max<difference_type>(0, n), end_,
+                 iterator_category());
+  }
+
+  /// @}
+
+  /// \name traversal from the end of the range
+  /// @{
+
+  /// Moves the end of the range earlier by one element.
+  ///
+  /// \par Ill-formed unless:
+  /// \c iterator_category is convertible to
+  /// \c std::bidirectional_iterator_tag.
+  ///
+  /// \par Complexity:
+  /// O(1)
+  ///
+  /// \pre \c !empty()
+  void pop_back() {
+    static_assert(std::is_convertible<iterator_category,
+                                      std::bidirectional_iterator_tag>::value,
+                  "Can only access the end of a bidirectional range.");
+    --end_;
+  }
+
+  /// Moves the end of the range earlier by \c n elements.
+  ///
+  /// \par Ill-formed unless:
+  /// \c iterator_category is convertible to
+  /// \c std::bidirectional_iterator_tag.
+  ///
+  /// \par Complexity:
+  /// O(1) if \c iterator_category is convertible to \c
+  /// std::random_access_iterator_tag, O(<code>n</code>) otherwise.
+  ///
+  /// \pre <code>n >= 0</code>, and there must be at least \c n
+  /// elements in the range.
+  void pop_back(difference_type n) {
+    static_assert(std::is_convertible<iterator_category,
+                                      std::bidirectional_iterator_tag>::value,
+                  "Can only access the end of a bidirectional range.");
+    advance(end_, -n);
+  }
+
+  /// Moves the end of the range earlier by <code>min(n,
+  /// size())</code> elements.  A negative argument causes no change.
+  ///
+  /// \par Ill-formed unless:
+  /// \c iterator_category is convertible to
+  /// \c std::bidirectional_iterator_tag.
+  ///
+  /// \par Complexity:
+  /// O(1) if \c iterator_category is convertible to \c
+  /// std::random_access_iterator_tag, O(<code>min(n,
+  /// <var>#-elements-in-range</var>)</code>) otherwise.
+  ///
+  /// \note Could be provided as a free function with little-to-no
+  /// loss in efficiency.
+  void pop_back_upto(difference_type n) {
+    static_assert(std::is_convertible<iterator_category,
+                                      std::bidirectional_iterator_tag>::value,
+                  "Can only access the end of a bidirectional range.");
+    advance_upto(end_, -std::max<difference_type>(0, n), begin_,
+                 iterator_category());
+  }
+
+  /// @}
+
+  /// \name creating derived ranges
+  /// @{
+
+  /// Divides the range into two pieces at \c index, where a positive
+  /// \c index represents an offset from the beginning of the range
+  /// and a negative \c index represents an offset from the end.
+  /// <code>range[index]</code> is the first element in the second
+  /// piece.  If <code>index >= size()</code>, the second piece
+  /// will be empty. If <code>index < -size()</code>, the first
+  /// piece will be empty.
+  ///
+  /// \par Ill-formed unless:
+  /// \c iterator_category is convertible to
+  /// \c std::forward_iterator_tag.
+  ///
+  /// \par Complexity:
+  ///   - If \c iterator_category is convertible to \c
+  ///     std::random_access_iterator_tag: O(1)
+  ///   - Otherwise, if \c iterator_category is convertible to \c
+  ///     std::bidirectional_iterator_tag, \c abs(index) iterator increments
+  ///     or decrements
+  ///   - Otherwise, if <code>index >= 0</code>,  \c index iterator
+  ///     increments
+  ///   - Otherwise, <code>size() + (size() + index)</code>
+  ///     iterator increments.
+  ///
+  /// \returns a pair of adjacent ranges.
+  ///
+  /// \post
+  ///   - <code>result.first.size() == min(index, this->size())</code>
+  ///   - <code>result.first.end() == result.second.begin()</code>
+  ///   - <code>result.first.size() + result.second.size()</code> <code>==
+  ///     this->size()</code>
+  ///
+  /// \todo split() could take an arbitrary number of indices and
+  /// return an <code>N+1</code>-element \c tuple<>. This is tricky to
+  /// implement with negative indices in the optimal number of
+  /// increments or decrements for a bidirectional iterator, but it
+  /// should be possible.  Do we want it?
