Support for function summary index bitcode sections and files

[tejohnson]

Summary

The bitcode format is described in this document:

https://drive.google.com/file/d/0B036uwnWM6RWdnBLakxmeDdOeXc/view

For more info on ThinLTO see:

https://sites.google.com/site/llvmthinlto

The first customer is ThinLTO, however the data structures are designed
and named more generally based on prior feedback. There are a few
comments regarding how certain interfaces are used by ThinLTO, and the
options added here to gold (and to a separate clang patch which will
be mailed along with this) currently have ThinLTO-specific names as the
behavior they provoke is currently ThinLTO-specific.

This patch includes support for generating per-module function indexes,
the combined index file via the gold plugin, and several tests
(more are included with the associated clang patch).

[tejohnson]

Rather than create a new clang patch, I have updated http://reviews.llvm.org/D11908.

[tejohnson]

• Fix function summary generation for anonymous function.

Don't try to generate a summary for them as they don't have a
VST entry/value Id, and ensure that any aliases get function summaries instead.

This was caught by running llvm test suite with thinlto enabled by
default in llvm-as.

Add anonymous function to thinlto test case.

[mehdi_amini]

Hi Teresa,

Please find some inline comments below. I had some minor naming suggestion, but more important is the ownership of some object instantiation.

include/llvm/Bitcode/BitcodeWriterPass.h

54

Rename ThinLTO with FunctionSummary to be coherent with the rest like hasFunctionSummary() for instance?

I think there are multiple places (not to say everywhere) where ThinLTO could be replaced with something more generic like FunctionSummary.

include/llvm/Bitcode/BitstreamWriter.h

24

Why?

528

Shouldn't be needed?

include/llvm/Bitcode/ReaderWriter.h

74

Might be more explicit to rename ParseFuncSummary into IsLazy.
(just a suggestion, do whatever seems better to you)

include/llvm/IR/FunctionInfo.h

46

s/disambiguiate/disambiguate/ ?

100

The "it" at the beginning of the line seems strange to me?

108

Who owns it?
(I may found out later)

109

(Missing empty line)

131

If this takes ownership, use a unique_ptr for the argument.

137

Any reason not to use unique_ptr for the Summary member?

144

This sounds suspicious: a copy ctor that does not really copy...
I have to see the use case first, I'm skeptical right now :)

167

The other setters name are beginning with set, why not this one?

191

remove

214

Why not using unique_ptr and remote the destructor?

224

If it was called "begin()" you could use for-range loop. An alternative would be to add another method functions() that would return an llvm::iterator_range.

233

Do you really intend to return a copy of the vector all the time?

include/llvm/Object/FunctionIndexObjectFile.h

85

indentation

lib/Bitcode/Reader/BitcodeReader.cpp

410

What about a variable name that contains Lazy?

437

The ownership/lifetime of FunctionSummary is not clear here?

3110

remove

5215

Should you just return now?

5241

I would expect clang-format to put the else on the same line as the {?

5313

I don't see the expected return for this function?

5343

The indentation seems incorrect, is it clang-formatted?

5382

Typo with ThihLTO.
(also I'm still not sure at which level where the "ThinLTO" concept should "leak").

5684

When is this memory released?

lib/Bitcode/Writer/BitcodeWriter.cpp

2304

For-range loops?

2323

for(auto P : FuncName) ?

2384

Ownership is not clear, what about having a DenseMap<const Function *, std::unique_ptr<FunctionInfo>> instead?

2693

for (auto P : MPSE.getKey()) ?

2720

For-range loop?

2757

Some refactoring seems possible between these two loops.

2780

for-range loop?

2969

(Side note) I read this so many times, I'm surprised there is not a better refactoring for reading/writing bitcode.

lib/IR/FunctionInfo.cpp

20

A mutable global is really to be avoided. I am really not convinced that it is desirable here.
Should it be passed by argument to the mergeFrom method? Or a member variable?

Maybe even mergeFrom() should not be a member function, but a separate object that would have a NextModuleId and you could add FunctionInfoIndex and get the merged result.

24

What about taking a unique_ptr here and "stealing" its content instead of doing a bunch of redundant memory allocation and copy?

30

This loop header is strange with the iterator declared outside. Turn into a for-range loop?

lib/Object/FunctionIndexObjectFile.cpp

90

I'm not sure why you would need to provide the fully qualified name for this function?

test/Bitcode/thinlto-function-summary.ll

18

This line is ignored

26

This line is ignored as well

test/tools/gold/X86/thinlto.ll

9

What is this line doing? Isn't the -a for the && operator?
Do you check that the file %t3 is not generated by gold and only the combine index is?

[tejohnson]

• Address Mehdi's review comments.

[mehdi_amini]

lib/Bitcode/Reader/BitcodeReader.cpp

5684

Why not having R as a stack local variable? Initially before the unique_ptr I thought you were intended to pass it outside but it seems to be a pure local variable.

test/Bitcode/thinlto-function-summary.ll

19

I didn't see the second RUN: line, forget this.

[tejohnson]

• Make FunctionIndexBitcodeReader a local variable.

[echristo]

Hi Teresa,

A couple of inline comments, I'll have more, but I just thought I'd send a couple out.

One general comment:

How expensive is the index (size and time to compute)? Is it worth always including in the module?

Thanks!

-eric

llvm/trunk/include/llvm/IR/FunctionInfo.h

57 ↗

(On Diff #36467)

Comment nit: "during the initial compile step" seems weird and not wholly accurate.

Relatedly: would a transform be expected to keep it up to date? Is it recomputed on every write? (Seems reasonable, but costly).

61–62 ↗

(On Diff #36467)

This comment seems a bit confusing. Could you elaborate more?

include/llvm/Bitcode/BitcodeWriterPass.h

 
 /// \brief Create and return a pass that writes the module to the specified
 /// ostream. Note that this pass is designed for use with the legacy pass
 /// manager.
 ///
 /// If \c ShouldPreserveUseListOrder, encode use-list order so it can be
 /// reproduced when deserialized.
 ModulePass *createBitcodeWriterPass(raw_ostream &Str,
-                                    bool ShouldPreserveUseListOrder = false);
+                                    bool ShouldPreserveUseListOrder = false,
+                                    bool EmitThinLTOIndex = false);
 
 /// \brief Pass for writing a module of IR out to a bitcode file.
 ///
 /// Note that this is intended for use with the new pass manager. To construct
 /// a pass for the legacy pass manager, use the function above.
 class BitcodeWriterPass {
   raw_ostream &OS;
   bool ShouldPreserveUseListOrder;
+  bool EmitThinLTOIndex;
 
 public:
   /// \brief Construct a bitcode writer pass around a particular output stream.
   ///
   /// If \c ShouldPreserveUseListOrder, encode use-list order so it can be
   /// reproduced when deserialized.
   explicit BitcodeWriterPass(raw_ostream &OS,
-                             bool ShouldPreserveUseListOrder = false)
-      : OS(OS), ShouldPreserveUseListOrder(ShouldPreserveUseListOrder) {}
+                             bool ShouldPreserveUseListOrder = false,
+                             bool EmitThinLTOIndex = false)
+      : OS(OS), ShouldPreserveUseListOrder(ShouldPreserveUseListOrder),
+        EmitThinLTOIndex(EmitThinLTOIndex) {}
 
   /// \brief Run the bitcode writer pass, and output the module to the selected
   /// output stream.
   PreservedAnalyses run(Module &M);
 
   static StringRef name() { return "BitcodeWriterPass"; }
 };
 
 }
 
 #endif

include/llvm/Bitcode/BitstreamWriter.h

 #define LLVM_BITCODE_BITSTREAMWRITER_H
 
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/Optional.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Bitcode/BitCodes.h"
 #include "llvm/Support/Endian.h"
+#include "llvm/Support/raw_ostream.h"
 #include <vector>
 
 namespace llvm {
 
 class BitstreamWriter {
   SmallVectorImpl<char> &Out;
 
   /// CurBit - Always between 0 and 31 inclusive, specifies the next bit to use.
public:
   ~BitstreamWriter() {
     assert(CurBit == 0 && "Unflushed data remaining");
     assert(BlockScope.empty() && CurAbbrevs.empty() && "Block imbalance");
   }
 
   /// \brief Retrieve the current position in the stream, in bits.
   uint64_t GetCurrentBitNo() const { return GetBufferOffset() * 8 + CurBit; }
 
+  /// \brief Retrieve the number of bits currently used to encode an abbrev ID.
+  unsigned GetAbbrevIDWidth() const { return CurCodeSize; }
+
   //===--------------------------------------------------------------------===//
   // Basic Primitives for emitting bits to the stream.
   //===--------------------------------------------------------------------===//
 
   /// Backpatch a 32-bit word in the output at the given bit offset
   /// with the specified value.
   void BackpatchWord(uint64_t BitNo, unsigned NewWord) {
     unsigned ByteNo = BitNo / 8;
private:
   void SwitchToBlockID(unsigned BlockID) {
     if (BlockInfoCurBID == BlockID) return;
     SmallVector<unsigned, 2> V;
     V.push_back(BlockID);
     EmitRecord(bitc::BLOCKINFO_CODE_SETBID, V);
     BlockInfoCurBID = BlockID;
   }
 
+public:
+
   BlockInfo &getOrCreateBlockInfo(unsigned BlockID) {
     if (BlockInfo *BI = getBlockInfo(BlockID))
       return *BI;
 
     // Otherwise, add a new record.
     BlockInfoRecords.emplace_back();
     BlockInfoRecords.back().BlockID = BlockID;
     return BlockInfoRecords.back();
   }
 
-public:
-
   /// EmitBlockInfoAbbrev - Emit a DEFINE_ABBREV record for the specified
   /// BlockID.
   unsigned EmitBlockInfoAbbrev(unsigned BlockID, BitCodeAbbrev *Abbv) {
     SwitchToBlockID(BlockID);
     EncodeAbbrev(Abbv);
 
     // Add the abbrev to the specified block record.
     BlockInfo &Info = getOrCreateBlockInfo(BlockID);

include/llvm/Bitcode/LLVMBitCodes.h

enum BlockIDs {
 
     VALUE_SYMTAB_BLOCK_ID,
     METADATA_BLOCK_ID,
     METADATA_ATTACHMENT_ID,
 
     TYPE_BLOCK_ID_NEW,
 
     USELIST_BLOCK_ID,
+
+    MODULE_STRTAB_BLOCK_ID,
+    FUNCTION_SUMMARY_BLOCK_ID
   };
 
 
   /// MODULE blocks have a number of optional fields and subblocks.
   enum ModuleCodes {
     MODULE_CODE_VERSION     = 1,    // VERSION:     [version#]
     MODULE_CODE_TRIPLE      = 2,    // TRIPLE:      [strchr x N]
     MODULE_CODE_DATALAYOUT  = 3,    // DATALAYOUT:  [strchr x N]
enum TypeSymtabCodes {
     TST_CODE_ENTRY = 1     // TST_ENTRY: [typeid, namechar x N]
   };
 
   // Value symbol table codes.
   enum ValueSymtabCodes {
     VST_CODE_ENTRY   = 1,   // VST_ENTRY: [valueid, namechar x N]
     VST_CODE_BBENTRY = 2,   // VST_BBENTRY: [bbid, namechar x N]
     VST_CODE_FNENTRY = 3,   // VST_FNENTRY: [valueid, offset, namechar x N]
+    VST_CODE_COMBINED_FNENTRY = 4    // VST_COMBINED_FNENTRY: [offset, namechar x N]
+  };
+
+  // The module path symbol table only has one code (MST_CODE_ENTRY).
+  enum ModulePathSymtabCodes {
+    MST_CODE_ENTRY   = 1,  // MST_ENTRY: [modid, namechar x N]
+  };
+
+  // The function summary section uses different codes in the per-module
+  // and combined index cases.
+  enum FunctionSummarySymtabCodes {
+    FS_CODE_PERMODULE_ENTRY   = 1,  // FS_ENTRY: [valueid, islocal, instcount]
+    FS_CODE_COMBINED_ENTRY    = 2,  // FS_ENTRY: [modid, instcount]
   };
 
   enum MetadataCodes {
     METADATA_STRING        = 1,   // MDSTRING:      [values]
     METADATA_VALUE         = 2,   // VALUE:         [type num, value num]
     METADATA_NODE          = 3,   // NODE:          [n x md num]
     METADATA_NAME          = 4,   // STRING:        [values]
     METADATA_DISTINCT_NODE = 5,   // DISTINCT_NODE: [n x md num]

include/llvm/Bitcode/ReaderWriter.h

 // This header defines interfaces to read and write LLVM bitcode files/streams.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_BITCODE_READERWRITER_H
 #define LLVM_BITCODE_READERWRITER_H
 
 #include "llvm/IR/DiagnosticInfo.h"
+#include "llvm/IR/FunctionInfo.h"
 #include "llvm/Support/Endian.h"
 #include "llvm/Support/ErrorOr.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include <memory>
 #include <string>
 
 namespace llvm {
   class BitstreamWriter;
namespace llvm {
   getBitcodeTargetTriple(MemoryBufferRef Buffer, LLVMContext &Context,
                          DiagnosticHandlerFunction DiagnosticHandler = nullptr);
 
   /// Read the specified bitcode file, returning the module.
   ErrorOr<std::unique_ptr<Module>>
   parseBitcodeFile(MemoryBufferRef Buffer, LLVMContext &Context,
                    DiagnosticHandlerFunction DiagnosticHandler = nullptr);
 
+  /// Check if the given bitcode buffer contains a function summary block.
+  bool hasFunctionSummary(MemoryBufferRef Buffer, LLVMContext &Context,
+                          DiagnosticHandlerFunction DiagnosticHandler);
+
+  /// Parse the specified bitcode buffer, returning the function info index.
+  /// If ParseFuncSummary is true, parse the entire function summary into
+  /// the index. Otherwise skip the function summary section, and only create
+  /// an index object with a map from function name to function summary offset.
+  /// The index is used to perform lazy function summary reading later.
+  ErrorOr<std::unique_ptr<FunctionInfoIndex>> getFunctionInfoIndex(
+      MemoryBufferRef Buffer, LLVMContext &Context,
+      DiagnosticHandlerFunction DiagnosticHandler,
+      bool ParseFuncSummary = true);
+
+  /// This method supports lazy reading of function summary data from the
+  /// combined index during function importing. When reading the combined index
+  /// file, getFunctionInfoIndex is first invoked with ParseFuncSummary=false.
+  /// Then this method is called for each function considered for importing,
+  /// to parse the summary information for the given function name into
+  /// the index.
+  std::error_code readFunctionSummary(
+      MemoryBufferRef Buffer, LLVMContext &Context,
+      DiagnosticHandlerFunction DiagnosticHandler, StringRef FunctionName,
+      std::unique_ptr<FunctionInfoIndex> Index);
+
   /// \brief Write the specified module to the specified raw output stream.
   ///
   /// For streams where it matters, the given stream should be in "binary"
   /// mode.
   ///
   /// If \c ShouldPreserveUseListOrder, encode the use-list order for each \a
   /// Value in \c M.  These will be reconstructed exactly when \a M is
   /// deserialized.
+  ///
+  /// If \c EmitThinLTOIndex, emit the function summary index for use
+  /// in ThinLTO optimization.
   void WriteBitcodeToFile(const Module *M, raw_ostream &Out,
-                          bool ShouldPreserveUseListOrder = false);
+                          bool ShouldPreserveUseListOrder = false,
+                          bool EmitThinLTOIndex = false);
+
+  /// Write the specified function summary index to the given raw output stream,
+  /// where it will be written in a new bitcode block. This is used when
+  /// writing the combined index file for ThinLTO.
+  void WriteFunctionSummaryToFile(const FunctionInfoIndex *Index,
+                                  raw_ostream &Out);
 
   /// isBitcodeWrapper - Return true if the given bytes are the magic bytes
   /// for an LLVM IR bitcode wrapper.
   ///
   inline bool isBitcodeWrapper(const unsigned char *BufPtr,
                                const unsigned char *BufEnd) {
     // See if you can find the hidden message in the magic bytes :-).
     // (Hint: it's a little-endian encoding.)

include/llvm/IR/FunctionInfo.h

+//===-- llvm/FunctionInfo.h - Function Info Index ---------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+/// @file
+/// FunctionInfo.h This file contains the declarations the classes that hold
+///  the function info index and summary.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_IR_FUNCTIONINFO_H
+#define LLVM_IR_FUNCTIONINFO_H
+
+#include "llvm/ADT/SmallString.h"
+#include "llvm/ADT/StringMap.h"
+#include "llvm/IR/Module.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/raw_ostream.h"
+
+namespace llvm {
+
+/// \brief Function summary information to aid decisions and implementation of
+/// importing.
+///
+/// This is a separate class from FunctionInfo to enable lazy reading of this
+/// function summary information from the combined index file during imporing.
+class FunctionSummary {
+ private:
+  /// \brief Path of module containing function IR, used to locate module when
+  /// importing this function.
+  ///
+  /// This is only used during parsing of the combined function index, or when
+  /// parsing the per-module index for creation of the combined function index,
+  /// not during writing of the per-module index which doesn't contain a
+  /// module path string table.
+  StringRef ModulePath;
+
+  /// \brief Used to flag functions that have local linkage types and need to
+  /// have module identifier appended before placing into the combined
+  /// index, to disambiguiate from other functions with the same name.
+  ///
+  /// This is only used in the per-module function index, as it is consumed
+  /// while creating the combined index.
+  bool IsLocalFunction;
+
+  // The rest of the information is used to help decide whether importing
+  // is likely to be profitable.
+  // Other information will be added as the importing is tuned, such
+  // as hotness (when profile available), and other function characteristics.
+
+  /// Number of instructions (ignoring debug instructions, e.g.) computed
+  /// during the initial compile step when the function index is first built.
+  unsigned InstCount;
+
+ public:
+  /// Construct a summary object from summary data expected for all
+  /// summary records.
+  FunctionSummary(unsigned NumInsts) :
+    InstCount(NumInsts) { }
+
+  /// Set the path to the module containing this function, for use in
+  /// the combined index.
+  void setModulePath(StringRef ModPath) {
+    ModulePath = ModPath;
+  }
+
+  /// Get the path to the module containing this function.
+  StringRef modulePath() const {
+    return ModulePath;
+  }
+
+  /// Record whether this is a local function in the per-module index.
+  void setLocalFunction(bool IsLocal) {
+    IsLocalFunction = IsLocal;
+  }
+
+  /// Check whether this was a local function, for use in creating
+  /// the combined index.
+  bool isLocalFunction() const {
+    return IsLocalFunction;
+  }
+
+  /// Get the instruction count recorded for this function.
+  unsigned instCount() const {
+    return InstCount;
+  }
+};
+
+/// \brief Class to hold pointer to function summary and information required
+/// for parsing it.
+///
+/// For the per-module index, this holds the bitcode offset
+/// of the corresponding function block. For the combined index,
+/// it, after parsing of the \a ValueSymbolTable, this initially
+/// holds the offset of the corresponding function summary bitcode
+/// record. After parsing the associated summary information from the summary
+/// block the \a FunctionSummary is populated and stored here.
+class FunctionInfo {
+ private:
+  /// Function summary information used to help make ThinLTO importing
+  /// decisions.
+  FunctionSummary *Summary;
+  /// \brief The bitcode offset corresponding to either the associated
+  /// function's function body record, or its function summary record,
+  /// depending on whether this is a per-module or combined index.
+  ///
+  /// This bitcode offset is written to or read from the associated
+  /// \a ValueSymbolTable entry for the function.
+  /// For the per-module index this holds the bitcode offset of the
+  /// function's body record  within bitcode module block in its module,
+  /// which is used during lazy function parsing or ThinLTO importing.
+  /// For the combined index this holds the offset of the corresponding
+  /// function summary record, to enable associating the combined index
+  /// VST records with the summary records.
+  uint64_t BitcodeIndex;
+
+ public:
+  /// Constructor used during parsing of VST entries.
+  FunctionInfo(uint64_t FuncOffset) :
+    Summary(nullptr),
+    BitcodeIndex(FuncOffset) { }
+
+  /// Constructor used for per-module index bitcode writing.
+  FunctionInfo(uint64_t FuncOffset,
+               FunctionSummary *FuncSummary) :
+    Summary(FuncSummary),
+    BitcodeIndex(FuncOffset) { }
+
+  ~FunctionInfo() {
+    if (Summary)
+      delete Summary;
+  }
+
+  /// Constructor used during combined index creation.
+  FunctionInfo(const FunctionInfo &Other) {
+    // Make sure we don't share the function summary object, so that
+    // destruction doesn't perform a double-delete.
+    Summary = nullptr;
+    BitcodeIndex = Other.BitcodeIndex;
+  }
+
+  /// Record the function summary information parsed out of the function
+  /// summary block during parsing or combined index creation.
+  void setFunctionSummary(FunctionSummary *FuncSummary) {
+    Summary = FuncSummary;
+  }
+
+  /// Get the function summary recorded for this function.
+  FunctionSummary *functionSummary() const {
+    return Summary;
+  }
+
+  /// Get the bitcode index recorded for this function, depending on
+  /// the index type.
+  uint64_t bitcodeIndex() const {
+    return BitcodeIndex;
+  }
+
+  /// Record the bitcode index for this function, depending on
+  /// the index type.
+  void bitcodeIndex(uint64_t FuncOffset) {
+    BitcodeIndex = FuncOffset;
+  }
+};
+
+/// List of function info structures for a particular function name held
+/// in the FunctionMap. Requires a vector in the case of multiple
+/// COMDAT functions of the same name.
+typedef std::vector<FunctionInfo*> FunctionInfoList;
+
+/// Map from function name to corresponding function info structures.
+typedef StringMap<FunctionInfoList> FunctionInfoMapTy;
+
+/// Type used for iterating through the function info map.
+typedef FunctionInfoMapTy::const_iterator const_funcinfo_iterator;
+
+/// String table to hold/own module path strings, which additionally holds the
+/// module ID assigned to each module during the plugin step. The StringMap
+/// makes a copy of and owns inserted strings.
+typedef StringMap<uint64_t> ModulePathStringTableTy;
+
+/// Class to hold module path string table and function map,
+/// and encapsulate methods for operating on them.
+class FunctionInfoIndex {
+ public:
+
+ private:
+  /// Map from function name to list of function information instances
+  /// for functions of that name (may be duplicates in the COMDAT case, e.g.).
+  FunctionInfoMapTy FunctionMap;
+
+  /// Holds strings for combined index, mapping to the corresponding module ID.
+  ModulePathStringTableTy ModulePathStringTable;
+
+  /// Used for assigning module ids when per-module index structures merged
+  /// to form combined index, necessary for consistent renaming of promoted
+  /// static (local) variables.
+  static uint64_t NextModuleId;
+
+ public:
+  FunctionInfoIndex() = default;
+
+  ~FunctionInfoIndex() {
+    for (const_funcinfo_iterator FMI = funcinfo_begin();
+         FMI != funcinfo_end(); ++FMI) {
+      for (std::vector<FunctionInfo*>::const_iterator FLI =
+           FMI->getValue().begin(); FLI != FMI->getValue().end(); ++FLI) {
+        delete *FLI;
+      }
+    }
+  }
+
+  // Disable the copy constructor and assignment operators, so
+  // no unexpected copying/moving occurs.
+  FunctionInfoIndex(const FunctionInfoIndex &) = delete;
+  void operator=(const FunctionInfoIndex &) = delete;
+
+  const_funcinfo_iterator funcinfo_begin() const {
+    return FunctionMap.begin();
+  }
+  const_funcinfo_iterator funcinfo_end() const {
+    return FunctionMap.end();
+  }
+
+  /// Get the list of function info objects for a given function.
+  FunctionInfoList getFunctionInfoList(StringRef FuncName) const {
+    return FunctionMap.lookup(FuncName);
+  }
+
+  /// Add a function info for a function of the given name.
+  void addFunctionInfo(StringRef FuncName,
+                       FunctionInfo *Info) {
+    FunctionMap[FuncName].push_back(Info);
+  }
+
+  /// Iterator to allow writer to walk through table during emission.
+  iterator_range<StringMap<uint64_t>::const_iterator>
+      modPathStringEntries() const {
+    return llvm::make_range(ModulePathStringTable.begin(),
+                            ModulePathStringTable.end());
+  }
+
+  /// Get the module ID recorded for the given module path.
+  uint64_t getModuleId(const StringRef ModPath) const {
+    return ModulePathStringTable.lookup(ModPath);
+  }
+
+  /// Create the combined function index/summary from multiple
+  /// per-module instances.
+  void mergeFrom(FunctionInfoIndex *Other);
+
+  /// Convenience method for creating a promoted global name
+  /// for the given value name of a local, and its original module's ID.
+  static std::string getGlobalNameForLocal(StringRef Name,
+                                           uint64_t ModId) {
+    SmallString<256> NewName(Name);
+    NewName += ".llvm.";
+    raw_svector_ostream(NewName) << ModId;
+    return NewName.str();
+  }
+
+  /// Add a new module path, mapped to the given module Id, and return StringRef
+  /// owned by string table map.
+  StringRef addModulePath(StringRef ModPath, uint64_t ModId) {
+    return ModulePathStringTable.insert(
+        std::make_pair(ModPath, ModId)).first->first();
+  }
+};
+
+} // End llvm namespace
+
+#endif

include/llvm/Object/Binary.h

protected:
 
   Binary(unsigned int Type, MemoryBufferRef Source);
 
   enum {
     ID_Archive,
     ID_MachOUniversalBinary,
     ID_COFFImportFile,
     ID_IR, // LLVM IR
+    ID_FunctionIndex, // Function summary index
 