+  std::pair<range, range> split(difference_type index) const {
+    static_assert(
+        std::is_convertible<iterator_category,
+                            std::forward_iterator_tag>::value,
+        "Calling split on a non-std::forward range would return a useless "
+        "first result.");
+    if (index >= 0) {
+      range second = *this;
+      second.pop_front_upto(index);
+      return make_pair(range(begin(), second.begin()), second);
+    } else {
+      return dispatch_split_neg(index, iterator_category());
+    }
+  }
+
+  /// \returns A sub-range from \c start to \c stop (not including \c
+  /// stop, as usual).  \c start and \c stop are interpreted as for
+  /// <code>operator[]</code>, with negative values offsetting from
+  /// the end of the range.  Omitting the \c stop argument makes the
+  /// sub-range continue to the end of the original range. Positive
+  /// arguments saturate to the end of the range, and negative
+  /// arguments saturate to the beginning.  If \c stop is before \c
+  /// start, returns an empty range beginning and ending at \c start.
+  ///
+  /// \par Ill-formed unless:
+  /// \c iterator_category is convertible to
+  /// \c std::forward_iterator_tag.
+  ///
+  /// \par Complexity:
+  ///   - If \c iterator_category is convertible to \c
+  ///     std::random_access_iterator_tag: O(1)
+  ///   - Otherwise, if \c iterator_category is convertible to \c
+  ///     std::bidirectional_iterator_tag, at most <code>min(abs(start),
+  ///     size()) + min(abs(stop), size())</code> iterator
+  ///     increments or decrements
+  ///   - Otherwise, if <code>start >= 0 && stop >= 0</code>,
+  ///     <code>max(start, stop)</code> iterator increments
+  ///   - Otherwise, <code>size() + max(start', stop')</code>
+  ///     iterator increments, where \c start' and \c stop' are the
+  ///     offsets of the elements \c start and \c stop refer to.
+  ///
+  /// \note \c slice(start) should be implemented with a different
+  /// overload, rather than defaulting \c stop to
+  /// <code>numeric_limits<difference_type>::max()</code>, because
+  /// using a default would force non-random-access ranges to use an
+  /// O(<code>size()</code>) algorithm to compute the end rather
+  /// than the O(1) they're capable of.
+  range slice(difference_type start, difference_type stop) const {
+    static_assert(
+        std::is_convertible<iterator_category,
+                            std::forward_iterator_tag>::value,
+        "Calling slice on a non-std::forward range would destroy the original "
+        "range.");
+    return dispatch_slice(start, stop, iterator_category());
+  }
+
+  range slice(difference_type start) const {
+    static_assert(
+        std::is_convertible<iterator_category,
+                            std::forward_iterator_tag>::value,
+        "Calling slice on a non-std::forward range would destroy the original "
+        "range.");
+    return split(start).second;
+  }
+
+  /// @}
+
+private:
+  // advance_upto: should be added to <algorithm>, but I'll use it as
+  // a helper function here.
+  //
+  // These return the number of increments that weren't applied
+  // because we ran into 'limit' (or 0 if we didn't run into limit).
+  static difference_type advance_upto(Iterator &it, difference_type n,
+                                      Iterator limit, std::input_iterator_tag) {
+    if (n < 0)
+      return 0;
+    while (it != limit && n > 0) {
+      ++it;
+      --n;
+    }
+    return n;
+  }
+
+  static difference_type advance_upto(Iterator &it, difference_type n,
+                                      Iterator limit,
+                                      std::bidirectional_iterator_tag) {
+    if (n < 0) {
+      while (it != limit && n < 0) {
+        --it;
+        ++n;
+      }
+    } else {
+      while (it != limit && n > 0) {
+        ++it;
+        --n;
+      }
+    }
+    return n;
+  }
+
+  static difference_type advance_upto(Iterator &it, difference_type n,
+                                      Iterator limit,
+                                      std::random_access_iterator_tag) {
+    difference_type distance = limit - it;
+    if (distance < 0)
+      assert(n <= 0);
+    else if (distance > 0)
+      assert(n >= 0);
+
+    if (abs(distance) > abs(n)) {
+      it += n;
+      return 0;
+    } else {
+      it = limit;
+      return n - distance;
+    }
+  }
+
+  // Dispatch functions.
+  difference_type dispatch_size(std::forward_iterator_tag) const {
+    return std::distance(begin(), end());
+  }
+
+  LLVM_CONSTEXPR difference_type dispatch_size(
+      std::random_access_iterator_tag) const {
+    return end() - begin();
+  }
+
+  std::pair<range, range> dispatch_split_neg(difference_type index,
+                                             std::forward_iterator_tag) const {
+    assert(index < 0);
+    difference_type size = this->size();
+    return split(std::max<difference_type>(0, size + index));
+  }
+
+  std::pair<range, range> dispatch_split_neg(
+      difference_type index, std::bidirectional_iterator_tag) const {
+    assert(index < 0);
+    range first = *this;
+    first.pop_back_upto(-index);
+    return make_pair(first, range(first.end(), end()));
+  }
+
+  range dispatch_slice(difference_type start, difference_type stop,
+                       std::forward_iterator_tag) const {
+    if (start < 0 || stop < 0) {
+      difference_type size = this->size();
+      if (start < 0)
+        start = std::max<difference_type>(0, size + start);
+      if (stop < 0)
+        stop = size + stop; // Possibly negative; will be fixed in 2 lines.