     // Object and children.
     ID_StartObjects,
     ID_COFF,
 
     ID_ELF32L, // ELF 32-bit, little endian
     ID_ELF32B, // ELF 32-bit, big endian
     ID_ELF64L, // ELF 64-bit, little endian
public:
   bool isCOFFImportFile() const {
     return TypeID == ID_COFFImportFile;
   }
 
   bool isIR() const {
     return TypeID == ID_IR;
   }
 
+  bool isFunctionIndex() const {
+    return TypeID == ID_FunctionIndex;
+  }
+
   bool isLittleEndian() const {
     return !(TypeID == ID_ELF32B || TypeID == ID_ELF64B ||
              TypeID == ID_MachO32B || TypeID == ID_MachO64B);
   }
 };
 
 /// @brief Create a Binary from Source, autodetecting the file type.
 ///

include/llvm/Object/FunctionIndexObjectFile.h

+//===- FunctionIndexObjectFile.h - Function index file implementation -----===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file declares the FunctionIndexObjectFile template class.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_OBJECT_FUNCTIONINDEXOBJECTFILE_H
+#define LLVM_OBJECT_FUNCTIONINDEXOBJECTFILE_H
+
+#include "llvm/Object/SymbolicFile.h"
+
+namespace llvm {
+class FunctionInfoIndex;
+
+namespace object {
+class ObjectFile;
+
+/// This class is used to read just the function summary index related
+/// sections out of the given object (which may contain a single module's
+/// bitcode or be a combined index bitcode file). It builds a FunctionInfoIndex
+/// object.
+class FunctionIndexObjectFile : public SymbolicFile {
+  std::unique_ptr<FunctionInfoIndex> Index;
+
+public:
+  FunctionIndexObjectFile(MemoryBufferRef Object,
+                          std::unique_ptr<FunctionInfoIndex> I);
+  ~FunctionIndexObjectFile() override;
+
+  // TODO: Walk through FunctionMap entries for function symbols.
+  // However, currently these interfaces are not used by any consumers.
+  void moveSymbolNext(DataRefImpl &Symb) const override {
+    llvm_unreachable("not implemented");
+  }
+  std::error_code printSymbolName(raw_ostream &OS, DataRefImpl Symb) const override {
+    llvm_unreachable("not implemented");
+    return std::error_code();
+  }
+  uint32_t getSymbolFlags(DataRefImpl Symb) const override {
+    llvm_unreachable("not implemented");
+    return 0;
+  }
+  basic_symbol_iterator symbol_begin_impl() const override {
+    llvm_unreachable("not implemented");
+    return basic_symbol_iterator(BasicSymbolRef());
+  }
+  basic_symbol_iterator symbol_end_impl() const override {
+    llvm_unreachable("not implemented");
+    return basic_symbol_iterator(BasicSymbolRef());
+  }
+
+  const FunctionInfoIndex &getIndex() const {
+    return const_cast<FunctionIndexObjectFile*>(this)->getIndex();
+  }
+  FunctionInfoIndex &getIndex() {
+    return *Index;
+  }
+  std::unique_ptr<FunctionInfoIndex> takeIndex();
+
+  static inline bool classof(const Binary *v) {
+    return v->isFunctionIndex();
+  }
+
+  /// \brief Finds and returns bitcode embedded in the given object file, or an
+  /// error code if not found.
+  static ErrorOr<MemoryBufferRef> findBitcodeInObject(const ObjectFile &Obj);
+
+  /// \brief Finds and returns bitcode in the given memory buffer (which may
+  /// be either a bitcode file or a native object file with embedded bitcode),
+  /// or an error code if not found.
+  static ErrorOr<MemoryBufferRef>
+  findBitcodeInMemBuffer(MemoryBufferRef Object);
+
+  /// \brief Looks for function summary in the given memory buffer,
+  /// returns true if found, else false.
+  static bool hasFunctionSummaryInMemBuffer(MemoryBufferRef Object,
+                                          LLVMContext &Context);
+
+  /// \brief Parse function index in the given memory buffer.
+  /// Return new FunctionIndexObjectFile instance containing parsed function
+  /// summary/index.
+  static ErrorOr<std::unique_ptr<FunctionIndexObjectFile>>
+      create(MemoryBufferRef Object, LLVMContext &Context,
+             bool ReadFuncSummaryData = true);
+
+  /// \brief Parse the function summary information for function with the
+  /// given name out of the given buffer. Parsed information is
+  /// stored on the index object saved in this object.
+  std::error_code findFunctionSummaryInMemBuffer(
+          MemoryBufferRef Object, LLVMContext &Context, StringRef FunctionName);
+};
+}
+}
+
+#endif

lib/Bitcode/Reader/BitcodeReader.cpp

 #include "llvm/IR/DiagnosticPrinter.h"
 #include "llvm/IR/GVMaterializer.h"
 #include "llvm/IR/InlineAsm.h"
 #include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/OperandTraits.h"
 #include "llvm/IR/Operator.h"
+#include "llvm/IR/FunctionInfo.h"
 #include "llvm/IR/ValueHandle.h"
 #include "llvm/Support/DataStream.h"
 #include "llvm/Support/ManagedStatic.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include "llvm/Support/raw_ostream.h"
 #include <deque>
 using namespace llvm;
private:
   std::error_code parseUseLists();
   std::error_code initStream(std::unique_ptr<DataStreamer> Streamer);
   std::error_code initStreamFromBuffer();
   std::error_code initLazyStream(std::unique_ptr<DataStreamer> Streamer);
   std::error_code findFunctionInStream(
       Function *F,
       DenseMap<Function *, uint64_t>::iterator DeferredFunctionInfoIterator);
 };
+
+/// Class to manage reading and parsing function summary index bitcode
+/// files/sections.
+class FunctionIndexBitcodeReader {
+  LLVMContext &Context;
+  DiagnosticHandlerFunction DiagnosticHandler;
+
+  /// Eventually points to the function index built during parsing.
+  FunctionInfoIndex *TheIndex = nullptr;
+
+  std::unique_ptr<MemoryBuffer> Buffer;
+  std::unique_ptr<BitstreamReader> StreamFile;
+  BitstreamCursor Stream;
+
+  /// \brief Used to indicate whether we are doing lazy parsing of summary data.
+  ///
+  /// If true, the summary section is fully parsed into the index during
+  /// the initial parse. Otherwise, if false, the caller is expected to
+  /// invoke \a readFunctionSummary for each summary needed, and the summary
+  /// section is thus parsed lazily.
+  bool ParseFuncSummaryData = true;
+
+  /// Used to indicate whether caller only wants to check for the presence
+  /// of the function summary bitcode section. All blocks are skipped,
+  /// but the SeenFuncSummary boolean is set.
+  bool CheckFuncSummaryPresenceOnly = false;
+
+  /// Indicates whether we have encountered a function summary section
+  /// yet during parsing, used when checking if file contains function
+  /// summary section.
+  bool SeenFuncSummary = false;
+
+  /// \brief Map populated during function summary section parsing, and
+  /// consumed during ValueSymbolTable parsing.
+  ///
+  /// Used to correlate summary records with VST entries. For the per-module
+  /// index this maps the ValueID to the parsed function summary, and
+  /// for the combined index this maps the summary record's bitcode
+  /// offset to the function summary (since in the combined index the
+  /// VST records do not hold value IDs but rather hold the function
+  /// summary record offset).
+  DenseMap<uint64_t, FunctionSummary*> SummaryMap;
+
+  /// Map populated during module path string table parsing, from the
+  /// module ID to a string reference owned by the index's module
+  /// path string table, used to correlate with combined index function
+  /// summary records.
+  DenseMap<uint64_t, StringRef> ModuleIdMap;
+
+public:
+  std::error_code error(BitcodeError E, const Twine &Message);
+  std::error_code error(BitcodeError E);
+  std::error_code error(const Twine &Message);
+
+  FunctionIndexBitcodeReader(MemoryBuffer *Buffer, LLVMContext &Context,
+                             DiagnosticHandlerFunction DiagnosticHandler,
+                             bool ParseFuncSummary = true,
+                             bool CheckFuncSummaryPresenceOnly = false);
+  FunctionIndexBitcodeReader(LLVMContext &Context,
+                             DiagnosticHandlerFunction DiagnosticHandler,
+                             bool ParseFuncSummary = true,
+                             bool CheckFuncSummaryPresenceOnly = false);
+  ~FunctionIndexBitcodeReader() { freeState(); }
+
+  void freeState();
+
+  void releaseBuffer();
+
+  /// Check if the parser has encountered a function summary section.
+  bool foundFuncSummary() {
+    return SeenFuncSummary;
+  }
+
+  /// \brief Main interface to parsing a bitcode buffer.
+  /// \returns true if an error occurred.
+  std::error_code parseSummaryIndexInto(std::unique_ptr<DataStreamer> Streamer,
+                                        FunctionInfoIndex *I);
+
+  /// \brief Interface for parsing a function summary lazily.
+  std::error_code parseFunctionSummary(std::unique_ptr<DataStreamer> Streamer,
+                                       FunctionInfoIndex *I,
+                                       size_t FunctionSummaryOffset);
+
+private:
+  std::error_code parseModule();
+  std::error_code parseValueSymbolTable();
+  std::error_code parseEntireSummary();
+  std::error_code parseModuleStringTable();
+  std::error_code initStream(std::unique_ptr<DataStreamer> Streamer);
+  std::error_code initStreamFromBuffer();
+  std::error_code initLazyStream(std::unique_ptr<DataStreamer> Streamer);
+};
 } // namespace
 
 BitcodeDiagnosticInfo::BitcodeDiagnosticInfo(std::error_code EC,
                                              DiagnosticSeverity Severity,
                                              const Twine &Msg)
     : DiagnosticInfo(DK_Bitcode, Severity), Msg(Msg), EC(EC) {}
 
 void BitcodeDiagnosticInfo::print(DiagnosticPrinter &DP) const { DP << Msg; }
case BitstreamEntry::SubBlock:
         // Suspend parsing when we reach the function bodies. Subsequent
         // materialization calls will resume it when necessary. If the bitcode
         // file is old, the symbol table will be at the end instead and will not
         // have been seen yet. In this case, just finish the parse now.
         if (SeenValueSymbolTable) {
           NextUnreadBit = Stream.GetCurrentBitNo();
           return std::error_code();
         }
+        //assert(false); // Temp assert to make sure we don't fall back to old code.
         break;
       case bitc::USELIST_BLOCK_ID:
         if (std::error_code EC = parseUseLists())
           return EC;
         break;
       }
       continue;
 
if (isBitcodeWrapper(buf, buf + 4)) {
     const unsigned char *bitcodeEnd = buf + 16;
     SkipBitcodeWrapperHeader(bitcodeStart, bitcodeEnd, false);
     Bytes.dropLeadingBytes(bitcodeStart - buf);
     Bytes.setKnownObjectSize(bitcodeEnd - bitcodeStart);
   }
   return std::error_code();
 }
 
+std::error_code FunctionIndexBitcodeReader::error(BitcodeError E,
+                                                  const Twine &Message) {
+  return ::error(DiagnosticHandler, make_error_code(E), Message);
+}
+
+std::error_code FunctionIndexBitcodeReader::error(const Twine &Message) {
+  return ::error(DiagnosticHandler,
+                 make_error_code(BitcodeError::CorruptedBitcode), Message);
+}
+
+std::error_code FunctionIndexBitcodeReader::error(BitcodeError E) {
+  return ::error(DiagnosticHandler, make_error_code(E));
+}
+
+FunctionIndexBitcodeReader::FunctionIndexBitcodeReader(
+    MemoryBuffer *Buffer, LLVMContext &Context,
+    DiagnosticHandlerFunction DiagnosticHandler,
+    bool ParseFuncSummary, bool CheckFuncSummaryPresenceOnly)
+    : Context(Context),
+      DiagnosticHandler(getDiagHandler(DiagnosticHandler, Context)),
+      Buffer(Buffer),
+      ParseFuncSummaryData(ParseFuncSummary),
+      CheckFuncSummaryPresenceOnly(CheckFuncSummaryPresenceOnly) {}
+
+FunctionIndexBitcodeReader::FunctionIndexBitcodeReader(
+    LLVMContext &Context, DiagnosticHandlerFunction DiagnosticHandler,
+    bool ParseFuncSummary, bool CheckFuncSummaryPresenceOnly)
+    : Context(Context),
+      DiagnosticHandler(getDiagHandler(DiagnosticHandler, Context)),
+      Buffer(nullptr),
+      ParseFuncSummaryData(ParseFuncSummary),
+      CheckFuncSummaryPresenceOnly(CheckFuncSummaryPresenceOnly) {}
+
+void FunctionIndexBitcodeReader::freeState() {
+  Buffer = nullptr;
+}
+
+void FunctionIndexBitcodeReader::releaseBuffer() { Buffer.release(); }
+
+// Specialized value symbol table parser used when reading function index
+// blocks where we don't actually create global values.
+// At the end of this routine the function index is populated with a map
+// from function name to FunctionInfo. The function info contains
+// the function block's bitcode offset as well as the offset into the
+// function summary section.
+std::error_code FunctionIndexBitcodeReader::parseValueSymbolTable() {
+  if (Stream.EnterSubBlock(bitc::VALUE_SYMTAB_BLOCK_ID))
+    return error("Invalid record");
+
+  SmallVector<uint64_t, 64> Record;
+
+  // Read all the records for this value table.
+  SmallString<128> ValueName;
+  while (1) {
+    BitstreamEntry Entry = Stream.advanceSkippingSubblocks();
+
+    switch (Entry.Kind) {
+    case BitstreamEntry::SubBlock: // Handled for us already.
+    case BitstreamEntry::Error:
+      return error("Malformed block");
+    case BitstreamEntry::EndBlock:
+      return std::error_code();
+    case BitstreamEntry::Record:
+      // The interesting case.
+      break;
+    }
+
+    // Read a record.
+    Record.clear();
+    switch (Stream.readRecord(Entry.ID, Record)) {
+      default:  // Default behavior: ignore (e.g. VST_CODE_BBENTRY records).
+        break;
+      case bitc::VST_CODE_FNENTRY: {
+        // VST_FNENTRY: [valueid, offset, namechar x N]
+        if (convertToString(Record, 2, ValueName))
+          return error("Invalid record");
+        unsigned ValueID = Record[0];
+        uint64_t FuncOffset = Record[1];
+        FunctionInfo *FuncInfo = new FunctionInfo(FuncOffset);
+        TheIndex->addFunctionInfo(ValueName, FuncInfo);
+        if (foundFuncSummary() && ParseFuncSummaryData) {
+          DenseMap<uint64_t, FunctionSummary*>::iterator SMI =
+              SummaryMap.find(ValueID);
+          assert(SMI != SummaryMap.end() && "Summary info not found");
+          FunctionSummary *FS = SMI->second;
+          FuncInfo->setFunctionSummary(FS);
+        }
+
+        ValueName.clear();
+        break;
+      }
+      case bitc::VST_CODE_COMBINED_FNENTRY: {
+        // VST_FNENTRY: [offset, namechar x N]
+        if (convertToString(Record, 1, ValueName))
+          return error("Invalid record");
+        uint64_t FuncSummaryOffset = Record[0];
+        FunctionInfo *FuncInfo = new FunctionInfo(FuncSummaryOffset);
+        TheIndex->addFunctionInfo(ValueName, FuncInfo);
+        if (foundFuncSummary() && ParseFuncSummaryData) {
+          DenseMap<uint64_t, FunctionSummary*>::iterator SMI =
+              SummaryMap.find(FuncSummaryOffset);
+          assert(SMI != SummaryMap.end() && "Summary info not found");
+          FunctionSummary *FS = SMI->second;
+          FuncInfo->setFunctionSummary(FS);
+        }
+
+        ValueName.clear();
+        break;
+      }
+    }
+  }
+}
+
+// Parse just the blocks needed for function index building out of the module.
+// At the end of this routine the function Index is populated with a map
+// from function name to FunctionInfo. The function info contains
+// either the parsed function summary information (when parsing summaries
+// eagerly), or just to the function summary record's offset
+// if parsing lazily (!ParseFuncSummaryData).
+std::error_code FunctionIndexBitcodeReader::parseModule() {
+  if (Stream.EnterSubBlock(bitc::MODULE_BLOCK_ID))
+    return error("Invalid record");
+
+  // Read the function index for this module.
+  while (1) {
+    BitstreamEntry Entry = Stream.advance();
+
+    switch (Entry.Kind) {
+    case BitstreamEntry::Error:
+      return error("Malformed block");
+    case BitstreamEntry::EndBlock:
+      return std::error_code();
+
+    case BitstreamEntry::SubBlock:
+      if (CheckFuncSummaryPresenceOnly) {
+        if (Entry.ID == bitc::FUNCTION_SUMMARY_BLOCK_ID)
+          SeenFuncSummary = true;
+        if (Stream.SkipBlock())
+          return error("Invalid record");
+        continue;
+      }
+      switch (Entry.ID) {
+      default:  // Skip unknown content.
+        if (Stream.SkipBlock())
+          return error("Invalid record");
+        break;
+      case bitc::BLOCKINFO_BLOCK_ID:
+        // Need to parse these to get abbrev ids (e.g. for VST)
+        if (Stream.ReadBlockInfoBlock())
+          return error("Malformed block");
+        break;
+      case bitc::VALUE_SYMTAB_BLOCK_ID:
+        if (std::error_code EC = parseValueSymbolTable())
+          return EC;
+        break;
+      case bitc::FUNCTION_SUMMARY_BLOCK_ID:
+        SeenFuncSummary = true;
+        if (!ParseFuncSummaryData) {
+          // Lazy parsing of summary info, skip it.
+          if (Stream.SkipBlock())
+            return error("Invalid record");
+        }
+        else if (std::error_code EC = parseEntireSummary())
+          return EC;
+        break;
+      case bitc::MODULE_STRTAB_BLOCK_ID:
+        if (std::error_code EC = parseModuleStringTable())
+          return EC;
+        break;
+      }
+      continue;
+
+    case BitstreamEntry::Record:
+      Stream.skipRecord(Entry.ID);
+      continue;
+    }
+  }
+}
+
+// Eagerly parse the entire function summary block (i.e. for all functions
+// in the index). This populates the FunctionSummary objects in
+// the index.
+std::error_code FunctionIndexBitcodeReader::parseEntireSummary() {
+  if (Stream.EnterSubBlock(bitc::FUNCTION_SUMMARY_BLOCK_ID))
+    return error("Invalid record");
+
+  SmallVector<uint64_t, 64> Record;
+
+  while (1) {
+    BitstreamEntry Entry = Stream.advanceSkippingSubblocks();
+
+    switch (Entry.Kind) {
+    case BitstreamEntry::SubBlock: // Handled for us already.
+    case BitstreamEntry::Error:
+      return error("Malformed block");
+    case BitstreamEntry::EndBlock:
+      return std::error_code();
+    case BitstreamEntry::Record:
+      // The interesting case.
+      break;
+    }
+
+    // Read a record. The record format depends on whether this
+    // is a per-module index or a combined index file. In the per-module
+    // case the records contain the associated value's ID for correlation
+    // with VST entries. In the combined index the correlation is done
+    // via the bitcode offset of the summary records (which were saved
+    // in the combined index VST entries). The records also contain
+    // information used for ThinLTO renaming and importing.
+    Record.clear();
+    uint64_t CurRecordBit = Stream.GetCurrentBitNo();
+    switch (Stream.readRecord(Entry.ID, Record)) {
+    default:  // Default behavior: ignore.
+      break;
+    // FS_PERMODULE_ENTRY: [valueid, islocal, instcount]
+    case bitc::FS_CODE_PERMODULE_ENTRY: {
+      unsigned ValueID = Record[0];
+      bool IsLocal = Record[1];
+      unsigned InstCount = Record[2];
+      FunctionSummary *FS = new FunctionSummary(InstCount);
+      FS->setLocalFunction(IsLocal);
+      FS->setModulePath(Buffer->getBufferIdentifier());
+      SummaryMap[ValueID] = FS;
+    }
+    // FS_COMBINED_ENTRY: [modid, instcount]
+    case bitc::FS_CODE_COMBINED_ENTRY: {
+      uint64_t ModuleId = Record[0];
+      unsigned InstCount = Record[1];
+      FunctionSummary *FS = new FunctionSummary(InstCount);
+      FS->setModulePath(ModuleIdMap[ModuleId]);
+      SummaryMap[CurRecordBit] = FS;
+    }
+    }
+  }
+  llvm_unreachable("Exit infinite loop");
+}
+
+// Parse the  module string table block into the Index.
+// This populates the ModulePathStringTable map in the index.
+std::error_code FunctionIndexBitcodeReader::parseModuleStringTable() {
+  if (Stream.EnterSubBlock(bitc::MODULE_STRTAB_BLOCK_ID))
+    return error("Invalid record");
+
+  SmallVector<uint64_t, 64> Record;
+
+  SmallString<128> ModulePath;
+  while (1) {
+    BitstreamEntry Entry = Stream.advanceSkippingSubblocks();
+
+    switch (Entry.Kind) {
+    case BitstreamEntry::SubBlock: // Handled for us already.
+    case BitstreamEntry::Error:
+      return error("Malformed block");
+    case BitstreamEntry::EndBlock:
+      return std::error_code();
+    case BitstreamEntry::Record:
+      // The interesting case.
+      break;
+    }
+
+    Record.clear();
+    switch (Stream.readRecord(Entry.ID, Record)) {
+    default:  // Default behavior: ignore.
+      break;
+      case bitc::MST_CODE_ENTRY: {
+        // MST_ENTRY: [modid, namechar x N]
+        if (convertToString(Record, 1, ModulePath))
+          return error("Invalid record");
+        uint64_t ModuleId = Record[0];
+        StringRef ModulePathInMap = TheIndex->addModulePath(ModulePath,
+                                                            ModuleId);
+        ModuleIdMap[ModuleId] = ModulePathInMap;
+        ModulePath.clear();
+        break;
+      }
+    }
+  }
+  llvm_unreachable("Exit infinite loop");
+}
+
+// Parse the function info index from the bitcode streamer into the given index.
+std::error_code
+FunctionIndexBitcodeReader::parseSummaryIndexInto(
+    std::unique_ptr<DataStreamer> Streamer, FunctionInfoIndex *I) {
+  TheIndex = I;
+
+  if (std::error_code EC = initStream(std::move(Streamer)))
+    return EC;
+
+  // Sniff for the signature.
+  if (Stream.Read(8) != 'B' ||
+      Stream.Read(8) != 'C' ||
+      Stream.Read(4) != 0x0 ||
+      Stream.Read(4) != 0xC ||
+      Stream.Read(4) != 0xE ||
+      Stream.Read(4) != 0xD)
+    return error("Invalid bitcode signature");
+
+  // We expect a number of well-defined blocks, though we don't necessarily
+  // need to understand them all.
+  while (1) {
+    if (Stream.AtEndOfStream()) {
+      // We didn't really read a proper Module block.
+      return error("Malformed ThihLTO IR file");
+    }
+
+    BitstreamEntry Entry =
+      Stream.advance(BitstreamCursor::AF_DontAutoprocessAbbrevs);
+
+    if (Entry.Kind != BitstreamEntry::SubBlock)
+      return error("Malformed block");
+
+    // If we see a MODULE_BLOCK, parse it to find the blocks needed for
+    // building the function summary index.
+    if (Entry.ID == bitc::MODULE_BLOCK_ID)
+      return parseModule();
+
+    if (Stream.SkipBlock())
+      return error("Invalid record");
+  }
+}
+
+// Parse the function information at the given offset in the buffer into
+// the index. Used to support lazy parsing of function summaries from the
+// combined index during importing.
+// TODO: This function is not yet complete as it won't have a consumer
+// until ThinLTO function importing is added.
+std::error_code
+FunctionIndexBitcodeReader::parseFunctionSummary(
+    std::unique_ptr<DataStreamer> Streamer,
+    FunctionInfoIndex *I,
+    size_t FunctionSummaryOffset) {
+  TheIndex = I;
+
+  if (std::error_code EC = initStream(std::move(Streamer)))
+    return EC;
+
+  // Sniff for the signature.
+  if (Stream.Read(8) != 'B' ||
+      Stream.Read(8) != 'C' ||
+      Stream.Read(4) != 0x0 ||
+      Stream.Read(4) != 0xC ||
+      Stream.Read(4) != 0xE ||
+      Stream.Read(4) != 0xD)
+    return error("Invalid bitcode signature");
+
+  Stream.JumpToBit(FunctionSummaryOffset);
+
+  BitstreamEntry Entry = Stream.advanceSkippingSubblocks();
+
+  switch (Entry.Kind) {
+    default:
+      return error("Malformed block");
+    case BitstreamEntry::Record:
+      // The expected case.
+      break;
+  }
+
+  // TODO: Read a record. This interface will be completed when ThinLTO
+  // importing is added so that it can be tested.
+  SmallVector<uint64_t, 64> Record;
+  switch (Stream.readRecord(Entry.ID, Record)) {
+    default:
+      return error("Invalid record");
+  }
+
+  return std::error_code();
+}
+
+std::error_code
+FunctionIndexBitcodeReader::initStream(std::unique_ptr<DataStreamer> Streamer) {
+  if (Streamer)
+    return initLazyStream(std::move(Streamer));
+  return initStreamFromBuffer();
+}
+
+std::error_code FunctionIndexBitcodeReader::initStreamFromBuffer() {
+  const unsigned char *BufPtr = (const unsigned char*)Buffer->getBufferStart();
+  const unsigned char *BufEnd = BufPtr+Buffer->getBufferSize();
+
+  if (Buffer->getBufferSize() & 3)
+    return error("Invalid bitcode signature");
+
+  // If we have a wrapper header, parse it and ignore the non-bc file contents.
+  // The magic number is 0x0B17C0DE stored in little endian.
+  if (isBitcodeWrapper(BufPtr, BufEnd))
+    if (SkipBitcodeWrapperHeader(BufPtr, BufEnd, true))
+      return error("Invalid bitcode wrapper header");
+
+  StreamFile.reset(new BitstreamReader(BufPtr, BufEnd));
+  Stream.init(&*StreamFile);
+
+  return std::error_code();
+}
+
+std::error_code
+FunctionIndexBitcodeReader::initLazyStream(std::unique_ptr<DataStreamer> Streamer) {
+  // Check and strip off the bitcode wrapper; BitstreamReader expects never to
+  // see it.
+  auto OwnedBytes =
+      llvm::make_unique<StreamingMemoryObject>(std::move(Streamer));
+  StreamingMemoryObject &Bytes = *OwnedBytes;
+  StreamFile = llvm::make_unique<BitstreamReader>(std::move(OwnedBytes));
+  Stream.init(&*StreamFile);
+
+  unsigned char buf[16];
+  if (Bytes.readBytes(buf, 16, 0) != 16)
+    return error("Invalid bitcode signature");
+
+  if (!isBitcode(buf, buf + 16))
+    return error("Invalid bitcode signature");
+
+  if (isBitcodeWrapper(buf, buf + 4)) {
+    const unsigned char *bitcodeStart = buf;
+    const unsigned char *bitcodeEnd = buf + 16;
+    SkipBitcodeWrapperHeader(bitcodeStart, bitcodeEnd, false);
+    Bytes.dropLeadingBytes(bitcodeStart - buf);
+    Bytes.setKnownObjectSize(bitcodeEnd - bitcodeStart);
+  }
+  return std::error_code();
+}
+
 namespace {
 class BitcodeErrorCategoryType : public std::error_category {
   const char *name() const LLVM_NOEXCEPT override {
     return "llvm.bitcode";
   }
   std::string message(int IE) const override {
     BitcodeError E = static_cast<BitcodeError>(IE);
     switch (E) {
llvm::getBitcodeTargetTriple(MemoryBufferRef Buffer, LLVMContext &Context,
   std::unique_ptr<MemoryBuffer> Buf = MemoryBuffer::getMemBuffer(Buffer, false);
   auto R = llvm::make_unique<BitcodeReader>(Buf.release(), Context,
                                             DiagnosticHandler);
   ErrorOr<std::string> Triple = R->parseTriple();
   if (Triple.getError())
     return "";
   return Triple.get();
 }
+
+// Parse the specified bitcode buffer, returning the function info index.
+// If ParseFuncSummary is true, parse the entire function summary into
+// the index. Otherwise skip the function summary section, and only create
+// an index object with a map from function name to function summary offset.
+// The index is used to perform lazy function summary reading later.
+ErrorOr<std::unique_ptr<FunctionInfoIndex>>
+llvm::getFunctionInfoIndex(MemoryBufferRef Buffer, LLVMContext &Context,
+                           DiagnosticHandlerFunction DiagnosticHandler,
+                           bool ParseFuncSummary) {
+  std::unique_ptr<MemoryBuffer> Buf = MemoryBuffer::getMemBuffer(Buffer, false);
+  FunctionIndexBitcodeReader *R =
+      new FunctionIndexBitcodeReader(Buf.get(), Context, DiagnosticHandler,
+                                     ParseFuncSummary);
+
+  std::unique_ptr<FunctionInfoIndex> Index =
+      llvm::make_unique<FunctionInfoIndex>();
+
+  auto cleanupOnError = [&](std::error_code EC) {
+    R->releaseBuffer(); // Never take ownership on error.
+    return EC;
+  };
+
+  if (std::error_code EC = R->parseSummaryIndexInto(nullptr, Index.get()))
+    return cleanupOnError(EC);
+
+  Buf.release(); // The FunctionIndexBitcodeReader owns it now.
+  return std::move(Index);
+}
+
+// Check if the given bitcode buffer contains a function summary block.
+bool llvm::hasFunctionSummary(MemoryBufferRef Buffer, LLVMContext &Context,
+    DiagnosticHandlerFunction DiagnosticHandler) {
+  std::unique_ptr<MemoryBuffer> Buf = MemoryBuffer::getMemBuffer(Buffer, false);
+  FunctionIndexBitcodeReader *R =
+      new FunctionIndexBitcodeReader(Buf.get(), Context, DiagnosticHandler,
+                                     false, true);
+
+  auto cleanupOnError = [&](std::error_code EC) {
+    R->releaseBuffer(); // Never take ownership on error.
+    return false;
+  };
+
+  if (std::error_code EC = R->parseSummaryIndexInto(nullptr, nullptr))
+    return cleanupOnError(EC);
+
+  Buf.release(); // The FunctionIndexBitcodeReader owns it now.
+  return R->foundFuncSummary();
+}
+
+// This method supports lazy reading of function summary data from the combined
+// index during ThinLTO function importing. When reading the combined index
+// file, getFunctionInfoIndex is first invoked with ParseFuncSummary=false.
+// Then this method is called for each function considered for importing,
+// to parse the summary information for the given function name into
+// the index.
+std::error_code llvm::readFunctionSummary(
+    MemoryBufferRef Buffer, LLVMContext &Context,
+    DiagnosticHandlerFunction DiagnosticHandler,
+    StringRef FunctionName,
+    std::unique_ptr<FunctionInfoIndex> Index) {
+  std::unique_ptr<MemoryBuffer> Buf = MemoryBuffer::getMemBuffer(Buffer, false);
+  FunctionIndexBitcodeReader *R =
+      new FunctionIndexBitcodeReader(Buf.get(), Context, DiagnosticHandler);
+
+  auto cleanupOnError = [&](std::error_code EC) {
+    R->releaseBuffer(); // Never take ownership on error.
+    return EC;
+  };
+
+  // Lookup the given function name in the FunctionMap, which may
+  // contain a list of function infos in the case of a COMDAT. Walk through
+  // and parse each function summary info at the function summary offset
+  // recorded when parsing the value symbol table.
+  FunctionInfoList FuncInfos = Index->getFunctionInfoList(FunctionName);
+  for (std::vector<FunctionInfo*>::iterator FII = FuncInfos.begin();
+       FII != FuncInfos.end(); FII++) {
+    size_t FunctionSummaryOffset = (*FII)->bitcodeIndex();
+    if (std::error_code EC = R->parseFunctionSummary(nullptr, Index.get(),
+                                                     FunctionSummaryOffset))
+      return cleanupOnError(EC);
+  }
+
+  Buf.release(); // The FunctionIndexBitcodeReader owns it now.
+  return std::error_code();
+}