+    }
+    stop = std::max<difference_type>(start, stop);
+
+    Iterator first = begin();
+    advance_upto(first, start, end(), iterator_category());
+    Iterator last = first;
+    advance_upto(last, stop - start, end(), iterator_category());
+    return range(first, last);
+  }
+
+  range dispatch_slice(const difference_type start, const difference_type stop,
+                       std::bidirectional_iterator_tag) const {
+    Iterator first;
+    if (start < 0) {
+      first = end();
+      advance_upto(first, start, begin(), iterator_category());
+    } else {
+      first = begin();
+      advance_upto(first, start, end(), iterator_category());
+    }
+    Iterator last;
+    if (stop < 0) {
+      last = end();
+      advance_upto(last, stop, first, iterator_category());
+    } else {
+      if (start >= 0) {
+        last = first;
+        if (stop > start)
+          advance_upto(last, stop - start, end(), iterator_category());
+      } else {
+        // Complicated: 'start' walked from the end of the sequence,
+        // but 'stop' needs to walk from the beginning.
+        Iterator dummy = begin();
+        // Walk up to 'stop' increments from begin(), stopping when we
+        // get to 'first', and capturing the remaining number of
+        // increments.
+        difference_type increments_past_start =
+            advance_upto(dummy, stop, first, iterator_category());
+        if (increments_past_start == 0) {
+          // If this is 0, then stop was before start.
+          last = first;
+        } else {
+          // Otherwise, count that many spaces beyond first.
+          last = first;
+          advance_upto(last, increments_past_start, end(), iterator_category());
+        }
+      }
+    }
+    return range(first, last);
+  }
+
+  range dispatch_slice(difference_type start, difference_type stop,
+                       std::random_access_iterator_tag) const {
+    const difference_type size = this->size();
+    if (start < 0)
+      start = size + start;
+    if (start < 0)
+      start = 0;
+    if (start > size)
+      start = size;
+
+    if (stop < 0)
+      stop = size + stop;
+    if (stop < start)
+      stop = start;
+    if (stop > size)
+      stop = size;
+
+    return range(begin() + start, begin() + stop);
+  }
+};
+
+/// \name deducing constructor wrappers
+/// \relates std::range
+/// \xmlonly <nonmember/> \endxmlonly
+///
+/// These functions do the same thing as the constructor with the same
+/// signature. They just allow users to avoid writing the iterator
+/// type.
+/// @{
+
+/// \todo I'd like to define a \c make_range taking a single iterator
+/// argument representing the beginning of a range that ends with a
+/// default-constructed \c Iterator.  This would help with using
+/// iterators like \c istream_iterator.  However, using just \c
+/// make_range() could be confusing and lead to people writing
+/// incorrect ranges of more common iterators. Is there a better name?
+template <typename Iterator>
+LLVM_CONSTEXPR range<Iterator> make_range(Iterator begin, Iterator end) {
+  return range<Iterator>(begin, end);
+}
+
+/// \par Participates in overload resolution if:
+/// \c begin(r) and \c end(r) return the same type.
+template <typename Range> LLVM_CONSTEXPR auto make_range(
+    Range &&r,
+    typename std::enable_if<detail::is_range<Range>::value>::type* = 0)
+    -> range<decltype(detail::adl_begin(r))> {
+  return range<decltype(detail::adl_begin(r))>(r);
+}
+
+/// \par Participates in overload resolution if:
+///   - \c begin(r) and \c end(r) return the same type,
+///   -  that type satisfies the invariant that <code>&*(i + N) ==
+///      (&*i) + N</code>, and
+///   - \c &*begin(r) has a pointer type.
+template <typename Range> LLVM_CONSTEXPR auto make_ptr_range(
+    Range &&r,
+    typename std::enable_if<
+      detail::is_contiguous_range<Range>::value &&
+      std::is_pointer<decltype(&*detail::adl_begin(r))>::value>::type* = 0)
+      -> range<decltype(&*detail::adl_begin(r))> {
+  return range<decltype(&*detail::adl_begin(r))>(r);
+}
+/// @}
+} // end namespace lld
+
+#endif