lib/Bitcode/Writer/BitcodeWriter.cpp

 #include "llvm/ADT/Triple.h"
 #include "llvm/Bitcode/BitstreamWriter.h"
 #include "llvm/Bitcode/LLVMBitCodes.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/DebugInfoMetadata.h"
 #include "llvm/IR/DerivedTypes.h"
 #include "llvm/IR/InlineAsm.h"
 #include "llvm/IR/Instructions.h"
+#include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/Operator.h"
 #include "llvm/IR/UseListOrder.h"
 #include "llvm/IR/ValueSymbolTable.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/Program.h"

 /// Emit names for globals/functions etc. The VSTOffsetPlaceholder,
 /// BitcodeStartBit and FunctionIndex are only passed for the module-level
 /// VST, where we are including a function bitcode index and need to
 /// backpatch the VST forward declaration record.
 static void WriteValueSymbolTable(
     const ValueSymbolTable &VST, const ValueEnumerator &VE,
     BitstreamWriter &Stream, uint64_t VSTOffsetPlaceholder = 0,
     uint64_t BitcodeStartBit = 0,
-    DenseMap<const Function *, uint64_t> *FunctionIndex = nullptr) {
+    DenseMap<const Function *, FunctionInfo *> *FunctionIndex = nullptr) {
   if (VST.empty()) {
     // WriteValueSymbolTableForwardDecl should have returned early as
     // well. Ensure this handling remains in sync by asserting that
     // the placeholder offset is not set.
     assert(VSTOffsetPlaceholder == 0);
     return;
   }
 
if (isa<BasicBlock>(Name.getValue())) {
       // have a VSTOffsetPlaceholder. The function-level VST should not
       // contain any Function symbols.
       assert(FunctionIndex);
       assert(VSTOffsetPlaceholder > 0);
 
       // Save the word offset of the function (from the start of the
       // actual bitcode written to the stream).
       assert(FunctionIndex->count(F) == 1);
-      uint64_t BitcodeIndex = (*FunctionIndex)[F] - BitcodeStartBit;
+      uint64_t BitcodeIndex =
+          (*FunctionIndex)[F]->bitcodeIndex() - BitcodeStartBit;
       assert((BitcodeIndex & 31) == 0 && "function block not 32-bit aligned");
       NameVals.push_back(BitcodeIndex / 32);
 
       Code = bitc::VST_CODE_FNENTRY;
       AbbrevToUse = FnEntry8BitAbbrev;
       if (Bits == SE_Char6)
         AbbrevToUse = FnEntry6BitAbbrev;
       else if (Bits == SE_Fixed7)
for (const ValueName &Name : VST) {
 
     // Emit the finished record.
     Stream.EmitRecord(Code, NameVals, AbbrevToUse);
     NameVals.clear();
   }
   Stream.ExitBlock();
 }
 
+/// Emit function names and summary offsets for the combined index
+/// used by ThinLTO.
+static void WriteCombinedValueSymbolTable(
+    const FunctionInfoIndex *Index, BitstreamWriter &Stream) {
+  Stream.EnterSubblock(bitc::VALUE_SYMTAB_BLOCK_ID, 4);
+
+  // 8-bit fixed-width VST_COMBINED_FNENTRY function strings.
+  BitCodeAbbrev *Abbv = new BitCodeAbbrev();
+  Abbv->Add(BitCodeAbbrevOp(bitc::VST_CODE_COMBINED_FNENTRY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // value id
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // funcoffset
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 8));
+  unsigned FnEntry8BitAbbrev = Stream.EmitAbbrev(Abbv);
+
+  // 7-bit fixed width VST_COMBINED_FNENTRY function strings.
+  Abbv = new BitCodeAbbrev();
+  Abbv->Add(BitCodeAbbrevOp(bitc::VST_CODE_COMBINED_FNENTRY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // value id
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // funcoffset
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 7));
+  unsigned FnEntry7BitAbbrev = Stream.EmitAbbrev(Abbv);
+
+  // 6-bit char6 VST_COMBINED_FNENTRY function strings.
+  Abbv = new BitCodeAbbrev();
+  Abbv->Add(BitCodeAbbrevOp(bitc::VST_CODE_COMBINED_FNENTRY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // value id
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // funcoffset
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Char6));
+  unsigned FnEntry6BitAbbrev = Stream.EmitAbbrev(Abbv);
+
+  // FIXME: We know if the type names can use 7-bit ascii.
+  SmallVector<unsigned, 64> NameVals;
+
+  for (const_funcinfo_iterator FII = Index->funcinfo_begin();
+       FII != Index->funcinfo_end(); ++FII) {
+    FunctionInfoList List = FII->second;
+    for (std::vector<FunctionInfo*>::const_iterator FLI =
+         FII->getValue().begin(); FLI != FII->getValue().end(); ++FLI) {
+      FunctionInfo *FI = *FLI;
+
+      NameVals.push_back(FI->bitcodeIndex());
+
+      StringRef FuncName = FII->first();
+
+      // Figure out the encoding to use for the name.
+      StringEncoding Bits =
+          getStringEncoding(FuncName.data(), FuncName.size());
+
+      // VST_COMBINED_FNENTRY: [funcsumoffset, namechar x N]
+      unsigned AbbrevToUse = FnEntry8BitAbbrev;
+      if (Bits == SE_Char6)
+        AbbrevToUse = FnEntry6BitAbbrev;
+      else if (Bits == SE_Fixed7)
+        AbbrevToUse = FnEntry7BitAbbrev;
+
+      for (const char *P = FuncName.data(),
+           *E = FuncName.data()+FuncName.size(); P != E; ++P)
+        NameVals.push_back((unsigned char)*P);
+
+      // Emit the finished record.
+      Stream.EmitRecord(bitc::VST_CODE_COMBINED_FNENTRY, NameVals, AbbrevToUse);
+      NameVals.clear();
+    }
+  }
+  Stream.ExitBlock();
+}
+
 static void WriteUseList(ValueEnumerator &VE, UseListOrder &&Order,
                          BitstreamWriter &Stream) {
   assert(Order.Shuffle.size() >= 2 && "Shuffle too small");
   unsigned Code;
   if (isa<BasicBlock>(Order.V))
     Code = bitc::USELIST_CODE_BB;
   else
     Code = bitc::USELIST_CODE_DEFAULT;
static void WriteUseListBlock(const Function *F, ValueEnumerator &VE,
   Stream.EnterSubblock(bitc::USELIST_BLOCK_ID, 3);
   while (hasMore()) {
     WriteUseList(VE, std::move(VE.UseListOrders.back()), Stream);
     VE.UseListOrders.pop_back();
   }
   Stream.ExitBlock();
 }
 
-/// WriteFunction - Emit a function body to the module stream.
-static void WriteFunction(const Function &F, ValueEnumerator &VE,
-                          BitstreamWriter &Stream,
-                          DenseMap<const Function *, uint64_t> &FunctionIndex) {
+/// \brief Save information for the given function into the function index.
+///
+/// At a minimum this saves the bitcode index of the function record that
+/// was just written. However, if we are emitting function summary information,
+/// for example for ThinLTO, then a \a FunctionSummary object is created
+/// to hold the provided summary information.
+static void SaveFunctionInfo(
+    const Function &F,
+    DenseMap<const Function *, FunctionInfo *> &FunctionIndex,
+    unsigned NumInsts, uint64_t BitcodeIndex, bool EmitFunctionSummary) {
+  FunctionSummary *FuncSummary = nullptr;
+  if (EmitFunctionSummary) {
+    FuncSummary = new FunctionSummary(NumInsts);
+    FuncSummary->setLocalFunction(F.hasLocalLinkage());
+  }
+  FunctionInfo *FuncInfo = new FunctionInfo(BitcodeIndex, FuncSummary);
+
+  FunctionIndex[&F] = FuncInfo;
+}
+
+/// Emit a function body to the module stream.
+static void WriteFunction(
+    const Function &F, ValueEnumerator &VE, BitstreamWriter &Stream,
+    DenseMap<const Function *, FunctionInfo *> &FunctionIndex,
+    bool EmitFunctionSummary) {
   // Save the bitcode index of the start of this function block for recording
   // in the VST.
   uint64_t BitcodeIndex = Stream.GetCurrentBitNo();
-  FunctionIndex[&F] = BitcodeIndex;
 
   Stream.EnterSubblock(bitc::FUNCTION_BLOCK_ID, 4);
   VE.incorporateFunction(F);
 
   SmallVector<unsigned, 64> Vals;
 
   // Emit the number of basic blocks, so the reader can create them ahead of
   // time.
static void WriteFunction(
   WriteFunctionLocalMetadata(F, VE, Stream);
 
   // Keep a running idea of what the instruction ID is.
   unsigned InstID = CstEnd;
 
   bool NeedsMetadataAttachment = F.hasMetadata();
 
   DILocation *LastDL = nullptr;
+  unsigned NumInsts = 0;
 
   // Finally, emit all the instructions, in order.
   for (Function::const_iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
     for (BasicBlock::const_iterator I = BB->begin(), E = BB->end();
          I != E; ++I) {
       WriteInstruction(*I, InstID, VE, Stream, Vals);
 
+      if (!isa<DbgInfoIntrinsic>(I))
+        ++NumInsts;
+
       if (!I->getType()->isVoidTy())
         ++InstID;
 
       // If the instruction has metadata, write a metadata attachment later.
       NeedsMetadataAttachment |= I->hasMetadataOtherThanDebugLoc();
 
       // If the instruction has a debug location, emit it.
       DILocation *DL = I->getDebugLoc();
static void WriteFunction(
   WriteValueSymbolTable(F.getValueSymbolTable(), VE, Stream);
 
   if (NeedsMetadataAttachment)
     WriteMetadataAttachment(F, VE, Stream);
   if (VE.shouldPreserveUseListOrder())
     WriteUseListBlock(&F, VE, Stream);
   VE.purgeFunction();
   Stream.ExitBlock();
+
+  SaveFunctionInfo(F, FunctionIndex, NumInsts, BitcodeIndex,
+                   EmitFunctionSummary);
 }
 
 // Emit blockinfo, which defines the standard abbreviations etc.
 static void WriteBlockInfo(const ValueEnumerator &VE, BitstreamWriter &Stream) {
   // We only want to emit block info records for blocks that have multiple
   // instances: CONSTANTS_BLOCK, FUNCTION_BLOCK and VALUE_SYMTAB_BLOCK.
   // Other blocks can define their abbrevs inline.
   Stream.EnterBlockInfoBlock(2);
// FIXME: This should only use space for first class types!
     if (Stream.EmitBlockInfoAbbrev(bitc::FUNCTION_BLOCK_ID, Abbv) !=
         FUNCTION_INST_GEP_ABBREV)
       llvm_unreachable("Unexpected abbrev ordering!");
   }
 
   Stream.ExitBlock();
 }
 
+/// Write the module path strings, currently only used when generating
+/// a combined index file.
+static void WriteModStrings(const FunctionInfoIndex *I,
+                            BitstreamWriter &Stream) {
+  Stream.EnterSubblock(bitc::MODULE_STRTAB_BLOCK_ID, 3);
+
+  // TODO: See which abbrev sizes we actually need to emit
+
+  // 8-bit fixed-width MST_ENTRY strings.
+  BitCodeAbbrev *Abbv = new BitCodeAbbrev();
+  Abbv->Add(BitCodeAbbrevOp(bitc::MST_CODE_ENTRY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 8));
+  unsigned Abbrev8Bit = Stream.EmitAbbrev(Abbv);
+
+  // 7-bit fixed width MST_ENTRY strings.
+  Abbv = new BitCodeAbbrev();
+  Abbv->Add(BitCodeAbbrevOp(bitc::MST_CODE_ENTRY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 7));
+  unsigned Abbrev7Bit = Stream.EmitAbbrev(Abbv);
+
+  // 6-bit char6 MST_ENTRY strings.
+  Abbv = new BitCodeAbbrev();
+  Abbv->Add(BitCodeAbbrevOp(bitc::MST_CODE_ENTRY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Array));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Char6));
+  unsigned Abbrev6Bit = Stream.EmitAbbrev(Abbv);
+
+  SmallVector<unsigned, 64> NameVals;
+  for (const StringMapEntry<uint64_t> &MPSE : I->modPathStringEntries()) {
+    StringEncoding Bits =
+        getStringEncoding(MPSE.getKey().data(), MPSE.getKey().size());
+    unsigned AbbrevToUse = Abbrev8Bit;
+    if (Bits == SE_Char6)
+      AbbrevToUse = Abbrev6Bit;
+    else if (Bits == SE_Fixed7)
+      AbbrevToUse = Abbrev7Bit;
+
+    NameVals.push_back(MPSE.getValue());
+
+    for (const char *P = MPSE.getKey().data(),
+         *E = MPSE.getKey().data()+MPSE.getKey().size(); P != E; ++P)
+      NameVals.push_back((unsigned char)*P);
+
+    // Emit the finished record.
+    Stream.EmitRecord(bitc::MST_CODE_ENTRY, NameVals, AbbrevToUse);
+    NameVals.clear();
+  }
+  Stream.ExitBlock();
+}
+
+/// Emit the per-module function summary section alongside the rest of
+/// the module's bitcode.
+static void WritePerModuleFunctionSummary(
+    DenseMap<const Function *, FunctionInfo *> &FunctionIndex,
+    const Module *M, const ValueEnumerator &VE, BitstreamWriter &Stream) {
+  Stream.EnterSubblock(bitc::FUNCTION_SUMMARY_BLOCK_ID, 3);
+
+  // Abbrev for FS_CODE_PERMODULE_ENTRY.
+  BitCodeAbbrev *Abbv = new BitCodeAbbrev();
+  Abbv->Add(BitCodeAbbrevOp(bitc::FS_CODE_PERMODULE_ENTRY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // valueid
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 1));  // islocal
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // instcount
+  unsigned FSAbbrev = Stream.EmitAbbrev(Abbv);
+
+  SmallVector<unsigned, 64> NameVals;
+  for (DenseMap<const Function *, FunctionInfo *>::iterator I =
+       FunctionIndex.begin(); I != FunctionIndex.end(); I++) {
+    FunctionSummary *FS = I->second->functionSummary();
+    assert(FS);
+
+    NameVals.push_back(
+        VE.getValueID(M->getValueSymbolTable().lookup(I->first->getName())));
+    NameVals.push_back(FS->isLocalFunction());
+    NameVals.push_back(FS->instCount());
+
+    // Emit the finished record.
+    Stream.EmitRecord(bitc::FS_CODE_PERMODULE_ENTRY, NameVals, FSAbbrev);
+    NameVals.clear();
+  }
+
+  Stream.ExitBlock();
+}
+
+/// Emit the combined function summary section into the combined index
+/// file.
+static void WriteCombinedFunctionSummary(const FunctionInfoIndex *I,
+                                         BitstreamWriter &Stream) {
+  Stream.EnterSubblock(bitc::FUNCTION_SUMMARY_BLOCK_ID, 3);
+
+  // Abbrev for FS_CODE_COMBINED_ENTRY.
+  BitCodeAbbrev *Abbv = new BitCodeAbbrev();
+  Abbv->Add(BitCodeAbbrevOp(bitc::FS_CODE_COMBINED_ENTRY));
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // modid
+  Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8)); // instcount
+  unsigned FSAbbrev = Stream.EmitAbbrev(Abbv);
+
+  SmallVector<unsigned, 64> NameVals;
+  for (const_funcinfo_iterator FII = I->funcinfo_begin();
+       FII != I->funcinfo_end(); ++FII) {
+    FunctionInfoList List = FII->second;
+    for (std::vector<FunctionInfo*>::const_iterator FLI =
+         FII->getValue().begin(); FLI != FII->getValue().end(); ++FLI) {
+      FunctionInfo *FI = *FLI;
+      FunctionSummary *FS = FI->functionSummary();
+      assert(FS);
+
+      NameVals.push_back(I->getModuleId(FS->modulePath()));
+      NameVals.push_back(FS->instCount());
+
+      // Record the starting offset of this summary entry for use
+      // in the VST entry. Add the current code size since the
+      // reader will invoke readRecord after the abbrev id read.
+      FI->bitcodeIndex(Stream.GetCurrentBitNo() + Stream.GetAbbrevIDWidth());
+
+      // Emit the finished record.
+      Stream.EmitRecord(bitc::FS_CODE_COMBINED_ENTRY, NameVals, FSAbbrev);
+      NameVals.clear();
+    }
+  }
+
+  Stream.ExitBlock();
+}
+
 /// WriteModule - Emit the specified module to the bitstream.
 static void WriteModule(const Module *M, BitstreamWriter &Stream,
                         bool ShouldPreserveUseListOrder,
-                        uint64_t BitcodeStartBit) {
+                        uint64_t BitcodeStartBit,
+                        bool EmitThinLTOIndex) {
   Stream.EnterSubblock(bitc::MODULE_BLOCK_ID, 3);
 
   SmallVector<unsigned, 1> Vals;
   unsigned CurVersion = 1;
   Vals.push_back(CurVersion);
   Stream.EmitRecord(bitc::MODULE_CODE_VERSION, Vals);
 
   // Analyze the module, enumerating globals, functions, etc.
static void WriteModule(const Module *M, BitstreamWriter &Stream,
   // Emit metadata.
   WriteModuleMetadataStore(M, Stream);
 
   // Emit module-level use-lists.
   if (VE.shouldPreserveUseListOrder())
     WriteUseListBlock(nullptr, VE, Stream);
 
   // Emit function bodies.
-  DenseMap<const Function *, uint64_t> FunctionIndex;
+  DenseMap<const Function *, FunctionInfo *> FunctionIndex;
   for (Module::const_iterator F = M->begin(), E = M->end(); F != E; ++F)
     if (!F->isDeclaration())
-      WriteFunction(*F, VE, Stream, FunctionIndex);
+      WriteFunction(*F, VE, Stream, FunctionIndex, EmitThinLTOIndex);
+
+  // Need to write after the above call to WriteFunction which populates
+  // the summary information in the index.
+  if (EmitThinLTOIndex)
+    WritePerModuleFunctionSummary(FunctionIndex, M, VE, Stream);
 
   WriteValueSymbolTable(M->getValueSymbolTable(), VE, Stream,
                         VSTOffsetPlaceholder, BitcodeStartBit, &FunctionIndex);
 
   Stream.ExitBlock();
 }
 
 /// EmitDarwinBCHeader - If generating a bc file on darwin, we have to emit a
static void EmitDarwinBCHeaderAndTrailer(SmallVectorImpl<char> &Buffer,
   // If the file is not a multiple of 16 bytes, insert dummy padding.
   while (Buffer.size() & 15)
     Buffer.push_back(0);
 }
 
 /// WriteBitcodeToFile - Write the specified module to the specified output
 /// stream.
 void llvm::WriteBitcodeToFile(const Module *M, raw_ostream &Out,
-                              bool ShouldPreserveUseListOrder) {
+                              bool ShouldPreserveUseListOrder,
+                              bool EmitThinLTOIndex) {
   SmallVector<char, 0> Buffer;
   Buffer.reserve(256*1024);
 
   // If this is darwin or another generic macho target, reserve space for the
   // header.
   Triple TT(M->getTargetTriple());
   if (TT.isOSDarwin())
     Buffer.insert(Buffer.begin(), DarwinBCHeaderSize, 0);
// Emit the module into the buffer.
     Stream.Emit((unsigned)'B', 8);
     Stream.Emit((unsigned)'C', 8);
     Stream.Emit(0x0, 4);
     Stream.Emit(0xC, 4);
     Stream.Emit(0xE, 4);
     Stream.Emit(0xD, 4);
 
     // Emit the module.
-    WriteModule(M, Stream, ShouldPreserveUseListOrder, BitcodeStartBit);
+    WriteModule(M, Stream, ShouldPreserveUseListOrder, BitcodeStartBit,
+                EmitThinLTOIndex);
   }
 
   if (TT.isOSDarwin())
     EmitDarwinBCHeaderAndTrailer(Buffer, TT);
 
   // Write the generated bitstream to "Out".
   Out.write((char*)&Buffer.front(), Buffer.size());
 }
+
+// Write the specified function summary index to the given raw output stream,
+// where it will be written in a new bitcode block. This is used when
+// writing the combined index file for ThinLTO.
+void llvm::WriteFunctionSummaryToFile(const FunctionInfoIndex *Index,
+                                      raw_ostream &Out) {
+  SmallVector<char, 0> Buffer;
+  Buffer.reserve(256*1024);
+
+  BitstreamWriter Stream(Buffer);
+
+  // Emit the bitcode header.
+  Stream.Emit((unsigned)'B', 8);
+  Stream.Emit((unsigned)'C', 8);
+  Stream.Emit(0x0, 4);
+  Stream.Emit(0xC, 4);
+  Stream.Emit(0xE, 4);
+  Stream.Emit(0xD, 4);
+
+  Stream.EnterSubblock(bitc::MODULE_BLOCK_ID, 3);
+
+  SmallVector<unsigned, 1> Vals;
+  unsigned CurVersion = 1;
+  Vals.push_back(CurVersion);
+  Stream.EmitRecord(bitc::MODULE_CODE_VERSION, Vals);
+
+  // Write the module paths in the combined index.
+  WriteModStrings(Index, Stream);
+
+  // Write the function summary combined index records.
+  WriteCombinedFunctionSummary(Index, Stream);
+
+  // Need a special VST writer for the combined index (we don't have a
+  // real VST and real values when this is invoked).
+  WriteCombinedValueSymbolTable(Index, Stream);
+
+  Stream.ExitBlock();
+
+  Out.write((char*)&Buffer.front(), Buffer.size());
+}

lib/Bitcode/Writer/BitcodeWriterPass.cpp

 #include "llvm/Bitcode/BitcodeWriterPass.h"
 #include "llvm/Bitcode/ReaderWriter.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/PassManager.h"
 #include "llvm/Pass.h"
 using namespace llvm;
 
 PreservedAnalyses BitcodeWriterPass::run(Module &M) {
-  WriteBitcodeToFile(&M, OS, ShouldPreserveUseListOrder);
+  WriteBitcodeToFile(&M, OS, ShouldPreserveUseListOrder, EmitThinLTOIndex);
   return PreservedAnalyses::all();
 }
 
 namespace {
   class WriteBitcodePass : public ModulePass {
     raw_ostream &OS; // raw_ostream to print on
     bool ShouldPreserveUseListOrder;
+    bool EmitThinLTOIndex;
 
   public:
     static char ID; // Pass identification, replacement for typeid
-    explicit WriteBitcodePass(raw_ostream &o, bool ShouldPreserveUseListOrder)
+    explicit WriteBitcodePass(raw_ostream &o, bool ShouldPreserveUseListOrder,
+                              bool EmitThinLTOIndex)
         : ModulePass(ID), OS(o),
-          ShouldPreserveUseListOrder(ShouldPreserveUseListOrder) {}
+          ShouldPreserveUseListOrder(ShouldPreserveUseListOrder),
+          EmitThinLTOIndex(EmitThinLTOIndex) {}
 
     const char *getPassName() const override { return "Bitcode Writer"; }
 
     bool runOnModule(Module &M) override {
-      WriteBitcodeToFile(&M, OS, ShouldPreserveUseListOrder);
+      WriteBitcodeToFile(&M, OS, ShouldPreserveUseListOrder, EmitThinLTOIndex);
       return false;
     }
   };
 }
 
 char WriteBitcodePass::ID = 0;
 
 ModulePass *llvm::createBitcodeWriterPass(raw_ostream &Str,
-                                          bool ShouldPreserveUseListOrder) {
-  return new WriteBitcodePass(Str, ShouldPreserveUseListOrder);
+                                          bool ShouldPreserveUseListOrder,
+                                          bool EmitThinLTOIndex) {
+  return new WriteBitcodePass(Str, ShouldPreserveUseListOrder,
+                              EmitThinLTOIndex);
 }

lib/IR/CMakeLists.txt

add_llvm_library(LLVMCore
   Metadata.cpp
   MetadataTracking.cpp
   Module.cpp
   Operator.cpp
   Pass.cpp
   PassManager.cpp
   PassRegistry.cpp
   Statepoint.cpp
+  FunctionInfo.cpp
   Type.cpp
   TypeFinder.cpp
   Use.cpp
   User.cpp
   Value.cpp
   ValueSymbolTable.cpp
   ValueTypes.cpp
   Verifier.cpp

lib/IR/FunctionInfo.cpp

+//===-- FunctionInfo.cpp - Function Info Index ----------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the function info index and summary classes for the
+// IR library.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/IR/FunctionInfo.h"
+#include "llvm/ADT/StringMap.h"
+using namespace llvm;
+
+uint64_t FunctionInfoIndex::NextModuleId = 0;
+
+// Create the combined function index/summary from multiple
+// per-module instances.
+void FunctionInfoIndex::mergeFrom(FunctionInfoIndex *Other) {
+
+  ++NextModuleId;
+
+  const_funcinfo_iterator OtherFuncInfoLists = Other->funcinfo_begin();
+  StringRef ModPath;
+  for (; OtherFuncInfoLists != Other->funcinfo_end(); OtherFuncInfoLists++) {
+    StringRef FuncName = OtherFuncInfoLists->getKey();
+    FunctionInfoList List = OtherFuncInfoLists->second;
+
+    // Assert that the func info list only has one entry, since we shouldn't
+    // have duplicate names within a single per-module index.
+    assert(List.size() == 1);
+    FunctionInfo *Info = List.front();
+
+    // Add the module path string ref for this module if we haven't already
+    // saved a reference to it.
+    if (ModPath.empty())
+      ModPath = addModulePath(Info->functionSummary()->modulePath(),
+                              NextModuleId);
+    else
+      assert(ModPath == Info->functionSummary()->modulePath());
+
+    FunctionSummary *NewFS = new FunctionSummary(*Info->functionSummary());
+    // Note the module path string ref was copied above and is still owned by
+    // the original per-module FunctionInfo. Reset it to the new module path
+    // string reference owned by the combined index.
+    NewFS->setModulePath(ModPath);
+    FunctionInfo *NewInfo = new FunctionInfo(*Info);
+    NewInfo->setFunctionSummary(NewFS);
+
+    // If it is a local function, rename it.
+    if (Info->functionSummary()->isLocalFunction()) {
+      // Any local functions are virtually renamed when being added to the
+      // combined index map, to disambiguate from other functions with
+      // the same name. The symbol table created for the combined index
+      // file should contain the renamed symbols.
+      FuncName = FunctionInfoIndex::getGlobalNameForLocal(FuncName,
+                                                          NextModuleId);
+    }
+
+    // Add new function info to existing list. There may be duplicates when
+    // combining FunctionMap entries, due to COMDAT functions. Any local
+    // functions were virtually renamed above.
+    addFunctionInfo(FuncName, NewInfo);
+  }
+}

lib/Object/CMakeLists.txt

add_llvm_library(LLVMObject
   IRObjectFile.cpp
   MachOObjectFile.cpp
   MachOUniversal.cpp
   Object.cpp
   ObjectFile.cpp
   RecordStreamer.cpp
   SymbolicFile.cpp
   SymbolSize.cpp
+  FunctionIndexObjectFile.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${LLVM_MAIN_INCLUDE_DIR}/llvm/Object
 
   DEPENDS
   intrinsics_gen
   )

lib/Object/FunctionIndexObjectFile.cpp

+//===- FunctionIndexObjectFile.cpp - Function index file implementation ----===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// Part of the FunctionIndexObjectFile class implementation.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/Object/FunctionIndexObjectFile.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/Bitcode/ReaderWriter.h"
+#include "llvm/IR/FunctionInfo.h"
+#include "llvm/MC/MCStreamer.h"
+#include "llvm/Object/ObjectFile.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/raw_ostream.h"
+using namespace llvm;
+using namespace object;
+
+FunctionIndexObjectFile::FunctionIndexObjectFile(
+    MemoryBufferRef Object, std::unique_ptr<FunctionInfoIndex> I)
+    : SymbolicFile(Binary::ID_FunctionIndex, Object), Index(std::move(I)) {
+}
+
+FunctionIndexObjectFile::~FunctionIndexObjectFile() {
+}
+
+std::unique_ptr<FunctionInfoIndex> FunctionIndexObjectFile::takeIndex() {
+  return std::move(Index);
+}
+
+ErrorOr<MemoryBufferRef> FunctionIndexObjectFile::findBitcodeInObject(
+    const ObjectFile &Obj) {
+  for (const SectionRef &Sec : Obj.sections()) {
+    StringRef SecName;
+    if (std::error_code EC = Sec.getName(SecName))
+      return EC;
+    if (SecName == ".llvmbc") {
+      StringRef SecContents;
+      if (std::error_code EC = Sec.getContents(SecContents))
+        return EC;
+      return MemoryBufferRef(SecContents, Obj.getFileName());
+    }
+  }
+
+  return object_error::bitcode_section_not_found;
+}
+
+ErrorOr<MemoryBufferRef> FunctionIndexObjectFile::findBitcodeInMemBuffer(
+    MemoryBufferRef Object) {
+  sys::fs::file_magic Type = sys::fs::identify_magic(Object.getBuffer());
+  switch (Type) {
+  case sys::fs::file_magic::bitcode:
+    return Object;
+  case sys::fs::file_magic::elf_relocatable:
+  case sys::fs::file_magic::macho_object:
+  case sys::fs::file_magic::coff_object: {
+    ErrorOr<std::unique_ptr<ObjectFile>> ObjFile =
+        ObjectFile::createObjectFile(Object, Type);
+    if (!ObjFile)
+      return ObjFile.getError();
+    return findBitcodeInObject(*ObjFile->get());
+  }
+  default:
+    return object_error::invalid_file_type;
+  }
+}
+
+// Looks for function index in the given memory buffer.
+// returns true if found, else false.
+bool FunctionIndexObjectFile::hasFunctionSummaryInMemBuffer(
+    MemoryBufferRef Object, LLVMContext &Context) {
+  ErrorOr<MemoryBufferRef> BCOrErr = findBitcodeInMemBuffer(Object);
+  if (!BCOrErr)
+    return false;
+
+  return hasFunctionSummary(BCOrErr.get(), Context, nullptr);
+}
+
+// Parse function index in the given memory buffer.
+// Return new FunctionIndexObjectFile instance containing parsed
+// function summary/index.
+ErrorOr<std::unique_ptr<FunctionIndexObjectFile>>
+llvm::object::FunctionIndexObjectFile::create(MemoryBufferRef Object,
+                                              LLVMContext &Context,
+                                              bool ReadFuncSummaryData) {
+  std::unique_ptr<FunctionInfoIndex> Index;
+
+  ErrorOr<MemoryBufferRef> BCOrErr = findBitcodeInMemBuffer(Object);
+  if (!BCOrErr)
+    return BCOrErr.getError();
+
+  ErrorOr<std::unique_ptr<FunctionInfoIndex>> IOrErr =
+      getFunctionInfoIndex(BCOrErr.get(), Context, nullptr,
+                           ReadFuncSummaryData);
+
+  if (std::error_code EC = IOrErr.getError())
+    return EC;
+
+  Index = std::move(IOrErr.get());
+
+  return llvm::make_unique<FunctionIndexObjectFile>(Object, std::move(Index));
+}
+
+// Parse the function summary information for function with the
+// given name out of the given buffer. Parsed information is
+// stored on the index object saved in this object.
+std::error_code FunctionIndexObjectFile::findFunctionSummaryInMemBuffer(
+    MemoryBufferRef Object, LLVMContext &Context, StringRef FunctionName) {
+  sys::fs::file_magic Type = sys::fs::identify_magic(Object.getBuffer());
+  switch (Type) {
+    case sys::fs::file_magic::bitcode: {
+      return readFunctionSummary(Object, Context, nullptr,
+                                 FunctionName, std::move(Index));
+    }
+    default:
+      return object_error::invalid_file_type;
+  }
+}

test/Bitcode/thinlto-function-summary.ll

+; RUN: llvm-as -thinlto < %s | llvm-bcanalyzer -dump | FileCheck %s -check-prefix=BC
+; Check for function summary block/records.
+
+; BC: <FUNCTION_SUMMARY_BLOCK
+; BC-NEXT: <PERMODULE_ENTRY
+; BC-NEXT: <PERMODULE_ENTRY
+; BC-NEXT: </FUNCTION_SUMMARY_BLOCK
+
+; RUN: llvm-as -thinlto < %s | llvm-dis | FileCheck %s
+; Check that this round-trips correctly.
+
+; ModuleID = '<stdin>'
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; CHECK: define i32 @foo()
+
+; Function Attrs: nounwind uwtable
+define i32 @foo() #0 {
+entry:
+  ret i32 1
+}
+
+; CHECK: define i32 @bar(i32 %x)
+
+; Function Attrs: nounwind uwtable
+define i32 @bar(i32 %x) #0 {
+entry:
+  ret i32 %x
+}

test/tools/gold/X86/Inputs/thinlto.ll

+define void @g() {
+entry:
+  ret void
+}

test/tools/gold/X86/thinlto.ll

+; RUN: llvm-as -thinlto %s -o %t.o
+; RUN: llvm-as -thinlto %p/Inputs/thinlto.ll -o %t2.o
+
+; RUN: %gold -plugin %llvmshlibdir/LLVMgold.so \
+; RUN:    --plugin-opt=thinlto \
+; RUN:    -shared %t.o %t2.o -o %t3
+; RUN: llvm-bcanalyzer -dump %t3.thinlto.bc | FileCheck %s --check-prefix=COMBINED
+; RUN: not test -a %t3
+
+; COMBINED: <MODULE_STRTAB_BLOCK
+; COMBINED-NEXT: <ENTRY
+; COMBINED-NEXT: <ENTRY
+; COMBINED-NEXT: </MODULE_STRTAB_BLOCK
+; COMBINED-NEXT: <FUNCTION_SUMMARY_BLOCK
+; COMBINED-NEXT: <COMBINED_ENTRY
+; COMBINED-NEXT: <COMBINED_ENTRY
+; COMBINED-NEXT: </FUNCTION_SUMMARY_BLOCK
+
+define void @f() {
+entry:
+  ret void
+}

tools/gold/gold-plugin.cpp

 #include "llvm/IR/DiagnosticPrinter.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/LegacyPassManager.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/Verifier.h"
 #include "llvm/Linker/Linker.h"
 #include "llvm/MC/SubtargetFeature.h"
 #include "llvm/Object/IRObjectFile.h"
+#include "llvm/Object/FunctionIndexObjectFile.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Support/Host.h"
 #include "llvm/Support/ManagedStatic.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include "llvm/Support/TargetRegistry.h"
 #include "llvm/Support/TargetSelect.h"
 #include "llvm/Transforms/IPO.h"
 #include "llvm/Transforms/IPO/PassManagerBuilder.h"
#ifdef NDEBUG
   static bool DisableVerify = true;
 #else
   static bool DisableVerify = false;
 #endif
   static std::string obj_path;
   static std::string extra_library_path;
   static std::string triple;
   static std::string mcpu;
+  // When the thinlto plugin option is specified, only read the function
+  // the information from intermediate files and write a combined
+  // global index for the ThinLTO backends.
+  static bool thinlto = false;
   // Additional options to pass into the code generator.
   // Note: This array will contain all plugin options which are not claimed
   // as plugin exclusive to pass to the code generator.
   // For example, "generate-api-file" and "as"options are for the plugin
   // use only and will not be passed.
   static std::vector<const char *> extra;
 
   static void process_plugin_option(const char *opt_)
static void process_plugin_option(const char *opt_)
     } else if (opt.startswith("obj-path=")) {
       obj_path = opt.substr(strlen("obj-path="));
     } else if (opt == "emit-llvm") {
       TheOutputType = OT_BC_ONLY;
     } else if (opt == "save-temps") {
       TheOutputType = OT_SAVE_TEMPS;
     } else if (opt == "disable-output") {
       TheOutputType = OT_DISABLE;
+    } else if (opt == "thinlto") {
+      thinlto = true;
     } else if (opt.size() == 2 && opt[0] == 'O') {
       if (opt[1] < '0' || opt[1] > '3')
         message(LDPL_FATAL, "Optimization level must be between 0 and 3");
       OptLevel = opt[1] - '0';
     } else if (opt.startswith("jobs=")) {
       if (StringRef(opt_ + 5).getAsInteger(10, Parallelism))
         message(LDPL_FATAL, "Invalid parallelism level: %s", opt_ + 5);
     } else if (opt == "disable-verify") {
static ld_plugin_status claim_file_hook(const ld_plugin_input_file *file,
   }
   std::unique_ptr<object::IRObjectFile> Obj = std::move(*ObjOrErr);
 
   Modules.resize(Modules.size() + 1);
   claimed_file &cf = Modules.back();
 
   cf.handle = file->handle;
 
+  // If we are doing ThinLTO compilation, don't need to process the symbols.
+  // Later we simply build a combined index file after all files are claimed.
+  if (options::thinlto)
+    return LDPS_OK;
+
   for (auto &Sym : Obj->symbols()) {
     uint32_t Symflags = Sym.getFlags();
     if (shouldSkip(Symflags))
       continue;
 
     cf.syms.push_back(ld_plugin_symbol());
     ld_plugin_symbol &sym = cf.syms.back();
     sym.version = nullptr;

 
 static void freeSymName(ld_plugin_symbol &Sym) {
   free(Sym.name);
   free(Sym.comdat_key);
   Sym.name = nullptr;
   Sym.comdat_key = nullptr;
 }
 
+static std::unique_ptr<FunctionInfoIndex>
+getFunctionIndexForFile(LLVMContext &Context, claimed_file &F,
+                        ld_plugin_input_file &Info) {
+
+  if (get_symbols(F.handle, F.syms.size(), &F.syms[0]) != LDPS_OK)
+    message(LDPL_FATAL, "Failed to get symbol information");
+
+  const void *View;
+  if (get_view(F.handle, &View) != LDPS_OK)
+    message(LDPL_FATAL, "Failed to get a view of file");
+
+  MemoryBufferRef BufferRef(StringRef((const char *)View, Info.filesize),
+                            Info.name);
+  ErrorOr<std::unique_ptr<object::FunctionIndexObjectFile>> ObjOrErr =
+      object::FunctionIndexObjectFile::create(BufferRef, Context);
+
+  if (std::error_code EC = ObjOrErr.getError())
+    message(LDPL_FATAL, "Could not read function index bitcode from file : %s",
+            EC.message().c_str());
+
+  object::FunctionIndexObjectFile &Obj = **ObjOrErr;
+
+  return Obj.takeIndex();
+}
+
 static std::unique_ptr<Module>
 getModuleForFile(LLVMContext &Context, claimed_file &F,
                  ld_plugin_input_file &Info, raw_fd_ostream *ApiFile,
                  StringSet<> &Internalize, StringSet<> &Maybe) {
 
   if (get_symbols(F.handle, F.syms.size(), F.syms.data()) != LDPS_OK)
     message(LDPL_FATAL, "Failed to get symbol information");
 

 /// native object file. Then, perform optimization and codegen.
 static ld_plugin_status allSymbolsReadHook(raw_fd_ostream *ApiFile) {
   if (Modules.empty())
     return LDPS_OK;
 
   LLVMContext Context;
   Context.setDiagnosticHandler(diagnosticHandler, nullptr, true);
 
+  // If we are doing ThinLTO compilation, simply build the combined
+  // function index/summary and emit it. We don't need to parse the modules
+  // and link them in this case.
+  if (options::thinlto) {
+    std::unique_ptr<FunctionInfoIndex> CombinedIndex(new FunctionInfoIndex());
+    for (claimed_file &F : Modules) {
+      ld_plugin_input_file File;
+      if (get_input_file(F.handle, &File) != LDPS_OK)
+        message(LDPL_FATAL, "Failed to get file information");
+
+      std::unique_ptr<FunctionInfoIndex> Index =
+          getFunctionIndexForFile(Context, F, File);
+      CombinedIndex->mergeFrom(Index.get());
+    }
+
+    std::error_code EC;
+    raw_fd_ostream OS(output_name + ".thinlto.bc", EC,
+                      sys::fs::OpenFlags::F_None);
+    if (EC)
+      message(LDPL_FATAL, "Unable to open %s.thinlto.bc for writing: %s",
+              output_name.data(), EC.message().c_str());
+    WriteFunctionSummaryToFile(CombinedIndex.get(), OS);
+    OS.close();
+
+    cleanup_hook();
+    exit(0);
+  }
+
   std::unique_ptr<Module> Combined(new Module("ld-temp.o", Context));
   Linker L(Combined.get());
 
   std::string DefaultTriple = sys::getDefaultTargetTriple();
 
   StringSet<> Internalize;
   StringSet<> Maybe;
   for (claimed_file &F : Modules) {

tools/llvm-as/llvm-as.cpp

 
 static cl::opt<bool>
 Force("f", cl::desc("Enable binary output on terminals"));
 
 static cl::opt<bool>
 DisableOutput("disable-output", cl::desc("Disable output"), cl::init(false));
 
 static cl::opt<bool>
+EmitThinLTO("thinlto", cl::desc("Emit function summary index for ThinLTO"),
+            cl::init(false));
+
+static cl::opt<bool>
 DumpAsm("d", cl::desc("Print assembly as parsed"), cl::Hidden);
 
 static cl::opt<bool>
 DisableVerify("disable-verify", cl::Hidden,
               cl::desc("Do not run verifier on input LLVM (dangerous!)"));
 
 static cl::opt<bool> PreserveBitcodeUseListOrder(
     "preserve-bc-uselistorder",
static void WriteOutputFile(const Module *M) {
   std::unique_ptr<tool_output_file> Out(
       new tool_output_file(OutputFilename, EC, sys::fs::F_None));
   if (EC) {
     errs() << EC.message() << '\n';
     exit(1);
   }
 
   if (Force || !CheckBitcodeOutputToConsole(Out->os(), true))
-    WriteBitcodeToFile(M, Out->os(), PreserveBitcodeUseListOrder);
+    WriteBitcodeToFile(M, Out->os(), PreserveBitcodeUseListOrder,
+                       EmitThinLTO);
 
   // Declare success.
   Out->keep();
 }
 
 int main(int argc, char **argv) {
   // Print a stack trace if we signal out.
   sys::PrintStackTraceOnErrorSignal();

tools/llvm-bcanalyzer/llvm-bcanalyzer.cpp

static const char *GetBlockName(unsigned BlockID,
   case bitc::PARAMATTR_GROUP_BLOCK_ID: return "PARAMATTR_GROUP_BLOCK_ID";
   case bitc::TYPE_BLOCK_ID_NEW:        return "TYPE_BLOCK_ID";
   case bitc::CONSTANTS_BLOCK_ID:       return "CONSTANTS_BLOCK";
   case bitc::FUNCTION_BLOCK_ID:        return "FUNCTION_BLOCK";
   case bitc::VALUE_SYMTAB_BLOCK_ID:    return "VALUE_SYMTAB";
   case bitc::METADATA_BLOCK_ID:        return "METADATA_BLOCK";
   case bitc::METADATA_ATTACHMENT_ID:   return "METADATA_ATTACHMENT_BLOCK";
   case bitc::USELIST_BLOCK_ID:         return "USELIST_BLOCK_ID";
+  case bitc::FUNCTION_SUMMARY_BLOCK_ID:
+                                       return "FUNCTION_SUMMARY_BLOCK";
+  case bitc::MODULE_STRTAB_BLOCK_ID: return "MODULE_STRTAB_BLOCK";
   }
 }
 
 /// GetCodeName - Return a symbolic code name if known, otherwise return
 /// null.
 static const char *GetCodeName(unsigned CodeID, unsigned BlockID,
                                const BitstreamReader &StreamFile,
                                CurStreamTypeType CurStreamType) {
default: return nullptr;
       STRINGIFY_CODE(FUNC_CODE, INST_GEP)
     }
   case bitc::VALUE_SYMTAB_BLOCK_ID:
     switch (CodeID) {
     default: return nullptr;
     STRINGIFY_CODE(VST_CODE, ENTRY)
     STRINGIFY_CODE(VST_CODE, BBENTRY)
     STRINGIFY_CODE(VST_CODE, FNENTRY)
+    STRINGIFY_CODE(VST_CODE, COMBINED_FNENTRY)
+    }
+  case bitc::MODULE_STRTAB_BLOCK_ID:
+    switch (CodeID) {
+    default: return nullptr;
+    STRINGIFY_CODE(MST_CODE, ENTRY)
+    }
+  case bitc::FUNCTION_SUMMARY_BLOCK_ID:
+    switch (CodeID) {
+    default: return nullptr;
+    STRINGIFY_CODE(FS_CODE, PERMODULE_ENTRY)
+    STRINGIFY_CODE(FS_CODE, COMBINED_ENTRY)
     }
   case bitc::METADATA_ATTACHMENT_ID:
     switch(CodeID) {
     default:return nullptr;
       STRINGIFY_CODE(METADATA, ATTACHMENT)
     }
   case bitc::METADATA_BLOCK_ID:
     switch(CodeID) {